The Cambridge Handbook of Germanic Linguistics 9781108421867, 9781108378291, 2018054844

1,418 59 5MB

English Pages [884] Year 2020

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

The Cambridge Handbook of Germanic Linguistics
 9781108421867, 9781108378291, 2018054844

Citation preview

The Cambridge Handbook of Germanic Linguistics Edited by Michael T. Putnam The Pennsylvania State University

B. Richard Page The Pennsylvania State University

University Printing House, Cambridge CB2 8BS, United Kingdom One Liberty Plaza, 20th Floor, New York, NY 10006, USA 477 Williamstown Road, Port Melbourne, VIC 3207, Australia 314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi – 110025, India 79 Anson Road, #06–04/06, Singapore 079906 Cambridge University Press is part of the University of Cambridge. It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning, and research at the highest international levels of excellence. www.cambridge.org Information on this title: www.cambridge.org/9781108421867 DOI: 10.1017/9781108378291 © Cambridge University Press 2020 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2020 Printed in the United Kingdom by TJ International Ltd, Padstow Cornwall A catalogue record for this publication is available from the British Library. Library of Congress Cataloging-in-Publication Data Names: Putnam, Michael T., editor. | Page, B. Richard, editor. Title: The Cambridge handbook of Germanic linguistics / edited by Michael T. Putnam, B. Richard Page. Description: New York, NY : Cambridge University Press, [2020] | Series: Cambridge handbooks in language and linguistics | Includes index. Identifiers: LCCN 2018054844 | ISBN 9781108421867 Subjects: LCSH: Germanic languages. | Germanic languages – Handbooks, manuals, etc. Classification: LCC PD25 .C36 2019 | DDC 430–dc23 LC record available at https://lccn.loc.gov/2018054844 ISBN 978-1-108-42186-7 Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

Contents

List of Figures List of Maps List of Tables List of Contributors Acknowledgments Germanic Languages: An Overview

page ix xi xii xiv xvi

B. Richard Page and Michael

T. Putnam Part I Phonology 1 Phonological Processes in Germanic Languages Tracy Alan Hall 2 Germanic Syllable Structure Marc van Oostendorp 3 The Role of Foot Structure in Germanic Laura Catharine Smith 4 Word Stress in Germanic Birgit Alber 5 Quantity in Germanic Languages B. Richard Page 6 Germanic Laryngeal Phonetics and Phonology

Joseph Salmons 7 Tone Accent in North and West Germanic Bjo¨rn Ko¨hnlein 8 Intonation in Germanic Mary Grantham O’Brien Part II Morphology and Agreement Systems 9 Verbal Inflectional Morphology in Germanic David Fertig 10 Inflectional Morphology: Nouns Damaris Nu¨bling 11 Principles of Word Formation Geert Booij 12 Grammatical Gender in Modern Germanic Languages

Sebastian Ku¨rschner 13 Case in Germanic Thomas McFadden 14 Complementizer Agreement Marjo van Koppen

1 9 11 33 49 73 97

119 143 167 191 193 214 238

259 282 313

viii

CONTENTS

Part III Syntax 15 VO-/OV-Base Ordering Hubert Haider 16 The Placement of Finite Verbs Sten Vikner 17 Germanic Infinitives Susi Wurmbrand and

Christos Christopoulos

337 339 365

389

18 The Unification of Object Shift and Object Scrambling 19 20 21 22 23

Hans Broekhuis Unbounded Dependency Constructions in Germanic Martin Salzmann The Voice Domain in Germanic Artemis Alexiadou and Florian Scha¨fer Binding: The Morphology, Syntax, and Semantics of Reflexive and Nonreflexive Pronouns Vera Lee-Schoenfeld Verbal Particles, Results, and Directed Motion Ida Toivonen Structure of Noun (NP) and Determiner Phrases (DP) Dorian Roehrs

413 436 461 493 516 537

Part IV Semantics and Pragmatics 24 Modality in Germanic Kristin Melum Eide 25 Tense and Aspect in Germanic Languages Kristin Melum Eide 26 Prepositions and Particles: Place and Path in English, German, and Dutch Joost Zwarts 27 Negative and Positive Polarity Items Doris Penka 28 Grammatical Reflexes of Information Structure in Germanic Languages Caroline Fe´ry

565 567 591

Part V Language Contact and Nonstandard Varieties 29 Second Language Acquisition of Germanic Languages

687

Carrie Jackson

615 639 661

689

30 Urban Speech Styles of Germanic Languages Pia Quist and Bente 31 32 33 34 35

A. Svendsen The West Germanic Dialect Continuum William D. Keel The North Germanic Dialect Continuum Charlotte Gooskens Heritage Germanic Languages in North America Janne Bondi Johannessen and Michael T. Putnam Minority Germanic Languages Mark L. Louden Germanic Contact Languages Paul T. Roberge

Index

714 736 761 783 807 833 865

Figures

7.1 Tonal contours in Cologne, focus, nonfinal position; stressed accent syllable nonshaded, overall post-tonic contour shaded page 145 7.2 Tonal contours in Cologne (Rule A) and Arzbach (Rule B), focus, nonfinal position; nuclear contour nonshaded, overall post-nuclear contour shaded 147 7.3 Analysis with lexical tone for Rule A (Gussenhoven and Peters 2004), Cologne dialect; focus, nonfinal position 149 7.4 Analysis with a foot-based approach (Ko¨hnlein 2016) for Rule A, Cologne dialect; focus, nonfinal position 150 7.5 Reconstructed accent contours before and after tone accent genesis in West Germanic (before = dashed contour; after = solid contour) 152 7.6 The development from the predecessor toward Rule A (before = dashed contour; after = solid contour) 153 7.7 The development from the predecessor toward Rule B (before = dashed contour; after = solid contour) 153 7.8 Idealized tonal contours in Central Swedish 154 7.9 Idealized tonal contours in different dialect areas, distinguished by number of peaks in Accent 2 (1 / 2) and alignment of the tonal melodies (A = early; B = late) 155 7.10 Competing analyses with lexical tone for Central Swedish 158 7.11 Basics of the North Germanic tone accent genesis in the accent-first, peak-delay approach (before = dashed contour; after = solid contour) 161 7.12 Development of two-peaked Accent 2 in the accent-first, peak-delay approach 162 8.1 Pitch track of the German utterance “auf gar keinen Fall” (“by no means”) 169 8.2 Tonal alignment in L1 German 173

x

LIST OF FIGURES

8.3 Tonal alignment in L2 English 8.4 A stylized Dutch hat pattern (“Peter comes never too late.”) recreated from ’t Hart (1998: 97) 8.5 Stylized intonation contour of Danish statements, whquestions, questions with word order inversion, and declarative questions (recreated from Grønnum 2009: 600) 8.6 Contour of a Dutch interrogative “Heeft Peter een nieuwe auto gekocht?” (“Has Peter bought a new car?”) reproduced from ’t Hart (1998: 103) 8.7 Norwegian utterance “Du aner ikke hvor dyr den er, vel?” (“You do not know how expensive it is, right?”) with rising intonation on the tag question 8.8 Norwegian utterance “Du aner ikke hvor dyr den er, vel.” (”Well, you do not know how expensive it is.”) with falling intonation on the tag 10.1 Complexity of nominal inflection of the GMC languages (based on Ku¨rschner 2008: 8) 10.2 Preservation, elimination and functionalization of umlaut in GMC languages 10.3 Direction of formal determination and influence between stem and suffix 12.1 Pragmatic hierarchy of gender agreement, from Ko¨pcke et al. (2010: 179) 12.2 The pronominal gender system of Danish and Swedish, translated and adapted from Braunmu¨ller (2007: 54) 25.1 Potential boundaries of an event 25.2 Ingressive, dynamic, eventive, inchoative 25.3 Progressive, imperfective, continuative, durative, stative, atelic 25.4 Egressive, perfective, telic 25.5 Iterative, generic, habitual 28.1 Afrikaans sentence with wide focus 28.2 Afrikaans sentence with a narrow focus 28.3 Afrikaans sentence with a topic 32.1 The Nordic language tree 32.2 The model of the modern Nordic languages (from Torp 2002: 19) 32.3 The mean results of the five investigations on mutual intelligibility in Scandinavia in Table 32.1 33.1 VOT-times in Wisconsin Ko¨lsch and varieties of English (from Geiger and Salmons 2006) 33.2 Differences in Wisconsin Heritage German case marking between NPs and pronouns 33.3 Frequency of the two word orders in the production of subordinate clauses in MSG according to clause type

173 177

178

180

182

182 231 232 233 271 273 598 598 599 600 602 670 671 674 763 764 774 786 790 794

Maps

31.1 Westgermanic Dialect Continuum: Low German varieties 1–11; Low Franconian (Dutch) varieties 12–20; Middle German varieties 21–28; Upper German varieties 29–32; Frisian varieties 33–35 page 738 31.2 Apocope of final schwa Atlas zur deutschen Alltagssprache 752 31.3 N-AD case systems to the east (Shrier 1965: 434) 754 31.4 NA-D case systems to the west (Shrier 1965: 435) 755 31.5 Bro¨tchen ‘bread roll’ (Elspaß and Mo¨ller 2003ff.) 757 31.6 Tag question gell? (Elspaß and Mo¨ller 2003ff.) 758 32.1 Characterization of the Nordic dialects (from Bandle 1973, map 22) 767

Tables

5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 6.1 7.1 7.2 7.3 9.1 9.2 10.1 10.2 10.3 10.4 10.5 10.6 10.7 10.8

Typology of quantity (revised later in this chapter) page 98 Distribution of Old Alemannic nonfricative obstruents 102 Stops in Thurgovian (Swiss German) 103 Neutralization of fortis-lenis contrast word-initially in Zu¨rich German 103 Vowels in American English 111 American English vowels in stressed open syllables 111 Distribution of vowels in stressed syllables 112 Dutch vowels in stressed syllables 113 Typology of quantity (revised version) 114 Summary of laryngeal phonetics, phonology, assimilation, and neutralization in Germanic 138 Some effects of affixes on the accent class of words 157 Basics of the North Germanic tone accent genesis in the “accent-first, double-peak” approach (Riad 1998) 160 Basics of the North Germanic tone accent genesis in the “stød-first” approach: development of stød versus no stød 162 Full paradigms of ‘give’ (strong) and ‘hear’ (weak) in Icelandic and German 194 Ablaut patterns of Germanic strong verbs 207 Declension class system in Proto-Germanic (based on Krahe 1969 and Ramat 1981) 216 Inflectional paradigm for Icelandic hestur ‘horse’ (m), a-class 218 The paradigm of Icelandic vo¨llur ‘field’ (m), u-class 219 The paradigm of Icelandic fjo¨rður ‘fjord’ (m), u-class 219 The Swedish declension class system 222 The Danish declension class system 222 The German declension class system (shaded rows: not productive) 224 The West Frisian declension class system (shaded rows: not productive) 227

List of Tables

10.9 The Dutch declension class system (shaded rows: not productive) 12.1 Overview of adnominal and pronominal genders in Germanic languages (adapted from Audring 2010: 697) 13.1 PGmc ‘day’ and ‘deed’ 13.2 OE ‘day’ and ‘deed’ 13.3 OE simple demonstrative / definite article 13.4 German ‘day’ and ‘cow’ 15.1 A synopsis of syntactic correlates of OV / VO 17.1 ECM 17.2 Splitting and IPP in Dutch 17.3 Splitting and long passive in German 17.4 Parasitic morphology in Swedish 17.5 Implicational restructuring hierarchy 17.6 ECM hierarchy 18.1 Linearization of VO-languages 18.2 Linearization of OV-languages 21.1 Anaphoric elements of English 21.2 Anaphoric elements of German 21.3 Typology of anaphoric expressions (Reinhart and Reuland 1993: 659) 21.4 Anaphoric elements of Dutch 21.5 Dutch and German pronoun systems (Bu¨ring 2005: 75) 21.6 Danish and Norwegian pronoun system (Bu¨ring 2005: 76) 21.7 Third person reflexive pronouns in Icelandic (Thra´insson 2007: 463) 23.1 Differences between North and West Germanic 26.1 Simple reflexive PathPs 26.2 Complex reflexive PathP 32.1 Results of five investigations on the mutual intelligibility between spoken and written Danish, Norwegian, and Swedish 33.1 Proportions of case marking on elements by definiteness 34.1 Classification of minority Germanic languages 34.2 Classification of minority Germanic languages and language endangerment

228 261 284 286 286 287 342 395 401 403 406 408 409 421 422 494 494 502 503 504 508 512 559 636 636

771 791 814 827

xiii

Contributors

Birgit Alber, University of Verona Artemis Alexiadou, Humboldt University of Berlin Geert Booij, Leiden University Hans Broekhuis, Meertens Institute Christos Christopoulos, University of Connecticut Kristin Melum Eide, Norwegian University of Science and Technology (NTNU) David Fertig, University at Buffalo (SUNY) Caroline Fe´ry, Goethe University Frankfurt Charlotte Gooskens, University of Groningen Hubert Haider, University of Salzburg Tracy Alan Hall, Indiana University Bloomington Carrie Jackson, The Pennsylvania State University Janne Bondi Johannessen, University of Oslo William D. Keel, University of Kansas Bjo¨rn Ko¨hnlein, The Ohio State University Marjo van Koppen, Meertens Institute and Utrecht University, Institute of Linguistics Sebastian Ku¨rschner, Katholische Universita¨t Eichsta¨tt-Ingolstadt Vera Lee-Schoenfeld, University of Georgia Mark L. Louden, University of Wisconsin–Madison Thomas McFadden, Leibniz-Zentrum Allgemeine Sprachwissenschaft (ZAS) Damaris Nu¨bling, Johannes Gutenberg University Mainz Mary Grantham O’Brien, University of Calgary Marc van Oostendorp, Meertens Institute and Radboud University of Nijmegen B. Richard Page, The Pennsylvania State University Doris Penka, University of Konstanz Michael T. Putnam, The Pennsylvania State University Pia Quist, University of Copenhagen Paul T. Roberge, University of North Carolina at Chapel Hill

List of Contributors

Dorian Roehrs, University of North Texas Joseph Salmons, University of Wisconsin-Madison Martin Salzmann, University of Pennsylvania Florian Scha¨fer, Humboldt University of Berlin Laura Catharine Smith, Brigham Young University Bente A. Svendsen, University of Oslo Ida Toivonen, Carleton University Sten Vikner, Aarhus University Susi Wurmbrand, University of Vienna Joost Zwarts, Utrecht University

xv

Acknowledgments

Anyone who has ever been involved in a massive undertaking such as editing a comprehensive handbook on a specialized domain of academic inquiry such as this requires the support of others in order to bring forth a volume of the highest possible quality. We are grateful to Cambridge University Press for the opportunity to publish this volume alongside other excellent volumes in the series Cambridge Handbooks in Language and Linguistics. Special praise and thanks are due to Andrew Winnard, Adam Hooper, and Ruth Durbridge, for assistance and consultation at various stages of the development of this handbook. In the initial stages of the development of this handbook, we benefited from detailed comments and suggestions from six reviewers. These comments undoubtedly helped shape the final product. Those who contributed chapters to this handbook are distinguished, internationally renowned scholars in various areas of Germanic Linguistics. Handbooks are truly a celebration of their brilliance and the insights they offer to students and established scholars alike. Without them and their willingness to participate in this project, this handbook would not exist, so we extend our gratitude to their hard work and continued dedication to the field. We also thank the many reviewers of these contributions who helped shape and improve them. All of you played an essential role in the success of this volume you are reading right now. At the local level, we would like to thank the Department of Germanic and Slavic Languages and Literatures (GSLL) at Penn State University, in particular Tom Beebee, for their generous financial support at various stages of this project. James Kopf and Chrisann Zuerner provided excellent editorial assistance in the intermediate phase of this project. Both of us count our blessings to be surrounded by colleagues in our home department and those associated and affiliated

Acknowledgments

with the Program of Linguistics and the Center for Language Science in creating a fun and stimulating environment to study (Germanic) Linguistics. Last and certainly not least, we’d like to thank our families (Mike: Jill & Abby / Richard: Katherine, Briar, Charlie, & Louise) for their unyielding love and support that finds their way into all of our work.

xvii

Germanic Languages An Overview B. Richard Page and Michael T. Putnam

I.1 Introduction The Germanic languages include some of the world’s most widely spoken and thoroughly researched languages. English has become a global language that serves as a lingua franca in many parts of the world and has an estimated 1.12 billon speakers (Simons and Fennig 2018). German, Dutch, Icelandic, Swedish, and Norwegian have also been studied and described extensively from both diachronic and synchronic perspectives. In addition to the standard varieties of these languages, there are available descriptions of many nonstandard varieties as well as of regional and minority languages, such as Frisian and Low German. There are several possibilities to consider when putting together a handbook of a language family. One would be to have a chapter devoted to each language, as in Ko¨nig and Auwera (1993). This would require the choice of which languages to include and would make cross-linguistic comparisons difficult. In the alternative, one could take a thematic, comparative approach and examine different aspects of the Germanic languages, such as morphology or syntax, as in Harbert (2007). We have chosen the latter approach for this handbook. The advantage of such an approach is that it gives the contributors in the volume an opportunity to compare and contrast the broad range of empirical phenomena found across the Germanic languages that have been documented in the literature. As a result, this handbook extends beyond treatments of the phonology, morphology, and syntax of the standard varieties of the Germanic languages to include the nonstandard varieties and topics such as semantics, second language acquisition, Germanic contact languages, and Germanic minority languages. The focus is on contemporary Germanic languages and synchrony. Earlier stages of the Germanic languages and historical developments are not treated in any kind of systematic way in this volume (see Fulk 2018 for an in-depth treatment of the early Germanic languages).

2

B . R I C H A R D PA G E A N D M I C H A E L T. P U T N A M

The Germanic languages are genetically related and constitute a branch of the Indo-European language family. Germanic is generally divided into three branches: East, North and West. The chief representative of East Germanic is Gothic, the earliest attested Germanic language. Wulfila, a bishop to the Goths, translated the Bible into Gothic in approximately 350 CE, and the earliest surviving manuscripts were transcribed around 500 CE. The East Germanic languages are now extinct though a dialect of Gothic known as Crimean Gothic was spoken in the Crimea into the eighteenth century (see Stearns 1978). Of the 47 Germanic languages spoken today and listed by Ethnologue, six belong to North Germanic and 41 to West Germanic. Louden (Chapter 34) provides a discussion of the difficulty of classifying some of the language varieties listed by Ethnologue as languages or dialects before proceeding to an overview of the Germanic languages spoken in minority language communities around the world. Examples of languages discussed in the chapter on minority languages by Louden include Pennsylvania German, different varieties of Frisian, Mennonite Low German (also called Plautdietsch), and many others. This chapter will focus on those Germanic languages which may be regarded as majority languages. Roberge (Chapter 35) discusses Germanic-based creoles and pidgins. Roberge follows Thomason’s (2001: 158) definition of a contact language as “any new language that arises in a contact situation. Linguistically, a contact language is identifiable by the fact that its lexicon and grammatical structures cannot all be traced back primarily to the same source language; they are therefore mixed languages in the technical historical linguistic sense; they did not arise primarily through descent with modification from a single earlier language.” The classification of a language as a majority language, a minority language or a contact language can be quite clear. For example, one can readily agree that Icelandic is the majority language in Iceland, Low German is a minority language in Germany, and Unserdeutsch is a contact language spoken in New Guinea. However, these classifications often overlap. For example, Swedish is the majority language in Sweden but a minority language in Finland. Moreover, it is sometimes debatable whether or not a language is in fact a contact language. Two examples discussed by Roberge (Chapter 35) as noncanonical contact languages are Afrikaans and English. Scholars often have conflicting opinions about how these two languages should be categorized. For instance, Bailey and Maroldt (1977) maintain that English is a French creole whereas Poussa (1982) claims that English is a Norse creole. Lutz (2012) argues that both analyses are wrong and that English is not a creole language. In this chapter, we provide a brief survey of Germanic languages that are commonly held to be majority languages followed by an overview of the chapters in the handbook. We begin with West Germanic followed by North Germanic. Majority languages in the West Germanic branch of the family are English, Dutch, German, and Luxembourgish. The North Germanic languages that may be classified as majority languages include

Germanic Languages

Icelandic, Faroese, Norwegian, Swedish, and Danish. Of course, these majority languages also have dialects. West Germanic dialects are addressed in Chapter 31 by William D. Keel, and Charlotte Gooskens discusses the North Germanic dialects in Chapter 32.

I.2 West Germanic English, the most widely spoken Germanic language, belongs to the West Germanic branch of the family. English is estimated to have 1.12 billion users including over 370 million speakers for whom English is their first language or L1 (Simons and Fennig 2018). The nation with the greatest number of speakers is the United States with more than 300 million followed by India with over 200 million. The two nations provide excellent examples of the global diversity of English. In the United States, the approximately 260 million L1 English speakers constitute the great majority of total English speakers in the country. The United States also has a great number of varieties of English, including African American Vernacular English and many regional and social dialects. In contrast, in India almost all English users acquired English as a second language (L2), and the language serves as a lingua franca. Indian English is a distinct variety of English and is one of many indigenized varieties of English found globally that are often referred to as World Englishes (see Kirkpatrick 2010 for a discussion of World Englishes). English has also played a crucial role in the development of many contact languages. Roberge (Chapter 35) provides a discussion of the many English-based pidgins and creoles. German is the second most widely spoken Germanic language and has approximately 130 million speakers. German is the statutory national language of Germany, Austria, and Liechtenstein. It is also a national language in Switzerland, co-equal with French and Italian on the federal level. Germany, Austria, and Switzerland each has its own variety of the standard language (Ammon 1995). In Luxembourg, German is one of three official languages. It is also a statutory provincial language in German-speaking areas of Belgium, in the Syddanmark region of Denmark, and in the South Tyrol region of Italy (Simons and Fennig 2018, see also Louden, Chapter 34). German has a great number of regional dialects that often diverge quite significantly from the standard varieties. Dutch has approximately 23 million speakers and is the national language of the Netherlands and Suriname. It is also the national language of Belgium alongside French. There are approximately 16 million Dutch speakers in the Netherlands, 6 million in Belgium, and 280,000 in Suriname (Simons and Fennig 2018). Dutch is also the principal lexifier of a number of creoles, and Afrikaans arose from Dutch in colonial southern Africa via language contact (Roberge, Chapter 35).

3

4

B . R I C H A R D PA G E A N D M I C H A E L T. P U T N A M

Luxembourgish has close to 400,000 L1 speakers. The great majority of Luxembourgish speakers (303,000) are in Luxembourg, where Luxembourgish is one of three national languages, alongside German and French. Luxembourgish is also spoken by 40,000 in the Grand-Est region of France, by 30,000 in parts of the Luxembourg province of Belgium, and by 18,200 in the Bitburg area of Rheinland-Pfalz in Germany (Simons and Fennig 2018).

I.3 North Germanic Swedish has approximately 9.6 million L1 speakers and 3.2 million L2 speakers and is the national language of Sweden. It is a minority language in Finland where there are 274,000 L1 Swedish speakers (Simons and Fennig 2018). Swedish has a number of dialects with significant differences in phonology. However, the dialects appear to be largely mutually intelligible. In fact, there is a large degree of mutual intelligibility across Swedish, Norwegian, and Danish, which make up the Scandinavian group of North Germanic languages (Gooskens, Chapter 32). Norwegian has slightly more than 5 million users, almost all of whom live in Norway. Norway is unusual in that there are two written standard languages, Bokmaº l and Nynorsk. Bokmaº l is used by the majority of Norwegians. Nynorsk has been declared the official written language for some municipalities in western areas of Norway that are largely rural (Simons and Fennig 2018). As is the case for Swedish dialects, there is much grammatical and phonological variation in Norwegian dialects. Swedish and Norwegian can be classified together as North Scandinavian in opposition to Danish, which has approximately 5.6 million speakers and is the national language of Denmark. Danish is a minority language in Germany, where it has official status in Schleswig-Holstein (see Louden, Chapter 34). Danish also displays a lot of regional, ethnic, and social variation though many dialects are moribund or extinct (Gooskens, Chapter 32). The insular group of North Germanic languages comprises Icelandic and Faroese. Icelandic has approximately 321,000 speakers, almost all of whom live in Iceland. Icelandic has relatively little dialectal variation in comparison to the languages in the Scandinavian group. Faroese is spoken on the Faroe Islands by approximately 48,000 speakers. It is an official language on the Faroe Islands, alongside Danish. Faroese and Icelandic are not mutually intelligible (Gooskens, Chapter 32).

I.4 The Organization of This Handbook This handbook is comprised of five subsections that address major themes in the study of Germanic linguistics; namely, (i) phonology, (ii) morphology and agreement systems, (iii) syntax, (iv) semantics and pragmatics, and

Germanic Languages

(v) language contact and nonstandard varieties. Part I (Phonology) begins with Tracy Alan Hall’s review of phonological processes in Germanic language, which provides an overview of synchronic phonological processes in both standard and nonstandard varieties of Germanic languages (Chapter 1). Hall’s review of phonological processes is followed by Marc van Oostendorp’s treatment of Germanic syllable structure (Chapter 2). Laura Catharine Smith provides an overview of the role of foot structure in Germanic in Chapter 3, which sheds light on the role of prosody in shaping the history of the (morpho)phonology of Germanic languages. In connection with Smith’s discussion of the role of the foot in Germanic phonology, Birgit Alber addresses the typology of word stress (i.e., metric parametrization) in Germanic phonology in Chapter 4. Richard Page’s contribution on quantity in Germanic (Chapter 5) focuses on the contrastive quality of both vowels and consonants, with a particular focus on geminate consonants. The phonetics and phonology of Germanic laryngeals is investigated in depth by Joseph Salmons in Chapter 6. Focusing primarily on assimilations and final neutralization, Salmons provides a detailed survey of selected patterns of alternations (in particular, the contrast between sounds written b versus p or v versus f), and presents a pan-Germanic view of these observed differences. This section finishes with two chapters dedicated to suprasegmental elements of Germanic phonology. In Chapter 7, Bjo¨rn Ko¨hnlein explores the role of tone accent in North and West Germanic. In this chapter, Ko¨hnlein concentrates primarily on tone-based oppositions within stressed syllables that are lexically contrastive. In contrast, Mary Grantham O’Brien’s chapter addresses the role of intonation at the postlexical, i.e., sentential, level in Germanic languages (Chapter 8). Part II (Morphology and Agreement Systems) begins with David Fertig’s treatment of verbal inflectional morphology in Germanic (Chapter 9). In this chapter, Fertig provides an overview of the morphosyntactic categories associated with the verb in Germanic and the various inflectional and periphrastic exponents of those categories. Damaris Nu¨bling’s contribution examines the inflectional morphology of nouns (Chapter 10). Nu¨bling engages in a detailed treatment of the various inflectional categories and patterns that persist in synchronic Germanic language nouns, while also incorporating a brief discussion of historical developments. Moving beyond inflectional categories, Geert Booij delivers an overview of the fundamental principles of word formation in Germanic (Chapter 11). Booij’s primary focus in this chapter is directed at the role of compounding and affixation in word formation in Germanic. Sebastian Ku¨rschner’s chapter on grammatical gender in modern Germanic languages (Chapter 12) serves as a detailed summary of the role and variation of grammatical gender as a classifying system in Germanic languages. Tom McFadden discusses the role that morphological case has played in both the development and classification of Germanic languages (Chapter 13). In addition to this overview, McFadden also highlights the significant impact

5

6

B . R I C H A R D PA G E A N D M I C H A E L T. P U T N A M

that morphological case has played in recent developments in linguistic theory. Marjo van Koppen’s treatment of complementizer agreement in contemporary Germanic languages concludes this section (Chapter 14). The initial focus in Part III (Syntax) concerns word order variation among modern Germanic languages. Hubert Haider’s contribution (Chapter 15) addresses the base order serialization of verbs in relation to their nominal object, with a focus on the primary distinction between VO-languages (North Germanic and English) and those that exhibit a base OV-ordering (Afrikaans, Dutch, Frisian, German, and various regional varieties). Sten Vikner discusses the placement of the finite verb in contemporary Germanic languages (Chapter 16). Making use of basic generative notation, Vikner captures the variation present in this language family by appealing to the placement of finite verbs in particular structure positions (i.e., head positions). Susi Wurmbrand and Christos Christopoulos summarize the distribution of infinitival constructions in Germanic languages and the important role they play in shaping linguistic theory (Chapter 17). In this chapter they restrict their treatment to two important phenomena; namely, (i) exceptional case marking (ECM) and (ii) restructuring. In his examination of object shift and object scrambling in Germanic (Chapter 18), Hans Broekhuis reviews evidence that suggests that both phenomena should be given a unified treatment. In Chapter 19, Martin Salzmann provides a detailed overview of filler-gap dependencies (i.e., unbounded dependency constructions) across Germanic languages. Turning to argument structure and the fundamental role of voice alternations, Artemis Alexiadou and Florian Scha¨fer review the syntactic and semantic properties of the verb’s external argument with an empirical focus on Germanic languages (Chapter 20). Vera Lee-Schoenfeld’s treatment of the morphological, syntactic, and semantic conditions that affect the binding relations of reflexives and nonreflexives appears in Chapter 21. In addition to a detailed treatment of the relevant data, Lee-Schoenfeld also ties these findings to controversies in the theoretical literature. Ida Toivonen’s contribution (Chapter 22) addresses the morphological, syntactic, and semantic factors that contribute to particle verb formation (and related structures, such as resultatives and directed motion predicates) in Germanic. The section on syntax concludes with Dorian Roehrs’s overview of the structure of noun (NP) and determiner (DP) phrases in Germanic (Chapter 23). Kristin Melum Eide leads off Part IV (Semantics and Pragmatics) with her contributions on modality (Chapter 24) and tense and aspect (Chapter 25) in contemporary Germanic languages. Joost Zwarts delivers an overview of the semantic properties of prepositions and particles in English, German, and Dutch, with a specific focus on the roles of place and path (Chapter 26). Negative and positive polarity items are the focus of Doris Penka’s contribution (Chapter 27). This section concludes with Caroline Fe´ry’s treatment of the roles of prosody, information structure, and syntax in modern Germanic languages.

Germanic Languages

The final section (Part V) of this handbook concentrates on language contact and nonstandard varieties. Carrie Jackson provides an exhaustive overview of themes related to L2 acquisition in Germanic languages (Chapter 29). Pia Quist and Bente A. Svendsen shift their focus to elements of grammar attested in urban speech styles of Germanic languages in Chapter 30. These two chapters stand to enrich our understanding of classroom-based research and emergent forms in naturalistic environments. Chapters 31 and 32 hone in on the dialectal variation found in West and Northern Germanic languages respectively. William D. Keel discusses the West Germanic dialect continuum in Chapter 31, while Charlotte Gooskens addresses the dialect continuum in Northern Germanic in Chapter 32. The final three contributions of this section provide an overview of contract varieties of Germanic languages. Janne Johannessen and Michael T. Putnam discuss developmental trends observed in heritage Germanic languages throughout the world, with a particular focus on the Midwestern United States (Chapter 33). To complement this previous chapter, Mark Louden’s treatment of Germanic minority languages, whose speakers, in contrast to heritage speakers, continue to pass their L1 minority language on to their children, reviews the sociolinguistic characteristics of these groups in Chapter 34. This section concludes with Paul Roberge’s contribution on Germanic contact languages, i.e., Germanic-based pidgins and creoles, in Chapter 35.

I.5 The Future Study of Germanic Languages As may be surmised by the survey of chapters in this volume, the study of Germanic languages continues to broaden and deepen. For example, the study of heritage Germanic languages is relatively new and has produced empirical data of theoretical significance. Similarly, the field of second language acquisition has seen rapid growth in this century and more than a single chapter could easily be devoted to the topic. In addition, the Germanic languages themselves continue to change and diversify as witnessed by the emergence of World Englishes globally and new urban vernaculars of German, Dutch, Norwegian, Danish, and Swedish in Europe. It is hoped that the contributions in this handbook spur more research, new theoretical insights, and further growth in the investigation of Germanic languages.

References Ammon, U. 1995. Die deutsche Sprache in Deutschland, O¨sterreich und der Schweiz: Das Problem der nationalen Varieta¨ten. Mouton de Gruyter: Berlin and New York.

7

8

B . R I C H A R D PA G E A N D M I C H A E L T. P U T N A M

Bailey, C-J. N. and K. Maroldt 1977. “The French lineage of English.” In J. M. Meisel (ed.), Langues en Contact – Pidgins – Creoles – Languages in Contact. Tu¨bingen: Gunter Narr: 21–53. Fulk, R. D. 2018. A Comparative Grammar of the Early Germanic Languages: Phonology and Inflectional Morphology. Amsterdam and Philadelphia: John Benjamins. Harbert, W. 2007. The Germanic Languages. Cambridge University Press. Kirkpatrick, A. (ed.) 2010. The Routledge Handbook of World Englishes. New York: Taylor & Francis. Ko¨nig, E. and J. van der Auwera 1993. The Germanic Languages. Routledge: London and New York: Routledge. Lutz, A. 2012. “Norse influence on English in the light of general contact linguistics.” In I. Hegedus and A. Fodor (eds.), English Historical Linguistics 2010: Selected Papers from the Sixteenth International Conference on English Historical Linguistics. Amsterdam and Philadelphia: John Benjamins: 15–42. Poussa, P. 1982. “The evolution of Early Standard English: The creolization hypothesis,” Studia Anglica Posnaniensia 14: 69–85. Simons, G. F. and C. D. Fennig 2018. Ethnologue: Languages of the World, 21st edn. Dallas: SIL International. www.ethnologue.com. Stearns, M. 1978. Crimean Gothic: Analysis and Etymology of the Corpus. Saratoga, CA: Anma Libri. Thomason, S. G. 2001. Language Contact: An Introduction. Washington: Georgetown University Press.

Part I

Phonology

Chapter 1 Phonological Processes in Germanic Languages Tracy Alan Hall

1.1 Introduction This chapter provides an overview of synchronic phonological processes in modern Germanic languages, including standard languages as well as nonstandard varieties. Sound changes as well as processes that might have been synchronically active at earlier stages are not taken into consideration. I refrain from discussing overly morphologized patterns – a restriction that is admittedly difficult to achieve because it is not always clear where to draw the line between phonology and morphology.1 Processes involving laryngeal features (e.g., voicing assimilation) are discussed in Salmons (Chapter 6). Since the article is intended to be accessible to readers from a variety of backgrounds, the overview I provide below is not overly technical. I stress description, although I also consider some of the important research questions for many of the processes discussed below. In the remainder of this chapter, I consider assimilation (Section 1.2), dissimilation (Section 1.3), epenthesis (Section 1.4), deletion (Section 1.5), coalescence, vowel reduction, strengthening and weakening (Section 1.6), and other processes (Section 1.7). In Section 1.8, I conclude.

1.2 Assimilation The most common type of assimilation in Germanic languages involves the place dimension among consonants; assimilation of manner features to vocalic or consonantal targets is not as well-attested and appears to be restricted to the spreading of nasality. The typical consonantal target for 1

Examples of once productive historical rules no longer analyzable in phonological terms include umlaut in German and breaking in Frisian.

12

TRACY ALAN HALL

assimilation is coronal (e.g., /n/, /t/), although dorsal targets are also attested (e.g., /x/). Consider first the various types of place assimilation affecting consonants. In all Germanic languages the place of articulation of a nasal consonant is generally predictable before a stop. For example, in German, the three-way nasal contrast between [m], [n], and [ŋ] is neutralized before noncoronal stops: [ŋ] is the only nasal possible before a velar stop, and [m] is likewise the only attested nasal before a labial stop; see (1a). By contrast, before a coronal stop, a nasal can be [m], [n], or [ŋ], as in (1b). In rare words, [m] can surface before [k], as in (1c). (1)

a. [kʀɑŋk] [lɑmpǝ] b. [ɑmt] [lɑnt] [zɪŋt] c. [ɪmkɐ]

‘sick’ ‘lamp’ ‘office’ ‘country’ ‘sings’ ‘bee-keeper’

(cf. [zɪŋə] ‘sing, 1 sg.’)

The noncontrastivity of the place of the nasals in (1a) is often captured by analyzing the place features for that underlying nasal (/n/) as underspecified. By contrast, the nasals are specified for place features before coronal stops in (1b) and before /k/ in the rare word in (1c), in which case the target for Regressive Nasal Place Assimilation (RNPA) is a nasal unspecified for place features. Alternative analyses without underspecification have also been proposed, e.g., in Optimality Theory (De Lacy 2006). The distribution of nasals before stops is essentially the same in Dutch (Booij 1995: 64–65), TPTL (see online reference), and Norwegian (Kristoffersen 2000: 319–327). That the target for RNPA is /n/ derives support from Yiddish, which has phonemic /m n/ and palatal /ɲ/ (transcribed as /n’/ in Jacobs 2005). According to Jacobs (2005: 113–114), /n/ inherits the place features from /k/ thereby creating the allophone [ŋ] in words like [bɑŋk] ‘bank’, but /m/ and /ɲ/ resist RNPA, e.g., za[m]d ‘sand’, [bɑɲkǝ] ‘cupping glass’. The latter example is significant because palatal nasals are usually analyzed as coronal (Hall 1997); hence, among coronal nasals only unmarked alveolar /n/ is the target. This fact derives support from markedness theory, which treats alveolar (/n/) as the least marked place among coronals (Paradis and Prunet 1991). The data in (1) show that the trigger for RNPA is the set of stops. Whether or not /n/ assimilates to the place of other segment types (e.g., fricatives, glides) is not as clear cut. According to Booij (1995: 64), the trigger for RNPA in Dutch is the entire class of consonants; hence, /n/ surfaces as [m] before the labial fricative /f/ and as palatal [ɲ] before the palatal glide [j]. However, the stop trigger is correct for Norwegian, as described by Kristoffersen (2000: 319), who notes that RNPA does not apply before /f/, e.g., the /n/ in inferno surfaces as [n] and not as [ɲ]. For Wiese (1996: 222), the assimilation of /n/ to

Phonological Processes in Germanic Languages

[ɱ] before /f/ in German words like Senf ‘mustard’ is only optional, but he denies that /n/ surfaces as palatal [ɲ] before [c¸] in words like [mɑnc¸] ‘some’. In the languages listed above, RNPA can also apply across word boundaries, although that assimilation is usually described as being dependent on rate or style of speech. For example, the /n/ in the German phrase in Kassel ‘in Kassel’ can surface as [ŋ], and in Yiddish, the /n/ in in bed ‘in bed’ can likewise be realized as [m] in rapid speech (Jacobs 2005: 114). In both languages the preposition surfaces in isolation as [ɪn]. Similar data from Norwegian and Dutch can be found in the sources cited above. See also the data for (West) Frisian in Tiersma (1985: 27) and TPTL, which involves the assimilation of /n/ to the place of a following stop in prefix+stem and compound junctures. Riad (2014: 88–91) motivates RNPA in Swedish on the basis of alternations involving a word-final [n], e.g., in the preposition [ɛn] ‘a’, which surfaces as [m] before bilabials and as [ŋ] before velars. According to Riad, /n/ surfaces as [ɱ] before labiodentals and as [ɲ] before the palatal glide. In German, a syllabic nasal ([n̩ ]) inherits the place features of a preceding stop (in 2a) or nasal (in 2b); see Wiese (1996: 222–224). In these examples, [n̩ ] surfaces as velar after [g], bilabial after [b], and alveolar after [d]. That the target for Progressive Nasal Place Assimilation (PNPA) is a syllabic nasal derives support from examples like the one in (2c) illustrating that nonsyllabic [n] fails to surface as bilabial [m] after a bilabial sound or as [ŋ] after a velar sound. (2)

a. [tʀɑːgŋ̍ ] [geːbm̩ ] [mɑidn̩ ] b. [kɛmm̩ ] [ʀɪŋŋ̍ ] [kɛnn̩ ] c. [hʏp.noː.zə] [hʏm.nə] [ɑk.nə]

‘carry’ ‘give’ ‘avoid’ ‘comb’ ‘wrestle’ ‘know’ ‘hypnosis’ ‘hymn’ ‘acne’

The triggers for PNPA are the set of stops and nasals. The status of fricatives as triggers is not as clear. According to Wiese’s intuitions, the assimilation of [n̩ ] to [ɱ̍ ] after /f/ (in [ʀaufn̩ ] ‘wrangle’) or to [ɲ̍ ] after [c¸] (in [ʀaic¸n̩ ] ‘reach’) is an optional fast-speech realization. Note that there is no assimilation of [n̩ ] to velar (or uvular) after the uvular rhotic [ʀ] in words like [keːʀn̩ ] ‘sweep’. PNPA is also active in Frisian (Tiersma 1985: 25, TPTL) and Yiddish (Jacobs 2005: 113). In contrast to German and Frisian, the set of triggers in Yiddish includes both stops and fricatives, e.g., the alveolar nasal /n/ assimilates to bilabial in [lebm] ‘live’ and labiodental in [gǝtrofɱ] ‘met’).2 2

Jacobs does not specify that the target for PNPA must be a syllabic nasal; syllabicity is not indicated in his phonetic representations.

13

14

TRACY ALAN HALL

Jacobs is also clear that /n/ surfaces as [ŋ] after the dorsal liquid in dialects with that sound, e.g., Central Yiddish [fuːʀŋ] ‘travel’ (cf. the equivalent German example [keːʀn̩ ] ‘sweep’ without assimilation.) The examples illustrated above show that among the nasals, /n/ is the one sound that undergoes both RNPA and PNPA. By contrast, other sonorant consonants (e.g., /l/ and /r/) do not assimilate before or after labials or velars, e.g., German [fɔlk] ‘people’. In many varieties of Central and Southeastern Yiddish (Jacobs 2005: 131–132), /l/ surfaces as secondarily palatalized ([l’] in Jacobs’s symbols) in the context of a front vowel or /k g/, e.g., the /l/ in /klajn/ ‘small’ is realized as [l’]. I interpret this change as an assimilation of the tongue height and backness features of the velar. In addition to [n], [t] and [d] are two sounds susceptible to place assimilation. For example, in German, Kohler (1977: 209) observes that a word-final [t] can optionally surface in fast speech as velar before a velar or labial before a labial, e.g., [hɑtkɑin]~[hɑkkɑin] ‘has no’, [hɑtmɪc¸]~[hɑpmɪc¸] ‘has me’; cf. [hɑt] ‘has’ in isolation. By contrast, [p] and [k] do not change, e.g., [zɑkbɪtə] ‘say please’ (cf. *[zɑpbɪtə]). In a Bavarian German dialect spoken in Northern Italy (Tyroller 2003: 92–95), /t/ inherits the place features of a preceding nasal. That process derives support from [t]~[p]~[k] alternations in the third person singular of verbs, e.g., /heːft/ [heːf-t] ‘lifts’ versus /bɛːrm-t/ [bɛːrmp] ‘warms’ versus /preŋ-t/ [preŋk] ‘brings’. In Dutch the /t/ of the diminutive suffix /-tjə/ likewise surfaces as [p] after [m], [t] after [n], and [k] after [ŋ], e.g., [raːmpjə] ‘window, dim.’, [baːntjə] ‘job, dim.’, [konɪŋkjə] ‘king, dim.’; see Trommelen (1984: 1), TPTL. The realization of alveolar coronals like [t]/[s] as the corresponding postalveolar sounds ([tʃ]/[ʃ]) in the neighborhood of front vowels is well-attested cross-linguistically (Bateman 2007). That type of change is attested in Dutch as an allophonic rule of Palatalization (Booij 1995: 95). For example, the fricative [ʃ] is an allophone of /s/ before the palatal glide [j] in examples like [tɑʃjə] ‘little bag’; cf. [tɑs] ‘bag’. The targets for Palatalization are the alveolar stops (/t d/), fricatives, (/s z/) and nasal (/n/), which all surface as the corresponding allophones before [j]. The process also applies across word boundaries, e.g., [bɛɲjə] ‘are you’ (from /bɛn jə/).3 A common place assimilation from the cross-linguistic perspective involves the fronting of a velar like [k] to a sound like [tʃ] in the neighborhood of front vowels (Bateman 2007). That type of change is not found as a synchronic rule in modern Germanic languages, although it was probably active at an earlier stage in North Germanic.4 However, the fronting of the velar fricative [x] to the palatal [c¸] in the front vowel context (Dorsal Fricative Assimilation [DFA]) is well-attested in standard German and in German dialects. DFA derives motivation from the complementary 3

/j/ can optionally delete ([bɛɲə]). Note that [bɛɲə] is opaque on the surface because the trigger is absent.

4

There are vestiges of that process in modern Scandinavian languages; see Kristoffersen (2000: 112) for Norwegian, Árnason (2011: 101) for Icelandic, and Riad (2014: 109) for Swedish.

Phonological Processes in Germanic Languages

distribution of [x] and [c¸]. The generalization is that [c¸] surfaces after any front vowel or sonorant consonant (in 3a) and [x] after any back vowel (in 3b). The sonorants referred to here are /n l ʀ/. (3)

a. [mɪc¸] [dɔlc¸] b. [bɑx]

‘me’ ‘dagger’ ‘stream’

Much discussion in the literature has concerned itself with the analysis of DFA. According to one approach (e.g., Robinson 2001), /x/ is the target, front vowels are [coronal], and palatals like [c¸] are complex segments that are marked for both [dorsal] and [coronal]. Given those featural assumptions, DFA spreads [coronal] from a front vowel to /x/. That approach also succeeds in deriving the palatal after /l/ and /n/, but the obvious challenge is how to account for the change from /x/ to [c¸] after dorsal /ʀ/. According to an alternative view, /c¸/ is the target, front vowels are [–back], back vowels are [+back], and palatals are [–back]. Given that treatment, the feature [+back] spreads from a back vowel to /x/, thereby creating the palatal. In many German dialects, there are two voiced dorsal fricatives (velar [ɣ] and palatal [ʝ]) whose distribution parallels the examples in (3); see Schirmunski (1962). For example, Moselle Franconian (Hommer 1910) has the same basic facts as in (3), in addition to examples like [fʊɣǝl] ‘bird’ versus [ɪʝǝl] ‘hedgehog’. In other dialects (e.g., Low German, Westphalian, Holthausen 1886), the facts involving [x] and [c¸] are essentially the same as in (3), but [ɣ] surfaces without change as [ɣ] after a back vowel or a front vowel, e.g., [vɑːɣn̩ ] ‘car’, [bʀʏɣə] ‘bridge’. In many Low German dialects, there is a mirror-image distribution of [x] and [c¸] in word-initial position. For example, in the Westphalian dialect referred to above, [c¸] surfaces word-initially before any front vowel (e.g., [c¸ɛɔs] ‘goose’) and [x] before any back vowel or consonant (e.g, [xuət] ‘good’, [xlʏkə] ‘fortune’). In Afrikaans, /x/ and /k/ undergo a similar fronting in initial position before a front vowel (Combrink and de Stadler [1987: 80]), e.g., gieter ‘watering can’ with [c¸]. Assimilations involving the manner dimension (nasality, continuancy) are not as pervasive as place assimilations in Germanic languages. One example involves the optional realization of [g] as [ŋ] in the context before a nasal consonant in German colloquial speech, e.g., [mɑgneːt]~[mɑŋneːt] ‘magnet’, [flɛgmɑ]~[flɛŋmɑ] ‘phlegm’; see Mangold (2005: 67).5 The target for g-Nasalization is /g/ and the trigger is a nasal consonant. It is not clear why /g/ is the only target for g-Nasalization. Thus, an explanatory account needs to account for why the /k/ in [ɑk.nə] from (2c) is never realized as [ŋ]. A similar question is why /d/ or /b/ are not targeted, e.g., [ɔʀdnɐ] ‘binder’.

5

A similar set of examples in Swedish is discussed in Riad (2014: 113–115), although he considers the pattern to be obsolete.

15

16

TRACY ALAN HALL

English is a language in which nasalized vowels occur as allophones of the corresponding oral vowels before nasal consonants, e.g., (4a, 4b) from Kager (1999: 27). The target for Vowel Nasalization (VNas) is any vowel and the trigger is any nasal consonant. (4)

a. [sæ̃ nd] [lɪŋ̃ k] b. [sæd] [lɪk]

‘sand’ ‘link’ ‘sad’ ‘lick’

VNas is also a very general process in Frisian (Tiersma 1985: 15–16, TPTL), e.g., the /in/ in /din/ ‘your’ surfaces as [ĩ] before continuants, e.g., [dĩfits] ‘your bicycle’ (cf. [fits] ‘bicycle’). Jacobs (2005: 97–99) describes data from Yiddish suggesting VNas is active. In general, /n/ triggers the nasalization of a preceding vowel, although Yiddish dialects impose various restrictions on the targets and triggers on VNas. For example, in some Yiddish dialects VNas only applies if the nasal consonant is followed by a fricative. It is often the case that the nasal consonant trigger is not present in the surface, thereby indicating opacity, e.g., [grin] ‘green’ ~ [grĩs] ‘vegetable’. Other dialects resolve fricative plus nasal sequences with the epenthesis of a stop, e.g., [gɑ̃ s]~[gɑnts] ‘entire’ (see section 1.4). According to Schirmunski (1962: 390), a nasal consonant can trigger the nasalization of a preceding vowel in German dialects, especially in Swabian and Middle Bavarian, e.g., [kʰɛn ̃ d] ‘child’. In some Swabian varieties, VNas is triggered by a preceding nasal consonant, e.g., [nɑ̃ s] ‘nose’. See also Noelliste (2017) for a formal account of VNas in a variety of Bavarian spoken in Austria. According to Combrink and de Stadler (1987: 72–75), nasalized vowels in Afrikaans are similarly derived from the corresponding oral vowel and nasal consonant.

1.3 Dissimilation There is agreement in the literature that dissimilations are far less common cross-linguistically than assimilations. Although some historical dissimilations are well-known (e.g., Grassmann’s Law in Sanskrit and Greek), there are very few clear synchronic dissmilations in Germanic languages. In some of the examples discussed below the rules involved are not obviously dissimilatory; instead, the treatment of the rule in question as a dissimilation is a function of how it is analyzed featurally. By definition, dissimilation involves a change whereby the target and trigger become less alike phonologically. In phonology, dissimilations are often analyzed as a repair to an OCP violation, e.g., in a hypothetical language /mb/ would violate a constraint *[labial][labial] and surface as [nb]; see Alderete and Frisch (2007) for discussion.

Phonological Processes in Germanic Languages

Words in Frisian ending in a sequence of two fricatives can optionally undergo a rule of Dissimilation that changes the first fricative into the corresponding stop, as in (5); see Tiersma (1985: 37) and TPTL. (5)

[vaːxs]~ [vaːks]

‘wax’

Dissimilation in Frisian is a general strategy the language employs for avoiding adjacent fricatives. In certain Southern Bavarian (Tyrolian) German dialects spoken in Austria (Schatz 1897) there are regular alternations between [ʀ] and [d], as in (6). Hall (2009) analyzes these alternations with an underlying /ʀ/ that surfaces as [d] before [l]. Since that change is argued to be motivated as a repair to adjacent liquids in the underlying representation, Hall analyzes it as a dissimilation of the feature [liquid]. (6)

a. [tiəʀ] b. [tiədlə]

‘animal’ ‘animal, dim.’

In Standard German the contrast between /s/ and /ʃ/ in intervocalic position is neutralized to [ʃ] in word-initial position before [p t m n l ʀ], but to [s] before [k], as in (7). On the basis of the complementary distribution of [ʃ] and [s] in word-initial preconsonantal position, it is sometimes assumed that both sounds derived from /s/, which surfaces as [ʃ] before all consonants with the exception of /k/; see Wiese (1996) and Alber (2001). If /s/ is [–high] and /ʃ/ is [+high], then s-Dissimilation is a change from [–high] (/s/) to [+high] ([ʃ]), but only before [–high] consonants. (7)

a. [ʃtɑt] b. [skɑt]

‘city’ ‘card game’

Note that [–high] subsumes labial (/p m/), alveolar (/t n l/), and uvular (/ʀ/), but not velar (/k/), which is [+high]. Given that featural treatment, s-Dissimilation is a dissimilation of the feature [high]. Changes involving consonants and vowels in German dialects are analyzed as dissimilations by Glover (2014) and Noelliste (2017). For example, in Kiel German, a high vowel lowers before a high vocoid [ɪ̯ ], which itself derives from /ʀ/, e.g., the /ʊ/ in /dʊʀx/ ‘through’ surfaces as [ɔ] in [dɔɪ̯ c¸]. Glover argues that Vowel Lowering is a dissimilation involving the change from [+high] (/ʊ/) to [–high] ([ɔ]) before [+high] ([ɪ̯ ]). In a Bavarian German dialect spoken in Austria, front vowels undergo a retraction before a front vocoid, which derives from /l/; see Noelliste (2017). For example, the /i/ in /ʃpil-n/ ‘play, inf.’ surfaces as [u] before [ɪ̯ ], i.e., [ʃpuɪ̯ n]. Noelliste analyzes Vowel Retraction as a dissimilation of the frontness feature ([coronal] in her treatment).

17

18

TRACY ALAN HALL

1.4 Epenthesis Epenthetic segments in Germanic languages can be vowels or consonants. If a vowel is inserted, then it is typically schwa ([ə]). The attested epenthetic consonants can be glottal ([ʔ]) or (alveolar) coronal (e.g., [d]). The typical function of either type of epenthesis is to repair an ill-formed syllable. I consider first the vowels and then the consonants. German has many morphemes with an alternation between schwa ([ə]) and zero, as in (8a). The schwa in morphemes like [hɪməl] is the product of epenthesis (Noske 1993, Wiese 1996), e.g., /hɪml/ [hɪməl]. Schwa Epenthesis is triggered by a sonorant consonant that cannot be syllabified for reasons of sonority; e.g., /ml/ in /hɪml/ cannot be parsed into the coda because liquids (/l/) are more sonorous than nasal (/n/). By contrast, schwa cannot be inserted between /m/ and /l/ in /hɪml-ɪʃ/ because those two sounds are parsed into the coda and onset respectively. Schwa Epenthesis also applies in words that do not have a schwaless alternant, as in (8b), the argument being that schwa must be present to make the underlying string (/hɑːkn/) pronounceable. That treatment derives support from the fact that syllables ending in such sequences without schwa are not even pronounceable, e.g., *[.hɑːkn.]. Schwa is phonemic in other contexts, e.g., word-finally in [kɑtsə] ‘cat’ and before a sonorant in words like [ɑləm] ‘all’ (cf. [ɑlm] ‘mountain pasture’).6 (8)

a. [hɪməl] [hɪmlɪʃ] b. [hɑːkən]

‘heaven’ ‘heavenly’ ‘hook’

An optional realization of words like [hɪməl] and [hɑːkən] with a syllabic sonorant is also possible, e.g., [hɪml̩ ]. That pronunciation is reflected in the phonetic transcriptions in (2) above for nasals. Pronunciation with a syllabic sonorant can be accounted for with a process converting [ə] plus nonsyllabic sonorant into a syllabic sonorant, e.g., /hɑːkn/ → hɑːkən → hɑːkn̩ → [hɑːkŋ̍ ]. In Icelandic there are regular [ur]~[r] alternations requiring a rule of u-Epenthesis, which has a motivation akin to German Schwa Epenthesis, e.g., [dagur] (/dag-r/) ‘day, nom sg’ versus [bær] (/bæ-r/) ‘farm, nom sg’. Following several earlier treatments, Kiparsky (1984) argues that the inflectional ending is /r/ and that u-Epenthesis applies before that consonant if it cannot be syllabified, i.e., after a consonant in /dag-r/ but not after a vowel in /bæ-r/. An interesting question not pursued here is why [u] and not some other vowel is epenthesized. In contrast to German, Icelandic has no schwa. An alternation between [n] and [ən] (e.g., in the plural suffix) occurs in German examples like [nɑdəln] ‘needles’ versus [fʀauən] ‘women’. The 6

In other Germanic languages examples like the ones in (8) are usually analyzed with syllabic sonorants and not with schwa plus nonsyllabic sonorant, e.g., Borowsky (1990) for English and Kristoffersen (2000) for Norwegian.

Phonological Processes in Germanic Languages

generalization is that words of a particular morphological category (e.g., noun plurals) must end in a trochaic foot (Wiese 2009). Given that prosodic requirement, [ən] must occur after a monosyllabic noun like [fʀau] ‘woman’, but not after a disyllabic trochee like [̍ nɑdəl] ‘needle’. Note too that since the schwa in the latter word is not present underlyingly (/nɑdl/), Schwa Epenthesis must apply before the correct plural ending can be appended. The epenthesis of a vowel as a strategy for avoiding an illegal sequence of underlying consonants can also be observed in English, e.g., after a sibilant final stem the regular plural suffix surfaces as [ɪz] in churches, kisses, houses, but as [s] after a voiceless nonsibilant (e.g., cats) and as [z] after a voiced nonsibilant (e.g., dogs). If the plural morpheme is underlyingly /-z/, then ɪ-Epenthesis only applies after a sibilant stem because no English syllable can end in a sequence of two sibilants.7 A similar analysis accounts for alternations between [ɪd], [d], and [t] in the past tense suffix, e.g., rented, learned, leased. German has schwa-zero alternations in inflectional suffixes motivating a rule of epenthesis similar in function to English ɪ-Epenthesis. One generalization is that schwa is present after a stem ending in a coronal stop like /t/ and before a verbal suffix beginning with a coronal obstruent (i.e., /-t/ or /-st/). Compare the presence of schwa in the suffixes -et and -est in (9a) and its absence in (9b). The motivation for epenthesis of schwa in an example like /ʀɛt-t/ is to avoid geminates, but that generalization cannot be extended to an example like /ʀɛt-st/ because the underlying consonant cluster surfaces in other examples, e.g., in inflected nouns in (9c).8 (9)

a. [ʀɛtət] [ʀɛtəst] b. [leːpt] [leːpst] c. [moːnaːts]

‘save, 3 sg.’ ‘save, 2 sg.’ ‘live, 3 sg.’ ‘live, 2 sg.’ ‘month, gen. sg.’

Clearly the epenthesis of schwa in (9a) has a condition on the consonants in the context and on lexical category (Giegerich [1989]). In Dutch (Trommelen [1989: 114], Booij [1995: 127–128], TPTL), schwa is inserted between a liquid and a final noncoronal (in 10a). Crucially, no schwa is epenthesized if the liquid and noncoronal are word-internal, e.g., [bal.kan] ‘Balkans’. A (coronal) consonant can follow the noncoronal (in 10b). Schwa is also inserted between /r/ and /n/ (in 10c), but not between a liquid and a coronal obstruent (in 10d) or between a nasal and a (homorganic) stop (in 10e). In contrast to the epentheses discussed above in German and English, epenthesis in (10a–10c) is optional. 7

The English examples have alternatively been analyzed with a suffix containing a vowel (/-ɪz/), which undergoes a rule of vowel deletion (syncope) after a stem ending in a nonsibilant; see Borowsky (1990).

8

Among inflected verbs [tst] is a possible sequence, e.g., [tʀɪtst] ‘kick, 2 sg’. The generalization is that the epenthesis of schwa as in (9a) is blocked if the stem vowel is nonidentical to the stem vowel of the infinitive, cf. [tʀeːtən] ‘kick, inf.’.

19

20

TRACY ALAN HALL

Epenthetic schwa cannot be the result of deletion because there are underlying (obligatory) schwas that would incorrectly delete, e.g., the in the proper name Willem. The implication is that the schwas in (10a–10c) cannot be underlying schwas that optionally undergo deletion, otherwise underlying (obligatory) schwas would also incorrectly be elided. (10)

a. [ɔlǝm] [ɑrǝm] [kɑlǝf] [fɔrǝk] b. [mɑrǝkt] c. [horǝn] d. [hɛlt] e. [lɑmp]

‘elm’ ‘arm’ ‘calf’ ‘fork’ ‘market’ ‘horn’ ‘hero’ ‘lamp’

Any treatment of Dutch needs to account for why the process applies in the contexts described above. For various views on this topic, the reader is referred to Trommelen (1984), Kager (1989), Booij (1995), and van Oostendorp (2000). The pattern of epenthesis in (10) is not restricted to Dutch. As noted by Schirmunski (1962: 401), that type of insertion can be observed to a certain degree in all High German dialects. Some of his examples (from Southern Franconian) include [dorǝf] ‘village’ (cf. Standard German [dɔʀf]), and [ɑrǝm] ‘arm’ (cf. Standard German [ɑrm]). Data similar to the ones in (10) obtain in Central Yiddish. Jacobs (2005: 121) notes that epenthesis eliminates final clusters of /rC/, where the C is a velar obstruent, e.g., /bɑrg/ ‘mountain’ surfaces as [bɑrǝk], but there is no epenthesis if the C is followed by a vowel, e.g., /berg-ǝ/ ‘mountains’ is realized as [bergǝ] and not as [berǝgǝ]. Visser (1997: 263–267) documents a pattern of word-final epenthesis in Frisian that is essentially the same as the one described for Dutch, e.g., [skɔlǝm] ‘link’, [arǝk] ‘tools’, [kɔrǝps] ‘corps’ versus [vilt] ‘wild’; see TPTL. The epenthesis of a vowel in a word-initial onset is rare in Germanic languages. Visser (1997: 258–263) discusses the optional epenthesis of schwa in Frisian between word-initial obstruent plus liquid clusters, as in (11a); see TPTL. The example in (11b) shows that a sequence of two initial obstruents resists epenthesis. (11)

a. [bliːt]~[bǝliːt] [slɪm]~[sǝlɪm] b. [spo¨l] (cf. *[sǝpo¨l])

‘glad’ ‘bad’ ‘gear, things; quarrel, jar’

In a Highest Alemannic variety of German spoken in Switzerland, there is a pattern of vowel epenthesis in word-initial position (Wipf 1910, Hall 2011). In contrast to the examples discussed above, the epenthetic vowel is [a] (which is nondistinct from phonemic /a/) and not [ǝ]. The generalization is that (unstressed) [a] is epenthesized in word-initial position

Phonological Processes in Germanic Languages

before the consonant [r], but only after a word ending in a stressed vowel. The word-initial [a] in (12) is the product of a-Prothesis; Wipf (1910) shows that there are no words beginning with [r] without a vowel. The reason schwa is not the epenthetic vowel is that the dialect in question has no schwa; however, an explanatory account must clarify why [a] and not some other phonemic vowel is epenthesized. (12)

a. [arad]~[rad] b. [ladu]

‘wheel’ ‘store’

Hall (2011) concludes that a-Prothesis only applies before [r] and not before other consonants because it has the function of repairing highly sonorous consonantal onsets. The process is inhibited after an unstressed vowel to avoid a surface hiatus sequence of unstressed vowels. In German the glottal stop ([ʔ]) predictably occurs before a word beginning with a vowel, as in (13a). Since there is no contrast between a word beginning with a glottal stop and one without a glottal stop (e.g., [ʔɑus] ‘out’ versus a nonoccurring word like [ɑus]), underlying representations do not include [ʔ], which is inserted by a rule of ʔ-Epenthesis. Epenthetic [ʔ] surfaces before any vowel, regardless of whether or not it is stressed or unstressed. In wordinternal position, ʔ-Epenthesis applies between two vowels, but only if the second one is stressed; see (13b). A similar set of data from Dutch is discussed in Booij (1995: 65–66) and from Frisian in Visser (1997: 45). (13)

a. [ˈʔɑlǝ] [ʔiˈdeː] b. [ˈkɑːɔs] [kɑˈʔoːtɪʃ]

‘everyone’ ‘idea’ ‘chaos’ ‘chaotic’

The challenge posed by (13) is how to state the prosodic context for ʔEpenthesis in a non-ad hoc way. It has been proposed that the rule applies at the left edge of a vowel-initial foot (e.g., Wiese 1996). Since feet in German are trochaic, the parsings assumed for the words in (13b) are F (kɑː.ɔs) and (kɑ F (ʔoː.tɪʃ)). The problem with the foot-based analysis is that a word like [ʔiˈdeː] in (13a) would have to be analyzed as a single foot (i.e., F (ʔiˈdeː)) even though the word exemplifies iambic stress. See Alber (2001) for discussion and an Optimality Theoretic treatment. According to Riad (2014: 275–276), Swedish has an optional rule epenthesizing the glottal stop before a vowel-initial word, e.g., [2ˈaŋːka]~ [2ˈʔaŋːka] ‘duck’. In contrast to German, there is no [ʔ] in intervocalic position even if the second vowel is stressed. The restriction of epenthetic [ʔ] to word-initial position before a vowel is also the regular pattern in the southern German dialects discussed by Alber (2001), e.g., Standard German [ʔo.ˈʔɑːzə] ‘oasis’ versus southern German [ʔo.ˈɑːzə] and supports the claim in cross-linguistic work that the place of an epenthetic consonant in wordinitial position before a vowel is glottal (Uffmann 2007).

21

22

TRACY ALAN HALL

In Dutch, Frisian, and Yiddish a rule of d-Epenthesis is active. In Dutch, d-Epenthesis derives motivation in the basic regular alternations, e.g., between the suffix [ər] and [dər], as in (14a). In these examples (Booij 1995: 73), the suffix forms the comparative of adjectives (e.g., 14b), but the same alternation is visible in the other functions for the same suffix. d-Epenthesis can be stated in phonological terms as the insertion of [d] after /r/ and before / ər/; the process is motivated as a strategy of avoiding the surface sequence [rVr], which is a sequence that tends to be avoided cross-linguistically. See Tiersma (1985: 36) and TPTL for similar data from Frisian, although d-Epenthesis in that language applies after /r l n/, e.g., brún~brúnder ‘brown~browner’. (14)

a. [rod] [rodər] b. [zyr] [zyːrdər]

‘red’ ‘redder’ ‘sour’ ‘sourer’

According to Jacobs (2005: 127) the diminutive in Yiddish is formed by appending a syllabic lateral to a noun ending in a consonant, as in (15a). If the base noun ends in /n/, [d] is epenthesized, as in (15b). (15)

a. [kop] [kepl̩ ] b. [bejn] [bejndl̩ ]

‘headline’ ‘headline, dim.’ ‘leg’ ‘leg, dim.’

Jacobs writes that d-Epenthesis is motivated as a strategy to eliminate sequences of [nl], although he notes that there are surface [nl] sequences in words like [pajn.ləx] ‘painful’ (from /pajn-ləx/), which are accounted for by treating d-Epenthesis as a cyclic rule. The epenthesis of a nonsyllabic sound (consonant or glide) often involves the assimilation of that sound to one of the segments in the context. Consider the following examples: One way English avoids hiatus ([V.V]) is by epenthesizing a glide if the first vowel is high (Uffmann 2007). The epenthetic glide is homorganic with the first vowel, e.g., [j] in key is ([kiːjɪz]), cf. [kiː] in isolation and [w] in zoo is ([zuːwɪz]), cf. [zuː] in isolation.9 Homorganic Glide Epenthesis (HGE) is also attested in Dutch (Booij 1995: 66–67), e.g., [dijet] (/diet/) ‘diet’, and Faroese (Taylor 1974: 69–86, A´rnason 2011: 81–82), e.g., [siːja] (/siːa/) ‘to say’. In those three languages HGE is not 9

In the context after a nonhigh vowel and before a vowel hiatus is tolerated in many dialects; however, in other varieties, [r] is attested as a hiatus breaker. Thus, a word ending in a nonhigh vowel like idea surfaces without [r] in isolation or before a word beginning with a consonant, but [r] is present if that word occurs before a vowel, e.g., in the phrase the idea is; see Uffmann (2007). English r-Epenthesis is a textbook case of a productive phonological process. r-Epenthesis is sometimes claimed to be active in German dialects (e.g., Bavarian) as well as in many non-Germanic languages, but a closer inspection of the original sources reveals that this claim is not correct and that the rule is English-specific (Hall 2013).

Phonological Processes in Germanic Languages

restricted to applying after high front vowels. For example, in Dutch, [j] is epenthesized after /i/ and /e/, front rounded [ɥ] after /y/ and /ø/ and back [ʋ] after /u/ and /o/. In Faroese [j] is inserted if the first vowel is high and front and unrounded and [w]/[ʋ] if it is high and rounded. After a nonhigh vowel the glide is determined by the following vowel: [j] before /i/ and the labiodental glide before /u/, e.g., [kleːjɪ] ‘gladness’, [leːʋʊ] ‘the (state) of lying down, a place to lie-dat.’. In many languages a stop can be inserted between a nasal and an obstruent that is homorganic with the nasal. For example, the /ms/ sequence in a word like Chomsky can be pronounced in English as [ms] or [mps]. See Clements (1987), who argued that Intrusive Stop Formation (ISF) in English is phonological and not phonetic. In the Yiddish dialects referred to in Section 1.2, a stop is optionally inserted after a homorganic nasal and before a fricative, e.g., [gɑ̃ s]~[gɑnts]. Booij (1995: 137–138) documents the same process in Dutch; see (16). As illustrated below, /kam-t/ and /zɪŋ-t/ are realized with intrusive stops, making them homophonous with ‘fight, 3 sg.’ and ‘sink, 3 sg.’. See Jacobs (2005: 128) for similar data from Yiddish and TPTL for Frisian. (16)

a. [kɑmpt] [kɑmpt] b. [zɪŋkt] [zɪŋkt]

/kam-t/ /kanp-t/ /zɪŋ-t/ /zɪnk-t/

‘comb, 3 sg.’ ‘fight, 3 sg.’ ‘sing, 3 sg.’ ‘sink, 3 sg.’

ISF is not a simple case of epenthesis because the segment that is inserted shares the place of articulation of the nasal. Hence, ISF also involves the spreading of the place features from the nasal to the epenthetic stop. The epenthesis of the stop [g] in Swedish is discussed at length in Riad (2014: 91–94). The generalization is that the velar nasal can only surface in coda position (e.g., [tɔŋː] ‘seaweed’). In intervocalic position, the velar nasal is in both coda position and in onset position, but only if the first vowel is stressed (e.g., [1ˈfɪŋːɛr] ‘finger’). If the velar nasal is situated between an unstressed vowel and a stressed vowel, then [g] is inserted by g-Excrescence. The function of g-Excrescence is to avoid a syllable-initial (or foot-initial) velar nasal. A few morphemes illustrating an alternation between [g] and zero are attested, as in (17a), but the bulk of the data show a static pattern of g-Excrescence, as in (17b). (17)

a. [dɪfˈtɔŋː] [dɪfˈtɔŋːɛr] [dɪfˈtɔŋːˈgeːra] b. [hɑŋˈgɑr]

‘diphthong’ ‘diphthongs’ ‘to diphthongize’ ‘hangar’

(cf. [ˈslaŋːar] ‘tubes’)

Like ISF in (16), g-Excrescence does not insert the velar stop [g]. Instead, the rule epenthesizes a stop which acquires its velar place of assimilation from the preceding sound (i.e., [ŋ]).

23

24

TRACY ALAN HALL

1.5 Deletion Deletion is a common strategy for eliminating sequences of consonants and/or vowels that arise via morpheme concatenation which would violate structural requirements if they were to surface without change. That type of deletion is typically obligatory. By contrast, the deletion of vowels and/ or consonants in fast and casual speech may be motivated as a way of reducing articulatory effort. Both obligatory and optional (fast speech) deletion are attested in Germanic languages. The typical vowel that undergoes deletion is schwa and the typical consonant is alveolar coronal, e.g., /t/, /n/, /r/. In Dutch, schwa is obligatorily deleted before a vowel (Booij 1995). Thus, if a word ending in schwa occurs before a vowel-initial suffix, then the schwa is elided, as in (18). Prevocalic Schwa Deletion only applies if the schwa plus vowel sequence belongs to the same prosodic word; hence, there is no deletion of /ǝ/ before the suffix -achtig, e.g., [zɛidǝɑxtǝx] ‘silky’, which forms an independent prosodic word. Prevocalic Schwa Deletion is one strategy Dutch employs to avoid hiatus. (18)

a. /romǝ-ɛin/ b. /zɛidǝ-ǝɣ/

[romɛin] [zɛidǝx]

‘Roman’ ‘silky’

Examples like the ones in (18) are common in German as well, e.g., the schwa in [zɑidǝ] ‘silk’ does not surface before the adjectival suffix /-ɪg/, cf. [zɑidɪc¸] ‘silky’. A´rnason (2011: 254–255) documents a similar deletion, I dub Prevocalic Vowel Deletion, which is active in both Icelandic and Faroese. The vowel undergoing deletion in those languages is any stemfinal full vowel in the context before a vowel-initial suffix, e.g., in Icelandic /ɪ/ surfaces before /r/ in [heiːtɪr̥ ] ‘is named’ (from /heiːtɪ-r/), but it deletes before the vowel-initial suffix /-um/ in [heiːtʏm] ‘we are named’. Vowel-zero alternations in Icelandic morphology as in (19) motivate the deletion of a vowel in a different context than the one described above. According to A´rnason (2011: 253–254), the second vowel in a disyllabic stem deletes before a vowel-initial inflectional or derivational suffix. For example, the second /a/ deletes in /kaman/ before /-ɪ/ in (19a), but it is retained in (19b) because no vowel follows. Additional examples in the original source reveal that there are no significant restrictions on the type of vowel that deletes. (19)

a. [kamnɪ] b. [kaːman] [kaːmans]

‘fun, dat. sg.’ ‘fun, nom. sg.’ ‘fun, gen. sg.’

If a Dutch word has two consecutive syllables headed by schwa, the first of those schwas is eliminated by Schwa Deletion, but only if the resulting onset cluster consist of an obstruent and a liquid. Two

Phonological Processes in Germanic Languages

examples (Booij [1995: 128–130]) are provided in (20a), where the underlined vowel is the schwa undergoing deletion. Schwa Deletion is an optional process dependent on rate/style of speech; it creates onset clusters that are not allowed by lexical syllabification, e.g., [tl] in (20b). The process does not apply if the resulting onset cluster is anything other than obstruent plus liquid; see (20c). Example (20d) shows that Schwa Deletion is only triggered by a following schwa and not by a following full vowel. (20)

a. [ˈsuplə] [ˈwɑndlən] b. [ˈkitlən] c. [ˈtekənən] [ˈrɑmələn] d. [bəˈlovən]

soupele wandelen kietelen tekenen rammelen beloven

‘smooth’ ‘to walk’ ‘to tickle’ ‘to draw’ ‘to rattle’ ‘to promise’

(*[teknən]) (*[rɑmlən]) (*[blovən])

Diacritics for stress are included, which are absent in Booij’s transcriptions. Schwa Deletion has the function of creating trochaic feet from underlyingly trisyllabic words. When sequences of adjacent consonants occur in underlying representations which cannot surface as such for phonotactic reasons, one of those consonants can be deleted. For example, neither German, Dutch nor Frisian allows for geminate consonants within a prosodic word. If a word ends in a consonant and a suffix is added to that word that is identical to the stem-final consonant, then Degemination applies. Examples from the verbal morphology of German and Dutch are illustrated in (21a) and (21b) respectively; see Booij (1995: 68–69) and TPTL. Note that Dutch employs deletion of /t/ after /t/ as a repair, but German epenthesizes a schwa in that context (recall 8a). Frisian is similar to Dutch, e.g., /sɪt-t/ ‘sit, 3 sg.’ is realized as [sɪt]; Tiersma (1985: 31). (21)

a. /kʏs-ə/ /kʏs-st/ b. /et-ən/ /et-t/

[kʏsə] [kʏst] [etən] [et]

‘kiss, 1 sg.’ ‘kiss, 2 sg.’ ‘eat, inf.’ ‘eat, 3 sg.’

Degemination operates optionally across prosodic word boundaries, e.g., across compound boundaries (cf. /hand-duk/ → [hantuk] ‘towel’). That type of degemination is also attested in Yiddish (Jacobs [2005: 130–131]), e.g., /on-nemən/ ‘accept’ surfaces as [onemən]. A similar set of data for Afrikaans are discussed in Combrink and Stadler (1987: 89–90). In Dutch, /t/ is prone to deletion (Booij 1995: 152–154). t-Deletion applies obligatorily after an obstruent and before the glide [j], thereby accounting for the [jə] allomorph of the diminutive suffix (see 22a); however, /t/ is retained after a sonorant (see 22b) even if [j] follows. The deletion of /t/ occurs as an optional, fast-speech process, but in that case the rule can also

25

26

TRACY ALAN HALL

apply after a sonorant, but the following consonant must be an obstruent, e.g., [tɑntsten]~ [tɑnsten] ‘tartar’. Similar data obtain in Frisian (Tiersma 1985: 31). (22)

a. [ɑpt]~[ɑpjə] b. [hɑrt]~[hɑrtjə]

‘abbot~abbot, dim.’ ‘heart~heart, dim.’

The context for fast-speech t-Deletion in Dutch is very similar to the context for the optional deletion of a stem-final stop or nasal in Norwegian (Kristoffersen 2000: 109–110): That type of sound is elided after a consonant and before a suffix beginning with a consonant, e.g., ‘to think’ surfaces as [2teŋk.tə] or [2tɛŋ.tə]. An optional deletion of /d/ is similarly attested in Swedish in the context after a nasal and before an obstruent (Riad 2014: 95–98), e.g., the /d/ in /land/ in the compound [ˈlanːsˌvæːɡ] ‘country road’. Interestingly, /t/ is retained in the same context, e.g., [2ˈpanːtˌsɛtːa] ‘to pawn’. A similar deletion in Afrikaans is discussed in Combrink and de Stadler (1987: 84–95) motivating the optional deletion of /d/ after a sonorant consonant and before a vowel, e.g., the /d/ in onder ‘amid’. Those authors note that the vowel following /d/ must be unstressed to account for its retention in words like soldaat ‘soldier’, in which the final syllable is stressed. In Dutch a word-final /n/ can optionally be deleted after schwa (Booij 1995: 139–141, TPTL). n-Deletion applies in all lexical categories, e.g., in singular nouns (in 23a), plural nouns (in 23b) or plural verbs (in 23c). However, the rule affects neither the /n/ at the end of a verbal stem (in 23d) nor the /n/ in the indefinite article [ǝn] ‘a’. (23)

a. b. c. d.

[molǝn]~ [molǝ] [blumǝn]~ [blumǝ] [lopǝn]~ [lopǝ] [tekǝn]~ *[tekǝ]

‘mill’ ‘flowers’ ‘walk’ ‘draw’

[r]~Ø alternations in Frisian (Tiersma 1985: 32) motivate r-Deletion, e.g., [jɛrǝ] ‘hear, inf.’, [jɛr] ‘hear, 1 sg.’ versus [jɛst] ‘hear, 2 sg.’. That process applies regularly before /t d n l s z/. However, if /r/ is final in a prefix or at the right edge of a compound, /r/ deletes before any consonant except /h/, e.g., [vɛkomǝ] ‘come again’ (from /vɛr-komǝ/). In Norwegian, /r/ optionally deletes depending on factors such as rate of speech if it is the final segment in a stressed syllable and the following segment is not coronal (Kristoffersen 2007: 311–315). Consider the following: (24)

a. [dǝ.bli.1ʈuː.2kɾuː.nɾ̩ ] b. [dǝ.bli.1fɛm.2kɾuː.nɾ̩ ]

‘that will be two kroner’ ‘that will be five kroner’

In (24), /r/ in /blir/ is elided before the noncoronal /f/ in (24b), but it is retained and undergoes the Retroflex Rule (see Section 1.6) in (24a).

Phonological Processes in Germanic Languages

1.6 Coalescence, Vowel Reduction, Strengthening, and Weakening Coalescence is the fusion of two adjacent sounds into one. Although not common in Germanic languages, that type of process is exemplified by the Retroflex Rule (RetR) in Norwegian (Kristoffersen 2000: 96–100, 315–319). RetR fuses a heteromorphemic rhotic (alveolar /r/ or retroflex /ɽ/) plus alveolar coronal (/t d s l n/) into a single apical coronal (retroflex) segment, as in (25). Kristoffersen treats RetR as the spreading of the feature [apical] from the rhotic to the following coronal. The remaining features of the rhotic (including the root node) are subsequently deleted. RetR affects coronals in inflectional and derivational suffixes (in 25a, 25b), but it also applies across word boundaries, as in compounds (see 25c). See Hamann (2003) for an alternative analysis. (25)

a. /sʉr-t/ b. /ʋor-li/ c. /ʋor-dag/

[1sʉːʈ] [2ʋoːɭi] [2ʋoːɖɑːg]

‘sour, infl.’ ‘spring+like’ ‘spring day’

Riad (2014: 73–86) discusses very similar data from Swedish motivating RetR, which he also analyzes as a coalescence. Vowel reduction is the neutralization of underlying vowel qualities in unstressed position (Crosswhite 2001). English is a language which suspends the contrast between most underlying full vowels in unstressed syllable to schwa; hence, there are many morphologically related word pairs such as Canada (with the full vowel [æ] in the initial stressed syllable) versus Canadian (with [ǝ] in the initial unstressed syllable). Vowel Reduction derives functional motivation if schwa requires less articulatory effort than a full vowel. The same process is attested in Dutch (Kager 1989: 297ff., Booij 1995: 130–137, De Lacy 2006), and Frisian (TPTL). In general, Dutch nonhigh vowels reduce to schwa in an unstressed syllable, as in (26a, 26b).10 (26)

a. [bəˈnan] b. [ˈdoməˌne] c. *[əˈrotis] *[həˈman]

‘banana’ ‘parson’ ‘erotic’ (cf. [eˈrotis]) ‘human’ (cf. [huˈman])

There are several contexts in which Vowel Reduction is blocked from applying, e.g., word-initial position or after a word-initial [h], as in (26c) because no prosodic word can begin with schwa or [h] plus schwa. Strengthening (fortition) and weakening (lenition) are two general labels for processes that alter the sonority of segments (see Lavoie 10

Evidence for underlying representations with full vowels (e.g., /banan/ ‘banana’) comes from a more formal register in which Vowel Reduction does not occur.

27

28

TRACY ALAN HALL

2001 for a cross-linguistic survey). In general, a fortition occurs when a sound becomes less sonorous, while a lenition involves the reverse. Both strengthenings and weakenings have occurred as historical processes in Germanic languages (Holsinger 2000), and many of those diachronic developments left regular alternations requiring synchronic rules. A typical weakening involves the change from a liquid to a vocoid (vowel or glide). Liquids are susceptible to that change in many Germanic languages, especially in coda position. For example, in German there are regular alternations between the consonant [ʀ] and [ɐ] which are captured with an underlying /ʀ/ that surfaces as [ɐ] in coda position (Hall 1993, Wiese 1996, Mangold 2005). r-Vocalization is obligatory after a long vowel (see 27a) and optional after a short vowel (see 27b). (27)

a. [tyːɐ] [tyː.ʀən] b. [ɪʀt]~[ɪɐt] [ɪ.ʀən]

Tu¨r Tu¨ren irrt irren

‘door’ ‘doors’ ‘is mistaken’ ‘to be mistaken’

The other liquid (/l/) does not vocalize, e.g., [fiːl] ‘much’. However, /l/ does weaken in many German dialects. It is often observed that /l/ vocalizes in some dialects, e.g., Bavarian, to [i] and in other dialects, e.g., Alemannic, to [u]; see Haas (1983), e.g., [fiu] /fil/ ‘much’ in (Alemannic) varieties of Swiss German (Christen 2001) versus [fɔɪ] /fɔl/ ‘full’ in Middle Bavarian varieties in Austria (Noelliste 2017). Other kinds of weakenings are attested in Germanic languages as well. Two segments that are particularly susceptible are [t] and [d], and the typical context is in the onset of an unstressed syllable (Holsinger 2000). A wellknown example from American English is the weakening of /d/ and /t/ to the flap [ɾ], e.g., eat [iːt] versus eating [ˈiː.ɾɪŋ] (Kahn 1976). A similar process is attested in Swedish, where /d/ is weakened to [r] between vowels (Riad 2014: 99–102). Riad notes that the change occurs in mostly unstressed pronouns, e.g., /vɑd do do/ ‘what then’ → [vɑˈdoːdo]~[vɑˈdoːro]. In Dutch, there are alternations between [t] (from /d/) and the palatal glide [j], motivating a lenition of /d/ to [j], e.g., [rot]~[ˈrojə] ‘red~red, attr.’ (/rod/), although Booij (1995: 90) notes that these alternations are lexically governed because there are also examples like /hud-ən/ ‘hats’ that surfaces as [ˈhudən] but not as [ˈhujən]. A common lenition in Germanic languages involves the change from stop to fricative. In German there are regular alternations between [g] and [c¸], as in (28). That type of example is captured with an underlying /g/ that shifts to the corresponding fricative ([ɣ]) after the vowel /ɪ/ in coda position. The derived fricative [ɣ] undergoes Final Devoicing (see Salmons, this volume) and DFA, surfacing as [c¸]. In the standard language, g-Spirantization only applies in coda position after /ɪ/, but in many regional dialects, there is

Phonological Processes in Germanic Languages

a larger set of triggers. For example, in Kiel German (Glover [2011]), g-Spirantization applies in coda position after all vowels and /ʀ/, e.g., [fluːx] ~[flyːgə] ‘flight~flights’ (from /fluːg/). (28)

a. [køːnɪc¸] ‘king’ b. [køːnɪgə] ‘kings’

The historical change from [x] to [h] in onset position has resulted in regular alternations between those two sounds in many German dialects. In some of those dialects (e.g., Southern Bavarian, Schatz [1897]), [x]~[h] alternations as in (29a) can only be captured with an underlying /h/ that undergoes Buccalization to [x] in the coda (Hall 2010). (29)

a. [lɑi.hə] [liːx] b. [ʀɑix] [ʀɑi.xə]

‘lend, inf.’ ‘lent’ ‘rich’ ‘rich, infl. adj.’

The reason morphemes like ‘lend’ require /h/ and Buccalization to [x] in coda position (and not /x/ with rule of Debuccalization to [h] in the onset) is that there are also morphemes like ‘rich’ in (29b) with a nonalternating [x] (from /x/) that also occurs in onset position.

1.7 Other Processes Some synchronic processes cannot be classified into one of the categories listed above. Examples involve the change from diphthong to monophthong (Monophthongization) and the reverse (Diphthongization). Both are well-attested diachronic developments in the history of Germanic. It is difficult to motivate either of those sound changes as synchronic rules, although some scholars present data motivating some type of Diphthongization as a synchronic process. Noelliste (2017) observes that in Middle Bavarian the four tense monophthongs /i e u o/ optionally surface as either monophthongs or as the diphthongs [iə eə uʊ oʊ], e.g., /vi/ ‘how’ → [vi]~[viə]. Her analysis accounts for that variation with a process of Diphthongization and a separate change accounting for the rounding of the second part of the derived diphthong after a rounded vowel. One cannot treat these data with a rule of epenthesis because it is not motivated as a repair to an illformed syllable (recall Section 1.4). Several Germanic languages are attested with changes altering the quality of vowels – changes that cannot easily be classified in terms of phonological processes. For example, in Norwegian there is a near complementary distribution between [eː ɛ] and [æː æ] (see 30) that motivate a process lowering the former to the latter (Kristoffersen 2000:105–109). Riad (2014: 84–85) observes that

29

30

TRACY ALAN HALL

Swedish has a similar process of lowering, but only in the context before retroflex sounds. (30)

a. [1ʋæj] [1hæwk] [1ʋæʈ]

‘road’ ‘hawk’ ‘host’

1 b. [ ʋɛg] [1ʋɛt]

‘wall’ ‘intelligence’

The difficulty regarding E-Lowering in (30) is that it applies before apical consonants (e.g., /ʈ/) or glides /j w/, two contexts that cannot be collapsed into a single natural class. Kristoffersen argues that there are two separate rules of E-Lowering, i.e., one before glides and the other before apicals. Hamann (2003) provides a phonetically grounded account for E-Lowering before apicals, but the motivation for E-Lowering before glides is unclear according to any treatment.

References Alber, B. 2001. “Regional variation and edges: Glottal stop epenthesis and dissimilation in standard and Southern varieties of German,” Zeitschrift fu¨r Sprachwissenschaft 20: 3–41. Alderete, J. D. and S. A. Frisch 2007. “Dissimilation in grammar and the lexicon.” In P. de Lacy (ed.), Cambridge Handbook of Phonology. Cambridge University Press: 379–398. A´rnason, K. 2011. The Phonology of Icelandic and Faroese. Oxford University Press. Bateman, N. 2007. A Crosslinguistic Investigation of Palatalization. Ph.D. dissertation: University of California San Diego. Booij, G. 1995. The Phonology of Dutch. Oxford University Press. Borowsky, T. 1990. Topics in the Lexical Phonology of English. New York: Garland. Christen, H. 2001. “Ein Dialektmarker auf Erfolgskurs: Die /L/-Vokalisierung in der deutschsprachigen Schweiz,” Zeitschrift fu¨r Dialektologie und Linguistik 68.1: 16–26. Clements, G. N. 1987. Phonological feature representation and the description of intrusive stops. Parasession on Autosegmental and Metrical Phonology. Chicago: Chicago Linguistic Society 29–50. Combrick, J. G. H. and L. G. de Stadler 1987. Afrikaanse Fonologie. Johannesburg: Oxford University Press. Crosswhite, K. 2001. Vowel Reduction in Optimality Theory. New York: Routledge. De Lacy, P. 2006. Markedness. Reduction and Preservation in Phonology. Cambridge University Press.

Phonological Processes in Germanic Languages

Giegerich, H. 1989. Syllable Structure and Lexical Derivation in German. Bloomington: Indiana University Linguistics Club Publications. Glover, J. 2011. “G-spirantization and lateral ambivalence in Northern German dialects,” Journal of Germanic Linguistics 23: 183–193. Glover, J. 2014. Liquid Vocalization and Underspecification in German Dialects. Ph.D. dissertation, Indiana University at Bloomington. Haas, W. 1983. “Vokalisierung in den deutschen Dialekten.” In W. Besch et al. (eds.), Dialektologie. Ein Handbuch zur deutschen und allgemeinen Dialektforschung, Vol. II. Berlin: Walter de Gruyter: 1111–1116. Hall, T. A. 1993. “The phonology of German /ʀ/,” Phonology 10.1: 83–105. Hall, T. A. 1997. The Phonology of Coronals. Amsterdam: John Benjamins. Hall, T. A. 2009. “Liquid dissimilation in Bavarian German,” Journal of Germanic Linguistics 21.1: 1–36. Hall, T. A. 2010. “On the status of [h]~[x] alternations in German dialects: The case for buccalization.” In S. Fuchs, P, Hoole, C. Mooshammer, and M. Zygis (eds.), Between the Regular and the Particular in Speech and Language. Berlin: Peter Lang Verlag: 29–56. Hall, T. A. 2011. “Vowel prothesis in Walliser German,” Linguistics 49.5: 945–976. Hall, T. A. 2013. “How common is r-epenthesis?,” Folia Linguistica 47.1: 55–87. Hamann, S. 2003. The Phonetics and Phonology of Retroflexes. Doctoral dissertation, University of Utrecht. Holsinger, D. 2000. Lenition in Germanic: Prosodic templates in sound change. Ph.D. dissertation, University of Wisconsin-Madison. Holthausen, F. 1886. Die Soester Mundart. Norden und Leipzig: Diedrich Soltau’s Verlag. Hommer, E. 1910. Studien zur Dialektgeographie des Westerwaldes. Marburg: R. Friedrich’s Universita¨ts-Buchdruckerei. Jacobs, N. 2005. Yiddish. A Linguistic Introduction. Cambridge University Press. Kager, R. 1989. A Metrical Theory of Stress and Destressing in English and Dutch. Dordrecht: Foris. Kager, R. 1999. Optimality Theory. Cambridge University Press. Kahn, D. 1976. Syllable-Based Generalizations in English Phonology. Ph.D. Dissertation, MIT. Kiparsky, P. 1984. “On the lexical phonology of Icelandic.” In C. Elert et al. (eds.), Nordic Prosody, Vol III. Stockholm: Almkvist and Wiksell: 135–164. Kristoffersen, G. 2000. The Phonology of Norwegian. Oxford University Press. Lavoie, L. 2001. Consonant Strength. Phonological Patterns and Phonetic Manifestations. New York: Garland. Mangold, M. 2005. Duden Aussprachewo¨rterbuch: Wo¨rterbuch der deutschen Standardaussprache. Duden, Band 6 5. Auflage. Noelliste, E. 2017. The Phonology of Sonorants in Bavarian German. Ph.D. dissertation, Indiana University.

31

32

TRACY ALAN HALL

Noske, R. 1993. A Theory of Syllabification and Segmental Alternation. With Studies on the Phonology of French, German, Tonkawa and Yawelmani. Tu¨bingen: Max Niemeyer Verlag. Oostendorp, M. van 2000. Phonological Projection. A Theory of Content and Prosodic Structure. Berlin: de Gruyter. Paradis, C. and J.-F. Prunet (eds.) 1991. The Special Status of Coronals: Internal and External Evidence. New York: Academic Press. Riad, T. 2014. The Phonology of Swedish. Oxford University Press. Robinson, O. 2001. Whose German? The ach / ich Alternation and Related Phenomena in “Standard” and “Colloquial”. Amsterdam: John Benjamins. Schatz, J. 1897. Die Mundart von Imst. Laut- und Flexionslehre. Strassburg: Tru¨bner. Schirmunski, V. M. 1962. Deutsche Mundartkunde. Vergleichende Laut- und Formenlehre der deutschen Mundarten. Berlin: Akademie-Verlag. Taylor, J. E. 1974. A Generative Phonology of Faroese Utilizing Unordered Rules. Ph.D. dissertation, Indiana University. Tiersma, P. M. 1985. Frisian Reference Grammar. Dordrecht: Foris. Trommelen, M. 1984. The Syllable in Dutch: With Special Reference to Diminutive Formation. Dordrecht: Foris. Trommelen, M. 1989. Lettergreepstructuur en woordcategorie. De Nieuwe Tallgids 82, 64–77. Tyroller, H. 2003. Grammatische Beschreibung des Zimbrischen von Lusern. Wiesbaden: Franz Steiner Verlag. Uffmann, C. 2007. “Intrusive [r] and optimal epenthetic consonants,” Language Sciences 29: 451–476. Visser, W. 1997. The Syllable in Frisian. Amsterdam Holland Institute of Generative Linguistics. Wiese, R. 1996. The Phonology of German. Oxford: Clarendon Press. Wiese, R. 2009. “The grammar and typology of plural noun inflection in varieties of German,” Journal of Comparative Germanic Linguistics 12: 137–173. Wipf, E. 1910. Die Mundart von Visperterminen im Wallis. Frauenfeld: Verlag von Huber and Co.

Online Reference TPTL: The Linguistics of Dutch, Frisian and Afrikaans Online (Taalportaal .org).

Chapter 2 Germanic Syllable Structure Marc van Oostendorp

2.1 Introduction Germanic syllables generally can have a rather complex syllable structure, allowing consonant clusters of sometimes considerable length, and also maintaining a contrast on vowels that is by some scholars considered as length (see also Page, Chapter 5). Germanic syllables are furthermore rather similar in the kinds of syllables they allow. This is sometimes obscured by the fact that the literatures on the individual languages tend to be separate, which means that very similar conclusions have been reached independently about different languages. It is of course regrettable that the wheel has to be reinvented separately several times, but at the same time, it may be an indication that there is something right about the idea of a wheel. In order to organize this chapter, I choose the following template – without this implying any claim about the “real” structure of the syllable in any individual Germanic language, or all Germanic languages considered together. (1)

O (s)(C1) (C2)

N

C

A

V

(C3) (C4)

(C*)

s here is a sibilant (/s/ or in some languages also /ʃ/), C1 and C2 are (the other) consonants of the onset (O). The vowel forms the nucleus (V), and C3 and C4 the coda (C). ‘A’ is an appendix (usually only found at the end of the word), containing usually voiceless coronal consonants. Generally the nuclear vowel is the only obligatory part of the syllable. This is indicated by the fact that it is the only part that is not bracketed. I use this template in this chapter as a descriptive tool, not because I would necessarily support any theoretical claim one can derive from it. The literature I use here has employed a variety of different representational

34

MARC VAN OOSTENDORP

means to describe phonotactics, but all Germanic languages seem to fit into this template. The only point of contention may be whether Scandinavian languages have an appendix; also not all positions can be filled in all languages equally easily. I also use (1) as a way to organize this chapter: I will discuss the onset in Section 2.2 and the coda and the appendix in Section 2.3, before turning to what happens at the boundaries between syllables in polysyllabic words in Section 2.4. The word is an important domain for syllabification in all Germanic languages. Syllabification rarely crosses word-boundaries, except in some cases of cliticization, and for this reason I have taken the word also as the domain of study in this chapter. My main focus will be on those modern Germanic languages that have a standardized form – Afrikaans, Danish, Dutch, English, Faroese, Frisian, German, Icelandic, Norwegian, Swedish – mainly because sources on those languages are more readily available, but occasionally I will also consider minority languages (like Yiddish), dialects, and older Germanic languages (like Gothic).*

2.2 Germanic Syllable Onsets 2.2.1 Onsetless Syllables and Simple Onsets Onsets – prevocalic consonant clusters – in Germanic contain zero to three consonants, at least at the beginning of the word. Onsetless syllables are always somehow marked. They do not often occur within the word, or more specifically in the middle of a foot. If they do occur in that position, we find various types of hiatus resolution, such as gliding, the insertion of a glide, or the insertion of some other segment, such as ‘linking’ [r], [l], [n], or [h]. The latter kind of process seems interestingly more widespread in nonstandardized languages than in standardized forms, with the possible exception of r insertion in British English. It is also typically linked to a deletion process of the same segment in the same language: (2)

Gliding /tiara/ > [tjara] ‘tiara’ (German, Hall 1992, Hamann 2003)

(3)

Glide insertion /dialyse/ > [dijalyːsə] ‘dialysis’ (Norwegian, Kristoffersen 2000: 139) /syanid/ > [syʋaniːd] ‘cyanide’

(4)

Linking segments [n] /wo i/ > [wo n i] ‘where I’ (Alemannic, Nu¨bling, and Schrambke 2004)

* Edoardo Cavirani, Björn Köhnlein, and editors and reviewers of this Handbook have provided useful comments on an earlier draft.

Germanic Syllable Structure

[r] /sofa is/ > [sofa r is] (English ‘non-rhotic’ varieties, Wells 1982) [l] /bra is/ > [bra l is] (Southern Pennsylvania English, Gick 2002) [h]/[ɦ] /idio:m/ > [idiho:m] ‘idiom’ (Afrikaans varieties, Den Besten 2012) Other hiatus avoidance strategies such as vowel deletion – with the exception of schwa deletion – or vowel merger are not well-attested in Germanic, and in particular not as regular processes. In word-initial position a segment is sometimes inserted, such as a glottal stop (e.g., Standard German) or an [h] (e.g., South African English, Lass 2002). This seems to happen always noncontrastively, i.e., these segments are only inserted in those languages in which glottal stop or [h] is not phonemic. But also in other Germanic languages in which we find these two segments, their occurrence is usually restricted to simplex onsets; they occur neither in coda position nor in more complex onsets. I am aware of very few restrictions on monosegmental onsets in Germanic. Except for the velar nasal, any legitimate consonant can occur in such a position in all Germanic languages. Some discussion would be possible on voiced and voiceless fricatives, and in particular sibilants, as some languages have a preference for one or the other in such a position, and historical changes may be a factor (English has [s]ea, German has [z]ee ‘lake’); however, this is mostly an issue of preference and not an absolute requirement in the synchronic state of the language (English has [z]any, German has [s]ent ‘cent’ so that synchronically voiced and voiceless sibilants are allowed in both languages; Van Oostendorp 2003, Fuchs et al. 2007). A final note concerns the occurrence of word-initial geminates. Although several Germanic languages have been analyzed as involving a quantity distinction on consonants (see also Page, Chapter 5), as far as I know only Swiss German has been claimed to also use geminate consonants in onsets: (5)

[pː]aar ‘pair’ – [p]aar ‘bar’ (Thurgovian Swiss German, Kraehenmann 2001) [tː]ankx ‘tank’ – [t]ankx ‘thank’ [kː]aar ‘coach’ – [k]aar ‘cooked’

The length distinction here replaces the voicing distinction we find elsewhere, as becomes clear when we compare these words to their cognates in other Germanic languages (like the English glosses for the first two words).

2.2.2 Complex Onsets It is interesting to try one’s hand at setting up a plausible typology of wordinitial biconsonantal clusters (disregarding for a moment the clusters starting with s, which will be discussed in Section 2.2.3). English is the least

35

36

MARC VAN OOSTENDORP

permissive of all Germanic languages (except maybe some Germanic-based creoles) in the range of complex onsets that are allowed (basically only obstruent-liquid clusters), and Southern German and Yiddish the most permissive. Like for non-Germanic languages, sonority is an important notion to understand the structure of complex onsets in Germanic languages. There is a general tendency to follow sonority sequencing restrictions, although there are some interesting differences as to how strictly these are implemented. We will use a rather simplex sonority hierarchy here, although more detailed ones have been suggested for individual languages (Parker 2011 gives an overview): (6)

Sonority hierarchy plosives < fricatives < nasals < liquids < glides < vowels

The sonority sequencing generalization on onsets holds that: (7)

Sonority Sequencing Generalization (SSG, see Parker 2011) Within an onset, consonant sequences should not display falling sonority.

Onsets in Germanic generally obey the SSG. In some languages, the restrictions on clusters might be stricter than this, for instance not allowing clusters a sonority plateau with a sonority distance that is too small (e.g., the two segments having the same sonority). All Germanic languages allow obstruent plus liquid clusters, such as [pl, kl, bl, gl, fl, vl, pr, kr, tr]. The clusters [tl] and [dl] are usually excluded from the list of possibilities, although some languages have exceptional words (e.g., Yiddish tliə ‘gallows’). The avoidance of these clusters is usually seen as an indication of the Obligatory Contour Principle (OCP, e.g., Kager and Shatzman 2007): [t], [d] and [l] are all specified as coronal and we cannot have onset clusters with consonants of the same place of articulation. This would entail that [r], which can freely co-occur with [t] or [d] is placeless, or at least not coronal (e.g., Kristoffersen 2000: 51 on Norwegian). Other kinds of clusters tend to be more restricted. [kn] is the only obstruent-nasal cluster with a nasal that seems quite widespread in both North- and West-Germanic. For example, Swedish has words like knekt ‘knight’ or knapp ‘scarce’. English is an exception, as etymological kn clusters have obviously been simplified to [n] (witness knee and gnome). Clusters with pn are rare also elsewhere (and typically are Greek loanwords, like Swedish pneumatisk); again such clusters are simplified in English. /tn/ may again be excluded completely for reasons of OCP. Clusters where the second nasal consonant is anything other than n (like m) are excluded everywhere. Certain clusters seem to have disappeared in most West Germanic varieties. This is true for clusters starting with h. Vennemann (1988: 46) gives the following historical developments for German:

Germanic Syllable Structure

(8)

Early OHG Late OHG NHG hnigan nigan neigen ‘to bow’ hlut lut laut ‘loud’ hruofan ruofan rufen ‘to call’ hwiz wiz weiss ‘white’ OHG = Old High German, NHG = New High German

As far as I am aware, other West Germanic languages have undergone the same development; only hw clusters have remained unchanged in some dialects of English. Germanic languages differ as to whether or not they allow clusters with a glide as a second element. They are considered more ‘normal’ in the literature on Northern Germanic; the examples which are given are usually derived from Old Norse. For instance, Basbøll (2005: 177) mentions mjød ‘the viking’s sweet beer’ and Njord (name). [nj] sequences are typical for names, and for this reason Vestergaard (1968) did not consider them, but Basbøll points out that such sequences are easily pronounced by Danish speakers, and he asserts that they therefore do belong to the Danish system. With an obstruent as its first member, j is definitely part of a complex onset in Scandinavian languages, although there might be some restrictions. For example, A´rnason (2011: 175) mentions the following forms for Faroese: (9)

pja´tra ‘to speak gently’ bjarga ‘to save’ fjall ‘mountain’ spjaldur ‘board, plaque’ stjo´rn ‘governing body, board’ mjo´lk ‘milk’ njo´ta ‘to enjoy’ rjo´ta ‘to snore’ ljo´tur ‘ugly’

(Faroese)

Most of these cases (with the exception of stj, nj, rj, and lj) have a labial consonant for reasons that need further inquiry. In Icelandic we also find occasional three-positional onsets with j, such as fljo´ta ‘to float’ or grjo´t ‘rocks’. I suppose that one could raise the issue in those cases whether j is not part of the nucleus rather than the onset, and whether sequences like lj or pj are not really monosegmental, which could explain why velars and coronal plosives seem to be absent (they would be fully palatalized) and why we have no clusters, but I know of no literature taking up this question. In West Germanic, such sequences are really considered exceptional and/or typical of loanwords and names (such as Piotr). English has sequences such as [kj]ute, com[pj]ute, and [tj]une, but such clusters are different from “ordinary” complex onsets in a number of ways, for instance in being rather unstable across dialects: American English

37

38

MARC VAN OOSTENDORP

generally does not have [tj]une, most dialects do not have [blj]ew or [glj]ue. Welsh English does, which means it has rather unusually long clusters, while East Anglian English does not permit any of these clusters (Szigetva´ri 2016). Even if we take this cross-dialectal instability into account, it is surprising that English, which otherwise seems to be the least permissive of all Germanic languages in terms of what are allowed onset clusters, appears to be most flexible here. Frisian also has wordinitial clusters like [pj]isk ‘peach’ or [kj]uw ‘gill’, but here the initial glide is most profitably seen as part of a complex nucleus (Visser 1997: 189). The other glide (be it /ʋ/ or /w/) seems to occur more happily in clusters in those languages that have it, at least after velar and coronal plosives (Dutch kwaad ‘angry’, twee ‘two’, dwaas ‘foolish’). After labials or after velar fricatives, [w] is not found, probably again for reasons of OCP; exceptions are sometimes loanwords from Romance (like pueblo). Southern German dialects also allow for obstruent-obstruent clusters. The following examples are from Tyrolean (Alber and Meneguzzo 2016: 33–34): (10)

kfollen ‘fallen’ (past participle) (Tyrolean German) pʃeid ‘news’ kxluan ‘small’

Yiddish is similarly permissive, and may be even more so in certain respects since it also allows for sonorant-sonorant clusters (Jacobs 2005: 123–124). (11)

tfilə ‘prayer’ (Yiddish) ptirə ‘prayer’ mloxəm ‘kings’

The historical origin of such forms is diverse. The Tyrolean examples (10) result from a process of vowel deletion within the cluster (the form for ‘fallen’ for instance has as its Standard German cognate gefallen, with a schwa between g and f, and ge). The Yiddish examples show that these kinds of clusters can also be found in loanwords (in this case, from Hebrew). Loanwords also play a role in the phonotactics of other Germanic languages. These will have (typically Greek-derived) words starting with pt-, ps-, ks-, or mn-. In each of these cases the second consonant is a coronal. The most important exception is again English, which simplified these clusters when it borrowed the same loanwords as the other languages. The Southern German and the Yiddish clusters are not completely unrestricted; for instance, we do not find examples of words starting with rising sonority clusters (rt), such as we find in Czech (Ra´cz 2010).

2.2.3 sClusters Germanic languages all have clusters consisting of a sibilant followed by an obstruent violating sonority restrictions on phonotactics, because the

Germanic Syllable Structure

fricative can be followed by a stop, as well as violating size restrictions, because s initial clusters can be three-consonantal clusters. Languages vary as to which sibilants can occupy the s position. Like the Scandinavian languages, some languages – Gothic, Dutch, Afrikaans, and Frisian – only allow for [s] to occur there. Others such as German and Yiddish allow for more variation. German (Wiese 1996, Alber 2007) and Yiddish (Jacobs 2005) have both [s]C and [ʃ]C clusters that are possibly contrastive. In German (12a), the common options are [ʃp, ʃt] and [sk], but [sp, st] and [ʃk] are also attested, and according to Wiese there is no strong tendency to assimilate those latter clusters: (12)

a. Spiel [ʃpil] ‘game’, Spezies [spezies] ‘species’ (German) b. svive ‘environment’, ʃvarc ‘black’ (Yiddish)

According to Hammond (1999: 35), for English the following distributional statement holds: (13)

a. [s] cannot precede [r]; b. [ʃ] can only precede an [r].

(14)

a. shrink, *srink (English) b. spy, star, screen, *shpy, *shtar, *shcreen. (English)

The statement in (13a) is virtually exceptionless for English and also seems to hold true of the other Germanic languages. It has sometimes been suggested that words such as schrijven ‘to write’ start with [sr] in Dutch (Kooij and Van Oostendorp 2004), but if that is the case, this cluster is indistinguishable from [sxr]. (13b) has exceptions in German and Yiddish loanwords in Dutch (schlemiel, shtick). One could therefore also say that all West Germanic languages have both [ʃc] and [sc]. In Yiddish, the sibilant seems furthermore to be subject to voicing assimilation to the following obstruent or (labial) nasal (Jacobs 2005): (15)

a. skarbovə ‘trite’, zgulə ‘remedy’, stazˇ ‘seniority’ (Yiddish) b. zman ‘semester’, zˇmenə ‘handful’

Yiddish is exceptional in this respect; in other Germanic languages, voiced obstruents are simply not allowed in such clusters. It is sometimes claimed that sC clusters are actually not clusters but single segments (see Goad 2011 for a cross-linguistic overview). This would allow us to assume that initial clusters are always maximally bisegmental, but it also would solve some other problems. For example, Visser (1997:109) points out that Frisian – like many other Germanic languages – has a process of degemination within e.g., compounds, and that this affects also sC clusters: (16)

list+stik > li[st]ik ‘piece of a frame’ fisk+skaal > fi[sk]aal ‘fish dish’

(Frisian)

39

40

MARC VAN OOSTENDORP

In Gothic, sC clusters reduplicate (e.g., in the preterite) as a whole when C is an obstruent, but only partly when C is a sonorant. This can be taken as an indication that the former cases are monosegmental (Voyles 1980, Keydana 2011). (17)

slepan ‘to sleep’ – saislep ‘to sleep – pret.’ (Gothic) staldan ‘possess’ – staistald ‘possess – pret.’

In some creole languages based on Germanic languages, sp clusters have been reduced completely, typically to the nonsibilant (Wells 1982). On the other hand, there are restrictions on the kinds of cluster that can follow s that may be more difficult to account for from the point of view of a monosegmental analysis. Kristoffersen (1999) posits constraints to rule out *skn and *stv in Norwegian, even though sk, st, kn, and tv are all possible. Jacobs (200) remarks that there are no triconsonantal voiced clusters (*zbr) in Yiddish. Such restrictions seem to point to these clusters being particularly marked.

2.3 Codas and Appendices 2.3.1 Monosegmental Codas Codas tend to be crosslinguistically more restrictive than onsets. As we have seen, monosegmental onsets can be filled by virtually any consonant, but this is not the case for codas. In the first place, many Germanic languages (and all of their dialects) have a process of final devoicing, so that voiced obstruents are excluded from this position in those languages (see Salmons, Chapter 6). Furthermore, as I indicated above, the ‘placeless’ segments [ʔ] and [h] are also avoided in this position. Some special attention should be paid to the velar nasal, of which the occurrence is also restricted. The segment is historically derived from an [ng] (or [ŋg]) cluster, and this cluster-like behavior still holds in the Germanic languages. In some dialects (like Northern German) it still surfaces as such: (18)

Di[ŋk] – Di[ŋg]e ‘thing- things’(Northern German, cf. Standard German Di[ŋ])

In most other languages, the cluster has been simplified to [ŋ], but its behavior still is that of a cluster, as it cannot occur after diphthongs or long vowels (see also Section 2.4.2), or in the onset: (19)

*ŋa, *baːŋ, *beiŋ, *baːŋ, baŋ ‘afraid’, bein ‘leg’(Standard German)

The ‘cluster’ origin of the velar nasal cannot be the only reason for this behavior, however. In some varieties, [ŋ] demonstrably has a different origin, but still shows the same behavior. For example, in Cologne German,

Germanic Syllable Structure

coronals in coda have been historically velarized. But where the preceding vowel is a diphthong (or a long vowel) it is shortened: (20)

braun > bruŋ ‘brown’(Cologne German, Scheer 1999)

The phonotactic constraint that [ŋ] does not occur after a long vowel is thus maintained, even though there is no reason to assume that this nasal corresponds to a cluster at any level of analysis.

2.3.2 Biconsonantal Codas All Germanic languages allow complex codas at least at the end of the word, although there generally is a restriction on the preceding vowels: clusters do not occur after long / tense / fortis vowels (see also Page, Chapter 5). Within the word, clusters tend to be more restricted, and unstressed syllables in particular usually allow for simpler codas only. The sonority scale is again an important tool for at least organizing our understanding of these clusters. Like all languages allow for obstruentliquid clusters in the onset, they also all allow for liquid-obstruent clusters in the coda. As a matter of fact, fewer restrictions seem to apply, in the sense that [lt] (and /ld/) clusters are unproblematic as codas, even if their mirror images are not as onsets. These clusters are sometimes broken up. Some varieties of Yiddish (Central Yiddish and West Yiddish, Jacobs 2005, Herzog et al. 1992) break up clusters of r followed by another consonant, but not when the cluster is followed by a vowel. One way to see this is of course that only codas are broken up this way (and when a vowel follows, the obstruent is put into the onset of the following syllable): (21)

ʃtark ‘strong’ > [ʃtarək] ʃtarkə ‘strong’ (PL) *[ʃtarəkə]

(Yiddish, Herzog et al. 1992: 134)

Dutch and Frisian have similar patterns of optional epenthesis also after /l/: (22)

skalm ‘link’ [skɔl(ə)m] (Frisian, Visser 1997: 76)

Epenthesis is prevented before a coronal consonant, so that Frisian held ‘hero’ is not pronounced as *[helət]. Similar kinds of observations can be made on sonorant-sonorant clusters: provided they have a following sonority profile (for instance, the first is a liquid and the second a nasal). For instance, Norwegian has kvalm ‘nauseated’. In Yiddish, Frisian, and Dutch dialects, such clusters can again be broken by a schwa. All Germanic dialects also allow for (some) nasal+obstruent clusters. Such clusters are never broken by a schwa. In monomorphemic forms, the nasal and the obstruent are always homorganic, and maybe that blocks the epenthesis of schwa: lamb and land are words of Icelandic, but *lanb or *lamd are not (A´rnason 2011: 166). If a morpheme boundary appears between the two consonants, more complexity is allowed so that we find

41

42

MARC VAN OOSTENDORP

lengd [leiŋt] ‘length’ (from langur ‘long’) or kı´mdi [chimtɪ] ‘smiled’ (from kima ‘smile’). In both cases, the obstruent is coronal, and that seems typical for this kind of structure. There seems a general tendency to avoid such clusters if the second consonant is a fricative or is voiced. I know of no examples in any variety of Germanic in which it can be a voiced fricative. Clusters with a voiceless fricative can be found (English triumph, dance and its cognates in other languages) although usually only in loanwords from Romance, while clusters with a voiced obstruent show some unexpected behavior. I have already indicated above that /ŋg/ has simplified to /ŋ/ in many languages; the same has often happened to historical /mb/ (as in English lamb). /nd/ clusters are an exception in English (land), but in most dialects of Norwegian, for example, they have also simplified (to [n]). Chapman (1962) reports that there are only a few ‘archaic’ dialects that have retained the [d]. There can be some discussion as to whether the Germanic languages have coda sequences of rising sonority (i.e., starting with an obstruent or a nasal, followed by a liquid). On the surface they generally do not allow for such sequences, but it is sometimes argued that underlyingly they do have them. For example, Rice (2002, 2004) points out that in Norwegian imperatives can occur before vowel-initial, but not before consonant-initial words: (23)

a. Sykl opp bakken Bike up the hill! (imperative) b. *Sykl ned bakken Bike down the hill

(Norwegian, Rice 2002)

We can draw two kinds of conclusions on these facts. On the one hand, we can conclude that the imperative form for ‘bike!’ in Norwegian ends in a rising sonority cluster (sykl), but at the same time, we can also see that such clusters only show up when they can be syllabified in an onset. Within the nominal domain in Norwegian, furthermore, rising sonority clusters can show up with an epenthetic schwa (pepər) or a syllabic consonant (pepr) (Kristoffersen 1999). Similar arguments can be made for other Germanic languages. For example, the Dutch noun filter [fɪltər] shows up without the schwa in the derived verb filtreer, leading Booij to suggest that the schwa is not underlying, but epenthesized. This epenthesis process should be distinguished from the one in sonorant-consonant clusters (such as kerk > kerrək), because the latter is optional whereas the former is obligatory (van Oostendorp 2000). In German, we find an interesting difference between consonant-liquid clusters and consonant-nasal clusters (Noske 1992). One seems to trigger schwa epenthesis within the cluster, the other at its edge: (24)

er zittert ([tsItəʀt – tsItʁt]) ‘he trembles’ er atmet ([atmət]) ‘he breathes’

(German)

Germanic Syllable Structure

Noske (1992) proposes that this is because liquids are lexically syllabic in these positions, but nasals are not (see also Issatschenko 1974). Notice that in German nouns, on the other hand, schwa also appears before the nasal; the relevant noun in this case is Atem ‘breath’. The issue is even more complicated, as Wiese (1999) points out, as /r/ and /l/ also behave differently in another paradigm, viz. before adjectival inflection: (25)

trocken+e ‘dry’ (inflected) sicher+e ‘certain’ (inflected) dunkl+e ‘dark’ (inflected)

(German)

In some way, different kinds of word-final sonorant consonants clearly interact differently with different kinds of morphological structures. Sometimes (for instance in English, Norwegian, or, for German, in [ʁ]) the final sonorant is realized as syllabic. In no case I know are rising sonority clusters ever realized as complex codas, even if we have reason to assume they are so ‘underlyingly’. Complex codas are often only allowed at the end of the word, or at least their distribution is much more limited within the word. For some authors (e.g., Kristoffersen 1999, van Oostendorp 2000) this has been a reason to assume that codas are really only monosegmental. The second consonant is then analysed as an onset of an otherwise empty syllable, which is only permitted word-finally. This idea also converges with analyses of vowel quantity: Typically we find only monosegmental codas after long (or tense) vowels. The reason could be that rhymes are maximally bisegmental in Germanic, the long / tense vowels occupy the rhyme on their own, so that there is only place for the following “onset”. This kind of analysis has to explain why the onsets of final empty nuclei are maximally monosegmental.

2.3.3 Appendices Word-final voiceless coronal obstruents behave slightly differently from other segments in all Germanic languages. First, these consonants can be the source of clusters that are much longer than bisegmental codas. Most often we find these in morphologically complex forms: (26)

a. b. c. d. e.

helped, waltzed, parks, boxed (English) glaubst ‘believe’ (2S), la¨chelnd ‘smiling’ (German) erft ‘inherits’ (3S), denkt ‘thinks’ (3S) (Dutch) hests ‘horse’ (gen.) (Icelandic) manst ‘you remember’ (Faroese)

But we find the same complexity occasionally also in words without a discernible morphological structure (e.g., English text, German Markt ‘market’, Dutch herfst ‘autumn’). Kristoffersen (1999) denies the existence

43

44

MARC VAN OOSTENDORP

of a coronal appendix in Norwegian. According to this author, we find only complex codas, which of course can have a coronal second consonant. If we analyze such sC clusters furthermore as monosegmental, we can also accommodate words like Norwegian kunst ‘art’ (as well as the Faroese example in (26e)). This seems typical for mainland Scandinavian languages; other kinds of words mentioned in (26) can not be accommodated, however, so that apparently other Germanic languages do have more complex clusters. In some loanwords, such as extra, we also find an s between [k] and [tr] as an extratemplatic consonant. In many of these languages there is also a tendency for simplification. Varieties of English and Dutch display tendencies to delete final coronals (Goeman 1999; Fabricius 2002; Guy 1980, 1991a, 1991b; Hinskens 1992; Kiparsky 1988; Labov 1969). The process is often variable, depending on many different sociolinguistic and phonological factors: (27)

gewirkt / gewirk ‘worked’ walked / walk

(Limburg Dutch, Hinskens 1992) (English varieties, Guy 1980)

In Afrikaans the process has affected many stems historically, resulting in alternations (Conradie 1981, 1982, Hinskens 2009): (28)

lig ‘light’ – ligte ‘lights’ hoof ‘head’ – hoofde ‘heads’

(Afrikaans, Conradie 1981)

The plural suffix is arguably -e, and the t (or d) only surfaces in front of it. In some cases, the t also shows up in the plural in cases where it is etymologically not warranted: (29)

graf ‘grave’ – grafte ‘graves’

(Afrikaans, Conradie 1981)

Voiceless coronal obstruents thus seem to play a role outside of the ordinary template; they can be added almost freely, especially at the end of the word, and they can also be more freely deleted than other consonants.

2.4 Syllable Boundaries and Syllable Contact The boundaries between syllables, in particular within a foot, are not always easy to discover. In particular in English, speakers seem to disagree on the precise position of syllable boundaries. In a famous experiment, Cutler et al. (1983) found that French speakers found it more difficult to find the sequence bal in balance than in balcon and inversely more difficult to find ba in balcon than in balance. English speakers basically did not make that difference, because syllable boundaries seemed less important to them. It is not entirely clear what the theoretical interpretation of this old fact should be.

Germanic Syllable Structure

Vennemann (1988) has pointed out the relevance of the Syllable Contact Law (SCL) for the typology of Indo-European languages: (30)

Syllable Contact Law (Vennemann 1988: 40, adapted) A syllable contact A.B is the more preferred, the stronger the sonority of the offset A and the less the sonority of the onset B; more precisely – the greater the characteristic difference CS(A)-CS(B) between the sonority of A and that of B.

The SCL is responsible for the difference between the following names in German: (31)

Wartha /var.ta/, Tatra /ta.tra/

(German)

In principle, /tat/ and /ra/ are well-formed German syllables, but the string /tatra/ will be syllabified as ta.tra in German. The SCL says that this is because /t.r/ involves a coda with smaller sonority than the following onset. This is dispreferred to /a.t/ in ta.tra in which the more sonorous /a/ is followed by the less sonorous /t/. The SCL does not always hold for multimorphemic forms, where alignment to morpheme boundaries seems to take precedence: (32)

ta¨g.lich ‘daily’ (< /tag/+/lich/)

(German)

Also clusters such as /tl/ which are not accepted word-initially will be separated by a syllable boundary word-internally: (33)

At.las (German)

Murray and Vennemann (1982) interpret West Germanic Consonant Gemination in this way. Before glides, obstruents geminate: (34)

satjan, skapjan settian, skeppian

(Gothic) (Old Saxon)

The Gothic examples have a bad syllable contact (sat.jan) and Old Saxon has a better contact (set.tian). /r/ before /j/ did not geminate, possibly because the syllable contact was not that bad, but according to Vennemann (1988) there still was a problem, which was solved by strengthening the glide to /g/ in Middle High German: (35)

far.io > ver.ie > Ferge ‘ferryman’

(Middle High German)

Vennemann provides a catalogue of other responses to violations of the Syllable Contact Law (SCL) in Germanic and other Indo-European languages. The fact that such repairs exist as historical processes also implies that the SCL states preferences and not absolute grammaticality. In a language like Norwegian eple is also still syllabified as ep.le, in apparent violation of the SCL.

45

46

MARC VAN OOSTENDORP

2.5 Conclusion Germanic syllable structure is remarkably stable across the various languages. The template in (1) basically suffices to describe all of them, with the possible exception of the appendix position, and some variation among languages as to how sonority restrictions regulate the possibility of adjacent segments. I have not been able to find a good overview of consonant cluster phonotactics in earlier stages of Germanic, but it seems safe to say that also over the course of time the system has remained fairly stable. It is not entirely clear what this means. As other chapters in this volume attest, the Germanic languages have shifted in all kinds of directions in other dimensions, and Germanic clearly has been in contact with other languages (Slavic, Romance) which work with very different templates. Consonant phonotactics thus seems to be subject to some kind of ‘macroparameter’ which is not very flexible. It might be interesting to also attempt a theoretical interpretation of the smaller differences which we do find in Germanic, for instance along the lines of Alber and Meneguzzo’s (2016) optimality-theoretic approach. This would have to account for the slightly different role that sonority sequencing and syllable contact play in the various languages. Why it is that these sonority-related issues are most variable, my knowledge of phonological theory has no answer to.

References Alber, B. 2007. Einfu¨hrung in die Phonologie des Deutschen. Verona: QuiEdit. Alber, B. and M. Meneguzzo, 2016. “Germanic and Romance onset clusters – how to account for microvariation.” In E. Bidese, F. Cognola, and M. C. Moroni (eds.), Theoretical Approaches to Linguistic Variation. Amsterdam: John Benjamins: 25–52. A´rnason, K. 2011. The Phonology of Icelandic and Faroese. Oxford University Press. Basbøll, H. 2005. The Phonology of Danish. Oxford University Press. Den Besten, H. 2012. “Speculations of [χ]-elision and intersonorantic [ʋ] in Afrikaans.” In T. van der Wouden (ed.), Roots of Afrikaans: Selected Writings of Hans Den Besten. Amsterdam: John Benjamins: 79–93. Chapman, K. G. 1962. “Icelandic-Norwegian linguistic relationships.” NTS, Suppl. 7. Conradie, C. 1981. Die ontwikkeling van die Afrikaanse voltooide deelwoord I: 1652–1875. Tydskrif vir Geesteswetenskappe 21: 275–284. Conradie, C. 1982. Die ontwikkeling van die Afrikaanse voltooide deelwoord I: 1875–1978. Tydskrif vir Geesteswetenskappe 22: 97–109.

Germanic Syllable Structure

Cutler, A. J. Mehler, and D. Norris 1983. “A Language-Specific Comprehension Strategy,” Nature 304.5922: 159–160. Fabricius, A. 2002. “Ongoing change in modern RP: Evidence for the disappearing stigma of t-glottaling,” English World-Wide 23.1: 115–136. Fuchs, S., J. Brunner, and A. Busler 2007. “Temporal and spatial aspects concerning the realizations of the voicing contrast in German alveolar and postalveolar fricatives.” Advances in Speech–Language Pathology 9.1: 90–100. Gick, B. 2002. “The American Intrusive L,” American Speech 77.2: 167–183. Goad, H. 2011. “sC Clusters.” In M. van Oostendorp, C. J. Ewen, E. Hume, and K. Rice (eds.), The Blackwell Companion to Phonology. Oxford: Wiley-Blackwell: 1123–1142. Goeman, T. 1999. T-deletie in Nederlandse dialecten. Kwantitatieve analyse van structurele, ruimtelijke en temporele variatie. Utrecht: LOT. Guy, G. 1980. “Variation in the group and the individual: The case of final stop deletion.” In W. Labov (ed.), Locating Language in Time and Space. New York: Academic Press: 1–36. Guy, G. R. 1991a. “Explanation in variable phonology: an exponential model of morphological constraints,” Language Variation and Change 3:1–22. Guy, G. R. 1991b. “Contextual conditioning in variable lexical phonology,” Language Variation and Change 3: 223–239. Hammond, M. 1999. The Phonology of English. A Prosodic Optimality-Theoretic Account. Oxford University Press. Hall, T. A. 1992. Syllable Structure and Syllable Related Processes in German. Tu¨bingen: Max Niemeyer Verlag. Hamann, S. 2003. “German glide formation functionally viewed,” ZAS Papers in Linguistics 32: 137–154. Herzog, M., U. Kiefer, R. Neumann, W. Putschke, A. Sunshine, V. Baviskar, and U. Weinreich 1992. The Language and Culture Atlas of Ashkenazic Jewry, Vol. I. Historical and Theoretical Foundations. Tu¨bingen: Max Niemeyer Verlag and YIVO Institute for Jewish Research. Hinskens, F. 1992. Dialect Leveling in Limburg: Structural and Sociolinguistic Aspects. Doctoral dissertation, University of Nijmegen. Hinskens, F. 2009. “Zuid-Afrika en het Afrikaans. Inleidende notities over geschiedenis, taal en letterkunde.” In H. den Besten, F. L. M. P. Hinskens, and J. Koch (eds.), Afrikaans. Een drieluik. Amsterdam and Mu¨nster: Stichting Neerlandistiek/ Nodus Publikationen: 9–33. Issatschenko, A. 1974. “Das ‘schwa mobile’ und ‘schwa constans’ im Deutschen.” In U. Engel and P. Grebe (eds.), Sprachsystem und Sprachgebrauch. Festschrift fu¨r Hugo Moser zum 65. Geburtstag. Du¨sseldorf: Schwann: 141–171. Jacobs, N. G. 2005. Yiddish: A Linguistic Introduction. Cambridge University Press. Kager, R. and K. Shatzman 2007. “Phonological constraints in speech processing.” In B. Los and M. van Koppen (eds.), Linguistics in the Netherlands 2007. Amsterdam: John Benjamins: 100–111.

47

48

MARC VAN OOSTENDORP

Keydana, G. 2011. “Evidence for non-linear phonological structure in IndoEuropean: The case of fricative clusters.” In B. N. Whitehead, T. Olander, B. A. Olsen, and J. E. Rasmussen (eds.), The Sound of Indo-European –Selected Papers from the Conference Held in Copenhagen, April 16–19, 2009, Copenhagen Studies in Indo-European 4. Copenhagen: Museum Tusculanum: 223–241. Kooij, J. and M. van Oostendorp 2004. Fonologie. Uitnodiging tot de klankleer van het Nederlands. Amsterdam University Press. Kraehenmann, A. 2001. “Swiss German stops: Geminates all over the word,” Phonology 18.1: 109–145. Kristoffersen, G. 1999. The Phonology of Norwegian. Oxford University Press. Labov, W. 1969. “Contraction, deletion and inherent variability of the English copula,” Language 45:715–762. Lass, R. 2002. “South African English.” In R. Mesthrie (ed.), Language in South Africa. Cambridge University Press. Murray, R. M. and T. Vennemann 1982. “Syllable contact change in Germanic, Greek and Sidamo.” Klagenfurter Beitra¨ge zur Sprachwissenschaft 8: 321–349. Noske, R. 1992. A Theory of Syllabification and Segmental Alternation. Ph.D. Thesis, Tilburg University. Oostendorp, M. van 2000. Phonological Projection. Berlin: Mouton de Gruyter. Oostendorp, M. van 2003. “Ambisyllabicity and fricative voicing in WestGermanic dialects.” In C. Fe´ry and R. van de Vijver (eds), The Syllable in Optimality Theory. Cambridge University Press: 304–337. Parker, S. 2011. “Sonority.” In M. van Oostendorp, C. J. Ewen, E. Hume, and K. Rice (eds.), The Blackwell Companion to Phonology. London: Blackwell-Wiley. Ra´cz, P. 2010. “On the phonotactic judgments of Czech native speakers.” In The Odd Yearbook 8: 79–86. Rice, C. 2002. “When nothing is good enough: Dialectal variation in Norwegian imperatives,” Nordlyd 31/2: 372–384. Szigetva´ri, P. 2016. “The curious case of Cj clusters in English.” In The Even Yearbook 12. Dept. of English Linguistics, Eo¨tvo¨s Lora´nd University, Budapest. http://seas3.elte.hu/even/2016/16sz.pdf. Vennemann, T. 1988. Preference Laws for Syllable Structure. Berlin: Mouton de Gruyter. Vestergaard, T. 1968. “Initial and final consonant combinations in Danish monosyllables,” Studia Linguistica 21: 37–66. Visser, W. 1997. The Syllable in Frisian. The Hague: HAG. Voyles, J. B. 1980. “Reduplicating Verbs in North-West Germanic,” Lingua 52: 89–123. Wells, J. 1982. Accents of English. Cambridge University Press. Wiese, R. 1996. The Phonology of German. Oxford University Press. Wright, J. 1910. Grammar of the Gothic Language. Oxford: Clarendon Press.

Chapter 3 The Role of Foot Structure in Germanic Laura Catharine Smith

3.1 Introduction For students and scholars of Germanic, it is difficult to deny prosody’s important role in shaping the history of the Germanic languages from their earliest written monuments to the modern varieties spoken today. Just as analyses of historical phenomena drew on the syllable (e.g. Murray and Vennemann 1983), stress (e.g. Verner’s Law), quantity and quantity changes, as well as the contrast between heavy (long) and light (short) stems, these same elements of prosody continue to shape the modern languages. Indeed, the role of these prosodic units is important enough that syllables, stress, accent, and quantity in Modern Germanic are each treated as separate chapters in this volume. One additional level of the prosodic hierarchy has in more recent decades shown promise in accounting for phonological and morphophonological patterns in Germanic, namely the foot, and more specifically the trochaic foot. Appealing to the trochee which ties the syllable, stress and in some cases quantity together highlights a common thread linking historical developments such as Siever’s Law in Gothic and high vowel deletion in Old English to seemingly unrelated patterns in the modern Germanic languages such as lenition and plural formation in Standard German and Dutch. Focusing on data from the modern Germanic languages and varieties, this chapter seeks to demonstrate the various ways by which the trochee plays a role in these languages ranging from a phonological role licensing segments and segmental features to a morphophonological role in which the trochee determines the stem shape of lexical classes. As such, the foot in Germanic serves as a prosodic template underpinning various aspects of the languages. This chapter does not set out to provide an exhaustive survey of all modern Germanic languages, nor does it seek to discuss the specifics of foot structure, extrametricality,

50

L A U R A C AT H A R I N E S M I T H

or the likes; instead this chapter is intended to highlight how the foot and foot-based templates can help us better understand linguistic phenomena in the modern German languages. By drawing primarily on data from West Germanic, the chapter’s intention is to invite other scholars to consider additional potential linguistic phenomena which may be better analyzed in terms of the foot and foot-based templates. To this end, the chapter begins in Section 3.2 by introducing the reader to the formation of both the moraic and syllabic trochees at work in the modern Germanic languages. Next in Section 3.3, I outline the role of the foot in phonological patterns including cluster simplification and consonant lenition in medial onsets as well as vowel reduction. Next, Section 3.4 explores how the trochee has shaped morphological patterns such as German and Dutch plurals as well as Dutch diminutive formation. Section 3.5 highlights how rethinking Frisian and Scandinavian vowel balance in terms of the trochee ties this phenomenon to other Germanic data. The chapter concludes in Section 3.6.

3.2 Foot Formation and the Foot-Based Template in Germanic To understand the foot’s role in the Germanic languages, it is critical to define the structure of the trochee and foot-based templates. As illustrated in the Prosodic Hierarchy in (1), individual sounds combine to form syllables which in turn come together to form feet. (1)

Prosodic hierarchy (Symbol) Feet (F) Syllables

(σ)

Moras

(µ)

Segments

(individual sounds)

As noted, the predominant foot in Germanic is the trochee. Although this foot type is typically defined in terms of a sequence of two syllables, the first of which is more prominent or salient (see van der Hulst 1999, Lo¨hken 1997, etc.), e.g., German Va´.ter ‘father’, Bru´. der ‘brother’ where the period represents the syllable break, there are actually two types of trochees at work in the modern Germanic languages, namely the moraic trochee and the bisyllabic trochee. Both types of trochees arise from the principle of foot binarity (see Prince and Smolensky 1993) which stipulates that trochaic feet must have either two syllables as in wı´n.ter, or they must have two moras as in German Kuh ‘cow’.

The Role of Foot Structure in Germanic

3.2.1 Moraic Trochees Since the moraic trochee appears in weight (or quantity) sensitive languages, we see it in all the modern Germanic languages except Faroese and Icelandic (Lahiri, Riad, and Jacob 1999, A´rnason 2011). Foot binarity emerges in this trochee in terms of bimoricity where either a single heavy syllable ending in a long vowel (V:) or a closed syllable with a short vowel (VC) can form a foot on its own as in (2) below. Foot boundaries are denoted using square brackets.1 (2)

a. Moraic trochee: single heavy syllables, e.g., German Ball ‘ball’ and Kuh ‘cow’ F Foot level (F=Foot) (ˈH) Syllabic level (H=heavy syllable of 2 moras) μμ Moraic level (μ=mora) (ˈbal) ‘ball’ VC Segmental level (ˈku:) ‘cow’ V: b.

F σS µµ ('bal) ('ku:)

In heavy syllables, the segmental material in the rhyme contributes to the weight. Each single short vowel and consonant is assigned a single mora as Ball illustrates, while long vowels are bimoraic as in Kuh [ku:] (see Broselow 1995). With their two moras, heavy syllables are capable of forming a foot. As depicted in (2b), the foot branches down to a single strong, i.e., stressed, syllable denoted by σS. The two moras necessary for moraic trochees, however, need not occur within the same syllable but can be spread across two adjacent syllables such that two light syllables, i.e., those ending in a short vowel, are seen as equivalent to the heavy footed syllable. This is illustrated in (3) for Old Frisian stidi ‘place’. (3)

1

a. Moraic trochee: light + light syllables, e.g., Old Frisian stidi ‘place’ F Foot level (F=Foot) (L L) Syllabic level (L=light syllable, 1 mora each) μμ Moraic level (μ=mora) [sti di] ‘place’ Segmental level

To remain as agnostic as possible regarding theoretical approaches to foot formation, the foot is frequently depicted using parentheses as in Alber (Chapter 4) except where I retain the formatting of the analyses of others. Interested readers are directed to Liberman (1975) on metrical trees, Liberman and Prince (1977) for metrical grid theory, and Halle and Vergnaud (1987) for bracket grids. Kager (1995) provides an informative overview of various approaches to metrical theory while McCarthy and Prince (1995) propose constraints for feet and related prosodic constituents.

51

L A U R A C AT H A R I N E S M I T H

F

b.

σS σW µ µ sti di

Here the short vowel in each syllable is monomoraic meaning that the resulting syllable is light. Although the first light syllable is unable to build a foot on its own, it can do so with the following light syllable resulting in a bimoraic foot (ˈLL) equivalent to that of the single heavy syllable (ˈH). As the prominent syllable, the first syllable in the (ˈLL) foot attracts stress as marked by the subscript “s” in (3b) while the unstressed syllable is marked as weak, i.e., σW.

3.2.2 Syllabic Trochees These moraic trochees of the shape (ˈLL) with initial stress lend themselves to reinterpretation as syllabic trochees of the shape (ˈσσ) where syllable weight plays no role. Consider German Va¨´.ter ‘fathers’ and Bru¨´. der ‘brothers’. Both conform to the pattern of a stressed syllable followed by an unstressed syllable, i.e., (ˈσσ) as found for the (ˈσσ) foot above. However, in these words, initial syllables with their long vowels are actually heavy and yet the weight of the first syllable is irrelevant. (4)

a. Syllabic trochee: (ˈσσ), e.g., German Va´¨ .ter ‘fathers’ F (ˈσ σ) μμ μ (fɛː tɐ) ‘fathers’ b.

Foot level (F = Foot) Syllabic level (L = light syllable, 1 mora each) Moraic level (μ = mora): irrelevant Segmental level

F σS σ W fε: t a

52

This heavy first syllable does not form a foot on its own. Instead, it is the sequence of two syllables, the first of which is stressed. This structure is critical in forming this trochee which plays a crucial role in the plural systems of Standard German and Dutch (see Section 3.4.1 below).

3.2.3 Moraic Versus Syllabic Trochees We can thus summarize the two trochaic types as follows:

The Role of Foot Structure in Germanic

(5)

Summary of trochaic foot types Moraic trochee = (ˈH) or (ˈLL) Syllabic trochee = (ˈσσ)

Interestingly, both Lo¨hken (1997) and Booij (1995, 1998) argue that the syllabic trochee is the most common or unmarked foot type for German and Dutch respectively. The heavy monosyllabic foot, Lo¨hken notes, is a second possible foot type for German. As data in this chapter will demonstrate, e.g., German and Dutch plural formation, there is a role for syllabic trochees which are blind to the weight of the individual syllables even in languages where stress is influenced by weight sensitivity (cf. Alber, Chapter 4). Indeed, both syllabic trochees and moraic trochees formed by heavy syllables play a role in those languages. This role comes in the form of prosodic templates governing how sounds and phonological features are licensed in certain prosodic positions and how different morphological processes are shaped by the foot. In each case, the specific foot type, i.e., the moraic trochee or the syllabic trochee will be outlined.

3.2.4 Foot-Based Templates Before continuing, it is critical to define what prosodic templates are. In many of the world’s languages, morphemes tend to have a “general phonemic shape” or “canonical form” which often matches or aligns with prosodic constituents such as syllables and feet (Downing 2006: 22). These canonical forms, or templates as they have come to be known, determine the shape of stems for morphological operations such as affixation, reduplication, croppings, or the shape of the output of these processes (cf. McCarthy and Prince 1995). Cross-linguistically, the canonical shape of lexical morphemes such as stems appears to be the foot (Downing 2006). However, these templates have also played a role beyond shaping morphological stems and patterns. Just as syllables have been shown to license certain segments and segmental features (cf. Itoˆ 1986), so, too, can the foot provide the basis for understanding the occurrence of some phonological patterns where for instance some processes such as lenition target certain foot positions which could not otherwise be explained using the syllable. It will be in the application of foot-based templates to the data which follows that we gain a better understanding of how both moraic and weight-insensitive syllabic trochees play a role in shaping the Germanic languages. With this in mind, I now turn to a discussion of the trochee’s role in phonological patterns in Germanic before examining its role in shaping morphophonological patterns.

53

L A U R A C AT H A R I N E S M I T H

3.3 The Role of the Foot in Phonological Patterns As noted, scholars have demonstrated the foot’s role in phonological patterns in West Germanic (e.g., Booij 1995, Holsinger and Houseman 1998, Holsinger 2000). In this section, I outline how divergent processes such as lenition and loss of medial consonants, the licencing of the phonological feature [spread glottis] and vowel reduction are best explained by appealing to one common structure: a foot-based template.

3.3.1 Cluster Simplification in Medial Onsets2 In a number of German and Dutch dialects, medial nasal + stop clusters simplify by losing the stop. In word final position, however, these clusters are retained. This pattern is illustrated by the data from Buttelstedt German in (6). (6)

Medial simplification versus final retention in Buttelstedt German (Thuringian, East Central German dialect, data from Ku¨rsten and Bremer 1910: 46, cited in Holsinger 2000) Buttelstedt German wind ~ winə ~ winɪχ khind ~ kinr̥ blind ~ blinr̩

Standard German Wind, Winde, windig Kind, Kinder blind, blinder

hund ~ hunǝχn̩

Hund, Hu¨ndchen

‘wind, winds, windy’ ‘child, children’ ‘blind, blind [masc.sg. strong adj.]’ ‘dog, doggy’

In Standard German (middle column in (6)), the nd sequence is pronounced in medial and final position. However, in Buttelstedt, this consonant sequence is only retained in word final position, e.g., wind, khind, blind. In medial position, the d has been lost entirely. Medial weakening is not unusual. Indeed, “the canonical weakening environment is medial” (Lavoie 2001: 7). The question then remains how to best account for lenition in medial position. From a purely syllabic perspective, lenition of a medial onset consonant is perplexing. As the onset of the second syllable, the d in a word like windig (cf. 7) should be less likely to undergo loss or lenition than a preceding coda consonant, e.g., the n: (7)

Syllable structure of Winde ‘winds’ σ

σ

O R O R (O = onset, R = rhyme) w

2

I

n d

e

54

The discussion in this section stems from work by Holsinger (2000, 2001) and is based on Smith et al. (2005) presented at the Germanic Linguistics Annual Conference, University of California-Davis.

The Role of Foot Structure in Germanic

The preference for the loss of coda consonants over onset consonants stems from the observation that the ideal syllable shape crosslinguistically is CV. According to Vennemann and others (e.g., Murray and Vennemann 1983, Vennemann 1988, also Hooper 1976), consonants in coda position are more likely to undergo lenition whereas onsets are more likely to be positions of consonantal strengthening, particularly in syllable contacts where a consonant in a syllable coda is immediately followed by another consonant in onset position as seen in (7). Thus, a syllabic perspective cannot adequately account for the medial plosive loss. Similar difficulties arise for analyses drawing on perceptual cues such as Licensing by Cue (cf. Coˆ te´ 2000, 2004) which have also been proposed to account for phonological patterns including consonant cluster simplification. This approach argues that consonants are more likely to be retained when adjacent to vowels which can convey their perceptual cues. Conversely, without an adjacent vowel to convey their cues, these consonants are more likely to undergo lenition and loss. Consider now the example Wind~Winde~windig in (6). When the nd cluster is in final position, the final d lacks a cuecarrying adjacent vowel. Conversely, when this same cluster appears in medial position, a following vowel i is available to carry the perceptual and acoustic cues of the d, meaning it would be less likely to be lost in medial position than in word final position. However, this is precisely the opposite to what happens: d is lost in the more favorable position in terms of perceptual cues. This theoretical approach thus fails to make accurate predictions for the West Germanic data. Seen through the lens of the syllabic trochee, these data find a straightforward explanation. As Holsinger (2000) argues, the loss or lenition of consonants in West Germanic correlates well with foot medial position. In short, in nd clusters, d is more likely to undergo lenition and outright loss in the onset of a weak branch of a bisyllabic foot, i.e., the second syllable of the syllabic trochee. This is reflected in (8) where arrows point to the onset of the weak foot branch targeted for lenition and loss: (8)

nd clusters: loss of d in onset of weak branch of foot F σS

σW

O R O R w

I

n d

‘winds’→wIn

As (8) illustrates, the voiced stop is unlicensed in the weak position of the trochaic foot in Buttelstedt. Such an approach is confirmed by

55

56

L A U R A C AT H A R I N E S M I T H

findings elsewhere. For instance, Raymond et al. (2006) argue that a strong head of a foot, i.e., a strongly stressed first syllable, is more likely to trigger consonant deletion or lenition in the onset of a following syllable. Strongly stressed syllables such as foot initial syllables permit consonants to be more fully articulated in that foot position than consonants found in less stressed syllables, i.e., nonheads of a foot. Similar alternations between nd~n in final and medial positions arising from word cluster simplification are found in a variety of other dialects such as North Saxon and Northern Bavarian shown in (9):3 (9)

nd~n alternations in German dialects: Kind-Kinder ‘child~children’ (Holsinger 2000) Dialect North Saxon, Holstein Northern Bavarian

Final ~ Medial Kind ~ Kinner ghind ~ ghinə

Despite differences in the pronunciation of the initial consonants and plural endings, the loss of d in the medial nd cluster remains constant across these dialect forms and can be summed up as follows: d is lost in onsets of the weak branch of the trochee. The same analysis can be applied to account for subtractive plurals in Hessian in which the plural form has less segmental material than the singular noun, e.g., hond (sg.) ~hon (pl.) ‘dog~dogs’. As the examples in (10) illustrate, it is not only the nd clusters which simplify, but rather sonorant + stop clusters. (10)

Medial cluster simplification and subtractive plurals in Hessian (LC, NC) (data from Holsinger and Houseman 1998) Original form hond+e vald+er bɛrg+e

Loss of medial stop → honne → vɛller → berre

Loss of plural ending → hon ‘dogs’ → vɛl ‘forest’ → ber ‘mountains’

In these examples, the double consonants indicate a preceding short vowel rather than the assimilation of the medial stop to the preceding sonorant. The subtractive plurals then arise from two processes. First, the failure of

3

It is not uncommon in dialects with foot-medial cluster simplification to find stop epenthesis word or foot finally: Excrescent -b, Buttelstedt (cited from Smith et al. 2005) Buttelstedt

Standard German

khaˉmn ~ khaˉmb ˙ naˉmɛ ~ nıˉmb

kamen ~ kam

‘they came, he / she / it came’

nehmen ~ nähme

‘to take, he / she / it would take’

That stop epenthesis is permitted foot-finally, but not foot-medially, underlines the fact that the stop is not licensed in foot-medial onsets.

The Role of Foot Structure in Germanic

the stop to be licensed in the onset of the weak branch of the foot leads to its loss which is next followed by the loss of the plural vowel ending. This is illustrated in (11). (11)

Hessian subtractive plurals F σS

σW

OS R OWR e

h on d

‘dogs’

Step 1: d in nd consonant clusters lost in onset of weak branch of foot Step 2: plural ending schwa is lost As a result of the two steps, the plural has less segmental material hon than the original singular form hond. Cluster simplification in other word forms in Hessian support the analysis above in which clusters simplify before schwa loss. Consider the word forms in (12) related to the root Hand in Westerwald. (12)

Cluster simplification in Hessian (examples from Westerwald, Hessian; Hommer 1910) Hessian hand ~ hɛn ~hanɛln

Standard German Hand~Ha¨nde~handeln

‘hand, hands, to treat, act’

As expected, the plural for hɛn lacks both the plural ending and consonant cluster found word-finally in the singular hand. The cluster has also been simplified word medially in the verb handeln, the loss of which is shown in (13). (13)

Loss of d in handeln in Westerwald German F [σS

σW]

OS R OWR h a n d εl n

‘to treat’

Since d is in the onset of the nonhead of the foot, i.e., the weak branch of the foot, then it is targeted for loss. Next, the n in the coda of the first syllable is resyllabified as the onset of the following syllable. In every example, the foot serves as a prosodic template where the realization of a consonant, i.e., fully realized versus lenited or completely lost, is dependent on the consonant’s position within the foot. Consonants within the head of the foot, i.e., the strong foot-initial

57

58

L A U R A C AT H A R I N E S M I T H

position are more strongly articulated with less variability than those found in the onset of the weak branch of the foot, i.e., the nonhead of the foot. Indeed, the medial stop lenition in the German dialects discussed above results from templatic constraints regarding the expression of features in specific positions of the foot, a result not restricted to consonant cluster simplification. Let us now turn to another type of word-medial consonant lenition.

3.3.2

Foot-Medial Consonant Lenition: Variability of Segmental Realization The cluster reduction examples just discussed demonstrated one potential outcome for lenition in foot-medial onsets: the complete loss of the consonant. Loss, however, is not the sole outcome. Consider the variable articulation of the nt in American English renter. An informal survey by Smith et al. (2005) asking linguistically aware colleagues who were unaware of the purpose of the inquiry revealed the following (in)variability in pronunciation of word medial /nt/ and /nd/ clusters in renter versus render: (14)

Reported pronunciation for /nt/ and /nd/ clusters (Smith et al. 2005) renter

[ɹɛ.nəɹ], [ɹɛ.̃ ɾə̃ ɹ]

render

[ɹɛn.dəɹ]

highly variable (deletion, flapping, full articulation) no variation

As the examples in (14) show, the /nd/ cluster in render lacks the variability of the /nt/ in renter. Since the two plosives are distinguished based on the unary feature [spread glottis] (cf. Holsinger 2000, and Salmons Chapter 6), the contrasting behavior can be explained in terms of the licensing of this feature drawing on Holsinger’s (2001) weak position constraint schema: (15)

Weak position constraint schema (Holsinger 2001: 103) “A feature is constrained in the non-head sector of a headed prosodic domain.”

As illustrated in (16), this means that [spread glottis] is constrained in the nonhead syllable, i.e., the weak or right branch of a trochaic foot. Consequently, since /d/ lacks the feature [spread glottis], then the appearance of this sound in the weak branch onset is left unaffected (16b). Conversely, since [spread glottis] applies to /t/, then its appearance in the nonhead, i.e., weak branch of the foot is impacted by the weak position constraint in (15) as highlighted by the arrows pointing to the parts of the weak branch syllable [σw]:

The Role of Foot Structure in Germanic

b.

F [σs

σw]

Os

R

Ow

/r

εn

t

F [σs

σw]

R

Os

R

/r e

a.

/r

εn

[spr gl] ⇒

Ow d

R /r e

(16)

/ / (no feature, no effect)

In other words, [spread glottis] is not licensed in the nonhead syllable of a foot. As such, the realization of /t/ is variable including the nasalized flap [ɾ]̃ and even the deletion of t. As these data illustrate, appealing to the syllable is inadequate since licensing of the feature [spread glottis] is done at the level of the foot. While the feature is explicitly licensed in head positions, i.e., the left branch of a trochee, the same is not true for its appearance in nonhead position. Lenition of fortis plosives is also found in German dialects such as Odenwald (Freiling 1929, cited in Holsinger 2000) where [spread glottis] is only licensed in the onset of a strong branch of a foot, i.e., head of a foot, when the stop is immediately followed by a vowel. Thus /t p k/ are realized in this position as [th ph kh], but in a trochee. In all other positions, e.g., syllable codas, consonant clusters, and onsets of nonhead syllables of feet, the allophone is the voiced or lenis counterpart, i.e., [d b g] respectively as in (17). (17)

/t/

/p/

/k/

[spread glottis] in Odenwald (Freiling 1929, cited in Holsinger 2000): Prevocalic onset In onset clusters Medial [tha:dəl], [the:dər] [fa:dər] Ta¨ter ‘perpetrator’ Vater ‘father’ versus /d/ = [d] in all positions [phebern] [blads] pappeln ‘to babble’ Platz ‘place’ versus /b/ = [b] in all positions [khabə] [globə] Kappe ‘hood’ klopfen ‘to knock’ versus /g/ = [g] in all positions

Final4 [khold] kalt ‘cold’ [khalb] Kalb ‘calf’ [jag] Jacke ‘jacket’

In sum, stops are fully realized as [spread glottis] in foot initial position. In all other positions, these underlyingly fortis stops are not fully realized as fortis and are subject to lenition.

4

Lenis obstruents are not permitted in syllable codas in Standard German and generally undergo final fortition. Conversely, the examples in (17) show fortis stops leniting in every position but foot initial.

59

60

L A U R A C AT H A R I N E S M I T H

This analysis can be extended to account for other types of word medial lenition beyond those involving [spread glottis]. Holsinger (2000) provides a catalogue of the types of lenition processes impacting the onset of the nonhead syllable of the syllabic trochee in German, Dutch, and even some Scandinavian dialects. These processes include voicing of stops and fricatives, spirantization of stops, sonorization of stops, and deletion of fricatives. The interested reader is directed to Holsinger (2000: 21–22) for more details. In each of these cases of lenition, the data can be analyzed in terms of a failure of the relevant segmental feature to be licensed in the nonhead foot position leading to lenition.

3.3.3 Vowel Reduction in Nonhead Branches of Feet Before concluding this section on the role of the foot-based template on segmental reduction, it is critical to acknowledge the role of the foot on vowels (see also Section 3.5). As for consonants, full articulation of vowels is more likely to be found in the head of a foot, i.e., the strong branch of a foot while vowels in the weak branch of a foot are subject to reduction or are already underlyingly centralized vowels such as [ǝ] or [ɐ] (cf. Booij 1995: 134). Examples from German illustrate this patterning: F

(18) [σS

σW]

OS R OW R

Since vowels can be fully articulated in the head of a foot, for tense vowels this means their long allophone, e.g., [u:], [a:]. However, when tense vowels occur in the weak branch of the foot, they surface as the short allophones, e.g., [i] in Uni, or they are underlyingly short or reduced, /ɐ/ or /ǝ/.

3.3.4

Shaping Phonological Patterns: The Trochee as a Prosodic Template To summarize, reduction processes impacting both consonants and vowels tend to stem from weak position constraints where features and the segments they define are fully expressed in the head of a foot, but weakly expressed – if at all – in other foot positions, e.g., the nonhead of the foot. Since the foot thus often serves to license features and segments in Germanic, it can consequently be viewed as a prosodic template stipulating the potential phonological form of words. Perhaps most interesting is that the trochee used in the consonantal and vocalic examples was a nonweightsensitive syllabic trochee. And yet the actual choice of tense-lax phonemes in German is guided by syllable weight and shape: lax vowels occur in closed

The Role of Foot Structure in Germanic

syllables while tense vowels occur in open syllables. This underscores the distinction between the syllabic trochee used in these prosodic templates regardless of syllable weight and the weight-sensitive moraic trochee at work in determining stress placement. Having shown the foot’s ability to license segments and segmental features, I next demonstrate its ability to shape morphological and lexical classes in West Germanic.

3.4 The Role of the Trochee in Shaping Lexical Classes and Patterns: German and Dutch Plurals In addition to the trochee’s role in shaping phonological patterns, this foot type has also been shown to shape lexical patterns in Germanic. For instance, plural formation in German and Dutch is strongly guided by the syllabic trochee, as is diminutive formation in Dutch. For the sake of space, this section focuses on plural formation in Standard German and Dutch while also highlighting the foot’s role in the various dialects. For a description of the foot’s role in Dutch diminutives, the interested reader is directed to Smith (2009).

3.4.1 Plurals in Standard German and Dutch Although the plural suffixes used in Dutch and German are different, plural formation in both languages is shaped by the syllabic trochee. Plurals of native words in both languages tend to end in a syllabic trochee thanks to the choice of suffix which is critical in aligning the plural form with the trochaic template. For Dutch, the choice between the plural endings -en and -s is determined by the shape of the singular noun (Booij 1998, van der Hulst and Kooij 1998). When the singular stem already ends in a trochee, then –s is added since it will not disturb the pre-existing trochaic pattern (19a). On the other hand, if the noun stem does not already satisfy the required trochaic shape, then the syllabic ending -en will be added to help the noun conform to the syllabic trochee (19b). (19)

Plural formation in Standard Dutch

a. Stem is a syllabic trochee, -s: (ˈσσ) ➔ (ˈσσ) vader – vaders ‘father—fathers’ natie – naties ‘nation—nations’

b. Stem is not a syllabic trochee, -en: (ˈH) ➔ (ˈσσ) boek – boeken ‘book—books’ non – nonnen ‘nun—nuns’

Thus, the plural suffix aligns the plural noun with the syllabic trochee template for the plural. A similar pattern emerges in German where a variety of plural endings are available to help align the plural with the syllabic template (see 20). As for Dutch, when the noun stem already ends in a syllabic trochee, the

61

L A U R A C AT H A R I N E S M I T H

nonsyllabic plural options -n or Ø (no ending) ensure the noun continues to conform to the final syllabic trochee (20a). Conversely, for nouns not already ending in a trochee, the syllabic suffixes -en, -e or -er add the necessary unstressed syllable needed for the plural to map to the trochee (20b).5 (20)

Plural formation in Standard German

a. Stem is a syllabic trochee: (ˈσσ) ➔ (ˈσσ)

b. Stem is not a syllabic trochee: (ˈσ) ➔ (ˈσσ)

Ta

Uh ren ‘clocks’ Frau

feln ‘tables’

-n

Tan ten ‘aunts’

-en

en ‘women’

Leh rer ‘teachers’ On kel ‘uncles’ Va¨ ter ‘fathers’

No ending

Freun de ‘friends’ Jah re ‘years’

No ending

Bu¨

cher ‘books’

(with umlaut) Bru¨ der ‘brothers’

-e -er (with or without umlaut)

Kin

der

The five choices of plural suffixes in German are thus delimited based on the prosodic shape of the stem. Trochaic stems select from -n or -Ø, while choices for nontrochaic stems are -en, -er, and -e. These specific choices of ending are further determined by other factors such as gender, e.g., feminine nouns are more likely to end in -n.6 Two additional points can be made regarding German and Dutch plurals. First, the template constraining the output of plural formation as a syllabic trochee without reference to syllable weight, and second the mapping of plurals to this trochaic template is done at the right edge of the word. Interestingly, these prosodically driven plurals end in schwa syllables in the nonhead syllable of the foot. This schwa will either be part of the stem as for the trochaic singular nouns, e.g., German Tafeln, Onkel,7 and Dutch vaders, or it will arise from the plural suffix itself. The plural template for German and Dutch can thus be formalized as in (21) (Smith 2007b: 362): (21)

Template for German and Dutch plurals Template for German and Dutch plurals (Ft)# Word final foot ('σσ) Syllabic trochee e

62

5

Must end in a schwa syllable

Plurals formed using the suffix -s “differ from all other plurals in a number of dimensions, that is, in phonological, morphological, lexical and processing properties” (Wiese 2001: 19). For instance, -s plurals tend to be borrowed words, e.g., Klub-Klubs ‘club-clubs’ or those ending in vowels, e.g., Uni-Unis ‘university-universities’. For this reason, s- is set aside in the discussion of the prosodic structure of German plurals.

6

Since umlauting is a distinct process from the prosodically driven choice of plural suffix, even though it co-occurs with some suffixes, it is beyond the scope of the chapter as is the role of gender in the selection of plural suffixes from the prosodically-driven choices.

7

For German, -er [ɐ] syllables suffice as schwa syllables as do other syllabic sonorants, e.g., Tafeln [ta:fl̩ n], which replace the schwa + sonorant combination especially after obstruents, e.g., [ǝl]→ [l̩ ], [ǝr]→ [ɐ].

The Role of Foot Structure in Germanic

As (21) illustrates, the template is mapped to the right edge of the word where the word must end in a bisyllabic foot, the last syllable of which must contain a schwa or syllabic sonorant. The reality of the trochaic template for plurals has been confirmed in a variety of studies for German. Comparing the acquisition of German plurals by native-speaking German children with and without language learning difficulties, Kauschke et al. (2013: 574) found that children without speech problems performed better than children with language deficits at forming prosodically appropriate plurals of real and nonsense words. Children with language deficit seemed to “show a reduced sensitivity to prosodic requirements.” In another study, Smith et al. (2016) found that native German speakers overwhelmingly produced plural forms of nonsense words ending in a syllabic trochee. Moreover, subjects tended to rate incorrect plural forms matching the trochee, e.g., Schlu¨ssels (incorrect but trochaic) more favorably than those which did not, e.g., Schlu¨sselen (incorrect but not trochaic). As these studies illustrate, the trochaic plural template is a part of a native speaker’s intuition which can be extended to new words. Indeed, even when natives failed to agree on endings in these studies, what was constant was their reliance on the trochaic plural form.

3.4.2 Plurals in German Dialects Plurals in the German dialects reveal some differences from the standard language. For instance, in north central German dialects, the plural suffixes may not always match those in Standard German, e.g., Fensters, but the syllabic trochee nevertheless continues to shape the plural forms in these dialects as in (22): (22)

Plural forms using the syllabic trochee (unexpected endings bolded) Dialect Sg. – Plural

Standard German

Glosses of singular

West- and Eastphalian (Durrell 1989)

Schaº aº p

– Aug(e) – Augen Fenster – Fensters

Schaf – Schafe Auge – Augen Fenster – Fenster

‘sheep’ ‘eye’ ‘window’

Upper Saxon (Bergmann 1989)

[dag] – [da:ɣə] [jast] – [jestə] [dorf] – [derfər]

Tag – Tage Gast – Ga¨ste Dorf – Do¨rfer

‘day’ ‘guest’ ‘town’

Schaº aº pe

Likewise in East Low German, plurals by and large conform to the syllabic trochee as in (23). However, some monosyllabic stems undergo vowel lengthening, e.g., [dax] ‘day’ – [daːːx] ‘days’ creating a single superheavy monosyllabic trochee. This results from the loss of final –e which is used to mark the Standard German plural forms for

63

64

L A U R A C AT H A R I N E S M I T H

words like Tage. Consequently, it is the lengthened stem vowel which denotes the plural form. These plurals nonetheless conform to a foot, but instead of the syllabic trochee (ˈσσ), they conform to a superheavy moraic trochee. Although it could be argued that the single form was already a (ˈH) moraic trochee, this lengthening draws on the superheavy syllable and foot at work in, for instance, Dutch (Booij 1998) and creates a greater distinction between the singular and suffixless plural. (23)

Plurals in East Low German: syllabic trochees and vowel lengthening in monosyllables Dialect Sg. – Plural

Standard German

Glosses of singular

East Low German (Scho¨nfeld 1989)

[slœːtl] – [slœːtls] [stɔk] – [stœkæ] [haːrt] – [haːrtn]

Schlu¨ssel – Schlu¨ssel Stock – Sto¨cke Herz – Herzen

‘key’ ‘stick’ ‘heart’

BUT:

[dax] – [daːːx] [slax] – [slεːːc¸]

Tag – Tage Schlag – Schla¨ge

‘day’ ‘hit’

Vowel lengthening has been found in the plurals of other dialects such as in North Saxon as well. In this dialect, some plurals which are formed using a suffix in Standard German, e.g., Schiff ~ Schiffe ‘ship~ships’, exhibit instead lengthened vowels in the plurals, e.g., Schipp ~ Scheep. This alternation between the vowels in the singular and plural forms reflects a similar vowel contrast in Dutch, e.g., schip ~ scheepen but which also includes a suffix as in (24). (24)

Plurals in North Saxon: vowel lengthening (Goltz and Walker 1989)

Vowel lengthening in some Dutch plurals

Dialect Sg. – Plural

Standard German

Glosses of singular

Schipp – Scheep Dag – Daag’ Breef – Breev’

Schiff – Schiffe Tag – Tage Brief – Briefe

‘ship’ ‘day’ ‘letter’

schip – scheepen [eː] glas – glazen [aː] blad – bladen [aː]

The alternation between the short vowel of the singular with the long vowel of the plural in the absence of a suffix yet again demonstrates that a heavy stem, i.e., one corresponding to a moraic trochee, satisfies the trochaic template in these dialects for plural formation. The effects of the trochaic plural in Standard German and Dutch have in some dialects, e.g., Low Alemannic, become obscured. For instance, the singular ~ plural pairs for Standard German day, night, and arm in (25a) all have plurals ending in –e. This vowel, however, has been lost in the Low Alemannic forms rendering a monosyllabic plural, i.e., one that clearly

The Role of Foot Structure in Germanic

does not correspond to the trochaic template in the standard language though which may be argued to conform to a moraic trochee. However, the (25b) plurals do indeed end in the expected syllabic trochee. (25)

Plurals in Low Alemannic (Philipp and Bothorel-Witz 1989) versus Standard German Low Alemannic Singular

Standard German

Plural

Singular

Plural

Gloss

a.

/tɑːj/ /nɑxt/

/taːj/ /naxt/

Tag Nacht

Tage Na¨chte

‘day’ ‘night’

b.

/plyːəm/ /manʃ/

/plyːəmə/ /manʃə/

Blume Mensch

Blumen Menschen

‘flower’ ‘people’

The failure of some forms to fit the syllabic trochee arises from the interplay of prosody and loss of final sounds in the dialect where the final sound of the plural marker is lost. If the plural marker is originally just schwa, as in Tage, then its loss leaves the plural without a suffix, /taːj/. However, for plurals originally ending in -en, Blumen, the loss of the final sound -n would still leave a plural marker on the noun, namely schwa, /plyːəmə/. Thus the loss of final sounds in the dialect obscures the trochaic plural pattern. Plural formation in the dialects reveals two ends of the spectrum. First, in some dialects, plurals are formed using the syllabic trochee, while in others, vowel lengthening can create a superheavy moraic foot which satisfies the trochaic template. And in yet others, the prosodically driven plural pattern has been obscured by subsequent sound changes at work in the dialect. Before closing, it is worth noting that West Germanic is not the only Germanic branch impacted by the trochee. Indeed, one pattern, namely vowel balance, appears across the history of both Frisian, a West Germanic language, and some East Scandinavian dialects of North Germanic. To date, this process has typically been associated with level stress rather than foot structure.8 In Section 3.5, I briefly describe how the foot provides a unified approach to vowel balance.

3.5 Reinterpreting Stress-Based Analyses in Terms of the Foot: Vowel Balance In vowel balance, the quantity and quality of the vowel in a second syllable depends upon the quantity of the first syllable (Kusmenko and 8

Riad (1992) is a notable exception; he does associate vowel balance in Scandinavian to level stress but explains level stress in terms of the foot.

65

66

L A U R A C AT H A R I N E S M I T H

Riessler 2000: 211, see also Riad 1992). In its earliest manifestation in Old Frisian (Riustring dialect), Old Swedish, and Old Eastern Norwegian manuscripts (cf. Kusmenko 2007), the full vowels a, i, and u appear after short syllables, while the reduced (or centralized, see Versloot 2008: 81) vowels æ (Scandinavian) / e (Frisian), e, and o occur after long syllables as in (26): (26)

Vowel balance in Old Frisian and Old Swedish

Old Frisian (Smith 2007a)

Old Swedish (Versloot 2008)

i and u after short stems (VC)

e and o after long (V:C and VCC) and polysyllabic stems

godi ‘God, dat. sg.’ wetir ‘water, nom./acc. sg. nom. pl.’ skipu ‘ship, nom. pl.’ himule ‘heaven, dat. sg.’

liode ‘people, nom. pl.’ himule ‘heaven, dat. sg.’

faÞir ‘father, nom.sg.’ faÞur ‘father, acc.sg.’

moˆÞer ‘mother, nom.sg.’ moˆÞor ‘mother, acc.sg.’

bokon ‘book, dat. pl.’ gerso ‘grass, nom. pl.’

As (26) illustrates for the old languages, the full vowels occur following short (or light) stems, but are reduced following long (or heavy) stems as well as polysyllabic stems. At this stage, the contrast is easily observable without other processes, e.g., vowel harmony, vowel lengthening, or even deletion, obscuring the pattern in the modern dialects. Accounts for vowel balance typically draw on moras and word stress (Perridon 2002, Kusmenko and Riessler 2000). In these accounts, the first two moras of originally long bisyllabic words are stressed leaving the third mora vulnerable to reduction and loss. For the sequence of two light syllables, however, the stress was evenly distributed across the two syllables as so-called ‘level stress.’ Level-stress vowels were argued to retain their quality thereby failing to reduce. This analysis can be reinterpreted in simpler terms via the moraic trochee. When the i, u, or a is footed with the preceding light syllable as in (27a), it is retained. Conversely, since heavy syllables are able to form [H] feet on their own without the following vowel (cf. 27b), then the following vowel is left subject to reduction and loss. (27)

A foot-based account of Old Frisian vowel balance

a. Foot structure of OFris stidi ‘place, acc./dat. sg.’

b. Foot structure of OFris gerso ‘grass, nom.pl.’

F (L L) μμ [sti. di]

F (H) L μμ μ [wral] de

Foot level Syllabic level Moraic level Segmental level

The Role of Foot Structure in Germanic

In the old period of these languages, a vowel’s inability or ability to be footed with the initial foot determined whether it would be lost or retained respectively. That the vowel was retained in the second syllable of a [LL] foot contrasts with the reduction of these vowels in modern German outlined in Section 3.3.1 above. Since level stress spread stress evenly across both syllables of the bisyllabic foot as opposed to the initial syllable in the modern German trochee, then the vowel in that second syllable was not as susceptible to loss (see also Riad 1992 for a similar analysis for Scandinavian). Vowel balance continues to persist in some modern dialects of Weser Frisian and northeastern Scandinavian. Consider the examples from Wursten and Wangeroog Frisian in (28). Following light stems, i and u are retained, while a appears as either e or schwa. The dialects differ, however, in their treatment of vowels following heavy stems. For Wursten, unfooted vowels reduce to schwa, while they undergo apocope in Wangeroog. (28)

Vowel balance in Weser Frisian dialects

Dialect

Light first syllable (level stress)

Heavy first syllable

Wursten Frisian (Smith and van Leyden 2007)

/(hood)-mikiir/ ‘hat maker’ /nukuude/ ‘naked’

/(kon)-jootǝr/ ‘pot moulder’

Wangeroog Frisian (Kock 1904)

hunǝ ‘chicken’ (OFr. hona) stidi ‘stead’ (OFr. stidi) sxypu ‘ship’ (OFr. skipu)

mo´un ‘moon’ (OFr. moˆna) Þuˆm ‘thumb’ (OFr. thuˆma)

The modern Scandinavian dialects have likewise inherited the historical pattern where the vowel following the heavy syllable has either reduced or apocopated: (29)

Vowel balance in modern Scandinavian dialects (Kusmenko 2007, unless otherwise stated)

Dialect

Light first syllable (level stress)

Heavy first syllable

Houtska¨r (Finland; Kusmenko 2007; Kock 1904)

bakaº ‘to bake’ hakur ‘chin’

kast ‘to cast’ fall (OSw. falla) ‘to fall’ tjeldor (OSw. kældor) ‘sources’

Norrland (Kusmenko 2007) Ormso¨/Nukko¨ (Kock 1904)

baka ‘to bake’ droepa ‘to strike dead’

kaste/kaˆst ‘to cast’ brinn (← brinna) ‘to burn’

67

68

L A U R A C AT H A R I N E S M I T H

Due to space constraints, the situation is admittedly simplified (cf. Kusmenko 2007, Kusmenko and Riessler 2000, Smith and van Leyden 2007). As in Frisian, final a tends to undergo either loss or reduction following a heavy syllable, while other vowels, e.g., u reduce to o. After light first syllables, however, this a (or aº ) has in some dialects undergone additional lengthening. Notably in some of the most archaic dialects, a type of vowel balance arises with a trimoraic requirement for bisyllabic words (Kusmenko and Riessler 2000). This rule has two consequences. First, the initial syllable may contain two moras followed by a monomoraic syllable. The vowel here is retained. Second, when the initial syllable is monomoraic, i.e., a light stems, the second mora, i.e., the vowel in the second syllable, is lengthened, e.g., viku: (OIcel. viku), stugu: (OIcel. stugu) in Vaº gaº (Norway). This trimoraic requirement could be interpreted as a trimoraic foot template applying to bisyllabic feet. With level stress, it could be argued that the equal intensity of the second syllable would lend itself well to lengthening. This appears substantiated by Kristoffersen’s (1990: 191–195, cited in Smith and van Leyden 2007) findings that the second syllable in levelstress environments is longer than the initial syllable in NordGudbrandsdalen. In sum, the moraic trochee can account for the alternations in the vowel balance data. First, the foot becomes the domain of stress, with a single heavy syllable foot attracting the stress to itself, or the syllabic trochee sharing the stress across the foot.9 Since the second syllable of a foot with level stress tends to be longer than the initial syllable, then the consequence is to not only offset the tendency to reduce vowels in the second syllable of the [σ́σ] trochees, but indeed to reinterpret the second vowel of the foot as long in some dialects. The real takeaway is that this phenomenon can be reinterpreted as arising from the interaction of stress with the foot.

3.6 Conclusion As the data presented in this chapter demonstrate, the foot does play a varied role in shaping the Germanic languages. Phonologically, this influence includes licensing features in the head of the foot template, but failing to do so even in onsets of weak branches of the trochee. At times the lenition that arises worsens phonetic cues. The trochee also shapes the output of morphological processes such as Dutch and German plural formation or the stem shape for suffixation as in 9

Riad (1992: 172) describes this slightly differently stating that a light stem is “dominated by a unipositional main-stress foot” such that “one single stress position dominates two syllables.” The outcome is nevertheless the same, that the stress is spread evenly across both syllables of the foot.

The Role of Foot Structure in Germanic

Dutch diminutive formation. And lastly, some processes which may be attributed to other factors, such as stress, may be reinterpreted in terms of the foot as with vowel balance in Frisian and Scandinavian thereby tying this phenomenon to the many others from Germanic’s history. This chapter has scratched the surface regarding the foot’s role in shaping the Germanic languages both historically and in modern times. The invitation is now extended to the reader to continue seeking out additional ways the foot has influenced (morpho)phonological patterns across the Germanic languages.

References A´rnason, K. 2011. The Phonology of Icelandic and Faroese. Oxford University Press. Bergmann, G. 1989. “Upper Saxon.” In Russ (ed.): 290–312. Booij, G. 1995. The Phonology of Dutch. Oxford: Clarendon Press. Booij, G. 1998. “Phonological output constraints in morphology.” In W. Kehrein and R. Wiese: (eds.), Phonology and Morphology of the Germanic Languages. Tu¨bingen: Max Niemeyer Verlag: 143–163. Broselow, E. 1995. “Skeletal positions and moras.” In J. Goldsmith (ed.), The Handbook of Phonological Theory. Cambridge, MA: Blackwell:175–205. Coˆte´, M-H. 2000. Consonant Cluster Phonotactics: A Perceptual Approach. Unpublished Ph.D. dissertation, MIT. Coˆte´, M.-H. 2004. “Consonant cluster simplification in Que´bec French,” Probus 16: 151–201. Downing, L. 2006. Canonical Forms in Prosodic Morphology. Oxford University Press. Durrell, M. 1989. “Westphalian and Eastphalian.” In Russ (ed.): 59–90. Freiling, P. 1929. Studien zur Dialektgeographie des hessischen Odenwaldes. Marburg: Elwert. Goldsmith, J. (ed.) 1995. The Handbook of Phonological Theory. Cambridge, MA: Blackwell. Goltz, R. and A. Walker 1989. “North Saxon.” In Russ (ed.): 31–58. Halle, M. and J-R. Vergnaud 1987. An Essay on Stress. Cambridge, MA: MIT Press. Holsinger, D. 2000. Lenition in Germanic: Prosodic Templates in Sound Change. Unpublished Ph.D. dissertation, University of Wisconsin–Madison. Holsinger, D. 2001. “Weak position constraints: The role of prosodic templates in contrast distribution,” Zentrum fu¨r Allgemeine Sprachwissenschaft Papers in Linguistics 19: 91–118. Holsinger, D. and P. Houseman 1998. “Lenition in Hessian: The case of subtractive plurals.” Yearbook of Morphology 1998. Dordrecht: Kluwer: 159–174.

69

70

L A U R A C AT H A R I N E S M I T H

Hommer, E. 1910. Studien zur Dialektgeographie des Westerwaldes. Dissertation, Philosophische Fakulta¨t der Universita¨t Marburg. Marburg: R. Friedrich’s Universita¨tsbuchdruckerei. Hooper, J. 1976. Introduction to Natural Generative Phonology. New York: Academic Press. Hulst, H. van der 1984. Syllable Structure and Stress in Dutch. Dordrecht: Foris. Hulst, H. van der 1999. “Word accent.” In H. van der Hulst (ed.), Word Prosodic Systems in the Languages of Europe. Berlin and New York: Mouton de Gruyter: 3–115. Hulst, H. van der (ed.) 1999. Word Prosodic Systems in the Languages of Europe. Berlin and New York: Mouton de Gruyter. Hulst, H. van der and J. Kooij 1998. “Prosodic choices and the Dutch nominal plural.” In W. Kehrein and R. Wiese (eds.), Phonology and Morphology of the Germanic Languages. Tu¨bingen: Max Niemeyer Verlag: 187–197. ˆ Ito, J. 1986. Syllable Theory in Prosodic Phonology. Unpublished Ph.D. Dissertation, University of Massachusetts Amherst. Kager, R. 1995. “The metrical theory of word stress.” In J. Goldsmith (ed.), The Handbook of Phonological Theory. Cambridge, MA: Blackwell: 367–402. Kauschke, C., L. Renner, and U. Domahs 2013. “Prosodic constraints on inflected words: An area of difficulty for German-speaking children with specific language impairment?” Clinical Linguistics & Phonetics 27: 574–593. Kehrein, W. and Wiese, R. (eds.) 1998. Phonology and Morphology of the Germanic Languages. Tu¨bingen: Max Niemeyer Verlag. Kock, A.1904. “Vocalbalance im Altfriesischen,” Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 29: 175–193. Kristoffersen, G. 1990. “East Norwegian prosody and the level stress problem.” Unpublished ms., University of Tromsø. Ku¨rsten, O. and Bremer, O. 1910. Lautlehre der Mundart von Buttelstedt bei Weimar. Leipzig: Breitkopf & Ha¨rtel. Kusmenko, J. 2007. “The origin of vowel balance in Swedish and Norwegian dialects.” In T. Bull, J. Kusmenko, and M. Rießler (eds.), Spraº k og spraº kforhold i Sa´pmi. Berlin: Nordeuropa-Institut der HumboltUniversita¨t 235–258. Kusmenko, J. and M. Riessler 2000. “Traces of Sa´mi-Scandinavian contact in Scandinavian dialects.” In D. G. Gilbers, J. Nerbonne, and J. Schaeken (eds.), Languages in Contact. Amsterdam and Atlanta: Rodopi: 209–224. Lahiri, A., T. Riad, and H. Jacobs 1999. “Diachronic prosody.” In H. van der Hulst (ed.), Word Prosodic Systems in the Languages of Europe. Berlin and New York: Mouton de Gruyter: 335–422. Lavoie, L. 2001. Consonant Strength: Phonological Patterns and Phonetic Manifestations. New York: Garland Publishing. Liberman, M. 1975. The Intonational System of English. Ph.D. dissertation, MIT, Cambridge, MA. Published in 1979 by Garland Press, New York.

The Role of Foot Structure in Germanic

Liberman, M. and A. Prince 1977. “On stress and linguistic rhythm,” Linguistic Inquiry 8: 249–336. Lo¨hken, S.1997. Deutsche Wortprosodie: Abschwa¨chungs- und Tilgungsvorga¨nge. Tu¨bingen: Stauffenburg Verlag. McCarthy J. and A. Prince 1995. “Prosodic morphology.” In J. Goldsmith (ed.), The Handbook of Phonological Theory. Cambridge, MA: Blackwell: 318–366. Murray, R. and T. Vennemann 1983. “Sound change and syllabic structure in Germanic phonology,” Language 59: 514–528. Perridon, H. 2002. “The quantity shift in North Germanic,” Amsterdamer Beitra¨ge zur a¨lteren Germanistik 56: 69–77. Philipp, M. and A. Bothorel-Witz 1989. “Low Alemannic.” In Russ (ed.): 313–336. Prince, A. and Smolensky, P. 1993. “Optimality theory: Constraint interaction in generative grammar.” Unpublished ms., Rutgers University and University of Colorado, Boulder. Raymond, W., R. Dautricourt and E. Hume 2006. “Word-internal /t,d/ deletion in spontaneous speech: Modeling the effects of extra-linguistic, lexical, and phonological factors,” Language Variation and Change 18: 55–97. Riad, T. 1992. Structures in Germanic Prosody: A Diachronic Study with Special Reference to the Nordic Languages. Dissertation, Stockholm University. Russ, C. (ed.) 1989. The Dialects of Modern German: A Linguistic Survey. Stanford University Press. Scho¨nfeld, H. 1989. “East Low German.” In Russ (ed.): 91–135. Smith, L. C. 2007. “Old Frisian vowel balance and its relationship to West Germanic apocope and syncope.” In R. Bremmer, S. Laker, and O. Vries (eds.), Advances in Old Frisian Philology, Amsterdamer Beitra¨ge zur a¨lteren Germanistik, Vol. 64 / Estrikken-Ålstraº ke, Vol. 81. Amsterdam and New York: Rodopi: 379–410. Smith, L. C. 2007. “The resilience of prosodic templates in the history of West Germanic.” In J. Salmons and S. Dubenion-Smith (eds.), Historical Linguistics 2005: Selected Papers from the 17th International Conference on Historical Linguistics, Madison, July 31–August 5, 2005. Amsterdam: John Benjamins: 351–365. Smith, L. C. 2009. “Dialect variation and the Dutch diminutive: Loss, maintenance and extension of prosodic templates.” In M. Dufresne, F. Dupuis, and E. Vocaj (eds.), Historical Linguistics 2007. Amsterdam: John Benjamins: 37–46. Smith, L. C., C. Champenois, C., and K. Schuhmann 2016. “The role of prosody in shaping German plurals.” Paper presented at the Germanic Linguistics Annual Conference (GLAC) 2016, Reykjavik, Iceland. Smith, L. C., D. Holsinger, and J. Salmons 2005. “The limits of perceptual distinctness: Evidence from West Germanic.” Paper presented at the

71

72

L A U R A C AT H A R I N E S M I T H

Germanic Linguistics Annual Conference (GLAC) 2005, University of California–Davis. Smith, N. and K. van Leyden 2007. “The unusual outcome of a level-stress situation: In the case of Wursten Frisian.” The North-western European Language Evolution 52: 31–66. Vennemann, T. 1988. Preference Laws for Syllable Structure and the Explanation of Sound Change: With Special Reference to German, Germanic, Italian, and Latin. Berlin, New York, and Amsterdam: Mouton de Gruyter. Versloot, A. 2008. Mechanisms of Language Change: Vowel Reduction in 15th Century West Frisian. Unpublished Ph.D. dissertation, Rijksuniversiteit Groningen. Wiese, R. 2001. “How prosody shapes German words and morphemes,” Interdisciplinary Journal for Germanic Linguistics and Semiotic Analysis 6: 155–184.

Chapter 4 Word Stress in Germanic Birgit Alber

4.1 Germanic Languages and the Typology of Word Stress Metrical theory and the typological literature based on it, identify a series of parameters according to which word stress varies in the world’s languages (see, among others, Liberman and Prince 1977; Hayes 1981, 1995; Prince 1983, 1990; Hammond 1984; Halle and Vergnaud 1987; Kager 1989, 1993, 2007; van der Hulst 1999, 2010; Hyde 2001, 2002, 2016; Gordon 2002; Alber 2005; McManus 2006). A summary of the main features of metrical parametrization might run as follows: (1)

Parameters classifying metrical systems a. position of main stress: leftmost or rightmost b. iterativity of metrical parsing: one foot or many feet c. edge orientation of secondary stress parsing: towards the left or the right edge of the word d. foot-type: trochaic or iambic e. strict binarity: strictly binary feet or also ‘degenerate’ feet f. quantity-sensitivity: stress placement is sensitive or insensitive to syllable quantity g. Nonfinality (extrametricality) effects: no / yes

Even though it has been shown that these parameters are not necessarily on or off, for a certain language, but that languages may e.g., be quantitysensitive in certain contexts, while they are quantity-insensitive in others (see, e.g., Section 4.2.1), the parameters still serve as a useful tool to describe the general pattern of word stress in a certain language or language family. Here, the contrasts implemented by the parameters are represented using a notation where Y stands for the main stress syllable, X for a secondarily stressed syllable, u for an unstressed syllable parsed into

74

BIRGIT ALBER

a foot, and o for an unparsed syllable, of any weight.1 Foot boundaries are indicated with round brackets. When syllable quantity is at play, we distinguish between L (=light) and H (=heavy) syllables, which can be stressed (ˈL, ˈH), or unstressed (L, H). Whenever necessary, we will distinguish between main stressed (ˈL, ˈH) and secondarily stressed (ˌL, ˌH) light and heavy syllables. Main stress in examples is indicated with an acute (´) accent, secondary stress with a grave (`) accent. (2) a. b. c. d. e. f.

Main typological distinctions and their representation

main stress: L / R iterativity: No / Yes edge orientation: L / R foot-type: T / I strict binarity: Yes / No quantity-sensitivity: No / Yes g. Nonfinality (extrametricality) effects: No /Yes

leftmost: no: left: trochaic: binary only: no:

(Yu) o o o o o o o (Yu) (Xu) (Xu) o (Yu) (Xu) (ˈLL) L (ˈLL) (ˈLH) L

rightmost: yes: right: iambic: also ‘degenerate’: yes:

o o o (Yu) (Xu) (Xu) (Yu) o (Xu) (Xu) (Yu) (uX) (ˈLL) (ˈL) (ˈLL) L (ˈHL)

no:

. . . o (Y) #

yes:

. . . (Y) o #

There are languages which place main stress on a syllable close to the left edge of the word, while others place it close to the right edge (2a). Some languages are described as assigning only one stress per word (the main stress), while others are reported to assign also one or more secondary stresses (2b). If a language parses a series of secondary stresses, these can be closer to the left or closer to the right edge of the word (2c). The orientation of feet toward one or the other edge has been analyzed in the framework of Optimality Theory in terms of “alignment” (McCarthy and Prince 1993, Hyde 2012) and this is how we will describe edge-orientation here. The type of feet that a language uses for metrical parsing typically is either left-headed (trochaic) or right-headed (iambic) (2d). Languages may either make use of strictly binary feet, i.e., feet consisting of two syllables (Xu or uX) or of two moras (H), or they may also allow for so-called ‘degenerate’ feet, consisting of a single light syllable (L) (2e).2 Languages may be quantitysensitive or quantity-insensitive (2f). In a quantity-sensitive language, heavy syllables typically interrupt the default metrical parse of the language attracting stress onto them. In a quantity-insensitive language, heavy syllables do not have any stress-attracting force and behave similarly to light syllables, with respect to metrical parsing. Languages also differ in what they consider to be a heavy syllable. In some quantity-sensitive languages only syllables containing a long vowel or a diphthong (CVV syllables) count as heavy, while in others also closed CVC syllables condition metrical parsing (see Hayes 1995). Finally, languages can exhibit Nonfinality (extrametricality) effects (2g). In traditional approaches (e.g., Hayes 1995), a syllable (or, in certain cases, a consonant or a foot) is considered 1

See Alber et al. (2016) for a similar representation of metrical structure.

2

We will assume here that feet are at most binary, and not consider the possibility of e.g., ternary feet.

Word Stress in Germanic

extrametrical if it is excluded from prosodic structure altogether and does not count for the placement of main stress. In more recent approaches in the framework of Optimality Theory, extrametricality effects are interpreted as the result of the activity of Nonfinality constraints requiring the head of the prosodic word (i.e., either the main stress syllable or the main stress foot, or both) not to be final (Prince and Smolensky 2004 [1993]; see Hyde 2007 for an extension of the family of Nonfinality constraints). Nonfinality effects are only attested at the right edge of the word, thus they become relevant only when main stress is rightmost. The parameters define the basic stress pattern of a language for underived words. Morphology may add an additional layer of complexity to a system, with affixes being stress neutral, pre/poststressing or stress attracting, but before considering the prominence system of morphologically complex words, the basic pattern in the underived words of a language has to be identified. All these parameters refer to systems in which stress assignment is fully predictable. There are, however, languages in which stress is assigned lexically, i.e., its position cannot be fully predicted by any rule or algorithm, but is (in part or completely) specified in the underlying representation of the word. The Germanic languages exhibit clear preferences for some of the metrical parameters, while there is still no complete consensus in the literature as to the setting of others. There is no doubt that with respect to the parameter of foot-type (2d), the Germanic languages have to be considered of the trochaic type (but see Burzio 1994, who argues for dactylic feet in the analysis of English). For most of them it has been claimed that there is a pattern of iterative secondary stress (2b), and although not much data is available for some of them, the available data point to a left edge-orientation of secondary stress feet (2c). Most Germanic languages have been described as quantitysensitive (2f), at least in certain contexts (cf. English, Dutch, German, Norwegian below), but for some of them this parameter has not been explored in any detail (e.g., Danish). No Germanic language allows for degenerate feet (but see the discussion of Icelandic below) (2e). As to parameter (2a), in most Germanic languages, main stress placement is not completely predictable. In the ancestor languages of the modern Germanic languages, main stress was placed on the leftmost syllable of roots, which most often corresponded to the first syllable of the word (Lahiri et al. 1999). After contact with languages of the Romance type and subsequent incorporation of vast numbers of loanwords into the lexicon, stress assignment in all Germanic languages except Icelandic and Faroese has to be considered of the rightmost type. Together with rightmost main stress, extrametricality effects have become part of the system (parameter 2g). In what follows we will assume as uncontroversial the setting of foottype (position of head): Trochaic and Degenerate feet: no for all Germanic

75

76

BIRGIT ALBER

languages. We will then concentrate on the remaining typological traits, the placement of main stress and concomitant Nonfinality effects, the presence and orientation of secondary stress, and the issue of quantity sensitivity. First, the metrical system of the Germanic languages that have opted for righmost main stress will be discussed (Danish, Dutch, English, German, Norwegian, Swedish, Section 4.2). We will discuss main stress placement and extrametricality effects (Section 4.2.1) in these languages, then secondary stress and its orientation (Section 4.2.2). Quantitysensitivity is discussed both under 4.2.1 and 4.2.2, as the issue arises. Second, we reserve a separate section for the two languages with leftmost main stress, Icelandic and Faroese (Section 4.3).

4.2 Germanic Languages with Rightmost Main Stress 4.2.1 Placement of Rightmost Main Stress All Germanic languages, to varying degrees, have undergone historical processes of weakening and deletion of unstressed vowels, which in large parts of native roots have led to erosion of all syllables except the initial, main stressed syllable, at most accompanied by a second, usually light syllable (for syllable structure in Germanic languages, see van Oostendorp, Chapter 2). The result are languages where, in the Germanic part of the lexicon, roots rarely are more than one or two syllables long and stress falls on the only syllable or the first of them. Taking German as an example, Golston and Wiese (1998) note that 79 percent of German root morphemes consist of a single, heavy syllable, i.e., a syllable containing a long vowel or closed in a consonant. Another 19 percent of roots consist of a heavy syllable followed by a second syllable containing schwa and, in some cases, a sonorant, which can become syllabic. Assuming that these second syllables are light, the typical German root morpheme has either the structure H or HL (see Basbøll 2005: 395 for similar observations with respect to Danish, Rice 2006 for Norwegian and Dresher 2013 for English; Alber 2001 for discussion of the template structure of German roots). Native affixes often have the same shape. (3)

German native root morphemes a. H: [zeː] ‘lake’, [huːt] ‘hat’, [praxt] ‘splendor’ b. HL: [aʊgə] ‘eye’, [beːzən/beːzn̩ ] ‘broom’, [zeːgəl/zeːgl̩ ] ‘sail’

The position of main stress placement in a trochaic language with mostly monosyllabic, at most bisyllabic morphemes is ambiguous. Main stress assignment will inevitably fall on the only, or on the first of two syllables. This (ˈH), (ˈHL) structure can then be interpreted as an instance of leftmost, but also as an instance of rightmost main stress assignment.3 3

If the first of the two syllables is heavy, the second light, as happens in German roots, an alternative (ˈH) L parse can be imagined. Still, also under such a parse, main stress can be interpreted as leftmost, but also as rightmost, in the limits of the restrictions of quantity-sensitive parsing and / or Nonfinality.

Word Stress in Germanic

The ambiguity with respect to parameter (2a) makes the Germanic languages particularly suited for an encounter with languages of the Romance type, where main stress is assigned rightmost. In all Germanic languages a large part of the lexicon is formed by loanwords, mostly borrowed from Latin, French, or Greek (the latter often via Latin). These loanwords are typically polysyllabic and have rightmost main stress. French loanwords contribute a pattern with final stress. Latin loanwords or loanwords from other Romance languages such as Italian come with the typical Latin pattern where stress falls on a heavy penultimate syllable, if there is one, else on the antepenultimate syllable. The latter pattern can be analyzed (as Latin stress usually is), as characterized by Nonfinality / extrametricality, banning the main stress foot from final position, and placement of a moraic trochee as close as possible to the right edge of the word. (4)

a. Typical Germanic pattern: (ˈH), (ˈHL) b. Typical Romance patterns: final stress: . . ..(ˈH)# penultimate (Latin) stress: . . ..(ˈH) o # antepenultimate (Latin) stress: . . ..(ˈHL) o / (ˈLL) o#

The encounter between Germanic and Romance patterns has had effects that differ across the Germanic languages. On the one hand there are Icelandic and Faroese, where initial, Germanic stress has been imposed also on (older) Romance loanwords, shifting the rightmost stress those words had in the donor languages. However, the other Germanic languages have chosen another possibility: They have reinterpreted the parameter of main stress placement as ‘rightmost’.4 A choice of this type preserves rightmost stress in loanwords. At the same time it is compatible with the stress patterns in the Germanic part of the lexicon. Linked to the integration of a rightmost main stress pattern is the question of Nonfinality (extrametricality) restrictions, a systematic feature in the major donor language Latin. By incorporating Romance patterns, Nonfinality effects became part of the metrical system of the Germanic languages. Yet, given the other major donor language, French, also words with final stress became part of the lexicon. Thus, due to the heterogeneity of the loanword vocabulary (some directly from Latin, some from Latin daughters such as French or Italian), Nonfinality effects did not become completely predictable, in the modern Germanic languages. The result is metrical systems where main stress is characterized by some clear regularities on the one side, but lexical idiosyncrasies on the other. The hybrid nature of main stress placement, partly governed by rules, partly lexicalized, has brought analysts to formulate algorithms for main 4

See Zonneveld et al. (1999), Speyer (2009), and Dresher (2013) for the same point. The alternative would be to assume that Germanic languages make use of two metrical grammars, one with leftmost main stress placement for the native lexicon, one with rightmost main stress placement for the Romance part of the lexicon (see discussion of the two possibilities in Jessen 1999: 516, with respect to German).

77

78

BIRGIT ALBER

stress placement which, inevitably, must leave room for exceptions, while it has convinced others that main stress placement should be considered altogether lexicalized (see e.g., Riad 2013 for Swedish underived words or Basbøl 2005: 386, 395f., who assumes that stress in Danish can be part of the lexical specification of the syllable). The partial unpredictability of main stress placement becomes particularly clear when considering variation within the same language. Thus, in regional varieties of German some lexical items vary as to placement of main stress inside a final three-syllable window. In most varieties of German, Mathematı´k has final stress, but in Austrian German it has penultimate stress (Mathema´tik). Te´lefon has mostly antepenultimate stress, except in Austria, where it is stressed on the final syllable (Telefo´n). Labo´r is predominantly stressed on the final syllable, except in Switzerland and the region of Vienna, where we also find penultimate stress (La´bor) (Kleiner 2011–2017). Other words, such as Kı´mono / Kimo´no are given by pronunciation dictionaries with variable stress, without indicating any regional distribution (Duden 2005). However, unpredictability has its limits. Among the clearest regularities characterizing all Germanic languages with rightmost main stress is the restriction of main stress to a final three-syllable window.5 In Danish, Dutch, English, German, Norwegian, and Swedish, main stress falls predominantly on the final, the penultimate or the antepenultimate syllable of the word. (5)

Main stress inside a final three-syllable window

Danish Dutch English

final balko´n lokomotı´ef ballyho´o

German Hermelı´n Norwegian orkide´ Swedish exege´t

penultimate memora´ndum heliko´pter horı´zon

antepenultimate le´ksikon ho´spitaal Ame´rica

Harpu´ne ko´kos pyja´mas

Rı´siko Ame´rika ele´ktriker

source Basbøl 2005 Kager 1989 Liberman and Prince 1977 Wiese 1996 Rice 2006 Riad 2013

Besides reference to a final three-syllable window, Nonfinality or extrametricality restrictions play a role in most analyses of Germanic main stress. If it is assumed that feet are at most binary, then antepenultimate stress can only be explained if the final syllable is not part of the main stress foot. However, accounts vary as to which prosodic units are considered to be extrametrical (the final syllable? the final foot? the final consonant?). Most analyses furthermore assume that syllable weight does play some role in main stress placement. But there is often no agreement as to which syllables have to be considered heavy and which light, in a certain language. While in the analysis of English stress, syllables containing a long 5

There are some known exceptions to this generalization, as e.g., preantepenultimate stress in English cá tamaran (Liberman and Prince 1977: 276) or German Ábenteuer ‘adventure’ (Jessen 1999: 520, 528f.).

Word Stress in Germanic

vowel or closed in a consonant are usually considered to be heavy, and all other syllables light, for other Germanic languages assumptions vary wildly. Thus, for instance, many analyses of Dutch and German assume that only syllables closed in a consonant count as heavy, but some propose that only superheavy syllables (e.g., VVC or VCC syllables), should count as heavy (Fe´ry 1995, 1998). Finally, given the undeniable lexical idiosyncrasies of main stress placement, accounts vary as to which (classes of) patterns are considered exceptional and which represent the default. Considering the different assumptions about Nonfinality / extrametricality, syllable weight and lexical exceptionality, comparison between accounts becomes difficult. Here we summarize Hayes’ (1981) analysis of English, as a representative of a rule-based analysis incorporating the three aspects of Nonfinality / extrametricality, quantity-sensitivity and lexical exceptionality observable in all Germanic languages discussed in this section. Similar accounts could probably be devised (and, in fact, have been proposed) for other Germanic languages.6 We will then summarize a recent study offering new insights into the structure of stress pattern data in Dutch, English, and German (Domahs et al. 2014), followed by two constraint-based proposals as to how lexical exceptionality can be incorporated into a single grammar, either by employing indexed constraints (Alber 1997a for German) or by referring to the structure of the input in terms of heavy and light syllables (Rice 2006 for Norwegian). Hayes (1981) proposes a set of ordered rules for main stress placement in English which falls into a class of general rules and a class of rules referring specifically to stress in nouns. Put differently, the analysis distinguishes between two lexical classes, nouns on the one side, all other words on the other, differing in their stress pattern. As a matter of fact, the general class includes almost all verbs, the specific class most, but not all nouns. Other words may fall into either of the two classes (see Burzio 1994: 43 for a summary in similar terms). The rules can be summarized as follows: (6)

General class (typical for verbs) – consonants are extrametrical – main stress falls on a heavy final, if there is one; else on the penult Specific class (typical for nouns) – syllables are extrametrical – stress falls on a heavy penult, if there is one; else on the antepenult

6

For description and analysis of the metrical system of Germanic languages assigning rightmost main stress, see, among others: for Danish: Basbøl (2005); for Dutch: van der Hulst (1984), Kager (1989), Booij (1999), Trommelen and Zonneveld (1999); for English: Chomsky and Halle (1968), Liberman and Prince (1977), Giegerich (1985), Halle and Vergnaud (1987), Kager (1989), Burzio (1994), Hammond (1999); for German: Wurzel (1980), Giegerich (1985), Vennemann (1990), Eisenberg (1991), Kaltenbacher (1994), Wiese (1996), Féry (1995, 1998), Alber (1997a, 1997b), Gaeta (1998), Jessen (1999), Janßen (2003), Knaus and Domahs (2009); for Norwegian: Lorentz (1996), Rice (1999, 2006), Kristoffersen (2000); for Swedish: Bruce (1999), Riad (2013); see Domahs et al. (2014) for stress in Dutch, English, and German.

79

80

BIRGIT ALBER

(7)

Examples of English main stress placement (Hayes 1981: 150f.) a. General class mole´s . . . (ˈH)

deve´lo

. . . (ˈLL)

obe´y . . . (ˈH)

Arizo´ . . . (ˈH)

age´nda . . . (ˈH)

b. Specific class la´by (ˈLL)

Hayes’ (1981) analysis incorporates all the typical features mentioned for Germanic rightmost main stress above. The rules determine that stress must fall on one of the last three syllables of the word. Syllable weight does play a role in attracting stress to a final or a penultimate syllable. Extrametricality is a prominent feature in the stress pattern of both lexical classes, assuming that final consonants do not contribute to syllable weight and, specifically for nouns, that final syllables are not part of the prosodic structure. Lexical idiosyncrasies are in part incorporated into the account by assuming that specific word classes (nouns) are subject to specific stress rules. Nevertheless, similarly to other analyses of English stress, the stress pattern of some words remains outside of the core rules of main stress placement. To mention just some: vanı´lla has penultimate stress on a light syllable, even though it is a noun; permı´t has final stress even though the final syllable should be considered light, given consonant extrametricality; in some cases stress is ‘retracted’ beyond a heavy penult as e.g., in de´signate.7 The mixed character of main stress placement becomes most evident in a recent study by Domahs et al. (2014), who carry out an experiment in which Dutch, German, and English speakers assign main stress to trisyllabic nonce words consisting of varying sequences of light, heavy, and superheavy (S, only in final position) syllables such as Binsakaf (HLH), Fekomot (LLH), or Rukolmenk (LHS) (examples of German pseudowords). Their experiment is designed under the assumption that in the languages under investigation open syllables are light, syllables closed in a consonant are heavy and syllables closed in a consonant cluster are superheavy. The possibility that open syllables may be interpreted as potentially heavy by speakers and that vowels might be lengthened under stress, is not considered. The mixed effects regression models and analyses using classification trees of the CHAID type show that although no single default stress rule emerges, there are some clear tendencies governing the assignment of main stress in nonce words. Similar tendencies are found when the same tools of analysis are applied to trisyllabic Dutch, English, and German words in the CELEX corpus. Stress on the penultimate is the overall preferred choice in all three languages, followed by antepenultimate, then final stress in Dutch and English and equally frequent antepenultimate and final stress in German. However, as soon as syllable weight is taken into 7

For discussion of stress retraction in English, see Liberman and Prince (1977), Hayes (1981), Kager (1989), and Burzio (1994).

Word Stress in Germanic

account, it becomes clear that the structure of the final, and to some extent the penultimate syllable, is of crucial importance in determining the probability of assigning stress to a certain position. The mixed effects regression models show that a heavy final syllable significantly increases the probability that stress be assigned to the final position. This effect is strongest in German and less strong in Dutch and English. The probability to assign stress to the final syllable increases even more, if the final syllable is superheavy. Probability of antepenultimate stress increases with the weight of the final syllable as well, even if it remains below 60 percent for the three languages. If, on the other hand, the final syllable is light, in Dutch and German there is a high probability that stress will fall on the penultimate syllable. English, on the other hand, is more sensitive to the weight of penultimate syllables: if the penultimate is heavy, the probability that stress falls on it, increases. We interpret Domahs et al.’s (2014) findings as the emergence of two basic patterns of main stress assigment, correlated to the weight of the final syllable: final or antepenultimate stress if the final syllable is heavy (a) and penultimate stress if the final syllable is light (b). In addition to these tendencies, there is a third pattern (c) in English assigning penultimate stress to the penult if it is heavy: (8)

Preferred patterns of main stress assignment in Dutch, German, and English: a. (Xu)(ˈH) or (Yu)(ˌH): main stress on final or antepenultimate b. o (YL): main stress on penult c. o (ˈHu): main stress on heavy penult

The emerging patterns first of all establish without doubt that the three languages show sensitivity to syllable weight, hence have to be considered quantity-sensitive. This is especially true for English, where a heavy penult can attract stress. Patterns a and b, rather than exhibiting stress attraction by heavy syllables, show that another principle active in metrical systems might be at play here, the principle of exhaustive parsing. A three-syllable window with a final heavy syllable can be parsed exhaustively into feet. If main stress falls on the final syllable, the two preceding syllables can be parsed into a secondary stress foot (German: Vı`tamı´n, (Xu) (ˈH)). If main stress falls on the antepenultimate, the final syllable can bear secondary stress and be parsed into a foot of its own (German: A´nana`s, (Yu)(ˌH)). This option of exhaustive parsing is not readily available to structures where the final syllable is light. Thus, e.g., the structures L (ˈLL) and L (ˈHL) (German: Bikı´ni, Age´nda) can parse only one foot, which, if placed rightmost, will result in penultimate stress.8 The fact that exhaustive parsing 8

Structures with final light, but heavy antepenultimate, such as HLL or HHL, could, in principle, be parsed exhaustively as (ˈH) (ˈLL) or (ˈH)(ˈHL), hence lead to penultimate or antepenultimate main stress. These parsings are excluded in Knaus and Domahs 2009 as possible outputs, assuming that they are ungrammatical, because they would create a stress clash. HLL words were not tested by Domahs et al. 2014, but HHL words were, and since they are reported to behave as other words with final light syllables, it can be assumed that indeed an exhaustive parsing causing a stress clash is not available for them.

81

82

BIRGIT ALBER

may play a role in the distribution of main stress is recognized in the analysis of Knaus and Domahs (2009: 1402), cast in the framework of Optimality Theory, where the constraint P A R S E - σ, requiring exhaustive parsing of syllables into feet, crucially prefers an LLH trisyllable to be parsed as (ˌLL)(ˈH) and not as L (ˈLH). If exhaustivity of parsing plays indeed a role in the emergence of the above patterns, the picture should change once four-syllable or longer words are considered and more possibilities to parse strings exhaustively into feet are available. Summarizing, Domahs et al. (2014) show that assignment of main stress is not completely lexicalized in Dutch, German, and English, but rather that certain preferences of stress placement correlated to the weight of the final syllable do emerge, both in nonce word experiments and in the analysis of the CELEX corpus. Besides other regularities already observed in the previous literature (e.g., quantity-sensitivity), their data add the insight that a requirement of exhaustive parsing might play a role in main stress placement, a hypothesis that remains to be tested with longer words. If one of the major patterns common to Germanic main stress placement can be summarized as (Xu)(H) (final or antepenultimate main stress if the final is heavy) and o (YL) (penultimate main stress if the final is light), the question arises as to whether there is a single grammar which can accommodate these patterns. Lexically conditioned stress which obeys certain metrical restrictions, while it violates others, has been analyzed successfully in the framework of Optimality Theory with the help of lexically indexed constraints (e.g., Pater 2000, for English secondary stress). Alber (1997a) analyzes main stress assignment in German as the interaction of the constraint R I G H T M O S T , penalizing every syllable intervening between the main stressed syllable and the right edge of the word, with constraints of the N O N F I N A L I T Y type, requiring the head of the prosodic word not to be final. The interaction between R I G H T M O S T and N O N F I N A L I T Y constraints guarantees the creation of a final threesyllable window. The fact that main stress is in part lexicalized is accounted for by indexing R I G H T M O S T for classes of words and ranking the indexed R I G H T M O S T constraints in different orders with respect to the N O N F I N A L I T Y constraints. An analysis of this type incorporates the ideas of previous analyses, such as Hayes’ (1981), that there are classes of words which differ in the degree to which they are subject to Nonfinality/extrametricality effects. Applying Alber’s (1997a) analysis to the findings of Domahs et al. (2014), along the lines of the analysis proposed in Knaus and Domahs (2009), we can derive the preferred patterns of Germanic main stress assignment within a single grammar containing two indexed rightmost constraints R I G H T M O S T X and R I G H T M O S T Y, ordered in different hierarchical positions with respect to a N O N F I N A L I T Y constraint penalizing structures where main stress falls on

Word Stress in Germanic

the final syllable.9 The interaction is illustrated in the following tableau, using examples from German: (9)

Germanic main stress assignment using indexed constraints

/Ananas/{class Y}

PARSE-σ RIGHTMOSTX NONFINALITY RIGHTMOSTY **

→ (a) A´nana`s (ˈLL) (ˌH)

(b) A`nana´s (ˌLL) (ˈH) (c) Ana´nas L (ˈLH)

*! *!

*

/Vitamin/ {class X} → (a) Vı`tamı´n (ˌLL) (ˈH)

*

(b) Vı´tamı`n (ˈLL) (ˌH) (c) Vita´min L (ˈLH)

**! *!

*

*

(*)

(*)

*

(**!)

(**!)

/Bikini/ {class X or Y} → (a) Bikı´ni L (ˈLL) (b) Bı´kini (ˈLL) L #

Every class of words is evaluated by the R I G H T M O S T constraints to which it is indexed. A´nana`s, of class Y, is indexed to low-ranked R I G H T M O S T Y. Domination of this constraint by N O N F I N A L I T Y will ensure the preference of antepenultimate over final stress. Penultimate stress is excluded because it prevents the possibility to exhaustively parse syllables into feet. Stress to the left of the final three-syllable window is excluded as well, since there is no constraint which could move stress further leftward than the antepenultimate. In the case of Vı`tamı´n, a class X word, the relevant R I G H T M O S T X constraint is ranked above N O N F I N A L I T Y , thus driving the preference for final stress. A word such as Bikı´ni will violate P A R S E - σ whichever way it is parsed, assuming that feet are trochaic and at most binary. Furthermore, given that the final syllable is light, stress cannot fall on it and N O N F I N A L I T Y will be obeyed in any case. This means that words of this type could be members of class X or class Y, yet they will always bear penultimate stress. 9

N O N F I N A L I T Y , in this case, is specified for the main stressed syllable, not for the main stress foot (cf. Prince and Smolensky 2004 [1993]).

83

84

BIRGIT ALBER

There are still some issues that an analysis of this type leaves open, foremost with respect to the role that exhaustive parsing may play in longer words. A quadrisyllabic string L L L H, in order to be exhaustively parsed, would prefer a metrical structure such as (ˌLL)(ˈLH), running against the generalization of Domahs et al. (2014) that words with a final heavy syllable are stressed on the final or antepenult. Whether the predicted penultimate pattern in quadrisyllabic words is real can only be determined once analyses of similar detail as carried out by Domahs et al. (2014) for trisyllabic words are replicated for longer words.10 The advantage of analyses in terms of indexed constraints is that within a single grammar they can account both for the overall regularity of a certain phenomenon and for its lexical idiosyncrasies. Thus, the analysis above accounts for the fact that main stress does not fall on any syllable in the word, but strictly remains inside a final three-syllable window and that the weight of the final syllable plays a crucial role in determining stress, while still incorporating the fact that part of the phenomenon is lexically determined. A different answer to the problem of where stress is placed inside the three-syllable-window is given by Rice (2006) in his analysis of Norwegian stress. Rice places the main burden of the choice between final, penultimate, and antepenultimate stress on faithfulness constraints preserving input moras (M A X - μ), which interact with a N O N F I N A L I T Y constraint (specified for the main stress syllable), the Weight-to-Stress principle (WSP: heavy syllables are stressed) and R I G H T M O S T . Rice (2006) assumes that in Norwegian syllables containing a long vowel and syllables closed in a consonant are heavy. Furthermore, he observes that, as in many Germanic languages, in Norwegian stressed syllables must be heavy, a requirement often described as Prokosch’s Law or, in terms of violable constraints, as the result of a dominant Stress-toWeight principle (SWP: stressed syllables are heavy; see, e.g., Lorentz 1996 for discussion). Finally, Rice assumes that in Norwegian word-final consonants are not syllabified as codas of the final syllable, but, rather, form the onset of a catalectic syllable. This means that the final syllable in a word such as e´ddi.k ‘vinegar’ is light. We will indicate extrasyllabic consonants of this type with a dot preceding them. Final stress is encountered in words such as orkide´. In Rice’s (2006) analysis, this word has final stress because the final vowel is specified in the input as long. The two moras of the final vowel have to be preserved in the output metrical parsing (M A X - μ) and the resulting heavy syllable has to receive stress (WSP), overriding such constraints as N O N F I N A L I T Y . Also in traf ´ı k.k, where the final syllable is heavy because closed in a consonant, stress has to be final. 10

Additional evidence that exhaustive parsing might indeed play a crucial role in determining the metrical structure of German comes from analysis of impaired speech. Janßen and Domahs (2008) report that a patient affected by primary progressive aphasia shows clear tendencies to “regularize” words by inserting or deleting syllables in such a way that exhaustive parsing is guaranteed. Thus, for instance, [maˈrɪlə], analyzable as L (ˈLL), is realized, in a repetition task, as [maləˈrɪlə], which can be exhaustively parsed as (ˌLL)(ˈLL).

Word Stress in Germanic

Final stress, however, can also be the result of an input where the two final syllables are light, as is assumed to be the case in toma´.t. In this case, the trigger for final stress is R I G H T M O S T , assigning stress as much to the right as possible. N O N F I N A L I T Y is obeyed, in this case, because the final foot in to(ma´:).t is supposed to be separated from the right edge by the final, extrasyllabic consonant. The final vowel is lengthened in order to obey the SWP. Penultimate stress arises when the penultimate syllable is heavy, but the final syllable is light, since it is more important to assign stress to the penultimate, heavy syllable (WSP) than to satisfy R I G H T M O S T . This is the case in e´ddi.k ‘vinegar’. Stress is placed on the penultimate also in ko´ko.s ‘coconut’, where the vowel of the penultimate syllable is assumed to be specified as long, underlyingly. Finally, penultimate stress is chosen for bisyllabic native words, as, e.g., hake ‘chin’, which always have the output structure HL. Although the input vowel could, in principle, be short, stress has to be penultimate, in this case, because of N O N F I N A L I T Y – the final syllable is not protected by a final, extrametrical C, in this case. Antepenultimate stress, as in Ame´rika, is the result of an HLL input, where stress is attracted to the antepenultimate heavy, whose weight has to be preserved, in order to obey the WSP. (10)

Germanic main stress assignment using faithfulness constraints to underlying moras

/orkiμdeμμ/

SWP MAX-μ WSP NONFINALITY

→ (a) orkiμde´μμ

RIGHTMOST

*

. . . L (ˈH)

(b) orkı´μμdeμμ

*!

*

. . . (ˈH) H

(c) orkı´μμdeμ

*

*

. . . (ˈH) L

/toμmaμt/ → (a) toμma´ μμ.t

*

L (ˈH).C

(b) to´μμmaμ.t

**!

(ˈH) L.C

/edμμdiμk/ → (a) e´dμμdiμ.k

**

(ˈH) L.C

(b) edμμdı´μμ.k

*!

*

*

H (ˈH).C

/haμkeμ/ → (a) haμμkeμ

*

(ˈH) L

(b) haμkeμμ L (ˈH)

*!

85

86

BIRGIT ALBER

Much of the burden of stress placement in Rice’s (2006) analysis is carried by the stress-attracting force of heavy syllables, which are either assumed to have an underlying long vowel or have a consonant closing the syllable. Words with final light syllables fall into two categories: If the final syllable is closed in a consonant, the final ‘extrametrical’ consonant protects it from incurring a N O N F I N A L I T Y violation and stress can be final (toma´t). If the final syllable is open, stress has to retreat to the penult (ha´ke). Words with final consonant clusters remain a problem for the analysis. In these words the final syllable is heavy, hence stress should be final, as in fact it often is (e.g., agu´r.k). But there are words with final syllables closed in a consonant cluster which have penultimate stress, such as bo´rak.s. Even if we assume that the penultimate syllable is underlyingly bimoraic (i.e., the words have an HH structure), in these cases, stress should still fall on the final syllable, optimizing R I G H T M O S T . Placement of main stress inside the final three-syllable window thus depends on the input structure of the word, in terms of strings of heavy and light syllables. (11)

Summary of input-output mappings in the analysis of Norwegian main stress (Rice 2006) interpretation of donor stress

input output

L (ˈH) orkide´, traf ı´k.k as H L (ˈH).C toma´.t not necessarily as H

final stress

LH LL.C

penultimate stress

HL.C (ˈH) L.C ko´ko.s e´ddi.k, as H LL (ˈH) L ha´ke –– HH.C H (ˈH).C agu´r.k, but not: as H bo´rak.s

antepen. stress

HLL

(ˈH)L L

Ame´rika

as H

Note that in the case of open syllables, the fact that a certain vowel is underlyingly long and thus contributes to syllable weight, is an assumption. In surface forms, all vowels in open syllables are long, since all stressed syllables must be heavy by the SWP. Thus nothing in the surface form tells us that orkide´ has an underlyingly long vowel in final position, while the underlying final vowel in toma´.t is short. Yet, the assumption that certain open syllables contain underlyingly long vowels, while others are light, is far from arbitrary, since it can be linked to stress in the donor language. Norwegian (as Danish, Dutch, German, and Swedish) are languages where stress is usually

Word Stress in Germanic

preserved in the same position as in the donor language.11 We can therefore assume that donor stress is interpreted as weight in Norwegian (and maybe in other Germanic languages as well). This assumption fits all of the above cases, except the word toma´.t, which is assumed by Rice to have an underlying light final syllable. However, final stress in toma´.t is also compatible with an underlying LH.C structure, which would be compatible with the hypothesis of a stress-weight correspondence. Summarizing, Rice (2006) attributes the placement of main stress in the three-syllable window to input specifications, in terms of weight, of the last three syllables of the word. This is equivalent to acknowledging that stress is lexically conditioned, to some extent, but it is not equivalent to saying that stress is prespecified. Rather, if we assume that stress in the donor language is reinterpreted as weight in Norwegian, the different resulting sequences of light and heavy syllables are evaluated in one single grammar, containing standard metrical constraints, which generate also the stress pattern of native words. Thus also under this approach there is a single grammar which nevertheless can account for the partially lexicalized nature of Germanic main stress. In the proposed grammar, as in the other accounts of main stress placement summarized above, constraints requiring a correspondence between weight and stress and constraints triggering Nonfinality effects play a dominant role. It remains to be seen whether an analysis of this type can integrate the insight of Domahs et al. (2014) that final heavy syllables favor not only final, but also antepenultimate stress – assuming that this generalization also holds for Norwegian.

4.2.2 Placement of Secondary Stress In the literature on stress in Germanic languages, iterativity and directionality of rhythmical secondary stress (parameters exemplified in 2b and 2c in Section 4.1) have been discussed in less detail than main stress placement. This may be due to the fact that phonological or acoustic correlates of secondary stress are often more difficult to determine than those of main stress, leading sometimes even to statements that secondary stress is absent in a certain language (e.g., Moulton 1962; Fe´ry 1995 for German). The major exception here is English, where secondary stress is detected more readily, given that unstressed vowels are systematically reduced. Correlates for secondary stress, however, are sometimes indicated also for other Germanic languages. In East Norwegian varieties, a stress shift can occur to an

11

This is not generally true for English. See Dresher (2013) for Middle English where, differently from later stages of English, stress of the donor language was not preserved. Neither did Icelandic and Faroese preserve main stress of the donor languages (see Section 4.3).

87

88

BIRGIT ALBER

initial, secondarily stressed syllable (pro`teste´re → pro´teste`re ‘to protest’), but not to an unstressed syllable (beto´ne → *be´tone ‘to accentuate’; Kristoffersen 2000: 165). In Dutch, unstressed syllables can be reduced to schwa, differently from secondarily stressed syllables (Kager 1989: 278). For German, Alber (1997b) mentions a process of optional insertion of glottal stops in hiatus contexts affecting secondarily stressed syllables, but not unstressed syllables; Jessen (1999 and references therein) observes that vowels bearing secondary stress preserve some contrast in vowel length, while unstressed vowels do not. Particularly convincing evidence for secondary stress feet comes from recent ERPstudies on German (Domahs et al. 2008, Knaus and Domahs 2009, Knaus et al. 2011). In a stress-evaluation experiment, Domahs et al. (2008) measured brain responses to incorrectly stressed trisyllabic words of the (Xu)(ˈH) (final stress, Vitamı´n) or (Yu)(ˌH) (antepenultimate stress, A´nanas) type. We will call this structure, where the trisyllabic string is parsed into two feet, the (oo)(H) structure. They furthermore tested words of the o (YL) type (Bikı´ni) with penultimate stress and a single parsed foot. They found that the strength of late positivity effects, encountered as a reaction to incorrectly stressed forms, was correlated to whether the violation resulted in restructuring of metrical structure, or not. Thus, wrongly stressed *Bı´kini (ˈLL)L triggered an enhanced late positivity effect, while shifting stress in an (oo)(H) structure from the final syllable to the antepenultimate (*Vı´tamin) or vice versa (*Anana´s) resulted in a much weaker effect. The difference between the two types of violations thus confirms the existence of foot-structure and therefore also the presence of a second foot besides the main stress foot in structures of the (oo)(H) type (see also the discussion in Knaus and Domahs 2009). In a similar study, Knaus et al. (2011) examine brain responses to German compounds such as Aro´ma-Sensibilit’a¨t, where the second member of the compound can be assumed to have a metrical structure of the type (ˌLL)(ˌLL)(Y). The authors observe positivity effects to stimuli wrongly stressed on the second syllable of the second element of the compound, which require restructuring of feet (*Aro´ma-Sensı´bilita¨t = L (YL) L (ˌH)), while no response occurs when main stress is simply shifted from final position to the head of the first secondary stress foot (*Aro´ma-Se´nsibilita¨t = (YL) (ˌLL)(ˌH)). Notwithstanding the fact that the few accounts on secondary stress in Germanic languages other than English and German are mostly based on grammaticality judgments, they still return a rather homogeneous picture as to the setting of parameters (see 2b and 2c in Section 4.1). All accounts contemplating the existence of secondary stress do assume that secondary stress is iterative and most claim that secondary stress feet are aligned to the left.

Word Stress in Germanic

(12)

Iterativity and edge-orientation of secondary stress in the Germanic languages iterativity

edge-orientation source

English

A`pala`chico´la (ˌLL)(ˌLL) Y

Ta`tamago´uchee (ˌLL) L Y

Liberman and Prince 1977

Danish

pa`nora´ma (ˌLL) Y

ma`teria´le (ˌLL) L Y

Ito and Mester 2015: 14

Dutch

o`noma`tope´e (ˌLL)(ˌLL) Y

lo`komotı´ef (ˌLL) L Y

Kager 1989

German

E`nzyklo`pa¨dı´e La`titu`dinarı´smus Alber 1997a, 1997b, 1998 (ˌHL)(ˌLL) Y (ˌLL)(ˌLL) L Y

Norwegian u`nive`rsite´t (ˌLL)(ˌHL) Y

Swedish

e`piste`mologı´ (ˌLL)(ˌLL) L Y u`niversa`lite´t (ˌLL) H (ˌLL) Y

de`mokra`tise´ra o`noma`topoe´tisk (ˌLL)(ˌLL) Y (ˌLL)(ˌLL) L Y

Lorentz 1996: 119, 120

Riad 2013: 14312

Iterativity of secondary stress is detected in sequences of at least four syllables before main stress (i.e., at least pentasyllabic words; no examples of this type have been found for Danish, yet). Only in structures of this type is there room for two secondary stress feet before the main stress foot. The edge-orientation of secondary stress is observed in odd strings of syllables before main stress, where left-alignment places secondary stress feet as close as possible to the left edge of the word. For Norwegian, Lorentz (1996) gives both left-aligning examples (e`piste`mologı´) as well as examples where an initial secondary stress foot is parsed, and a second one is aligned to the right (u`niversa`lite´t). Lorentz (1996) proposes that stress on the third syllable of e`piste`mologı´ is due to influence of the adjective e`piste´misk, i.e., main stress of e`piste´misk ‘epistemic’ would be preserved as secondary stress in e`piste`mologı´. Note, however, that this would be a rather unusual case of stress-preservation across words not linked through cyclic affixation. Usually, main stress of an underived word is preserved as secondary stress if a suffix is attached to the underived word and main stress is shifted. This could indeed be the case in u`niversa`lite´t, derived from u`niversa´l, which, after suffixation by -itet maintains the original final main stress as a secondary stress. We therefore think that 12

Riad describes these Swedish examples as displaying “rhythmical prominences”, distinct from “phonological stress”. He remains agnostic about whether there is a foot structure beyond that of the main stress foot. Stress on the third syllable in ònomàtopoétisk is described as “doubtful or optional” (Riad 2013: 143).

89

90

BIRGIT ALBER

secondary stress can be assumed to be left-aligning also in Norwegian and that e`piste`mologı´ is an example that displays the basic pattern. Besides establishing the main parameters of iterativity and edgeorientation, studies on secondary stress have also contributed to the understanding of the role of quantity-sensitivity and the relevance of lexical stress assignment in secondary stress placement. Pater (2000) shows that while secondary stress in English is clearly quantitysensitive, it is not so in all contexts. This ‘non-uniformity’ of quantitysensitivity can be captured best in a framework like Optimality Theory, assuming that constraints favoring stress on heavy syllables, such as the WSP, dominate some metrical constraints, while they are dominated by others. Thus, in English a heavy syllable can disrupt left-alignment of secondary stress feet (compare Mono`ngahe´la, L (ˌHL) Y to Ta´tamago´uchee (ˌLL) L Y; see also Liberman and Prince 1977: 276). But the WSP cannot impose itself if exhaustive parsing is at stake. Thus, A`lexa´nder, (ˌLH) Y is not stressed on the second, heavy syllable, because the initial syllable otherwise would remain unparsed.13 A similar, context-dependent sensitivity to quantity is observed also in German (Alber 1997a, 1997b), where word-medial heavy syllables can attract main stress (A`utodete`rminı´smus, (ˌHL) L (ˌHL) Y, compare with La`titu`dinarı´smus, (ˌLL)(ˌLL) L Y), while peninitial syllables usually do not (a`malgamı´eren (ˌLH) L Y). Quantity-sensitive parsing, in violation of left-edge orientation, is reported also for Dutch, where peninitial heavy syllables optionally bear secondary stress (kale`idosco´op, L (ˌHL) Y, Kager 1989: 280), violating left edge-orientation. Similarly, in Norwegian, stress is attracted by the second, heavy syllable in giga`ntomanı´e L (ˌHL) L Y (compare with e`piste`mologı´, (ˌLL)(ˌLL) L Y, Lorentz 1996). Pater (2000) observes that not only quantity-sensitivity, but also lexical exceptionality is context-sensitive, in secondary stress placement. He analyzes lexical exceptionality as the result of a faithfulness constraint to underlying stress, which occupies an intermediate position in the constraint hierarchy. For instance, faithfulness to prespecified lexical stress can lead to violations of left-alignment (apo`theo´sis, L(ˌLL)Y). However, there are no cases where a lexically prespecified stress would be preserved, e.g., on the third syllable of a pretonic LLL sequence (*Ta`tama`go´uchee). Such cases of preservation of lexically specified stress are banned because of the dominance of metrical constraints such as F T B I N , requiring strict binarity of metrical feet. Interestingly, the same contexts in which preservation of a lexically specified stress are possible are also those in which preservation of stem stress is possible under suffixation. Thus main stress of ima´gine is preserved as secondary stress in the suffixed form ima`gina´tion, leading to a disruption in the left-alignment of secondary stress feet. 13

Note that parsing cannot be optimized in the case of Monongahela by shifting stress to the initial syllable, since one light syllable will be left unparsed under any (binary) foot-structure that can be assigned to this word. Furthermore, stress on the initial syllable in Àlexá nder cannot be attributed to an avoidance of stress clash, since stress clash is in fact tolerated in other words, such as Hàlicàrná ssus, (ˌLL)(ˌH)Y (see Pater 2000 for discussion).

Word Stress in Germanic

However, stress cannot be preserved on the last syllable of a pretonic LLL string, as, e.g., in ma`thema´tic → *ma`thema`tı´cian. Summarizing, secondary stress in most Germanic languages is still not studied in great detail, even though evidence for its existence increases as new experimental methodologies are applied to metrical structure. The descriptions that we do have point to a pattern of iterative secondary stress characterized by left edge orientation. The more detailed analyses available for English reveal that at least in certain contexts secondary stress is quantity-sensitive, an observation confirmed also for Dutch, German, and Norwegian, and that as main stress, secondary stress is subject to lexical idiosyncrasies.

4.3 Germanic Languages with Leftmost Main Stress: Icelandic and Faroese Icelandic most closely resembles the original Germanic stress pattern, placing main stress consistently on the leftmost syllable of the word in native words and also in older loanwords. It displays an iterative secondary stress pattern, parsing trochaic, left-aligned feet. Syllable quantity does not seem to influence stress placement, hence Icelandic can be classified as a quantity-insensitive system. (13)

Icelandic stress (A´rnason 2011: 271) [ˈaːkvaˌrɛlˑa] (Yu) (Xu) [ˈpijːouˌkraˑfiˌjaˑ] (Yu) (Xu) o

akvarella

‘aquarelle’

bı´o´graf ı´a

‘biography’

There has been some discussion as to whether Icelandic allows for socalled ‘degenerate feet’, consisting of a single, light syllable (Hayes 1995, A´rnason 1999, A´rnason 2011). A´rnason (2011: 273) does assign secondary stress to final syllables of odd-numbered strings, as in bı´ografı´a. However, he acknowledges that the perceived stress might be some form of postlexical or rhythmical strengthening rather than the instance of a secondary stress foot. Hayes (1995: 188f.) attributes prominence in this position to a process of final lengthening. He shows that even though penultimate secondary stress in Icelandic is preserved after metrical restructuring in compounding, final stress in odd-strings is not. This, he argues, is evidence against interpreting final prominence as the head of a degenerate, monosyllabic foot. Faroese has a pattern very similar to that of Icelandic, with main stress placed leftmost and secondary stress of the iterative, leftaligning type. The main difference with respect to Icelandic is that loanwords more readily bear non-initial main stress. According to

91

92

BIRGIT ALBER

A´rnason (2011: 278), loanwords in Faroese are mostly borrowed via Danish, where, as in other Germanic languages, stress falls inside a final three-syllable window. (14)

Faroese stress (A´rnason 2011: 275f.) a. Stress in native words to´mur (Yu) [ˈkʰʊnlaiˌka] kunnleika ( Yu) o [ˈnʊifoeˌjɩɲʧɩˌnʊn] ny´jføðinginum (Yu) (Xu) o

‘empty’ ‘acquaintance-O B L ’ ‘the newly-born- D A T ’

b. Stress in loanwords (A´rnason 2011: 278) [tɩskoˈtʰeːk] diskotek ‘discotheque’ . . . (Y) # [stuˈtɛn̥ tʊɹ] studentur ‘student’ . . . (Yu) # [amɛɹɩˈkʰaːnaɹɩ] amerikanari ‘an American’ . . . (Yu) o # A´rnason notes prominence on the final syllable of odd strings also for Faroese, but given the strong overall resemblance to Icelandic there remain doubts that final prominence should not be considered a correlate of final lengthening also in this case. Icelandic and Faroese thus share with other Germanic languages the setting of the parameters regarding foot-type (trochaic) as well as iterativity (yes) and edge-orientation (left) of secondary stress. As in other Germanic languages, the existence of degenerate feet is at least doubtful. They differ from other Germanic languages in that the position of main stress is leftmost, rather than rightmost and stress in general does not seem to be quantity-sensitive. Since main stress is leftmost, Nonfinality effects, which are attested only at the right edge of the word, do not play any role. In loanwords, Faroese seems to have integrated the rightmost Romance pattern encountered in other Germanic languages, while such a pattern is attested in Icelandic only for recent loans.

References Alber, B. 1997a. Il sistema metrico dei prestiti del tedesco. Aspetti e problemi della teoria prosodica. Ph.D. thesis, University of Padova. Alber, B. 1997b. “Quantity sensitivity as the result of constraint interaction.” In G. Booij and J. van de Weijer (eds.), Phonology in Progress – Progress

Word Stress in Germanic

in Phonology, HIL Phonology Papers III. Holland Academic Graphics: The Hague: 1–45. Alber, B. 1998. “Stress preservation in German loan words.” In Kehrein and Wiese (eds.): 113–114. Alber, B. 2001. “Maximizing first positions.” In C. Fe´ry, A. D. Green, and R. van de Vijver (eds.), Proceedings of HILP 5, University of Potsdam: 1–19. Alber, B. 2005. “Clash, lapse and directionality,” Natural Language and Linguistic Theory 23: 485–542. Alber, B., N. DelBusso, and A. Prince 2016. “From intensional properties to universal support,” Language 92.2: e88–e116. ´Arnason, K. 1999. “Icelandic and Faroese.” In van der Hulst (ed.): 567–596. A´rnason, K. 2011. The Phonology of Icelandic and Faroese. Oxford University Press. Basbøll, H. 2005. The Phonology of Danish. Oxford University Press. Booij, G. 1999. The Phonology of Dutch. Oxford University Press. Bruce, G. 1999. “Swedish.” In van der Hulst (ed.): 554–567. Burzio, L. 1994. Principles of English Stress. Cambridge University Press. Chomsky, N. and M. Halle 1968. The Sound Pattern of English. Harper and Row: New York. Deaton, K., M. Noske, and M. Ziolkowski (eds.), CLS 26-II: Papers from the Parasession on the Syllable in Phonetics and Phonology. Chicago: Chicago Linguistic Society. Domahs, U., I. Plag, and R. Carroll 2014. “Word stress assignment in German, English and Dutch: Quantity-sensitivity and extrametricality revisited,” The Journal of Comparative Germanic Linguistics 17: 59–96. Domahs, U., R. Wiese, I. Bornkessel-Schlesewsky, and M. Schlesewsky 2008. “The processing of German word stress: Evidence for the prosodic hierarchy,” Phonology 25: 1–36. Dresher, E. 2013. “The influence of loanwords on Norwegian and English stress,” Nordlyd 40.1: 55–43. Duden Aussprachewo¨rterbuch 2005. 6. Auflage. Mannheim, Leipzig, Wien, and Zu¨rich: Dudenverlag. Eisenberg, P. 1991. “Syllabische Struktur und Wortakzent: Prinzipien der Prosodik deutscher Wo¨rter,” Zeitschrift fu¨r Sprachwissenschaft 10: 37–64. ´ Fery, C. 1995. “Alignment, syllable and metrical structure in German,” SfSReport-02–95, University of Tu¨bingen. Fe´ry, C. 1998. “German word stress in Optimality Theory,” Journal of Comparative Germanic Linguistics 2: 101–142. Gaeta, L. 1998. “Stress and loan words in German,” Rivista di Linguistica 10.2: 355–392. Giegerich, H. 1985. Metrical Phonology and Phonological Structure: German and English. Cambridge University Press.

93

94

BIRGIT ALBER

Golston, C. and R. Wiese 1998. “The structure of the German root.” In W. Kehrein and R. Wiese (eds.), Phonology and Morphology of the Germanic Languages. Tu¨bingen: Max Niemeyer Verlag: 165–187. Gordon, M. 2002. “A factorial typology of quantity-insensitive stress,” Natural Language and Linguistic Theory 20: 491–552. Halle, M. and J-R. Vergnaud 1987. An Essay on Stress. Cambridge, MA: MIT Press. Hammond, M. 1984. Constraining Metrical Theory: A Modular Theory of Rhythm and Destressing. Bloomington: Indiana University Linguistics Club. Hammond, M. 1999. The Phonology of English: A Prosodic Optimality-Theoretic Approach. Oxford University Press. Hayes, B. 1981. A Metrical Theory of Stress Rules. Bloomington: Indiana University Linguistics Club. Hayes, B. 1995. Metrical Stress Theory: Principles and Case Studies. University of Chicago Press. Hulst, H. van der 1984. Syllable Structure and Stress in Dutch. Dordrecht: Foris. Hulst, H. van der (ed.) 1999. Word Prosodic Systems in the Languages of Europe. Berlin: Mouton de Gruyter. Hulst, H. van der, J-R. Goedemans, and E. van Zanten (eds.) 2010. A Survey of Word Accentual Patterns in the Languages of the World. Berlin: Mouton de Gruyter. Hyde, B. 2001. Metrical and Prosodic Structure in Optimality Theory. Doctoral dissertation, Rutgers University. ROA-476. Hyde, B. 2002. “A restrictive theory of stress,” Phonology 19: 313–360. Hyde, B. 2007. “Non-finality and weight-sensitivity,” Phonology 24: 287–334. Hyde, B. 2012. “Alignment constraints,” Natural Language and Linguistic Theory 30: 1–48. Hyde, B. 2016. Layering and Directionality: Metrical Stress in Optimality Theory. London: Equinox. Itoˆ, J. and A. Mester 2015. “The perfect prosodic word in Danish,” Nordic Journal of Linguistics 38.1: 5–36. Janßen [Domahs], U. 2003. Untersuchungen zum Wortakzent im Deutschen und Niederla¨ndischen. Doctoral dissertation. University of Du¨sseldorf. Janßen [Domahs], U. and F. Domahs 2008. “Going on with optimised feet: Evidence for the interaction between segmental and metrical structure in phonological encoding from a case of primary progressive aphasia,” Aphasiology 22.11: 1157–1175. Jessen, M. 1999. “German.” In van der Hulst (ed.): 515–545. Kager, R. 1989. A Metrical Theory of Stress and Destressing in English and Dutch. Ph.D. dissertation. Foris Publications, Dordrecht. Kager, R. 1993. “Alternatives to the iambic-trochaic law,” Natural Language and Linguistic Theory 11: 381–432. Kager, R. 2007. “Feet and metrical stress.” In P. de Lacy (ed.), The Cambridge Handbook of Phonology. Cambridge University Press: 195–227.

Word Stress in Germanic

Kaltenbacher, E. 1994. “Typologische Aspekte des Wortakzents: Zum Zusammenhang von Akzentposition und Silbengewicht im Arabischen und Deutschen,” Zeitschrift fu¨r Sprachwissenschaft 13: 20–55. Kehrein, W. and R. Wiese (eds.) 1998. Phonology and Morphology of the Germanic Languages. Tu¨bingen: Max Niemeyer Verlag. Kleiner, S. 2011–2017. Atlas zur Aussprache des deutschen Gebrauchsstandards (AADG). Unter Mitarbeit von Ralf Kno¨bl. http://pro wiki.ids-mannheim.de/bin/view/AADG/. Knaus, J. R. Wiese, and U. Domahs 2011. “Secondary stress is distributed rhythmically within words: an EEG study on German.” In Proceedings of the 17th International Conference of the Phonetic Sciences 2011. Hong Kong: 1114–1117. Knaus, J. and U. Domahs 2009. “Experimental evidence for optimal and minimal metrical structure of German word prosody.” Lingua 119.10: 1396–1413. Kristoffersen, G. 2000. The Phonology of Norwegian. Oxford University Press. Lacy, P. de 2007. The Cambridge Handbook of Phonology. Cambridge University Press. Lahiri, A., T. Riad, and H. Jacobs 1999. “Diachronic prosody.” In van der Hulst (ed.): 335–422. Liberman, M. and A. Prince 1977. “On stress and linguistic rhythm,” Linguistic Inquiry 8: 249–336. Lorentz, O. 1996. “Length and correspondence in Scandinavian,” Nordlyd 24: 111–128. McCarthy, J. and A. Prince 1993. “Generalized alignment,” Yearbook of Morphology 1993: 79–153. McManus, H. 2006. Stress Parallels in Modern OT. Doctoral dissertation, Rutgers University. Moulton, W. G. 1962. The Sounds of English and German. University of Chicago Press. Pater, J. 2000. “Non-uniformity in English secondary stress: The role of ranked and lexically specific constraints,” Phonology 17: 237–274. Prince, A. 1983. “Relating to the Grid,” Linguistic Inquiry 14: 19–100. Prince, A. 1990. “Quantitative consequences of rhythmic organization.” In K. Deaton, M. Noske, and M. Ziolkowski (eds.), CLS 26-II: Papers from the Parasession on the Syllable in Phonetics and Phonology. Chicago: Chicago Linguistic Society: 355–398. Prince, A. and P. Smolensky 2004 [1993]. Optimality Theory: Constraint Interaction in Generative Grammar. Oxford: Blackwell. Riad, T 2013. The Phonology of Swedish. Oxford University Press. Rice, C. 1999. “Norwegian.” In van der Hulst (ed.): 545–553. Rice, C. 2006. “Norwegian stress and quantity: The implications of loanwords,” Lingua 116: 1171–1194. Speyer, A. 2009. “On the change of word stress in the history of German,” Beitra¨ge zur Geschichte der deutschen Sprache und Literatur (PBB) 131.3: 413–441.

95

96

BIRGIT ALBER

Trommelen, M. and W. Zonneveld 1999. “Dutch.” In van der Hulst (ed.): 492–515. Vennemann, T. 1990. “Syllable structure and simplex accent in Modern Standard German.” In K. Deaton, M. Noske, and M. Ziolkowski (eds.), CLS 26-II: Papers from the Parasession on the Syllable in Phonetics and Phonology. Chicago: Chicago Linguistic Society: 399–412. Wiese, R. 1996. The Phonology of German. Oxford University Press. Wurzel, W. U. 1980. “Der deutsche Wortakzent: Fakten – Regeln – Prinzipien. Ein Beitrag zu einer natu¨rlichen Akzenttheorie,” Zeitschrift fu¨r Germanistik 3: 299–318. Zonneveld, W. 1999. “Word stress in West-Germanic and North-Germanic Languages: Introduction.” In van der Hulst (ed.): 476–478.

Chapter 5 Quantity in Germanic Languages B. Richard Page

5.1 Introduction Quantity may refer to either segments or syllable weight (Davis 2011). With regard to segments, a contrast in quantity is realized as a difference in phonetic duration. In the literature, a quantity opposition on the segmental level may be represented by the feature [+/- long], by association with moras, or by association with one or two positions on a skeletal tier. A moraic representation of quantity is given in (1) and a skeletal representation is given in (2). In (1), a short vowel is associated with one mora and a long vowel is associated with two moras. Therefore, we can refer to a short vowel as monomoraic and a long vowel as bimoraic. A short (or singleton) consonant is not associated with a mora underlyingly, and a long (or geminate) consonant is associated with a mora underlyingly. In (2), short segments are associated with one position on the skeletal tier and long segments are associated with two positions on the skeletal tier. (1)

Underlying moraic representation of segmental quantity (Hayes 1989) µ

µ

a = /a /

a = /a/

(2)

µ

µ t = /t /

t = /t/

Underlying skeletal representation of segmental quantity X a = /a/

X

X a = /a /

X t = /t/

X

X t = /t /

While some researchers refer to segmental duration with the term “quantity” (Riad 1995; Kraehenmann 2001; Davis 2011), others prefer the term “length” (e.g., Odden 2011). This chapter will primarily focus on

98

B . R I C H A R D PAG E

contrastive segmental duration with references to syllable weight. The terms “quantity” and “length” will be used interchangeably. Unless noted otherwise, it is assumed that a long vowel is also bimoraic. Historically, all Germanic languages had quantity contrasts for both vowels (long versus short) and consonants (geminate versus singleton). With regard to syllable weight, quantity refers to a distinction between heavy and light syllables. If the distinction between heavy and light syllables plays a role in stress assignment, the language is said to be quantity sensitive. If not, the language is quantity insensitive (see Alber, Chapter 4, for further discussion). In addition, a distinction between light and heavy syllables may play a role in verse. For example, the basic unit of early Germanic verse, known as a lift, could be realized as a heavy syllable (H) or as a sequence of two light syllables (LL) (Suzuki 1995). The earliest attested Germanic languages are usually analyzed as forming moraic trochees from left to right beginning with the initial syllable of the root. Thus, H and LL were metrically equivalent in early Germanic and a foot could consist of either possibility (Suzuki 1995; see Dresher and Lahiri 1991, Riad 1992, Lahiri et al. 1999 for overviews of the diachronic development of stress and foot structure in Germanic). A wide range of quantity changes, including open syllable lengthening and degemination, occurred in the medieval period and are known as the Germanic quantity shift. The quantity changes were not uniform throughout Germanic. For example, many varieties of German underwent degemination, but Swiss German (High Alemannic) and Bavarian did not. Depending upon the extent to which they were affected by the quantity shift, some Germanic languages retained both vowel and consonant quantity, some retained vowel quantity but lost consonant quantity, and some lost vowel quantity but retained consonant quantity. The Germanic languages can be divided typologically into three groups with regard to quantity, as in Table 5.1 (adapted from Riad 1995: 165; see also Lahiri et al. 1999). Table 5.1 Typology of quantity (revised later in this chapter) Vowel and consonant quantity

Consonant quantity

Vowel quantity

Älvdalsmål (Swedish dialect) Nord Gudbrandsdalska (Norwegian dialect) Western Nyland (Fenno-Swedish dialect) High Alemannic (Swiss German dialect) Highest Alemannic (Swiss German dialect) South Bavarian (Tyrolean German dialect)

Icelandic Faroese Swedish Norwegian Central Bavarian (BavaroAustrian German dialect)

English Dutch German Danish

Quantity in Germanic Languages

It is interesting to note that varieties of North and West Germanic are found in all three categories. The first category comprises nonstandard varieties found on the periphery of the area where Germanic languages are spoken in Europe. These conservative varieties, which have preserved quantity contrasts for both vowels and consonants, will be discussed in Section 5.2. From a diachronic perspective, the dialects that preserve quantity for both vowels and consonants did not undergo either open syllable lengthening or degemination. From a synchronic perspective, they permit a stressed syllable to be light or heavy. Section 5.3 provides an overview of languages in which quantity is only contrastive for consonants. Historically, these languages have undergone open syllable lengthening but not degemination and now require that the stressed syllable be heavy. Language varieties in this category include Central Bavarian and the standard varieties of North Germanic languages spoken on the Scandinavian Peninsula and in Iceland and the Faroe Islands. The languages that have preserved vowel quantity according to Riad (1995) are all located centrally in the Germanic language area and include the standard varieties of Danish, English, German, and Dutch. Standard varieties of English, Danish, German, and Dutch have undergone both open syllable lengthening and degemination. As will be discussed in more detail in Section 5.4, whether or not vowel quantity is in fact contrastive in these languages is not always clear and has been the subject of debate. Section 5.5 concludes the discussion of quantity in Germanic.

5.2 Contrastive Quantity for Consonants and Vowels Contrastive vowel and consonant quantity can be found in Upper German dialects as well as in dialects of Norwegian and Swedish. In all languages in this category, long and short vowels can be found in stressed open syllables. However, they differ in whether light monosyllables exist and whether superheavy syllables are permitted.

5.2.1 Varieties of Swedish There are three quantitative types in Swedish as shown in (3) by illustrating possible vowel plus consonant (VC) sequences (Riad 2013: 163, based on Schaeffler 2005). (3)

Finland Swedish: VːCː, VːC, VCː, VC Northern Swedish: *V:Cː, VːC, VCː, VC Southern Swedish: *VːCː, VːC, VCː, *VC

As indicated by (2), contrastive vowel and consonant quantity is found in both Finland Swedish and in Northern Swedish. In both cases, quantity is

99

100

B . R I C H A R D PAG E

not predictable. In other words, a stressed short vowel may be followed by either a singleton or a geminate consonant, and a singleton consonant may be preceded by either a short stressed vowel or a long stressed vowel. In Southern Swedish, which displays complementary quantity, a phenomenon that will be discussed in Section 5.2.2, quantity appears to be predictable. Thus, in stressed syllables a short vowel is always followed by a geminate consonant and a long vowel is always followed by a singleton consonant, or, from a different perspective, a singleton consonant is always preceded by a long vowel and a geminate consonant is always preceded by a short vowel. The Swedish dialects in Finland show four possible VC sequences in both monosyllabic and bisyllabic stems. Long vowels and long consonants are always part of the stressed syllables, and long consonants are ambisyllabic if the stressed syllable is nonfinal (see Kiparksy 2008 for a typology of quantity in different Fenno-Swedish dialects). FennoSwedish permits monosyllables of the shape CVC as well as light stressed syllables in disyllables, i.e., CV.CV. The contrast in vowel and consonant quantity in Fenno-Swedish is shown in (4) (based on Harling-Kranck 1998 as reported in Kiparsky 2008: 186). (4)

a. [baka] ‘bake’, [baaket] ‘after’, [bakkan] ‘the hill’ b. [vaten] ‘water’, [maaten] ‘the food’, [natten] ‘the night’ c. [beta¨r] ‘better’, [fleetor] ‘braids’, [tvettar] ‘washes’

Unlike most Swedish dialects, at least some dialects of Fenno-Swedish also permit superheavy syllables consisting of a sequence of long vowel plus geminate consonant in the stressed syllable of both disyllabic and monosyllabic words, as in (5) (based on Harling-Kranck 1998 as reported in Kiparksy 2008: 192). (5)

a. /loo-dde-s/ [looddes] ‘pretended’, /dreett-en/ [dreetten] ‘the shaft’ b. /smoo-tt/ [smoott] ‘little one’, /haa-dd/ [haadd] ‘have’, /ruu-dd/ [ruudd] ‘rowing’

Northern Swedish dialects show a three-way system permitting sequences of VC, VːC, VC: but disallowing VːCː. Thus, Northern Swedish dialects allow both light and heavy stressed syllables, but superheavy stressed syllables are largely disallowed. In Svea varieties, spoken in Sweden just south of the area where northern varieties are spoken, superheavy stressed syllables (sequences of VːCː) are reported in some dialects of Dalarna by Levander (1925). The sequence VC is also reported in both monosyllables (i.e., CVC) and disyllables (i.e., CV.CV) in some Dalarna dialects (Levander 1925: 65–74; discussion in Schaeffler 2005: 45). In varieties of Swedish spoken south of Dalarna dialects, there is complementary quantity, and only sequences of VCː and VːC are reported (Schaeffler 2005: 45).

Quantity in Germanic Languages

5.2.2 Varieties of Norwegian The Norwegian dialects of North Gudbrandsdal and Mid Gudbrandsdal each show contrastive vowel and consonant quantity in stressed syllables. Examples from North Gudbrandsdal are given in (6) (adapted from Kristoffersen 2011: 53). The consonants that are enclosed in parentheses are extrametrical. Extrametrical consonants do not affect the weight of the syllable and therefore do not affect the assignment of stress (see Alber, Chapter 4). (6)

Distribution of vowel and consonant quantity in North Gudbrandsdal Monosyllabic words CV CVC CVV(C) CVCC CVVCC

/1le/ le ‘opening in fence’ (N O M . /1ʂen/ skin ‘shine’ /1fiːn/ fin ‘fine’ /1ʂinn/ skinn ‘skin’ No longer exists

Disyllabic words CV.CV CVV.CV CVC.CV CVVC.CV

/2bo.ko/ baka ‘to bake’ /2ʋiː.se/ vise ‘song’ /1ʂin.nə/ skinnet ‘the skin’ No longer exists

SG.)

In Mid Gudbrandsdal, contrastive quantity for vowels and consonants is present in disyllabic words. In monosyllabic words, there is complementary quantity, as shown in (7) (adapted from Kristoffersen 2011: 55). (7)

Distribution of vowel and consonant quantity in Mid Gudbrandsdal CV CV(C) CVV(C)

CVVCC

No longer exists No longer exists /1leː/ le ‘opening in fence’ (N O M . /1fiːn/ fin ‘fine’ /1ʂinn/ skinn ‘skin’ /1ʂenn/ skin ‘shine’ No longer exists

Disyllabic words CV.CV CVV.CV CVC.CV CVVC.CV

/2bo.ko/ baka ‘to bake’ /2ʋiː.se/ vise ‘song’ /1ʂin.nə/ skinnet ‘the skin’ No longer exists

CVCC

SG.)

5.2.3 Swiss German (High and Highest Alemannic) Both contrastive vowel quantity and contrastive consonant quantity are also found in Swiss German dialects. In the Swiss German dialect literature, the distinction between geminate and singleton obstruents is often

101

102

B . R I C H A R D PAG E

labeled a fortis-lenis distinction (e.g., Sievers 1876, Winteler 1876, Dieth 1950, Fleischer and Schmid 2006). Several studies have found that closure duration is the primary phonetic correlate of the fortis-lenis distinction (Fulop 1994; Kraehenmann 2001, 2003; Willi 1995, 1996). Kraehnmann (2001) argues that the distinction is best analyzed as a contrast between geminates and singletons since there are significant differences in closure duration but no significant differences in voice onset time between the two series (see Salmons, Chapter 6, for an overview of laryngeal contrasts in Germanic obstruents). In Swiss German, the laryngeal contrast between the reflexes of Gmc. *p, *t, *k and Gmc. *b, *d, *g was eliminated by the High German Consonant Shift (Second Sound Shift), which shifted Gmc. *p, *t, *k to voiceless affricates word-initially, in geminates, and after consonants, and to fricatives after vowels. The shift reached its greatest extent in the Upper German dialects of High and Highest Alemannic and South Bavarian (see Keel, Chapter 31, for a discussion of the West Germanic dialect continuum). In the aftermath of the High German Consonant Shift, nonfricative obstruents in Old Alemannic, the precursor to Swiss German, were distributed as in Table 5.2 (Kraehenmann 2001: 139). In contemporary Swiss German, the singleton-geminate contrast can also be found in initial position, as shown by the Thurgovian examples in Table 5.3 (from Kraehenmann 2001: 140). As noted by Kraehenmann (2001: 110), initial geminates are typologically rare, and the initial geminates in Swiss German are particularly notable because they are lexical. Many researchers have noted that realizations of stops word-initially in modern Swiss dialects often display a sandhi phenomenon known as Heusler’s Law (Heusler 1888, Penzl 1955). In a description of this sandhi phenomenon, Moulton (1986) shows that the contrast between singleton and geminate stops is neutralized in word-initial position when the preceding segment is voiceless in Zurich German. It is important to note that all obstruents are voiceless and all other segments are voiced in Zurich German as in other Swiss German dialects. In addition to being voiceless, all stops are also unaspirated in Zurich German and other Swiss German dialects. Following a long tradition in Swiss German dialectological research, Moulton refers to the geminate stops as “fortis” and transcribes them as /pp tt kk/, and he calls Table 5.2 Distribution of Old Alemannic nonfricative obstruents initial p pf

t ts

medial k kx

p pp pf

t tt ts

final k kk kx

p pp pf

t tt ts

k kk kx

Quantity in Germanic Languages

Table 5.3 Stops in Thurgovian (Swiss German) a. Original voiced stops are voiceless singletons (except */d/ > /tt/ word-initially) initial /poonə/ /ttak/ /kʀap/

‘bean’ ‘day’ ‘grave’

medial /ʃ tupe/ /vatə/ /pokə/

‘living room’ ‘calf’ ‘bow’

final /krap/ /rat/ /ttak/

‘grave’ ‘wheel’ ‘day’

b. Voiced stops in loanwords are realized as voiceless singletons initial /poott/ /tattlə/ /kalopp/

‘boat’ ‘date’ ‘gallop’

medial /piplə/ /ʀataʀ/ /tɛkə/

‘bible’ ‘radar’ ‘sword’

final /snop/ /faat/ /sook/

‘snob’ ‘tasteless’ ‘suction’

c. Voiceless stops in loanwords are realized as voiceless geminates initial /ppaaʀ/ /ttuʀttə/ /kkomfi/

medial

‘pair’ /xappə/ ‘layer cake’ /mattə/ ‘confiture’ /jakkə/

‘cap’ ‘mat’ ‘jacket’

final /kalopp/ /fett/ /paŋkk/

‘gallop ‘fat’ ‘bank’

Table 5.4 Neutralization of fortis-lenis contrast word-initially in Zürich German Singleton or geminate Initial position /b/

/pp/

/d/ /tt/ /g/ /kk/

/baxxə/ ‘backen’

After a sonorant

[ʃoː baxxə] ‘schon backen’ /ppaxxə/ ‘gebacken’ [ʃoː ppaxxə] ‘schon gebacken’ /doːrff/ ‘Dorf’ [im doːrff] ‘im Dorf’ /ttaːg/ ‘Tag(e)’ [tsveː ttaːk] ‘zwei Tage’ /geː/ ‘geben’ [au geː] ‘auch geben’ /kkeː/ ‘gegeben’ [au kkeː] ‘auch gegeben’

After a pause

After an obstruent

[paxxə] ‘backen’

[muəs paxxə] ‘muss backen’

[paxxə]‘gebacken’

[häp paxxə] ‘hat gebacken’

[toːrf] ‘Dorf’

[s toːrf] ‘das Dorf’

[taːk] ‘Tag(e)’

[säxs taːk] ‘sechs Tage’ [öppis keː] ‘etwas geben’ [öppis keː] ‘etwas gegeben’

[keː] ‘geben’ [keː] ‘gegeben’

singleton stops “lenis stops” and transcribes them as /b d g/. After a voiceless segment, geminate and singleton stops are realized as “half-fortis” stops, which are represented as [p t k]. Examples of the neutralization of the geminate-singleton contrast in word-initial position are given in Table 5.4 (based on Moulton 1986: 387).

103

104

B . R I C H A R D PAG E

The same pattern of neutralization found in Zurich German is also present in Thurgovian. As is the case in Zurich German, the contrast between word-initial singletons and geminates is neutralized after all obstruents. Examples are given in (8). (8)

Neutralization of initial singleton-geminate contrasts in Thurgovian after obstruents a. After sonorants After stops (Kraehenmann 2001: 135) au /k/ern [au kəʀn] ‘gladly too’ no¨/t/ /k/ern au /kk/nau [au kknau] ‘exactly too’ no¨/t/ /kk/nau [no¨ kkəʀn] ‘not gladly’ [no¨ kknau] ‘not exactly’ b. After sonorants [kxai paːʀ] ‘no bar’ [tsvai ppaʀ] ‘two pairs’

After fricatives (Lahiri and Kraehenmann 2004: 33) [fyːf paːʀə] ‘five bars’ [fyːf paːʀ] ‘five pairs’

5.3 Complementary Quantity Germanic languages that have complementary quantity include Icelandic, Faroese, Swedish, Norwegian, and Central Bavarian. From a historical perspective, all of these languages have undergone open syllable lengthening but not gemination. From a synchronic perspective, they all require stressed syllables to be heavy. Complementary quantity refers to the distribution of vowel length and consonant length in sequences of vowel plus consonant in a stressed syllable, as illustrated by the Swedish data given in (9). (9) Complementary quantity in Swedish (Riad 2013: 159–160) a. Disyllables la¨cka [2ˈlɛ̝ kːa] ‘to leak’ la¨ka [2ˈlɛːka] ‘to heal’ 2 º soppa [2ˈsɔpːa] ‘soup’ sapa [ ˈsoːpa] ‘soft soap’ 2 puta [ ˈpʉːta] ‘to pout; bulge’ putta [2ˈpɵtːa] ‘to putt; push’ tiga [2ˈtiːga] ‘to keep silent’ tigga [2ˈtɪgːa] ‘to beg’ b. Monosyllables ful [ˈfʉːl] ‘ugly’ full [ˈfɵlː] ‘full’ vin [viːn] ‘wine’ vinn [vɪnː] ‘win’ lob [luːb] ‘lobe’ lobb [lɔbː] ‘lob’ lam [lɑːm] ‘lame’ lamm [lamː] ‘lamb’ The data in (9) show that sequences of vowel plus consonant in a stressed syllable are always realized as VːC or as VCː In Standard Swedish (and more generally in southern varieties of Swedish), sequences of *VC and *VːCː are

Quantity in Germanic Languages

not allowed in stressed syllables (see example 3). A similar pattern can be found in varieties of Norwegian (Kristoffersen 2000), Icelandic (A´rnason 2011), Faroese (A´rnason 2011), and Central Bavarian (Kleber 2017). In all cases, there is an element of predictability. If one knows the length of the stressed vowel, one can predict the length of the following consonant. Similarly, if one knows the length of the consonant, one can predict the length of the immediately preceding vowel. There has been much debate in the literature as to the best interpretation of these sequences (see overviews in Riad 2013 for Swedish, A´rnason 2011 for Icelandic, Kristoffersen 2000 and 2011 for Norwegian, Kleber 2017 for Central Bavarian). In what follows, we will focus on arguments put forth by Riad (2013) that consonantal length in Swedish is phonemic and vowel length is derived. We will then discuss evidence presented by Kleber (2017) that complementary quantity is best viewed as a prosodeme in Central Bavarian.

5.3.1 Complementary Quantity in Swedish Riad (2013) provides an excellent overview of the previous literature on the phonemic status of the vowel and consonant length in Swedish. In his analysis, vowel length is always derived whereas consonantal length has three sources: underlying length, assimilation of dental stops, and weight by position. Examples are given in (10). (10) Sources of consonantal length (Riad 2013: 166) a. Underlying: vinna /vinμ-ɑ2/ [2ˈvɪnːa] ‘to win’ katt /kɑtμ/ [katː] ‘cat’ b. Assimilated: bytte /byt-de2/ [2bʏtːɛ] ‘changed’ (byta [2ˈbyːta] ‘to change’) vitt /vit-t/ [vɪtː] ‘white, neut.’ (vit [viːt] ‘white, common gender’) vitt /vid-t/ [vɪtː] ‘wide, neut.’ (vid [viːd] ‘wide, common gender’) c. Weight by pos.: kasta /kɑst-ɑ2/ [2ˈkasːta] ‘to throw’ mjo¨lk /mjølk/ [mjølːk] ‘milk’ In Riad’s analysis, a long consonant1 is one that carries a mora. Long consonants from all three sources listed in (10) add weight to the syllable and are attached to a mora. He notes that the consonants that are long because of weight by position are longer than their nonmoraic counterparts but they are shorter than long consonants from the other two sources (Riad 2013: 167). In cases where an intervocalic cluster has a rising sonority profile, the cluster is syllabified with the following vowel and the preceding vowel is 1

Riad (2013) makes a distinction between a long consonant and a geminate. In Riad’s view, an underlying long consonant is associated with a mora and may be represented as /tμ/ as is the case for katt in (10a). In contrast, a geminate is underlyingly a sequence of two consonants and would be represented as /tt/. Therefore, a geminate is phonologically the equivalent of a consonant cluster.

105

106

B . R I C H A R D PAG E

lengthened if stressed. However, the vowel remains short if the initial consonant in the cluster is underlyingly moraic, as shown in (11). (11)

Intervocalic consonant clusters in Swedish (Riad 2013: 167) stapla /stɑpl-ɑ2/ [2ˈstɑːpla] ‘to heap’ stappla /stɑpμl-ɑ2/ [2ˈstapːla] ‘to stagger’

In cases of stress alternation, long consonants are shortened when stress does not fall on the syllable containing the underlyingly long consonants. Examples are in (12). (12)

Long consonants in stress alternations (Riad 2013: 176) tyrann /tyrɑnμ/ [tʏˈranː] ‘tyrant’ attack /ɑtɑkμ/ [aˈtakː] ‘attack’

tyranni [tʏraˈni] ‘tyranny’ attackera [ata1ˈkeːra] ‘to attack’

In Riad’s analysis, vowel length is always predictable in contrast to consonant length, which can be contrastive. Stressed vowels are always long in open syllables, as exemplified in (13). (13)

Long vowels in stressed open syllables (Riad 2013: 168–169) tre /tre/ [ˈtreː] ‘three’ tobak /tubɑk/ [1ˈtuː.bak] ‘tobacco’

Stressed vowels are also long before morpheme-final single consonants. In such cases, the consonant is considered to be extrametrical (Riad 2013). Final consonants that are underlyingly long are not extrametrical. Examples are in (14). (14)

Long vowels before final extrametrical consonants (Riad 2013: 169) kamin /kɑmin/ [kaˈmiːn] ‘stove’ (cf. tyrann /tyrɑnμ/ [tʏˈranː] ‘tyrant’) spel /spel/ [ˈspeːl] ‘game’ (cf. kam /kɑmμ/ [kamː] ‘comb’)

Before intervocalic consonant clusters, syllabification determines vowel length. If the entire cluster is syllabified with the following vowel (V. CCV), the preceding stressed vowel is in an open syllable and therefore long. If the cluster is heterosyllabic (VC.CV), the preceding stressed vowel is in a closed syllable and therefore short. The consonant that closes the stressed syllable becomes long because of weight by position (Riad 2013). Vowel length before intervocalic consonant clusters is illustrated in (15). (15)

Vowels before intervocalic clusters (Riad 2013: 170) V.CCV ta¨vla [2ˈtɛː.vla] ‘to compete’ va¨gra [2ˈvɛː.gra] ‘to refuse’

VC.CV ta¨nda [2ˈtɛnː.da] ‘to light’ va¨lta [2vɛlːta] ‘to overturn’

In Riad’s view, vowel length is fully predictable in Swedish and therefore not contrastive. Vowels are long when stressed and in open syllables (as well as before final consonants, which are extrametrical). Otherwise, vowels are

Quantity in Germanic Languages

short. Thus, vowel lengthening in Swedish is a synchronic process that is motivated by syllable structure and word stress (Riad 2013: 177).

5.3.2 Complementary Quantity in Central Bavarian Complementary quantity in Central Bavarian was first identified by Pfalz (1911, 1913). Pfalz (1911: 260, 1913: 9) stated that there was a relationship between vowel quantity and consonantal strength such that obstruents must be fortis after short vowels and lenis after long vowels. Before liquids and nasals, which are either vocalized or realized as singletons, vowels are always long (Pfalz 1913: 9). An example of complementary quantity in monosyllables in Central Bavarian is given in (16). In disyllables, stressed vowels are long in open syllables and short before geminates, as shown in (17). (16) (17)

[v̥ iːʒ] ‘fish (sg.)’ versus [v̥ iʃʃ] ‘fish (pl.)’ (Seiler 2009: 243, based on Pfalz 1913) Complementary quantity in disyllables (Seiler 2009: 244, based on Pfalz 1913) a. [leː.d̥ ɒ] ‘leather’ [v̥ oː.g̊ l̩ ] ‘bird’ b. [ʋɔs.sɒ] ‘water’ [ʋiʃ.ʃn̩ ] ‘wipe’

Unlike in Swedish, vowels are also long in Central Bavarian before consonant clusters if the cluster does not contain a geminate. As discussed above, Swedish may assign a mora to a postvocalic coda consonant via weight by position. These derived moraic consonants in Swedish undergo lengthening. In contrast, Central Bavarian does not have weight by position, and the postvocalic consonant remains a singleton. In accordance with the pattern of complementary quantity, the stressed vowel is long before the nonmoraic singleton (Seiler 2009). Examples are given in (18). (18)

[nɔːχd̥ ] ‘night’ (Seiler 2009: 245) [khĩːnd̥ ] ‘child’

Kufner (1957) points to evidence from sandhi to argue that fortis obstruents are better described as long and lenis obstruents are better described as short in Central Bavarian. An example of sandhi is given in (19). (19)

Sandhi in Central Bavarian (Kufner 1957: 178) a. geht er [ge´˛ e˛dæ] ‘goes he ’ b. geht der Bauer [ge˛´ttæba`ua] ‘goes the farmer’

In (19a) geht [ge˛e˛d] ‘go (3sg.)’ is combined with er [æ] ‘he’. The vowel in geht [ge˛e˛d] ‘go (3sg.)’ is long and is followed by a lenis obstruent. In (19b), geht is followed by der [dæ] ‘the (M A S C . N O M .)’. The lenis /d/ at the end of geht comes

107

108

B . R I C H A R D PAG E

into contact with lenis /d/ at the beginning of der in the string geht der Bauer, and this results in a derived geminate. The derived geminate in (19b) is preceded by a short vowel as is the case before obstruents traditionally labeled fortis. Kufner argues that both derived geminates and underlying fortis obstruents should therefore be analyzed as geminates. Just as a vowel will shorten before a sandhi-derived geminate, a vowel will also shorten before geminates created by affixation as in (20). (20)

/red+∅/ [iː reːd̥ ] ‘I talk’ versus /red̥ +d̥ / [eə rett] ‘he talks’ (Rowley 1989: 423; cited in Seiler 2009: 244)

Kufner (1957), Hinderling (1980), Scheutz (1985), Wiesinger (1990), and Seiler (2005, 2009) argue that difference in consonantal length is the underlying contrast in Central Bavarian and that vowel length is derived (cf. Riad 2013 for complementary quantity in Swedish). In a perceptual experiment, Bannert (1976, 1977) found that Central Bavarian users identified vowel and consonant length on the basis of their relative duration regardless of absolute segment durations. Thus, Bannert (1976, 1977) argues that complementary length itself is contrastive, that is, Central Bavarian has a contrast between VːC and VCː. Kleber (2017) tests the competing theories about the phonological basis for complementary quantity in Central Bavarian by analyzing Central Bavarian speakers’ production of the Standard German vowel length contrast before sonorant consonants.2 Vowels are always long before sonorant consonants in Central Bavarian, whereas long and short vowels may occur before sonorant consonants in Standard German, which has no length or fortis-lenis contrast for sonorant consonants. If complementary length is contrastive, Kleber predicts that Central Bavarian speakers will show longer sonorant consonant durations after lax (short) vowels than after tense (long) vowels. On the other hand, if consonantal length is contrastive and vowel length is allophonic, Central Bavarian speakers should not produce a contrast between tense (long) and lax (short) vowels because Central Bavarian has no contrast between long and short sonorant consonants. Kleber (2017) found that Central Bavarian speakers did not neutralize the difference between long and short vowels before sonorants and that sonorant consonants were longer after lax (short) vowels than after tense (long) vowels. Her finding supports Bannert’s position that complementary length is contrastive. The difference in duration for sonorants was greater among older Central Bavarian speakers than among younger ones, suggesting that Central Bavarian is undergoing dialect leveling under the influence of Standard German.

2

The discussion of Kleber (2017) follows her use of the terms “length,” “long”, and “short” instead of the terms often used in discussions of quantity by researchers such as Riad, e.g., “quantity,” “geminate”, and “singleton.”

Quantity in Germanic Languages

5.4 Contrastive Quantity for Vowels Only According to Riad’s typology of quantity in Germanic outlined in Table 5.1, the standard varieties of Danish, English, German, and Dutch all have a quantity contrast for vowels but not for consonants. These languages have all undergone open syllable lengthening and degemination. It is widely accepted that they no longer have contrastive length for consonants. An example of contrastive vowel length in Danish can be found in the minimal pair hule [ˈhuːlǝ] ‘cave’ ~ hulde [ˈhulǝ] ‘benign (pl., def.)’ (Basbøll 2005: 79). The existence of contrastive vowel length in German, English, and Dutch has been the subject of much debate. In these languages, the vowel systems show a great redundancy between vowel tenseness and vowel length. In most cases (but not always), tense vowels are long and long vowels are tense; lax vowels are short and short vowels are lax. For this reason, many researchers have tried to determine whether the underlying contrast in a minimal pair like German Beet [beːt] ‘(flower) bed’ – Bett [bɛt] ‘bed’ is based on a difference in vowel length or a difference in vowel quality that stems from a contrast in vowel tenseness. In the discussion that follows, we will examine evidence that vowel length is indeed contrastive in German and American English followed by evidence that vowel length is not contrastive in Dutch.

5.4.1 Vowel Quantity in German Minimal pairs involving vowel length are shown in (21). In most cases, there is also a contrast in vowel quality. The exceptions are the contrast between [aː] and [a] and between [ɛː] and [ɛ]. (21)

Minimal pairs involving vowel quantity (Wiese 1996: 11) bieten ‘offer’ ~ bitten ‘request’ Hu¨te ‘hats’ ~ Hu¨tte ‘hut’ Beeten ‘(flower) bed’ ~ Bett ‘bed’ sehen ‘see’ ~ sa¨en ‘sow’ a¨ße ‘eat (1sg.subj.)’ ~ esse (1sg.ind.) Ho¨hle ‘cave’ ~ Ho¨lle ‘hell’ Schal ‘scarf’ ~ Schall ‘sound’ Ofen ‘oven’ ~ offen ‘open’ spuken ‘spook’ ~ spucken ‘spit’

[iː] ~ [ɪ] [yː] ~ [ʏ] [eː] ~ [ɛ] [eː] ~ [ɛː] [ɛː] ~ [ɛ] [øː] ~ [œ] [aː] ~ [a] [oː] ~ [ɔ] [uː] ~ [ʊ]

Seven of the nine minimal pairs listed in (21) show a difference in both vowel length and vowel quality. The difference between [aː] ~ [a] and [ɛː] ~ [ɛ] involves only length. However, many researchers have not accepted the existence of the phoneme /ɛː/ in standard varieties of German (Moulton 1962, Sanders 1972, Reis 1974). They argue that German dialects do not have a contrast between [eː] and [ɛː] and that [ɛː] is just a spelling pronunciation. They also cite

109

110

B . R I C H A R D PAG E

forms such as those in (22) to support the view that there is no contrast between [eː] and [ɛː] since [eː] never occurs before [ʀ] (Wiese 1996: 17): (22)

[ɛː] Speeren ‘spears (dat.)’ Na¨hren ‘nourish’ Hehren ‘noble’

[ɛ] sperren ‘block’ Na¨rrin ‘(female) fool’ Herren ‘gentlemen’

Wiese (1996: 17) argues that the examples in (22) are not dispositive because the distinction between /eː/ and /ɛː/ is neutralized before /ʀ/. In other positions, the distinction is quite robust in the variety of Standard German spoken in northern Germany as in the minimal pair sa¨en ‘sow’ with [ɛː] versus sehen ‘see’ with [eː] (Wiese 1996: 17). Even if [eː] and [ɛː] are not contrastive in all varieties of German, it appears clear that the contrast does exist and is stable in at least some varieties of German. As opposed to /ɛː/, the existence of the phonemes /aː/ and /a/ is uncontroversial, but there is widespread disagreement about the proper description of the contrast. Wiese (1996: 21–22) summarizes well the different descriptions of the contrast, which we recapitulate here. Moulton (1962) states that [a] is more central than [aː]. Some have maintained that [aː] is further back than [a] and should therefore be transcribed as [ɑː] (Wa¨ngler 1974, Krech et al. 1982, Basbøll and Wagner 1985). Wa¨ngler (1974: 105) argues that [a] is tense and [aː] (which he transcribes as [ɑː]) is lax, and others claim [a] is lax and [aː] is tense. Many describe the vowels as having the same quality (Martens and Martens 1961, Siebs 1969, Duden 1990, Kohler 1990, Wiese 1996). Wiese (1996: 22) points out that if /a/ and /aː/ undergo umlaut, they are realized as [ɛ] and [ɛː] respectively. Vowels that undergo umlaut are fronted but do not change with regard to vowel tenseness or vowel length. Therefore, the umlauting of /a/ and /aː/ to [ɛ] and [ɛː] supports the view that /a/ and /aː/ contrast only with respect to length (Wiese 1996: 22). There is also a long tradition in German phonology to use the notion of syllable cut to describe the putative contrast between long and short vowels (Sievers 1876, Trubetzkoy 1939, Vennemann 1994, Becker 1998; Spiekermann 2000, Venneman 2002, Restle 2003, Auer and Murray 2004). In this view, long vowels are found in syllables that have smooth cut, and short vowels are found in syllables that have abrupt cut. The terms smooth cut and abrupt cut are intended to describe the transition from the vowel to the following consonant. Trubetzkoy (1939) maintained that languages with syllable cut do not have contrastive length and that syllable cut is a prosodic phenomenon. Therefore, in a syllable cut language vowel length is predictable. Vowels are short in syllables with abrupt cut and long in syllables with smooth cut. According to this view, German is a syllable cut language and does not have vowel quantity (Vennemann 2000). Like the supporters of syllable cut, Riad (1995), Wiese (1996), and Lahiri et al. (1999) also treat vowel length as a prosodic phenomenon. Riad (1995)

Quantity in Germanic Languages

and Lahiri et al. (1999) represent vowel length via association with moras. A short vowel is associated with one mora and a long vowel is associated with two moras. Wiese (1996) represents vowel length by association with positions on the skeletal tier. A short vowel is associated with one position on the skeletal tier and a long vowel is associated with two positions.

5.4.2 Vowel Quantity in English The vowel inventory in American English is given in Table 5.5 (adapted from Hammond 1997: 2). As can be seen from Table 5.5, there is no need to posit a contrast of vowel length. The tense vowels are always long, and the lax vowels are always short. It is therefore not surprising that there is disagreement about the existence of vowel quantity (length) in English (Odden 2011). Chomsky and Halle (1968) treat the contrasts illustrated in Table 5.5 as a tense / lax distinction. Halle (1977) analyzes the vowel in words like meat and mate to be long in contrast to short vowels in mitt and met. In stressed open syllables, only tense vowels and diphthongs can occur. Examples are in Table 5.6 (adapted from Hammond 1997:2). If the final Table 5.5 Vowels in American English Tense (Long) i e u o ɔ ɑ

meat mate food moat caught balm

[mit] [met] [fud] [mot] [kɔt] [bɑm]

Lax (Short) mitt met good mutt cat bomb

ɪ ɛ ʊ ʌ æ a

Diphthongs [mɪt] [mɛt] [gʊd] [mʌt] [kæt] [bam]

aj aw oj ju

ice louse voice puce

Table 5.6 American English vowels in stressed open syllables Monosyllables

Polysyllables tense

i e u o ɔ ɑ

bee bay two toe paw spa

[bí] [bé] [tú] [tó] [pɔ́] [spɑ́ ]

Tennessee delay kangaroo hello macaw bourgeois

[tɛ̀ nəsí] [dəlé] [kæ̀ ŋgərú] [hɛ̀ ló] [məkɔ́] [bʊ̀ ʒwɑ́ ]

ally allow employ review

[əlay] [əlaw] [ɛ̀ mplój] [rəvjú]

diphthongs aj aw oj ju

buy bow boy cue

[baj] [baw] [bój] [kjú]

[ajs] [laws] [vojs] [pjus]

111

112

B . R I C H A R D PAG E

Table 5.7 Distribution of vowels in stressed syllables vowel tense diphthong lax schwa

´___# √ √ Ø Ø

´___C# √ √ √ Ø

stressed syllable is closed, the vowel may also be lax. The distribution of vowels in final stressed syllables is given in Table 5.7 (adapted from Hammond 1997: 3). The distribution of vowels can be interpreted as a requirement that the final stressed syllable consist of at least two moras. Therefore, Hammond (1997: 3) treats tense vowels and diphthongs as bimoraic and lax vowels as monomoraic.

5.4.3 Vowel Quantity in Dutch Our discussion of vowel quantity in Dutch follows Page (2006). Evidence from stress assignment in Dutch indicates that Dutch is weight sensitive but tense vowels are light, i.e., monomoraic. Lahiri and Koreman (1988: 222) describe stress assignment in Dutch as in (23) and provide the examples in (24). (23)

Stress assignment rule in Dutch (quoted from Lahiri and Koreman 1988: 222): Stress the penultimate syllable unless this syllable is open and the final syllable is closed; then stress the antepenultimate syllable.

(24)

Examples of stress assignment in Dutch (Lahiri and Koreman 1988: 222, Kager 1989: 232): a. Penultimate: VV VV VV valuta VC VV VV bombarie VV VC VV kwitanitie VC VC VV injectie VV VC VC detector VC VC VC Wilhelmus VV ViVj VC Poseidon b. Antepenultimate: VV VV VC monitor VC VV VC festival

vaː luː taː bom baː riː kwiː tan siː in jek siː de: tek tor wil hel mus po sei don

‘currency’ ‘hullabaloo’ ‘receipt’ ‘injection’ ‘detector’ ‘William’ ‘Poseidon’

moː niː tor fes tiː val

‘monitor’ ‘festival’

Lahiri and Koreman (1988: 222) observe that “stress usually falls on the penultimate syllable; however, if the final syllable is closed (-VC rhyme) stress falls on the penultimate syllable only if it is closed, otherwise it falls on the antepenultimate syllable. Clearly the stress assignment rule makes

Quantity in Germanic Languages

Table 5.8 Dutch vowels in stressed syllables Short (lax)

Long or short (tense) Long (loanwords only)

Diphthong

ɪʏ ɛɑ ç

i y u eː. øː oː aː

ɛi œy ʌu

iː yː uː ɛː œː ɔː ɛː̃ ɔ̃ː ɑː ɑ̃ ː

reference to weight distinctions between closed and open syllables.” Since the medial syllables in (24b) are not stressed even though the vowel is long, Lahiri and Koreman conclude that long vowels in Dutch are not bimoraic. Phonetic evidence supports Lahiri and Koreman’s (1988) conclusion that tense vowels are not bimoraic. The inventory of vowels that appear in stressed syllables is shown in Table 5.8 (based on van Oostendorp 1995, Gussenhoven 2000). As is the case in English, lax vowels are only found in closed syllables in Dutch. Tense vowels are phonetically long only when stressed. Unstressed tense vowels have the same duration as lax (short) vowels. High tense vowels are always short, even when stressed, unless they are followed by /r/. If a high tense vowel is followed by /r/, it is almost always long. Thus, the length of vowels in stressed positions depends upon vowel height and vowel tenseness as shown in (25). Note that the lax vowel in Dutch zit ‘sit 3 S G ’ and the high tense vowel in ziet ‘see 3 S G ’ have approximately the same duration. Tense high vowels are also short in open monosyllables as in (26). (25)

Length of vowels in stressed positions dependent on vowel height and tenseness (Gussenhoven 2000: 3, Nooteboom 1972: 25–47) [zit] ‘see 3S G ’

(26)

[zɪt] ‘sit 3S G ’

[zaːt] ‘seed’

High tense vowels remain short in open syllables (Mees and Collins 1983: 64, 68) i zie ‘see’ eː zee ‘sea’

y nu ‘now’ øː reu ‘dog (male)’ aː ja ‘yes’

u moe ‘tired’ oː zo ‘so’

Dutch appears to be different from Danish, English, and German. Tense vowels in Dutch do not appear to be underlyingly bimoraic. Because tense vowels do not have underlying quantity but Dutch appears to be weight sensitive for stress assignment, Hayes (1995) surmises that Dutch may have syllable cut as described by Trubetzkoy (1939). Page (2006) presents an analysis of Dutch as a syllable cut language. Van Oostendorp (2000: 43) notes that some dialects of Dutch have a contrast between short and long lax vowels. He gives the example of Tilburg, which has the following vowel inventory: i, y, u, e, ø, o, ɪ, ʏ, ɔ, ɛ, œ, ɒ, ɪː, ʏː, ɔː, ɛː, œː, ɒː.

113

114

B . R I C H A R D PAG E

5.5 Conclusion After our survey, we can now present a revised typology of quantity in Table 5.9 (adapted from Riad 1995: 165). We can identify differences among the languages within each category. Among the languages that preserve vowel and consonant quantity, Swiss German has geminates in initial position whereas the Scandinavian varieties do not. Among the languages that were originally classified as having consonant quantity, all have complementary quantity. Riad (2013) presents compelling evidence that consonant quantity is underlying and vowel quantity is derived in Swedish. However, Bannert (1976, 1977) and Kleber (2017) provide experimental perception and production data indicating that consonant quantity is not underlying in Central Bavarian but rather the contrast is based on the prosody of complementary length. In other words, they argue that VːC contrasts with VCː in Central Bavarian. Whether or not Central Bavarian should be classified as having underlying consonant quantity is unclear, and the quantity system of the dialect is undergoing change (Kleber 2017). The main change in the typology is the movement of Dutch to its own column as a language that has no underlying quantity for either vowels or consonants. It should be noted that some dialects of Dutch, such as Tilburg Dutch, do in fact have contrastive vowel length. Similarly, not all Swiss German dialects have retained both vowel and consonant quantity (see Seiler 2009 for an overview of German dialects). Further investigation will undoubtedly uncover more variation in the phonetics and phonology of quantity in Germanic.

Table 5.9 Typology of quantity (revised version) Vowel and consonant quantity Älvdalsmaë l

(Swedish dialect) Nord Gudbrandsdalska (Norwegian dialect) Western Nyland (FennoSwedish dialect) High Alemannic (Swiss German dialect) Highest Alemannic (Swiss German dialect) South Bavarian (Tyrolean German dialect)

Consonant quantity

Vowel quantity

Icelandic Faroese Swedish Norwegian Central Bavarian (BavaroAustrian German dialect)

English German Danish Tilburg (Dutch dialect)

Neither vowel nor consonant quantity Dutch

Quantity in Germanic Languages

References A´rnason, K. 2011. The Phonology of Icelandic and Faroese. Oxford University Press. Auer, P. and R. W. Murray 2004. “Bavarian isochrony without mora-counting,” Paper presented at Germanic Linguistics Annual Conference 10. Ann Arbor, MI. Bannert, R. 1976. Mittelbairsiche Phonologie auf akustischer und perzeptorischer Grundlage. Ph.D. dissertation, Lund University. Bannert, R. 1977. “Quantita¨t im Mittelbairischen: Komplementa¨re La¨nge von Vokal und Konsonant.” In W. U. Dressler and O. E. Pfeffer (eds.), Phonologica 1976: Akten der dritten Internationalen Phonologie-Tagung, Wien, 1–4. September 1976. Institut fu¨r Sprachwissenschaft der Universita¨t Innsbruck: 261–270. Basbøll, H. 2005. The Phonology of Danish. Oxford University Press. Basbøll, H. and J. Wagner 1985. Kontrastive Phonologie des Deutschen und Da¨nischen: Segmentale Wortphonologie und -phonetik. Tu¨bingen: Max Niemeyer Verlag. Becker, T. 1998. Das Vokalsystem der deutschen Standardsprache. Frankfurt am Main: Peter Lang Verlag. Chomsky, N. and M. Halle 1968. The Sound Pattern of English. New York: Harper & Row. Davis, S. 2011. “Quantity.” In J Goldsmith, J. Riggle, and A. C. L. Yu (eds.), The Handbook of Phonological Theory, 2nd edn. Malden, MA: Blackwell:103–140. Dieth, E. 1950. Vademekum der Phonetik. Phonetische Grundlagen fu¨r das wissenschaftliche und praktsiche Studium der Sprachen Unter Mitwirkung von Rudolf Brunner. Bern: Francke. Dresher, B. E. and A. Lahiri 1991. “The Germanic Foot: Metrical Coherence in Old English,” Linguistic Inquiry 22: 251–286. Duden 1990. Duden Aussprachewo¨rterbuch: Wo¨rterbuch der deutschen Standardaussprache, 3. edn. Mannheim: Dudenverlag. Fleischer, J. and S. Schmid 2006. “Zurich German,” Journal of the International Phonetic Association 36: 243–253. Fulop, S. 1994. Acoustic correlates of the fortis/lenis contrast in Swiss German plosives. Calgary Working Paper in Linguistics 16: 55–63. Gussenhoven, C. 2000. “Vowel duration, syllable quantity and stress in Dutch,” ROA-381. Halle, M. 1977. “Tenseness, vowel shift, and the phonology of back vowels in Modern English,” Linguistic Inquiry 8: 611–626. Hammond, M. 1997. “Vowel quantity and syllabification in English,” Language 73: 1–17. Harling-Kranck, G.1998. Fraº n Pyttis till Nedervettil. Helsingfors. Hayes, B. 1989. “Compensatory lengthening in moraic phonology,” Linguistic Inquiry 20: 253–306. Hayes, B. 1995. Metrical Stress Theory. University of Chicago Press.

115

116

B . R I C H A R D PAG E

Heusler, A. 1888. Der alemannische Konsonantismus in der Mundart von Baselstadt. Strassburg: Tru¨bner. Hinderling, R. 1980. “Lenis und Fortis im Bairischen,” Zeitschrift fu¨r Dialektologie und Linguistik 47: 23–51. Kager, R. 1989. A Metrical Theory of Stress and Destressing in English and Dutch. Dordrecht: Foris. Kiparsky, P. 2008. “Fenno-Swedish quantity: Contrast in Stratal OT.” In B. Vaux and A. Nevins (eds.), Rules, Constraints, and Phonological Phenomena. Oxford University Press: 185–219. Kleber, F. 2017. “Complementary length in vowel-consonant sequences: Acoustic and perceptual evidence for a sound change in progress in Bavarian German,” Journal of the International Phonetic Association: 1–22. doi:10.1017/S0025100317000238. Kohler, K. J. 1990. “German,” Journal of the International Phonetic Association 20. 1: 48–50. Kraehenmann, A. 2001. “Swiss German stops: Geminates all over the word,” Phonology 18: 109–145. Krech, E.-M., E. Kurka and H. Stelzig 1982. Großes Wo¨rterbuch der deutschen Aussprache. Leipzig: VEB Bibliographisches Institut. Kristoffersen, G. 2000. The Phonology of Norwegian. Oxford University Press. Kristoffersen, G. 2011. “Quantity in Old Norse and modern peninsular North Germanic,” Journal of Comparative Germanic Linguistics 14: 47–80. Kufner, H. L. 1957. “Zur Phonologie einer mittelbairischen Mundart,” Zeitschrift fu¨r Mundartforschung 25: 175–184. Lahiri, A. and J. Koreman 1988. “Syllable weight and quantity in Dutch,” Proceedings of the West Coast Conference on Formal Linguistics 7: 217–228. Lahiri, A. and A. Kraehenmann 2004. “On maintaining and extending contrasts: Notker’s Anlautgesetz,” Transactions of the Philological Society 102: 1–55. Lahiri, A., T. Riad, and H. Jacobs 1999. “Diachronic prosody.” In H. van der Hulst (ed.), Word Prosodic Systems in the Languages of Europe. Berlin and New York: Mouton de Gruyter: 335–442. Levander, L. 1925. Dalmaº let: Beskrivning och historia, Volume I. Uppsala: Appelbergs boktryckeri. Martens, C. and P. Martens 1961. Phonetik der deutschen Sprache. Munich: Hueber. Mees, I. and B. Collins 1983. “A phonetic description of Standard Dutch,” Journal of the International Phonetic Association 13: 64–75. Moulton, W. G. 1962. The Sounds of English and German. University of Chicago Press. Moulton, W. G. 1986. “Sandhi in Swiss German dialects.” In H. Andersen (ed.), Sandhi Phenomena in the Languages of Europe. Berlin and New York: Mouton de Gruyter: 385–392. Nooteboom, S. 1972. Production and Perception of Vowel Duration: A Study of Durational Properties of Vowels in Dutch. Rijksuniversiteit Utrecht dissertation.

Quantity in Germanic Languages

Odden, D. 2011. “The representation of vowel length.” In M. van Oostendoorp, C. J. Ewen, E. Hume, and K. Rice (eds.), The Blackwell Companion to Phonology. Malden, MA: Blackwell. Blackwell Reference Online, www.companiontophonology.com. Oostendorp, M. van 1995. Vowel Quality and Phonological Projection. Tilburg University dissertation. Oostendorp, M. van 2000. Phonological Projection: A Theory of Feature Content and Prosodic Structure. Studies in Generative Grammar 47. Berlin and New York: Mouton de Gruyter. Page, B. R. 2006. “The diachrony and synchrony of vowel quantity in English and Dutch,” Diachronica 23: 61–104. Penzl, H. 1955. “Zur Erkla¨rung von Notkers Anlautgesetz 2. Zeitschrift fu¨r deutsches Altertum und deutsche Literatur 86: 196–210. Pfalz, A. 1911. “Phonetische Beobachtungen an der Mundart des ¨ sterreich,” Zeitschrift fu¨r deutsche Mundarten. Marchfeldes in Nieder-O Pfalz, A. 1913. Die Mundart des Marchfeldes. Vienna: Ho¨lder. Sitzungsberichte der kaiserlichen Akademie der Wissenschaften in Wien. Philosophisch-historische Klasse, Vol. 170, 6th Abhandlung. Reis, M. 1974. Lauttheorie und Lautgeschichte. Untersuchungen am Beispiel der Dehnungs- und Ku¨rzungsvorga¨nge im Deutschen. Munich: Wilhelm Fink Verlag. Restle, D. 2003. Silbenschnitt – Quantita¨t – Kopplung. Munich: Wilhelm Funk. Riad, T. 1992. Structures in Germanic Prosody: A Diachronic Study with Special Reference to the Nordic Languages. Ph.D. dissertation, Stockholm University. Riad, T. 1995. “The quantity shift in Germanic: A typology,” Amsterdamer Beitra¨ge zur a¨lteren Germanistik 42: 159–184. Riad, T. 2013. The Phonology of Swedish. Oxford University Press. Rowley, A. A. 1989. “North Bavarian.” In C. V. J. Russ (ed.), The Dialects of Modern German: A Linguistic Survey. London: Routledge: 417–437. Sanders, W. 1972. “Hochdeutsch /a¨/ — ‘Ghostphonem’ oder Sprachpha¨nomen?,” Zeitschrift fu¨r Dialektologie und Linguistik 93: 37–58. Schaeffler, F. 2005. Phonological Quantity in Swedish Dialects: Typological Aspects, Phonetic Variation and Diachronic Change. Ph.D. dissertation, Umeaº University, Umeaº , Sweden. Scheutz, H. 1985. Strukturen der Lautvera¨nderung: Variationslinguistische Studien zur Theorie und Empirie sprachlicher Wandlungsprozesse am Beispiel des Mittelbairischen von Ulrichsberg/Obero¨sterreich. Schriften zur deutschen ¨ sterreich 10. Vienna: Wilhelm Braunmu¨ller. Sprache in O Seiler, G. 2005. “On the development of the Bavarian quantity system,” Interdisciplinary Journal of Germanic Linguistic and Semiotic Analysis 10: 102–129. Seiler, G. 2009. “Sound change or analogy? Monosyllabic lengthening in German and some of its consequences,” The Journal of Comparative Germanic Linguistics 12: 229–272.

117

118

B . R I C H A R D PAG E

Siebs, T. 1969. “Siebs. Deutsche Aussprache: Reine und gema¨ßigte Hochlautung mit Aussprachewo¨rterbuch.” In H. de Boor, H.Moser, and C. Winkler (eds.), 19. edn. Berlin: Mouton de Gruyter. Sievers, E. 1876. Grundzu¨ge der Lautphysiologie zur Einfu¨hrung in das Studium der Lautlehre der indogermanischen Sprachen. Leipzig: Breitkopf & Ha¨rtel. Spiekermann, H. 2000. Silbenschnitt in deutschen Dialekten. Tu¨bingen: Max Niemeyer Verlag. Spiekermann, H. 2002. “Ein akustisches Korrelat des Silbenschnitts: Formen des Intensita¨tsverlauf in Silbenschnitt- und Tonakzentsprachen.” In P. Auer, P. Gilles, and H. Spiekermann (eds.), Silbenschnitt und Tonakzente, Linguistische Arbeiten 463. Tu¨bingen: Max Niemeyer Verlag: 181–200. Suzuki, S.1995. “The decline of the foot as a supersyllabic mora-counting unit in Early Germanic,” Transactions of the Philological Society 93: 227–272. Trubetzkoy, N. S. 1939. Grundzu¨ge der Phonologie. Prague: Travaux du cercle linguistique de Prague 7. Vennemann, T. 2000. “From quantity to syllable cut: On so-called lengthening in the Germanic languages,” Italian Journal of Linguistics / Rivista di Linguistica 12: 251–282. Wa¨ngler, H-H. 1974. Grundriß einer Phonetik des Deutschen: Mit einer allgemeinen Einfu¨hrung in die Phonetik. Marburg: Elwert. Wiese, R. 1996. The Phonology of German. Oxford University Press Wiesinger, P. 1990. “The central and southern Bavarian dialects in Bavaria and Austria.” In C. V. J. Russ (ed.), The Dialects of Modern German. London: Routledge: 438–519. Willi, U. 1995. “‘Lenis’ und ‘fortis’ im Zu¨richdeutschen aus phonetischer Sicht.” In H. Lo¨ffler (ed.), Alemannische Dialektforschung. Bilanz und Perspektiven. Tu¨bingen and Basel: Francke: 253–265. Willi, U. 1996. Die segmentale Dauer als phonetischer Parameter von ‘fortis’ und ‘lenis’ bei Plosiven im Zu¨richdeutschen. Eine akustische und perzeptorische Untersuchung. Stuttgart: Steiner. Winteler, J. 1876. Die Kerenzer Mundart des Kantons Glarus in ihren Grundzu¨gen dargestellt. Leipzig: Winter.

Chapter 6 Germanic Laryngeal Phonetics and Phonology Joseph Salmons

6.1 Introduction This chapter examines Germanic laryngeal phonetics and phonology, focusing on assimilations and final neutralization. The contrast between sounds written b versus p or v versus f and their various assimilations (like English -s plurals, possessive -s or -ed past tense forms) and final neutralizations (in German or Dutch) are textbook examples in phonology, yet on closer examination these inform fundamental issues in the study of sound systems. The chapter is organized as follows: Section 6.2 provides definitions and examples; Section 6.3 presents theoretical underpinnings, including Laryngeal Realism; Section 6.4 surveys selected patterns across Germanic; Section 6.5 notes open issues and directions for future research; and Section 6.6 summarizes laryngeal phonetic and phonological differences across Germanic.

6.2 The Basics This section describes the basic physiology of laryngeal phonetics and phonology and introduces laryngeal contrasts. Human beings use the larynx (‘voice box’) in speaking. This “complex structure composed of cartilages, muscles and various related tissues” was first described in detail by Leonardo Da Vinci (Kent 1997: 100, a good source on the relevant human anatomy). It contains I thank the editors for the opportunity to develop ideas Greg Iverson and I explored over two decades. For feedback, I owe Brent Allen, Jackson Crawford, Stefan Dollinger, Patrick Honeybone, Pavel Iosad, Greg Iverson, Neil Jacobs, Seetha Jayaraman, James Kirby, Sam[antha] Litty, Monica Macaulay, David Natvig, Tom Purnell, Juho Pystynen, Eric Raimy, and Paul Roberge. The usual disclaimers apply. Cited examples retain original orthography.

120

JOSEPH SALMONS

the vocal folds, pieces of mucous membrane on either side of the throat which can be opened to allow free airflow for breathing or brought together in different configurations to shape speech sounds. The opening between the folds is called the glottis. For present purposes, two states of the glottis are important. The first is when the folds are held close to one another so that, with sufficient airflow from the lungs, they vibrate steadily. This position is typical of vowels and sonorants in the world’s languages, an effect called ‘voicing’. The second is when the folds are held apart to inhibit vibration, creating, among other things, aspiration after word-initial /p, t, k/ in languages like Norwegian, German and English. (Usage varies by tradition, but many people use ‘phonation’ for how these states of the glottis shape speech sounds.) I focus on two phonological features, those traditionally associated with voicing and aspiration.1 Henton et al. (1992: 96) observe that “some contrasts, most notably those for phonation, use a large number of cues for each distinctive feature.” In Germanic and elsewhere, those cues include spectral information like changes in vowel formants next to obstruents (see Kingston and Diehl 2008 on a ‘low-frequency property’ associated with voicing) and various patterns of durational difference (Henton et al. 1992). Lisker (1986) lists 16 acoustic properties associated with voicing in English /b/ and expressly notes that the list is incomplete. In short, even where only one or two phonological features are involved, they can have a plethora of phonetic correlates. The main phonetic measurement, especially since Lisker and Abramson 1964, has been Voice Onset Time (VOT). For stops, many researchers take the release of closure as a point of reference, when the lips part at the beginning of bad or when the tongue tip moves away from the alveolar ridge at the beginning of times. With aspiration, a gesture opening or spreading the glottis can inhibit vibration from the vocal folds coming together, creating a lag between release and voicing, or positive VOT. Vocal fold vibration before stop release is called prevoicing or negative VOT. If we have a gesture spreading the glottis, that can be timed variably relative to stop closure and release. The gesture can start before closure, creating preaspiration, for instance, systematically in several North Germanic varieties and variably in some West Germanic ones. This discussion assumes regular, robust vocal fold vibration, but the physics and physiology are complicated. First, voicing can take place by virtue of the active configuration of the vocal folds. But with sufficient airflow and no particular laryngeal gesture, we can get vibration, ‘passive voicing’, such as when an English stop without laryngeal specification comes between two vowels, e.g., in hobby. 1

I leave aside discussion of glottalization in English, Danish, and West Jutlandic stød, creaky voice, and relationships between laryngeal systems and quantity, such as the Bavarian Quantity Law.

Germanic Laryngeal Phonetics and Phonology

Second, oral closure downstream from the glottis makes voicing difficult: Voicing requires a pressure drop and enough air flow to create reliable vibration, yet air flow is by definition stopped downstream – pushing air through the glottis without an oral escape route. Finally, note that other configurations are possible and used in Germanic, if not generally for contrast, e.g., creaky voice (‘vocal fry’). I focus on obstruents, consonants with enough constriction above the glottis (i.e., in the mouth, ‘supralaryngeally’) that airflow is impeded to cause frication (fricatives) or to stop airflow entirely (oral stops). In some languages, the larynx is not phonologically relevant to obstruents: In Hawaiian and Menominee, only one stop series exists, so /p/ lacks a counterpart /b/ or any other labial oral stop. Other languages – Otomı´, Igbo, and Beja – have up to six different laryngeal contrasts, involving voicing, aspiration and glottalization, and combinations thereof. Almost all Germanic languages show two series, crosslinguistically the most common pattern, found in 162 of 317 languages surveyed by Maddieson (1984: 26). These are usually written in the Latin alphabet as versus , etc., and versus , etc. As discussed momentarily, we have reasons to analyze the distinction in different Germanic and other languages in two different ways: Using traditional symbols for the nonce, some languages actively spread the glottis on /p, t, k, f, s/ to contrast with plain or unmarked /b, d, g, v, z/; others instead use vocal fold vibration on /b, d, g, v, z/ to produce the distinction, leaving the other series unmarked. Here, I refer to both systems as laryngeal contrasts and the former as involving Glottal Width or [spread], yielding aspiration, and the latter Glottal Tension or [slack], yielding voicing.2 The three types – ‘plain’, aspirated and voiced – are again the most widely reported series cross-linguistically (Maddieson 1984: 27). Assimilation is possibly the most common process in sound systems, where sounds, typically adjacent ones, become more like each other (see Hall, Chapter 1). In laryngeal contrasts, we find consistent agreement in ‘static’ (i.e., non-alternating) clusters within syllables, where English allows [sp] and marginally allows [zb] (spot, Sbarro) but not *[zp] or *[sb]. Alternations provide clearer evidence. Famously, English’s limited morphological paradigms show very systematic laryngeal alternations, illustrated here with plurals. I use phonetic brackets since English orthography does not capture the pattern, writing [s] for sounds specified as [spread glottis] and [z] for those unmarked, written out as empty brackets [ ] (explained below):

2

The names for features vary by tradition and phonetic or phonological assumptions. In particular, some use “fortis” for [spread], contrasting it with “lenis”, i.e., [

].

121

122

JOSEPH SALMONS

(1)

English plurals a. slip slip[s] slit slit[s] slick slick[s] b. bib bib[z] bid bid[z] big big[z]3 c. lease leas[əz] Liz Liz[əz] leash leash[əz]

Analyses vary but the most common approach assumes that the underlying representation of the plural marker is /z/ (Zwicky 1975), in traditional terms ‘voiced’ (= [ ] for us). The vowel insertion in (c) is not a direct concern but it presumably prevents sequences of similar consonants. For now, in our terms, the sibilant becomes [spread] by assimilation following an obstruent marked for [spread], which extends into the empty [ ] (in 1a). German and many other languages allow a distinction between the two laryngeal series in initial and medial position, but not finally, as in (2). (2)

Positional laryngeal contrasts in Standard German a. Gabel ‘fork’ ≠ Kabel ‘cable’ b. Kragen ‘collar’ ≠ Kraken ‘octopuses’ c. Sack [zak] ‘bag’ = Tag [taːk] ‘day’4 d. Sa¨cke ‘bags’ ≠ Tage ‘days’

Here too, alternations provide key evidence; both Sack and Tag end with a surface [k] in the standard language, but differ when inflection adds a final vowel: Ta[g]e. Let us now contextualize this background within current phonological theory.

6.3 Theory and Analysis We first need to take a stance on some fundamental issues in laryngeal phonology. The first issue is what features are involved and the second the nature of featural representation. All phonologically relevant states of the glottis for human languages can be captured using only three phonological dimensions: Glottal Width, Glottal Tension, Larynx Height (not discussed here) in the various attested combinations (Iverson and Salmons 1995, others). This system is developed by Avery and Idsardi (2001: 42, passim) as in (3). 3

The noun big refers to “inside” players in basketball, e.g., center or power forward.

4

For historical reasons, few German pairs have underlying laryngeal contrasts with the same preceding vowel length.

Germanic Laryngeal Phonetics and Phonology

(3)

Avery and Idsardi’s 2001 model of laryngeal phonology Articulators

Dimensions

Gestures [spread]

Glottal Width [constricted] [stiff] Laryngeal

Glottal Tension [slack] [raised] Larynx Height [lowered]

This model seeks to make minimal assumptions about contrast, based on P H O N O L O G I C A L A C T I V I T Y , defined by Dresher and Zhang (2005: 52) as cases “for which we have positive evidence based on participation in phonological processes.” Consider an example of place features: Many see /t, d, n/ as unmarked for place ([ ]) but /p, b, m/ as marked for [labial]. On that view, [labial] can spread into [ ] but not vice versa, accounting for patterns like in+possible = impossible. In contrast, where we see m becoming n, these are instances of feature loss, like the widespread reduction of Germanic -m to -n in final position historically. Activity crucially involves the addition of structure while reduction involves its loss. I follow Avery and Idsardi’s terminology (except when discussing authors who use [voice]) and transcribe to reflect this phonological analysis. That is, /p, t, k, s/ refer to laryngeally unmarked segments, /b, d, g, z/ imply the presence of a [slack] gesture, and where a [spread] gesture is posited, I use /pʰ, tʰ, kʰ, sʰ/. For example, English thank is rendered as /θʰæːŋkʰ/ and do as / tuː/, German danke as /tɑŋkʰə/ and Dutch bedankt as /bədaŋkt/. Here I will draw on assimilations, but there is other evidence of activity in Germanic (see discussion of Scots, below in this section). Co-occurrence restrictions and other patterns provide evidence in other languages. Gestures are added during ‘completion’, and for English or German, Glottal Width is typically completed with [spread]. From there, contrasts can be enhanced or overdifferentiated phonetically by using a different dimension (predicted in the quote from Henton et al., Section 6.2). Avery and Idsardi develop the example of Japanese, a Glottal Tension language, enhancing with Glottal Width (2001: 53–54). This is captured in Vaux’s Law (Vaux 1998), which “entails that the unmarked fricative will acquire the dimension of Glottal Width, the default gesture of which is [spread]” (Iverson and Salmons 2003b). Similarly, Glottal Width systems can enhance laryngeal contrasts with Glottal Tension on the unmarked member, as exemplified below. This enhancing phonetic activity is layered on top of the basic contrast, which remains simply Glottal Tension. Enhancement is by definition not necessary and languages like Icelandic employ extremely

123

124

JOSEPH SALMONS

little glottal pulsing in obstruents. (See Keyser and Stevens 2006 on phonetic enhancement and D. C. Hall 2011 on the phonology.) Our second issue is the nature of phonological representation, how dimensions or features are deployed. In many classic works, many or all features are binary (‘equipollent’), that is, have plus or minus values, e.g., traditional [+voiced] versus [-voiced]. Modern work pursues various kinds of underspecification, where some values are left blank. Much work since Lombardi (1995) treats laryngeal features as ‘privative’, either having a value or lacking one. This ‘unmarked’ or ‘plain’ value means that any specification is absent. (Wetzels and Mascaro´ 2001 argue against this, countered in Iverson and Salmons 2003a, Brown 2016.) Hestvik and Durvasula (2016) provide neurolinguistic evidence that phonological representation is sparser than phonetic representation, see also Park et al. 2010. In a language like Dutch, whose laryngeal system is built fundamentally around Glottal Tension, this yields a contrast between a [slack] gesture on /b/ and literally no laryngeal specification on /p/, written as [ ]. In formal systems, economy and parsimony are valued, speaking for this approach. In developing Dresher’s Contrastive Hierarchy (2009), Oxford (2015: 311) argues that “Privativity . . . makes the model maximally restrictive, since it predicts that only the marked values of contrastive features will be phonologically active.” We will see evidence in favor of this view below. With those notions in hand, let us consider Laryngeal Realism, a term coined by Honeybone (2002, more widely known from Honeybone 2005) to capture the distinction made above, that languages like German (and around the world in Cantonese, Somali, and others) use Glottal Width completed with [spread] while languages like Dutch (and Hungarian, most Slavic and Romance languages) use Glottal Tension completed with [slack].5 This contrasts with the tradition, associated with work after Lisker and Abramson 1964 through Kingston and Diehl 1994 and beyond, which treats at least all of the Germanic languages as employing one feature, [voice], analyzed as unified phonologically with differences between our two types all carried on the back of the phonetics. More generally, Laryngeal Realism connects phonological representation with phonetic realization, but in a mediated way; the two levels neither completely disconnected from one another nor one predetermined by the other. As Eric Raimy put it recently, “Phonetics provides the menu from which phonology chooses.” Fully understanding sound patterns requires attention to both. If we accept the above-described set of laryngeal gestures, their privativity and the notion of Laryngeal Realism, the picture for many Germanic systems can be characterized as in (4). Note the mismatch with orthographic representation in the first column. Again, /p, t, k, s/ mean laryngeally unmarked, /b, 5

The roots of Laryngeal Realism go back far earlier than the work usually cited. Van Rooy and Wissing (2001) call attention to Jakobson’s 1949 “narrow” interpretation of [voice], while the long German tradition of using “fortis” and “lenis” connects here as well (e.g., Sievers 1893).

Germanic Laryngeal Phonetics and Phonology

d, g, z/ imply the presence of a [slack] gesture, and where a [spread] gesture is posited, I use /pʰ, tʰ, kʰ, sʰ/. (4)

Realist view of [spread] and [slack] systems in Germanic English Dutch

/p/ [ ]

[ ]

/b/

/ph/

[spread]

[slack]

What is the justification for this distinction? We’ve already seen that there is a P H O N E T I C correlation in terms of actual aspiration and voicing: English-type languages show consistent lack of pulsing on [spread] obstruents and presence of aspiration on stops, with systematic phonological nuances discussed below. In contrast, the unmarked series shows great variation. English speakers vary regionally, stylistically, and otherwise in pulsing on these, while Icelandic generally lacks pulsing. Dutchtype languages show consistent pulsing on the [slack] segments (save for final devoicing and such). For P H O N O L O G Y , I have already mentioned the notion of ‘active’ features. An old-school [voice] analysis of English or German requires an aspiration rule for /p, t, k/ in the onset of many stressed syllables. Yet there is little or no aspiration in clusters of s +stop (p[h]ot versus spot, German P[h]ass ‘passport’ versus Spaß ‘fun’), which must be excepted from the rule, and no aspiration in stops followed by a sonorant, where instead the sonorant is devoiced: p[l̥ ]an, t[r̥ ]ain, s[n̥ ]eeze or P[l̥ ]an ‘plan’, t[r̥ ]inken ‘drink’, Sch[n̥ ]ee ‘snow’. That is, these forms require an additional rule of sonorant devoicing. Beyond saving us aspiration and sonorant devoicing rules, a [spread] analysis allows for sharing of the spread glottis gesture between the /s/ and stop, and sonorant devoicing falls out from the presence (and timing) of [spread] in the initial obstruents (Iverson and Salmons 1995). These types of phonological activity should, on many views, be encoded in the specification of the obstruents. While early arguments were built on speaker intuitions of such patterns, they hold up in phonetic investigation. In German, voicelessness spreads progressively (rightward) over word boundaries where the first word ends in a [spread] obstruent and the second begins with an unspecified one. Taking data from test sentences containing (traditional) /t/ + /z/ or /v/ (e.g., “Benno hat Wa¨lder und Seen gemalt”), Kuzla et al. (2007: 316) conclude that “A preceding voiceless obstruent (e.g., /t/) triggers assimilatory devoicing of /v, z/.” (See Jessen 1998 for a fuller accounting of patterns and previous research.) In German and English, laryngeal assimilations change unmarked segments to [spread] and only where two unmarked obstruents are adjacent do we find what are traditionally called ‘voiced’ clusters. In plurals, in (1) above, past tense and possessive forms, in (5), or other English inflection, that is the case, based on the traditional analysis positing /d/ or /z/ rather than /t/ or /s/ (Zwicky 1975 and work since).

125

126

JOSEPH SALMONS

(5)

Past tense and possessive forms a. walk walked [kt] jog jogged [gd] skate skated [təd] b. the book’s title the blog’s title the issue’s title

[ks] [gz] [ʃuz]

Here a simple progressive (rightward) spread of [spread] yields the correct outcome and the patterns comport with contractions (let us call this cliticization) like what’s and it’s, both with [ts] where the fricative is clearly underlyingly /z/. In privative terms on a traditional ‘voice’ account, as argued by Iverson and Salmons (1999), these facts “cannot even be described, because there is no way to refer to the absence of a feature under a privative theory of representation.” English also shows limited patterns of voicelessness spreading leftward (regressively), as in these examples from derivational morphology and connected speech (again starting from Iverson and Salmons 1999): (6)

Regressive assimilations in English a. In derivation describe description recede recession b. Over word boundaries have to [ft] has to [st]

The direction of assimilation in English and German is progressive, but the same phenomenon spreads. All this strongly suggests that [spread] is the active feature. Abercrombie (1967: 135–136) provides an interesting counterexample from “speakers of educated Scots,” who show regressive assimilation to [slack] in blackboard [gb] and birthday [ðd]. He concludes that “Such regressive assimilations of voice appear to be found in no other kind of English, though they are the regular thing in French, Dutch, and several other languages.”6 Iverson and Salmons 1999 analyze Scots as a Glottal Tension variety based on these assimilation patterns supported by basic phonetics, such as unaspirated /p, t, k/. Very different evidence comes from Youssef (2010), where a [slack] specification (his [Lowered Larynx]) blocks a lowering harmony. Van Rooy and Wissing (2001: 325) raise regressive voice assimilations to the status of a general principle: 6

This claim seems to hold up in recent phonetic work, though with some variability, as Snoeren et al. (2006: 262) conclude for French that “We consistently found that underlyingly voiceless word-final stops more easily shift toward voiced stops than voiced stops devoice.” For Dutch, see contributions to van der Weijer and van der Torre 2007.

Germanic Laryngeal Phonetics and Phonology The distinctive feature [voice] automatically entails the presence of regressive voicing assimilation in a particular language, provided that the feature [voice] is interpreted in the narrow sense of the actual presence of vocal fold vibration. This is evident from the widespread occurrence of regressive voicing assimilation in languages of the world that all happen to share the distinctive feature voice.

This generalization provides a key indication that languages using the gestures [spread] and [slack] differ phonologically as well as phonetically, and in a way that is not intuitively obvious: Our notion of phonological activity predicts what should spread, but not the direction of spread in and of itself. Starting from Lombardi’s (1995) analysis, Zonneveld (2007) summarizes Dutch data on regressive voice assimilation, including forms like those in (7). (7)

Dutch regressive voicing assimilation (Zonneveld 2007: 10) voi | LAR | strop-das [b – d] ‘necktie’

voi voi voi Voice tier | | | LAR LAR LAR Laryngeal node | | | bloed bank hand - palm [d – b] [t – p] ‘blood-bank’ ‘palm of the hand’

These data can be accounted for privatively by first applying final devoicing (in our terms, removing the gesture [slack]) to forms like bloed or hand and then applying ‘Spread-voice’ regressively, thus voicing and revoicing the first two examples. Numerous languages, including Dutch, show stop-fricative asymmetries in their laryngeal behavior and the behavior of the preterit marker -de is especially difficult (e.g., Zonneveld 2007). While Dutch /b, d/ show expected phonetic voicing (again, there is no /g/) and assimilation patterns, fricative /z/ shows little pulsing and no regressive voice assimilation. /ɣ/ has so little pulsing that Allen (2016) suggests that contrast with /x/ may have been neutralized for some speakers. Moreover, he and others find relatively long VOTs on /p, t, k/ and some sonorant devoicing in clusters. Allen lays out a role for enhancement in the [ ] series and the set of cues exploited for the laryngeal distinction in Dutch appears particularly rich. While Netherlandic language history motivates the established asymmetries, work like Pinget 2015 suggests that sound change may be underway at present. It should be clear by now that by no means was Laryngeal Realism posited simply on surface phonetic data, that is, that languages with aspirated /p, t, k/ are [spread] and those with modally voiced /b, d, g/ are [slack]. Phonetic evidence alone would be woefully inadequate for deciding the question. Some Germanic varieties, like my own Southern US English, ‘overmark’ the contrast with prevoicing on phonologically

127

128

JOSEPH SALMONS

unspecified /b, d, g/ as phonetic enhancement (see Jacewicz et al. 2009, or Docherty 2011 on British English). More important are patterns of laryngeal assimilation. My system of contrast is the same as that of other English speakers, as made clear by the fact that my assimilation patterns (in plurals and past tense forms, etc.) are those of the broader community. The case for Realism has been built on a broad set of evidence, including phonetic evidence beyond aspiration and glottal pulsing (like sonorant devoicing), phonology (assimilation patterns in particular), and patterns of variation, and integrated into typology (as already introduced), e.g., Iverson and Salmons 1995. Diachronic evidence also strongly supports a realist interpretation of Germanic laryngeal history (see Honeybone 2002, 2005; Iverson and Salmons 2003a, 2007, 2008, others). This chapter underscores the breadth of evidence involved. Turning to final neutralization, we can define it simply as the loss of phonological distinction at some right edge, often word finally or in codas. German spelling reflects underlying forms, as in (2c) and (2d), or familiar examples like Rad ‘wheel, bike’ versus Rat ‘advice’, both with [t]. In the traditional broad interpretation of [voice], the name ‘final devoicing’ is warranted across the family, i.e., the analysis is removal of a specification of [voice]. Realism provides a richer set of options.7 Neutralizing Glottal Tension varieties do involve removal of [slack] (that is, in some sense actual devoicing), but Glottal Width varieties can neutralize either to the unmarked, by removal of [spread], or by the addition of [spread], a final fortition and an analysis implied in the indigenous German name for the process in that language, Auslautverha¨rtung, literally ‘final sound hardening.’8 It is clear that some languages of the world neutralize by insertion of [spread] to unmarked segments, including Kashmiri, Eastern Armenian, and Kashaya. Iverson and Salmons (2011) argue that German shows this pattern. (Some Germanic varieties show aspiration of final /p, t, k/, though not with neutralization, e.g., Swedish, per Riad 2014: 47.) Today, some varieties of American English are developing patterns of final laryngeal neutralization, see Purnell et al. (2005a, 2005b), so that his comes to merge with hiss, apparently by removal of [spread] (like Odenwald German, Section 6.4.3). Harris (2009) argues that final devoicing is lenition, a reasonable position for feature deletion but more difficult for insertion of [spread]. On this view, then, Germanic final neutralization falls into three distinct types, shown in (8), one feature addition and two feature removal.

7

Whether neutralization by final voicing exists in human language is unclear and debated between Kiparsky 2006 and Blevins 2006 (connected pieces in the same journal issue), but it appears irrelevant to Germanic. See Iverson and Salmons 2011 on the typology of final neutralization.

8

The claim that a prosodically weak position like word-final position universally prohibits feature insertion is empirically incorrect (see Iverson and Salmons 2011).

Germanic Laryngeal Phonetics and Phonology

(8)

Typology of Germanic final laryngeal neutralization Dutch German Upper Midwestern English

Removal of [slack] Addition of [spread] Removal of [spread]

The history of German Auslautverha¨rtung has been investigated in most detail by Mihm (2004, 2007). Into the modern period, some varieties showed final distinctions and some varieties neutralize by removal of [spread] (see Section 6.4.3).

6.4 Comparative Overview Issues of laryngeal phonetics and phonology have been extensively studied for all standardized modern Germanic languages as well as many dialects, even if rigorous comparative work is still rare (though see Allen 2016 for a comparison of Norwegian, Dutch, and Swiss German). This section sketches how particular varieties fit into the larger picture, covering each of the three branches of Germanic (East, North, and West) and illustrating both well-known and more recently investigated patterns in several languages: [spread] in Gothic, Swedish and Norwegian, Icelandic, and [slack] in Yiddish, West Frisian, and other systems found in some German dialects and Swiss German.

6.4.1 East Germanic For practical purposes East Germanic is only represented by Gothic. Based on comparative evidence, Gothic is treated as a [spread] language with aspiration, though there is no evidence for this beyond spirant alternations in codas, like giban ‘to give’ ~ fragift ‘betrothal’, though another passage spells the same cluster (Miller 2019). We lack, for instance, clear evidence of progressive assimilation of [spread]. Gothic fricatives are written as in medial positions versus in final position (leaving aside the velar). This is interpreted first as reflecting a stop-spirant alternation, where stops /b, d/ were realized as fricatives [β, ð] intervocalically, and secondly as a fricative fortition or devoicing process in final position as in (9a–9c), examples and discussion after Roberge 1983. (9)

Gothic fricative alternations and exceptions a. bidjan ‘to ask for’, inf. baþ 1–3.sg.pret. b. riqizis ‘darkness’, gen.sg. riqis nom.sg. c. hlaiba ‘bread’, dat.sg. hlaifs nom.sg. d. twalib ‘twelve’ for expected twalif haubid ‘head’ for expected haubiþ minz ‘less’ for expected mins

129

130

JOSEPH SALMONS

This pattern is widespread but shows over 200 exceptions, concentrated in particular parts of the small corpus, illustrated in (d). The exceptions always involve reflexes of *b, *d, *z, rather than underlyingly ‘voiceless’ segments, so that qiþan ‘to speak’, inf., only shows the 1–3.sg.pret. form qaþ, never *qad (Roberge 1983: 110). Roberge concludes that the exceptional spellings are “vestiges of an older, presumably Wulfilian usage” (1983: 149). That is, Wulfila’s phonology had a laryngeal contrast in final position and neutralization took place in the scribal period, where scribes incompletely reconciled the differences between the Vorlage they worked from and their own Gothic phonology. Teasing out this chronology of change between the translation and the later manuscript we have today shows a nuanced form of familiar patterns of final neutralization, affecting fricatives but not stops. Far more exotic in Gothic is the apparent laryngeal dissimilation in nonadjacent obstruents, known as Thurneysen’s Law, illustrated in (10) (after Collinge 1985: 183 and Suzuki 1992) with variants of the suffix reconstructed by Lehmann (1986: 110) as *-uƀni̯ a. (10)

Thurneysen’s Law fastubni ‘fasting’ fra´istubni ‘temptation’ witubni ‘knowledge’

waldufni ‘might’ wundufni ‘wound(ing)’

One workable formulation of the generalization is Suzuki’s (1992: 29), that fricatives “following an unstressed vowel . . . dissimilated in a handful of derivational suffixes with the preceding root final consonant with respect to the feature [± voice].” That is, fricatives in certain suffixes changed their laryngeal specification to D I S A G R E E with the specification of the last consonant in the root. While some examples certainly follow this pattern, there are many exceptions and what we see in the Gothic evidence reflects at best a shadow or compromised pattern. Suzuki reasonably stresses the interaction of the process with the fricative pattern discussed above as well as the (again, partial) loss of Verner’s Law, further complicated by morphological transparency/opacity. Collinge (1985: 183) opens his discussion of Thurneysen’s Law by declaring that “Gothic is maddening in the variability in respect of voicing shown by its spirant consonants.” As Roberge shows, sufficient care in sifting the evidence can provide real progress, though much remains to be explained.

6.4.2 North Germanic All North Germanic systems appear to be [spread]. Two examples show informative patterns, Swedish and Icelandic. Swedish patterns are illustrated in (11) with preterit forms and past participles below, assuming underlying forms /-de2/ and /-te/ respectively (Riad 2014: 102–103).

Germanic Laryngeal Phonetics and Phonology

(11)

Assimilation in Swedish (slightly altered) Verb

Inflected

Form

Pattern

a.

bygga /bygμ-a2/ [2'bʏgːa]

bygg-de [2'bʏgːdɛ]

pret.

agreement

b.

‘to build’ ko¨pa /ɕøp-ɑ2/ [2'ɕøːpha] ‘to buy’

bygg-t [bʏkːt] ko¨p-te [2'ɕø̞ hpːtɛ] ko¨p-t [ɕø̞ hpːt]

p. part. pret. p. part.

regr. assim. agreement regr. assim.

While this looks like a simple case for [spread], another view treats Swedish as overmarked, requiring both active [spread] and [voice] to capture the evidence. Helgason and Ringen (2008) present data that Central Standard Swedish stops are realized with aspiration on one series and voicing on the other, leading them to conclude that Swedish contrasts involve both [spread] and [voice] as active features: “If prevoicing implicates the feature [voice] in phonetic representations and aspiration implicates [sg] [=[spread]], then Swedish would appear to have both” (2008: 625). However, recall the discussion in Section 6.3 regarding enhancement, where a laryngeally unmarked series can be enhanced with phonetic voicing. If we move beyond the basic phonetics, Helgason and Ringen report laryngeal assimilations typical of simple [spread] systems, testing clusters of various configurations like those above, covering, in their terms, fortis + fortis, fortis + lenis, lenis + fortis and lenis + lenis: “Stop clusters that have underlying (or historically) mixed voicing are entirely voiceless. Only clusters in which both stops are lenis are voiced” (2008: 622). In short, [spread] is phonologically active and voicing inert, a phonetic enhancement of the contrast (see also Riad 2014: 102).9 Beckman et al. (2011) provide another kind of argument for using both features for Swedish. Previous work shows that changes in speaking rate correlate with active features: “Slowing down causes longer aspiration in aspirating languages and longer prevoicing in voicing languages but no change in short-lag stops” (2011: 39). They find that both aspiration and prevoicing increase in Swedish slow speech, from which they conclude that both features are active. What has not been shown yet is that this is a phonological rather than a phonetic effect. That is, is the correlation with speaking rate an effect of phonological specification or a phonetic enhancement? Norwegian patterns much like Swedish, though it seems that there is less phonetic voicing. Notable is the bidirectional spread of voicelessness, discussed by Kristoffersen (2000: 74–87). Icelandic obstruents are voiceless phonetically, with stops distinguished by the presence or absence of pre- or postaspiration (Hansson 2003: 49–50, A´rnason 2011: 99). Aside from /v/ and /j/, likely better

9

See Pavel (2017: 208–225) for related arguments with Breton data.

131

132

JOSEPH SALMONS

analyzed as approximants, fricatives do not show a laryngeal contrast initially (A´rnason 2011: 106–107). Phonetic voicing is found only intervocalically and even there, /s/ does not normally voice. This presumably means that even the broadest understanding of [voice] cannot be invoked for this language. Example (12a) shows the contrast in initial position, while (12b) shows neutralization differing by dialect, going to the unaspirated in the Southern (or “soft”) dialect but to the aspirated in the Northern (or “hard”) dialect. Intervocalic (here inter-sonorant) context is a common one for neutralization (e.g., in many German dialects), but the phonological neutralization going by dialect in two different directions, seems less common, reminiscent of the distinction drawn above between final neutralization by feature removal versus insertion. (12)

Aspiration in Icelandic (from Hansson 2003: 51) a. ['thiː.na] tı´na ‘to pick’ ['tiː.na] dy´na ‘mattress’ kra´ ‘pub’ ['khrau] ['krau] gra´ ‘gray’ (fem.) b. Southern Northern ['svɪːpa] ['svɪːpha] svipa ‘whip’ hetja ‘hero’ ['hɛːtja] ['hɛːthja] ['sɪːkrɪ] ['sɪːkhrɪ] sykri ‘sugar’ (dat.sg.)

A´rnason writes that the “clear opposition between fortis (hard) and lenis (soft) consonants is only fully realized before a vowel and in the onset of word-initial syllables” (2011: 104). A´rnason goes on to describe a pattern of ‘final devoicing’ to capture alternations like dagur ‘day’ nom.sg. versus dag ‘day’ acc.sg.: [taːɣʏr̥ ] versus [taːx] (2011: 236–237, elsewhere). More interesting though is prepausal context which can affect final obstruents (2011: 237, 302–303), and which A´rnason sees as a marker of the right edge of prosodic boundaries. Perhaps the opposite pole within North Germanic is Danish, a clear [spread] language with rampant lenition (featural reduction) in its unmarked obstruents.

6.4.3 West Germanic This branch shows wide diversity in its laryngeal phonetics and phonology. As sketched above, much of West Germanic uses Glottal Width as its dimension and a [spread] gesture (German and most kinds of English, along with Low German, North and East Frisian). Here I sketch some varieties that are Glottal Tension with [slack] (Yiddish, West Frisian) and then two varieties without a laryngeal contrast at all.

Germanic Laryngeal Phonetics and Phonology

6.4.3.1 Yiddish Slavic languages are classic Glottal Tension languages and Yiddish is as well, possibly due to intense, long-term contact with Slavic, perhaps involving language shift where Slavic-speaking Jews acquired Yiddish and imposed their native phonology and phonetics on an originally Germanlike [spread] system. Such a history leads us to expect complexity. Kleine (2003: 262) describes the standard language as one where “Voiceless plosives are unaspirated or might occur with slight aspiration, mainly in final position, whereas voiced plosives are fully voiced in all positions.” Assimilation patterns are more complex than in many other [slack] languages, perhaps driven by historical language contact. (13)

Yiddish regressive laryngeal assimilation (Jacobs 2005: 120–121, 129–130) a. Northeast Yiddish Central and Southeast Yiddish dos bux ‘the book’ do[z] bux du[s] biːəx fusbenkl ‘bench’ fu[z]bejnkl fiː[s]benkl b.

Alternations in Standard Yiddish sˇi[d]əx ‘match’ sˇa[t]xn ‘matchmaker’ [z]okn ‘old man’ [s]keijnəm ‘old men’

c.

/v/ s[v]ivə ‘environment’ sˇ[v]uə ‘oath’ d[v]ojrə ‘Deborah’ s[f]orəm ‘holy books’ t[f]ilə ‘prayer’

In (13a) we see an expected regressive voicing assimilation, with differences across dialects, while (13b) shows regressive devoicing (which would be formally captured by delinking of a feature). Jacobs indicates “apparent exceptions” with regard to the facts in (13c). In various Germanic and other languages (including Slavic, key contact languages), /v/ behaves as an approximant rather than as a fricative, which may play a role here. More generally, Iverson and Salmons (2003a) argue that wrinkles in Dutch laryngeal assimilations are traceable to the language contact which led to the language becoming a [slack] language, with traces of its heritage still visible. Perhaps such a case could be made for Yiddish. For final devoicing, Weinreich (2008: 435–436) argues that “Through the German determinant, facts of Verha¨rtung came into the German component of Yiddish.” King (1980) shows that Northeast Yiddish lost the rule, restoring the earlier state of affairs, so that veg ‘way’ surfaces with [g] in the singular and plural: veg ~ vegən. Jacobs (2005: 78) reviews work arguing for a cline from Western Yiddish with regular final devoicing through (apparently ill-understood) Central Yiddish patterns of variation.

133

134

JOSEPH SALMONS

Following earlier work, Jacobs (2005: 130) summarizes the major dialect difference as between final neutralization in Central Yiddish versus Eastern Slavic-like anticipatory or regressive voicing assimilation in Northeast Yiddish. Thus tog-student ‘day student’ devoices in Central via final devoicing and in Northeast via voicing assimilation. But Southeast Yiddish lacks either process and /g/ surfaces as [g].

6.4.3.2 West Frisian This language patterns with [slack] varieties. For instance, Cohen et al. (1959: 116) see voicing as essentially the same as in Dutch. [slack] is the active phonological gesture, spreading regressively, including across word boundaries (Riemersma 1979: 62–70, Tiersma 1985: 27–28) into stop (14a) and fricative (14b) word-final forms as well as spreading into laryngeally unspecified /s/ in a reduced form of the pronoun for ‘she’ between sonorants (14c). (14)

West Frisian regressive laryngeal assimilations a. op + dwaan ‘put on’ [obdwaːn] net dwaan ‘don’t do’ [ned dwaːn] b. of buˆter ‘or butter’ [ov butər] thu´s we´ze ‘be home’ [tyz vɛːzə] c. wie se ‘was she’ [viə zə] foel se ‘fell she’ [fuəl zə] d. leˆz-t ‘read + 3S G ’ [lɛːst] hoeg-t ‘have to + 3S G ’ [huəxt] e. op dy ‘on that’ [op ti] wolst do ‘do you want’ [vost to]

Obstruent voicing is not licensed in final position so that, as in (d), clusters are resolved toward the voiceless. A more substantive wrinkle is illustrated in (e), where in a small set of function words beginning in d following a word ending in a voiced stop, “both often become voiceless” (1985: 27–28). This optional process looks like a reduction (removal of laryngeal specification) of the function word, i.e. it does not appear to be assimilation in the relevant sense. Instead, like the change of m to n introduced in Section 6.3, that is presumably a case of reduction or feature loss. As just noted, the contemporary language shows final devoicing. Tiersma (1985: 30) gives evidence on the history: [R]ecords indicate that final devoicing in Frisian is a phenomenon of recent origin. In the phonetic study of the language of Grou by Eijkman (1907: 19), b and d are said to be largely voiced at the end of a word. But in Sipma’s grammar (1913), there are signs that devoicing has started to set in. His transcriptions, although not entirely consistent, suggest that devoicing had taken place following long vowels, falling diphthongs, and liquids, but not after short vowels or rising diphthongs.

Germanic Laryngeal Phonetics and Phonology

Finally, we come to Germanic varieties without any contrast, in one instance by historical loss of laryngeal contrast and in the other due to transphonologization, a sound change followed by reinterpretation of an existing length contrast.

6.4.3.3 Central German Dialects In some German dialects, a widespread lenition process has eliminated laryngeal contrasts entirely or left contrasts only in very limited contexts. For instance, Odenwald German (Holsinger 2008) maintains a distinction in simple onsets before vowels where fortis can occur ([khabə] Kappe ‘hood’) but lenite elsewhere, including onset clusters and final position ([blug] ‘plow’). In others, like some varieties of East Franconian, there is no distinction at all, with all obstruents phonologically unmarked (‘lenis’ in the German tradition) and passive voicing occurring intervocalically (Rowley 1989: 400). 6.4.3.4 Swiss German In many varieties of Swiss German, there is no laryngeal distinction in consonants (see Schifferle 2010 on increasing aspiration in Zu¨rich). Kraehenmann (2001: 111) writes, “As has been found for other Swiss German dialects . . ., laryngeal features such as voicing or aspiration do not take part in the underlying opposition or in the phonetic manifestation of the stops.” Instead, she argues, the distinction is carried entirely by duration (Kraehenmann 2003, Allen 2016). (I interpret length not as a segmental feature at all, but a structural pattern, represented here by two consonants following Purnell et al. 2019.) Contrasts are illustrated in (15). (15)

Alemannic singleton versus geminate contrasts (Kraehenmann 2001: 122) a. initial b. medial c. final

/pp/omfrit ‘French fries’ /tt/urte ‘layer cake’ Su/pp/e ‘soup’ Tol/kk/e ‘smudge’ schla/pp/ ‘limp’ al/tt/ ‘old’

/p/ohne ‘bean’ /t/otter ‘egg yolk’ Stu/p/e ‘living room’ fol/k/a¨ ‘to obey’ Raa/p/ ‘raven’ Wal/t/ ‘forest’

The historical development of these patterns (especially Kraehenmann 2003) is noteworthy. Briefly, the High German or Second Consonant Shift eliminated old fortis stops (creating affricates or geminate fricatives) and old lenis contrasted singleton versus geminate, a distinction that was extended to initial position with the integration of Romance loanwords. In short, this is not a transphonologization of old laryngeal distinctions, but a maintained (and expanded) contrast.

135

136

JOSEPH SALMONS

To conclude this section, we see overall rich diversity over time and space but patterns that appear to allow straightforward analyses under the tenets of both Laryngeal Realism and privativity. More importantly, this survey suggests how much work remains to be done with the tools available.

6.5 Further Issues Some issues are unresolved or require further work, for instance: 1. Phonological activity contrasts with reduction, or loss of structure, asymmetries that warrant further investigation. Similarly, underspecified segments pattern distinctly, not only as targets of activity, but likely showing greater phonetic variability and probably different patterns of reduction. 2. In phonetics, articulatory evidence is needed. It would be particularly helpful to establish whether German final obstruents show an actual spread glottis gesture. Magnetic Resonance Imaging work underway at the University of Wisconsin–Madison should provide evidence. 3. Whether final neutralization is complete or incomplete has long engendered discussion (Fourakis and Iverson 1984, Piroth and Janker 2004, Warner et al. 2006, especially Kharlamov 2014 on Russian). 4. Passive voicing – vocal fold vibration in voicing-friendly environments without a phonological specification – is variable in Germanic. English and German have it, while Icelandic generally does not. What can we learn from these differences about phonetic enhancement? 5. Phonologically, are there overmarked systems? Mixed systems do seem to exist, as argued by Iverson and Salmons (2003a) for Dutch. Cyran (2011) on Polish, and Ramsammy and Strycharczuk (2016) on Portuguese, argue expressly for more complexity in marking, beyond the discussion of Swedish in Section 6.4. 6. Why do we have the patterns of directionality of assimilation we find? One might suspect that the duration of a glottal spreading gesture lasts longer than most single segments and that the variable timing here yields assimilations parallel to those of post- versus pre-aspiration. Those phonetic seeds could drive phonologization of assimilation. But regressive voice assimilation requires a different account, perhaps a tendency toward anticipatory voicing. 7. Issues of fricative versus approximant were alluded to above for /v/ in Yiddish and /v, j/ in Icelandic (with parallels in Slavic and elsewhere). Those examples show nonparticipation in laryngeal assimilations reflecting likely approximant status. In other instances, we see what are usually approximants behaving like fricatives, in particular the devoicing of a phonetically fricative /r/ in some Yiddish dialects

Germanic Laryngeal Phonetics and Phonology

(Jacobs 2005: 121). This has close parallels in some German varieties, e.g., those with velar [ɣ] for /r/, found in some west central areas, or uvular fricatives (see Hall 1993). 8. Social and historical context matters. For instance, language contact is often suspected in changes in laryngeal phonetics and phonology, as noted above for the Dutch and Yiddish switches to [slack], the creation of some symmetry in Swiss German geminate distributions, but also in final devoicing in West Frisian, which Hoekstra (2001: 731) suspects “might be a case of Dutch interference.” Simon 2011 shows how bilinguals may promote such changes. Second, there is the possibility that variation in unspecified segments allows them to be used more readily to mark social or regional speech patterns, an idea being explored in follow-up work to Purnell et al. 2005a, 2005b. In short, there is plenty to be done.

6.6 Conclusions What have often been presented as simple issues of ‘voicing’, ‘voicing assimilation’, and ‘final devoicing’ have moved to the center of much discussion in phonetics, phonology, sound change, and other areas. I have argued that despite considerable diversity in the surface patterns, we actually have solid understanding of most fundamental issues, often in ways quite different from what was communis opinio only a decade or two ago. The table below summarizes some findings of this chapter for selected Germanic languages and dialects. The first columns indicate whether there is robust aspiration of /p, t, k/ and allied fricatives, and whether there is full voicing of /b, d, g/ and allied fricatives. The third indicates whether there is assimilation to [spread] (i.e., voicelessness). The fourth reports whether there is regressive voice assimilation and the fifth shows whether there is final neutralization. To sum up: • There is a clear connection between phonetics and phonology: Glottal Tension languages show full voicing and Glottal Width languages show aspiration. Phonetic overmarking by enhancement is not uncommon but broadly variable as we might expect for a facultative pattern. • Phonetics suggests where to look but phonological activity is the key, and we see clear differences between activity and reduction. • We see a robust correlation between specification and assimilation patterns: Active features spread and unspecified values do not. Still, deletion, e.g., by reduction, can remove features. • Directionality matters: Regressive assimilation appears broadly in Glottal Tension varieties, if not universally, while [spread] systems often show some bidirectionality. (This could be connected to the time

137

138

JOSEPH SALMONS

Table 6.1 Summary of laryngeal phonetics, phonology, assimilation, and neutralization in Germanic Glottal Width languages/dialects Phonetics

Phonology

asp.

assimilation voicing to [spread]

English



dial.

Std. German



Swedish



Norwegian Icelandic

✔ ✔

Gothic

?



Regr. voice Final assim. neutral. Notes



dial.





✔ ✔ ✔

Regr. some progr. assim. Regr. some progr. assim. Regr. some progr. assim. Progr. and regr. assim.

✔ fric.

Glottal Tension languages/dialects Dutch







Yiddish W. Frisian

✔ ✔

✔ ✔

dial. ✔

‘Educated Scots’





Stop/fricative asymmetries Dialectal variation Possible progr. [slack] assim.; del. of [slack] in reduction

Other E. Franconian Alemannic

No distinction Duration, not laryngeal

needed for a spread glottis gesture, which is longer than the duration of most single segments.) • There appears to be no particular relationship between laryngeal contrast system and whether final neutralization takes place, just how. There is no phonological unity across Germanic with regard to laryngeal patterns but two distinct and coherent sets, Glottal Width with [spread] and Glottal Tension with [slack]. Final neutralization occurs in both systems in a variety of different ways.

References Abercrombie, D. 1967. Elements of General Phonetics. Edinburgh University Press. Allen, B.C. 2016. Laryngeal Phonetics and Phonology in Germanic. Ph.D. dissertation, University of Wisconsin–Madison. ´Arnason, K. 2011. The Phonology of Icelandic and Faroese. Oxford University Press.

Germanic Laryngeal Phonetics and Phonology

Avery, P. and W. J. Idsardi 2001. “Laryngeal dimensions, completion and enhancement.” In T. A. Hall (ed.), Distinctive Feature Theory. Berlin: Mouton de Gruyter: 41–70. Beckman, J., P. Helgason, B. McMurray, and C. Ringen 2011. “Rate effects on Swedish VOT: Evidence for phonological overspecification,” Journal of Phonetics 39:39–49. Blevins, J. 2006. “A theoretical synopsis of evolutionary phonology,” Theoretical Linguistics 32: 117–165. Brown, J. 2016. “Laryngeal assimilation, markedness and typology,” Phonology 33: 393–423. Cohen, A., C. L. Ebeling, P. Eringa, K. Fokkema, and A. G. F. van Holk 1959. Fonologie van het Nederlands en het Fries: Inleiding tot de modern klankleer. The Hague: Nijhoff. Collinge, N. E. 1985. The Laws of Indo-European. Amsterdam: John Benjamins. Cyran, E. 2011. “Laryngeal realism and laryngeal relativism: Two voicing systems in Polish?,” Studies in Polish Linguistics 6: 45–80. Docherty, G. J. 2011. The Timing of Voicing in British English Obstruents. Berlin: Mouton de Gruyter. Dresher, B. E. 2009. The Contrastive Hierarchy in Phonology. Cambridge University Press. Dresher, B. E. and Zhang, Xi 2005. “Contrast and phonological activity in Manchu vowel systems,” The Canadian Journal of Linguistics / La Revue Canadienne de Linguistique 50.1: 45–82. Fourakis, M. and G. K. Iverson 1984. “On the ‘incomplete neutralization’ of German final obstruents,” Phonetica 41: 140–149. Hall, D. C. 2011. “Phonological contrast and its phonetic enhancement: Dispersedness without dispersion,” Phonology 28: 1–54. Hall, T. A. 1993. “The phonology of German R,” Phonology 10: 83–105. ´ . 2003. “Laryngeal licensing and laryngeal neutralization in Hansson, G. O Faroese and Icelandic,” Nordic Journal of Linguistics 26: 45–79. Harris, J. 2009. “Why final obstruent devoicing is weakening.” In K. Nasukawa and P. Backley (eds.), Strength Relations in Phonology. Berlin: Mouton de Gruyter: 9–46. Helgason, P. and C. Ringen 2008. “Voicing and aspiration in Swedish stops,” Journal of Phonetics 36: 607–628. Henton, C., P. Ladefoged, and I. Maddieson 1992. “Stops in the world’s languages,” Phonetica 49: 65–101. Hestvik, A. and K. Durvasula 2016. “Neurobiological evidence for voicing underspecification in English,” Brain and Language 152: 28–43. Hoekstra, J. F. 2001. “An outline history of West Frisian.” In H. H. Munske and N. Århammer (eds.), Handbuch des Friesischen. Tu¨bingen: Max Niemeyer Verlag: 722–734. Holsinger, D. J. 2008. “Germanic prosody and consonantal strength.” In J. Branda˜o de Carvalho, T. Scheer, and P. Se´ge´ral (eds.), Lenition and Fortition. Berlin: Mouton de Gruyter: 273–300.

139

140

JOSEPH SALMONS

Honeybone, P. 2002. Germanic Obstruent Lenition: Some Mutual Implications of Theoretical and Historical Phonology. Ph.D. dissertation, University of Newcastle upon Tyne. Honeybone, P. 2005. “Diachronic evidence in segmental phonology: The case of obstruent laryngeal specifications.” In M. van Oostendorp and J. van de Weijer (eds.), The Internal organization of Phonological Segments, Berlin: Mouton de Gruyter: 319–354. Iverson, G. K. and J. Salmons 1995. “Aspiration and laryngeal representation in Germanic,” Phonology 12: 369–396. Iverson, G. K. and J. Salmons 1999. “Glottal spreading bias in Germanic,” Linguistische Berichte 178: 135–151. Iverson, G. K. and J. Salmons 2003a. “Legacy specification in the laryngeal phonology of Dutch,” Journal of Germanic Linguistics 15: 1–26. Iverson, G. K. and J. Salmons 2003b. “Laryngeal enhancement in Early Germanic,” Phonology 20: 43–72. Iverson, G. K. and J. Salmons 2007. “Domains and directionality in the evolution of German final fortition,” Phonology 24: 1–25. Iverson, G. K. and J. Salmons 2008. “Germanic aspiration: Phonetic enhancement and language contact,” Sprachwissenschaft 33: 257–278. Iverson, G. K. and J. Salmons 2011. “Final devoicing and final laryngeal neutralization.” In M. van Oostendorp, C. Ewen, B. Hume, and K. Rice (eds.), Companion to Phonology, Vol. III. Oxford: Wiley-Blackwell. Blackwell Reference Online. May 31, 2012: 1622–1643. Jacewicz, E., R. A. Fox, and S. Lyle 2009. “Variation in stop consonant voicing in two regional varieties of American English,” Journal of the International Phonetic Association 39: 313–334. Jacobs, N. G. 2005. Yiddish: A Linguistic Introduction. Cambridge University Press. Jakobson, R. 1949. “On the identification of phonemic entities.” In Selected Writings I: Phonological Studies.The Hague: Nijhoff: 418–425. Jessen, M. 1998. Phonetics and Phonology of Tense and Lax Obstruents in German. Amsterdam: John Benjamins. Kent, R. D. 1997. Speech Sciences. San Diego: Singular. Keyser, S. J. and K. N. Stevens 2006. “Enhancement and overlap in the speech chain,” Language 82: 33–63. Kharlamov, V. 2014. “Incomplete neutralization of the voicing contrast in word-final obstruents in Russian: Phonological, lexical, and methodological influences,” Journal of Phonetics 43: 47–56. King, R. D. 1980. “The history of final devoicing in Yiddish.” In M. I. Herzog, B. Kirshenblatt-Gimblett, D. Miron, and R. Wisse (eds.), The Field of Yiddish: Studies in Language, Folklore, and Literature, Fourth Collection. Philadelphia: Institute for the Study of Human Issues: 371–430. Kingston, J. and R. L. Diehl 1994. “Phonetic knowledge,” Language 70: 419–454. Kiparsky, P. 2006. “The Amphichronic Program vs. Evolutionary Phonology,” Theoretical Linguistics 32: 217–236.

Germanic Laryngeal Phonetics and Phonology

Kleine, A. 2003. “Standard Yiddish,” Journal of the International Phonetic Association 33: 261–265. Kraehenmann, A. 2001. “Swiss German stops: Geminates all over the word,” Phonology 18: 109–145. Kraehenmann, A. 2003. Quantity and Prosodic Asymmetries in Alemannic: Synchronic and Diachronic Perspectives. Berlin: Mouton de Gruyter. Kristoffersen, G. 2000. The Phonology of Norwegian. Oxford University Press. Kuzla, C., T. Cho, and M. Ernestus 2007. “Prosodic strengthening of German fricatives in duration and assimilatory devoicing,” Journal of Phonetics 35: 301–320. Lehmann, W. P. 1986. A Gothic Etymological Dictionary. Leiden: Brill. Lisker, L. 1986. “‘Voicing’ in English: A catalogue of acoustic features signaling /b/ versus /p/ in trochees,” Language and Speech 29: 3–11. Lisker, L. and A. Abramson 1964. “A cross-language study of voicing in initial stops: Acoustical measurements,” Word 20: 527–565. Lombardi, L. 1995. “Laryngeal neutralization and syllable wellformedness,” Natural Language and Linguistic Theory 13: 39–74. Maddieson, I. 1984. Patterns of Sounds. Cambridge University Press. Mihm, A. 2004. “Zur Geschichte der Auslautverha¨rtung und ihrer Erforschung,” Sprachwissenschaft 29: 133–206. Mihm, A. 2007. “Theorien der Auslautverha¨rtung im Spannungsverha¨ltnis zwischen Normsetzung und Sprachwirklichkeit,” Deutsche Sprache 35: 95–118. Miller, D. G. 2019. The Oxford Gothic Grammar. Oxford University Press. Oxford, W. 2015. “Patterns of contrast in phonological change: Evidence from Algonquian vowel systems,” Language 91: 308–358. Park, H., H-J. Park, and G. K. Iverson 2010. “The frontal and temporal lobe in the identification of laryngeal contrasts,” NeuroReport 21: 474–478. Pinget, A-F. C. H. 2015. The Actuation of Sound Change. Ph.D. dissertation, Universiteit Utrecht. Piroth, H. G. and P. M. Janker 2004. “Speaker-dependent differences in voicing and devoicing of German obstruents,” Journal of Phonetics 32: 81–109. Purnell, T., E. Raimy, and J. Salmons 2019. “Old English vowels: Diachrony, privativity and phonological representations,” Language, Research reports. 95(4). e447–e473. Purnell, T. C., J. Salmons, and D. Tepeli 2005a. “German substrate effects in Wisconsin English: Evidence for final fortition,” American Speech 80: 135–164. Purnell, T. C., J. Salmons, D. Tepeli, Dilara, and J. Mercer 2005b. “Structured heterogeneity and change in laryngeal phonetics: Upper Midwestern final obstruents,” Journal of English Linguistics 33: 307–338. Ramsammy, M. and P. Strycharczuk 2016. “From phonetic enhancement to phonological underspecification: Hybrid voicing contrast in European Portuguese,” Papers in Historical Phonology 1: 285–315. www .journals.ed.ac.uk/pihph/article/view/1704. Riad, T. 2014. The Phonology of Swedish. Oxford University Press.

141

142

JOSEPH SALMONS

Riemersma, T. 1979. Sylabysjerring, nazzeljerring, assymljerring. Ljouwert: Koperative Utjowerij. Roberge, P. T. 1983. “Those Gothic spirants again,” Indogermanische Forschungen 88: 109–155. Rooy, B. van and D. Wissing 2001. “Distinctive [voice] implies regressive voicing assimilation.” In T. A. Hall (ed.), Distinctive Feature Theory. Berlin: Mouton de Gruyter: 295–334. Rowley, A. R. 1989. “East Franconian.” In C. V. J. Russ (ed.), The Dialects of Modern German: A Linguistic Survey. Stanford University Press: 394–416. Schifferle, H. P. 2010. “Zunehmende Behauchung: Aspirierte Plosive im modernen Schweizerdeutsch.” In H. Christen, S. Germann, W. Haas, N. Montefiori, and H. Ruef (eds.), Alemannische Dialektologie: Wege in die Zukunft. Stuttgart: Steiner: 43–55. Sievers, E. 1893. Grundzu¨ge der Phonetik: Zur Einfu¨hrung in das Studium der Lautlehre der indogermanischen Sprachen. Leipzig: Breitkopf & Ha¨rtel. Simon, E. 2011. “Laryngeal stop systems in contact: Connecting present-day acquisition findings and historical contact hypotheses,” Diachronica 28: 225–254. Snoeren, N. D., P. A. Halle´, and J. Segui 2006. “A voice for the voiceless: Production and perception of assimilated stops in French,” Journal of Phonetics 34: 241–268. Suzuki, S. 1992. “Toward an explanatory account of Thurneysen’s Law in Gothic,” Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 114: 28–46. Tiersma, P. M. 1985. Frisian Reference Grammar. Dordrecht: Foris. Vaux, B. 1998. “The laryngeal specifications of fricatives,” Linguistic Inquiry 29: 497–511. Warner, N., E. Good, A. Jongman, and J. Sere´no 2006. “Orthographic versus morphological incomplete neutralization effects,” Journal of Phonetics 34: 285–293. Weijer, J. van de and E. J. van der Torre (eds.) 2007. Voicing in Dutch: (De)voicing – Phonology, Phonetics, and Psycholinguistics. Amsterdam: John Benjamins. Weinreich, Max 2008. History of the Yiddish Language. Edited by P. Glaser, translated by S. Noble. New Haven: Yale University Press/New York: YIVO. Wetzels, W. L. and J. Mascaro´ 2001. “The typology of voicing and devoicing,” Language 77: 207–244. Wiese, R. 1996. The Phonology of German. Oxford: Clarendon Press, 2nd edn. 2000. Youssef, I. 2010. “Laryngeal assimilation in Buchan Scots,” English Language and Linguistics 14: 321–345. Zonneveld, W. 2007. “Issues in Dutch devoicing: Positional faithfulness, positional markedness, and local conjunction.” In J. van de Weijer and E. J. van der Torre (eds.), Voicing in Dutch: (De)voicing – Phonology, Phonetics, and Psycholinguistics. Amsterdam: John Benjamins: 1–40. Zwicky, A. M. 1975. “Settling on an underlying form: The English inflectional endings.” In D. Cohen and J Wirth (eds.), Testing Linguistic Hypotheses. Washington: Hemisphere: 129–185.

Chapter 7 Tone Accent in North and West Germanic Björn Köhnlein

7.1 Introduction All Germanic languages use intonational tone to signal information status, sentential prominence, and prosodic boundaries (Fe´ry, Chapter 28; O’Brien, Chapter 8). In most varieties, this usage of tone is considered to be purely postlexical, in the sense that tone does not distinguish lexical items. The alignment of intonational pitch accents can be a correlate of word stress, as in the isolation forms for the English noun import versus the verb import; yet there are other correlates of stress (at least duration, vowel quality, intensity) that still make it possible to distinguish lexical items in the absence of pitch cues. Furthermore, such contrasts are always between syllables ( versus ) but never within them. In some varieties of North and West Germanic, however, (primarily) tone-based oppositions exist within stressed syllables. The phenomenon is typically referred to as tonal accent. In North Germanic (often also called Scandinavian), tonal accent occurs in most varieties of Norwegian and Swedish, as well as in some varieties of Danish. In West Germanic, tone accent contrasts can be found in dialects spoken in Belgium, Germany, and the Netherlands. The West Germanic tone accent area does not coincide with any of the traditional West Germanic dialect areas but covers Ripuarian, Moselle Franconian, and Limburgian dialects. In the English literature on the subject, the area as a whole is often referred to with the cover term Franconian.1 Intuitively, we can say that the tone accent oppositions in question are located “somewhere between tone and stress”. While tone refers to the distinctive function of pitch, the relationship with stress derives I would like to thank Pavel Iosad, Nina Hagen Kaldhol, and the editors Michael T. Putnam and B. Richard Page for their valuable suggestions. This work was supported by the National Science Foundation [BCS-1845107]. 1

This term is somewhat problematic, though, as there are Franconian dialects that are located outside of the tone accent area.

144

BJÖRN KÖHNLEIN

from the fact that the tonal opposition is restricted to syllables carrying primary word stress (see Alber, Chapter 4, for a discussion of word stress in Germanic). In virtually all Germanic tone accent dialects that have been described so far, the opposition is between two accents (and never more), typically called Accent 1 and Accent 2 (Kristoffersen 2010, however, discusses the possibility of a third accent in Sollero¨n Swedish). A well-known minimal pair from Swedish is [1andən] ‘the duck’ versus [2andən] ‘the spirit’ (accents are indicated with superscripts). The two items are segmentally identical but can be distinguished on the basis of their respective tonal contours. In Central Swedish, [1andən] has a rising-falling (LHL) contour in isolation, while [2andən] is pronounced with an additional high tone at the beginning of the stressed syllable (HLHL). Furthermore, the rising-falling part of the contour (LHL) occurs in Accent 2 later than in Accent 1. As we shall see in the respective sections on North and West Germanic, however, the precise tonal realization of Accent 1 and Accent 2 can differ across dialects, sometimes in nontrivial ways. This chapter will examine some general synchronic and diachronic properties of the Germanic tone accent systems (Section 7.2 on West Germanic, Section 7.3 on North Germanic). Note, however, that many central aspects concerning the synchronic and diachronic analysis of these systems are still under debate. Due to space limitations, I can only provide the very basics of some opposing synchronic and diachronic approaches. Furthermore, to avoid influence of personal bias, I will not provide detailed evaluations of different approaches. I restrict the discussion to varieties of Germanic where tonal accent is uncontroversially attested in modern systems. For North Low Saxon, it has sometimes been claimed that tonal oppositions exist alongside a ternary quantity contrast. So far, detailed acoustic studies have been unable to substantiate these claims, which is not meant to suggest that there might not be a tonal opposition in at least some dialects (see, e.g., Prehn 2012, Ho¨der 2014 for discussion). Furthermore, some scholars have argued that Frisian (Smith and van Leyden 2007, Versloot 2008) and Icelandic (Myrvoll and Skomedal 2010, Haukur Þorgeirsson 2013) had tonal accent at some point of their respective developments, but modern systems do not display accentual contrasts.

7.2 Tone Accent in West Germanic 7.2.1

The Present-Day Accent Contrast

7.2.1.1 Basic Patterns The tone accent opposition in West Germanic serves to distinguish lexical and grammatical units. Some minimal pairs from the Mayen dialect (Schmidt 1986) are provided in (1):

Tone Accent in North and West Germanic

(1)

Tone accent minimal pairs from Mayen a. [man1] ‘basket’ [man2] ‘man’ 1 b. [tɔuvə ] ‘pigeons’ [tɔuvə2] ‘baptisms’

(1a) shows a monosyllabic lexical minimal pair, while (1b) demonstrates that disyllabic minimal pairs exist as well. In most dialects, the opposition is restricted to syllable rhymes with two sonorant units (long vowels, diphthongs, or short vowels plus sonorants; see van Oostendorp, Chapter 2, for a more detailed discussion of syllable structure in Germanic).2 Within dialects, the precise tonal realization of the accents varies considerably depending on three factors: first, whether the item is in focus or occurs in a prefocal or postfocal context; second, whether the item occurs in phrase-final or nonfinal position; third, the expression of illocutionary force (declarative, interrogative, etc.). Figure 7.1 shows realizations of Accent 1 and Accent 2 in Cologne (e.g., Gussenhoven and Peters 2004), a fairly prototypical West Germanic tone accent dialect, in nonfinal focus position. Dialects with Cologne-like realizations of the accents are commonly referred to as Rule A, a dialect classification based on Wiesinger (1970); see Section 7.2.1.2 for discussion of Rule B, another dialect type. In Rule A, declarative contours are generally falling, and interrogative contours are generally rising (often followed by a low boundary tone). The difference between Accent 1 and Accent 2 can usually be expressed as a contrast in the timing of the pitch movements – Accent 1 typically moves earlier than Accent 2. In nonfinal position, Accent 1 is realized with an early fall in the stressed syllable of declaratives, and as an early rise in interrogatives; Accent 2 has a high level tone with a late, post-tonic fall in declaratives, and a low level tone with a late, post-tonic rise in interrogatives. While the contours in nonfinal positions (Figure 7.1) are similar across most Rule-A dialects, there is more variation in phrase-final realizations of Accent 2 (Accent 1 varies to a much smaller degree). For instance, in some dialects, such as present-day Cologne, Accent 2 will be realized with a high level tone and a relatively late fall in declaratives, similar to phrase-medial Context

Accent 1

Accent 2

Declarative, nonfinal position Interrogative, nonfinal position Figure 7.1 Tonal contours in Cologne, focus, nonfinal position; stressed accent syllable nonshaded, overall post-tonic contour shaded

2

For some varieties, contrasts have also been reported for sequences of short vowels plus obstruents (e.g., Jongen 1972 for Moresnet).

145

146

BJÖRN KÖHNLEIN

realizations. In other dialects, Accent 2 will have a two-peaked, falling-rising contour, such as in Mayen (Schmidt 1986) or Roermond (Gussenhoven 2000). Since phrase-final tone accent syllables typically interact with intonational boundary tones, it is unsurprising that we find more variation in this context. Along these lines, we can regard nonfinal occurrences (as depicted in Figure 7.1 for Rule A) as default realizations, since they are unaffected by boundary signals. (Traditional dialectological descriptions, however, are typically restricted to realizations in isolation, which mimic the realizations in phrase-final position.) Considerable variation also occurs in prefocal and postfocal position, where we either find neutralization of the accent contrast or can observe diverse strategies to maintain the opposition across dialects – I will not treat such nonfocal realizations any further here; discussion of realizations out of focus for certain dialects can be found in, e.g., Schmidt (1986) for Mayen, Gussenhoven and Peters (2004) for Cologne, or Ko¨hnlein (2011) for Arzbach. While pitch is usually the main correlate of the accent opposition (e.g., Werth 2011), the tonal opposition between the accents is usually accompanied by other contrasts. The most widespread additional correlate of accent is duration (e.g., Gussenhoven and Peters 2004 for Cologne, Schmidt 1986 for overview). In modern systems, tone accent syllables with contour tones (falling, rising) typically have shorter duration than accent syllables with level tone (high, low); this correlation is discussed in detail in Ko¨hnlein (2015a). Furthermore, Accent 1 can be accompanied by glottalization in some dialects; at least in Cologne, however, glottalization is restricted to realizations with falling tone (Gussenhoven and Peters 2004).

7.2.1.2 Synchronic Tonal Typology As indicated in Section 7.2.1.1, there is some variation in the realization of the accent contrast across dialects, such as differences in the phrase-final realization of Accent 2. There are, however, also more substantial crossdialectal differences. Here, I focus on a particularly striking case of variation – a reversal of tonal melodies found in some dialects in the South East of the tone accent area (Rule B). Rule B has first been described in Bach (1921). Bach claimed that in the Arzbach dialect (a small village in the German Westerwald), speakers systematically pronounce items that have falling tone in other dialects with a high level tone, and items that have a high level tone in other dialects with falling tone. In my own fieldwork (reported in Ko¨hnlein 2011), I have been able to confirm Bach’s claims, although they had to be relativized. As shown in Figure 7.2, the Rule-B dialect of Arzbach indeed has reversed declarative contours in comparison to Rule A. That is, as observed by Bach, Accent 1 falls earlier than Accent 2 in Rule A, while Accent 1 falls later than Accent 2 in Rule B. In interrogatives, however, which Bach did not mention, the contours are not reversed, but resemble each other in the two areas: in both

Tone Accent in North and West Germanic

Cologne (Rule A) Accent 1 Accent 2

Context

Arzbach (Rule B) Accent 1 Accent 2

Declarative, nonfinal Interrogative, nonfinal Figure 7.2 Tonal contours in Cologne (Rule A) and Arzbach (Rule B), focus, nonfinal position; nuclear contour nonshaded, overall post-nuclear contour shaded

Rule A and Rule B, the interrogative rise occurs earlier in Accent 1 than in Accent 2. In light of these novel facts, I have suggested in my work to refer to the difference between Rule A and Rule B as a semi-reversal of tonal contours. From a typological perspective, this semi-reversal is remarkable in various ways. On the one hand, tonal reversals between related language varieties are exceedingly rare across languages (Kingston 2011 for a case from Athabaskan). Furthermore, to the best of my knowledge, the Arzbach dialect is the only language variety described so far where we find an “alignment reversal” of word accents within the same system: Accent 1 falls later than Accent 2 in declaratives, but rises earlier than Accent 2 in interrogatives.

7.2.1.3

Synchronic Lexical Distribution

By and large, the synchronic distribution of tonal accent can be derived from a Middle High German reference system (or some other, comparable reference system), which will be discussed in Section 7.2.1.4; yet later changes have obscured the original distribution, and the accent of individual items is often synchronically unpredictable (see also Hall, Chapter 1, for a discussion of synchronic segmental processes in Germanic). For instance, the difference between [tɔuf1] ‘pigeon’ and [tɔuf2] ‘baptism’ must presumably be learned by present-day speakers. Yet there still are valid synchronic generalizations, two of which I mention below. A strong generalization holds with regard to morphological alternations. As first explicitly stated in van Oostendorp (2005), grammatical accent minimal pairs always switch from Accent 2 in the simplex form to Accent 1 in the more complex form across dialects; I am not aware of a single counterexample. Two minimal pairs from Arzbach are given in (2); (2a) shows an alternation between an Accent-2 singular and an Accent-1 plural, and (2b) provides an example for the adjective meaning ‘broad’: the predicative form has Accent 2, and the corresponding attributive feminine and neuter singular forms have Accent 1 (here: neuter). (Attributive forms can be regarded as more complex than predicative forms since they sometimes also come with additional segmental material, as in the Arzbach form [brai1də], which is used for masculine singulars and all plurals.)

147

148

BJÖRN KÖHNLEIN

(2)

a. [ʃtaːn2] ‘stone’ b. [ət is brait2] ‘it is broad’

[ʃtaːn1] ‘stones’ [ən brait1 haus2] ‘a.N O M broad.N O M house’

Furthermore, some modern dialects also show vocalic differences between the two accents. The most widespread pattern is that syllables with contour tones (typically Accent 1) tend to display more diphthongal realizations of vowels, whereas items with level tone (typically Accent 2) tend to have more monophthongal realizations (e.g., Dols 1953, Wiesinger 1970, Cajot 2006, Gussenhoven 2012, Ko¨hnlein 2017). Two examples from the diphthongization of long high vowels under Accent 1 in Maastricht are given in (3). Such developments appear to be still ongoing in various dialects. This can be inferred from the discussion in Cajot (2006), as well as from Gussenhoven (2012), who describes an ongoing vowel split in Maastricht that affects certain mid vowels and that had not been mentioned in previous literature on the dialect. (3)

a. [blɛif1] ‘stay.1.PS.SG’ [bliː2və] ‘stay-INF’ [duː2və] ‘pigeon-PL’ b. [dɔuf1] ‘pigeon.SG’

Notably, there are many aspects that still require more investigation across dialects. For instance, the accentuation patterns in compounds and loanwords are well studied for different varieties of North Germanic (see Section 7.3 above), but less so for West Germanic. It seems to be the case that members of compounds typically retain their accent in West Germanic, but deviant patterns may have escaped our attention so far (e.g., Jongen 1972 for Moresnet). We also know relatively little about the accent patterns in loanwords (yet see, e.g., Hermans 2012).

7.2.1.4 Theoretical Analysis The “traditional” theoretical approach to the analysis of West Germanic tonal accent, proposed in a series of papers by Gussenhoven and collaborators, operates under the assumption that there is a lexical tone contrast between the two accents. Accent 2 carries a lexical tone, and Accent 1 is the default. As a prototypical example, consider the tenets of the tonal analysis of the Cologne dialect, as provided in Gussenhoven and Peters (2004). The authors assume that Accent 1 is lexically toneless, which means that its surface structure will consist of intonational tones only. In phrase-medial position, this will be H*L for declaratives and L*H for interrogatives. The first tone links to the first mora of the accent syllable; the second tone goes to the second mora. As shown in Figure 7.3 (left side), this leads to a falling contour in declaratives and a rising contour in interrogatives. Accent 2, then, has a lexical tone, which is realized on the first mora of the accent syllable. Since the ideal association of tones to moras is one-toone, there is only space for one intonational tone on the second mora of the accent syllable. This is shown in Figure 7.3 (right side); the leftover

Tone Accent in North and West Germanic

Accent 1

Accent 2

H∗

L

HLex

H∗

µ

µ

µ

µ

L∗

H

LLex

L∗

µ

µ

µ

µ

Figure 7.3 Analysis with lexical tone for Rule A (Gussenhoven and Peters 2004), Cologne dialect; focus, nonfinal position

tone will be realized after the accent syllable. The lexical tone surfaces as HLex in declaratives and as LLex in interrogatives. According to Gussenhoven and Peters, the lexical tone changes its quality because it copies the value of the following intonational tone. As a consequence, we will get two identical tones in the accent syllables, which are realized with high level pitch in declaratives and with low level pitch in interrogatives, respectively. In recent years, an alternative approach has emerged, where it is assumed that the essential difference between the accents is metrical, rather than tonal (e.g., Hermans 2012, Ko¨hnlein 2016, Kehrein 2017). The main idea is that the tonal contours of Accent 1 and Accent 2 differ because metrical differences between the accents lead to diverse associations of intonational tones. Here, I briefly sketch the tenets of the analysis proposed in Ko¨hnlein (2011, 2016, 2017); the precise implementation and representational assumptions differ in other approaches, but the general idea is similar. Essentially, Ko¨hnlein claims that tonal accent emerges because the two accents have different types of trochaic feet: Accent 1 is a syllabic trochee, and Accent 2 is a moraic trochee (the general concepts of syllabic and moraic trochees go back to Hayes 1995; see Smith, Chapter 3, for a more detailed discussion of foot structure in Germanic). The head of Accent 1, the syllabic trochee, is the syllable node, which dominates both moras in the accent syllable. Ko¨hnlein assumes that being linked to the foot head makes these two moras metrically “strong.” In Figure 7.4, this is represented with two superscript pluses (µ+). The head of Accent 2, the moraic trochee, is not the syllable node but the first mora of the accent syllable; the second mora is the foot dependent. Therefore, only the first mora is metrically strong (µ+), but the second mora, which is not linked to a foot head, is weak (µ-).

149

150

BJÖRN KÖHNLEIN

Accent 1

Accent 2

H*

L

H*

µ+

µ+

µ+

L*

H

L*

µ+

µ+

µ+

µ–

µ–

Figure 7.4 Analysis with a foot-based approach (Köhnlein 2016) for Rule A, Cologne dialect; focus, nonfinal position

For Rule A dialects (such as Cologne), Ko¨hnlein assumes that only strong moras can receive their “own” tone. Therefore, in Accent 2, the metrically strong first mora can license a tone while the weak second mora cannot support its own tone. Instead, the tone from the first mora will spread to the second mora, which results in a high level tone in declaratives, and a low level tone in interrogatives (the respective trailing tones are realized after the accent syllable, similar to the tonal approach). In Accent 1, each of the two moras in the tone accent syllable is strong, so each can license its own tone. Therefore, both tones (H*L in declaratives, L*H in interrogatives) can be realized in the accent syllable, leading to a falling and a rising tone, respectively (see Ko¨hnlein 2016: section 4 for a discussion of how the analysis accounts for predictable quantity differences between the accents in some dialects and categorical differences in vowel/consonant quality).

7.2.2 Diachronic Development As first established in No¨rrenberg (1884), Accent 1 arose in stressed syllables under two conditions: either the syllable contained an originally long nonhigh vowel / opening diphthong, or the item was originally disyllabic with a voiced intervocalic consonant (here, “original” refers to a Middle High German reference system). All other contexts (monosyllabic words without an originally long nonhigh vowel / opening diphthong, disyllabic words with a voiceless intervocalic consonant) received Accent 2. (There is some variation across dialect areas, which we can ignore for our purposes; here I focus on Rule A.) Bach (1921) discovered that the natural classes underlying No¨rrenberg’s description are based on differences in intrinsic duration: Intrinsically longer vowels in the original system received Accent 1, whereas intrinsically shorter vowels received Accent 2. Bach bases his claim on two observations. On the one hand, vowels tend to be longer before lenis consonants

Tone Accent in North and West Germanic

than before fortis consonants.3 This accounts for the differences in accentuation for long high vowels, closing diphthongs, short vowels plus sonorants, and lengthened vowels depending on the voicing quality of the following consonant (for a discussion of obstruent voicing in Germanic, see Salmons, Chapter 6; vowel quantity in Germanic is discussed in Page, Chapter 5). Furthermore, nonhigh vowels tend to be phonetically longer than high vowels, which captures the fact that all nonhigh long vowels have Accent 1. Bach’s generalization has been accepted by most scholars, and has formed the basis for what I refer to as the duration-based approach to the tone accent genesis (e.g., Schmidt 2002; Ko¨hnlein 2011, 2015a,2015b; Boersma 2017; see also de Vaan 1999). The most well-known alternative has been proposed by Gussenhoven (e.g., 2000, 2004, in press), whose scenario invokes multiple factors to account for different aspects of the distribution (homonym avoidance in noun paradigms, intrinsic F0 differences, among others); for critical evaluations of this approach see, e.g., Schmidt (2002), Boersma (2006), Ko¨hnlein (2015a, 2015b); see Gussenhoven (2018) for a response to some of the criticisms. Consider Figure 7.5 (adapted from Ko¨hnlein 2015a: 243), which demonstrates how intrinsic durational differences may have facilitated the emergence of the tone accent contrast. The precise details of this scenario are based on my speculations, but they are perfectly compatible with other duration-based approaches. The stage before the genesis of the contrast is displayed with dashed rising contours in the stressed syllable of Accent 1 (left) and Accent 2 (right). The overall rising-falling intonational contour is chosen since this is the general interrogative intonation across all known modern dialects; in comparison, the declarative contours differ substantially across dialects (as, e.g., between Rule A and Rule B). This might be seen as an indication that the interrogative contours reflect a common old stage (Ko¨hnlein 2013). As a matter of fact, there still are certain dialects that only display rising-falling contours in declaratives and interrogatives, resembling those shown in Figure 7.5 (e.g., Hasselt, Peters 2008). Before the accent genesis (here referred to as “Pre-Accent” contours), the rise in the stressed syllable presumably started approximately at the same point for Pre-Accent 1 and Pre-Accent 2 and continued outside of the accent syllable. Since there was more time to realize a rising movement in relatively longer Pre-Accent-1 syllables (wider initial box in Figure 7.5) than in Pre-Accent-2 syllables (narrower box in Figure 7.5), the rise reached a higher end point in Pre-Accent-1 syllables than in PreAccent-2 syllables. These intrinsic pitch differences were at some point reinterpreted as a categorical tonal contrast, which extended the 3

This is comparable to the different vowel durations of English bat and bad; yet the correlation in West Germanic tone accent systems only applies to originally intervocalic consonants because the systems had final devoicing.

151

152

BJÖRN KÖHNLEIN

Accent 1

Accent σ

Postnuclear

Accent 2

Accent σ

Postnuclear

Figure 7.5 Reconstructed accent contours before and after tone accent genesis in West Germanic (before = dashed contour; after = solid contour)

differences between the contours (phonologization, Hyman 1976); this led to a rising tone for Accent 1 versus a low level tone for Accent 2. In Figure 7.5, this newly emerged tone accent contrast is represented with solid contours. The contrast became lexically distinctive after further developments. For instance, the Arzbach contrast between [ʃ tain1] ‘stones’ and [ʃ tain2] ‘stone’ derives from the fact that the plural originally ended in schwa. If we assume that dialects with contours such as in Figure 7.5 (as attested across the board in Hasselt, which is located at the North-Western border of the tone accent area) are the predecessor of modern Rule-A and Rule-B dialects, the question remains how the tonal reversal in declaratives may have come into existence. To the best of my knowledge, the only account that explicitly discusses the full set of relevant data is Ko¨hnlein (2013, 2015b).4 In a nutshell, Ko¨hnlein argues that the tonal reversal between Rule A and Rule B emerged when the two varieties changed the generally rising contours in stressed accent syllables (as displayed in Figure 7.5) to generally falling contours in declaratives; yet Rule A and Rule B used different adaptation strategies. The change from rising to falling intonation is in line with the observation that cross-linguistically (as well as in West Germanic), declaratives are more likely to have falling or high-level pitch accents, rather than low-level or rising pitch. Along these lines, the change created more prototypical intonation systems (see Ko¨hnlein 2013 for a detailed discussion). In Rule A, the alignment of the tonal contours was shifted leftwards, which moved the low pitch targets before the accent syllables and the high targets into the accent syllables for both Accent 1 and Accent 2. The process is displayed in Figure 7.6. Once more, the original contours are indicated with a dashed line, the modern Rule A contours with a solid line. The diachronic leftwards shift is indicated with an arrow. 4

Previous accounts (e.g., Bach 1921, Schmidt and Künzel 2006) have attempted to explain a full reversal, on the basis of the correct but incomplete facts provided in Bach (1921). Gussenhoven (2013) proposes an alternative account to Köhnlein’s reconstruction. As demonstrated in Köhnlein (2015b), however, the account only provides a scenario for phrase-final contexts while phrase-medial contexts can hardly be attributed to the mechanisms discussed in the paper. Lastly, Boersma (2017) describes a scenario that can only account for a full reversal but acknowledges that his approach cannot account for the full set of facts.

Tone Accent in North and West Germanic

Accent 1

µ PreAccent

Accent 2

µ

µ PostAccent

Accent σ

PreAccent

µ

Accent σ

PostAccent

Figure 7.6 The development from the predecessor toward Rule A (before = dashed contour; after = solid contour)

Accent 1

µ

Accent 2

µ Accent σ

µ

µ Accent σ

Figure 7.7 The development from the predecessor toward Rule B (before = dashed contour; after = solid contour)

Instead of shifting the tonal contours leftward, Rule B introduced a high tone by raising the pitch at the beginning of stressed syllables. The process is shown in Figure 7.7. Dashed lines indicate the predecessor contours, solid lines display the outcome of the change; again, the arrow indicates the direction of the change. The different adaptation strategies, namely, the leftward shift of tonal contours (Rule A) versus syllable-initial pitch raising (Rule B) led to opposite tonal contours in the respective accent syllables. That is, in Rule A, we now find a fall for Accent 1 versus a high level tone for Accent 2, while Rule B has a high level tone for Accent 1, and a low level tone for Accent 2.

7.3 Tone Accent in North Germanic 7.3.1

The Present-Day Accent Contrast

7.3.1.1 Basic Patterns North Germanic tone accent systems can be found in varieties of Norwegian, Swedish, and some Danish dialects. The opposition is usually restricted to nonfinal syllables of polysyllabic words – with the exception of a few dialects, final stressed syllables (including all monosyllabic words) always receive Accent 1. Two surface minimal pairs from Central Swedish are provided in (4): (4)

a. [1andən] ‘duck-the’ [2andən] ‘spirit-the’ b. [1taŋkən] ‘tank-the’ [2taŋkən] ‘thought-the’

153

BJÖRN KÖHNLEIN

A related phenomenon occurs in most varieties of Danish: A group of words, which largely correlate with Accent 1 from an etymological perspective, show glottalization in stressed syllables, the so-called stød (see Ejskjær 2005 for a typology and distributional variation across dialects). The occurrence of stød is restricted to the second mora of stressed syllables, but only if that mora is sonorous, i.e., a sonorant consonant or the second part of a long vowel. An example from Standard Danish is the difference between [tsæ:lə] ‘to count’, which does not have stød, and the corresponding imperative [tsæ:ʔl] ‘count!’, which has stød (indicated with a glottal-stop superscript). Over the years, there have been debates as to whether stød should be regarded as a tonal phenomenon. Recent experimental investigations by Grønnum et al. (2013), however, suggest that pitch is not a reliable correlate of stød, whereas glottalization is. Due to space constraints, I will not discuss synchronic aspects of stød any further (yet see, e.g., Basbøll 2005, Itoˆ and Mester 2015, Iosad 2016a for discussion and analysis). I return to stød in section Section 7.3.2, where I discuss the diachronic development of tonal accent in North Germanic. In North Germanic varieties with tonal accent, the main correlate of the opposition between Accent 1 and Accent 2 is pitch. Figure 7.8 displays prototypical realizations of the accents from Central Swedish. In isolation, Accent 1 is realized with a one-peaked rising-falling contour. If a nonaccent syllable precedes Accent 1 (indicated with grey shading), we find a pitch fall into the accented syllable. Accent 2 has a two-peaked contour from the stressed syllable onward and a later timing of the rise-fall. The first L in Accent 1 and the first H in Accent 2 are realized early in the respective accent syllables. The timing of the rest of the contour depends on various factors, such as the position of accent items in the intonational phrase. Furthermore, some tones are not always present, as first explicitly discussed in Bruce (1977). The prominence tone (HP) only occurs when accent items are realized in prominent phrasal positions, such as (but not limited to) focus (see also Myrberg and Riad 2015). The final low boundary tone (L%) surfaces only if the accent item occurs in the final position of an intonational phrase. The first L of Accent 1 and the initial HL sequence in Accent 2 are always realized. Consequently, there is no neutralization in nonprominent phrasal positions, unlike what we sometimes find in West Germanic (see Section 7.2.1.1). Note also that most North Germanic tone

[1and- n] ‘duck-the’ σ Accent 1 e

Minimal pair Accent type

[2and -n] ‘spirit-the’ Accent 2 e

154

σ

Tonal contour Syllable structure

σ (H)

σ LHp

σ L%

Figure 7.8 Idealized tonal contours in Central Swedish

σ

σ HL

σ HpL%

Tone Accent in North and West Germanic

accent systems do not use different pitch accents to signal pragmatic differences between, say, declaratives and interrogatives; yet phrase boundaries can be signaled with different boundary tones (e.g., Riad 2013: 257).

7.3.1.2 Synchronic Tonal Typology There has been extensive research on the synchronic typology of the North Germanic tone accents. Most of this research has been informed by Meyer (1937, 1954), who conducted a large number of phonetic studies on the realization of tonal accent across different varieties (with a focus on Swedish dialects); more recent collections of tone accent realizations in various Norwegian dialects can be found in the project Norsk tonelagstypologi “Norwegian tonal accent typology” (2000– 2002) and in Hognestad (2012). Building on Meyer’s work, Gaº rding (1977) created a widely used Swedish accent typology. Gaº rding’s typology distinguishes two main factors: First of all, dialects are classified on the basis of whether Accent 2 is realized with one peak (symbolized with ‘1’) or with two peaks (symbolized with ‘2’). Furthermore, Gaº rding distinguishes dialects according to the timing of the melodies. The two main types are dialects where the realization of the accent melodies starts relatively early in accent syllables (symbolized with ‘A’) versus those where they start later (symbolized with ‘B’). A few examples of prototypical realizations of four dialect types are provided in Figure 7.9. 7.3.1.3 Synchronic Distribution An important aspect of the synchronic distribution of North Germanic tonal accent is that there are few, if any, lexical minimal pairs. Surface

Accent type Syllables

Accent 1 σ σ

Accent 2 σ σ

Type 2B (e.g., Urban East Norwegian) Type 2A (e.g., Central Swedish) Type 1B (e.g., Dala Swedish) Type 1A (e.g., Malmö Swedish) Figure 7.9 Idealized tonal contours in different dialect areas, distinguished by number of peaks in Accent 2 (1 /2) and alignment of the tonal melodies (A = early; B = late)

155

156

BJÖRN KÖHNLEIN

minimal pairs typically also differ in their morphological structure. Consider the opposition between [1and-ən] “duck-the” versus [2andə-n] “spirit-the” as an example (hyphens indicate morpheme boundaries). These forms can be distinguished only on the basis of their accent; yet the base form for [1and-ən] is monosyllabic [1and] “duck”, whereas the base form for [2andə-n] is disyllabic [2andə] “spirit”. Since monosyllables as well as polysyllabic words with final stress always take Accent 1 (see Section 7.2.1.1), this has led to claims in the (mostly earlier) literature that Accent 2 is a “signal” of polysyllabicity. This indeed works well for some parts of the vocabulary; yet the correlation is far from perfect. For instance, as pointed out in Riad (2013), there are various initially stressed disyllabic (loan) words that have Accent 1 in Swedish, such as 1fa¨nrik ‘lieutenant’ and or 1brandy ‘id.’. The accent class of initially stressed disyllabic words interacts with the phonological make-up of the second, unstressed syllable. In Swedish, disyllabic words ending in unstressed /ə/ or /a/ (which correspond to the inflectional endings used in the language) typically have Accent 2 (e.g., 2 ande, 2manga), while other endings trigger Accent 1 (e.g., Bruce 1977). In Norwegian, only word-final unstressed /ə/ triggers Accent 2 in disyllabic words while other endings generally lead to Accent 1 (Kristoffersen 2000), an example being 2eple ‘apple’. Furthermore, monomorphemic trisyllabic words with initial stress typically receive Accent 1, examples from Swedish being 1gigolo ‘id.’ or 1polio ‘id.’ A particularly strong factor in determining the present-day lexical distribution of the accents is morphology. Arguably the most wellknown morphological property of North Germanic tone accent systems is the tendency to assign Accent 2 to compounds. The degree to which this happens, however, varies across dialects. In Eastern dialects (including Central Swedish), compounds always have Accent 2, independent of the accent of the individual members. This is often referred to as the “compound rule”, a term I shall adopt here.5 A few examples from Riad (2013) are given in (5). (5a) shows a compound where the first member has Accent 2. (5b) demonstrates that even inputs with only Accent-1 items result in Accent 2. (5c) shows that affixes which form an independent prosodic word (here: [heːt]) will also trigger the compound stress rule, because they have an independent stress. (5)

5

Compound accent in Stockholm Swedish (always Accent 2) a. [2sɔmːar] ‘summer’ + [1loːv] ‘holiday’ → [2sɔmːarloːv] ‘summer break’ b. [1jʉːl] ‘Christmas’ + [1loːv] ‘holiday’ → [2jʉːlloːv] ‘Christmas break’ c. [ɡrʏmː] ‘cruel’ + [heːt] ‘N O M I N A L I Z E R ’ → [2ɡrʏmːheːt] ‘cruelty’

Riad (2013: 127) points out that the term “compound rule” is somewhat misleading, since the rule is prosodically and not morphologically driven, applying to constituents with more than one stress.

Tone Accent in North and West Germanic

Varieties in the West and South of the tone accent area tend to have more complex compound rules, in the sense that compounds (or compoundlike words) can have Accent 1 under certain circumstances. Here, I will restrict myself to a discussion of the compound rule in Urban East Norwegian, as given in (6). The examples are taken from Lahiri et al. (2005); see, e.g., Bruce (1973) for a discussion of compound stress in Malmo¨ Swedish. In Urban East Norwegian, initially stressed compounds will always have Accent 2 if the first member has Accent 2, similar to Central Swedish (6a). Some first members that have Accent 1 in isolation, however, trigger Accent 1 in compounding, counter to the compound rule in Eastern dialects. This is true in particular for polysyllabic Accent 1 words, where the resulting compound will typically have Accent 1 (6b). The situation is more complicated for monosyllabic Accent -1 items, some of which trigger Accent 1 in compounds, while others do not. A striking example is the behavior of the homonyms 1ball ‘ball (dance event)’ versus 1ball ‘ball (object)’, both of which have predictable Accent 1 in isolation. When these forms surface as the first member of a compound, however, 1ball ‘dance event’ triggers Accent 1 (6c), whereas 1ball ‘object’ triggers Accent 2 (6d). (6)

Norwegian compounds (sometimes Accent 1, sometimes Accent 2) a. 2kirke ‘church’+ 1taº rn ‘tower’ → 2kirketaº rn ‘church tower’ b. 1aksje ‘stock’ + 2marked ‘market’ → 1aksjemarked ‘stock market’ c. 1ball ‘ball (dance event)’ + 1sal ‘room’ → 1ballsal ‘ballroom’ d. 1ball ‘ball (object)’ + 1spill ‘game’ → 2ballspill ‘ball game’

When it comes to affixation, North Germanic tone accent systems show certain prosodic effects that we also find in other Germanic languages, i.e., stress neutral, stress shifting, or stress attracting affixation. In addition, affixes in tone-accent systems can be accent neutral or accent inducing. Accent inducing suffixes can either induce Accent 1 or Accent 2. For instance, the verbal prefix be-1 always triggers Accent 1 in Swedish and Norwegian, even if a verb stem has Accent 2 when used in imperatives. be-1 also overrides the tonal specifications typically assigned by the Table 7.1 Some effects of affixes on the accent class of words IMPERATIVE

INFINITIVE

1

stäm be-1stäm

2

1

2

GLOSS

Swedish stämma be-1stämma

‘tune’ ‘decide’

Norwegian stem be-1stem

stemme be-1stemme

‘tune’ ‘decided’

157

BJÖRN KÖHNLEIN

infinitive markers 2-a (Swedish) and 2-e (Norwegian). This is shown in Table 7.1 for the verb stems sta¨m (Swedish) and stem (Norwegian); data are from Lahiri et al. (2005).

7.3.1.4 Theoretical Analysis Over the years, various analyses of the North Germanic tone accent opposition have been proposed. In the most widespread approach, at least one of the accents is specified with a lexical tone. Some proponents of the tonal analysis have claimed that Accent 2 is the marked member of the opposition and thus carries a lexical tone (Accent 1 is then toneless, e.g., Riad 2013). Others have argued that Accent 1 is marked in the lexicon, and Accent 2 is lexically toneless (e.g., Lahiri et al. 2005). I will briefly describe the tonal mapping under these two privative approaches; again, I take Central Swedish as an example. Riad (2013) argues that Accent 2 has a lexical high tone Hlex that precedes the intonational tones. As indicated in Section 7.2.1.1, these intonational tones consist of a word tone L, which is always realized, followed by a prominence tone HP, which is restricted to prominent positions in the phrase, and a low boundary tone L%, which only occurs at the end of an intonational phrase. In isolation, Accent 2 will have a HlexLHPL% tonal contour. Accent 1, which lacks the lexical tone, will accordingly have a LHPL% contour without the initial lexical H. In the analysis by Lahiri et al. (2005), Accent 1 in Central Swedish is lexically marked with a lexical tone LLex. In isolation, LLex combines with HP and L%; this results in an LHL contour. Accent 2 starts with an intonational H* in the input, whose assignment is apparently blocked by LLex in Accent 1 words. This H* then combines with the high prominence tone and the boundary L – since the resulting H*HL contour would violate the Obligatory Contour Principle, an epenthetic low tone Lepen is inserted, which results in a H*LepenHL-contour. The two analyses are schematized in Figure 7.10. A second type of analysis is the “timing approach”, where it is argued that the shape of the melodies is essentially identical, but their timing differs. In his groundbreaking work, Bruce (1977) has proposed that both accents are marked tonally in the lexicon, which does not necessarily provide a mechanism for default accentuation. Kristoffersen (2006)

[1and- n] ‘duck-the’ σ Accent 1

[2and -n] ‘spirit-the’ σ Accent 2

σ (H) (H)

σ

e

Minimal pair Accent type

e

158

Tonal contour Syllable structure Lahiri et al. (2005) Riad (2013)

σ LLexHp L*Hp

σ L% L%

σ H*Lepen HlexL*

Figure 7.10 Competing analyses with lexical tone for Central Swedish

σ HpL% HpL%

Tone Accent in North and West Germanic

proposes a related approach for urban East Norwegian. He regards Accent 1 as marked (in line with Lahiri et al. 2005), arguing that parts of a default intonational contour are prelinked in the lexicon for Accent 1 in a way that differs from the default tone assignment for Accent 2. An approach that tries to derive accent assignment (largely) by rule has been proposed in More´n-Duollja´ (2013). In a nutshell, More´n-Duollja´ argues that Accent 1 is a monosyllabic foot, and Accent 2 a disyllabic foot. More´n-Duollja´’s main claim is that Accent 2 typically occurs when there are two underlying vowels, while Accent 1 occurs when there is one underlying vowel. Along these lines, [1and-ən] (one underlying vowel, epenthetic schwa) will have Accent 1, and [2andə-n] will have Accent 2 (two underlying vowels). As pointed out by More´n-Duollja´ (2013: 251), this metrical approach is comparable to Ko¨hnlein’s analysis of West Germanic tonal accent (see Section 7.2.1.4). He notes, however, that his study concentrates on nominal paradigms and still needs to be extended to other parts of the grammar.

7.3.2 Diachronic Development Opinions on how North Germanic tonal accent arose still vary widely. Roughly, we can identify two major debates. A first debate concerns the relationship of tonal accent and stød. While there is no doubt that the two phenomena are historically related, the question which of the two phenomena is older is still disputed. Over the past decades, most scholars seem to have converged on the idea that stød developed out of tonal accent (e.g., Ringgaard 1960, Riad 2000, Bye 2004, Iosad 2016b); yet Wetterlin and Lahiri (2015) have recently proposed a novel scenario in which they assume stød to be the predecessor of tonal accent, inspired by Liberman’s (1982) ‘stød-first’ approach. A second debate, discussed mostly among proponents of the “accent-first” approach, is between what Iosad (2016b) refers to as the “double-peak” approach (e.g., Riad 1998, 2000) and the “peak-delay” approach (e.g., Bye 2004, Iosad 2016b). Given space restrictions, I will only be able to briefly discuss the accent-first, double-peak approach (Riad 1998), the accent-first, peak-delay approach (Bye 2004, Hognestad 2007, Iosad 2016b), and the ‘stød-first’ approach by Wetterlin and Lahiri (2015).6 Riad’s accent-first, double-peak scenario builds on a general approach by Kock (1901), and more specifically on a paper by D’Alquen and Brown

6

In addition to the scenarios mentioned here, a largely ignored explanation of the diachronic typology of (Swedish) tone accent dialects has been proposed in Meyer (1937: 231–240). Building on his own fieldwork, Meyer argues that one-peaked dialects of the type ‘1B’ display the oldest stage of the contrast. Type ‘1A’ dialects with generally falling contours are derived by moving the pitch contours leftward; two-peaked dialects are derived by introducing an additional high tone at the beginning of Accent 2 syllables. As discussed in detail in Köhnlein (2013: section 3.1), the mechanisms suggested by Meyer are very similar to Köhnlein’s proposed developments of Rule A and Rule B in West Germanic, which was sketched in Section 7.2.1.3.

159

160

BJÖRN KÖHNLEIN

Table 7.2 Basics of the North Germanic tone accent genesis in the “accentfirst, double-peak” approach (Riad 1998) Stage 1 doomijan H H wordoo H H gastiz H

Stage 2 Syncope

døøman H H Clash resolution wordu H Syncope gæstr H

Stage 3 Clash resolution Syncope Epenthesis

2

døøma H H 1 ord H 1 gæster H

(1992). D’Alquen and Brown argue that Proto-North-Germanic sequences of a stressed syllable plus two light syllables or two heavy syllables correlate with Accent 2. Originally monosyllabic words and words with stressed heavy syllables followed by one light syllable correlate with Accent 1. Along these lines, Riad (1998: 69) proposes that syncope (deletion of word-medial vowels, ca. 8th century AD) and the resolution of stress clashes (deletion of the second stress in adjacent stressed syllables) led to the modern-day difference between Accent 1 and Accent 2. The derivation of three representative words is shown in Table 7.2. As is typical of earlier Germanic, all items shown in Table 7.2 had main stress on the (stem-)initial syllable, which, according to Riad, was marked with a high tone. Riad furthermore assumes that *doomijan and *wordoo at some point had secondary stresses, which received an additional high tone. *doomijan had a secondary stress on the third syllable (Stage 1). The two high tones were separated by a low tone, which resulted in a fallingrising HLH tonal contour. After syncope – a process that led to a change from *doomijan (Stage 1) to døøman (Stage 2) – the word had two adjacent stressed syllables. In line with the cross-linguistic tendency that adjacent stressed syllables are dispreferred, the resulting stress clash was resolved by making the second syllable light (deletion of coda consonant, Stage 2 døøman to Stage 3 2døøma). Riad assumes that after resolving the stress clash, the HLH melody was retained, which led to the double-peaked Accent 2 (2døøma). *wordoo, which had an original secondary stress on the heavy second syllable, resolved the stress clash between Stage 1 and Stage 2. Since the word later received Accent 1 (1ord, Stage 3), we have to assume that the high tone of the secondary stress was not retained after the stress clash was resolved. Lastly, *gastiz did not have a secondary stress in Stage 1, and therefore only one high tone throughout its development. This means that resulting gæster received Accent 1 at Stage 3. One-peaked systems evolved later when the accent contours shifted leftward (comparable to the development of West Germanic Rule A, as discussed in Section 7.2.1.3). Stød, then, developed out of one-peaked systems when the falling pitch movement of Accent 1 was accompanied by a glottal closure, or possibly even replaced by it (Riad 2000).

Tone Accent in North and West Germanic

Pre-Accent 1

Pre-Accent 2

σ – HL akr – Accent 1

σ σ H L ly kill Accent 2

σ HL a

σ H ly

σ ker

σ L kill

Figure 7.11 Basics of the North Germanic tone accent genesis in the accent-first, peak-delay approach (before = dashed contour; after = solid contour)

The accent-first, one-peaked approach, defended in, e.g., Oftedal (1952), Bye (2004), Hognestad (2007), and Iosad (2016b), maintains that the development of tonal accent occurred in Old Scandinavian (ca. AD 1000–1200). The main assumption is that monosyllabic and disyllabic words originally had a functionally redundant difference in the timing of intonational contours. An originally falling tonal contour HL was realized within one syllable in monosyllabic words, but spread across two syllables in disyllabic words; this corresponded most closely to modern dialects of type 1A. As a consequence of this alignment difference, the pitch peak in monosyllabic words (Figure 7.11, left side, Pre-Accent 1) occurred later than in disyllabic words (Figure 7.11, right side, Pre-Accent 2). When some originally monosyllabic words became disyllabic through schwa epenthesis (akr > aker), they retained the original fall in the first syllable of the original monosyllabic form (Figure 7.11, left side, Accent 1). The tonal melody in ‘new’ disyllables thus differed from that of old disyllables (Figure 7.11, right side, Accent 2), which led to the tone-accent contrast. Under the one-peaked approach, two-peaked Accent 2 would be the result of two independent processes, as depicted in Figure 7.12. First, in some varieties, the peaks of the two accents were delayed, and the peak of Accent 2 ended up in the second syllable; this led to 1B dialects (Figure 7.12, 1A to 1B). At some point, some varieties introduced an additional high tone in the stressed syllable, which created double-peaked Accent 2 (Figure 7.12, 1A to 2A / B). Similar to Riad’s approach, the development of stød can be attributed to a reinterpretation of the falling tone. A third proposal I will briefly discuss regards the emergence of tonal accent as a development out of stød. The general idea, originally put forward in Liberman (1982), has recently been elaborated on in Wetterlin and Lahiri (2015). In a nutshell, Wetterlin and Lahiri assume that after

161

162

BJÖRN KÖHNLEIN

Table 7.3 Basics of the North Germanic tone accent genesis in the “stød-first” approach: development of stød versus no stød Stage 1

Stage 2

ʔ

ʔ

armr= inn himinn=hinn

arm inn himinn=inn

Stage 3 ʔ

arm en himmelen

Gloss arm=the sky=the

Development from 1A to 1B Accent 1 Accent 2

σ σ σ – Development from 1B to 2A/B Accent 1 Accent 2

σ a

σ ker

σ ly

σ kill

Figure 7.12 Development of two-peaked Accent 2 in the accent-first, peak-delay approach

encliticization of the definite articles hinn / hitt, the initial /h/ of these articles was glottalized following monosyllabic bimoraic nouns, but not after polysyllabic nouns. In Table 7.3 the overall developments are shown for the nouns armʔen ‘arm’ (monosyllabic stem, stød develops) and himmelen ‘sky’ (disyllabic stem, no stød develops). Wetterlin and Lahiri (2015: 61) assume that glottalization reinforced the syllable boundary between the stem and the clitic, as a defense mechanism against the “danger of resyllabification and incorporation into the foot for directly preceding stressed syllables.” Since feet are maximally disyllabic, no such “danger” was present after unstressed (second) syllables of disyllabic stems. At some point, the newly emerged glottal stop was reinterpreted as part of the stem (Stage 2, 3), and was then extended to all monosyllabic nouns. The opposition eventually became phonemic when some monosyllabic stød items became disyllabic as a consequence of vowel epenthesis (e.g., fingʔr ‘finger’ > fingʔer ‘finger’). Over time, stød was used in more morphological environments. Tonal accent came into existence when speakers of some varieties interpreted the glottal stop as either a high or a low tone. Thus, in this scenario, one-peaked and double-peaked tone accent dialects could have come into existence as (at least partially) independent developments. Wetterlin and Lahiri (2015: 64) point out, however, that they do not wish to make specific claims regarding the typology of tonal dialects.

Tone Accent in North and West Germanic

References Bach, A. 1921. “Die Scha¨rfung in der moselfra¨nkischen Mundart von Arzbach (Unterwesterwaldkreis),” Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 45: 266–290. Basbøll, H. 2005. The phonology of Danish. Oxford University Press. Boersma, P. 2017. “The History of the Franconian Tone Contrast.” In Kehrein et al. (eds.): 27–98. Botma, B. and R. Noske (eds.) 2012. Phonological Explorations: Empirical, Theoretical and Diachronic Issues. Berlin: Mouton de Gruyter. Bruce, G. 1973 “Tonal accent rules for compound stressed words in the Malmo¨ dialect,” Working papers 7: 1–35. Lund: Phonetics laboratory, Lund University. Bruce, G. 1977. Swedish Word Accents in Sentence Perspective. Lund: Gleerup. Bye, P. 2004. Evolutionary Typology and Scandinavian Pitch Accent. Dordrecht: Kluwer Academic Publishers. Cajot, J. 2006. “Phonologisch bedingter Polytonieverlust – eine tonlose Enklave su¨dlich von Maastricht.” In de Vaan (ed.): 11–24. d’Alquen, R. and K. Brown 1992. “The origin of Scandinavian accents I and II.” In I. Rauch, G. F. Carr, and R. L. Kyes (eds.), On Germanic Linguistics: Issues and Methods. Berlin: Mouton de Gruyter: 61–79. de Vaan, M. (ed.) 2006. Germanic Tone Accents. Proceedings of the First International Workshop on Franconian Tone Accents, Leiden, June 13–14, 2003. Zeitschrift fu¨r Dialektologie und Linguistik – Beiheft 131. Stuttgart: Franz Steiner Verlag. de Vaan, M. 1999. “Towards an explanation of the Franconian tone accents,” Amsterdamer Beitra¨ge zur a¨lteren Germanistik 51: 3–44. Dols, W. 1953. Sittardse diftongering. Sittard. Ejskjær, I. 2005. “Dialects and regional linguistic varieties in the 20th century, Vol. III: Denmark.” In O. Bandle, K. Braunmu¨ller, E. Jahr, A. Karker, H-P. Naumann, U. Telemann, L. Elmevik, and G. Widmark (eds.), The Nordic Languages. An International Handbook of the History of the North Germanic Languages, Vol. 2. Berlin and New York: Mouton de Gruyter: 1721–1741. Gaº rding, E. 1977. The Scandinavian Word Accents. Lund: Gleerup. Grønnum, N., M. Vazquez-Larruscain, and H. Basbøll, 2013. “Danish stød: Laryngealization or tone,” Phonetica 70.1–2: 66–92. Gussenhoven, C. 2000. “The lexical tone contrast of Roermond Dutch in Optimality Theory.” In M. Horne (ed.), Prosody: Theory and Experiment. Amsterdam: Kluwer: 129–167. Gussenhoven, C. 2004. The Phonology of Tone and Intonation. Cambridge University Press. Gussenhoven, C. 2012. “Quantity or durational enhancement of tone: the case of Maastricht Limburgian high vowels.” In B. Botma and R. Noske (eds.): 241–254.

163

164

BJÖRN KÖHNLEIN

Gussenhoven, C. 2013. “From Cologne to Arzbach: An account of the Franconian ‘tone reversal’.” In E-L. Asu and P. Lippus (eds.), Nordic Prosody. Proceedings of the XIth Conference, Tartu 2012. Frankfurt am Main: Peter Lang Verlag: 11–24. Gussenhoven, C. 2018. “In defense of a dialect-contact account of the Central Franconian tonogenesis.” In H. Kubozono and M. Giriko (eds.), Tonal Change and Neutralization. Berlin and New York: Mouton de Gruyter: 350–379. Gussenhoven, C. and J. Peters (2004). “A tonal analysis of Cologne Scha¨rfung,” Phonology 21.2: 251–285. Hart, J. ‘t 1998. “Intonation in Dutch.” In D. Hirst and A. Di Cristo (eds.): 96–111. Haukur, Þ. 2013. Hljo´ðkerfi og bragkerfi. Doctoral dissertation, University of Iceland. Hayes, B. 1995. Metrical Stress Theory: Principles and Case Studies. University of Chicago Press. Hermans, B. 2012. “The phonological representation of the Limburgian tonal accents.” In B. Botma and R. Noske (eds.): 227–244. Ho¨der, S. 2014. “Low German: A profile of a word language.” In J. C. Reina and R. Szczepaniak (eds.), Syllable and Word Languages. Berlin and New York: Mouton de Gruyter: 305–326. Hognestad, J. K. 2007. “Tonelag i Flekkefjord bymaº l,” Norsk lingvistisk tidsskrift 25.1: 57–88. Hognestad, J. K. 2012. Tonelagsvariasjon i norsk. Ph.D. thesis. Kristiansand, University of Agder. Hyman, L. M. 1976. “Phonologization.” In A. Juilland (ed.), Linguistic Studies Presented to Joseph H. Greenberg. Saratoga: Anma Libri: 407–418. Iosad, P. 2016a. “Prosodic structure and suprasegmental features: Short-vowel stød in Danish,” Journal of Comparative Germanic Linguistics 19.3: 221–268. Iosad, P. 2016b. “Tonal stability and tonogenesis in North Germanic.” In I. Giles, L. Chapot, C. Cooijmans, R. Foster, and B. Tesio (eds.), Beyond Borealism: New perspectives on the North. London: Norvik Press: 80–98. Itoˆ, J. and A. Mester 2015. “The perfect prosodic word in Danish,” Nordic Journal of Linguistics 38.1: 5–36. Jongen, R. 1972. Phonologie der Moresneter Mundart; Eine Beschreibung der Phonologie der Moresneter Mundart. Eine Beschreibung der segmentalen und prosodischen Wortformdiakrise. Assen: Van Gorcum. Kehrein, W. 2017. “There’s no tone in Cologne: Against tone segment interactions in Franconian.” In Kehrein et al. (eds.): 147–195. Kehrein, W, B. Ko¨hnlein, P. Boersma, and M. van Oostendorp (eds.) 2017. Segmental Structure and Tone. Berlin: Walter de Gruyter. Kingston, J. 2011. “Tonogenesis.” In M. van Oostendorp, C. J. Ewen, E. Hume, and K. Keren (eds.), The Blackwell Companion to Phonology. Oxford: Blackwell: 2304–2333.

Tone Accent in North and West Germanic

Kock, A. 1901. Die alt- und neuschwedische Accentuierung. Strassburg: Karl J. Tru¨bner. Ko¨hnlein, B. 2011. Rule Reversal Revisited: Synchrony and Diachrony of Tone and Prosodic Structure in the Franconian Dialect of Arzbach. Utrecht: LOT. Dissertation Series 274. Ko¨hnlein, B. 2013. “Optimizing the relation between tone and prominence: Evidence from Franconian, Scandinavian, and Serbo-Croatian tone accent systems,” Lingua 131: 1–28. Ko¨hnlein, B. 2015a. “An asymmetry in the interaction of pitch and duration,” Diachronica 32.2: 231–267. Ko¨hnlein, B. 2015b. “A tonal semi-reversal in Franconian dialects: Rule A vs. Rule B,” North-Western European Language Evolution (NOWELE) 68.1: 81–112. Ko¨hnlein, B. 2016. “Contrastive foot structure in Franconian tone-accent dialects,” Phonology 33.1: 87–123. Ko¨hnlein, B. 2017. “Synchronic alternations between monophthongs and diphthongs in Franconian: a metrical approach.” In Kehrein et al. (eds.): 211–236. Kristoffersen, G. 2000. The Phonology of Norwegian. Oxford University Press. Kristoffersen, G. 2006. “Markedness in Urban East Norwegian tonal accent,” Nordic Journal of Linguistics 29.1: 95–135. Kristoffersen, G. 2010. Fra jamvekt til etterleddstrykk og tonelag 3: Kvantitetsomleggingen in Ovansiljan. Maal og Minne 2010.2: 1–5. Lahiri, A., A. Wetterlin, and E. Jo¨nsson-Steiner 2005a. “Lexical specification of tone in North Germanic,” Nordic Journal of Linguistics 28: 61–96. Liberman, A. S. 1982. Germanic Accentology. Vol. 1: The Scandinavian Languages. Minneapolis: University of Minnesota Press. Meyer, E. A. 1937. Die Intonation im Schwedischen I: Die Sveamundarten. Stockholm University. Meyer, E. A. 1954. Die Intonation im Schwedischen II: Die Norrla¨ndischen Mundarten. Stockholm University. More´n-Duollja´, B. 2013. “The prosody of Swedish underived nouns: No lexical tones required,” Nordlyd 40.1: 196–248. Myrberg, S. and T. Riad 2015. “The prosodic hierarchy of Swedish,” Nordic Journal of Linguistics 38.2: 115–147. Myrvoll, K. J. and S. Trygve 2010. “Tonelagsskilnad i islendsk i Tridje grammatiske avhandling,” Maal og Minne 2010.1: 68–97. No¨rrenberg, K. 1884. “Ein niederrheinisches Accentgesetz,” Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 9: 402–412. Oftedal, M. 1952. “On the origin of the Scandinavian tone distinction,” Norsk tidsskrift for sprogvidenskap 16: 201–225. Oostendorp, M. van 2005. “Expressing inflection tonally,” Catalan Journal of Linguistics 4.1: 107–127. Oostendorp, M. van 2017. “Tone, Final Devoicing and Assimilation in Moresnet.” In Kehrein et al. (eds.): 237–252.

165

166

BJÖRN KÖHNLEIN

Peters, J. 2008. “Tone and intonation in the dialect of Hasselt,” Linguistics 46: 983–1018. Prehn, M. 2012. Vowel Quantity and the fortis-lenis Distinction in North Low Saxon. Utrecht: LOT Dissertation series. Riad, T. 1998. “The origin of Scandinavian tone accents,” Diachronica 15.1: 63–98. Riad, T. 2000. “The origin of Danish stød.” In A. Lahiri (ed.), Analogy, Levelling, Markedness: Principles of Change in Phonology and Morphology. Berlin: Mouton de Gruyter: 261–300. Riad, T. 2013. The phonology of Swedish. Oxford University Press. Ringgaard, K. 1960. Vestjysk stød. Aarhus: Universitetsforlag. Schmidt, J. E. 1986. Die Mittelfra¨nkischen Tonakzente (Rheinische Akzentuierung). Mainzer Studien zur Sprach- und Volksforschung 8. Stuttgart: Franz Steiner Verlag. Schmidt, J. E. 2002. “Die sprachhistorische Genese der mittelfra¨nkischen Tonakzente.” In P. Auer, P. Gilles, and H. Spiekermann, (eds.), Silbenschnitt und Tonakzente, Linguistische Arbeiten 463. Tu¨bingen: Max Niemeyer Verlag: 201–233. Schmidt, J. E. and H. J. Ku¨nzel 2006. “Das Ra¨tsel lo¨st sich: Phonetik und sprachhistorische Genese der Tonakzente im Regelumkehrgebiet (Regel B).” In de Vaan (ed.): 135–163. Smith, N. and K. van Leyden 2007. “The unusual outcome of a level-stress situation: The case of Wursten Frisian,” NOWELE. North-Western European Language Evolution 52.1: 31–66. Versloot, A. P. 2008. Mechanisms of Language Change: Vowel Reduction in 15th Century West Frisian. Doctoral Dissertation, Netherlands Graduate School of Linguistics. Werth, A. 2011. “Perzeptionsphonologische Grundlagen der Prosodie Eine Analyse der mittelfra¨nkischen Tonakzentdistinktion,” Zeitschrift fu¨r Dialektologie und Linguistik, Beiheft 143. Stuttgart: Franz Steiner Verlag. Wetterlin, A., and A. Lahiri 2015. The diachronic development of stød and tonal accent in North Germanic. In: D.T.T. Haug (ed.), Historical Linguistics 2013: Selected papers from the 21st International conference on historical linguistics (pp. 53–67). Oslo: University of Oslo. Wiesinger, P. 1970. Phonetisch-phonologische Untersuchungen zur Vokalentwicklung in den deutschen Dialekten. Bd. 1: Die Langvokale im Hochdeutschen. Bd. 2: Die Diphthonge im Hochdeutschen. Studia Linguistica Germanica. 2. Berlin: Mouton de Gruyter.

Chapter 8 Intonation in Germanic Mary Grantham O’Brien

8.1 Introduction This contribution focuses on intonation, which listeners perceive as the tune – or the rises and falls – of an utterance. The unit of analysis is the intonation phrase (IP), which can range in length from a single word to a complete sentence (Benware 1986). In Germanic languages intonation is used for a range of linguistic functions including phrasing (i.e., dividing the speech stream into chunks), signaling sentence mode (i.e., distinguishing, for example, declaratives from yes / no questions), and highlighting information (i.e., focus). Pitch does not perform its work in isolation, but it functions in combination with loudness and lengthening cues. The Swedish example from Bruce and Granstro¨m (1993: 70) in (1) demonstrates how the insertion of an IP boundary, marked via ||, works to disambiguate otherwise ambiguous utterances. (1)

a. Fast man offrade bonden || och lo¨paren ha¨lsade kungen. But you sacrificed pawn and the bishop greeted the king ‘But we sacrificed the pawn, and the bishop greeted the king.’ b. Fast man offrade bonden och lo¨paren || ha¨lsade kungen. ‘Though we sacrificed the pawn and the bishop, the king greeted us.’

Intonation may also be used paralinguistically, for example, to express emotions. Consider how an English speaker might modulate intonation to show emotion when uttering the statement “It’s snowing again.” If the speaker recently purchased new skis, s/he might indicate excitement by raising the pitch on “snowing.” On the other hand, if the speaker must move a car from the street to enable the snow plow to clear the street, s/he might express frustration by lowering overall pitch and using monotone intonation. For the purposes of the present

168

MARY GRANTHAM O’BRIEN

contribution, I will not devote much attention to the paralinguistic uses of intonation. This chapter will concentrate primarily on the grammatical role of intonation in West Germanic and North Germanic languages. Although the chapter mainly addresses standard varieties, some attention will be paid to dialectal variation. The chapter is organized as follows: Section 8.1 provides a basic overview of intonation and the ways in which it is investigated; Section 8.2 concentrates on declaratives across the languages in question; Section 8.3 focuses on the intonation of questions; Section 8.4 looks at special intonation contours; Section 8.5 provides a summary of illustrative research into listener judgments; and Section 8.6 offers an outlook for future research. The discussion of intonation presented here complements that in other chapters of this Handbook of Germanic Linguistics, especially Fe´ry’s treatment of information structure and Ko¨hnlein’s discussion of tone accents. Examination of both topics is therefore limited in this chapter.

8.2 The Fundamentals of Intonation Research 8.2.1 Analyzing Intonation Bolinger (1978) described intonation as a “half-tamed savage.” Gussenhoven (2004: 49) explains this metaphor by distinguishing between the “discretely represented prosodic structure” (the tamed half) and the “unusually generous scope” that speakers have “in the implementation of fundamental frequency” when they speak (the untamed half). Consider yes / no questions in English. Although the default intonation contour ends in a rise, speakers may choose to produce this type of question with level pitch or even with a fall. The available research enables us to make generalizations about both the linguistic forms and functions of intonation. It is possible to analyze intonation from different perspectives. Phonetic analyses examine the details of the acoustic correlate of pitch, fundamental frequency (F0). F0 is a measure of the frequency of vocal fold vibration, measured in Hertz (Hz). Pitch is perceived to be higher when a speaker’s vocal folds vibrate more quickly. Phonetic analyses of intonation often make use of pitch tracks like that in Figure 8.1, which is presented together with its corresponding spectrogram. Pitch tracks, the black lines in the Figure 8.1, present frequency (y axis) over time (x axis). In Figure 8.1, we see that the pitch rises, remains level for some time, and then falls. Researchers performing phonetic analyses use measurements including F0 minima and maxima and pitch range, which is the difference between a speaker’s highest F0 and the baseline within an F0 contour. Phonological studies of

Intonation in Germanic

Frequency (Hz)

5000

0

0

1.937 Time(s)

Figure 8.1 Pitch track of the German utterance “auf gar keinen Fall” (“by no means”)

intonation focus on intonation systems and therefore make use of more abstract categories that capture the overall pitch contour and relative pitch height.

8.2.2 The Form and Meaning of Intonation It is possible to make broad generalizations about roles of intonation contours in signaling sentence mode. Across the Germanic languages there is a tendency for declarative utterances and wh-questions to exhibit falling intonation and for yes / no questions to be produced with a rising contour (e.g., Benware 1986, Fe´ry 1993, A´rnason 1998, House 2005, Ambrazaitis 2009, Grønnum 2009). Continuation, for example, in the production of utterances containing a subordinate clause followed by an independent clause, is often signaled via a level or a rising contour (e.g., Delattre 1965, Fe´ry 1993, A´rnason 1998, Gibbon 1998, ’t Hart 1998). There is one constituent in Germanic utterances that receives the primary emphasis. It is spoken more loudly, and its duration is longer. In the case of neutral utterances spoken “out of the blue” or in response to a question like “What’s happening?”, speakers tend to emphasize the last content word of the utterance. It is possible, however, to highlight another unit in an utterance through the use of intonational cues. Consider the Icelandic question-answer pairs from Dehe´ (2009: 16) in (2) and (3). In (2), the question requires the speaker to emphasize, or focus, the direct object, whereas in (3), the question requires focus on the verb. The emphasized constituent in both instances receives the sentence stress, the phonetic manifestation of which is referred to as the nuclear pitch accent.

169

170

MARY GRANTHAM O’BRIEN

(2)

Direct object focus Q: Hvað skrifaði Marı´a upp? ‘What did Marı´a write up?’ A: Marı´a skrifaði [so¨guna]Foc upp. Marı´a wrote story.D E F up ‘Marı´a wrote the story up.’

(3)

Verb focus Q: Hvað gerði Marı´a við so¨guna? ‘What did Marı´a do with the story?’ A: Marı´a [skrifaði]Foc so¨guna [upp]Foc.

Discussion of intonation’s role in focus marking is taken up further in Fe´ry, Chapter 28.

8.2.3 Central Terms Together with stress and rhythm, intonation is classified as a suprasegmental, or prosodic, aspect of speech. Units of suprasegmental analysis can range from syllables to entire texts.1 Of primary concern in linguistic analyses of intonation are the IP and the pitch movements associated with both the prominent syllable and the end of an IP. The IP tends to be delimited – especially in read speech – by a pause and lengthening of the final syllable. In addition, the IP tends to drop in amplitude at its end and is often followed by pitch resetting (e.g., Cruttenden 1986, Grabe 1998). In spontaneous utterances, however, it can be more difficult to delimit IPs (Cruttenden 1986), and they may or may not coincide with syntactic units (Benware 1986: 114). Phonological models of intonation tend to fall into one of two categories: holistic and compositional. Holistic models such as the Fujisaki model (e.g., Fujisaki 1983) are based on the assumption that complete intonation contours carry meaning. Many of the compositional phonological models of intonation are based on the early work of Bruce (1977) on Stockholm Swedish, in which he demonstrates hierarchical structure of suprasegmental features. Importantly, he shows that intonation peaks are used to mark focus and that the basic units of intonation include word accents2 (i.e., accent 1 and accent 2 for Swedish), sentence accent (i.e., the syllable in an IP that is emphasized), and terminal juncture (i.e., boundary signals). An example of the interplay between the levels is provided in (4). Here Bruce demonstrates that the difference in meaning of stegen between 1

See Alber (Chapter 4) for additional information on word stress and van Oostendorp (Chapter 2) for more information on syllables in Germanic.

2

Word accents are distinctive prosodic patterns associated with stressed syllables in Swedish (accent 1 versus accent 2), Norwegian (toneme 1 versus toneme 2), and Dutch (stød versus non-stød). Accents are assigned to words, and they interact with the higher-level phenomena including focus and sentence mode, to name just a few. See Köhnlein (Chapter 7) for a more detailed discussion.

Intonation in Germanic

(a) vi bru`kade kla`ra ste´gen ‘we used to manage the steps’, (b) vi bru`kade kla`ra ste`gen ‘we used to manage the ladder’, and (c) vi bru`kade kla`raˌstegen ‘we used the steps / the ladder of Klara’ is distinguished at a particular level in the hierarchy. (4) stress word accent accent II sentence accent

(a) vi bru`kade kla`ra ste´gen - +- +- +- +- +- +- +- +-- +- ---

stress word accent accent II sentence accent

(b) vi bru`kade kla`ra ste`gen - +- +- +- +- +- +- +- +- +- +- ---

stress word accent accent II sentence accent

(c) vi bru`kade kla`raˌstegen - +- +- +- +- +-- +- +-- +- --Bruce (1977: 15)

An examination of all three utterances indicates that the stress assignment is the same. Utterances (a) and (b) are distinguished at the level of accent II, such that kla`ra ste`gen (‘ladder’) is realized as accent II + accent II in utterance (b). Utterance (c) is distinguished from the others at the level of word accent, where kla`raˌstegen is realized as one unit. Bruce’s work inspired Pierrehumbert (1980), who posited a connection between the melody of the utterance on the one hand, and prominence and phrasing on the other (Arvanti in press). Ladd (1996) coined the term autosegmental-metrical (AM) to refer to descriptions of this type. Importantly, AM models distinguish between the phonological abstraction (i.e., the tones) and the phonetic realization of an utterance. Within AM models every IP must minimally contain a nuclear pitch accent, that is, the most prominent syllable of that phrase. The nuclear pitch accent is usually accompanied by a pitch change. The syllable receiving the nuclear pitch accent is annotated as T*, where T is a tone that corresponds to a high (H) or low (L) target in the pitch contour (Fe´ry 1993, A´rnason 1998). The notions of high and low are assigned based on the extent to which a pitch excursion fits within a given speaker’s range. Each H* or L* is often either preceded or followed by a pitch movement. For example, L+H* indicates that the pitch preceding the high accented syllable is low, whereas an L*+H accent is low on the stressed syllable and rises following the low pitch. Finally, an IP is characterized by a boundary tone, that is the pitch at the end of the

171

172

MARY GRANTHAM O’BRIEN

contour, which can either be high (H%) or low (L%)3. An example of an AM annotation of a German utterance is provided in (5). (5)

H* L% | | [Weil er sich an mir vollgefressen hat.] Because he himself on me stuffed has ‘Because he stuffed himself on me.’ von Heusinger (2008: 278)

The goal of AM annotations is to both model meaningful distinctions such as modality (e.g., declaratives versus interrogatives) and capture similarities within a given melody that may differ in terms of phonetic realization (e.g., the “hat pattern” described in Section 8.3.1) within a language or potentially across languages (Arvanti in press). There is a strong tradition of AM-based modelling of Germanic languages (e.g., GToBI for German, Grice and Baumann 2002; ToDI for Dutch, Gussenhoven 2010; ToBI for English, Beckman et al. 2005; the Lund Model for Swedish, Bruce et al. 1998). The annotations in this chapter are not attached to a particular annotation model. Common topics of investigation in phonetic analyses of intonation include the timing and scaling of F0. Research investigating timing focuses on the alignment of a pitch high or low with a particular segmental anchor in the speech stream. For example, Atterer and Ladd (2004) and O’Brien and Gut (2011) investigated how German speakers align prenuclear pitch troughs (F0 minima) and peaks (F0 maxima). The results of Atterer and Ladd (2004) demonstrate that speakers of Northern and Southern German show different alignment patterns, and those of O’Brien and Gut (2011) show that German speakers tend to align pitch peaks differently in their first language (L1) and their second language (L2) English, with the alignment occurring later in English than in German. Participants in the study read sentences such as those in (6) and (7), from Atterer and Ladd (2004). (6)

Die mollige Dame bezauberte durch ihr La¨cheln. ‘The plump woman charmed with her laughter.’

(7)

I need a monosyllabic word for my crossword puzzle.

The locations of F0 minima and maxima within the syllables in bold in (6) and (7) were averaged over 15 German sentences and 13 English sentences, all from Atterer and Ladd (2004). Figures 8.2 and 8.3 (revised from O’Brien 2013: 48–49) present differences in tonal alignment for German speakers from the North (Potsdam) and the South (Augsburg) in German and in English. 3

Although it is also possible to annotate prenuclear information, the examination thereof goes beyond the scope of the current contribution.

Intonation in Germanic

North South

C0

C1

V0

V1

Figure 8.2 Tonal alignment in L1 German

North South

C0

V0

C1

V1

Figure 8.3 Tonal alignment in L2 English

Dehe´ (2010) examined F0 alignment in Icelandic, and Van de Ven and Gussenhoven (2011) and Lickley et al. (2005) investigated F0 alignment in Dutch. The results of studies investigating tonal alignment demonstrate salient cross-linguistic and regional variation. Research into tonal scaling looks at the lowering and raising of high and low tones. Fox (1984) notes that wider pitch ranges can be used to attract attention and signal emphasis. Fe´ry and Ku¨gler (2008) found that German high tones are raised in narrow focus conditions, and they are lowered in prenuclear position. Gussenhoven and Rietveld (2000) looked at how F0 variations affect perceived prominence in Dutch, and they found evidence that listeners perceive a difference between low and high rises. Although they all make use of intonation, not all Germanic languages have been classified as intonation languages, that is, languages in which intonation is used to convey meanings at the level of the utterance. According to many accounts, Swedish and Norwegian belong to another class of languages known as pitch accent languages, in which lexical tone is used to distinguish the meanings of words (Bailey 1990: 80; Kristoffersen 2003). Danish makes use of a kind of lexically specified creaky voice or glottal stop known as stød, to modify stress group patterns (Grønnum 1998). In these three languages as well as in certain dialects of Dutch (Gussenhoven and van der Vliet 1999) there is an interaction between intonation and these other prosodic features. Given these differences

173

174

MARY GRANTHAM O’BRIEN

across systems, researchers have failed to agree on an annotation system that captures the meaningful units and distinctions of all Germanic languages (e.g., Grønnum 1998, 2009). Annotations in this chapter therefore focus primarily on nuclear pitch accents and boundary tones.

8.2.4 Data Sources Studies investigating intonation make use of a range of data types. Production data can include controlled, read speech samples to ensure both that speakers produce the kinds of utterances a researcher wishes to analyze and that utterances can be directly compared across speakers. Speakers may read sentences out of context (Atterer and Ladd 2004), sentences preceded by a particular context (Fe´ry and Ku¨gler 2008), or longer texts (Peters 2006). Grønnum (2009) notes that the data obtained from read speech samples may serve as the basis for studies investigating more spontaneous speech, and researchers’ careful preparation of read texts may enable meaningful cross-linguistic comparisons (Grabe 1998). At the other end of the spectrum are spontaneous utterances that are representative of casual speech produced by speakers in the real world. These can take the form of monologues (Grønnum 2009) or dialogues (House 2005). In order to minimize the variation in unscripted speech samples, researchers often choose to have speakers complete goal-oriented tasks such as map tasks that contain a series of previously identified landmarks (Lickley et al. 2005). Given the complexities inherent in examinations of intonation, it is necessary to limit the scope of analyses, both in terms of the languages covered and the types of analyses presented. The languages chosen for this chapter include standard varieties and some regional varieties of German, Dutch, English, Icelandic, Danish, Norwegian, and Swedish. This contribution primarily focuses on the linguistic use of intonation in complete utterances, and it relies on research into both production and perception.

8.3 Declaratives This section deals with phrase-final intonation in all new utterances. Although Fe´ry (2008: 363) notes that broad focus utterances are uttered in an “informational vacuum,” they are important for providing the baseline for additional analyses. Fox (1984: 60) emphasizes the importance of establishing what is “normal” intonation and then analyzing other patterns on the basis thereof. As a first step we will focus on commonalities across the Germanic languages and will then move to specifics within each of the languages. In the case of some languages, especially German, we will highlight work that demonstrates regional variation.

Intonation in Germanic

Across Germanic languages, there is a general tendency for a fall in fundamental frequency over the course of an utterance, or longer text.4 This has been demonstrated for German5 (Truckenbrodt 2004), Dutch (’t Hart 1998), English (Pierrehumbert 1980), Icelandic (Dehe´ 2009), Danish (Grønnum 2009), Swedish (Gaº rding 1998), and Norwegian (Kristoffersen 2000). An example of a downward trend in Danish is provided in (8). (8)

Ammerne i Alabama var i strejke. LH∗L H∗L H∗L L ‘The nurses in Alabama were on strike.’

(Gussenhoven 2004: 225)

It is also common to deaccent given information (i.e., the information that is recoverable from discourse, presupposed, or that the speaker believes is in the interlocutor’s foreground, Fe´ry 1993: 17). This has been demonstrated for German (Baumann 2006), Dutch (Gussenhoven 2004), English (Ladd 1996), and Swedish (Myrberg and Riad 2015), although deaccenting is not obligatory in Icelandic (Dehe´ 2009, A´rnason 2011).

8.3.1 German Researchers investigating German intonation tend to agree that two tone levels, one high and the other low, are sufficient for an accurate description (e.g., Isacˇenko and Scha¨dlich 1970, Fe´ry 1993, Baumann 2006). If we combine the tones, we can describe contours as rising, falling, or level. Within German broad focus utterances, the last lexical word is emphasized, most often via a H*+L pitch accent, and a plain fall is the most common utterance-final movement (Fe´ry 2008; Fox 1984). Although researchers concur that most intonation contours lack consistent meaning (e.g., Kohler 1992, Fe´ry 1993: 106), there is general agreement that falling intonation signals finality, and it is used primarily with both unmarked statements and categorical assertions (Uhmann 1991, Kohler 1992, Gibbon 1998). Continuation is marked within German with a rising slope (Delattre 1965, Pu¨rschel 1975, Fe´ry 1993) or level pitch (Fox 1984, Gibbon 1998). Kohler (1992), who proposes that rising intonation is also used in incomplete utterances and requests, equates rising intonation with an appeal to 4

A range of terms has been used to describe the different types of downward trends at the ends of utterances. These include declination, downdrift, downstep, and final lowering (Hirst and Di Cristo 1998: 21). A more thorough discussion of the differences among the terms goes beyond the scope of the current contribution.

5

Ulbrich (2004) did not find evidence for declination in the read speech of standard Swiss speakers.

175

176

MARY GRANTHAM O’BRIEN

the listener. When a speaker wishes to combine two IPs in a single German utterance, there are two main tendencies. Utterances made up of two independent clauses, both of relatively equal communicative value, which are conjoined with a coordinate conjunction, usually take the form of two declarative utterances. That is, the two independent IPs each end in a final fall (Fox 1984: 91). Utterances consisting of a subordinate clause followed by an independent clause take the form of a high boundary tone on the first IP followed by a fall in the second (Fox 1984: 92–93). A number of recent studies have examined regional variation in intonation. Much of the research focuses on the realization of a particular feature in a specific variety, and Peters (2006) points to the importance of examining not only the forms, but also the functions of intonation across varieties. Ku¨gler’s (2004) study on nuclear rises in Swabian German provides evidence for Sievers’ (1912) claim that the tonal system of Swabian has “inverted” nuclear accents as compared to those of Standard German. That is to say, whereas the default pitch accent for Standard German is H*+L, Swabian German’s default is L*+H. Bergman’s (2009) investigation of the “hat pattern” (i.e., a rise-fall contour) in Cologne German, based on 14 hours of spontaneous dialogue, unearthed 51 instances of the pattern. Although she found some overlap between the use of the contour in Standard German and Cologne German, Bergmann also discovered a great deal of variation in the phonetic realization of the contour as well as syntactic and information structural differences between its use in Standard and Cologne German. Leeman and Zuberbu¨hler’s (2010) phonetic study of declarative contours in Swiss German dialects demonstrates that the primary differences across dialects is due to timing in simple declaratives. Ulbrich (2004) compares intonation across speech samples produced by model speakers (i.e., newscasters) of Standard Swiss and German German. Acoustic analyses demonstrate slower speech and articulation rates, larger F0 intervals, and differences in peak location for the Swiss speakers. Ulbrich’s (2006) follow-up study looking at differences between spontaneous and semi-spontaneous data, demonstrates that the differences in pitch range between Swiss and German intonation may be due to the fact that German spoken in Germany has fewer and smaller syllable-internal F0 movements. Fitzpatrick-Cole’s (1999) findings surrounding the differences in default nuclear accents in Bern Swiss German and Northern Standard German are in line with those of Ulbrich (2004) and Ku¨gler (2004). Importantly she found that the default accent in Bern is a rising accent (L*+H), whereas that of Northern Standard German is falling (H*+L). She points to this categorical difference as being phonological. Like Ulbrich (2004), she also found evidence for differences in peak alignment, with Bern speakers aligning F0 peaks later than Northern German speakers.

Intonation in Germanic

8.3.2 Dutch Analyses of Dutch declaratives show similar results to those of German (Gibbon 1998), although Dutch intonation may demonstrate relatively less F0 movement than German intonation (’t Hart 1998). ’t Hart’s (1998) comprehensive analysis, based on a range of data sources including isolated words in citation form, radio interviews, news bulletins, university lectures, and theater plays, suggests that the default intonation pattern of non-emphatic Dutch declaratives is a rise-fall. The most common pattern is the hat pattern, which may have either one or two pitch accents. Utterances with two pitch accents are often realized with high pitch on the intervening syllables such that the F0 contour between the stressed syllables is level, as in Figure 8.4. Longer Dutch utterances may be broken up in various ways (’t Hart 1998). As in German, Dutch speakers often use the continuation rise, which involves a sustained high pitch after the nuclear pitch accent, followed by a falling boundary tone. Another option is a sustained low pitch followed by a continuation rise after a falling pitch accent. ’t Hart notes that this second option is common when a speaker wishes to create boundaries between two main clauses, whereas the first option (i.e., with the continuation rise) can be used to separate a subordinate and a main clause. 8.3.3 English Among the West Germanic languages, English is usually described as having larger pitch movements than both German and Dutch (Gibbon 1998, Grabe 1998, ’t Hart 1998). This may be because English relies more on intonation than other Germanic languages do, due to its relatively set word order (Fox 1984, Gibbon 1998). In her comparison of Southern Standard British English and Northern Standard German, Grabe (1998) found similarities in the use of falling pitch accents, which she categorizes as H*+L in both languages. As with other Germanic languages, the basic contour of English declaratives is a fall with nuclear pitch accent on the final content word (Hirst 1998). Bolinger (1998) posits that Standard American English and Standard British English share a single intonation system, but a few of the important differences are taken up below. 8.3.4 Icelandic Like other Germanic languages, Icelandic generally exhibits rightmost sentence stress in neutral declaratives (Dehe´ 2009). Falling pitch accents (H*L) are the most frequent, although rising accents (L*H) and monotonal

Peter

komt

nooit

te

laat

Figure 8.4 A stylized Dutch hat pattern (“Peter comes never too late.”) recreated from ’t Hart (1998: 97)

177

MARY GRANTHAM O’BRIEN

pitch accents (H*) and (L*) also exist (Dehe´ 2009). A´rnason (2011) proposes that the use of the marked L*H contour indicates that a speaker is making a friendly suggestion. Declarative utterances tend to end with a low boundary tone, L% (A´rnason 1998, Dehe´ 2009). A´rnason (2011) reports that speakers from northern Iceland are more likely to use H% boundary tones, which may give the impression that speakers from the north show more anger or irritation (p. 325). In utterances composed of two IPs, the first tends to end with rising pitch, which can take the form of either L*H H% or H*L H% (A´rnason 1998, 2011). This is used as a mark of nonfinality, as shown in (9).6 (9)

Jón er SKEMMTILEGUR, pótt hann sé varhugaver∂ur H∗L H% !HL L% ‘John is entertaining, although he is not to be trusted’

(A´rnason 1998, 321)

8.3.5 Danish Although the overall contour is assigned at the IP, Danish intonation is directly influenced by the stress group or foot, which is composed of the stressed syllable and any following unstressed syllables (Grønnum 1998). The F0 pattern within a stress group tends to consist of a low stressed syllable, which is followed by a high, unstressed syllable and a series of low unstressed syllables. Although Grønnum (1998) notes that Copenhagen Danish does not have a compulsory default contour, the nuclear pitch accent of the IP tends to fall on the final element in Bornholm. An important feature of nuclear stress assignment is that instead of exhibiting a raised F0 peak, a nuclear pitch accent has the impact of lowering, shrinking, or deleting the F0 peaks in the surrounding stress groups. Figure 8.5, reproduced from Grønnum (2009: 600), demonstrates the relative slopes of

declarative questions 6 semitones

178

questions with word order inversion wh-questions

statements Figure 8.5 Stylized intonation contour of Danish statements, wh-questions, questions with word order inversion, and declarative questions (recreated from Grønnum 2009: 600) 6

The !HL annotation indicates that the rising intonation in the second IP is downstepped in relation to the H*L in the first IP.

Intonation in Germanic

declaratives as opposed to wh-questions, questions with word order inversion, and declarative questions. Overall we see that the intonation pattern of declaratives is a relatively steep falling slope, especially when compared to the other utterance types.

8.3.6 Norwegian Norwegian differs from the other Germanic languages in that there are unofficial norms, with East Norwegian intonation being widely studied (Fretheim and Nilsen 1989, Kristoffersen 2000). The options for differentiating meaning via intonation are rather limited in Norwegian, given the need to maintain lexical contrasts via tone accents (Kristoffersen 2000: 274). Kristoffersen (2000) and Fretheim and Nilsen (1989) indicate that focused elements are marked via a rising pitch movement, and this is most often followed by a falling movement. A feature that Norwegian shares with other Germanic languages is the use of a rising contour in nonfinal position (Fretheim and Nilsen 1989).

8.3.7 Swedish Like Norwegian, Swedish makes use of tone to distinguish lexical items, and intonation and word accents interact (Meyer 1937). Gaº rding (1994) points out that it is important to distinguish between global intonation that stretches over a phrase and local accents and tones. It has been proposed that the interaction between tone and intonation limits the potential intonation contours in Swedish, but Ambrazaitis (2009) and Myrberg and Riad (2015) argue that because it is used for similar purposes including phrasing and accentuation, the intonation of Swedish is actually quite similar to that of West Germanic languages. According to Bailey (1990) and Bruce and Granstro¨m (1993), Swedish declaratives begin and end with L tones, and the last content word in the utterance is stressed by default (Gaº rding 1998). Declarative intonation contours tend to rise on the last stressed vowel, and the F0 maximum is realized in the vowel following the stressed syllable (Bailey 1990: 59). Ambrazaitis (2009) indicates that continuation is marked as a rise (LH%), which is similar to that of other Germanic languages.

8.4 Interrogatives Interrogatives across the Germanic languages may take three forms: those that have a question word, those that are syntactically marked as yes / no questions, and those that make use of declarative word order, most often with an accompanying rise in pitch (Gunlogson 2002). German questions tend to follow the basic pattern in Germanic: wh-questions tend to end in falling pitch, and interrogatives without

179

180

MARY GRANTHAM O’BRIEN

Heef

Peter

een nieuwe

auto

gekoeht?

Figure 8.6 Contour of a Dutch interrogative “Heeft Peter een nieuwe auto gekocht?“ (“Has Peter bought a new car?”) reproduced from ’t Hart (1998: 103)

a question word tend to rise (Kohler 2004). The system is certainly much more complex than this, and there are exceptions to the general rules. Benware (1986: 112) indicates that wh-questions tend to rise in pitch if a speaker wishes the speaker to repeat a previous utterance, and Fox (1984) and Kohler (2004) note that rising patterns in such questions can indicate a speaker’s interest, friendliness, or openness. Interrogatives without question words may be produced with falling patterns to signal that a speaker is being “assertive” (Fox 1984: 62) or that he or she lacks interest (Kohler 2004). ’t Hart (1998) posits that there is no specific contour for questions in Dutch. He indicates that although they may end with a final rise, his corpus analysis indicates that speakers made use of rising intonation about half of the time. As in declaratives, the hat pattern is common in Dutch questions, as demonstrated in Figure 8.6. The results of a larger corpus analysis carried out by Rietveld et al. (2002) examined a total of 600 Dutch questions, 85 percent of which ended in H%. English questions follow the same basic pattern as German and Dutch. Hirst (1998) notes that although it appears that rising intonation is generally used when speakers utter a question with declarative word order, this use of rising intonation is instead use to indicate that “a syntactic statement is being used pragmatically as a request for information” (1998: 65). See Bartels (1997) for a thorough discussion of intonation in English questions. A´rnason (2011) proposes that L*H L% is the form of neutral questions in Icelandic. Dehe´’s (2009) study found that speakers of Icelandic tend to use H% in both yes/no- and wh-questions, a finding that aligns with A´rnason’s (1998) proposal that H% is an indicator of non-finality. Dehe´’s (2009: 27) analysis indicates that speakers of Icelandic may produce yes/no-questions with either an H%, to indicate a “friendly suggestion”, or an L% to indicate that the question is “matter of fact.” Grønnum (1998: 140) describes the intonation contours of unmarked questions in Danish as “horizontal.” She proposes that intonation is assigned in questions in a trade-off relationship with syntax. That is to say, questions containing more syntactic (yes/no-questions) or lexical (question words) information tend to be produced with more falling intonation. This general finding is represented in the stylized intonation contour in Figure 8.5. In a study of 18 speakers from Cophenhagen, Grønnum (2009) found a relatively large number (N=139) of questions produced with declarative word order. Almost one third of those utterances (N=41) were produced with intonation that could not be perceptually distinguished from true declaratives.

Intonation in Germanic

Fretheim and Nilsen (1989: 164) posit that there is no single question intonation in Norwegian, but they note that speakers tend to make use of “non-falling intonation” when they produce questions. The intonation that the speaker chooses signals to the listener that a speaker is either providing (falling intonation) or requesting (rising intonation) information (Fretheim and Nilsen 1989: 162). As in other Germanic languages it is possible to turn a Norwegian statement into a question via rising intonation. Questions in Swedish optionally end with a final rise (Bailey 1990, House 2005), and this may be due to the fact that interrogatives are usually expressed via word order or lexical means (Gaº rding 1998, House 2005). Intonation can be used to mark interrogatives not marked via word order (Gaº rding 1998). Bailey (1990) found that Swedish questions exhibit an overall higher pitch range than statements, in line with the [raised peak] feature proposed by Ladd (1983). House (2005) investigated the use of phrase-final rises in Swedish wh-questions in computer-directed spontaneous speech. The August database (Gustafson et al. 1999), which served as the corpus for the study, included over 10,000 utterances produced primarily by speakers from central Sweden at a kiosk located in central Stockholm. The kiosk was set up to provide answers to questions about local attractions. The analyses of wh-questions indicated that 22 percent were produced with a final rise. Children produced the largest percentage of rises (32 percent), and women and men produced relatively fewer rises on questions (27 percent and 17 percent, respectively).

8.5 Special Contours There is a tendency across Germanic languages for imperatives to be spoken with falling contours. Tag questions, which generally take the form of negative tags after positively worded statements and positive tags after negatively worded statements (Heinemann 2010: 2707), tend to be produced with rising intonation (e.g., ’t Hart 1998 for Dutch, Fe´ry 1993 for German, Fretheim and Nilsen 1989 for Norwegian). Consider the Norwegian example in Figure 8.7 (Fretheim and Nilsen 1989: 168). Fretheim and Nilsen (1989: 167) note that Norwegian tag questions may also be uttered falls, as in Figure 8.8. In this case, they are interpreted as “pessimistic” or implying finality. Although there is a tendency across Germanic languages for imperatives to be spoken with falling contours, Kohler (1992) and Fox (1984) note that the meaning assigned to German imperatives depends on the boundary tone. Whereas a falling contour is the default, a slightly rising imperative is taken as a polite request, and when spoken with a high rise, an imperative takes on a note of insistence, urgency, or encouragement.

181

MARY GRANTHAM O’BRIEN

Pitch (Hz)

300

70 1.845

3.406 Time (s)

Figure 8.7 Norwegian utterance “Du aner ikke hvor dyr den er, vel?” (“You do not know how expensive it is, right?”) with rising intonation on the tag question

300

Pitch (Hz)

182

70 0.7932

2.815 Time (s)

Figure 8.8 Norwegian utterance “Du aner ikke hvor dyr den er, vel.” (“Well, you do not know how expensive it is.”) with falling intonation on the tag

Incomplete sentences such as greetings or those used when taking leave can be produced with a range of intonation patterns. Fox (1984) indicates that it is difficult to determine what the “normal” pattern in German is, since these utterances are incomplete sentences. He notes, for example, that the meaning of a greeting like Morgen! ‘Good morning!’ depends on the contour. If it spoken with a falling contour, it is interpreted as an assertion (e.g., as an answer to a question). When spoken with a rising

Intonation in Germanic

contour, it can be interpreted as a question or a challenge. Gibbon (1998) notes that many greetings are spoken in a level, chanting (or call) contour. This contour is often interpreted as unmarked when used for greetings, but the same contour can be used to signal a request for a repetition or “discourse repairs caused by mishearing” (Gibbon 1998: 91). Gibbon (1998: 91–92) provides examples of these, reproduced in (10). (10) German stylized patterns a. Call

e— la—! Manu

b. Leave taking

Wie– der– sehen—! Auf

c. Request for louder repetition

Lau— ter—!

d. Repetition after mishearing

“Jo— hann” — ge —sagt —! Ich habe

This description is similar to Kristoffersen’s (2000) findings regarding the East Norwegian calling contour, which is often used with names. It consists of two tones, the first one high and the following somewhat lower (H* on the stressed syllable, and !H on the following syllable). A relatively recent trend in varieties of English intonation is the use of the high-rise terminal (HRT) contour, or “uptalk”, in declaratives. Transcribed as L*H-H% or H* H-H% (Barry 2007, Fletcher et al. 2002), it is characterized by either a substantially higher F0 or a rise in F0 in the final syllable of an utterance (Britain and Newmann 1992). In addition, HRTs tend to occur in succession, thereby functioning as a cohesive device (McLemore 1991: 100, Britain and Newmann 1992). Although early descriptions of HRTs tended to associate them with question intonation (Ching 1982), they have more recently been associated with continuation (Fletcher et al. 2002). Britain and Newmann’s (1992) study on the use of HRT among English speakers in New Zealand demonstrates that it tends to be used more frequently by women and younger speakers. Barry’s (2007) investigation into the use of HRTs among speakers of Southern Californian and Southern British English indicates that the pragmatic function of the HRT differs across the varieties: whereas London speakers used it to request verification of a listener’s comprehension at the end of an utterance, speakers from Southern California used it to extend turns. A´rnason (2011: 321) indicates, as noted above, that an H% can be used utterance internally to delimit IPs in Icelandic. He notes that the second IP

183

184

MARY GRANTHAM O’BRIEN

can also be left off of the utterance, leaving an H%-final IP. Although the meaning of the utterance remains the same, utterances of this type are interpreted as ironic statements. The example in (11) is an incomplete version of (9) above, and it can be interpreted ironically, meaning “John is (the opposite of) entertaining.” (11) Jón er SKEMMTILEGUR H∗L H% ‘John is entertaining’

(A´rnason 2011, 321)

8.6 Listener Judgments Much of what we know about the meanings of intonation contours comes from listener judgments. That is, researchers investigate listeners’ ability to both assign meaning to particular contours and to distinguish among dialects on the basis of intonation cues. Some studies have looked at listeners’ sensitivity to F0 differences. Remijsen and Van Heuven (2003) investigated whether listeners perceive Dutch statements and questions categorically, like they do segments, or whether differences in intonation are more continuous. Participants in the study completed both a classification task and a discrimination task. Although the results were characterized by a great deal of between-subject variation, the authors conclude that participants did indeed perceive the contrast categorically. In spite of what appears to be discriminatory precision, the authors conclude that intonation is not well suited for categorical perception research, given the range of meanings that can be assigned to a particular contour as well as the importance of assigning intonational meaning syntagmatically. Some intonation research investigates the attitudinal meanings. Listeners in Rietveld et al. (2002) assigned ratings to Dutch questions that varied in pitch accents and final boundary tones. The results point to the importance of pitch levels in the assignment of scores, such that those scales that are associated in one way or another with dependence (e.g., surprise, appeal) contained the highest proportion of H tones throughout the utterances (i.e., %H, H*, H%). In addition to providing the analysis of Swedish wh-questions described above, House (2005) also carried out a perception study to determine whether questions produced with rises are intended for social interaction, as opposed to merely seeking information. Participants were presented with pairs of manipulated stimuli and were asked to choose the friendlier sounding of the pair of utterances. Those questions produced with a rise were judged as friendlier and indicated a higher level of social interest. He interprets the findings by appealing to Bell and Gustafson’s (1999) proposal that final rises are used to socialize and that final falls are used to

Intonation in Germanic

seek information (p. 273). Fretheim and Nilsen (1989) tested listeners’ responses to East Norwegian imperatives. They found a difference in attitudinal meaning assigned to falling and nonfalling terminals. Those produced with a rising tone addressed to an adult are interpreted as threats, and those produced with a fall are interpreted as a “friendly offer performed by a benevolent agent” (Fretheim and Nilsen 1989: 177). Given the extent to which intonation features vary by language and by language variety, some perception research investigates whether listeners are able to classify speech as belonging to a particular variety. Grønnum (1994) investigated whether listeners could distinguish among six dialects of Danish. Although she was primarily interested in rhythmic differences across the varieties, the author found that the most important cue that the listeners relied upon in their judgments was F0 differences in nuclear pitch accent across the varieties. Gooskens (2005) explored how well Norwegians could identify Norwegian dialects when they were presented with two types of speech samples: complete samples with both segments and intonation intact and monotonized samples with intonation removed. Given participants’ difficulty with the task and their ability to better identify speech samples containing both segments and intonation, she concludes that Norwegian listeners rely primarily on intonation when identifying dialects.

8.7 Outlook Improvements in speech recording and analysis software such as Praat (Boersma and Weenink 2019) make it possible for more robust data collection and acoustic analyses, and the use of similar annotation schemes makes it possible for researchers to compare languages and language varieties with more confidence than ever before. I expect that developments in text-to-speech systems like those described by, for example, Siebenhaar et al. (2004), and those included in Burkhardt (2017) will continue to provide researchers with important insights into the form and functions of intonation both within and across languages. Intonation research is growing within the field of second language acquisition. Jilka (2000: 58) describes “intonational foreign accent” as a particular type of deviation from nativelike speech in which “intonation in the speech of a nonnative speaker must deviate to an extent that it is clearly inappropriate for what is considered native.” Gut’s (2009) corpus analysis provides important insights into the differences in the intonation systems of native and nonnative speakers of German and English. Importantly it also points to some of the more general issues that come with the analysis of nonnative intonation (e.g., the limits of AM annotations, the role of fluency, and the differences in task types). Although much of the research in the field of second language pronunciation is less interested in foreign accent and more concerned with what makes the speech of language learners intelligible

185

186

MARY GRANTHAM O’BRIEN

(Levis 2005), research has shown that prosodic deviations do indeed affect the ability of nonnative speakers to be understood (Hahn 2004). Nonetheless, there is a dearth of studies systematically investigating the extent to which a range of deviations in intonation produced by language learners affect understanding. When the insights gained from studies comparing the breadth of features and associated meanings of native speaker intonation and that of language learners (see, e.g., Grice and Baumann 2007) are combined with the technological developments described above, there is great potential for improvements in the tools that second language learners may use to become more comprehensible speakers.

References Ambrazaitis, G. 2009. Nuclear Intonation in Swedish: Evidence from ExperimentalPhonetic Studies and a Comparison with German. Travaux De l’Institut De Linguistique De Lund, 49. A´rnason, K. 1998. “Toward an analysis of Icelandic intonation.” In Werner (ed.), Nordic Prosody: 7th Conference. Frankfurt, Berlin, and New York: Peter Lang: 49–62. A´rnason, K. 2011. The Phonology of Icelandic and Faroese. Oxford University Press. Arvanti, A. in press. “The autosegmental metrical model of intonational phonology.” In S. Shattuck-Hufnagel and J. Barnes (eds.), Prosodic Theory and Practice. Cambridge, MA: MIT Press. Atterer, M. and D. R. Ladd 2004. “On the phonetics and phonology of ‘segmental anchoring’ of F0: Evidence from German,” Journal of Phonetics 32: 177–97. Bailey, L. M. 1990. A Feature-Based Analysis of Swedish Pitch Accent and Intonation. Unpublished Ph.D. dissertation, University of Delaware. Bartels, C. 1997. “The pragmatics of wh-question intonation in English,” University of Pennsylvania Working Papers in Linguistics 4: 1–17. Barry, A. S. 2007. The Form, Function and Distribution of High Rising Intonation in Southern Californian and Southern British English. Unpublished Doctoral dissertation, University of Sheffield. Baumann, S. 2006. “Information structure and prosody: Categories for spoken language annotation.” In S. Sudhoff, D. Lenertova´, R. Meyer, S. Pappert, P. Augurzky, I. Mleinek, N. Richter, and J. Schließer (eds.), Methods in Empirical Prosody Research. Berlin and New York: Mouton de Gruyter: 153–180. Beckman, M. E., J. Hirschberg, and S. Shattuck-Hufnagel 2005. “The original ToBI system and the evolution of the ToBI framework.” In S.-A. Jun (ed.), Prosodic Typology: The Phonology of Intonation and Phrasing: Oxford University Press: 9–54. Bell, L. and J. Gustafson 1999. “Utterance types in the August dialogues.” In Interactive Dialogue in Multi-Modal Systems: 81–84. www.isca-speech.org /archive_open/archive_papers/ids_99/ids9_081.pdf.

Intonation in Germanic

Benware, W. A. 1986. Phonetics and Phonology of Modern German. Georgetown University Press. Bergmann, P. 2009. “Regional variation in intonation: Conversational instances of the ‘hat pattern’ in Cologne German.” In F. Ku¨gler, C. Fe´ry, and R. van de Vijver (eds.), Variation and Gradience in Phonetics and Phonology. Berlin: Mouton de Gruyter: 377–404. Boersma, P. and D. Weenink 2019. Praat: Doing Phonetics by Computer (version 6.1.01). www.fon.hum.uva.nl/praat/. Bolinger, D. 1978. “Intonation across languages.” In J. H. Greenberg (ed.), Universals of Human Language. Stanford University Press: CSLI: 471–524. Bolinger, D. 1998. “Intonation in American English.” In D. Hirst and A. Di Cristo (eds.): 45–55. Britain, D. and J. Newman 1992. “High rising terminals in New Zealand English,” Journal of the International Phonetic Association 22: 1–11. Bruce, G. 1977. Swedish Word Accents in Sentence Perspective. Lund: Gleerup. Bruce, G. and B. Granstro¨m 1993. “Prosodic modeling in Swedish speech synthesis,” Speech Communication 13: 63–73. Bruce, G., J. Frid, B. Granstro¨m, K. Gustafson, M. Horne, and D. House 1998. “Prosodic segmentation and structuring of dialogue.” In Nordic Prosody – Proceedings of the VIIth Conference, Joensuu, 1996. Frankfurt: Peter Lang Verlag: 63–72. Burkhardt, F. 2017. German text-to-speech. http://ttssamples.synthetic speech.de/. Ching, M. K. L. 1982. “The question intonation in assertions,” American Speech 57: 95–107. Cruttenden, A. 1986. Intonation. Cambridge University Press. Dehe´, N. 2009. “An intonational grammar for Icelandic,” Nordic Journal of Linguistics 32: 5–34. Dehe´, N. 2010. “The nature and use of Icelandic prenuclear and nuclear pitch accents: Evidence from F0 alignment and syllable/segment duration,” Nordic Journal of Linguistics 33: 31–65. Delattre, P. 1965. Comparing the Phonetic Features of English, French, German and Spanish. Heidelberg: Julius Gross Verlag. Fe´ry, C. 1993. German Intonational Patterns. Tu¨bingen: Max Niemeyer Verlag. Fe´ry, C. 2008. “Information structural notions and the fallacy of invariant correlates,” Acta Linguistica Hungarica 55.3/4: 361–379. ´ Fery, C. and F. Ku¨gler 2008. “Pitch accent scaling on given, new and focused constituents in German,” Journal of Phonetics 36: 680–703. Fitzpatrick-Cole, J. 1999. “The Alpine intonation of Bern Swiss German.” Proceedings of the XIVth International Conference of the Phonetic Sciences (ICPhS), San Francisco, 941–44. Fletcher, J., L. Stirling, I. Mushin, and R. Wales 2002. “Intonational rises and dialog acts in the Australian English map task,” Language and Speech 45: 229–253. Fox, A. 1984. German Intonation. Oxford: Clarendon Press.

187

188

MARY GRANTHAM O’BRIEN

Fretheim, T. and R. A. Nilsen. 1989. “Terminal rise and rise-fall tunes in east Norwegian intonation,” Nordic Journal of Linguistics 12: 155–181. Fujisaki, H. 1983. “Dynamic characteristics of voice fundamental frequency in speech and singing.” In P. F. MacNeilage (ed.), New York: Springer: 39–55. Gaº rding, E. 1994. “Prosody in Lund,” Speech Communication 15: 59–67. Gaº rding, E. 1998. “Intonation in Swedish.” In D. Hirst and A. Di Cristo (eds.): 112–130. Gibbon, D. 1998. “Intonation in German.” In D. Hirst and A. Di Cristo (eds.): 78–95. Gooskens, C. 2005. “How well can Norwegians identify their dialects?” Nordic Journal of Linguistics 28: 37–60. Grabe, E. 1998. Comparative Intonational Phonology: English and German. Wageningen: Ponsen and Looijen. Grice, M. and S. Baumann 2002. “Deutsche Intonation und GToBI,” Linguistische Berichte 191: 267–298. Grice, M. and S. Baumann 2007. “An introduction to intonation – functions and models.” In J. Trouvain and U. Gut (eds.), Non-native Prosody: Phonetic Description and Teaching Practice. Berlin: Mouton de Gruyter: 25–51. Grønnum, N. 1994. “Rhythm, duration and pitch in regional variants of Standard Danish,” Acta Linguistica Hafniensia 27: 189–218. Grønnum, N. 1998. “Intonation in Danish.” In D. Hirst and A. Di Cristo (eds.): 131–51. Grønnum, N. 2009. “A Danish phonetically annotated spontaneous speech corpus (DanPASS),” Speech Communication 51: 594–603. Gunlogson, C. 2002. “Declarative questions.” In B. Jackson (ed.), Proceedings of Semantics and Linguistic Theory 12, Ithaca, NY: CLC Publications: 124–143. Gussenhoven, C. and P. van der Vliet 1999. “The phonology of tone and intonation in the Dutch dialect of Venlo,” Journal of Linguistics 35: 99–135. Gussenhoven, C. 2004. The Phonology of Tone and Intonation. Cambridge University Press. Gussenhoven, C. 2010. “Transcription of Dutch intonation.” In S-A. Jun (ed.), Prosodic Typology: The Phonology of Intonation and Phrasing. Oxford Scholarship Online. Gussenhoven, C. and T. Rietveld 2000. “The behavior of H* and L* under variations in pitch range in Dutch rising contours,” Language and Speech 43: 183–203. Gustafson, J., N. Lindberg, and M. Lundeberg 1999. “The August spoken dialogue system.” In Proceedings of Eurospeech 9. Budapest: 1151–1154. Gut, U. 2009. Non-native Speech: A Corpus-based Analysis of Phonological and Phonetic Properties of L2 English and German. Frankfurt: Peter Lang Verlag. Hahn, L. D. 2004. “Primary stress and intelligibility: Research to motivate the teaching of suprasegmentals,” TESOL Quarterly 38: 201–223.

Intonation in Germanic

Hart, J. ’t 1998. “Intonation in Dutch.” In D. Hirst and A. Di Cristo (eds.): 96–111. Heinemann, T. 2010. “The question-response system of Danish,” Journal of Pragmatics 42: 2703–2725. Hirst, D. 1998. “Intonation in British English.” In D. Hirst and A. Di Cristo (eds.): 56–77. Hirst, D. and A. di Cristo (eds.) 1998. Intonation Systems: A Survey of Twenty Languages. Cambridge University Press. House, D. 2005. “Phrase-final rises as a prosodic feature in wh-questions in Swedish human-machine dialogue,” Speech Communication 46: 268–283. Isacˇenko, A. and H-J. Scha¨dlich 1970. A Model of Standard German Intonation. The Hague: Mouton de Gruyter. Jilka, M. 2000. The Contribution of Intonation to the Perception of Foreign Accent: Identifying Intonational Deviation by Means of F0 Generation and Resynthesis. Arbeitspapiere des Instituts fu¨r Maschinelle Sprachchverarbeitung, Vol. 6.3. Dissertation, Universita¨t Stuttgart. Kohler, K. J. 1992. Einfu¨hrung in die Phonetik des Deutschen, 2nd edn. Berlin: Erich Schmidt Verlag. Kohler, K. J. 2004. “Pragmatic and attitudinal meanings of pitch patterns in German syntactically marked questions.” In G. Fant, H. Fujisaki, J. Cao, and Y. Xu (eds.), From Traditional Phonology to Modern Speech Processing. Foreign Language Teaching and Research Press. Beijing: Foreign Language Teaching and Research Press: 205–214. Kristoffersen, G. 2000. The Phonology of Norwegian. Oxford University Press. Kristoffersen, G. 2003. “The tone bearing unit in Swedish and Norwegian tonology.” In H. G. Jacobsen, D. Bleses, T. O. Madsen, and P. Edltem (eds.), Take Danish – for Instance: Linguistic Studies in Honour of Hans Basbøll Presented on the Occasion of his 60th Birthday July 12, 2003. Odense, Denmark: U Press Southern Denmark: 189–197. Ku¨gler, F. 2004. “The phonology and phonetics of nuclear rises in Swabian German.” In P. Gilles and J. Peters (eds.), Regional Variation in Intonation, Tu¨bingen: Max Niemeyer Verlag: 75–98. Ladd, D. R. 1983. “Phonological features of intonational peaks,” Language 59: 721–759. Ladd, D. R. 1996. Intonational Phonology. Cambridge University Press. Leeman, A. and L. Zuberbu¨hler 2010. “Declarative sentence intonation patters in 8 Swiss German dialects,” Proceedings of Interspeech 2010, Makuhari, Japan, September 26–30, 2010, 1768–1771. Levis, J. 2005. “Changing contexts and shifting paradigms in pronunciation teaching,” TESOL Quarterly 39: 369–377. Lickley, R. J., A. Schepman, and D. R. Ladd 2005. “Alignment of ‘phrase accent’ lows in Dutch falling rising questions: Theoretical and methodological questions,” Language and Speech 48: 157–183. McLemore, C. A. 1991. The Pragmatic Interpretation of English Intonation: Sorority Speech. Unpublished Ph.D. dissertation, The University of Texas at Austin.

189

190

MARY GRANTHAM O’BRIEN

Meyer, E. A. 1937. Die Intonation im Schwedischen. Erster Teil: Die Sveamundarten. Stockholm: Fritzes. Myrberg, S. and T. Riad 2015. “The prosodic hierarchy of Swedish,” Nordic Journal of Linguistics 38: 115–147. O’Brien, M. G. 2013. “Investigating second language pronunciation.” In P. Siemund, I. Gogolin, M. E. Schulz, and J. Davydova (eds.), Multilingualism and Language Contact in Urban Areas. Amsterdam: John Benjamins: 39–62. O’Brien, M. G. and U. Gut 2011. “Phonological and phonetic realisation of different types of focus in L2 speech.” In M. Wrembel, M. Kul, and K. Dziubalska-Kolaczyk (eds.), Achievements and Perspectives in SLA of Speech: New Sounds 2010, Vol. 1. Frankfurt am Main: Peter Lang Verlag: 205–216. Peters, J. 2006. Intonation deutscher Regionalsprachen. Berlin and New York: Walter de Gruyter. Pierrehumbert, J. B. 1980. The Phonology and Phonetics of English Intonation. Unpublished Ph.D. dissertation, Massachusetts Institute of Technology. Pu¨rschel, H. 1975. Pause und Kadenz: Interferenzerscheinungen bei der englischen Intonation deutscher Sprecher. Tu¨bingen: Max Niemeyer Verlag. Remijsen, B. and V. J. van Heuven 2003. “On the categorical nature of intonational contrasts: An experiment on boundary tones in Dutch.” In J. van de Weijer, V. J. van Heuven, H. van der Hulst (eds.), The Phonological Spectrum. Volume II: Suprasegmental Structure. Amsterdam: John Benjamins: 225–246. Rietveld, T., J. Haan, L. Heijmans, and C. Gussenhoven 2002. “Explaining attitudinal ratings of Dutch rising contours: Morphological structure vs. the frequency code,” Phonetica 59: 180–194. Siebenhaar, B., M. Forst, and E. Keller 2004. “Prosody of Bernese and Zurich German: What the development of a dialectal speech synthesis system tells us about it.” In P. Gilles and J. Peters (eds.), Regional Intonation in Variation. Tu¨bingen: Max Niemeyer Verlag: 219–238. Sievers, E. 1912. Rhythmisch-melodische Studien. Heidelberg: Carl Winter. Truckenbrodt, H. 2004. “Final lowering in non-final position,” Journal of Phonetics 32: 313–348. Uhmann, S. 1991. Fokusphonologie: Eine Analyse deutscher Intonationskonturen im Rahmen der nicht-linearen Phonologie. Tu¨bingen: Max Niemeyer Verlag. Ulbrich, C. 2004. “A comparative study of declarative intonation in Swiss and German standard varieties.” In P. Gilles and J. Peters (eds.), Regional Intonation in Variation. Tu¨bingen: Max Niemeyer Verlag: 99–122. Ulbrich, C. 2006. “Pitch range is not pitch range.” Speech Prosody 2006 Proceedings. Dresden. www.isca-speech.org/archive/sp2006/papers/ sp06_041.pdf. Ven, M. van de and C. Gussenhoven 2011. “On the timing of the final rise in Dutch falling-rising intonation contours,” Journal of Phonetics 39: 225–236.

Part II

Morphology and Agreement Systems

Chapter 9 Verbal Inflectional Morphology in Germanic David Fertig

9.1 Introduction Although the present-day Germanic languages show a great deal of diversity in their verbal inflectional systems, a number of characteristics are widely shared across the family. Grammatical categories generally relevant to the description of Germanic verbal inflection are: tense (present [P R S ] versus past [P S T ]) (Section 9.2.1); subject agreement for person (1, 2, 3) and number (singular [S G ], plural [P L ]) (Section 9.2.2), mood (indicative [I N D ], subjunctive [S B J V ], imperative [I M P ]) (Section 9.2.3), and voice (active, passive, middle [M I D ]) (Section 9.2.4). Nonfinite forms include an infinitive [I N F ] (which generally serves as the citation form), a present participle [P R E S P ] , and a past participle [P P ] (Section 9.2.7). The most fundamental inflectional-class distinction is between strong verbs, which generally mark the present–past distinction with a root-vowel alternation (ablaut), and weak verbs, which rely mainly on a “dental” suffix (generally -d-, -t-, or -ð-) (Section 9.5). Among the modern languages, synthetic forms reflecting most or all of these categories and distinctions are best represented in Icelandic and Standard German, as shown in Table 9.1, with reconstructed ProtoGermanic endings in the rightmost column (based on Ringe 2006; two endings separated by a slash represent strong versus weak differences). In addition to synthetic forms, all of the languages make use of periphrastic constructions consisting of an auxiliary plus a nonfinite main verb (Section 9.4). These constructions figure prominently in the marking of tense, mood, and voice distinctions in all modern Germanic languages. Some varieties have also developed grammaticalized constructions that mark aspectual distinctions (Section 9.4.4).

194

DAVID FERTIG

Table 9.1 Full paradigms of ‘give’ (strong) and ‘hear’ (weak) in Icelandic and German

weak

PGmc ending(s) (str/wk)

heyra ‘hear’ geben ‘give’

hören ‘hear’

-Vną

gef gefur gefur gefum gefið gefa gefi gefir gefi gefum gefið gefi

heyri heyrir heyrir heyrum heyrið heyra heyri heyrir heyri heyrum heyrið heyri

gebe gibst gibt geben gebt geben gebe gebest gebe (geben) (gebet) (geben)

höre hörst hört hören hört hören (höre) hörest höre (hören) (höret) (hören)

-oˉ -zi/si -di/þi -maz -d/þ -ndi -a-ų -ai-z/s -ai -ai-m -ai-d -ai-n

gaf gafst gaf ga fum ga fuð ga fu gæfi gæfir gæfi gæfum gæfuð gæfu gef(ðu) gefinn gefið gefandi gefast, etc.

heyrði heyrðir heyrði heyrðum heyrðuð heyrðu heyrði heyrðir heyrði heyrðum heyrðuð heyrðu heyr(ðu) heyrður heyrt heyrandi heyrast, etc.

gab gabst gab gaben gabt gaben gäbe gäb[e]st gäbe gäben gäb[e]t gäben gib gegeben —— gebend ——

hörte hörtest hörte hörten hörtet hörten —— —— —— —— —— —— hör(e) gehört —— hörend ——

∅/-ǭ -t/-ēz ∅/-ē -u-m -u-d -u-n -ij-ų (?) -ıˉ -z -ıˉ -ıˉ -m -ıˉ -d -ıˉ -n ∅ -ana-/-da—— -and——

Icelandic strong

weak

gefa ‘give’ 1SG 2SG 3SG 1PL 2PL 3PL 1SG 2SG 3SG 1PL 2PL 3PL 1SG 2SG 3SG 1PL 2PL 3PL 1SG 2SG 3SG 1PL 2PL 3PL 2SG

INF

German strong

PRS IND

SBJV

PST IND

SBJV

IMP PP

supine PRESP MID

9.2 Morphosyntactic Categories of Germanic Verbal Inflection 9.2.1 Tense All of the Germanic languages, modern and older, have a tense distinction between present (or nonpast) and past. Most of the modern languages still mark this opposition ́ primarily with reflexes of inflectional patterns inheŕ ited from Proto-Germanic. In addition, all of the modern languages have ́ a periphrastic “perfect” construction (she has eaten dinner). In most of the present-day languages the perfect contrasts functionally with the synthetic “simple past” (or “preterite”), but in some it has become the primary or even the only way of expressing past tense. Many grammars classify the

Verbal Inflectional Morphology in Germanic

perfect as a tense – or a set of tenses: present perfect, past perfect, future perfect. Quite a few linguists, however, regard the perfect – where it stands in opposition to a simple past – as something other than (merely) a tense (Huddleston 2002: 116, Harbert 2007: 307–315). The form of the perfect construction is discussed in Section 9.4.1. Similarly, grammars often characterize at least one auxiliary + infinitive construction in each language as a “future tense” (e.g., English he will leave tomorrow). For every Germanic language, it can be shown that (1) these constructions have a modal sense (generally involving probability or evidentiality); and (2) “present” forms are commonly used to talk about future events (he leaves/is leaving tomorrow). Many linguists thus argue that no Germanic language has a true future tense, and the “present” is better understood as a “non-past” (Booij 2002: 70, Huddleston 2002: 208–210, Thra´insson 2007: 15–16). The “future” constructions are described in Section 9.4.2.

9.2.2 Subject Agreement In all of the modern (standard) West Germanic languages except Afrikaans and in the insular North Germanic languages (Icelandic and Faroese) – as well as in the older languages of all branches – verbs are inflected for agreement with the subject in person and number (e.g., English 3 S G P R S I N D walks versus non-3 S G walk). The usual number of distinct forms today is everywhere fewer than the six that would be possible with 3 persons x 2 numbers, ranging from five in most cases in Icelandic down to at most two in standard English. The continental North Germanic languages (Swedish, Norwegian, Danish), along with Afrikaans, have lost subject agreement entirely, as have many nonstandard varieties of English. Patterns of syncretism are largely language or branch specific, but one widespread pattern, found everywhere except early North Germanic, is exceptionless identity of the 1 S G and 3 S G P S T I N D – and of the 1 S G and 3 S G P R S I N D of preterite-present verbs (see Section 9.5.4). In all of the modern languages – and in older varieties of West Germanic – this 1 S G =3 S G pattern holds for both tenses of the subjunctive as well, making an inflectional difference between 1 S G and 3 S G characteristic of the P R S I N D . A 2 S G =3 S G pattern in the P R S I N D has been characteristic of North Germanic since Old Norse times, although this has been disrupted in one set of verbs in modern Icelandic and Faroese by the accretion of a final -t or -ð in the 2 S G (Oresˇnik 1980). Modern Dutch has also developed 2 S G =3 S G syncretism in the present tense. Exceptionless identity of 1 P L and 3 P L is a feature of modern standard German and closely related varieties, including Yiddish and Pennsylvania German. Regular sound change from Proto-Germanic resulted in identity of the 3 S G P R S I N D with the 2 P L . This pattern is maintained in Gothic and partially in the older West Germanic languages and in modern German. One of the main defining characteristics of the

195

196

DAVID FERTIG

Ingvaeonic subgroup within West Germanic – which includes English, Frisian, and most dialects of Low German – is a uniform plural (1 P L =2 P L = 3 P L ). Dutch, Faroese, and quite a few dialects of German have joined the uniform plural group in modern times. This is consistent with the typological generalization that languages tend to make more inflectional distinctions in morphosyntactically “unmarked” contexts (e.g., singular) than in the corresponding marked contexts (plural) (Haspelmath 2006: 38). Similarly, we often find fewer agreement distinctions in the past and / or subjunctive than in the present indicative. Dutch, for example, has a three-way distinction in the present but only two distinct forms in the past. Exceptions to the typological generalization, however, are found in Icelandic, which has three distinct forms everywhere in the plural but usually only two in the singular (see Table 9.1), and Gothic, with one fewer distinct forms in the present indicative active than in the subjunctive or in either mood of the past. Gothic is the only attested Germanic language with a three-way (singular– dual–plural) number distinction in verbal agreement. Accordingly, Gothic has eight distinct forms in its subjunctive active paradigms – the maximum possible since there is no third person dual – and seven in the indicative active. Old English, Old Saxon, and Old Norse maintain a three-way number distinction in their personal pronouns but use plural verb forms with dual subjects. In the second person, all of the present-day languages except English have conventions for distinguishing formal from informal address. In languages with subject agreement, this has implications for verbal inflection. Yiddish uses its regular 2 P L pronouns and the corresponding agreement endings for formal address. Icelandic uses special pronouns with 2 P L agreement. Faroese and Frisian are similar, except that the ending is the undifferentiated P L . German is unique in using 3 P L pronouns and agreement. Dutch – like Icelandic, Faroese, and Frisian – has a special pronoun (U) for second person formal but uses it with a S G verb form, which is ambiguous between 2S G and 3 S G for most verbs (Booij 2002: 56–57). Like English, Dutch has completely lost its original Germanic 2 S G pronouns and agreement endings (Middle Dutch du -s; archaic English thou -st) and replaced them with what had been 2 P L forms. Dutch also has a syntactically conditioned alternation in its 2 S G agreement. The ending -t occurs when the verb is not immediately followed by the subject pronoun jij / je. When the pronoun is present, the agreement ending is ∅. Similar alternations occur in the 1 P L in medieval and early modern German (Fertig 2000: 47–49, Axel and Weiß 2011: 38–40) and in both the 1 P L and 2 P L in Old English (Campbell 1959: 296–297) as well as in a number of nonstandard continental West Germanic varieties today, where we also find cases in which the agreement ending on the verb depends on whether it is in a syntactic position typical of main or of subordinate clauses (Weiß 2005). This bears some resemblance to the

Verbal Inflectional Morphology in Germanic

Northern Subject Rule in northern England and Scotland, where verbs with non-3 S G subjects can take the ending -s – like the 3 S G – so long as they are not immediately adjacent to a pronominal subject, whereby it makes no difference whether the pronoun precedes or follows the verb (Pietsch 2005). Some of these “double agreement” patterns clearly originated in phonological interaction between the agreement ending and the subject pronoun, but synchronically the conditioning is generally not purely phonological (Booij 2002: 58). In a number of nonstandard modern dialects, certain subject agreement endings occur not only on the verb, but also on the complementizer in subordinate clauses. See van Koppen, Chapter 14, for discussion and examples.

9.2.3

Mood

9.2.3.1 The Imperative All of the Germanic languages have at least one second person imperative construction, usually characterized by the absence of an overt subject pronoun – although in present-day Icelandic an enclitic pronoun has become the norm in the imperative singular, and in the German formal imperative the subject pronoun (Sie) is obligatory. The use of the bare verb stem for the 2 S G imperative goes back to Proto-Indo-European, and the distinctness of this form in each modern language depends largely on how well other endings have been preserved. In Icelandic, Faroese, and standard German, bare stems of many verbs are still used for no function other than the 2 S G imperative. At the other end of the spectrum is Afrikaans, where the bare stem is used throughout the present-tense system. For all verbs other than be, standard English differs only in the 3 S G . In languages that distinguish between 2 S G and 2 P L imperative, the 2 P L is usually identical to the indicative. This was the case in all of the older Germanic languages, but in modern Faroese and archaic modern Dutch, an old 2 P L form has survived in the imperative that has been lost to analogical leveling in the indicative. Gothic is the only attested Germanic language with third person imperative forms. 9.2.3.2 The Subjunctive The older languages all maintained an inflectional distinction, in both tenses, between indicative and subjunctive – the latter often called “optative” based on its Indo-European origins. The conventional labels “present” and “past subjunctive” are based more on morphological form than on function (see Table 9.1). Today, the subjunctive is most fully intact in Icelandic. In West Germanic, the present subjunctive’s best chances for survival were always in the 3S G , where the indicative has a distinctive consonantal ending (-t, -th, -s, etc.). Standard German

197

198

DAVID FERTIG

still uses 3S G present subjunctive forms. Their main function today is for reported speech, and they are particularly common in journalistic texts, as in (1). (1)

Angela Merkel behauptet sie distanziere sich nicht Angela Merkel claims she distance-3 S G . P R S . S B J V herself not von der Resolution . . . from the resolution . . . “Angela Merkel claims she is not distancing herself from the resolution . . .” (www.wiwo.de/politik/deutschland/14492128.html)

Some speakers of (especially American) English also maintain a 3 S G subjunctive, which distinguishes itself from the indicative by its lack of -s, e.g., it is very important that he arrive on time. In standard German and the relevant varieties of English, present subjunctive forms of the verb ‘to be’ are used for all persons and numbers. Most of the other modern languages – and many varieties of English – retain nothing more than a few traces of old P R S S B J V forms in lexicalized idioms, e.g., Dutch het zij so ‘it may be so’, Swedish Leve konnungen! ‘Long live the king’. The current situation in the past subjunctive is similar. Outside of Icelandic and German, English 1 / 3 S G were (I wish it were Friday) is probably the most robust relic. Swedish has a cognate in vore, along with a handful of other rarely used forms. The other languages retain at most a few uses in fixed idioms. At the other extreme, many southern dialects of German make heavy use of synthetic past subjunctive forms of all verbs. These are dialects that supposedly lost the simple past (see Section 9.3.1), but – unlike Yiddish, Afrikaans, and Pennsylvania German (and southern colloquial varieties that are closer to standard German than to the traditional dialects) – they actually only lost the past indicative while retaining a synthetic past subjunctive. In many places, a weak past suffix -ɒt or -ɒd has been generalized to all verbs, e.g., ı¯ zˇraiwɒd ‘I would write’. In some dialects, this suffix can co-occur with ablaut of the root vowel in strong verbs: ¯ı zˇrı¯wɒd (Wiesinger 1989: 56–61). The limited use of the synthetic past subjunctive in standard German is described in Section 9.4.3 below.

9.2.4 Voice Gothic has synthetic passive forms cognate with mediopassives in other branches of Indo-European. Only a few traces of such forms survive elsewhere in Germanic. North Germanic has developed new synthetic forms, sometimes referred to as mediopassives, from what were originally enclitic reflexive pronouns (see Lee-Schoenfeld, Chapter 21). The resulting suffix – -st in Icelandic, Faroese, and Nynorsk; -s in Bokmaº l, Swedish, and Danish – occurs after any other inflectional endings and can be added to both finite and nonfinite forms. In Icelandic, it behaves in a number of respects more like a derivational than an inflectional affix, making its

Verbal Inflectional Morphology in Germanic

position outside of all inflectional formatives highly unusual (Anderson 1990, Thra´insson 1994: 161–162). In the modern continental North Germanic languages, especially Swedish, -s(t) is used more productively than in Icelandic to form passives and is generally regarded as an inflectional category of the verb (Harbert 2007: 327–329, Alexiadou and Scha¨fer, Chapter 20).

9.2.5 Aspect The absence of any inflectional marking of aspectual distinctions, such as perfective–imperfective, is a defining characteristic that sets Germanic apart from many other branches of Indo-European. Today, modern English and Icelandic stand out among the (standard) Germanic languages for their fully grammaticalized opposition between “simple” (they dance) and “progressive” (they are dancing) aspect. The English and Icelandic progressive constructions, along with less grammaticalized counterparts in other Germanic languages and a few related constructions, are described in Section 9.4.4. Many linguists also regard the perfect (Section 9.4.1) as (partly) an aspect.

9.2.6 Negation Negation has come to be marked inflectionally on auxiliaries in modern (informal) English. That the formative represented orthographically as is an inflectional suffix rather than an enclitic is most apparent in forms like won’t and don’t, whose phonological form is not predictable from the components will + not and do + not (Huddleston 2002: 90–92).

9.2.7 Nonfinite Verb Forms All Germanic languages have nonfinite forms used in verbal constructions with auxiliaries as well as in substantival or adjectival functions. Across the entire family, it is common to speak of an infinitive, a present participle, and a past participle. The term “infinitive” is widely used to refer to: (1) a synthetic form often called the “bare infinitive”, whose functions include complement of a modal auxiliary; (2) a construction, sometimes known as the “prepositional infinitive”, in which the verb form in (1) is preceded by a particle derived historically from an allative preposition (see Wurmbrand and Christopoulos, Chapter 17). The extent to which (1) is distinct from other forms of the verb varies from language to language. Modern English and Afrikaans arguably have no bare-infinitive form but rather a polyfunctional “plain form” (Huddleston 2002: 83). The other modern languages – and all of the older languages – have a synthetic infinitive with a nonnull reflex of the Proto-Germanic suffix *-Vną that makes it distinct from most or all finite present forms.

199

200

DAVID FERTIG

The particle used for the prepositional infinitive is a cognate of English to throughout West Germanic, a cognate of at in North Germanic, and an etymologically mysterious du in Gothic. Modern Frisian has an additional nonfinite form often called the “gerund”, generally formed by adding -n to the infinitive. Cognate forms with the suffix -Vnne (-Vnnia in Old Saxon) were present everywhere in earlier West Germanic. The gerund provided dative and (occasionally) genitive forms to complement the inherently nominative / accusative infinitive. The prepositional infinitives in the modern West Germanic languages actually descend from a to + gerund construction. Yiddish has a special nonfinite form known as the “tautological infinitive”, used for clauseinitial topics and formed by adding -n to a finite form of the verb. With regular verbs, this yields a form identical to the infinitive, but many irregular verbs have distinct tautological infinitives (Jacobs 2005: 212–213). Everywhere except Yiddish and modern English, the present-participle suffix for all verbs has the form -(V)nd(V), inherited from Proto-IndoEuropean. Yiddish has added an additional suffixal element, yielding -endik. In English, the present participle has fallen together with a verbal noun in -Vng(V), resulting in the -ing form that is used in both noun-like and adjectival/participial functions today. The widespread nonstandard pronunciation of -ing with /n/ rather than /ŋ/ reflects the original present participle. Labov (1994: 583–584) has shown that many English speakers today still tend to use /n/ more often in participial function versus /ŋ/ for the verbal noun. In most of the Germanic languages, the present participle plays little or no role in periphrastic verbal constructions, the use of English -ing for progressive aspect being the major exception (Section 9.4.4). The past participle has always been treated as one of the principle parts of the verbal paradigm in every Germanic language. Reflexes of the ProtoGermanic P P suffixes – -Vn(-) for strong verbs; -d(-) or -t(-) for weak – are largely intact in some modern varieties, such as standard German, Yiddish, and written Dutch. In other languages and dialects, one or both of the suffixes has been lost or altered beyond recognition in at least a large subset of verbs. Grammars of the modern North Germanic languages identify an additional, closely related form known as the “supine”, which is used primarily in the perfect construction (Section 9.4.1) and is generally identical or very similar to the (nominative / accusative S G ) neuter form of the past participle.

9.3 Types of Exponents Suffixes are the most heavily used type of exponent in Germanic verbal inflection, marking morphosyntactic distinctions in every category. In the past tense of weak verbs, it is possible in the older and the more conservative languages to identify distinct tense and agreement affixes. In Icelandic heyrðum, for example, the -ð- can be analyzed as an exponent of P S T and the

Verbal Inflectional Morphology in Germanic

-um as an exponent of 1 P L agreement. Otherwise, as in other branches of Indo-European, cumulative exponence is the norm (Matthews 1972). The -t in German gibt ‘gives’, for example, might be regarded as a 3 S G agreement ending, but since it does not occur in the past tense or the subjunctive, it is also an exponent of present and indicative, much like English 3 S G -s. Prefixes play a much more limited role in Germanic verbal inflection, but one important one is the ge- that occurs on most past participles in many West Germanic varieties. This prefix is sometimes analyzed as part of a circumfix, together with the strong or weak P P suffix (Simmler 1998: 85–86). Afrikaans lacks these suffixes, leaving ge- as the only exponent of the P P . The Gothic reduplicant is another important prefixal element (Section 9.5.1). In some Germanic languages, a case can also be made for analyzing the infinitive particle as a prefix. Whereas English to and Swedish att can be separated from the verbal part of the infinitive (to boldly go . . .), this is completely impossible in German, Dutch, and Afrikaans, as it was in Old English (Harbert 2007: 338–339). In colloquial English forms such as wanna, gonna, hafta, by contrast, to has been reanalyzed as part of the auxiliary (Palmer et al. 2002: 1616–1618). Some inflectional affixes display patterns of phonologically predictable allomorphy. Assimilatory /t/–/d/ alternations in the weak P S T /P P suffix occur in several of the modern West and North Germanic languages. Modern Icelandic has a related but more complex pattern of allomorphy in the weak P S T suffix. Standard German breaks up tt and tst sequences through schwa-epenthesis. Many of the other languages instead deal with such sequences through degemination and cluster simplification. In spoken English, the weak past suffix shows a three-way alternation involving both voicing assimilation and epenthesis: /t/–/d/–/ɨd/. The parallel /s/–/z/–/ɨz/ alternation in the 3 S G P R S I N D occurs in both verbal and nominal inflection (noun plurals and possessives). The suffix of the I N F , the 1 / 3 P L and the strong P P in standard German and Yiddish alternates between /ən/ and /n̩ /, depending mainly on the preceding segment. This allomorphy is reflected in Yiddish – but not German – orthography. The same suffix shows more complex patterns of allomorphy in many nonstandard varieties of German. The presence of the ge- P P prefix is prosodically conditioned in German, Yiddish, and Afrikaans. When the main stress is not on the first syllable of the verb, ge- is not added, e.g., German stuˈdieren [I N F ]–stuˈdiert [P P ] ‘study’. In standard Dutch, by contrast, the conditioning is morphological: The prefix is added unless the verb already begins with an unstressed prefix: stuˈderen–gestuˈdeerd, but beˈginnen–beˈgonnen ‘begin’ (Booij 2002: 74). Morphologized root-vowel alternations are the second most important exponent type in Germanic verbal inflection. One set of alternations shared across the entire family – except Afrikaans – goes back to the IndoEuropean phenomenon known as ablaut (Section 9.5.2). Ablaut alternations are often the sole or primary marker of the morphosyntactic

201

202

DAVID FERTIG

distinctions with which they correlate (sing–sang–sung). In West and North Germanic languages, the assimilatory effects triggered by certain vowels and glides in following unstressed syllables – which were subsequently lost in many cases – have given rise to numerous additional root-vowel alternations. One important class of these is commonly referred to as umlaut or mutation. Such alternations often accompany suffixal marking (e.g., tell–told, German trage [1 S G P R S ]–tra¨gt [3 S G P R S I N D ] ‘carry’). In modern German and Icelandic, umlaut alternations serve as primary marker of the past indicative–subjunctive distinction in important subsets of verbs (see Table 9.1). Historical lengthenings and shortenings are also reflected in some modern root-vowel alternations (e.g., keep–kept; German nehme [1 S G P R S ]–nimmt [3 S G P R S I N D ] ‘take’). Morphologized or lexicalized root-final consonant alternations are of less systematic importance but are fairly widespread in irregular verbs. They often go hand-in-hand with root-vowel alternations (e.g., was–were; teach–taught; German ziehe [1 S G P R S ]–zog [1 / 3 S G P S T I N D ] ‘pull’, Swedish faº [I N F ]–fick [P S T ] ‘get’). Consonant alternations occasionally serve as sole exponent of a morphosyntactic distinction, e.g., send–sent. Most Germanic languages have a small number of suppletive verb paradigms (Fertig 1998, Corbett 2007). It is common to distinguish between strong suppletion, where forms of a lexical item bear no apparent resemblance to each other, and weak suppletion, where all the forms have some segments in common, but the inflectional pattern is nevertheless idiosyncratic. In most of the Germanic languages, strong suppletion in verbal inflection is restricted to the verb ‘be’. English and Gothic also have strong suppletion in ‘go’: Modern English go–went; Old English ga¯n–e¯ode; Gothic gaggan–iddja. Weak suppletion is much more widespread (see Section 9.5.3 for a few examples).

9.4 Periphrastic Constructions Throughout their histories, the Germanic languages have relied on a combination of synthetic inflection and periphrastic constructions to mark the grammatical categories of the verb. This division of labor has tended to shift over time in favor of periphrasis. The periphrastic constructions generally consist (minimally) of an auxiliary plus a nonfinite form of a main verb. In colloquial English, a system of grammaticalized clitic forms of auxiliaries has developed, e.g., would/had → ’d, is/has → ’s (Palmer et al. 2002:1614–1616; see also Sections 9.2.6, 9.4.2).

9.4.1 The Perfect Perfect constructions consisting of auxiliary + past participle or supine are common to all of the modern Germanic languages. A cognate of English

Verbal Inflectional Morphology in Germanic

have is the most common auxiliary in every language today. Present-day English and Swedish, and – with the exception of main verb ‘be’ – Afrikaans use have exclusively. The other modern languages sometimes use ‘be’, generally with intransitive predicates that express change of location (e.g., ‘go/ride to/from/through’) or change of state (e.g., ‘become’, ‘freeze’, ‘die’). In Icelandic, Faroese, and to some extent in Nynorsk, ‘have’ combines with the supine, but ‘be’ with an inflected past participle. In English and throughout North Germanic, the general function of the present perfect is to express present relevance of events that occurred or situations that began in the past. In German, Dutch, and Frisian, by contrast, the use of the perfect has shifted and expanded considerably at the expense of the simple past. This development has reached its ultimate conclusion in Yiddish, and for the most part in Afrikaans, Pennsylvania German, and most southern varieties of German, where the synthetic past (indicative) has been lost and replaced by the perfect for (nearly) all verbs (Schirmunski 1962: 489–492). Wherever simple past forms of ‘have’ – and, where relevant, ‘be’ – still exist, a past perfect (pluperfect) can be formed with a past-tense auxiliary + past participle, as in English I had seen the film. Especially where the simple past forms of the auxiliaries have been lost, a threepart construction is possible with present-tense auxiliary + past participle of auxiliary + past participle of main verb, e.g., German ich habe den Film gesehen gehabt.

9.4.2 The Future Most languages use a modal verb + bare infinitive for their “future” constructions. Cognates of shall are the most common auxiliary (Dutch, Afrikaans, Frisian, Swedish). Norwegian, Danish, and Faroese use both shall and will, with a semantic difference. Some varieties of English favor shall in the first person (S G and P L ) versus will in the second and third, a pattern well-known from prescriptive grammars, but many speakers – especially outside of England – use only will (’ll, won’t). The future auxiliary munu is unique to Icelandic, as is vel to Yiddish. German uses werden ‘become’, the same auxiliary that it uses with the past participle for the passive (Section 9.4.5). Motion-verb auxiliaries used in secondary future constructions include cognates of go in Dutch and Afrikaans and of come in Swedish and Norwegian. The English be going to construction is unique in that it requires the progressive form of the auxiliary.

9.4.3 The Conditional All of the modern languages have an auxiliary + bare-infinitive construction (in Yiddish more often auxiliary + past participle), often called the “conditional”, whose functions overlap with those of the past subjunctive (see

203

204

DAVID FERTIG

Section 9.2.3.2). The auxiliary is the past (subjunctive) form of the future auxiliary in each language (compare Section 9.4.2): English would – and in some varieties should – Icelandic mund-, Swedish skulle, Dutch zou; Afrikaans sou, Frisian soe, Yiddish volt, standard German wu¨rde. In most of the presentday languages, past subjunctive functions are split more or less systematically between the periphrastic conditional and the synthetic past tense. The English pattern with P S T in the protasis of (nonpast) counterfactual conditionals and the would/should construction in the apodosis illustrates this division of labor: if I knew the answer, I would be happy to tell you. Earlier stages of English and other Germanic languages could use past subjunctive in both clauses of counterfactual conditionals, as can modern Icelandic. In languages such as Yiddish and Afrikaans, which have lost the synthetic P S T , the periphrastic conditional is used in all contexts. For most verbs, modern standard colloquial German patterns with Yiddish and Afrikaans in this respect. German reference grammars often show past subjunctive paradigms for all verbs, but this is misleading. As in the other modern languages, the past subjunctive forms of regular weak verbs are identical to the past indicative. Unlike most of the other languages, however, German generally does not use past forms that are not unambiguously subjunctive for any subjunctive functions. In fact, use of the past subjunctive is only common in present-day standard German for a handful of high-frequency verbs, including sein ‘be’, haben ‘have’, and the modals. For these verbs, especially sein and haben, the synthetic past subjunctive forms (wa¨re, ha¨tte) are used almost exclusively in all contexts, including those where other modern Germanic languages would use a conditional construction (2a). Conversely, for the vast majority of German verbs, the wu¨rde + infinitive construction is normally used everywhere (2b). (2)

a. Wenn die Ku¨che gro¨ßer wa¨re, ha¨tten wir Platz fu¨r alle. “If the kitchen were larger, we would have room for everyone.” b. Wir wu¨rden uns o¨fter sehen, wenn wir in der selben Stadt wohnen wu¨rden. “We would see each other more often if we lived in the same town.”

9.4.4 The Progressive and Other Aspectual Constructions The English progressive construction combines auxiliary be with the -ing form of the main verb. The Icelandic progressive uses vera ‘be’ with an aðinfinitive. The grammaticalization of the English and Icelandic progressives has greatly reduced the use of the simple present, robbing it of the central function of referring to events ongoing at the moment of speaking. For most verbs, the simple present is now used mainly to express habitual aspect. The consequences of the rise of the progressive are less drastic in

Verbal Inflectional Morphology in Germanic

the past tense, where the simple past remains the normal way of expressing perfective aspect. English has an additional construction specifically for past habitual: he used to play basketball. Yiddish has a functionally similar construction with the auxiliary flegn. The other languages have ways of expressing meanings similar to the English and Icelandic progressives, but at least in the standard languages their use is never obligatory. Several languages use a locative preposition (+ definite article) + infinitive to express a progressive-like meaning, e.g., German ich bin beim / am schreiben ‘I’m writing’ [literally ‘I am at the writeI N F ’]. This construction is limited in standard German in that the infinitive cannot take any (unincorporated) arguments. Some colloquial varieties, however, have a more flexible construction similar to Dutch: Hans is zijn moeder aan het opbellen ‘Hans is phoning his mother’ (Booij 2002: 206, Krause 2002). Another widespread pattern uses motion and / or posture verbs, either as auxiliaries with an infinitive, as in Dutch hij zit te zeuren ‘he is nagging’ [literally ‘he sits to nag’], or in a coordinate structure, as in Nynorsk han dreiv og las ‘he was reading’ [literally: ‘he moved around and read’] (Askedal 1994: 246), Afrikaans Ek sit en lees ‘I’m reading’ [literally: ‘I sit and read’] (Donaldson 1993: 220).

9.4.5 Passive Constructions All of the modern Germanic languages have at least one passive construction consisting of auxiliary + past participle (or, rarely, supine). In North Germanic, these constructions overlap in function with the -s(t) suffix described in Section 9.2.4. The most common passive auxiliaries are ‘be’ (Icelandic, Swedish, English), ‘become’ (Faroese, Norwegian, Swedish, Danish, German, Dutch, Afrikaans, Yiddish, Frisian), and ‘get’ (Norwegian, colloquial English). Many languages have more than one passive construction (see Alexiadou and Scha¨fer, Chapter 20).

9.5 Inflectional Classes The older Germanic languages all had several relatively well-defined inflectional classes of verbs. Inflectional-class structure has broken down to some degree in most of the modern languages. Grammars of modern English, German, and Dutch typically identify one “regular” pattern, which the vast majority of verbs follow, and a number of “irregular” patterns, most of which apply to at most a handful of verbs.

9.5.1 Strong Versus Weak Verbs The distinction between ablauting strong verbs (sing–sang–sung) and suffixing weak verbs (walk–walked) is one of the main defining characteristics of

205

206

DAVID FERTIG

the Germanic family and is maintained to some degree in all of the modern languages except Afrikaans. Weak verbs have a (reflex of a) “dental” suffix in the past tense and participle; strong verbs instead have a root-vowel alternation as the primary marker of the P R S –P S T opposition and – where it has survived – a -(V)n suffix in the P P . The rightmost column in Table 9.1 shows a number of places where strong and weak verbs also had different agreement endings in Proto-Germanic. The most important of these is the P S T I N D S G , where strong verbs had no ending in the 1/3S G and -t in the 2S G , whereas weak verbs had 1 S G -o¯˛ , 2 S G -e¯z, 3 S G -e¯. In spite of analogical and sound changes that have altered the details in various ways, the modern German and Icelandic forms in Table 9.1 still reflect these strong–weak differences. Proto-Germanic also had a class of verbs that formed its past tense through reduplication. This pattern is well-represented in Gothic, but only a few traces survive in the (older) North and West Germanic languages. In their inflectional endings, the reduplicating verbs pattern with the strong verbs. The traditional defining criterion for the strong versus weak distinction – whether the past tense is marked by ablaut or by a dental suffix – breaks down somewhat in some of the modern languages. On the weak side, the dental suffix has been lost in one of the major classes of verbs in Frisian and some varieties of continental North Germanic (Section 9.5.3.1), and on the strong side Yiddish and those varieties of German that have lost the simple past are left with quite a few strong verbs that have no ablaut alternation since some classes have always had the same ablaut grade in the present and the past participle (Section 9.5.2), e.g., Yiddish forn [I N F ]–geforn [P P ] ‘travel’; this leaves the P P suffix -(e)n versus -t as the only remaining consistent difference between strong and weak verbs in these varieties.

9.5.2 Strong-Verb Ablaut Classes In grammars of the older languages and of modern Icelandic, it is customary to characterize strong verb classes in terms of the ablaut grade of four principle parts: (1) the infinitive, whose stem serves as the base for the entire present system; (2) the 1 / 3 S G P S T I N D ; (3) the 1 P L P S T I N D , which has the same ablaut grade as all P L P S T I N D and P S T S B J V forms; and (4) the past participle. The majority of Germanic strong verbs have an ablaut pattern that can ultimately be traced back to a proto-Indo-European e–o–∅–∅ alternation, which becomes e–a–∅–∅ or e–a–u–u in Proto-Germanic. Two other patterns can be regarded as variations on this one: e–o–e¯–∅ (PGmc e– a– e¯1–∅) and e–o–e¯–e (PGmc e–a–e¯1–e). The remaining patterns are all twoway alternations, with one vowel in principle parts (1) and (4) and the other in (2) and (3). The most important of these is a Germanic a–o¯ alternation (Class VI). Verbs following the e–o–∅–∅ pattern generally have roots ending in -VRC, where R stands for any sonorant (j, w, l, r, m, or n). These verbs are

Verbal Inflectional Morphology in Germanic

traditionally divided into three classes depending on which sonorant they contain: j (Class I); w (Class II); or one of the liquids or nasals (Class III). The glides, j and w, form diphthongs with preceding e and a, and in the ∅-grade their vocalic counterparts – i and u – become the syllable nucleus. The liquids and nasals do not interact with preceding ablauting vowels; in the ∅-grade, where some other Indo-European languages have syllabic sonorants, Germanic inserts a u, yielding ur, um, etc. The ablaut patterns of the six traditional classes of Germanic strong verbs are shown in Table 9.2. The remaining strong verbs are usually grouped together in a Class VII, but they do not really constitute a coherent class comparable to I-VI. These are the verbs with reduplication in Gothic. In West and North Germanic, they all follow a (1)=(4)≠(2)=(3) pattern like Class VI. The ablaut distinction between (2) and (3) in Classes I–V has been largely leveled in all of the modern (standard) languages except Icelandic and Faroese. English was–were is an isolated relic. Dutch has a more systematic holdover in verbs from Classes IV and V, where the P S T S G has short a and the P S T P L long aː (the regular reflex of PGmc e¯1). This ablaut alternation probably owes its survival to reanalysis as open-syllable lengthening (Fertig 2005). There is considerable variation – across languages and within individual languages – as to which vowel has won out in the leveling of the (2)–(3) distinction. In English sing–sang–sung (Class III, with e > i raising before a nasal), the a of (2) has prevailed, whereas in spin–spun–spun, we have the u of (3). There is variation and uncertainty among English speakers for several verbs of this class, e.g., shrank~shrunk. The tidy picture in Table 9.2 is a reconstruction that does not correspond to any attested language. The older languages already had subclasses and exceptions, mostly attributable to conditioned sound changes. The modern languages show further splintering of some classes and consolidation of others. Generally, the (sub)classes that were relatively large and homogeneous in the older languages have remained robust, while the small classes have gotten even smaller. Two classes that have fared well in many of the modern languages are Class I and the subset of Class III with stemfinal nasal + consonant. Class I is represented in modern English by about 12 verbs, 6 of which have vowels reflecting ablaut grades (1), (2) and (4) (drive–drove–driven), while the others have grades (1), (3) and (4) (bite–bit– Table 9.2 Ablaut patterns of Germanic strong verbs Class

(1) I N F

(2) 1 / 3 S G

I II III IV V VI

-ei-eu-e-e-e-a-

-ai-au-a-a-a-oˉ -

PST IND

(3) 1 P L -i-u-u-ē1-ē1-oˉ -

PST IND

(4) P P -i-u-u-u-e-a-

207

208

DAVID FERTIG

bitten) or use the vowel of (2) for both the P S T and the P P (shine–shone). Class I has done even better in Swedish (28 members), German (39), and Dutch (51). The subset of Class III with e > i/____N is the largest of all strong classes in modern English. Between the sing–sang–sung and the spin–spun– spun patterns, present-day English counts 21 verbs. Modern German has 19 that all follow the i–a–u pattern, Swedish 22, and the corresponding class in Dutch 25. Class II has remained strong in several modern languages. Its two largest subclasses have 15 and 11 members in Swedish, 22 and 3 in German, and 23 and 12 in Dutch. In English, by contrast, this class has nearly disappeared. Of the handful of former members that have not been completely lost or regularized, no two follow the same pattern: freeze– froze–frozen; choose–chose–chosen; fly–flew–flown; shoot–shot; lose–lost; creep– crept. Once we move beyond these classes, the numbers fall off sharply. In modern Swedish, for example, no other class has more than four members, and most have only one or two. Overall, the number of basic strong verbs in English has been reduced from more than 290 in Old English to about 70 today. Most of the lost verbs – about 120 – have simply disappeared from the language; about half this many have undergone straightforward regularization. The attrition has been much less extreme in some of the other languages (Harbert 2007: 277). Dutch still has about 180 basic strong verbs (Booij 2002: 60–64), German about 170 (Drosdowski 1984: 127), and Swedish more than 100 (Holmes and Hinchliffe 1994: 258–263). Almost all of the regularizations in English occurred between the fourteenth and sixteenth centuries. Outside of this period, the small number of strong→weak changes is roughly matched by movements in the opposite direction, e.g., dig, ring, stick, string, strive, wear. Examples of weak→strong changes in other languages include German weisen–wies–gewiesen ‘show’; Dutch schenken–schonk–geschonken ‘give (as a gift)’; Yiddish meldn–gemoldn ‘announce’.

9.5.3 Weak Verbs The basic form of the weak past suffix is -t- in German and -d(-) in most of the other modern languages, with phonologically conditioned allomorphy as described in Section 9.2 above. Gothic has a morphosyntactically conditioned alternation between two forms of the P S T suffix, with a distribution parallel to that described above for principle parts (2) and (3) of strong verbs (see Table 9.2): -d- in the P S T I N D S G versus reduplicated -de¯d- in the remaining finite past forms. No traces of this alternation are present in any North or West Germanic language. In modern English the weak P P is systematically identical in form to the weak P S T . This applies both to regular verbs and to those following any irregular weak pattern (bring, creep, hit, send, have, etc.). This identity also holds for the largest class of weak verbs in Frisian and Norwegian, e.g.,

Verbal Inflectional Morphology in Germanic

Frisian helle [1 / 3 S G P S T ]–helle [P P ] ‘fetched’. Elsewhere, the relationship between the weak P S T and P P is slightly more complicated. Most Germanic languages have something, at least a -ə, following the -t- or -d(-) suffix in weak finite past forms. The only thing that can follow the dental suffix segment in a weak P P , by contrast, is an adjectival agreement ending. The “uninflected” form of the weak P P can generally be derived from the basic P S T form by removing the final (/ə/) and in German and Dutch prefixing ge- where conditions allow, e.g., German machte [1 / 3 S G P S T ]– gemacht [P P ] ‘made’, Swedish kallade [P S T ]–kallad [P P ] ‘called’. This systematic relationship between the weak P S T and P P forms generally holds for all classes of weak verbs, including most irregulars, e.g., German bringen– brachte–gebracht ‘bring’, Swedish go¨ra–gjorde–gjord ‘make, do’ but does break down in a handful of cases, e.g., German haben–hatte–gehabt ‘have’, Swedish sa¨ga–sa(de)–sagd ‘say’. The uninflected form of the P P plays an important role in West Germanic as it is used in all periphrastic verbal constructions, where North Germanic more often uses the supine. Several languages have a small set of verbs with weak finite P S T forms but a strong P P . Some of these are originally strong verbs that have undergone regularization in the P S T only, e.g., German backen [I N F ]–backte [1 / 3 S G P S T I N D ]–gebacken [P P ] ‘bake’, Dutch wreken–wrekte–gewroken ‘avenge’. A few result from conflation of forms of near-synonymous strong-weak pairs of verbs that had long co-existed, e.g., Dutch lachen–lachte–gelachen ‘laugh’. There are also cases where (classes of) originally weak verbs have adopted the strong P P suffix, e.g., Icelandic telja [I N F ]–taldi[1 / 3 S G P S T I N D ]–talinn [P P ] ‘believe’. Many Frisian speakers systematically add -en to the participles of otherwise regular weak verbs with stems ending in t or d (Tiersma 1985: 70). Verbs with strong finite past forms but a weak participle are less common but do occur, e.g., Dutch vragen–vroeg–gevraagd ‘ask’.

9.5.3.1 Weak Verb Classes All of the modern North Germanic languages, in addition to Frisian, have distinct classes of weak verbs. The other modern languages do not have weak verb classes in the same sense, although the irregular weak verbs of English can be grouped into several small classes. In standard German, Dutch, and Yiddish, by contrast, all but a handful of weak verbs follow exactly the same pattern, aside from phonologically predictable differences. The closest thing to a class in these languages is the small set of German verbs with the P R S –P S T root-vowel alternation known as Ru¨ckumlaut, as in kennen–kannte–gekannt ‘know’. In modern North Germanic, there is a general distinction between one large, productive class of weak verbs with a stem-final “thematic” vowel, -a or -e depending on the language, and two or three smaller classes which lack this vowel. In Icelandic, Faroese, and Swedish, the presence of the thematic vowel is apparent in some P R S forms; it appears in all North Germanic languages in the P S T and the P P /supine, generally occurring between the

209

210

DAVID FERTIG

root and the dental suffix (e.g., Swedish kalla [inf]–kallade [pst]–kallad [pp] ‘call’) and thus distinguishing this class from the other weak classes where the suffix is attached directly to the root (Swedish fylla–fyllde–fylld ‘fill’). In some continental North Germanic varieties, the dental suffix itself has been lost in the thematic-vowel class, as it has in modern Frisian. Frisian and the North Germanic languages also have substantial numbers of irregular weak verbs that do not fit neatly into any of the larger classes. Present-day English has several dozen irregular weak verbs that can be grouped into about eight classes. The largest of these, with about 20 members, is the no-change class – where the basic P R S , P S T , and P P forms are identical, as in hit. All of these verbs have stems ending in t or d. Other classes include: vowel shortening with phonologically regular -t: creep– crept (9 members); with phonologically irregular -t: mean–meant (8); vowel shortening with no suffix (stem-final t or d): feed–fed (8); stem-final d→t: send–sent (6); Ru¨ckumlaut (only): sell–sold (2). Idiosyncratic (weakly suppletive) weak verbs include: have–has–had, make–made, say–says–said. One weakly suppletive verb that is difficult to categorize is do–does–did–done. Historically the di- in did (and its West Germanic cognates) is a reduplicative prefix. Formerly regular verbs and loanwords that have joined one of the irregular weak classes in modern English include kneel, cost, hurt, quit, bet, and split. Among verbs that have varied between regular and irregular weak conjugation for centuries, some – such as burned~burnt – have shown an increasing preference for regular forms (burned) over recent generations, whereas for others, such as leaped~leapt, the irregular form has been gaining ground (Davies 2010). English also shows remarkable robustness in one small group of irregular weak verbs that goes back to Proto-Germanic: bring–brought; buy–bought; seek–sought (beseech–besought); teach–taught; think–thought. This group has even attracted a new member in the loanword catch. Only Gothic can match present-day English for the number of verbs in this class.

9.5.4 Preterite-Present Verbs The small class of verbs traditionally referred to as preterite-presents, whose finite present forms look more or less like the past-tense forms of ordinary strong verbs, are another characteristic feature of Germanic. The preterite-presents sometimes preserve older properties of the strong P S T pattern that have otherwise been lost in the language, such as the S G –P L ablaut alternation in German darf [1 / 3 S G I N D ]–du¨rfen [1 / 3 P L I N D ] ‘be allowed to’, Dutch kan–kunnen ‘can’, etc. The membership of this class overlaps greatly with the semantically/syntactically defined class of modal auxiliaries: can, may, shall, must, etc., and this association has strengthened over the centuries as verbs are attracted to or repelled from the class; will, for example, was not originally a preterite-present but has

Verbal Inflectional Morphology in Germanic

taken on the class’s morphological characteristics in several Germanic languages. Conversely, originally preterite-present verbs such as English owe and German taugen ‘be suitable’ and go¨nnen ‘not begrudge’ are now conjugated as regular weak verbs. Variation between preterite-present and weak inflection can be found both in original preterite-presents such as English dare, Frisian doare ‘dare’, doge ‘be good’, and in weak verbs that have been partially attracted into the preterite-present class, such as English need and German brauchen ‘need’ (Wurzel 1989: 145). As mentioned in Section 9.2.2 above, the preterite-presents show 1 S G = 3 S G syncretism in all Germanic languages – older and modern – which is otherwise foreign to the present indicative in every language that preserves any subject agreement inflection at all. In the modern continental North Germanic languages, the class is distinguished by the absence of the present indicative -(V)r ending. The preterite-presents formed their P S T in Proto-Germanic by attaching the weak dental suffix directly to the root. The suffix is still present in many cases in the modern languages, e.g., English should, German sollte, but has been lost, for example, in cognate forms such as Swedish skulle, Dutch zou, Frisian soe. In a number of cases, the P S T (subjunctive) of modals has undergone some degree of lexicalization, so that it is no longer clear whether speakers regard pairs such as English may–might or German mo¨gen–mo¨chte as forms of the same lexical item. The English modals have a number of other distinctive properties: They lack participle forms, for example, and cannot be used in any context that calls for a nonfinite form (*I’m sorry to must go). These reflect more recent, English-specific developments. In other Germanic languages, the inflection and syntactic behavior of preterite-presents is more verb-like (Harbert 2007: 287–292).

9.6 Derived Verbs Prefixed verbs such as become, forget, withdraw and compound verbs such as babysit generally inherit the inflectional properties of the basic verbs from which they are derived. The number of lexicalized prefixed verbs is much larger in some languages, such as German, than it is in English, although if we include productive prefixation with mis-, out-, over-, under-, re-, un-, etc., this phenomenon is quite important in English as well. To account for the inflectional properties of such verbs, some linguists propose that the “head” of a morphologically complex word bears and determines its inflection (Booij 2002: 65). In Germanic languages, the head is generally the rightmost morpheme in a lexical stem – be that a derivational suffix or a root. In the modern West Germanic languages, the primary productive native way of creating new verbs from other parts of speech involves “conversion” or “zero-derivation”, as in English to fish, German fischen, Dutch vissen. Verbs formed in this way are almost always inflectionally regular, at least initially (Pinker 1999: 157–174, Booij, Chapter 11). In

211

212

DAVID FERTIG

North Germanic languages, productive derivation of new verbs is often understood to involve suffixation of the “thematic vowel” -a or -e (Andersson 1994: 278, Thra´insson 1994: 142, Allan et al. 1995: 540), although the presence of this suffix is not apparent in all forms in the paradigm, reminding us of the fine line between affixal derivation and conversion.

References Allan, R., P. Holmes, and T. Lundskær-Nielsen 1995. Danish: A Comprehensive Grammar. London: Routledge. Anderson, S. R. 1990. “The grammar of Icelandic verbs in -st.” In J. Maling and A. Zaenen (eds.), Modern Icelandic Syntax, New York: Academic Press: 235–273. Andersson, E. 1994. “Swedish.” In Ko¨nig and van der Auwera (eds.): 271–312. Askedal, J. O. 1994. “Norwegian.” In Ko¨nig and van der Auwera (eds.): 219–270. Axel, K. and Weiß, H. 2011. “Pro-drop in the history of German.” In M. Wratil and P. Gallmann (eds.), Null Pronouns, Berlin: Mouton de Gruyter: 21–52. Booij, G. 2002. The Morphology of Dutch. Oxford University Press. Campbell, A. 1959. Old English Grammar. Oxford: Clarendon. Corbett, G. 2007. “Canonical typology, suppletion, and possible words,” Language 83: 8–42. Davies, M. 2010. The Corpus of Historical American English (COHA): 400 million words, 1810–2009. Available online at http://corpus.byu.edu/coha/. Donaldson, B. C. 1993. A Grammar of Afrikaans. Berlin: Mouton de Gruyter. Drosdowski, G. 1984. Duden Grammatik der deutschen Gegenwartssprache, 4th edn. Mannheim: Bibliographisches Institut. Fertig, D. 1998. “Suppletion, Natural Morphology, and Diagrammaticity,” Linguistics 36: 1065–91. Fertig, D. 2000. Morphological Change Up Close. Tu¨bingen: Max Niemeyer Verlag. Fertig, D. 2005. “Review of The Morphology of Dutch by Geert Booij,” Journal of Germanic Linguistics 17: 141–148. Harbert, W. 2007. The Germanic Languages. Cambridge University Press Haspelmath, M. 2006. “Against markedness (and what to replace it with),” Journal of Linguistics 42: 25–70. Holmes, P. and I. Hinchliffe 1994. Swedish: A Comprehensive Grammar. London: Routledge. Huddleston, R. 2002. “The verb.” In Huddleston and Pullum: 71–212. Huddleston, R. and. G. K. Pullum 2002. The Cambridge Grammar of the English Language. Cambridge University Press.

Verbal Inflectional Morphology in Germanic

Jacobs, N. 2005. Yiddish: A Linguistic Introduction. Cambridge University Press. Ko¨nig, E. and J. van der Auwera (eds.) 1994. The Germanic Languages. London: Routledge. Krause, O. 2002. Progressiv im Deutschen. Tu¨bingen: Max Niemeyer Verlag. Labov, W. 1994. Principles of Linguistic Change, Vol. I: Internal Factors. Cambridge, MA: Blackwell. Matthews, P. H. 1972. Inflectional Morphology: A Theoretical Study Based on Aspects of Latin Verb Conjugation. Cambridge University Press. Oresˇnik, J. 1980. “On the dental accretion in certain 2nd p. sg. verbal forms of Icelandic, Faroese, and the old West Germanic languages,” I´slenskt ma´l 2: 195–211. [Reprinted in Oresˇnik 1985. Studies in the Phonology and Morphology of Modern Icelandic, ed. by M. Pe´tursson. Hamburg: Helmut Buske: 191–211.] Palmer, F., R. Huddleston and G. K. Pullum 2002. “Inflectional morphology and related matters.” In Huddleston and Pullum: 1565–1619. Pietsch, L. 2005. Variable Grammars: Verbal Agreement in Northern Dialects of English. Tu¨bingen: Max Niemeyer Verlag. Pinker, S. 1999. Words and Rules. New York: HarperCollins. Ringe, D. 2006. From Proto-Indo-European to Proto-Germanic, A Linguistic History of English, Vol. I. Oxford University Press. Schirmunski, V. M. 1962. Deutsche Mundartkunde. Berlin: Akademie-Verlag. Simmler, F. 1998. Morphologie des Deutschen. Berlin: Weidler. Thra´insson, H. 1994. “Icelandic.” In Ko¨nig and van der Auwera (eds.): 142–189. Thra´insson, H. 2007. The Syntax of Icelandic. Cambridge University Press. Tiersma, P. M. 1985. Frisian Reference Grammar. Dordrecht: Foris. Weiß, H. 2005. “Inflected complementizers in continental West Germanic dialects,” Zeitschrift fu¨r Dialektologie und Linguistik 72: 148–166. Wiesinger, P. 1989. Die Flexionsmorphologie des Verbums im Bairischen. Vienna: ¨ sterreichische Akademie der Wissenschaften. O Wurzel, W. U. 1989. Inflectional Morphology and Naturalness. Boston: Kluwer.

213

Chapter 10 Inflectional Morphology Nouns Damaris Nübling

10.1 Introduction Today there are more than a dozen Germanic languages which derive from the same source, Proto-Germanic (Proto-GMC). While Proto-GMC still had a rather complex nominal inflectional system, the modern languages simplified it to different degrees ranging from rather conservative Icelandic with a broad variety of declension classes, four cases, and three genders to English with only one kind of declension, without cases (except the possessive ’s, a clitic which attaches to the last element of the possessor phrase) and without nominal gender (see Ku¨rschner, Chapter 12). The only surviving category is number consisting of singular and plural (the dual had already disappeared in Proto-GMC). Therefore, in many languages the plural allomorphs are the last relics of the former declension classes. Due to initial stress, all GMC languages underwent rather radical reductions of unstressed word parts, above all the weakening of unstressed vowels to [ǝ] (except for Icelandic and Faroese) or even their loss. From a bird’s eye view the following phenomena developed in different ways and therefore deserve special attention: • Today, some languages show many declension classes (Icelandic), others few (Danish) or even no classes (English). This is connected to the amount of plural allomorphy. • Some languages preserved all three genders (Icelandic, Faroese, German, Luxembourgish), others only two (Swedish, Danish, Dutch, Frisian), and some completely lost this category (English, Afrikaans). • In some languages, gender is crucial for noun inflection, in others it is not. Distributional criteria for declension can be found on different linguistic levels (Neef 2000a, 2000b): (a) Phonology, i.e., a plural allomorph may depend on the stem final sound, the number of syllables or

Inflectional Morphology













rhythmic principles (stress position); (b) Morphology, e.g., derivational affixes; (c) Semantics, i.e., animacy can be relevant for inflection; (d) Gender (located between grammar and lexicon); (e) Lexicon, i.e., specific nouns. In sum, these principles can be combined which leads to rather complex assignment systems. Some languages only use concatenative techniques (affixes), others nonconcatenative ones (modification), where plural (or case) expression affects the lexical part. Some languages extensively use umlaut and even grammaticalized it to indicate plural (German, Luxembourgish); others eliminated it partially (Swedish, Danish) or completely (English, Dutch). Some languages still express case and number in a fused portmanteau unit (Icelandic, Faroese); others separate these categories (most GMC languages). Some languages show (older) stem-based inflection (Icelandic, Faroese), others base-form inflection (the West-GMC languages), and others have both (Swedish). Some languages use zero plurals, especially in the neuter (North-GMC languages), others do not (most of the West-GMC languages); here, number is expressed by the whole noun phrase. In some languages, the inflectional endings caused modifications of the (lexical) stem (Icelandic, German, Luxembourgish), in others, it is the stem which influences the shape of the inflectional endings (Dutch, Frisian, English); thus the direction of determination differs.

In the following, we start with the common source, Proto-GMC (Section 10.2). Then, the majority of the GMC languages will be sketched briefly starting with some North-GMC (Scandinavian) languages (Section 10.3) and proceeding with the most West-GMC ones (Section 10.4). For lack of space, Faroese (which resembles Icelandic), Norwegian (which resembles Danish and Swedish), Yiddish, and Afrikaans, as well as all dialects, cannot be addressed. As proper names often develop differently, they have to be ignored in this chapter as well a foreign words. Section 10.5 presents the most important findings and draws some conclusions.

10.2 The Proto-Germanic Noun System The GMC noun as it is supposed to have existed in the millennium BC showed stem-based inflection, i.e., the lexeme itself never occurred on its own (see Table 10.1). The noun had a tripartite structure: lexical root + primary suffix + grammatical ending: *fisk-a-z ‘fish’ + a-class + masc.nom. sg. suffix. Only athematic nouns had no primary suffix: mann-Ø-z (m.) ‘man’ or *burg-Ø-s (f.) ‘castle, town’. The primary suffix developed out of IE derivational morphology and determined the declension class of the noun. It is

215

+

+

+

1.2 oˉ-class (+ joˉ /woˉ) (IE ā -cl.)

1.3 i-class (IE i-cl.)

1.4 u-class (IE u-cl.)

+

+

(+)



+

+



+



2.2 r-stems (IE ter-stems + ablaut)

2.3 nd-stems (IE -nt-cl.)

2.4 athematic stems (as well as in IE)

2.5 iz/az-stems (IE s- or es/os-class)

+

+

+



+

m.

2.1 n-stems “weak inflection” (as well as in IE)

2. consonantal stems



f.

gender

1.1 a-class (+ ja/wa) (IE o-cl.)

1. vocalic stems “strong inflection”

declension class

+

(+)



-

+

(+)

(+)



+

n.

smaller class (almost no neuter nouns remain); merger with other classes: m. with a-class, f. with oˉ-class f. pl. has already changed to other classes; small class with many losses; already disappearing in GMC

* dē ðiz (f.) – *ðedı̄ z ‘action’ *gastiz (m.) – *gastı̄ z ‘guest’

*handuz (f.) ‘hand’ – (other class) *sunuz (m.) – *suni(we)z ‘son’ *fehu (n.) ‘cattle’ – (no pl.)

only neuter nouns; very small class denoting predominantly animals

small class; in GMC almost only feminine nouns

*burgs (f.) – *burgiz ‘castle’ *mann-z (m.) – *manniz ‘man’

*lambiz (n.) – *lambizoˉ ‘lamb’

originally participles; small class

small class with kinship terms; unproductive

*moˉðer (f.) – *moˉð(e)riz ‘mother’ *broˉþar (m.) – *broˉþ(e)riz ‘brother’

*frijoˉnd (m.) – *frijoˉnd(e)z ‘friend’

large class; highly productive; later many syncretisms in Old GMC languages

*tungo (f.) – *tungon(e)z ‘tongue’ *hanoˉn (m.) – *hanan(e)z ‘cock’ *hertoˉ (n) (n.) – *hertoˉno ‘heart’

consonantal primary suffix or ending on consonant (VC)

large, productive class; (joˉ-stems with different endings)

large, productive class; (ja-/wa-stems with different endings)

vocalic primary suffix in IE + Germ.

assignment principles / remarks

*geboˉ (f.) – *geboˉz ‘gift’

*wulfaz (m.) – *wulfoˉz ‘wolf’ *wurðan (n.) – *wurðo ‘word’

examples nom. sg. – nom. pl.

Table 10.1 Declension class system in Proto-Germanic (based on Krahe 1969 and Ramat 1981)

Inflectional Morphology

assumed that the IE declension classes were semantically motivated which became obsolete in GMC. This IE system can be best observed when looking at the GMC r-stems (see class no. 2.2 in Table 10.1) going back to IE ter-stems which always formed kinship terms (‘sister, mother, father’; today they do not constitute a declension class anymore). Other classes seem to have refunctionalized primary suffixes as the neuter nouns of the GMC iz/azclass (< IE de-verbal es/os-stems) which denoted animate objects (young domestic animals); this was not the case in IE. Most of the GMC noun classes were semantically empty. They also interacted with gender which is the second classification system inherited from IE. Declension is expressed on the noun itself whereas gender is a typical agreement category which is marked on specific targets (pronouns, adjectives, articles). Thus, gender was sometimes connected to specific declension classes; e.g., the o¯-class exclusively contained feminine nouns. Subsequently, the Proto-GMC declension class assignment was formalized. Several classes were closed and many nouns changed their class: Some i-nouns moved to the a-class, while some athematic nouns shifted to the u-class. Later, nouns from the u-class often changed to the a- or the i-class etc. (Ramat [1981: 61]). This can be explained (in addition to the loss of meaning) by the decay of unstressed syllables at that time, which obscured class membership due to vocalic weakenings and reductions. Ramat (1981: 62) describes this collapse as a “crisis of the inflectional system in Germanic.” GMC had six cases. The vocative and the instrumental later merged with the nominative and the dative, respectively. The case/number-suffix always occurred as a fused portmanteau. Case/number morphology was characterized by some syncretisms which increased considerably at later stages. In contrast to most of the modern GMC languages, the nominative singular is also expressed, i.e., zero-marked singulars versus marked plurals developed much later. Table 10.1 provides a survey of the most important Proto-GMC declension classes. The first column indicates the declension class, the second one lists the genders corresponding to these classes. The third column contains examples in nom.sg. and nom.pl., respectively. The last column provides some additional remarks. In the following sections we cannot provide similarly detailed overviews for every language mentioned due to special constraints.

10.3 The North-Germanic Noun Systems Proto-GMC did not have definite and indefinite articles yet. Their grammaticalization started in the Middle Ages. A principal difference between North- and West-GMC languages is the position of the definite article: In West-GMC, it precedes the noun, whereas in North-GMC, it originally

217

218

DAMARIS NÜBLING

followed the noun and was later attached to it, i.e., after grammaticalization, Scandinavian nouns have an additional inflectional category, definiteness:1 West-Germanic Dutch: het huis German: das Haus DE F . SG . N. house

North-Germanic Swedish: huset Icelandic: hu´sið house-DE F . SG . N.

10.3.1

Icelandic – Increase of Complexity by Accumulated Phonological Change Icelandic and Faroese preserved and developed the most complex noun systems among the GMC languages which cannot be presented and examined in detail in this chapter. Number is expressed both in the singular and in the plural, and it is fused with case (underlined in Table 10.2). The attached definite article increased the overall complexity as can be seen in Table 10.2 with hestur (m., a-class) ‘horse’. The definite suffix (bold) also inflects for gender, case, and number. This kind of double-inflection (Icelandic hest-s-in-s horse-G E N . S G . - D E F - G E N . S G . ‘of the horse’: the genitive is expressed twice) was strongly reduced in languages of the Scandinavian mainland (Swedish ha¨st-en-s horse- D E F - G E N ‘of the horse’). Icelandic preserved most GMC declension classes, except the neuter iz/az-class (no. 2.5 in Table 10.1) which has been preserved only in some West-GMC languages. In addition, Icelandic developed a high number of subclasses due to the fact that assimilations between the grammatical endings (including the fused primary suffixes) and the lexical part caused many stem allomorphs (mostly occurring between AD 500 and 700). In contrast to many other languages most of these allomorphs have not been leveled out during the following centuries. Thus Icelandic is characterized by the accumulation of Table 10.2 Inflectional paradigm for Icelandic hestur ‘horse’ (m), a-class bare noun: ‘a horse’

nom. gen. dat. acc.

1

noun + definite suffix: ‘the horse’

Sg.

Pl.

Sg.

Pl.

hestur hests hesti hest

hestar hesta hestum hesta

hesturinn hestsins hestinum hestinn

hestarnir hestanna hestunum hestana

With regard to the indefinite article, it was only grammaticalized in the singular and precedes the noun in all GMC languages; however, in Icelandic indefinite articles are completely absent. Indefiniteness is expressed by zero: Ø hús ‘a house’, húsið ‘the house’.

Inflectional Morphology

regular phonological processes that were going on for centuries and which were neither functionalized to mark specific information (such as plural umlaut in German and Luxembourgish) nor eliminated. This led to a high amount of morphological diversity and even irregularity. Table 10.3 and Table 10.4 provide two frequent nouns of the u-class, Icelandic vo¨llur (m) ‘field’ and fjo¨rður (m) ‘fjord’. Whereas i- and u-umlaut (the latter is restricted to North-GMC) changed the stem vowel a of Proto-Norse *valþ- (Table 10.3), Proto-Norse *ferþ- was affected by rising of e > i before -i and by breaking, i.e., stressed short e before u > Old Norse jo˛ [jɔ] > Icelandic jo¨ [jø] and before a > Old Norse / Icelandic ja (Table 10.4; hyphens indicate the border between stem and inflectional ending). Except for acc.pl., no morphological change has occurred during language history. Until today, the consequences are rather heterogeneous paradigms which usually are explained by the fact that Icelandic has not been affected by language contact (and was hardly ever acquired as a second language). Faroese experienced a similar scenario, however it is less complex than Icelandic and lost the inflectional genitive.

Table 10.3 The paradigm of Icelandic völlur ‘field’ (m), u-class Modern Old Norse > Icelandic (since (since eighth c.) sixteenth c.)

N G D A

valþ-uR valþ-oˉR > -aR valþ-iu valþ-u

vǫll-r/ vall-ar/ vell-i/ vǫll-Ø/

völl-ur vall-ar vell-i völl

u-umlaut

N G D A

valþ-iuR valþ-oˉ > -a valþ-umR valþ-un(n)

vell-ir/ vall-a/ vǫll-um/ vǫll-u/

vell-ir vall-a völl-um vell-i

i-umlaut

number case sg.

pl.

phonological processes AD 500–700

stem + ending in Proto-Norse (AD 200–500)

i-umlaut u-umlaut

u-umlaut u-umlaut, today leveled out

Table 10.4 The paradigm of Icelandic fjörður ‘fjord’ (m), u-class stem + ending in Proto-Norse (AD number case 200–500)

Old Norse > (since eighth c.)

phonological Modern Icelandic (since processes AD 500–700 sixteenth c.)

sg.

N G D A

*ferþ-uR *ferþ-oˉR > -aR *ferþ-iu > firþ-iu *ferþ-u

fjǫrð-r fjarð-ar firð-i fjǫrð

fjörð-ur fjarð-ar firð-i fjörð

u-breaking a-breaking rising u-breaking

pl.

N G D A

*ferþ-iuR > firþ-iuR *ferþ-oˉ > -a *ferþ-umR *ferþ-un(n)

firð-ir fjarð-a fjǫrð-um fjǫrð-u

firð-ir fjarð-a fjörð-um firð-i

rising a-breaking u-breaking u-breaking, today leveled out

219

220

DAMARIS NÜBLING

In sum, there are three options of what can happen to the results of phonological change, the most important being umlaut: They can be preserved (Icelandic), they can be partially (Swedish, Danish) or completely eliminated (English, Frisian, Dutch), or they can be used morphologically and even made productive (analogical plural umlaut in German and Luxembourgish) (see Section 10.4.1). In principal, the same applies to verbal inflection (see Fertig,Chapter 9).

10.3.2 Swedish – Decrease of Complexity In Swedish, a two-gender-system evolved and it distinguishes neuter from common gender (“cg.”). The same holds for Danish and Norwegian (Bokmål). Furthermore, number and case are dissociated. The dative and accusative endings have been lost on the noun, and the genitive (more accurately the possessive marker) is developing into a clitic (-s) which is conditioned syntactically and attaches to the last element of the possessor phrase, here the pronoun mig ‘me’ (see Norde 1997, 2006): [en va¨n till mig]=s fo¨retag [a friend to me]=G E N company → ‘a friend of mine’s company’ In the singular, the definite suffix occurs after the stem (hund-en ‘the dog’), in the plural after the plural morpheme and in front of the possessive -s: hundar-na-s ‘dog- P L . - D E F . P L . - G E N . ’→ ‘of the dogs’. If the noun is preceded by an adjective, another definite article appears in the beginning of the NP: den stora hunden ‘ D E F . S G . C G . big-D E F . dog- D E F . S G . C G . ’ → ‘the big dog’. In sum, definiteness (underlined) is expressed three times. The younger adjectivearticle den derives from the GER d- article and is a loan from Middle Low GER, which was widely spoken in continental Scandinavia during the Hanse period. This double determination, “double definiteness” (Roehrs, Chapter 23) or “over-determination” (Dahl 2004) is the result of two competing grammaticalization areas, the older suffixation in North-GMC (Icelandic, Faroese) and the West-GMC model (German, English) of preposed determiners in the south. Swedish combined both strategies. Since Old Norse, the fused case and number suffix was separated in all Mainland Scandinavian languages and definiteness was morphologized. In these cases, an important serialization order evolved in correspondence to the relevance principle described by Bybee (1985, 1994): The less relevant case information is expressed in the rightmost periphery of the word form whereas the more relevant definiteness marker which marks that the denoted object is familiar is expressed closer to the stem.2 The most relevant 2

Booij (1996) differentiates between inherent (number) and contextual categories (case). However, the strictly relevance-based scale of Bybee (1985, 1994) is more subtle and fine-graded and allows for the scaling of different inherent and contextual categories (see Dammel and Gillmann 2014).

Inflectional Morphology

information is number because it affects the meaning of the noun most strongly (it carries the information about the number of the denoted referents): All GMC languages with separated number and case affixes placed number (the plural) in front of case, i.e., directly after the stem – and even alter the vowel quality of the stem if we consider plural umlaut (Swedish stad – sta¨der ‘city’).3 The likelihood of the most relevant categories to fuse with the stem increases if the lexeme has a high token- frequency. This relevancedriven order is best observable in Swedish hund-ar-na-s ‘dog- P L . - D E F . P L . - G E N . ’: L E X E M E - N U M B E R - D E F I N I T E N E S S - C A S E (see Ku ¨ rschner and Dammel 2013: 50). The Old Norse (and Icelandic) double-inflection of case, e.g., hund-s-in-s ‘dogG E N . S G . - D E F . - G E N . S G ’ → ‘of the dog’ has been reduced in Swedish: hund-en-s ‘dog-D E F . S G . - G E N . ’ → ‘of the dog’. The same holds for Danish and Norwegian. As we can see, Swedish is a language that is more agglutinative. However, it preserved some stem-based inflection (e.g., in class 3: flick-a – flick-or) and about 35 nouns with umlauting plurals which are not productive anymore and protected by their high token frequency: man – ma¨n ‘man’, stad – sta¨der ‘city’, bok – bo¨cker ‘book’, dotter – do¨ttrar ‘daughter’, tand – ta¨nder ‘tooth’, etc. The dominating inflection principle is base-form inflection, i.e., the complete singular form is part of the plural. Thus, the five plural allomorphs -ar, -(e)r, -or, -Ø and -n are the last reminders of the former inflectional system (the possessive marker consists of uniform -s; see Norde 1997, 2006 and Bo¨rjars 2003 for further details). This is also due to the fact that Swedish preserved three unstressed vowels which facilitate the first three classes in Table 10.5. As can be seen, most plural allomorphs are highly sensitive to gender, i.e., the declension class system is tightly connected to gender. There are however some exceptions such as common gender nouns in class 5 ending in -are and -ande: en lärare – två lärare ‘a teacher – two teachers’, en studerande – två studerande ‘a student – two students’. The last two neuter classes are conditioned by the final sound: Neuter nouns ending in a vowel belong to class 4 (the plural -n is the result of a reanalysis of the definite plural suffix), those ending in a consonant take zero plural. Here, formal plural assignment rules have emerged.

10.3.3

Danish – Representing the Most Simplified Scandinavian Noun Inflection In comparison to Swedish, Danish is characterized by the development of the most simplified noun system and only uses base-form inflection. Similar to Swedish, the Danish genitive developed into a phrasal possessive clitic. Only three noun classes are left represented by three plural allomorphs (see Table 10.6). This simplification was supported by phonological

3

In Old Norse and Icelandic, there are also case umlauts (especially for datives), e.g., Icelandic dagur (m.nom.sg.) ‘day’ – degi (m.dat.sg.) or akrar (m.nom.pl.) ‘acre’ – ökrum (m.dat.pl.). These case umlauts were eliminated at an early stage in the Mainland Scandinavian languages.

221

222

DAMARIS NÜBLING

Table 10.5 The Swedish declension class system4 gender class

plural allomorph

cg.

n.

1

-ar

+



2

-(e)r

+



3

-or

+



4

-n



+

5





+

Examples – definite singular – plural

+ definite singular – plural

hund – hundar ‘dog’ dag – dagar ‘day’ minut – minuter ‘minute’ hustru – hustrur ‘wife’ flicka – flickor ‘girl’ pizza – pizzor ‘pizza’ arbete – arbeten ‘job’ piano – pianon ‘piano’ å r – å r ‘year’ barn – barn ‘child’

hunden – hundarna dagen – dagarna minuten – minuterna hustrun – hustrurna flickan – flickorna pizzan – pizzorna arbetet – arbetena pianot – pianona å ret – å ren barnet – barnen

Table 10.6 The Danish declension class system plural class allomorph

1

-(e)r [ɐ]

2

-e [ə]

3



Examples – definite singular – plural

+ definite singular – plural

bil – biler ‘car’ bi – bier ‘bee’ dame – damer ‘lady’ dag – dage ‘day’ hund – hunde ‘dog’ hus – huse ‘house’

bilen – bilerne bien – bierne damen – damerne dagen – dagene hunden – hundene huset – husene

å r – å r ‘Jahr’ barn – børn ‘child’

å ret – å rene barnet – børnene

Remarks

mono- and polysyllabics, often with final vowel monosyllabics with final consonant; increasingly animates entering this class many (but not exclusively) neuter nouns

developments of Danish as all stressed vowels have been reduced to [ǝ]. In contrast to Swedish, there is a pure e-ending whose diachronic base is the acc.pl.-ending -a of the Old Norse masculine a-class (in Swedish, it was the Old Norse nom.pl. -ar which led to -ar). Although Danish equally preserved two genders, gender is irrelevant for plural assignment. It has been dissociated from the noun classes. Instead, the form of the noun is becoming more important (length, final sound). Many monosyllabic neuter and common gender nouns ending in a consonant shifted from the first to the second class, i.e., the number of syllables and the final sound are becoming relevant for the plural assignment. Thus, the high number of Old Norse neuter nouns with zero plurals has diminished: Many neuter nouns adopted the plural ending -e (hus – huse ‘house’, land – lande ‘country’) and -er if polysyllabic and/or ending in a vowel (bi – bier ‘bee’). In contrast to 4

The following tables exhibit only the biggest and most relevant classes and assignment principles. More information is provided by Kürschner (2008).

Inflectional Morphology

Icelandic and Faroese, in Danish (and to a lesser extent also in Swedish) the category of number underwent remarkable upgrading. In Danish as well as in Swedish, the preposed as well as the suffixed definite article have specific plural forms, which explains the fact that zero marked neuter plural nouns are rather common (even more in Swedish than in Danish). Furthermore, animate nouns enter the -e plural class (2). There are still about 25 nouns showing (unproductive) umlaut in the plural: barn – børn ‘child’ can be traced back to u-umlaut, other plurals to i-umlaut: stad – stæder ‘city’, bog – bøger ‘book’, hånd – hænder ‘hand’, datter – døtre ‘daughter’, etc. Unlike Swedish, Danish did not develop double determination: As soon as an adjective appears, the definite suffix disappears and is replaced by the preposed adjective-article den (or det): den store hund ‘def.cg. big-def. dog’ → ‘the big dog’. In Danish a division of labor is exercised by using the WestGMC type in case of modifiers (den store hund) and otherwise the North-GMC type (hunden ‘the dog’) (Dahl 2004).

10.4 The West-Germanic Noun Systems 10.4.1 German and Luxembourgish – Morphological Umlaut German preserved and developed a quite complex noun system consisting of about nine different noun classes. First of all, the GMC three-gender system has been continued. Gender is rather tightly connected to declension. In the course of time, the inflection of feminine nouns diverged more and more from that of masculine and neuter nouns. Most important is the fact that due to homophony of the feminine and the plural article (both die), there is no feminine noun with zero plural (die Schu¨ssel (f) ‘the bowl’ – die Schu¨sseln) (see Table 10.7, class no. 1), but many disyllabic masculine and neuter nouns with zero plural (der Schlu¨ssel (m) ‘the key’ – die Schlu¨ssel-Ø, see Table 10.7, class no. 6). This demonstrates the relevance of the entire NP: As the inflection of nouns decreased in all GMC languages during the last two thousand years, a good deal of number and case information is managed by the NP, the most important unit carrying information being the definite and indefinite article. Thus, a rather complex division of labor has taken place which also includes the adjective. German clearly shows the connection between determiners and noun inflection (so-called noun group inflection, see Roehrs, Chapter 23). A reason for the complexity of the German declension class system is the fact that case still has to be considered, above all the three genitive allomorphs; furthermore some noun classes display specific dative and accusative endings (which will not be discussed any further in this chapter). There is also a specific dative plural suffix -n which attaches to every noun unless its plural form ends in -n or -s. As this phonological rule does not contribute to the morphological noun class system, only the genitive singular suffix in combination with the plural suffix is taken into account for declension. Table 10.7 contains the most relevant declension classes. As there is a rather

223

224

DAMARIS NÜBLING

Table 10.7 The German declension class system (shaded rows: not productive) gen.sg. / plural allomorph

f.

Gender m. n.

examples (gen.sg. – nom.pl.) Notes

1

Ø / -(e)n

+





Dame – Damen ‘lady’ Schrift – Schriften ‘script’

biggest f. class; final [ə]→ -n, final consonant → -en, (pl. output: trochee)

2

-(e)n / -(e)n



+



Menschen – Menschen ‘human’ Kurden – Kurden ‘kurd’, Affen – Affen ‘ape’

only humans and higher animals, (mostly) trochees with final -[ə] → -n; (seldom) final consonant → -en

3

-s / -(e)n



+

+

Staats – Staaten (m) ‘state’, Hemds – Hemden (n) ‘shirt’

small mixed (new) class: strong gen. + weak pl.

4

-Ø / uml. + -e

+





Stadt – Städte ‘town’, Maus – Mäuse ‘mouse’, Kuh – Kühe ‘cow’

small unproductive class; ~35 token frequent monosyllabic f. nouns

5

-s / uml. (+ -e)



+



Hahns – Hähne ‘cock’, Ladens – Läden ‘shop’

monosyllabics with -e, disyllabics with zero plural (pl. output: trochee)

6

-s / (+ -e)



+

+

Tags – Tage (m) ‘day’, Brunnens – Brunnen (m) ‘fountain’, Ufers – Ufer (n) ‘coast’

trochees with zero plural, monosyllabics with -e (pl. output: trochee); former m. a-class, later opened for n.

7

-s / uml. + -er



+

+

Manns – Männer (m) ‘man’, originally small n. class Lamms – Lämmer (n) ‘lamb’, (GMC iz/az), later opened Hauses – Häuser (n) ‘house’ for m.

8

-s / -s



+

+

9

-Ø / -s/

+





Zoos – Zoos (m) ‘zoo’, Klos – foreign words, short Klos (n) ‘toilet’, Kommas – words, proper names Kommas (n) ‘comma’ (words with phonologically foreign structures) Pizza – Pizzas ‘pizza’

big group of phonologically deviating words which prefer -s in the plural (-s attaches to the lexeme but does not require any additional changes, thus the word remains more easily recognizable), classes 8 and 9 were added (in the North-GMC languages, -s plurals are not as common as in German). Most feminine nouns belong to the first class and always form a wellmarked plural while they never exhibit case markers (see also class nos. 4 and 9). They most clearly represent the diachronic upgrading of number, which always won when competing against case (Dammel and Gillmann 2014).5 Case is typically realized on the determiner. Another means of 5

To give an example: Homophonous case and plural affixes experience different developments as can be seen with final -e occurring with strong masculine nouns in the dative singular (dem Tage ‘to the day’) and in the plural (die Tage ‘the days’). In the first position, -e has already become extremely rare, in the second one, it is absolutely stable. This is due to the higher relevance of number in contrast to case for the meaning of the noun.

Inflectional Morphology

upgrading number is the umlauting plural which occurs most frequently in the masculine (class nos. 5 and 7) and also in the neuter (class 7); the feminine nouns only preserved a small number out of the group of umlauting plurals (class 4) most members of which changed to the first (weak) class. The remaining nouns comprise about 35 types, which have been rather stable for many centuries. Number upgrading becomes most evident if we consider class no. 7, the successor of the small neuter iz/az-class in Proto-GMC which has been completely lost in the North-GMC languages. In Old High German, this class had consisted of less than ten members, which denoted young animals (lamb ‘lamb’, kalb ‘calf’, ei ‘egg’, huon ‘hen’). Since then it has had a notable career by attracting more neuter nouns from the Proto-GMC a-class with zero markers in the nominative and accusative singular and plural. Later, some masculine nouns shifted to this class (Mann ‘man’, Wald ‘forest’, Gott ‘god’ etc. See Nu¨bling 2018). Today, about 80 neuter and 15 masculine nouns with high token-frequencies belong to this highly number-profiled class which is characterized by forming its plural with {umlaut + -er}. Although today unproductive, the members of this class are stable, which shows that productivity should not be the only criterion for a “good” inflection class. The most important development of the German inflection system is the rise of analogical (or morphological) umlaut. In this case, the results of phonological umlaut of the Proto-GMC i- (f., m.) and iz/az-class (n.) (classes 1.3 and 2.5 in Table 10.1) have been connected with plural meaning (grammaticalization) and subsequently attracted a lot of nouns which adopted this kind of explicit plural marking. Before becoming productive, phonological umlaut in the genitive and dative of the singular paradigms was eliminated. Especially masculine ({umlaut (+ -e)}, {umlaut + -er}) and neuter nouns ({umlaut + -er}) use analogical umlaut. In many disyllabic nouns, umlaut even progressed to become the only plural marker, e.g., Apfel – A¨pfel ‘apple’, Boden – Bo¨den ‘floor’. In contrast, the feminine nouns withdrew early from plural umlaut, which is only preserved in a small class of remnants, no. 4 in Table 10.7. In sum, the inflectional behavior of feminine nouns diverged more and more from that of nonfeminine nouns. Still, in present-day German, some umlauting plurals are competing against nonumlauting ones (e.g., Wagen/Wa¨gen ‘cars’, Erlasse/Erla¨sse ‘decrees’, Bogen/Bo¨gen ‘sheets’) although plural umlaut is not fully productive anymore. In the last centuries, numerous nouns shifted their class. Comparing masculine -e plurals with and without umlaut, Ko¨pcke (1994: 83–86) showed that animacy favors umlaut (Arzt/A¨rzte ‘doctor’, Wolf/Wo¨lfe ‘wolf’ versus Tag/Tage ‘day’, Dolch/Dolche ‘dagger’). Besides animacy, token frequencies and formal schemas are the driving forces behind the reorganization of the inflectional system and transitions to other classes (for further details, see Ko¨pcke 1988, 1993, 1994, 1995, 2000, Ku¨rschner 2008, Nu¨bling 2008). Today, the famous weak masculine nouns (class no. 2) are restricted to human males and some mammals, furthermore to trochaic structures and final -[ə]. Many inanimate nouns

225

226

DAMARIS NÜBLING

left this class. The most recent group that shifted classes were nouns denoting birds. They moved to class no. 5, cf. former Hahn ‘cock’ – Hahnen (also Schwan ‘swan’, Storch ‘stork’, etc.) with the current plural Ha¨hne (Schwa¨ne, Sto¨rche). Another option for former weak masculine nouns was class no. 6, e.g., Brunnen, Balken, Kuchen ‘fountain, beam, cake’; they changed their phonological schema by adopting -n in the nom.sg.: der Brunne > der Brunnen, der Schade > der Schaden ‘damage’. Therefore, any possibility of overtly distinguishing singular and plural has been lost. However, the distinctive plural article diepl. (against derm.sg.) compensates for this deficit. Luxembourgish as a small language closely related to German is characterized by apocope, which does not allow for plurals in -e. In this language, three compensations can be observed: First, the originally neuter class {umlaut + -er} has considerably more members than the German one (in Luxembourgish it is productive to this day): compare Luxembourgish De¨sch – De¨scher ‘table’ with German Tisch – Tische ‘table’. Second, (e)n-plural became extremely productive (see Dammel 2018). Third, umlaut is even more common than in German and spread to more noun classes: Luxembourgish Aarm – A¨erm ‘arm’ versus German Arm – Arme, Luxembourgish Rass – Re¨ss ‘crack’ versus German Riss – Risse, Luxembourgish Numm – Nimm ‘name’ versus German Name – Namen.6 Even foreign words adopt umlaut: Luxembourgish Mount – Me´int ‘month’ versus German Monat – Monate, Luxembourgish Clubb – Clibb ‘club’ versus German Club – Clubs. While German umlauting plural forms with a¨, u¨, o¨, or a¨u clearly refer to singular forms with a, u, o, or au, Luxembourgish has cut this connection. There is a broad variety of (former) umlauts far away from a 1:1-relation to a specific vowel in the singular. Thus, umlaut has become rather arbitrary. In principle, this is comparable to ablaut in the verbal system: Umlaut is not fully predictable from the singular anymore (Nu¨bling 2006, 2013; Dammel et al. 2010: 604–607; Dammel 2018).

10.4.2 Frisian: Conservation of Breaking West Frisian, historically most closely related to ENG, is spoken in the Netherlands. As most speakers are bilingual, there are many contactinduced influences from Dutch, mostly on the lexical level. Like most GMC languages, Frisian reduced case on the noun so that plural class is identical to noun class. Frisian also reduced one gender. Its two-gender system consists of neuter and common gender (cg.) which is expressed on the definite article it [ət] (n.sg.) and de (cg.sg.) the last being homophonous to the plural article. This led to the fact that there are no zero plurals except for some animates usually occurring in groups such as bern – bern ‘child’, skiep – skiep ‘sheep’, etc. The same principle is known from English. Interestingly, gender is completely detached from plural formation. 6

Similar extensions of umlaut into other classes are reported from Yiddish (Weissberg 1988).

Inflectional Morphology

Overall, a rather simple two-class noun system has evolved (Tiersma 2 [1999]; Hoekstra and Tiersma [1994]). Plural formation follows a simple, productive, and rather outputoriented rule: monosyllabic words take -en in the plural, disyllabic words ending in -[ə] take -n (this is the successor of the Proto-GMC n-stems or “weak inflection”, see no. 2.1 in Table 10.1). Other words including diminutives take -s. Evidently, plurals should form trochees. In the first class, however, there are many plurals with stem alternations going back to shortened vowels or broken diphthongs, a phonological rule from Middle Frisian (sixteenth–seventeenth century) where an additional syllable caused these alternations (De Graaf and Tiersma [1980], Tiersma [1983]). About half of the originally broken plurals are still preserved and concern the most frequent nouns; the others have been leveled out analogically, mostly resulting in the singular vowel. For some West-Frisian dialects, Tiersma (1982) described the other direction of leveling: Nouns which occur more frequently in the plural than in the singular used the broken plural forms as the base form for the new singular. Originally, earm ‘arm’ had the broken plural jermen ‘arms’. Today the singular is remodeled by using the plural stem jerm – jermen. The same holds for goes – gwozzen > gwos – gwozzen ‘goose – geese’; kies – Table 10.8 The West Frisian declension class system (shaded rows: not productive) class

plural allomorph

stem alternation

Examples

notes

no (productive)

tsjerk – tsjerken ‘church’ kroade – kroaden ‘pushcart’

monosyllabics: en; disyllabics ending in -[ǝ]: -n

1b

shortening

[a:] – [a]: hân– hannen ‘hand’ [u:] – [u]: hûs– huzen ‘house’

unproductive

breaking

[iǝ] – [jɪ]: stien – stiennen ‘stone’ [ɪǝ] – [jɛ]: beam – beammen ‘tree’ [uǝ] – [wo]: foet – fuotten ‘foot’ [oǝ] – [wa]: soan – soannen ‘son’

unproductive

-(e)n 1c

output: trochees

1a

2

-s

no

wurker – wurkers ‘worker’ words ending in woartel – woartels ‘carrot’ [ǝ] + sonorant; boatsje – boatsjes ‘small diminutives boat’

3

-ens -e

no

lears – learz-ens ‘boot’ bean – beane ‘bean’

lexically and semantically restricted; unproductive

227

DAMARIS NÜBLING

kjizzen > kjizze – kjizzen ‘tooth – teeth’; hoas – vjazzen > vjazze – vjazzen ‘stocking – stockings’. In Standard Frisian, real umlaut plurals have not survived, except ko – kij ‘cow’. It should be emphasized that all GMC languages underwent phonological i-umlaut (palatalization), NorthGMC additionally u-umlaut (labialization of a > o˛ [ɔ], later > o¨). However, vowel shortening and breaking was a typical Frisian phenomenon. There are some exceptions to the productive two-class system (no. 3 in Table 10.8) which go back to former noun classes (-e). The ending -ens is a double suffix composed of en+s and attaches to nouns for objects which typically occur in pairs or groups: reed – redens ‘skate’, lears – learzens ‘boot’, trep – treppens ‘stair’.

10.4.3 Dutch – Elimination of Umlaut, Prosodic Constraints The Dutch plural system is quite similar to the Frisian one. Dutch is also characterized by the development of a two-gender system and the dissociation of gender from declension which is restricted to plural formation, as case is not expressed on the noun anymore. Dutch also has two definite article forms: neuter het in the singular and de for cg. singular and cg./n. plural. Therefore Dutch also lacks zero plurals (Table 10.9). Just like in Frisian, Dutch plural formation is mostly conditioned by phonology and prosody. Only morphologically complex words such as Table 10.9 The Dutch declension class system (shaded rows: not productive) class

plural allomorph

stem alternation

Examples

Notes

1a

no

1b

voicing of final [s] – [z]: huis – huizen many native words; consonant ‘house’ modifications of 1b and [f] – [v]: graf – graven 1c can be combined ‘tomb’

-(e)n

stoel – stoelen ‘chair’ prosodisc/phonological gave – gaven ‘gift’ (monosyllabic: -en; disyllabic:- n); productive

vowel [ɔ] – [o:]: hof – hoven ~ 30 nouns lengthening ‘court’ [ɛ]– [e:]: weg – wegen ‘way’

1c

2

-s

no

vogel – vogels ‘bird’ phonologically fietser – fietsers conditioned: words ‘cyclist’ ending in [ǝ] + sonorant; hoogte – hoogtes morphologically ‘height’ conditioned: many word worstje – worstjes formations, especially ‘small sausage’ diminutives

3

-eren

no

kind – kinderen ‘child’

~ 15 neuters; unproductive

Output: trochees

228

Inflectional Morphology

hoogte ‘height’ and the numerous diminutives contradict these rules by taking -s (instead of -n). Umlaut has been eliminated completely, the only exception is stad – steden ‘town’ (see Figure 10.2 in Section 10.5.2). Class no. 1 shows another peculiarity, the voicing of the stem final consonant in front of the plural marker -en and/or the lengthening of the stem vowel. This applies to a rather big group; however, loanwords do not undergo these processes (see ANS I: 165–175). In contrast to German to which Dutch is closely related, Dutch noun inflection was dramatically simplified, formalized, and down-graded, i.e., the allomorphs -(e)n and -s are governed by prosodic and phonological rules (Ku¨rschner 2008, chapter 3). Old Dutch had a rather complex noun inflection system, which was gradually simplified and eventually shifted to the weak declension class with -en after a consonant and -n after a vowel. In the thirteenth century, the plural allomorph -s emerged whose origin remains opaque and needs more research. It seems to be of Ingweonian origin (Marynissen 1994) or may result from a reanalysis of the genitive singular (Nu¨bling and Schmuck 2010, Schmuck 2011). It started to occur exclusively with animate masculine nouns ending in -er (ridders ‘knights’, wevers ‘weavers’). Later, it expanded to nouns ending in -ier (portiers) and -eur (gouverneurs). Subsequently, animacy and gender lost in importance and were eventually given up (bevers ‘bibers’, kamers ‘rooms’). Today, every noun ending in [ə] + sonorant (-er, -en, -el, -em) takes the highly productive s-ending irrespective of the meaning of the word. Asyllabic -s preserves the trochaic structure of the word best. As the definite plural article is homophonous to the common gender singular article every noun obligatorily inflects for plural. Unlike German, Dutch does not display zero plurals. The s-allomorph after a final schwa-syllable avoids an additional syllable and thus guarantees the trochaic output. Another peculiarity concerns class no. 3: Contrary to Frisian, Dutch preserved a small group of neuter nouns of the Proto-GMC iz/az-class, which enlarged their former plural ending -er by the weak ending -en to -eren (the so-called “stapelsuffix”). This leads to dactyls (kind – kinderen ‘child’, ei – eieren ‘egg’). The former phonological umlaut has also been eliminated: lam – lammeren ‘lamb’, kalf – kalveren ‘calf’. Although being an unproductive and small (although semantically related) group (they refer to animate concepts), these plurals are quite stable (ANS I: 176). For the Afrikaans noun system, see Donaldson (1993, 1994) and Dammel et al. (2010: 610–612).

10.4.4 English: The Simplest Noun Inflection System English is characterized by the development of the simplest plural formation using -s as the only plural morpheme. Dependent on the preceding sound, some phonologically conditioned allomorphs exist, besides [s] after voiceless consonants (cats), [z] after voiced sounds (cows), and [ɪz] after

229

230

DAMARIS NÜBLING

sibilants (horses). Thus, the only plural marker is phonologically influenced by the context to its left. Formerly, it was the plural marker which affected the stem, i.e., the direction of determination has changed (see Section 10.4). As the definite article the does not distinguish between singular and plural, there are no zero plurals comparable with German or the Scandinavian languages. Only for nouns denoting animals usually occurring in groups there is a possible zero expression (many sheep, deer, fish, lion). English is (besides Afrikaans) the only language without nominal gender. Similar to Dutch, about 20 nouns display voicing of their final consonant which is a remainder from Old and Middle English when the plural marker still was syllabic (-es). These alternations have been preserved until today in token-frequent words, e.g., between [s] and [z] in house – houses [haʊs – haʊzɪz], [f] and [v] in calf – calves (wolf – wolves, wife – wives etc.) and – with [ð] in the plural – clothes. Some of these alternations are not obligatory anymore. Other very frequent and furthermore animate nouns preserved their lexicalized plurals. They are relics of the former noun classes: mouse – mice, louse – lice, and goose – geese once belonged to the strong feminine nouns (class no. 1.3 in Table 10.1); ox – oxen still shows the weak masculine ending -en (class no. 2.1); and foot – feet, tooth – teeth, man – men (including woman – women) were once members of the masculine i-class (no. 1.3 in Table 10.1) albeit some of them originally stem from other classes. The plural form child-ren (< Old English child-er-en) corresponds to Dutch kind-eren as it combines two former plural endings, -er from the GMC iz/az-class and the weak ending -en. As can be clearly seen, high tokenfrequency and animacy support irregular plurals. The English plural -s seems to originate from the former masculine a-class: Old English dagas developed to Middle English daw-es and was then syncopated to day-s (Krahe 1969, Vol. II: 28).

10.5 Findings and Further Topics As real comparative studies of more than two GMC languages are rather rare and nominal inflection is hardly investigated compared to verbal inflection we have to refer to the limited existing research literature on various West- and North-GMC languages (Dammel and Ku¨rschner 2008, Ku¨rschner 2008, Ku¨rschner and Dammel 2008, Dammel et al. 2010). In Romance and Slavic studies, however, the contrastive perspective is frequently employed. The comparison of a variety of GMC languages and their development from one and the same diachronic source reveals insightful implications and interrelations between different parameters. Some of them will be addressed in the following section.

Inflectional Morphology

Icelandic Faroese

German Luxembourgish

Swedish Frisian Yiddish Danish Dutch

high complexity many plural allomorphs / classes fused number-case morphemes stem-based inflection stem-internal inflection (umlaut) high degree of irregularity three genders gender-driven allomorphy

English Afrikaans

low complexity

two genders

few plural allomorphs separated number & case base-form inflection external inflection low degree of irregularity no gender formal pl.assignment

Figure 10.1 Complexity of nominal inflection of the GMC languages (based on Kürschner 2008: 8)

10.5.1 Overall Complexity There are different parameters for nominal complexity which are discussed in Ku¨rschner (2008) and Dammel and Ku¨rschner (2008). As we saw, not only the number of nominal classes or plural allomorphs is decisive. It is important whether case is still expressed on the noun and contributes to distinguishing different noun classes or if number and case still are fused in one portmanteau morpheme. Furthermore, the number of gender distinctions is relevant and whether there is a connection between gender and declension. Last but not least, the interaction between lexical stem and grammatical ending is an important factor. Apart from umlaut, some languages exhibit other phonologically determined stem alternations. Of course it is difficult to consider every factor. Nonetheless, from a bird’s eye view the GMC languages can be represented on a scale as in Figure 10.1. Icelandic is the most complex language with regard to noun inflection, English the least complex one. The presence or absence of an agglutinated article has been disregarded in this case.

10.5.2 How to Deal with Umlaut – Different Strategies Every GMC language experienced a period where phonological umlaut was productive. However, the length of this period was different as well as the number of affected vowels. Nonetheless, the solutions to deal with the products of umlaut are quite varied: Some languages (Frisian, Dutch, Afrikaans, English) eliminated umlaut, others retained it more or less as it was when it emerged (Icelandic, Faroese), whereas German and Luxembourgish employed it systematically for expressing plural (and in word formation) and even made it productive (see Figure 10.2 based on Nu¨bling 2013: 22). While German preserved a systematic relation between base (velar) vowel and umlaut (palatal) vowel, Luxembourgish cut this connection and exhibits a wide variety

231

232

DAMARIS NÜBLING

preservation Icelandic

Swedish

English

elimination

German

functionalization

Figure 10.2 Preservation, elimination and functionalization of umlaut in GMC languages

of quite (although not absolutely) arbitrary vowel correspondences between singular and plural forms. After all, the degree of morphological productiveness is highest in Luxembourgish as it also includes foreign words. Figure 10.2 shows the spectrum in these three strategies, with Icelandic, English, and German as representatives for different options. The whole space within the triangle can be used to locate any of the Germanic languages discussed in this paper. To present an example, Swedish should be located between preservation and elimination (most umlauts have disappeared), but the loss of umlaut was influenced functionally: If umlaut was preserved, then it was for functional purposes, i.e., for plural formation of highly tokenfrequent nouns. However, umlaut never became morphologized and thus productive. High token-frequency always has a conserving effect. Therefore, Swedish is situated within the triangle. Danish lies between Swedish and English. Dutch and Afrikaans can be located nearby English while Frisian preserved breaking as another phonological result. Luxembourgish (and apparently Yiddish) behaves in an even more extreme way than German.

10.5.3 Relative Strength of Lexical and Grammatical Morphemes In Proto-GMC the grammatical endings together with the primary suffix were attached to the lexical base. If there was any phonological interaction it was always from right to left, i.e., the grammatical part affected the lexical one (regressive assimilations such as rising of e > i before i, u or j, i-umlaut, u-umlaut, breaking, vowel lengthening, or vowel shortening). This relative strength of the grammatical ending has been abolished and the direction of assimilation has even been reversed in some languages. In Dutch, Frisian, and Afrikaans (in parts also in German), phonological and prosodic properties of the stem govern the plural allomorph, i.e., the selection is determined by the stem. In English, it is exclusively the stem which determines one and the same plural morpheme -s by transferring phonetic properties from left to right (progressive assimilation). Here, it is

Inflectional Morphology

right-to-left

left-to-right progressive determination of suffix progressive assimilation of suffix

other languages and former language stages

Dutch Frisian

Afrikaans

English

Figure 10.3 Direction of formal determination and influence between stem and suffix

the stem which dominates over the grammatical ending. This can be interpreted as inflectional weakness and as a possible pre-stage of deflection. Interestingly, inflectional weakness is connected to few or even no allomorphs. Figure 10.3 (based on Dammel and Ku¨rschner 2008: 255) illustrates this profound diachronic change. Again, the same well-known languages turned the original right-to-left into a left-to-right determination by formalizing and down-grading the plural assignment.7

10.5.4 Gender As a Determinant of Noun Inflection It seems to be a principle that three-gender languages tightly connect gender and inflectional morphology. In Icelandic, German, and also Faroese (albeit to a lesser extent) plural and case allomorphs interact with gender. In German, gender has even been strengthened as it became increasingly important for declension: Feminine nouns concentrated on the former weak declension (plurals in -(e)n) whereas the neuter nouns left this class and the masculine restricted them consequently to the above mentioned phonosemantic schema, i.e., animate nouns (denoting male humans) ending in -[ǝ]; most stable are those with an additional pretonic syllable (xXx): Gese´lle, Matro´se ‘companion, sailor’ (see Section 10.3.1). On the other hand, the plural {umlaut + -e} was restricted to masculine and {umlaut + -er} primarily to neuter nouns. If languages lose one gender, they typically dissociate the remaining two genders from inflection. Only Swedish is an exception, whereas Danish, Dutch, and Frisian detached their plural formation from gender (see Ku¨rschner, Chapter 12). Principally, the relative strength of the two nominal classification systems, declension and gender, can differ. Enger (2004) addresses the question “[o]n the relation between gender and declension” and states that “DeclensionFirst” exists if declension dominates and determines gender (“Gender is predicted on the base of declension”) while in the case of “GenderFirst” gender is ranked higher than declension (“Declension is predicted on the base of gender”). In German, both interactions can be 7

Dutch can be situated between English and German even concerning many aspects of nominal inflectional morphology other than the one discussed here (see Kürschner 2006).

233

234

DAMARIS NÜBLING

observed. Examples for “GenderFirst” have been provided already. ”DeclensionFirst” applied when inanimate nouns of the weak masculine class became feminine and thus remained in the same class but not in the same gender.8 Interestingly, nouns for objects of low animacy such as fish, insects, reptiles, molluscs, furthermore body parts and plants changed their gender (Schnake ‘crane fly’, Schnecke ‘slug’, Drohne ‘drone’, Schlange ‘snake’, Hode ‘testicle’, Niere ‘kidney’, Lilie ‘lily’, Traube ‘grape’). Masculine nouns for completely inanimate objects formed a new class by adopting final -n in the nominative singular which used to be present only in the plural before (see Section 10.3.1). Thus, they take zero plurals (or pure umlaut plurals) and -s in the genitive singular: der Brunnennom.sg. – des Brunnensgen.sg. – die Brunnen-Øpl. ‘fountain’ < Middle High German brunne; der Schadennom.sg. – des Schadensgen.sg. – die Scha¨denpl. ‘damage’ < Middle High German schade.

10.6 Conclusion The modern GMC languages have simplified the Proto-GMC complex noun inflection system to very different degrees. Icelandic preserves the most complex structures whereas English gave up all declension classes as well as grammatical gender and case except the possessive clitic. Languages with three genders use them for the organization of their noun inflection (Icelandic, Faroese, German, Luxembourgish), whereas those with two genders tend to dissociate them from declension (Danish, Dutch, Frisian) except Swedish with full unstressed vowels and a preserved five-class system. The loss of gender is always connected to or more specifically preceded by a decrease of phonological distinctiveness. Interestingly, gender loss correlates with the loss of case. Thus, the declension class system of many languages is identical with the number of their plural allomorphs. Most languages formalized their plural expression, i.e., morphological, phonological, or prosodic features of the stem determine the plural allomorph. Only lexical plurals such as English men, children, mice, oxen are based on semantics and have a high token-frequency. German retained and further developed a rather complex noun class system, which is based on semantics, gender, and form. Moreover, the transition to other declension classes is mostly determined by semantics, by formal properties of the stem, and by gender. Thus, declension caused a clear feminine/nonfeminine distinction in German, a distinction which has even been sharpened in the last centuries. Zero inflection can only be found in languages with distinct singular and plural articles. If these are homophonous, overt plural inflection is obligatory (Dutch, Frisian, English). These interrelations show that the noun is integrated into the NP and that all its components contribute to the expression of the nominal categories. Most important is the article, 8

Today there are, however, slight inflectional differences between the masculine and feminine weak class.

Inflectional Morphology

which was grammaticalized after the Proto-GMC period. The Scandinavian languages even agglutinated it to the noun. Umlaut (and further stem-affecting phonological effects) was treated highly differently: Some languages preserved it, resulting in rather heterogeneous paradigms, other languages eliminated it, and a third group developed it as a systematic means for expressing plural. In these languages, grammatical information is fused to the lexical stem. Animacy and high token-frequency mostly correlate with umlaut or other stem alternations. In these languages the dominance of the grammatical ending over the lexical stem still applies. Other languages reversed this direction: The stem governs the choice of the plural allomorph, which means that inflection has been weakened and could probably result in deflection.

References ANS = Haeseryn, W. et al. (eds.) 21997. Algemene Nederlandse Spraakkunst. Leiden: Nijhoff. Booij, G. 1996. “Inherent versus contextual inflection and the split morphology hypothesis,” Yearbook of Morphology 1995: 1–16. Bo¨rjars, K. 2003. “Morphological status and (de)grammaticalisation: The Swedish possessive,” Nordic Journal of Linguistics 26: 133–163. Bybee, J. 1985. Morphology: A Study of the Relation between Meaning and Form. Amsterdam: John Benjamins. Bybee, J. 1994. “Morphological universals and change.” In R. Asher (ed.), The Encyclopedia of Language and Linguistics, Vol. V. Oxford: Pergamon Press: 2557–2562. ¨ . 2004. “Definite articles in Scandinavian: Competing grammaticaliDahl, O zation processes in standard and non-standard varieties.” In B. Kortmann (ed.), Dialectology meets Typology. Berlin: Mouton de Gruyter: 147–180. Dammel, A. 2018. “Warum eigentlich nicht Worter? Ein Beitrag zur ¨ kumene des Umlauts.” In K. Kazzazi, K. Luttermann, S. Wahl, and O T. Fritz (eds.), Worte u¨ber Wo¨rter. Tu¨bingen: Stauffenburg: 65–98. Dammel, A. and M. Gillmann 2014. “Relevanzgesteuerter Umbau der Substantivflexion im Deutschen. Spiegelt Diachronie Typologie?” Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 136.2: 173–229. Dammel, A and S. Ku¨rschner 2008. “Complexity in nominal plural morphology: A contrastive survey of ten Germanic languages.” In M. Miestamo, S. Kaius, and F. Karlsson (eds.), Language Complexity: Typology, Contact, Change. Amsterdam: John Benjamins: 243–262. Dammel, A., S. Ku¨rschner, and D. Nu¨bling 2010. “Pluralallomorphie in den germanischen Sprachen: Konvergenzen und Divergenzen in Ausdrucksverfahren und Konditionierung.” In A. Dammel, S. Ku¨rschner, and D. Nu¨bling (eds.), Kontrastive germanistische Linguistik. Hildesheim: Olms: 587–642.

235

236

DAMARIS NÜBLING

Donaldson, B. 1993. A Grammar of Afrikaans. Berlin: Mouton de Gruyter. Donaldson, B. 1994. “Afrikaans.” In E. Ko¨nig and J. van der Auwera (eds.), The Germanic Languages. London: Routledge 478–504. De Graaf, T. and P. Tiersma 1980. “Some phonetic aspects of breaking in West Frisian,” Phonetica 37: 109–120. Enger, H-O. 2004. “On the relation between gender and declension: A diachronic perspective from Norwegian,” Studies in Language 28.1: 51–82. Hoekstra, J. and P. Tiersma 1994. “Frisian.” In E. Ko¨nig and J. van der Auwera (eds.), The Germanic Languages. London: Routledge: 505–531. Ko¨pcke, K-M. 1988. “Schemas in German plural formation,” Lingua 74: 303–335. Ko¨pcke, K-M. 1993. Schemata bei der Pluralbildung im Deutschen. Versuch einer kognitiven Morphologie. Tu¨bingen: Narr. Ko¨pcke, K-M.1994. “Zur Rolle von Schemata bei der Pluralbildung monosyllabischer Maskulina.” In K-M. Ko¨pcke (ed.), Funktionale Untersuchungen zur deutschen Nominal- und Verbalflexion. Tu¨bingen: Max Niemeyer Verlag: 81–95. Ko¨pcke, K-M. 1995. “Die Klassifikation der schwachen Maskulina in der deutschen Gegenwartssprache,” Zeitschrift fu¨r Sprachwissenschaft 14.2: 159–180. Ko¨pcke, K-M. 2000. “Starkes, Schwaches und Gemischtes in der Substantivflexion des Deutschen. Was weiß der Sprecher u¨ber die Deklinationsparadigmen?” In R. Thieroff, M. Tamrat, N. Fuhrhop, and O. Teuber (eds.), Deutsche Grammatik in Theorie und Praxis. Tu¨bingen: Max Niemeyer Verlag: 155–170. Krahe, H. 1969. Germanische Sprachwissenschaft, Vol. II: Formenlehre. Berlin: Mouton de Gruyter. Ku¨rschner, S. 2006. “De Nederlandse meervoudsallomorfie tussen Duitse complexiteit en Engelse eenvoud.” In M. Hu¨ning, V. Matthias, U. Vogl, T. van der Wouden, and A. Verhagen (eds.), Nederlands tussen Duits en Engels. Leiden: Stichting Neerlandistiek Leiden: 103–122. Ku¨rschner, S. 2008. Deklinationsklassenwandel. Eine diachron-kontrastive Studie zur Entwicklung der Pluralallomorphie im Deutschen, Niederla¨ndischen, Schwedischen und Da¨nischen. Berlin: Walter de Gruyter. Ku¨rschner, S. and A. Dammel 2013. “Flexionsklassenwandel im Vergleich. Nominale und verbale Entwicklungen in vier germanischen Sprachen.” In J. Fleischer and H. Simon (eds.), Sprachwandelvergleich – Comparing Diachronies. Tu¨bingen: Max Niemeyer Verlag: 43–71. Marynissen, A. 1994. “Het -s-meervoud in het vroegste ambtelijke Middelnederlands,” Amsterdamer Beitra¨ge zur A¨lteren Germanistik 40: 63–105. Neef, M. 2000a. “Phonologische Konditionierung.” In G. Booij, C. Jehmann, J. Mudgan, and S. Skopetas (eds.), Morphology. An International Handbook on Inflection and Word-Formation, Vol. I. Berlin: Mouton de Gruyter: 463–473.

Inflectional Morphology

Neef, M. 2000b. “Morphologische und syntaktische Konditionierung.” In G. Booij, C. Jehmann, J. Mudgan, and S. Skopetas (eds.), Morphology. An International Handbook on Inflection and Word-Formation, Vol. I. Berlin: Mouton de Gruyter: 473–484. Norde, M. 1997. The History of the Genitive in Swedish. A Case Study in Degrammaticalization. Amsterdam: John Benjamins. Norde, M. 2006. “Demarcating degrammaticalization: The Swedish s-genitive revisited,” Nordic Journal of Linguistics 29.2: 201–238. Nu¨bling, D. 2006. “Zur Entstehung und Struktur ungeba¨ndigter Allomorphie: Pluralbildungsverfahren im Luxemburgischen.” In C. Moulin and D. Nu¨bling (eds.), Perspektiven einer linguistischen Luxemburgistik. Heidelberg: Winter: 107–128. Nu¨bling, D. 2008. “Was tun mit Flexionsklassen? Deklinationsklassen und ihr Wandel im Deutschen und seinen Dialekten,” Zeitschrift fu¨r Dialektologie und Linguistik 753: 282–330. Nu¨bling, D. 2012. “Auf dem Wege zu Nicht-Flektierbaren: Die Deflexion der deutschen Eigennamen diachron und synchron.” In B. Rothstein (ed.), Nicht-flektierende Wortarten. Berlin: Mouton de Gruyter: 224–246. Nu¨bling, D. 2013. “Zwischen Konservierung, Eliminierung und Funktionalisierung: Der Umlaut in den germanischen Sprachen.” In J. Fleischer and H. Simon (eds.), Comparing Diachronies. Berlin: Mouton de Gruyter: 15–42. Nu¨bling, D. 2018. “Worte versus Wo¨rter: Zur Genese und zur semantischen Differenzierung einer Pluraldublette.” In K. Kazzazi, K. Luttermann, S. Wahl, and T. Fritz (eds.), Worte u¨ber Wo¨rter. Tu¨bingen: Stauffenburg, 385–407. Nu¨bling, D. and M. Schmuck 2010. “Die Entstehung des s-Plurals bei Eigennamen als Reanalyse vom Kasus- zum Numerusmarker. Evidenzen aus der deutschen und niederla¨ndischen Dialektologie,” Zeitschrift fu¨r Dialektologie und Linguistik 77.2: 145–182. Ramat, P. 1981. Einfu¨hrung in das Germanische. Tu¨bingen: Max Niemeyer Verlag. Schmuck, M. 2011. “Vom Genitiv- zum Pluralmarker: Der s-Plural im Spiegel der Familiennamengeographie.” In R. Heuser, D. Nu¨bling, and M. Schmuck (eds.), Familiennamengeographie. Berlin: Walter de Gruyter: 285–304. Tiersma, P. 1982. “Local and general markedness,” Language 58.4: 832–849. Tiersma, P. 1983. “The nature of phonological representation: evidence from breaking in Frisian,” Journal of Linguistics 19: 59–78. Tiersma, P. 21999. Frisian Reference Grammar. Leeuwarden: Fryske Akademy Ljouwert. Weissberg, J. 1988. Jiddisch. Eine Einfu¨hrung. Berlin: Peter Lang Verlag.

237

Chapter 11 Principles of Word Formation Geert Booij

11.1 Introduction Germanic languages share their main principles of word formation. The two most important processes of word formation that are used are compounding and affixation. In addition, new words can also be formed by means of conversion, the transposition of a word to another syntactic category without overt morphological marking. Apophony or vowel alternation, traditionally referred to as Ablaut in Indo-European linguistics (see Fertig, Chapter 9), only occurs in closed sets of related established words, and is unproductive: it cannot be used productively for word formation anymore. However, established words of this type can be used as constituents of compounds. Reduplication, the copying of words or parts thereof, is cross-linguistically a very common mechanism of word formation, but plays a minor role in Germanic languages. Finally, there are various ways of abbreviating words into shorter ones, for instance in order to express endearment, and blending of words also occurs in Germanic. These various word formation processes will be discussed in the following sections. Word formation processes are mainly used to create new words of lexical categories: nouns, verbs, adjectives, and adverbs. Words of grammatical categories such as prepositions and conjunctions may be complex as well, but this is due to the diachronic process of univerbation, word sequences becoming words. An example of univerbation is the English preposition notwithstanding. The similarities between Germanic languages in the domain of word formation originate mainly from their having a common ancestor. In addition, language contact between the individual languages also boosted similarity. For instance, Danish and Norwegian (through Danish) have borrowed lots of complex words from Middle Low German and High German, Frisian is heavily influenced by Dutch, Faroese by Danish, and recently all (other) Germanic languages borrowed words from English.

Principles of Word Formation

Most Germanic languages were also influenced by French. Yiddish is a language with a German base, but with extensive influence of HebrewAramaic, Romance, and Slavic languages. Finally, Germanic languages borrowed words and word formation patterns from Greek and Latin, in some cases through French.

11.2 Compounding Compounding is the process in which two or more words are combined into a compound. The dominant pattern in Germanic is that of the rightheaded compound: The rightmost constituent functions as the semantic and formal head, and the left constituent functions as a modifier. The general schema for Germanic compounds is therefore the following (Booij 2010, chapter 2): (1)

form: meaning:

[[a]X [b]Y]Y [b]Y with some relation R to [a]X

In this schema a and b are variables for the phonological contents of words, and X and Y are variables for the syntactic category of the word constituents. Y is a variable for the major word classes Noun, Adjective, Adverb, and Verb. In the modifier position with the variable X we may find other word categories such as prepositions and numerals as well. As the schema indicates, the syntactic category of the whole compound word is the same as that of the rightmost constituent, the head. The relation R is unspecified, and is determined for each individual compound by the semantics of the two components and general knowledge. Compounds in which the syntactic category of one of the constituents determines the category of the compound as a whole, such as the rightheaded compounds of Germanic, are called endocentric compounds. Exocentric compounds are rare in Germanic languages. An example is English cut-throat, which does not denote a throat but a person who cuts throats, a murderer. The most productive category of compounds is formed by N+N compounds. Here is an example of the compound for ‘blood pressure’ in a number of the Germanic languages: (2)

Afrikaans Danish Dutch English Frisian German Icelandic Norwegian Swedish

bloed-druk blod-tryk bloed-druk blood-pressure bloed-druk Blut-druck blo´ð-þry´stingur blod-trukk blod-tryck

239

240

GEERT BOOIJ

When we compare these compounds, we will see that English is exceptional in that its word for ‘pressure’ is not of Germanic but of Romance origin. This reflects the fact that English has been influenced much earlier and more profoundly by French than the other Germanic languages, an effect of the Norman conquest in 1066. The morphosyntactic properties of the head of a compound are transferred to the compound as a whole. For instance, when the head is a noun, the compound as a whole is a noun as well. In languages with gender, the gender of the head noun is identical to the gender of the compound. For instance, German distinguishes three genders for nouns, neuter, masculine, and feminine, and thus we see the following pattern of gender transfer, with the same choice of definite singular article: (3)

das Recht ‘the right’ (neuter) der Mann ‘the man’ (masculine) die Heizung ‘the heating’ (feminine)

das Menschenrecht ‘the human right’ der Hauptmann ‘the head-man, captain’ die Zentralheizung ‘the central heating’

The N+N compound schema is very productive because the N constituents can also be compounds themselves. Hence, we get recursivity in this type of compounding. German is famous for its richness of recursive compounds (Gaeta and Schlu¨cker 2012). Here are some examples: (3)

Lebens-mittel-farb-stoff-zulassungs-verordnung ‘lit. life-means-colorstuff-approval-regulation, food coloring approval regulation’ Straßen-ausbau-beitrags-gesetz ‘lit. street-improvement-contributionlaw, law that requires a financial contribution for the street in which one lives’

Such long compounds may also occur in other Germanic languages such as Danish and Dutch. The left constituent of Germanic compounds, in particular nominal compounds, can also be a phrase, and even a sentence: (4)

English Dutch Faroese German

all-you-can-eat buffet, first-in-first-out policy, a let-it-happen -attitude, the eat- your-spinach-approach to education buiten-de-deur-eters ‘lit. outside-the-door-eaters, restaurant visitors’ gamla-manna-dansur ‘old men’s dance’ Bitte-greifen-Sie-zu-Schu¨ssel ‘lit. please-take-you-to-bowl, please help yourself bowl’

Adjectival compounding is also a productive process in Germanic languages, as illustrated by the compound word for ‘environmentfriendly’:

Principles of Word Formation

(5)

Afrikaans Danish Dutch Frisian German Nor wegian Swedish

omgevings-vriendelik miljø-venlige milieu-vriendelijk miljeu-freonlik umwelt-freundlich miljø-vennlig miljo¨-va¨nlig

However, not all types of compounding are productive. A notable exception is that in Germanic languages verbal compounding, that is, compounding with a verbal head tends to be marginal. Such verbal compounds do exist, but mainly arise through back formation. A textbook example is the English verb to babysit that was formed from the N+N compound babysitter by omitting the suffix -er of the deverbal nominal head. A similar Dutch example is the verb beeld-houwen ‘to sculpture’ coined on the basis of the N+N compound beeld-houw-er ‘lit. image hewer, sculptor’. Frisian has a remarkable number of such N+V compounds, for instance syk-helje ‘to breath-take’, which are used without being split in main clauses, as in Hy sykhelle amper ‘lit. He breath-takes hardly, he hardly breathes’. Instead of verbal compounds, Germanic languages prefer another option, phrasal combinations of a verb with another word. A dominant type is that of the particle verb, in which a particle (often related to a preposition or an adverb) combines with a verb (Los et al. 2012): (6)

Danish Dutch German English Frisian Swedish Yiddish

af-vise ‘to reject’ op-bellen ‘to phone up’ an-rufen ‘to phone up’ to phone up op-skriuwe ‘to write down’ stiga upp ‘to rise up’ ahinter loyfn ‘lit. back walk. to run back’

In English and the Scandinavian languages the particle follows the verb, in Afrikaans, Dutch, Frisian, German, and Yiddish the particle precedes the verb. The phrasal nature of the lexical expressions can be deduced from the fact that verb and particle can be separated: (7)

Dutch: German: English: Nor wegian:

Jan belde zijn moeder op ‘Jan phoned his mother up’ Der Johann rief seine Mutter an ‘Johann phoned his mother up’ John phoned her up Mann-en har drikket vinn-en opp ‘lit. Man-D E F . S G has drunk wine-D E F . S G up, the man has drunk the wine’

In the case of N+V and A+V combinations Germanic languages may use socalled quasi-incorporation (Booij 2010, chapter 3):

241

242

GEERT BOOIJ

(8)

Dutch German Frisian Norwegian

koffie zetten ‘lit. coffee make, to make coffee’ schoon maken ‘lit. clean make, to clean’ Kino gehen ‘lit. cinema go, go to the cinema’ Not landen ‘lit. emergency land, to emergency land’ klear komme ‘lit. ready become, to become ready’ spille piano ‘lit. play piano, to play the piano’

These (phrasal) expressions are qualified as cases of quasi-incorporation because they still consist of two words, and are split in main clauses, as in Dutch Hij speelt piano ‘He plays the piano’. Yet, they behave like close syntactic units, for instance in verb clusters, as in the following Dutch sentence: (9)

Ik hoorde dat Jan {kan piano spelen / piano kan spelen} I heard that John {can piano play / piano can play} ‘I heard that John can play the piano’

In the first variant of this sentence the modal verb kan precedes the cluster piano spelen. Both variants are grammatical. Germanic languages also have copulative compounds of the type blue-green. The compound adjective qualifies entities that are both blue and green, and hence, there is no semantic relationship of modification between the left and the right constituent. Instead, the compound blue-green means ‘blue and green’. Yet, from a formal point of view the rightmost constituent may still be considered as the head of such compounds, when this is the only part that carries inflectional endings. This can be observed in German ein blau-gru¨n-er Stuhl ‘a blue-green chair’, where only gru¨n carries the proper inflectional ending -er. On the other hand, in Swedish both adjectival parts agree with the head noun, at least when the head noun is neuter singular, as in ett Engelsk-t-Svensk-t lexicon ‘an English-Swedish lexicon’. There are also nominal and verbal copulative compounds such as German Fu¨rst-Bischof ‘prince-bishop’ and Dutch roer-bakken ‘to stir-fry’, respectively. In addition to compounds that consists of two words, in most Germanic languages we also find compounds with one or two nonnative bound roots. For instance, the Greek bound root bio ‘life’ appears in German Bio-loog ‘biologist’, Bio-laden ‘bio-shop’, in bio-aktiv ‘bio-active’, and in Dutch bioscoop ‘cinema’, bio-industry ‘bio-industry’, and bio-akker ‘organic garden’. There are hundreds of such roots, and thus we get many parallel words of this type in Germanic languages. These roots are also called confixes or combining forms: They are bound forms, like affixes, but are similar to words in the kind of meaning they carry, which is more lexical than grammatical in nature. Germanic languages may share a certain word formation process but differ in the degree to which they make use of it. A well-known case is that

Principles of Word Formation

both Dutch and German can make an [A+N]N compound such as German Gelbsucht and Dutch geelzucht, both meaning ‘jaundice’. Yet, German often has an [A+N]N compound where Dutch uses an A+N phrase. Compare, for example German Rot-wein ‘red wine’ and Weiß-wein ‘white wine’ with the Dutch phrases rode wijn and witte wijn, with the same meanings. As the English glosses indicate, Dutch is more similar to English in this respect than to German. In Germanic compounds, words of different origin can be combined. For instance, many English loanwords appear in Dutch compounds, as in Dutch brood-shop ‘bread-shop’ and zomer-sale ‘summer sale’. In addition, we find lots of loan translations. Frisian has a high number of loan translations of Dutch words, due to the intensive contact between these two languages. For instance, Frisian sondesskoalle is a loan translation of Dutch zondags-school ‘Sunday school’ and foetbalje ‘to play football’ is a loan translation of the Dutch verb voetballen ‘to play football’; compare Frisian fuotbalje (Dutch voet = Frisian fuot). In Yiddish we find compounds with a Hebrew component, and the Hebrew left-headed compound structure, as in sof-wokh ‘lit. end-week, weekend’, consisting of Hebrew sof ‘end’ and Yiddish wokh ‘week’, hence a loan translation of German Wochenende ‘weekend’.

11.2.1 Linking Elements As stated above, compounds are combinations of two or more words. In most Germanic languages (with the exception of English), compounds may have a linking element between the two constituents of a Germanic compound. These linking elements often have the form /ə/, /ər/, /ən/, or /s/. This is a reflection of the fact that these elements derive historically from case endings. For instance, the Dutch compound koning-s-zoon ‘king’s son’, has a linking element /s/ between the constituent words koning and zoon. This compound derived from the phrase koning-s zoon ‘king’s son’, in which -s is a genitive case ending. The reinterpretation of these endings as linking elements can be concluded from their distribution. The /s/, originally a genitive case ending for singular masculine and neuter German nouns, now also appears with feminine nouns, as in Scho¨nheit-s-ideal ‘beauty ideal’. It also appears after verb stems, as in Dutch scheid-s-rechter ‘referee’, with the verbal stem scheid ‘to sort out’. Here are some examples of N+N compounds with linking elements from German, which has the largest range of such linking elements of all Germanic languages: (10)

Bad-e-tuch ‘bath towel’ Scho¨nheit-s-ideal ‘beauty ideal’

243

244

GEERT BOOIJ

Tag-es-form ‘day form’ Decke-n-leuchte ‘ceiling light’ President-en-wahl ‘president election’ Kind-er-dorf ‘children village’ Herz-ens-wunsch ‘heart wish’ Phonologically, the linking elements form one phonological word with the preceding word stem, and hence these stem + linking element combinations might be considered as compound-bound allomorphs of left constituent words. In some cases, these morphemes may still have a morphological function, the marking of plurality, as plural nouns can function as left constituents of compounds: (11)

Dutch stad-s-raad ‘city council’ German Lande-s-konferenz ‘conference of a state’

sted-en-raad ‘cities’ council’ La¨nd-er-konferenz ‘conference of states’

This shows that inflected forms of words can feed word formation, in particular compounding (Booij 1993). This applies in particular to plural forms of nouns, participles, and infinitives.

11.2.2

Affixoids

Constituents of compounds may acquire specific meanings that are more abstract than the meaning of the corresponding word when used on its own. As left constituents they often express an evaluative or intensive meaning. For instance, the following compounds express ‘action as fast as lightning, very fast action’, and ‘very big’ respectively: (12)

Dutch bliksem-actie, German Blitz-aktion, Swedish blix-action Dutch reuze-groot, German riesen-groß, Swedish ja¨tte-stor

The words for lightning and giant have acquired a more general abstract meaning of great speed and size in these languages. Words used as right constituents may also have meanings that differ from their meaning when used as separate words. Some of these compound constituents have developed a more general meaning of denoting persons of a certain type. (13)

German Papst ‘pope’ Dutch boer ‘farmer’

Literatur-papst ‘lit. literature pope, literature authority’ kranten-boer ‘lit. newspaper farmer, newspaper seller’

Principles of Word Formation

Adjectives may also have acquired a more specific meaning. For instance, the adjective for ‘free’ has acquired the meaning ‘without’ in adjectival compounds such as the following: (14)

Dutch vet-vrij, English fat-free, German fett-frei, Swedish fett-fri

In this use, these constituents are referred to as affixoids, because they are similar to affixes in that they have a meaning that is bound to their occurrence in complex words (Hu¨ning and Booij 2014). These affixoids are also the diachronic source of affixes. An example is the English suffix -wise, originally an English word with the meaning ‘manner’, and now used in words like money-wise with the meaning ‘from the point of view’, and with the meaning ‘in the manner of’ in frog-wise. Germanic languages have a number of intensifier affixoids. Here are some examples from Frisian: (15)

dea ‘death’ dea-bang ‘very afraid’, dea-gewoon ‘very common’ poer ‘pure’ poer-lilk ‘very ugly’, poer-verlegen ‘very shy’ troch ‘through’ troch-kaˆld ‘very cold’, troch-waarm ‘very warm’

Like in Dutch, the intensifier affixoid can be repeated in Frisian by means of coordination with the conjunction en ‘and’, or the older form of this conjunction, ende: (16)

poer-en(de)-poer-swart ‘pure-and-pure-black, very, very black’ troch-en(de)-troch-kaˆld ‘through-and-through cold, very, very cold’

Such compounds with an intensifier left constituent are also referred to as elative compounds (Hoeksema 2012).

11.3 Derivation In Germanic languages affixation, both prefixation and suffixation, is used predominantly for the derivation of new words. These processes are used for the derivation of all major categories: nouns, verbs, adjectives, and adverbs. Examples of derived words from these categories in English are un-rest, be-little, un-able (prefixation), and bak-er, stabil-ize, beauti-ful, and nice-ly (suffixation). The input words are usually also words of major categories (except adverbs), but in some cases they can also be words of minor lexical categories such as prepositions and numerals. Phrases also occur as bases of derivation. This is illustrated here by the derivation of Dutch diminutives; the last two examples have a phrase as their base:

245

246

GEERT BOOIJ

(17)

huis ‘house’ blond ‘blond’ speel ‘to play’ tien ‘ten’ onder ons ‘between us’ twaalf uur ‘12 o’clock’

huis-je ‘small house’ blond-je ‘blond woman’ speel-tje ‘toy’ tien-tje ‘ten-guilder-note’ onderons-je ‘private chat’ twaalfuur-tje ‘lunch packet’

Very often, an affix in one Germanic language will have a cognate in a number of others. This is illustrated here for a number of prefixes and suffixes. For instance, the prefix be- is used in various Germanic languages to create verbs. Another illustration is that the following Danish suffixes all have a cognate in Dutch and German, and some of the English glosses are also cognates: (18)

Danish -bar -er -ing -hed -lig -som

Dutch -baar -er -ing -heid -lijk -zaam

German -bar -er -ung -heit -lich -sam

English gloss -able -er -ing -ness -ly -some

The prefix of prefixed verbs often derives from a particle, and Germanic languages may therefore differ in that a particle morpheme in one language has a prefix cognate in another one. For instance, the Dutch particle verb over-laten ‘to leave to’ (a separable complex verb) has the German prefixed (inseparable) verb u¨ber-lassen ‘to leave to’ as its cognate. In the Dutch particle verbs, main stress is on the particle, but in the German prefixed verb main stress is on the verb stem: o´ver-laten versus u¨ber-la´ssen. The change from particle to prefix illustrated here is a case of grammaticalization, the diachronic process in which lexical morphemes become grammatical morphemes, and grammatical morphemes become even more grammatical (Hopper and Traugott 1993). Besides prefixes and suffixes, we find a few circumfixes, combinations of a prefix and a suffix. Example are the Dutch affix combination ge- . . . -te and the German affix combination Ge- . . . -e that serve to create nouns that denote collectives and nominalizations, as in: (19)

Dutch German:

boom ‘tree’ boef ‘crook’ Berg ‘mountain’ lach(en) ‘to laugh’

ge-boom-te ‘collection of trees’ ge-boef-te ‘collection of crooks’ Ge-birg-e ‘mountains’ Ge-lach-e ‘laughter’

Such circumfixes emerged due to the occurrence of an affix combination in a set of complex words. The prefix be- and the suffix -ig co-occur in a number of Dutch and German words (there are no adjectives or verbal stems edig and scho¨nig):

Principles of Word Formation

(20)

Dutch German

eed ‘oath’ scho¨n ‘beautiful’

be-ed-ig(en) ‘to swear in’ be-scho¨n-ig(en) ‘to make more beautiful’

Thus, these affixes are not only used separately, but also in combination, to derive verbs from nouns and adjectives. Suffixes tend to determine the syntactic category of the derived words that they create. For instance, German -ung creates nouns from verbs, and hence, -ung determines that the resulting complex word is a noun (e.g., bemerk(en) ‘to remark’ – Bemerkung ‘remark’). Moreover, nominalizing suffixes also determine nominal gender: All words in -ung such as Bemerkung are feminine. Prefixes vary in this respect: Most prefixes are category-neutral, and do not change the category of the base word. For instance, both pleasant and un-pleasant are adjectives. This suggested the generalization for Germanic compounds and affixed words that the rightmost constituent determines the category of the complex word as a whole: the so-called Right-hand Head Rule (Williams 1981). However, there are exceptions to this rule: Verbalizing prefixes may change the category of their base word, as can be seen in the English verbs be-jewel and en-large, with a nominal and an adjectival base respectively. The phonological form of Germanic affixes with a proto-Germanic origin may betray their status of bound morpheme. Whereas words of lexical categories must have at least one full vowel (i.e., no schwa), grammatical morphemes, including affixes, do not have to comply with this requirement. For instance, the prefix be- /bə/ has schwa as its only vowel, just like the suffix -er /ər/. Germanic suffixes divide into two classes: Some of them form one domain of syllabification with the stem, whereas others form a phonological word of their own, and are syllabified separately. The first type is referred to as “cohering suffix”, the second type as “noncohering suffix”. A minimal pair to illustrate this comes from Dutch (the dot in the phonetic transcription denotes a syllable boundary): (21)

rod-ig ‘redd-ish’ [ro.dəx] versus rood-achtig ‘redd-ish’ [rot.ɑx.təx]

Noncohering suffixes, just like the constituents of compounds, are phonological words by themselves, even though they are not independent grammatical words. This shows that there might be an asymmetry between the morphological structure and the phonological structure of Germanic complex words. This asymmetry can also be observed in the phenomenon of word-internal gapping: Parts of complex words may be omitted as a kind of ellipsis provided that the deleted part is a phonological word of its own. This is illustrated here by data from German which illustrate that in this respect derived words with a noncohering suffix behave just like compounds:

247

248

GEERT BOOIJ

(22)

derived words: compounds:

frucht- und freudlos ‘fruit- and joyless’ Arbeits- und Sozialverhalten ‘labor- and social behavior’; Nicht nur im Schul- sondern auch in Berufsleben ‘not only in school-, but also in professional life’

As mentioned in Section 11.1, complex words may also have been formed by means of vowel alternation, referred to as Ablaut or apophony. This system is an old Indo-European system, and gradually disappeared from Germanic. Many deverbal nouns have once been formed by means of Ablaut, sometimes in combination with suffixation: (23) Dutch German

verb stem bind ‘to bind’ vind ‘to find’ bind ‘to bind’

noun band ‘book volume’ vond-st ‘finding’ Bund ‘union’ Band ‘book volume’

These Ablaut patterns which are also well-known from the past tense and past participle forms of the so-called strong verbs, can no longer be used productively. However, when a new particle verb is formed with a verb with a corresponding ablauting noun, this noun can also be used for the nominalization of such particle verbs (Booij 2015): (24)

Dutch

komen ‘to come’- komst

German

sehen ‘to see’ – Sicht

aan-komen ‘to arrive’ – aankomst an-sehen ‘to look at’ – An-sicht

These ablauting nouns must be stored lexically, and can be used as the head of compounds with a particle as modifier. Thus, such compounds can function as the nominalization of particle verbs.

11.3.1

Nonnative Affixation

The word formation system of Germanic languages has been influenced strongly by contact with the classical languages Greek and Latin, and with French. The influence of Latin was due to its role as lingua franca in Europe in the domains of science and religion. The specific influence of French on English dates back to the eleventh century when the Normans conquered England. For centuries, French was a dominant language in European culture and politics, and thus affected all Germanic languages. Hence, we see many nonnative affixes in Germanic languages as the effect of borrowing. Borrowing did not take place of affixes in isolation: Complex words were borrowed, and once a sufficient number of words with a certain affix had been borrowed, the affix could be identified on the basis of a systematic form-meaning correspondence in a set of similar words, and also used productively in the borrowing languages. For instance, French

Principles of Word Formation

deadjectival nouns ending in -ite´ such as liberte´ ‘freedom’ were borrowed in Dutch, English, and German, and subsequently, new nouns of this type were coined. The suffix itself was slightly adapted as to phonological form: -iteit in Dutch, -ity in English, -ita¨t in German, and -itet in Norwegian: (25)

actief active aktiv aktiv

activiteit activity Aktivita¨t aktivitet

These nonnative suffixes are shared in some form or another by most Germanic languages, and often carry the main stress of the complex words that they created. Germanic languages do not only feature a substantial number of nonnative affixes, but also nonnative prefixes, in particular of Greek and Latin origin, such as: (26)

anti-, co-, contra-, de-, ex-, hyper-, infra-, meta-, neo-, non-, pre-, pro-, sub-, super-, turbo-, ultra-

A remarkable property of many of these affixes is that they do not combine with all kinds of base words, but only with words of nonnative origin. For instance, the Dutch suffix -iteit is mainly combined with nonnative adjectives, and the same applies to its English counterpart -ity witness the ungrammaticality of greenity: (27)

absurd ‘absurd’ stabiel ‘stable’ groen ‘green’

absurd-iteit ‘absurd-ity’ stabil-iteit ‘stabil-ity’ *groen-iteit ‘green-ity’

The effect of this constraint on nonnative suffixation is that in complex words with more than one suffix, native suffixes will appear peripherally to nonnative suffixes. So we can coin a word like absurd-iteits-loos ‘absurdity-less’ in Dutch, but einde-loos-iteit ‘end-less-ity’ is ill-formed. However, this constraint on nonnative affixes does not apply to the same extent to nonnative pan-Germanic (or even pan-European) prefixes such as ex- and super- that are easily attached to words of Germanic origin: (28)

Dutch

ex-vrouw super-rijk

‘ex-wife’ ‘super-rich’

A more recent borrowing in English is the German prefix u¨ber- with the meaning ‘to an excessive degree.’ Subsequently, this prefix was borrowed in Dutch from English: (29)

English: Dutch:

u¨ber-cool, u¨ber-gay, u¨ber-nerd, u¨ber-parent, u¨ber-hacker u¨ber-nicht ‘excessive homosexual’, u¨ber-ouder ‘u¨berparent’, u¨ber-lijp ‘very smart’

249

250

GEERT BOOIJ

This type of borrowing reflects the fact that English has taken over from French as the main source language for borrowings in Germanic languages, and worldwide.

11.3.2

Affix Substitution

Derivation of words in which affixes are involved may also takes place by means of affix substitution instead of by affix addition. For instance, nominalization of Dutch verbs of which the stem ends in the suffix -iseer can be performed by replacing the part -eer by the suffix -atie: (30)

kanal-is-eer ‘to canalize’ organ-is-eer ‘to organize’ stabil-is-eer ‘to stabilize’

kanal-is-atie ‘canalization’ organ-is-atie ‘organization’ stabil-is-atie ‘stabilization’

Consider also the following nouns in English, where a noun ending in -ism that denotes an ideology or disposition correlates with a noun ending in -ist denoting a person related to that ideology or disposition: (31)

aut-ism Marx-ism solips-ism

aut-ist Marx-ist solips-ist

These words exhibit a systematic form-meaning correspondence without one of them forming a part of the other. Hence, they cannot be described in terms of affix addition. These patterns arose due to the massive borrowing of nonnative complex words in Germanic languages. We can account for such systematic patterns by means of schemas that specify paradigmatic relations between classes of complex words. For instance, the schema for the words in (31) could have the following form (Booij and Masini 2015): (32)

form: meaning:

[x-ism]Ni [Ideology/Disposition Y]SEMi

≈ [x-ist]Nj [Person involved in SEMi]SEMj

The symbol ≈ denotes the paradigmatic relationship between two morphological constructions, and SEM stands for the meaning of a word or constituent. The symbol x is a variable for a phonological string, and Y is a variable for a meaning component of the individual words ending in -ism. This schema specifies that the meaning of the word ending in -ist is a compositional function of the word ending in -ism, even though it has the same degree of formal complexity.

11.4 Conversion A very productive process of word formation in Germanic languages is conversion, the process by which a word is going to belong to another

Principles of Word Formation

syntactic category without any overt morphological marking. English is famous for its great potential of converting nouns into verbs, as illustrated here by some recent coinings such as to skype and to whatsapp, derived from the nouns skype and whatsapp respectively. Conversion of nouns into verbs is also productive in other Germanic languages. Here are a number of conversion of various types of word into verbs from Frisian: (33)

keal ‘calf’ grean ‘groen’ hinne-en-wer ‘to and fro’ ien-en-tweintig ‘twentyone’

keal-je ‘to calve’ grean-je ‘to become green’ hinne-en-wer-je ‘to go to and fro’ ien-en-tweintig-je ‘to play twentyone’

The verbs derived by conversion always belong to the default inflectional class. In Frisian, verbs with an infinitive in -je form the default conjugation. Dutch conversion verbs are never conjugated as strong, ablauting verbs, but always as weak verbs (the default conjugation), with the past tense stem ending in -te or -de. Thus, we get minimal pairs like the following: (34)

prijs ‘to praise’ prijs ‘to price’ (conversion of prijs ‘price’)

past tense stem: prees past tense stem: prijs-de

Metalinguistic usage may also lead to conversion of nonlexical categories, as in German: das Nein ‘the no’ and kein Wenn und Aber ‘no when and if’. There is an extensive debate in the linguistic literature as to how to describe conversion, because there is no overt morphological marking of the category change. For that reason, some linguists have postulated a zerosuffix that is assigned category-changing power. Another solution is to consider this kind of word formation as being paradigmatic in nature. For instance, the verb to skype is paradigmatically related to its base noun skype. This can be expressed by a schema that specifies this paradigmatic relationship: (35)

form: meaning:

[x]Ni SEMi



[x]Vj [Event with SEMi involved]SEMj

where x stands for the phonological string of the related words, SEMi for the meaning of the noun, and SEMj for the meaning of the verb. The nonfinite forms of Germanic verbs, participles and infinitives, can be used systematically as adjectives and nouns respectively. The participles in prenominal position are inflected as prenominal adjectives. (36)

German:

Die verkauft-e Braut The.F E M . N O M . S G sold-F E M . N O M . S G bride.F E M . S G ‘The sold bride’ Das sing-end-es Ma¨dchen.N E U T . S G The.N E U T . N O M . S G sing-ing-N E U T . N O M . S G . girl ‘The singing girl’

251

252

GEERT BOOIJ

Dutch:

Het eten staat op tafel The.N E U T . S G eat-I N F . N E U T . S G stands on table ‘The food is on the table’

In German, present participles can also be used as nouns, and even then they are inflected as adjectives, as illustrated by the phrases ein Vorsitz-end-er ‘a chairman’ and eine Vorsitz-end-e ‘a chairwoman’, in which the present participle Vorsitz-end- ‘chairing’ is used as a noun, but inflected as a prenominal adjective. Prenominal adjectives can be used as nouns in elliptic contexts, as in Danish de gamle (mennesker) ‘the old (people)’, and en bærbar (computer) ‘a portable (computer)’ (Go¨tzsche 2016). Specific syntactic contexts may coerce words into another word class (Audring and Booij 2016), as shown by the following example from Dutch: (37)

Nederland van smerig tot schoon Netherlands from dirty to clean ‘Netherlands from being dirty to being clean’

Here the adjectives smerig and schoon are coerced into nouns, as they are the complements of prepositions. The particle verb construction may also have the effect of converting a noun or an adjective into a verb, as in English to buckle up and to pretty up, derived from the noun buckle and the adjective pretty respectively. This illustrates once more that conversion may be dependent on specific syntactic or morphological constructions.

11.5 Reduplication Reduplication is a widespread process of word formation in the languages of the world. In its simplest form reduplication is total reduplication, in which a word is copied completely. In other words, this type of word formation consists of the doubling of a word. The copying configuration often evokes semantic notions such as intensity, high degree, and repetition. In Germanic languages reduplication is a marginal phenomenon. Repetition to express intensity or repetition does occur on the syntactic level (as in He walked, and walked, and walked), but not so much on the word level. However, there are two classes of exceptions. The first type is the doubling of nouns in order to express the notion ‘real, very good specimen of’ (Gomeshi et al. 2004): (38)

English: Dutch:

salad-salad ‘very good salad, salad as it should be’ coffee-coffee ‘very good coffee, real coffee’ meisje-meisje ‘a very girlish type of girl, a real girl’ vakantie-vakantie ‘a real holiday without work’

Principles of Word Formation

An exceptional language in this respect is Afrikaans. This sisterlanguage of present-day Dutch was heavily influenced by Malay, the native language of servants/slaves in South Africa, an Austronesian language. Austronesian languages exhibit a lot of reduplication. In Afrikaans, reduplication has become a very productive word formation process and applies to all kinds of base words. Here is a sample: (39)

knip ‘to cut’ klop ‘to knock’ bal ‘ball’ brul ‘to roar’ dik ‘thick’ tien ‘ten’

knip-knip ‘to cut continously’ klop-klop ‘to knock continuously’ bal-bal ‘ball game’ brul-brul ‘roaring’ dik-dik ‘very thick’ tien-tien ‘ten by ten’

A second Germanic language with reduplication is Yiddish. Examples are kukn-kukn ‘to look’ and loyf-loyf ‘to run’. In Yiddish, reduplication can be combined with insertion of an emphatic infix, as in mayster-shebe-mayster ‘master’ and vunder-shebe-vunder ‘miracle’. Instead of full reduplication, the second word can also be adapted by replacement of its initial consonants by shm-, thus evoking a disparaging interpretation, as in gold-shmold ‘sort of gold’ and kidnep-shidnep ‘sort of kidnapping’. This process of creating words with a disparaging flavor has been borrowed in American English. The second type of reduplication is exemplified by the following compounds: (40)

German

Dutch

Wirr-warr ‘tangle’ Tick-tack ‘clock’ tag-ta¨glich ‘daily’ Bla-bla ‘empty talk’ dag-dagelijks ‘daily’ hotel-de-botel ‘madly’ holder-de-bolder ‘helter-skelter’ wisse-wasje ‘trifle’

In these words there is some form of repetition in form, but no complete doubling. These kinds of words are used as expressive language, or by children (example Ticktack).

11.6 Prosodic Morphology Besides the formal mechanisms discussed so far, there is another way of coining new words, in which the base word is reduced to one or two syllables. This is used for coining endearment forms of first names. Here are some examples from Dutch:

253

254

GEERT BOOIJ

(41)

Albert Emmy Herry Suzanne

Al Em Her Suus

This type of morphology is called prosodic morphology because a prosodic template determines the shape of the endearment form: It consists of one syllable, closed by a consonant. The consonant is taken from the first syllable of the base word, if available, and otherwise from the onset of the second syllable. Like in the case of conversion, we can express this regularity by means of a schema with a paradigmatic relationship between the base word and the shortened word: (42)

form: meaning:

[xVCy]Ni Namei



[xVC]Nj [Namei with Endearment]SEMj

For instance, the name Albert /ɑlbərt/ conforms to schema xVCy (x = ø, V = / ɑ/, C= /l/, and y = /bərt/. Hence we get xVC = Al /ɑl/ as the endearment form. In English the truncated hypocoristic XCV output may be affixed with -y, as in Patty and Tristy, and daffy for daffodil (Lappe 2007), and in German with -i, as in Ossi ‘former inhabitant of East Germany’ and Wessi ‘former inhabitant of West Germany’. In Dutch, -o is used for this purpose: Brab-o ‘inhabitant of Brabant’. Prosodic morphology with monosyllabic outputs is also applied to common nouns, as in Dutch mees < meester ‘teacher’ and bieb < bibliotheek ‘library’.

11.7 Abbreviation and Blending Prosodic morphology is one way of creating shorter words from longer ones, that is, it is a form of abbreviation. Abbreviation is a common process of word formation in Germanic languages. The effect is that long words become much shorter, which will give processing advantages. On the other hand, they may create difficulties because the meaning of an abbreviated word cannot be deduced from its form. That is, its meaning can only be recovered on the basis of its paradigmatic relation to another word or phrase. In Germanic languages we find lots of abbreviations of common nouns such as: (43)

Dutch:

German

fysio-therapie ‘physiotherapy’ > fysio doctoraal-examen ‘MA exam’ > doctoraal kandidaats-examen ‘bachelor exam’ > kandidaats scheids-rechter ‘referee’ > scheids Abitur ‘high school final exam’ > Abi Demonstration ‘demonstration’ > Demo Tachometer ‘id.’ > Tacho Universita¨t ‘university’ > Uni

Principles of Word Formation

English

Scandinavian

celebrity > celeb microphone > mike prefabricated > prefab automobil ‘car’ > bil

Names of institutions lend themselves very well to abbreviation in the form of acronyms, such as Frisian FNP standing for Fryske Nasjonale Party ‘Frisian National Party’. This name has the form of a phrase, but as an acronym it behaves as a noun. Examples of very common and widely used German abbreviations are the following: (44)

Kinder-tages-sta¨tte ‘children-day-centre’ > Kita Aus-zu-bildend-er ‘trainee’ > Azubi Vereinigte Dienstleistungs-gewerkschaft ‘united service union’ > Verdi

Again, it is obvious that these abbreviations make the expression of such meanings much shorter. In blending, two words are blended by combining the first part of one word with the second part of another. A classic example is English smog from smoke and fog. Brunch < breakfast + lunch is another well-established one.

11.8 Numerals Germanic languages construct numerals in a number of ways. Cardinal numerals above 20 show systematic properties. Consider the following numerals of Dutch, English, and German: (45) 6 16 60 61 106

Dutch zes zes-tien zes-tig een-en-zestig honderd (en) zes

English six six-teen six-ty sixty-one hundred (and) six

German sechs sech-zehn sech-zig ein-und-sechzig hundert-(und)-sechs

Most numerals of Dutch and English, and of all Germanic languages are complex linguistic expressions, formed by a recursive system of rules that enables the language user to form an (in principle) infinite set of numerals. In Dutch, as in English and German, all numerals above the number 12 are complex expressions. The numerals for 16 in (45) have the shape of a compound consisting of two lexeme constituents, whereas the words for 60 have the shape of a derived word with a suffix. The words for 61 and 106 have the appearance of phrases, formed by means of coordination with a conjunction. As to the expression of cardinal numbers from 21 to 99, the Germanic languages can be divided into two groups. Afrikaans, Danish, Dutch,

255

256

GEERT BOOIJ

Faroese, Frisian, and German place the lower digit before the higher one, whereas in English and Swedish, the order is reversed. Before 1951, Norwegian complex numerals like ‘sixty-one’ etc. followed the pattern with the lower number before the ten, for instance to-og-femti ‘52’. In 1951, a reform was carried through, prescribing the opposite, Swedish and English order, hence femti-to (Askedal 2016). Even though the words for 61 and 106 have the form of phrases, they can function as bases of word formation, for the formation of ordinal numerals by means of suffixation: (46)

Dutch een-en-zes-tig-ste, German ein-und-sechzig-ste ‘one-and-sixty-th, sixty-first’ Dutch honderd (-en) zes-de, German hundert-und-sechste ‘hundred (and) six-th’

The formation of words for numerals is a complicated subsystem of its own for the various Germanic languages, and for details, the reader is referred to the literature at the end of this chapter.

11.9 Sources of Information A general study of the Germanic lexicon including its morphology is Harbert (2007: Chapter 2). There is a chapter on word formation for each of the Germanic languages German, English, Dutch, Frisian, Yiddish, Faroese, Danish, Norwegian, Swedish and Icelandic, as spoken in Europe, in Mu¨ller et al (eds.). For Dutch, Afrikaans, and Frisian, there is a website called Taalportaal, with detailed information in English on the morphology of these languages, and further references: www.taalportaal.org. For German, see the website of the Institut fu¨r deutsche Sprache, https://grammis.ids-man nheim.de/, with extensive information on morphology, and further references. The best reference work for English word formation is Bauer et al. (2013). The Duden volume on Grammatik is the main handbook on the grammar of German, with extensive information on word formation. Hentschel (2016) is a volume on various aspects of German word formation. The reference work for Afrikaans is Carstens and Bosman (2014), for Dutch morphology it is Booij (2019), and for Frisian morphology Hoekstra (1998). Josefsson (1997) is a theoretical study of Swedish word formation. Bandle et al. (2002/ 2005) contains chapters on various synchronic and diachronic aspects of word formation in Nordic languages. Thra´insson et al. (2004) is a reference grammar of Faroese with information on its word formation. Jacobs (2005) is a good introduction to the grammar of Yiddish.

Principles of Word Formation

References Askedal, J. O. 2016. “Norwegian.” In P. O. Mu¨ller, I. Ohnheiser, S. Olsen, and F. Rainer (eds.), Word Formation. An International Handbook of the Languages of Europe, Vol. 4. Berlin and New York: Mouton De Gruyter: 2525–2554. Audring, J. and G. Booij 2016. “Cooperation and coercion,” Linguistics 54: 617–637. Bandle, O., K. Braunmu¨ller, E. H. Jahr, A. Karker, H. P. Naumann, and U. Teleman (eds.), 2002/2005. The Nordic Languages: An International Handbook of the History of the North Germanic Languages. Berlin: Mouton De Gruyter, Vol. 1 (2002), Vol. 2 (2005). Bauer, L., R. Lieber, and I. Plag 2013. The Oxford Reference Guide to English Morphology. Oxford University Press. Booij, G. 1993. “Against split morphology.” In G. Booij and J. van Marle (eds.), Yearbook of Morphology 1993. Dordrecht: Kluwer: 27–49. Booij, G. 2010. Construction Morphology. Oxford University Press. Booij, G. 2015. “The nominalization of Dutch particle verbs: Schema unification and second order schemas,” Nederlandse Taalkunde 20: 285–314. Booij, G. 2019. The Morphology of Dutch. 2nd edition. Oxford University Press. Booij, G. and F. Masini 2015. “The role of second order schemas in word formation.” In L. Bauer, L. Ko¨rtve´lyessy, and P. Sˇtekauer (eds.), Semantics of Complex Words. Dordrecht: Springer: 47–66. Carstens, W. and N. Bosman (eds.) 2014. Kontemporeˆre Afrikaanse taalkunde. Pretoria: Van Schaik. Duden 2009. Die Grammatik. Mannheim, Wien and Zu¨rich: Dudenverlag. Gaeta, L. and B. Schlu¨cker (eds.) 2012. Das Deutsche als kompositionsfreudige Sprache. Strukturelle Eigenschaften und systembezogene Aspekte. Berlin and New York: Mouton de Gruyter. Ghomeshi, J., R. Jackendoff, N. Rosen, and K. Russell 2004. “Contrastive focus reduplication in English (the salad-salad paper),” Natural Language and Linguistic Theory 22: 307–357. Go¨tzsche, H. 2016. “Danish.” In P.O. Mu¨ller, I. Ohnheiser, S. Olsen, and F. Rainer (eds.), Word Formation: An International Handbook of the Languages of Europe, Vol. 4. Berlin and New York: Mouton de Gruyter: 2505–2524. Harbert, W. 2007. The Germanic Languages. Cambridge University Press. Hentschel, E. (ed.) 2016. Wortbildung im Deutschen. Tu¨bingen: Narr Francke Attempto Verlag. Hoeksema, J. 2012. “Elative compounds in Dutch: Properties and developments.” In G. Oebel (ed.), Intensivierungskonzepte bei Adjektiven und Adverben im Sprachenvergleich / Crosslinguistic comparison of intensified adjectives and adverbs. Hamburg: Verlag dr. Kovacˇ: 97–142. Hoekstra, J. 1998. Fryske wurdfoarming. Ljouwert: Fryske Akademy.

257

258

GEERT BOOIJ

Hopper, P. and E. Traugott 1993. Grammaticalization. Cambridge University Press. Hu¨ning, M. and G. Booij 2014. “From compounding to derivation. The emergence of derivational affixes through ‘constructionalization’,” Folia Linguistica 48: 579–604. Jacobs, N. G. 2005. Yiddish. A Linguistic Introduction. Cambridge University Press. Josefsson, G. 1997. On the Principles of Word Formation in Swedish. Lund: Lund University Press. Lappe, S. 2007. English Prosodic Morphology. Dordrecht: Springer. Los, B., C. Blom, G. Booij, M. Elenbaas, and A. van Kemenade 2012. Morphosyntactic Change: A Comparative Study of Particles and Prefixes. Cambridge University Press. Mu¨ller, P.O., I. Ohnheiser, S. Olsen, and F. Rainer (eds.) 2016, Word Formation. An International Handbook of the Languages of Europe. Berlin and New York: Mouton De Gruyter. Thra´insson, H., H. P. Petersen, J. Jacobsen, and Z. S. Hansen 2004. Faroese. An Overview and Reference Grammar. To´rshavn: Føroya Fro´ðskaparfelag. Williams, E. 1981. “On the notions ‘lexically related’ and ‘head of a word’,” Linguistic Inquiry 12: 245–275.

Chapter 12 Grammatical Gender in Modern Germanic Languages Sebastian Kürschner

12.1 Grammatical Gender The term ‘grammatical gender’ is used for specific sorts of noun classes. These noun classes are not usually indicated by the morphological behavior of the noun itself, but are defined by agreement, reflected in the inflectional morphology of associated items such as articles, adjectives, or pronouns. Gender is thus grammatical variation “reflected in the behavior of associated words” (Hockett 1958: 231). The noun is the controller of the invariant gender information, and the associated elements in the noun phrase or in surrounding anaphoric elements are the targets reflecting the gender information in their variable inflectional morphology. Some examples of gender agreement in articles and adjectives from German are found in (1): (1)

Definite große Straße, die . . . a. die the.F big street, that.F . . . ‘the big street that . . . ’ b. der große Platz, der . . . the.M big square, that.M ‘the big square that . . . ’ c. das große Fest, das . . . the.N big party, that.N ‘the big party that . . . ’

Indefinite eine große Straße, die . . . a-F big-F street, that.F ‘a big street that . . . ’ ein großer Platz, der a.N O N -F big-M square, that.M ‘a big square that . . . ’ ein großes Fest, das . . . a big.N party, that.N ‘a big party that . . . ’

While gender is generally reflected in associated words, the Scandinavian languages also grammaticalized a marker of gender as a noun clitic which in some languages turned into an inflectional suffix. In Danish, e.g., an unmodified noun is marked for definiteness in a noun

260

SEBASTIAN KÜRSCHNER

suffix (2a–2b), while definiteness is marked in an article word if the noun is premodified by an adjective (2c–2d). Since definiteness markers are main markers of gender in Germanic languages, gender has become part of noun morphology in the definiteness markers of unmodified nouns, whereas it is reflected in associated words only in premodified nouns.1 (2)

Definite a. gaden street-D E F . CG ‘the street’ b. landet country-D E F . N ‘the country’ gade c. den store the.CG big-D E F street ‘the big street’ store land d. det the.N big-D E F country ‘the big country’

Indefinite en gade a.CG street ‘a street’ et land a.N country ‘a country’ en stor gade a.CG big street ‘a big street’ et stort land a.N big-N country ‘a big country’

Articles are generally used for determining the adnominal gender of a noun, but gender is also reflected in other items showing agreement with nouns. A prominent target is found in pronouns. Table 12.1, adapted from Audring (2010: 697), presents an overview of the modern Germanic languages according to their gender systems as reflected in attributive elements (adnominal gender) and pronouns (pronominal gender).2 Gender is inherited from an early Indo-European noun classification. Reconstructional linguistics believes that it arose as a semantically based classification system going back to a distinction of two classes probably based on animacy, agentivity, or the capability of subject marking (MeierBru¨gger 2002: 80), and developed into a three-way distinction in later Proto-Indo-European.3 These three genders, masculine (M), feminine (F), and neuter (N), existed in Proto-Germanic and can still be found in some modern Germanic languages, e.g., German or Icelandic. Gender turned from a classification assigned to nouns for distinguishing between differing semantic classes into a grammaticalized system in which the semantic basis was being blurred and new semantic and formal assignment criteria became more relevant. In the further development new transparent semantically based classifications were added to parts of the gender 1 2

See Nübling, Chapter 10, for more detailed information on definiteness marking in Germanic. For a more detailed overview including a more fine-grained selection of target groups, e.g., different pronoun types (personal, demonstrative, possessive, interrogative, relative, reflexive), attributive and predicative adjectives, numerals, and the difference between grammatical and semantic or referential pronominal gender, see Duke (2009: 82).

3

Leiss (2000) presents a view of late Indo-European gender based on a paradigm of alternative quantifications, namely (1) collective and abstract semantics in feminines, (2) countable, singulative, and concrete in masculines, and (3) masses in neuters.

Grammatical Gender in Modern Germanic Languages

Table 12.1 Overview of adnominal and pronominal genders in Germanic languages (adapted from Audring 2010: 697) Language

Adnominal gender

Pronominal gender

German Yiddish Luxembourgish Faroese

M



F

Icelandic Norwegian (Nynorsk and radical Bokmål)

N M

Norwegian (moderate Bokmål)

N CG

Swedish Danish

F

CG −

Dutch West Frisian English



Afrikaans

system (e.g., the marking of males and females that is still transparent today; cf. Wurzel [1986] on repeated classification). The grammaticalized system is reflected in Modern Germanic adnominal gender. While we regard the agreement system as ‘grammatical gender’, we will follow Dahl (2000) in distinguishing between ‘lexical’ and ‘referential gender’. Referential gender refers to the referent of a noun phrase and applies, e.g., in sex-specific pronoun use. For example, the word doctor can be referred to by feminine or masculine pronouns in virtually any of the Germanic languages depending on the actual referent’s sex. Lexical gender, by contrast, is a lexical property of the noun. For example, there is no referential reason to use a masculine marker with a spoon in German (der Lo¨ffel), but the gender is part of the lexical information. Lexical gender can be traced back to semantic and formal gender assignment (see Section 12.3) and is stored in the lexical entry. While adnominal gender usually reflects lexical gender, pronominal gender is more open to referential uses, resulting in variable pronominal gender e.g., for the Dutch word meisje ‘girl’ which is lexically N (cf. het meisje ‘the girl’ – formally assigned because of the diminutive suffix -je) but referentially female, triggering variation between lexical (het ‘it’) and referential (zij ‘she’) pronoun use. English is an example of a language that lost lexical (and adnominal) gender and developed a purely referential pronominal gender system.

261

262

SEBASTIAN KÜRSCHNER

While languages like German retained a three-gender distinction up to their modern form (see Table 12.1), some other Germanic languages – like Danish – reduced their lexical gender system to a two-way distinction, mostly by a merger of feminine and masculine nouns into a common gender (CG), e.g., Danish or Dutch. Some of the modern Germanic languages even totally lost lexical gender in the course of their history, namely English and Afrikaans. We will discuss these different types of lexical gender systems in Section 12.2. Table 12.1 also shows that three pronominal gender markers are retained even when lexical gender is reduced, and the Mainland Scandinavian languages even enlarged the pronominal system by developing an additional pronominal marker (Section 12.4). In languages with gender reduction, this results in a mismatch for which different solutions can be found. Section 12.3 looks at factors relevant in the assignment of genders to nouns, and Section 12.4 looks into differing developments of adnominal (lexical) and pronominal (lexical or referential) gender more specifically, discussing the semantic reorganization of gender systems. Gender is strongly connected with the inflectional morphology of nouns in Germanic languages. For example, declension classes in modern Germanic languages are often predictable based on gender information and vice versa. Section 12.5 treats aspects of the interaction of gender and noun inflection. A short conclusion is found in Section 12.6.

12.2 Lexical Gender Systems in Germanic 12.2.1 Three-Gender Systems Three-gender systems exist in many standard and nonstandard varieties of Germanic today. Concerning the West-Germanic languages, Standard German retained three genders, just as most dialects of German, including all the High German and many Low German dialects (Section 12.2.2). Luxembourgish and Yiddish, two languages based on High German dialects, retain three genders as well (with exceptions in North-Eastern varieties of Yiddish, Section 12.2.2). In Dutch, where the standard variety reflects a two-gender system, many dialects are still characterized by the old three-gender system, especially in the Southern part of the Dutchspeaking area. East Frisian Saterlandic retains three genders, just as the mainland varieties of North Frisian, while North Frisian insular varieties reduced genders. In North Germanic, three genders are retained in the group of the West Nordic so-called Insular Scandinavian languages, consisting of Faroese and Icelandic. In the group of Mainland Scandinavian languages, the more conservative Norwegian standard variety of New Norwegian (Nynorsk) reflects the old three-gender system while the more progressive standard variety of Dano-

Grammatical Gender in Modern Germanic Languages

Norwegian (Bokmaº l) shows variability between three- and two-gender systems (Section 12.2.2). While Standard Danish and Standard Swedish reflect a merger between masculine and feminine gender, many dialects of the Mainland Scandinavian languages still retain the classic three-gender system. The fact that some of the Germanic languages retained all three genders while others reduced or even totally lost lexical gender, gives reason to ask if languages with three-gender systems have functionalized gender to a stronger extent than languages reducing genders. Reasons for a linguistic function of gender in Modern Germanic have been sought on various grounds. First, gender is useful in reference tracking in all Germanic languages, since ambiguities in anaphoric reference to nouns can be solved based on the gender information. This is useful when multiple nouns can be referred to, see (3) from German. However, with only three genders available, disambiguation is frequently likely to fail because of multiple nouns with identical genders (see example 4), so it is doubtful whether this functionality justifies the high learning and memorability costs of lexical gender. (3)

Die Katze la¨uft dem Hund hinterher. Sie ist schneller. the.F cat runs the.M dog . . . behind she.F is faster ‘The cat is running behind the dog. It is faster.’

(4)

Die Katze la¨uft der Maus hinterher. Sie ist schneller. the.F cat runs the.F mouse behind she.F is faster ‘The cat is running behind the mouse. It is faster.’

Ronneberger-Sibold (2007) suggests another reason for retaining a strong gender system. According to her view, the retention of three genders in German is connected to a morphosyntactic property highly characteristic of German, namely the use of framing constructions. In such constructions, an element providing morphosyntactic information about a corresponding, morphosyntactically agreeing element provides the start of a frame of optional length that does not end before the element in agreement is found. German is shaped by such framing constructions both in its verb phrases and noun phrases, with gender playing an important role in noun phrases. NPs usually start with a determiner and end with the corresponding noun, and they can integrate further NPs shaped by the same characteristics. See (5) for an example where the frame opened by eine (nominative case) is closed by feminine Wendung, and the included frame opened by dem (dative case) is closed by masculine Leser. (5)

Eine dem geschulten Leser ha¨ufig begegnende Wendung a.F the.M/N trained reader.M often encountering collocation.F ‘A collocation that the trained reader often runs across’

263

264

SEBASTIAN KÜRSCHNER

Together with case and number information, the stable gender information plays an important role in German NP framing. Since German typologically developed into a framing language, Ronneberger-Sibold argues that gender cues are highly important, motivating the retention of a threegender distinction. Other functional aspects of gender systems that were suggested include, e.g., processing advantages (aid in lexical retrieval both in production and perception, cf. Ko¨pcke and Zubin 2009) and the disambiguation of homonyms (German der Kiefer ‘the jaw’ versus die Kiefer ‘the pine tree’).

12.2.2 Two-Gender Systems In Germanic, nearly all varieties that reduced the gender system to two genders retain the neuter gender and perform a merger between masculines and feminines. This is in sharp contrast with some other IndoEuropean languages which reduced genders, e.g., with Romance languages retaining masculine and feminine gender (e.g., French, Italian). Exceptions in Germanic are found, e.g., in the North-Frisian dialect of Fo¨hring-Amring where feminines and neuters merge against masculines (Wahrig-Burfeind 1989, Nu¨bling 2017: 198–203), or in North-Eastern varieties of Yiddish, where a feminine-masculine distinction is retained, but a subdivision of the feminine into three subgenders adds complexity to the gender system (see Jacobs et al. 2002, 402). In the West Germanic group, the merger of masculines and feminines is found in Standard Dutch and West Frisian. While southern Low German dialects retain three genders, some northern Low German dialects also reflect the merger, e.g., in the region of East Friesland in the NorthWestern part of Germany, bordering with Dutch, and in the Slesvig region close to Denmark (see Wahrig-Burfeind 1989). In North Germanic, the standard languages of Danish and Swedish reflect the merger, and the Norwegian standard variety of Dano-Norwegian mainly shows the twogender system as in Danish and Swedish. In variants of Dano-Norwegian, however, the distinction between masculine and feminine has been reintroduced by re-establishing feminine markers in definiteness and adjective inflection. These are mainly used with nouns denoting individuals (persons and some animals), and according to Braunmu¨ller (2000: 27) also with “a couple of more or less frequent words . . ., the feminine gender of which is commonly well known by native Norwegians with a good command of their local dialects.” Examples of such words are sol ‘sun’, tid ‘time’, or bok ‘book’. Example (6) shows the variation between variants of Bokmaº l reflecting feminine gender (so-called radical variants of Bokmaº l, see 6a) and variants using only the common gender (so-called moderate variants, 6b). Given the intermediary status of the Bokmaº l system between a three-gender and a two-gender system, Braunmu¨ller (2000) classifies it as a hybrid-gender system.

Grammatical Gender in Modern Germanic Languages

(6)

Definite

Indefinite

a. den lille boka ei lita bok the.CG little book-D E F . F a.F little-F book ‘the little book’ ‘a little book’ b. den lille boken en lille bok the.CG little book-D E F . C G a.CG little book We will take examples from West Germanic (Dutch) and, after that, from North Germanic (Danish and Swedish) to study the backgrounds of the merger. In Modern Dutch, the two genders are marked in the singular morphology of definite articles and in strong adjective declension (7). (7)

a. het boek the.N book ‘the book’ b. een goed boek a good book ‘a good book’

de tafel the.CG table ‘the table’ een grote tafel a large-CG table ‘a large table’

In Old Dutch, the demonstrative – the form from which the definite article was grammaticalized – still differed for masculine (theˆ / thie) and feminine gender (thiu [?]) in the nominative singular (Quak and van der Horst 1997: 53, question mark original), in addition to neuter that. Toward Middle Dutch, unstressed vowels were reduced. The vowel quality of feminine and masculine articles merged, cf. emphatic die and unemphatic de versus neuter emphatic dat and unemphatic ’t. At this point, a difference between masculines and feminines is still found in the nonnominative cases, e.g., in the accusative (M dien / den versus F die / de). However, the case system was lost in the further development, and only the merged form survived (the unemphatic form being generalized). In strong adjective declension, however, the Middle Dutch system showed no consistent merged use, yet, and sound change cannot explain the merger. Duke (2010: 660) therefore suggests that adjective declension may have been restructured as “a product of analogy with the definite article acting as a model.” The dialects of Dutch reflect the merger in the Northern part (including the parts most relevant in Dutch standardization), while Southern and Eastern dialects mostly stick to the old three-gender system. Turning to the Mainland Scandinavian languages, some Danish dialects still retain a three-gender system while the standard language shifted towards a two-gender system. This is also the case in Swedish and some variants of Dano-Norwegian. Most of the dialects of Norwegian, however, retain the old three-gender system – the dialects in the region around the city of Bergen actually being the only dialects reflecting the two-way distinction. As for Dutch, we take a closer look at

265

266

SEBASTIAN KÜRSCHNER

definiteness marking and strong adjective declension to trace the development of the merger and possible reasons for it. In Old Swedish, e.g., masculines and feminines were distinguished both by case/number markers (masc. –er versus no marker in fem.) and by definiteness markers (resulting from an enclitic demonstrative marker, cf. masc. – inn versus fem. –in) in the nominative singular, with considerable variation found already at this time. The masculine nom.sg. marker -er (fisk-er -inn > fisk-inn) was eventually lost, and the definiteness suffixes merged (-inn > -in). The latter can be explained by the vulnerability of a morphological difference marked only by consonant length in the coda of an unstressed syllable (/nː/ versus /n/). From this starting point, the definiteness marker for CG nouns turned into the identical form -en after vowel reduction in unstressed position. This is the common form of Modern Swedish, Danish, and Dano-Norwegian, as opposed to N -et. Regarding strong adjectives, neuters were marked by a final -t in the Old Scandinavian dialects which is still in use today (cf. Danish et stor-t hus ‘a large house’). Masculines were distinguished from feminines by a nom.sg.-marker -er, which disappeared just as quickly as in noun inflection and left the two genders merged in adjective marking (cf. Danish en stor gade ‘a large street’). There is an ongoing discussion about the question whether the merger was “only” the result of phonological mergers and morphological restructuring, or if language contact was involved in the process. The most influential contact situation in the Scandinavian Middle Ages was caused by the strong economic influence of the Hanseatic league that brought speakers of Low German to the coasts, and especially the economic centers, of the Scandinavian countries (c. 1300–1550). Pedersen (1999) points out that the gender merger appeared in groups of dialects that were surrounded by three-gender dialects, and these groups are located precisely in the trade centers of Copenhagen, Stockholm, and Bergen. This supports the idea that contact between the closely related varieties of Low German and Scandinavian dialects resulted in gender reduction, probably due to imperfect learning and accommodation effects. However, data suggests that gender reduction is not only an effect of language contact. Ringgaard (1986) shows that tendencies of deflection can be found in Jutish villages far away from the contact zone at the coast, and that coastal regions were affected only later than that, suggesting that the loss of categories (including gender) developed independently from language contact. The oldest law documents of Danish, dating from 1250 to 1300, also partly reflect fluctuating gender marking in the demonstrative pronoun (thænn versus thæn; cf. Duke 2010 for a discussion based on Ringgaard 1986). A likely conclusion is that language contact was not the primary trigger of the merger, but played a role in accelerating the process.

Grammatical Gender in Modern Germanic Languages

12.2.3 Gender Loss There are two standard varieties of Germanic languages which completely lost lexical gender, English and Afrikaans. Regarding dialectal varieties, some insular North Frisian (Syltring, Helgolandic) dialects reflect gender loss, just as West Jutish varieties of Danish do. All these varieties continue to have pronominal gender tracking, but adnominal gender is lost: The English definite article the and its indefinite counterpart a(n) occur with every noun, with pronunciation differences based on phonological grounds only (cf. the [ðə] house versus the [ði] owl, a house versus an owl). In Afrikaans, the definite article is invariable die and the indefinite article is invariable ’n. Gender differences have no effect on strong adjective declension either. In the following section, we will have a closer look at how the two languages lost their lexical gender systems. In English, the gender system went through significant changes at some point between 1000 and 1400, changing the system from a lexical gender system to a purely referential system. In current English, gender is only marked pronominally in the third person sg., see the pronouns she/her, he/his, and it/its, while adnominal gender has been lost. The modern gender system depends on the opposition of animate versus inanimate (N), with nouns denoting animates being marked for M or F, which are thus used to mark the sexes male and female.4 In Old English, however, there was still a lexical gender system distinguishing between the three genders, as observable in the demonstrative nom.sg. se¯o (F), se (M), þæt (N) that still had a complex paradigm of inflectional forms in four cases and two numbers. In Middle English, there is considerable variation among the dialects, some of which still retain the marking of the lexical gender system, and some of which only keep an invariant marker þe as a definite article for nouns of all genders, which finally formed the basis for the modern English system. In the strong adjective paradigm, feminine marking still differed from nonfeminine marking by a suffix -u (> ME -e) in nom.sg. In Middle English, the distinction fell out of use, leaving adjectives unmarked as in Modern English. There have been several recent studies suggesting multiple factors involved in gender loss in Old and Middle English, e.g., Curzan (2003), Stenroos (2008), Siemund and Dolberg (2011), Dolberg (2012, 2014). The studies differ regarding the question whether the first step consisted of pronominal gender changing to referential gender and providing a basis for the reanalysis of the adnominal gender system (supported by Dolberg, and Siemund and Dolberg, but contradicted by Curzan and Stenroos). It seems clear that referential gender was expressed in pronouns already in Old English for animates, and that gender was overall less frequently marked in the stage of transition towards a purely referential gender

4

Some exceptions are found in vessels, e.g., ships, and countries, which can be referred to using the female pronoun she.

267

268

SEBASTIAN KÜRSCHNER

system. The data support that mass nouns and abstracts were the first items transferred from lexical to referential gender. Several reasons might have played a role in the loss of gender in English. It is unlikely that gender loss had purely phonological reasons. While some determiners and adjective declensions are rather similar, the overall differences between gender markers are too big to suggest a full phonological merger. However, gender loss goes parallel with significant de-flection in other parts of the inflectional system, thus pointing towards a possible typological drift. It is a matter of discussion whether language contact has been involved. English went through heavy stages of intense language contact, of which contact with Scandinavian varieties is most likely to have played a role here. English was mostly affected by the contact situation in its northern dialects, in regions in which Scandinavians settled over a considerable period of time (the Danelaw area from around AD 870 to the time around the Norman conquest 1066). Dolberg (2014) shows that (at least parts of) the Danelaw area is the first to attest a change towards the referential system and loss of lexical gender, while the southern dialects retain the lexical gender system for a longer time. Moreover, a number of other innovations pointing in the direction of modern Standard English are shared early on among the northern dialects, while southern dialects do not adopt the innovations quickly. The strong changes in the northern dialects might be connected with the contact situation. We can assume that Old English and Old Norse were similar enough at that time to be used in a situation of receptive multilingualism, with speakers using their mother tongue and trying to understand speakers of another (closely related) variety. This kind of conversation usually includes accommodation processes. For mutual understanding to be possible, Dolberg (2014: 437) assumes that “aiming for simple syntactic structures, de-emphasis of inflections, and contrast-maximisation of paradigmatic closed-class items bearing high functional load, i.e. personal pronouns” are likely strategies. The contact situation might thus have accelerated or even caused the loss of English lexical gender. Although this is attested only much later in written language, the changes might have been transported orally, being reflected literally only when the scribes lost competence of the lexical gender system. Turning to Afrikaans, we are dealing with a different sociolinguistic background. Dutch settlers at the Cape brought mainly Hollandic dialects with them that were probably in the process of merging masculines and feminines at that point of time (second half of the seventeenth century). The result was the Dutch two-gender system marking a difference between N (article het and demonstratives dit and dat) and CG (article de and demonstrative die), and this is reflected in early written evidence from the Cape. Current Afrikaans only uses the definite marker die (a demonstrative form in Modern Dutch) with all nouns. As Duke (2010) points out, no

Grammatical Gender in Modern Germanic Languages

phonological reasons play a role in the second merger of N and CG, since the markers are saliently different. We must assume that neuters were reclassified as CG nouns. Thinking of the high language contact situation at the Cape involving imperfect learning, indifferent learning of gender agreement with nouns might have resulted in a generalization of the CG marker. Written documents from Cape Dutch show uncertainty in gender assignment from early on, i.e., already in the second half of the seventeenth century, and through the whole history. Only the late standardization in the nineteenth and twentieth centuries made it clear that no gender distinction is used in Standard Afrikaans nouns. It is therefore likely that differing varieties were used over the centuries, some closer to the original Dutch dialects, some reflecting the (emerging) Dutch standard variety, some more progressive in the direction of Modern Afrikaans, and Roberge (2002) assumes that speakers were used to accommodating according to context conditions. Gender was probably lost early on in basilectal varieties, while it was – at least partly – maintained in acrolectal varieties over a long period and into the twentieth century (Duke 2010).

12.3 Gender Assignment While it is often said that gender is arbitrarily assigned to nouns in modern Germanic languages, there are reasons to argue that large parts of the lexicon show (partly) transparent assignment: There are several groups of nouns based on common formal (phonological and morphological) behavior and semantics that mostly appear with the same gender. In-depth studies of the Modern German gender system were provided by Ko¨pcke and Zubin (e.g., 1996, 2009). They found that gender assignment is far from arbitrary, and that based on formal and semantic clusters, nouns are found with predictable genders over large parts of the lexicon. The following levels are relevant in gender assignment according to their studies: a)

Formal levels i) sound-based: first and final sound(s), number of syllables Example: monosyllabic nouns ending in /ʃ/ → masculine ii) morphological: derivational suffixes Examples: nouns ending in -heit → feminine, -ling → masculine, -chen → neuter

b)

Semantic levels i) referential: reference object must be known and classifiable exophoric reference to persons based on sex; name classes: ship names → feminine, car names → masculine, city names → neuter

269

270

SEBASTIAN KÜRSCHNER

ii) semantic: based on word semantics women → feminine, men → masculine; hyperonyms → neuter (Tier N ‘animal’), hyponyms → non-neuter (Kuh F ‘cow’, Hund M ‘dog’); fruits → feminine, etc. Apart from these levels, there is lexical assignment that is arbitrary and cannot be deduced from any of the criteria named above. Nu¨bling et al. (2014) complement the assignment factors by a sociopragmatic factor that they found relevant in the gender assignment on womens’ first names in some Western German dialects. This additional factor is introduced because in these dialects, female’s first names are used with hybrid F and N gender, according to the social and emotional relation between the speaker and the referent: For closely related women, a neuter gender target is used, while the use of a feminine target is the more likely, the less close the relation between the speaker and the woman is. Marital status is also relevant: Unmarried women are usually referred to with a neuter gender target, while a feminine target is used for married women (cf. Nu¨bling 2017 for a broader analysis of Germanic languages regarding the use of N for persons).5 Sometimes there are also mismatches between adnominal and pronominal gender. Such cases of mismatches between adnominal and pronominal gender are not uncommon in modern Germanic – recall that there is a higher degree of specification in pronominal than in adnominal gender in many Germanic languages (see Table 12.1 above). Examples from German show that mismatches even occur in languages with an identical number of three adnominal and pronominal genders. Panther (2009) and Ko¨pcke et al. (2010) tried to capture the different uses of adnominal and pronominal genders by making a difference between grammatical and conceptual gender. Conceptual gender is not an inherent lexical characteristic, but assigned based on the semantics of a word (e.g., girl → F) or by characteristics of a referent (e.g., the doctor → F or M according to the referent’s sex). Note that, although the authors’ terminology differs slightly from Dahl’s introduced in Section 12.1, ‘grammatical gender’ corresponds roughly with what we call ‘lexical gender’, and ‘conceptual gender’ with what we call ‘referential gender’. Corbett (1979: 204) defined the Agreement Hierarchy that makes predictions about gender agreement and looks like the following: attributive – predicate – relative pronoun – personal pronoun. The possibility of syntactic agreements decreases monotonically from left to right. The further left an element on the hierarchy, the more likely syntactic agreement is to occur, the further right, the more likely semantic agreement. 5

Nübling (2017) provides an interesting example of gender reduction in her discussion of the dialect of Föhring-Amring (Section 12.2.2) as an example of a F-N merger). According to her analysis, affective pejorative uses of neuter gender targets with words denoting women (former feminine words) might have been the starting point for generalizing neuter gender over most feminine words – then losing the affective connotation – and causing the merger in neuter forms.

Grammatical Gender in Modern Germanic Languages

specifying

modifying

grammatical gender agreement

predicating

referent tracking

conceptual gender agreement

Figure 12.1 Pragmatic hierarchy of gender agreement, from Köpcke et al. (2010: 179)

Concerning gender, the elements further to the left on the hierarchy are more likely to use lexical gender as a basis for agreement, and the elements further to the right are likely to use conceptual gender deriving from semantics or actual referents. Based on data from German, Ko¨pcke et al. (2010) suggest that agreement varies between different gender target classes according to their pragmatic functions. For this reason, they adapt Corbett’s (1979) agreement hierarchy into a pragmatic hierarchy of gender agreement (cf. Figure 12.1). The pragmatic functions are based on Speech Act Theory and correspond with prototypical parts of speech. The specifying function is mostly established by determiners – they lay an anchor for successful reference tracking in the following parts of the discourse. Attributive adjectives fulfill a modifying function and usually reflected the grammatical (lexical) gender. The predicating function is found in predicative adjectives or nouns, with adjectives being gender sensitive only in the modern North Germanic languages (cf. Danish bilen er rød [ʁœð̞ ] ‘the car is red’ versus huset er rødt [ʁœd̥ ] ‘the house is red’). The strongest basis for referent tracking is found in pronouns. While relative pronouns in German tend to reflect the grammatical (lexical) gender, personal and possessive pronouns are more flexible and show varying use of grammatical (lexical) and conceptual gender agreement. The pragmatic functions are interrelated with (and constrained by) other formal determinants like the linear distance between source and target, the syntactic domains in which source and target are situated and the grammatical category / function of the target. As Ko¨pcke et al.’s observations show, grammatical (lexical) and conceptual genders both play a role in reference tracking. It is therefore interesting to ask what relevance conceptual factors have in the development of gender systems and whether conceptual gender as reflected in pronominal gender might be relevant for the restructuring of adnominal gender. These questions are discussed in Section 12.4.

12.4 Pronominal Gender and Semantic Reorganization of Gender Systems The names of the genders ‘masculine’ and ‘feminine’ reflect a sex-based difference marked by genders that is only relevant in a small part of the lexicon, denoting people and larger (individuated) animals. It is assumed that the three genders of Indo-European developed independently, and the

271

272

SEBASTIAN KÜRSCHNER

marking of sex differences was only later applied to the existing marking differences (see Section 12.1). As discussed in Section 12.3, modern Germanic languages with three-gender systems reflect the sex-based marking in relevant words, with the determiner and relative pronoun system (8a) reflecting lexical gender and the system of personal pronouns (8b) being more semantically flexible, see example (8) using the German word Ma¨dchen ‘girl’. (8)

a. Da dru¨ben steht ein Ma¨dchen, das / *die auf jemanden wartet. Over there stands a girl.N who.N / who.F for someone waits. ‘There is a girl standing over there who is waiting for someone.’ b. Da dru¨ben steht ein Ma¨dchen. Es / Sie wartet auf jemanden. Over there stands a girl.N. It / she waits for someone. ‘There is a girl standing over there. She is waiting for someone.’

While cases of mismatch between lexical and referential gender, such as German Ma¨dchen, are relatively rare in three-gender languages, the situation in Germanic languages with gender reduction and loss is especially interesting: Recall that in most two-gender languages it is precisely the masculine and feminine genders that merge, leaving sex-based differences unreflected in adnominal gender marking (cf. Danish pige-n ‘the girl’, en pige ‘a girl’, dreng-en ‘the boy’, en dreng ‘a boy’). Obviously, the same goes for the two languages with gender loss. However, as Table 12.1 shows, all Germanic languages retain pronoun systems that make it possible to distinguish between F, M, and N. In the following, we will take a closer look at the interplay of gender agreement in articles and pronouns in languages with gender reduction and at the use of pronouns in languages that lost lexical gender. We will specifically look at gender reflected in the third person sg. pronominal system, since this is where we find variation today across all Germanic languages. It is, however, noteworthy that gender has been part of the third person pl. pronominal system in older stages of Germanic, and still is in modern Icelandic and Faroese, with consequences for referential gender marking: While groups of males are referred to by the masculine form (Icelandic þeir / Faroese teir), groups of females are referred to by the female form (þær / tær), and mixed groups or groups in which the sexes of the members are unknown are referred to using the neutral form (þau / tey). Let us now turn to languages with lexical gender reduction. In Danish (just as in Swedish), the two lexical genders are reflected in the third person sg. pronominal system, see CG den and N det. Additionally, the pronominal system provides two pronouns that are used when a noun denotes a person or an individuated animal, cf. masculine han and feminine hun (Swedish hon).6 So while there are only two grammatical genders 6

Swedish provides an additional pronoun hen that is used for gender-neutral reference to persons.

Grammatical Gender in Modern Germanic Languages

left, and this is reflected in two pronouns for nonanimates, two additional sex-specific pronouns exist for nouns in which this distinction is relevant. In these cases, sex-based marking is even chosen when the grammatical gender is neuter but the sex is known (cf. barn N ‘child’, anaphorically referred to with han or hun if the sex is known, otherwise with neutral det). Danish and Swedish thus developed two formally different pronominal subsystems, one for syntactic agreement depending on lexical gender, and one for referential sex-based agreement. The choice for lexical versus referential agreement depends on semantics, namely, the question of whether the sex information is relevant to the concept of the noun. Interestingly, in the Norwegian standard variety of Dano-Norwegian, the same system of pronominal reference as in Danish and Swedish is used, although the feminine gender has been restituted, i.e., nonanimate feminine words like dør ‘door’ are referred to using the pronoun den, not feminine hun. Braunmu¨ller (2007) summarizes the pronominal gender system as adapted in Figure 12.2. In Dutch, there are pronouns for masculine (hij) and feminine (zij), but in contrast with Danish and Swedish, there is no CG pronoun to be used with nouns in which the sex information is irrelevant. Resulting from this, there are three pronominal targets for gender information (including neuter het) while there are only two lexical genders. The masculine and feminine pronouns therefore needed to be redistributed for nouns in which the sex information plays no role. Audring (2006) shows that in current spoken Dutch it is not the old, inherited gender information that is referred to when choosing a pronoun but that a new semantically based assignment has developed that involves all three pronominal gender markers and results in a high number of mismatches between determiner and pronominal gender. This new system is characterized by (a) using the feminine pronoun only for females, (b) using the masculine pronoun for males and for countable nonanimate entities, and (c) using the neuter pronoun for mass nouns. An alternative is provided by syntactic agreement for which either the old lexical gender information must be available

Pronouns

[+animate]

sg.

[–animate]

[male]

[female]

[CG]

[N]

han

hun/on

den

det

Figure 12.2 The pronominal gender system of Danish and Swedish, translated and adapted from Braunmüller (2007: 54)

273

274

SEBASTIAN KÜRSCHNER

(which is usually not the case in CG) or demonstratives are used, because these (deze ‘this’, die ‘that’) do not provide different forms for masculine and feminine agreement (reflecting the CG) but clearly differ from neuter demonstratives (dit, dat). Regarding languages with gender loss, Standard English was introduced in Section 12.2.3, showing a simple system of referentially based pronominal gender. The case is different with Afrikaans: While the feminine pronoun sy is used for sex-based reference to females only, masculine hy is used for most concrete nouns apart from mass nouns, while neutral dit has developed into a marker of abstracts and mass nouns. This distribution is not stable but varying depending on factors of medium, style, and affectivity of the speech situation (Ponelis 1979: 585ff., see Audring 2010: 700ff. for discussion). Although Dutch retained two grammatical genders and Afrikaans lost this category, the organization of pronominal gender is thus rather similar in both languages. A look at dialectal varieties shows that an alternative development has also been possible for English, resembling the Dutch / Afrikaans pictures: As Siemund (2008) shows in a contrastive study, varieties of English organize pronominal gender differently. While sex-based differences are used referring to persons (including the use of anthroponyms) and (some or all) animals as in the standard variety, the masculine and neuter pronouns can be distributed differently in dialects, e.g., in the use of masculine he for nouns with inanimate concrete denotations, while it is reserved for abstract and mass nouns, e.g., in the South-Western variety of West Somerset English. In all the systems discussed, quantification and individuation play a role, so differences according to very general and cognitively relevant semantic categorization are marked using the “overdistinguished” pronouns left from former three-gender systems. The Individuation Hierarchy (Sasse 1993, Siemund 2008) captures the most relevant categories in this respect. It expands the Animacy Hierarchy according to individuation and views proper names as most individuated, and mass nouns as least individuated, with the following steps on the cline as presented in (9): (9)

Individuation Hierarchy

proper names human beings animal most individuated

physical objects

abstracts

mass nouns least individuated

We saw that the parameters of the Individuation Hierarchy played a role in several of the examples provided above. There are also examples from North Germanic that fit well with this picture. In Western Jutish dialects in which grammatical gender was lost (see above) the distinctions are reflected by use of the (former) CG pronoun den ‘it’ for concrete and countable nouns versus the (former) N pronoun det ‘it’ for abstract and

Grammatical Gender in Modern Germanic Languages

noncountable nouns (see Braunmu¨ller 2000: 28–29). It is especially interesting that the marking difference can even occur at the level of determiners (compare den hus ‘the house’ with pronoun den versus det mælk ‘the milk’ with pronoun det), thus providing the start for reshaping even the adnominal gender system. Semantic reorganization at the level of determiners has also been observed, e.g., for the Dutch dialects of Groningen (cf. Audring and Booij 2009: 24–26), based on the features [±human] or [±animate], adding up nicely to the picture that the Individuation Hierarchy is most relevant for the reorganization of gender systems. To this point, we looked at the reorganization of gender systems based on mismatches between grammatical and semantic gender and on the overspecification of pronominal gender in systems with gender reduction and loss. Gender can also be used to reorganize parts of the noun lexicon in which no gender information is naturally available, i.e., the part of the onomasticon with nonanimate referents. As Fahlbusch and Nu¨bling (2014, 2016) show, proper names (like product names) are purely referential and do not provide semantic characteristics (or features). They refer to exactly one entity (or a class of identical entities in the case of product names), and the assignment of a gender is accordingly referentially based. Vice versa, the gender provides information about proper referents to the listener, as Fahlbusch and Nu¨bling (2016: 108) exemplify by the German word Admiral: A feminine form (die Admiral) refers to a ship, a motor cycle, or a plane, a neuter form (das Admiral) to a hotel, restaurant, or beer, and the masculine form, unless the appellative ‘admiral’ is used, to a car or a mountain. So in a language like German, there are classes of referents associated with the adnominal gender of names, providing a new class system of onymic genders. Starting from this observation, Nu¨bling (2015, in press) interprets the gender system of proper names as an emerging classifier system consisting of six classes: The three classes named above, and three classes in which names do not usually occur with an article word but can be ascribed to a gender class (as in German place names like Hamburg, which are neuter (das scho¨ne Hamburg ‘beautiful Hamburg’), and articleless names for women and men which are female or male, respectively). Nu¨bling argues that an exaptation of the article and a degrammaticalization of the gender system form the background of this process.

12.5 The Interaction of Grammatical Gender and Inflection Interaction between grammatical genders and the inflectional system of nouns (for details on the history of declensions, see Nu¨bling, Chapter 10) can be encountered in all the Germanic languages, at least in their first documented historical stages. In the strong declension classes, e.g., an s-suffix marking the genitive has been found on masculines and neuters,

275

276

SEBASTIAN KÜRSCHNER

but never on feminines.7 Certain other case-number-affixes are also primarily conditioned by genders. This distribution is still characteristic across a couple of languages, mainly those which kept three-gender systems, e.g., Icelandic, Faroese, and German. In some other languages, the interaction between grammatical gender and declension is almost lost. In a study of a number of Germanic languages and dialects of German, Ku¨rschner and Nu¨bling (2011) found that the interaction of gender and declension can be grouped into four types, with (1) and (4) presenting the poles of a continuum: (1) Gender and declension class merge, i.e., there are exactly as many declension classes as genders, and each gender corresponds to exactly one declension class (and, naturally, vice versa). New Norwegian tends to use this distribution, e.g., where more or less all feminines use the er-plural, masculines the ar-plural, and neuters show no plural marking (see Enger 2004). The same tendency is found in Alsatian dialects of German. (2) There are more declension classes than genders. Each declension is bound to one specific gender. A system of this kind is found in Swedish, where the plural allomorphs -ar, -er, -or, and -r tend to occur with CG nouns, and -n and zero plurals are mostly reserved to neuters. This type is also found in dialectal three-gender systems, e.g., in the Swiss German dialect of Fribourg. (3) There are exactly three genders and more than three declension classes. Each declension is bound to one or two out of the three genders. This is the case in Modern Standard German, where, e.g., strong declension with er-plural is found with neuters and masculines, the so-called mixed class (with strong declension in the singular and weak declension in the plural) is found in feminines and masculines, and the strong umlaut plural (Hafen – Ha¨fen ‘harbor(s)’) is almost exclusively found on masculines. (4) Gender and declension are fully dissociated (attested only in twogender systems). Declensions are distributed according to factors other than gender. This is the case in Dutch in which the plural allomorphs -s and -(e)n are distributed on prosodic grounds (a plural form ends in a trochee, cf. hand-en ‘hand(s)’ versus bezem-s ‘brooms’, gave-n ‘gifts’), and the class characterized by eren-plural is the only reminder of the old gender-based assignment (restricted to 15 neuters). A comparable tendency has also been found for the Low German dialect of East Friesland. Gender has thus had an impact on organizational changes in declension classes in the history of Germanic languages, at least in types (1) to (3). To illustrate this, the history of German feminines is revealing (see 7

The only exception being women’s names, compare German Julias / Omas Auto and ‘Julia’s / grandma’s car’.

Grammatical Gender in Modern Germanic Languages

Nu¨bling, Chapter 10 for more details): While in Old High German, feminines were found in high numbers in the o¯ -stems, i-stems and the weak n-stems, they were reorganized in Early New High German, resulting in most feminines assembling in a mixed class characterized by the plural n-marker (cf. Standard German nom.sg. Tasche – nom.pl. Taschen ‘bag’). In this process, the link between feminine gender and “its” declension class was strengthened by abandoning the class of old o¯ -stems (which had developed a zero plural in the important nominative and accusative cases in Middle High German) and retaining only a limited set of approximately 35 feminines in the old i-stems (today: e-plural with umlaut, cf. Kunst – Ku¨nste ‘arts’). Several reasons for the interaction of genders and declensions have been discussed. A particularly interesting discussion has been started by Wurzel (2001) who points out that declension classes existed in high numbers in Proto-Germanic but lacked a functional motivation, resulting in a low memorability. Declension classes were therefore linked to other, noninflectional structural characteristics of nouns to enhance memorability. One of these characteristics is the gender information (in addition to, e.g., word-formation affixes, prosodic structures, final sounds). Since the functional load of gender is relatively low as well, both categorization systems might have supported each other by being bound to each other. Starting from this view, it is striking that the important (or ‘relevant’, in the sense of Bybee 1985) number information is profiled using both classification systems for each of the two numbers (singular and plural) in the further development of the modern Germanic languages and dialects (cf. Nu¨bling 2008 on German and its the dialects): While the most reliable markers of declension class tend to be found in the plural allomorphy (with singular being unmarked and case allomorphy becoming less and less marked or even obsolete in some languages), gender is usually neutralized in the plural but is most clearly profiled in the singular (cf. the German der, die, das markers in nominative singular versus the neutralized die marker for all genders in nominative plural). This observation might also offer a starting point for explaining why some languages tend to retain a strong link between genders and declension classes: In the plural, gender information can be deduced from the inflectional morphology, while in the singular, possible declension classes can be deduced (and excluded) from the gender markers.

12.6 Conclusions Gender in Germanic languages can be studied from a range of different perspectives. The lexical gender systems, reflected in adnominal markers, show a rather strong diversity across varieties of Modern Germanic, ranging from three- and two-gender systems to systems with no grammatical

277

278

SEBASTIAN KÜRSCHNER

gender. Apart from lexical gender, which is assigned based on semantic or formal characteristics, referential gender is found mainly in pronominal gender, sometimes causing mismatches with adnominal gender markers. Different patterns of reorganization resulted from a mismatch between the number of adnominal and pronominal genders in two-gender or genderless languages. A topic for further investigation is the extent to which the conceptual reorganization of pronominal gender has an influence on lexical gender. Gender also interacts with other class systems, like declension classes. Both class systems share a long history of interrelations in the Germanic languages, and this might be due to increased memorability of both categories and/or the profiling of the highly relevant number category in nouns. The many directions into which gender systems developed in a rather small language family leave a lot of questions to be dealt with in future (comparative) research, e.g., regarding the possible functions of gender and the reasons for reduction and loss, and still support the characterization that Corbett (1991: 1) gave in his seminal study of gender: “Gender is the most puzzling of the grammatical categories.”

References Audring, J. 2006. “Genusverlies en de betekenis van naamwoorden.” In M. Hu¨ning, U. Vogl, T. van der Wouden, and A. Verhagen (eds.), Nederlands tussen Duits en Engels. Handelingen van de workshop op 30 september en 1 oktober 2005 aan de Freie Universita¨t Berlin. Leiden: SNL: 71–88. Audring, J. 2010. “Deflexion und pronominales Genus.” In A. Dammel, S. Ku¨rschner, and D. Nu¨bling (eds.), Kontrastive germanistische Linguistik. Hildesheim: Olms: 693–717. Audring, J. and G. Booij 2009. “Genus als probleemcategorie,” Taal en tongval 22: 13–37. Braunmu¨ller, K. 2000. “Gender in North Germanic: A diasystematic and functional approach.” In B. Unterbeck, M. Rissanen, T. Nevalainen, and M. Saari (eds.), Gender in Grammar and Cognition. Berlin and New York: Mouton de Gruyter: 25–53. Braunmu¨ller, K. 2007. Die skandinavischen Sprachen im U¨ berblick, third edn. Tu¨bingen and Basel: Francke. Bybee, J. L. 1985. Morphology. A Study of the Relation between Meaning and Form. Amsterdam: John Benjamins. Corbett, G. G. 1979. “The agreement hierarchy,” Journal of Linguistics 15: 203–395. Corbett, G. G. 1991. Gender. Cambridge University Press. Curzan, A. 2003. Gender Shifts in the History of English. Cambridge University Press.

Grammatical Gender in Modern Germanic Languages

¨ . 2000. “Animacy and the notion of semantic gender.” In Dahl, O B. Unterbeck, M. Rissanen, T. Nevalainen, and M. Saari (eds.), Gender in Grammar and Cognition. Berlin and New York: Mouton de Gruyter: 99–115. Dolberg, F. 2012. “Gender change from Old to Middle English.” In I. Hegedu˝s and A. Fodor (eds.), English Historical Linguistics 2010. Selected Papers from the Sixteenth International Conference on English Historical Linguistics (ICEHL 16), Pe´cs, 23–27 August 2010. Amsterdam: John Benjamins: 263–288. Dolberg, F. 2014. Gender Variation, Change, and Loss in Mediaeval English. Evidence from the Anglo-Saxon Chronicle. Doctoral dissertation, University of Hamburg. Duke, J. 2009. The Development of Gender as a Grammatical Category: Five Case Studies from Germanic. Heidelberg: Winter. Duke, J. 2010. “Gender reduction and loss in Germanic.” In A. Dammel, S. Ku¨rschner, and D. Nu¨bling (eds.), Kontrastive germanistische Linguistik. Hildesheim: Olms: 643–672. Enger, H-O. (2004). “On the relation between gender and declension. A diachronic perspective from Norwegian,” Studies in Language 28: 51–82. Fahlbusch, F. and D. Nu¨bling 2014. “Der Schauinsland – die Mobiliar – das Turm. Das referentielle Genus bei Eigennamen und seine Genese,” Beitra¨ge zur Namenforschung 49.3: 245–288. Fahlbusch, F. and D. Nu¨bling 2016. “Genus unter Kontrolle. Referentielles Genus bei Eigennamen – am Beispiel der Autonamen.” In A. Bittner and C. Spieß (eds.), Formen und Funktionen. Morphosemantik und grammatische Konstruktion: 103–125. Hockett, C. F. 1958. A Course in Modern Linguistics. New York: Macmillan. Jacobs, N. G., E. F. Prince, and J. van der Auwera 2002. “Yiddish.” In E. Ko¨nig and J. van der Auwera (eds.), The Germanic Languages. London and New York: Routledge: 406–436. Ko¨pcke, K-M., K-U. Panther and D. Zubin 2010. “Motivating grammatical and conceptual gender agreement in German.” In H-J. Schmid and S. Handl (eds.), Cognitive Foundations of Linguistic Usage Patterns. Berlin: Mouton de Gruyter: 171–194. Ko¨pcke, K-M. and D. Zubin 1996. “Prinzipien fu¨r die Genuszuweisung im Deutschen.” In E. Lang and G. Zifonun (eds.), Deutsch: Typologisch. IDS-Jahrbuch 1995. Mannheim: IDS: 473–491. Ko¨pcke, K-M. and D. Zubin 2009. “Genu.” In E. Hentschel and P. M. Vogel (eds.), Deutsche Morphologie. Berlin: Mouton de Gruyter: 132–154. Ku¨rschner, S. and D. Nu¨bling 2011. “The interaction of gender and declension in Germanic languages,” Folia Linguistica 45.2: 355–388. Leiss, E. 2000. “Gender in Old High German.” In B. Unterbeck, M. Rissanen, T. Nevalainen, and M. Saari (eds.), Gender in Grammar and Cognition. Berlin and New York: Mouton de Gruyter: 237–258. Meier-Bru¨gger, M. 2002. Indogermanische Sprachwissenschaft, eighth edn. Berlin and New York: Mouton de Gruyter.

279

280

SEBASTIAN KÜRSCHNER

Nu¨bling, D. 2008. “Was tun mit Flexionsklassen? Deklinationsklassen und ihr Wandel im Deutschen und seinen Dialekten,” Zeitschrift fu¨r Dialektologie und Linguistik 75.3: 282–330. Nu¨bling, D. 2015. “Die Bismarck – der Arena – das Adler. Vom Drei-Genuszum Sechs-Klassen-System bei Eigennamen im Deutschen: Degrammatikalisierung und Exaptation,” Zeitschrift fu¨r Germanistische Linguistik 43.2: 306–344. Nu¨bling, D. 2017. “Funktionen neutraler Genuszuweisungen bei Personennamen und Personenbezeichnungen im germanischen Vergleich,” Linguistische Berichte Sonderheft 23: 173–211. Nu¨bling, D. in press. “Die Bismarck – der Arena – das Adler. The emergence of a classifier system for proper names in German.” In J. Flick and R. Szczepaniak (eds.), Walking on the Grammaticalization Path of the Definite Article in German: Functional Main and Side Roads. Nu¨bling, D., S. Busley, and J. Drenda 2013. “Dat Anna und s Eva – Neutrale Frauenrufnamen in deutschen Dialekten und im Luxemburgischen zwischen pragmatischer und semantischer Genuszuweisung,” Zeitschrift fu¨r Dialektologie und Linguistik 80.2: 152–196. Panther, K-U. 2009. “Grammatische versus konzeptuelle Kongruenz. Oder: Wann siegt das natu¨rliche Geschlecht?” In R. Brdar-Szabo´, E. Knipf-Komlo´si, and A. Pe´teri (eds.), An der Grenze zwischen Grammatik und Pragmatik. Frankfurt am Main: Peter Lang: 67–86. Pedersen, K. M. 1999. ”Genusforenklingen i københavnsk,” Danske folkemaº l 41: 79–105. Ponelis, F. 1979. Afrikaanse syntaksis. Pretoria: Van Schaik. Quak, A. and J. M. van der Horst 1997. “Oudnederlands (tot circa 1200).” In M. van den Toorn, W. J. J. Pijnenburg, J. A. van Leuvensteijn, and J. M. van der Horst (eds.), Geschiedenis van de Nederlandse taa. Amsterdam University Press: 19–36. Ringgaard, K. 1986. “Fleksionssystemets forenkling og middelnedertysk,” Arkiv fo¨r nordisk filologi 101: 173–183. Roberge, P. 2002. “Convergence and the formation of Afrikaans,” Journal of Germanic Linguistics 14.1: 57–59. Ronneberger-Sibold, E. 2007. “Typologically motivated over- and underspecification of gender in Germanic languages,” Language Typology and Universals 60: 205–218. Sasse, H-J. 1993. “Syntactic categories and subcategories.” In J. Jacobs, A. von Stechow, W. Sternefeld, and T. Vennemann (eds.), Syntax: An International Handbook of Contemporary Research. Berlin: Mouton de Gruyter: 646–686. Siemund, P. 2008. Pronominal Gender in English: A Study of English Varieties from a Cross-Linguistic Perspective. New York: Routledge. Siemund, P. and F. Dolberg 2011. “From lexical to referential gender: An analysis of gender change in medieval English based on two historical documents,” Folia Linguistica 45.2: 489–534.

Grammatical Gender in Modern Germanic Languages

Stenroos, M. 2008. “Order out of chaos? The English gender change in the southwest Midlands as a process of semantically based reorganization,” English Language and Linguistics 12: 445–473. Wahrig-Burfeind, R. 1989. Nominales und pronominales Genus im su¨dlichen Nordseegebiet: eine areallinguistische Untersuchung. Munich: Tuduv. Wurzel, W. U. 1986. “Die wiederholte Klassifikation von Substantiven. Zur Entstehung von Deklinationsklassen,” Zeitschrift fu¨r Phonetik, Sprachwissenschaft und Kommunikationsforschung 3.1: 76–96. Wurzel, W. U. 2001. Flexionsmorphologie und Natu¨rlichkeit. Ein Beitrag zur morphologischen Theoriebildung, second edn. Berlin: Akademie-Verlag.

281

Chapter 13 Case in Germanic Thomas McFadden

13.1 Introduction The Germanic (henceforth Gmc) languages provide an extremely interesting backdrop for a discussion of case phenomena, shedding light on the interactions between morphology, syntax and semantics, as well as a number of diachronic issues. Indeed, Gmc languages have played a central role in the development of theoretical treatments of case, especially within the (broadly) generative tradition. The oldest attested members of the family, and in particular what we can reconstruct as ProtoGermanic (henceforth PGmc), were highly inflecting languages with rich systems of morphological case. Within their recorded histories, however, all of the Gmc languages have reduced the extent to which cases are distinguished morphologically, though with significant differences in the details of how much has been lost and when. This has led to a contemporary situation where some members of the family (especially Icelandic) have retained most of the older distinctions, others (e.g., German) have retained the basic category distinctions but heavily restricted the ways in which they are expressed, and still others (English, Dutch, Afrikaans, and most of the mainland Scandinavian varieties) have reduced morphological case to a few vestiges in the pronominal system. This means that any treatment of case in Gmc must be comparative and historical, but also that the family can provide interesting opportunities for developing and testing theories of how the properties of morphological case systems might interact with other characteristics of a language. In any theoretically minded discussion of case, it is of course important to clarify what precisely is intended by the term. There are well-established traditions for using case to refer to several distinct morphological, syntactic and semantic concepts relating to nouns and NPs, often in some combination with each other. For example in the Case Grammar tradition following Fillmore (1968), it refers essentially to the semantic or thematic

Case in Germanic

role played by an NP, while within generative grammar in the Principles and Parameters tradition since the late 1970s, there is a notion of abstract syntactic Case, roughly an idealized abstraction of traditional morphological case categories spliced with a version of traditional grammatical functions as redefined in purely structural terms. Such theory-specific semantic and syntactic notions of case will not be my focus here, though they will frequently come up in the discussion. Rather, when I use the term “case” without qualification, what I mean is essentially the traditional morphologically grounded understanding. This has the advantage of being rather more theory-neutral, which allows me to delineate the subject matter in a relatively uncontroversial way and provides a simple basis for setting up comparisons. Typologically speaking, the Gmc languages, both historical and contemporary, uniformly show nominative-accusative patterns of case marking, apparently inherited from Proto-Indo-European (henceforth PIE), with no hints of ergative-absolutive or any other case systems. Case markers in Gmc generally take the form of inflectional suffixes, usually fusional markers also indicating number, information about inflectional class and (indirectly) gender, sometimes complemented by semi-regular mutations of the stem to which they are applied (i.e. Umlaut). In the pronouns and demonstratives, case marking often involves a mixture of suffixation and stem suppletion. The morphological details of case inflection can thus be quite complex, especially in the older and more conservative languages. For relevant details, see Nu¨bling (Chapter 10). The remainder of this chapter is organized as follows. In Section 13.2, I will discuss the different case inventories found in the various Gmc languages, starting with the six-case inventory that can be reconstructed for PGmc and working my way down to the successively smaller inventories and ultimately the (near) loss of case in some of the modern languages. This will include the necessary presentation of the relevant morphological details and basic discussion of the diachronic developments that lead from one inventory to another. Section 13.3 will be concerned with the distribution of the different cases in each language across grammatical environments, i.e., how the various cases are “assigned.” Of central importance here will be questions of how the distribution of the cases is determined by and related to details of the syntactic structure, the lexical identity of verbs and other predicates, and semantic issues. Then in Section 13.4 I will turn to a number of interesting empirical and theoretical questions about how case interacts with other grammatical phenomena in the Gmc languages. The issues here arise in any language with case marking, but the patterns observed in the Gmc languages have played an outsized role in theoretical discussions of case. Picking up from the discussion of how the cases are distributed, we will turn to the long-standing question of how case interacts with grammatical functions and in particular the existence of oblique subjects and the notion of quirky case.

283

284

T HO M A S M CFA DD E N

This will lead into a more general discussion of how morphological case interacts with syntax, including ideas of abstract Case and how richness of morphological marking may influence other syntactic properties of a language.

13.2 Inventories PGmc, as best we can reconstruct it, had at least six morphologically distinct cases.1 We can exemplify these with singular forms of the word for ‘day’ (a masculine a-stem) and ‘deed’ (a feminine i-stem) as in Table 13.1, following the reconstruction of Ringe (2006).2 Morphological case distinctions were clearly marked in PGmc on nouns, adjectives, pronouns, demonstratives, and numerals up to four, though the details in how they were marked in these categories differ considerably. Already in PIE, the pronouns in general had their own special inflection which differed in several points from the nominal inflection, including not just distinct sets of case endings but also rampant suppletion and other irregular stem alternations. As for the adjectives, in PIE their inflection had been essentially identical to that of nouns, but in the prehistory of PGmc, it underwent significant modifications. First, every adjective gained the ability to alternate between two distinct inflectional classes – the so-called ‘strong’ and ‘weak’ inflections – with the choice being sensitive to the definiteness of the NP. Second, the strong adjective forms began to diverge considerably from the originally identical strong noun forms by taking over distinctly pronominal case and number endings (see McFadden 2003 [2009], Ringe 2006, Ratkus 2015). This is relevant for our concerns primarily in that the innovative case-number endings of the strong inflection were morphophonologically more substantial than the original nominal ones

Table 13.1 PGmc ‘day’ and ‘deed’

Nominative Vocative Accusative Genitive Dative Instrumental

1

‘day’

‘deed’

*dagaz *dag *daga˛ *dagas *dagai *dago¯

*de¯diz *de¯di *de¯d˛i *de¯d¯ı z *de¯d¯ı *de¯d¯ı

It is also plausible that a distinct locative had not yet fallen together with the dative and/or the instrumental, but the evidence is equivocal at best. See Ringe (2006), Ringe and Taylor (2014, especially p. 379) and Harðarson (2017) for discussion.

2

As is standard in historical linguistics, the asterisk preceding a form here indicates that it is reconstructed. PGmc was spoken by a preliterate society, thus everything we know about the language has been reconstructed on the basis of evidence from its attested daughter languages.

Case in Germanic

(compare e.g., masculine singular accusative strong adjective *-ano˛¯ to nominal *-a˛ ). They were thus more resistant to loss, so that in many Gmc languages certain case distinctions have held on longer in adjectives than in nouns. See Nu¨bling (Chapter 10) for more details. Even though the PGmc case system was robust, a non-trivial number of syncretisms can already be observed. For example, dative and instrumental are syncretic in the inflection of ‘deed’, and nominative and accusative are systematically syncretic in neuters throughout the language. Furthermore, the case system of PGmc was already a reduction compared with PIE, which had distinguished at least eight cases including an ablative and a locative. A gradual reduction in the number of distinct case categories is characteristic of the further development of Gmc, and in fact none of the actually attested languages distinguishes all six cases. Many of the older languages distinguish five, but differ in which one they have lost. There are four cases which all retain – nominative, accusative, genitive, and dative – and which remain as a relatively stable core for a considerable amount of time. Gothic has lost the instrumental, having assimilated it to the dative, but preserves the vocative quite clearly. The West Gmc languages have all lost the vocative by the time of their earliest attestations, but maintain distinct instrumental forms, at least in certain masculine and neuter paradigms. The situation in North Gmc is a bit more complicated. The earliest stage attested in the early Runic inscriptions is quite fragmentary, and there is disagreement over whether certain forms should be interpreted as vocatives or not (Nielsen 1998). By the time of the rather well-attested Old Icelandic (OIce), the vocative is gone. The instrumental is also no longer distinguished from the dative but has left behind clear traces, e.g., in the shape of certain neuter dative forms and in a number of otherwise surprising uses of the dative in modern Icelandic.3 The older languages also maintain case-inflectional distinctions on all of the categories where these were marked in PGmc. However, the overall frequency with which the full number of case distinctions available in a language as a whole are in fact expressed by particular forms is already significantly reduced compared to the Proto-language. Take e.g., the Old English (OE) forms in Table 13.2, descended from the PGmc ones in Table 13.1. The vocative has disappeared entirely from the language, and the instrumental is no longer distinguished in the nouns. Furthermore, nominative and accusative, while distinguished with some nouns, have become syncretic for these two, and genitive and dative have fallen together for ‘deed’. Indeed, while the four cases are still distinguished by noun inflection as a whole, there is no individual noun in the language that distinguishes all of them. More distinctions are found in the OE adjectives, at least in the strong inflection. In fact, the masculine singular has distinct forms for all 3

See Barðdal (2001), Svenonius (2002), Thrá insson (2007) for some relevant discussion of “‘instrumental” uses of the dative in Modern Icelandic.

285

286

T HO M A S M CFA DD E N

Table 13.2 OE ‘day’ and ‘deed’

Nominative Accusative Genitive Dative

‘day’

‘deed’

dæ g dæ g dæges dæge

dæ¯ d dæ¯ d dæ¯ de dæ¯ de

Table 13.3 OE simple demonstrative / definite article Sg N A G D I

M

N

F

Pl

se¯ Þone þæs þæ¯ m þy¯ /þon

þæ t þæ t

se¯ o þa¯ þæ¯ re þæ¯ re þæ¯ re

þa¯ þa¯ þa¯ ra þæ¯ m þæ¯ m

five cases including the instrumental. However, even here fewer distinctions are made in the other genders and the plural, and the weak adjective declension has undergone quite massive syncretism. The case system is most extensively preserved in the pronouns, and especially in the demonstratives, exemplified in Table 13.3, though even here clear reductions relative to the Proto-language can be discerned. OE can stand in reasonably well as a model for most of the older Gmc languages. The details of where exactly clear distinctions are to be found differ quite widely across the languages, but the overall patterns are remarkably similar. Significant divergences among the languages first appear later in the Middle Ages. Though the timing differs for the various stages of the development, a similar diachronic pattern can be discerned across all of the languages (with the possible exception of Icelandic) – a gradual but persistent weakening and ultimately reduction of case distinctions. Initially, this amounts to an increase in the incidence of syncretism on individual word forms, with retention of the distinctions in a language as a whole. This development typically hits the nouns earliest and hardest, then the adjectives, and only in the last stages the pronouns and demonstratives, with the result that, during intermediate periods, the case of an NP as a whole will often still be clearly identifiable by summing up the various bits of information contributed by its different parts. As an example, consider the NP der besten Spieler in Modern German. The form of the definite article der could be masculine singular nominative, dative or genitive singular feminine, or genitive plural; besten ‘best’ in this context could be anything but nominative singular of any gender or accusative singular neuter; the noun Spieler ‘player’ is itself masculine, and its form could be anything but

Case in Germanic

Table 13.4 German ‘day’ and ‘cow’ ‘day’

Nominative Accusative Genitive Dative

‘cow’

Sg

Pl

Sg

Pl

Tag Tag Tages Tag

Tage Tage Tage Tagen

Kuh Kuh Kuh Kuh

Kühe Kühe Kühe Kühen

a genitive singular or dative plural. Each of these three word forms is thus multiply ambiguous for case, but the three together as one NP can only be genitive plural (‘of the best players’). In the more advanced stages of this development, as distinctions disappear even in pronouns and demonstratives, the case categories as a whole begin to collapse, and the system is reduced from four to three and then two cases, and in some of the languages arguably to a complete loss of morphological case. Let us consider some specific examples of these stages of development. Standard German presents a straighforward example of the first intermediate stage. The four central cases are still distinguished, but the categories where distinctions are marked have been heavily reduced, and syncretism has increased significantly. Very few distinctions are now marked on nouns themselves, as seen in Table 13.4. Distinctions have also been mostly lost in the numerals beyond “one” and reduced to the bare minimum in the weak adjective inflection. Case marking is, however, alive and well in the pronouns, the strong adjectives and especially the demonstratives. Even in these categories, however, far fewer than the full set of potential distinctions are actually made. The nominative / accusative distinction is especially tenuous, restricted entirely to the masculine singular and the personal pronouns. This situation, where the four-case system is preserved, but only barely, is characteristic of historical stages of several of the other Gmc languages, including late OE / early Middle English (ME), Middle Dutch, and late Medieval forms of the mainland Scandinavian languages. The next stage can be exemplified with Yiddish, many modern German dialects, transitory stages of ME, and many Scandinavian varieties. This is when one of the four core cases ceases to be a productive part of the grammar, yielding a three-case system. In most of the languages it seems to have been the genitive which was lost first. In Yiddish, for example, similar patterns to those found in the closely related German are found in the pronouns, strong adjectives and demonstratives, but only for these three cases, and only a very small class of nouns still make any distinctions at all (Jacobs 2005). Faroese also seems to be moving in this direction, as the use of its genitive is increasingly restricted (Thra´insson et al. 2004). It should be noted here that there are some complications regarding the status of the genitive in several of the modern languages, including

287

288

T HO M A S M CFA DD E N

English, Mainland Scandinavian, Yiddish, and colloquial German as well as many of its dialects. Forms that are in some sense the descendants of older clear genitives persist in these languages, but demonstrate peculiar behaviors suggesting that they should no longer be analyzed as true case marking. For example, English has generalized the suffix ’s – historically the genitive singular of a large and productive class of masculines and neuters – to be a general possessive marker with nouns of all kinds. However, it is clearly no longer a nominal case suffix, but rather a phrasal affix or clitic on the entire nominal phrase, as shown by (1): (1)

I really like [the woman over there]’s theory.

In colloquial German, an -s suffix has similarly spread from masculines and neuters to appear on feminines as well, but is largely restricted to names and namelike nominals and shows distinct syntactic behavior from the uncontroversial genitives of the modern standard language.4 This suggests that in these languages, the relevant suffixes no longer signal a general genitive case, but rather specific possessive structures. The next distinction to be lost is that between accusative and dative. In English this occurred over the course of approximately two hundred years in the ME period, with significant differences between dialects, as chronicled in great detail by Allen (1995). In Mainland Scandinavian it happened significantly later, in most dialects by 1500 or 1600, though distinctly dative forms are retained even now in a significant number of varieties of Swedish and especially Norwegian (see Eyþo´rsson et al. 2013 and references there). In a number of German dialects we can observe this change in progress. Merkle (1993), e.g., describes details of the collapse of some dative and accusative forms in Bavarian. Yager et al. (2015) explore interesting developments in Heritage German varieties, where the dative is not so much being lost as being repurposed in a way reminiscent of differential object marking. Note that the form retained as the collapsed oblique of a given word in a given language is sometimes what was historically the accusative and sometimes what was the dative. In English, e.g., oblique forms for animates like him and her generally go back to old datives, whereas those for inanimates like it and what reflect old accusatives. The reduction to a two-case system, where a nominative is distinguished from a general oblique, has generally been accompanied in the Gmc languages by a radical reduction in the categories where case distinctions are expressed. Thus the Mainland Scandinavian languages, English, Dutch, and Afrikaans have all completely eliminated case-marking in nouns, adjectives, articles, and numerals. The only place where nominative and oblique are distinguished is in the pronouns, and even here this has generally been restricted to the personal pronouns, with relative pronouns, demonstratives, 4

Yiddish shows a pattern – what Jacobs (2005) refers to as the possessive – that is similar in its abstract outlines, but a bit more complex in the morphological details.

Case in Germanic

and wh-pronouns having become indeclinable. Such heavily reduced patterns are sometimes referred to by the term vestigial case or impoverished case. Logically speaking, an even further reduction is of course possible, where the nominative-oblique distinction is given up, and a language simply ceases to have morphological case. Now, all Gmc languages that I am aware of continue to have distinct forms of at least some pronouns that historically reflect case distinctions.5 However, in some of the vestigial case languages, including at least English, Danish, and some Norwegian varieties, it can be argued that the contrasts here no longer really involve case (see especially Parrott 2007, 2009). The arguments are based on the distribution of the different forms, thus I postpone discussion to Section 13.3. Our understanding of the causes for this pattern of gradual loss of case distinctions remains surprisingly incomplete. The traditional and perhaps still most popular idea is that it was driven primarily by regular phonological developments which predictably obliterated most of the case distinctions via reductions of the unstressed final syllables that contained the various endings (see Trask 1996, Blake 2001, Delsing 2002 for representative presentations). However, while it is uncontroversial that certain case markers would have disappeared due to the operation of well-established sound laws, phonological change cannot explain the loss of case distinctions on its own. There are several distinctions in the individual languages that should have remained intact after all known regular sound changes, yet were lost anyway. For example, as discussed above, the genitive has been lost as a productive case in many German dialects, but there are no sound changes operating in these dialects that would have deleted the final -s characteristic of masculine and neuter genitive markers in the language. The fact that the genitive is so clearly on the retreat must, thus, have a non-phonological explanation. Figuring out what such non-phonological causes there might have been for the gradual dismantling of case distinctions in the Gmc languages turns out to be rather difficult. Barðdal (2009) and Enger (2013) offer extensive and convincing critiques of traditional proposals, especially the purely phonological ones, though their own alternative proposals remain largely suggestive. What exactly has driven the consistent reduction of the case systems of the Gmc languages must thus still be regarded as an open question.

13.3 Distributions Let us now consider how the various cases discussed above are employed in the Gmc languages. I cannot hope to cover the distinct details of all of the historical stages of the various languages, and so I will focus here on a few representative snapshots. Let us start with a brief, pre-theoretical description 5

I am setting aside here creole languages with English or some other Gmc language as a lexifier, some of which lack such distinctions.

289

290

T HO M A S M CFA DD E N

of what the six cases must have been used for in PGmc. Since PGmc is prehistoric, and we care here about abstract patterns rather than specific forms, I will use English examples, annotated with the cases that would be expected. The nominative was the unmarked case for subjects in prototypical finite clauses, and for nominals predicated of the subject (2). The vocative was used for direct address (3). The accusative was the unmarked case for direct objects and was used for certain adverbial relations, including those describing extent direction notions, sometimes accompanied by prepositions (4). The genitive was used to mark NPs occurring inside of other NPs, including possessors and complements of nouns, as well as for certain adverbial relations (5). The dative prototypically marked experiencer, recipient, and beneficiary objects, as well as various adverbial relations including locations, usually with local prepositions (6). The instrumental, as its name implies, was used to mark instruments and a range of other adverbial relations, often with prepositions (7). (2)

This tree.NOM is an oak.NOM.

(3)

I.NOM am here, friend.VOC.

(4)

We.NOM baked bread.ACC all day.ACC.

(5)

Robin’s.GEN child.NOM watched the baking.ACC of the bread.GEN.

(6)

Joe.NOM baked Hannah.DAT bread.ACC in the oven.DAT.

(7)

We.NOM sliced the bread.ACC with a sharp knife.INST.

This description covers only the broadest outlines, and the actual details will have been far more interesting. In particular, many subjects would have been accusative, dative, or genitive, and objects could similarly be nominative, dative, or genitive rather than accusative, with the choice depending on a combination of thematic and lexical issues. Let us move then to the actually attested languages, where these details can be explored based on real data. The discussion will be based primarily on Modern Icelandic and German as these have been (by far) the most extensively studied and can together be taken as representative in the broad strokes of what can be found across the family. Against this background, I will then discuss interesting points of variation and deviation from these patterns in specific languages. It is generally agreed that the determination of the case of an NP, both in Gmc and cross-linguistically, depends on the interaction of syntactic, semantic, and lexical factors. A first cut can be made between argument NPs and adjunct / modifer ones. Case must carry a particularly heavy load with non-argument NPs, since their relationship to the rest of the clause cannot be determined on the basis of their relationship to a selecting predicate. In languages without rich case systems, they are typically introduced by adpositions that indicate the kind of modification they are involved in. In case-rich languages, including the older Gmc ones, such essentially semantic information can

Case in Germanic

instead be supplied by the choice of case, e.g., with instrumental case signalling that the NP describes the means or instrument by which an event is carried out. In such instances the semantic aspect is clearly primary, thus we typically refer to case assigned in this way as semantic case (see, e.g., Nikanne 1993, McFadden 2004, chapter 3). Turning to arguments, certain initial generalizations can be stated either in terms of the place of the NP within the argument structure of the verb, or in terms of its grammatical function, without reference to semantic factors or the lexical identity of the verb. These generalizations have important exceptions, which depend on semantic or lexical information (or a far more subtle syntactic analysis), but the case of the majority of argument NPs is covered by these purely syntactic or structural factors, which is thus usually referred to as structural case. There are several different ways in which the rules of structural case assignment can be stated, based on distinct theoretical conceptions of case assignment. Here I will try to abstract away from these differences and adopt a reasonably theory-neutral presentation of the basic patterns, coming back to the theoretical questions that are particularly interesting for Gmc later in this section and the next.6 As nominative-accusative systems, the Gmc languages have two generally agreed upon structural cases: the nominative and the accusative.7 The highest argument NP in each finite clause (what can be thought of as the subject or the thematically most prominent NP), receives the nominative. If there is an additional, lower argument NP (the object, lower in the thematic hierarchy), it receives the accusative. This means that the structural accusative case will only be assigned in clauses with at least two argument NPs, since the nominative takes precedence. Thus the sole argument of all kinds of intransitives, including passivized transitives, will receive nominative. Nonfinite complement clauses introduce interesting complications. In contexts where they can have an overt subject, this often counts for case purposes as belonging to the matrix finite clause. The notional subject or highest argument of the nonfinite clause then counts as being thematically lower than the highest argument of the matrix finite clause. Thus, in the German example in (8), the highest argument wir ‘we’ gets nominative, while the notional subject dich ‘you’ of the nonfinite clause gets accusative, as does the embedded object deinen Aufsatz ‘your paper’. (8)

6

Wir lassen [dich deinen Aufsatz u¨berarbeiten]. we.NOM let you.ACC your paper.ACC revise ‘We’re going to have you revise your paper.’

Questions of substance include the extent to which grammatical functions must be assumed to play a role, whether structural case assignment targets specific structural positions, and whether it is assigned by verbs and other heads in the clausal structure, or rather on the basis of hierarchical relationships between NPs within a structural domain. See among many others Zaenen et al. (1985), Yip et al. (1987), Marantz (1991), Wunderlich (1997), Reuland (2000), Stiebels (2002), McFadden (2004), Sigurðsson (2006), Baker and Vinokurova (2010), Baker (2015), Levin and Preminger (2015), Baker and Bobaljik (2017).

7

Ergative-absolutive and other systems work differently, though the building blocks involved are arguably the same (see e.g., Stiebels 2002, Baker 2015).

291

292

T HO M A S M CFA DD E N

Another complication arises when a predicate takes two arguments that are necessarily equated, i.e., with copulas and verbs with meanings like ‘become’. In such instances, both argument NPs generally get nominative case. An important property of structural cases, much discussed in the literature on Gmc, is their ability to alternate according to the argument structure of the clause in which they appear (see Alexiadou and Scha¨fer, Chapter 20). Thus, e.g., we see the accusative on the embedded subject dich in (8) alternating with the nominative du in (9), when the relevant clause appears unembedded and finite. Similarly, the accusative object deinen Aufsatz in (8) and (9) becomes nominative dein Aufsatz under passivization in (10). (9)

Du u¨berarbeitest deinen Aufsatz. you.NOM revise your paper.ACC ‘You’re revising your paper.’

(10)

Dein Aufsatz wird u¨berarbeitet. your paper.NOM becomes revised ‘Your paper is being revised.’

This is precisely what we should expect given the description of structural nominative and accusative above. Accusative is only assigned to an NP that is thematically and structurally lower than some other NP in the same finite clause.8 The highest NP, i.e., the subject, will always get nominative. If we do something to the argument structure of a clause to add or subtract an argument at the top, this can then affect the case of an NP below it. Things get trickier when a simple clause has more than two arguments. By far the most common pattern in the Gmc languages is that we still have one nominative and one accusative, with the addition of a dative. However, it is not clear whether the assignment of this dative should be regarded as a structural case. First, while it is usually fairly clear in transitive clauses which of the two arguments is structurally higher, and this relative height matches reliably with the assignment of nominative and accusative, things are often more difficult with ditransitives. Even where we do have good arguments regarding the hierarchy of the arguments, we do not find a consistent relationship with the cases they are assigned. In German, for example, even though most ditransitive verbs involve a nominative subject above a dative object, which is in turn above an accusative object, there is also a class (including verbs like aussetzen ‘expose’) where the accusative object is higher than the dative one (Wegener 1991, Haider 1993, Meinunger 2000, McFadden 2004, Cook 2006, McIntyre 2006), and similar patterns have been reported for Icelandic (see Thra´insson 2007, section 3.1.2.3, with references). Case assignment in such ditransitives thus cannot be determined purely on the basis of structural hierarchy, but must make 8

This is essentially the import of Burzio’s Generalization (Burzio, 1986), which has been incorporated into several diverse approaches to case assignment. See the contributions in Reuland (2000) for discussion.

Case in Germanic

reference to other information. One option is that case assignment in such instances is simply regulated by lexical information, i.e., it would be part of the lexical specification of aussetzen that it assigns accusative to its higher object and dative to its lower one, while geben ‘give’ would be specified to assign case the other way around. A more interesting possibility is that case assignment in these instances is guided by semantic or thematic factors. Indeed, there are clear – if tendential – generalizations about the semantics of the arguments that will be realized as high versus low datives (see the sources cited above). An important question then is how we should understand the assignment of these various datives theoretically. One widespread idea is that they should have a status distinct from the structural cases nominative and accusative, which are assigned on purely syntactic grounds, independent of semantic and lexical concerns. The dative, on the other hand, is regarded as an inherent case. It is assigned to specific arguments of specific (classes of) verbs, in a way that is independent of the presence or absence of other arguments in the same argument structure. This accounts for the fact that such inherent cases are generally unaffected by argument structure alternations like passivization, as shown by the German examples in (11): (11) a. Die Chefin glaubte mir den Bericht nicht. the boss-FEM.NOM believed me.DAT the report.ACC not ‘The boss didn’t believe my report.’ b. Die Chefin glaubte mir nicht. the boss-FEM.NOM believed me.DAT not ‘The boss didn’t believe me.’ c. Mir wurde der Bericht nicht geglaubt. me.DAT was the report.NOM not believed ‘My story wasn’t believed.’ d. Mir wurde nicht geglaubt. me.DAT was not believed ‘I wasn’t believed.’ (11a) shows a ditransitive with a nominative subject, dative indirect object, and accusative direct object. The verb glauben ‘believe’ has the useful property that its accusative object can be left off, but the indirect object remains and is crucially still assigned dative, as in (11b). If the assignment of the dative were, e.g., dependent on the presence of a lower accusative argument, we would expect it to change to an accusative in this context, but it does not. We learn something else interesting if we passivize clauses with glauben, as in (11c) and (11d). Again, the dative remains dative, even though the nominative subject above it has been removed – in stark contrast to the structural accusative which, as we saw in (10) above, changes to nominative in such contexts. Indeed, an argument does become nominative in (11c) as well, but it is the accusative lower object

293

294

T HO M A S M CFA DD E N

den Bericht ‘the report’ from (11a). All of this suggests that the dative is independent of and irrelevant to the assignment of nominative and accusative. This is further reinforced by (11d), where we leave off the lowest accusative object and passivize, getting rid of the highest nominative argument, leaving just the dative argument, which remains dative. Additionally relevant here is the fact that there are a number of verbs in German, Icelandic, and the various older Germanic languages which are not ditransitive and yet assign dative to an argument. Well-known examples in German include folgen ‘follow’, geho¨ren ‘belong to’ and helfen ‘help’, shown in (12a). Such structures make it especially clear that the dative is independent of any additional accusative object, and here too, passivization has no effect on the dative, as in (12b). (12)

a. Die Chefin hat mir geholfen. the boss-FEM.NOM has me.DAT helped ‘The boss helped me.’ b. Mir ist geholfen worden. me.DAT is helped been ‘I was helped.’

The broadly agreed-upon theoretical understanding of facts like these is that the structural and inherent cases are assigned in distinct ways that render them independent of each other to a considerable extent, and furthermore that in instances where the conditions for either a structural or an inherent case would be met, it is the inherent that takes precedence. The basic idea, again abstracting away from many important theoretical differences, is that any inherent cases are assigned to the various NPs first, based on their thematic relationships with particular predicates. Because they come first, they supersede any structural cases, and because they are determined by thematic relationships with predicates, they are not sensitive to other syntactic factors, including the presence or absence of other arguments, and thus will be unaffected by things like passivization. Then the rules for structural case get to apply, based on the syntactic positions and relationships with relevant functional heads of any arguments that have not received an inherent case. Due to their sensitivity to such structural relationships, the structural cases can be affected by things like passivization. There are several ideas about how the structural and inherent cases are kept orthogonal from each other. The Case in Tiers approach (Yip et al. 1987, etc.) pursues the idea, inspired by autosegmental phonology and morphology, that they operate on distinct representational tiers, much like consonants and vowels in accounts of Semitic templatic morphology. Work in GB (e.g., Chomsky 1981, Haider 1985) associated inherent case (however indirectly) with θ-role assignment, and thus with lexical heads and D-structure, whereas structural case was independent of θ-roles and could be assigned at or on the way to S-structure. Some work in the

Case in Germanic

Dependent Case theory tradition initiated by Marantz (1991) explores the possibility that inherent case creates a kind of structural opacity, making an NP it is assigned to invisible for the subsequent computation of structural case, e.g., Richards (2010), Baker (2015), McFadden (2014). An important question is the extent to which inherent case is predictable, e.g., based on semantic or thematic roles. The dative assigned to traditional indirect objects, in particular, shows clear and consistent semantic correlations, is mostly predictable and is typically productive.9 Simplifying a bit, if a verb takes two objects, one a theme or patient and the other a beneficiary, recipient or goal, the latter will quite reliably receive dative (and the former accusative) in all Gmc languages with a solid dative-accusative distinction. Moreover, there are clear generalizations that can be made about when the sole object of a verb receives dative rather than the expected accusative. Such datives are commonly beneficiaries (helfen ‘help’ and cognates), experiencers (gefallen ‘please’) or locations/goals (folgen ‘follow’ or begegnen ‘run into’). Nonetheless, the correlations between cases and semantic or thematic roles are notoriously approximate. It is not even particularly difficult to find pairs of verbs with fairly close semantics but rather different case frames, e.g., German unterstu¨tzen ‘support’ takes an accusative object in contrast to the dative-assigning helfen ‘help’.10 These facts have always been taken to show that the relationship between cases and semantic/ thematic roles cannot be direct. Indeed, the traditional Government and Binding view was that inherent case was associated with θ-roles in that an NP could only be assigned inherent case by the lexical item that assigned its θ-role, but it was not thought that specific cases were tied to specific θ-roles (see Haider 1985 for explicit and representative discussion of this issue). The quite traditional idea, then, is that inherent case – unlike the semantic case we discussed above – is not tied directly to an NP’s semantics, but is assigned by the lexical verb. We can state generalizations about the kinds of verbs that take dative objects, but in the end it is a fact about the verb helfen that it assigns dative to its beneficiary argument whereas unterstu¨tzen does not. Once we think in these terms, an interesting analytic possibility arises. If we have recourse to lexically specific statements, we can use these to take care of true exceptions, allowing us to be bolder about building the general patterns into the theory. We can propose that there actually are real and direct connections between particular θ-roles and particular cases, e.g., that recipient arguments are, by default, assigned dative case. It just happens that specific lexical verbs can be specified with exceptional patterns that supersede the general thematic patterns. It is thus frequently suggested that in addition to structural and inherent case, we should recognize lexical case as a distinct type. Woolford (2006: 112), who 9

The discussion to follow could also be applied to inherent genitives and accusatives, but the patterns there are far less consistent and well-understood, so I will leave them aside here.

10

See Sigurðsson (2009: 270) for a list of similar pairs from Icelandic.

295

296

T HO M A S M CFA DD E N

explores these issues in depth, puts things rather nicely: “It has often been noted that some instances of nonstructural Case are truly idiosyncratic, while others are quite regular and predictable . . . There is general agreement that the truly idiosyncratic Cases . . . are lexically selected by individual verbs; it is thus appropriate to label these as instances of lexical Case. Other instances of nonstructural Case . . . are much more regular and predictable, and fit the notion of inherent Case as Case that is inherently associated with θ-marking.” This could then mean that the divide between the structural cases and the more regular and predictable inherent ones is not so sharp as was thought. Indeed, a number of researchers have argued that at least some datives in Gmc should be analyzed as structural in some sense (see Wegener 1991 for insightful discussion and foundational empirical findings from German). Broadly speaking, we can distinguish two different “structural” approaches to Gmc datives. The first is exemplified especially by work in Lexical Decomposition Grammar (Wunderlich 1997, Stiebels 2002, etc.), in which the contexts for assignment of the different cases are expressed using features that make reference to the presence of additional higher or lower arguments in an argument structure hierarchy. The accusative is [+hr] (“there is a higher role”), while the nominative is simply underspecified. The dative is then specified as [+hr, +lr], meaning that it is assigned to an argument position that has additional arguments both above and below in the structure.11 This characterizes the highly regular and predictable dative found with ditransitives. What about the various types of monotransitive verbs with dative arguments? With helfen, e.g., the object should be [+hr] because of the higher subject, but should lack [+lr] since there is no lower argument, and thus we incorrectly expect an accusative. Again, this can be addressed by additional lexical specifications. Wunderlich (1997) proposes that, on top of the values for [+hr/+lr] supplied to arguments based on their place in the argument structure, lexical verbs can specify additional features for particular θ-roles. Thus helfen has a lexical [+lr] on its lower θ-role, which combines with the [+hr] based on position, yielding an argument that will be linked as a dative. Another concern for this approach is why datives do not seem to undergo alternations with other cases the way that accusatives become nominative under passivization. It has been pointed out, however, that datives in German do seem to undergo alternations in other circumstances. Most famously, while the dative remains fixed in the standard passive, a distinct ‘recipient passive’ can be constructed (at least in colloquial varieties) with ditransitives using kriegen ‘get’ or bekommen ‘receive’ as the auxiliary (see Wegener 1991, Abraham 1995, Cook 2006, and many others), as shown by the alternation in (13).

11

Baker (2015) proposes a somewhat similar view of certain kinds of datives within the framework of Dependent Case theory, but it is not clear that he would extend this to any of the Gmc languages.

Case in Germanic

(13)

a. Die Chefin schickt mir den Bericht per Post. the boss-FEM.NOM sends me.DAT the report.ACC by post ‘The boss will send me the report by (snail) mail.’ b. Ich kriege/bekomme den Bericht per Post geschickt. I get/receive the report by post sent ‘I’ll be sent the report by (snail) mail.’

There has also been extensive discussion of dative alternating with nominative in Icelandic under certain circumstances. While Icelandic datives are generally retained in the passive, the dative assigned to direct objects of certain monotransitives alternates with the nominative in what is sometimes termed the middle or anticausative, which is formed with an -st suffix (Svenonius 2006, Sigurðsson 2009, etc.), exemplified in (14) from Wood (2015): (14)

a. A´sta splundraði ru´ðunni. ´Asta.NOM shattered window.the.DAT ‘A´sta shattered the window.’ b. Ru´ðunni var splundrað. window.the.DAT was shattered ‘The window was shattered.’ c. Ru´ðan splundraðist. window.the.NOM shattered-ST ‘The window shattered.’

Icelandic also has a ‘get’-passive that shows a dative-to-nominative alternation similar to the German recipient passive (Sigurðsson and Wood 2013). Such alternations can be taken as evidence that the relevant datives are not assigned purely based on a thematic relationship to a particular verb as traditionally assumed. That relationship is presumably unmodified in the recipient passives or the Icelandic anticausative, and so we would expect the dative to be preserved. Rather, the dative seems to be assigned based on the structural status of the argument, which is related to the thematic properties of the relevant verb but can be modified by the right type of argumentstructural operation. However, it has also been argued that the alternations here do not have the same status as those between accusative and nominative. Sigurðsson and Wood (2013), e.g., propose that in the Icelandic ‘get’passive, the nominative does not actually correspond to the dative verbal argument in transitive and normal passive constructions, but rather is introduced as an argument of ‘get’ itself. Thus what we see would not constitute evidence that the dative is a structural case like the accusative. The second broad approach to the dative as a “structural” case is a bit more of a compromise. Like the traditional inherent case approach, it recognizes that even these datives are assigned differently than structural accusatives, accounting for their restricted ability to alternate. But like the approach just described, it also recognizes that many datives are regular and predictable based on argument-structural and thematic considerations, and thus posits

297

298

T HO M A S M CFA DD E N

a means of assignment that is more anchored in the abstract syntax than in lexical peculiarities of specific verbs. Approaches in this vein typically rely on the more nuanced structural analyses of the traditional verb phrase in recent work, with additional functional heads responsible for various aspects of argument structure and thematic detail. The most popular proposal is that recipient and beneficiary arguments are introduced not by the lexical verb itself, but by a special Applicative functional head, following Marantz (1993) and Pylkka¨nen (2002). Note how this approach can accommodate the intermediate status of predictable datives. Dative can be assigned to the DP in SpecApplP in a way that essentially follows traditional ideas about inherent case: It is purely on the basis of the thematic or first-merge relationship of the DP to the relevant syntactic head, does not depend on any other properties of the clause, and thus will take precedence over the structural cases and be unaffected by things like passivization. However, the fact that it comes from the Appl head and not from the lexical verb accounts for the thematic predictability and productivity. Typical datives are not assigned by specific lexical verbs with their various idiosyncrasies, but by a functional head with a consistent syntactic and semantic contribution. Properties of the lexical verb only come in for the compatibility of various verbs with the Applicative structure. Cross-lingustic differences in the availability of certain kinds of datives can then be attributed to parametric differences in both the availability of specific Applicative and other argument-introducing heads, and in which of these are specified to assign dative to their arguments. See, e.g., McFadden (2004, 2006), McIntyre (2006), Scha¨fer (2008), Sigurðsson (2009), Wood (2015) for applications of this type of analysis to datives in Germanic. Against this background, we can now say a bit more about how specific Gmc languages differ in the particulars. For space reasons I will go into very little detail here, restricting myself to pointing out broad patterns of distinction and referring the reader to the vast existing literature for details. What we know about Gothic is relatively limited. This is largely because of the limited size of the corpus of surviving Gothic texts and the fact that they are essentially all translations, meaning that it is not always clear what reflects actual native Gothic grammar as opposed to the Greek original.12 For basic description of the use of the cases in Old High German, see Schrodt (2004). The situation is considerably better for Old English (OE). Van Kemenade (1987) and especially Allen (1995) provide useful overviews, including discussion of whether particular uses of particular cases should be regarded as structural or inherent as well diachronic developments. For Icelandic, both Old and Modern, a considerable volume of high quality work is available. Thra´insson (2007) provides systematic coverage of the case patterns in the context of a general description of the syntax of the language. Barðdal (2001, 2009) gives extensive discussion of the use of 12

See, e.g., the discussion of the absolutive construction (Dewey and Syed 2009, with references), the topic that has received the most attention when it comes to the distribution of cases in Gothic.

Case in Germanic

the different cases in both Old and Modern Icelandic, with useful discussion of the observed changes. Sigurðsson’s work (Sigurðsson 1989, 2006, 2009, etc.) contains a wealth of insightful observations and discussion on the distribution of the cases and how these interact with various important syntactic details. Maling (2001, 2002) catalogues the syntactic, thematic, and lexical patterns in the use of the dative in particular. Some notable peculiarities of Icelandic case distribution include: the dative of objects of ballistic motion (Svenonius 2002, Jo´nsson 2012); the so-called ‘new passive’ or ‘new impersonal’ construction, where accusative is unexpectedly retained on objects after demotion of the subject (Maling and Sigurjo´nsdo´ttir 2002); so-called ‘fake accusatives’, which appear unexpectedly on the sole arguments of unaccusatives under certain circumstances (Sigurðsson 1989, 2006, Scha¨fer 2008, Wood 2017); the retention of an especially large number of verbs with one or more inherent case-marked arguments (see especially Barðdal 2009 for comparison with other Gmc languages as well as Yip et al. 1987, Sigurðsson 1989, Maling 2002, Jo´nsson 2003 among many others); the evidence (from agreeing secondary predicates) that null even DPs including PRO subjects of nonfinite clauses are assigned case (Sigurðsson 1991 and much following work). Careful discussion of case distribution in German can be found, e.g., in Haider (1985, 1993), Czepluch (1996), Maling (2001), Haider (2010), and several contributions in Hole et al. (2006). Some points of interest include the fact that the overt expletive es in certain contexts is sufficiently argumental to license accusative on a lower argument, e.g., in the expression es gibt ‘there is’, and in constructions that are vaguely akin to Icelandic fate accusatives, see Haider (2000), Scha¨fer (2008). There is also evidence for assignment of case to silent PRO, but it is far more restricted than in Icelandic (Haider 2010: 293ff.). A moderate number of verbs still assign inherent case to one of their arguments, but far fewer than in Icelandic, with a far more restricted set of patterns retained – inherent accusative is restricted to a handful of verbs, and inherent genitive is almost entirely gone (Haider 1985, Yip et al. 1987, Maling 2001, McFadden 2004, Barðdal 2009). Note that this latter point is part of a broader historical tendency across Gmc of the gradual decrease in the frequency and variety of inherent and lexical case patterns, which correlates only approximately with the general reduction in case distinctions in a language. See especially Barðdal (2009) and Eyþo´rsson et al. (2013) for comparative discussion and Allen (1995) on diachronic patterns in English, in particular the surprising longevity of oblique subject constructions well after the loss of the dative-accusative distinction in the language. One further point of variation in Gmc that is relevant to the preceding theoretical discussion is what structural case will appear on the lower argument of a monotransitive verb when the higher argument receives an inherent case. In German and Icelandic, it is uniformly nominative, as in the German example in (15a), which supports the idea that inherent case-

299

300

T HO M A S M CFA DD E N

marked NPs are invisible for the calculation of structural cases. Famously, however, in Faroese they receive structural accusative, as in (15b): (15)

a. Mir gefa¨llt dieser Song. me.DAT please this song.NOM 'I like this song’ b. Mær da´mar væl hasa bo´kina. me.DAT like well this book.ACC ‘I like this book.’

Barðdal (2009) reports analogous patterns for stages of Middle English (ME) and Old Swedish, so this is clearly something available as a general option. In such languages, things must be set up so that inherent case-marked NPs are visible after all for the assignment of structural accusative. This brings us finally to the general distribution of cases in the Gmc languages which retain only the two-way distinction between nominative and general oblique. We can distinguish here two broad groups (Sigurðsson 2006). In the first, including Swedish, Dutch, Afrikaans, and some varieties of Norwegian, the use of the two cases resembles fairly closely what is observed in the languages with richer case systems, except that the undifferentiated oblique takes over roles played elsewhere by both accusative and dative. In particular, pronouns are in their nominative form when used predicatively and in various types of fragmentary or elliptical contexts when the corresponding pronoun in a nonfragmentary version would have been nominative, as seen in the Swedish examples in (16). (16)

a. Det a¨r vi it is we.NOM ‘It is us.’ b. Vem gjorde detta? Jag. who did this? I.nom ' Who did this? Me.’

For these languages, it seems plausible to adopt essentially a simplified version of the analysis used for German and Icelandic, with a reduced inventory and no need for inherent or lexical case. The second group, which includes at least English, Danish, and other varieties of Norwegian, looks like the first group as long as we look at simple subject and object pronouns in full clausal environments. However, pronouns are oblique when used predicatively or in fragmentary contexts, irrespective of what shape they would have taken in the nonfragmentary analogues: (17)

a. It’s us/*we. b. Who did this? Me/*I.

Similarly, unlike in the case-rich Gmc languages and the first group of casepoor ones, we find oblique forms of pronouns in these languages when

Case in Germanic

they are conjoined, modified, left-dislocated, or appear as complements of comparatives (see, e.g., Schu¨tze 2001, Quinn 2005, Parrott 2007, 2009 for data and discussion).13 This has led to rather different accounts of case assignment in these languages. A central insight developed by Schu¨tze (2001), building on ideas from Marantz (1991) is that the two groups of languages differ in terms of which case functions as the default. In the first group, as in the case-rich languages, nominative is the default case, while oblique has to be explicitly assigned where certain conditions are met. In the second group, this relationship is reversed: Oblique is the default case, and nominative is assigned by special rule. The odd and limited distribution of the nominative forms in these languages comes out of the fact that this rule is highly specific, apparently making reference to a tight relationship with the finite subject position. Schu¨tze (2001) retained at least the outlines of a nominative-accusative system behind the scenes in these languages, but subsequent researchers have pursued the idea that they have actually moved to a different type of system. McFadden (2004) and McFadden and Sundaresan (2011) argue that while all nominatives in languages like German and Icelandic are default case, all obliques in English are default case as well. That is, there is no structural oblique in the second group of case-poor languages anymore, and since the “nominative” is assigned according to rather different principles than those observed elsewhere, vestigial case languages like English and Danish should no longer be classified as nominative-accusative in typological terms. Parrott (2007, 2009) goes one step further, arguing that in fact the alternations that we see in such languages no longer reflect case features at all. Rather, the alternations in pronominal forms should be analyzed as instances of contextual allomorphy. This explains why the distribution of the forms no longer conforms to what is expected for nominative-accusative languages as well as a series of quite peculiar restrictions on the distribution of the forms that make reference to things like linear order and adjacency.

13.4 Interactions In this final section I turn to a selection of topics where case interacts with other grammatical issues and the Gmc languages have played an important role in related theoretical debates. The discussion will work through three broad, interconnected themes. I will start by taking up in more detail the relationship between case and grammatical functions and in particular the discussion of so-called quirky case. This will naturally lead into 13

The relevant facts in both English and Danish are subject to heavy inter- and intra-speaker variation, conditioned by strong prescriptivist pressure toward something a bit more like the Swedish system. For details and discussion I refer the reader to the excellent work by Quinn (2005) and Parrott (2007, 2009) and further sources cited there.

301

302

T HO M A S M CFA DD E N

a consideration of the notion of abstract Case, and in particular how morphological case may – or may not – relate to conditions on the licensing of NPs in different syntactic positions. Resulting questions about how syntax and morphology interact will bring us to a discussion of how richness of morphological case in a language correlates with other syntactic properties. In Section 13.3 I repeatedly made informal reference to grammatical functions like subject and object. Note, however, that in my more careful formulations, I spoke in terms of thematically or syntactically higher and lower arguments. There are two reasons for this. First, most syntactic work in Chomskyan varieties of the Generative tradition does not admit grammatical functions as theoretical primitives. Subject and object are at most derivative notions, defined in terms of structural positions, and used as descriptive labels for convenience rather than components playing a crucial role in the formulation of analyses. Thus, in early Minimalist terms a DP would get nominative not by virtue of being the subject, but because it occupies Spec-TP, finite T bears a nominative feature, and case is assigned via Spec-head agreement. Second, as a matter of fact, it is rather difficult to develop operationalizable characterizations of the grammatical functions that work well across languages – even just the Gmc languages – and even when we think we can get a grip on subjects in a particular language, the relationship between grammatical functions and cases rarely turns out to be one-to-one. The working out of these issues took place in large part on the basis of Gmc languages and in particular the phenomenon of quirky case. The assumption from traditional grammar was that subjects are assigned nominative case and trigger agreement on the finite verb. Thus, by and large, if an argument was nominative, it could be regarded as the subject, and if it had some other case, it would have to be some sort of object. Certain constructions, found in Gmc but also many other languages, have always presented a challenge for these assumptions, because the case marking we see is out of whack with our expectations about what it is to be a subject. The German example in (18) is typical: (18)

Mir gefallen deine neuesten Aufsa¨tze. me.DAT please.3PL your newest papers.PL.NOM ‘I like your most recent papers.’

The verb gefallen ‘please’ or ‘like’ takes a dative experiencer argument and a nominative theme. The nominative also triggers person and number agreement on the verb. We would thus be apt to take the theme as the subject here, and the experiencer as the object. Now, if we define subject and object purely in terms of case, and have no further expectations about their properties, then grammatical functions will be of no real use, entirely redundant with the cases they receive. For them to be worthwhile theoretical notions, they must be predictive of other properties or behaviors along the lines proposed in works like Keenan (1976). It turns out then that

Case in Germanic

if we develop such independent criteria for identifying subjects and objects and then look to see how these match up with the cases that DPs receive, we get a surprise. The remarkably clear result, at least for Icelandic, is that case and subjecthood are orthogonal to each other (Andrews 1976, Thra´insson 1979, Andrews 1982, Zaenen et al. 1985, and much subsequent work). If we take an Icelandic example like (19) from Zaenen et al. (1985), roughly parallel to German (18), a series of diagnostics show that the dative argument henni ‘her’ behaves like the subject, while the nominative Olafur is the object. (19)

´ lafur Henni hefur alltaf þo´tt O leiðinlegur. her.DAT has always thought Olaf.NOM boring.NOM ‘She has always thought Olaf boring.’

This dative – but not this nominative – undergoes raising, can bind a reflexive, appears immediately after the finite verb in a verb-second clause when another constituent is fronted, can appear in initial position in nonV2 embedded clauses, can be elided in certain conjunct structures and can be null PRO in control infinitives. In all of these respects, the dative behaves like the uncontroversial subjects and unlike the uncontroversial objects of prototypical, highly transitive clauses. All of this means that the notion of subject must be independent of case, at least in Icelandic. It has usually been thought, however, that this is a point of difference with German. The same set of tests applied in the latter language to examples like (18) do not clearly indicate that the dative argument should be analyzed as the subject and the nominative as the object (Zaenen et al. 1985ff.). In fact, the classical treatment is that inherent cases like the dative in (18) actually prevent an NP from becoming the subject. Under this view, the fact that such datives could be subjects in Icelandic presented a challenge, which engendered a series of interesting debates centered around comparative Gmc data, ultimately leading to proposals that radically alterered how case is understood within syntactic theory. To understand this discussion, we need to step back and consider the theory of abstract Case that constituted a central module of Government and Binding theory. Following a suggestion made by Jean-Roger Vergnaud (Vergnaud 1977), Chomsky (1980, 1981) proposed that Case – as an abstract syntactic notion, with distinctive capitalization – is universal and plays a central role in driving and constraining syntactic derivations, though languages may differ in whether and how it may translate into overt morphological realization (see Lasnik 2008 for a detailed history of relevant ideas). The central assumption that gets the system off the ground is the Case Filter, which states that an NP is only licensed to appear overtly if it has received abstract Case. If an NP starts out in a position where it cannot receive Case, then either it must move to a position where it can, or it must remain silent (perhaps requiring further licensing), or the result

303

304

T HO M A S M CFA DD E N

will not be a well-formed sentence. Case Theory in this form provided a unified approach to a series of disparate facts about the overt distribution of NPs, including conditions on A-movement and the behavior of different types of infinitival clauses. The fundamentals were developed on the basis of English, which again shows little overt morphology, but with the idea that it should be applicable as well to case-rich languages where the different cases would be more easily identifiable on the surface. Indeed, at least in outline, the patterns observed in languages like German seemed to confirm this. One central aspect of classical Case theory is that movement of NPs to a derived subject position like Spec-TP is driven by their need for Case. Given that inherent Cases are determined based on the thematic position of an NP and are unaffected by A-movement and passivization, we make a straightforward prediction. If we passivize a normal transitive verb, structural accusative on its object will disappear, and the object will be forced to move to subject position to receive nominative Case. If, however, we have a verb that assigns inherent dative to its object, then passivization will leave the dative Case unaffected, and the object will not be driven to move to subject position for nominative Case. This fits in nicely then with the evidence that datives and other inherent case-marked NPs do not behave like subjects in German, even under passivization. But now we see why Icelandic presents a problem. Arguments that receive dative or any other inherent Case from the verb should not have to move to Spec-TP for Case, and thus should not behave like subjects. Initially, attempts were made to minimize the import of the Icelandic patterns by emphasizing the idea that abstract Case has an indirect relationship with overt morphological case and regarding the degree of that indirectness as a quirk of Icelandic. The standard analysis was that oblique subjects in Icelandic actually get structural nominative abstract Case, but then get a quirky morphological dative (or accusative or genitive) case on top of this (see, e.g., Freidin and Sprouse 1991, Chomsky 2000). German, on the other hand, is more transparently well-behaved, in that the morphological case matches up with the abstract inherent Case. This view has subsequently come under intense fire. For one thing, it has been put forward that the contrast between Icelandic and German has been overstated or misinterpreted. Barðdal (2002 and Barðdal and Eytho´rsson (2003), e.g., argue that, contrary to prior claims, the relevant German datives do behave as subjects. McFadden (2004, 2006) suggests that the difference between the two languages is not the status of the dative arguments themselves but in the existence of a clear syntactic subject position that plays an important role in clausal syntax and thus can be easily tested for. The crucial point for this work is that most of the subjecthood diagnostics for Icelandic are simply not applicable in German, i.e., they don’t single out nominative subjects either. Subsequent research has furthermore identified quirky case-marked subjects in a number of languages around the world (see, e.g., the contributions in Bhaskararao and Subbarao 2004). It is thus not viable to treat the mismatch between grammatical functions and syntactic positions

Case in Germanic

on the one hand, and morphological cases on the other, as a quirky phenomenon restricted to a few languages like Icelandic. It may still be useful to adopt some version of abstract Case theory to model the syntactic distribution of overt DPs, but it has become increasingly clear that the distribution of specific morphological cases in languages that have them will be at best indirectly related to this. Thus we might wonder whether Case is really an appropriate label for whatever licensing requirement drives DP distribution (see, e.g., McFadden 2004, Sigurðsson 2009). This has led to a divide in theoretical treatments of case. Those who are primarily interested in the syntactic distribution of DPs usually maintain updated versions of abstract Case theory (typically following Chomsky 2001), where specific (abstract) Cases are assigned via Agree with specific functional heads. Those who are concerned with the distribution of actual morphological cases, on the other hand, tend to adopt theories with a pedigree going back to Yip et al. (1987), Marantz (1991), Kiparsky (1992, 2001), based on the insight that at least the structural cases are assigned in a way that depends on how a DP relates to other DPs in a local domain, not specific positions or functional heads.14 This includes work in the Case in Tiers framework (Yip et al. 1987), under the linking theory of Lexical Decomposition Grammar (Wunderlich 1997, Stiebels 2002), and more recently especially Marantz (1991)’s Dependent Case approach (McFadden 2004, Sigurðsson 2006, McFadden 2009, Baker and Vinokurova 2010, Baker 2015, Levin and Preminger 2015, Levin 2017, Baker and Bobaljik 2017). The final topic for this chapter pertains to how morphological case relates to the syntax from a concrete and comparative perspective. Abstract Case theory posits certain universal points of contact between Case and syntactic derivations, but as we have discussed extensively in our consideration of morphological case in Gmc, specific languages differ considerably in the cases they distinguish and how they are concretely realized. We might reasonably ask then whether and how such differences in morphological case correlate with differences in the syntactic properties of the languages, and by extension whether there are causal connections between the loss of case distinctions that we have discussed in the languages and any observable syntactic changes. We can start with the traditional observation that languages with rich case marking, e.g., Latin, tend to allow greater word-order freedom than case-poor ones, e.g., English. There have been a number of attempts to derive such observations from deep theoretical principles, often based on data from Gmc languages. For example, Neeleman and Weerman (1999), building on Lamontagne and Travis (1987), propose that abstract Case requires close syntactic proximity between a DP and its Case-assigner, and thus relatively rigid ordering, unless the Case is overtly realized. They use this to account for why case-poor modern Dutch has more 14

Indeed, a commonly repeated witticism in recent years (originating, as far as I know, with Mark Baker) is that “‘Standard Case Theory” is primarily assumed by people who do not work on case.

305

306

T HO M A S M CFA DD E N

restricted positioning of objects than either modern German or older stages of Dutch that still had a richer case system. Kiparsky (1997) proposes that a DP can be associated with a θ-role either via morphological case, or due to canonical ordering relative to the verb. While both options are available to a language with rich case marking, a case-poor language is forced to rely on canonical ordering. This can explain why case-poor modern English requires IO-DO order in double object constructions, while German, which distinguishes dative from accusative, allows both IODO and DO-IO. Fittingly, OE was like German in this respect, and the loss of DO-IO orders in ME coincides at least approximately with the collapse of the dative-accusative distinction (McFadden 2002). It has also been argued that richness of case morphology correlates not just with freedom of word order but with the availability of specific orders. One common observation is that, while Icelandic and the mainland Scandinavian languages have broadly similar syntax in many respects, they differ in the availability of object shift. Object shift is (mostly) obligatory with pronouns in all of Scandinavian, but it is possible with nonpronominal objects only in Icelandic. Holmberg and Platzack (1995) tie this difference directly to the fact that Icelandic distinguishes morphological case in both pronouns and nonpronominal NPs, while the mainland Scandinavian languages have vestiges of case only in their pronouns. Roberts (1997) proposes an extension of this idea to account for a change in English, from the largely OV OE to the rigidly VO modern language. He observes that crucial stages of this change occurred during ME, around when the case system collapsed. He then adopts Kayne (1994)’s antisymmetry approach to surface OV orders as derived by movement, specifically object shift driven by the need for Case (following Chomsky 1995). He then develops an account where rich case-marking forces object shift to be overt, yielding OV order in OE, while the lack thereof allows it to default to being covert, yielding VO from ME onwards. As we can see, there was a move to tie particular syntactic properties to the presence of morphological case through the 1990s and early 2000s. This was part of a larger trend of seeing syntactic operations as morphologically driven (see also the tradition of work connecting verb movement to richness of tense and agreement inflection inspired by Pollock 1989) and it motivated a good deal of interesting and productive comparative work. However, subsequent research has cast doubt on such direct relationships, with comparative Gmc data again playing an important role. While the correlation between rich case marking and word-order freedom is fairly clear, it turns out to be only tendential. Even restricting ourselves to modern Gmc languages, Icelandic has the richest system of case marking, but has considerably more rigid word order than German. Even Dutch, one of the vestigial case languages, allows certain freedoms lacking in Icelandic, e.g., with scrambling (McFadden 2004, 2005). Turning to object shift, a largescale study by Sundquist (2002) shows that diachronic developments of OS

Case in Germanic

in mainland Scandinavian do not track with changes in their case systems as they should if the two were tightly connected in the grammar. As for the connection between OV order and case marking, Icelandic and Dutch present immediate and obvious problems – the former has rich case marking yet is VO, while the latter has lost case and yet is OV. Approaches like Roberts (1997) can deal with such mismatches, but only by resorting to mechanisms that significantly weaken their predictive power. Diachronic work by Hro´arsdo´ttir (2000) and Sundquist (2002) on Scandinavian and by Pintzuk (2002) and McFadden (2005) on English has also shown that the timing of changes does not match up as tightly as it should if word order were directly tied to the richness of case morphology. Finally, as discussed by McFadden (2004, 2005) for case and Alexiadou and Fanselow (2000), Bobaljik (2002) for verbal inflection, there are considerable methodological and theoretical problems, both with the concept of “rich inflection” and with the idea that it could directly drive syntactic computation. All of these authors come to the conclusion that what drives the observed correlations must be related to how language acquisition and change are sensitive to inflectional distinctions, rather than dependencies within the grammar.

References Abraham, W. 1995. Deutsche Syntax im Sprachenvergleich. Tu¨bingen: Gunter Narr Verlag. Alexiadou, A. and G. Fanselow 2000. “On the correlation between morphology and syntax: The case of V-to-I.” In C. J-W. Zwart and W. Abraham (eds.), Studies in Comparative Germanic Syntax: Proceedings of GLOW 2000. Amsterdam and Philadelphia: John Benjamins: 219–242. Allen, C. 1995. Case Marking and Reanalysis. Oxford: Clarendon Press. Andrews, A. 1976. “The VP complement analysis in modern Icelandic.” In Papers from the Sixth Annual Meeting of the North Eastern Linguistic Society, Recherches linguistiques a` Montre´al, Vol. 6: 1–21. Andrews, A. 1982. “The representation of case in modern Icelandic.” In J. Bresnan (ed.), The Mental Representation of Grammatical Relations. Cambridge, MA: MIT Press: 427–503. Baker, M. 2015. Case: Its Principles and Its Parameters. Cambridge University Press. Baker, M. and J. D. Bobaljik 2017. “On inherent and dependent theories of ergative case.” In J. Coon, D. Massam, and L. deMena Travis (eds.), The Oxford Handbook of Ergativity. Oxford University Press: 111–134. Baker, M. and N. Vinokurova 2010. “Two modalities of Case assignment: Case in Sakha,” Natural Language and Linguistic Theory 28: 593–642. Barðdal, J. 2001. Case in Icelandic: A Synchronic, Diachronic and Comparative Approach. Lund: Dept. of Scandinavian Languages, Lund University. Barðdal, J. 2002. “ ‘Oblique subjects’ in Icelandic and German,” Working Papers in Scandinavian Syntax 70: 61–99.

307

308

T HO M A S M CFA DD E N

Barðdal, J. and T. Eytho´rsson 2003. “The change that never happened: The story of oblique subjects,” Journal of Linguistics 39: 439–472. Barðdal, J. 2009. “The development of case in Germanic.” In J. Barðdal and S. Chelliah (eds.), The Role of Semantic, Pragmatic, and Discourse Factors in the Development of Case. Amsterdam: John Benjamins: 123–159. Bhaskararao, P. and K. V. Subbarao (eds.) 2004. Non-Nominative Subjects. Philadelphia: John Benjamins. Blake, B. 2001. Case, second edn. Cambridge University Press. Bobaljik, J. D. 2002. “Realizing Germanic inflection: Why morphology does not drive syntax,” Journal of Comparative Germanic Linguistics 6: 129–167. Burzio, L. 1986. Italian Syntax. Boston: Reidel. Chomsky, N. 1980. “On binding,” Linguistic Inquiry 11:1–46. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. 2000. “Minimalist inquiries: The framework.” In R. Martin, D. Michaels, and J. Uriagereka (eds.), Step by Step: Essays on Minimalism in Honor of Howard Lasnik. Cambridge, MA: MIT Press: 1–52. Chomsky, N. 2001. “Derivation by phase.” In M. Kenstowicz (ed.), Ken Hale: A Life in Language. Cambridge, MA: MIT Press: 1–52. Cook, P. 2006. “The datives that aren’t born equal: Beneficiaries and the Dative Passive.” In D. Hole, A. Meinunger, and W. Abraham (eds.), Datives and Similar Cases. Amsterdam: John Benjamins: 141–184. Czepluch, H. 1996. Kasus im Deutschen und Englischen: ein Beitrag zur Theorie des abstrakten Kasus. Tu¨bingen: Max Niemeyer Verlag. Delsing, L-O. 2002. “The morphology of Old Nordic II: Old Sw. and Old Dan.” In O. Bandle, K. Braunmu¨ller, E. H. Jahr, A. Karker, H-P. Naumann, U. Telemann, L. Elmevik, and G. Widmark (eds.), The Nordic Languages, Vol. 1. Berlin: Mouton de Gruyter: 925–940. Dewey, T. K. and Y. Syed 2009. “Case variation in Gothic absolute constructions.” In J. Barðdal and S. Chelliah (eds.), The Role of Semantic, Pragmatic, and Discourse Factors in the Development of Case. Amsterdam: John Benjamins: 3–21. Enger, H-O. 2013. “Inflectional change, ‘sound laws’ and the autonomy of morphology,” Diachronica 30: 1–26. Eyþo´rsson, Þ., J. B. Johannessen, S. Laake, and T. Åfarli 2013. “Dative case in Norwegian, Icelandic and Faroese: Preservation and non-preservation,” Nordic Journal of Linguistics 35: 219–249. Fillmore, C. 1968. “The case for case.” In E. Bach and R. T. Harms (eds.), Universals in Linguistic Theory. New York: Holt, Rinehart, and Winston, Inc.: 1–25. Freidin, R. and R. Sprouse 1991. “Lexical case phenomena.” In R. Freidin (ed.), Principles and Parameters in Comparative Grammar. Cambridge, MA: MIT Press: 392–416. Haider, H. 1985. “The case of German.” In J. Toman (ed.), Studies in German Grammar. Dordrecht: Foris: 65–101.

Case in Germanic

Haider, H. 1993. Deutsche Syntax – generativ. Tu¨bingen: Gunter Narr Verlag. Haider, H. 2000. “The license to license: Licensing of structural Case plus economy yields Burzio’s Generalization.” In E. Reuland (ed.), Arguments and Case: Explaining Burzio’s Generalization. Amsterdam: John Benjamins: 31–56. Haider, H. 2010. The Syntax of German. Cambridge University Press. Harðarson, J. A. 2017. “The morphology of Germanic.” In J. Klein, B. Joseph, and M. Fritz (eds.), Handbook of Comparative and Historical Indo-European Linguistics, Vol. 2. Berlin: Mouton de Gruyter: 913–95. Hole, D., A. Meinunger, and W. Abraham (eds.) 2006. Datives and Other Cases. Amsterdam: John Benjamins. Holmberg, A. and C. Platzack 1995. The Role of Inflection in Scandinavian Syntax. Oxford University Press. Hro´arsdo´ttir, Þ. 2000. “Interacting movements in the history of Icelandic.” In G. Tsoulas, S. Pintzuk, and A. Warner (eds.), Diachronic Syntax: Models and Mechanisms. Oxford University Press: 296–321. Jacobs, N. 2005. Yiddish: A Linguistic Introduction. Cambridge University Press. Jo´nsson, J. G. 2003. “Not so quirky: On subject case in Icelandic.” In E. Brandner and H. Zinsmeister (eds.), New Perspectives on Case Theory. Stanford, CA: CSLI Publications: 127–163. Jo´nsson, J. G. 2012. “Dative versus accusative and the nature of inherent case.” In B. Fernandez and R. Etxepare (eds.), Variation in Datives. Oxford University Press: 144–160. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Keenan, E. 1976. “Towards a universal definition of ‘subject’.” In C. Li (ed.), Subject and Topic. New York: Academic Press: 303–333. Kemenade, A. van 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kiparsky, P. 1992. Structural Case. Ms., Institute for Advanced Study, Berlin. Kiparsky, P. 1997. “The rise of positional licensing.” In A. van Kemenade and N. Vincent (eds.), Parameters of Morphosyntactic Change. Cambridge University Press: 460–494. Kiparsky, P. 2001. “Structural case in Finnish,” Lingua 111: 315–376. Lamontagne, G. and L. Travis 1987. “The syntax of adjacency.” In Proceedings of WCCFL, 6: 173–186. Lasnik, H. 2008. “On the development of Case theory: Triumphs and challenges.” In R. Freidin, C. Otero, and M. L. Zubizarreta (eds.), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. Cambridge, MA: MIT Press: 17–41. Levin, T. 2017. “Successive-cyclic case assignment: Korean nominative- nominative case-stacking,” Natural Language and Linguistic Theory 35: 447–498. Levin, T. and O. Preminger 2015. “Case in Sakha: Are two modalities really necessary?” Natural Language and Linguistic Theory 33 :231–250.

309

310

T HO M A S M CFA DD E N

Maling, J. 2001. “Dative: The heterogeneity of the mapping among morphological case, grammatical functions, and thematic roles,” Lingua 111: 419–464. Maling, J. 2002. “Icelandic verbs with dative objects,” Working Papers in Scandinavian Syntax 70: 1–60. Maling, J. and S. Sigurjo´nsdo´ttir 2002. “The ‘new impersonal’ construction in Icelandic,” Journal of Comparative Germanic Linguistics 5: 97–142. Marantz, A. 1991. “Case and Licensing.” In ESCOL ’91: Proceedings of the Eighth Eastern States Conference on Linguistics: 234–253. Marantz, A. 1993. “Implications of asymmetries in double object constructions.” In S. Mchombo (ed.), Theoretical Aspects of Bantu Grammar. Stanford: CSLI Publications: 113–150. McFadden, T. 2002. “The rise of the to-dative in Middle English.” In D. Lightfoot (ed.), Syntactic Effects of Morphological Change. Oxford University Press: 107–130. McFadden, T. 2003 [2009]. “On the pronominal origins of the Germanic strong adjective inflection,” Mu¨nchener Studien zur Sprachwissenschaft 63: 53–82. McFadden, T. 2004. The Position of Morphological Case in the Derivation: A Study on the Syntax-Morphology Interface. Doctoral dissertation, University of Pennsylvania. McFadden, T. 2005. “OV-VO in English and the role of case marking in word order,” English Language and Linguistics 9: 1–20. McFadden, Thomas 2006. “German inherent datives and argument structure.” In D. Hole, A. Meinunger, and W. Abraham (eds.), Datives and Similar Cases: Between Argument Structure and Event Structure. Amsterdam: John Benjamins: 49–77. McFadden, T. 2009. “Structural case, locality and cyclicity.” In K. Grohmann (ed.), Explorations of Phase Theory: Features and Arguments. Berlin: Mouton de Gruyter: 107–130. McFadden, T. 2014. Deducing the structural /inherent / quirky case distinction from competing theories of case. Presentation at CGSW 29, University of York. McFadden, T. and S. Sundaresan 2011. Nominative Case is Independent of Finiteness and Agreement. Ms., Tromsø and Tromsø/Stuttgart. http://ling .auf.net/lingbuzz/001350. McIntyre, A. 2006. “The interpretation of German datives and English have.” In D. Hole, A. Meinunger, and W. Abraham (eds.), Datives and Similar Cases. Amsterdam: John Benjamins: 185–211. Meinunger, A. 2000. Syntactic Aspects of Topic and Comment. Amsterdam: John Benjamins. Merkle, L. 1993. Bairische Grammatik. Mu¨nchen: Hugendubel. Neeleman, A. and F. Weerman 1999. Flexible Syntax: A Theory of Case and Arguments. Dordrecht: Kluwer.

Case in Germanic

Nielsen, H. F. 1998. “The linguistic status of the early runic inscriptions of Scandinavia.” In K. Du¨wel (ed.), Runeninschriften als Quellen interdisziplina¨rer Forschung. Berlin: Walter de Gruyter: 539–55. Nikanne, U. 1993. “On assigning semantic cases in Finnish.” In A. Holmberg and U. Nikanne (eds.), Case and Other Functional Categories in Finnish Syntax. New York: Mouton de Gruyter: 75–88. Parrott, J. 2007. Distributed Morphological Mechanisms of Labovian Variation in Morphosyntax. Doctoral dissertation, Georgetown University. Parrott, J. 2009. “Danish vestigial case and the acquisition of vocabulary in Distributed Morphology,” Biolinguistics 3: 270–304. Pintzuk, S. 2002. “Morphological case and word order in Old English,” Language Sciences 24: 267–299. Pollock, J-Y. 1989. “Verb movement, Universal Grammar, and the structure of IP,” Linguistic Inquiry 20: 365–424. Pylkka¨nen, L. 2002. Introducing Arguments. Doctoral dissertation. Cambridge, MA: MIT. Quinn, H. 2005. The Distribution of Pronoun Case Forms in English. Amsterdam: John Benjamins. Ratkus, A. 2015. “Gothic possessives, adjectives, and other modifiers in -ata,” Journal of Germanic Linguistics 27: 238–307. Reuland, E. (ed.) 2000. Arguments and Case: Explaining Burzio’s Generalization. Amsterdam: John Benjamins. Richards, N. 2010. Uttering Trees. Cambridge, MA: MIT Press. Ringe, D. 2006. A Linguistic History of English, Vol. I: From Proto-Indo-European to Proto-Germanic. Oxford University Press. Ringe, D. and A. Taylor 2014. The Development of Old English: A Linguistic History of English, Vol. II. Oxford University Press. Roberts, I. 1997. “Directionality and word order change in the history of English.” In A. van Kemenade and N. Vincent (eds.), Parameters of Morphosyntactic Change. Cambridge University Press: 397–426. Scha¨fer, F. 2008. The Syntax of (Anti-)Causatives: External Arguments in Changeof-State Contexts. Amsterdam: John Benjamins. Schrodt, R. 2004. Althochdeutsche Grammatik II: Syntax. Tu¨bingen: Max Niemeyer Verlag. Schu¨tze, C. 2001. “On the nature of default case,” Syntax 4: 205–238. Sigurðsson, E. F. and J. Wood 2013. “Case alternations in Icelandic ‘get’passives,” Nordic Journal of Linguistics 35: 269–312. Sigurðsson, H. A´. 1989. Verbal Syntax and Case in Icelandic. Doctoral dissertation, Lund University. Sigurðsson, H. A´. 1991. “Icelandic case-marked PRO and the licensing of lexical arguments,” Natural Language and Linguistic Theory 9: 327–363. Sigurðsson, H. A´. 2006. “The nom/acc alternation in Germanic.” In J. Hartmann and L. Molna´rfi (eds.), Comparative Studies in Germanic Syntax. Amsterdam: Benjamins: 13–50.

311

312

T HO M A S M CFA DD E N

Sigurðsson, H. A´. 2009. “The No Case generalization.” In A. Alexiadou et al. (eds.), Advances in Comparative Germanic Syntax. Amsterdam: John Benjamins: 249–280. Stiebels, B. 2002. Typologie des Argumentlinkings: O¨konomie und Expressivita¨t. Berlin: Akademie Verlag. Sundquist, J. D. 2002. Morphosyntactic Change in the History of the Mainland Scandinavian Languages. Doctoral dissertation, Indiana University. Svenonius, P. 2002. “Icelandic case and the structure of events,” Journal of Comparative Germanic Linguistics 5:197–225. Svenonius, P. 2006. “Case alternations in the Icelandic passive and middle.” Ms., University of Tromsø, available at ling.auf.net/lingbuzz/ 000124. Thra´insson, H. 1979. On Complementation in Icelandic. New York: Garland. Thra´insson, H. 2007. The Syntax of Icelandic. Cambridge University Press. Thra´insson, H., H. Petersen, J. ı´ L. Jacobsen, and Z. S. Hansen 2004. Faroese: An Overview and Reference Grammar. To´rshavn: Føroya Fro´ðskaparfelag. Trask, R. L. 1996. Historical Linguistics. London: Arnold. Vergnaud, J-R. 1977. “Letter to Noam Chomsky and Howard Lasnik on ‘Filters and Control’.” In R. Freidin et al. (eds.), 2008. Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. Cambridge, MA: MIT Press. Wegener, H. 1991. “Der Dativ – ein struktureller Kasus?” In G. Fanselow and Sascha Felix (eds.), Strukturen und Merkmale syntaktischer Kategorien. Tu¨bingen: Narr: 70–103. Wood, J. 2015. Icelandic Morphosyntax and Argument Structure. Dordrecht: Springer. Wood, J. 2017. “The accusative subject generalization,” Syntax 20:249–291. Woolford, E. 2006. “Lexical case, inherent case and argument structure,” Linguistic Inquiry 37: 111–130. Wunderlich, D. 1997. “Cause and the structure of verbs,” Linguistic Inquiry 28: 27–68. Yager, L., N. Hellmold, H-A Joo, M. T. Putnam, E. Rossi, C. Stafford, and J. Salmons 2015. “New structural patterns in moribund grammar: Case marking in heritage German,” Frontiers in Psychology Vol. 6, Article 1716. www.frontiersin.org/article/10.3389/fpsyg.2015.01716. Yip, M., J. Maling, and R. Jackendoff 1987. “Case in tiers,” Language 63: 217–250. Zaenen, A., J. Maling, and H. Thra´insson 1985. “Case and grammatical functions: The Icelandic passive,” Natural Language and Linguistic Theory 3: 441–483.

Chapter 14 Complementizer Agreement Marjo van Koppen

14.1 Introduction This chapter deals with a unique phenomenon found in a subset of the Germanic Languages: Complementizer Agreement (henceforth CA).1,2 CA is attested in (dialects of) Frisian and in a subset of the Dutch and the German dialects. It is, as far as I know, not found in (dialects of) English or the North Germanic languages.3 There is a large descriptive literature on CA in the continental West-Germanic dialects (see Barbiers et al. (2005, 2006), Weiß (2005), van Koppen 2017 for an elaborate list of references). In Germanic languages / dialects with CA (henceforth CA-varieties), complementizers introducing embedded clauses show agreement for φfeatures (in particular in person and / or number) with the embedded subject. Consider the examples below:

1

An extended version of this chapter is published as van Koppen (2017). The text of the current contribution partly overlaps with that paper, focusing mainly on the empirical generalizations concerning CA. The theoretical debates about this phenomenon and CA outside of the Germanic area are discussed in van Koppen (2017).

2

Outside Germanic we also find languages / dialects displaying agreeing complementizers. There are for instance languages that agree with wh-elements (in Irish and Chamorro, see McCloskey 2001 and Chung 1998, but potentially also in French, see Rooryck 2000) or with the subject of the matrix clause (in Bantu-languages like Lubukusu, see Diercks 2010, 2013 and Carstens 2016). See also van Koppen (2017) for discussion about CA for ᵠ-features in non-Germanic languages.

3

A potential case of CA in English is discussed by Putnam and van Koppen (2011). They show that MidWestern American English seems to show CA in a particular type of pseudoclefts, which they dub the alls-construction: (i) All-s Greg and Marsha want to do is kiss each other when no one else is around. (ii) All*(-s) you know about Cindy is that she likes to tattle on her siblings. The s-inflection which appears on all in these pseudocleft sentences is sensitive to the φ-features of the embedded subject (in some dialects) just like the agreement on the complementizer in the dialects with CA. The s-inflection on all cannot appear when the subject is second person. This inflection is restricted to these pseudocleft sentences and it is not found on regular complementizers in these dialects.

314

MARJO VAN KOPPEN

(1) a. K peinzen da I

think

/ *dan

dienen student nen buot gekocht eet.

that-3p.sg / that-3p.pl that

student a

boat bought

has

‘I think that that student has bought a boat.’

b. K peinzen dan

/ *da

die

studenten nen buot gekocht een.

I think that-3p.pl / that-3p.sg those students ‘I think that those students have bought a boat.’

a

boat bought

have

(West Flemish, Haegeman 2000)

The complementizer in the West-Flemish example in (1) has to carry a third person plural affix, -n, when the subject of the embedded clause is third person plural as in (1b), but it crucially cannot carry this affix when the subject is third person singular as in (1a). CA is obligatory in some varieties, like the West-Flemish variety Haegeman (1992, 2000) reports on, but there are also varieties in which CA is optional. CA does not have an effect on the meaning of a sentence. It is reported that there is a register effect, though: Some speakers experience CA as belonging to another (more archaic) register (cf. Hoekstra & Smits 1997: 12); there is no thorough sociolinguistic research about this aspect of CA, however. CA is also a phenomenon that is typically found in nonstandard languages, i.e., in dialects or regiolects, i.e., it is not a property of the standard varieties of Dutch or German, although it is a property of Frisian, which is recognized as an official language in the Netherlands. This chapter discusses the empirical observations that have been made about CA; the reader is referred to van Koppen (2017) for an in-depth discussion of the theoretical accounts of CA. The chapter is organized as follows: Section 14.2 provides a detailed discussion about the various types of CA-paradigms. Section 14.3 explores the relation between verbal agreement and CA, by diving into so-called double agreement. Section 14.4 discusses the interaction of CA with clitics and pro-drop. Section 14.5 goes into the question whether CA is sensitive to linear adjacency and, finally, Section 14.6 goes into the syntactic distribution of CA.

14.2

The Complementizer Agreement-Paradigm

The affixes used on the complementizer in CA-varieties are the same as the ones used as the verbal agreement affixes in these dialects. The CAparadigm is quite often defective, however, in the sense that less overt agreement reflexes appear on the complementizer than on the finite verb in the same variety. A good example is Frisian. As the Frisian dialect of Heerenveen (2) illustrates, there is only an affix for the second person singular on the complementizer as ‘if’ and possibly for the third person singular (depending on the analysis of the t-ending), whereas the finite verb gaen ‘go’ also shows overt agreement reflexes for, at least (again

Complementizer Agreement

depending on the analysis of these verb forms) first person plural and the third person plural.4 (2)

a.

...

b. . . . c.

...

d. . . . e.

...

f.

...

as ik if I ast if-2p.sg at er if he as we if we as jimme if you as se if they

a’.

...

b’.

...

c’.

...

d’.

...

e’

...

f ’.

...

gaen ik go I giest go-2p.sg giet er goes he geane we go we gean jimme go you gaene sij go they (Heerenveen Frisian, Barbiers 2006)

There are dialects that have a full paradigm. Haegeman (1992), for instance, argues that West Flemish has a full, i.e., nondefective, CA-paradigm: (3)

a.

...

b.

...

c.

...

d.

...

e.

...

f.

...

da-n-k that-1p.sg-1p.sg da-j that-2p.sg da-se that-3p.sg da-me that-1p.pl da-j that-2p.pl da-n-ze that-3p.pl-3p.pl

(ik) (I) (gie) (you) (zie) (she) (wunder) (we) (gunder) (you) (zunder) (they)

morgen tomorrow morgen tomorrow morgen tomorrow morgen tomorrow morgen tomorrow morgen tomorrow

goa-n. go-1p.sg goa-t. go-2p.sg goa-t. go-3p.sg goa-n go-1p.pl goa-t. go-2p.pl goa-n go-3p.pl

(West-Flemish, Haegeman 1992: 49)

This paradigm is complete in the sense that there is an element agreeing with the subject in each person/number and gender combination. The agreement morphology appearing on the complementizer in this dialect consists of two parts. The first part can be classified as inflectional morphology expressing at least agreement for number (i.e., the n-ending in the first person singular and third person plural). The second part can be analyzed as a clitic pronoun (i.e., the elements k, j, se, me, j, ze in the

4

Note that Frisian also displays subject drop in the second person singular. The second person subject pronoun do ‘you’ can be present, but this leads to a focused reading on the subject.

315

316

MARJO VAN KOPPEN

examples above), which conveys the person, number, and gender information of the subject (cf. Haegeman 1992: 68–69). We will come back to the relation between clitics and CA in Sections 14.3 and 14.4. De Vogelaer (2006: 99–101) shows that the West- and East-Flemish dialects as well as some other Dutch dialects, like the Eastern Dutch dialect of Groenlo (data from Barbiers et al. 2005 / 2006), have a CA-paradigm that is completely parallel to the verbal agreement paradigm and can hence be argued to be nondefective as well. This becomes clear if we compare the inflection on the complementizer with the inflection on the verb in the verb-subject order: (4)

a.

...

b.

...

c.

...

d.

...

e.

...

f.

...

as-ik if I a(s)-je if you ast-e if he azze-wie if we as-jullie if you at-ze if they

a’.

...

b’.

...

c’.

...

d’.

...

e’

...

f ’.

...

was-ik wash I was-je wash you wast-e washes he wasse-wie wash we was(t)-jullie wash you leeft-ze live they

(Groenlo Dutch, De Vogelaer 2006: 100)

Example (4) illustrates that the complementizer shows the same inflectional pattern as the verb wassen ‘to wash’. This pattern does not hold for all CA-varieties, however. When we consider the CA-varieties, we can roughly divide the Germanic dialects and languages with CA into two groups. The first one includes Frisian, the Saxonian and Limburgian dialects, and the German dialects with CA. These dialects all at least show agreement with second person singular subjects. The agreement affix in these dialects is typically -s or -st and the subject pronoun is (a variant of) du / do ‘you’. Consider the examples in (2) from Frisian and in (5) from the Bavarian dialect of Gmunden (Gruber 2008). (5)

I I

woas know

net, not

ob-st if-2.p.sg

du yousg

des that

mocha do

kaun-st. can-2p.sg

‘I don’t know if you can do that.’ (Gmunden Bavarian, Gruber 2008: 6)

There are also dialects within this group with a more elaborate CA-paradigm. The variety of Bavarian discussed by Bayer (1984), for instance, just like the

Complementizer Agreement

Gmunden dialect discussed in (5), has CA for the second person singular and plural.5 (6)

a. Dui bis dass-st ti kummst . . . you are that-2p.sg come-2p.sg ‘You are the one that comes . . . ’ b. Ihri/esi bis dass-ts ti kummts . . . you are that-2p.pl come-2p.pl ‘You are the ones that come . . . ’ (Bavarian, Bayer 1984: 234)

The same holds for several Eastern Dutch and Limburgian Dutch dialects (see Barbiers et al. 2005: 221). Some dialects within this group, like Lower Bavarian (see Bayer 1984), additionally have CA in the first person plural – an example is provided in (15) in the next section – and some German dialects have CA in the third person plural in addition to the second person singular and plural and the first person plural. An example of such a dialect is Egerlandish (also a dialect of Bavarian): (7)

a. b. c. d.

wal-st because-2p.sg das-n that-1p.pl wenn-ts when-2p.pl daa-n-s that-3p.pl-they

mer we diaz you

(Egerlandish, Schiepek 1899/1908, Weise 1907, as cited in Weiß 2005: 151)

The second group of CA-dialects consists of the western Dutch dialects (including Flemish), but it also includes some Dutch dialects spoken in the east of the country, like Groenlo (see example (4) above). These dialects typically show plural agreement on the complementizer: (8)

a.

. . . as when ‘ . . . when b. . . . as-e when-pl ‘ . . . when

ik/jij/hij I/you/he I/you/he we/jollie/ze we/you/they we/you/they

hoor(t) . . . hear(s) hear(s) . . . ’ hore . . . hear hear . . . ’

(Katwijk Dutch, Barbiers et al. 2006)

5

The examples from Bayer (1984) display movement of the subject to show that the endings (i.e., -st/-ts) on the complementizer are inflectional affixes rather than subject clitics. Subject clitics cannot be stranded when the subject moves.

317

318

MARJO VAN KOPPEN

Katwijk Dutch shows plural CA with all plural subjects, but there are also dialects in this region that only show CA with one or two of the plural persons. Within this group there are also dialects that additionally show agreement in one or more of the singular persons: See, for instance, the West-Flemish example in (3) for agreement with the first person singular or the Groenlo example in (4) for agreement with the third person singular. The West-Frisian dialects furthermore show agreement in second person singular (see Hoekstra and Smits 1997): (9)

a.

. . . datte je zwemme. you swim-2p.sg that-2p.sg ‘ . . . that you are swimming.’

(West-Frisian, Hoekstra & Smits 1997: 18) b.

. . . datte we ’ier burgers en visscherlui that-pl we here civilians and fishermen ‘ . . . that we have civilians and fishermen here.’

ewwe have-pl

(West-Frisian, West-Friesland’s Oud en Nieuw: Jaarboek 1929) It is clear from this section that CA occurs in many continental WestGermanic varieties and that there is a close connection between the CAparadigm and the verbal paradigm. An intriguing question is why some varieties within this continuum do have CA whereas it is absent in others, and why some person / number combinations on the subject trigger CA in the CA-dialects, but not others. Before we can go into these questions, however, we first have to explore the relation between CA-paradigms and verbal paradigms in a bit more detail and discuss the so-called Double Agreement.

14.3 A Closer Inspection of the Relation between the CA-Paradigm and the Verbal Agreement Paradigm: Double Agreement 14.3.1 Double Agreement CA is closely related to another phenomenon, namely so-called Double Agreement. The term Double Agreement (henceforth DA) refers to the pattern of agreement in which the affix on the finite verb differs depending on its position in the sentence (see also among others Van Haeringen 1958, Bayer 1984, Zwart 1993, Hoekstra and Smits 1997, Zwart 1997, 2001, van Koppen 2005, Weiß 2005). The agreement is double in the sense that there is more than one affix for the finite verb representing the same set of phi-features. The affix appearing on the complementizer is the same as one of these DA-affixes. There are (at least) two patterns of DA. In the first one, the ending on the finite verb depends on the relative position of subject and verb: If the finite verb follows the subject (in subject-initial main clauses or embedded

Complementizer Agreement

clauses) the verb has a different inflection than when the subject follows the verb (in inverted main clauses or V1-clauses). The frequently discussed DA pattern of the Dutch dialect of Hellendoorn is exemplified in (10)–(12). This dialect only shows a DA pattern with first person plural subjects. (10)

a.

Wiej bin-t den besten! the best we are-agr1 ‘We are the best!’ b. * Wiej binn-e den besten! we are-agr2 the best (Hellendoorn Dutch, van Koppen 2005: 125–126)

(11)

a. * Bin-t wiej den besten? are-agr1 we the best b. Binn-e wiej den besten? we the best are-agr2 ‘Are we the best?’ (Hellendoorn Dutch, van Koppen 2005: 125–126)

When the verb follows a first person plural subject in Hellendoorn Dutch it ends in a t-affix (see example (10), when the verb precedes the subject it ends in a e-affix (see example (11)). The agreement on the complementizer always patterns with that on the finite verb in VS-clauses. This is known in the literature as the inversion generalization (see Hoekstra and Smits 1997, Barbiers et al. 2005: 19–34).6 (12)

. . . darr-e / *dat wiej den besten bin-t / *binn-e! that-agr2 / that we the best are-agr1 / are-agr2 ‘ . . . that we are the best!’ (Hellendoorn Dutch, van Koppen 2005: 125–126)

A similar case of DA in which the ending on the verb is dependent on the relative order of verb and subject is found in Brabantic, see (13).7

6

It is not the case that all verbs in DA-dialects necessarily pattern the same. There is for instance sometimes a difference between auxiliaries and main verbs or between monosyllabic and polysyllabic verbs. This has led to a debate in the literature on exactly which verbs are relevant for the inversion generalization. Goeman (1980, 2000) argues that the complementizer copies the agreement affix of monosyllabic verbs. De Vogelaer (2006) compares CA to the inflection on inverted monosyllabic and polysyllabic verbs to see which paradigm matches the CA paradigm best. He reaches the conclusion that the CA paradigm resembles the verbal paradigm of the verbs that have the same morpho-phonological shape as the complementizer: polysyllabic complementizers match the inflection on polysyllabic verbs and monosyllabic complementizers that on monosyllabic verbs. Hoekstra and Smits (1997, 1999) argue that auxiliaries are the relevant group of verbs. Van Craenenbroeck and van Koppen (2002) claim that CA resembles the inflection on the verb to be in inversion in the present tense.

7

As has been pointed out by Bayer (1984), Weiß (1998, 2005), the DA ending /-ma/ replaces the regular verbal agreement ending for the first person plural, /-a(n)/, only with a couple of polysyllabic verbs (e.g., laffa ‘to run’, gengan ‘to go’, soucha(n) ‘to seek’).

319

320

MARJO VAN KOPPEN

(13)

a.

b.

c.

. . . da-de gullie kom-t / *kom-de. that-2p youpl come-2p.pl / come-2p ‘ . . . that you will come.’ Gullie kom-t / *kom-de. come-2p.pl / come-2p youpl ‘You will come.’ Wanneer kom-de / *kom-t gullie. when come-2p / come-2p.pl youpl ‘When do you come?’ (Brabantic, Zwart 1997:140)

The Brabantic second person singular and plural de-affix appears on the finite verb when it precedes the subject and on the complementizer, but not on the finite verb in subject-initial main clauses and embedded clauses. Zwart (1997) argues that this de-ending is an affix, but it has also been analyzed as a clitic pronoun which necessarily appears in the Wackernagel position (see, for instance, Stroop 1987). Clitic pronouns and inflection on the verb in the verb-subject order or the inflection on the complementizer are closely related and it is very often hard to tease the two apart, as we will discuss in more detail in Section 14.3.2. It is not always easy to distinguish between the two (see for instance Gruber [2008] for an extensive discussion). For this reason, it is not always clear if we are dealing with a pattern of different affixes, for instance -t and -de in Brabantic (i.e., DA), or with the same affix, for instance -t in Brabantic which disappears because of assimilation with the clitic pronoun -de. The same issue plays a role in Frisian. De Haan (1997) shows that Frisian also has a DA pattern similar to Hellendoorn Dutch and Brabantic in the second person singular. He provides the examples in (14). (14)

a.

. . . dat-st do moarn komme soe-ste. that-2p.sg you tomorrow come would-2p.sg ‘ . . . that you would come tomorrow.’ b. Do soe-ste moarn komme. tomorrow come you would-2p.sg ’You would come tomorrow.’ c. * Moarn soe-ste (*do) komme. Tomorrow would-2p.sg you come d. * . . . dat-ste (do) moarn komme soe-ste. that-2p.sg you tomorrow come would-2p.sg (Frisian, De Haan 1997: 65)

Frisian appears to have two agreement affixes -st and -ste. De Haan (1997) shows that -ste displays DA-behavior: It can only co-occur with the subject pronoun do if the finite verb follows the subject. If the ste-affix appears on a

Complementizer Agreement

finite verb preceding the subject – see (15c), or the complementizer, see (15d) – do cannot appear. De Haan concludes from this that Frisian has two different ste-endings. One that cannot co-occur with an overt subject pronoun and one that can. He argues that the first one is comparable to the Brabantic de-ending and the Hellendoorn Dutch e-ending because it is also restricted to complementizers and verbs in subject inversion contexts. Here the issue of pro-drop also plays a role. The ste-ending appearing in the verb / complementizer subject order is only compatible with a pro-drop sentence. We will explore the relation between CA/DA, clitics and pro-drop more in depth in Section 14.3.2. Double agreement does not necessarily follow the pattern displayed in (10)–(12) and (13), however. The second DA-pattern is found in Lower Bavarian. This dialect shows a double agreement pattern in the first person plural, but with a slightly different distribution than the patterns discussed above, see (15). (15)

a.

b.

c.

. . . daβ-ma mia noch Minga fahr-n / *fahr-ma. Munich go-1p.pl / go-1p.pl that-1p.pl we to ‘ . . . that we are going to Munich.’ Mia fahr-ma / *fahr-n noch Minga. / go-1p.pl to Munich we go-1p.pl ‘We are going to Munich.’ Fahr-ma/ *fahr-n mia noch Minga? to Munich go-1p.pl / go-1p.pl we ‘Are we going to Munich?’ (Lower Bavarian, Zwart 1997: 140)

This example shows that the affix on the finite verb in Lower Bavarian is not so much dependent on the position of the finite verb relative to the subject, but to the position of the finite verb within the clause. The verb carries a ma-affix if it is in the left periphery of the clause, whereas it ends in a n-affix if it is in the right periphery. In short, the DA-dialects show that CA is similar to the inflection on the verb in the left periphery for dialects like Lower Bavarian and to the inflection on the verb in the verb-subject order in others, like Hellendoorn Dutch, Brabantic, and Frisian.

14.3.2 The Defectivity of the CA-Paradigm As I have already mentioned, a lot of dialects with CA make fewer distinctions in the CA-paradigm than in the verbal paradigm. The question arises, of course, why this would be the case. Weiß (2012) and Fuß (2004, 2005) argue that this defectivity of the CA-paradigm in Bavarian can be explained by an accidental reanalysis of a clitic pronoun to an agreement ending. They assume this reanalysis takes place for a limited number of subject

321

322

MARJO VAN KOPPEN

pronouns. As disscussed above, there are several cases of CA that seem to have a pronominal origin, in particular the second person singular CA of for instance Frisian in (14), Brabantic in (13), and Bavarian in (5), but also the first person plural CA in Bavarian in (15). This analysis does not provide an explanation however for the CA-endings that are unlikely to have this origin, like the dialect of Katwijk in (8), Groenlo in (4), and parts of the West-Flemish paradigms in (3). The endings in these latter dialects are, at least not straightforwardly, analyzable as clitic pronouns. Hoekstra and Smits (1997, 1999) have proposed a different explanation for the defectivity of CA-paradigms. Their Identity Generalization states that the CA-paradigm only uses those verbal affixes that are identical in the present and past tense. Put differently, CA affixes are those affixes of the verbal paradigm that do not express tense information and are hence pure phi-affixes. The discussion on DA makes clear that we should look at the verbal paradigm in inversion contexts, at least for the dialects with the Hellendoorn Dutch pattern. The example below illustrates the identity generalization: (16)

a.

b.

c.

Present tense:

Wol-st do komme? want-2p.sg you come ‘Do you want to come?’ Preterit: Woe-st do komme? wanted-2p.sg you come ‘Did you want to come?’ Complementizer: . . . dat-st do komme sil-st will-2p.sg . . . that-2p.sg you come ‘ . . . that you will come’ (Frisian, van Craenenbroeck and van Koppen 2002)

The st-affix appears both in the present tense and the past tense, compare (16a–16b), and hence does not convey tense information. It can be used as CA, see (16c). Frisian does not have CA in the plural. This is also predicted by the identity generalization. Consider the examples in (17). (17)

a.

b.

c.

Present tense:

moatt-e must-1p.pl ‘must we’ Preterit: moast-en must-1p.pl ‘must we’ Complementizer: dat-(*e/*en) that-1p.pl

wy we wy we wy we

moatt-e must-1p.pl (Frisian, Hoekstra p.c.)

The first person plural does not have an ending in the verbal paradigm that is the same in the present tense and the past tense, compare (17a–17b), and

Complementizer Agreement

hence Frisian does not have CA in the first person plural, see (17c). As such the Identity Generalization provides a potential tool to explain why some dialects have CA and others lack it. It also gives us a handle on the defectivity of the CA-paradigm. Hoekstra and Smits (1997, 1999) base their generalization on the agreement patterns in seven Dutch dialect areas. It is unclear, however, if this generalization can be extended to the German CA dialects, especially since most of these dialects lack a past tense form. The generalization has also not been checked in more detail for the Dutch dialects (for instance within the SAND-project, Barbiers et al. 2005). Some deviations from this generalization have been reported for the Dutch dialects, however. Hoekstra & Smits (1997) themselves for instance show that there are dialects in the Dutch province of Limburg that have a verbal agreement ending that is identical in the present tense and the preterit, but that do not have CA. So, at the very least, identity seems to be a necessary but not a sufficient condition to get CA. Van Craenenbroeck and van Koppen (2002) provide a counterexample against this generalization from NieuwkerkenwaasDutch, see (i) below: (i)

a.

Present tense:

Will-e want-pl

zunder they

komm? come

‘Do they want to come?’ b.

c.

Preterit:

Wou-n zunderkomm? wanted-plthey come ‘Did they want to come?’ Complementizer: da-n zunder zulle that-pl they will ‘that they will come’.

komm. come

(Nieuwkerken-Waas Dutch, van Craenenbroeck and van Koppen 2002: 5)

In this example, the verbal agreement on the auxiliary willen ‘want’ is not the same in the present tense and the preterit, but CA is still an option. This shows that a more fine-grained investigation of this generalization is necessary.

14.3.3 Summary This section has discussed the questions surrounding the CA-paradigm. We have seen that the CA-affixes mirror the verbal affixes on the verb in the left periphery (in the case of Lower Bavarian) or on the verb in the verbsubject order (in the cases of the other dialects). This is usually referred to as the inversion-generalization. Furthermore, we have explored two potential explanations for the defectivity of CA-paradigms. One potential

323

324

MARJO VAN KOPPEN

reason is that CA-affixes are often accidentally reanalyzed clitic pronouns. Another explanation can be found in the Identity Generalization.

14.4

CA, Clitics and Pro-Drop

We have seen that there is a close connection between CA, clitics, and prodrop. We will further explore this connection in this section. As discussed in Section 14.3, we can distinguish two types of CAaffixes. There are affixes that are pronominal in origin and affixes that have a verbal origin (see among others Weiß 2012, Goeman 1997). Dialects can have both types of CA affixes within one paradigm. There appears to be an especially close relation between clitic pronouns and CA in the second person. For instance, Bayer (1984), Fuß (2004), Gruber (2008), and Weiß (2012) argue that CA in the second person singular is the result of reanalysis of second person subject clitics in the verb-subject order as inflectional markers. Hoekstra (1997) also addresses the question as to why pro-drop is found exclusively with first and second person in the WestGermanic languages. He argues that because first and second person subjects are necessarily pronominal in contrast to third person subjects, there is a strong tendency for cliticization to the verb in the verb-subject order in these persons. This means that reanalysis of a first and second person pronoun into an inflectional affix is much more likely than of a third person pronoun (cf. also Weiß 2012).8 These verbal inflectional markers then also appear on the complementizer by extension. The origin of verbal CA-affixes is much less clear (see Goeman 1997 for a detailed discussion of the attestations of verbal CA in older stages of Dutch). One difference between pronominal and verbal CA affixes is that the former tend to license pro-drop, but the latter do not (see among many others Bayer 1984, Hoekstra 1997, Fuß 2005, Weiß 2012).9 Consider the examples in (18a) and (18b) respectively.

8

The question arises how we can determine whether the element attached to the complementizer is a clitic or an inflectional ending. There are various tests proposed in the literature to distinguish between these two options. Zwicky (1977) and Zwicky and Pullum (1983) provide several tests. De Haan (1997) gives specific tests to distinguish between clitics and inflectional endings in Frisian. Fuß (2005) discusses several tests applied to Bavarian. Gruber (2008) also discusses the issue of the status of the CA-endings in the Austrian Bavarian dialect of Gmunden. She provides several tests and reaches the conclusion that in her dialect the CA-affix has both properties of an inflectional ending and of a clitic pronoun. She provides an analysis for the dual status of this type of ending. It would lead us too far afield to discuss these tests in detail here. The reader is referred to the literature given in this footnote and the references therein for the complete debate.

9

Hoekstra and Mará cz (1989) argue that pro-drop in the West-Germanic dialects is always licensed by CA and vice versa. However, as is shown by Zwart (1993) and as can be derived from the examples in the main text, this generalization is not correct.

Complementizer Agreement

(18)

a.

b.

wenn-sd will-sd. want-2sg if-2p.sg ‘ . . . if you want.’ ...

(Bavarian, Weise 1907, as cited in Weiß 2005: 154) . . . waal-n *(mer) graad besamn senn. at_the_moment together are-1p.pl because-1p.pl We ‘ . . . because we are together at the moment.’ (Bavarian, Weise, 1907, as cited in Weiß 2005: 154)

The d-affix in (18a) derives from a subject clitic, which has become part of the verbal inflection. In this example, pro-drop can take place. The first person plural agreement in (18b) does not contain such a pronominal part and pro-drop is not an option.10 The same opposition is found in the Dutch dialects and Frisian. Hoekstra (1997) argues that the Frisian st-ending has a pronominal origin and licenses pro-drop. The West Flemish n-ending which is not pronominal does not license pro-drop. (19)

a.

b.

. . . dat-st (do) juˆn kom-st. that-2p.sg (you) tonight come-2p.sg ‘ . . . that you will come tonight.’ (Frisian, Weiß 2005: 156) . . . da-n *(ze) goan werk een. that-3p.pl they go3p.pl work have ‘ . . . that they have gone to work.’ (West Flemish, Weiß 2005: 156)

To summarize, there is a close connection between clitics, CA, and pro-drop. Clitics in the first and second person have a tendency to attach to the verb and complementizer. This then leads to reanalysis of the clitic into a complementizer agreement affix and a (double) verbal agreement affix. Since these affixes still have a pronominal component (or because they are in between a proper affix and a clitic, as Gruber [2008] argues), they are able to license pro-drop.

14.5 CA and Linear Adjacency As discussedin 14.4, the regular CA-pattern in the Germanic languages is one in which the subject of the embedded clause agrees with the complementizer. There are a couple of CA-patterns that deviate from 10

There appears to be an exception to this rule discussed by Zwart (1993: 167), namely Zürich German that allows prodrop without a clear (pronominal) agreement affix: (i)

. . . öb whether

(d/du)

nach

Züri

you

to

Zürich come2p.sg.

chunnsch.

‘ . . . whether you come to Zürich.’ However, Weiß (2005) argues that there might be a pronominal zero-affix in this example that licenses pro-drop.

325

326

MARJO VAN KOPPEN

this general pattern and where the CA seems to target a directly adjacent noun phrase: CA with coordinated subjects and CA with external possessors. I discuss these two patterns in Section 14.5.1. Another issue, discussed in Section 14.5.2, related to adjacency concerns the question of whether an intervener between the complementizer and subject has an effect on CA.

14.5.1

Agreement with Coordinated Subjects and External Possessors The complementizer normally agrees with the subject. However, there are two cases in which it agrees with only part of the subject, rather than with the complete subject. In both these cases CA is with a noun phrase which is linearly closest to the complementizer, which might lead to the conclusion that CA is the result of linear adjacency between complementizer and subject. The first example concerns cases in which the subject is a coordination and the second involves an external possessor. Let us first look at the patterns with coordinated subject. Van Koppen (2005) investigates the behavior of CA with coordinated subjects in varieties of Germanic. She shows that there are two patterns: Either the complementizer agrees with the coordinated subject as a whole, see (20), or it agrees with the first conjunct only, see (21). (20)

Kpeinzen da-n [Vale`re en Pol] morgen goa-n. I.think that-3p.pl [Vale`re and Pol]3p.pl tomorrow go-P L . ‘I think that Vale`re and Pol will go tomorrow.’ (Lapscheure Dutch, from Haegeman 1992: 49)

(21)

. . . de-s doow en ich oˆs treff-e. that-2P . S G [youS G and I]1 P . P L each.other1 P . P L know-1 P . P L ‘ . . . that you and I know each other.’ (Tegelen Dutch, van Koppen 2005: 174)

In example (20) the complementizer and the finite verb are both inflected, agreeing with the third person plural feature of the subject Vale`re en Pol. The agreement on the complementizer in (21) is different from that on the finite verb. The complementizer agrees with second person singular features of the first conjunct of the coordinated subject, whereas the verb agrees with the first person plural features of the complete coordinated subject. Due to reasons of space the reader is referred to Van Koppen (2005, 2012) for an elaborate discussion of these patterns. CA can also target a part of the subject rather than the complete subject in the case of external possession, see Haegeman and van Koppen (2012).

Complementizer Agreement

(22)

. . . omda-n die venten toen juste underen computer because-3 P . P L those guys then just their computer kapot was broken was ‘ . . . because those guys’ computer broke just then.’ (Lapscheure Dutch, Haegeman and van Koppen 2012: 444)

The subject, die venten underen computer ‘those guys’ computer’, is discontinuous: The possessor die venten ‘those guys’ and the possessee underen computer ‘their computer’ are interrupted by the adverb toen juste ‘just then’.11 These patterns might give the impression that CA simply targets a right adjacent noun phrase. This is not the case, however, as exemplified in (23). (23) . . . omdat/*omda-n [Andre´ en Vale`re underen because-3p.sg /because-3p.pl Andre´ and Vale`re their computer] kapot was. computer broken was ‘. . . because Andre´ and Vale`re’s computer was broken.’ (Lapscheure Dutch, Haegeman and van Koppen 2012)

In this example, the complementizer cannot agree with the immediately adjacent third person plural possessor noun phrase Andre´ en Vale`re, but it has to agree with the complete subject Andre´ en Vale`re underen computer which is third person singular. That adjacency is not the trigger for CA is also confirmed by example (24) in which a scrambled object zukken boeken intervenes between complementizer and subject. This third person plural direct object cannot agree with the complementizer, as illustrated in (24c), although it is adjacent to it. (24) a.

kpeinzen dat

zelfs

Vale`re zukken boeken niet

leest.

I.think

even

Vale`re

reads

that

b. ?? kpeinzen dat I.think

c. *

that

kpeinzen da-n I.think

such

zukken boeken zelfs such

books

even

zukken boeken zelfs

that.pl such

books

even

books

not

Vale`re niet Vale`re

leest.

not

reads

Vale`re niet

leest.

Vale`re

reads

not

‘I think that even Vale`re would not read such books.’

(Lapscheure Dutch, Haegeman and van Koppen 2012)

14.5.2 Interveners between Complementizer and Subject Another issue related to the question of whether CA is sensitive to adjacency concerns intervening adverbs between the complementizer and the subject. Although agreement relations are generally not sensitive to

11

The reader is referred to Haegeman and van Koppen (2012) for an in-depth discussion of this construction.

327

328

MARJO VAN KOPPEN

intervening adverbs, CA seems to be sensitive to this in some CA-varieties, but not in all of them. Ackema and Neeleman (2004) show that adjacency plays a role in the West-Flemish dialect of De Panne, see (25). (25) a. . . . da / da-n zunder op den warmste dag van t jaar on the hottest day of the year that / that-3 P . P L they tegen under wil gewerkt en. against their will worked have b. . . . da / *da-n op den warmste dag van t jaar zunder that / that-3p.pl on the hottest day of the year they tegen under wil gewerkt en. against their will worked have ‘ . . . that they have worked against their will at the hottest day of the year.’ (De Panne Dutch, Ackema, and Neeleman 2004: 240)

CA is optional in this dialect. However, when an adverb intervenes between the complementizer and the subject pronoun, it is obligatorily absent. Fuß (2008) shows that the same pattern can be observed in Bavarian: (26) a. Obwoi-st du ins Kino ganga bist . . . although-2 P . S G you to-the movies gone are2 P . S G ‘Although you went to the movies . . . ’ b. Obwoi(*-st) woartscheints du ins Kino ganga bist . . . you to-the movies gone are2 P . S G although(-2 P . S G ) probably ‘ Although you probably went to the movies . . . ’ (Bavarian, Grewendorf p.c., as cited in Fuß 2008: 85)

However, there are also ample examples in the literature that show that intervention does not have an effect on CA. Gruber (2008), for instance, shows that adverb or argument intervention in her Bavarian variety does not seem to play a role. (27) a. Warum-st sein Freind uns du net vorgsto¨ht ho-st, Why-2 P . S G his friend us you not introduced have-2 P . S G vasteh i a net. understand I too not ‘Why you didn’t introduce his friend to us, I don’t understand either.’ b. Wos hot da Hannes gsogt, wo-st morgn What has the Hannes said, that-2 P . S G tomorrow du mitbringa soid-st? you with-bring should-2 P . S G ‘What did Hannes say that you should bring along tomorrow?’ (Bavarian, Gmunden dialect, Gruber 2008: 54)

Complementizer Agreement

The example at (27a) shows the intervention of a scrambled direct and indirect object between the CA and the subject. Example (27b) illustrates intervention of an adverb.12 Haegeman and van Koppen (2012) discuss similar examples from Lapscheure Dutch, a West-Flemish dialect; just like De Panne Dutch illustrated in (25). Lapscheure Dutch has CA when an adverb intervenes, see example (28):13 (28) a. . . . da-n

/

?

*dat toen juste men twee broers

that- P L / that

then just

my

two

kwamen.

brothers came

‘ . . . that my two brothers came just then.’

b. . . . da-n

/ ?*dat juste ip dienen moment men twee broers

that- P L / that just

at that

moment my

two

kwamen.

brothers came

‘ . . . that my two brothers came just at that moment.’

(Lapscheure Dutch, Haegeman and van Koppen 2012: 447)

Weiß (2005) provides an example from Zu¨rich German where an indirect object intervenes between the subject and the CA (see Haeberli 1999 for similar data from this dialect): (29)

. . . wie-t mer du(u) how-2 P . S G to-me you ‘ . . . what you have told me.’

gsa¨it said

ha¨sch have2 P . S G

(Zu¨rich German, Weber 1964, as cited in Weiß 2005: 57)

To summarize, linear adjacency seems to plays a role in the distribution of CA in some dialects, but clearly not in all dialects. More research is necessary to further investigate this aspect of CA.14

14.6 Syntactic Distribution of CA Up till now, we have only looked at the morphological aspects of CA, e.g., the paradigm, the origin of the affixes, the relation of CA with verbal agreement, the licensing of pro-drop. This section briefly addresses the matter of the syntactic distribution of CA. CA is normally attached to the complementizer of finite clauses as the examples given have amply illustrated. Syntactically, CA is restricted to finite clauses, i.e., it is never found in infinitival clauses on the infinitival complementizer. Furthermore, CA necessarily expresses agreement with

12

Helmut Weiß (p.c.) informs me that his own Bavarian dialect, which is a variant of Middle Bavarian and spoken in Bavaria, has the same properties as the variant reported in Gruber (2008).

13

In previous work Haegeman did argue that the subject and the complementizer should be adjacent (see Haegeman 1992). Given the right pragmatic context, intervention is possible in this dialect, however.

14

Van Koppen (2017) shows that intervention of a subject modifier has some other interesting effects in some but not all CA-varieties.

329

330

MARJO VAN KOPPEN

the embedded subject and not with any other argument. The category of the subject generally does not affect the possibility to have CA. Full NPs and pronouns can both trigger it in the dialects that have CA in the third person (see, for instance, De Vogelaer 2006: 32–33). Barbiers et al. (2005: 33) have investigated this in depth for the Dutch dialects and show that the Flemish dialects have CA both with pronominal and nonpronominal subjects, but that there are dialects in the Hollandic area (i.e., the northwestern part of the Netherlands) in which CA exclusively occurs with pronominal subjects.15,16 It can also be found on other elements in the left periphery when the complementizer is absent. Consider for instance the data in (30) where the inflection surfaces on the wh-word warum ‘why’ or the data in (31) where the inflection is attached to a wh-phrase:17 (30)

. . . warum-sd des ned mochsd. this not make-2p.sg why-2p.sg ‘ . . . why you do not make this.’ (Bavarian, Weiß 2005: 148)

(31)

Du sollst song [[ an wa¨ichan Schuah]i -st] du ti wui-st]. you should say the which-one shoe 2p.sg you want-2.SG ‘You should say which one of the shoes you want.’ (Bavarian, Bayer 1984: 235)

Similar examples with CA on elements other than the complementizer can be found in many Dutch and German dialects (see for instance Zwart 1993:171, Weiß 2005). A possible hypothesis about these cases is that the inflection appears to be on the wh-word or wh-phrase. Another hypothesis, however, is that it is hosted by an empty complementizer immediately following this left peripheral item (see Zwart 1993 among many others). A comparable case of this pattern in which the inflection is not on the complementizer but on another left peripheral element is discussed by Cremers and van Koppen (2008). They show that CA in the Dutch dialect of Tegelen can appear on coordinating conjunctions like of ‘or’ or en ‘and’. An example is given in (32). 15

Goeman (1980, 2000) argues that CA is sensitive to rhythm in the sense that a weak pronoun or an unstressed determiner following the complementizer triggers CA. Hoekstra and Smits (1997) confirm that there is a tendency on the basis of a West-Frisian corpus (a dialect spoken in the Dutch province of North Holland) to be more easily available with NPs if the NP is unstressed.

16

Vanacker (1949: 38) discusses a Flemish dialect where CA is obligatory with pronominal subjects, but optional with full NPs. This observation has not been investigated systematically for other dialects.

17

Bayer (2014) questions whether the example in (30) is indeed grammatical. However, there are many examples of this type in the literature. Hoekstra and Smits (1997), for instance, provide this example: (i)

Iech

wil

waete

wievöl

geld-s

te

hob-s.

I

want

know

how.much

money-2p.sg

you

have-2p.sg

‘I want to know how much money you have.’ (Maastricht Dutch, van Ginneken 1939 as cited by Hoekstra and Smits 1997)

Complementizer Agreement

(32)

Ich ving det Marie of-s toow d’n ierste moˆs sien. I find that Marie or-2 P . S G you the first must be ‘I think that Marie or you should be the first.’ (Tegelen Dutch, Cremers, and van Koppen 2008: 1065)

Cremers and van Koppen show that the inflection on conjunctions can only appear in cases that allow coordination of complete embedded clauses, which are normally introduced by a complementizer. They also pursue the hypothesis that the agreement ending on the conjunction of ‘or’ in this example is actually present on an empty complementizer introducing an embedded finite clause. Finally, CA is also found on the complementizer in comparative clauses. Bayer (1984) and Fuß (2008) report that CA in Bavarian is only possible in clausal comparatives, but not in phrasal ones, see the data in (33). (33) a. D‘Resl is gresser [ als The-Resl is taller than ‘Resl is taller than you are.’ b. D’Resl is gresser [ als The-Resl is taller than ‘Resl is taller than you.’

wia-st du bist] as-2p.sg you are wia(*-st) du] as-2p.sg you (Bavarian, Bayer 1984: 269)

A subset of the Dutch dialects with CA does allow CA with phrasal comparatives, see (34). See Cremers and van Koppen (2008), and Barbiers et al. (2005: 36) for a map showing the exact distribution of this pattern. (34)

Du geloofst zeker niet dat er sterker is wie-st-u. you believe-2p.sg surely not that he stronger is than-2p.sg-you ‘You probably don’t believe that he is stronger than you.’ (Nieuwenhagen Dutch, Barbiers et al. 2006)

Movement out of the embedded clause (either by the subject or any other element) does not seem to affect the presence of CA in most dialects: (35)

DOOW denk ik de-s / *det de wedstrijd winnen zal-s. win will-2 P . S G youS G think I that-2 P . S G / that the game ‘YOU, I think will win the game.’ (Tegelen Dutch, van Koppen 2012: 137)

A similar example can be given for Frisian: (36)

Do, tink ik dat-st / *dat moarn komme sil-st you, think I that-2 P . S G / that tomorrow come shall-2 P . S G ‘I think YOU will come tomorrow.’

331

332

MARJO VAN KOPPEN

An exception to this generalization is Hellendoorn Dutch (see van Koppen 2012 for an explanation): (37)

WIEJ denkt Jan dat / *darr-e die pries ewo¨nnen hebt, have we think Jan that / that-1 P . S G that prize won nie ZIEJ. not they ‘WE John thinks won that prize, not THEM.’ (Hellendoorn Dutch, van Koppen 2012: 138)

Boef (2013) and Mayr (2010) propose a close connection between CA and subject extraction in the Dutch dialects and in Bavarian respectively. They argue that CA makes subject extraction possible. Mayr (2010) gives a particularly convincing argument for this, illustrated in (38). (38) a. [Es Kinda]1 hot da Hauns gfrogt [t1 ob-s t1 you children has the John asked if-2p.pl ‘John asked if you children will come home.’ b. *[Es Kinda]1 hot da Hauns gfrogt [t1 ob-ø t1 you children has the John asked if-ø

hamkummts] home_come hamkummts] home_come

(Bavarian, Mayr 2010: 121)

Mayr shows that CA is optional for some speakers of Bavarian. However, when the subject is extracted, the only grammatical option is the one with CA (i.e., example (38a)) and not the one without CA (i.e., example (38b)). A final point about the distribution of CA concerns the clauses it appears in. The dialects and languages that display CA are all asymmetric Verb Second languages, which means that the finite verb is in second position in main clauses and in final position in embedded clauses (Zwart 1993). There are several analyses of CA that make use of this observation (see, for instance, Zwart 1993, 1997, 2001). That the link between embedded Verb Second and the presence of CA is not coincidental can be seen in Frisian. This language (in contrast to most other Germanic languages and dialects) allows embedded Verb Second. Interestingly, CA cannot appear in these clauses (cf. De Haan and Weerman 1986, Zwart 1993, De Haan 2001). Consider the sentences in (39). (39) a. Heit sei dat-st do soks net leauwe dad said that-2p.sg you such not believe b. Heit sei dat(*-st) do moa-st soks dad said that(-2p.sg) you must-2p.sg such ‘Dad said that you should not believe such things.’

moa-st. most-2p.sg net leauwe. not believe

(Frisian, van der Meer 1991, as cited in Zwart 1993: 198)

Example (39a) is a regular embedded clause with CA and the verb in final position. The embedded Verb Second clause in (39b) cannot have CA however. De Haan (2001) convincingly shows that these clauses with

Complementizer Agreement

what seems to be embedded Verb Second are actually coordinated root clauses. The complementizer dat ‘that’ is not a subordinating conjunction in these clauses, but it is a coordinating conjunction. If these sentences are indeed root clauses, then we do not expect CA to occur, as it is restricted to embedded contexts (see Zwart 1993, 1997, 2001, Carstens 2003 for alternative accounts of these data).

14.7 Summary To summarize, this chapter has provided an overview of the empirical properties of CA in the West-Germanic languages. It focused on the morphological aspects of CA, like the properties of the paradigm, the properties of the affixes, but also on morphosyntactic aspects, like the effect of adjacency on CA, the relation between CA, clitics and pro-drop, and CA’s syntactic distribution.

References Ackema, P. and A. Neeleman 2004. Beyond Morphology. Interface Conditions on Word Formation. Oxford University Press. Barbiers, S., H. Bennis, G. De Vogelaer, M. Devos, and M. van der Ham 2005. Syntactic Atlas of the Dutch Dialects, Part 1. Amsterdam University Press. Barbiers, S., H. Bennis, G. De Vogelaer, M. Devos, and M. van der Ham 2006. “Dynamische Syntactische Atlas van de Nederlandse Dialecten (DynaSAND).” www.meertens.knaw.nl/sand/. Bayer, J. 1984. “COMP in Bavarian XE ‘Bavarian’ syntax,” The Linguistic Review, 3: 209–274. Bayer, J. 2014. “Syntactic and phonological properties of wh-operators and wh- movement in Bavarian.” In G. Grewendorf and H. Weiß (eds.), Bavarian Syntax. Contributions to the Theory of Syntax. Amsterdam: John Benjamins: 23–50. Boef, E. 2013. Doubling in Relative Clauses. Aspects of Morphosyntactic Microvariation in Dutch. Ph.D. dissertation, Meertens Instituut (KNAW) / Universiteit Utrecht: LOT Dissertation Series 317. Carstens, V. 2003. “Rethinking complementizer agreement: Agree with a case-checked goal,” Linguistic Inquiry 34.3: 393–412. Carstens, V. 2016. “Delayed valuation: A reanalysis of ‘upwards’ complementizer agreement and the mechanics of case,” Syntax, 1–42. Chung, S. 1998. The Design of Agreement: Evidence from Chamorro. University of Chicago Press. Craenenbroeck, J. van and M. van Koppen 2002. “The locality of agreement and the CP-domain,” Handout Glow 2002, Amsterdam.

333

334

MARJO VAN KOPPEN

Cremers, C. and M. van Koppen 2008. “Boolean Agreement in Tegelen Dutch,” Lingua 118.8: 1064–1079. Diercks, M. 2010. Agreement with Subjects in Lubukusu. Ph.D. dissertation, Georgetown University. Diercks, M. 2013. “Indirect agree in Lubukusu complementizer agreement,” Natural Language & Linguistic Theory 31.2: 357–407. Fuß, E. 2004. “Diachronic clues to pro-drop and complementizer agreement in Bavarian.” In E. Fuß and C. Trips (eds.), Diachronic Clues to Synchronic Grammar. Amsterdam: John Benjamins: 59–100. Fuß, E. 2005. The Rise of Agreement: A Formal Approach to the Syntax and Grammaticalization of Verbal Inflection. Amsterdam: John Benjamins. Fuß, E. 2008. “Multiple agreement and the representation of inflection in the c-domain,” Linguistische Berichte 213: 77–106. Ginneken, J. van 1939. “De vervoeging der onderschikkende voegwoorden en voornaamwoorden,” Onze Taaltuin, 8:1–11. Goeman, T. 1980. “Comp-Agreement?” In W. Zonneveld and F. Weerman (eds.), Linguistics in the Netherlands 1977–1979. Dordrecht: Foris: 291–306. Goeman, T. 1997. “Historiografie van het onderzoek naar voegwoordvervoeging: een bibliografisch overzicht (1821–1997).” In E. Hoekstra and C. Smits (eds.), Vervoegde voegwoorden. Amsterdam: Cahiers van het P. J. Meertensinstituut 9: 112–145. Goeman, T. 2000. “Structurele aspecten van de morfologie van voegwoordvervoeging: mogelijkheden en beperkingen, morfologisch gewicht en MCGG.” In V. de Tier, M. Devos, and J. van Keymeulen (eds.), Nochtans was scherp van zin. Huldealbum Hugo Ryckeboer. Een bundel artikelen aangeboden aan Hugo Ryckeboer voor zijn 65e verjaardag. Gent-Deinze: Vakgroep Nederlandse Taalkunde van de Universiteit Gent-Van Daele: 269–294. Gruber, B. 2008. Complementizer Agreement: New Evidence from the Upper Austrian Variant of Gmunden. Master’s thesis, University of Vienna. Haan, G. de 1997. “Voegwoordcongruentie in het Fries.” In E. Hoekstra and C. Smits (eds.), Vervoegde voegwoorden. Amsterdam: Cahiers van het P.J. Meertensinstituut 9: 50–67. Haan, G. de 2001. “More is going on upstairs than downstairs: Embedded root phenomena in West Frisian,” Journal of Comparative Germanic Linguistics 4: 3–38. Haan, G. de and F. Weerman 1986. “Finiteness and verb fronting in Frisian.” In H. Haider and M. Prinzhorn (eds.), Verb Second Phenomena in Germanic Languages. Dordrecht: Foris: 77–110. Haeberli, E. 1999. Features, categories and the syntax of A-positions. Ph.D. dissertation, University of Geneva. Haegeman, L. 1992. Theory and Description in Generative Syntax. Cambridge University Press. Haegeman, L. 2000. “The external possessor construction in West Flemish,” GG@G1:1–20.

Complementizer Agreement

Haegeman, L. and M. van Koppen 2012. “Complementizer Agreement and the relation between T and C,” Linguistic inquiry 43.3: 441–454. Haeringen, C. van 1958. “Vervoegde voegwoorden in het Oosten,” Driemaandelijkse bladen 19: 115–124. Hoekstra, E. and C. Smits 1997. “Vervoegde voegwoorden in de Nederlandse dialecten.” In E. Hoekstra and C. Smits (eds.), Vervoegde voegwoorden. Amsterdam: Cahiers van het P. J. Meertensinstituut 9: 6–30. Hoekstra, E. and C. Smits 1999. “Everything you always wanted to know about Complementizer Agreement.” In E. van Gelderen and V. Samiian (eds.), Proceedings of WECOL 10. California State University: Fresno: 189–200. Hoekstra, J. 1997. “Pro-drop, cliticering en voegwoordcongrue¨ntie in het Westgermaans.” In E Hoekstra and C. Smits (eds.), Vervoegde voegwoorden. Amsterdam: Cahiers van het P. J. Meertensinstituut 9: 68–86. Hoekstra, J. and L. Mara´cz 1989. “On the position of inflection in West Germanic,” Working papers in Scandinavian Syntax, 44:75–88. Koppen, M. van 2005. One Probe, Two Goals: Aspects of Agreement in Dutch Dialects. Ph.D. dissertation, Leiden University. Utrecht: LOT Dissertations 105. Koppen, M. van 2012. “The distribution of phi-features in pronouns,” Natural Language and Linguistic Theory, 30: 135–177. Koppen, M. van 2017. “Complementizer agreement.” In M. Everaert and H. van Riemsdijk (eds.), Companion to Syntax. Malden, MA: Blackwell-Wiley. Mayr, C. 2010. “On the necessity of phi-features: The case of Bavarian subject extraction.” In P. Panagiotidis (ed.), The Complementizer Phase: Subjects and wh-dependencies. Oxford University Press: 117–142. McCloseky, J. 2001. “The morphosyntax of WH-extraction in Irish,” Journal of Linguistics 37: 67–100. Meer, G. van der 1991. “The ‘conjugation’ of subclause introducers: Frisian -st,” North-Western European Language Evolution (NOWELE), 17:63–84. Putnam, M. T. and M. van Koppen 2011. “All there is to know about the allsconstruction,” Journal of Comparative Germanic Linguistics 14.2: 81–109. Rooryck, J. 2000. Configurations of Sentential Complementation: Perspectives from Romance Languages. London: Routledge. Schiepek, J. 1899 / 1908. Der Satzbau der Egerla¨nder Mundart. 2 Teile. Prag: Verlag des Vereines fu¨r Geschichte der Deutschen in Bo¨hmen. Stroop, J. 1987. “Enclitische verschijnselen in het Westbrabants,” Taal en Tongval 39:121–140. Vanacker, V-F.1949. “Over enkele meervoudsvormen van voegwoorden,” Taal en Tongval 1: 32–45, 77–93, 108–112. Vogelaer, G. de 2006. Subjectsmarkering in de Nederlandse en Friese Dialecten. Ph.D. dissertation, Ghent University. Weber, A. 1964. Zu¨richdeutsche Grammatik. Ein Wegweiser zur guten Mundart. Zu¨rich: Schweizer Spiegel Verlag. Weise, O. 1907. “Die sogenannte Flexion der Konjunktionen,” Zeitschrift fu¨r Deutsche Mundarten 2: 199–205.

335

336

MARJO VAN KOPPEN

Weiß, H. 2005. “Inflected complementizers in Continental West Germanic Dialects,” Zeitschrift fu¨r Dialektologie und Linguistik 72: 148–166. Weiß, H. 2012. “The diachrony of complementizer agreement.” Paper presented at the Complementizer Agreement Workshop, October 17, University of Gent. www.gist.ugent.be/file/366. Westfries Genootschap (ed.)1929. West-Friesland’s Oud en Nieuw Jaarboek. Zwart, J-W. 1993. Dutch Syntax: A Minimalist Approach. Ph.D. dissertation, University of Groningen. Zwart, J.-W. 1997. A Minimalist Approach to the Syntax of Dutch. Dordrecht, Boston, and London: Kluwer. Zwart, J.-W. 2001. “Syntactic and phonological verb movement,” Syntax 4.1: 34–62. Zwicky, A. 1977. “On clitics,” Indiana University Linguistics Club. Zwicky, A. and G. Pullum 1983. “Cliticization vs. inflection: The case of English n’t,” Language 59.3: 502–513.

Part III

Syntax

Chapter 15 VO-/OV-Base Ordering Hubert Haider

15.1 Introduction The base order serialization of verbs in relation to their nominal objects partitions the contemporary Germanic languages into two major groups. The VO group consists of the North Germanic languages plus English and their regional varieties. The OV group comprises Afrikaans, Dutch, Frisian, German, and their regional varieties.1 Yiddish is the only Germanic language that does not neatly fit into this partitioning since objects may precede as well as follow a nonfinite main verb. The assignment of Yiddish to one of the two base order types is controversial: See the examples (3) below. With respect to the internal syntax of the verb phrase, the Scandinavian languages and English are fairly uniform. The objects follow the main verb, as exemplified by Swedish (1). For checking the VO characteristics of a given language, the relevant verb positions are the positions of nonfinite verbs such as infinitives (1a, 1c) and supines (1b, 1d, 1e). Finite verbs (Vikner, Chapter 16) are subject to superimposed ordering constraints since the Germanic languages, except English, are V2-languages, that is, languages in which finite verbs are fronted in main clauses and in certain embedded contexts in North-Germanic languages. (1)

1

a. Jag ska ko¨pa en bil / den. (Swedish, Lundquist 2014a) I shallPRES buyINF a car / it. b. Jag har gett mannen boken. (Swedish, Lundquist 2014b) I have givenSUP manDEF bookDEF c. *Jag ska en bil / den ko¨pa. (Swedish) I shallPRES a car / it buyINF

For the purpose of this chapter is not essential to decide whether Swiss German varieties or Luxembourgish, to name just two cases, ought to be filed as different languages or different varieties within the German language continuum since the base order characteristics of a main verb relative to its objects is OV, just like in standard German.

340

HUBERT HAIDER

d. *Jag har mannen boken gett. (Swedish) e. *Jag har mannen gett boken. (Swedish) In the Germanic OV languages, the nonfinite verb follows its nominal objects. The relative order of the objects in VO and OV is identical (1b, 2b). The verb positions are different. The ungrammatical orders of VO, viz. (1c, 1d), are grammatical orders in OV languages (2a, 2b, 2d), except for (1e). The grammatical orders of Swedish are ungrammatical serializations in OV languages. In the Germanic OV languages, nominal arguments obligatorily precede the VP-final base position of the main verb. The neutral order for two nominal objects of verbs with an experiencer theta-role for the indirect object is IO-DO-V (2b, 2d). Pronominal objects, however, are subject to a separate ordering template. Their relative order is DO-IO-V (2c, 2e). In German, this amounts to accusative before dative. (2)

a. Heute wird er das Auto / es kaufen today shall he the car / it buy b. Heute habe ich dem Mann das Buch gegeben today have I the manDAT the bookAKK given c. Heute habe ich es ihm gegeben today have I itAKK himDAT given d. Vandaag heeft de man zijn broer een boek today has the man his brotherIO a bookDO gegeven given e. De man heeft het hem niet gegeven the man has itDO himIO not given

(German) (German) (German) (Dutch)

(Dutch)

The pattern with the nonfinite main verb sandwiched by its objects (1e) is ungrammatical in OV as well as in VO, but not in Yiddish. According to Diesing (1997: 402), this word order is an acceptable order variant (3c, 3d). This is confirmed by Kroch and Santorini (2016) with corpus data.2 Given the serialization properties illustrated in (3), Yiddish cannot be unequivocally filed as either VO or OV. In Haider (2014, 2015), it is argued that Yiddish has preserved word order properties common to all older stages of Germanic languages before the diachronic split into an OV and a VO group has become manifest, namely unspecified directionality of the verbal head in the VP. The following examples are quoted from Diesing (1997: 402). The patterns are grammatically well-formed, but the variants differ with respect to pragmatics, that is, information structure effects. These are the by now familiar preferences for serializing ‘given’ before ‘new’, ‘background’ before (postverbal) ‘focus’, ‘topic’ before ‘comment’, just as in other 2

Based on a Yiddish Corpus of Santorini’s, Kroch and Santorini (2016) report the following figures of a corpus search for double object patterns. The counts have considered only nonfinite VPs. After 1900: IO-V-DO: 6; DO-V-IO: 6; V-NP-NP: 25; NP-NP-V: 2. 1800–1900: IO-V-DO: 3; DO-V-IO: 1; V-NP-NP: 8; NP-NP-V: 10.

VO-/OV-Base Ordering

languages in which grammar does not strictly constrain serialization and thereby allows for pragmatic factors to get superimposed, taking advantage of the flexibility that syntax admits. (3)

a. Maks hot Rifken dos bukh nit gegebn Max has Rebecca the book not given b. Maks hot nit gegebn Rifken dos bukh c. Maks hot Rifken nit gegebn das bukh d. Maks hot das bukh nit gegebn Rifken

OV-like order VO-like order IO-V-DO order DO-V-IO order

These three word-order patterns of Germanic languages, namely VO, OV, and “unspecified” can be synopsized as parametric variants of VP structuring. In North Germanic and English, the VP is head initial; in the continental West-Germanic languages and in Afrikaans it is head final, while it appears to be flexible in Yiddish. The Germanic VO languages are head initial across all categories. In the OV group, however, the head positioning co-varies with the lexical category of the head. Verb and adjective phrases are head final, while noun phrases and PPs are head initial. For NPs and VPs it is particularly easy to provide minimal pairs since infinitival verb forms can be converted into nominal forms without any further morphological modifications. The direct object of the verb then turns into the object of the noun, either as a PP object (4b) or as a genitive object (4d). The object precedes the verb (4a, 4c) but follows the noun (4b, 4d) and these are the only licit linearizations. (4)

a. [een container naar Madagaskar sturen]VP a container to Madagascar send b. het [sturen van een container naar Madagaskar]NP the sendINF of a container to Madagascar c. [einen Container nach Madagaskar transportieren]VP a container to Madagascar transport d. das [Transportieren eines Containers nach the transportINF aGEN containerGEN to Madagaskar]NP Madagascar

(Dutch) (Dutch) (German) (German)

Functional lexical heads such as articles3 (see Roehrs, Chapter 23) and complementizers precede their complements in all Germanic languages. Arguably, there are no functional heads at all that follow their complements, as for instance a clause-final position for the finite verb (see Haider 2010, chapter 2). The position of the head in a phrase correlates with a number of syntactic properties that will be reviewed in the following section. Since the VP is 3

The suffixed definiteness marker of nouns in Scandinavian is an affix, not an article in a D°-position, such as Swedish flask-an (the bottle) or brev-et (the letter). The definite article is den, det (neuter), and de (plural).

341

342

HUBERT HAIDER

Table 15.1 A synopsis of syntactic correlates of OV / VO OV

VO

i. ii.

particle verbs resultatives

particle ← V resultative ←V

V → particle V → resultative

iii. iv.

order of auxiliaries VP-medial adverbs

. . . (XP) V ← Aux ☑ [DP adverb V]VP

. . . Aux → V (XP) . . . ☒ [V adverb DP]VP

v. vi.

left-adjoined adjuncts VP-internal scrambling

unconstrained ☑ [DPi DP ei V]VP

head-adjacency ☒ [V DPi DP ei]VP

vii. viii.

V-V-complementation expletive or quirky subject

. . . [. . . [V° V°]verbal cluster]VP4 excluded

. . . [V° [V° . . .]VP]VP obligatory expletives

a core constituent of a clause, its syntactic properties are reflected in the syntactic properties of clauses. The first three properties (i.–iii.) are subinstances4 of the directionality property that determines V-O and O-V order, respectively. The arrows symbolize the directionality relation between the head and the element it combines with. “Particles” refers to those particles of particle verbs that are obligatorily stranded when the finite verb occurs in the fronted position, viz. in V2 and V1 clauses; see also Toivonen (Chapter 22) and Zwarts (Chapter 26). For auxiliaries, the directionality-sensitive property is their selection effect on the form of the dependent verb. A main verb selects category and case of its nominal complements; an auxiliary selects a verbal form5 of the dependent verb such as a bare infinitive, an infinitive with infinitival particle, a participle or an aspectual form as for instance the English progressive form. Moreover, in VO languages, the relative order of auxiliaries is strict. Germanic OV languages, on the other hand, are known for order variations. The second triplet (iv.–vi.) lists the effects of syntactic constraints that are unique for head-initial VPs. These restrictions are absent in head-final (verb) phrases. Head-initial VPs (and NPs as well) are “compact.” There is no room for interveners between the head and the nominal objects, neither for adverbials nor for scrambled arguments. Adjuncts in head-initial VPs that precede the verb are subject to an adjacency constraint. The head of the adjunct must be adjacent to the VP. This constraint is absent in headfinal structures. The very same constraint applies to adnominal attributes (for a theoretical account see Haider [in press]). 4

This two-verb cluster here merely serves as an example. Verb clusters may easily consist of up to five verbs. i. da man ihn [liegen bleiben lassen können wird]

(German, Bech 1957 / 1983: 64)

since one him [lie remain let can will] (‘since one shall be entitled to let him stay recumbent’) ii. dat hij de chauffeur [willen laten blijven wachten had] (Dutch Augustinus 2015: 8) that he the driver want let stay wait had (‘that he had wanted to keep the driver waiting’) 5

The analogy between case government and “status government” (“Statusrektion”) has been emphasized first by Bech (1957, 1983: 17): “statusrektion und statuskongruenz [. . .] in genauer analogie mit der terminologie der kasuslehre” (= status government and status agreement in precise analogy to the terminology of the case systems). The verbal status categories are bare infinitive, supine (past participle), and infinitive with particle ‘to’.

VO-/OV-Base Ordering

The final two properties concern follow-up properties of the head-initial versus head-final structure of VPs. The Germanic languages with head-final VPs employ verb clustering instead of stacking VPs as in VO languages. In verb clusters, the main verb plus the auxiliaries and quasi auxiliaries of a simple clause form a syntactic unit. In VO languages, the auxiliaries and quasi-auxiliaries each select a VP, with the VP of the main verb as the most deeply embedded one. The final property is the hallmark of the SVO type. There are two types of languages, in which the main verb precedes the objects, namely SVO and VSO. Only in the SVO type, the subject is outside of the directionality domain of the verb. It is assigned to a special position outside of the VP. This position precedes the main verb while the object positions follow the main verb. In OV language, any nominal argument precedes the main verb. In SVO, the structural subject position is obligatory and it is obligatorily lexicalized. In the absence of a subject argument, an expletive element lexicalises this position. In OV languages, there is arguably no VP-external structural subject position and therefore no room for an expletive subject in an otherwise subjectless clause (Haider 2010: 36).

15.2 Syntactic Correlates of the Base-Ordering Types 15.2.1 Particle Verbs Particle verbs are highly productive combinations of a particle and a verb in Germanic languages (see Dehe´ 2015, Toivonen, Chapter 22). Particle verbs are syntactic units consisting of two head-level elements, namely a particle and a verb.6 Most of these particles are homophonous with prepositions of the particular languages. The linearization of the particles relative to the verb matches the OV and VO order, respectively. In VO languages, the particles follow the verb: In OV languages they precede, unless they are stranded when the finite verb is in a displaced position, such as the V2-position. Here is an illustrative cross-linguistic sample. (5)

look up (VO) a. b. c. d. e.

6

kig op sla´a upp fletta upp slaº opp slaº upp

‘up-look’ (OV) (Danish) (Faroese) (Icelandic) (Norwegian) (Swedish)

opkyk opzoeken opsykje aufschlagen oyfzukhn

(Arikaans) (Dutch) (Frisian) (German) (Yiddish)

Particle verbs are syntactically complex. They must be distinguished from morphologically complex verbs. In German, for instance, ‘umfahren’ (bypass versus run somebody down) or ‘übersetzen’ (ferry across versus translate) are ambiguous in their written form. Stress disambiguates. If the verb is the particle verb, the particle carries the main stress. If the initial morpheme is a derivational prefix, the verbal root carries the stress. When the verb is finite and gets fronted, the particle is stranded. The prefix could not be stranded since it is part of the word structure of the verb.

343

344

HUBERT HAIDER

In VO languages, the verb and the particle may occur in nonadjacent positions within the verb phrase. The Scandinavian languages display several distinct patterns, as illustrated by the examples (7) from Thra´insson (2007: 34, 142). Faroese, Icelandic, and Norwegian pattern like English (6). The particle may be stranded after an object unless the object is a pronoun. In Danish, the particle must be stranded behind an object (7a), while in Swedish it must not (7d). For more details, see Svenonius (1996), Vikner (2017). (6)

(7)

a. He looked something/it up7 b. he looked up something/*it c. Hann gjørdi upp snøri/*ta. he made up fishing-line-DEF/*it ‘He wound up the fishing line.’ d. Hann gjørdi snøri/ta upp. e. E´g skrifaði niður sı´manu´merið/*það. I wrote down phone-number-DEF/it f. E´g skrifaði sı´manu´meriðniður/það niður. g. Han spiste opp tørrfisken/*den. he ate up dryfish-DEF/*it h. Hann spiste tørrfisken/den opp. a. Jeg skrev nummeret/det op. I wrote number-DEF /it down b. Jeg skrev op *nummeret/*det. I wrote up *number-DEF/*it c. Hon kastade ut Johan/honum. she threw out J./him d. *Hon kastade Johan/honom ut.

(English) (Faroese)

(Faroese) (Icelandic) (Icelandic) (Norwegian) (Norwegian) (Danish) (Danish) (Swedish) (Swedish)

In VO languages with optional particle stranding, but not in OV languages, particles may appear in a position between two objects (8a, 8b). Such a particle position is ungrammatical in any OV language. Only within a verbal cluster could a particle be separated from its verb, as in Dutch (8c) in comparison with (8d, 8e). In this case, the only intervening items are verbs of the same verbal cluster. (8) a. John sent the stockholders out a schedule. hafa þeir sent stra´kunum upp b. gær yesterday have they sent boys-DEF up peningana. money-DEF

7

(Neeleman 2002: 141) (Collins and Thra´insson 1996: 435)

This example with an indefinite object is documented in the British National Corpus. For full noun phrases, superimposed information structure effects result in a preference for definite noun phrases preceding the particle, but big corpora contain entries of the pattern with an indefinite preceding a particle, too, for this and other particle verbs, but more than twice as many with definite noun phrases.

VO-/OV-Base Ordering

c. omdat de rapper niet mee heeft gewerkt aan het onderzoek (De Telegraaf Sept. 15, 2016) since the rapper not with has worked at the investigation (with-work = cooperate) d. dat hij aan die tweede inbraak niet heeft meegewerkt (Het Laatste Nieuws July 11, 2016) that he at the second break-in not has participated e. Dokter Moser [. . .] die niet meegewerkt heeft aan de studie (VRT Nieuws July 9, 2015) doctor Moser [. . .] who not collaborated has on the study

15.2.2 Resultatives In all Germanic languages, particles, adjectives as well as PPs are employed for resultative readings, that is, for denoting a property of the resulting state. What Broekhuis (2013:135–136) affirms for Dutch is true for all Germanic OV languages: “Both resultatives and particles must be leftadjacent to the verbs in clause-final position.” (9)

a. als er es aufmachte when he it up-made (‘up-make’ = open) b. als er es sauber machte when he it clean made c. als er es zu einem Problem machte when he it to a problemDAT made ‘when he turned it into a problem’

(German) (German) (German)

Broekhuis’ statement applies not only to the base position of these items. Just like particles (10a), resultative PPs (10b, 10c), unlike locative PPs (10d) or prepositional objects (10e), do not extrapose and they cannot be stranded by V(P) topicalization (11). (10) a. dat Jan het boek {neer} legde {*neer} (Dutch, Broekhuis 2013: 136) that Jan the book down put b. dat Jan het boek {op de tafel} legde {*op de tafel}.8 (Dutch) that Jan the book on the table put c. Jan hat alles {in kleine Stu¨cke} (German) Jan has everything in small pieces geschnitten {*in kleine Stu¨cke} cut

8

In (i), the extraposed PP is not directional but locative. In German, this is reflected by dative (locative) instead of accusative (directional) case assigned by the preposition ‘auf’ (on). i. dat Jan het boek {op de tafel} neer legde {op de tafel}. (Dutch) that Jan the book on the table down put ii. dass Jan das Buch auf denAKK Tisch legte – dass Jan das Buch {auf demDAT Tisch} ablegte {auf dem Tisch} that Jan the book on the table put – that Jan the book on the table down-put

345

346

HUBERT HAIDER

(Dutch, Broekhuis d. Jan heeft op zijn vader gewacht op het station. 2013: 368) Jan has for his father waited at the station e. Jan heeft op het station gewacht op zijn vader. (Dutch) Participles or infinitival verbs may be topicalized (11a), but topicalization of the verb must not strand a directional or resultative PP (11b). These expressions are pied-piped (11c). (11) a. Gewartet hat er auf sie. waited has he for her b. *Gestellt hat er es auf den Tisch. put has he it on the table

Warten wird er auf sie. (German) wait shall he for her *Getrunken hat er das Glas drunk has he the glass leer emptyresultative c. [Auf den Tisch gestellt] hat er es [Leergetrunken] hat er das Glas on the table put has he it empty-drunk has he the glass

In VO languages, resultatives of any category – particles, adjectives, PPs – follow the verb (12). The relation between the verb and these elements is tight. There is no room for intervening adverbs (12a–12c). In languages with optional particle stranding, a parallel pattern is attested for resultative adjectives (12d, 12e). If the object is pronominal, it precedes the particle as well as a resultative. (12)

a. b. c. d. e.

He wiped it (*yesterday) up (English) He wiped it (*yesterday) clean He wiped it (*yesterday) under the door handle Then they cut {loose} the craft {loose} Vi vaska {reint} golvet {reint} (Norwegian, A˚ farli 1985: 97) we washed clean floor-DEF f. Þeir dældu hana fulla (Icelandic, Whelpton 2007) they pumped her-fem-ACC-sg full-fem-ACC-sg af lyfjum of drugs

15.2.3 The Order of Auxiliaries In all Germanic VO languages, the relative order of auxiliaries and quasiauxiliaries is uniform and invariant. The dependent verb follows the verb it is dependent on in terms of status selection (see fn. 5). In (13a), ‘skulle’ selects an infinitival form which in turn selects a supine. The auxiliary order in (13a,b) is representative of the Germanic VO languages; see Thra´insson (2007: 459), Bentzen (2005: 156). Frisian (13c) illustrates the OV order of verbs, which is a mirror image of the VO order. The most frequent order of verbs in a cluster in Dutch (13d) is the inverted order of

VO-/OV-Base Ordering

the Frisian example. The German example (13e) appears like a mixture of the Dutch and the Frisian order. The first two verbs follow the Dutch order while the other four verbs are serialized as in Frisian. This is the effect of a special constraint that is operative in (13e), namely the IPP9 constraint discussed below (see also Wurmbrand and Christopoulos, Chapter 17). In general, the verbs in German are serialized in the OV order, that is, the order illustrated for Frisian (13c). (13) a. Du skulle ha

kunnet temme løver naº . (Norwegian, Hauge 2003: 63) you should have canSUP tame lions now (Icelandic, Thra´insson 2007: 56)

b. Það munu aldrei margir hafa

lokið there will never many haveINF finishedSUP verkefninu the assignment-DEF c. omdat ik dy deˆr wol ris stean bliuwen sjen (Frisian, Hoekstra 1998: 155) that I you there M P M P stand stay see wollen hawwe soe want-Participle have would

d. dat ik je daar wel eens zou

hebben that I you there M P M P would have willen zien blijven staan want see stay stand

(Dutch, Hoekstra 1998: 155)

e. dass ich dich da

ja wu¨rde haben stehen bleiben sehen that I you there M P would have stand stay see wollen want

(German)

Compared to the invariant relative order of auxiliaries in VO, the grammar of auxiliaries is more complex in the Germanic OV languages. First, there is order variation within a given language (14a, 14c), and second, there is cross-linguistic order variation (14c, 14d). In Dutch, there is a slight NorthSouth gradient for the order (14a) over (14c), but both orders are frequent in The Netherlands and in Belgium.10 (14)

9

a. dat niemand iets gedaan heeft that nobody anything done has b. dass niemand etwas getan hat that nobody anything done has c. dat niemand iets heeft gedaan

(Dutch) (German) (Dutch)

IPP (Infinitivus pro participio = infinitive for a participle; “Ersatzinfinitiv”) refers to the following phenomenon: An auxiliary that selects a participial form on the dependent verb is fronted across the verb in the verb cluster and the verb switches into the infinitival form. In German, IPP is triggered by modal verbs and other quasi-auxiliaries like ‘lassen’ (let), ‘brauchen’ (need) and perceptions verbs like ‘sehen’ (see) and ‘hören’ (hear). In Dutch and Afrikaans, the set of IPP triggers comprises more kinds of verbs than in German. i. *dass er es essen gemusst hat )IPP dass er es hat essen müssen / dass er das essen hat müssen that he it eat mustParticiple has )IPP that he it has eat mustINF / that he it eat has mustINF (= ‘that he had to eat it’)

10

A google search for “gedaan heeft” produced 662, 000 hits, and 13,700 for news sites: “heeft gedaan” 565,000 and 43,800, respectively.

347

348

HUBERT HAIDER

d. *dass niemand etwas hat getan (German) e. dat hy later sou/moet gekom het (Afrikaans, Donaldson that he later would / must come have 1993: 366) f. of hy miskien sou kan kom help (Afrikaans) whether he perhaps would can come help A complicating factor is the IPP phenomenon. In German (15a), ‘wu¨rde haben’ (would have) precedes the main verb in the verbal cluster because of the IPP effect triggered by ‘ko¨nnen’. In Dutch, (15b), all auxiliaries are in the inverted order, which would be ungrammatical in German. Afrikaans11 optionally allows an in situ switch to the infinitival form (15c, 15d). (15)

a. dass er jetzt Lo¨wen wu¨rde haben za¨hmen ko¨nnen (German) that he now lions would have tame canInf. (= be able to) b. dat hij nu leeuwen zou hebben kunnen temmen (Dutch) that he now lions woul have canInf. tame c. Ons het kom kuier (Afrikaans, Augustinus we have comeINF visitINF and Dirix 2013: 221) d. ’n vragmotor wat aangery gekomParticiple het(Afrikaans) a lorry which driving-along come has

But even without IPP, there is variation in the order of verbs in the cluster as (16) illustrates for Dutch and (17) for German four-verb clusters. The numbers are the google hits with the filter “News” followed by the number of google hits in toto. The variant (16d) is ungrammatical in Dutch. All variants are synonymous. (16)

a. gebeurd zou kunnen zijn happenedParticiple would canINF beINF ‘would be possible to have happened’ b. zou kunnen gebeurd zijn c. zou kunnen zijn gebeurd d. *zou gebeurd zijn kunnen

(111; 24,000)

(32; 3,000) (4; 60,000) (0; 0)

(17a) is ungrammatical, because the modal is a trigger of IPP. Therefore, the auxiliary ‘have’ gets fronted, but there are several alternative positions, namely (17b–17d). (17)

11

a. *passiert sein gekonnt ha¨tte happened be canParticiple hadPast-SUBJ ‘could have had happened’

(0; 0)

An in-situ switch is characteristic of Eastern Austrian vernacular (i). The standard versions are as in (ii). i. dass sie nicht nachgeben müssen / dürfenINF hätte that she not give-in must / may had ii. dass sie nicht hätte nachgeben müssen / dürfen – dass sie nicht nachgeben hätte müssen / dürfen

VO-/OV-Base Ordering

b. passiert sein ha¨tte ko¨nnen c. passiert ha¨tte sein ko¨nnen d. ha¨tte passiert sein ko¨nnen

(2; 1030) (0; 508) (8; 2350)

15.2.4 VP-Medial Adverbs The positioning of adverbs sharply differentiates between VO and OV. In VO, adverbs either precede the VP or follow the verb plus its nominal arguments, but they do not intervene. The pattern (18a), which Engdahl et al. (2003: 43) provide for Swedish, is representative of North Germanic languages and English, modulo V2. The Icelandic example (18c) illustrates the general pattern (Thra´insson 2007: 37). In OV, however, adverbs may intervene between the verb and nominal objects (18d, 18e). (18)

a. XP VFIN Subj S-ADV VNNON-FIN ObjIND ObjDIR ADV b. She had {often} read {*often} the instructions {often}. c. Hu´n hafði {oft} lesið {*oft} leiðbeiningarnar {oft}. (Icelandic) she had {often} read {often} instructions-DEF {often} d. wenn man jemandem absichtlich etwas mehrmals erkla¨rt (German) if one somebodyDAT intentionally somethingACC repeatedly explains e. of iemand {toevallig} iets {toevallig} gezien heeft (Dutch) whether someone {accidentally}something {accidentally} seen has

If an adverb appears to intervene between nominal objects in a North Germanic language, this is the effect of an interfering condition, namely “object shift”; for details see Thra´insson (2010) and Broekhuis (Chapter 18). Object shift applies only if the VP is headless, that is, the verb is in a fronted position (19a, 19b). In addition, no lexical material must precede the shifted item in the VP, that is neither verb (19c) nor a stranded particle nor any other VP-internal item. Fronting a pronominal object across an adverbial or a negation particle is ungrammatical in all these instances; see Engdahl et al. (2003: 45) on Swedish. (19)

a. Bo gavi hendej aldrig [ei ej bogen]VP (Danish) Bo gave her never book-DEF b. Eva gavi honomj fo¨rmodligen inte [ei ej naº gra pengar]VP (Swedish) any money Eva gave himACC probably not c. *Eva har honomi fo¨rmodligen inte [gett ei naº gra pengar]VP (Swedish) Eva has him probably not [given any money]

In OV, the availability of the VP-medial positions as in (18d, 18e) for adverbs is a cross-linguistically valid property. It is available in a language with DP

349

350

HUBERT HAIDER

scrambling such as German but also in Dutch, which forbids the scrambling of a DP object across another object but scrambles PP objects.

15.2.5

A Constraint on Phrases Left-Adjoined to Head-Initial Phrases Adjunction to head-initial phrases is constrained in a way that adjunction to head-final phrases is not. The head of the adjunct must be adjacent to the phrase the adjunct is adjoined to. This is true for adjunction to VPs as well as for adjunction to NPs. English is representative of the Germanic VO languages in this respect. Since VP and NP are head initial, the effect shows in each case, as (20). (20) a. They [[much more often (*than the controls)] guessed the result correctly]VP b. an [[extremely happy (*with his score)] candidate]NP In the Germanic OV languages, the VP is head final and the NP head initial. So, adjunction to the VP is unconstrained (21a, 21b), while adjunction to the NP (21c, 21d) is subject to the very same adjacency constraint as in English and the North-Germanic languages. (21) a. als hij [verder dan vijf kilometer] had moeten rijden (Dutch)12 When he [more-far than five kilometres]had have-to ride b. So kann man Zu¨ge [viel genauer als (German)13 that-way can one trains [much more-precisely than bisher] lokalisieren to-date] localize c. eine [viel genauere (*als bisher)]A P Lokalisierung (German) a much more-precise (than to-date) localization (Dutch) d. een [minder intelligente (*dan/als Els]AP persoon a less intelligent (than Els) person This phenomenon has an immediate impact on the theoretical modeling of structures with adverbial or attribute phrases. It is a genuine effect of adjunction and it is absent with phrases in spec-positions, be it a subject or a phrase fronted to the clause initial-position. Hence, any claim that situates adverbial phrases or attributive APs in spec-positions of functional heads fails to capture this effect (Haider 2015 and in press).

15.2.6 VP-Internal Scrambling Variable ordering of argumental DPs (viz. “scrambling”) is a property of the containing phrases, namely head-final phrases. This is a necessary but not sufficient property. In addition, the scrambled DPs must be 12

Stephen King, 1989, De duistere kant.

13

Deutschlandfunk (website), Dec. 13, 2016.

VO-/OV-Base Ordering

morphologically identifiable, either by case or by relational particles. German meets both preconditions. As a consequence, the order of nominal arguments, subject included, is variable in head-final phrases, that is, in VPs (22a, 22b) and APs. On the other hand, head-initial phrases such as NPs (22c, 22d) are as rigidly serialized as in all Germanic VO languages. (22)

a. [SiegernDAT PokaleAKK u¨berreichen]VP (German) winners cups hand-over b. [PokaleAKK SiegernDAT u¨berreichen]VP c. das [U¨berreichen von Pokalen/der PokaleGEN an Sieger]NP the hand-overINF of cups/the cups’ to winners ¨ d. *das [Uberreichen an Sieger der PokaleGEN/von Pokalen]NP the hand-overINF to winners of cups/the cups’

In Dutch, nonpronominal DPs are morphologically indistinct with respect to subject versus object case and so they do not scramble. Unlike German, neither the nouns nor the head of the attributes nor the articles signal case distinctions and unlike in German “direct objects must appear to the right of indirect objects, while both of them must appear to the right of the subject” (Neeleman 1994a: 416). The German counterparts of (23b, 23c) are fully grammatical. (23)

a. dat Jan de mannen deze film toont (Neeleman 1994a: 416) that John the men this movie shows b. ??dat Jan deze film de mannen toont c. *Dat de mannen Jan deze film toont

The only distinctly marked arguments in Dutch are prepositional objects and these objects may be scrambled, as Algemene Nederlandse Spraakkunst (Geerts et al. 1984: 989f.), the standard grammar of Dutch, witnesses (24c). This is an essential difference between OV and VO in languages without morphological case-marking. In VO languages, prepositional objects cannot be scrambled (24d) although their grammatical function would be easily detectable. (24) a. *Toen hebben de autoriteiten het kind de moeder (Dutch) then have the authorities the child the mother teruggegeven. back-given b. Toen hebben de autoriteiten het kind aan de moeder then have the authorities the child to the mother teruggegeven back-given c. Toen hebben de autoriteiten aan de moeder het kind teruggegeven d. *Then, the authorities returned to the mother the child Icelandic and Faroese are testimonies for the rigidity of word order in VO even in the presence of morphologically clearly distinguishable objects. The

351

352

HUBERT HAIDER

Germanic languages provide immediate counterevidence for a popular generalization, namely the direct correlation between explicit case-marking and ‘free’ word order. Icelandic confirms insights already gained from German NP-internal word order (23d): The word order in head-initial phrases is strict and this is independent of rich case marking, which would guarantee clearly recoverable grammatical relations under any order. Dehe´ (2004) sums up her experimental study14 as follows: “The order of the objects in Icelandic double object constructions is much more restricted than one would expect. The unmarked order is by far the preferred one even in contexts where the inverted order is expected to be equally acceptable.” Thra´insson (2007: 98) indicates: “For a relatively small number of ditransitive verbs it is possible to reverse the ordering of the indirect and direct object” and these verbs “correspond roughly to the English variants where the goal follows the direct object, except that in English the goal would be prepositional (to the king, to the parents).” The last observation is particularly instructive since there exists a corresponding set of facts in German. For a small class of double-object verbs in German, A C C - D A T is the neutral order. The dative object of these verbs is a goal relation and not an experiencer relation as with the majority class of D A T - A C C verbs, and in addition, this goal-relation cannot be alternatively expressed by means of a PP object. However, in the majority class of D A T - A C C verbs, there is a subset of verbs whose dative object can be construed as an experiencer or a goal relation (25c–25e). (25) a. Sie setzten KandidatenACC TemparaturenDAT (German) they exposed candidates temperatures u¨ber 45° aus above 45° ‘They exposed candidates to temperatures above 45°’ b. ??Sie setzten Temperaturen u¨ber 45° Kandidaten aus c. Er u¨bergab alle Dokumente an die Polizei he over-gave all documents to the police d. Er u¨bergab alle DokumenteACC der PolizeiDAT he over-gave all documents the police e. Er u¨bergab der PolizeiDAT alle DokumenteACC The prepositional object variant (25c) is reserved for pure goal relations. The dative alternates with the PP option when the relation can be construed either as goal or experiencer-like relation. If the verb provides only an experiencer relation (26a), the PP option is not available (26b). (26)

14

a. Sie erkla¨rte ihmDAT das ProblemACC (German) she explained him the problem b. *Sie erkla¨rte das ProblemACC an ihnACC she explained the problem to him

18 participants; 36 target items plus 25 filler items; acceptability judgments on a 1–4 scale.

VO-/OV-Base Ordering

In Dutch, however, the alternation between indirect object and a prepositional object with aan’ (to) is available for a much large class of verbs than in German, as the contrast between (26b) and (27) illustrates. This larger class of double-object verbs is congruent with the English class of verbs that display the so-called dative-alternation, viz. an alternation between the indirect object variant and a prepositional object with the preposition ‘to’. (27)

Hij verklaarde het aan Constantijn Huygens. he explained it to C. H.

In Icelandic, the class of verbs that allow an A C C - D A T order is highly restricted. It is worthwhile pointing out that Dehe´’s (2004) findings call for a reassessment of frequently cited examples for a DOACC-IODAT order in Icelandic. Thraı´nsson (2007: 131) approvingly follows Ro¨gnvaldsson’s (1990) judgments of (28a, 28b): “Some double object verbs in Icelandic allow a DO–IO order of their arguments in addition to the normal IO–DO order.” Dehe´ (2004: 94), however, reports for the very same stimuli (28a, 28b) that “the inverted order was rejected” by all her informants. (28) a. Hann gaf konunginum amba´ttina. He gave king-DEF-DAT maidservant-DEF-ACC b. Þau sy´ndu foreldrunum krakkana. They showed parents-D E F - D A T kids-DEF-ACC In sum, even distinctly case-marked objects do not scramble in head-initial languages such as in Icelandic while morphologically poorly case-marked arguments may be scrambled in OV languages such as German. In the following example, only the dative of the indirect object is morphologically coded. The other two arguments in (29) can only be identified as nondatives. In principle, they could either be nominative or accusative. Nevertheless, scrambling is available. (29)

dass solche SpieleNOM/ACC KindernD A T alle ElternNOM/ACC verbieten sollten that such games children all parents forbid should ‘that all parents should forbid children such games’

(German)

The crucial difference between German and Dutch seems to be the following. In Dutch, the grammar does not provide any case distinctions for nonpronominal subjects. In German, there are morphologically coded case distinctions in all four cases, especially for articles and the heads of attributes. Even if these distinctions are morphologically neutralized in quite a few contexts, this does not preclude scrambling. What seems to bar scrambling is the principled morphological indistinctness of DPs in an OV language with respect to their grammatical functions, as in Afrikaans or Dutch. That scrambling must not change the linear order of object noun phrases in Dutch is standard wisdom. Nevertheless, “scrambling” has been invoked for Dutch in another respect, namely the serialization of

353

354

HUBERT HAIDER

adverbials relative to objects. The analysis rests on a doubtful premise, though, namely the premise of an exact parallel between adverbial placement in VO and OV. This parallel is arguably mistaken (see property iv. in Table 15.1). Adverbials are assumed to either precede or follow the VP. If the adverbial ‘gisteren’ in (30) has to precede (or follow) the VP just like in English, the objects have to be assigned to positions outside of the VP, as indicated in (30b). This is “Dutch scrambling”. This analysis has originally been proposed by Kerstens (1975). It has gained broad reception in generative accounts of Dutch clause structuring, but it has not reached full acceptance. Substantive counterarguments have been put forward in Neeleman and Weerman (1999), Neeleman (1994a, 1994b), and Neeleman and van de Koot (2008). In the analysis they argue for, the objects in (30a, 30b) are in their base-positions and the adverb in (30b) intervenes. This pattern is typical of OV structures, universally. (30) a. dat Jan de mannen de film gisteren toonde that Jan the men the movie yesterday showed b. dat Jan de manneni de filmj gisteren [ei ej toonde] The fact that the relative order of nominal arguments is rigid in Dutch prompted researchers to equate the Dutch phenomenon with Scandinavian object shift, as for instance Broekhuis (2008) and in Chapter 18. Vikner (2007: 411) has listed several independent syntactic contexts in which scrambling and Scandinavian object shift differ. The overarching difference is the fact that object shift must not cross any VPinternal material such as a verb, a stranded particle or a co-argument. “Only object shift requires verb movement, and only object shift is restricted to DPs.” (Vikner 2007: 393). This is not true of the OV-kind of scrambling, and it is not true for the alleged equation of scrambling and object shifting in Dutch. A clear case of scrambling in Dutch that plainly differs from object shift is the scrambling of PP objects. As noted by standard grammars, PP objects may be scrambled across an object in Dutch,15 just like in German, but they cannot be object-shifted in Scandinavian languages (Thra´insson 2007: 64) and they cannot be scrambled within a headinitial VP, as exemplified by English (Haider 2010:14) or in a head-initial NP, as illustrated in (22). Finally, in every OV language, and therefore in Dutch, too, adverbs may be placed in between objects that precede an adverb.16 This distribution is incompatible with object shift but it straightforwardly follows from the noncompactness property of headfinal phrases (see Section 15.2.4). 15

i. Toen hebben de autoriteiten aan de moederi het kind ei teruggegeven (Geerts et al. 1984: 989)

16

In (i), with a scrambled PP-object, the “%” sign marks positions for adverbials.

then have the authorities to the mother the child back-given. i. voor het slapengaan % aan kinderen % griezelige verhaaltjes % vertellen before the sleep-go(ing) % to children % creepy fairy-tales % tell

VO-/OV-Base Ordering

15.2.7 Verbal Clusters in OV – Stacked VPs in VO If in a VO-language a single clause contains more than one verb, then each verb heads a VP and so the VPs are stacked (31a). Each auxiliary is head of a VP, with another VP serving as its complement (31a). Stacking is the source of the rigid relative order of the auxiliaries in VO. In the Germanic OVlanguages, the verbs of a simple clause form a verb cluster (31b), as illustrated by Frisian (example 13c, in Section 15.2.3) or German: See Haider (2010, chapter 7.2), Wurmbrand (2017), and Wurmbrand and Christopoulos (Chapter 17). The variable verb order in verb clusters is the topic of Section 15.2.3 above. The verb order variations as well as the impenetrability for nonverbal material17 are characteristic properties of verbal clusters. (31) a. . . . [VP-1 V1 [VP-2 V2 [VP-3 V3 . . .]]] b. . . . [VP . . . [V-cluster V3 V2 V1]]

e.g., ‘shallV1 haveV2 answeredV3’ e.g., ‘beantwortetV3 habenV2 wirdV1’ German

The following example (32a) with an adverbial left-adjoined to each VP illustrates the stacked V-VP complementation structure of English. The Dutch examples (32b, 32c) illustrate the clustering phenomenon and one of its order variations: See Section 15.2.3 for verb order variation. (32)

a. It [certainly [VP may [possibly [VP have [indeed [VP been [badly [VP formulated]]]]]]]] (Quirk et al. 1985: section 8.20, 495) b. Hij zei dat hij de chauffeur [had willen laten blijven wachten] he said that he the driver has wantI N F letINF stayINF waitINF ‘He said that he wanted to let the driver wait’ (Augustinus 2015: 8) c. Hij zei dat hij de chauffeur [willen laten blijven wachten had] (Augustinus 2015: 8)

Clustering is not restricted to auxiliaries and quasi-auxiliaries. For a subset of control verbs, a clustering variant (33a, 33b) optionally alternates with the clausal infinitival complementation (33c). A serialization as in (33b) is a particularly unequivocal indication of a cluster since the infinitival main verb that otherwise heads the VP of the infinitival complement clause (33c) is flanked by the main verb and the auxiliary of the finite clause. The clustering variants (33a, 33b) entail clause union. The clause structure in the clustering variant is monoclausal while the control construction is biclausal, that is, it consists of a matrix clause with an embedded infinitival clause (33c). Obviously, this structural difference is the source of numerous predictable 17

The term “verbal material” includes particles of particle verbs (i) and resultative adjectives as in (ii.): i. dat hij haar had op moeten bellen. (Haeseryn et al. 1997: 1357) that he her had up must ring – ‘that he had to ring her up’ (opbellen = ring up) ii. dat hij zich niet [zal laten bang maken] (Haeseryn et al. 1997: 1358) that he himself not will let afraid make – ‘that he will not let (somebody / something) frighten him’

355

356

HUBERT HAIDER

follow-up effects. As for German, Haider (2010: 211–213) lists and discusses 16 syntactic collateral effects of clustering, confirming the structural difference. In Dutch, the IPP effect applies to these clustering control verbs as well,18 whence the infinitive ‘proberen’ in the clustering variant instead of the participle ‘geprobeerd’ in the clausal complementation variant (33c), that is, the familiar infinitival control construction. (33)

a. dat Jan Marie [heeft proberen te (Dutch, Augustinus 2015: 25) kussen]cluster that Jan Marie has tryINF to kiss b. dat Jan Marie [proberen te kussen (Dutch, Augustinus 2015: 25) heeft]cluster c. dat Jan geprobeerdParticiple heeft [Marie te kussen]clause

The clustering variant is restricted to a subclass of control verbs in the Germanic OV language. (34a, 34b) illustrates the parallel construction in Frisian and German. In the cluster (34b) the infinitival main verb of the clausal variant (34c) is sandwiched in the cluster, similar to Dutch (33b), due to the IPP effect triggered by the modal ‘ko¨nnen’ (can). (34) a. omdat er har mei har wurk [besocht te helpen]cluster (Frisian, Hoekstra 2016) because he her with her work tried to help b. dass Jan Marie [ha¨tte zu ku¨ssen versuchen ko¨nnen]cluster that Jan Marie had to kiss tryI N F canI N F ‘that Jan could have tried to kiss Marie’ c. dass Jan ha¨tte versuchen ko¨nnen [Marie zu ku¨ssen]clause that Jan had tryINF canINF Marie to kiss

(German)

(German)

15.2.8 Subject Expletives and Quirky Subjects The following examples (35) of passivized intransitive verbs illustrate a crucial contrast between Germanic OV and VO languages with respect to subjectless clauses. In the VO group, in the absence of a subject argument, the subject position is obligatorily lexicalized by means of an expletive element (Vikner 1995: 209). The Germanic OV languages do not require or admit an expletive (see Haider 2019). (35)

18

a. at der er blevet danset that E X P L has been danced b. att det arbetades that E X P L worksPass c. dat wordt gewerkt that is worked d. dass gearbeitet wird that worked is

(Danish) (Swedish) (Dutch) (German)

In German, the domain of IPP effects is more restricted than in Dutch. For instance, IPP does not affect these alternatively clustering control verbs.

VO-/OV-Base Ordering

Without an expletive, (35a, 35b) would be ungrammatical, such as the English passive of intransitive verbs. German, on the other hand, clearly demonstrates that an expletive is not admitted. The candidate for serving as expletive is ‘es’ (it), which is homophonous with the third person neuter pronoun. (36a) is unacceptable. The appreciation of Dutch is complicated by the fact that the candidate for the very same expletive function is ‘er’, which is homophonous with the locative adverbial particle, corresponding to English ‘there’. So, ‘er’ in (36b) could be an expletive or an adverbial. (36)

a. *dass es getanzt / gearbeitet wird that it danced / worked is b. dat er wordt gewerkt that there is worked

(German) (Dutch)

The test case is obvious. If it is an adverbial, it may be absent. If it is an expletive subject, it must be present, just as in (35a, 35b). The research literature on this issue is controversial. For Dutch syntacticians such as Hoekstra and Mulder (1990), Neeleman and Weerman (1999: 210–213) or Koeneman (2000: 192), ‘er’ in (36b) does not qualify as an expletive subject. Nowadays, such judgments are easy to confirm by corpus searches. A Google search for the strings in (37), restricted to news sites, produced the following results. The numbers in brackets are the results for the unrestricted search.19 (37)

a. b. c. d.

‘dat wordt gewerkt’ ‘dat er wordt gewerkt’ ‘dat wordt gesproken over’ ‘dat er wordt gesproken’ over’

594 1780 176 800

(345,000) (819,000) (162,900)20 (312,000)

Backed by such easily reproducible results, it is safe to agree with the above-mentioned grammarians and arrive at the following generalization, which is not restricted to Germanic languages. Germanic VO languages obligatorily lexicalize a VP-external structural subject position. No such requirement holds in the Germanic OV languages. Arguably, there is no VP-external subject position in OV languages, for principled reasons (see Section 15.3), because of the following structural difference. 19

A search for the following German string on news sites produced the following results; the unrestricted search hits (Feb. 21, 2017) are in brackets. i. ‘dass gearbeitet wird’ 77 (4,780) (that worked is) ii. * ‘dass es gearbeitet wird’ 0 (7) (that it worked is) iii. ‘dass da gearbeitet wird’ 1 (3,000) (that there worked is)

20

Note that Dutch does not strand prepositions except for combinations of er+P°. So, the examples without ‘er’ are genuine cases of subjectless clauses: i. dat wordt gesproken over een nieuwe crisis (that is talked about a new crisis) ii. de eerste keer dat wordt gesproken over een fusie (the first time that is talked about a fusion)

357

358

HUBERT HAIDER

In the SOV clause structure (38a), the head of the VP follows all of its arguments. They are allocated within the same directionality domain. In SVO, there is a mismatch (38b). One argument of the verb is not within the canonical directionality of the verbal head, namely the would-be subject. This argument gets reassigned to the pre-VP functional subject position, which is obligatory. Expletive subjects are a reflex of the need of lexicalizing it. The arrows in (38) indicate the directional licensing requirement that holds between a head and its dependents. See Section 15.3 for more detail. You may also wish to consult Haider (2010: 36, 2015: 86). (38)

a. [VP XPSubj ← [V’ ZP ←V°]] b. [FP XPj [F´ F°→ [VP ej [V° → ZP]]]]

If, as in the case of Icelandic, a nonnominative DP may occur in the VP-external subject position, viz. the position of XP in (38b), it gains the subject properties that are correlated with this position (Thra´insson 2007: 161–165). Typical instances involve verbs whose highest-ranking argument in the argument structure is a dative (39a). Passive is another source for an argument structures in which a dative argument is ranked higher than the derived nominative candidate (39b). (39) a. O¨llum

lı´kar þessi forseti. (Icelandic, Thra´insson 2007: 258) everybody-DAT-PL likes-3-SG president-DEF-NOM-SG verið sendir peningarnir? (Thra´insson 2007: 162) b. Hafa henni havePL her-D A T been sent-N O M - P L - M money-D E F - N O M P L -m ‘Has the money been sent to her?’

In (39b), passive turns the direct object into nominative in situ. This shows that nominative assignment is not restricted to the VP-external subject position. The dative argument as the highest-ranking argument is placed into the XP position of (38b). In German, the closest OV-counterpart of Icelandic, the relative order of arguments is the same as in Icelandic, but the preceding dative does not acquire any unequivocal subject properties. Example (40a) illustrates verbs with a dative-nominative base order. Example (40b) shows that the passive nominative stays in situ, which is confirmed by VPtopicalization (40c). (40) a. dass Journalisten dieser Pra¨sident gefa¨llt/missfa¨llt (German) that journalists-DAT this president-NOM pleases/displeases b. dass dem Angestellten das Geld u¨berwiesen wurde was that the employeeDAT the money-NOM sent c. [Gelder u¨berwiesen]VP wurden dem Angestellten. money-NOM-PL transferred were-PL the employee-DAT

VO-/OV-Base Ordering

The grammatical contrast between Icelandic and German with respect to nonnominative subjects is an immediate collateral effect of the OV- versus VO-based clause structure. Only in the VO-clause structure (38b), there is a VP-external structural subject position which awards subject properties to any phrase in this position.21

15.3 Coming to a Theoretical End What are the regulating grammatical principles responsible for the contrasting patterns reviewed above? The first subset – 15.2.1 to 15.2.3 – merely reflects the directional implementation of the head-dependent relation. Any dependent element of the head has to appear in the canonical direction. Therefore, in the base order, nominal objects, particles, resultatives, and dependent verbal projections either precede (OV) or follow (VO) the verbal head they depend on. The second subset – 15.2.4 to 15.2.6 plus 15.2.8 – is the challenging set for any theoretical modeling. Why should these properties cluster in correlation with the canonical head-complement order? Space limitations forbid a broader discussion. Here is a directionality-based account of licensing based on Haider (2010, 2013, 2015, in press). This account covers the compactness properties of head-initial structures as well as the mandatory raising of a subject to the VP-external structural subject position in SVO languages. The theoretical core assumptions are as follows (Haider 2010: 26; 2013: 3f.). (41) a. Projection lines are universally right-branching22 and endocentric. b. A dependent phrase is licensed in the canonical direction. c. The position of a dependent phrase P is licensed =Def. a (projection of the) phrase head h and P minimally and mutually c-command each other. It is the minimal & mutual c-command condition (41c) that is directly causal for compactness, scrambling, and the need of a functional subject position in VO. In order to directionally license YP in (42b), the verb must be reinstantiated. Note that the empty verb position between YP and ZP is a position for a stranded particle. The resulting shell structure is compact (42c), because of the minimality condition (41c). Any intervening item, be it an adjunct or a scrambled object, would destroy the minimality requirement of the licensing relation (41c). 21

Even English displays specimens of nonnominative subjects. The absence of do-support in locative-inversion constructions such as (ii) indicates that the PP ‘into the valley’ occupies the structural position of the XP in (38b).

22

i. Into the valley of death rode the six hundred

The six hundred rode into the valley of death.

ii. Into which valley rode the six hundred?

Into which valley did the six hundred ride?

In other words, the direction of merger within a phrase is universally to the left.

359

360

HUBERT HAIDER

(42) a. . . . [YP [V´ V°→ ZP]]VP b. . . . [Vi → [YP [V´ ei → ZP]]] c. . . . [Vi → [(*π) [YP [(*π) [V´ ei → ZP]]]]] In OV, the situation is different because the canonical directionality of licensing is congruent with the directionality of merger (43). Hence not only the head but any projection node on the projection line can serve as a licensing node. As a consequence, interveners do not matter. This opens room for VP-internal adverbs and for scrambled objects. (43)

a. . . . [. . . [YP ← [V´ ZP ←V°]]] b. . . . [. . . [YP ← [V´ π ← [V´ ZP ← [V´ π V°]]]]]

The hallmark of SVO languages is the VP-external functional subject position. In SVO, the highest argument in the VP is not in the directionality domain of the verbal head. Neither the verb nor a projection node can provide directional licensing. Therefore, a functional head has to do it. The mutual c-command condition (41c) is the trigger for raising the subject to the specposition. The functional head c-commands the position of the trace of the subject, and the raised subject c-commands the functional head, so mutual c-command is effected. The adjacency effect of adjuncts on the left side of a head-initial phrase is directionality-triggered, too. A pre-VP adjunct is not directionally licensed in VO since the canonical directionality of the verbal head in VO is to the right. Pre-NP attributes are not licensed directionally since in any Germanic language, NPs are head initial, that is, N° licenses to the right. In OV Germanic, on the other hand, the pre-head adjuncts of VPs and APs are within the licensing domain of either the verbal head or the adjectival head or its projection. Adjuncts that are not canonically licensed by the head of the host phrase must be “properly attached,” which entails that each node on the projection line of the adjoined phrase is adjacent to the host phrase (see Haider [in press]). The result is an overall adjacency effect. The final property – verb clustering in OV – is an indirect effect of directionality. In OV, embedding by VP-stacking in the canonical direction would result in central embedding, which is strongly disfavored by any parser (44a). Grammars are parser friendly. Clustering opens an alternative way of structuring and reduces the unbounded domain of unwanted left-branching structures to a local domain of clustered verbs (44b). Even in this domain, left-branching structures tend to get reordered into right-branching ones (44c). In VO (44d), the stacked VPs yield a right-branching structure right away, which does not pose the parsing problems the unpredictable number of opening brackets in (44a) would create.

VO-/OV-Base Ordering

(44)

a. * . . . [VP [VP [VP XP V3] V2] V1] b. . . . [VP XP [[V3 V2]V° V1]V°-cluster] (e.g., Frisian, German) c. . . . [VP XP [V1 [V2 V3]V°]V°-cluster] (e.g., Dutch) d. . . . [VP V1 [VP V2 [VP V3 XP]]]

(e.g., English)

In sum, the set of contrasting grammatical properties between OV and VO can be reduced to a single major factor, namely the directionality of the head within a universally right-branching phrasal architecture.

Bibliography A˚ farli, T. A. 1985. “Norwegian verb particle constructions as causative constructions,” Nordic Journal of Linguistics 8: 75–98. Augustinus, L. 2015. Complement Raising and Cluster Formation in Dutch. Doctoral dissertation. KU Leuven. Augustinus, L. and P. Dirix 2013. “The IPP effect in Afrikaans: A corpus analysis,” Proceedings of the 19th Nordic Conference of Computational Linguistics. Linko¨ping Electronic Conference Proceedings 85: 213–225. Bech, G. 1957/1983.2 Studien u¨ber das Deutsche Verbum infinitum. Tu¨bingen: Max Niemeyer Verlag. First edition in: Historisk-filologiske Meddelelser udgivet af Det Kongelige Danske Videnskabernes Selskab 35: 2 (1955) and 36: 6 (1957). Bentzen, K. 2005. “What’s the better move? On verb placement in Standard and Northern Norwegian,” Nordic Journal of Linguistics 28.2: 153–188. Broekhuis, H. 2006. “The universal base hypothesis: VO or OV?” In J. van de Weijer and B. Los (eds.), Linguistics in the Netherlands 2006. Amsterdam: John Benjamins: 28–39. Broekhuis, H. 2008. Derivations and Evaluations: Object Shift in the Germanic Languages. Berlin: Mouton de Gruyter. Broekhuis, H. 2013. Syntax of Dutch: Adpositions and Adpositional Phrases. Amsterdam University Press. Collins, C. and H. Thra´insson 1996. “VP-internal structure and object shift in Icelandic,” Linguistic Inquiry 27.3: 391–344. Dehe´, N. 2004. “On the order of objects in Icelandic double object constructions,” UCL Working Papers in Linguistics 16: 85–108. Dehe´, N. 2015. “Particle verbs in Germanic.” In P. O. Mu¨ller, I. Ohnheiser, S. Olsen, and F.Rainer (eds.), Word Formation. An International Handbook of the Languages of Europe. Berlin: Mouton de Gruyter: 611–626. Diesing, M. 1997. “Yiddish VP order and the typology of object movement in Germanic,” Natural Language and Linguistic Theory 15.2: 369–427.

361

362

HUBERT HAIDER

Donaldson B. C. 1993. A Grammar of Afrikaans. Berlin: Mouton de Gruyter. Engdahl, E., M. Andre´asson, and K. Bo¨rjars 2003. “Word order in the Swedish midfield – an OT approach.” in M. Butt and T. HollowayKing (eds.), Proceedings of the LFG 03 Conference. Stanford: CSLI Publications: 43–58. Geerts, G., W. Haeseryn, J. de Rooij, and M. C. van den Toorn 1984. Algemene Nederlandse Spraakkunst. Groningen: Wolters-Noordhoff. Haeseryn, W., K. Romijn, G. Geerts, J. de Rooij, and M. van den Toorn. 1997. Algemene Nederlandse Spraakkunst2. Groningen and Deurne: Martinus Nijhoff and Wolters Plantyn. Haider, H. 2010. The Syntax of German. Cambridge University Press. Haider, H. 2013. Symmetry Breaking in Syntax. Cambridge University Press. Haider, H. 2014. “The VO-OV split of Germanic languages – a T3 & V2 production,” Interdisciplinary Journal for Germanic Linguistics and Semiotic Analysis 19.1: 57–79. Haider, H. 2015. “Head directionality – in syntax and morphology.” In A. Fa´bregas, J. Mateu, and M. T. Putnam (eds.), Contemporary Linguistic Parameter. London: Bloomsbury Academic: 73–97. Haider, H. 2019. “On absent, expletive and non-referential subjects.” In A. C. Wolfsgruber, B. Po¨ll, and P. Herbeck (eds.), Semantic and Syntactic Aspects of Impersonality. Linguistische Berichte (Sonderheft 26). Hamburg: Helmut Buske Verlag: 11–46. Haider, H. (in press). “The Left-Left Constraint – a structural constraint on adjuncts.” In U. Freywald and S. Horst (eds.), Headedness and / or Grammatical Anarchy? Berlin: Language Science Press. Hauge, H. 2003. “Towards a unified representation of English and Norwegian auxiliaries,” Nordic Journal of English Studies 2: 53–74. Hoekstra, E. 1998. “Analysing linear asymmetries in the verb clusters of Dutch and Frisian and their dialects.” In D. Beerman, D. LeBlanc, and H. C. van Riemsdijk (eds.), Rightward Movement. Amsterdam: John Benjamins: 153–169. Hoekstra, E. 2016. Frisian syntax – control verbs. Taalportaal. The Linguistics of Dutch, Frisian, and Afrikaans online: www.taalportaal.org/taalpor taal/topic/pid/topic-14127700752502220. Hoekstra, T. and R. Mulder 1990. “Unergatives as copular verbs: Locational and existential predication,” The Linguistic Review 7: 1–79. Kerstens, J. G. 1975. Over afgeleide structuur en de interpretatie van zinnen. Doctoral dissertation, University of Amsterdam. King, S. 1989. The dark half. New York, NY: Viking. Koeneman, O. 2000. The Flexible Nature of Verb Movement. Doctoral dissertation, University of Utrecht. Lundquist, B. 2014a. “The verb phrase: Argument structure and particle placement,” The Nordic Atlas of Language Structures Journal 1: 107–109.

VO-/OV-Base Ordering

Lundquist, B. 2014b. “Double object constructions: Active verbs,” The Nordic Atlas of Language Structures Journal 1: 136–145. Kroch, A. and B. Santorini 2016. “Evidence for OV word order in Older French, Icelandic, and Yiddish.” Presentation, University of Pennsylvania. www.ling.upenn.edu/~kroch/handouts/fwav3.pdf (Feb. 3, 2017). Neeleman, A. 1994a. “Scrambling as a D-structure phenomenon.” In N. Corver and H. C. van Riemsdijk (eds.), Studies on Scrambling: Movement and Non-movement Approaches to Free Word-Order Phenomena. Berlin: Mouton de Gruyter: 387–430. Neeleman, A. 1994b. Complex Predicates. Ph.D. dissertation, University of Utrecht. Neeleman, A. 2002. “Particle placement.” In N. Dehe´, R. Jackendoff, A. McIntyre, and U. Silke (eds.), Verb-Particle Explorations. Berlin: Mouton de Gruyter: 141–164. Neeleman, A. and H. van de Koot 2008. “Dutch scrambling and the nature of discourse templates,” Journal of Comparative Germanic Linguistics 11.2: 137–189. Neeleman, A. and F. Weerman 1993. “The balance between syntax and morphology: Dutch particles and resultatives,” Natural Language & Linguistic Theory 11: 433–475. Neeleman, A. and F. Weerman 1999. Flexible Syntax. Dordrecht: Kluwer. Quirk, R., S. Greenbaum, G. N. Leech, and J. Svartvik 1985.4 A Comprehensive Grammar of the English Language. London: Longman. Ro¨gnvaldsson, E. 1990. Um orðaro¨ð og færslur ´ı ı´slensku. Institute of Linguistics, University of Iceland, Reykjavı´k. [1982 M.A. thesis, University of Iceland, Reykjavı´k] Svenonius, P. 1996. “The optionality of particle shift,” Working Papers in Scandinavian Syntax 57: 47–75. Thra´insson, H. 2007. The Syntax of Icelandic. Cambridge University Press. Thraı´nsson, H. 2010. “Predictable and unpredictable sources of variable verb and adverb placement in Scandinavian,” Lingua 120: 1062–1088. Vikner, S. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford University Press. Vikner, S. 2007. “Object Shift.” In M. Everaert and H. C. van Riemsdijk (eds.), The Blackwell Companion to Syntax. Oxford: Blackwell: 392–436. Vikner, S. (2017). “Germanic verb particle variation.” In E. Aboh, E. Haeberli, G. Puska´s, and M. Scho¨nenberger (eds.), Elements of Comparative Syntax – Theory and Description. Berlin: Mouton de Gruyter: 371–398. Whelpton, M. 2007. “Building resultatives in Icelandic.” In E. Bainbridge and B. Agbayani (eds.), Proceedings of the 34th

363

364

HUBERT HAIDER

Western Conference on Linguistics 17. Fresno, CA: California State University: 478–486. Wurmbrand, S. (2017). “Verb clusters, verb raising, and restructuring.” In M. Everaert and H. C. van Riemsdijk (eds.), The Wiley Blackwell Companion to Syntax, 2nd edn. Oxford: Blackwell.

Chapter 16 The Placement of Finite Verbs Sten Vikner

16.1 Introduction Language abbreviations Af. Afrikaans Be. Swiss German from Bern Da. Danish Du. Dutch En. English Fa. Faroese Fs. Frisian Ge. Standard German Hd. Hallingdalen

Ic. ME. ¨ d. O SG. St. WF. Yi. Zu¨.

Icelandic Middle English ¨ vdalian (A¨lvdalen) O Swiss German from Sankt Gallen Swabian German from Stuttgart West Flemish Yiddish Swiss German from Zu¨rich

In a clause in a Germanic language, there are three different positions in which the finite verb may occur: (1) a. The position immediately before the (this position will be subject called C°) b. The position immediately after the (this position will be subject called T°) c. The base position next to, e.g., the (this position will be object called V°) There is a choice associated with each of these positions, and this chapter will show how the exact position of the finite verb in

For helpful comments and suggestions, I am grateful to Theresa Biberauer, Ken Ramshøj Christensen, Jamie Douglas, Eva Engels, Constantin Freitag, Hans-Martin Gärtner, Hubert Haider, Johannes Kizach, Anne Mette Nyvad, Richard B. Page, Erin Pretorius, Michael T. Putnam and Johanna Wood as well as to audiences at the universities of Aarhus, Cambridge, Konstanz, and Lund. The work presented here was partly supported by Forskningsrådet for Kultur og Kommunikation (Danish Research Council for Culture and Communication).

366

STEN VIKNER

a particular type of clause in a given Germanic language depends on these three choices. The first choice is whether or not the finite verb occurs in the position called C° (i.e., in the position immediately before the subject). This choice can be seen as one between having what is called V2 (which will be analyzed below as involving V°-to-T°-to-C° movement), as in (2a), where the finite verb is the second constituent, or not having V2, as in (2b): C° Subject V° (2) a. Danish Den mulighed tænkte vi desværre aldrig paº . b. English That possibility we unfortunately never thought of. This first choice is only made once for each Germanic language, and it holds for all finite verbs in all main clauses (and in some embedded ones). The second choice is whether or not the finite verb occurs in the position immediately after the subject (i.e., in the position called T°). This choice can be seen as one between having V°-to-T° movement, as in (3a), or not having V°-to-T° movement, as in (3b): Subject T° (3)

a. Icelandic Hu´n spurði b. Danish

hvers vegna við

Hun spurgte hvorfor

vi

She

we

asked

why

V° til I´slands.

flyttum ekki

ikke flyttede til Island. (moved) not

(moved) to Iceland

This second choice is also only made once for each Germanic language, and it holds for all finite verbs in all clauses (even if its effect can only be observed when V2 does not apply). The third and last choice is the one also discussed in Chapter 15, namely whether the base order is VO or OV, i.e., whether the verb (when it is in V°) comes before its complement, as in (4a), or after it, as in (4b): Verb (4) a. English Many linguists

Object

who already know this

b. German Viele Linguisten, die schon

book

find

it useful.

dieses Buch kennen, finden es nu¨tzlich. Object

Verb

This third choice, between VO or OV, (4), is also only made once for each Germanic language, and it holds for all verbs in all clauses (even if its effect can only be observed for verbs which have not undergone movement either to C° [V2] or to T°). These three binary choices can maximally result in eight different types of Germanic languages, but not all of these types are actually attested:

The Placement of Finite Verbs

finite verb finite in C ° ( V 2 ) verb in T °

(5) Languages

¨ vdalian + a. Icelandic, O (A¨lvdalen) b. Danish, Faroese, Hallingdalen, Norwegian, Swedish c. Yiddish

d. Afrikaans, Dutch, Frisian, German, Swabian, Swiss German, West Flemish e. Middle English f. English g. h. -

+

VO/OV

VO

+



VO

+

+

OV

+



OV

– – – –

+ – + –

VO VO OV OV

g

SCANDINAVIAN

g

CONTINENTAL

(=

NORTH

GERMANIC)

WEST GERMANIC

Only one of the Germanic languages spoken today is not V2, namely English, (5f). In order to maximize the number of non-V2-languages in the table, I have included a language no longer spoken, namely (late) Middle English, (5e), see Fischer et al. (2001: 132). Even so, there are still two possible types of non-V2-languages not attested among the Germanic languages, (5g, 5h). To give an idea of the (simplified) analysis behind the use of the labels C°, T° and V°, here is what I take to be the structure of a clause (irrespective of whether it is a main or an embedded clause): (6)

A clause is a CP, the complement of the CP’s head (= C°) is a TP, and the complement of the TP’s head (= T°) is a VP.

For a clause in a VO-language with no auxiliary verbs and with a monotransitive main verb, the structure looks as follows: CP

(7)

topic

C° subj

TP T°

VP

adv



DP

367

368

STEN VIKNER

As will be illustrated below, the subject occurs at the left edge of TP and if there is a sentence-medial adverbial, it occurs at the left edge of VP. Furthermore, if the language in question had been OV, the sequence between the verb and its complement would be reversed.

16.2 Verb Second (V2) 16.2.1 V2 in All Main Clauses In most Germanic languages, not including Middle English and Modern English, all main clauses are V2. This means that the finite verb occupies the second position in the clause, irrespective of which constituent occupies the first position: (8)

Verb second = V2 1 one constituent



2 the finite verb

3 the rest of the clause



It might appear that also in English, the finite verb occupies the second position: (9) a. Da. Peter b. Ic.

har

sandsynligvis

Pe´tur hefur sennilega

c. Ge. Peter

hat

læst den her bog. lesið þessa

bo´k.

wahrscheinlich

dieses

d. Af. Pieter het

waarskynlik

hierdie boek gelees.

e. En. Peter

probably

has

read this

Buch gelesen. book.

This is an illusion, though, and for declarative examples like (9), it only holds when we consider subject-initial main clauses with a finite auxiliary verb (see the discussion in Section 16.3 below, especially footnote 4). If we look at declarative clauses that have, e.g., an initial object, (10), or an initial adverbial, (11), we see that English clearly differs from the V2-languages. However, there are two sets of very specific circumstances, where also English has V2, namely after an initial wh-constituent, (12), or after an initial negative constituent, (17b) further below: C° (10) a. Da.

Den her

bog

har

b. Ic.

Þessa

bo´k

hefur Pe´tur lesið.

c. Ge.

Dieses

Buch hat

Peter

gelesen.

d. Af.

Hierdie

boek het

Pieter

gelees.

book has

Peter read.

e. En. * This

Peter læst.

The Placement of Finite Verbs

(11) a. Da.

Nu

har

b. Ic.

Nu´

hefur Pe´tur lesið þessa

bo´k.

c. Ge.

Jetzt

hat

Peter

dieses

Buch gelesen.

d. Af.

Nou

het

Pieter

hierdie boek gelees.

e. En. * Now

has

Peter read this

har

Peter læst?

(12) a. Da.

Hvad for en bog

Peter læst den her bog.

book.

b. Ic.

Hvaða

bo´k

hefur Pe´tur lesið?

c. Ge.

Welches

Buch hat

Peter

gelesen?

d. Af.

Watter

boek het

Pieter

gelees?

e. En.

Which

book has

Peter read?

The discussion of V2 goes back to at least Wackernagel (1892) and Fourquet (1938). A common analysis of V2 which goes back to den Besten (1977, published as 1983) and Thiersch (1978) and found its canonical form in Platzack (1985) and Chomsky (1986: 6) is that the finite verb in V2-main clauses occupies the same position that the subordinating conjunction (also called the complementizer, e.g., that, if, because) occupies in an embedded clause. This position is called C°: (13) Da. a. . . . b. Denne film (14) Ic.

a. . . . b. Þessa mynd

C° at har

børnene børnene

har ___

set set

den her film. __________.

að hafa

bo¨rnin bo¨rnin

hafa se´ð ___ se´ð

þessa mynd. _________.

(15) Ge. a. . . . b. Diesen Film gesehen gesehen

dass die Kinder haben die Kinder haben. _____.

diesen Film _________

(16) Af. a. . . . b. Hierdie film gesien gesien

dat het het. _____.

hierdie film _________

(17) En. a. . . . that b. None of these films have

die kinders die kinders

the children have seen this film. the children ___ seen _________.

and here is the tree structure for such V2-clauses (for German and other OV-languages, the order inside the two VPs is reversed with V° being rightmost):

369

370

STEN VIKNER

(18)

CP

XP

C'

TOPIC



V2 = V°-TO-T°-TO-C° MOVEMENT

TP

FINITE VERB

DP

T'

SUBJECT



VP



VP V° NONFINITE VERB

DP OBJECT

The finite verb moves to C°, and some other constituent (e.g., the topic) moves into the specifier position of CP (CP-spec). This constituent in the first position can be, e.g., the subject, the object, an adverbial, or an embedded clause. If the first constituent is not the subject, then the subject has to occur in the third position (i.e., the DP that is the specifier of TP). If a clause does not have V2, then either the finite verb does not move at all (i.e., it stays in V°) or it moves to T° and stays there, see Section 16.3 below. Supporting evidence for the assumption that the finite verb (in a V2main clause) occupies the same position that the complementizer occupies (in an embedded clause) may be found in conditional clauses, where the subject is preceded either by a complementizer (e.g., if) or by the finite verb (e.g., had), but not by both, see den Besten (1983: 117): C° (19) Da. a.

Hvis

jeg havde

haft mere tid,

...

Ic. b.

Ef

e´g hefði

haft meiri tı´ma,

...

Ge. c.

Wenn

ich

Af. d.

As

ek

En. e.

If

I

Havde

(20) Da. a.

mehr Zeit gehabt ha¨tte, . . . meer tyd had

gehad het,

had more time,

...

jeg ____

haft mere tid,

...

haft meiri tı´ma,

...

Ic. b.

Hefði

e´g ____

Ge. c.

Ha¨tte

ich

Af. d.

Het

ek

En. e.

Had

I ____

mehr Zeit gehabt ____, . . . meer tyd had more time,

gehad ____, ...

The Placement of Finite Verbs

(21) Da. a.

*

Havde hvis jeg ____ haft mere tid,

Ic. b.

*

Hefði ef

Ge. c.

*

Ha¨tte wenn ich

mehr Zeit gehabt ____, . . .

Af. d.

*

Het as

ek

meer tyd

I ____

...

e´g ____ haft meiri tı´ma,

... gehad ____,

En. e.

*

Had if

(22) Da. a.

*

Hvis havde jeg ____ haft mere tid,

Ic. b.

*

Ef hefði

Ge. c.

*

Wenn ha¨tte ich

mehr Zeit gehabt ____, . . .

Af. d.

*

As het

meer tyd

En. e.

*

If had

Da. . . . ville

jeg

Ic. . . . myndi e´g ek

En. . . .

I

... ...

e´g ____ haft meiri tı´ma, ek I ____

... gehad ____,

had more time,

...

have lavet et

endnu længere hand-out.

hafa gert

ennþa´ lengri

Ge. . . . ha¨tte ich Af. . . . sou

had more time,

u´thendu.

ein noch la¨ngeres Thesenpapier gemacht. ‘n nog

langer

uitdeelstuk

would have made an even longer

gemaak het.

hand-out.

Consider finally the following examples from two V2-languages, Danish and German: CP-spec



TP-spec

AdvP





jeg

desværre

burde

spise

jeg

desværre

c. * Derfor

burde spise jeg

desværre

d. * Derfor

spise jeg

desværre

(23) Da. a. * Derfor b.

Derfor

burde

Therefore

(ought) (eat)

I

CP-spec



TP-spec

(24) Ge. a. * Deswegen b.

ich

spise

mindre chokolade. mindre chokolade. mindre chokolade

burde

unfortunately (ought) AdvP leider

DP

mindre chokolade. (eat)

less

DP

chocolate V°

weniger Schokolade essen

Deswegen sollte

ich

leider

weniger Schokolade essen.

c. * Deswegen sollte

essen ich

leider

weniger Schokolade

d. * Deswegen

essen ich

leider

weniger Schokolade

Therefore

(should) (eat)

I

unfortunately less

chocolate

V° sollte.

sollte. (eat)

(should)

(23a, 23b) and (24a, 24b) again show that Danish and German are V2languages, which is why the finite verb cannot occur to the right of the subject. Examples (23c) and (24c) show that only one verb may undergo V2. Examples (23d) and (24d) show that only a finite verb may undergo V2 (and not the infinitive spise / essen ‘eat’). For a less simplified analysis of V2 than the one presented here, see, e.g., Holmberg (2015 in press).

16.2.2 V2 in English Main Clauses As already mentioned, it is assumed that English is not a V2-language, but nevertheless, there are two sets of very specific circumstances, where even English has V2.

371

372

STEN VIKNER

One V2-context is (nonsubject-initial) interrogative main clauses (i.e., direct questions): (25) a. En. Which book

C° has

Peter ___ read _____

____?

Peter has read _____

____?

Peter ___ læst _____

____?

hefur Pe´tur ___ lesið _____

____?

b. En. * Which book c. Da. Hvad for en bog har d. Ic.

Hvaða bo´k

e. Ge. Welches Buch hat

Peter

_____

____ gelesen ___?

g. Af. Watter boek

het

Pieter

_____

____ gelees ___?

has

Peter ___ read this

book?

Peter has read this

book?

(26) a. En. Why b. En. * Why c. Da. Hvorfor

har

d. Ic.

hefur Pe´tur ___ lesið þessa

Af hverju

Peter ___ læst den her bog? bo´k?

e. Ge. Warum

hat

Peter

dieses Buch gelesen ___?

g. Af. Waarom

het

Pieter

hierdie boek gelees ___?

The other English V2-context is when there is an initial negative constituent: C° (27) a. En. Never have the children ____ seen such a

bad

film.

b. En. * Never

the children have seen such a

bad

film.

c. Da. Aldrig har

børnene

____ set saº dan en

daº rlig

film.

d. Ic.

bo¨rnin

____ se´ð svona

slæma

mynd.

Aldrei hafa

e. Ge. Nie

haben die Kinder

g. Af. Nooit het

die kinders

(28) a. En. Only in America b. En. * Only in America c. Da. Kun i Amerika d. Ic. Aðeins ı´ Bandarı´kjunum e. Ge. Nur in Amerika g. Af. Net in Amerika

so

einen schlechten Film

gesehen ___.

so

‘n

gesien

slegte

film

___ nie.

C° could such a thing _____ happen. such a thing could happen. kunne saº dan noget _____ ske. gæti eitthvað ___ gerst. svona ko¨nnte so etwas passieren ____. kon so ‘n ding gebeur ____.

Notice that in the above examples, the finite verb is an auxiliary (see the discussion in Section 16.3 below, especially footnote 4). When the finite verb is a main verb, do-insertion is necessary: C° (29) En. a. Which book did Peter ___ read _____ ____ ? b. * Which book read Peter ____ _____ ____ ? c. * Which book Peter read _____ ____ ?

The Placement of Finite Verbs

C° (30) En. a. Never did the children ____ see such a bad film. b. * Never saw the children ___ such a bad film. c. * Never the children saw such a bad film. As seen in (9) and (31a), subject-initial main clauses in English may look as if they are V2, but if an element is inserted to the left of the subject, the verb does not insist on ocurring in second position, (31b) (see also (2) above): (31) En. a.

The city council has unfortunately never considered this possibility.

b. Unfortunately, the city council has

never considered this possibility.

Furthermore, at least for those cases where the finite verb is a main verb, it can be shown that subject-initial main clauses are not V2, because the finite verb has to follow any sentence-medial adverbials, and because doinsertion is not necessary: V° (32) En. a.

The

city council never considered this possibility.

b.

Which city council never considered this possibility?

c.

No

city council ever

d. Unfortunately, the

considered this possibility.

city council never considered this possibility.

It would thus seem that in English, V2 is restricted to interrogative main clauses and to clauses with negative preposing, provided in both cases that the initial element is not the subject. No other main clauses in English are V2,1 see also (10e) and (11e).

16.3 V°-to-T° Movement Let us now consider what happens in situations where V2 cannot apply, i.e., where the finite verb cannot occur in the position immediately left of the subject. This leaves two options, namely that the finite verb occurs either in the position immediately right of the subject (i.e., in the position called T°), as in (33), or in its base position (i.e., the one called V°), as in (34). C°

TP-spec T°

(33) a. En. * That John

1

eats

AdvP V°

DP

often

tomatoes (surprises most people.)

b. Da. * At

Johan

c. Fa. * At

Jo´n

etur

tomatir

(kemur o´vart a´ tey flestu.)

d. Ic.



Jo´n

borðar oft

to´mata

(kemur flestum a´ o´vart.)

e. Yi.

Az

Jonas

est

pomidorn (iz a khidesh far alemen.)

spiser ofte ofta oft

tomater

(overrasker de fleste.)

What Huddleston and Pullum (2002: 97) call subject-auxiliary inversion are thus cases of V2, whereas what Huddleston and Pullum (2002: 97, 1385) call subject postposing or subject-dependent inversion (including locative inversion and quotative inversion) are not cases of V2, i.e, that the former but not the latter allows the subject to occur between a finite auxiliary and a nonfinite main verb.

373

374

STEN VIKNER



AdvP V°

DP

That John

often eats

tomatoes (surprises most people.)

b. Da.

At

Johan

ofte

spiser tomater

(overrasker de fleste.)

c. Fa.

At

Jo´n

ofta

etur

(kemur o´vart a´ tey flestu.)

d. Ic. * Að

Jo´n

oft

borðar to´mata

e. Yi. * Az

Jonas

oft

est

(34) a. En.

TP-spec T°

tomatir

(kemur flestum a´ o´vart.)

pomidorn (iz a khidesh far alemen.)

A common analysis of this difference (which goes back to Emonds 1978 and Pollock 1989) is that two languages have V°-to-T° movement, namely Icelandic and Yiddish, as opposed to English, Danish, and Faroese. In Icelandic and Yiddish, the finite verb is therefore taken to always move from its position in V° to a position further left, namely T°. This movement can only be detected if something occurs to the right of T° but to the left of V°, in this case the medial adverbial often: CP

(35) XP

C'



TP

SUB. CONJ.

DP

V°-TO-T° MOVEMENT

T'

SUBJECT



VP

FINITE VERB

AdvP MEDIAL ADVERBIAL

VP



DP OBJECT

The following examples from Middle English and from two conservative Mainland Scandinavian dialects display the same difference. In embedded clauses, the finite verb precedes the medial adverbial or ¨ vdalian (from negation in Middle English and in the Swedish dialect O ¨Alvdalen, see, e.g., Garbacz 2010), whereas the finite verb follows the medial adverbial or negation in the Norwegian dialect from Hallingdalen: T° AdvP V° (36) a. ME. . . . and he swore that he talkyd neuer wyth no man. . . . and he swore that he talked never with no man

The Placement of Finite Verbs

(This example is from 1460, William Paston I, Letter to John Paston I, May 2, 1460, Davis 1971: 164) T° AdvP V° ¨ b. Od. Ba fo dye at ig uild int fy om Just because that I would not follow him (This example is from Levander 1909: 123, see also Platzack and Holmberg 1989: 70) T° AdvP V° c. Hd.

. . . fisk, jammvært om støræls’n paº o

ikki

va myky skrytæ taº

. . . fish, although

not

was much brag

size-the

of them

about

(This example is from Venaº s [1977: 243], see also Trosterud 1989: 91 and Platzack and Holmberg 1989: 70) Among the Germanic VO-languages, the ones without V°-to-T° movement are modern English2 and five of the seven Scandinavian variants: Danish, Faroese, Hallingdalen Norwegian, Norwegian, and Swedish, see (34) above. Among the Germanic OV-languages, only one language, Yiddish,3 (34e) above, seems to clearly have V°-to-T° movement. The discussion of the other Germanic OV-languages will therefore have to wait until Section 16.6 below. Consider finally the following examples from the two languages with V°to-T° movement, Icelandic and Yiddish: C° (37) Ic. a. * Að

TPsp



Jo´n

AdvP V°



DP

oft

hafi borðað to´mata

oft

borðað to´mata

(kemur flestum a´ o´vart.)

b.



Jo´n

hafi

(kemur flestum a´ o´vart.)

c. * Að

Jo´n

hafi borðað oft

d. * Að

Jo´n

borðað oft

to´mata

(kemur flestum a´ o´vart.)

hafi

to´mata

(kemur flestum a´ o´vart.)

2

For English, it must be added that only finite main verbs stay in V°, whereas finite auxiliary verbs seem to occur in T°, cf.

3

As opposed to all the other Germanic languages, it is not immediately obvious whether Yiddish is a VO- or OV-language.

the difference between (31) and (32) above. The reason is that both VO- and OV-orders are possible, and that it can be independently shown that Yiddish has both scrambling (which moves objects leftward) and extraposition (which moves objects rightward). Other properties therefore have to be taken into account, and a number of such other properties would seem to receive a more satisfactory analysis if Yiddish is not VO, but OV. These include the behavior of verb particles, the lack of agreement on predicative adjectives, the variation in verb sequences, and coordinated VPs, cf. Hall (1979), Geilfuß (1991), Vikner (2001a, 2001b, 2003), Haider (2005), and Wallenberg (2009, 2013). Another option might be that Yiddish (as the only Germanic language) is indeterminate with respect to the VO / OVchoice, as suggested in Chapter 16 above and in Haider (2010:7, 2013:102). Finally, a number of analyses have taken Yiddish to be VO, e.g., den Besten and Moed-van Walraven (1986: 113), Diesing (1997: 388), Vikner (1995, 1997), Sadock (1998).

375

376

STEN VIKNER

(38) Yi. a. * Az

Jonas

oft

hot gegesn pomidorn (iz a khidesh far

Jonas hot

oft

gegesn pomidorn (iz a khidesh far

alemen.) b.

Az

alemen.) c.

??

Az

Jonas hot gegesn oft

pomidorn (iz a khidesh far alemen.)

d. * Az

Jonas

gegesn oft

hot

pomidorn (iz a khidesh far alemen.)

That John

(has) (eaten) often

(has) (eaten) tomatoes

(surprises most people.)

Examples (37a, 37b) and (38a, 38b) again show that Icelandic and Yiddish have V°-to-T° movement, and this is why the finite verb cannot remain in V°. Examples (37c) and (38c) show that only one verb may undergo V°-to-T° movement. Examples (37d) and (38d) show that only a finite verb may undergo V°-to-T° movement. For a less simplified analysis of V°-to-T° movement than the one presented here, see, e.g., Bobaljik (2003) and Koeneman and Zeijlstra (2014).

16.4 Differences Between V°-to-T° Movement and V2 There are two main differences between V°-to-T° movement and V2: CP

(39) XP

C' V2 = V°-TO-T°-TO-C° MOVEMENT C°

TP

V°-TO-T° MOVEMENT T'

DP SUBJECT



VP AdvP

VP

ADVERBIAL



DP OBJECT

V°-to-T° movement applies to all finite verbs, whereas V2 only applies to finite verbs in main clauses (and some embedded clauses). In other words, although V2 is not completely restricted to main clauses, it is only possible in a subset of finite embedded clauses, whereas V°-to-T° movement is obligatory for all finite verbs. In a clause with V°-to-T° movement but without V2, the first element is the subject and the second element is the finite verb. In a clause with V2,

The Placement of Finite Verbs

the second element is also the finite verb, but the first element can be any constituent (subject, object, adverbial, embedded clause, . . .). The reason why the embedded clauses in (33)–(34) above are subject clauses is that this is a context where main clause word order (i.e., V2) is NOT allowed, see (40), and also (33b, 33c). This is relevant because there are also many embedded contexts where both main, (41) and (42), and embedded clause word orders, (43), are possible: C° (40) a. Da. *





(At) tomater spiser Johan ofte

b. Fa. *

(At) tomatir etur

Jo´n

ofta

kemur o´vart a´ tey flestu.

That tomatoes eats

John

often

surprises most people

(41) a. Da.

Hun siger at

tomater spiser Johan ofte.

b. Fa.

Hon sigur at

tomatir etur

Jo´n

ofta.

that tomatoes eats

John

often

She

says

overrasker de fleste.

(42) a. Da.

Hun siger at

Johan

spiser

ofte

tomater.

b. Fa.

Hon sigur at

Jo´n

etur

ofta

tomatir.

that John

eats

often

tomatoes

She (43) a. Da. b. Fa.

says

Hun siger at

Johan ofte spiser tomater.

Hon sigur at

Jo´n

ofta etur

tomatir.

She

John

often eats

tomatoes

says

that

Provided the special conditions for V2 in English are observed, the judgments are very similar here: C° (44) En. a. * b. (45) En.



That in no way could we That

be held responsible must now be clear.

we could in no way be held responsible must now be clear.

The judge emphasised . . . a.

. . . that in no way could we

b.

. . . that

be held responsible.

we could in no way be held responsible.

In, e.g., German, embedded V2 is only possible if the subordinating conjunction is left out. This is not the case in Scandinavian, Yiddish, and English, where the subordinating conjunction has to precede the embedded V2-clause, cf. (41), (42) and (45a). In other words, the complementary distibution shown in (19)–(22) does not hold here, and we therefore need two C°-positions (sometimes called CP-recursion), one for the conjunction and one for the finite verb (see e.g., Julien [2015] or Nyvad et al. [2017] and references there for more detailed analyses).4 4

The idea is thus that (41), (42) and (45a) must have two CPs whereas the standard cases as discussed in Section 16.3 above only need one CP. Walkden (2017) suggests a different way of having two CPs instead of one. Taking his point of departure in cartographic analyses of the left edge of the clause, in particular Frascarelli and Hinterhölzl (2007), he suggests that one-CP analysis of the standard cases should be seen as a syncretion of the six (or more) projections in the left periphery (viz. ForceP, ShiftP, ContrP, FocP, FamP, and FinP) into one single CP. Based on data with not one but two phrases preceding the finite verb from modern urban vernacular Danish, German, Norwegian, and Swedish (as

377

378

STEN VIKNER

According to Vikner (2001b: 226), three conditions seem to be necessary for embedded V2 to be possible5 (whereas the non-V2-options are always possible, even when these three conditions are not observed, as shown below): (46)

a. An embedded V2-clause requires certain matrix verbs (verbs of saying and believing). b. An embedded V2-clause requires the matrix verb not to be negated. c. An embedded V2-clause has to occur in object position.

16.5 Deriving V°-to-T° Movement Both V2 and VO/OV seem to ‘run in the family’, i.e., these features are often shared by closely related languages, cf. (5) above. However, this is clearly not the case with V°-to-T° movement, since Icelandic (which has V°-to-T° movement) is much more closely related to the other Scandinavian languages (almost all of which do not have V°-to-T° movement) than it is related to Yiddish or French (both of which have V°-to-T° movement). Chomsky (1995: 222) says about the ability of constituents to move in the syntax: “Minimalist assumptions suggest that this property should be reduced to morphology-driven movement.” This is the objective of quite a number of accounts, including the one in Vikner (1997, 1999), where finite verb movement is linked to verbal inflectional morphology in the following way: (47)

An SVO-language has V°-to-T° movement if and only if person morphology is found in all tenses. (Vikner 1997: 207, example (23))

There are many alternatives to this particular implementation of a link between V°-to-T° movement and a rich verbal inflectional system, e.g., Bobaljik and Thra´insson (1998), Rohrbacher (1999), Thraı´nsson (2009), Biberauer and Roberts (2010), Koeneman and Zeijlstra (2014). These suggestions (since Bobaljik [2003] commonly subsumed under the label RAH, also discussed in Section 30.3.1 of Chapter 30 below), Walkden (2017) goes on to suggest these are cases with a different structure, where the first four projections (ForceP, ShiftP, ContrP, and FocP) are syncretized into one CP-like projection, and the last two (FamP and FinP) are syncretized into a different CP-like projection. Thus clear predictions are made as to which types of phrases can be in either the first or the second preverbal position in these modern urban vernaculars. 5

Even though the following three contexts do not involve two CPs, the conditions in (46) also seem to hold for (i) a. embedded V2 in German b. embedded non-V2-topicalization in English, and c. optional that / at as subordinating conjunction in English and Danish.

The Placement of Finite Verbs

i.e., Rich Agreement Hypotheses) are all based on the fact that all the Germanic VO-languages without V°-to-T° movement (i.e., Danish, English, Faroese, Hallingdalen, Norwegian, and Swedish) have a much poorer verbal inflectional system than the Germanic (and Romance) languages that have V°-to-T° movement, e.g., Icelandic and Yiddish (and French). Furthermore, these six languages also have in common both that they have a relatively poor verbal inflectional system, which was much richer not that long ago, and that they used to have V°-to-T° movement which they only lost relatively recently. In most of them, this change took place between 1450 and 1700, whereas in Faroese, it is much more recent (see Heycock et al. 2012). The suggested link between V°-to-T° movement and a rich verbal inflectional system has also received a large amount of criticism, e.g., Sprouse (1998), Alexiadou and Fanselow (2002), Bobaljik (2003), Bentzen et al. (2007), Hrafnbjargarson and Wiklund (2010), Holmberg (2010), Anganty´sson (2011), Heycock and Wallenberg (2013), Harbour (2016), Heycock and Sundquist (2017). Some of these criticisms interestingly suggest alternative derivations which then lead to sets of predictions different from the predictions derived by the RAH, even if they frequently end up including revised versions of the RAH. An example of this is Heycock and Wallenberg (2013), which suggests a link with the availability of embedded V2, and given that embedded V2 is and was possible in all Scandinavian languages including Icelandic, which has not lost V°-to-T° movement, Heycock and Wallenberg (2013: 151–154) have to include a version of the RAH in their analysis. Other criticisms do not link the difference to any other properties of the languages in question, and so no new predictions can be derived. This is of course not a completely unknown situation, in fact, it is rather like the situation concerning V2, where it is difficult to see which properties of Danish and English could be directly linked to the former, but not the latter, being V2. Given that there are a great many accounts available as to which properties of Icelandic and Danish could be directly linked to the former, but not the latter, still having V°-to-T° movement, such accounts still merit serious consideration.

16.6 V°-to-T° Movement and the OV-Languages So far all Germanic OV-languages except Yiddish have been left out of the discussion of V°-to-T° movement. Some formulations of the RAH, including (47), explicitly only cover the VO-languages, whereas other formulations, e.g., the one in Koeneman and Zeijlstra (2014), apply to VO- and OV-languages alike.

379

380

STEN VIKNER

If any of the above mentioned versions of the RAH, including (47), would apply to the nine Germanic OV-languages / dialects in (48) below, then we would expect only Dutch and Afrikaans not to have V°-to-T° movement, whereas West Flemish, Frisian, German, Swabian German from Stuttgart, and the Swiss German variants from Sankt Gallen, Zu¨rich, and Bern should all have V°-to-T° movement. However, in all of these languages, the finite verb does not precede the sentential adverb in those embedded clauses where main clause word order is not possible. In fact, the finite verb does not even precede its own object in any of these cases: Adv (48) a. Du. Dat Johan

Object Verb

vaak

tomaten eet

(verrast de meeste mensen.)

b. Af. Dat Johan

gereeld tamaties eet

(verras die meeste mense.)

c. WF. Da

dikkerst tematen eet

(verwondert de meeste mensen.)

d. Fs. Dat Johan

faak

tomaten yt

(die de measte minsken nij.)

e. Ge. Dass Johann

oft

Tomaten isst

(u¨berrascht die meisten Leute.)

f. St. Dass dr Johann oft

Tomada isst

Johan

(ieberrascht der maschde Leid.)

g. SG. Dass de Johann o¨pedie Tomaate a¨st

(u¨berascht di meischte Lu¨t.)

h. Zu¨. Dass de Johann hu¨u¨fig

(u¨berrascht di ma¨ischte Lu¨u¨t.)

i. Be. Dass dr Johann hu¨u¨fig That John

often

Tomaten isst Tomaten isst

(u¨berrascht di meischte Lu¨t.)

tomatoes eats

(surprises most people)

Assuming that all of these languages are OV (see also Chapter 15 above), there are still two open questions, namely whether T° precedes VP, as in (49), or follows it,6 as in (50), and whether there is V°-to-T° movement, as in (49) and (50) with the arrows, or not, as in (49) and (50) without the arrows: (49)

CP XP

C' C°

TP

SUB. CONJ.

V°-TO-T° MOVEMENT OV-LANGUAGE WITH MEDIAL T°

IN AN

DP

T'

SUBJECT



VP

(FINITE VERB)

AdvP MEDIAL ADVERBIAL

6

VP DP



OBJECT

(FINITE VERB)

As opposed to what has been assumed so far, this part of the discussion introduces a further complication, namely that OV-languages may not only be V°-final, but also T°-final.

The Placement of Finite Verbs

(50)

CP XP

C' C°

TP

SUB. CONJ.

V°-TO-T° MOVEMENT IN AN OV-LANGUAGE WITH FINAL T°

DP

T'

SUBJECT



VP

(FINITE VERB)

AdvP

VP

MEDIAL ADVERBIAL

DP



OBJECT

(FINITE VERB)

Let us go through the different options, referring to the German versions of (33), (34), (48): (51) Ge. a. * Dass Johann isst oft

Tomaten

(u¨berrascht die meisten Leute.) b. * Dass Johann oft isst Tomaten (u¨berrascht die meisten Leute.) c. Dass Johann oft Tomaten isst (u¨berrascht die meisten Leute.) That John (eats) often (eats) tomatoes (eats) (surprises most people)

As this is a context where embedded V2 is excluded, (51a) would have to have the structure in (49) with the arrow. The ill-formedness of (51a) could then be due to T° being final in German and / or to German not having V°-to -T° movement. (The corresponding example in Yiddish, (33e), is grammatical.)7 The ill-formedness of (51b) must be caused by German being an OVrather than a VO-language, i.e., the order inside the German VP is DP-V° (and not V°-DP as in English or in the Scandinavian languages). As for the well-formedness of (51c), it may either be the result of V°-to-T° movement provided T° is final, as in (50) with the arrow, or it may be the result of the lack of V°-to-T° movement, in which case we do not know 7

The view that Yiddish has V°-to-T° movement, as shown by (33e), taken together with the analysis of verb-final embedded clauses in German etc. as not having V°-to-T° movement at all – as discussed in connection with (52)–(54) below – leads to the following reinterpretation: Where Santorini (1992, 1993) and Wallenberg (2012, 2013) suggest that between 1400 and 1800, Yiddish changed from having V°-to-T° movement to a final T° to having V°-to-T° movement to a medial T°, I would like to suggest that Yiddish instead changed from not having V°-to-T° movement at all to having V°-to-T° movement to a medial T°.

381

382

STEN VIKNER

whether T° is medial or final, i.e., the structure could be either of (49) and (50) but crucially without the arrows. Many analyses have taken German – and by extension many of the other examples in (48) – to have V°-to-T° movement to a final T°, e.g., den Besten (1986: 247), Grewendorf (1990: 87), Webelhuth (1992: 73), Vikner (1995: 153). However, there is a growing number of arguments against this and in favor of not having V°-to-T° movement at all, cf. also e.g., Haider (1997a, 1997b, 2010, 2013, 2015), of which I will only mention two.8 As discussed in, e.g., Vikner (2005), a number of complex verbs in German and Dutch have a peculiar distribution. They occur as nonfinite verbs in both main and embedded clauses, (52a, 52b), but as finite verbs, they only occur in embedded clauses, (52c), and not in main clauses, (52d, 52e): (52)

Ge.

a.

b.

c.

Sie will bausparen. She wants (to) building-save (‘She wants to save money with a building society’) . . . weil sie bausparen will. . . . because she building-save wants (‘ . . . because she wants to save money with a building society’) . . . weil sie bauspart. . . . because she building-saves (‘ . . . because she saves money with a building society’)

(Example (52a)–(52c) adapted from Eisenberg 1998: 226, 324, (16a)) d. * Erst jetzt spart sie bau. e. * Erst jetzt bauspart sie. Only now (building-)saves she (building) (Intended: ‘Only now does she save money with a building society.’) These data support the view that clause-final finite verbs do not undergo V°-to-T° movement. What (52a–52c) have in common is that the verbs here are all in V°, i.e., these verbs are unable to leave V°. Vikner (2005) suggests that the reason could be that they then would have to be treated either as separable or as nonseparable verbs, and the special property of these verbs is that they have to fulfill the conditions on verbs of both types. The second argument for the view that in most OV-languages, clause-final finite verbs do not undergo V°-to-T° movement concerns the high amount of variation in the sequence of verbs found in embedded clauses like 8

There is yet another logical possibility, namely V°-to-T° movement to a medial T°, followed by movement of the remnant VP to a position further left than T°, see Haegeman (2001) and references there. Such an analysis would, however, not seem to be compatible with data of the types discussed in (52)–(54) below.

The Placement of Finite Verbs

(53)

a. b.

Du. Ge.

... ... ...

dat dass that

hij er he

haar sie her

rufen (shout)

hoort ho¨rt. hears

roepen. (shout)

both across the nine different Germanic OV-languages/dialects already discussed in (48) above and across six different constructions (perfect, passive, durative, causative, perception verbs, and modal verbs), as discussed in Vikner (2001b: 66–99) (see also Chapter 17 below on infinitival structures). This variation in embedded clauses where one of the two verbs is finite, as in (53a, 53b), is almost identical to the variation in the sequence of the verbs in main clauses where none of the two verbs in question are finite, (54a, 54b): (54)

a. Du.

Hij

zal

haar

b. Ge.

Er He

wird will

sie her

rufen (shout)

horen

roepen.

ho¨ren. hear

(shout)

In other words, it makes no significant difference whether the higher of the two verbs concerned is finite, as hoort / ho¨rt in (53a, 53b), or nonfinite, as horen / ho¨ren in (54a, 54b), which again would seem to indicate that in embedded clauses in the nine OV-Germanic languages in (48), there is no obligatory movement that involves only finite verbs. Again the conclusion is that there is no V°-to-T° movement in the nine Germanic OV-languages in (48).9 Consider first the consequences for the derivation of V°-to-T° movement by means of the RAH. The data above present us with the following problem: If the RAH is valid for both VO- and OV-languages in Germanic, then seven of the nine Germanic OV-languages in (48), including German, ought to have V°-to-T° movement, but the data discussed in connection with (52)–(54) speak against this. If the RAH is valid for only for VO-languages in Germanic, then we have nothing to say about the difference between German (48e) and Yiddish (33e), even though this difference would seem to be related to V°-to-T° movement. Let us turn to the consequences for the clause structure in Germanic. If German and the other OV-languages in (48) lack V°-to-T° movement, then we have no evidence as to whether T° is medial or final in these languages, i.e., the structure could be either (49) or (50) but crucially without the arrows. This again makes it a distinct possibility that T° precedes VP, (49), which would have the interesting 9

Notice that although Koeneman and Zeijlstra (2014: 604, example 60) assume ArgP to be head-final, so that Arg° (which corresponds to T° in the present paper) occurs to the right of the VP: (i) [ArgP [vP [VP subject object V°] v°] affix Arg°]. this does not have to imply a rightwards V°-to-Arg° movement. Koeneman and Zeijlstra (2014: 604) further say that their analysis is both compatible with a rightward V°-to-Arg° movement and with an analysis where no such movement is necessary, as the verb is already string-adjacent to Arg°.

383

384

STEN VIKNER

consequence that the only difference between the clause structure of Germanic OV-languages and that of Germanic VO-languages is the position of V°, compare (49) to e.g., (39).

16.7 Conclusion I have proposed an analysis where Germanic clauses consist of (among other things) CPs, TPs, and VPs. Within the Germanic languages, we find variation with respect to all three in relation to the position of the finite verb: The CP is the locus of the difference between V2-languages and nonV2-languages, where in the former group, the finite verbs in all main clauses (and in some embedded ones) move to C°. The TP is the locus of the difference between languages with V°-to-T° movement and languages without it, where in the former group, all finite verbs in all clauses move to T°. Given that if the clause in question is V2, the verb will then move on from T° to C°, the effect of V°-to-T° movement can only be observed in clauses where V2 does not apply. Finally, the VP is the locus of the difference between VO- and OVlanguages. The VO- or OV-property holds for all verbs in a language, even if its effect can only be observed for verbs which have moved neither to C° / V2, nor to T°. Here is a summary of these three properties:

(55)

finite verb

finite verb

Languages ¨ vdalian a. Icelandic, O (A¨lvdalen)

in C ° ( V 2 )

in T °

VO/OV

+

+

VO

b. Danish, Faroese,

+



VO

c. Yiddish

+

+

OV

d. Afrikaans, Dutch,

+



OV

e. Middle English



+

VO

f. English





VO

SCANDINAVIAN

g

(=

NORTH

GERMANIC)

Hallingdalen, Norwegian, Swedish

Frisian, German, Swabian, Swiss German, West Flemish

g. -



+

OV

h. -





OV

g

CONTINENTAL WEST GERMANIC

The Placement of Finite Verbs

References Alexiadou, A. and G. Fanselow 2002. “On the correlation between morphology and syntax: The case of V-to-I.” In J-W. Zwart and W. Abraham (eds.), Studies in Comparative Germanic Syntax. Amsterdam: John Benjamins: 219–242. Anganty´sson, A´. 2011. The Syntax of Embedded Clauses in Icelandic and Related Languages. Ph.D. dissertation, University of Iceland. Bentzen, K., G. H. Hrafnbjargarson, T. Hro´arsdo´ttir, and A-L. Wiklund 2007. “The Tromsø guide to the Force behind V2,” Working Papers in Scandinavian Syntax 79: 93–118. Besten, H. den 1983. “On the interaction of root transformations and lexical deletive rules.” In W. Abraham (ed.), On the Formal Syntax of the Westgermania. Amsterdam: John Benjamins: 47–131. Reprinted 1989 in H. den Besten (ed.), Studies in West Germanic Syntax. Amsterdam: Rodopi: 14–100. Besten, H. den 1986. “Decidability in the syntax of verbs of (not necessarily) West-Germanic languages,” Groninger Arbeiten zur Germanistischen Linguistik 28: 232–256. Reprinted 1989 in H. den Besten (ed.), Studies in West Germanic Syntax. Amsterdam: Rodopi: 137–160. Besten, H. den and C. Moed-van Walraven 1986. “The syntax of verbs in Yiddish.” In H. Haider and M. Prinzhorn (eds.), Verb Second Phenomena in Germanic Languages. Dordrecht: Foris: 111–135. Biberauer, T. and I. Roberts 2010. “Subjects, tense and verb-movement.” In T. Biberauer, A. Holmberg, I. Roberts, and M Sheehan (eds.), Parametric Variation: Null Subjects in Minimalist Theory. Cambridge University Press: 263–303. Bobaljik, J. 2003. “Realising Germanic inflection: Why morphology does not drive syntax,” Journal of Comparative Germanic Linguistics 6: 129–16. Bobaljik, J. and H.Thra´insson 1998. “Two heads aren’t always better than one,” Syntax 1.1: 37–71. Chomsky, N. 1986. Barriers. Cambridge MA: MIT Press. Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Davis, N. (ed.) 1971. Paston Letters and Papers of the Fifteenth Century. Oxford: Clarendon Press. Diesing, M. 1997. “Yiddish VP order and the typology of object movement in Germanic,” Natural Language and Linguistic Theory 15.2: 369–427. Eisenberg, P. 1998. Grundriß der deutschen Grammatik I: Das Wort. Stuttgart: J. B. Metzler. Emonds, J. 1978. “The verbal complex of V’-V in French,” Linguistic Inquiry 9: 151–175. Fischer, O., A. van Kemenade, W. Koopman, and W. van der Wurff 2001. The Syntax of Early English. Cambridge University Press. Fourquet, J. 1938. L’Ordre des e´le´ments de la phrase germanique ancien – Etudes de syntaxe de position. Paris: Les Belles Lettres.

385

386

STEN VIKNER

Frascarelli, M. and R. Hinterho¨lzl 2007. “Types of topics in German and Italian.” In K. Schwabe and S. Winkler (eds.), On Information Structure, Meaning and Form: Generalizations across Languages. Amsterdam: John Benjamins: 87–116. Garbacz, P. 2010. Word Order in O¨vdalian. A Study in Variation and Change. Ph.D. dissertation, University of Lund. Geilfuß, J. 1991. “Jiddisch als SOV-Sprache,” Zeitschrift fu¨r Sprachwissenschaft 9.1/2: 170–183. Also published in 1991 in Working Papers of Sonderforschungsbereich 340 (Universities of Stuttgart and Tu¨bingen) 11: 3–17. Grewendorf, G. 1990. “Verb-Bewegung und Negation im Deutschen,” Groninger Arbeiten zur germanistischen Linguistik 20: 57–125. Haegeman, L. 2001. “Antisymmetry and verb-final order in West Flemish,” The Journal of Comparative Germanic Linguistics 3: 207–232. Haider, H. 1997a. “Extraposition.” In D. Beerman, D. LeBlanc, and H.van Riemsdijk (eds.), Rightward Movement. Amsterdam: John Benjamins: 115–151. Haider, H. 1997b. “Precedence among predicates,” Journal of Comparative Germanic Linguistics 1.1: 3–41. Haider, H. 2005. “How to turn German into Icelandic – and derive the VOOV contrasts,” Journal of Comparative Germanic Linguistics 8: 1–53. Haider, H. 2010. The Syntax of German. Cambridge University Press. Haider, H. 2013. Symmetry Breaking in Syntax. Cambridge University Press. Haider, H. 2015. “Head directionality – in syntax and morphology.” In A. Fa´bregas, J. Mateu, and M. T. Putnam (eds.), Contemporary Linguistic Parameters. London: Bloomsbury Academic: 73–97. Hall, B. 1979. “Accounting for Yiddish word order, or what’s a nice NP like you doing in a place like this?” In J. Meisel and M. Pam (eds.), Linear Order and Generative Theory. Amsterdam: John Benjamins: 253–287. Harbour, D. 2016. “Parameters of poor pronoun systems,” Linguistic Inquiry 47.4: 706–722. Heycock, C., A. Sorace, Z. S. Hansen, S. Vikner, and F. Wilson 2012. “Detecting the late stages of syntactic change: The loss of V-to-T in Faroese,” Language 88.3: 558–560. Heycock, C. and J. Sundquist 2017. “Don’t rush to rehabilitate: A remark on Koeneman and Zeijlstra 2014,” Linguistic Inquiry 48.1: 173–179. Heycock, C. and J. Wallenberg 2013. “How variational acquisition drives syntactic change – The loss of verb movement in Scandinavian,” Journal of Comparative Germanic Linguistics 16: 127–157. Holmberg, A. 2010. “Parameters in minimalist theory: The case of Scandinavian,” Theoretical Linguistics 36.1: 1–48. Holmberg, A. 2015. “Verb second.” In T. Kiss and A. Alexiadou (eds.), Syntax: An International Handbook of Contemporary Syntactic Research, 2nd edn. Berlin: Walter de Gruyter: 343–384.

The Placement of Finite Verbs

Holmberg, A. (in press). “On the bottleneck hypothesis of V2 in Swedish.” In T. Biberauer, S. Wolfe, and R. Woods (eds.), Rethinking Verb Second. Cambridge University Press. Hrafnbjargarson, G. H. and A-L. Wiklund, 2010. “AGR and V2,” Theoretical Linguistics 36: 57–68. Julien, M. 2015. “The force of V2 revisited,” Journal of Comparative Germanic Linguistics 18.2: 139–181. Koeneman, O. and H. Zeijlstra 2014. “The rich agreement hypothesis rehabilitated,” Linguistic Inquiry 45.4: 571–615. Levander, L. 1909. A¨lvdalsmaº let i Dalarna. Stockholm: P. A. Norstedt and So¨ner. Levander, L. 1925. Dalmaº let: Beskrivning och historia, Vol. 1. Uppsala, Sweden: Appelbergs noktryckeri. Nyvad, A. M., K. R. Christensen, and S. Vikner 2017. “CP-recursion in Danish: A cP/CP-analysis,” The Linguistic Review 34.3: 449–477. Platzack, C. 1985. “A survey of generative analyses of the Verb Second phenomenon in Germanic,” Nordic Journal of Linguistics 8.1: 49–73. Platzack, C. and A. Holmberg 1989. “The Role of AGR and finiteness,” Working Papers in Scandinavian Syntax 43: 51–76. Pollock, J.-Y. 1989. “Verb movement, universal grammar, and the structure of IP,” Linguistic Inquiry 20.3: 365–424. Rohrbacher, B. 1999. Morphology-driven syntax. Amsterdam: John Benjamins. Sadock, J. 1998. “A vestige of verb final syntax in Yiddish,” Monatshefte fu¨r deutschsprachige Literatur und Kultur 90: 220–226. Santorini, B. 1993. “Jiddish als gemischte OV/VO-Sprache.” In W. Abraham, and, J. Bayer (eds.), Dialektsyntax, Sonderheft 5, Linguistische Berichte. Opladen: Westdeutscher Verlag: 230–245. Sprouse, R. 1998. “Some notes on the relationship between inflectional morphology and parameter setting in first and second languages acquisition.” In M-L. Beck (ed.), Morphology and Its Interfaces in Second Language Knowledge. Amsterdam: John Benjamins: 41–67. Thiersch, C. 1978. Topics in German Syntax. Ph.D. dissertation, MIT. Thra´insson, H. 2010. “Predictable and unpredictable sources of variable verb and adverb placement in Scandinavian,” Lingua 120: 1062–1088. Trosterud, T. 1989. “The null subject parameter and the new mainland Scandinavian word order: A possible counter example from a Norwegian dialect.” In J. Niemi (ed.), Papers from the 11th Scandinavian Conference of Linguistics 1: 87–100. Venaº s, K. 1977. Hallingmaº let. Oslo: Det Norske Samlaget. Vikner, S. 1995. Verb movement and expletive subjects in the Germanic languages. New York: Oxford University Press.

387

388

STEN VIKNER

Vikner, S. 1997. “V°-to-I° movement and inflection for person in all tenses.” In L. Haegeman (ed.), The New Comparative Syntax. London: Longman: 189–213. www.hum.au.dk/engelsk/engsv/papers/vikn97b.pdf. Vikner, S. 1999. “V°-til-I°Flytning og personfleksion i alle tempora,” Islenskt ma´ l 19: 81–128. www.hum.au.dk/engelsk/engsv/papers/ vikn99b.pdf. Vikner, S. 2001a. “Predicative adjective agreement.” In K. Adamzik and H. Christen (eds.), Sprachkontakt, Sprachvergleich, Sprachvariation: Festschrift fu¨r Gottfried Kolde. Tu¨bingen: Max Niemeyer Verlag: 399–414. www .hum.au.dk/engelsk/engsv/papers/vikn01b.pdf Vikner, S. 2001b. Verb Movement Variation in Germanic and Optimality Theory. Habilitationsschrift, University of Tu¨bingen. www.hum.au.dk/engelsk/ engsv/papers/viknhabi.pdf. Vikner, S. 2003. “Null objects under coordination in Yiddish and Scandinavian.” In L-O. Delsing, C. Falk, G. Josefsson, and H. A´. Sigurðsson, (eds.), Grammar in Focus: Festschrift for Christer Platzack, Vol. II. Dept. of Scandinavian Languages, University of Lund: 365–375. www .hum.au.dk/engelsk/engsv/papers/vikn03a.pdf Vikner, S. 2005. “Immobile complex verbs in Germanic,” Journal of Comparative Germanic Linguistics 8: 83–115. www.hum.au.dk/engelsk/ engsv/papers/vikn05b.pdf. Wackernagel, J. 1892. “U¨ber ein Gesetz der indogermanischen Wortstellung,” Indogermanische Forschungen 1: 333–436. Walkden, G. 2017. “Language contact and V3 in Germanic varieties new and old,” Journal of Comparative Germanic Linguistics 20: 49–81. Wallenberg, J. 2009. Antisymmetry and the Conservation of C-Command: Scrambling and Phrase Structure in Synchronic and Diachronic Perspective. Ph.D. dissertation, University of Pennsylvania. Wallenberg, J. 2012. “Language acquisition in German and phrase structure change in Yiddish.” In C. Galves, S. Cyrino, R. Lopes, F. Sandalo, and J. Avelar (eds.), Parameter Theory and Linguistic Change. Oxford University Press: 60–76. Wallenberg, J. 2013. “Scrambling, LF, and phrase structure change in Yiddish,” Lingua 133: 289–318. Webelhuth, G. 1992. Principles and Parameters of Syntactic Saturation. New York: Oxford University Press.

Chapter 17 Germanic Infinitives Susi Wurmbrand and Christos Christopoulos

17.1 Introduction Infinitival constructions have been central in Germanic for several reasons and they have been at the core of theory-building syntactic research. This paper presents an overview of two research areas – exceptional case marking (ECM) and restructuring. While in-depth discussions of these rich topics are not possible, we aim to highlight some of the major discoveries and remaining puzzles. The first part of the paper discusses the distinction between control and ECM and provides an overview of the distribution of ECM in six Germanic languages (English, Icelandic, Swedish, Norwegian, Dutch, and German). The second part of the paper summarizes phenomena falling under the rubric of restructuring. Although the literature on restructuring has focused predominantly on German and Dutch, phenomena showing sensitivity to restructuring have also been observed in the Mainland Scandinavian languages. Finally, in the last part we combine the observations regarding ECM and restructuring and suggest a possible connection.

17.2 Control and ECM Infinitival complements in English come in two basic forms – control and ECM. As shown in (1), a predicate like decide must combine with a control complement in which the embedded subject is obligatorily silent (unless an embedded for complementizer is used), whereas a predicate like believe requires a complement with an overt embedded subject which is case

We thank the following native speakers for providing judgments for the data given without references in this paper: Kajsa Djarv (Swedish), Tonje Andersen, Sigurd Hanserud, Adrian Kristing Ommundsen, Henrik T. Torgersen (Norwegian), and Fenna Bergsma, Paula Fenger, Beata Moskal (Dutch).

390

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

dependent on the matrix predicate (hence “exceptionally” case marked since case and theta-licensing apply in different predicates). The distinction between control and ECM in English has been a prominent research area in Government and Binding and Minimalist approaches. Among others, it led to the postulation of the Case Filter (Chomsky 1980, 1981; see Bobaljik and Wurmbrand 2008 for an overview), and contributed fundamentally to other core syntactic developments. The traditional analysis of control and ECM infinitives is based on the assumption that the two types of infinitives involve different structures. ECM infinitives are reduced clauses, e.g., IPs, which allow a case dependency between the matrix case licenser and the embedded subject in Spec,IP. Control infinitives, on the other hand are CPs which shield the embedded subject from being case marked by the matrix case licenser, which, assuming the Case Filter, prevents overt DP subjects and only allows PRO which is exempt from the Case Filter. (1)

a. We decided (*Leo) to go swimming. control b. We believed *(Leo) to have gone swimming. ECM

While the English control / ECM distinction has been majorly involved in shaping syntactic theories, some recalcitrant puzzles remain which we summarize in the next sections.

17.2.1 Predicting Control Versus ECM? A long-standing question is whether the control/ECM choice can be predicted from the matrix predicate and be derived from general principles of grammar, or whether it boils down to a lexical diacritic of the selecting predicate. A major step towards deriving the ECM versus control distinction has been Stowell’s (1982) observation, developed in detail in Pesetsky (1992), that control infinitives are typically future irrealis infinitives, whereas ECM infinitives are typically propositional infinitives. Stowell suggested that ECM and control infinitives differ semantically, and that this difference is a difference in tense: ECM infinitives are tenseless, whereas control infinitives are tensed. Adopting this claim, the so-called case theoretic view of control (see Chomsky and Lasnik 1993, 1995; Bosˇkovic´ 1996, 1997; Martin 1996, 2001) proposed that the control/ECM distinction can be derived from Case Theory as detailed in (2). (2)

a. Only [+tense] Infl assigns case i. [+tense, +finite] Infl assigns nominative ii. [+tense, −finite] Infl assigns a special Null Case b. All DPs must be case-marked (Case Filter) i. ii.

Overt DPs receive nominative or accusative, but not Null Case PRO can only receive Null Case

Germanic Infinitives

The assumption that PRO has to receive case and can only be assigned Null Case (which, in turn, can only be assigned by [+tense, −finite] Infl) guarantees that objects and subjects of finite clauses cannot be PRO (e.g.,*John saw PRO; *PRO left). Furthermore, these assumptions, together with the assumption that believe infinitives are [−tense], exclude a control structure under believe (the ungrammatical version of (1b)), since PRO would not be properly case-marked. Lastly, the assumption that Null Case can only be assigned to PRO excludes the ungrammatical case of (1a), since decide infinitives involve a [+tense, −finite] Infl which can only assign Null Case. While reducing the ECM / control distinction to case theory is an attractive enterprise, this view has also been challenged on both empirical and theoretical grounds, see Landau (2000), Baltin and Barrett (2002), Hornstein (2003), Cecchetto (2004), Wurmbrand (2014c). One challenge for case theoretic treatments of the ECM / Control distinction (problematic for both the Null Case view and the original case theoretic analysis) comes from languages in which the subject of control infinitives receives a detectable morphological case. Icelandic has played an important role in the discussion since morphological case on PRO is detectable via elements that show case agreement with the subject position, see Sigurðsson (1991). A different aspect of this problem (noted already in Chomsky 1981: 140, example 25) is posed by languages that allow overt, case-marked subjects of nonfinite, non ECM clauses, such as accusative subjects in Latin and Greek. Another challenge concerns the definition of [±tense]. In Wurmbrand (2014c) it is shown that based on their temporal and aspectual properties, infinitival constructions fall into three classes: future infinitives as in (3), which are argued to be tenseless but involve a syntactically present future modal woll; propositional attitude infinitives as in (4), which impose the N O W of the propositional attitude holder as the reference time of the embedded event; and simultaneous infinitives as in (5), which involve a reduced structure and form a single temporal domain with the matrix clause in that their references time corresponds to the reference time of the matrix predicate. (3)

a. Leo decided to read a book (tomorrow). b. The bridge is expected to collapse (tomorrow).

(4)

a. Leo claimed to be reading a book (*tomorrow), when . . . control b. Leo believed Lewis to be reading a book (*tomorrow), when . . . ECM

(5)

a. Leo tried to read a book (*tomorrow). b. Leo saw Lewis read a book (*tomorrow).

control ECM

control ECM

As can be seen in (3) through (5), all three types of infinitives come as control and ECM configurations. Thus, the distribution of control versus ECM does not correlate with the semantic tense and aspect properties, see Wurmbrand (2014c) for further tests and details, making a classification of

391

392

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

control as [+tense] and ECM as [−tense] essentially a syntactic diacritic equivalent to a selectional property. Furthermore, a [−tense] classification would also not be sufficient to predict ECM, neither in English nor crosslinguistically. As discussed in detail in Pesetsky (1992), in English only a subclass of [−tense] infinitives, namely those combining with a nonagentive matrix predicate (such as (4) or (6a)), allows ECM. If the matrix verb selects an agent argument as in (6b), ECM is blocked.1 (6) a. Mary wanted Bill to read the book. (Pesetsky 1992: 17, example 63a) b. *Mary demanded Bill to read the book. (Pesetsky 1992: 17, example 68a) The situation is (in part) different in Icelandic, however. As shown in (7), agentive verbs like say cannot combine with a control infinitive but require ECM, in contrast to English infinitives combining with agentive speech predicates. (7)

a. *Jo´nas sagði að hafa ´ı bı´o´. farið Jonas said to have gone to cinema ‘Jonas said / claimed to have gone to the cinema.’ b. Jo´nas sagði Garp hafa farið ´ı bı´o´. Jonas said Garpur.A C C have gone to cinema ‘Jonas said that Garpur has gone to the cinema.’

In the Section 17.2.2, we show that the distribution of ECM is subject to further restrictions in other Germanic languages, which will enforce the conclusion that in addition to the possible factors of agentivity and the semantic type of infinitives, other language-specific factors are at work to determine the choice of ECM predicates.2

17.2.2 ECM in Germanic In Germanic infinitival constructions, ECM comes in two types: ECM1 exemplified in (8a), where the matrix verb takes tenseless events, and ECM2, exemplified in (8b), where the matrix verb takes full propositions.3 Across Germanic, ECM1 always occurs without infinitival marker, which, following most of the literature, we take to show that ECM1 involves reduced complements (e.g., bare vPs). ECM2 shows variation regarding the infinitival marker: Some languages (English) require the infinitival marker; others (Icelandic) prohibit it, and yet others (Swedish) require or allow it 1

See the Agent / ECM correlation: For α, β and γ in E, if α assigns Agent to γ in E and requires γ to be animate as a lexical property, then α Case-marks β only if α θ-marks β (Pesetsky 1992: 21, example 83).

2

Here we concentrate on Germanic. However, it is worth noting that once we move beyond Germanic, in many languages ECM is not constrained by tense nor by finiteness (see Wurmbrand 2017a).

3

ECM also occurs in constructions where a verb (e.g., find, call, consider) combines with an AP / NP small-clause, as in We consider / find ice cream nutritious / the greatest invention. In this paper, we set these aside to focus on constructions involving ECM infinitives. We also set aside constructions which utilize a preposition (e.g., I want for him to like ice cream).

Germanic Infinitives

optionally depending on the matrix verb (see Christensen 2007, Thra´insson 1993). Given this variation, we set aside the intricate distribution of infinitival markers including the question of whether the marker is a T or C element (see Holmberg 1986, Platzack 1986, Johnson and Vikner 1994). a. I saw/let him steal my ice cream. ECM1 b. Leo believes/expects me to like ice cream. ECM2

(8)

The distribution of the two types of ECM across Germanic differs in that ECM1 occurs in all Germanic languages with a stable class of matrix predicates (causatives/perception verbs), whereas the class of verbs that allow for ECM2 seems to be variable. ECM1 is illustrated for five Germanic languages in (9). (9)

a. Eg I

let

Jon

fara.

let

Jon

go

Icelandic

‘I let Jon go’

b. Vi We

(Holmberg 1986: 158, example57a)

ho¨rde

dem

komma.

heard

them

come

Swedish

‘We heard them come.’

c. Vi We

(Holmberg 1986:158, example57b)

lot

/

hørte

Jon

synge

i dusj-en.

let

/

heard

Jon

sing

in shower-the

Norwegian

‘We let / heard Jon sing in the shower.’ (Holmberg 1986: 155, example49a)

d. dat that

Jan

haar

een lied

hoorde / liet zingen. Dutch

Jan

her

a song

heard / let sing

‘that Jan heard / let her sing a song.’ (Broekhuis and Corver 2015: 614, example 39a, 39b)

e. Leo hat den Prinzen

einen Frosch essen lassen / (ge)sehen. German

Leo has the.A C C prince a.A C C frog eat let

/ see(n)

‘Leo let/saw the prince eat a frog.’

Once we move to ECM2, the situation is less stable across Germanic: in some languages, certain predicates allow for ECM infinitives, whereas their counterparts in other languages do not. We illustrate this with the verbs expect, consider and believe. As shown in (10), expect can occur in ECM (-like) configurations in Swedish and possibly Norwegian, but not in Dutch and German.4 Icelandic also does not allow ECM in this context (see (10e); 4

A notorious difficulty with expect-statements is to determine whether a configuration involves ECM or object control/prolepsis (see among others Bresnan 1972, Pesetsky 1992). In the latter, theta-role and Case both come from the matrix predicate, hence technically no ECM would take place. While (10a, 10b) may thus not be instances of true ECM, the contrasts in (10) are nevertheless noteworthy – although German and Dutch allow object control with verbs like convince, beg and many others, the configurations in (10c, 10d) with an

ACC

DP are impossible. Thus, if (10a, 10b) are explained away as object control, the variation across

languages would have to be located in the matrix verbs’ potential to license a direct object (e.g., via Case). However, this would make it a suspicious coincidence that none of the verbs that are potentially ambiguous between ECM and object control in English can license a direct object in German and Dutch (note also that the verb erwarten assigns

ACC

when it combines only with a DP object and no infinitive). On the other

393

394

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

ECM is also impossible without að; however, verbs like expect do not select the infinitive directly but require a pronominal complement (þess). We suggest that in such cases þess is a correlate pronoun in complement position of the verb, associated with the dislocated infinitive. ECM is then excluded independently since the infinitive is not in the required sisterhood configuration with the matrix verb. (10) a. Jag fo¨rva¨ntade mig henne *(?att) svika I

expected

me

her

*(?to)

sina

va¨nner.

Swedish

let.down her.own friends

‘I expected her to let her friends down.’

b. Ingen

forventer lærere

aº være perfekte.

Norwegian5

nobody expects teachers to be perfect ‘Nobody expects teachers to be perfect.’

c. *dat that

(Lødrup 2002: 3, example10)

ik

Peter

de deur

verwachtte

te sluiten.

I

Peter

the door

expected

to lock

Dutch

‘that I expected Peter to lock the door.’ (Broekhuis 1992: 211, example 33)

d. Sie She

erwartet

(*ihn)

Goldfische

zu mo¨gen.

expects

(*him.A C C )

goldfish

to like

German

‘She expects (*him) to like goldfish.’

e. E´g vænti I

*(þess)

(*Astridar)

að hoppa ´ı sjo´inn

a´ morgun.

Icelandic

expect *(it.G E N ) (*Astrid.G E N ) to jump in sea.the on morning

‘I expect (*Astrid) to jump into the sea tomorrow.’

A similar distribution is found for the matrix verb consider – ECM is possible in English / Icelandic / Swedish, but shows speaker variation and additional restrictions in Norwegian, as shown in (11) (German and Dutch consider do not occur with infinitival complements). Furthermore, the frequency of ECM with similar verbs (find and assume) is strikingly lower in Norwegian than in Swedish.

hand, if (10a, 10b) involve an ECM configuration, the language variation can be attributed to a general lack of ECM2 in German and Dutch. 5

Lødrup (2008) shows that there is a strong tendency in Norwegian for ECM subjects to be dislocated (e.g., via verb second movement or other movement operations); if the subject remains in the regular object position, the structure is often degraded or unacceptable. This Derived Object Constraint (Postal 1974) is also found with verbs like wager in English (Pesetsky 1992). Furthermore, ECM in Norwegian is restricted to embedded individual-level predicates (or usages) as in (10c) in Norwegian, and eventive interpretations are impossible (see also Lødrup 2008): i. *Jeg I

forventer

ham

å drepe

mus-en

expect

him

to kill

mouse-the

‘I expect him to kill the mouse.’ Similar restrictions have been noted for other languages. For instance, in Korean and Japanese ECM is generally disallowed with embedded stage-level predicates (this is simplified – see, among many others, Yoon 2007 and Horn 2008 for detailed semantic restrictions). However, it is not well understood yet what the semantic restrictions on ECM in individual languages are and when they do and do not apply.

Germanic Infinitives

(11) a. Eg tel I

hann vera heimskan.

consider him

be

‘I consider him stupid.’ b. Jag anser

I

(Holmberg 1986: 159, example 60b)

henne ?(att) ha

consider her

Icelandic

stupid

?(to)

svikt

sina

va¨nner. Swedish

have let.down her.own friends

‘I consider her to have let her friends down.’ c. Internett-brukerne

internet-users.D E F

anser

dette

aº være

en fordel.

consider

this

to be

an advantage

‘The internet users consider this to be an advantage.’ (Lødrup 2008: 162, example 26)

Finally, believe takes ECM infinitives in English/Icelandic but not in Swedish / Norwegian / Dutch and German, as illustrated in (12). (12) a. Pe´tur taliði

Marı´u

ekki hafa

vaskað

upp diskana

Peter believed Maria.A C C not have washed up

Icelandic

dishes-the

‘Peter believed Mary not to have washed up the dishes.’ (Christensen 2007: 156, example 25a) b. *Jag

I

tror

henne

(att) vara begaº vad

believe

her

(to) be

Swedish

gifted

‘I believe her to be gifted.’ c. *Jeg

I

tror

ham



believe

him

to be

være dum

Norwegian

stupid

‘I believe him to be stupid.’ d. Ik geloof

I

(*haar) slim

believe (*her)

te zijn.

Dutch

smart to be

‘I believe (her) to be smart.’ e. Sie

glaubt

(*ihn)

Goldfische zu mo¨gen.

She believes (*him.A C C ) goldfish

German

to like

‘She believes him to like goldfish.’

The distribution of ECM in the six Germanic languages discussed here is summarized in Table 17.1. The existence of ECM1 shows that all of these languages allow “exceptional case” in principle. What seems to vary across Germanic is ECM2, in particular the list of verbs that participate in ECM2. Table 17.1 ECM ECM2 Language

ECM1

say / claim

believe

consider

expect

Icelandic English

✓ ✓

✓ ✗

✓ ✓

✓ ✓

N/A ✓

Swedish Norwegian

✓ ✓

✗ ✗

✗ ✗

✓ restricted

✓ restricted

German / Dutch ✓









395

396

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

Interestingly, however, the distribution of ECM infinitivals brings out a markedness hierarchy for matrix predicates: If a language allows ECM with its believe predicate, it also allows it with its consider and expect predicates (modulo idiosyncratic lexical restrictions). (13)

most marked ⇠ say – believe – consider – consider to be – expect – see ⇢ least marked

The distribution of ECM in German and Dutch poses another puzzle for theories which derive clausal ECM via a truncated CP-less structure. As we will see in Section 17.3, German and Dutch display a range of restructuring phenomena, which have been taken to indicate reduced (in particular CPless) structures. Yet, as we have seen, in these languages, ECM2 is not possible at all. Thus, CP-omission is at best a necessary condition for ECM and not a sufficient one. In the next section, we discuss restructuring in Germanic, and return to ECM in Section 17.4, suggesting a possible connection between ECM and restructuring.

17.3 Restructuring The second research area related to infinitives which has produced a wealth of works across Germanic is restructuring (also known as clause union, reanalysis, coherence, verb complexes, verb clusters). Restructuring refers to the phenomenon that certain verbal complements are nonclausal in that they allow dependencies which are otherwise characterized as clause-bound to span the embedded and matrix predicates. Although the phenomena through which restructuring can be diagnosed vary across languages, the broad distribution of restructuring can be shown to follow the same basic pattern. One important observation already made in Wurmbrand (2001) based on German is that the phenomenon of restructuring is not a simple on/off property, but rather there are different degrees of restructuring, both within single languages and across languages (see Wurmbrand (2014a, 2014b, 2015b, 2016) for motivation and documentation). More specifically, comparing restructuring cross-linguistically, we find a stable split of embedded clauses (both finite and nonfinite) into three types of complements which are defined semantically and form the following transparency scale (from least to most transparent): propositional attitude, speech » future, irrealis » tenseless, aspectual. While differences between these types of complements can be neutralized in a language, which then makes restructuring less visible in that language, what is nevertheless observable in all languages investigated so far is the following implicational dependency: If a property shows transparency restrictions (i.e., is not possible from all types of complements), the transparency potential increases from left to right on the scale. Thus, there are no languages / properties for which a future complement is transparent but a tenseless complement is

Germanic Infinitives

not. In Section 17.3.1, we will see effects of the implicational restructuring hierarchy in Dutch and German. In Section 17.3.2, we show that although Swedish and Norwegian are not typical restructuring languages, effects of restructuring can be observed even in these languages.

17.3.1 West Germanic Restructuring 17.3.1.1 Splitting and IPP in Dutch In German and Dutch, infinitival complements of lexical verbs can occur in three configurations. In this subsection, we concentrate on Dutch and return to German in the next subsection. The examples in (14) illustrate two main phenomena. First, as shown in (14a, 14b), a matrix element (in this case the boxed verb) can occur between material from the embedded complement (underlined). Following Broekhuis and Corver (2015), we will refer to such configurations as splitting to stay neutral about how exactly the discontinuous order of the embedded predicate is achieved. Second, as shown in (14c), the matrix verb which is selected by the perfect auxiliary have can be realized as an infinitive instead of the expected participle. This phenomenon is called Infinitivus Pro Participio ‘Infinitive for participle’ [IPP]. While both splitting and the IPP are optional in tenseless try-complements (see (14a, 14c)), there is nevertheless an implicational relation between the two phenomena: As illustrated in (14d), when splitting does not occur (or to be more accurate, when the embedded verb does not end up next to the matrix verb, the IPP cannot occur either. (14) a. dat

Jan dat boek

heeft geprobeerd te lezen

that Jan that book has

try.P A R T

splitting, no IPP

to read

‘that Jan has tried to read that book’ (Broekhuis and Corver 2015: 869, example 594b)

b. dat

Jan dat boek

heeft proberen te lezen

that Jan that book has

try.I N F

splitting, IPP

to read

‘that Jan has tried to read that book’ (Broekhuis and Corver 2015: 869, example 594c)

c. dat

Jan heeft geprobeerd (om) dat boek

that Jan has

try.P A R T

te lezen no splitting, no IPP

(O M ) that book to read

‘that Jan has tried to read that book’ (Broekhuis and Corver 2015: 869, (594a)]

d. *dat

Jan

heeft

proberen

(om)

dat boek

that

Jan

has

try.I N F

(O M )

that book to read

te lezen *no splitting, IPP

‘that Jan has tried to read that book’ (Broekhuis and Corver 2015: 623, example 59a)

The research on restructuring in West Germanic has centered around the question of how splitting is derived, in particular which of the underlined

397

398

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

parts in (14a, 14b) is moved away from the other part. The question is complicated by another issue arising for the West Germanic languages, namely the question of whether the underlying VP and TP structures are head final (the complement precedes the matrix verb), or head-initial (the complement follows the matrix verb). Thus, the configuration in (14c) may be treated as a derived or underived order. As shown in detail in Wurmbrand (2006, 2017b), depending on one’s commitments to the directionality issue, the syntactic toolkit offers a multitude of ways for how to derive the word orders in (14a, 14b). Some basic and highly simplified options are given (15), which will be sufficient for the main point to be made in this paper, namely that splitting requires some form of movement across the boundary (the bold parts in (15)) of the embedded predicate (whatever it’s size and label), which thus has to be transparent for such movement. (15)

a. [matrix [embedded the book tV] has tried [V to read]] V movement b. [matrix [the book] tembedded XP has tried remnant extraposition [embedded tNP to read]] [the book] has tried [embedded tNP to read]] scrambling [ c. matrix

The earliest approaches to restructuring assume that examples like (14a, 14b) involve a verb cluster – a complex head formed by the matrix and embedded verbs. As indicated in (15a), the infinitival complement originates to the left of the matrix verb and the embedded verb undergoes some form of verb raising which dislocates the embedded verb from the rest of the embedded clause, see for instance Evers (1975), den Besten and Edmondson (1983). While this approach has been successful in accounting for contexts in which the verbs involved form a compact, often impenetrable unit, it cannot extend to configurations in which the right part of the split is clearly phrasal as in (16a), a configuration which has been named the third construction (reflecting that it is in-between verb raising and a full clausal complement).6 As shown in (16b), the third construction is not compatible with the IPP, thus a verb raising derivation as in (15a) may still be possible for (14b), see Wurmbrand 2007 for other issues arising for complex head approaches. (16)

6

a. ?dat Elsje hem heeft geprobeerd een brief te schrijven that Elsje him has try.P A R T a letter to write ‘that Elsje has tried to write him a letter’ (Rutten 1991: 68, example 6d) b. *dat Elsje hem heeft proberen een brief te schrijven that Elsje him has try.I N F a letter to write ‘that Elsje has tried to write him a letter’ (Rutten 1991: 68, example 7d)

Dutch speakers vary regarding whether they allow the third construction.

Germanic Infinitives

The second option in (15b) has a similar underlying configuration as in (15a) but instead of verb raising, the entire embedded predicate moves to the right via extraposition (see, among others, den Besten and Rutten (1989), den Besten and Broekhuis (1992)). To derive the discontinuous order, the left part of the split has to first vacate the embedded XP, e.g., via scrambling. Thus extraposition only affects the remnant of the embedded predicate, which readily derives the third construction. Lastly, the third option in (15c) is based on a head-initial base-order. Since in that case, the right part of the split is already in the “right” position, all that is needed in this scenario is movement (e.g., scrambling or object shift) of the left part of the split (see Zwart (1996)). We will not decide between these options in this paper. All we wish to maintain is that to account for restructuring as a whole (and not just for verb clusters) at least one of the movement operations in (15), possibly a combination thereof, is necessary. Let us return to different types of matrix predicates and see which of the complementation options in (14) are available. As we have seen in this section, the IPP is not possible when the matrix and embedded verbs are not adjacent (cf. (14d) or (16b)). Although adjacency is a necessary condition for the IPP, it is not a sufficient one, as we see in (14a). This observation in fact reflects a property of restructuring in West Germanic in general – namely, the possibility of the IPP entails the possibility of splitting.7 In other words, whenever a matrix verb appears in the IPP, the complement is transparent. As a consequence, since both splitting and the IPP require (some form of) restructuring, the IPP constructions single out only a subset of the restructuring contexts, which indicates, as we lay out in more detail below, that there are different degrees of restructuring. Further evidence for a nuanced notion of restructuring comes from matrix predicates which select future oriented complements (e.g., decide now to do something later). In such contexts, splitting is usually possible, perhaps marked for some speakers, whereas the IPP is entirely excluded, even in cases that, superficially, look identical to the IPP context in (14b). This is illustrated in (17) (our paraphrases). 7

Recall that splitting is defined as a case where a matrix element (not just the matrix verb) is interspersed with the embedded predicate. Thus, splitting is not restricted to constructions or languages in which the embedded verb can follow the matrix verb, but is also observed in cases where an element of the embedded clause precedes the matrix subject. This is the case, for instance, in German IPP constructions in i. which prohibit extraposition of the modal complement; however, as indicated splitting is still observable. i. dass that

das Buch

keiner

hat

lesen

wollen

the book

nobody

has

read.I N F

want.I N F

‘ that nobody has wanted to read the book’ Note also that it cannot be assumed that the IPP entails actual splitting (or verb cluster formation) since as summarized in Wurmbrand (2006, 2017a), in certain languages and contexts, the IPP also occurs when the complement is unsplit and fully precedes the matrix verb.

399

400

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

(17) a. dat zij

de vergadering hebben besloten

that they the meeting

have

te bezoeken splitting, no IPP

decide.P A R T to attend

‘that they decided to attend the meeting’ (den Besten and Rutten 1989: 46, example 16c)

b. *dat that

zij

de vergadering

hebben

besluiten

te bezoeken *splitting, IPP

they

the meeting

have

decide.I N F to attend

‘that they decided to attend the meeting’ (den Besten and Rutten 1989: 46, example 16b)

Lastly, the configuration in (14c) is revealed, as the sole or preferred option, in contexts where a matrix verb selects a propositional complement. Such complements are not transparent, as shown in (18a–18c) (see below for a comment on variation). In comparison, splitting with a remnant that is larger than just the embedded verb is more acceptable with matrix verbs selecting a future complement, as in (18d).8 (18) a. ??dat that

Hans

Carla

beweerde

een brief te hebben gestuurd *splitting

Hans

Carla

claimed

a letter to have

sent

‘that Hans claimed to have sent Carla a letter’ (Rutten 1991: 46, example 15a)

b. ??dat that

Piet

het probleem

geloofde

wel te begrijpen

Piet

the problem

believed

too to understand

‘that Piet believed that he understood the problem’ (Rutten 1991: 46, example15b)

c. Jan heeft {*het vliegtuig} Jan has

beweerd

{het vliegtuig}

niet te kunnen zien

{*the airplane} claimed {the airplane} not to can

see

‘Jan claimed not to be able to see the airplane’ (B. Moskal, F. Bergsma, p.c.)

d. ?dat Anna

haar broer

heeft

besloten

een brief te schrijven

that Anna his brother the letter decide.P A R T a letter to write ‘that Anna has decided to write her brother a letter’ (Rutten 1991: 43, example 8)

As noted above, from a cross-linguistic perspective, the three-way split of complements in tenseless (try-type), future (decide-type), and propositional (claim-type) complements is not surprising since this basic split is very robust across languages. However, as shown in the examples, the judgments are not absolute but seem to be on a continuum. We suggest that this is the case because the classification of matrix verbs can sometimes be vague and ambiguous. For instance, examples such as (19a) are marked as possible for some speakers in Broekhuis and Corver (2015). On closer inspection, however, splitting with a verb like beweren seems to only be possible when the infinitive is minimal and does not contain other material as in (18a–18c). One 8

Although, as noted in footnote 7, the third construction is marked for many Dutch speakers, speakers generally find a contrast between (18d) (better) and (18a, 18b) (worse).

Germanic Infinitives

possibility suggested in Rutten (1991: 76) is that the split version in cases such as (19a) is erroneously analyzed as verb raising in analogy with constructions in which splitting is possible. Alternatively, split cases with beweren could be taken as future constructions rather than as propositional ones, which the translation given for (19a) seems to suggest already. In contrast to infinitival complements of claim in English, which in the absence of perfect have only allow a simultaneous interpretation of embedded episodic predicates (cf. *Jan claimed to buy the house tomorrow, see Wurmbrand 2014c), the complement of Dutch beweren appears to also allow a future interpretation (cf. (19b)). It may thus be the case that, at least for some speakers, when beweren combines with a future infinitive, it does not function as a (full) propositional verb. (19) a. dat Jan {%het huis} heeft beweerd {het huis} te kopen that Jan {%the house} has claimed {the house} to buy ‘that Jan has claimed to buy the house’ (Broekhuis and Corver 2015: 848, example 546a) b. Jan beweerde morgen te vertrekken. Jan claimed tomorrow to leave ‘Jan claimed to leave tomorrow.’ (Broekhuis and Corver 2015: 776, example 369) A similar phenomenon has been noted in Haeseryn et al. (1997) for the verb menen ‘mean / think’ which can be used like an epistemic modal verb meaning ‘to wrongly suppose’. Importantly, transparency effects only arise when menen functions as a modal and not when it is used as a true propositional verb. Assuming that the variation encountered regarding the classification of certain propositional verbs can be attributed to subtle semantic differences, we arrive at the distribution of Dutch infinitival complements in Table 17.2.9 Table 17.2 Splitting and IPP in Dutch Control Structure

IPP

Splitting (possible)

Config A





Config B Config C

✗ ✗

✓ ✗

Unattested ✓

9



Transparency Class I (try)

Class II (decide)

most ↕ least

✓ ✓

Class III (claim)

✓ ✓ ✓

most ⇠ transparency

✓ potential ⇢ least

Broekhuis and Corver (2015) note another difference between tenseless and future complements versus propositional complements: The former allow control by implicit passive agents whereas the latter do not – i.e., Visser’s generalization is only observed in propositional contexts. We depart from the authors mentioned and assume that this difference can be related to the lack of transparency of propositional complements if implicit control requires a dependency between PRO and the implicit matrix agent and/or matrix T (see van Urk 2013), which is blocked across nontransparent propositional complements but possible across transparent tenseless and future complements.

401

402

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

17.3.1.2 Splitting and Long Passive in German Turning to German, the distribution of splitting is the same as in Dutch, which is illustrated in (20). Since the infinitival complement can (and for some speakers is preferred) to occur to the left of the matrix verb, we use pronoun fronting across the matrix subject in (20a). Complements of propositional verbs typically occur after the matrix verb; however, in neither configuration, splitting is possible, as shown in (20b, 20c). (20) a. weil since

ihn

der Hans

zu reparieren

versuchte

/

beschloss

it.A C C

the John

to repair

tried

/

decided

German

‘since John tried / decided to repair it’ (Wurmbrand 2001: 268, examples 215a, 216b)

b. *weil ihn

der Hans {repariert zu haben} bedauerte {repariert zu haben}

since it.A C C the John {repaired to have} regretted {repaired to have} ‘since John regretted having repaired it’

c. *weil since

sie

der Hans

{zu mo¨gen}

behauptete

{zu mo¨gen}

her

the John

{like}

claimed

{to like}

‘since John claimed to like her’

(Wurmbrand 2001: 336, example 269ix)

German differs from Dutch in not allowing the IPP with lexical restructuring verbs such as try. The IPP only occurs with modals, causative lassen ‘let’, and perception verbs. The three-way split can nevertheless be observed if the property of long passive is included. Long passive refers to constructions in which the matrix predicate is passivized (or unaccusative), but the embedded predicate is active. As shown in (21a), in tenseless restructuring contexts, the embedded object can then be promoted to matrix subject. Future and propositional contexts, (21b), on the other hand, do not allow this operation.10 (21) a. dass der Traktor zu reparieren versucht wurde that the tractor.N O M to repair tried was Lit. ‘that the tractor was tried to repair’ ‘that they tried to repair the tractor’ (Wurmbrand 2001: 19, example 6a) b. *dass der Traktor zu reparieren beschlossen / behauptet wurde that the tractor.N O M to repair decided / claimed was ‘that they decided to repair the tractor’ (Wurmbrand 2001: 267, 332, examples 214c, 265ix)

10

It is sometimes claimed that long passive is a marked operation. While this may be the case for certain speakers, it is frequent in corpora and has been shown experimentally to be clearly acceptable to native speakers (see Bader and Schmid 2009). Furthermore speakers have very clear judgments about the scope properties of such constructions (see Bobaljik and Wurmbrand 2005), which would be surprising if this construction is not part of the core grammar of German speakers.

Germanic Infinitives

Table 17.3 Splitting and long passive in German Control Structure

Long passive

Splitting (possible)

Class II Transparency Class I (try) (decide)

Config A Config B

✓ ✗

✓ ✓

most ↕ least

Config C ✗ ✗ Unattested possible impossible

✓ ✓

Class III (claim)



✓ ✓ ✓ most ⇠ transparency potential ⇢ least

German thus also displays a three-way split of infinitival constructions as summarized in Table 17.3.

17.3.2 Scandinavian Restructuring The Scandinavian languages differ from West Germanic in that, among others, they lack verb clusters, the IPP, and scrambling. This has often been taken to indicate that restructuring is not active in Scandinavian. While it is certainly true that restructuring is less visible in these languages, it can nevertheless be shown that restructuring effects exist. In this section, we summarize the main observations that have been made regarding restructuring in Norwegian and Swedish. 17.3.2.1 Parasitic Morphology in Swedish (and beyond) The existence of restructuring has been substantiated by a rich set of morphological evidence observed in certain varieties of Swedish and Norwegian. As illustrated in (22), the embedded verb in a complement of a perfect modal construction can occur in the usual infinitival form (accepted by all speakers) or as a participle (accepted in many Norwegian and Swedish varieties). I will refer to such constructions as parasitic morphology.11 (22) a. Jeg hadde villet I

had

lese

/ lest

boka. Norwegian (varieties)

want.P A R T read. I N F / read.P A R T book.D E F

‘I would have liked to read the book.’ (Wiklund 2001: 201, examples 2a, 2b)

b. Jag hade velat I

la¨st

/ la¨sa

boken.

Swedish (varieties)

had want.P A R T read. I N F / read.P A R T book.D E F

‘I would have liked to read the book.’ (Wiklund 2001: 200f, examples 1a, 1b)

11

Parasitic participles are also found in many West Germanic dialects (see Dikken and Hoekstra 1997, Zwart 1995, Wurmbrand 2012).

403

404

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

Parasitic participles, and as we will see below parasitic morphology in general, is always optional – the parasitic forms alternate with infinitives. Furthermore, the parasitic morphology is semantically vacuous; the embedded participles in (22) are not interpreted as perfectives, but rather the meaning is identical to the meaning of the infinitival construction (see Wiklund 2005, 2007 for a caveat about covert auxiliaries and arguments against a necessary counterfactual interpretation). As shown in Wiklund 2005, 2007, in addition to parasitic participles, present and past tense as well as imperative morphology can occur parasitically. In (23a), the matrix past tense is copied onto the embedded verb, and in (23b), the matrix imperative morphology. The continuation ‘but did not succeed’ in (23a) shows that the embedded clause is interpreted like an irrealis infinitive, i.e., the past morphology is semantically vacuous since otherwise the after-clause would yield a contradiction as in a true past statement ‘#He fried a fish and didn’t succeed’. (23) a. Han pro¨vade {att steka } / { o stekte He

} en fisk men lychades

inte.

try.P A S T{to fry. I N F } / { and fry. P A S T } a fish but succeeded not

‘He tried to fry a fish but did not succeed’ (Wiklund 2007: 19, examples 13a, 13b)

b. Bo¨rja

{att la¨sa

Start.I M P {to ‘Start reading!’

} / {o

la¨s

} !

read. I N F } / { and read. I M P } (Wiklund 2007: 50, examples 37a, 3b)

Whether and what type of parasitic morphology is possible in a complement depends on two factors: the type of matrix predicate and the presence/absence of the infinitival marker. If the complement in the nonparasitic form involves a bare infinitive as in (22), only participial morphology can be copied from the matrix clause. If the infinitival complement occurs with an infinitival marker as in (23), tense and imperative morphology can be parasitic as well. Since we will not discuss details of the structure of infinitival complements, we will set this property aside (see Wiklund 2005, 2007). The property that is relevant for the discussion here is the connection between parasitic morphology and the choice of matrix verb. The examples with successful parasitic morphology all involve verbs from the tenseless try-class. As shown in (24), parasitic morphology is excluded in future, (24a), and propositional complements, (24b, 24c). Wiklund shows that this is systematic and hence concludes that parasitic morphology is constrained by restructuring. (24) a. Han beslutade {att steka } / {*o stete } en fisk. He decide.P A S T {to fry. I N F } / { * and fry. P A S T } a fish ‘He decided to fry a fish.’ (Wiklund 2007: 25, examples 24b, 25b)

Germanic Infinitives

b. De

hade ansett

honom {bli

They had considered.P A R T him

} / { * blivit

} oartig.

{become. I N F } / { * become. P A R T } impolite

‘They had considered him to become impolite.’

(Wiklund 2007: 45, examples 21a, 21b)

c. Tro

dig {vara

} / { * var

} vacker

och du blir

vacker!

believe.I M P R E L F {be. I N F } / { * be. I M P } beautiful and you become beautiful! ‘Believe yourself to be beautiful and you

(Wiklund 2007: 45,

become beautiful!’

examples 20a, 20b)

In addition to the sensitivity to the type of matrix predicate, complements with parasitic morphology also display other restructuring diagnostics, even when occurring with tense morphology: They do not involve semantic tense, do not allow overt subjects, (25a), nor floating quantifiers associated with an embedded subject, and cannot occur with embedded TP and vP-modifiers, (25b). (25) a. Lars pro¨vade o (*han) slutade tidigt. Lars try.P A S T o (*he) finish. P A S T early ‘Lars tried to finish early.’ (Wiklund 2007: 169, example 21) b. Han pro¨vade {att (alltid) sluta } / {o (*alltid) slutade } tidigt He try.P A S T { to (always) finish.I N F } / { and (*always) finish.P A S T } early ‘He tried to (always) finish early.’

(Wiklund 2007: 79, examples 26a, 26b)

Although parasitic morphology provides evidence for the existence of restructuring in Swedish (and other Scandinavian languages),12 it only singles out one class of restructuring – the tenseless class. There are, however, some speakers who also allow (at least some) parasitic copying in future complements. For such speakers, examples such as (26) are possible. But as Wiklund notes, the phenomenon of copying into future complements is rather limited. (26)

%Han hade planerat o kommit hem. He had plan.P A R T o come.P A R T home ‘He had planned to come home.’ (Wiklund 2007: 65, example 90)

Parasitic morphology thus leads to the distribution of restructuring in Table 17.4.

17.3.2.2 Long Passive in Norwegian The last set of restructuring effects arise as a combination of parasitic morphology and long passive. Holmberg (2002) suggests, based on complex passives as in (27), that Danish and a variety of Norwegian involve restructuring. As shown in (27a, 27b), complements of try – i.e., a 12

Wiklund 2007: 188–191 reports that parasitic morphology is also found in Norwegian varieties (Solør Norwegian, the dialects spoken in Romsdal and Gudbrandsdalen), Danish varieties, and Faroese varieties.

405

406

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

Table 17.4 Parasitic morphology in Swedish Control Structure

Transparency

Class I (try)

Config A

most ↕ least



Config B Config C

Class II (decide)

Class III (claim)

✗/ % ✗

prototypical restructuring predicate of the tenseless class – allow these constructions, whereas complements of a typical nonrestructuring verb, hate in (27c) do not. (27)

a. Bilen ble forsøkt reparert. Norwegian 1 car.the was tried repaired ‘There was an attempt to repair the car.’ (Holmberg 2002: 119, example 67a) b. Bilen blev forsøgt repareret. Danish car.the was tried repaired ‘There was an attempt to repair the car.’ (Holmberg 2002: 119, example 67b) c. Bilen ble hatet reparert. car.the was hated repaired ‘They hated to repair the car.’ (Holmberg 2002: 122, example 78)

Although further research is needed to see what exactly the distribution of this form of complex passive is, these facts lend initial plausibility to the hypothesis that long passive exists in Norwegian, which in cases such as (27a, 27b) is paired with copying of the matrix participial morphology onto an underspecified embedded verb (note again that the meaning of the embedded predicates in (27a, 27b) is irrealis and not perfect). In a recent corpus study, Lødrup (2014) confirms this hypothesis and shows that, at least in informal contexts, long passive is clearly attested in Norwegian. Interestingly, it can be observed that Norwegian allows in fact two forms of long passive, both of which have been attested abundantly cross-linguistically. As shown in (28a), a passive matrix predicate (in this case the familiar restructuring verb try) can trigger long object movement in that the embedded object is promoted to matrix subject. Like in the German case in (21a), the embedded verb occurs as an active infinitive. In addition, Norwegian also allows the long passive configuration in (28b), which differs from (28a) in that the embedded verb is also passive. Double passive configurations are common in other languages (see Wurmbrand 2015a, Wurmbrand and Shimamura 2017 and references therein) and also restricted to restructuring contexts. The analysis proposed there is that in

Germanic Infinitives

both cases in (28), the embedded predicate is underspecified for voice features and enters a dependency with the matrix voice head. The source of difference of the two types of long passive lies in morphology: In (28a) (which is the only option in German), the lower verb is spelled out as a default form (the infinitive), whereas in (28b), voice matching takes place in that the passive feature of the matrix verb is also spelled out on the embedded verb. Thus, (28b) is again a case of vacuous parasitic morphology. Languages usually allow only one of these options, but Norwegian allows both (although Lødrup 2014 notes that speakers may have a preference for voice matching). (28) a.

b.

aº gjøre Norwegian to do.I N F (Lødrup 2014: 371, example 13) Slike ting forsøkes ofte aº gjøres such things try.P R E S . P A S S often to do.I N F . P A S S ‘One often tries to do such things.’ (Lødrup 2014: 371, example 13) Slike ting forsøkes ofte such things try.P R E S . P A S S often ‘One often tries to do such things.’

That long and double passive are constrained by restructuring is supported again by the type of matrix verbs that can occur in these constructions. The verbs attested in Lødrup’s corpus are given in (29), which by and large correspond to the verbs typically found to combine with tenseless complements. (29)

begynne ‘begin’, forsøke ‘try’, fortsette ‘continue’, fraraº de ‘advise against’, klare ‘manage’, lære ‘learn’, makte ‘manage’, nekte ‘deny’, plikte ‘have a duty to’, prøve ‘try’, paº legge ‘instruct’, slutte ‘stop’, starte ‘start’, tillate ‘allow’, tørre ‘dare’, taº le ‘stand, tolerate’, unngaº ‘avoid’, unnlate ‘neglect’, velge ‘choose’, ynde ‘like’, ønske ‘wish’ (Lødrup 2014: 369, example 7)

In conclusion, restructuring phenomena clearly exist in Swedish and Norwegian. In contrast to West Germanic, the evidence available so far mostly singles out the tenseless class of complements (only some Swedish speakers allow partial copying of matrix morphology into future complements, thus distinguishing all three types of complements). To see whether in general all three classes of infinitival complements are structurally distinguished in Swedish and Norwegian, further research is required.13

13

One possibility would be to investigate the processing of covert phenomena, for which restructuring hierarchy effects have been observed in English. In Wurmbrand (2018), for instance, it is suggested based on experimental evidence that quantifier raising in English shows sensitivity to the three-way split of complements: The higher a complement is on the transparency scale, the easier it is for an embedded quantifier to be interpreted in the matrix clause.

407

408

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

17.4 Restructuring ECM The findings of the last section have shown that at least two, in many languages three types of control infinitives are structurally distinguished, as diagnosed via diverse syntactic and morphological properties (possibly also processing differences, see footnote 14). To account for the different types of complements various options have been proposed in the literature. One approach is to tie the differences to different sizes of the complements, e.g., vP, TP, CP (see Wurmbrand 2001 et seq.). Alternatively, all complements could have uniform CP structures, but involve underspecified TP and/or CP projections (see Wiklund 2007), or some form of full or partial “pruning” of structure (Evers 1975; see Mu¨ller 2016 for a recent revival of structure removal). In this chapter, we will not go into detail about what exactly the structural differences are but simply use the descriptive labels [±max(imal)] and [±min(imal)] as in Table 17.5, which characterize restructuring properties (e.g., long passive is restricted to [+min] complements; splitting to [−max] contexts), as well as the types of complements. As we have seen, complements are characterized by a (possibly universal) minimum configuration (e.g., decide-complements cannot be [+min]) and may involve language-specific further options, as summarized in the last row of Table 17.5 for German and Dutch.14 As we have seen in Section 17.3, although there is evidence for the separation of the tenseless class in Swedish and Norwegian, the difference between the future class and the propositional class is less visible via restructuring diagnostics. Returning to the distribution of ECM, however, we seem to see an effect resembling the restructuring hierarchy, despite significant variation. As shown in Table 17.6, the least permissive languages, Dutch and German, only allow ECM in contexts that fall within the [+min] class (causatives and perception verbs take tenseless complements). Norwegian and Swedish are more permissive in allowing ECM in some [−min] contexts (with Table 17.5 Implicational restructuring hierarchy

Structure

Property Property A [+min] B [−max] Transparency

Config A Config B

✓ ✗

Config C ✗ Minimal structure Dutch, German

14

✓ ✓ ✗

Class I (try)

Class II (decide)

Class III (claim)

[+min] [−max] [−min] [−max]

most ✓ ↕ ✓ ✓ [−min] [+max] least ✓ ✓ ✓ [+min] [−min, −max] [+max] [±min] [−min]

[+max]

Note that the optionality of the tenseless class in Dutch and German is not a general property of restructuring. As shown in Wurmbrand (2016, 2017a), there are languages where tenseless complements are obligatorily [+min].

Germanic Infinitives

Table 17.6 ECM hierarchy ECM Structure

Lang Type I Du, Ge

Lang Type II No, Sw

Lang Type III E, Ice

Config A [+min] Config B [−min]

✓ *

✓ ✓

✓ ✓

Config C [+max: believe]

*

*



different restrictions and degrees of acceptability), but generally not in believe configurations. Lastly, English and Icelandic are the most permissive in allowing ECM also in clear [+max] contexts. A generally open question concerns the finer-grained details of the distribution, and, as noted in Section 17.2.1, the determination of possible ECM verbs within the three classes. For instance, in English, we have seen that ECM verbs are restricted to nonagentive verbs; however, in Icelandic, in addition to allowing believe-type ECM, the verb say/claim also allows (in fact requires) ECM. It also remains to be seen whether the two hierarchies in Table 17.5 and Table 17.6 can be unified (e.g., whether verbs like consider which allow ECM in Type II languages, albeit with additional restrictions, can be classified as [−max] since they are in between the clear tenseless and the strong believe classes). Lastly, if the ECM–restructuring connection exists, which the markedness hierarchy in (13) suggests, this raises interesting questions for the structure of ECM infinitives. Given that Configuration C is characterized as a full clausal configuration, the standard accounts of Type III ECM (involving reduced structures) would need to be re-thought, which, as noted in footnote 3, is also the case if English-type ECM is to be extended to languages which allow ECM into finite CP complements, see Wurmbrand 2017a for a preliminary direction.

References Bader, M. and T. Schmid 2009. “Minimality in verb-cluster formation,” Lingua 119.10: 1458–1481. Baltin, M. and L. Barrett 2002. The Null Content of Null Case. Ms., New York University. Basilico, D. 2003. “The topic of small clauses,” Linguistic Inquiry 34.1: 1–35. Besten, H. den and H. Broekhuis 1992. “Verb projection raising in het Nederlands,” Spektator 21: 21–34. Besten, H. den and J. A. Edmondson 1983. “The verbal complex in continental West Germanic.” In A. Werner (ed.), On the Formal Syntax of the Westgermania. Amsterdam: John Benjamins: 155–216.

409

410

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

Besten, H. den and J. Rutten 1989. “On verb raising, extraposition and free word order in Dutch.” In D. Jaspers, W. Klooster, Y. Putseys, and P. Seuren (eds.), Sentential Complementation and the Lexicon: Studies in Honour of Wim de Geest. Dordrecht: Foris: 41–56. Bobaljik, J. D. and S. Wurmbrand 2005. “The domain of agreement,” Natural Language and Linguistic Theory 23.4: 809–865. Bobaljik, J. D. and S. Wurmbrand 2008. “Case in GB / Minimalism.” In A. Malchukov and A. Spencer (eds.), Handbook of Case. Oxford University Press: 44–58. Bosˇkovic´, Ž. 1996. “Selection and the categorial status of Infinitival Complements,” Natural Language and Linguistic Theory 14: 269–304. Bosˇkovic´, Ž. 1997. The Syntax of Nonfinite Complementation: An Economy Approach. Cambridge, MA: MIT Press. Bresnan, J. 1972. Theory of Complementation in English. Doctoral dissertation, Cambridge, MA:MIT. Broekhuis, H. 1992. Chain-Government: Issues in Dutch Syntax. Doctoral dissertation, University of Amsterdam. Broekhuis, H. and N. Corver 2015. Syntax of Dutch: Verbs and Verb Phrases, Vol. 2. Amsterdam University Press. Cecchetto, C. 2004. “A challenge to Null Case theory,” Linguistic Inquiry 35.1: 141–149. Chomsky, N. 1980. “On binding,” Linguistic Inquiry 11: 1–46. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. and H. Lasnik 1993. “Principles and parameters theory.” In J. Jacobs, A. von Stechow, W. Sternefeld, and T. Vennemann (eds.), Syntax: An International Handbook of Contemporary Research. Berlin and New York: Mouton de Gruyter: 506–569. Chomsky, N. and H. Lasnik 1995. “Principles and parameters theory.” In N. Chomsky (ed.), The Minimalist Program. Cambridge, MA: MIT Press: 13–127. Christensen, K. R. 2007. “The infinitive marker across Scandinavian,” Nordlyd 34.1. Dikken, M. den and E. Hoekstra 1997. “Parasitic participles,” Linguistics 35: 1057–1089. Evers, A. 1975. The Transformational Cycle of Dutch and German. Doctoral dissertation, University of Utrecht. Haeseryn, W., K. Romijn, G. Geerts, J. De Rooij, and M. C. Van den Toorn 1997. Algemene Nederlandse spraakkunst. Groningen: Nijhoff. Holmberg, A. 1986. Word Order and Syntactic Features in the Scandinavian Languages and English. Doctoral dissertation, University of Stockholm. Holmberg, A. 2002. “Expletives and agreement in Scandinavian passives,” Journal of Comparative Germanic Linguistics 4.2: 85–128. Horn, S. W. 2008. Syntax, Semantics, and Pragmatics of Accusative-Quotative Constructions in Japanese. Doctoral dissertation, The Ohio State University.

Germanic Infinitives

Hornstein, N. 2003. “On control.” In R. Hendrick (ed.), Minimalist Syntax. Oxford: Blackwell: 6–81. Johnson, K. and S. Vikner 1994. “The position of the verb in Scandinavian infinitives: In V˚ or C˚ but not in I˚,” Working Papers in Scandinavian Syntax 53: 61–84. Landau, I. 2000. Elements of Control: Structure and Meaning in Infinitival Constructions. Dordrecht, Boston, and London: Kluwer Academic Publishers. Lødrup, H. 2002. “Infinitival complements in Norwegian and the form– function relation.” In M. Butt and T. Holloway King (eds.), Proceedings of the LFG02 Conference. Stanford, CA: CSLI Online. Lødrup, H. 2008. “Raising to object in Norwegian and the derived object constraint,” Studia Linguistica 62.2: 155–181. Lødrup, H. 2014. “Long passives in Norwegian: Evidence for complex predicates,” Nordic Journal of Linguistics 37.3: 367–391. Martin, R. A. 1996. A Minimalist Theory of PRO and Control. Doctoral dissertation, University of Connecticut, Storrs. Martin, R. A. 2001. “Null case and the distribution of PRO,” Linguistic Inquiry 32.1: 141–166. Mu¨ller, G. 2016. “Rethinking restructuring.” Ms., Universita¨t Leipzig. Pesetsky, D. 1992. “Zero syntax II: An essay on infinitives.” Ms., Cambridge, MA: MIT. Platzack, C. 1986. “The structure of infinitive clauses in Danish and Swedish,” Scandinavian Syntax: 123–137. Postal, P. 1974. On Raising: One Rule of English and Its Theoretical Implications. Cambridge, MA: MIT Press. Rutten, J. 1991. Infinitival Complements and Auxiliaries. Doctoral dissertation, University of Amsterdam. Sigurðsson, H. A´. 1991. “Icelandic case-marked PRO and the licensing of lexical arguments,” Natural Language and Linguistic Theory 9: 327–364. Stowell, T. 1982. “The tense of infinitives,” Linguistic Inquiry 13:561–570. Thra´insson, H. 1993. “On the structure of infinitival complements.” In H. Thra´insson, S. D. Epstein, and S. Kuno (eds.), Harvard Working Papers in Linguistics 3. Cambridge, MA: Harvard University, Department of Linguistics: 181–213. Urk, C. van 2013. “Visser’s Generalization: The syntax of control and the passive,” Linguistic Inquiry 44.1: 168–178. Wiklund, A.-L. 2001. “Dressing up for vocabulary insertion: The parasitic supine,” Natural Language and Linguistic Theory 19.1: 199–228. Wiklund, A.-L. 2005. The Syntax of Tenselessness: On Copying Constructions in Swedish. Doctoral dissertation, Umeaº University. Wiklund, A.-L. 2007. The Syntax of Tenselessness: Tense / Mood / Aspect-Agreeing Infinitivals. Berlin: Mouton de Gruyter. Wurmbrand, S. 2001. Infinitives: Restructuring and Clause Structure. Berlin and New York: Mouton de Gruyter.

411

412

S U S I WU R M B R A N D AN D CH R I STO S CH R I STO P O U LO S

Wurmbrand, S. 2006. “Verb clusters, verb raising, and restructuring.” In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax, Vol. V. Oxford: Blackwell: 227–341. Wurmbrand, S. 2007. “How complex are complex predicates?” Syntax 10:243–288. Wurmbrand, S. 2012. “Parasitic participles in Germanic: Evidence for the theory of verb clusters,” Taal en Tongval 64.1: 129–156. Wurmbrand, S. 2014a. “Restructuring across the world.” In L. Veselovska´ and M. Janebova´ (eds.), Complex Visibles Out There. Proceedings of the Olomouc Linguistics Colloquium 2014: Language Use and Linguistic Structure. Olomouc: Palacky´ University: 275–294. Wurmbrand, S. 2014b. “Restructuring across the world – Data summary.” http://wurmbrand.uconn.edu/Papers/RATW-data.pdf. Wurmbrand, S. 2014c. “Tense and aspect in English infinitives,” Linguistic Inquiry 45.3: 403–447. Wurmbrand, S. 2015a. “Complex predicate formation via voice incorporation.” In L. Nash and P. Samvelian (eds.), Approaches to Complex Predicates. Leiden: Brill: 248–290. Wurmbrand, S. 2015b. “Restructuring cross-linguistically.” In T. Bui and D. ¨ zyıldız (eds.), Proceedings of the North Eastern Linguistics Society Annual O Meeting 45. Amherst: University of Massachusetts, GLSA: 227–240. Wurmbrand, S. 2016. “Restructuring as the Regulator of Clause Size.” Invited plenary talk at the workshop Shrinking trees, Universita¨t Leipzig. Wurmbrand, S. 2017a. Discoveries and Puzzles in Syntax: ECM and Restructuring. Lecture series given at the Institute of Linguistics, Cognition and Culture (NYI), St. Petersburg, Russia. Wurmbrand, S. 2017b. “Verb clusters, verb raising, and restructuring.” In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax, 2nd edn. Oxford: Wiley-Blackwell. Wurmbrand, S. 2018. “The cost of raising quantifiers,” Glossa 3.1: 19. Wurmbrand, S. and K. Shimamura 2017. “The features of the voice domain: Actives, passives, and restructuring.” In R. D’Alessandro, I. Franco, and A´. Gallego (eds.), The Verbal Domain. Oxford University Press: 179–204. Yoon, J. H. 2007. “Raising of major arguments in Korean and Japanese,” Natural Language & Linguistic Theory 25.3: 615–653. Zwart, C. J-W. 1995. “A note on verb clusters in the Stellingwerf dialect.” In M. den Dikken and K. Hengeveld (eds.), Linguistics in the Netherlands 199. Amsterdam: John Benjamins: 215–226. Zwart, C. J-W. 1996. “Verb clusters in continental West Germanic dialects.” In J. Black and V. Motapanyane (eds.), Microparametric Syntax and Dialect Variation. Amsterdam and Philadelphia: John Benjamins: 229–258.

Chapter 18 The Unification of Object Shift and Object Scrambling Hans Broekhuis

18.1 Introduction The examples in (1) illustrate for Icelandic and Dutch that in specific Germanic languages the order of nominal objects and clause adverbials is relatively free; the angled brackets indicate the alternative positions of the direct object. I will follow the linguistic literature in referring to this phenomenon in the Scandinavian and West-Germanic languages as object shift and object scrambling (henceforth: OS and scrambling), respectively.1 Note that we will see later that the notion of scrambling is actually too wide, as it is normally used to refer to a set of rules; see, e.g., Mahajan (1994), Neeleman (1994a), and Haider and Rosengren (1998). (1)

Jo´n las ekki . Jo´n read this book not ‘Jo´n didn’t read this book.’ b. Jan las waarschijnlijk . Jan read this book probably ‘Jan probably read this book.’

a.

[OS]

[scrambling]

This review of OS and scrambling takes a different point of departure than earlier ones like Thra´insson (2001) and Vikner (2006) in that it adopts as a null hypothesis that the two phenomena should be given a unified treatment. More specifically, I will assume that we are dealing with leftward movement of the direct object into some, yet to be determined, designated I am indebted to Frits Beukema for proofreading an earlier version of this chapter; his comments have led to a large number of clarifications and other improvements. I thank Michael T. Putnam for fine-tuning the final version. 1

Dutch and German differ from the Scandinavian languages (and English) in that the subject moves into the subject position (SpecTP) only optionally. The restrictions on subject movement are similar to those on object scrambling so that we may conclude that scrambling may also apply to subjects; due to space limitations I have to refer the reader to, e.g., Haider and Rosengren (1998) and Broekhuis (2007, 2008: section 4.2).

414

HANS BROEKHUIS

position external to the lexical projection of the verb, vP in the current version of mainstream generative grammar. (2)

a. . . . ADV [vP . . . [VP . . . object . . .]] b. . . . objecti . . . ADV [vP . . . [VP . . . ti . . .]]

The derivation sketched in (2b) seems to be fairly standard for OS, although the details of the analyses proposed since Holmberg (1986) differ considerably; see Thra´insson (2001), Vikner (2006), and Engels and Vikner (2014: chapter 1) for details. Extending the derivation of OS in (2b) to scrambling, however, is highly controversial because there is hardly any agreement on the analysis of scrambling in the literature; see the contributions in Grewendorf and Sternefeld (1990) and Corver and Van Riemsdijk (1994) as well as the reviews in, e.g., Haider (2000, 2006), Chocano (2007: chapter 3), and Putnam (2007: chapter 3). However, the unification of OS and scrambling has been advocated in more recent works like Broekhuis (2000, 2008), Chocano (2007: chapter 4), and Engels and Vikner (2014). This chapter discusses some pros and cons of unification and explores the hypothesis in (2) in more detail.

18.2 Effects on Output One reason for unifying OS and scrambling is that they have similar effects on the output: shifted/scrambled objects cannot be part of the information focus (discourse-new information) of the clause and they cannot be assigned noncontrastive accent; cf., e.g., Verhagen (1986), Diesing (1997), and Holmberg (1999). (3)

Effect on output: a. A shifted/scrambled object is part of the presupposition of the clause. b. A shifted/scrambled object cannot be assigned neutral clause accent.

The information-structural effect (3a) can be readily accounted for by assuming that the information focus is prototypically located within the lexical domain of the clause (vP), while the presupposition (discourse-old information) is preferably located in the functional domain. That the postadverbial objects in (1) are indeed construed as part of the focus is clear from the fact that clauses with this order can be used as answers to questions like “What happened?”, and that the preadverbial objects are construed as part of the presupposition is clear from the fact that the clauses with this order would rather be construed as answers to questions like “What happened to the book?”. The prosodic effect in (3b) can be made to follow from Cinque’s (1993) hypothesis that neutral clause accent is assigned to the most deeply

Object Shift and Object Scrambling

embedded phrase in the clause: If the object occupies its base position within vP, as in (2a), it counts as the most deeply embedded phrase and, consequently, it must be assigned clause accent; if it is moved into the functional domain, it no longer counts as the most deeply embedded phrase and the accent will therefore shift to some other phrase that counts as most deeply embedded in the derived structure. That the effects in (3) can be derived from the claim that the object is moved out of the lexical domain of the clause can be seen as support for a movement approach, especially since Haider (2000, 2006) argues that a similar simple account of (3) is not available in base-generation approaches like Bayer and Kornfilt (1994), Neeleman (1994a, 1994b), Fanselow (2001), and Sells (2001).2

18.3 Pronominal Versus Nonpronominal Noun Phrases The hypothesis that vP constitutes the focus domain of the clause also accounts for the fact illustrated in (4) that OS/scrambling cannot apply to nonspecific indefinite objects because these belong to the information focus by definition, while weak (unstressed / phonetically reduced) definite pronouns must undergo OS/scrambling because they are always presuppositional. Note that the hash signs in (4) are used to express that indefinites may occur in preadverbial position if they are interpreted specifically or generically, and that the asterisks are used to indicate that weak pronouns are always unacceptable in postadverbial position. (4)

a. Jo´n las ekki . Jo´n read them/books not b. Jan las waarschijnlijk . Jan read them/books probably

The data in (1) and (4) are only partly captured by the formulation of the hypothesis in (3a); it correctly predicts that the preadverbial objects in (1) are part of the presupposition of the clause but does not account for the fact that the postadverbial objects must be construed as part of the information focus; it also correctly predicts that the indefinite bare-plurals in (4) must follow the adverbial, but it does not account for the fact that the definite pronouns must be shifted/scrambled. Of course, this can be easily remedied by adding that presuppositional objects must be shifted, but this would clearly be an undesirable step in view of the fact that the Mainland Scandinavian languages categorically prohibit OS of nonpronominal 2

Fanselow (2003: section 2.4) maintains that this does not hold for his proposal, in which scrambling requires the main verb to undergo V-to-v. This correlation is difficult to test for OV-languages, however, because it entails that vP and VP are both head-final so that V-to-v always applies string-vacuously: if vP were head-initial, we would wrongly predict that scrambling forces movement of the clause-final verb in front of VP-adverbials and VP-internal material in embedded clauses.

415

416

HANS BROEKHUIS

objects. This is illustrated for Danish in (5), taken from Vikner (1994: 502): the definite object artiklen ‘the article’ occurs in postadverbial position regardless of whether it belongs to the focus or the presupposition of the clause, which shows that presuppositional objects can sometimes remain in the lexical domain of the verb. (5)

a. Hvorfor læste studenterne ikke ? why read the students the.article not b. Hvorfor læste studenterne ikke ? why read the students it not

The contrast in (5) raises another issue: Can leftward movement of nonpronominal and pronominal objects be treated on a par? It is clear that the two forms of object movement exhibit different properties in Dutch; example (6) shows that object pronouns normally occur right-adjacent to either the subject or the finite verb in subjectinitial main clauses, while scrambled nonpronominal objects may also occur in a lower, that is, more rightward position; see van Bergen and de Swart (2010). (6)

a. Dat Jan gisteren waarschijnlijk las. that Jan this book/it yesterday probably read ‘that Jan probably read this book/it yesterday.’ b. Jan las gisteren waarschijnlijk. Jan read this book/it yesterday probably ‘Jan probably read this book/it yesterday.’

This is a first illustration of the fact that scrambling is often used as a cover term for different movement types: it refers to the movement deriving (7b) from (7a), but in addition it refers to the movement in (7c) that obligatorily places object pronouns right-adjacent to the regular subject position, SpecTP. (7)

a. . . . ADV [vP . . . [VP . . . object . . .]] b. . . . objecti . . . ADV [vP . . . [VP . . . ti . . .]] c. . . . objecti . . . [ti . . . ADV [vP . . . [VP . . . ti . . .]]]

Scandinavian definite object pronouns are like the Dutch ones in that they also obligatorily occur right-adjacent to either the subject or the finite verb in subject-initial main clauses (Holmberg 1991), but they do not differ in this respect from shifted non-pronominal objects in Icelandic, which must occur in the same position. This is shown by the Icelandic examples in (8), taken from Vikner (1994, 2006). (8) I gær las Pe´tur < bo´kina/hana> eflaust ekki. yesterday read Pe´tur the.book/it doubtlessly not ‘Yesterday Pe´tur undoubtedly did not read the book.’

Object Shift and Object Scrambling

This suggests that OS and scrambling differ in that movement of weak object pronouns should be distinguished from other cases of object movement in the case of scrambling only (but see footnote 9 below). This can be further supported by means of the contrast between the Danish and Dutch double object constructions in (9) and (10). The (a)-examples show that in both languages the indirect object precedes the direct object if they occupy their base positions. Example (9b), taken from Vikner (1989), shows that OS of the corresponding pronouns does not affect this order, while (10b) shows that scrambling of the pronouns resembles Romance clitic movement in that it inverts the object order. (9)

(10)

a. Peter viste jo Marie bogen. Peter showed indeed Marie the.book b. Peter viste hende jo. Peter showed it her indeed

(IO>DO)

a. Jan geeft Jan gives b. Jan geeft Jan gives

(IO>DO)

waarschijnlijk Marie het boek. probably Marie the book ’r waarschijnlijk. it her probably

(IO>DO)

(DO>IO)

However, scrambling of pronominal objects cannot be fully assimilated with Romance clitic placement because these objects do not syntactically cliticize to a verbal head. It is in fact quite doubtful that they target any designated head position in the functional domain of the clause. If we adopt the claim discussed in footnote 2, that Dutch subjects do not obligatorily move into SpecTP but optionally remain in their base position within vP following the clause adverbials, the examples in (11b–11c) show that in-situ subjects block scrambling of weak pronouns; they do not obligatory move into some specific head position in the functional domain of the clause but seem to be phonologically supported by the subject; see Broekhuis and Corver (2016: chapter 13), and Haider and Rosengren (1998: section 5.4) for similar data from German. (11)

a. Dat Jan ’t ’r waarschijnlijk gaf. That Jan it her probably gave ‘that Jan probably gave it her.’ b. dat waarschijnlijk Jan ’t ’r gaf. c. *dat ’t ’r waarschijnlijk Jan gaf.

I conclude from the discussion above that while there are good reasons for assuming that scrambling of weak object pronouns differs syntactically from other forms of object scrambling, there is no reason to make a similar distinction in the case of OS. This raises the question as to why the Mainland Scandinavian languages prohibit nonpronominal OS. It is not a priori clear that this question should receive a syntactic answer, given that there is also variation in mainland Scandinavian with respect to

417

418

HANS BROEKHUIS

pronominal OS: Holmberg (1986, 1999) observes that while this is obligatory in Danish, it is optional in most varieties of Swedish and excluded in Finnish-Swedish, without there being any obvious syntactic property to which this variation can be attributed.

18.4 Verb Movement Section 18.3 has shown that the formulation of the information-structural effect on the output in (3a) is consistent with the fact that the Mainland Scandinavian languages do not allow nonpronominal OS. It is also needed to account for the fact that even Icelandic definite objects can be interpreted as part of the presupposition if OS is blocked for independent reasons. For instance, OS is categorically blocked if the main verb remains within its lexical projection, as in perfect-tense constructions such as Icelandic (12), taken from Holmberg (1986) and Thra´insson (2001); (3a) correctly predicts that the definite noun phrase / pronoun can be presuppositional in such cases. (12)

a. Jo´n hefur ekki keypt . Jo´n has the.book not bought b. Nemendurnir hafa ekki lesið . the.students have it not read

The fact that OS cannot cross the participle in (12) is a special case of a more general condition stating that the canonical word order within the lexical projection of the verb cannot be affected by OS, formulated in Holmberg (1999) as in (13). We will only discuss verbs here but return to the case of arguments in Section 18.6. (13)

Holmberg’s Generalization (HG): Object shift cannot apply across a phonologically visible category within VP: verbs, arguments, particles, etc. (but excluding adverbials).

HG also accounts for the contrast between the Icelandic and Danish examples in (14), taken from Vikner (1994). The position of the finite verb relative to negation reveals that while Icelandic requires leftward movement of the finite verb in embedded clauses (V-to-I), the finite verb remains vP-internal in Danish; HG therefore correctly predicts that while Icelandic allows OS in embedded clauses, as in (14a), this is blocked by the main verb in Danish, as in (14b). (14)

a. að hann keypti

ekki that he bought the book not ‘that he didn’t buy the book.’ b. at Peter ikke købte . that Peter it not bought ‘that Peter didn’t buy it.’

.

Object Shift and Object Scrambling

HG exploits the fact that OS should cross specific vP-internal material preceding the objects such as the main verb in perfect-tense constructions or Danish embedded clauses. If HG were a universal, this would correctly predict that scrambling is possible in perfect-tense constructions as well as embedded clauses in OV-languages like Dutch, where the main verb follows the objects in the canonical word order. (15) a. Jan heeft waarschijnlijk gekocht. Jan has the book/it probably bought b. dat Jan waarschijnlijk < het boek/*het> koopt. that Jan the book/it probably buys The assumption that HG-effects depend on the VO/OV-status of the language has given rise to the claim that OS is found in VO-languages only, while scrambling is only found in OV-languages. This neat picture is disturbed by the fact that Yiddish seems to allow object movement across nonfinite verbs. The judgments are similar to the ones given earlier for Icelandic and Dutch: Definite noun phrases move leftward (into preadverbial position) depending on whether they belong to the focus or the presupposition of the clause, nonspecific indefinites do not allow movement, and weak definite pronouns must be moved; see Diesing (1997: section 5.1). (16)

a. Maks hot Maks has b. Maks hot Maks has c. Maks hot Maks has

mistome nit the book probably not mistome nit a book probably not gekent . us known

geleyent . read geleyent . read

The data in (16) revived the discussion initiated in den Besten and Moedvan Walraven (1986) whether Yiddish should be considered a VO or a OVlanguage; Diesing (1997) advocates a VO-analysis and claims that Yiddish should be seen as an exception to the rule that scrambling occurs in OVlanguages only, while Haider and Rosengren (1998) argue that Yiddish is an OV-language but obligatorily moves the nonfinite verb to the left of the objects.3 The outcome of this debate seems less relevant to me: What is crucial is that (16) shows that the canonical order of the main verb and direct object is VO and that leftward object movement therefore violates HG in Yiddish. If correct, this entails that HG is a violable constraint and that attempts to derive it from inviolable conditions on movement such as the equidistance principle in Chomsky (1995: chapter 3) are doomed to fail. The conclusion that HG is not a universal has as a desirable effect that it 3

Broekhuis (2008: chapter 2) argues that participle movement is a property of the Germanic languages in general; Yiddish may actually be an exception to this rule. Be this as it may, it seems that Haider and Rosengren’s proposal does not solve the problem: The participle follows negation, which demarcates the boundary between the functional and the lexical domain of the clause, so that we must conclude that the participle remains part of the lexical domain of the clause.

419

420

HANS BROEKHUIS

allows object movement in examples like (15) and (16a and 16c) without the need of postulating “invisible” movement of formal verbal features, as in Zwart (1997), or the phonetically empty light verb v, as in Broekhuis (2000).

18.5 Remnant VP-Topicalization The conclusion that HG is not an inviolable condition on derivations entails that it is a language-specific constraint on output representations. This claim is empirically supported by the fact that OS is possible in Remnant VP-topicalization constructions; cf. Fox and Pesetsky (2005) and Engels and Vikner (2013, 2014). Consider the Swedish examples in (17a– 17b), taken from Holmberg (1999) and Engels and Vikner (2014). (17)

a. Jag har

inte kysst . I have her not kissed b. Kysst har jag inte . kissed have I her not

Holmberg claims that (17b) is not derived by Remnant VP-topicalization but by V-topicalization followed by OS. However, if both of these movements were syntactic, OS should precede V-topicalization and this would make the derivation anticyclic. This makes Holmberg conclude that OS is not a syntactic but a stylistic rule that applies after the syntactic derivation has concluded. A problem for this proposal pointed out by Engels and Vikner is that OS is obligatory in (17b) while it is normally optional for Swedish speakers. This problem does not arise for Fox and Pesetsky, who derive (17b) by means of Remnant VP-topicalization, in accordance with the standard assumption that the sentence-initial position must be occupied by a phrase (and not a head). Their derivation proceeds as in (18), with the crucial steps of OS in (18b) and VP-topicalization in (18d); I placed the auxiliary in v only for concreteness’ sake as I believe that the actual structure of perfect-tense constructions is slightly more complex than suggested here. The crucial thing is that the unacceptable order in (17b) cannot be derived: VP-topicalization with OS must strand the object in preadverbial position, while VP-topicalization without OS results in obligatory pied piping of the object, as in the fully acceptable sentence [Kysst henne] har jag inte tVP, taken from Holmberg (1999: 7). (18)

a. b. c. d.

inte [vP jag har [VP kysst henne]] henne inte [vP jag har [VP kysst thenne]] [TP jag har [henne inte [vP tjag thar [VP kysst thenne]]]] [CP [VP kysst thenne] har [TP jag thar henne inte [vP tjag thar tVP]]]

Object Shift and Object Scrambling

Table 18.1 Linearization of VO-languages OS

V-to-T/C

Structure

VP

CP

— —

— +

.. ADV [vP .. [VP Vmain O]] [CP .. Vmain .. ADV [vP .. [VP tV O]]]]

V > O V > O

V > O V > O

+ +

— +

*.. O .. ADV [vP .. [VP Vmain tO]] [CP .. Vmain .. O ADV [vP .. [VP tV tO]]]]

V > O V > O

O > V V > O

Fox and Pesetsky propose that HG results from the conditions on linearization of syntactic structures in (19), where at least VP and CP are considered Spell-out domains. (19)

a. The relative ordering of words is fixed at the end of each Spellout domain. b. Ordering established in an earlier phase may not be revised or contradicted in a later Spell-out domain.

Table 18.1, in which the two final columns provide the relative orders of the main verb and the object at the relevant Spell-out domains, shows that (19) provides the correct prediction for OS. The first two rows show that movement of the main verb to T or C is independent of OS because it does not change the linearization at the CP-level. The last two rows show, however, that OS crucially depends on verb movement: OS inverts the linearization at the CP-level, but this can be “repaired” by subsequent V-to-T/C. The acceptability of VP-topicalization constructions such as (17b) with OS follows from the linearization approach because VP-topicalization has the same effect as V-to-T/C: (20b) shows again that OS normally results in a violation of (19b), while (20c) shows that this violation is “repaired” by VP-topicalization. (20)

a. Jag har inte [VP kysst henne]. (V>O;V>O) b. *Jag har henne inte [VP kysst thenne] (V>O;O>V) c. [VP kysst thenne] har jag henne tVP (V>O;V>O)

The linearization approach provides the novel prediction that VPtopicalization is only possible if the object is base-generated in the right periphery of the VP; if not, VP-topicalization will not be able to restore the contradictory ordering resulting from OS. Consider the Danish example in (21a) from Engels and Vikner (2013, 2014): since the object is followed by a predicative PP, the ordering at the VP-level is V>O>PP. VP-topicalization of the full VP, as in (21b), is allowed as it results in the same ordering at the CP-level, but Remnant VP-topicalization in (21c) is excluded because it results in the contradictory ordering V>PP>O.

421

422

HANS BROEKHUIS

Table 18.2 Linearization of OV-languages OS

V-to-T/C

Structure

VP

CP

— —

— +

.. ADV [vP .. [VP O Vmain]] *[CP .. Vmain [.. ADV [vP .. [VP O tV]]]]

O > V O > V

O > V V > O

+ +

— +

.. O ADV [vP .. [VP tO Vmain]] *[CP .. Vmain .. O ADV [vP .. [VP tO tV]]]]

O > V O > V

O > V V > O

(21)

a. Jeg har ikke [VP stillet det paº bordet]. I have not put it on the.table ‘I haven’t put it on the table.’ b. [VP Stillet det paº bordet] har jeg ikke tVP. c. *[VP Stillet tdet paº bordet] har jeg det ikke tVP

Despite its obvious description success in the case of OS, the linearization approach faces severe problems with OV-languages with V-second. Table 18.2 shows that (19) wrongly predicts that V-movement is categorically disallowed because it results in a contradictory ordering statement (regardless of the position of the object); see Mu¨ller (2007).4 Since VP-topicalization has the same effect as V-to-T/C, it is also incorrectly predicted that it is incompatible with scrambling: The Dutch example in (22c) is fully acceptable despite the contradictory orderings at the VP and CP-level. (22) a. Jan heeft vaak [VP geholpen]. Jan has Marie often helped ‘Jan has often helped Marie.’ b. [VP Marie geholpen] heeft Jan vaak. c. [VP tMarie geholpen] heeft Jan Marie vaak.

(O>V;O>V)

(O>V;O>V) (O>V;V>O)

The failure of the linearization approach can be attributed to the presumed fact that (19) is universal; it suggests an optimality-theoretic approach to HG.5 Broekhuis (2008), for instance, proposes the order preservation constraint H - C O M P L in (23), which favors preservation of the universal underlying base order verb-object, predicted by Kayne’s (1994) Linear Correspondence Axiom. (23)

4

H E A D - C O M P L E M E N T (H - C O M P L ) : A head precedes all terminals originally dominated by its complement.

Müller’s revision of Fox and Pesetsky’s proposal is claimed to allow V-second in main clauses without scrambling, as in his example (47), which corresponds to the second representation in Table 18.2. Unfortunately, he does not discuss similar examples with scrambling (corresponding to the fourth representation), which still seem problematical for his approach.

5

Optimality Theory (OT) defines different languages by postulating a universal set of violable constraints {A, B, C, . . .}, that are ranked in a language-specific manner. If A outranks B, and A and B impose conflicting requirements on the output of the grammar, an output that satisfies A but violates B is preferred to an output that satisfies B but violates A. The ranking A≫B thus gives rise to a different language than B≫A. See Archangeli (1997) for more details.

Object Shift and Object Scrambling

This violable OT-constraint is argued to be ranked relatively high in the Germanic VO-languages but fairly low in the Germanic OV-languages, so that its effects will be readily observable in the former only. Broekhuis argues that this constraint accounts for two things: (1) a high ranking of H - C O M P L results in an unmarked VO-order when the verb remains within its lexical domain, while a low ranking results in an unmarked OV-order; (2) languages with a high ranking of H - C O M P L will exhibit HG-effects while languages with a low ranking will not. We can now add that the ranking of this constraint also accounts for the VP-topicalization examples discussed in this section: A high ranking of H - C O M P L ensures that VP-topicalization preserves the unmarked word within VP, while a low ranking allows this order to be changed.6

18.6 Argument Order Preservation HG, in tandem with the standard assumption that a nominal indirect object (IO) canonically precedes a direct object (DO), correctly predicts that OS of the latter cannot cross the former. This is illustrated in (24) for Icelandic; the objects may both stay in postadjectival position, they can both shift, the IO may shift alone but this is impossible for the DO. The examples in (25) show that the same holds for scrambling in Dutch. (24)

a. Pe´tur sy´ndi oft Marı´uIO bo´kinaDO. Pe´tur showed often Marı´u the book b. Pe´tur sy´ndi Marı´u oft tIO bo´kina. c. Pe´tur sy´ndi Marı´u bo´kina oft tIO tDO. d. *Pe´tur sy´ndi bo´kina oft Marı´u tDO.

(25)

a. Dat Peter waarschijnlijk Marie het boek getoond heeft. that Peter probably Marie the book shown has b. dat Peter Marie waarschijnlijk tIO het boek getoond heeft. c. dat Peter Marie het boek waarschijnlijk tIO tDO getoond heeft. d. *dat Peter het boek waarschijnlijk Marie tDO getoond heeft.

The same can be shown for cases with a pronominal DO although the effect on the output is radically different in this case: OS of the 6

Space limitations bar me from providing the formal proof. Broekhuis (2008) did not discuss languages such as Yiddish with an unmarked VO-order, which nevertheless do not exhibit HG-effects. For readers familiar with my work, it suffices to say that Yiddish can be derived by means of a specific ranking of H - C O M P L and the E P P -constraints E P P ( φ) and E P P (case). First, Broekhuis (2008: chapter 2) established that the relative ranking of H - C O M P L and E P P ( φ) determines whether we are dealing with a VO or an OV-language (provided, of course, that the ranking of the economy constraint S T A Y allows movement). Second, Broekhuis (2008: chapter 3) showed that the ranking in (ii) is responsible for the HG-effects found in the Scandinavian languages; the alternative ranking in (iii) therefore derives OV-languages without HG-effects. (i) a. Unmarked OV order: E P P (φ) ≫ H - C O M P L b. Unmarked VO order: H - C O M P L ≫ E P P ( φ) (ii) HG-effects: H - C O M P L ≫ E P P (case) (iii) no HG-effects: E P P (case) ≫ H - C O M P L

423

424

HANS BROEKHUIS

DO-pronoun must apply and seems to push the IO up into preadverbial position regardless of its informational-structural status; see Matthews (2000: 158–159) and Broekhuis (2007, 2008: 171). Judgments on OS in ditransitive constructions are unfortunately not always clear, as is evident by comparing Rustick (1991: section 3.1.2) and Vikner (1989), but the crucial thing is that the order in (26c) is clearly the preferred one. (26)

a. *Pe´tur sy´ndi oft Marı´u hana. Pe´tur showed often Marı´u it b. *Pe´tur sy´ndi Marı´u oft tIO hana. c. Pe´tur sy´ndi Marı´u hana oft tIO tDO. d. *Pe´tur sy´ndi hana oft Marı´u tDO.

The examples in (27) provide similar cases for Dutch; recall that the judgments only hold for noncontrastive intonation patterns and that the DO-pronoun may also precede the scrambled IO in (27c) as a result of weak pro-form shift. (27)

a. * dat Peter waarschijnlijk Marie het getoond heeft. that Peter probably Marie it shown has b. *dat Peter Marie waarschijnlijk tIO het getoond heeft. c. dat Peter Marie het waarschijnlijk tIO tDO getoond heeft. d. *dat Peter het waarschijnlijk Marie tDO getoond heeft.

The Dutch examples are important because they show that HG is not exclusively applicable to the Scandinavian languages but at least partly also to specific scrambling constructions. We can account for this by assuming that Icelandic and Dutch both obey constraint (28); cf., e.g., Mu¨ller (2000, 2001), Williams (2003), and Broekhuis (2008). (28)

D E F A U L T A R G U M E N T O R D E R P R E S E R V A T I O N (D A O P ): OS/scrambling does

not affect the underlying order of arguments. Like H - C O M P L , D A O P cannot be a locality condition on derivations because standard assumptions on movement ensure that the (c)-examples in (25)–(27) are derived by first moving the DO across the IO and subsequent movement of the IO across the derived position of the DO. VP-topicalization provides support for assuming that D A O P is a constraint on representations; the Danish examples in (29), taken from Engels and Vikner, show that VPtopicalization may pied pipe the two pronominal objects, the IO without the DO but not the DO without the IO, or strand the two objects (the judgments are originally due to Anders Holmberg). (29)

a. [VP Givet hende den] har jeg ikke tVP. given her it have I not b. ?[VP Givet hende tden] har jeg den ikke tVP. c. *[VP Givet thende den] har jeg hende ikke tVP. d. [VP Givet thende tden] har jeg hende den ikke tVP.

Object Shift and Object Scrambling

The Danish data in (29) are somewhat surprising because pronominal OS is suddenly optional, but the crucial fact is the contrast between the two examples in (29b) and (29c): example (29b), in which the DO has been shifted across the IO, is markedly better than (29c), in which OS does not involve crossing. Example (29b) illustrates once more that remnant VP-topicalization may repair a violation of D A O P created by DO shift, but (29c) shows that it may in fact also invoke a violation of D A O P in a structure that would otherwise satisfy it. The Dutch examples in (30) illustrate essentially the same as (29): although (30b) and (30c) are clearly marked compared to the two other examples, there is a clear difference in acceptability between them.7 (30)

a. [ V P Marie het boek gegeven] heeft hij zeker. Marie the book given has he certainly b. ? [ V P Marie thet boek gegeven] heeft hij het boek zeker. c. *[ V P tMarie het boek gegeven] heeft hij Marie zeker. d. [ V P tMarie thet boek gegeven] heeft hij Marie het boek zeker.

Fox and Pesetsky’s linearization approach and Engels and Vikner’s OTapproach differ in that the former takes (28) to be inviolable and thus predicts that OS/scrambling should respect D A O P in all languages, while the latter allows cross-linguistic variation here. That the OT-approach is to be preferred is clear from the fact illustrated in (31d) and (32d) that German and Yiddish allow DO to move across IO in violation of (28); see Vikner (1994), Diesing (1997), Haider and Rosengren (1998, section 7.3.2), Fanselow (2003), and many others. (31)

a. Dass Peter wirklich Maria das Buch gezeigt hat. that Peter probably Maria the book shown has b. dass Peter Maria wirklich tIO das Buch gezeigt hat. c. dass Peter Maria das Buch wirklich tIO tDO gezeigt hat. d. dass Peter das Buch wirklich Maria tDO gezeigt hat.

(32)

a. Maks hat nit gegebn Rifken dos bukh. Maks has not given Rifken the book b. Maks hat Rifken nit gegebn dos bukh. c. Maks hat Rifken dos bukh nit gegebn. d. Maks hat dos bukh nit gegebn Rifken.

Engels and Vikner suggest that the constraint H - C O M P L in (23) and the constraint D A O P in (28) can be unified into the more general constraint O R D E R P R E S E R V A T I O N ; however, this is clearly not possible if one would like to give a unified account of OS and scrambling: the fact that Dutch is insensitive to the default order of the verb and its object while it is sensitive to the default order of arguments shows that the constraints are both needed. 7

Order preservation constraints are normally also applicable to subjects, and favor preservation of the order S>IO>DO. In order to evaluate the proposal in Engels and Vikner (2013, 2014) in full, more discussion will be necessary to account for the fact that VP-topicalization affects this order in (29a) and (30a), while OS and Dutch scrambling normally respect it.

425

426

HANS BROEKHUIS

18.7 Categorial Restrictions This section discusses the kind of movement OS/scrambling instantiates. I adopt the standard assumption that two main types of XP-movement can be distinguished: A and A′-movement. OS is generally taken to be A-movement because it affects nominal phrases only: PP-arguments, selected AP/PPpredicates and clauses do not shift.8 This is illustrated by the Icelandic examples in (33), taken from Vikner (2006: 404). (33)

a. E´g borgaði ekki . I paid for it not b. Pe´tur er aldrei . Pe´tur is ill never

That OS cannot be applied to object clauses strongly suggests that the A-movements involved are triggered by features responsible for structural case assignment. This conclusion has various controversial consequences. First, it implies that dative case is a structural case, which was independently concluded in Broekhuis and Cornips (1994, 2012). Second, phrases lexically case-marked with e.g., genitive or dative case must also be assigned structural case since such noun phrases can undergo OS, as is shown by the Icelandic examples in (34), adapted in a slightly simplified form from Engels and Vikner (2013: 20). That lexically case-marked phrases are also assigned structural case has been proposed on independent grounds in Jo´nsson (1996: 146) and Chomsky (2000: 127) for quirky subjects. (34) a. Pe´tur leitiða sennilega ekki . Pe´tur looked.for this bookgen probably not b. Pe´tur ly´sti sennilega ekki . not Pe´tur described this bookdat probably If OS does target the specifier position of a head responsible for structural case assignment, we would expect to find similar movements in other, typologically different, languages; the conjecture that OS and scrambling are instantiations of the same movement type should therefore be considered the null hypothesis. The null hypothesis is problematical, however, as it is generally assumed that scrambling is not restricted to noun phrases; while predicative complements of the verb such as groen in (35b) are wellbehaved with respect to the hypothesis in that they resist scrambling, PPobjects such as naar Peter in (35a) seem to refute it by allowing scrambling. 8

I did not find any examples specifically illustrating the ban on clausal OS. Locational adverbial pro-forms meaning “here” and “there” constitute an exception to this generalization. Since Icelandic does not allow shift of complex locational adverbials, Vikner (2006: 421–422) argues that the exceptional behavior of the adverbial pro-forms supports the cliticization approach to pronominal object shift discussed in Section 18.3, especially because the corresponding French pro-form y ‘there’ evidently behaves as a clitic as well.

Object Shift and Object Scrambling

(35) a. Dat Jan nauwelijks luisterde. that Jan to Peter hardly listened ‘that Jan hardly listened to Peter.’ b. dat Jan de deur waarschijnlijk verft that Jan the door green probably paints ‘that Jan will probably paint the door green.’ If we want to maintain the null hypothesis, we must conclude that the movement placing the PP in preadverbial position cannot be A-movement. This conclusion is not bizarre since it is widely appreciated that the clauseinternal order can be affected by various movement operations that should be distinguished from scrambling of the sort discussed so far; see Neeleman (1994a, 1994b), Haider and Rosengren (1998), Mu¨ller (1998), and references cited there. Broekhuis and Corver (2016, chapter13) propose that minimally the three main types in (36) should be distinguished in Dutch; recall that weak pro-form movement was already discussed in Section 18.3. (36)

a. A-movement: object shift and scrambling b. A′-movement: Negation movement; contrastive focus / topic movement c. Weak pro-form movement

Object shift / scrambling should be seen as a specific instantiation of Amovement. If the movement of the PP in (35a) is indeed of a different kind, it should instantiate some type of clause-internal A′-movement such as contrastive focus / topic movement. Before illustrating that this might really be the case, I will briefly discuss the clause-internal A′-movements in (36b): The main conclusion will be that they resemble wh-movement in that they do not exhibit HG-effects. That wh-movement does not exhibit HG-effects is clear from the fact that the Scandinavian languages allow it to cross the main verb, as illustrated by the Danish example in (37) from Jo´nsson (1996: 17). (37)

Hvilken which

film har movie have

børnene set? the.children seen

The same holds for Neg-movement, as is illustrated for Danish by (38a): The negative direct object obligatorily shifts leftward while violating the constraints on OS: It violates H - C O M P L in ( 23) because the direct object precedes the main verb and D A O P in (28) because the direct object precedes the indirect object. Example (38b) shows that pronominal OS restores the base-order of the arguments due to the fact that it clearly targets a structurally higher position than the negative phrase: While shifted objects precede clause adverbials such as faktisch negative phrases must follow them; (38b) thus clearly establishes that OS and Neg-movement target different positions and confirm the standard assumption that Neg-movement is

427

428

HANS BROEKHUIS

A´-movement. The examples in (38) are taken from Christensen (2005: 163–165). (38)

. a. Jeg har

laº nt hende I have no books lent her ‘I haven’t lent her any books. b. Jeg laº nte hende faktisch ingen bøger I lent her actually no books

Neg-movement is less easy to illustrate in OV-languages such as Dutch because it normally does not cross other vP-internal material such as the verb, but Haegeman (1995: 130–131) has shown that Neg-movement is obligatory in such languages by means of PP-complement of predicative adjectives: Example (39a) shows that while such PP-complements may normally precede or follow the adjective, they obligatorily precede the adjective if they are negative. Example (39b) serves the same purpose as (38b) by showing that scrambling targets a higher position than Negmovement. (39)

a. Jan is erg trots . Jan is of no.one/him very proud ‘Jan is quite proud of Peter/not quite proud of anyone.’ b. Jan heeft Marie waarschijnlijk geen boeken geleend. Jan has Marie probably no books lent

The examples in (40) show that contrastive foci can also be moved into a position following the clause adverb, while contrastive topics are rather placed in front of them; the two cases differ in meaning and intonation (indicated by “/” for rising and “\” for falling) but for reasons of space I cannot discuss this here; see, e.g., Haider and Rosengren (1998, section 6), Molna´rfi (2003), Neeleman and van de Koot (2008) or Broekhuis and Corver (2016, section 13.3) for relevant discussion. (40) a. dat Jan waarschijnlijk op P E T E R \ erg trots is. [focus-movement] that Jan probably of Peter very proud is b. dat Jan op /P E T E R waarschijnlijk erg trots is. [topic-movement] that Jan of Peter probably very proud is The crucial thing for our present purpose is that the nominal part of the PP op Peter cannot be a weak pronoun due to the fact that it must be accented. Let us now return to example (35a), repeated here as (41a), which motivated the claim that PPs may be scrambled as well. Neeleman (1994a) argues that we are not dealing with contrastive focus / topic movement because the nominal part of the PP need not be accented. Example (41b) shows, however, that PP-movement is excluded if the PP contains a weak pronoun, which indicates that the complement of the scrambled PP is obligatorily assigned accent. Because this is a typical property of

Object Shift and Object Scrambling

contrastive foci / topics but not of scrambled objects, which are typically destressed, we conclude that the movement of PP should be distinguished from scrambling. Another reason for assuming this is that leftward movement of the PP is possible with a restricted set of adverbial phrases only: While object scrambling across a time adverbial like gisteren is easily possible, (41c) shows that this gives rise to a severely degraded result in the case of PP-movement under a noncontrastive intonation pattern. (41)

a. dat Jan that Jan b. dat Jan that Jan c. dat Jan that Jan

nauwelijks luisterde. to Peter hardly listened nauwelijks luisterde. to him hardly listened gisteren luisterde. to Peter yesterday listened

18.8 A′ and A′-Scrambling of Objects Section 18.7 has shown that scrambling of the type intended is what we may call A-scrambling: scrambling of an object into a structural case position. A-scrambling should be carefully distinguished from cases of A′-scrambling such as contrastive focus /topic-movement. Making the distinction between A and A′-scrambling may provide a better understanding of the longstanding problem that scrambling often seems “Janus-faced”, in the sense that it exhibits properties of both A and A′-movement; see Webelhuth (1992). Consider the examples in (42): Example (42a) shows that scrambling may feed anaphor binding, which is considered a typical A-movement property; (42b) on the other hand shows that it can also license parasitic gaps, which is normally considered a typical property of A´-movement. (42) a. dat Jan namens elkaar

feliciteert.

that Jan the rivals on.behalf.of each.other congratulates ‘that Jan congratulates the rivals on behalf of each other.’ b. dat Jan [zonder pgi aan te kijken] feliciteert. that Jan the rivals

without

prt. to look

congratulates

Example (43a) shows that a scrambled object may in fact exhibit both properties simultaneously: The object de gasten binds the reciprocal elkaar and licenses a parasitic gap. Webelhuth assumes that the object is moved into its surface position in one fell swoop, as in (43b), and concludes that the target position of scrambling has mixed A/A′-properties. An alternative possibility is that the scrambled object is moved in two steps, as in (43b´): The object is first moved into an A-position, from where it binds the reciprocal, and is subsequently moved into an A′-position, from where it licenses the parasitic gap.

429

430

HANS BROEKHUIS

(43) a. Hij had de gasten [zonder pg te bekijken] aan elkaar

voorgesteld. he had the guests without to look-at to each.other introduced ‘He had introduced the guests to each other without looking (at them).’ b. Hij had de gasteni [zonder pg te bekijken] aan elkaar ti voorgesteld. b´. Hij had de gasteni [zonder pg te bekijken] t′i aan elkaar ti voorgesteld.

The alternative analysis in (43b´) is in line with the finding from Section 18.3 that scrambled nonpronominal objects may occupy (at least) two positions in the clause: one more deeply embedded A-position to the left of the modal adverbs and a higher one right-adjacent to the regular subject position. If we assume that the second position is an A′-position, the facts in (42) and (43) follow. A nice bonus is that it may also give us a handle on the fact that OS does not license parasitic gaps, as shown by the Icelandic example in (44), taken from Thra´insson (2001); this is expected if the conclusion from Section 18.3 that OS is uniformly A-movement is indeed correct. (44)

Pe´tur bauð Marı´ui aldrei ti [a´n þess að sækja hanai/*pgi] Pe´tur invited Marı´u never without it to fetch her ‘Pe´tur never invited Marı´u without picking her up.’

I am aware that the parasitic-gap analysis of the Dutch examples in (42b) and (43) is controversial. Broekhuis (1987), Neeleman (1994a), and de Hoop and Kosmeijer (1995), for instance, claimed that the presumed parasitic gaps differ from “true” parasitic gaps in that they can be licensed by subjects in passive constructions. Haider and Rosengren (1998) and Fanselow (2001: 412) added other properties that are not expected in parasitic-gap constructions.9 However, even if the parasitic-gap analysis should be abandoned, the logic of the argument still provides a useful guide for addressing the many other “mixed” properties of object scrambling, which are frequently assumed.

18.9 More on the Information-Structural Effect on Output Section 18.7 has argued that OS is triggered by structural case features. If we adopt the controversial but generally accepted claim that case features are checked under Agree, and that as a result of this, the subsequent Amovement of the noun phrase into the specifier of the case-assigning head is not needed for syntactic reasons, we must conclude that the derivations deriving the two representations in (2), repeated as (45), are both convergent. This does not entail, however, that the two representations are acceptable in all languages, due to language-specific filters: Icelandic 9

It is not clear whether Fanselow’s argument can be replicated for Dutch; Dutch parasitic-gap constructions comparable to his examples in (22) with nonreferential reflexive antecedents and (23) with multiple gaps have an unacceptable feel about them.

Object Shift and Object Scrambling

allows (45b) with nonpronominal as well as pronominal noun phrases, Danish allows it with pronouns only, and Finnish-Swedish does not allow it at all. Additionally, the representation in (45b) is subject to HG in all OSlanguages. Furthermore, the assumption that HG functions as a languagespecific filter on representations predicts that there are languages like Dutch, German, and Yiddish which do not exhibit (all) HG-effects on Ascrambling. (45)

a. . . . ADV [vP . . . [VP . . . object . . .]] b. . . . objecti . . . ADV [vP . . . [VP . . . ti . . .]]

What we have not yet accounted for, however, is the informationstructural effect of object movement. One possibility, already suggested by Haider and Rosengren (1998) but fully exploited in Broekhuis (2000, 2008) and Chomsky (2001), is that alternate representations tend to undergo meaning specialization. I will follow Chomsky in assuming that (45b) is subject to an effect-on-output condition: The shifted/scrambled object in (45b) must be construed as presuppositional. This proposal allows the object in (45a) to be construed as either part of the focus or the presupposition of the clause, which is needed for cases in which (45b) is blocked for languagespecific reasons; see section 18.2. This leaves us with the fact that the object in (45a) is normally construed as part of the information focus. We can account for this in a Gricean (1975) manner: The Maxim of Quantity requires the speaker to select (45b) as the more restrictive option if the object is part of the presupposition of the clause, and representation (45a) will consequently be used only if the object provides new information. Chomsky (2001) formalized this by means of an output filter, which can be expressed in a more sophisticated way (with wide empirical ramifications) in an OT-fashion by appealing to constraints like H - C O M P L and D A O P discussed earlier.

18.10 Remaining Problems and Consequences I have argued that OS and A-scrambling can both be considered A-movement triggered by structural case features. This is problematic under the current mainstream generative view that accusative case features are located on the light verb v, and that these features attract the theme argument into the outer specifier of vP; this leads to representation (46), where the displaced object follows the clause adverbial. (46)

. . . ADV [vP objecti [vP . . . [VP . . . ti . . .]]]

Chomsky (2001) repaired this deficiency by assuming that the object must undergo an additional phonological dislocation movement that places it in front of the clause adverbial; this proposal was extended to A-scrambling in

431

432

HANS BROEKHUIS

Chocano (2007). Since the nature of phonological movement is still unexplored, this solution has an ad hoc flavor and, in my view, should not be resorted to, if possible. This is all the more pressing because the problem results from the theory-specific assumption of multiple specifiers: the predecessor of this theory with AgrPs, for instance, seemed to fare much better in this respect, as will be clear from the representation in (47a). Other alternatives are available that result in more or less the same structure; Broekhuis (2008), for instance, follows the line of thinking in Grimshaw (1997) by claiming there is no functional head Agr but that the object crosses the adverbial by moving into the specifier of an extended projection of vP as in (47b). (47)

a. . . . [AgrP objecti Agr [vP ADV [vP . . . v [VP . . . ti . . .]]]] b. . . . [vP objecti v [vP ADV [vP . . . tv [VP . . . ti . . .]]]]

Another issue that should receive attention is the fact that Dutch and German differ from the Scandinavian languages in that definite noun phrases cannot follow negation if they are assigned neutral clause accent; Schaeffer (2000). Given that moved objects may either precede or follow the clause adverbial with the by now familiar effect-on-output, it seems that we must conclude that placement of the object in postadverbial position cannot be the result of A-scrambling of the type discussed earlier. (48) a. *dat Jan waarschijnlijk niet het boek gelezen heeft. b. dat Jan waarschijnlijk niet gelezen heeft. that Jan the book probably not read has ‘that Jan probably hasn’t read the book.’ Finally, the hypothesis that A-scrambling is like OS in being triggered by structural case features defines a research program aiming at systematically distinguishing A-scrambling from the various types of A´-scrambling; a first attempt for Dutch can be found in Broekhuis and Corver (2016: chapter 13) and it is clear that the existing literature on German contains abundant material bearing on this issue.

References Archangeli, D. and T. Langendoen 1997. Optimality Theory. An Overview. Malden and Oxford: Blackwell. Bayer, J. and J. Kornfilt 1994. “Against scrambling as an instance of Movealpha.” In N. Corver and H. van Riemsdijk (eds.), Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order Phenomena. Berlin and New York: Mouton de Gruyter: 17–60. Bergen, G. van and P. de Swart 2010. “Scrambling in spoken Dutch: Definiteness versus weight as determinants of word order variation,” Corpus Linguistics and Linguistic Theory 6: 267–295.

Object Shift and Object Scrambling

Besten, H. den and C. Moed-van Walraven 1986. “The syntax of verbs in Yiddish.” In H. Haider and M. Prinzhorn (eds.), Verb Second Phenomena in Germanic Languages. Dordrecht: Foris: 111–135. Broekhuis, H. 1987. “Chain-government,” The Linguistic Review 4: 297–374. Published in 1991. Broekhuis, H. 2000. “Against feature strength: the case of Scandinavian object shift,” Natural Language & Linguistic Theory 18: 673–721. Broekhuis, H. 2007. “Subject shift and object shift,” Journal of Comparative Germanic Linguistics 10: 109–141. Broekhuis, H. 2008. Derivations and Evaluations: Object Shift in the Germanic Languages. Berlin and New York: Mouton de Gruyter. Broekhuis, H. and L. Cornips 1994. “Undative constructions,” Linguistics 32: 173–190. Broekhuis, H. and L. Cornips 2012. “The verb krijgen ‘to get’ as an undative verb,” Linguistics 50: 1205–1249. Broekhuis, H. and N. Corver 2016. Syntax of Dutch: Verbs and Verb Phrases, Vol. 3. Amsterdam University Press. Chocano, G. 2007. Narrow Syntax and Phonological Form. Amsterdam and Philadelphia: John Benjamins. Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. 2000. “Minimalist inquiries: The framework.” In R. Martin, D. Michaels, and J. Uriagereka (eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasik. Cambridge, MA: MIT Press: 89–155. Chomsky, N. 2001. “Derivation by phase.” In M. Kenstowicz (ed.), Ken Hale: A Lfe in Language. Cambridge, MA: MIT Press: 1–52. Christensen, K. R. 2005. Interfaces. Negation – Syntax – Brain. Ph.D. thesis, University of Aarhus. Cinque, G. 1993. “A null theory of phrase and compound stress,” Linguistic Inquiry 24: 239–297. Corver, N. and H. van Riemsdijk 1994. Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order Phenomena. Berlin and New York: Mouton de Gruyter. Diesing, M. 1997. “Yiddish VP order and the typology of object movement in Germanic,” Natural Language & Linguistic Theory 17: 369–427. Engels, E. and S. Vikner 2013. “Derivation of Scandinavian object shift and remnant VP-topicalization.” In H. Broekhuis and R. Vogel (eds.), Linguistic Derivations and Filtering: Minimalism and Optimality Theory. Sheffield, UK, and Bristol, USA: Equinox Publishing: 193–220. Engels, E. and S. Vikner 2014. Scandinavian Object Shift and Optimality Theory. Basingstoke: Palgrave Macmillan. Fanselow, G. 2001. “Features, theta-roles, and free constituent order,” Linguistic Inquiry 32: 405–437. Fanselow, G. 2003. “Free constituent order: A minimalist interface account,” Folia Linguistica 37: 191–231.

433

434

HANS BROEKHUIS

Fox, D. and D. Pesetsky 2005. “Cyclic linearization of syntactic structure,” Theoretical Linguistics 31: 1–45. Grewendorf, G. and W. Sternefeld (eds.) 1990. Scrambling and Barriers. Amsterdam: John Benjamins. Grice, H. P. 1975. “Logic and conversation.” In P. Cole and J. L. Morgan (eds.), Speech Acts: Syntax and Semantics, Vol. 3. New York: Academic Press: 41–58. Grimshaw, J. 1997. “Projection, heads and optimality,” Linguistic Inquiry 28: 373–422. Haegeman, L. 1995. The Syntax of Negation. Cambridge University Press. Haider, H. 2000. “Scrambling – What’s the state of the art?” In S. M. Powers and C. Hamann (eds.), The Acquisition of Scrambling and Cliticization. Springer: Dordrecht: 19–40. Haider, H. 2006. “Mittelfield phenomena.” In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax, Vol. III. Malden, MA, and Oxford: Blackwell: 204–274. Haider, H. and I. Rosengren 1998. “Scrambling,” Sprache und Pragmatik 49: 1–104. Holmberg, A. 1986. Word Order and Syntactic Features in the Scandinavian Languages and English. Ph.D. thesis, University of Stockholm. Holmberg, A. 1991. “The distribution of Scandinavian weak pronouns.” In H. van Riemsdijk and L. Rizzi (eds.), Clitics and Their Hosts (Eurotyp Working Papers 8). Tilburg University: 155–173. Holmberg, A. 1999. “Remarks on Holmberg’s generalization,” Studia Linguistica 53: 1–39. Hoop, H. de and W. Kosmeijer 1995. “Case and scrambling: D-structure versus S-structure.” In H. Haider, S. Olsen, and S. Vikner (eds.), Studies in Comparative Germanic Syntax. Dordrecht, Boston, and London: Kluwer Academic Publishers: 139–158. Jo´nsson, J. G. 1996. Clausal Architecture and Case in Icelandic. Ph.D. thesis, University of Massachusetts: GLSA. Kayne, R. S. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Mahajan, A. 1994. “Toward a unified theory of scrambling.” In N. Corver and H. van Riemsdijk (eds.), Studies on Scrambling. Movement and NonMovement Approaches to Word-Order Phenomena. Berlin and New York: Mouton de Gruyter: 301–330. Matthews, K. 2000. The Syntax of Object Shift in Icelandic. Memorial University of Newfoundland. Molna´rfi, L. 2003. “On optional movement and feature checking in West Germanic,” Folia Linguistica 28: 129–162. Mu¨ller, G. 1998. Incomplete Category Fronting. A Derivational Approach to Remnant Movement in German. Dordrecht, Boston, and London: Kluwer Academic Publishers.

Object Shift and Object Scrambling

Mu¨ller, G. 2000. “Shape conservation and remnant movement.” In A. Hirotani, N. Hall Coetzee, and J.-Y. Kim (eds.), Proceedings of NELS 30. Amherst, MA: GLSA: 525–539. Mu¨ller, G. 2001. “Order preservation, parallel movement, and the emergence of the unmarked.” In G. Legendre, J. Grimshaw, and S. Vikner (eds.), Optimality-Theoretic Syntax. Cambridge, MA: MIT Press: 113–142. Mu¨ller, G. 2007. “Towards a relativized concept of cyclic linearization.” In U. Sauerland and H-M. Ga¨rtner (eds.), Interfaces + Recursion = Language? Berlin and New York: Mouton de Gruyter: 61–114. Neeleman, A. 1994a. “Scrambling as a D-structure phenomenon.” In N. Corver and van H. Riemsdijk (eds.), Studies on Scrambling. Movement and Non-Movement Approaches to Free Word-Order Phenomena. Berlin and New York: Mouton de Gruyter: 387–429. Neeleman, A. 1994b. Complex Predicates. Ph.D. thesis, University of Utrecht. Neeleman, A. and H. van de Koot 2008. “Dutch scrambling and the nature of discourse templates,” The Journal of Comparative Germanic Linguistics 11: 137–189. Putnam, M. T. 2007. Scrambling and the Survive Principle. Amsterdam and Philadelphia: John Benjamins. Rustick, S. M. 1991. Verb Second and Object Shift in Germanic. University of Wisconsin-Madison. Schaeffer, J. 2000. The Acquisition of Direct Object Scrambling and Clitic Placement. Amsterdam: John Benjamins. Sells, P. 2001. Structure Alignment and Optimality in Swedish. Stanford: CSLI Publications. Thra´insson, H. 2001. “Object shift and scrambling.” In M. Baltin and C. Collins (eds.), Contemporary Syntactic Theory. Malden: Blackwell: 148–202. Verhagen, A. 1986. Linguistic Theory and the Function of Word Order in Dutch: A Study on Interpretive Aspects of the Order of Adverbials and Noun Phrases. Dordrecht: Foris. Vikner, S. 1989. “Object shift and double objects in Danish,” Working Papers in Scandinavian Syntax 44: 141–155. Vikner, S. 1994. “Scandinavian object shift and West Germanic scrambling.” In N. Corver and H. van Riemsdijk (eds.), Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order Phenomena. Berlin and New York: Mouton de Gruyter: 487–517. Vikner, S. 2006. “Object shift.” In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax, Vol. III. Malden, MA and Oxford: Blackwell: 392–436. Webelhuth, G. 1992. Principles and Parameters of Syntactic Saturation. New York: Oxford University Press. Williams, E. 2003. Representation Theory. Cambridge, MA: MIT Press. Zwart, J-W. 1997. Morphosyntax of Verb Movement: A Minimalist Approach to the Syntax of Dutch. Dordrecht: Kluwer Academic Publishers.

435

Chapter 19 Unbounded Dependency Constructions in Germanic Martin Salzmann

19.1 Introduction Unbounded dependency constructions canonically involve displacement of a phrase across a finite clause-boundary. In the following English example, the wh-phrase which book originates as a complement of like and surfaces at the beginning of the clause:1 (1)

Which book1 do you think Mary likes __1?

Such dependencies are thus usually filler-gap dependencies with the fronted XP being (usually, but see Sections 19.2.2., 19.4.4 and 19.4.5 below) overt and the position it originates from being phonologically empty. The fronted XP occupies an operator/discourse-related position, and constructions involving such fronting are termed instances of long A′-movement within Generative Grammar. All Germanic languages have unbounded dependency constructions of the type in (1); however, long A′-dependencies in Germanic can come in different forms that deviate in various ways from the variant in (1). Given that long A′-dependencies are one of the major research topics in syntactic theory, Germanic languages represent a particularly fruitful area for the exploration of the nature and limits of these dependencies. This chapter is organized as follows: In Section 19.2, I will introduce the major types of A′-dependencies and some of their salient properties before addressing basic analytical issues in Section 19.3. In Section 19.4, I discuss alternatives to long extraction, and Section 19.5 deals with various remarkable locality-related properties of long extraction in Germanic.

1

In what follows, the filler and the gap are coindexed by means of numerical subscripts. For coreference relationships alphabetical subscripts are used.

Unbounded Dependency Constructions in Germanic

19.2 Types of (Long) A’-Dependencies and Their Major Properties The three major types of long A′-dependencies are wh-movement as in (1) and relativization and topicalization as illustrated in (2): (2)

a. This book1, I think Mary likes __1. topicalization b. the book which1 I think Mary likes __1 relativization

A brief terminological remark: The term topicalization does not imply that the fronted elements have the function of topics. In all Germanic languages, topicalized constituents can have various topic functions (aboutness, discourse, contrastive) as well as different foci functions (information focus and contrastive focus). For information-structural differences between short and long topicalization, see Section 19.3.2.

19.2.1

Shared Properties of Long A′-Dependencies

Since Chomsky (1977: 555) the three constructions have been grouped together as instances of A′-movement because they share a number of important properties: First, they are potentially unbounded, i.e., can, in principle, span an unlimited number of clauses; second, they are subject to the same locality restrictions, viz., the island constraints discovered by Ross (1967). For instance, the antecedent/filler of these constructions cannot be related to a gap within an adjunct clause:2 (3) a. *Which politician1 did you leave the room because John praised __1? b. *This politician1, you left the room because John praised __1. c. *the politician who1 you left the room because John praised __1 Third, these unbounded dependencies are usually analyzed as involving a sequence of smaller steps, viz., the filler is assumed to make stopovers in intermediate (SpecCP and possibly SpecvP) positions. Finally, information contained within the filler can be interpreted in lower positions, i.e., the constructions display so-called reconstruction effects. In the following example, the variable contained in the filler is bound by the embedded subject: (4)

[Which of hisi pictures] 1 do you think every boyi likes best __1?

All Germanic languages have these three types of unbounded dependencies (for restrictions in German with respect to clause-boundedness, see Section 19.5.1 below). In what follows, I will focus on these three types as 2

While extraction from so-called strong islands like the adjunct island in (3), relative clauses and complement clauses to nouns as well as (sentential) subjects is prohibited across the board, so-called weak islands (like embedded questions) allow argument extraction / extraction of referential/D-linked elements but block extraction of adjuncts / nonreferential / non-D-linked elements. See Szabolcsi (2006) for discussion.

437

438

MARTIN SALZMANN

they have received the most attention in the literature. For discussion of further A′-constructions in English, such as comparative deletion, clefts, exclamatory and negative NP-extraction, see, e.g., Postal (1998); for longdistance scrambling in German, see Frey (2006).

19.2.2 Unbounded Dependencies – The Left Periphery The three constructions differ in a number of interesting respects both language-internally as well as cross-linguistically with respect to inversion and the shape of the left periphery (form of the displaced element, overtness of attracting heads, co-occurrence restrictions). 19.2.2.1 Unbounded Dependencies and Inversion in the Final Clause The three major constructions show certain differences with respect to subject-verb inversion in the clause targeted by movement when a nonsubject is fronted. In English matrix clauses, only wh-movement shows inversion, while subject and verb occur in canonical order in the other two constructions. The other Germanic languages pattern differently in this respect: Both topicalization and wh-movement involve inversion in matrix clauses, while relativization does not.3 (5) a. Wen1 glaubst du, dass Maria __1 liebt? who.A C C believe.2S G you that Mary love.3S G ‘Who do you believe that Mary loves?’ b. [Den du, dass Maria __1 Maler]1 glaubst the.A C C painter believe.2S G you that Mary ‘The painter, you believe that Mary loves.’ c. ein Maler, den1 du glaubst, dass Maria __1 a painter who.A C C you believe.2 S G that Mary ‘a painter who you believe that Mary loves’

liebt. love.3S G liebt love.3S G German

With embedded wh-clauses and topicalization, there is no inversion in Standard English. With embedded questions, there is generally no subjectverb inversion in Germanic, except in Irish English, African American and Appalachian English, and (optionally) in Afrikaans (Harbert 2007: 475); in Yiddish and Icelandic there is inversion with expletive subjects, see Vikner (1995: 76) and Vikner (Chapter 16). With embedded topicalization, subjectverb inversion is possible in all Germanic languages except English (and perhaps Dutch, where embedded verb-second is not generally available, see Vikner 1995: chapter 3). Embedded topicalization is illustrated by the following German example (in all other Germanic languages, including English, embedded topicalization follows the declarative complementizer, while complementizers and fronted verbs are in complementary distribution in German): 3

For so-called V2-relatives in German, see Gärtner (2000).

Unbounded Dependency Constructions in Germanic

(6) Ich glaube,

[dieses Buch]1 denkt

der

Hans, dass Maria __1 mag

I believe.1S G this book think.3 S G the.N O M John that Mary ‘I believe this book, John thinks that Mary likes.’

like.3S G German

19.2.2.2 Unbounded Dependencies and the Shape of the Left Periphery While topicalization simply involves a fronted unmarked XP all across Germanic, wh-movement and relative clauses display a certain variation with respect to the expression of operators and functional heads. Starting with wh-movement, while matrix questions generally involve fronting of an XP plus inversion (when a nonsubject is fronted), embedded questions usually do not involve any inversion and the C-position remains empty; in some Germanic languages, though, the whphrase can be followed by a complementizer. While impossible in Standard English because of the doubly filled Comp filter (see Chomsky and Lasnik 1977), in nonstandard English (and Belfast English, see Harbert 2007: 435) the wh-phrase can co-occur with the declarative complementizer: (7)

They discussed a certain model, but they didn’t know which model that they discussed. (Baltin 2010: 331)

The situation in German is similar. While impossible in the standard language, many, especially the southern dialects like Bavarian or Alemannic, allow doubly filled Comps, see, e.g., Bayer (1984), and Bayer and Brandner (2008). The same is also possible in Flemish and various other Dutch varieties, Middle English, Middle Danish, and Colloquial Copenhagen Danish (Vikner 1995: 119f.) as well as in Norwegian and Swedish (Harbert 2007: 427, Larsson 2014). In Dutch varieties, one can even find co-occurrence of the wh-phrase with two complementizers, viz., ‘if’ and ‘that’ (Zwart 2000: 361). Turning to headed relative clauses (which are postnominal and headexternal in Germanic4), in Standard English, relativization normally involves either a relative pronoun (who/which) / a null operator, or the declarative complementizer that (for a reanalysis of that as a relative pronoun, see Broekhuis and Dekkers 2000 and Kayne 2010). Cooccurrence of both elements can be found in nonstandard varieties, as well as in Old and Middle English (Harbert 2007: 434): (8) It’s down to the community in which that the people live. (Gelderen 2013: 59) 4

English and Dutch appositive relatives like the following have been argued to represent double-headed types, see Bhatt (2015: 712); for Dutch data, see Vries (2002: 189). i) I read the New Yorker yesterday, [which magazine is one of the finest in the country]. For free relatives, see Harbert (2007: 466ff.).

439

440

MARTIN SALZMANN

Doubly-filled-Comp violations of this type are also found in Gothic, Frisian (Harbert 2007: 434) and in Dutch varieties (with both wh- and D-pronouns, Zwart 2000: 357f.). Dutch varieties that use a wh-pronoun in relativization additionally allow the combination of relative pronoun and ‘if’ (Zwart 2000: 358ff., Boef 2012: 141). Additionally, Dutch varieties also allow that-relatives, cf. Boef (2012: 141); in Afrikaans, relatives involve invariant wat for subjects and objects, while for oblique positions the relative pronouns wie/wat are used (Harbert 2007: 443ff.). While relatives in Standard German use the relative D-pronoun only, many German varieties allow the co-occurrence of the relative D-pronoun with da ‘there’/was ‘what’/wo ‘where’, see Weise (1916), Bayer (1984), Fleischer (2004). The particle wo lit. ‘where’ (to be distinguished from the homophonous relative adverb) also occurs on its own as a relative marker in Alemannic varieties, see van Riemsdijk (1989, 2008), Salzmann (2006): (9)

d Frau, wo immer the woman C always ‘the woman who is always late’

z too

spaat late

chunt come.3 S G Zurich German

Another invariant complementizer found in German varieties is was ‘what’ (see also Yiddish vos). Wh-relatives (based on welch- ‘which’) are restricted to the standard language (and Yiddish, which has velkh and ver) and never co-occur with a complementizer, see Fleischer (2004). Relative clauses without any kind of marking, viz., zero relatives, do not seem to exist in either German or Dutch. They are prominent in Standard English nonsubject relatives; in nonstandard English, we also find contact relatives for subject relativization, see Bhatt (2015: 737ff.): (10)

There’s a girl (who) wants to see you.

In Scandinavian languages relative clauses are primarily introduced by an invariant complementizer som/sem ‘as’. It can be omitted in (restrictive) object (but not subject) relatives in Mainland Scandinavian (Harbert 2007: 450, 464f.) but not in Icelandic (Thra´insson 2007: 447). In Danish and Norwegian, we alternatively find der in subject relativization. Swedish and Danish also have wh-pronouns for oblique/ prepositional relations (Harbert 2007: 425, 447f.). Icelandic allows the combination of sem + ‘that’ (Vikner 1995: 122), as does Danish(Nyvad et al. 2017: 463). Faroese allows zero relatives for both subjects and objects (Harbert 2007: 456, 465). In some languages, a formal difference between what look like subject and object relative pronouns may in fact rather result from the subject form being an agreeing complementizer (Harbert 2007: 430ff.).

Unbounded Dependency Constructions in Germanic

19.3 Basic Analytical Issues 19.3.1 The Nature of the Landing Site The classical analysis of unbounded dependencies involves A′-movement to a position in the left periphery via a stopover in the intermediate SpecCP position. For English wh-movement, this looks as follows (for evidence for an additional stopover in SpecvP, see Fox 1999): (11)

[CP Who1 did you say [CP __1 that Mary likes __1]?

The landing site of English wh-movement and relativization is usually identified with SpecCP (accompanied by T-to-C-movement in the former), while topicalization is generally assumed to target a lower position between TP and CP. While earlier work took this to be an adjunction position (adjunction to TP), in more recent approaches, the landing site of topicalization is identified with the specifier of an additional functional head in the left periphery, e.g., SpecTopP, see Branigan (2011). Apart from cartographic considerations that may favor distinct landing sites for different types of A′-movement, see Rizzi (1997), Branigan (2011), evidence for positing different landing sites for topicalized material than wh-/ relative operators comes from the fact that topicalization inside relative clauses is permitted (to some extent, Douglas 2016: 71): (12)

a man [to whom] 2, liberty1, we should never grant __1 __2

Unlike wh-movement and relativization, topicalization in English can affect several constituents, in which case the first XP functions as a topic and the second one as a focus (Douglas 2016: 85): (13)

That book2, to JOHN1 Mary gave __2 __1 in 1979.

In the other Germanic languages, it has generally been assumed that unbounded dependencies all target the same position in the left periphery because unlike in English they are in complementary distribution with each other and multiple topicalization is impossible. This position is traditionally identified with SpecCP. In cartographic work, however, whmovement and topicalization usually target different projections even in these languages (e.g., Branigan 2011, but see also Mu¨ller 1995 for an early noncartograpic proposal to this effect).

19.3.2 Differences between Local and Long-Distance A′-Movement Traditionally it has been assumed that the landing sites of local and longdistance movement are the same, viz., SpecCP. There is some reason to believe, though, that this characterization is too simple. At the very least, there are asymmetries between local and long-distance topicalization that need to be accounted for somehow. It has been observed for

441

442

MARTIN SALZMANN

German that next to information-structurally marked fronting like proper topic or focus fronting there are also cases where the fronted element is information-structurally neutral. This may involve fronting of unstressed elements and sentence adverbials as in Frey (2006: 240): (14)

hat __1 keiner dem alten Mann geholfen. Leider1 unfortunately have.3S G no.one the.D A T old man helped ‘Unfortunately, nobody helped the old man.’ German

According to Frey, such elements reach the prefield by means of a different operation, so-called formal fronting, which affects the highest element in the middle field. This may include sentence adverbials, the subject5 and scrambled nonsubjects. Crucially, long-distance movement cannot be information-structurally neutral but rather necessarily involves contrast, according to Frey (2006). This can be shown by the fact that long-distance topicalization of sentence adverbs, which cannot be contrasted, is ungrammatical, see (15) (the sentence is, of course, acceptable with matrix construal of the adverb; see also Fe´ry, Chapter 28 for the relationship between information structure and syntax): (15)

*Leider1

sagte

unfortunately said

Karl, dass __1 Carl that

keiner dem

alten Mann geholfen hat.

no.one the.D A T old man

‘Unfortunately, Carl said that no one helped the old man.’

helped

have.3S G German

Because of this asymmetry it has been argued that formal fronting targets a different position than long A′-movement, viz. the neutral SpecFinP rather than SpecCP/ForceP/KontrP in a cartographic system (Frey 2006).

19.4 Alternatives to Long-Distance Movement Arguably the most interesting aspect of unbounded dependencies in Germanic is the fact that there is a large number of alternative constructions. These alternatives have largely the same meaning but usually differ in important syntactic respects from the canonical unbounded constructions.

19.4.1 Extraction from Verb-Second Clauses In German, wh-movement and topicalization can also take place from verbsecond clauses next to extraction across dass (which entails verb-final): (16)

5

du, dass sie seit 1985 mit ihrem Kind __1 wohnt? a. Wo1 glaubst where believe.2S G you that she since 1985 with her child live.3S G ‘Where do you believe that she has been living with her child since 1985?’ verb-final

For this reason, subject initial verb-second clauses have sometimes been analyzed as TPs rather than CPs (Zwart 1997). See also the debate whether local subject questions in English constitute CPs or TPs, e.g., Grimshaw (1997).

Unbounded Dependency Constructions in Germanic du, wohnt sie seit 1985 mit ihrem Kind __1? b. Wo1 glaubst where believe.2S G you live.3S G she since 1985 with her child verb-second

(17)

a. [In Bonn]1 meint in Bonn

Franz,

dass

sie

think.3S G Franz

that

she since 1985

seit

1985 __1

wohnt. live.3S G

‘Franz thinks that she has been living in Bonn since 1985.’ b. [In Bonn]1, in Bonn

meint

Franz,

think.3S G Franz

wohnt

sie

live.3S G she

seit

verb-final

1985 __1.

since 1985

verb-second

While the long movement analysis of extraction from V-final structures is uncontested, the analysis of structures with extraction from verb-second clauses has been controversial. An obvious alternative to long extraction is to treat everything of the main clause after the wh-word as a V1parenthetical. Evidence for this reanalysis comes from the fact that V1parentheticals are independently possible in various positions of the clause, see Reis (1995: 46): (18)

Wo1 wohnt sie {glaubst du} seit 1985 {glaubst du} where live.3S G she believe.2sg you since 1985 believe.2S G you du}? mit ihrem Kind __1 {glaubst with her child believe.2S G you

Consequently, (16b) would just be a variant of this with the parenthetical right after the wh-phrase. An argument for a long-distance extraction analysis comes from the observation that the class of verbs that allow long extraction across dass, viz., the so-called bridge verbs (like think, believe, say etc.), is largely identical to the class of verbs that allow for embedded verb second, see Featherston (2004). One of the challenges for an extraction analysis comes from what Reis calls the V2 route restriction: Extraction may occur via V2 and into V2-clauses only; these conditions are violated in the following examples (adapted from Reis 1995: 50): (19)

a. *Ich weiß nicht, wo1 er meint, wohnt sie __1. I know.1S G not where he think.3S G live.3S G she ‘I don’t know where he thinks she lives.’ b. *Wo1 meint er, dass Peter geglaubt hat, wohnt sie __1? where think.3S G he that Peter believed have.3S G live.3S G she?

This follows straightforwardly under a parenthetical analysis because in both cases the material between the fronted wh-phrase and the final V1 clause is not a felicitous parenthetical. Accounting for this restriction under an extraction analysis has generated a large number of proposals; for a recent contribution with further references, see Mu¨ller (2010). Another possible argument against an extraction analysis comes from the fact that in present day Standard German, long extraction across dass is unacceptable for many speakers, as discussed in Section 19.5.1 below, while extraction from V2-clauses is readily available for all speakers. This may be unexpected if both constructions involve the same extraction

443

444

MARTIN SALZMANN

mechanism. Still, the issue is far from settled, and the reader is referred to the above-mentioned publications for further theoretical discussion; for empirical evidence in favor of the parenthetical analysis, see Kiziak (2010). A parenthetical analysis has also been proposed for extraction from V2clauses in Danish and Swedish, see Hrafnbjargarson et al. (2010: 301f.). In this case, the structures look exactly as in German, while normally embedded verb second occurs in the presence of a complementizer in these languages. For the Faroese equivalent of the German construction, however, the authors propose an extraction analysis because unlike in Swedish/Danish, variable binding by a quantifier inside the alleged parenthetical is grammatical, suggesting that the V1-clause is syntactically visible: (20)

[Mamma

hansarai/j ]1

segði

hvør

ein

mother

his

said

each

one man

maðuri __1 hevði ikki had

not

lisið bo´kina. read book.the Faroese

‘Hisi mother, each mani said (that) had not read the book.’

As for extraction from V2 after the complementizer, it is blocked in all Scandinavian languages if the V2-clause is nonsubject initial (e.g., involves an embedded topic). This is unsurprising as these are classical topic islands. Interestingly, however, extraction from subject-initial embedded V2-clauses is grammatical in Faroese, Icelandic, and (certain varieties of) Norwegian. Vikner (1995) accounts for the extractability in Icelandic with the assumption that there is no proper V2 (i.e., V-to-C-movement) in these cases but rather only V-to-I movement; the subject is thus in SpecTP so that no topic island intervenes (while in the other languages there is V-to-C-movement with the subject in the topic position). While a possibility for Icelandic, where V-to-I has been independently argued for, this is not an obvious solution for Norwegian, where this movement is not independently attested. However, if the verb occupies the same position in all Scandinavian embedded V2clauses, it is no longer clear why some languages allow for extraction, while others do not. See Hrafnbjargarson et al. (2010) for more empirical details.

19.4.2 Scope Marking / Partial Movement Another alternative to long extraction across that is the partial movement / scope-marking construction. In this construction, an invariant scope marker occupies the scope position of the wh-dependency, while the wh-phrase (which has exactly the same form as wh-phrases in long extraction and can thus be of greater complexity than in the example below) remains in an intermediate SpecCP position. It has been most prominently discussed in German but can be found in typologically diverse languages (see Fanselow 2017 for a recent overview): (21)

Was glaubst du, wen1 Maria __1 what believe.2S G you whom Mary ‘Who do you believe Mary loves?’

liebt? love.3S G

Unbounded Dependency Constructions in Germanic

In German the construction is only found in wh-movement, but depending on the implementation, partial movement may also exist in relative clauses, see Boef (2012) on Dutch and the discussion of the copy construction below. Unlike in other languages, partial movement in German does not allow the associate of was to be a yes-no-question. While semantically very similar to long extraction, there are certain asymmetries that are noteworthy (Felser 2001: 11f.): First, unlike long extraction, the scope marking construction is sensitive to negative island effects: (22) a. Wen1 glaubst du nicht dass Maria __1 whom think.2S G you not that M. ‘Who don’t you think Maria has met?’ b. *Was glaubst du nicht wen1 Maria __1 what think.2S G you not whom M.

getroffen hat? met have.3S G getroffen hat? met have.3S G

Second, the class of predicates that permits the scope marking construction is not coextensive with the class of verbs allowing long extraction; that is, not every long-distance movement structure has a scope marking correlate, and not every scope-marking sentence has a long-distance movement counterpart. The following example illustrates the latter asymmetry: (23) a. Was hat

Hans entschieden wer1 __1 morgen

what have.3S G H. b. *Wer1 hat

decided

who

Hans entschieden dass __1 morgen

who have.3S G H.

decided

that

kommen soll?

tomorrow come.I N F

should.3S G

kommen soll?

tomorrow come.I N F should.3S G

There are two main approaches to partial movement: the direct dependency approach and the indirect dependency approach (which both come in various flavors). In the direct dependency approach (McDaniel 1989) the invariant was and the wh-phrase are part of the same chain; was marks the scope of the wh-phrase, which is assumed to replace was at LF (was is thus assimilated to an expletive like there in A-chains). There is also a version of the direct dependency approach that treats it as a subcase of the copy construction discussed in Section 19.4.3. In the indirect dependency approach (simplifying somewhat), was is analyzed as the standard wh-word for clausal complements. The relationship between was and the wh-CP is then similar to that between a propositional pro-form and its clausal associate; i.e., it can be considered a monosentential counterpart of (18): (24)

What do you think? Who came?

A comparison of the two approaches is beyond the scope of this paper; while both have their pros and cons, both generally seem capable of capturing the relevant data, see Fanselow (2017).

445

446

MARTIN SALZMANN

19.4.3 Wh-Copying In the so-called copy construction, which is found in German, Frisian, Afrikaans, and Dutch varieties, the wh-phrase appears multiple times, viz., in the landing site as well as in all intermediate SpecCP positions: (25)

tinkst weˆr1’t se __1 Weˆr1 where think.2S G where-C O M P she ‘Where do you think she lives?’

wennet? live.3S G Frisian

There is a long tradition going back to Hiemstra (1986) to analyze the copy construction as involving the realization of several links of a movement dependency. This seems straightforward under an approach to movement that involves copying, as has become standard since Chomsky (1995). However, since normally only the highest (or, in wh-in-situ, the lowest) link is phonetically realized, extra assumptions are necessary to allow for copying. According to one popular approach (Nunes 2004), multiple copyspell-out results from reanalysis of intermediate copies with intermediate C-heads. As a consequence of this, the wh-phrase is no longer visible for the linearization mechanism so that it escapes chain reduction (which is otherwise necessary to obtain nonconflicting ordering statements). An argument for this view comes from the observation that copying tends to be restricted to simple wh-pronouns, while doubling of full wh-phrases is usually rejected (Felser 2004: 550): (26)

*[Welchen which

Mann]1 man

glaubst du, believe.2S G you

[welchen Mann]1 sie __1 which man she

‘Which man do you believe that she loves?’

liebt? love.3S G German

There is, admittedly, a gray area where judgments vary. Light PPs, for instance, are often found acceptable: (27)

%[An wen]1 glaubst du, [an wen]1 sie __1 of whom believe.2S G you of whom she ‘Who do you believe that she thinks of?’

denkt? think.3S G German

Given that in Afrikaans copying of full wh-phrases is possible (Plessis 1977), however, the reanalysis approach cannot be generally correct. For more discussion about the restrictions on copying, see Fanselow and Mahajan (2000) and Felser (2004). There is a line of research, starting with Hiemstra (1986), that argues that scope marking and wh-copying should receive a unified account. The basic idea is that the two constructions differ only in the amount of material that is copied of a wh-phrase. The scope marking construction results if only the wh-feature is copied (which is then spelled-out by the most unmarked wh-word), see Cheng (2000) for a revival of this idea. The most systematic approach to wh-copying and scope marking is proposed by Barbiers et al. (2010), who focus on variation in wh-chains in Dutch varieties. They observe that the first member of the wh-chain is always less

Unbounded Dependency Constructions in Germanic

specific than the second one. They argue that this generalization follows from partial copying; for the proposal to work, pronouns must be spellouts of phrases. They assume that pronouns have rich internal structure: (28)

[DP D [PhiP Phi [QP Q]]]

Depending on how much of that phrase is affected by copying, different types of wh-chains emerge. It is assumed that QP is realized as wat, PhiP as wie, and DP as die. This correctly predicts the possibilities wat – wie, wat – die and wie – die: The first option results if the wh-pronoun is only a PhiP from which the QP is extracted; the other options obtain if the wh-word is a DP and either PhiP or QP are extracted. Because of partial copying, the reverse order of wh-words is ruled out: *wie – wat. Interestingly, relative clauses in Dutch varieties display a similar pattern. Boef (2012) convincingly demonstrates that the partial copying approach can be fruitfully extended to this empirical domain as well. Partial copying also seems to be at work in certain German patterns identified by Fanselow and Cavar (2001), where the wh-phrase in the scope position is a proper subset of the lower wh-phrase (rather than just the scope marker was): (44)

Wen denkst du [wen von den Studenten]1 sie __1 mag? whom think.2S G you whom of the students she likes? ‘Who of the students do you think she likes?’ German

Simplifying somewhat, approaches in terms of copying essentially assume that wh-copying and scope marking are syntactically very similar to regular long extraction. We have already seen evidence that there are certain nontrivial differences between long extraction and scope marking that cast doubts on this view. Importantly, at least in German there are also systematic differences between the copy construction and long extraction on the one hand as well as differences between the copy construction and scope marking on the other. This calls a unification / a parallel treatment of the three constructions into question. As for the copy construction and long extraction: As discussed in Felser (2004: 555ff.), the two constructions differ in the verb classes they occur with (copying is, e.g., impossible with factive verbs) and their sensitivity to negative islands: (29)

a. Wen1 glaubst du nicht, dass sie __1 liebt? whom believe.2S G you not that she love.3S G ‘Who don’t you believe that she loves?’ b. *Wen1 glaubst du nicht, wen1 sie __1 liebt? whom believe.2S G you not whom she love.3S G

Furthermore, the copying construction is incompatible with wh-in-situ and displays different scope properties (in the interaction with other quantifiers).

447

448

MARTIN SALZMANN

As for differences between partial movement and the copy construction, they do not occur with the same set of verbs (Felser 2004: 552): (30)

a. *Was scheint what seem.3S G b. ?Wen1 scheint who seem.3S G

es, it es, it

wen1 whom wen whom

Hans __1 H. Hans __1 H.

geschlagen hit geschlagen hit

hat? have.3S G hat? have.3S G

Furthermore, while copying of full phrases is rather restricted in the copy construction, the wh-phrases in the scope marking construction can be arbitrarily complex: (31)

Was glaubst du, [welchen Mann] sie __1 liebt? what believe.2S G you which man she love.3S G ‘Which man do you believe that she loves?’ German

Perhaps the most serious problem for the (partial) copying account is the fact, stressed in Pankau (2013), that there are cases in German where the upper copy is more complex than the lower one: (32) [Mit welchem Mann] glaubst

du [mit wem]

sie

with which man believe.2S G you with who.D A T she ‘With which man do you believe she dances?’

tanzt? dance.3S G ? German

19.4.4 Resumption In some Germanic languages, unbounded dependencies can involve resumption (for a discussion of Hanging Topic and Contrastive Left Dislocation, which bear similarities to resumption, see Harbert 2007: 478–481). Resumptives occur in (spoken) English, but it is usually thought that they do not represent a fully grammatical phenomenon. One can distinguish two types: First, so-called intrusive pronouns occur in positions from where regular movement (leaving a gap) is not possible, i.e., they are a repair device that alleviates the violation of island / locality constraints (a wh-island / adjunct island in the following examples), but the result is usually not considered fully grammatical, see Chao and Sells (1983: 48): (33)

a. This is the man that Mary couldn’t remember if she had seen him before. b. Which man do they think that if Mary marries him, then everyone will be happy?

A different repair device are so-called processing resumptives. They occur in positions from where movement is generally possible (and thus a gap would be expected). Their acceptability increases the further away they are from their antecedent, see, e.g., Erteschik-Shir (1992: 89):

Unbounded Dependency Constructions in Germanic

(34)

a. This is the girl that John likes __/*her. b. This is the girl that Peter said that John thinks that Bob likes __/?her.

It should be pointed out that the claims about the ameliorating effect of intrusive pronouns and the general acceptability of processing resumptives has been called into question by various experimental work, see, e.g., Asudeh (2012: chapter 10), the contributions in Sprouse and Hornstein (2013) and McCloskey (2017) for discussion and references. Next to these possibly extragrammatical resumptives, one does find resumptives in nonstandard English in positions from where extraction is usually unproblematic and where processing cannot be at stake (from Prince 1990): (35)

I have a friend who she does all the platters.

While frowned upon by prescriptivists, such examples are frequent in colloquial speech. Whether they need to be distinguished from the two previous types is in fact not fully clear. For resumption in earlier stages of English, see Harbert (2007: 457). Among the Scandinavian languages, resumptives are systematically found in Swedish, where they only occur in embedded subject position (Engdahl 1985: 8): (36)

Vilket ord visste ingen hur *(det) stavas? which word knew no.one how it is.spelled lit.: ‘Which word did no one know how it is spelled?’

Swedish

Most Swedish examples with resumptives involve wh-islands; they are much rarer in declarative complements, see Engdahl (1986: 98–99). However, one cannot generally say that Swedish resumptives amnesty locality violations. In fact, the object resumptives that one can find in the literature are probably best analyzed as processing resumptives. The discussion in Asudeh (2012: 35–36) (who discusses data from Zaenen et al. 1981: 681) and Engdahl 1982: 154, 165) suggests that they are not more acceptable than gaps and Engdahl (1985: 10) shows that they do not improve locality violations. Subject resumptives have the properties of gaps in that they can occur in ATB-contexts (Zaenen et al. 1981: 681) and license parasitic gaps (Engdahl 1985: 7, 38–39): (37)

Det var den faº ngen som la¨karna inte kunde avgo¨ra om It was that prisoner that doctors.the not could decide if han verkligen var sjuk [utan att tala med __ personligen] he really was ill without to talk with in.person. lit.: ‘This was the prisoner that the doctors couldn’t determine if he really was ill without talking to in person.’ Swedish

Since Swedish resumptives do not seem to have any influence on the syntax, one obvious possibility is to analyze them as the spell-out of a trace (note that

449

450

MARTIN SALZMANN

the other Scandinavian languages usually have gaps in embedded subject position). But why Swedish has resumptives in the first place is not clear. Perhaps the resumptives serve to satisfy an EPP-requirement of the embedded clause (viz., that the subject position, SpecTP, be filled). See Salzmann (2017) for detailed discussion about possible causes of resumption. A fully productive resumption system is only found in Alemannic varieties of German and in Yiddish. In Alemannic, resumptives are found in relativization only (but not in topicalization and wh-movement). They have a somewhat peculiar distribution in that in local relativization they only occur in the relativization of dative objects and more oblique relations, while in long-distance relativization they occur across the board, i.e., also for subjects and direct objects, see van Riemsdijk (1989), Salzmann (2017):6 (38)

a. es Bild, wo niemert (*s) cha zale a picture C nobody (it) can.3 S G pay.I N F ‘a picture that nobody can afford’ b. de Bueb, wo mer *(em) es Velo versproche ha¨nd the boy C we (he.D A T ) a bike promise.P T C P have.1P L ‘the boy who we promised a bike’ c. s Bild, wo t gsa¨it ha¨sch, dass *(es) de the picture C you say.P T C P have.2S G that it the Peter wett verchauffe Peter want.3 S G sell.I N F ‘the picture that you said Peter wants to sell’ Zurich German

Resumption in Swiss German is island-insensitive but displays certain reconstruction effects. For a base-generation analysis that reconciles these properties, see Salzmann (2017). Long-distance relativization has been reanalyzed as an instance of resumptive prolepsis, discussed in Section 19.4.5. Resumption is also systematic in Yiddish relative clauses. What is remarkable about Yiddish is that resumptives can (optionally) occur in subject and direct object position, even though resumptives are crosslinguistically less frequent in these positions (Prince 1990): (39)

mentshn vos a shlang hot (zey) gebisn people C a snake has them bitten ‘people whom a snake bit’

Yiddish

19.4.5 Resumptive Prolepsis German and Dutch have an alternative to long extraction where the extracted constituent is governed by a functional preposition (usually corresponding to of) and where the position that the fronted constituent is related to is occupied by a coreferential pronoun. This construction, termed resumptive prolepsis in Salzmann (2006, 2017), occurs with all types of 6

Some dialects use gaps for the relativization of datives, see Salzmann (2017: chapter 5) for details.

Unbounded Dependency Constructions in Germanic

unbounded dependencies, although it is most frequent with relativization. This is arguably not a coincidence because long-distance relativization is ungrammatical in modern Standard German (and restricted for some speakers of Dutch, see Section 19.5). At least in relativization resumptive prolepsis can thus be considered to be fully grammaticalized: (40)

a. Von [welchem Maler]i glaubst du, dass Maria of which.D A T painter believe.2 S G you that Mary ihni mag? him like.3 S G ‘Of which painter do you believe that Mary likes him?’ b. ein Maler, von [dem]i ich glaube, dass Maria a painter of who.D A T I believe.1 S G that Mary ihni mag him like.3 S G ‘a painter of whom I believe that Mary likes him’ c. Von [dem Maler]i glaube ich, dass Maria ihni mag. of the.D A T painter believe.1 S G I that Mary him like.3 S G ‘Of the painter, I believe that Mary likes him.’ German

(41)

Kort geleden had ik deze droom waarivan ik denk Recently have.P S T .1S G I this dream which.of I think.1S G dat hiji belangrijk is. that he important be.3S G ‘Not long ago I had this dream of which I think that it is important.’ Dutch www.jancvanderheide.com/vragen-aan-jan/144-mb-te-a-kort-geledenhad-ik-deze-droom-waarvan-ik-denk- dat-hij-belangrijk-is, found on January 20, 2015

At first sight, one may take the proleptic constituent (the ‘of’-PP) to be an argument of the matrix verb with the pronoun establishing an anaphoric dependency. However, upon closer inspection, this simplistic view fails: The construction is compatible with a very wide range of matrix verbs, casting doubt on the argumenthood of the object. Second, a coreferential element is obligatory, which argues against a mere aboutness relationship. In Salzmann (2017), building on van Riemsdijk (2008), it is proposed that the proleptic object is licensed by predication; the coreferential pronoun is analyzed as a resumptive, and the relationship between the two crucially involves ellipsis.

19.5 Locality In this section, I will discuss various locality-related phenomena within Germanic that have received prominent discussion in the literature.7 7

For reasons of space, I cannot address superiority effects, viz., the rigidity of the order of wh-phrases in multiple questions. While most Germanic languages require fronting of the subject wh-phrase with the object wh-phrase

451

452

MARTIN SALZMANN

19.5.1 Clause-Bound Unbounded Dependencies in German One rather curious property of unbounded dependencies in German is that they are degraded for many speakers once they cross a finite clause boundary (the effect is strongest with relativization). This was noted in Andersson and Kvam (1984) and experimentally verified in Featherston (2005). It has frequently been claimed that the degradation of long A′-movement is characteristic of more Northern varieties of German, including the standard language, while there is no such restriction in Southern German varieties (Haider 1983: 97–99, Grewendorf 1988: 126). Indeed, both Fanselow and Weskott (2010) and Salzmann et al. (2013) find experimental evidence that long extraction is less acceptable in the North than in the South. It is still not quite clear whether a dialectal contrast is involved or whether the unacceptability is just a phenomenon of the standard language. What is clear is that long A′-movement is well-attested in dialects. For instance, one can find examples in traditional descriptions of the dialects of Basel, Bern, Lucerne, and Zurich, see Salzmann (2017: 259) for references. Since several of the linguists who have worked on the topic happen to come from the South and additionally are dialect speakers, what they describe as Southern varieties of German (or Southern versions of the standard language) may be heavily influenced by their native dialect. Evidence that it may be a restriction of the standard language comes from Lu¨hr (1988: 79), who cites several sources that suggest that there was explicit prescriptive pressure in the nineteenth century in Germany against long A′-movement: It was considered “illogical” or “sloppy”. This accords nicely with the fact that long A′-movement is more productive in dialects, which are usually immune to such pressure. In line with that is the fact that long distance movement (of all three types) is welldocumented for earlier stages of the language (roughly before 1850–1900). For an extensive collection of examples, see Behaghel (1928: 547ff.): (42)

[An was]1 verlangt ihr, dass ich __1 glauben soll? at what demand.2P L you that I believe.I N F should.1S G ‘What do you expect me to believe in?’ (Schiller)

I am not aware of similar restrictions in other Germanic languages, perhaps with the exception of certain varieties of Dutch: Although long relativization is generally possible in Dutch, there are certain indications that it may be restricted nevertheless. Many speakers accept or in fact prefer the prolepsis construction for long-distance dependencies. The situation in the dialects is particularly interesting, see Boef (2012). Some of the observed patterns could perhaps be analyzed as not involving long remaining in-situ (in non-D-linked questions), e.g., who said what versus *What did who say, German has been claimed to freely allow both orders. This asymmetry has been related to the possibility of scrambling in German: Assuming that in wh-movement the closest wh-phrase is fronted, scrambling the wh-object across the wh-subject is a strategy to make an object-initial multiple question possible. Empirical work has confirmed the characterization in the theoretical literature by and large, even though crossing movement also leads to lower acceptability ratings in German, see Häussler et al. (2015) and references cited there.

Unbounded Dependency Constructions in Germanic

extraction. Barbiers et al. (2005) explicitly suggest that long-distance relativization is impossible in many dialects.

19.5.2

Intervention Effects – Topic- versus Wh-Islands

As discussed in Section 19.4.1, extraction from complement clauses with a filled SpecCP-position is blocked in most Germanic languages. Topic islands also strongly degrade extraction in English, see Mu¨ller and Sternefeld (1993: 481), but the relativization data discussed in Section 19.3.1 suggest that this is not an absolute restriction. In German topic islands are categorical, while wh-islands are more permeable, at least for topicalization, as suggested by the following contrast (Mu¨ller and Sternefeld 1993: 485): (43)

ich [C P gestern1 hat Ede __1 __2 repariert]. a. *Radios2 glaube radios believe.1S G I yesterday has Ede repaired ‘Radios, I believe Ede repaired yesterday.’ ich nicht [C P wie1 man __1 __2 repariert]. b. ??Radios2 weiß radios know.1S G I not how one repair.3S G ‘Radios, I don’t know how to repair.’

19.5.3 Absence of Island Effects in Mainland Scandinavian There is a large body of literature suggesting that Mainland Scandinavian languages are peculiar from a cross-linguistic perspective in that they fail to obey the Complex Noun Phrase Constraint. Concretely, they allow extraction from relative clauses and complement clauses to nouns (but usually not from other strong islands like adjunct islands and subject islands; however, see Nyvad et al. 2017: 460f. for grammatical extraction from adverbial clauses in Danish). Corresponding examples in other Germanic languages are sharply ungrammatical, see Heinat and Wiklund (2015: 37) and references cited there for earlier work: (44)

a. [Saº na blommor]1 ka¨nner jag [en man [som sa¨ljer __1]]. Such flowers know.1S G I a man that sells Swedish b. *?[Those flowers]1 I know [someone [who sells __1]].

Given that the CNPC constraint is usually considered to hold universally, the behavior of the Mainland Scandinavian languages is rather puzzling. A grammatical account of the cross-linguistic asymmetry is provided in Nyvad et al. (2017): They assume extraction from CNPCislands in Danish to be generally grammatical and relate it to the possibility of CP-recursion; this option, which they argue is independently motivated in the language, provides an escape hatch for the filler.

453

454

MARTIN SALZMANN

Kush et al. (2013), on the other hand, argue that the acceptability of extraction from strong islands in Swedish is illusory; relative clauses that allow extraction have a different structure (a small clause structure) from those that block extraction (true relatives). According to this hypothesis, the head noun of the relative clause is reanalyzed as the subject of a PredP, while the relative complementizer som is identified with Pred (Som can indeed independently function as the head of an SC). Since small clauses do not block extraction, the acceptability of relative clause extraction is unsurprising: (45)

[De blommorna]1 ka¨nner jag [PredP en man [Pred som] sa¨ljer __1]. those flower.P L . D E F know.1S G I a man who sells lit.: ‘Those flowers, I know a man who sells.’

Extraction is predicted to be blocked with verbs that cannot take such a small clause complement. According to Kush et al. (2013), this prediction is borne out with verbs like “meet” or “kiss.” However, empirical studies on Danish (Christensen and Nyvad 2014) and Swedish (Mu¨ller 2015) refute these claims: They found no difference in acceptability between extraction with verbs that are compatible with SCs and extraction with verbs that are not; furthermore, in both studies, extraction from relative clauses led to significant degradation, which contrasts with the claims in the earlier literature. A different perspective has been to assume that extraction is indeed severely constrained even in Scandinavian languages but that there are factors that make extraction more felicitous (factors, which can also play a role in other languages). Such factors/conditions include semantic properties of the head noun of the RC, the nature of the gap inside the RC (subject versus nonsubject), the nature of the matrix subject, properties of the extracted element (referentiality) and information-structural factors. While these factors indeed make extraction from strong islands in languages like English more acceptable (but not grammatical), Heinat and Wiklund (2015) show that none of these factors is crucial in Scandinavian languages: Relative clause extractions are acceptable even if these conditions are not met; thus, extraction in MSc is indeed exceptional in their view. In a recent acceptability study, Kush et al. (2017) found that extraction from complex noun phrases in Norwegian is just as degraded as extraction from adjunct islands and subject islands; furthermore, extraction was as degraded as in languages like English and Italian, which are generally thought to comply with the CNPC. Thus, these empirical facts suggest that Norwegian does not differ in its locality properties from the more ‘well-behaved’ languages (although the authors stress that their results do not necessarily imply that extraction from complex DPs is ungrammatical in Norwegian). As this short summary of the literature shows, the status of the locality properties of Mainland Scandinavian unbounded dependencies remains controversial; trying to determine whether extraction is really

Unbounded Dependency Constructions in Germanic

different in these languages will therefore constitute an intriguing topic for future research.

19.5.4 That-Trace Effects In many languages there is a curious subject-object asymmetry in the extraction across that: While unproblematic for objects, it is often degraded for subjects, which is why the effect has been termed that-trace effect: (46)

a. Who1 did you say that Mary likes __1? b. *Who1 did you say that __1 came?

That-trace effects seem to be found in most Germanic languages to some extent; there are subtle differences, and it is not fully clear to what extent the differences reflect deeper grammatical asymmetries; perhaps, some of the variation reported in the literature is due to the fact that the data were obtained by different methodological means. That-trace effects in English seem undisputed even though there also seems to be some variation, see Pesetsky (2017) and references cited there. Among the Scandinavian languages, Danish, Faroese, Swedish, and some varieties of Norwegian are claimed to display the that-trace effect, while Icelandic and certain (Northern) varieties of Norwegian do not (Hrafnbjargarson et al. 2010: 301). In some of the Scandinavian languages, as in English, that-trace effects can be voided by means of an expletive (der in Danish) or a subject resumptive (Swedish), see Lohndal (2009: 216f.). It generally seems difficult to relate the observed differences to independent cross-linguistic differences; Lohndal (2009) argues that the feature content of C is crucial in this respect. For German, the existence of that-trace effects has long been denied, see Haider (1983), Grewendorf (1988), Mu¨ller (1995). Subject and object extraction were taken to be equally acceptable (in those varieties that permit extraction across dass, see above). Featherston (2005), however, provided experimental evidence for that-trace effects in German as well. These results were confirmed in a more detailed study by Kiziak (2010). More recently, Bayer and Salzmann (2013) have shown that that-trace effects are unrelated to subjecthood; they argue that the degradation always obtains when the sentence topic undergoes long movement and relate it to improper movement. Salzmann et al. (2013) go a step further and show that the degradation observed in German is unrelated to extraction, because it obtains more generally when the complementizer and the finite verb are adjacent. They propose that such structures are ruled out because of a phonological EPP that requires the specifier of a high functional head to be overtly occupied. The status of that-trace effects in Dutch is rather murky. Bennis (1987: 238) argues that long extraction as such is unproblematic but that it is restricted by pragmatic constraints, which boils down to having presuppositional material in the middle field, with traces of

455

456

MARTIN SALZMANN

movement counting as well. While relativization and topicalization involve topical presupposed material, wh-movement (often) does not; as a consequence, the dummy pronoun er surfaces under long whextraction to satisfy the pragmatic constraint. The facts reported in Bennis are largely confirmed in a questionnaire study by den Dikken (2007). Boef (2012) documents subject-object asymmetries in long relativization in Dutch dialects. She argues that the form die that occurs in subject extraction is an agreeing complementizer akin to French qui, which makes extraction possible.

References Andersson, S-G. and S. Kvam 1984. Satzverschra¨nkung im heutigen Deutsch. Eine syntaktische und funktionale Studie unter Beru¨cksichtigung alternativer Konstruktionen. Tu¨bingen: Gunter Narr. Asudeh, A. 2012. The Logic of Pronominal Resumption. Oxford University Press. Baltin, M. 2010. “The nonreality of doubly filled comps,” Linguistic Inquiry 41: 331–335. Barbiers, S., H. Bennis, M. Devos, G. de Vogelaer, and M. van der Ham 2005. Syntactic Atlas of the Dutch Dialects (SAND), Vol 1. Amsterdam University Press. Barbiers, S., O. Koeneman, and M. Lekakou 2010. “Syntactic doubling and the structure of wh-chains,” Journal of Linguistics 46: 1–46. Bayer, J. 1984. “Comp in Bavarian syntax,” The Linguistic Review 3: 209–274. Bayer, J. and E. Brandner 2008. “On wh-head-movement and the doublyfilled-comp filter.” In C. B. Chang and H. J. Haynie (eds.), Proceedings of the 26th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Proceedings Project: 87–95. Bayer, J. and M. Salzmann 2013. “That-trace effects and resumption – how Improper Movement can be repaired.” In P. Brandt and E. Fuss (eds.), Repairs: The Added Value of Being Wrong. Berlin: Mouton de Gruyter: 275–334. Behaghel, O. 1928. Deutsche Syntax: eine geschichtliche Darstellung. Band 3: Die Satzgebilde. Heidelberg: Winter. Bennis, H. 1987. Gaps and Dummies. Dordrecht: Foris. Bhatt, R. 2015. “Relative clauses and correlatives.” In A. Alexiadou and T. Kiss (eds.), Syntax – Theory and Analysis: An International Handbook, Vol. 1. Berlin, Munich, and New York: Mouton de Gruyter: 708–749. Boef, E. 2012. Doubling in Relative Clauses: Aspects of Morphosyntactic Microvariation in Dutch. Utrecht: LOT. Branigan, P. 2011. Provocative Syntax. Cambridge, MA: The MIT Press. Broekhuis, H. and J. Dekkers 2000. “The minimalist program and optimality theory: Derivations and evaluations.” In J. Dekkers, F. van der Leeuw, and J. van der Weijer (eds.), Optimality Theory: Phonology, Syntax, and Acquisition. Oxford University Press: 386–422.

Unbounded Dependency Constructions in Germanic

Chao, W. and P. Sells 1983. “On the interpretation of resumptive pronouns,” Proceedings of NELS 13: 47–61. Cheng, L. 2000. “Moving just the feature.” In U. Lutz, G. Mu¨ller, and A. von Stechow (eds.), Wh-Scope Marking. Amsterdam: John Benjamins: 77–99. Chomsky, N. 1977. “On wh-movement.” In P. Culicover, T. Wasow, and A. Akmajian (eds.), Formal Syntax. New York: Academic Press: 71–132. Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. and H. Lasnik 1977. “Filters and control,” Linguistic Inquiry 8: 425–504. Christensen, K. R. and A. M. Nyvad 2014. “On the nature of escapable relative islands,” Nordic Journal of Linguistics 37: 29–45. Dikken, M. den 2007. Questionnaire study on Dutch that-trace effects: Stimuli and results. Ms., CUNY. Douglas, J. 2016. The Syntactic Structures of Relativisation. Ph.D. dissertation, Cambridge University. Engdahl, E. 1982. “Restrictions on unbounded dependencies in Swedish.” In E. Engdahl and E. Ejerhed (eds.), Readings on Unbounded Dependencies in Scandinavian Languages. Stockholm: Almquist & Wiksell International: 151–174. Engdahl, E. 1985. “Parasitic gaps, resumptive pronouns, and subject extractions,” Linguistics: An Interdisciplinary Journal of the Language Sciences 23: 3–44. Engdahl, E. 1986. Constituent Questions: The Syntax and Semantics of Questions with Special Reference to Swedish. Dordrecht: Reidel. Erteschik-Shir, N. 1992. “Resumptive pronouns in islands.” In H. Goodluck and M. Rochemont (eds.), Island Constraints. Theory, Acquisition and Processing. Dordrecht: Kluwer: 89–108. Fanselow, G. 2017. “Partial wh-movement.” In M. Everaert and H. van Riemsdijk (eds.), The Wiley Blackwell Companion to Syntax, Second Edition. doi: 10.1002/9781118358733.wbsyncom088. Fanselow, G. and A. Mahajan 2000. “Towards a minimalist theory of whexpletives, wh-copying, and successive cyclicity.” In U. Lutz, G. Mu¨ller, and A. von Stechow (eds.), Wh-Scope Marking. Amsterdam: John Benjamins: 195–230. Fanselow, G. and D. Cavar 2001. “Remarks on the economy of pronunciation.” In G. Muller and W. Sternefeld (eds.), Competition in Syntax. Berlin: Mouton de Gruyter: 107–150. Fanselow, G. and T. Weskott 2010. “A short note on long movement in German,” Linguistische Berichte 222: 129–140. Featherston, S. 2004. “Bridge verbs and V2 verbs – the same thing in spades?” Zeitschrift fu¨r Sprachwissenschaft 23: 181–209. Featherston, S. 2005. “that-trace in German,” Lingua 115: 1277–1302. Felser, C. 2001. “Wh-expletives and secondary predication: German partial wh-movement reconsidered,” Journal of Germanic Linguistics 13: 5–38.

457

458

MARTIN SALZMANN

Felser, C. 2004. “Wh-copying, phases, and successive cyclicity,” Lingua: International Review of General Linguistics 114: 543–574. Fleischer, J. 2004. “A typology of relative clauses in German dialects.” In B. Kortmann (ed.), Dialectology Meets Typology: Dialect Grammar from a Cross-Linguistic Perspective. Berlin: Mouton de Gruyter: 211–243. Fox, D. 1999. “Reconstruction, binding theory, and the interpretation of chains,” Linguistic Inquiry 30: 157–196. Frey, W. 2006. “Contrast and movement to the German prefield.” In V. Molnar and S. Winkler (eds.), The Architecture of Focus. Berlin: Mouton de Gruyter: 235–264. Ga¨rtner, H-M. 2000. “Are there V2 relative clauses in German?” The Journal of Comparative Germanic Linguistics 3: 97–141. Gelderen, E. van 2013. Clause Structure. Cambridge University Press. Grewendorf, G. 1988. Aspekte der deutschen Syntax. Eine Rektions-BindungsAnalyse. Tu¨bingen: Narr. Grimshaw, J. 1997. “Projection, heads, and optimality,” Linguistic Inquiry 28: 373–422. Haider, H. 1983. “Connectedness effects in German,” Groninger Arbeiten zur Germanistischen Linguistik 23: 83–119. Harbert, W. 2007. The Germanic Languages. New York: Cambridge University Press. Ha¨ussler, J., M. Grant, G. Fanselow, and L. Frazier 2015. “Superiority in English and German: cross-language grammatical differences?” Syntax 18: 235–265. Heinat, F. and A-L. Wiklund 2015. “Scandinavian relative clause extractions: Apparent restrictions,” Working Papers in Scandinavian Syntax 94: 36–50. Hiemstra, I. 1986. “Some aspects of wh-questions in Frisian,” Nowele 8: 97–110. Hrafnbjargarson, G. H., K. Bentzen, and A-L. Wiklund 2010. “Observations on extraction from V2 clauses in Scandinavian,” Nordic Journal of Linguistics 33: 299–309. Kayne, R. 2010. “Why isn’t this a complementizer?” In R. Kayne, Comparison and Contrasts. Oxford University Press: 190–227. Kiziak, T. 2010. Extraction Asymmetries: Experimental Evidence from German. Amsterdam: John Benjamins. Kush, D., A. Omaki, and N. Hornstein 2013. “Microvariation in islands?” In J. Sprouse and N. Hornstein (eds.), Experimental Syntax and Island Effects. Cambridge University Press: 239–264. Kush, D., T. Lohndal, and J. Sprouse 2017. “Investigating variation in island effects,” Natural Language & Linguistic Theory 36: 743–779. https://munin .uit.no/handle/10037/13111. Larsson, I. 2014. “Double complementizers,” Nordic Atlas of Language Structures (NALS) Journal 1.

Unbounded Dependency Constructions in Germanic

Lohndal, T. 2009. “Comp-t effects: variation in the position and features of C*,” Studia Linguistica 63: 204–232. Lu¨hr, R. 1988. “Zur Satzverschra¨nkung im heutigen Deutsch,” Groninger Arbeiten zur Germanistischen Linguistik 29: 74–87. McCloskey, J. 2017. “Resumption.” In M. Everaert and H. van Riemsdijk (eds.), The Wiley-Blackwell Companion to Syntax, Second Edition. Oxford: WileyBlackwell. https://onlinelibrary.wiley.com/doi/10.1002/9781118358733. wbsyncom105. McDaniel, D. 1989. “Partial and multiple wh-movement,” Natural Language & Linguistic Theory 7: 565–604. Mu¨ller, C. 2015. “Against the small clause hypothesis: evidence from Swedish relative clause extractions,” Nordic Journal of Linguistics 38: 67–92. Mu¨ller, G. 1995. A-Bar Syntax: A Study in Movement Types. Berlin: Mouton de Gruyter. Mu¨ller, G. 2010. “Movement from verb-second clauses revisited.” In T. Hanneforth and G. Fanselow (eds.), Language and Logos. A Festschrift for Peter Staudacher. Berlin: Akademieverlag: 97–128. Mu¨ller, G. and W. Sternefeld 1993. “Improper movement and unambiguous binding,” Linguistic Inquiry 24: 461–507. Nunes, J. 2004. Linearization of Chains and Sideward Movement. Cambridge, MA: MIT Press. Nyvad, A. M., K. R. Christensen, and S. Vikner 2017. “CP-recursion in Danish: A cP/CP-analysis,” The Linguistic Review 34: 449–477. Pankau, A. 2013. Replacing Copies: The Syntax of Wh-Copying in German. Utrecht: LOT. Pesetsky, D. 2017. “Complementizer-trace effects.” In M. Everaert and H. C. Riemsdijk (eds.), The Wiley-Blackwell Companion to Syntax, Second Edition. Oxford: Wiley-Blackwell. https://onlinelibrary.wiley.com/doi/ abs/10.1002/9781118358733.wbsyncom108. Plessis, H. du 1977. “Wh movement in Afrikaans,” Linguistic Inquiry 8: 723–726. Postal, P. M. 1998. Three Investigations of Extraction. Cambridge, MA: MIT Press. Prince, E. F. 1990. “Syntax and Discourse: A look at resumptive pronouns,” BLS 16: 482–497. Reis, M. 1995. “Extractions from verb-second clauses in German?” In U. Lutz and J. Pafel (eds.), On Extraction and Extraposition in German. Amsterdam: John Benjamins: 45–88. Riemsdijk, H. van 1989. “Swiss relatives.” In D. Jaspers, W. Klooster, Y. Putseys, and P. Seuren (eds.), Sentential Complementation and the Lexicon. Berlin: Foris: 343–354. Riemsdijk, H. van 2008. “Identity avoidance: OCP effects in Swiss relatives.” In R. Freidin, C. P. Otero, and M. L. Zubizarreta (eds.), Foundational Issues in

459

460

MARTIN SALZMANN

Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. Cambridge, MA: MIT: 227–250. Rizzi, L. 1997. “The fine structure of the left periphery.” In L. Haegeman (ed.), Elements of Grammar. Dordrecht: Kluwer: 281–337. Ross, J. R. 1967. Constraints on Variables in Syntax. Doctoral dissertation, MIT. Salzmann, M. 2006. Resumptive Prolepsis: A study in Indirect A′-Dependencies. Utrecht: LOT. Salzmann, M. 2017. Reconstruction and Resumption in Indirect A′-Dependencies: On the Syntax of Prolepsis and Relativization in (Swiss) German and Beyond. Berlin: Mouton de Gruyter. Salzmann, M., J. Ha¨ussler, M. Bader, and J. Bayer 2013. “That-trace effects without traces: An experimental investigation,” Proceedings of NELS: 149–162. Sprouse, J. and N. Hornstein 2013. Experimental Syntax and Island Effects. Cambridge University Press. Szabolcsi, A. 2006. “Strong and weak islands.” In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax. 1st edn. Vol. 4. Cambridge: Blackwell, 479–531. Thra´insson, H. 2007. The Syntax of Icelandic. Cambridge University Press. Vikner, S. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. New York: Oxford University Press. Vries, M. 2002. The Syntax of Relativization. Utrecht: LOT. Weise, O. 1916. “Die Relativpronomina in den deutschen Mundarten,” Zeitschrift fu¨r Deutsche Mundarten 12: 64–71. Zaenen, A., E. Engdahl, and J. M. Maling 1981. “Resumptive pronouns can be syntactically bound,” Linguistic Inquiry 12: 679–682. Zwart, J-W. 1997. Morphosyntax of Verb Movement: A Minimalist Approach to the Syntax of Dutch. Dordrecht: Kluwer. Zwart, J-W. 2000. “A head raising analysis of relative clauses in Dutch.” In A. Alexiadou, P. Law, A. Meinunger, and C. Wilder (eds.), The Syntax of Relative Clauses. Amsterdam: John Benjamins: 348–385.

Chapter 20 The Voice Domain in Germanic Artemis Alexiadou and Florian Schäfer

20.1 Introduction In this chapter, we will discuss the core Voice alternations in Germanic languages. Different Voice alternations have in common that the argument structure of the underlying verb, in particular the semantic and syntactic properties of the verb’s external argument, are manipulated and this often goes along with changes in the morpho-syntax of the verb or the verbal phrase. In Section 20.2, we discuss inner-Germanic variation in the formation of analytic eventive and adjectival passives. In Section 20.3, we discuss the Mainland Scandinavian s-passives and in Section 20.4 we discuss non-canonical get-passives and related constructions. In Section 20.5, we turn to the morpho-syntactic variation in anticausative formation. Finally, in Section 20.6, we discuss how dispositional statements known as generic middles are realized across the Germanic languages.

20.2

Analytic Passives

The most productive Voice alternation is the one between the active and the analytic passive (cf. 1a, 1b). In analytic passives (also called periphrastic passive), the lexical verb appears as a past participle and the inflectional properties of the clause (agreement with a nominative DP and/or Tense) are realized on a passive auxiliary. In passives of transitive verbs, the verb’s external argument is demoted and the internal argument is promoted to the subject position. Even though the demoted logical subject typically remains unexpressed (“short passive”), it is clearly present in the semantic representation of the clause: in other words, it is generally agreed upon that passives have an implicit external argument. Implicit external arguments of passives can, however, also be overtly expressed via

462

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

an adjunct PP, the so-called by-phrase, in all Germanic languages (“long passive”). (1)

a. John wrote the novel. b. The novel was written (by John).

(active) (analytic passive)

The presence of an implicit external argument in short passives can best be demonstrated with transitive causative verbs which, next to the passive (2a, 3a), can also form an anticausative variant (2b, 3b). Only the passive but not the anticausative variant has an implicit external argument in its semantic representation. Standard tests indicating this are the (non-)licensing of agentive by-phrases, instrumental PPs, purpose clauses (2a, 2b) and, in particular, the (non-)licensing of the adjunct phrase ‘by itself’ (3a, 3b) (see, e.g., Levin and Rappaport Hovav 1995, Alexiadou et al. 2015 for further discussion). (2)

a. The vase was broken (by Mary/with a hammer/on purpose/in order to impress Mary). b. The vase broke (*by Mary/*with a hammer/*on purpose/*in order to impress Mary).

(3)

a. The vase was broken (*by itself). b. The vase broke (by itself).

20.2.1

Morphological Properties

20.2.1.1 The Passive Participle Germanic passive participles are always identical in form with the participle used in the periphrastic perfect. However, the agreement properties can be different: While perfect participles never agree with a DP in the clause, the passive participle shows agreement in some Germanic languages. In Icelandic (cf. 4a, 4b) and Faroese, the passive participle obligatorily agrees in number, gender and case with the verb’s internal argument DP if this DP is marked with structural case (nominative, or accusative in ECM constellations). In the absence of such a DP, i.e., in impersonal passives or if the internal argument has inherent case (cf. 4), the passive participle appears in a default form (neuter, singular) (Thra´insson 2007). (4)

a. Þrja´r bækur voru skrifaðar. three books.Nom.pl.f. were.pl. written.Nom.pl.f. b. Það voru skrifaðar þrja´r bækur. there were.pl. written.Nom.pl.f. three books.Nom.pl.f.

While in Icelandic participial agreement shows up irrespectively of any word order variation, in Swedish (cf. 5), the passive participle

The Voice Domain in Germanic

agrees with the internal argument DP only if the DP precedes the participle. (5)

a. Tre bo¨cker blev skrivna /*skrivet. three book.pl. were written.pl. written.sg. b. Det blev skrivet / *skrivna tre bo¨cker. it was written.sg. written.pl. three book.pl.

In West Germanic languages as well as in Danish (cf. 6), passive participles do not agree, irrespective of word order. In Norwegian, we find dialectal variation, certain dialects pattern with Icelandic, while others pattern with Swedish, and yet others with Danish (see Holmberg 2001, Sigurðsson 2004, Å farli 2009). (6)

Han/De he/they

bliver are

dømt. judged

20.2.1.2 The Auxiliary: Verbal versus Adjectival Passives The passive auxiliary shows the ordinary agreement properties of finite verbs. In impersonal passives (see Section 20.5), the auxiliary shows up in a default form. English uses the copula ‘be’ as the auxiliary in canonical passives. Short passives (The book is written) are thus ambiguous between an eventive and a stative reading. These are called verbal and adjectival passives respectively. Different modifiers disambiguate between the two. For example, by-phrases and instrumental PPs are productively available only in the verbal passive (7a). Similarly, the progressive and temporal in-adverbials enforce the eventive reading. Finally, only adjectival passive participles appear in adjectival contexts such as modification with very or prefixing with negative un- (7b). (7)

a. Harry was (being) beaten (by an expert) (with a computer) (in three days). b. Harry was (*being) unbeaten (*by an expert) (*with a computer) (*in three days).

Adjectival passives have been controversially discussed in the literature, the main question in earlier work being their lexical versus phrasal status. More recent work suggests that adjectival passives are not so different from eventive passives, see, e.g., McIntyre (2013), Bruening (2014), Alexiadou et al. (2014b, 2015). German uses two different auxiliaries for verbal and adjectival passives (werden ‘become’ versus sein ‘be’). The two passives are therefore easily kept apart (8a, 8b). The Mainland Scandinavian languages make the same distinction.

463

464

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

(8)

a. Das Buch wird

(von Peter) (langsam) geschrieben. (verbal passive)

the book becomes by

Peter

slowly

written

b. Das Buch ist (??von Peter) (*langsam) geschrieben. (adjectival passive) the book is by

Peter) slowly

written

Dutch and Afrikaans also have two passive auxiliaries worden (become) and zijn (be), which allow to differentiate in many cases between verbal and adjectival passives. However, worden does normally not show up in the perfect tense. Thus, while the perfect tense of passives in German is derived by participle + werden + sein, in Dutch and Afrikaans only zijn appears in the perfect tense of passives (cf. 9). Most researchers analyze zijn as being ambiguous between the perfect form of the passive auxiliary (forming verbal passives) and a copula (used in adjectival passives), see Schlu¨cker (2009) for further discussion and references. Consequently, the following string is ambiguous, but it can be disambiguated, for example by adding a by-phrase. (9)

De winkel is gesloten. ‘The store is closed.’ ‘The store has been closed.’

(adjectival passive) (verbal passive)

While Icelandic (as Faroese) has two passive auxiliaries vera ‘be’ and verða ‘become’, vera can be used in verbal passives of all tenses. Consequently, we have the same ambiguity as in English; (10) can either express that the breaking event occurred yesterday, or that the state of being broken held yesterday. (10)

Ru´ðan var brotin ı´ gær. window.the was broken in yesterday ‘The window was broken yesterday.’

20.2.2 Implicit Arguments and the By-Phrase All Germanic languages have a dedicated preposition to optionally realize the verb’s external argument in the passive (“long passive”). Such prepositions are often taken from the locational domain. English uses by, German von (from) (and in restricted circumstances durch [through]), Dutch uses door (through), but the Scandinavian languages use af/av (of). The choice of the particular preposition is rather idiosyncratic and subject to grammaticalization. For example, while earlier stages of Dutch used bi (by) and van (from), these became fully replaced by door. In most languages, passivization absorbs the external argument irrespective of its exact thematic role (agent, causer, experiencer, holder, etc.) (see, e.g., Fox and Grodzinsky 1998, Collins 2005, Bruening 2013). Icelandic passives are exceptional here in that the implicit argument (or

The Voice Domain in Germanic

the argument expressed in the by-phrase) can only be an agent and not an inanimate effector such as a natural force (Thra´insson 2007: 250ff.).1

20.2.3 Syntactic Variation with Analytic Passives Across frameworks, active and passive clauses are assumed to be closely related. While Chomsky (1957) formulated to this effect a particular passive rule, later work in the government-and-binding-tradition related the properties of passives to the presence of the passive (participial) morphology in combination with general properties of the grammar such as the Case Filter and generalized XP-movement. In particular, the presence of the participial morpheme -en on the lexical verb was assumed to have the following two effects (e.g., Jaeggli 1986; Baker et al. 1989): (11)

a. The passive suffix -en absorbs accusative Case. b. The passive suffix -en absorbs the external theta role.

In more recent work, passives are assumed to involve a particular Voice (or little v) head which (i) does not project a specifier and (ii) does not assign accusative (cf. Chomsky 1995, Kratzer 1996).2 Both treatments are meant to bring about that (a) the verb’s external argument is not syntactically projected as a DP, and (b) the internal argument does not receive accusative case in its base position. In order to obey the Case Filter, the internal argument thus has to A-move to the structural subject position, Spec,IP/TP, where it receives nominative case.

20.2.3.1 Variation in Case Shift According to the above view, the verb’s internal argument obligatorily becomes a derived subject under passivization. In English, this is reflected in word order and in the change in the morphological case of the internal argument from accusative in the active to nominative in the passive. As in most Germanic languages, this case shift is observable only on pronouns. (12)

a. Mary/she called the mayor/him. b. The mayor/He was (being) called (by her/Mary).

German and Icelandic still show case morphology on lexical DPs including two inherent or lexical cases, genitive and dative. These languages show that only structural object case (accusative) is absorbed under canonical passivization (cf. 13a, 13b) while inherent or lexical case is not affected (cf. 14a, 14b). Note that Icelandic verbal and adjectival passives differ here (Svenonius 2006); inherent/lexical case is in fact turned to nominative in 1

The picture with experiencers is more complicated in that they are sometimes but not always accepted as implicit argument of Icelandic passives.

2

For more recent developments in the theoretical debate on passives, see, e.g., Collins 2005, Bruening 2013, Alexiadou et al. 2015, 2018.

465

466

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

adjectival passives (cf. 14c). In German, on the other hand, lexical case is kept in verbal and adjectival passives (15): (13)

a. Þeir seldu bo´kina. they.Nom sold book.the.Acc ‘They sold the book.’ var seld. b. Bo´kin book.the.Nom was sold ‘The book was sold.’

(14)

a. Marı´a bauð honum. Marı´a invited him.Dat ‘Marı´a invited him’ b. Honum var boðið (af Marı´u). him.Dat was invited by Marı´a ‘He was invited (by Marı´a).’ c. Hann var (o´-)boðinn (*af Marı´u). he.Nom was un-invited by Marı´a ‘He was (un-)invited.’

(15)

(verbal passive)

(adjectival passive)

Ihm wurde/war schon lange geholfen. he.Dat became/was already long helped ‘He was helped for a long time/since a long time’. (verbal/adjectival passive)

While the Faroese passive is very similar to its Icelandic counterpart, Faroese verbal passives are special in that they often do not even preserve lexical/inherent dative (cf. 16). However, as the contrast with (17) shows, this depends on the particular verb involved (Thra´insson et al. 2004, Thra´insson 2007, Eytho´rsson and Thra´insson 2017: 78f.). (16) a. Teir hja´lptu honum. they.Nom helped him.Dat b. Hann / *Honum varð hja´lptur. he.Nom him.Dat was helped (17)

(lexical case not preserved)

a. Teir takkaðu honum. they.Nom thanked him.Dat b. Honum /*Hann Varð takkað. him.Dat he.Nom was thanked

(lexical case preserved)

Note that the datives in (14b) and (17b) behave as quirky subjects, i.e., they move from their base position to Spec,TP (and then on to Spec,CP) and thereby acquire all subject properties known in these languages except for triggering verbal agreement (Zeanen et al. 1985, Sigurðsson 1989, Thra´insson 2007). Icelandic and Faroese passives show then that A-movement and case advancement do not go hand in hand (see McFadden 2004, Chapter 13).

The Voice Domain in Germanic

20.2.3.2

Variation in DP-Movement

According to the standard GB-treatment, the internal argument should obligatorily move to Spec,TP in passives. Such movement is, however, not a universal property of passives. In German and Dutch as well as Icelandic and Faroese, the internal argument can stay in situ in passives. This is illustrated in the Icelandic example in (4b) above where the internal argument follows the participle and, therefore, is inside the verbal phrase. In the German example in (18), the internal argument follows the adverbials ja or noch nie which are taken to indicate the vP/VP-border (e.g., Wurmbrand 2006).3 (18)

weil as

ja/noch well/yet

nie never

ein a

Buch book.Nom

gelesen read

wurde. was

These languages show that A-movement of the internal argument is not a universal or obligatory property of passives, and they suggest, again, that there must be an alternative explanation for the obligatory movement of the internal argument in English passives.

20.2.3.2.1 DP-Movement in Passives of Double Object Verbs In passives of English double object verbs, the indirect object moves to subject position (cf. 19a–19c). In the terminology of Baker (1988), English is then an asymmetric double object language.4 Theoretically, the phenomenon can be described as a minimality effect; if the order of the objects is changed, the direct object moves (20a–20b). (19)

a. John sent Mary a book. b. Mary was sent a book. c. *A book was sent Mary.

(20)

a. John sent the boy to Munich. b. The boy was sent to Munich.

Icelandic also forms asymmetric passives of double object verbs in that only the closest object moves to subject (irrespective of the case of this object – recall that Icelandic allows for quirky subjects) (cf. 21a–21c).5 Again, movement to subject seems to be driven by a sort of a minimality effect as a DO can in fact rise to subject if the IO is (allowed to be) omitted. 3

Another context showing that the nominative theme can stay in situ in Dutch and German passives are passives of double object verbs (cf. (i)) where the nominative theme canonically follows the dative indirect object; see e.g., den Besten 1985, Haider 1993, Wurmbrand 2006, or Haider, Chapter 15: (i) dass dem that the

4

Angestellten

das Geld

überwiesen wurde.

employee.Dat the money.Nom transferred was

However, there are English dialects that form symmetric double object passives, in that either the IO or the DO can rise to subject (e.g., Haddican and Holmberg 2012).

5

With a few double object verbs such as gefa (give), two variants of the passive are possible. However, the relevant verbs already allow reordering of their objects in the active.

467

468

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

(21)

a. Þeir sviptu manninn vinnunni. (Icelandic) they deprived man.the.Acc work.the.Dat ‘They deprived the man of the work.’ var sviptur vinnunni. b. Maðurinn man.the.Nom was deprived work.the.Dat ‘The man was deprived of his work.’ c. *Vinnunni var sviptur maðurinn. work.the.Dat was deprived man.the.Nom

Interestingly, Faroese differs here from Icelandic. Almost all Faroese double object verbs are of the pattern Nom < Dat < ACC, and only the second object can be raised to subject straightforwardly, though definiteness and heaviness of the DO play a role (Thra´insson et al. 2004: 269ff., Thraı´nsson 2007: 304): (22)

a. Ku´gvin varð seld bo´ndanum. cow.the.Nom was sold farmer.the.Dat ‘The cow was sold to the farmer.’ b. ??Bo´ndanum varð seld ku´gvin. farmer.the.Dat was sold cow.the.Nom c. ?Bo´ndanum varð seld ein ku´gv. farmer-the.Dat was sold a cow.Nom ‘The farmer was sold a cow.’

The Mainland Scandinavian languages, on the other hand, are symmetric double object languages in that either the first or the second object may be promoted to subject under passivization (though promotion of indirect objects is reported to be much more frequent) (Platzack 2005). This is illustrated in (23a–23c) for Swedish. (Swedish s-passives discussed in Section 20.3 show the same symmetric behavior.) (23)

a. Dom erbjo¨d honom ett nytt jobb. they offered him a new job b. Han blev erbjuden ett nytt jobb. he was offered a new job c. Ett nytt jobb blev erbjudet honom. a new job was offered him

German and Dutch lack canonical goal-raising passives of the type in (23b) (see, e.g., Anagnostopoulou 2003). Haegeman (2016) presents evidence from West Flemish that this dialect has goal passives.

20.2.3.3 Expletive Passive Constructions The Mainland Scandinavian languages allow the internal argument to stay inside the verbal phrase if the canonical subject position is filled by an expletive. However, this construction is restricted by

The Voice Domain in Germanic

a definiteness effect on the side of the internal argument as shown in (24) for Norwegian (Å farli 1992): (24)

Det vart mjølka ei ku /*kua. it became milked a cow/cow.the ‘A cow/the cow was milked.’

While English does not allow the direct object to stay in situ in passives, raising to subject is also not always obligatory. In certain sentences featuring the locative expletive there in Spec,TP, the internal argument moves to a position at the edge of the verbal phrase, immediately to the left of the participle (e.g., Chomsky 2001, Holmberg 2001, Julien 2006). As in the Scandinavian languages, the DP in the expletive passive construction cannot be definite. (25)

a. *?There was placed a large book on the table. b. There was a large book placed on the table c. *There was the book placed on the table.

In most varieties of Norwegian (cf. 26a, 26b) as well as in Danish, this short movement of the internal argument is disallowed and the internal argument must follow the participle (Holmberg 2001). (26)

a. Det har blitt skrevet tre bøker om dette. Expl have been written three books about this b. *Det har blitt tre bøker skrevet om dette. Expl have been three books written about this

Swedish (27a, 27b) and some varieties of Norwegian allow the order expletive < DP < participle. Crucially, the Scandinavian languages and varieties that admit this order are the ones that have participle agreement (cf. the discussion about (5); the participle agrees with the DP if and only if the DP precedes the participle). (27)

a. Det har blivit skrivet/*skrivna tre bo¨cker om detta. Expl have been written.n.sg/written.pl three books about this b. Det har blivit tre bo¨cker *skrivet/skrivna om detta. Expl have been three books written.n.sg/written.pl about this

Short A-movement to the immediate left of the participle can also be found in passives of double object verbs as is illustrated in (28) for Swedish and in (29) for English: (28)

Det Expl

blev was

tre three

generaler generals

tilldelade assigned.pl

en a

medalj medal

(29)

There were three soldiers awarded the badge of honor.

var. each

Holmberg (2001) argues that short A-movement to the immediate left of the participle has a semantic effect in that the moved DP should preferably be quantified and not be bare plural. Julien (2006) on the other hand

469

470

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

accepts bare plurals but argues that bare singular nominals are rejected by most speakers in preparticipial position. It turns out that Icelandic does not allow short A-movement to the edge of the verbal phrase. In (30), the nominative DP has moved to Spec,TP and the expletive is a CP-expletive inserted for V2-reasons. This is revealed by the observation that in Icelandic the fronted internal argument must precede all nonfinite auxiliaries, while in Swedish and in English the fronted internal argument follows such auxiliaries (cf. (31a–31c); e.g., Holmberg 2001, but cf. Thra´insson 2007). (30)

Það hafa þja´r bækur þegar verið Expl have three books already been ‘Three books have already been written.’

skrifaðar. (Icelandic) written.pl

(31) a. Það

munu þja´r bækur hafa (*þja´r bækur) Expl will three books have verið (*þja´r bækur) skrifaðar. been written

(Icelandic)

b. Det borde (*tre bo¨cker) ha (*tre bo¨cker) blivit tre bo¨cker Expl should (*three books have been skrivna. (Swedish) written ‘Three books should have been written.’ c. There should (*three books) have (*three books) been three books written. A further difference between Mainland Scandinavian and Icelandic is found in passives of double object verbs. Above, we saw that Icelandic passives allow the internal argument of transitive verbs to stay inside the VP. However, with passives of double object verbs one object must leave the VP and move to Spec,TP (cf. 32a–32c) (Holmberg 2001).6 No such restriction holds in the Mainland Scandinavian languages (cf. Swedish in (33)). (32)

(33)

6

a. *Það voru gefnar stra´knum gjafir. (Icelandic) Expl were given boy.the.Dat presents.Nom b. Stra´knum voru gefnar gjafir. boy.the.Dat were given presents.Nom c. Gjafir voru gefnar stra´knum. presents.Nom were given boy.the.Dat Det Expl

blev was

givet given

pojken boy.the

presenter. presents

Recall from footnote 5 that gefa (give) allows either object to move to Spec,TP as it allows two base orders in the active. In fact, Holmberg reports that both objects can stay inside the VP if the DO precedes the IO. i. Það voru gefnar gjafir öllum litlu krökkunum. Expl were given gifts all little the-kids ‘All the little kids were given gifts.’

The Voice Domain in Germanic

20.2.4 Impersonal Passives In contrast to the expletive passives discussed above, we call passives without an argument DP that could trigger verbal agreement impersonal passives. Prototypical cases are passives of one-place verbs (cf. (34a)). Passives of transitive verbs with a lexically case-marked internal argument as in (34b) also form impersonal passives under this definition. Impersonal passives in the Germanic languages differ in whether they feature an expletive in Spec,TP and in whether they license by-phrases. German impersonal passives lack a subject expletive in Spec,TP and they license by-phrases: (34) a. dass (*es) auf der Party (nur von Maria) getanzt wurde. that it at the party (only by Mary) danced was.sg ‘since people (only Mary) danced at the party.’ b. weil ja den Kindern (von niemandem) geholfen as well the chidren.Dat by no-one helped ‘because the children were helped by no-one.’

wurde. was.sg

Dutch and Afrikaans optionally insert an expletive in impersonal passives. However, the status of Dutch er in (35) as an expletive is disputed; see Haider (Chapter 15) for arguments that Dutch er is not a real expletive but an adverb. For the Afrikaans counterpart daar, however, it has been claimed that it qualifies as a real expletive (e.g., Wurmbrand 2006). (35)

dat (er) (door iedereen) gedanst werd. that there by everyone danced was ‘that everyone danced.’

(Dutch)

West Flemish (also an OV language) uses an expletive obligatorily (Haeberli 2002: 206): (36)

dat *(er) overal that there everywhere ‘that everyone danced’

gedanst danced

wier. was

Turning to the VO languages, Icelandic, like German, disallows an expletive in impersonal passives, while Faroese optionally inserts a pronominal expletive (Vikner 1995: 150): (37)

I´ gær var (*Það) dansað. yesterday was Expl danced ‘Yesterday, people were dancing.’

(38)

Blı´vur (tað) ofta becomes (Expl) often ‘Do people work hard?’

arbeitt worked

(Icelandic)

hart? hard

(Faroese)

471

472

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

Finally, all Mainland Scandinavian languages (as well as Yiddish; Vikner 1995: 209) obligatorily insert a pronominal expletive in their impersonal passives. This is illustrated for Norwegian in (39). (39)

I gaº r ble *(det) danset. yesterday was Expl danced ‘Yesterday, people were dancing.’

While impersonal passives in all West Germanic languages license byphrases, by-phrases are typically rejected in all Scandinavian impersonal passives. Example (42) shows that this also holds for the so-called s-passive discussed in more detail in Section 20.3. (40)

Það var dansað alla no´ttina (*af fo´lkinu). it was danced all night by people

(41)

Tað varð dansað alla na´ttina (*av teimum). there was danced all night.the by them

(42)

Det dansades hela natten (*av dom). there was-danced whole night.the by them

(Icelandic)

(Faroese) (Swedish)

However, in Icelandic, the rejection of by-phrases seems to be a more general property of passives without structural subject as it even holds for passives with VP-internal argument DP (e.g., Thra´insson 2007, Jo´nsson 2009) (cf. 43a versus 43b). In all remaining Scandinavian languages, this type of expletive passive actually allows by-phrases (cf. 44a, 44b) and only real impersonal passives reject by-phrases. (43) a. Einhverjum nemanda var hja´lpað (af kennaranum). (Icelandic) some student.Dat was helped by teacher.the b. Það var hja´lpað einhverjum nemanda (?*af kennaranum). Expl was helped some student.Dat by teacher.the ‘Some student was helped.’ (44) a. Tað bleiv etið eitt su´repli (av næmingunum). (Faroese) Expl was eaten an apple by students.the b. Der blev spist et æble (af studenterne). (Danish) Expl. was eaten an apple by students.the) It is often claimed (in particular for Dutch, German, and Icelandic) that the implicit external argument of impersonal passives must be a human agent. Thus, examples like (45) could not describe situations where the implicit external argument is a bird, a whistle or a train (Frajzyngier 1982). (45)

Es wird gebellt/gepfiffen. it is barked/ whistled ‘Someone barked/whistled.’

(German)

The Voice Domain in Germanic

While correct for Icelandic (Thra´insson 2007, Primus 2011),7 at least for German and Dutch this claim seems too strong. Primus (2011) provides many Dutch (46) and German (47) corpus examples of impersonal passives where the implicit argument is either an animal or even a nonhuman technical device. (46)

Toch hoorden wij dat er but

heard

gemiauwd werd, dus de poes moest er

we that there meowed was, so

the cat

nog

in zitten.

must there still in sit

‘But we heard that there was meowing, so that the cat must have been still in there.’

Finally, even passives of unaccusative verbs have been reported in Dutch and German. However, as Primus (2011) points out, passives of unaccusative verbs always involve coercion of the event structure from telic to homogeneous events; whether this shift involves a change from an unaccusative into an unergative syntax remains an open question. (47)

In Bosnien wird weiter in Bosnia is further ‘In Bosnia dying is going on.’

20.2.5

gestorben. died

The Icelandic New Passive and Passives of Reflexive Verbs

Rather recently Icelandic developed a construction that looks, at least at the surface, like a noncanonical passive, the so-called New Passive (alternatively also called New Impersonal, see Maling and Sigurjo´nsdo´ttir 2002, Thra´insson 2007, Eytho´rsson 2008, Jo´nsson 2009, Sigurðsson 2011; for more detailed references, see also Wood 2015 and Thra´insson 2017). The construction is mainly found within the younger generation and the %-sign on the examples below indicates that a sentence is accepted only by speakers of the New Passive variety. The New Passive involves the ordinary passive participle and auxiliary, and the external argument is not overtly expressed. The following properties set the New Passive apart from canonical passive in Icelandic: (i) Structural accusative on internal arguments is preserved and the auxiliary and the participle show the kind of default agreement known from impersonal passives (48a). (ii) Contrary to VP-internal arguments in canonical passives (48b), the VPinternal argument in the new passive is not subject to a definiteness effect (48a). (iii) Internal argument DPs cannot A-move out of the VP to Spec,TP (48c). (48)

7

a. %Það Expl

var was

lamið beaten.Nom.sg.n.

stu´lkuna girl.the.Acc.sg.f.

ı´ in

klessu. mess

But recall that already personal passives in Icelandic only allow human agents as their implicit external arguments.

473

474

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

stu´lkan ı´ klessu. b. *Það var lamin Expl was beaten.Nom.sg.f. girl.the.Nom.sg.f in mess ‘The girl was badly beaten up.’ c. *Var stu´lkuna lamið ı´ klessu? was girl.the.Acc.sg.f. beaten.Nom.sg.n. In mess With inherently/lexically case-marked internal arguments, the only surface difference between canonical passives involving an expletive and the New Passive pertains to the absence of a definiteness effect. (49)

Það Expl

var was

hrint pushed

%henni/einhverri stelpu her.Dat/ some girl.Dat

ı´ in

sko´lanum. school.the

Two alternative analyses have been proposed for the New Passive. Maling and Sigurjo´nsdo´ttir (2002) suggest an analysis as an active impersonal construction involving an impersonal pro-subject. Such an analysis must stipulate that the particular impersonal pro can appear only as an external argument and it obligatorily appears in the context of passive morphology. Other authors argued for an analysis involving an ordinary passive Voice (with an implicit/absorbed external argument), which, however, triggers special case and A-movement properties; here the question is how to derive the latter properties without stipulation (see Sigurðsson 2011, Scha¨fer 2012, Legate 2014). A phenomenon that figured prominently in the discussion of the New Passive is passives of reflexive verbs as in (50), where an SE-reflexive pronoun is semantically co-valued with the verb’s implicit external argument. (50)

Það var baðað sig a´ Expl was bathed SE.ACC on ‘There was bathing on Saturdays.’

laugardo¨gum. Saturdays (Sigurðsson 1989: 355)

The acceptability of (50) could be seen as support for an analysis of the New Passive as involving a pro-subject that antecedes the reflexive pronoun (Maling and Sigurjo´nsdo´ttir 2002). However, this conclusion is weakened by the observation that sentences like (50) are accepted by most speakers of Icelandic including those that reject (48a) and (49) (e.g., Eytho´rsson 2008). Furthermore, German (and some Norwegian speakers) also allow passives of reflexive verbs as in (50) but German speakers strongly reject an accusative internal argument of the type in (48a) (Scha¨fer 2012).8 8

German allows passives as in (i) where the auxiliary shows default agreement although the internal argument is either plural or sometimes even overtly marked with accusative (Abraham 1986). This is possible, however, only in particular verb-object collocations where the object must be adjacent to the verb and arguably undergoes semantic incorporation into the verb. (i) Jetzt wird

Karten

gespielt/die Zähne geputzt.

now becomes.sg cards.pl played/ the teeth.pl brushed ‘Now one plays cards/brushes the teeth.’

The Voice Domain in Germanic

20.3 Scandinavian S-Passives The Mainland Scandinavian languages are the only Germanic languages with a synthetic passive where a reflexive affix -s is attached to the inflected verb (see Laanemats 2009 for an overview and references).9,10,11 Similar passives formed with reflexive morphology exist in the Slavic and Romance languages.12 (51)

a. En expertgrupp utarbetade rapporten. (Swedish) an expert-group prepare.past report.the ‘A group of experts prepared the report.’ b. Rapporten utarbetades (av en expertgrupp). report.the prepare.past.SE by an expert-group ‘The report was prepared by a group of experts.’

In Swedish, the s-passive can be used in all tenses. In Danish and Norwegian, the s-passive has considerable gaps in the past tense in that compound past tense forms are ungrammatical and only regular verbs of particular conjugation classes can form s-passives in the preterite tense (see Langemaats 2009, Lundquist 2016 for details). Engdahl (2006) investigates the semantic and syntactic factors favoring the s-passive and the periphrastic passive respectively and concludes that the s-passive is the unmarked passive in Swedish but not in Danish and Norwegian.

20.4 Get-Passives Within Germanic, there is a further passive-like construction, whose status as a real passive has been controversially discussed in the literature. As 9

While Icelandic has a very similar reflexive affix -st, this affix does not form passives (see also footnote 14). However, Wood (2015, chapter 6.2) discusses the potential exception of modal contexts where st-marked verbs receive more passive-like interpretations. Nath (2009) reports a reflexive passive in Yiddish, developed under Slavic influence, which is formed with the full SE-reflexive pronoun and does not license by-phrases.

10

Note that -s attaches outside tense and agreement morphology (as well as participial morphology). For this reason, some researchers analyze -s (as well as its Icelandic counterpart -st) as a clitic (Julien 2007, Wood 2015; but see, e.g., Lundquist 2016 for an alternative view).

11

Mainland Scandinavian -s (as well as its Icelandic counterpart -st) derived diachronically from the free-standing SEreflexive pronoun. Different from the SE-reflexive pronouns, the -s affix does not agree in phi-features with any DP in the clause. Furthermore, the -s affix seems to lack reflexivizing force as it is not used to form canonically reflexive verbs, which involve, instead, the full SE-reflexive pronoun. Only some (but not all) inherently reflexive verbs are formed with the -s affix. The -s affix is also found with a small set of basically transitive verbs to form the so-called ‘absolute construction’ in (i) (Lyngfelt and Solstad 2006, Engdahl 2006). Example (i) reminds of antipassives as the object is demoted and is interpreted as being existentially bound. While this construction often has a generic or habitual interpretation, eventive interpretations are possible. (i)

Hunden

bit-s.

the.dog

bite-S E

‘The dog tends to bite/The dog is biting someone right now.’ 12

As analytic passives, s-passives form impersonal passives with a subject expletive and then they reject by-phrases, see (42) above.

475

476

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

ordinary passives, the construction involves a passive participle. However, as this construction differs, among other things, with respect to the auxiliary used, it is sometimes referred to as a noncanonical passive. In English, the verb used in this type of passive is get instead of be, see (52a) versus (52b). (52)

a. John was killed in the war. b. John got killed in an accident.

The English get-passive is special in that it can be applied to ordinary transitive verbs. The other Germanic languages can use a counterpart of the verb ‘get’ (kriegen / bekommen in German, krijgen in Dutch, kry in Afrikaans, or faº in Norwegian) to form a passive only with ditransitives or dative verbs (see, e.g, Wood and Sigurðsson 2014). A Norwegian example from Lødrup (1996) is given in (53). The range of thematic roles that the subject has in these get-passives corresponds to those available for dative arguments across languages. Typically, they are that of beneficiary, maleficiary or recipient. Thus in (53), Per is interpreted as the beneficiary of the repairing event, a thematic role realized in the active with dative. Crucially, while canonical passives preserve dative case on internal arguments (with some exceptions in Faroese discussed in Section 20.2.3.1), in get-passives the dative is always shifted to nominative. This is illustrated for Icelandic in (54a) and (54b), the former a canonical passive and the latter a get-passive: (53)

(54)

Per fikk reparert bilen. Per got repaired car.the ‘Per got the car repaired.’ a. Me´r var borgað. me.Dat was paid ‘I was paid.’ b. E´g fe´kk borgað. I.Nom got paid ‘I got paid.’

There is much dialectal variation with respect to the availability of this type of passivization. As Alexiadou and Scha¨fer (2013) summarize, the variation relates to whether the get-passive is possible both with monotransitive and di-transitive verbs taking dative arguments, the type of auxiliary used, e.g., kriegen versus bekommen in German, and the type of predicate, e.g., it is possible with some ditransitive predicates but not all. As an illustration of the first point of variation, we note that in the dialect of German spoken in Luxemburg a monotransitive verb taking a dative object such as drohen (to threaten) can form a get-passive, while this is not possible in standard German: (55)

De the

Mann man

kritt gets

gedroot. threatened

The Voice Domain in Germanic

Assuming that case alternation in passivization suggests the structural status of the alternating case, the above examples point to the structural nature of dative. However, the status of (53), (54b) or (55) as a passive construction has been controversially discussed in the literature. Several researchers hold that these are not truly passive constructions. We refer the reader to Lødrup (1996), Reed (2011), Alexiadou (2012), Alexiadou et al. (2014a), or Wood and Sigurðsson (2014) for discussion of this debate. In particular, it has been argued that in all Germanic languages, the get-verb involved in this type of passive does not pattern with passive auxiliaries. For example, in English, it does not invert in yes-no questions (Haegeman 1985), in Norwegian and Icelandic it can also undergo a Voice alternation itself, i.e., appear with -s/-st (Lødrup 1996; Wood and Sigurðsson 2014). For this reason, it seems to fit the label of a semifunctional head (see the contributions in Corver and van Riemsdijk 2001). English get-passives have been argued to be ambiguous between an eventive passive reading and a stative passive interpretation (Reed 2011, Alexiadou 2012). Alexiadou (2012) proposes that get in English realizes a v-head introducing event implications, and the structure of its complement can include either a fully fledged verbal participial construction or simply an adjectival participle, cf. McIntyre (2012). Matters are different in the other Germanic languages. For instance, German and Dutch get-verbs have been argued to be realizations of a complex structure involving v and a P incorporating into v (Alexiadou et al. 2014a, building on Taraldsen 2010). This P heads an Applicative Phrase that introduces the dative argument, explaining why the surface subject has a thematic role that is typically associated with the dative. Scandinavian get-verbs are analyzed similarly in Wood and Sigurðsson (2014), i.e., as spelling out a v head together with its applicative complement. This explains why in all Germanic languages with the exception of English get-passives are only acceptable with ditransitive or dative verbs. Wood and Sigurðsson actually generalize their analysis to all Germanic get-verbs and propose that the reason why all other Germanic languages lack the counterpart of (52b) is because in English get is ambiguous between the ‘general Germanic get’ (containing Appl) and a meaning close to become.

20.5

Anticausatives

The term “anticausative verb” is meant to refer to the intransitive version of verbs undergoing the causative alternation (56a, 56b). Anticausative verbs express a change of state of their sole argument. In their transitive use, these verbs are called lexical causative verbs as the transitive use can roughly be paraphrased with a periphrastic causative construction “cause to V-intransitive.”

477

478

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

(56)

a. The boy/The high pressure broke the window. (lexical causative) b. The window broke. (anticausative)

As with passives, the subject of an anticausative verb bears the same semantic relation to the verbal event as the object in the corresponding causative verb (i.e., anticausatives are prototypical unaccusative verbs). Crucially, anticausatives differ from passives in that they lack an implicit external argument. This has already been illustrated with a number of tests in Section 20.2, such as the nonlicensing of by-phrases, of instrumental PPs or purpose clauses. It is worth mentioning in this connection that, across languages, anticausatives can combine with PPs introducing nonhuman causer arguments (Alexiadou et al. 2015 and references there). (57)

The window broke (from the high pressure/*from the boy).

Based on (57), one might think that anticausatives involve an implicit argument, which must, however, be a nonhuman causer (Kalluli 2007). However, PPs such as from the high pressure in (57) differ from passive byphrases in that they do not take up an implicitly represented external argument but they add a causal modifier to a one-place change-of-state event. One test to show that anticausatives lack any kind of implicit external argument is the so-called by-itself test (Alexiadou et al. 2015, Scha¨fer and Vivanco 2016). As mentioned in Section 20.2, anticausatives but not passives allow modification with by itself. Crucially, even passives of verbs like ‘wash ashore’ or ‘blow away’ in (58b), which select for nonhuman causer external arguments, reject modification with by itself. The anticausative versions of such verbs (cf. 58b’), on the other hand, allow by itself. If an anticausative were to involve an implicit causer, (58b’) should behave on a par with (58b). (58)

a. But you didn’t take care of the trampoline. b. That’s not true! I did everything I was supposed to ??but the trampoline was blown away by itself. b’. That’s not true! I did everything I was supposed to but the trampoline blew away by itself.

Across languages, anticausatives qualify semantically and syntactically as unaccusative verbs (e.g., Scha¨fer 2009, Scha¨fer and Vivanco 2016). They vary, however, in their morphosyntactic shape. Most English anticausatives are morphologically identical with their causative counterparts (unmarked anticausatives, cf. 56a, 56b above). However, with a small set of verbs the causative alternation is reflected morphologically by a change in the verb’s stem vowel (ablaut; 59a, 59b). The remaining West Germanic languages also have a few verbs with this marking pattern; it is more frequent in the Scandinavian languages.

The Voice Domain in Germanic

(59)

a. They raise/*rise the prices. b. The prices rise/*raise.

While German, Dutch, and the Mainland Scandinavian languages have unmarked anticausatives of the English type, they also have a set of anticausatives that obligatorily combine with a reflexive pronoun (60), an SEreflexive pronoun in the terminology of Reinhart and Reuland (1993). This is, from a morphological perspective, the very same reflexive pronoun that is used to form semantically reflexive verbs such as body care verbs in (61). (60)

Die Tu¨r hat *(sich) the door has SE ‘The door opened.’

geo¨ffnet. opened

(61)

Das Kind hat sich the child has SE ‘The child washed.’

gewaschen. washed

(SE-marked anticausative)

(semantically reflexive verb)

SE-marked anticausatives are frequent, actually dominant in German and the Mainland Scandinavian languages. Dutch, on the other hand, has only very few SE-marked anticausatives (Everaert 1986), and Frisian and Afrikaans lack them as they do not have an SE-reflexive pronoun. Languages with unmarked and SE-marked anticausatives typically have a small set of anticausatives that optionally come with or without the SEreflexive pronoun; this is illustrated below for Dutch: (62)

a. De suiker is/∗heeft opgelost. the sugar is/has dissolved b. De suiker heeft/∗is zich opgelost. the sugar has/is SE dissolved ‘The sugar has dissolved.’

Based on optional pairs as in (62a,62b), a number of observations can be made: First, in languages with a nuanced system of auxiliary selection such as German or Dutch, unmarked anticausatives, as all plain unaccusative verbs, select be, while SE-marked anticausatives select have. (In the Scandinavian languages, be-selection is either absent or restricted to unaccusatives expressing change of location.) Second, the two versions of the anticausative in (62a, 62b) do not differ in meaning (Ackema and Sorace 2017, see Scha¨fer 2008, Alexiadou et al. 2015 and Scha¨fer and Vivanco 2016 for further discussion). More precisely, anticausatives express pure unaccusative one-place predications, irrespective of their morphological make up. Note that this means that the SE-reflexive pronoun in anticausatives does not trigger any reflexive semantics (Scha¨fer and Vivanco 2016, though see Lundquist et al. 2016 for a different view). The main theoretical challenge is then to explain why SE produces reflexive semantics in (61) but not in (60 and 62b) (cf. Scha¨fer 2008, Alexiadou et al. 2015 for discussion).

479

480

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

The morphological variation in the domain of anticausatives is more nuanced in Icelandic and Faroese (Thra´insson 2004, 2007, Wood 2015 and the references there.) First, Icelandic has unmarked anticausatives and anticausative formed by a stem alternation. Next, Icelandic has reflexively marked anticausatives. However, Icelandic does not use its free-standing SE-reflexive pronoun sig to mark anticausatives, but it uses an affix -st, which is historically derived from the SE-reflexive pronoun.13,14 (63)

a. Tru´ðurinn opnaði hurðina. clown.the.Nom opened door.the.Acc ‘The clown opened the door.’ b. Hurðin opnaðist. door.the.Nom opened.SE ‘The door opened.’

Finally, some Icelandic anticausatives are marked with the affix -na which appears inside tense and agreement morphology. (64)

a. Jo´n hitaði vatnið. John.Nom heat.3.sg.pst water.the.Acc ‘John heated the water.’ b. Vatnið hitnaði. water.the.Nom heat.na.3.sg.pst ‘The water heated.’

While causative verbs typically assign structural accusative to their internal argument, some Icelandic causative verbs mark their object with dative (cf. 65a); crucially, while this dative is preserved under passivization (65b), it shifts to nominative in the corresponding anticausatives (65c) (just as it does in adjectival passives, cf. 14c above) (Zaenen and Maling 1990):

13

There is an interesting difference between Icelandic and the Mainland Scandinavian languages. While both have an SEreflexive pronoun and a reflexive -s/-st affix, Icelandic only uses the affix to form anticausatives while the Mainland Scandinavian languages typically use the SE-reflexive pronoun to form anticausatives. However, it is unclear whether causative verbs in Mainland Scandinavian languages necessarily receive a passive interpretation when they combine with the -s affix or whether anticausative semantics are possible, too (p.c. Björn Lundquist).

14

While the set of verbs undergoing the causative alternation is relatively stable across languages, Icelandic allows forming st-marked anticausatives from verbs that would not form anticausatives in other Germanic languages. An example is given in (ib). One might be tempted to analyze examples as (ib) as s-passives of the type found in the Mainland Scandinavian languages (Section 20.3). However, these Icelandic examples differ from s-passives in that they do not pass any tests for implicit agentivity such as the licensing of by-phrases, agentive adverbs or purpose clauses (cf. also footnote 9). Still, since examples like (ib) stand somewhere between prototypical anticausatives and prototypical passives, the literature typically uses the term “st-middle” for them. See Thrá insson (2007) and Wood (2015) for detailed discussion. (i) a. b.

Útlendingar veiddu laxana í Laxá . foreigners caught salmons.the in Salmon River Laxarnir veiddust í Laxá (*af útlendingum). salmons.the caught.SE in Salmon River by foreigners

The Voice Domain in Germanic

(65)

a. A´sta splundraði ru´ðunni. A´sta.Nom shattered window.the.Dat ´ ‘Asta shattered the window.’ b. Ru´ðunni var splundrað. window.the.Dat was shattered ‘The window was shattered.’ c. Ru´ðan splundraðist. window.the.Nom shattered.SE ‘The window shattered.’

However, with some verbs undergoing the causative alternation, the object case of the transitive causative variant is preserved in the anticausative variant. The phenomenon can be found both, with dative objects (66a, 66b) and with accusative objects (67a, 67b). Examples as in (67b) received the term “fate accusatives” as such strings “always have a narrow, semi-idiomatic fate meaning, absent from the transitive and the passive” of the respective verbs (Sigurðsson 2006: 25). It remains unclear whether the same fate semantics (ever) hold in cases where dative is preserved, as in (66b). Note finally that case preserving anticausatives are never marked with -st. For different theoretical treatments of case preserving anticausatives, see Haider (2001), Svenonius (2002, 2006), Sigurðsson (2006), or Scha¨fer (2008, 2012). (66)

a. Jo´n lauk so¨gunni. John.Nom ended story.the.Dat ‘John ended the story.’ b. So¨gunni lauk. story.the.Dat ended ‘The story ended.’

(67)

a. Stormurinn rak ba´tinn storm.the.Nom drove boat.the.Acc ‘The storm drove the boat ashore.’ b. Ba´tinn rak a´ land. boat.the.Acc drove on land ‘The boat drifted ashore.’

20.6

Middles

a´ on

land. land

20.6.1 Canonical Middles Canonical generic middles ascribe a dispositional property to a verb’s understood object (e.g., Lekakou 2005, Ackema and Schoorlemmer 2006). Generic middle sentences are thus nonepisodic and do not make reference to an actual event. The verb’s external agent argument is demoted and receives an arbitrary interpretation similar to a generic indefinite one. The implicit argument of middles, while semantically understood to be

481

482

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

present, is less active than the one in passives: Generic middles do not license by-phrases, agentive adverbials, or control into purpose clauses; instrumental PPs, on the other hand, are good in middles. (68)

a. This wall paints easily (*on purpose) (*by Harry) (with a pertinent brush). b. *Walls paint best to protect them against the rain.

It has long been noted in the literature that English, Dutch, and German middles typically feature an adverb like easily. However, the adverb is not obligatory as adverbless middles greatly improve in the presence of negation, a modal, or focus on the verb (69a–69d) (e.g., Ackema and Schoorlemmer 2006). (69)

a. b. c. d.

This book reads *(easily). This book doesn’t read. Bureaucrats may bribe, but you never know. This rock DOES cut after all!

By far, not every verb allows middle formation. Roughly activity and accomplishment verbs are a legitimate input to middle formation in Germanic, while achievements and stative verbs are illicit (Fagan 1992, Lekakou 2005, Ackema and Schoorlemmer 2006). Since the subject of middles is thematically an internal argument, it is interesting to compare them with anticausatives. Just like anticausatives, English middles are morphologically unmarked. Interestingly, if we form a middle of the small class of verbs showing stem alternation in the causative alternation, we see that the stem of the transitive form is used in the middle (Fellbaum 1986, Lekakou 2005). (70)

a. John raises his kids very strictly. (causative) b. The sun rises from the East. (anticausative) c. Obedient daughters raise/*rise more easily than disobedient sons. (middle)

Dutch also forms its middles without any special morphological marking (though there are Dutch dialects that form their middles as German below). However, different from unmarked anticausatives which select be, Dutch middles select have in the perfect. (71)

Dit soort boeken heeft/*is altijd this sort books has/is always ‘This kind of books always sold well.’

goed well

verkocht. sold

Afrikaans and Frisian also form morphologically unmarked middles. (72)

Dit boek leˆst maklik. this book reads easily ‘This book reads easily.’

(Frisian, Lekakou 2005)

The Voice Domain in Germanic

German marks its middles obligatorily with the SE-reflexive pronoun just as SE-marked anticausatives. (73)

Das the

Buch book

liest reads

*(sich) SE

leicht. easily

There is lots of discussion as to whether middles are unergative or unaccusative predicates. Based on their morphological behavior and auxiliary selection, Dutch and English middles seem to behave as unergatives (e.g., Lekakou 2005, Ackema and Schoorlemmer 2006). Scha¨fer (2008) and Pitteroff (2014) show that German middles are unaccusative in that the full DP is merged as an internal argument; thereby they behave exactly as SEmarked anticausatives discussed in Section 20.5. The theoretical question is how to make sense of this difference between languages. Lekakou (2005) argues that “middle” is not a particular Voice, but it is a particular interpretation that is parasitic on a particular syntax. In English and Dutch, the middle interpretation is parasitic on an unergative syntax; in German, it could be parasitic on the syntax of SE-marked anticausatives. Both German and Dutch, but not English, allow middles of unergative verbs, i.e., impersonal middles. These differ from impersonal passives in that a pronominal expletive is obligatory; recall that German impersonal passives reject the expletive and Dutch impersonal passives optionally use the locative expletive. (74)

In dieser Disco tanzt *(es) sich in this club dances it SE ‘It is pleasant to dance in this club.’

(75)

Het /*er loopt prettig op deze it/there walks comfortably on these ‘These shoes are comfortable to walk in.’

angenehm. pleasantly

schoenen. shoes

As observed in Lekakou (2005), impersonal middles obligatorily come with some kind of adjunct PP (temporal or locational – cf. in this club / in these shoes) and the dispositional description that characterizes the middle semantics is applied to the DP inside of these PPs (club/shoes). As an alternative to impersonal middles of the type in (75), Dutch, Frisian, and Afrikaans, but not German or English, form so-called adjunctmiddles as in (76), where the complement of the preposition is raised to subject and the preposition itself is not realized anymore. Ackema and Schoorlemmer (2006) show that the promoted DP in “adjunct-middle formation” originates from a PP that actually behaves more like an argument. That not any PP, in particular not any adjunct PP can participate in this kind of middle is shown in (77a, 77b): (76)

Deze schoenen lopen prettig. (Dutch) these shoes walk comfortably ‘These shoes are comfortable to walk in.’

483

484

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

(77)

a. Het rijdt niet prettig met grote it rides not comfortably with great ‘It’s not comfortable to drive in a great hurry.’ b. *Grote haast rijdt niet prettig. Great hurry drives not comfortably

haast. hurry

Icelandic forms generic middles with the same st-affix that is also used in anticausatives. As with middles in other Germanic languages the Icelandic middle does not license a by-phrase. (78)

Rafmagnsbı´lar seljast (vel) he´r (*af so¨lumo¨nnum). electric.cars.Nom sell.SE well here by salesmen ‘Electric cars sell well here (*by salesmen).’

As with anticausatives, in Icelandic middles a dative object can raise to nominative: (79)

a. Við læsum dyrunum. we.Nom lock door.the.Dat ‘We’re locking the door.’ (Svenonius 2006) b. Þessi hurð læsist ekki this door.Nom locks.SE not ‘This door doesn’t lock.’ (Maling 2001)

However, this option is restricted to a particular type of dative, the dative assigned to the theme of some causative verbs. Other noncausative verbs with lexically marked object like “help” do not allow shifting the dative to nominative. In this case, middle formation is not possible. (80)

*Gamalt fo´lk hja´lpast illa. old people.Dat help.SE poorly ‘It is difficult to help old people.’

(Maling 2001)

German is fundamentally different here as shown in Steinbach (2002). A verb like helfen (help), which marks its object with lexical dative, can form a middle even though the dative is not shifted to nominative. Just as in impersonal middles above, the expletive becomes obligatory. (81)

Einem Obdachlosen hilft a homeless-person.Dat helps ‘It is easy to help a homeless person.’

*(es) it

sich SE

leicht. easily

Descriptively, the difference between German and Icelandic seems to be that while both languages allow impersonal passives (under exclusion of an expletive), only German allows impersonal middles (which must feature an expletive, cf. 74); Icelandic simply seems to lack impersonal middles where no nominative DP is available to trigger verbal agreement, an interesting point of variation without any immediate explanation.

The Voice Domain in Germanic

(82)

?*He´r dansast þægilega. here dances.SE nicely ‘One can dance well here’ (p.c. Halldo´r A´rmann Sigurðsson)

Middles in Swedish and Norwegian are discussed in Klingvall (2007) and Fa´bregas and Putnam (2014). It turns out that these languages neither form unmarked middles of the English and Dutch type nor reflexively marked middles of the German or Icelandic type. As Fa´bregas and Putnam show, Norwegian s-passives as in (83a, 83b) (diagnosed as passives by the presence of a by-phrase) can receive a dispositional meaning only for few speakers. Most speakers can interpret such examples only as habitual passives. The Swedish counterparts of (83a, 83b) only receive a habitual passive interpretation. (83) a. Denne typen hus gjen-opp-bygge-s lett av alle. this type house again-up-build-SE easy of everyone ‘This kind of house can easily be rebuilt by anyone.’ b. Denne bandasjen fjerne-s lett fra huden. this bandage.the removes-SE easily from skin.the ‘This bandage is easy to remove from the skin.’ In order to express the dispositional meaning associated with the term “middle”, the Mainland Scandinavian languages use alternative means. Swedish and Norwegian use the construction in (84) involving a participle compounded with an adjective like ‘easy’, ‘difficult’, ‘fast’, or ‘slow’; byphrases are not possible. (84)

Den this

ha¨r here

boken book.the

a¨r is

la¨tt-la¨st easy-read

(*av nunnor). by nuns

(Swedish)

Finally, in Danish, a tough-construction must be used to express dispositional middle semantics (p.c. Sten Vikner). (85)

Den this

her here

bog book

er is

nem easy

at to

læse. read

20.6.2 Let-Middles A number of Germanic languages form so-called let-middles (see Pitteroff 2014, 2015 for a detailed discussion). In (86a) we see a German canonical middle; (87b) features the corresponding let-middle. (86)

a. Das Buch liest sich the book reads SE b. Das Buch la¨sst sich the book lets SE ‘The book reads easily.’

leicht. easily leicht lesen. easily read

485

486

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

As the translation shows, (86a, 86b) have basically the same meaning. In fact, let-middles share many properties with canonical middles: They typically express dispositional statements about their subject DP. The subject corresponds to the theme argument of the embedded predicate (lesen ‘to read’), and the external argument of this predicate is suppressed. However, let-middles differ from canonical middles in at least three ways, as illustrated by the contrast between (87a) and (87b): (i) Adverbial modification is not necessary, (ii) By-phrases are allowed, (iii) in impersonal let-middles, the expletive is not obligatory: (87)

a. Hier tanzt *(es) sich (*von jedem) *(gut). here dances Expl SE by everyone well b. Hier la¨sst (es) sich (von jedem) (gut) tanzen. there lets Expl SE by everyone well dance ‘One can dance well there.’

Furthermore, while canonical middles are obligatorily generic, let-middles allow episodic readings without any problem; they seem to behave then more like passives. (88)

Das Problem ließ sich endlich lo¨sen. the problem let SE finally solve ‘Finally, the problem could be solved.’

Finally, let-middles are lexically less restricted than canonical middles (though they are not as productive as passives (cf. Pitteroff 2014)). In (89a, 89b) we see that an achievement predicate rejects ordinary middle formation but allows the formation of a let-middle. (89)

a. *Diese Krankheit erkennt sich leicht. this sickness recognizes SE easily b. Diese Krankheit la¨sst sich leicht erkennen. this sickness lets SE easily recognize ‘This sickness can be recognized easily.’

Some of these differences between canonical and let-middles become less surprising if one understands that the verb ‘let’ is canonically used to form periphrastic causative constructions. We illustrate this here with Dutch examples (for let-middles in Dutch, see Everaert 1986, Ackema and Schoorlemmer 2006, Pitteroff 2014, 2015, Broekhuis and Corver 2015). In the periphrastic causative construction in (90a), ‘let’ takes a human causer subject and embeds a transitive infinitive whose subject is case marked in ECM-style. In (90b), we see that the embedded infinitive may also lack an overt subject; however, the subject is implicitly understood to be present and it can be taken up with a by-phrase; thus, the complement infinitive of ‘let’ behaves

The Voice Domain in Germanic

semantically as a passive. In the let-middle in (90c) finally the subject of the causative matrix verb is deleted and the reflexive pronoun appears instead. In Pitteroff (2014, 2015) it is argued at length that (90c) actually features an SE-marked anticausative version of ‘let’ and the embedded infinitive is syntactically and semantically identical in (90c) and (90b). The licensing of by-phrases then is not a particular property of let-middles, but it is a general property of the complement of the verb ‘let’, whether it is used as an active causative verb or as a middle/anticausative verb. (90)

a. Marie laat Jan de truitje wassen. Marie lets Jan the sweater wash ‘Marie made Jan wash the sweater.’ b. Marie laat de truitje (door Jan) wassen. Marie lets the sweater by Jan wash ‘Marie made Jan wash the sweater.’ c. Het truitje laat zich gemakkelijk wassen. this sweater lets SE easily wash ‘The sweater washes easily.’

While both German (91a) and Dutch (91b) allow impersonal canonical middles, only German allows impersonal let-middles (Pitteroff 2014): (91)

a. Es la¨sst sich gut auf diese Musik tanzen. it lets SE well on that music dance b. *Het laat zich lekker dansen op die muziek. it lets SE nicely dance on that music ‘It is nice to dance to that music.’

The Mainland Scandinavian languages also have let-middles. Danish (Pitteroff 2014) and Norwegian (p.c. Bjo¨rn Lundquist) let-middles are identical to those in German and Dutch: (92)

Bogen lader sig let book.the lets SE easily ‘The book reads easily.’

læse. read

(Danish)

Swedish differs minimally in that it realizes the embedded infinitive as an infinitival s-passive (Klingvall 2007). (93)

Boken laº ter sig la¨sas book.the lets SE read.SE ‘The book reads easily.’

utan without

svaº righet. difficulty

(Swedish)

A less productive Icelandic construction which shares some but not all properties of let-middles is discussed in Wood (2015: 267ff.).

487

488

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

References Abraham, W. 1986. “Unaccusatives in German,” GAGL: Groninger Arbeiten zur germanistischen Linguistik 28: 1–72. Ackema, P. and M. Schoorlemmer 2006. “Middles.” In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax, Vol. III. Oxford: Blackwell: 131–203. Ackema, P. and A. Sorace 2017. “Auxiliary Selection.” In M. Everaert and H. van Riemsdijk (eds.), The Wiley-Blackwell Companion to Syntax, 2nd edn. Oxford: Wiley-Blackwell: https://doi.org/10.1002/9781118358733 .wbsyncom072. Å farli, T. A. 1992. The Syntax of Norwegian Passive Constructions. Amsterdam: John Benjamins. Å farli, T. A. 2009. “Passive participle agreement in Norwegian dialects,” Groninger Arbeiten zur germanistischen Linguistik 49: 167–181. Alexiadou, A. 2012. “Non-canonical passives revisited: Parameters of non-active voice,” Linguistics 50: 1079–1110. Alexiadou, A., E. Anagnostopoulou, and C. Sevdali 2014a. “Opaque and transparent datives, and how they behave in passives,” Journal of Comparative Germanic Linguistics 17: 1–34. Alexiadou, A., B. Gehrke, and F. Scha¨fer 2014b. “The argument structure of adjectival participles revisited,” Lingua 149: 118–138. Alexiadou, A., E. Anagnostopoulou, and F. Scha¨fer 2015. External Arguments in Transitivity Alternations: A Layering Approach. Oxford University Press. Alexiadou, A., E. Anagnostopoulou, and F. Scha¨fer 2018. “Passive.” In N. Hornstein, H. Lasnik, P. Patel-Grosz, and C. Yang (eds.), ‘Syntactic Structures’ 60 Years On. The Impact of the Chomskyan Revolution in Linguistics. Berlin: Mouton de Gruyter. Alexiadou, A. and F. Schafer 2013 “Non-canonical passives.” In A. Alexiadou and F. Scha¨fer (eds.), Non-canonical Passives. Amsterdam and Philadelphia: John Benjamins: 1–19. Anagnostopoulou, E. 2003. The Syntax of Ditransitives: Evidence from Clitics. Berlin: Mouton de Gruyter. Baker, M. 1988. Incorporation: A Theory of Grammatical Function Changing. University of Chicago Press. Baker, M., K. Johnson, and I. Roberts 1989. “Passive arguments raised,” Linguistic Inquiry 20: 219–252. Besten, H. den 1985. “The Ergative hypothesis and free word order in Dutch and German.” In J. Toman (ed.), Studies in German Grammar. Dordrecht: Foris: 23–64. Broekhuis, H., N. Corver, and R. Vos 2015. Syntax of Dutch: Verbs and Verb Phrases. Amsterdam University Press. Bruening, B. 2013. “By-Phrases in Passives and Nominals,” Syntax 16: 1–41.

The Voice Domain in Germanic

Bruening, B. 2014. “Word Formation is Syntactic: Adjectival Passives in English,” Natural Language and Linguistic Theory 32: 363–422. Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton de Gruyter. Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. 2001. “Derivation by phase.” In M. Kenstowicz (ed.), Ken Hale: A Life in Language. Cambridge, MA: MIT Press: 1–52. Collins, C. 2005. “A smuggling approach to the passive in English,” Syntax 8: 81–120. Corver, N. and H. van Riemsdijk 2001. Semi-Lexical Categories: The Content of Function Words and the Function of Content Words. Berlin: Mouton de Gruyter. Engdahl, E. 2006. “Semantic and syntactic patterns in Swedish passives.” In B. Lyngfelt and T. Solstad (eds.), Demoting the Agent – Passive, Middle and Other Voice Phenomena. Amsterdam: John Benjamins: 21–46. Everaert, M. 1986. The Syntax of Reflexivization. Dordrecht: Foris. Eytho´rsson, T. 2008. “The New Passive in Icelandic really is a passive. In T. Eytho´rsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers. Amsterdam: John Benjamins: 173–219. Eytho´rsson, T. and H. Thra´insson 2017. “Variation in oblique subject constructions in Insular Scandinavian.” In H. Thra´insson, C. Heycock, H. Petersen, and Z. Hansen (eds.), Syntactic Variation in Insular Scandinavian. Amsterdam: John Benjamins: 54–90. Fa´bregas, A. and M. T. Putnam 2014. “The emergence of middle voice readings with and without agents,” The Linguistic Review 31: 193–240. Fagan, S. 1992. The Syntax and Semantics of Middle Constructions: A Study with Special Reference to German. Cambridge University Press. Fellbaum, C. 1986. On the Middle Construction in English. Bloomington: Indiana University Linguistics Club. Fox, D. and Y. Grodzinsky 1998. “Children’s passive: A view from the by phrase,” Linguistic Inquiry 29: 311–332. Frajzyngier, Z. 1982. “Indefinite agent, passive and impersonal passive. A functional study,” Lingua 58: 267–290. Haddican, W. and A. Holmberg 2012. “Object movement symmetries in British English dialects: Experimental evidence for a mixed case/locality approach,” Journal of Comparative Germanic Linguistics 15: 189–212. Haeberli, E. 2002. Features, Categories and the Syntax of A-Positions: CrossLinguistic Variation in the Germanic Languages. Dordrecht: Kluwer. Haegeman, L. 1985. “The get passive and Burzio’s Generalization,” Lingua 66: 53–77. Haegeman, L. 2016. The canonical goal passive in Dutch and its dialects. Ms., lingbuzz/003105. Haider, H. 1993. Deutsche Syntax - Generativ. Tu¨bingen: Gunter Narr. Haider, H. 2001. “How to stay accusative in insular Germanic,” Working Papers in Scandinavian Syntax 68: 1–14.

489

490

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

Holmberg, A. 2001. “Expletives and Agreement in Scandinavian Passives,” Journal of Comparative Germanic Linguistics 4: 85–128. Jaeggli, O. 1986. “Passive,” Linguistic Inquiry 17: 587–622. Jo´nsson, J. G. 2009. “The new impersonal as a true passive.” In A. Alexiadou, J. Hankamer, T. McFadden, J. Nuger, and F. Scha¨fer (eds.), Advances in Comparative Germanic Syntax. Amsterdam: John Benjamins: 281–306. Julien, M. 2006. “On argument displacement in English and Scandinavian,” Working Papers in Scandinavian Syntax 77: 1–69. Julien, M. 2007. “On the relation between morphology and syntax.” In G. Ramchand and C. Reiss (eds.), The Oxford Handbook of Linguistic Interfaces. Oxford University Press: 209–238. Kallulli, D. 2007. “Rethinking the passive/anticausative distinction,” Linguistic Inquiry 38: 770–780. Klingvall, E. 2007. (De)composing the middle: A Minimalist Approach to Middles in English and Swedish. Doctoral dissertation, Lund University. Kratzer, A. 1996. “Severing the external argument from its verb.” In J. Rooryck and L. Zaring (eds.), Phrase Structure and the Lexicon. Dordrecht: Kluwer: 109–137. Laanemets, A. 2009. “The passive voice in written and spoken Scandinavian.” In M. Fryd (ed.), The Passive in Germanic Languages. GAGL: Groninger Arbeiten zur germanistischen Linguistik 49: 144–166. Legate, J. A. 2014. Voice and v: Lessons from Acehnese. Cambridge, MA: MIT Press. Lekakou, M. 2005. In the Middle, Somewhat Elevated: The Semantics of Middles and Its Cross-linguistic Realization. Ph.D. dissertation, University of London. Levin, B. and M. Rappaport Hovav 1995. Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge, MA: MIT Press. Lødrup, H. 1996. “The theory of complex predicates and the Norwegian faº ‘get’,” Working Papers in Scandinavian Syntax 57: 76–91. Lundquist, B. 2016. “The role of tense-copying and syncretism in the licensing of morphological passives in the Nordic Languages,” Studia Linguistica 70: 180–220. Lundquist, B., M. Corley, M. Tungseth, A. Sorace, and G. Ramchand 2016. “Anticausatives are semantically reflexive in Norwegian but not in English,” Glossa: A Journal of General Linguistics 1: 1–30. Lyngfelt, B. and T. Solstad 2006. “Perspectives on demotion: Introduction to the volume.” In B. Lyngfelt and T. Solstad (eds.), Demoting the Agent – Passive, Middle and Other Voice Phenomena. Amsterdam: John Benjamins: 1–20. Maling, J. 2001. “Dative: The heterogeneity of the mapping among morphological case, grammatical functions, and thematic roles,” Lingua 111: 419–464. Maling, J. and S. Sigurjo´nsdo´ttir 2002. “The ‘New Impersonal’ construction in Icelandic,” Journal of Comparative Germanic Linguistics 5: 97–142.

The Voice Domain in Germanic

McFadden, T. 2004. The Position of Morphological Case in the Derivation: A Study on the Syntax-Morphology interface. Ph.D. dissertation, University of Pennsylvania. McIntyre, A. 2012. “The become = cause hypothesis and the polysemy of get,” Linguistics 50: 1251–1281. McIntyre, A. 2013. “Adjectival passives and adjectival participles in English.” In A. Alexiadou and F. Scha¨fer (eds.), Non-canonical Passives. Amsterdam: John Benjamins: 21–42. Nath, H. 2009. “The passive in Soviet Yiddish.” In M. Fryd (ed.), The Passive in Germanic Languages. GAGL: Groninger Arbeiten zur germanistischen Linguistik 49: 182–199. Pitteroff, M. 2014. Non-canonical lassen-middles. Doctoral dissertation, Universita¨t Stuttgart. Pitteroff, M. 2015. “Non-canonical middles: A study of personal let-middles in German,” Journal of Comparative Germanic Linguistics 18: 1–64. Platzack, C. 2005. “Cross-Germanic Promotion to Subject in Ditransitive Passives – a Feature-Driven Account.” In M. Vulchanova and T. A. Å farli (eds.), Grammar and Beyond. Essays in honour of Lars Hellan. Oslo: Novus Forlag: 135–161. Primus, B. 2011. “Animacy and telicity: Semantic constraints on impersonal passives,” Lingua 121: 80–99. Reed, L. 2011. “Get-passives,” The Linguistic Review 28: 41–78. Reinhart, T. and E. Reuland 1993. “Reflexivity,” Linguistic Inquiry 24: 657–720. Scha¨fer, F. 2008. The Syntax of (Anti-)Causatives: External Arguments in Changeof-State Contexts. Amsterdam: John Benjamins. Scha¨fer, F. 2009. “The Causative Alternation,” Language and Linguistics Compass 3: 641–681. Scha¨fer, F. 2012. “The passive of reflexive verbs and its implications for theories of binding and case,” Journal of Comparative Germanic Linguistics 15: 213–268. Scha¨fer, F. and M. Vivanco 2016. “Anticausatives are weak scalar expressions, not reflexive expressions,” Glossa: A Journal of General Linguistics 1.18: 1–36. Schlu¨cker, B. 2009. “Passive in German and Dutch: The sein / zijn + past participle construction.” In M. Fryd (ed.), The Passive in Germanic Languages. GAGL: Groninger Arbeiten zur germanistischen Linguistik 49: 96–124. Sigurðsson, H. A´. 1989. Verbal Syntax and Case in Icelandic. Doctoral dissertation, University of Lund. Sigurðsson, H. A´. 2004. “Agree and agreement: Evidence from Germanic.” In W. Abraham (ed.), Focus on Germanic Typology. Berlin: Akademie Verlag: 61–103.

491

492

ARTEM I S ALEXIADOU AND FLOR IAN SCHÄFER

Sigurðsson, H. A´. 2006. “The nom/acc alternation in Germanic.” In J. Hartmann and L. Molna´rfi (eds.), Comparative Studies in Germanic Syntax. Amsterdam: John Benjamins: 13–50. Sigurðsson, H. A´. 2011. “On the new passive,” Syntax 14: 148–178. Steinbach, M. 2002. Middle Voice. Amsterdam: John Benjamins. Svenonius, P. 2002. “Icelandic case and the structure of events,” Journal of Comparative Germanic Linguistics 5: 197–225. Svenonius, P. 2006. Case alternations and the Icelandic passive and middle. Ms., University of Tromsø, available at ling.auf.net/lingbuzz/000124. Taraldsen, K. T. 2010. “Unintentionality out of control.” In M. Duguine, S. Huidobro, and N. Madariaga (eds.), Argument Structure and Syntactic Relations: A Cross-linguistic Perspective. Amsterdam: John Benjamins: 283–302. Thra´insson, H. 2007. The Syntax of Icelandic. Cambridge University Press. Thra´insson, H. 2017. “On quantity and quality in syntactic variation studies.” In H. Thra´insson, C. Heycock, H. P. Petersen, and Z. S. Hansen (eds.), Syntactic Variation in Insular Scandinavian. Amsterdam: John Benjamins: 20–52. Thra´insson, H., H. Petersen, J. Jacobsen, and Z. Hansen 2004. Faroese: A Handbook and Reference Grammar. To´rshavn: Faroe University Press. Vikner, S. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford University Press. Wood, J. 2015. Icelandic Morphosyntax and Argument Structure. Dordrecht: Springer. Wood, J. and H. A´. Sigurðsson 2014. “Get-passives and Case alternations: The view from Icelandic,” Proceedings of WCFFL 31: 494–503. Wurmbrand, S. 2006. “Licensing Case,” Journal of Germanic Linguistics 18: 175–234. Zaenen, A. and J. Maling 1990. “Unaccusative, passive and quirky case.” In J. Maling and A. Zaenen (eds.), Modern Icelandic Syntax. San Diego: Academic Press: 137–153. Zaenen, A., J. Maling, and H. Thra´insson 1985. “Case and grammatical functions: The Icelandic passive,” Natural Language and Linguistic Theory 3: 441–483.

Chapter 21 Binding The Morphology, Syntax, and Semantics of Reflexive and Nonreflexive Pronouns Vera Lee-Schoenfeld

21.1 Introduction This chapter discusses the occurrence of anaphoric elements, that is, reflexive and nonreflexive pronouns, in particular, their syntactic distribution, their morphological structure, and how they get their reference (their semantics). While the main goal is to give a descriptive overview, I will also mention and roughly explain two landmark theories proposed to account for the distribution of reflexive and nonreflexive pronouns: Chomsky’s (1981, 1986) classic Binding Theory and Reinhart and Reuland’s (1993) “Reflexivity” approach. Section 21.2 covers anaphoric elements in English and German, more specifically, the basic complementarity of reflexive and nonreflexive pronouns (21.2.1), how Chomsky (1981, 1986) accounts for the facts (21.2.2), and how certain instances of non-complementarity, including logophoric uses of reflexive pronouns, are best accounted for in Reinhart and Reuland’s (1993) system (21.2.3). Section 21.3 covers anaphoric elements going beyond English and German, focusing on so-called SELF versus SE anaphors and, where applicable, also on possessive reflexives in Dutch (21.3.1), Norwegian, Danish, and Swedish (21.3.2), and in Icelandic (21.3.3). Section 21.4 concludes the chapter.

21.2 Anaphoric Elements in English and German I start this overview with some basic observations about anaphoric elements in English and German because (i) I assume that English is the language that most readers have in common and (ii) the distribution of anaphoric elements in German is very similar to that in English. This latter point is noteworthy because, morphologically, the German reflexive pronoun sich looks much more like the Dutch reflexive zich than the English self-anaphors (himself, herself, themselves, etc.). Despite this superficial

494

VERA LEE-SCHOENFELD

Table 21.1 Anaphoric elements of English1 Number

Person

Singular

1. 2. 3.

MASC FEM NEUT

Plural

1. 2. 3.

Nonreflexive (NOM / ACC)

Reflexive (A C C /

I / me you he / him she / her it we / us you they / them

myself yourself himself herself itself ourselves yourselves themselves

DAT)

Possessive my your his her its our your their

Table 21.2 Anaphoric elements of German Number

Person

Singular

1. 2. 3.

MASC FEM NEUT

Plural

1. 2. 3.

Nonreflexive ( N O M / ACC / DAT)

Reflexive (A C C / D A T )

ich / mich / mir du / dich / dir er / ihn / ihm sie / sie / ihr es / es / ihm wir / uns / uns ihr / euch / euch sie / sie / ihnen

mich / mir dich / dir sich

Possessive mein dein sein ihr sein unser euer ihr

uns / uns euch / euch sich

similarity between German and Dutch, we will see that German patterns more like English than Dutch when it comes to the kinds of anaphoric elements it has and the way they are distributed. The tables above list the reflexive and nonreflexive pronouns of English and German, respectively.2 Note that, in both languages, possessives can be used reflexively or nonreflexively. There are no special reflexive possessive pronoun forms. Furthermore, the German third person singular and plural reflexive is invariant – it is not inflected for case, number, or gender.

21.2.1 Basic Complementarity As Safir (2004) explains in his introductory chapter of The Syntax of Anaphora, any theory of syntactically conditioned anaphora must capture the complementary distribution of reflexive and nonreflexive pronouns shown in (1) and (2).3

1

I use the term “anaphoric elements” to include both reflexive and nonreflexive pronouns although the latter do not require an antecedent in the sentence.

2

“N O M /

ACC

/

D AT ”

stands for nominative, accusative, and dative case, and “M A S C /

FEM

/

NEUT”

for the

grammatical genders masculine, feminine, and neuter. In Table 21.2, the German formal second person singular is not included, and the possessive pronouns are given without case inflection. 3

The examples in (1), (2), and (3) are adapted from Safir (2004: 9), his examples (3), (4), and (5).

Binding

(1)

a. The meni praised themselvesi. b. *The meni expected that themselvesi would win.

(2)

a. *The meni praised themi. b. The meni expected that theyi would win.

Notice that the English reflexive pronouns in (1), the so-called “SELF anaphors” (Reinhart and Reuland 1993), must have an antecedent (coindexed nominal) “nearby”, whereas the nonreflexive pronouns in (2) may not have an antecedent “nearby.” What exactly is meant by “nearby” will be specified in Section 21.2.2. For now, let us say that it stands for “in their clause.” Notice also that the SELF anaphor in (1b) has no way of being grammatical, but the pronoun in (2a) is grammatical as long as it does not refer to the subject of its clause the men, that is, as long as it is not coindexed with this nominal. This is shown in (3), where them refers to some plural entity not mentioned in the sentence. (3)

The meni praised themj.

In other words, while the SELF anaphors in the data thus far cannot be used grammatically without an antecedent in their clause, the pronouns cannot be used grammatically with an antecedent in their clause. And, unlike SELF anaphors, pronouns do not need an antecedent in the sentence at all. These generalizations can also be made about German reflexive and nonreflexive pronouns. Though their morphology differs, their syntactic distribution in examples corresponding to (1), (2), and (3) is the same as in English. While English reflexive pronouns consist of pronoun + self (myself, yourself, himself, herself, etc.), German reflexive pronouns are identical in form to nonreflexive pronouns (mich/mir ‘me’, dich/dir ‘you’, uns ‘us’, euch ‘you’) except in the third person singular and plural, where the reflexive is simply sich. Note that, while English SELF anaphors serve as both reflexive pronouns (as in (1)) and intensifiers (as in The chancellor herself did it), the German reflexive sich cannot be used as an intensifier (see, e.g., Bergeton 2004 and Gast 2006). The German equivalents to (1), (2), and (3) are given here as (4), (5), and (6).4 (4)

a. Die Ma¨nneri lobten sichi. (German) the men praised self ‘The men praised themselves.’ b. *Die Ma¨nneri erwarteten, dass sichi gewinnen wu¨rden. (German) the men expected that self win ‘*The men expected that themselves would win.’

(5)

4

a. *Die Ma¨nneri lobten siei. the men praised them ‘*The meni praised themi.’

would

(German)

Unless otherwise noted, the German examples provided in this chapter are my own, based on my judgments as a native speaker and on my previous work on binding (see references in footnote 9).

495

496

VERA LEE-SCHOENFELD

b. Die Ma¨nneri erwarteten, dass siei

gewinnen wu¨rden. (German) the men expected that they win would ‘The meni expected that theyi would win.’

(6)

Die Ma¨nneri lobten siej. the men praised them ‘The meni praised themj.’

(German)

Unlike the other Germanic languages to be discussed in this chapter, English and German have no more than these two types of anaphoric elements. In English, there are only SELF anaphors normally needing a local antecedent and pronouns not allowing a local antecedent. And correspondingly in German, there are only the elements sich, needing a local antecedent, and nonreflexive pronouns, not allowing a local antecedent. Neither language has possessive reflexive pronouns or so-called “long-distance” reflexives that are morphologically distinct from the “normal” reflexive and nonreflexive pronouns. Instances of noncomplementarity, i.e., contexts in which either a reflexive or nonreflexive pronoun can be used grammatically, which complicate the picture presented thus far for English and German quite a bit, will be discussed in Section 21.2.3.

21.2.2

A Brief Overview of Chomsky’s (1981, 1986) Binding Theory

The complementary distribution shown in the data thus far is captured by the classic Chomskyan Binding Theory, briefly summarized here. The occurrence of reflexive pronouns (“anaphors” in Chomsky’s terminology) is restricted by Condition (or Principle) A, and the occurrence of nonreflexive pronouns (“pronominals” in Chomsky’s terminology) is restricted by Condition (or Principle) B.5 (7)

Condition A: An anaphor must be bound in Domain D. Condition B: A pronominal must be free in Domain D.

Domain D stands for some structurally local domain, for example, the clause containing the anaphoric element, as tentatively suggested in Section 21.2.1. Instead of saying that a reflexive must have an antecedent in this local domain, and a nonreflexive may not, Conditions A and B use the terms “bound” and “free”, respectively. Being “bound” means being coindexed with and c-commanded by another nominal, and not being bound means being free. Let us walk through these definitions with the help of examples.6 (8)

a. Johni loves himselfi. b. *Johni’s mother loves himselfi.

5

The wording of the binding conditions here is adapted from Safir (2004: 9), his (7) and (8).

6

Examples (8) and (9) are adapted from Zwart (2002: 270), his (4) and (5).

Binding

(9)

a. *Johni loves himi. b. Johni’s mother loves himi.

The reflexive in (8a) is used grammatically because it is coindexed with John and, crucially, it is c-commanded by John. Informally stated, a node in a syntactic tree structure c-commands its sister and everything its sister dominates. In other words, from the subject position of the sentence, the nominal John in the (a)-examples c-commands the predicate loves himself / him, which contains the anaphoric element. This causes the reflexive in (8a) to be bound in Domain D, satisfying Condition A. It also causes the nonreflexive in (9a) to be bound in Domain D, violating Condition B, because, if it is bound, it is not free. On the other hand, looking at the (b)-examples, from the possessor position within the subject John’s mother, John is too deeply embedded to c-command the predicate. The subject as a whole does c-command the predicate, but the possessor phrase within the subject does not. This causes the reflexive in (8b) not to be bound in Domain D, violating Condition A. It also causes the nonreflexive in (9b) not to be bound, satisfying Condition B, because it is now free. Since there are binding domains that are smaller than the full-blown clause containing the anaphoric element (e.g., reduced infinitive clauses, complex nominal expressions, and certain prepositional phrases, see, e.g., Lee-Schoenfeld 2004 and 2007), and the subject position of certain infinitive clauses counts as part of the higher clause, the definition of Domain D is complicated and has undergone many revisions in the Government and Binding phase of mainstream generative syntax. We will return to this issue when needed in the following sections.

21.2.3

Noncomplementarity: Reflexive Pronouns Exempt from the Binding Theory? Notice that examples like those in (10), with English SELF anaphors that are used grammatically without an antecedent, seem to be a blatant violation of the need for reflexives to have a local antecedent (Condition A). (10)

a. They nominated Dr. Miller, Dr. Smith, Dr. Jones, and myself. b. There were five tourists in the room besides myself. (Reinhart and Reuland 1993: 669) c. Max boasted that the queen invited Lucie and himself for a drink. (Reinhart and Reuland 1993: 670)

In (10a) and (b), myself has no antecedent in the sentence at all, and in (c), himself has Max as its antecedent, but Max is in a higher clause than himself and therefore not a local binder. Reflexive pronouns in these contexts have been analyzed as exempt anaphors or logophors (see, e.g., Pollard and Sag 1992, Reinhart and Reuland 1993, and a very helpful overview in Bu¨ring 2005). Pollard and Sag (1992) suggest that first pronoun reflexives do not

497

498

VERA LEE-SCHOENFELD

need a linguistic antecedent because the speaker is necessarily a designated participant. Interestingly, the reflexive in all three of these examples can be replaced by the corresponding nonreflexive pronoun, me in (10a–10b) and him in (10c). The use of me instead of myself changes the utterances to sound less formal, so in these instances, it seems that the motivation for the use of the SELF anaphor could be the avoidance of the potentially impolite use of me. This is somewhat similar to people’s use of the subject form of the first person singular pronoun, I, instead of the object form, me, to sound more polite or educated, even when the pronoun occurs in an object position (e.g., in utterances like That is between you and I). In (10c), however, the SELF anaphor is the third person singular himself and is therefore less likely to be an exception motivated by politeness or hypercorrection. In addition, as we will see in Section 21.3, the Dutch equivalent of (10b) also has a reflexive pronoun as the complement to the preposition besides. This suggests that there is a pattern here that needs to be accounted for by something other than hypercorrection, and this brings us back to the analysis of these unbound or nonlocally bound reflexives as logophors. Unlike in Chomsky’s Binding Theory, logophoric use of SELF anaphors is not a violation of the binding conditions in Reinhart and Reuland’s (1993) system because their Condition A only applies to reflexives that are “direct” arguments of a predicate. Their abbreviated version of Conditions A and B and crucial definitions are given in (11). (11)

Condition A: A reflexive-marked predicate is reflexive. Condition B: A reflexive predicate is reflexive-marked. A predicate is reflexive if two of its arguments are coindexed. A predicate is reflexive-marked if either it is lexically reflexive or one of its arguments is a SELF anaphor. Reinhart and Reuland (1993: 670–671)

In (10a) and (10c), the anaphor is not an argument of the predicate (nominated and invited, respectively) – it is only one of the conjuncts of the coordinated argument (Dr. Miller, Dr. Smith, Dr. Jones, and myself in (12a) and Lucie and himself in (12b)) – and therefore does not reflexive-mark the predicate, which in turn means that Condition A does not apply to it or restrict its occurrence. The fact that not only the reflexive but also the corresponding nonreflexive pronouns are grammatical in (10) means that we are dealing with instances of noncomplementarity. The examples are repeated here with both the reflexive and the nonreflexive pronoun options in (12). (12)

a. They nominated Dr. Miller, Dr. Smith, Dr. Jones, and myself/me. b. There were five tourists in the room besides myself/me. c. Maxi boasted that the queen invited Lucie and himselfi/himi for a drink.

Binding

Because the bound reflexive in (12c) behaves like a nonreflexive pronoun in that it does not have a local antecedent, it can also be thought of as a “long-distance” reflexive, especially when it comes to similar data from Germanic languages other than English.7 We will return to this point in Section 21.3. Other instances of non-complementarity discussed in the literature involve so-called “picture NPs” (13a–13b) and prepositional phrases (PPs) that are thematically independent from the verb or noun they modify (13c). (13)

a. Luciei saw a picture of herselfi/heri. b. Maxi likes jokes about himselfi/himi. c. Maxi saw a gun near himselfi/himi. (Reinhart and Reuland 1993: 661)

The nominal in (13a), which is literally a “picture NP”, has the potential to describe a situation including participants like agent (external argument, e.g., the person having taken the picture) and patient (internal argument, e.g., the person being depicted), so that they can resemble a whole clause consisting of subject and predicate. Similarly, the noun jokes in (13b) potentially describes a situation involving someone who is telling jokes about someone else, as in people’s jokes about Max, which has the same core lexical meaning as the clause People tell jokes about Max. Finally, the PP near himself / him in (13c) is selected by neither the verb see nor the noun gun, which means that the object of the preposition is not thematically dependent on the verb or noun preceding it. The P near assigns its own thematic role to its complement regardless of the head it modifies. Though inherently subjectless (i.e., necessarily lacking an external argument), this kind of PP is similar to the nominals in (12a) and (12b), then, in that it has fully fleshed-out argument structure (see Hestvik 1991).8 Chomsky’s (1986) solution to the apparent violations of Condition B in (13a–13c) – the nonreflexive pronouns seem to be in the same local domain as their binder – was to propose that nominals with a potential possessor 7

See Büring 2005 for an overview of how the terms “exempt anaphor”, “logophor”, and “long-distant reflexive” are used in the literature. In the Germanic languages that distinguish between complex SELF and simplex SE anaphors, only the latter can be used as long-distance reflexives.

8

Notice that the argument structure of picture and jokes in (13a) and (13b) is not overtly fully fleshed out because there is no syntactically expressed possessor. In fact, in (13a), the indefinite D a can be argued (contra Chomsky 1986) to express that not even an implicit external argument is possible (it is just some picture, not one taken by someone). The picture NP in (13a) not having its full argument structure realized and therefore not constituting its own binding domain goes with Büring’s (2005) grammaticality judgments given in (i). The judgments here indicate that, contra Reinhart and Reuland 1993, a nonreflexive pronoun embedded in a nominal without a possessor cannot refer to the subject of the clause. (i) John5 saw [NP a picture of himself5/*him5]. (Büring 2005: 50) In the case of (13b), it can be argued more straightforwardly that jokes allows for a null possessor, i.e., an implicit external argument, because there is no overt D indicating indefiniteness (the jokes could be somebody’s, meaning they could be told by somebody).

499

500

VERA LEE-SCHOENFELD

(agent / external argument) as well as thematically independent PPs constitute their own binding domains (i.e., their own Domain D – see (7)), so that a nonreflexive pronoun can be free in this domain. In order to account for the use of a reflexive pronoun also being a possibility, the proposal was an extension of the reflexive binding domain if it did not include a potential binder, that is, if there was no coindexation possibility with a c-commanding nominal. This is justifiable via an appeal to the nature of reflexive pronouns, which is very different from that of nonreflexive pronouns. While reflexives come with the restriction that they must find a binder in the sentence, nonreflexive pronouns only come with the restriction of not being bound locally – otherwise they are free to occur as either bound or not. In a scenario like (13a), for example, the nominal, a picture of herself/her, constitutes a binding domain separate from the bigger, clausal domain. The nonreflexive pronoun is free within the smaller domain of the nominal, and, at the same time, the reflexive pronoun can be properly bound because its binding domain gets extended to the bigger clausal domain, which includes Lucie. This is because the domain of the nominal does not include a potential antecedent for the reflexive. The non-complementarity in (13b) can be explained in the same way. In a scenario like (13c), it is the thematically independent PP near himself/him that constitutes a binding domain separate from the clause as a whole, so that, again, the nonreflexive pronoun is free inside the smaller domain. For the reflexive, which does not have a potential antecedent within the PP, the binding domain gets extended to the clause, so that Max can properly bind the reflexive. As mentioned above, Reinhart and Reuland’s (1993) solution to instances of noncomplementarity like (13) is to treat the reflexives in these scenarios as exempt anaphors, i.e., as logophors, according to their terminology. Roughly speaking, since the reflexive pronouns in (13) are not arguments of reflexive predicates (neither the verbs of the sentences nor the nouns or prepositions have an external argument that the reflexive pronoun in the respective sentence is coindexed with), Reinhart and Reuland’s Condition A does not apply to these reflexive pronouns. Interestingly, German is unlike English and Dutch in that it does not have exempt anaphors like those in (10) and (12) but does resemble English and Dutch in having instances of noncomplementarity like (13). This seems to suggest that the English reflexives in (13) should not be treated as exempt from the binding conditions either. The following are the German equivalents of (12) and (13),9 showing that, unlike English SELF anaphors, German sich cannot occur without a local binder but that, like in English, both German reflexive and nonreflexive pronouns are grammatical in potentially complex nominals and thematically independent PPs. 9

Examples (12a–12b) do not have an equivalent in German because the first person singular reflexive has the same form as the corresponding nonreflexive pronoun (see Table 21.2).

Binding

(14)

Maxi prahlte, dass die Ko¨nigin Lucie und *sichi/ihni Max boasted that the Queen Lucie and self/him auf einen Drink eingeladen hatte. (German) on a drink invited had ‘Maxi boasted that the queen invited Lucie and himselfi/himi for a drink.’

(15)

a. Luciei Lucie b. Maxi Max c. Maxi Max

sah saw mag likes sah saw

ein Bild a picture Witze u¨ber jokes about eine Waffe a weapon

von sichi/ihr?i. of self/her sichi/ihni. self/him neben sichi/ihmi. next-to self/him

Given that German does not allow exempt/logophoric use of the reflexive sich, Chomsky’s reflexive domain extension account captures the parallel between the English and German facts in (13) and (15) better than Reinhart and Reuland’s system.10 Looking a little bit further, however (e.g., in Bu¨ring’s (2005) chapter 3), examples like (16) and (17)11 tell us that the reflexive domain extension approach is not sufficient to explain the facts. (16)

Johni believes that pictures of himselfi/himi are on sale.

(17)

Martelli hofft, dass eine Reportage u¨ber ihni/*sichi Martell hopes that a report about him/ self im Radio gespielt wird. (German) in-the radio played is.P A S S ‘Martell is hoping that a report about himself is going to be aired.’

While both reflexive and nonreflexive are grammatical in English, even when the picture NP is the subject of a finite clause, as in (16), the reflexive is completely impossible in this context in German, as shown in (17). If we want to account for the English non-complementarity in (16) with a reflexive domain extension, we would need to allow this extension to cross a finite clause boundary. If we did this, however, we would not be able to explain the ungrammaticality of the German reflexive in (17). Following Reinhart and 10

In Lee-Schoenfeld 2004, 2007, and 2008, I argue for a formal account of Chomsky’s domain extension based on Safir’s (2004) treatment of sich as a Romance reflexive clitic and Chomsky’s (2001) “Derivation by phase”, arguing that reflexives can gain access to the higher phase (i.e., reach an antecedent outside their coargument domain by covertly moving to the phase-edge). This is how reflexives embedded in theta-complete DPs and PPs, as well as Accusativus cum Infinitivo (AcI) complement clauses (see (i)) can be bound by the matrix clause subject without violating Chomsky’s classic Condition A, as stated in (7). (i)

Die Mutteri lässt [AcI die Kleinej the mother lets

sichi/j/ihri/*j die Schokolade in den Mund stecken].

(German)

the little-girl self / her the chocolate in the mouth stick

‘The motheri lets the little onej stick the chocolate in heri/j mouth.’ (Lee-Schoenfeld 2007: 118) 11

Examples (16) and (17) are adapted from Büring (2005: 52–53). They are his (3.20) and (i) in footnote 7, respectively.

501

502

VERA LEE-SCHOENFELD

Reuland in treating English reflexive pronouns that do not have a coindexed coargument as exempt from the binding conditions and admitting that the binding facts in English and German are no longer parallel at this point seems to be the best solution then. Reinhart and Reuland’s (1993) approach is also attractive when it comes to Dutch, Norwegian, and Danish to be discussed in Section 21.3 because, in addition to locally bound reflexive pronouns and nonreflexive pronouns, these languages have a third type of anaphoric element, namely reflexive pronouns that must be bound non-locally. Conditions A and B of the classic Binding Theory (see (7) above) are not designed to deal with an anaphoric element like this. Obviously, a reflexive that cannot be bound locally violates Condition A, and it is also not fully captured by Condition B because in addition to being free in its local domain, it is unlike a nonreflexive pronoun in that it must be bound within the sentence.

21.3 Anaphoric Elements Beyond English and German: SELF Versus SE and Possessive Anaphors When it comes to the inventory of anaphoric elements in the Germanic languages, we can group together Dutch, Norwegian, Danish, Swedish, and Icelandic because they all have in common that, besides the one reflexive and the one nonreflexive pronoun type found in English and German, they each have an additional anaphoric element. Dutch, Norwegian, and Danish have not just either a SELF anaphor like that found in English or a so-called “SE anaphor” with the shape of the reflexive sich in German, but both a complex SELF and a simplex SE anaphor. The following table nicely shows the “inbetween” status of SE anaphors, having in common with nonreflexive pronouns that they are not reflexivizers (i.e., do not reflexive-mark a predicate) and having in common with SELF anaphors that they are not referentially independent (i.e., cannot refer to an entity in the world without being bound). The extra type of anaphoric element in Icelandic and Swedish is a reflexive possessive pronoun, distinct in form from the nonreflexive possessive pronoun. Norwegian and Danish distinguish between reflexive and nonreflexive possessives as well, in addition to having both SELF and SE anaphors, which makes them the Germanic languages with the biggest inventory of anaphoric elements. This section looks at each of these Table 21.3 Typology of anaphoric expressions (Reinhart and Reuland 1993: 659)

Reflexivizing function Referential independence

SELF

SE

Pronoun

+ −

− −

− +

Binding

languages and their binding facts individually, often in comparison to English and/or German.

21.3.1 Dutch The following lays out the inventory of anaphoric elements in Dutch.12,13 The anaphoric elements in parts of Table 21.4 line up perfectly with those of German (see Table 21.2). Dutch and German both have a simplex third person anaphor – zich in Dutch and sich in German – and both have nonreflexive pronouns which are also used as reflexives in the first and second person. Furthermore, given Bu¨ring’s (2005) overview of Germanic pronoun systems, the Dutch and German parts of which are shown here in Table 21.5, even the Dutch complex SELF anaphors, zichzelf (third person) and pronoun + zelf (first and second person) seem to have a counterpart in German, namely the combination of sich or pronoun plus selbst. “P-form” stands for “pronominal”, i.e., nonreflexive. Taking a look at Dutch binding data, however, we see that things do not match up as neatly as Table 21.5 seems to suggest. The Dutch SELF anaphor, which must be bound in its coargument domain (18a) or can be used logophorically, i.e., as exempt from the binding conditions (19a), corresponds more closely to the English SELF anaphor than to any element in German. This is because, in German, the addition of the intensifier selbst to Table 21.4 Anaphoric elements of Dutch

Number

Person

Singular

1. 2. 3.

MASC FEM NEUT

Plural

1. 2. 3.

MASC FEM NEUT

12

Nonreflexive pronoun (full, reduced; NOM/ACC)

Reflexive Reflexive SE SELF anaphor anaphor

Possessive (full, reduced)

ik/mij, ‘k/me jij/jou, je/je hij/hem, ie/‘m zij/haar, ze/d‘r het/het, ‘t/‘t wij/ons jullie/jullie zij/hen, ze zij/die, ze zij/hen, ze

me je zich

mezelf jezelf zichzelf

ons je zich

onszelf jezelf zichzelf

mijn, m’n jouw, je zijn, z’n haar, d’r zijn, z’n ons jullie, je hun

This table was assembled with the help of Shetter and Van der Cruysse-Van Antwerpen (2004: 35, 66, 79) and Oosterhoff (2015: 30–31). Like in Table 21.2 (for German), the formal second person singular is left out.

13

Jan-Wouter Zwart (2011 and p.c.) explains that some dialects of Dutch still use the regular third person pronoun instead of the special reflexive zich, as was common in Dutch before the seventeenth century standardization of the language. Also, since possessive pronouns (such as the reduced form z’n ‘his’) are not marked for anaphoricity, the element eigen ‘own’ can be added to them for disambiguation. In fact, in certain dialects as well as spoken Dutch more generally, z’n eigen is used instead of the reflexive zich(zelf). Furthermore, the reduced (or weak) pronouns (e.g., hem ‘him’) can combine with zelf and occur in picture NP constructions like (16) in Section 21.2, where neither zichzelf nor zich can be used, but judgments on the distribution of weak pronoun + SELF are murky. Notice that weak hemzelf is included in Table 21.5 (taken from Büring 2005) but will not be discussed further in this chapter.

503

504

VERA LEE-SCHOENFELD

Table 21.5 Dutch and German pronoun systems (Büring 2005: 75)

bare

Dutch German Dutch German

+ ‘self’

SE-form

P-form

zich sich zich zelf sich selbst

hem ihn hem zelf ihn selbst

sich or a nonreflexive pronoun is merely emphatic, see the b-examples of (18) and (19),14 and, as established in Section 21.2, German reflexives cannot be used logophorically at all. (18) a. *Vijf touristen praatten met mezelf. (Dutch) b. ✓Fu¨nf Touristen sprachen mit mir selbst (die anderen five tourists spoke with myself (the others mit meinem Vertreter). (German) with my representative) (19)

a. Er waren vijf toeristen in de kamer behalve mezelf. (Dutch) b. Es waren fu¨nf Touristen im Zimmer außer mir (selbst). (German)

Also, unlike Dutch zich, which cannot be bound in its coargument domain and therefore functions as a long-distance reflexive, German sich must be bound locally. The Dutch examples in (20)15 showing zich bound in its coargument domain and thus used ungrammatically are contrasted with their German equivalents, which show sich used grammatically. (20) a. *Max haat

zich / ✓Max hasst sich.

(Dutch / German)

Max hates self

b. *Max praat

met

zich / ✓Max spricht mit sich. (Dutch / German)

Max speaks with self

The Dutch version of these examples can only be made grammatical by replacing zich with zichzelf. In contrast, the German version is grammatical with sich, and the addition of selbst is optional. This is shown in (21).

14

The Dutch data in (18) and (19) are adapted from Reinhart and Reuland (1993: 669), their (22c–22d). The German use of mir selbst in (18b) needs the context added in parentheses to sound acceptable. However, the point that mir selbst, unlike Dutch mezelf, is not a SELF anaphor and is therefore acceptable without a coindexed coargument can easily be made with the help of examples built around other two-place predicates. This is shown in (i) and (ii), where selbst is glossed as “I N T ” for ‘intensifier’: (i)

Das schmeckt mir selbst am besten. that tastes

(ii)

me I N T

(German)

the best

‘I’m the one who likes this the most.’ Mich selbst hat noch keiner erwischt. me

INT

has yet

(German)

nobody caught

‘As for me personally, nobody has caught me yet.’ 15

The Dutch data in (18) and (19) are adapted from Reinhart and Reuland (1993: 661, 665), their (17a–17b), and (10).

Binding

(21) a. ✓Max haat zichzelf / ✓Max hasst sich (selbst). Max hates himself b. ✓Max praat met zichzelf / ✓Max spricht mit sich (selbst). Max speaks with himself

(Dutch / German)

(Dutch / German)

Similarly, as shown in (22) and (23),16 while the use of Dutch zich versus zichzelf is strictly complementary with inherently (or intrinsically) reflexive verbs like zich schamen ‘be ashamed’, this is not the case in German. In Reinhart and Reuland’s (1993) terms, intrinsically reflexive verbs come from the lexicon as reflexive-marked and therefore do not need a SELF anaphor to satisfy their Condition B. In the German equivalent of (23), selbst is glossed as “I N T ” for ‘intensifier’. (22) (23)

✓Max schaamt Max shames *Max schaamt Max shames (am meisten). (the most)

zich / ✓Max scha¨mt sich. (Dutch / German) self zichzelf / ✓Max scha¨mt sich selbst himself / Max shames self INT (Dutch / German)

Thus, while Dutch strictly disallows the use of zichzelf with inherently reflexive verbs (which cannot also be used as normal transitives), German allows the use of sich selbst with such verbs given the right context. This further confirms that Dutch has two distinct reflexive pronouns – the SE anaphor zich and the SELF anaphor zichzelf – while German only has one reflexive pronoun, namely sich, which can be combined with selbst for emphasis (see also Bergeton 2004 and Gast 2006). Reinhart and Reuland (1993), citing Truckenbrodt 1992, claim that German does resemble Dutch in exhibiting complementary distribution of sich and sich selbst at least in the dative case. They provide the following data set arguing that German sich, like Dutch zich, cannot grammatically refer to a coargument in the context of a ditransitive verb that is not intrinsically reflexive (“– intr. refl.”). (24) a. *Peter1 vertraute sich1 seine *Peter1 vertrouwde zich1 zijn Peter1 entrusted to-himself1 his NOM ( – intr.refl.) D A T ACC b. Peter1 vertraute seine Tochternur

Tochter an. dochter toe. daughter P R T sichselbst1

an. (German)

Peter1 vertrouwde zijn

dochterslechts zichzelf1

Peter1 entrusted

his

daughter only to-himself1 P R T

ACC

DAT

NOM

( – intr. refl.)

(German) (Dutch)

toe. (Dutch)

(Reinhart and Reuland 1993: 667) 16

The Dutch example in (20) is again adapted from Reinhart and Reuland (1993: 666), their (19b). Example (21) is based on Reinhart and Reuland’s (1993: 666) observation that intrinsically reflexive verbs “allow or even require” zich, from which I conclude that zichzelf would be ungrammatical in their examples of intrinsically reflexive verbs.

505

506

VERA LEE-SCHOENFELD

Notice that, in (24b), the use of German sich selbst (which, unlike in Dutch, consists of two separate words – Reinhart and Reuland misrepresented the spelling here) goes with the noncanonical word order of A C C > D A T and the word nur ‘only’, which provides a context of focus. Given a context of contrastive focus, where both the indirect (D A T ) and the direct (A C C ) object are being contrasted, as in (25), the use of dative sich sounds much better than in (24a). (25)

Peter Peter vertraut entrusts

vertraut entrusts nur only

seiner his sichi self

Ex-Frau ex-wife seine his

seinen Sohn an, aber eri his son P R T but he Tochter an daughter P R T

This version of the use of the ditransitive predicate jemandem(D A T ) jemanden(A C C ) anvertrauen ‘entrust somebody to someone’ with a dative reflexive is still a bit awkward and needs emphatic stress on sich, which usually comes with the addition of selbst, but it is not ungrammatical. More importantly, as shown in (26), there are plenty of other noninherently reflexive ditranstive verbs that sound perfectly fine with a dative reflexive. The predicate in (26a) is jemandem(D A T ) etwas zu Essen oder Trinken(A C C ) machen ‘make somebody something to eat or drink’; in (26b) it is jemandem(D A T ) etwas(A C C ) geben ‘give somebody something’; and in (26c) it is jemandem (D A T ) etwas(A C C ) schicken ‘send somebody something’. (26) a. Peteri Peter b. Peteri Peter c. Peteri Peter

machte sichi einen Kaffee. made self a coffee gab sichi einen Kuss auf die gave self a kiss on the schickte sichi eine Postkarte. sent self a postcard

(German) Hand. (German) hand (German)

We can conclude, then, that the distinction in Dutch between the SELF anaphor zichzelf, which must be bound by a coargument if it is not used logophorically, and the SE anaphor zich, which must not be bound by a coargument unless it is an argument of an inherently reflexive verb, does not exist in German. Given the data we have seen thus far, German only has the reflexive sich, which must be locally bound and cannot be used logophorically, i.e., is never exempt from the binding conditions. The next part of the Dutch inventory of anaphoric elements to be tackled is the distinction between SE anaphors and nonreflexive pronouns. If SE anaphors cannot be bound in their coargument domain, then how are they different from nonreflexive pronouns? Crucially, SE anaphors lack number and gender features so that they can never refer to an entity in the world without being bound (i.e., having a c-commanding antecedent in the sentence). In contrast, as was established for English and German in Section 21.2, nonreflexive pronouns can have a referent not mentioned in the sentence.

Binding

This is shown in example (27). Furthermore, unlike nonreflexive pronouns, SE anaphors are subject-oriented in that they need to refer to the closest subject outside of their coargument domain (Bu¨ring 2005: 59, 70), unless this coargument domain has an inherently reflexive verb as its predicate. Example (28) illustrates that, like an SE anaphor, a nonreflexive pronoun is ungrammatical when bound in its coargument domain, and (29) shows that, unlike an SE anaphor, a nonreflexive pronoun is also ungrammatical when bound within the coargument domain of an inherently reflexive predicate. In the former kind of example, a SELF anaphor is the only grammatical option, while the latter kind of example only allows an SE anaphor. (27)

Jani Jani

(28)

Jan1 haat zichzelf1/*zich1/*hem1. (Dutch) Jan1 hates himself1/*SE1 /*him1 (Reinhart and Reuland 1993: 661, 665)

(29)

Willem1 schaamt zich1/*hem1. Willem1 shames SE1/*him1 (Reinhart and Reuland 1993: 691)

haat hates

*zichj/hemj. *SEj/ himj

(Dutch)

(Dutch)

As for the subject orientation of SE anaphors, it is a characteristic we should add to the list of distinctions between Dutch zich and German sich. Although the latter has the shape of an SE anaphor, it is not subjectoriented. As shown in (30), German sich, just like English SELF anaphors, can be bound by a coargument that is an object. (30)

Die Friseurin

zeigt

den Kundeni sichi im

the hair-stylist.F E M shows the client

self

Spiegel. (German)

in-the mirror

‘The hair sylist showed the clienti himselfi in the mirror.’

Getting back to Dutch zich versus nonreflexive pronouns, another context in which these two types of anaphoric elements are in noncomplementary distribution is given in (31), where the anaphoric element is embedded in a PP, complementing a theta-independent preposition, i.e., in a PP that is not dependent on the verb for theta-assignment and therefore constitutes its own predicate and coargument domain. (31)

Klaas1 duwde de kar voor Klaas1 pushed the cart before (Reinhart and Reuland 1993: 690)

zich1/hem1 SE1 /him1

uit. out

(Dutch)

In Reinhart and Reuland’s system, this noncomplementarity is accounted for because, as explained in Section 21.2, their Conditions A and B only restrict the occurrence of anaphoric elements that are SELF anaphors and therefore reflexive markers (see Condition A in (11a)) or have a coindexed coargument and are therefore in the domain of a reflexive predicate (see Condition B in (11b)). Since the anaphoric element in (31) is

507

508

VERA LEE-SCHOENFELD

not a SELF anaphor, Condition A cannot apply. And since the coindexed nominals (anaphoric element and antecedent) are arguments of different predicates, there is no reflexive predicate in this sentence and therefore no chance for Condition B to apply either. There are other types of Dutch binding facts (e.g., involving so-called Exceptional Case Marking [ECM] constructions, see Reinhart and Reuland 1993: 680, 691) that give an edge to Reinhart and Reuland’s (1993) “Reflexivity” approach over Chomsky’s (1981, 1986) classic Binding Theory, but for the purposes of this broad overview, the provided data shall suffice.

21.3.2 Norwegian, Danish, and Swedish According to Reinhart and Reuland’s (1993) references to Norwegian, the Norwegian SELF and SE anaphors seg selv and seg pattern just like Dutch zichzelf and zich. In my discussion here, I will group Norwegian and Danish together because, according to Bu¨ring (2005), who cites Vikner’s (1985) work, Danish also has distinct SELF and SE anaphors that pattern as in Dutch, but see Bergeton (2004) for a different view.17 What Swedish has in common with Norwegian and Danish that sets these three languages apart from Dutch, German, and English is that they have possessive reflexives. And what sets Norwegian and Danish even further apart is that their nonreflexive pronouns (e.g., ham ‘him’), when combined with selv ‘self’ (e.g., ham selv), have to be locally bound by a nonsubject and therefore exhibit anti-subject orientation. We will start with Norwegian and Danish and then take a brief look at Swedish. The following is an overview of the nonpossessive anaphoric elements in Norwegian and Danish taken from Bu¨ring 2005. The best way to explain and illustrate the content of this table is probably to follow Bu¨ring and immediately provide examples of the distribution of Table 21.6 Danish and Norwegian pronoun system (Büring 2005: 76)

Bare (free in coargument domain) + SELF (bound in subject domain)

17

SE-form (bound to subject in tense domain)

P-form (free from subject in coargument domain)

Danish: sig Norwegian: seg

Danish: ham (M A S C ), hende (F E M ) Norwegian: ham Danish: ham selv (M A S C ), hende selv (F E M ) Norwegian: ham selv

Danish: sig selv Norwegian: seg selv

Bergeton (2004) convincingly argues for an account of Danish sig and the intensifier selv in a system that treats binding and intensification as two independent modules of the grammar. My thanks to Line Mikkelsen (p.c.) for this reference.

Binding

the four types of anaphoric elements. The a-examples in (32)–(33)18 show Danish, and the b-examples Norwegian. Note that the only difference between the ways that the anaphoric elements in Danish and Norwegian pattern is that Danish has both a masculine and a feminine third person pronominal form (ham and hende) and Norwegian only has one form (ham). (32) a. Susan1 fortalte Anne2 om

*hende1/*hende selv1/*sig1/sig selv1.

Susan told Anne about *hende2/*hende selv2/*sig2/sig selv2. ‘Susan told Anne about her/herself.’ (Danish) b. Harald1 fortalde Jon2 om *ham1/*ham selv1/*seg1/seg selv1. Harald told Jon about *ham2/ham selv2/*seg2/*seg selv2. ‘Harald told Jon about him/himself.’ (Norwegian) In this first set of examples, the coargument, subject, and tense domains are the same. There is only one coargument domain in this sentence because the only predicate with argument structure is the verb (fortalte/fortalde) which has an external argument (Susan/Harald) and two internal arguments (the object Anne/Jon and the theta-dependent om-phrase). Since there is only one predicate with argument structure, there can also only be one subject, namely the external argument of the verb. Finally, because there is only one clause, there can also only be one tense domain. For our four anaphoric elements, this means that only the SE-form + SELF (sig selv / seg selv) can be used when referring to the subject, and only the P-form + SELF (hende selv / ham selv) can be used when referring to the object. The bare SE and P-forms (sig/seg and hende/ham) cannot be used because they must be free in their coargument domain. The bare SE-form can only be used with a binder in its coargument domain when the predicate is an inherently reflexive verb. This parallels the Dutch facts discussed in Section 21.3.1, see examples (22) and (23). In the next set of examples, the coargument and subject domain of the anaphoric elements is distinct from their tense domain. We have two clauses, a main clause, and an embedded infinitive clause. The main clause verb bad ‘asked’ is an object control verb, which means that the unpronounced subject of the infinitive clause is the same as the object of the main clause. (33) a. Susan1 bad Anne om at ringe til hende1/*hende selv1/sig1/*sig selv1. Susan asked Anne for to ring to ‘Susan asked Anne to call her.’ (Danish) b. Jon1 bad oss snakke om ham1/*ham selv1/seg1/*seg selv1. Jon asked us to-talk about ‘Jon asked us to talk about him.’ (Norwegian) In (33), the coargument domain of the listed anaphoric elements is the embedded infinitive clause, and this is also their subject domain because 18

Examples (32) and (33) are adapted from Büring (2005: 76), his (3.74) and (3.75). Büring in turn cites Dalrymple (1993) in his discussion of Norwegian and attributes his Danish examples to Vikner (1985).

509

510

VERA LEE-SCHOENFELD

this infinitive clause has an unpronounced (PRO)-subject. Now, since the binder (Susan/Jon) is the main clause subject and therefore not included in the smaller subject domain of the anaphoric elements, the SE-form + SELF (sig selv/seg selv) cannot be used – it needs to be bound by the subject of the smallest clause containing it. The P-form + SELF (hende selv/ham selv) cannot be used either because it also needs to be bound within its subject domain (albeit by an argument other than the subject). So, we are left with the bare forms, the SE-form (sig/seg) and the P-form (hende/ham). Why are they both marked as grammatical, i.e., why do they occur in noncomplementary distribution here? Because the SE-form is properly bound by the subject of the smallest tensed domain containing it, namely the main clause, and the P-form is free from the subject of its coargument domain, i.e., it is not bound by the unpronounced (object-controlled PRO) subject of the infinitive clause. As looking back at Table 21.6 will confirm, this is exactly how SE and P-forms are supposed to be distributed. Like the Dutch data featuring zich in Section 21.3.1, example (33) serves as an illustration of the longdistance binding characteristic of SE anaphors. While they must be bound by a subject, they can be bound across a clause boundary, as long as this clause boundary is a nonfinite one. Still following Bu¨ring’s sequence of examples, let us compare the binding facts in the object-control construction in (33) with those in the subject control construction in (34).19 The main clause verb in (34) is lovede ‘promised’, and the unpronounced subject of the infinitive clause is the same as the subject of the main the clause. (34)

Susan1

lovede

Anne2

at

ringe

Susan promised Anne to ring ‘Susan promised Anne to call her.’

til

hende2/*hende selv2/ *sig2/*sig selv2.

to (Danish)

Here, the binder of the listed anaphoric elements, Anne, which is the object of the main clause, is not also the unpronounced subject of the infinitive clause. This means that, like in (33), the anaphoric elements do not have their binder in their coargument/subject domain, but, unlike in (33), their binder is an object. The use of an SE-form, whether combined with SELF or not, is therefore ruled out. Both the SE-form + SELF (sig selv) and the bare SE-form (sig) are subject-oriented and thus need to be bound by a subject. The P-form + SELF (hende selv) is also ruled out because it needs its binder to be in its subject domain, which corresponds to the embedded infinitive clause here. The bare P-form (hende) is the one grammatical option in (34) because, as long as it is not bound by the subject of its coargument domain, is it free to occur. On to possessive reflexive pronouns, the type of anaphoric element that the Scandinavian languages (and, as we will see in Section 21.3.3, also 19

Example (34) is again adapted from Büring (2005: 77), his (3.76).

Binding

Icelandic) have, but English, German, and Dutch lack. According to Dalrymple (1993), the Norwegian reflexive possessive sin ‘self’s’, unlike the nonreflexive possessive hans ‘his’, must be bound by a subject. This is shown in (35). (35)

Jon beundrer sin /*hans Jon admires self’s/ his ‘Jon admires his mother.’ (Dalrymple 1993: 32)

mor. mother

(Norwegian)

As Bu¨ring (2005) points out, the possessive sin behaves like the bare SE anaphor seg in that it must be bound in its tense domain. Often, of course, the tense domain coincides with the subject domain, but when sin and seg are embedded in complex nominals or infinitive clauses (as shown for seg in (33b) above), they can be bound across a closer subject by the subject of the tensed verb. The nonreflexive possessive can be used with a binder in the same domain as the reflexive possessive, as long as this binder is not a subject. This is shown in (36) and indicates that, like the P-form + SELF, the Norwegian nonreflexive possessive exhibits antisubject orientation. (36)

Vi fant Joni under sengen we found Jon under bed ‘We found Jon under his bed.’ (Dalrymple 1993: 33)

hansi. his

(Norwegian)

The Swedish repertoire of anaphoric elements also includes the SE anaphor sig and the possessive reflexive sin, and Swedish is often discussed as patterning like Norwegian and Danish when it comes to the distribution of SE versus SELF anaphors (see, e.g., Gast 2006: 184, who cites Kiparski 2002). Holmes and Hinchliffe (1994), however, argue that the addition of the SELF morpheme sja¨lv / sja¨lvt (neuter) / sja¨lva (plural) to sig is used only for emphasis. This would make nonpossessive reflexive pronouns in Swedish pattern more like in German. Reflexive pronouns have the typical shape of a SE anaphor, with the special uninflected reflexive form (sig) only showing up in the third person and the other forms being identical to the nonreflexive ones, but they are claimed by Holmes and Hinchliffe not to be distinct from SELF anaphors. In (37), there is no need for the addition of the SELF morpheme sja¨lv to the reflexively used pronoun mig despite the fact that the anaphoric element is bound by its coargument. (37)

Jag skar mig. (Swedish) I cut me ‘I cut myself.’ (Holmes and Hinchliffe 1994: 146)

511

512

VERA LEE-SCHOENFELD

21.3.3 Icelandic Icelandic has in common with the Mainland Scandinavian languages that is has possessive reflexive pronouns in its repertoire of anaphoric elements, and it has in common with English and German that, for the majority of predicates, there is only one type of reflexive pronoun, not both an SE and a SELF anaphor. As is typical of SE anaphors, the Icelandic reflexive pronoun sig has a special reflexive form for the third person, but its first and second person forms are identical to the corresponding nonreflexive pronouns. Unlike Danish and Swedish sig, Icelandic sig is inflected for case, and the Icelandic reflexive possessive is additionally inflected for person, number, and gender. Table 21.7 shows the third person forms of both the nonpossessive and possessive reflexive pronouns of Icelandic. For most predicates, the SELF morpheme sja´lfur (which needs to be inflected to agree with its antecedent) is added optionally for emphasis. An example of this in the context of a typically reflexive verb (what Hyams and Sigurjo´nsdo´ttir [1990] call a shave-class verb) is given in (38).20 (38)

Hanni rakaði (sja´lfan) shaved E M P H hei ‘He shaved himself.’

sigi. himselfi

(Icelandic)

Interestingly, like in the languages that distinguish between SE and SELF anaphors, the use of the SELF morpheme sja´lfur is ungrammatical with an inherently reflexive verb. This makes Icelandic different from German, which allows the SELF morpheme selbst for emphasis, even with an inherently reflexive verb, see example (23) in Section 21.3.1 on Dutch. It seems that the use of sja´lfur in Icelandic falls somewhere in between that of selbst in German and that of the SELF morpheme in languages that distinguish between SE and SELF anaphors, like -zelf in Dutch, because, according to Thra´insson (2007), Table 21.7 Third person reflexive pronouns in Icelandic (Thráinsson 2007: 463) Possessive Reflexive

NOM ACC DAT GEN

20

Singular

Plural

SE Anaphor

MASC

FEM

NEUT

MASC

FEM

NEUT

– sig sér sín

sinn sinn sínum síns

sín sína sinni sinnar

sitt sitt sínu síns

sínir sína sínum sinna

sínar sínar sínum sinna

sín sín sínum sinna

Example (38) is adapted from Thrá insson (2014: 21), his (40).

Binding

there is, in fact, a certain class of predicates in Icelandic that requires the addition of sja´lfur to the reflexive pronoun. This class consists of verbs that would not normally have a coreferential argument, like ‘help’ or ‘talk to’. An example of ‘help’ used reflexively is given in (39), where the simplex reflexive pronoun is glossed as “R E F L ” and the SELF morpheme as “S E L F ” . (39)

Marı´ai getur ekki hja´lpað Maria can not help ‘Maria cannot help herself.’ (Thra´insson 2007: 264)

*se´ri / sja´lfri se´ri. / SELF REFL

(Icelandic)

REFL

Another characteristic that Icelandic sig shares with the Dutch SE anaphor zich, but not with German sich, is that it is subject-oriented, i.e., must be bound by a subject. Perhaps the most noteworthy property of the Icelandic simplex reflexive pronoun is that, unlike SE or SELF anaphors in any other language discussed thus far, it can be bound across a finite clause boundary. However, as shown in (40), the verb in this finite clause may not be in the indicative mood (glossed as “I N D ”), at least not if the main clause verb introducing the embedded finite clause is a factual one, like ‘know’. (40)

*Jo´n veit [að e´g hef logið Jon knows that I have.I N D lied Intended: Johni knows that I lied to himi. (Thra´insson 2007: 467)

að to

se´r].

(Icelandic)

REFL

It is with nonfactual attitude verbs, like ‘believe’, that the longdistance, cross-clausal use of the reflexive is grammatical and in noncomplementary distribution with the corresponding nonreflexive pronoun. This is shown in (41). (41)

Jo´n heldur [ad þu´ hatir Joni believes that you hate ‘John believes you hate him.’ (Thra´insson 2007: 467)

sig /hann]. R E F L i/himi

(Icelandic)

The optionality between the reflexive and the nonreflexive may be attributed to the perception and attitude of the referent of the main clause subject. Since the reflexive must refer to the main clause subject, while the nonreflexive could refer to an entity not mentioned in the sentence, the version of (41) with the reflexive can be paraphrased as John believes “you hate me.” If only subjunctive verbs intervene between the subject antecedent and the reflexive pronoun, there is a possibility for even longer distance binding. As shown in (42), the antecedent can be several sentences away from the reflexive. The subjunctive mood is glossed as “S B J ”.

513

514

VERA LEE-SCHOENFELD

(42)

Sumar væru Jo´ni sagði að hann he´lidi margar ræður. Jon said that he held.S B J many speeches some were.S B J um efnahafmalin, about economics aðrar fjo¨lluðu um tru´ma´l eða fjo¨lskylduma¨l. Samt kæmi others dealt with religion or family-values yet came.S B J e´g aldei til að hlusta a´ sigi. I never for to listen to R E F L ‘Jon said he held many speeches. Some were about economics, others dealt with religion or family values. Yet, I would never come to listen to them.’ (Thra´insson 2007: 472) (Icelandic)

21.4 Conclusion Wrapping up this overview of anaphoric elements and binding facts in (many of) the Germanic languages, we have seen that the only observation holding for all of them is that there are at least two types of anaphoric elements, a reflexive and a nonreflexive pronoun. Beyond this, there is a perhaps surprising amount of variation. Some have both a complex SELF and a simplex SE anaphor (e.g., Dutch, Norwegian, and Danish), some have possessive reflexives (e.g., Norwegian, Danish, Swedish, and Icelandic), and some include exempt anaphors, “logophors” (e.g., English, Dutch, and Icelandic). It turns out that German has the most basic inventory of anaphoric elements, with only one type of reflexive, which cannot be used logophorically.

References Bergeton, U. 2004. The Independence of Binding and Intensification. Ph.D. thesis, University of Southern California. Bu¨ring, D. 2005. Binding Theory. Cambridge University Press. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986. Knowledge of Language. New York: Praeger. Dalrymple, M. 1993. The Syntax of Anaphoric Binding. Stanford: Center for the Studies of Language and Information. Gast, V. 2006. The Grammar of Identity: Intensifiers and Reflexives in Germanic Languages. London: Routledge. Hestvik, A. 1991. “Subjectless binding domains,” Natural Language and Linguistic Theory 9.3: 455–496. Holmes, P. and I. Hinchliffe 1994. Swedish: A Comprehensive Grammar. London: Routledge.

Binding

Hyams, N. and S. Sigurjo´nsdo´ttir 1990. “The development of ‘long-distance anaphora’: A cross-linguistic comparison with special reference to Icelandic,” Language Acquisition 1.1: 57–93. Kiparski, P. 2002. “Disjoint reference and the typology of pronouns.” In I. Kaufmann and B. Stiebels (eds.), More than Words: A Festschrift for Dieter Wunderlich. Berlin: Akademie Verlag: 179–226. Lee-Schoenfeld, V. 2004. “Binding by phase: (Non-)complementarity in German,” Journal of Germanic Linguistics 16.2: 111–171. Lee-Schoenfeld, V. 2007. Beyond Coherence: The Syntax of Opacity in German. Amsterdam: John Benjamins. Lee-Schoenfeld, V. 2008. “Binding, phases, and locality,” Syntax 11.3: 281–298. Oosterhoff, J. 2015. Modern Dutch Grammar: A Practical Guide. London: Routledge. Reinhart, T. and E. Reuland 1993. “Reflexivity,” Linguistic Inquiry 24.4: 657–720. Safir, K. 2004. The Syntax of Anaphora. Oxford University Press. Shetter, W. Z. and I. Van der Cruysse-Van Antwerpen 2004. Dutch: An Essential Grammar. London: Routledge. Thra´insson, H. 2007. The Syntax of Icelandic. Cambridge University Press Thra´insson, H. 2014. On Complementation in Icelandic. London: Routledge. Reprint of 1979 dissertation originally published by Garland. Truckenbrodt, H. 1992. Clitics and the Chain Condition. Ms., Cambridge, MA: MIT. Vikner, S. 1985. “Parameters of binder and of binding category in Danish,” Working Papers in Scandinavian Syntax 23: 1–61. Zwart, J.-W. 2002. “Issues relating to a derivational theory of binding.” In S. D. Epstein and T. D. Seely (eds.), Derivation and Explanation in the Minimalist Program. Malden: Blackwell: 269–304. Zwart, J.-W. 2011. The Syntax of Dutch. Cambridge University Press.

515

Chapter 22 Verbal Particles, Results, and Directed Motion Ida Toivonen

22.1 Introduction Particles (such as out) and particle verbs (such as call out) have received much attention in the literature on Germanic linguistics, as evidenced by the following sample of book-length treatments: Fraser (1976), Svenonius (1994), den Dikken (1995), Zeller (1999), Lu¨deling (2001), Zeller (2001), McIntyre (2001), Toivonen (2001), Dehe´ (2002), Dehe´ et al. (2002), Mu¨ller (2002), Toivonen (2003), Blom (2005), Cappelle (2005), Los et al. (2012), Thim (2012), Walkova´ (2013), and Larsen (2014). Recognizing that it is not easy to posit a clear definition of Germanic particles, Dehe´ et al. (2002: 3) suggest the following characterization: A particle is an accented element which is formally (and, often, semantically) related to a preposition, which does not assign case to a complement and which displays various syntactic and semantic symptoms of what may informally be called a close relationship with a verb, but without displaying the phonological unity with it typical of affixes.

The particles have attracted interest for a variety of reasons, but perhaps especially because verb-particle combinations seem to have intermediate status between words and phrases. Since verbal particles have already received many in depth treatments (see the references above) and they have also been described in several excellent general overviews (see, e.g., the introduction to Dehe´ et al. 2002, chapter 2 of Los et al. 2012, and Dehe´ 2015), I will not attempt a comprehensive review here. Instead, after providing a brief summary of their main characteristics and also some of the main topics of debate regarding particles in the literature in Section 22.2, I will turn to interactions between verbal particles and other constructions. Section 22.3 focuses on resultative constructions such as hammer the metal flat. Such constructions occur in various forms across Germanic, and they share

Verbal Particles, Results, and Directed Motion

many characteristics in common with verb-particle constructions. I will in particular focus on what appears to be a cross-Germanic observation about resultative particles in transitive constructions: Post-object particles may receive a resultative interpretation predicated of the direct object, but preobject particles necessarily receive such an interpretation. Section 22.4 provides an overview of directed motion constructions in Germanic, often referred to as way-constructions, exemplified by elbow one’s way across the room. The section points out how verbal particles and directed motion constructions interact.

22.2 The Syntax of Verbal Particles 22.2.1 Morphology or Syntax? Particle verbs raise the important question of what distinguishes morphology from syntax. It is not clear whether particles combine with verbs morphologically or syntactically (Booij 1990, Zeller 2001, Lu¨deling 2001, Los et al. 2012, Dehe´ 2015). Is throw up a single compound word, or is it a syntactic phrase made up of two words? Verb-particle combinations are word-like in that they often carry idiomatic or semi-idiomatic meaning that is not completely compositional, as for example in the expressions hang out meaning socialize and put down meaning humiliate or criticize, and also kill (for an animal). Particles also participate in derivational morphology unlike full prepositional phrases. For example, particle verbs such as German vorwerfen ‘reproaches (lit. in front throw)’ can be nominalized: Vorwurf (Zeller 2001: 2): Furthermore, modification of particles is restricted. For example, pre-object particles in English, Norwegian, Icelandic, and Swedish cannot be modified: *throw right out the garbage. This supports the morphological view of particles, as does the fact that pre-object particles in many cases cannot be separated from the verb. Compare the (a) examples below to the (b) examples (all examples are from McCawley 1988): (1)

a. John picked up the money and picked out a coin. b. *John picked up the money and out a coin.

(2)

a. John picked up, and Mary hoisted up, some heavy weights. b. *John picked, and Mary hoisted, up some heavy weights

Examples (1) and (2) illustrate the tight connection between pre-object particles and the verb, which fits nicely with the view that the verbparticle connection is morphological. There are also data points that speak against a morphological analysis. Particles are not always adjacent to the verb: (3)

John picked the money up.

517

518

I DA TO IVO N EN

(4)

Sara kastade aldrig bort brevet. S. threw never away letter.the ‘Sara never threw away the letter.’

(Swedish)

The examples above show that English post-object particles are separated from the verb by the direct object, and Swedish pre-object particles can be separated from the verb by an adverb. Numerous other examples of nonadjacent verbs and particles are included below from several Germanic languages; see, for example, the German example in (8). This separability is the main difference between particle verbs and so-called prefix verbs (Stiebels and Wunderlich 1994 list a number of differences between German prefix verbs and particle verbs). In general, the separability of the verb and the particle depends on a variety of factors, such as the syntactic configuration in which the particle occurs, and the meaning contributed by the particle (Zeller 2001). There are also cross-linguistic differences. For example, Swedish pre-object particles are less closely connected to the verb than their English counterparts (Toivonen 2003). Finally, even though modification of particles is restricted, modification is not prohibited across the board. For example, many postobject particles in English, Norwegian, Icelandic, and Swedish can be modified, as in throw the garbage right out, kick the ball straight up. Data points such as the ones reviewed here (see also Los et al. 2012 and others) make it difficult to determine whether particle verbs are words or phrases. We return to this question in Section 22.2.5 below.

22.2.2 Word Order English, the Scandinavian languages, and Yiddish1 are SVO languages, and German, Dutch and Afrikaans are (S)OV in subordinate clauses. All Germanic languages except English are verb-second in declarative main clauses. The placement of particles interacts interestingly with the ordering of the verbs and objects in the different languages. In German, Dutch, and Afrikaans, the particle follows the main verb and the direct object in V2 structures but immediately precedes the verb elsewhere. This is illustrated with the German examples in (5) to (9), taken from Lu¨deling (2001): (5)

Das Ma¨rchen fa¨ngt the fairy.tale catches ‘The fairy tale begins’

(6)

daß das Ma¨rchen an-fa¨ngt that the fairy.tale on-catches ‘that the fairy tale begins’ *Das Ma¨rchen an-fa¨ngt.

(7)

1

an. on

See Dehé (2015) for references and discussion of Yiddish word order.

Verbal Particles, Results, and Directed Motion

(8)

Der Prinz ruft Dornro¨schen the prince calls Sleeping.Beauty ‘The prince calls up Sleeping Beauty.’

an. up

(9)

daß der Prinz Dornro¨schen an-ruft. that the prince Sleeping.Beauty up-calls ‘that the prince calls up Sleeping Beauty.’

Yiddish patterns with German, Dutch, and Afrikaans, except that particles precede the direct object in V2 structures: (10)

az er zogt nit ois dem sod. that he says not out the secret ‘that he does not reveal the secret.’ (Yiddish, den Besten and van Walraven 1986)

Particle verbs are normally inseparable in OV structures in these languages. However, in Dutch (Neeleman and Weerman 1993, Evers 2003), some varieties of German (Mu¨ller 2002, section 6.1.3.3), and marginally in Afrikaans (Le Roux 1988), the particle can appear in several possible positions if the clause contains more than one verbal element: (11)

dat Jan de informatie op wilde Zoeken / dat Jan de that J. the information up wanted Search / that J. the informatie wilde op-zoeken. information wanted up-search ‘that John wanted to look up the information.’ (Dutch, Blom 2005,7)

Standard German and Afrikaans normally require that the particle appears immediately before the verb in verb-final structures. In English, Norwegian, and Icelandic, the particle appears before or after the direct object, both in main and subordinate clauses: (12)

John sparka (ut) hunden J. kicked out dog.the ‘John kicked out the dog.’

(ut). out (Norwegian; Åfarli 1985)

The particle typically precedes the direct object in Swedish: (13)

Maja tryckte ner M. pressed down ‘Maja pushed the button down.’

knappen. button.the (Swedish)

In Danish, the particle follows the object: (14)

Han knugede sine hænder he clasped his hands ‘He clasped his hands.’

sammen. together. (Danish, Platzack 1998)

The above description covers the basic facts about the word order in clauses containing particles, see also the summary in (17), as well as Haider’s

519

520

I DA TO IVO N EN

contribution to this volume (Chapter 16). However, it is important to note that the description here shies away from several important details. For example, discourse considerations such as focus can play a role. Also, modification of particles may alter the regular ordering. In English, Swedish, Norwegian, and Icelandic, modified particles necessarily follow the direct object: (15)

Meg hurled the wallet right out.

(16)

*Meg hurled right out the wallet.

In (15), the particle out is modified by right and it must follow the object. Furthermore, it is in some languages relevant whether the direct object is a pronoun. In English, for example, particles must follow pronominal objects: *pick up it. These characteristics are discussed at length in the books and articles cited above. (17)

MAIN CLAUSE

SUBORDINATE CLAUSE German, Dutch, Afrikaans S V O Prt S O Prt V Yiddish S V Prt O S O Prt V English, Icelandic, S V { Prt } O { Prt } S V { Prt } Norwegian O { Prt } Swedish S V Prt O S V Prt O Danish SV O Prt S V O Prt

22.2.3 Word Class In addition to the phrase-structural status and semantics of verbal particles, the word class of particles has also been a topic of discussion in the literature. Traditional grammars often note that particles are not easy to characterize as a particular part of speech and therefore often resort to calling them “adverbial particles” or “adverb particles”, suggesting that they are a special type of adverb. Linguists tend to refer to particles as (intransitive) prepositions (Emonds 1972, van Riemsdijk 1978, den Dikken 1995), and the majority of particles do seem to be homophonous with and otherwise similar to prepositions. Compare the particles in the (a) examples to the transitive prepositions in the (b) examples of (18) and (19): (18)

a. John put his socks on. b. John’s socks are on his feet.

(19)

a. Jenna satte paº skorna. J. put on shoes.the ‘Jenna put the shoes on.’ paº fo¨tterna. b. Jenna satte skorna J. put shoes.the on feet.the ‘Jenna put the shoes on her feet.’

(Swedish)

Verbal Particles, Results, and Directed Motion

It is reasonable to assume that the words on and paº in (18) and (19) are intransitive and transitive prepositions, respectively: They are identical in form and close in meaning. This is indeed also true for many other particles: in, over, etc. However, even though prepositional particles dominate, it has been pointed out that certain members of other word classes also pattern with particles. For example, free and loose can sometimes appear before the object, just as particles do (e.g., tear loose the ligaments). Focusing on German, Stiebels and Wunderlich (1994), Zeller (2001), and Mu¨ller (2002) discuss adjectival particles such as krank-feiern ‘sick-celebrate’ (skip school) and nominal particles such as teil-nehmen ‘part-take’. Booij (1990) and van Marle (2002) provide similar examples from Dutch; for example, hard-lopen ‘fast-walk’ (race), buik-spreken ‘belly-speak’ (ventriloquize). Toivonen (2001, 2003) suggests that there are Swedish adjectives, nouns and verbs that seem to pattern with particles. Åfarli (1985) and Svenonius (1996) offer non prepositional particle examples from Norwegian, such as the adjectival example in (20): (20)

Vi gjorde klar bilen. we made ready car.the ‘We made the car ready.’

/ Vi / we

gjorde made

bilen klar. car.the ready (Åfarli 1985)

Most theoretical work thus assumes that particles are elements pulled from various word classes that have specific morpho-syntactic properties. Prepositional particles have received much more attention than particles from other word classes (Lu¨deling 2001, Larsen 2014). The traditional analysis that particles form their own word class is not widely adopted in the theoretical literature (but see, e.g., Nore´n 1996), mostly because there is such a clear overlap between particles and prepositions. However, the view that particles are simply special cases of prepositions, etc., does not explain why there are so many particles that only function as particles. That is, there are words that consistently pattern with particles and that do not pattern with nonparticle members of the relevant word class. For example, English away and apart do not seem to pattern with regular prepositions (or adverbs), and Swedish ner ‘down’ and kvar ‘left, remaining’ are not obviously prepositions (or adjectives or adverbs).

22.2.4 Argument Structure Particles can alter the argument structure of verbs, as in examples (21) and (22) taken from Larsen (2014, chapter 3): (21)

a. John shut *(his mouth). b. John shut (*his mouth) up.

(22)

a. John waited (*the storm). b. John waited out *(the storm).

521

522

I DA TO IVO N EN

The verb shut (21) normally requires an object, but the addition of the particle up prohibits a direct object. Conversely, wait (22) is intransitive, but wait out is transitive. The fact that the addition of particles can alter the argument structure of a verb is generally true across the Germanic languages and not unique to English. For example, the Swedish verb slaº ‘to hit’ requires an object, but the particle verb slaº runt ‘to party’ (literally ‘to hit around’) cannot take an object. In many cases, the argument alternation induced by particles can be compared (and perhaps equated) to argument additions that can be observed in resultative constructions; compare, for example, dream oneself away and dream oneself across the sea. Resultatives and particles are the topic of Section 22.3 below. It has also been argued that some aspectual particles (for example continuative on) restrict the argument structure. Examples and references are given in Section 22.3. For thorough discussion of particles and argument structure, see Stiebels (1996), McIntyre (2003, 2004), and Cappelle (2005, chapter 7).

22.2.5 Theoretical Analyses Numerous theoretical analyses have been proposed for the morpho-syntax of Germanic particles but the field has not yet reached a consensus; see Larsen (2014) for a thorough and recent review of different theoretical approaches to particles and particle verbs. This might be because a unified analysis is not correct, even though a unified analysis would seem intuitively desirable given the diachronic relatedness and synchronic similarity of particles across the Germanic languages. One major point of disagreement is whether verbs and particles form complex predicates or whether particles and the direct objects form small clauses (for a good overview on this debate, see Blom 2005). The syntactic arguments for each view tend to rely on assumptions about the relationship between syntax and semantics. Scholars who adopt frameworks that assume a direct and universal connection between specific syntactic configurations and semantic interpretation will of course reach different conclusions than scholars who assume that the syntax-semantics mapping is more flexible. The fact that particles do not correspond to a single semantic role complicates the matter. Germanic particles roughly divide into three semantic classes: They can be interpreted as spatial locations or directionality markers as in (23), or as aspectual markers as in (24), or else their use is idiomatic (25): (23)

a. Carla took out the food. b. Barbara put down her books.

(24)

a. Susie walked on. b. Mark wiped up the crumbs.

Verbal Particles, Results, and Directed Motion

(25)

a. Mona gave in. b. Sam threw up.

Especially the locative use (as in (23)) has led to the small clause analysis mentioned above, whereas the aspectual and idiomatic uses may seem more compatible with a complex predicate analysis. Representative examples of each type of analysis are den Dikken (1995) for a small clause analysis and Blom (2005) for a complex predicate analysis. Several researchers have argued that particles can be part of lexical compounds or fully projecting words, but also something in between, namely nonprojecting words (Toivonen 2001, Neeleman 2002, Toivonen 2003, Elenbaas 2007, Larsen 2014). That is, particles can occur in the following distinct morphosyntactic guises across and also within languages (note that the word order may differ as particles can also precede the verb): (26)

V0 verb-prt

V0 V0

Prt

VP V

PP

Allowing for the possibilities that particles can be bound morphemes, nonprojecting words or fully projecting words explains some of the diversity and seemingly incompatible characteristics described in the literature. For example, Toivonen (2003) argues that the fact that pre-object particles cannot be modified while post-object particles can is due to their distinct phrase-structural realization: Pre-object particles are nonprojecting words whereas post-object particles project full phrases.

22.3 Resultatives The Germanic languages display an array of different resultative constructions, see Goldberg and Jackendoff (2004), Beavers (2012), and Christie (2015) for overviews of resultatives in English. A typical English example is she paints the door black, and similar examples from German, Dutch, and Swedish are given in (27–29): (27)

(28)

(29)

Sie Streicht die Tu¨r schwarz. she paints the door black ‘She paints the door black.’ (German, Mu¨ller 2002) dat Jan de stok in stukken breekt that J. the stick in pieces breaks ‘that Jan breaks the stick into pieces’ (Dutch, Neeleman and Weerman 1993) Jag kokar moro¨tterna mjuka. I boil carrots.the soft ‘I boil the carrots soft.’ (Swedish)

523

524

I DA TO IVO N EN

There are also plain intransitive resultatives (30), as well as resultatives with fake objects (31) and fake object reflexives (32):2 (30)

The cup shattered into pieces.

(31)

Joe danced his shoes to shreds.

(32)

I sang myself into a finer state of mind.

Verb-particle constructions and resultative constructions share certain properties in common and are often compared in the literature (e.g., Neeleman and Weerman 1993, Svenonius 1996, Lu¨deling 2001, Zeller 2001, Mu¨ller 2002, Larsen 2014). One striking similarity is that particles can be secondary predicates that arguably denote a result: he kicked the ball away roughly means ‘he kicked the ball and as a result, the ball was away.’ It is somewhat controversial whether locational endpoints count as true results, and by extension, it is controversial whether examples such as (33) and (34) with locational endpoints count as resultatives: (33)

Sam kicked the ball into the net.

(34)

Tim fell down.

If locational examples of this type are included in the class of resultatives, then the parallel between resultatives and verb-particle constructions is very clear: Compare Sue drove the truck back/home to Sue drove the truck to her house. The parallel is also clear in property results, especially if we accept that there are adjectival particles (see Section 22.2.3 above): Yvonne broke open the coconut is a straightforward example of a resultative construction. Resultatives are generally assumed to be telic (but see Goldberg and Jackendoff 2004 for discussion). Setting aside aspectual particles and idiomatic particle verbs, it seems that verb-particle combinations are telic as well. Adverbs or prepositions that denote a path and are inherently incompatible with an endpoint interpretation do not function as particles, as evidenced by the fact that they cannot appear in the pre-object particle position:3 (35)

a. Marly rolled the ball in/inwards. b. Marly rolled in/*inwards the ball.

(36)

a. They pushed the agenda further. b. *They pushed further the agenda.

Words like inwards, downwards, outwards and further do not denote endpoints and they are dispreferred in the pre-object position characteristic for particles in English. (Although they can, of course, appear before the object if the object has undergone heavy-NP shift.)

2

These are attested examples found on the internet with simple Google searches.

3

Zwarts (2015, Chapter 26) discusses different nuances of locational meanings of prepositions, particles, and adverbs.

Verbal Particles, Results, and Directed Motion

The following examples indicate that the same generalization holds for Swedish: (37)

a. Malin Sparkade fram/upp bollen. M. kicked forth/up ball.the ‘Malin kicked out/up the ball.’ b. *Malin sparkade bollen fram/upp.

(38)

a. Malin sparkade bollen framaº t/uppaº t. M. kicked ball.the forwards ‘Malin kicked the ball forwards/upwards.’ b. *Malin sparkade framaº t/uppaº t bollen.

The words fram and upp can denote an endpoint and must occur in the typical Swedish particle position before the object. The words framaº t and uppaº t, on the other hand, denote a direction or a path without an endpoint and are not acceptable in the particle position. The absence of atelic directional particles is explained if we assume that particles necessarily denote results, and results are telic. Particles also pattern with resultatives in that they obey the direct object restriction (DOR) (Simpson 1983, Levin and Rappaport Hovav 1995). The DOR states that resultative predicates are predicated of the object. It has been suggested that it follows from the DOR that intransitive resultatives are restricted to unaccusatives, where the result is predicated of an objectlike subject (30). In order to add a secondary resultative predicate to an unergative intransitive, a fake reflexive must be added, as in sing oneself hoarse or laugh oneself to death, see also (32). Like resultative predicates, particles follow the DOR and must be predicated of the object. In Sarah threw out the garbage, the garbage (not Sarah) is out as a result, and in Timothy took down the bag, the bag (not Timothy) is down as a result. Remarkably, particles (or words that can normally function as particles) cannot appear in the pre-object particle position if they are not predicated of the object (Toivonen 1999, 2001):4 (39)

a. Sarah took the path back. b. #Sarah took back the path.

(40)

a. They took the elevator down. b. #They took down the elevator.

(41)

a. Caudia followed the stranger home. b. *Caudia followed home the stranger.

The generalization that pre-object particles must be predicated of the object appears to hold in Swedish, Norwegian, and Icelandic as well. If the ability to occur before the object in English, Swedish, Icelandic, and 4

The examples in (39b) and (40b) are grammatical, but have readings which are not intended here.

525

526

I DA TO IVO N EN

Norwegian is taken to be a crucial criterion for particlehood, then the words back, down, and home are not functioning as true particles in (39–41). They are instead fully projecting prepositions (or perhaps adverbs) with an adverbial use. It seems especially reasonable that the pre-object position should be considered a criterion for particlehood in Swedish, where postobject particles are generally barred: (42)

a. Sixten plockade upp plaº nboken. S. picked up wallet.the ‘Sixten picked up the wallet.’ b. *Sixten plockade plaº nboken upp.

In (42a), upp is a regular resultative particle, predicated of the object. The particle in (42a) occurs in the regular particle position, before the object; it cannot follow the object (42b). However, in example (43a), where upp is predicated of the subject (Ramona), the particle must follow the object: (43)

a. Ramona tog hissen upp. R. took elevator.the up. ‘Ramona took the elevator up.’ b. *Ramona tog upp hissen.

How is the fact that true, pre-object particles must be predicated of the object relevant to the question of whether verb-particle constructions are resultatives? If we adopt the DOR (resultatives are object-predicated), and we assume that the ability to appear before the object is a criterion for particlehood (that is, the words synonymous to particles in (39–41) and (43) are not in fact true particles), then the generalization that pre-object particles are objectpredicated can be taken as evidence that predicative verb-particle constructions are resultatives. Results must be predicated of objects and pre-object particles must be predicated of objects: this follows if (pre-object) particles denote results. However, does the DOR hold over all (and only) resultatives? The examples in (39–41) are similar to examples without particles such as those in (44) and (45), which are argued in Verspoor (1997) and Wechsler (1997) to be counterexamples to the DOR: (44)

John swam laps to exhaustion.

(45)

The wise men followed the star out of Bethlehem.

Examples like (44) and (45) indicate that resultative predicates can be predicated of subjects: John and the wise men end up exhausted and out of Bethlehem, laps and the star do not. If it is indeed generally possible for resultative predicates to be predicated of subjects, then (39–41) are orthogonal to the question of whether true particles (that is, particles that can precede the object) are resultatives. They may or may not be, and there might be some independent constraint that requires that pre-object

Verbal Particles, Results, and Directed Motion

particles be predicated of the object, regardless of whether they are resultatives or not. A clue that may help answer this question comes from aspectual particles such as up, on, along, and over: (46)

John tidied up the hotel room.

(47)

Tonya drove on.

(48)

Things moved along.

(49)

She looked the paper over.

Are these particles predicated of anything at all, or do they simply add aspectual information to the verb (telicity / completion in the case of up, atelicity/continuation in the case of on)? These elements are not clear predicates and therefore should perhaps not in principle be predicated of anything at all. However, in Toivonen (2006), I suggest that a peculiar argument structure constraint imposed by continuative on can be explained if it is assumed that a version of the DOR applies. The constraint that has been proposed is that continuative on is restricted to intransitive verbs (Fraser 1976, McIntyre 2001, Jackendoff 2002, McIntyre 2004). Consider (50) and (52): (50)

Norah ate on.

(51)

Norah ate burgers.

(52)

*Norah ate burgers on.

Note that the example with a direct object is not allowed. A sentence like (51) is atelic and seems to be perfectly compatible with a continuative action (as in Norah kept on eating burgers), but on is not compatible with Norah ate burgers (52). This is consistent with the claim that continuative on is restricted to intransitive verbs. However, in a limited set of cases the particle on can, in fact, co-occur with direct objects:5 (53)

They encouraged the contestants on.

(54)

. . . by means of his songs he urged on the Lacedaemonians in their war with the Messenians . . . Relief pilots have arrived in Harare to fly the plane on to its destination.

(55)

It seems that objects are allowed only when on can be predicated of the object. For example, in (53) the contestants are supposed to continue. In contrast, no continuation is implied for burgers in (51) and (52). It thus seems that continuative on is predicated of the lowest available argument function: the 5

Example (54) is lifted from the Wikipedia article on Tyrtaeus, and example (55) is also an attested example, retrieved in May 2017 from www.pressreader.com/south-africa/daily-news/20160216/281509340243420.

527

528

I DA TO IVO N EN

object if there is one, otherwise the subject. Examples such as *Tom played guitar on is ungrammatical because an interpretation where the guitar continues is not available. Parallel examples occur in Swedish with the Swedish equivalent paº (Toivonen 2006). It is worth noting that the comparable Swedish particle paº occurs in its characteristic pre-object position. It is unclear whether this generalization holds for other aspectual particles as well. Consider completive up, for example: There does not seem to be any obvious difference in predication between intransitive Sam tidied up and transitive Sam tidied up the room. I will not discuss other aspectual particles further here, but see Section 22.2.4 and references cited there for more examples of how particles influence argument structure. Assuming that it is correct that on must be predicated of the lowest available grammatical function, how is this relevant for the question of whether particles are resultative or not? Continuative on is not a resultative, yet it seems to be covered by the same predication criteria as resultatives: The DOR restricts the distribution of both resultatives and on. The fact that the DOR holds over locational particles that precede the object therefore does not constitute evidence that those particles are resultatives. The predication requirements should be considered separately from the question of whether locational particles are results. Before concluding this discussion, let us consider one further generalization from Dutch. Bennis (1991) points out that Dutch particles can follow the modal in cases where the locative particle denotes a location that is arrived at as a result of the action denoted by the verb (56): (56)

a. dat Jan niet boven wil komen that J. not upstairs wants come ‘that Jan doesn’t want to come upstairs’ b. dat Jan niet wil boven komen

The same particle boven cannot follow the modal if the verb that the particle modifies is stative and the particle denotes a stative location (57): (57)

a. dat Jan niet boven wil wonen that J. not upstairs wants live ‘that Jan doesn’t want to live upstairs’ b. *dat Jan niet wil boven wonen

These Dutch examples are similar to the examples we saw above (e.g., (35)) in that certain positions in the clause are reserved for telic, resultative particles. However, these examples differ from (39–41) in that they are intransitive, which makes it difficult to intuit what they are predicated over. I will leave it an open question whether the relevant point is telicity/ resultativity (boven is a telic result in (56) but not in (57)), or secondary predication (boven is a secondary predicate predicated of the subject in (56), and not a secondary predicate but an adverbial modifier in (57)).

Verbal Particles, Results, and Directed Motion

To summarize this section, verb-particle constructions are interestingly similar to resultatives in several ways, which has led many researchers to posit that verb-particles are resultatives. See, for example, Bolinger (1971), Svenonius (1994), but see also Neeleman and Weerman (1993), Cappelle (2005), who note several differences between the constructions. This section has pointed to some relevant data points concerning telicity, the DOR and aspect-marking particles. The connection between the resultatives and particle verbs is tight, interesting, and deserving of further investigation.

22.4 Directed Motion Constructions In addition to the resultatives discussed in the previous section, Germanic languages have constructions that resemble resultatives but express directed motion towards a goal instead of the achievement of a result.6 All Germanic languages seem to have a way of expressing directed motion with a semi-fixed combination of a verb, some kind of reflexive element, and a prepositional phrase, even though the verb itself is not a motion verb. In English, Icelandic, Dutch, and German, the directed motion construction can contain a word equivalent to way: (58) (59)

Lilah laughed her way across the room.

(60)

Der Regenwurm frisst sich seinen Weg durch das Erdreich. the earthworm eats REFL its way through the soil ‘The earthworm eats its way through the ground.’ (German, Ludwig 2005) Zij worstelde zich een weg door de menigte. she struggled REFL a way through the crowd ‘She struggled her way through the crowd.’ (Dutch, van Egmond 2009)

(61)

Það var eittvað að e´ta se´r leið inn i eyrað a´ me´r. there was something to eat REFL.DAT way.ACC in to ear on me ‘Something was eating its way into my ear.’ (Icelandic, Barðdal et al. 2011)

German examples such as (60) are unusual, as are the equivalent examples in Swedish and Norwegian, which also display occasional examples of expressions denoting directed motion with a word like va¨g, roughly equivalent to ‘way’ (or ‘road’), but these expressions do not appear to be productive or particularly common. Instead, Swedish, Norwegian, and German make more frequent use of a construction with a slightly different format, a construction that looks very similar to a fake reflexive resultative (62–64). Dutch also exhibits this type of construction (65): 6

I use the term construction in a pre-theoretical sense in this paper. See Asudeh et al. (2013) for a proposal of how to analyze constructions in a lexicalist framework.

529

530

I DA TO IVO N EN

(62)

Mamma Anne Grete trente seg gjennom svangerskapet. mummy A. G. exercised SELF through pregnancy ‘Mummy Anne Grete exercised her way through the pregnancy.’ (Norwegian, Seland 2001)

(63)

Storebror Hugo sparkar sig fram o¨ver big.brother H. kicks SELF forth across ‘Big brother Hugo kicks his way across the asphalt.’

(64)

Die Helfer wu¨hlen sich durch den Schutt. the helpers dig themselves through the rubble ‘The helpers dig their way through the rubble.’ (German, Perek and Hilpert 2014)

(65)

Hugo zong zich in de finale. H. sang REFL in the final ‘Hugo sang his way into the final.’ (Dutch, Marie-Elaine Van Egmond, p.c.)

asfalten. asphalt (Swedish)

These brief examples of course do not do justice to all the intricate details involved in the syntax and semantics of these constructions. For further examples and discussion, see van Egmond (2006, 2009), Verhagen (2003, 2007) for Dutch; Oya (1999), Ludwig (2005), Perek and Hilpert (2014) for German; Toivonen (2002), Lyngfelt (2007) for Swedish; Barðdal et al. (2011) for Icelandic; Seland (2001) for Norwegian; and Asudeh et al. (2013), Pedersen (2013) for typological overviews. The motion constructions, especially the ones without way in (62– 65) are quite similar to fake reflexive resultatives such as (32) above, but they differ in several ways: (i) Resultatives emphasize the end result and motion constructions emphasize motion along a path. (ii) Fake object resultatives are telic and can take adjectival results, whereas directed motion constructions are not necessarily telic, as illustrated with the atelic example John elbowed his way towards the exit. (iii) Directed motion constructions take PPs and not APs: *Sam drank his way silly. Resultatives typically take APs, even though there are also examples with PP results: drink oneself into a drunken stupor. (iv) The AP/PP in resultatives is typically a property (as in hammer the metal flat), and the PP in directed motion constructions typically expresses a path or location. However, there are examples where the result in a fake object resultative is a location (66), and there are also examples where the PP in a motion construction is a (figurative) path or goal that actually expresses a property (67): (66)

Kim danced himself across the room.

(67)

Kim sang his way into happiness.

Verbal Particles, Results, and Directed Motion

Proto-typical motion constructions and resultatives are easy to tell apart, but there are many examples that are very similar in meaning. Some attested examples are given in (68) and (69): (68)

a. A mouse chewed its way into our room. b. A rat has chewed itself into the home.

(69)

a. Two five-year-old boys use spades to dig their way out of their kindergarten and set off on a mission to buy a Jaguar sports car. b. Dantes and Priest hatched a plan to dig themselves out of prison.

It is difficult to pinpoint the exact difference in meaning between the (a) and (b) examples in (68) and (69). I will set aside here the interesting question of the meaning difference between the constructions, but see Christie (2011) and Mondorf (2011) for discussion. See also van Egmond (2009) for a discussion of the equivalent Dutch constructions. As described in Section 22.3, Germanic particles interact with resultatives in an interesting way: A particle in pre-object position necessarily denotes an object-predicated result. It appears that the opposite holds for motion constructions, at least for English and Swedish: Even if the path/ goal is expressed with a particle (or particle-like word) like out, up, in, or back, that word cannot appear in the typical pre-object position. The following examples illustrate that particles can follow but not precede one’s way in English: (70)

a. Sara danced her way out. b. *Sara danced out her way.

(71)

a. Clara worked her way back. b. *Clara worked back her way.

The Swedish directed motion construction does not include a word equivalent to the English way but instead just a plain reflexive, see example (63). In Swedish, fake reflexive resultatives are therefore very difficult to distinguish from directed motion constructions. However, when the examples include particles, the two constructions differ in word order: Resultatives take preobject particles and directed motion constructions take post-object particles. This is illustrated with the resultative examples in (72) and the directed motion examples in (73). The particles are boldfaced: (72)

a. Filip stressade so¨nder sig. F. stressed broken REFL (Literally) ‘Filip stressed himself broken.’ b. De maº lade in sig i ett ho¨rn. they painted in REFL in a corner ‘They painted themselves into a corner.’

531

532

I DA TO IVO N EN

(73)

a. Margareta jobbade sig upp. M. worked REFL up ‘Margareta worked her way up.’ b. De in i ett ho¨rn. maº lade sig they painted REFL in in a corner ‘They painted their way into a corner.’

The examples in (70–73) suggest that regular, pre-object particles are not allowed in Germanic motion constructions in Swedish and English. As far as I am aware, the question of whether particles are restricted in directed motion constructions in other Germanic languages has yet to be investigated.

22.5 Concluding Remarks The Germanic verbal particles provide many interesting puzzles for linguists. Even the basic issue of categorizing particles as a coherent class is a struggle. Nouns are recognized by their ability to take number and definiteness marking, subjects agree with verbs and appear in a specific phrase structural position, etc. Verbal particles are harder to pin down, especially if we attempt to determine defining characteristics that hold across the Germanic languages. The particles do not take inflectional morphology. It is difficult to pinpoint their phrase structural position, and they do not appear in a set position even within a given language. Moreover, as pointed out in Sections 22.1 and 22.2, particles have long puzzled linguists with their mixed morphological and syntactic properties. Despite all this, “verbal particles” does mean something to linguists working on Germanic. Using the term rarely leads to confusion, and for a scholar versed in the structure of one of the Germanic languages, it is not difficult to identify examples of particles in another. Why should it then be so difficult to state their defining characteristics? And what exactly is it that enables researchers to distinguish particles from other words/morphemes? Perhaps there is no uniform particle category that shares certain characteristics in common across the Germanic languages. Perhaps the particles are connected across the languages merely by their diachronic affiliation and not by any synchronically real formal similarity. Particles do seem cross-linguistically similar in that their ability to take modifiers or complements is heavily restricted. However, this is also true for sentential adverbs. Particles do not take inflection, but neither do prepositions, conjunctions, and complementizers in many languages. One interesting feature that particles do seem to have in common across the Germanic languages is their ability to alter the argument structure of the verb. Compare, for example, the grammatical look up a word to the ungrammatical *look

Verbal Particles, Results, and Directed Motion

a word. However, particles are not unique in this regard, as resultatives can also introduce “fake objects” (Section 22.3). The very fact that it is difficult to state defining properties that cover all particles and exclude other words and phrases means that they are a worthwhile topic of study. Identifying defining characteristics and distinguishing features in order to delimit classes and subclasses of linguistic categories is an important part of what linguists do. Particles are therefore of interest precisely because they are difficult to characterize. And if we decide that it is in fact not possible to define and delimit a coherent class of verbal particles, then other important questions arise: How are the particles different? How did the classes of particles drift apart diachronically? In these pages, I have described the main properties of Germanic particles. I have also pointed to some of the theoretical controversies in the literature, as well as some of the reasons why a unified analysis has proved elusive. I have also connected particles to two Germanic constructions with which they interact in interesting ways: resultatives and directed motion (‘way’) constructions. I would like to end by suggesting that explorations of the use of particles in various constructions might lead to interesting novel discoveries about verbal particles, various interesting constructions (such as those mentioned here), and Germanic syntax more generally.

References Åfarli, T. 1985. “Norwegian verb particle constructions as causative constructions,” Nordic Journal of Linguistics 8: 75–98. Asudeh, A., M. Dalrymple, and I. Toivonen 2013. “Constructions with Lexical Integrity,” Journal of Language Modelling 1.1: 1–54. Barðdal, J., K. E. Kristoffersen, and A. Sveen 2011. “West Scandinavian ditransitives as a family of constructions: With a special attention to the Norwegian ‘V-REFL-NP’ Construction,” Linguistics 49.1: 53–104. Beavers, J. 2012. “Resultative Constructions,” In R. I. Binnick (ed.), The Oxford Handbook of Tense and Aspect. Oxford University Press: 908–933. Bennis, H. 1991. “Theoretische aspekten van partikelvooropplaatsing II.” In TABU 21: 89–95. Besten, H. den and C. M. van Walraven 1986. “The syntax of verbs in Yiddish.” In H. Haider and M. Prinzhorn (eds.), Verb Second Phenomena in Germanic Languages. Dordrecht: Foris: 111–135. Blom, C. 2005. Complex Predicates in Dutch: Synchrony and Diachrony. Ph.D. thesis, Vrije Universiteit Amsterdam. Bolinger, D. 1971. The Phrasal Verb in English. Berlin: Mouton de Gruyter. Booij, G. 1990. “The boundary between morphology and syntax: Separable complex verbs in Dutch.” In G. E. Booij and J. van Marle (eds), Yearbook of Morphology, Vol. 3. Dordrecht: Foris: 45–63.

533

534

I DA TO IVO N EN

Cappelle, B. 2005. Particle Patterns in English: A Comprehensive Coverage. Ph.D. thesis, Katholieke Universiteit Leuven. Christie, E. 2015. The English Resultative. Ph.D. thesis, Carleton University, Ottawa. Christie, L. 2011. “Investigating the differences between the English wayconstruction and the fake reflexive resultative construction.” In L. Armstrong (ed.), CLA Conference Proceedings, Canadian Linguistic Association: 1–14. Dehe´, N. 2002. Particle Verbs in English: Syntax, Information Structure and Intonation. Amsterdam and Philadelphia: John Benjamins. Dehe´, N. 2015. “Particle verbs in Germanic.” In P. O. Mu¨ller, I. Ohnheiser, S. Olsen, and F. Rainer, eds., Word Formation: An International Handbook of the Languages of Europe. Berlin: Mouton de Gruyter. Dehe´, N., S. Urban, A. McIntyre, and R. Jackendoff (eds.) 2002. Explorations in Verb-Particle Constructions. Berlin: Mouton de Gruyter. Dikken, M. den 1995. Particles: On the Syntax of Verb-Particle, Triadic, and Causative Construction. Oxford University Press. Egmond, M-E. 2006. Two Way-Constructions in Dutch. Master’s thesis, University of Canterbury. Egmond, M-E. van 2009. Two Way-Constructions in Dutch. Saarbru¨cken: VDM Verlag. Elenbaas, M. 2007. The Synchronic and Diachronic Syntax of the English VerbParticle Combination. Ph.D. thesis, Radboud Universiteit, Nijmegen. Emonds, J. 1972. “Evidence that indirect object movement is a structurepreserving rule,” Foundations of Language 8: 546–561. Evers, A. 2003. “Verbal clusters and cluster creepers.” In P. Seuren and G. Kempen (eds.), Verb Constructions in German and Dutch, Vol. 242. Current Issues in Linguistic Theories. Amsterdam: John Benjamins: 43–89. Fraser, B. 1976. The Verb-Particle Combination in English. New York: Academic Press. Goldberg, A. and R. Jackendoff 2004. “The English resultative as a family of constructions,” Language 80.3: 532–568. Jackendoff, R. 2002. “English particle constructions, the lexicon and the autonomy of syntax.” In Dehe´ et al. 2002, pp. 67–94. Larsen, D. 2014. Particles and Particle-Verb Combinations in English and Other Germanic Languages. Ph.D. thesis, University of Delaware, Newark, DE. Le Roux, C. 1988. On the Interface of Morphology and Syntax: Evidence from VerbParticle Combinations in Afrikaans. Master’s thesis, University of Stellenbosch. Levin, B. and M. Rappaport Hovav 1995. Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge, MA: MIT Press. Los, B., C. Blom, G. Booij, M. Elenbaas, and A. van Kemenade 2012. Morphosyntactic Change: A Comparative Study of Particles and Prefixes. Cambridge University Press.

Verbal Particles, Results, and Directed Motion

Lu¨deling, A. 2001. On Particle Verbs and Similar Constructions in German. Stanford, CA: CSLI Publications. Ludwig, I. 2005. The Way-Construction in German. Ms., University of Canterbury. Lyngfelt, B. 2007. “Mellan polerna. Reflexiv- och deponenskonstruktioner i svenskan,” Spraº k och stil 17: 86–134. Marle, J. van 2002. “Dutch separable compound verbs: Words rather than phrases?” In Dehe´ et al. 2002, pp. 211–232. McCawley, J. 1988. The Syntactic Phenomena of English. Chicago, IL: University of Chicago Press. McIntyre, A. 2001. German Double Particles as Preverbs: Morphology and Conceptual Semantics. Tu¨bingen: Stauffenburg. McIntyre, A. 2003. “Preverbs, argument linking and verb semantics: Germanic prefixes and particles.” In G. Booij and J. van Marle (eds.), Yearbook of Morphology. Dordrecht: Springer: 119–144. McIntyre, A. 2004. “Event paths, conflation, argument structure and VP shells,” Linguistics 42: 523–571. Mondorf, B. 2011. “Variation and change in English resultative constructions,” Language Variation and Change 22: 397–421. Mu¨ller, S. 2002. Complex Predicates: Verbal Complexes, Resultative Constructions, and Particle Verbs in German. Stanford, CA: CSLI Publications. Neeleman, A. 2002. “Particle placement.” In Dehe´ et al. 2002, 141–164. Neeleman, A., and F. Weerman 1993. “The balance between syntax and morphology: Dutch particles and resultatives,” Natural Language and Linguistic Theory 11: 433–475. Nore´n, K. 1996. Svenska partikelverbs semantik. Gothenburg: Acta Universitatis Gothoburgensis. Oya, T. 1999. “Er bettelt sich durchs Land: Eine ‘one’s way’ Konstruktion im Deutschen?” Deutsche Sprache 27: 356–369. Johan, P. 2013. “The Way-Construction and Cross-Linguistic Variation in Syntax: Implications for Typological Theory.” In C. Paradis, J. Hudson, and U. Magnusson (eds.), The Construal of Spatial Meaning: Windows into Conceptual Space. Oxford University Press: 236–262. Perek, F. and M. Hilpert 2014. “Constructional tolerance: Are argument structure constructions equally powerful across languages?” Constructions and Frames 6.2: 266–304. Platzack, C. 1998. Svenskans inre grammatik: Det minimalistiska programmet. Lund, Sweden: Studentlitteratur. Riemsdijk, H. van 1978. A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases in Dutch. Dordrecht: Foris. Seland, G. 2001. The Norwegian Reflexive Caused Motion Construction: A Construction Grammar Approach. Cand. Philol. thesis, Department of Linguistics, University of Oslo. Simpson, J. 1983. “Resultatives.” In L. S. Levin, M. Rappaport, and A. Zaenen (eds.), Papers in Lexical-Functional Grammar. Bloomington: Indiana University Linguistics Club: 143–157.

535

536

I DA TO IVO N EN

Stiebels, B. (ed.) 1996. Lexikalische Argumente und Adjunkte. Zum semantischen Beitrag von verbalen Pra¨fixen und Partikeln. Berlin: Akademie Verlag. Stiebels, B. and D. Wunderlich 1994. “Morphology feeds syntax,” Linguistics 36: 1–39. Svenonius, P. 1994. Dependent Nexus. Subordinate Predication Structures in English and the Scandinavian Languages. Ph.D. thesis, University of California at Santa Cruz, Santa Cruz, CA. Svenonius, P. 1996. “The optionality of particle shift.” Working Papers in Scandinavian Syntax 57: 47–75. Thim, S. 2012. Phrasal Verbs: The English Verb-Particle Construction and its History. Berlin and New York: Mouton de Gruyter. Toivonen, I. 1999. “Swedish place expressions.” In P. Tamanji, M. Hirotani, and N. Hall (eds.), Proceedings of the Twenty-Ninth Annual Meeting of the North East Linguistic Society. Somerville, MA: Cascadilla Press: 367–379. Toivonen, I. 2001. The Phrase Structure of Non-Projecting Words. Ph.D. thesis, Stanford University, Stanford, CA. Toivonen, I. 2002. “The directed motion construction in Swedish,” Journal of Linguistics 38.2: 313–345. Toivonen, I. 2003. The Phrase Structure of Non-projecting Words: A Case Study of Swedish Particles. Studies in Natural Language and Linguistic Theory, Vol. 58. Dordrecht: Kluwer. Toivonen, I. 2006. “On continuative on,” Studia Linguistica 60.2: 181–219. Verhagen, A. 2003. “The Dutch way.” In A. Verhagen and J. van de Weijer (eds.), Usage-based Approaches to Dutch. Utrecht: LOT: 27–57. Verhagen, A. 2007. “English constructions from a Dutch perspective: Where are the differences?” In M. Hannay and G. J. Steen (eds.), StructuralFunctional Studies in English Grammar. In honour of Lachlan Mackenzie. Amsterdam and Philadelphia: John Benjamins: 257–274. Verspoor, C. M. 1997. Contextually-Dependent Lexical Semantics. Ph.D. thesis, University of Edinburgh. Walkova´, M. 2013. The Aspectual Function of Particles in Phrasal Verbs. Ph.D. thesis, University of Groningen. Wechsler, S. 1997. “Resultative predicates and control,” Texas Linguistics Forum 38: 307–321. Zeller, J. 1999. Particle Verbs, Local Domains, and a Theory of Lexical Licensing. Ph.D. thesis, Johann-Wolfgang-Goethe-Universita¨t zu Frankfurt am Main. Zeller, J. 2001. Particle Verbs and Local Domains. Amsterdam: John Benjamins. Zwarts, J. 2015. “Out of phase: Form-meaning mismatches in the prepositional phrase.” In I. Toivonen, P. Csu´ri, and E. van der Zee (eds.), Structures in the Mind: Essays on Language, Music and Cognition in Honor of Ray Jackendoff. Cambridge, MA: MIT Press: 63–78.

Chapter 23 Structure of Noun (NP) and Determiner Phrases (DP) Dorian Roehrs

23.1 Introduction Following Abney’s (1987) seminal dissertation, many more phenomena in the noun phrase have been discovered and discussed. In this brief survey, I point out some recent results, some debated topics, and some new avenues for future research. I focus on structural aspects of the noun phrase discussing different nominal elements and how they relate to one another. I use contemporary German as a representative language. This is a highly inflected language, a fact that is often taken to provide clues about the analysis of certain linguistic phenomena. Other Germanic languages are discussed if they show an important relevant difference from German. I also point out some differences between North and West Germanic (East Germanic, e.g., Gothic, no longer exits). I start with some general facts about the noun phrase that the Germanic languages instantiate and then move on to some Germanic-specific phenomena. Note that a number of discoveries of the structure of the noun phrase have been made on the basis of non-Germanic languages. These are only very briefly discussed here. Finally, a note on nomenclature is in order. If the syntactic structure is not relevant, I label the nominal string as noun phrase; if the structure is important, I specify it as, for instance, NP or DP.

23.1.1 A First Overview Noun phrases vary in their complexity. Minimally, they consist of a head noun (1a). One might suggest that they simply project a NP (1b): (1)

a. Bu¨cher sind wichtig. books are important ‘Books are important.’

538

DORIAN ROEHRS

NP

b.

Bücher

In addition to the head noun, noun phrases may also involve quantifiers, determiners, numerals, adjectives, possessives, and other argumental elements. Typically, these elements occur in the fairly fixed order in (2a). The example (2b) is the equivalent of (2a) in the dative. One very common way to analyze these strings is the structure in (2c). The noun forms the head of the Noun Phrase (NP), the arguments of the noun are in the complement and specifier positions of the noun, number is mitigated by the Number Phrase (NumP), adjectives are in recurring Agreement Phrases (AgrP), numerals are in the Cardinal Phrase (CardP), articles and demonstratives are in the Determiner Phrase (DP), and elements preceding determiners are in – what I call here – XP. The head noun undergoes partial N-raising to Num: (2)

a. All die 25 interessanten Bu¨cher meines Bruders all the 25 interesting books my-G E N brother u¨ber Geschichte sind wichtig. about history are important ‘All the 25 interesting books of my brother’s about history are important.’ b. mit allen diesen 25 interessanten Bu¨chern meines with all these 25 interesting books my-G E N Bruders u¨ber Geschichte brother about history c. XP

X’ all

DP D’ CardP

die 25

Card’ Card

AgrP

interessanten Agr’ Agr

NumP Num’ Bücherk

NP

meines Bruders N’ tk über Geschichte

Structure of Noun (NP) and Determiner Phrases (DP)

According to Grimshaw (1991), the noun, the lexical head, projects structure from the bottom up yielding an extended projection of the noun topped off by functional structure (for some parallels between the nominal and clausal domains, see Grohmann and Haegeman 2003). All relevant elements are nominal in category and this categorial identity indicates one nominal domain. This single nominal domain becomes evident when the phrase (2a) is compared to (2b). With the exception of the two arguments of the noun, all elements share the same gender, number, and case. In other words, quantifiers, determiners, numerals, adjectives, and head nouns show agreement in features, something usually referred to as concord. With the arguments of the noun not participating in concord, they are considered not to be part of the nominal proper. In fact, given their own features for gender, number, and case, they make up their own nominal domain and are analyzed as embedded. It is probably fair to state that most linguists working on the Germanic noun phrase agree that there is a DP on top of NP (except, e.g., Payne and Huddleston 2002). However, a consensus about the intermediate structure has not been reached but it seems to be emerging for some parts. Before I turn to a more detailed discussion, I should mention that there are some other important intermediate phrases not seen in (2c). For instance, scholars have proposed a Possessor Phrase (PossP), which is located between DP and CardP, and another phrase with a variety of labels (e.g., nP, DefP, ArtP), which is between AgrP and NumP. Following the extended projection line of the noun in (2c), I comment on the individual phrases moving bottom up.

23.1.2 NP: Noun Phrase Nouns are an open, lexical class. In German, they may involve features for gender, number, and case (Nu¨bling, Chapter 10). Gender is mostly an abstract lexical feature (Ku¨rschner, Chapter 12). It is usually assumed to be specified on the noun itself but with the noun having no exponence of it, gender becomes visible only on higher elements like adjectives and determiners. In contrast, number is typically not inherent with a noun. With a few exceptions, the singular form of the noun corresponds to singular semantics and the plural form to plural semantics. Number is often taken to originate in NumP. Similarly, nouns are not fixed for morphological case but may vary according to their syntactic context. Sometimes, this is claimed to be mitigated by a Kase Phrase (KP). I return to both NumP and KP. Exhibiting varying morpho-syntactic properties, nouns are categorized in different ways. I focus on the syntactic characteristics discussing three types of distinctions. Two differences suggest varying positions of the noun itself and one indicates different argument structures of the noun.

539

540

DORIAN ROEHRS

The most basic distinction is between common noun and proper noun. When in argument (e.g., subject) position, common count nouns must have an article (3a). While proper nouns cannot take an article in many languages, certain dialects of German allow an optional article (3b): (3)

a. Das Auto ist klein. the car is small ‘The car is small.’ b. (Der) Peter ist klein. the Peter is small ‘Peter is short.’

Based on cross-linguistic data, Longobardi (1994) argues that noun phrases involving proper nouns like Peter are also complex. Rather than head-tospecifier movement (i.e., N moves to Spec,NP), he argues that nouns move to the DP-level. Specifically, Longobardi (1994: 653) proposes that German has N-to-D movement (at LF) for cases like Peter but if an expletive article is present as in der Peter, a CHAIN is built. This analysis provides evidence for two head positions inside the noun phrase, one in DP and one in NP. One way to subcategorize common nouns is along the opposition of lexical versus semi-lexical. Lexical nouns are the lowest elements in the extended projection. They determine the gender of the noun phrase, they may undergo compounding, and all agreeing adjectives precede the (compound) noun (4a). Semi-lexical nouns, sometimes also called classifiers, are different. For instance, in pseudo-partitives, they may occur with another (lower) noun but they themselves determine the gender of the determiner (4b). When in the singular, they may combine with a nonsingularity numeral (4c). Furthermore, considering that all elements in (4d) are in the nominative case, most clearly marked on the adjectives, one can state that adjectives can both precede and follow semi-lexical nouns: (4) a.

das /*der neue (Wein)glas the.N / the.M new wine.M - glass.N ‘the new wine glass’ b. das /*der Glas Wein the.N / the.M glass.N wine.M ‘the glass of wine’ c. zwei Glas Wein two glass.S G wine ‘two glasses of wine’ d. (?) ein großes Glas roter Wein a big-N O M glass red-N O M wine ‘a big glass of red wine’

Lo¨bel (1990) proposes that both nouns in (4b–4d) are in one and the same extended projection where nouns like Glas are in an intermediate position

Structure of Noun (NP) and Determiner Phrases (DP)

between the determiner and the lexical noun, in the head position of CardP. However, given the structure (2c), this does not immediately explain adjectives preceding this noun. Furthermore, although (4d) indicates one nominal domain with regard to case, gender (4b) and number (4c) may differ. Given these issues, van Riemsdijk (1998) refines the concept of extended projection proposing that these intermediate elements are semilexical nouns. Unlike (2c), his structure exhibits only one phrasal projection (at the top). In addition to agreement in case (4d), the (lower) lexical noun and its related adjective may also exhibit other morphological cases (e.g., genitive). This means that at least for the latter instances, one typically argues for binominal structures where the second adjective and noun are not part of the same nominal as the first noun. Furthermore, while there is variation with regard to (4c), all Germanic languages allow the intermediate noun to be in the plural as in zwei Gla¨ser Wein ‘two glasses of wine.’ Both strings have different interpretations: (4c) involves a quantity reading (wine in the amount of two glasses); the plural counterpart has a container reading (two glasses filled with wine). Subtypes for both constructions have been identified. Again, different structural proposals have been made (see, e.g., Section 23.2.4). One way to subcategorize lexical common nouns is as concrete (5a), relational (5b), and deverbal (5c). The latter can be further subdivided into nominals with a result or event/process reading: (5)

a. Auto, Wein, Seele car, wine, soul b. Gesicht, Mutter face, mother c. Eroberung, Beteiligung conquest, participation

Differences between these nouns emerge when they combine with possessors (broadly defined) as in Peters Auto ‘Peter’s car.’ Cases with the nouns in (5a) involve alienable possession where the possessor can have a variety of interpretations that, in some way, associate the possessor with the possessum head noun. In contrast, nouns in (5b), that is, names for body parts and kinship terms, present instances of inalienable possession where the possessor and the possessum are linked by an inherent (e.g., part-whole) relation between them. Alexiadou (2003) proposes that possessors in alienable possession are located higher in the structure than possessors in inalienable possession. The first are in PossP and the latter undergo complex predicate formation with the possessum noun. Although possessors seem to be optional with nouns of both (5a–5b), they are assumed to be present at least covertly with the nouns in (5b).

541

542

DORIAN ROEHRS

This is different for the nouns in (5c) under the (complex) event reading. As argued by Grimshaw (1990), these types of nouns take obligatory (overt) arguments whose theta roles are similar to those of their corresponding verbs (e.g., agent and theme). Taking a cross-linguistic view, Alexiadou (2001) refines this proposal by arguing that variation in type and number of verbal projections like AspectP and vP above the lexical root explain the differences between these result and event/process nouns. Given the difference in obligatoriness of the dependents of the noun, it has been proposed that optional elements with concrete nouns are more like adjuncts and obligatory elements with process nouns are closer to arguments. Relational nouns and picture-nouns seem to have intermediate properties. As discussed by Broekhuis and Keizer (2012: chapter 2), this is a complex topic, complicated by the fact that dependents of the noun can be left out if implied or contextually recoverable. There are word order restrictions as regards the dependents of nouns. Genitive DPs are usually required to stay syntactically close to the head noun: (6)

a.

das Buch meines Bruders u¨ber Chomsky the book my-G E N brother about Chomsky ‘the book of my brother’s about Chomsky’ b. * das Buch u¨ber Chomsky meines Bruders the book about Chomsky my-G E N brother

Replacing the genitive DP by von meinem Bruder ‘of my brother’s’ shows a similar restriction although the contrast is less strong. Given partial N-raising (next section), the possessor in (6a) is in a higher position than the u¨ber-PP, in the specifier and complement of N in (2c), respectively. This is similar to deverbal nouns where agent arguments have to precede themes (7a). Theme arguments can only be in a higher position if the agent is part of a durch-PP (7b): (7)

a. Ca¨sars Eroberung Galliens/von Gallien dauerte Jahre. Caesar’s conquest Gaul’s /of Gaul took years ‘Caesar’s conquest of Gaul took years.’ b. Galliens Eroberung durch Ca¨sar dauerte Jahre. Gaul’s conquest by Caesar took years ‘Gaul’s conquest by Caesar took years.’

Other phenomena (binding, extraction, etc.) confirm that dependents of nouns are merged in certain positions subject to their theta roles. If the arguments are reordered as in (7b), there is debate as to whether the theme is base-generated in the higher position or moves there. However, there is also some word order variation. Although the u¨bercomplement has a closer semantic relation to the head noun than the aufadjunct, both word order possibilities exist (8a–8b). A relative clause follows a PP element (8c):

Structure of Noun (NP) and Determiner Phrases (DP)

(8) a. das Buch u¨ber Chomsky auf Franzo¨sisch the book about Chomsky in French ‘the book about Chomsky in French’ b. das Buch auf Franzo¨sisch u¨ber Chomsky the book in French about Chomsky ‘the book in French about Chomsky’ c. das Buch u¨ber Chomsky/auf Franzo¨sisch, das ich gelesen habe the book about Chomsky/in French, that I read have ‘the book about Chomsky/in French that I have read’ Similarly, although relative clauses seem to be preferred to occur closer to the head noun than complement clauses, both orders are in principle fine (9a–9b). PP elements typically precede clausal dependents (9c): (9) a. ? die Behauptung, dass Madonna kommt, die in der Zeitung the claim

that Madonna comes

steht

that in the newspaper is

‘the claim that Madonna is coming that is in the newspaper’ b.

die Behauptung, die the claim

in der Zeitung

steht, dass Madonna kommt

that in the newspaper is

that Madonna comes

‘the claim that is in the newspaper that Madonna is coming’ c.

die Behauptung in der Zeitung, the claim

dass Madonna kommt

in the newspaper that Madonna comes

‘the claim in the newspaper that Madonna is coming’

Thus, while word order restrictions are related to the theta role of the dependents, word order variation is subject to heaviness. Nouns can be elided. Traditionally, there are two strategies to support NP-ellipsis. While some languages license the elided noun with adjectival inflection (10a), others exhibit one-insertion (10b). (10) a. Sie hat ein großes Auto, und er hat ein kleines. she has a big-I N F L car and he has a small-I N F L b. She has a big car and he has a small one. The difference between these two strategies is often reduced to adjectival inflections, present in German but absent in English. Corver and van Koppen (2011) argue that there are languages where the licensing factor appears to be adjectival inflection but turns out to be a pronoun. Finally, one-insertion has been used as a diagnostic to distinguish between arguments and adjuncts of the head noun but there is growing evidence that this is not a reliable test.

23.1.3 NumP: Number Phrase Ritter (1991) argues that the locus of the number specification of the noun phrase is in NumP. Borer (2005) proposes that all noun roots are mass in interpretation and that number brings about countability deriving the

543

544

DORIAN ROEHRS

traditional distinction of mass versus count nouns. In the languages with a discrete plural suffix, this suffix is often taken to be in Num and the head noun is suggested to move there to combine with it (Julien 2005a). This was indicated in a simplified way in (2c). There is evidence from DP-internal binding that the head noun moves to (at least) NumP. Anaphors and pronouns have to be c-commanded by their antecedent to be bound (Lee-Schoenfeld, Chapter 21). We can state then that the genitive possessor in (11a) is higher than the PP-element: the possessor is in Spec, NP and the PP is the complement of N. Furthermore, it is a standard assumption that selection occurs in a local relation. However, while the head noun Wut selects the preposition auf, both elements are separated by the possessor. Similar argumentation applies to (11a) when des is replaced by jedes ‘each’ yielding a bound variable reading. These observations extend to (11b), where the head noun is separated from its clausal complement: (11) a. die riese Wut des Sohnesi auf seineni Vater the huge rage the-G E N son at his father ‘the huge rage of the son at his father’ b. die Behauptung des Kriminelleni, dass eri unschuldig ist the claim the-G E N criminal that he innocent is ‘the claim of the criminal that he is innocent’ If we assume that the head noun originates in N and moves up to precede the possessor, then we can state that the noun locally selects its complement in its base position. This provides a strong argument for partial N-raising of common lexical nouns.

23.1.4 AgrP: Agreement Phrase There are different kinds of adjectives, sometimes proposed to have different structural analyses (Bernstein 1993). Furthermore, different structural analyses have also been proposed for adjectives with restrictive versus nonrestrictive/appositive interpretation accounting for properties such as differences in word order and intonation (Alexiadou 2013: 478) as well as inflection (Pfaff 2015: chapter 3). I focus on restrictive adjectives. In the Germanic noun phrase, adjectives are typically prenominal and inflected (12) (for exceptions in both regards, see Payne and Huddleston’s 2002: 445 discussion of English). Depending on the syntactic context, the inflection on the adjective can alternate between strong (S T ) and weak (W K ) but adjacent adjectives typically have the same endings. In Section 23.2.2, I devote a separate discussion to this inflectional alternation. Second, if heavy as in coordinations, adjectives can also directly follow the noun (13), where the adjectives are uninflected in West Germanic but inflected in North Germanic. Third, adjectives can also be used predicatively in

Structure of Noun (NP) and Determiner Phrases (DP)

corpular constructions (14). Again, adjectives are uninflected in West Germanic but inflected in North Germanic. (12)

a. das groß-e teur-e Haus (German) the big-W K expensive-W K house ‘the big expensive house’ b. det stor-e dyr-e huset (Norwegian) that big-W K expensive-W K house-D E F

(13)

a. das Haus, groß und teuer (German) the house big and expensive ‘the house, big and expensive’ b. det huset, stor-t og dyr-t (Norwegian) that house-D E F big-S T and expensive-S T

(14)

a. Das Haus ist groß und teuer. (German) the house is big and expensive ‘The house is big and expensive.’ b. Det huset er stor-t og dyr-t. (Norwegian) that house-D E F is big-S T and expensive-S T

In both types of languages, adjectives in (13) and (14) pattern similarly, the same language-internally but in a different way cross-linguistically. It is sometimes proposed that not only (13) but also (12) is based on (14). In other words, adjectives are assigned a clausal analysis (Kayne 1994). Note though that both (13) and (14) must have a strong ending in North Germanic while (12) may alternate between a strong and a weak ending depending on the context (Section 23.2.2). This inflectional alternation does not immediately follow from a clausal analysis. Similarly, a clausal analysis leaves the very presence of any inflection on prenominal adjectives to be explained in West Germanic. Another argument against a clausal analysis is that not all prenominal adjectives can be used predicatively and vice versa (Alexiadou and Wilder 1998). Consequently, many scholars assume that adjectives have a nonclausal analysis. There are three main structural analyses for restrictive adjectives. Traditionally, adjectives are assumed to be adjoined to some projection in the noun phrase (Jackendoff 1977). Second, Abney (1987) proposes that adjectives are in head positions in the extended projection line of the noun. Third, Cinque (2010) argues that they are in specifier positions of recurring AgrPs. Abney (1987) proposes that adjectives are in head positions. One argument for this structural analysis comes from English. Adjectives can take arguments: The father is proud of his son. English allows no arguments with prenominal adjectives (15). This follows if the adjective selects NP as its complement and degree words are in Spec,AP. However, in formal style, the German counterpart allows arguments to precede the adjective (16). To

545

546

DORIAN ROEHRS

maintain the basic proposal, one could suggest that degree words are in a different, higher position. This simple structural adjustment does not account for the equivalent in Yiddish, which allows both preceding and following arguments (17): (15)

a. * the [of his son] proud father b. * the proud [of his son] father

(16)

a. der [auf seinen Sohn] stolze Vater b. * der stolze [auf seinen Sohn] Vater

(17)

a. der [mit zayn zun] shtoltser tate (Yiddish) b. der shtoltser [mit zayn zun] tate

The very existence of (17b) undermines this argument for adjectives as heads. A second argument for adjectives as heads is made by Bosˇkovic´ (2005). Assuming phrasal movement, adjectives as heads cannot move out of the noun phrase thus explaining the ban on Left-branch Extraction of adjectives (18a). Importantly, demonstratives have to be analyzed as heads as well (18b). However, without further assumptions, arguments of adjectives, as phrases, should be able to move out of the noun phrase, contrary to fact (18c): (18)

a. *

Stolze habe ich proud have I b. * Diese habe ich these have I c. * [Auf ihre So¨hne] of their sons

Va¨ter fathers Va¨ter fathers habe have

gesehen. seen gesehen. seen ich stolze Va¨ter gesehen. I proud fathers seen

In Section 23.1.6, we see evidence that demonstratives are more likely to be in a specifier position; that is, both (18b) and (18c) raise issues for an account postulating adjectives are heads. Corver (1997) argues that degree elements are inside the phrase containing the adjective. As pointed out by Svenonius (1994: 445–446), degree elements such as sehr take scope only over the adjective immediately on their right (# indicates absence of reading): (19)

sehr heißer schwarzer Kaffee very hot-S T black-S T coffee ‘very hot black coffee’ #‘very hot, very black coffee’

This observation is only compatible with adjectives in phrasal positions, either adjoined or as specifiers. Bosˇkovic´ (2016) provides a new account of the ban on Left-branch Extraction of adjectives. Crucially, it is based on adjectives as adjuncts to NP. Assuming phase theory, elements have to move through the edge to vacate a phase. Adjectives cannot move out of

Structure of Noun (NP) and Determiner Phrases (DP)

the noun phrase as movement from a NP-adjoined position to Spec,DP, the edge, is anti-local (i.e., the movement is too short as it must cross more than phrasal segments). However, there is also data that suggests that adjectives are in specifier positions. Julien (2005a: 9) observes that prenominal adjectives in some Scandinavian dialects can be separated by indefinite articles (20). Julien proposes that the article is in α (our Agr) and the adjectival inflection is part of the complex specifier of the adjective: (20)

?

eit stor-t eit styg-t a big-S T an ugly-S T ‘a big ugly house’

eit a

hus house

(Norwegian)

Adjectives may co-occur. Scott (2002) observes that they typically appear in a certain fixed sequence; for instance, size adjectives precede color adjectives. In accounts with adjoined adjectives, this is argued to be a semanticocognitive phenomenon (Sproat and Shih 1991); in analyses with adjectives in specifiers, this is taken to be syntactic (Cinque 2010). Under certain conditions, adjectives can be reordered (capitalization indicates focus stress). This reordering does not change the inflections on the adjectives: (21)

a. der groß-e rot-e Ballon the big-W K red-W K balloon ‘the big red balloon’ b. der ROT-E groß-e Ballon the red-W K big-W K balloon

It has been proposed that the focused adjective is in a Focus Phrase (FocP) above AgrP, although there is debate as to whether it is base-generated in FocP or moves there (for general discussion of DP-internal information structure, see Aboh et al. 2010).

23.1.5 CardP: Cardinal Phrase There are two types of quantifiers. Simplifying here, so-called weak quantifiers have an existential interpretation and strong quantifiers have a universal interpretation. Weak quantificational elements consist of cardinal numerals and quantifiers like viel. They precede adjectives and follow determiners (22a–22b). Typically and in stark contrast to adjectives, only one of these elements is possible inside the DP (22c): (22)

a. die zwanzig netten Studenten the twenty nice students ‘the twenty nice students’ b. die vielen netten Studenten the many nice students ‘the many nice students’

547

548

DORIAN ROEHRS

c. *

Die the

zwanzig twenty

vielen / vielen many / many

zwanzig twenty

Studenten students

Assuming that numerals and weak quantifiers are in the same positions, their complementary distribution follows. There are many parallels between numerals / weak quantifiers and adjectives. First, although inflection is not fully productive or obligatory in most cases, the endings of numerals/quantifiers are the same as those on adjectives. While the singularity numeral is inflected in all morphological cases, the numerals for ‘two’ and ‘three’ occur with inflection only in the genitive (23a). As for quantifiers, the inflection is often morphologically optional (23b): (23)

a. der Verkauf zwei-er Ha¨user the sale two-G E N houses ‘the sale of two houses’ b. mit viel(em) Interesse with much(-D A T ) interest ‘with a lot of interest’

Second, numerals and quantifiers can also be modified by a degree word: (24)

a. fast hundert Leute almost hundred people ‘almost one hundred people’ b. sehr viele Leute very many people ‘very many people’

Third, with contrastive stress on the adjective, cardinal numbers can follow: (25)

a. die zwei roten Autos the two red-W K cars ‘the two red cars’ b. die ROTEN zwei Autos the red-W K two cars

Fourth, numerals and weak quantifiers have the same inflection as a following adjective: (26)

a. viel-e nett-e Leute many-S T nice-S T people ‘many nice people’ b. die viel-en nett-en Leute the many-W K nice-W K people ‘the many nice people’

Structure of Noun (NP) and Determiner Phrases (DP)

All elements in (26) agree in gender, number, and case making up one nominal domain. Besides this possibility, quantifiers can also be followed by a DP in the genitive (27a) or a PP containing a DP (27b). These two strings are labeled partitive constructions. As recently discussed by Pfaff (2015: 84), Icelandic has a third possibility where a quantifier may be followed by a DP in concord (27c): (27)

a. viele der netten Leute (German) many-S T the-G E N nice-W K people ‘many of the nice people’ b. viele von den netten Leuten (German) many-S T of the-D A T nice-W K people ‘many of the nice people’ c. margar þessar bækur (Icelandic) many-S T these books ‘many of these books’

Quantifiers like viel involve different interpretations: in the cardinal reading, (26a) asserts the existence of a large number of members of a set restricted by the adjective and noun; in the proportional reading, (26a) denotes a large number of members of a pre-established set restricted by the adjective and noun. The examples in (27) are only proportional in interpretation. There are two main structural analyses of quantifiers: heads or specifiers. Their exact position has been argued for using different criteria. On the one hand, some scholars focus on the morpho-syntactic properties of these elements emphasizing the commonalities with or differences from adjectives. While Abney (1987) locates these elements in an intermediate head position, Julien (2005a) puts them in a specifier position. Danon (2012) argues for both. In contrast, Cardinaletti and Giusti (2006) propose that quantifiers are lexical heads at the bottom of the nominal structure. On the other hand, in the semantic literature, quantifiers have been argued to be in different positions depending on their reading: Quantifiers with a cardinal interpretation are in an intermediate position; quantifiers with a proportional reading are at the bottom of the matrix nominal. Unlike adjectives, quantifiers are not proposed to be in an adjoined position. Taking the morpho-syntactic parallels to adjectives seriously, quantifiers can be located in the specifier of CardP in both (26) and (27a– 27b). For the latter, one often assumes an elided noun and the quantified DP or PP is in the complement position of that noun. In Section 23.1.7, I return to the Icelandic pattern in (27c). Strong (universal) quantifiers like alle and jeder have different properties from adjectives. Resembling determiners, they cannot be preceded by a definite article (although the equivalent of (28a) is possible in Yiddish: di ale kleyne shtetlekh):

549

550

DORIAN ROEHRS

(28)

a. *

die alle(n) kleinen Sta¨dte the all-S T ( W K ) small-W K towns b. * der jede gute Student the every good student

Second, strong quantifiers have an influence on the inflection of following adjectives bringing about a weak ending: (29)

a. alle klein-en Sta¨dte all-S T small-W K towns ‘all small towns’ b. jeder gut-e Student every-S T good-W K student ‘every good student’

One might claim that the difference between the existential quantifiers in (26) and the universal quantifiers in (29) has to do with definiteness such that only the definite elements in (29) trigger a weak ending on the following adjective. However, such a claim does not pan out for German (Section 23.2.2). Rather, strong quantifiers are suggested to be of a different category. They are in a higher surface position in the noun phrase, a structural level I turn to next.

23.1.6 DP: Determiner Phrase Abney (1987) argues in detail for the DP-level, the location of determiners. Narrowly defined, determiners consist of definite articles and demonstratives (30a) as well as indefinite articles (30b) (note that Icelandic does not have the latter): (30)

a. der / dieser nette Freund the / this nice-W K friend ‘the/this nice friend’ b. ein netter Freund a nice-S T friend ‘a nice friend’

More broadly defined, one can also include in this group strong quantificational elements and question words (31a) as well as possessive pronouns and Saxon Genitives (31b). Typically, these elements are in complementary distribution with articles and demonstratives. As such, they are often referred to as determiner-like elements. With the exception of Saxon Genitives in German, these elements have to precede adjectives: (31)

a. jeder / welcher nette Freund every / which nice-W K friend ‘every/which nice friend’

Structure of Noun (NP) and Determiner Phrases (DP)

b. sein / Peters netter Freund his / Peter’s nice-S T friend ‘his/Peter’s nice friend’ Basically all scholars agree that the definite article der is a head. It cannot be modified and is thus assumed to be in D. There is some controversy with demonstratives. While a few scholars analyze it as a head in D, most argue that it is a phrasal element in Spec,DP. Diachronically, they form the source of the definite article whereby the latter has undergone specifierhead reanalysis (van Gelderen 2007). Synchronically, they may take reinforcers, a kind of modifier that reinforces or specifies the deixis of these elements (Roehrs 2010). While the combination of demonstrative and reinforcer is, for present purposes, not illuminating in German, other languages are more telling. Specifically, multiple reinforcers can follow in Eastern Norwegian or reinforcers can both precede and follow in Yiddish. No other element can intervene between the determiner and the reinforcer (reinforcers without a straightforward translation are glossed as R E I N F ) : (32)

a. [den herre her] klokka (Eastern Norwegian) this here-I N F L here watch-D E F ‘this watch’ b. [ot-o di dozike] froy (Yiddish) here-R E I N F this here- I N F L woman ‘this woman’

Leu (2015) proposes that similar to adjectives, demonstratives project complex structures. There is also disagreement about the location of the indefinite article. While it is traditionally assumed to be in D, more recent work questions this simple analysis (Sections 23.2.2 and 23.2.4). Strong quantifiers were discussed in Section 23.1.5; possessors will be taken up again in Section 23.2. For the discussion of constructions like solch ein ‘such a’, see Wood and Vikner (2011). Postal (1966) argues that personal pronouns are determiners. Often, these elements are interpreted as determiners of the first or second person. These pronouns can, in the plural, take a weak adjective – a hallmark of a determiner: (33)

wir dumm-en Idioten we stupid-W K idiots ‘we stupid idiots’

Pronouns may have various functions and different morpho-syntactic properties. Consequently, different structures have been proposed (Cardinaletti and Starke 1999, De´chaine and Wiltschko 2002). For the discussion of indefinite pronouns like something, see Roehrs (2008).

551

552

DORIAN ROEHRS

Besides distributional evidence for the DP-level, Abney (1987) highlights some parallels between the clausal and nominal domains, most clearly seen in some non-Germanic languages. Currently, there is no agreement as to whether DP corresponds to IP, as evidenced by agreement between the possessor/subject and the head noun, or CP, as suggested by subextraction of out DP. A third line of investigation tries to relate the two by proposing that subextraction indicates a DP/CP and possessor/subject agreement suggests PossP/IP.

23.1.7 XP: Pre-determiners The nominal domain does not end at DP. All Germanic languages allow all(e) to precede a definite DP (34a). With some variation (Cirillo 2016), many languages also allow the combinations in (34b–34d): (34)

a. all(?e) die Autos all(-I N F L ) the cars ‘all the cars’ b. all(e) meine Autos all(-I N F L ) my cars ‘all my cars’ c. all(e) diese Autos all(-I N F L ) these cars ‘all these cars’ d. all(e) diese meine Autos all(-I N F L ) these my cars ‘all these cars of mine’

It is fairly uncontroversial that possessors in (34b) are in Spec,DP and for many linguists demonstratives in (34c) are in that position too. If so, all(e) must be outside the DP proper. Considering (34d), there might be several DP-external positions. As briefly shown in the introduction, all these elements participate in concord; that is, they all belong to one nominal domain. In structure (2c), I labeled this phrasal level as XP. In Section 23.1.5, we have seen that Icelandic allows other quantifiers to precede an agreeing DP. The question arises as to whether we can specify XP as a Quantifier Phrase (QP) on top of DP. Given our current understanding, the answer appears to be negative. It seems clear that nonquantificational elements can also occur in front of a DP; for instance, strongly inflected adjectives in Icelandic (Pfaff 2015). One might then suggest a category-neutral label like the Kase Phrase (KP), often argued to top off the DP (Bayer et al. 2001). Recalling that all elements in the extended projection line show concord in gender, number, and case, let us assume for a moment that this is correct. With KP at the top, case features have to spread downwards; with number and gender low in the

Structure of Noun (NP) and Determiner Phrases (DP)

structure, the latter two features have to spread upwards. There is currently no consensus about the right way to analyze this type of agreement. For instance, Schoorlemmer (2009) argues for an Agree-based account and Norris (2014) proposes feature spreading with local feature copying. Thus, I retain the label XP. That there are more positions above the DP is confirmed by other facts. For instance, PPs of various meanings can be topicalized from postnominal position. Note that the Verb-Second constraint in German indicates that these PPs form one constituent with the following DP: (35)

a. Vor OSTERN die Woche geht nicht. before Easter the week goes not ‘The week before Easter does not work.’ b. Von PETER das Auto ist klein. of Peter the car is small ‘The car of Peter’s is small.’

The topicalized PP has to precede XP: (36)

a. (?)

b. ?*

von PETER all die Bu¨cher of Peter all the books ‘all the books of Peter’s’ alle von PETER die Bu¨cher all of Peter the books

Interestingly, these topicalized PPs are not possible in North Germanic here illustrated with Norwegian (data provided by Marit Julien): (37)

a. *

Før PÅSKE uka gaº r before Easter week-D E F goes b. * Til PETER bilen er liten. to Peter car-D E F is small

ikke. not

(Norwegian)

This empirical contrast between the two language groups aligns with differences in nominal left dislocation and possessor-related floating quantifiers (Grohmann and Haegeman 2003: 56–59). Compare the (a)-examples and the (b)-examples, respectively: (38)

a. Verhofstadt den dienen zen Verhofstadt the that.M his ‘Verhofstadt’s mistakes’ b. djoengers al under us the-kids all their house ‘all the kids’ house’

(39)

a. *

Per, han sin Per him his ‘Per’s plan’

plan plan

fouten mistakes

(Norwegian)

(West Flemish)

553

554

DORIAN ROEHRS

b. *

barna alle sitt kid-P L . D E F all his ‘all the kids’ house’

hus house

While more work is needed here, we may tentatively state that West Germanic projects more structure, call it Topicalization Phrase (TopP), than North Germanic: (40)

TopP von Peter

XP DP

all die



23.2 Germanic-Specific Phenomena In this section, I single out some other interesting facts. While not exclusively restricted to Germanic, they have generated a lot of discussion among Germanic linguists and have been very influential in the discussion of the structure of the DP. Both Double Definiteness and the strong/weak alternation of adjective endings suggest the presence of another intermediate phrase. Discontinuous noun phrases, spurious indefinite articles, and doubly filled DPs are also discussed.

23.2.1 Double Definiteness All Germanic languages have free-standing determiners, articles and demonstratives. In addition, the North Germanic languages also have suffixed determiners. Focusing on the main Scandinavian languages, bare nouns are followed by this suffixal determiner (41a). In contrast, modified nouns have a variety of distributions. They may be preceded by a determiner (41b), they may be sandwiched by two determiner elements (41c), or they may be followed by a suffixal determiner (41d). The distribution in (41c) is referred to as Double Definiteness and its two elements have been claimed to involve different semantics (Julien 2005a). I gloss the suffixal element as D E F : (41)

a. mand-en (Danish, [Norwegian, Swedish, Icelandic]) man-D E F ‘the man’ b. den gamle mand (Danish) the old-W K man ‘the old man’

Structure of Noun (NP) and Determiner Phrases (DP)

c. den gamle mann-en (Norwegian, [Swedish]) the old-W K man-D E F ‘the old man’ d. gamli maður-inn (Icelandic) old-W K man-D E F ‘the old man’ A host of analyses have been proposed. To mention just a few milestones, Delsing (1993) proposes that the noun raises to D in (41a). Adjectives block this head movement in (41b–41d) and a variety of other operations explain these distributions. Julien (2005a) argues that determiners can be spelled out in two different positions depending on the language. In addition to DP, nP can host definiteness features. Located between AgrP and NumP in (2c), these features can be spelled out by the suffixal determiner. Simplifying somewhat, Schoorlemmer (2012) proposes that these two positions (our DP and nP) are related by movement of the determiner and some late language-specific copy-deletion rules account for the distributions in (41b–41d). The above empirical picture becomes more complicated when restrictive relative clauses or possessors are added (Julien 2005a). Focusing on the latter and acknowledging some variation, if the possessor follows, both determiner elements are present (42a); if the possessor precedes, both disappear (42b): (42)

a. det store hus-et the big-W K house-D E F ‘the big house of mine’ b. mitt store hus my big-W K house ‘my big house’

mitt my

(Norwegian)

While the absence of the free determiner follows from the general complementary distribution of prenominal possessors and other determiners, the disappearance of the suffixal determiner is less clear. Julien (2005b) proposes that on their way to the DP-level, possessors move through the intermediate nP, the locus of the suffixal determiner, thus preventing its appearance. Possessors show the greatest variation in the Germanic noun phrase. Possessors can take the forms of pronouns, proper names, full DPs, PPs, or even a combination of these, and they may occur in different positions depending on those forms. Harbert (2007: 156) observes that overall possessive pronouns and proper names seem to pattern together (versus genitive DPs and PPs). Some of this variation is discussed by Delsing (1998), who adopts PossP, a phrase between DP and CardP in (2c).

555

556

DORIAN ROEHRS

23.2.2 Inflectional Alternation With the exception of English, all Germanic languages have inflections on prenominal adjectives alternating between a strong or a weak ending. The set of strong endings is bigger thus providing more information about gender, number, and case than that of the weak endings. The basic patterns are as follows: a definite article is followed by a weak adjective (43a), an indefinite article by a strong adjective (43b), and an unpreceded adjective by a strong ending (43c): (43)

a. das kalt-e Bier the cold-W K beer ‘the cold beer’ b. ein kalt-es Bier a cold-S T beer ‘a cold beer’ c. kalt-es Bier cold-S T beer ‘cold beer’

There are two basic analyses. On the one hand, definite elements like das cause the adjective to be weak (cf. Julien 2005a). On the other, if determiners like das have (or are interpreted as having) an inflection, then the adjective is weak (Esau 1973). Both the semantic and lexico-inflectional analysis work for (43). Harbert (2007: 135) points out that the Germanic languages do not pattern the same. Among others, this is evident with possessors: (44)

a. Peters kalt-es Bier Peter’s cold-S T beer ‘Peter’s cold beer’ b. Pers kald-e øl Peter’s cold-W K beer ‘Peter’s cold beer’

(Norwegian)

Roehrs (2015) argues that the North Germanic languages have a semantic account. Definiteness features in Julien’s nP bring about a weak ending on the adjective. Dutch patterns with North Germanic. The (remaining) West Germanic languages have a lexical account such that only certain determiner elements bring about a weak ending. In all languages, the strong endings present the elsewhere case. Adjectival inflections can also appear on determiners (broadly defined). Leu (2015) proposes that the (strong) inflection is on an Agr head inside the adjectival projection. If the adjective stem raises to Agr, (43c) obtains; if the adjective stem stays in situ, the inflection in Agr is supported by d- resulting in (43a); that is, the strong ending winds up on the determiner. Cases like (43b) have the analysis of (43c) with the indefinite article located

Structure of Noun (NP) and Determiner Phrases (DP)

outside the adjectival projection. A challenge for this type of account is that not all Germanic languages are as transparent in their inflections as German (Roehrs 2013).

23.2.3 Discontinuous DPs As briefly discussed in Section 23.1.4, Left-branch Extraction of adjectives is not possible. In contrast, dependents of nouns can be extracted (45a). Additionally, nouns as well as adjectives in combination with nouns can topicalize stranding higher elements of the noun phrase (45b–45c): (45)

a. Von wem hast du ein Bild gemalt? of whom have you a picture painted ‘Of whom have you painted a picture?’ b. Bu¨cher habe ich interessante gelesen. books have I interesting read ‘As for books, I have read interesting ones.’ c. Interessante Bu¨cher habe ich keine gelesen. interesting books have I none read ‘As for interesting books, I have read none.’

The cases in (45b–45c) are often labeled Split Topicalization. Unlike (45a), this presumably does not involve a simple movement operation whereby the lower part of the noun phrase undergoes displacement. This becomes evident by the possibility of two determiners (46a) or even two lexical nouns (46b): (46)

a. Ein / *Das Buch habe ich keins gelesen. a / the book have I none read ‘As for a book, I have read none.’ b. Bu¨cher habe ich Romane gelesen. books have I novels read ‘As for books, I have read novels.’

Ott (2015) proposes that the two related nominals are base-generated low in the structure, separately but in a symmetric fashion. Assuming that syntax does not tolerate symmetry, one nominal has to undergo movement. Another case of a discontinuous noun phrase is Quantifier Float where strong quantifiers like alle and jeder appear below their related noun phrase. While the inflection on alle is obligatory (cf. [34]), jeder exhibits a mismatch in number: (47)

a. Diese Sta¨dte sind all*(e) klein. these towns are all-I N F L small ‘These towns are all small.’ b. Die Kinder haben jedes ein Eis the children have each-S G an ice-cream ‘The children have each eaten an ice-cream.’

gegessen. eaten

557

558

DORIAN ROEHRS

Bosˇkovic´ (2004) argues that alle and its related DP form a constituent at a certain point in the derivation and movement of the DP may strand the quantifier. As for jeder, given the mismatch in number, the quantifier and its related nominal are often argued to involve binomial structures.

23.2.4 Spurious Indefinite Articles Bennis et al. (1998) discuss – what they call – spurious indefinite articles in Dutch, singular indefinite articles that are compatible with plural nouns: (48)

a. Wat *(een) jongens! (Dutch) what a boys ‘What boys!’ b. idioten van (een) mannen idiots of a men ‘idiots of men’

These linguists propose a small clause structure, represented by YP in (49), with some further functional positions on top. The nominal jongens is assumed to be the subject, wat is the predicate, and een is the head of the small clause. While een is proposed to move to D, the predicate wat undergoes fronting to Spec,DP: (49)

DP D’

Watk

YP

eeni

Y’

jongens ti

tk

This type of analysis, usually referred to as Predicate Inversion, has been extended to constructions like that idiot of a doctor (den Dikken 2006) and, as briefly discussed in section 23.1.2, pseudo-partitives (Corver 1998). It is fundamentally different from (2c) in that it does not involve the extended projection of a noun.

23.2.5 Doubly Filled DPs There is an interesting contrast between Yiddish and Danish possessives. In Yiddish, there is no agreement in definiteness between the possessor and the indefinite article. This indefinite article is obligatory when the possessive pronoun has an inflection (50a). In Danish, the possessor and the definite article agree in definiteness but the article is possible only when an adjective is present (50b):

Structure of Noun (NP) and Determiner Phrases (DP)

(50)

a. mayn-er a (guter) khaver (Yiddish) my-I N F L a good friend ‘a good friend of mine’ b. min (den) sorte kat (Danish) my the black cat ‘my black cat’

Most likely, the Yiddish construction does not involve a spurious indefinite article as all elements in (50a) have to be in the singular. In fact, this construction is restricted to singular. Considering that Germanic has null indefinite articles in the plural, one might suggest that mayne khaveyrim is the plural counterpart of (50a) meaning ‘friends of mine.’ However, it is only definite in interpretation ‘my friends.’ Given the nonagreement in definiteness in (50a), one could suggest that the possessor is in TopP in (40). In contrast, the Danish construction has been suggested to involve a doubly-filled DP, with the demonstrative in Spec,DP and the article in D. More generally, it appears that West Germanic does not allow (definite) doubly filled DPs (51a) but North Germanic does (51b): (51)

a. * b.

Poss/Dem Art Adj N dette (det) høje hus this the high house ‘this tall house’

(West Germanic) (Danish)

23.3 Summary Taking the structure in (2c) as a starting point, we have arrived at the following hierarchy where phrases to the left are higher in the structure: (52)

(TopP)

XP

DP

PossP

CardP

AgrP

nP

NumP

NP

Table 23.1 summarizes the five differences between the North and the West Germanic DPs (– indicates that no language has that property; √ means that at least some languages have that property):

Table 23.1 Differences between North and West Germanic

TopP above XP Inflection on postnominal and predicative adjectives Suffixal determiners Weak/strong alternation relates to definiteness (Definite) doubly-filled DP

North Germanic

West Germanic

− √ √ √ √

√ − − − (except Dutch) −

559

560

DORIAN ROEHRS

Currently, it is not clear why North Germanic does not seem to project TopP but West Germanic does. Second, weak endings on adjectives occur only inside DPs. In North Germanic, they require definiteness. Strong endings present the elsewhere case, which includes postnominal adjectives. Diachronically, West Germanic has lost inflection on postnominal adjectives. Finally, there is work in progress that attempts to relate the last three differences in Table 23.1. Assuming that definiteness consists of subcomponents, it is suggested that in North Germanic, individual components are distributed across the higher levels of the DP and can be spelled out separately. In contrast, West Germanic has all these components in one feature bundle in the DP-level.

References Abney, S. 1987. The English Noun Phrase in Its Sentential Aspect. Doctoral dissertation, MIT. Aboh, E. O., N. Corver, M. Dyakonova, and M. van Koppen 2010. “DP internal information structure: some introductory remarks,” Lingua 120: 782–801. Alexiadou, A. 2001. Functional Structure in Nominals. Nominalization and Ergativity. Amsterdam: John Benjamins. Alexiadou, A. 2003. “Some notes on the structure of alienable and inalienable possessors.” In M. Coene and Y. D’hulst (eds.), From NP to DP. Volume 2: The expression of possession in noun phrases. Amsterdam: John Benjamins: 167–188. Alexiadou, A. 2013. “Adverbial and adjectival modification.” In Dikken, M. den (ed.), The Cambridge Handbook of Generative Syntax. Cambridge University Press: 458–484. Alexiadou, A. and Wilder, C.1998. “Adjectival modification and multiple determiners.” In A. Alexiadou and C. Wilder (eds.), Possessors, Predicates and Movement in the Determiner Phrase. Amsterdam: John Benjamins: 303–332. Bayer, J., M. Bader, and M. Meng 2001. “Morphological underspecification meets oblique case: Syntactic and processing effects in German,” Lingua 111: 465–514. Bennis, H., N. Corver, and M. den Dikken 1998. “Predication in nominal phrases,” The Journal of Comparative Germanic Linguistics 1: 85–117. Bernstein, J. B. 1993. Topics in the Syntax of Nominal Structure across Romance. Doctoral dissertation, The City University of New York. Borer, H. 2005. In Name Only: Structuring Sense, Vol. I. Oxford University Press. Bosˇkovic´, Ž. 2004. “Be careful where you float your quantifiers,” Natural Language & Linguistic Theory 22: 681–742. Bosˇkovic´, Ž. 2005. “On the locality of left-branch extraction and the structure of NP,” Studia Linguistica 59: 1–45.

Structure of Noun (NP) and Determiner Phrases (DP)

Bosˇkovic´, Ž. 2016. “Getting really edgy: On the edge of the edge,” Linguistic Inquiry 47: 1–33. Broekhuis, H. and E. Keizer 2012. Syntax of Dutch: Nouns and Noun Phrases, Vol. 1. Amsterdam University Press. Cardinaletti, A. and G. Giusti 2006. “The syntax of quantified phrases and quantitative clitics.” In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax, Vol. V. Oxford: Blackwell: 23–93. Cardinaletti, A. and M. Starke 1999. “The typology of structural deficiency: A case study of the three classes of pronouns.” In H. van Riemsdijk (ed.) Clitics in the Languages of Europe. New York: Mouton de Gruyter: 145–233. Cinque, G. 2010. The Syntax of Adjectives: A Comparative Study. Cambridge, MA: MIT Press. Cirillo, R. 2016. “Why all John’s friends are Dutch, not German: On the determiner-like characteristics of the inflection on the universal quantifier in West Germanic,” Journal of Germanic Linguistics 28: 179–218. Corver, N. 1997. “The internal structure of the Dutch extended adjectival projection,” Natural Language & Linguistic Theory 15: 289–368. Corver, N. 1998. “Predicate movement in pseudopartitive constructions.” In A. Alexiadou and C. Wilder (eds.), Possessors, Predicates and Movement in the Determiner Phrase. Amsterdam: John Benjamins: 215–257. Corver, N. and M. van Koppen 2011. “NP-ellipsis with adjectival remnants: a micro-comparative perspective,” Natural Language & Linguistic Theory 29: 371–421. Danon, G. 2012. “Two structures for numeral–noun constructions,” Lingua 122: 1282–1307. De´chaine, R-M. and M. Wiltschko 2002. “Decomposing pronouns,” Linguistic Inquiry 33: 409–442. Delsing, L-O. 1993. The Internal Structure of Noun Phrases in the Scandinavian Languages: A Comparative Study. Doctoral dissertation, University of Lund. Delsing, L-O. 1998. “Possession in Germanic.” In A. Alexiadou and C. Wilder (eds.), Possessors, Predicates and Movement in the Determiner Phrase. Amsterdam: John Benjamins: 87–108. Dikken, M. den. 2006. Relators and Linkers: The Syntax of Predication, Predicate Inversion, and Copulas. Cambridge, MA: MIT Press. Esau, H. 1973. “Form and function of German adjective endings,” Folia Linguistica 6: 136–145. Gelderen, E. van. 2007. “The definiteness cycle in Germanic,” Journal of Germanic Linguistics 19: 275–308. Grimshaw, J. 1990. Argument Structure. Cambridge, MA: MIT Press. Grimshaw, J. 1991. “Extended projections.” Ms., Brandeis University. Grohmann, K. K. and L. Haegeman 2003. “Resuming reflexives,” Proceedings of the 19th Scandinavian conference on linguistics. Nordlyd 31: 46–62. Harbert, W. 2007. The Germanic Languages. Cambridge University Press.

561

562

DORIAN ROEHRS

Jackendoff, R. 1977. X’ Syntax: A Study of Phrase Structure. Cambridge, MA: MIT Press. Julien, M. 2005a. Nominal Phrases from a Scandinavian Perspective. Amsterdam: John Benjamins. Julien, M. 2005b. “Possessor licensing, definiteness and case in Scandinavian.” In M. den Dikken and C. Tortora (eds.), The Function of Function Words and Functional Categories. Amsterdam: John Benjamins: 217–249. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Leu, T. 2015. The Architecture of Determiners. Oxford University Press. Lo¨bel, E. 1990. “D and Q als funtionale Kategorien in der Nominalphrase,” Linguistische Berichte 127: 232–264. Longobardi, G. 1994. “Reference and proper names: A theory of N-movement in syntax and logical form,” Linguistic Inquiry 25: 609–665. Norris, M. 2014. A Theory of Nominal Concord. Doctoral dissertation, University of California, Santa Cruz. Payne, J. and R. Huddleston 2002. “Nouns and noun phrases.” In R. Huddleston and G. K. Pullum (eds.), The Cambridge Grammar of the English Language. Cambridge University Press: 323–523. Pfaff, A. 2015. Adjectival and Genitival Modification in Definite Noun Phrases in Icelandic: A Tale of Outsiders and Inside Jobs. Doctoral dissertation, University of Tromsø. Postal, P. M. 1966. “On the so-called pronouns in English.” In F. Dinneen (ed.), Nineteenth Monograph on Language and Linguistics. Washington DC: Georgetown University Press: 177–206. Riemsdijk, H. van 1998. “Categorial feature magnetism: The endocentricity and distribution of projections,” The Journal of Comparative Germanic Linguistics 2: 1–48. Ritter, E. 1991. “Two functional categories in noun phrases: Evidence from modern Hebrew.” In S. D. Rothstein (ed.), Perspectives on Phrase Structure: Heads and Licensing. San Diego: Academic Press: 37–62. Roehrs, D. 2008. “Something inner- and cross-linguistically different,” The Journal of Comparative Germanic Linguistics 11: 1–42. Roehrs, D. 2010. “Demonstrative-reinforcer constructions,” The Journal of Comparative Germanic Linguistics 13: 225–268. Roehrs, D. 2013. “The inner makeup of definite determiners: The case of Germanic,” Journal of Germanic Linguistics 25: 295–411. Roehrs, D. 2015. “Inflections on pre-nominal adjectives in Germanic: main types, subtypes, and subset Relations,” The Journal of Comparative Germanic Linguistics 18: 213–271. Schoorlemmer, E. 2009. Agreement, Dominance and Doubling: The Morphosyntax of DP. Doctoral dissertation, University of Leiden. Schoorlemmer, E. 2012. “Definiteness marking in Germanic: Morphological variations on the same syntactic theme,” The Journal of Comparative Germanic Linguistics 15: 107–156.

Structure of Noun (NP) and Determiner Phrases (DP)

Scott, G.-J. 2002. “Stacked adjectival modification and the structure of nominal phrases.” In G. Cinque (ed.), Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol. I. Oxford University Press: 91–120. Sproat, R. and C. Shih 1991. “The cross-linguistic distribution of adjective ordering restrictions.” In C. Georgopoulos and R. Ishihara (eds.), Interdisciplinary Approaches to Language: Essays in Honor of S.-Y. Kuroda. Dordrecht: Kluwer: 565–593. Svenonius, P. 1994. “The structural location of the attributive adjective.” In E. Duncan, D. Farkas, and P. Spaelti (eds.), The Proceedings of the Twelfth WCCFL, Stanford Linguistics Association: 439–454. Wood, J. L. and S. Vikner 2011. “Noun phrase structure and movement: A cross-linguistic comparison of such/saº dan/solch and so/saº /so.” In P. Sleeman and H. Perridon (eds.), The Noun Phrase in Romance and Germanic: Structure, Variation, and Change. Amsterdam: John Benjamins: 89–109.

563

Part IV

Semantics and Pragmatics

Chapter 24 Modality in Germanic Kristin Melum Eide

24.1 Introduction The realm of modality in language is also the realm of imagination. Modality deals with the description and qualification of what is perceived as less than real in one sense or another, and illustrates how humans can use language to think about and talk about entities and situations that are not proven to exist, perhaps even positively known to be counter-to-fact. Recent theories and discoveries tie the emergence of the human language ability to socalled Offline Brain Systems. These are systems of neurons that allow the human brain to create cognitive symbols for entities and events even in the absence of the actual entities and events themselves, and to manipulate these symbols cognitively; in short, to imagine entities and events that are not necessarily current, present or even real. According to, e.g., Bouchard (2013) and a range of research quoted therein, these offline brain systems are in fact what separate human brains from the brains of other primates. Comparative studies show that some neuronal systems differ in an important way in the brains of humans and other primates. The human brain has a system of neurons that can be activated in absentia: the individual does not have to see or hear an action for these neuronal systems to be activated. These Offline Brain Systems (OBS) are triggered by representations of events instead of the events themselves, and produce representations of events with no brain-external realization. (Bouchard 2013: 107).

That is, these neural systems allow humans to evoke, represent and respond to symbols of entities and events the way, e.g., primates and other intelligent animals respond to external stimuli – actual events – without actually simultaneously perceiving them, but simply by calling I thank the editors for their support, and Michael T. Putnam and Tor Åfarli for useful comments on the manuscript. All usual disclaimers apply.

568

KRISTIN MELUM EIDE

them to their mind’s eye. Language, according to this view, is the externalization of these cognitive operations, allowing humans to express and refer to these imagined entities and events through externalized symbols made visible to other humans, mainly via sounds (or other signs). Although all linguistic expressions hence depict reality in an indirect manner, most languages flag specific events as modal or less than real either through specific lexical items or through designated grammatical markers.1 These expressions are flagged to make the hearer aware that the speaker does not grant the current correspondence between the situations depicted in the predication or proposition flagged by these markers and situations in the real world.2 Instead, the speaker signals by these means that certain situations and events are, e.g., possible, probable, hypothetical, evidenced by someone else, or counterfactual; this is known as the domain of epistemic, evidential or – as an overarching term – propositional modality (Palmer 2001). In this case, the correspondence between the proposition and the depicted real world situation is flagged as “not one to one.” Other modal expressions describe how the real world could or should be manipulated to resemble the situation depicted by the proposition. The latter is achieved, e.g., by committing oneself or others to specific future behaviors via, e.g., promises, directives, commands, or statements about an individual’s potential, ability, or intentions. This is the domain of root modality (including deontic, dynamic and various types of “agent-oriented modality”), or in Palmer’s (2001) terms, event modality. Palmer (2001, unlike Palmer 1986) also acknowledges what has been called speech-act modality as a third subtype of modality, where the speaker carries out a performative act through the act of speaking. Asserting, asking, and giving orders are types of speech act modalities often expressed via designated structural means; as declaratives, interrogatives, and imperatives respectively (cf. Lyons 1977, 745 ff., also Nordstro¨m 2010: 269 ff.).

24.2 Defining and Delimiting Modality Providing a definition of modality that would encompass and suit all types of approaches to the subject seems utopic, and any definition should always be based on a thorough and sound theoretical analysis. However, 1

It is a well-known paradox that adding modality markers, even markers of definite certainty such as “certainly”, “necessarily” and “must”, does not strengthen the speaker´s commitment to the truth of the proposition; here, Lyons (1977: 809): “the very fact of introducing must, necessarily, certainly, etc., into the utterance has the effect of making our commitment to the factuality of the proposition explicitly dependent upon our perhaps limited knowledge.” Thus in uttering It is raining the speaker seems more committed to the statement´s factuality than in uttering It must be raining. He states that ”there is no epistemically stronger statement than the categorical assertion”, i.e., a nonmodalized declarative.

2

Von Fintel (2006: 1): “Together, modality and temporality are at the heart of the property of ‘displacement’ (one of Charles F. Hockett’s design features of human language) that enables natural language to talk about affairs beyond the actual here and now.”

Modality in Germanic

a working definition is part and parcel for any linguist trying to describe a grammatical category; in describing “the conceptual domain covered by the category” and “how this domain is divided by the members of the category” (Bybee 1985: 191). Portner (2009: 1) reluctantly provides the following definition: Modality is the linguistic phenomenon whereby grammar allows one to say things about, or on the basis of, situations which need not be real.

As van der Auwera and Aguilar (2016) demonstrate, the reluctance to provide a definitive description is common to the entire literature on mood and modality, extending from antiquity to the modern age (p. 9). There have been thousands of grammatical discussions of mood and modality and [. . .] the field as a whole cannot be said to have come to grips with these notions. This avowal has a negative and a positive side. The negative side is obvious: It is sad that after more than 2000 years our discipline has not reached a better understanding of what is fundamental to modality and mood. On the positive side, one gets the feeling that the subject matter of modality and mood is a fascinatingly difficult one [. . .].

Modality is often compared to and contrasted with the categories tense and aspect; cf., e.g., Palmer (2001: 1): Modality differs from tense and aspect in that it does not refer directly to any characteristic of the event, but simply to the status of the proposition. One possible approach to its analysis is to make a binary distinction between ‘non-modal’ and ‘modal’ or ‘declarative’ and ‘non-declarative’, and to associate this distinction with the notional contrast of ‘factual’ and ‘non-factual’, or ‘real’ and ‘unreal’. [..I]n recent years, ‘realis’ and ‘irrealis’ have been used for this distinction. For Mithun (1999: 173) the distinction is that “The realis portrays situations as actualized, as having occurred or actually occurring, knowable through direct perception. The irrealis portrays situations as purely within the realm of thought, knowable only through imagination.”

The attempt to draw a sharp distinction between tense and aspect on one side and mood and modality on the other is, however, impeded firstly by the fact that tense markers have clear modal uses in many Germanic languages. It is widely recognized that tense markers substitute for mood and modality markers in languages where mood is an obsolete category;3 compare, for instance, the Swedish counterfactual in (1) from Falk (2010: 4), to the German counterfactual in (1b). In the latter, a modal marker, “past” subjunctive mood (SUBJ2)4 is used where Swedish uses a preterit tense marker. Compare also the embedded Dutch quotative in (1c), from Patard (2014: 76) to the 3

Mezhevich (2008: 330): [I]f a language does not have a piece of morphology reserved specifically for Mood, a morpheme can be “borrowed” from Tense provided it licenses the right semantic relation.

4

Cf. below for an explication and discussion of these terms.

569

570

KRISTIN MELUM EIDE

Icelandic embedded nominal clause in (1d) from Sigurðsson (2010: 47). The former uses a tense marker; the latter uses mood.5,6 (1)

(a) I saº fall stod jag inte ha¨r. (Swedish) in that case stand.PRET. I not here ‘In that case; I would not stand here.’ (b) Dann stu¨nde ich jetzt nicht hier. (German) then stand.SUBJ2 I now not here ‘Then I would not be standing here.’ (c) Marie zei dat Jan morgen kwam. (Dutch) Mary said that John tomorrow come.PRET. ‘Mary said that John was coming tomorrow.’ (d) Jo´n vissi að Marı´a kæmi heim. (Icelandic) John knew that Mary come.SUBJ2 home ‘John knew that Mary would come home.’ (i.e., ‘John was confident that Mary would come home’)

Secondly, constructions most often analyzed as belonging to the aspectual domain (like present perfects) are also used as expressions of modality in some Germanic varieties. A case in point is counterfactual readings of present perfects. The sentence in (2) was part of a dialectal survey presented to dialectal users all over Scandinavia in the project ScanDiaSyn7, and the responses revealed that this is a common counterfactual in numerous dialects in a wide belt across Norway and Sweden; cf. Eide (2017); cf. also Larsson (2014). (2) Ha/har jeg vært yngre, (Norwegian / Swedish / ti aº r have I be.PERF. ten years younger Finno-Swedish) ha/har jeg studert fysikk. have I study.PERF physics ‘If I were/had been ten years younger, I would have studied physics.’ The term inferential perfect (term due to Haugen, 1972) likewise describes a quotative or evidential use of the present perfect; another type of modal reading of an aspectual construction attested in Mainland Scandinavian (cf., e.g., Kinnander 1974, Izvorski 1997, Elsness, 2003). Unlike purely temporal-aspectual uses of the present perfect this inferential use co-occurs with punctual adverbials; i.e., no Present Perfect

5

Note that even Icelandic may employ, e.g., a preterit indicative verb here (kom), with a different interpretation as

6

Worth mentioning here is the emotive preterit in Mainland Scandinavian, where the preterit occurs in exclamatives

a result (’John was aware of the fact that Mary was coming.’) as an immediate response to a perception: Dette var en god vin! Lit. ´This was a good wine.´ 7

http://www.tekstlab.uio.no/nota/scandiasyn/. This sentence is ambiguous in many dialects (although not in FinnoSwedish) between a present and a past reading of the counterfactual, since in the relevant dialects there is syncretism between the present and the preterit form of the auxiliary.

Modality in Germanic

Puzzle-effect (Klein 1992) arises in these instances, cf. this example from Faarlund et al. (1997: 632). (3)

ned i redaksjonen (Norwegian) Saº har han gaº tt rett so has he gone straight down into newsroom.DEF etter at han var hos henne. after that he was with her ‘Then he evidently went straight down into the newsroom after he was with her.’

A third complicating factor is the observation that markers of future tenses in some Germanic languages are modal markers formally;8 a case in point is the English modal will, and a fourth potentially problematic point is the existence of so-called quantificational (or existential) readings of modal expressions (Carlson 1977, von Wright 1951); cf. (4) from Leech (1969: 223), which could be paraphrased as “Lions are sometimes dangerous.” This reading is also referred to as weak epistemic or potential. (4)

Lions can be dangerous.

Quantificational readings of modal expressions are clearly semantically akin to aspectual and (a)temporal constructions like habituals, generic statements and universal expressions, hence this is another area where the line between modality, tense and aspect is blurred.9 Apart from the evident problems related to clearly separating mood and modality from certain uses of tenses and aspects, to isolate “an area of meaning” encompassing all and only linguistic modality has proven notoriously difficult. Instead, authors typically resort to listing the specific semantic and syntactic notions they will include as the relevant ingredients in their approach. Thus, Barbiers (2002: 1) and Barbiers and van Dooren (2017) use this list, explicitly stated to be nonexhaustive: Modality is a semantic concept and covers, inter alia, notions such as possibility, necessity, probability, obligation, permission, ability and volition. Modality can be expressed syntactically by modal verbs, imperatives, verbal inflection, modal adverbs and modal particles.

Von Fintel (2006: 1) likewise states that the following list of expressions carrying modal meanings constitutes only a “subset of the variety one finds in English”: 8

In some sense the future is always modal more than temporal, and in various languages the mood marker for irrealis is used to refer to future situations, e.g., Dyirbal and Burmese (Comrie 1985: 45).

9

To many logicians (ever since Aristotle) all modality is quantificational; necessity being a type of a universal quantifier and possibility an existential operator. Aristotle saw modality as quantifying over times (Possibly, Socrates is sitting means that Socrates does sit down sometimes). To others, modality quantifies over possible worlds (e.g., Leibniz, Kripke, Montague), models of reality, model sets, etc. See Brennan (2004:13) and Eide (2005: 27 ff.) for discussion.

571

572

KRISTIN MELUM EIDE

Modal auxiliaries Semimodal Verbs Adverbs Nouns Adjectives Conditionals

Sandy must / should / might / may / could be home. Sandy has to / ought to / needs to be home. Perhaps, Sandy is home. There is a slight possibility that Sandy is home. It is far from necessary that Sandy is home. If the light is on, Sandy is home.

These two lists illustrate that modality can be expressed through lexical means (nouns, adjectives), semilexical elements (semimodal verbs, adverbs), syntactic elements (modal auxiliaries, verbal inflection like subjunctive mood, modal particles) and certain designated constructions (imperatives, conditionals). The comparison between the two lists also illustrates that authors differ greatly in regards to the specific entries they will choose to include in their list of modal expressions. The present chapter focuses mainly on syntactic exponents of modality, although other ways of expressing modality will be discussed where relevant. Modal verbs and auxiliaries will be at center stage in a substantial part of this chapter. These specific modality exponents have always played a major role throughout the existing, and quite comprehensive, literature on Germanic modality.

24.3 The Deontic (Root) – Epistemic Distinction “In the field of modality, traditionally two kinds are differentiated, deontic and epistemic modality” (Thieroff 2010: 2), and almost every work on modality is, to some extent, concerned with this distinction. In Section 24.1 this was referred to via Palmer’s (2001) more overarching terms event modality versus propositional modality, but the terms “deontic” versus “epistemic” are more widespread, have a much longer tradition, and are hence more familiar. The former set of terms terms (deontic, event, or root modality) regards the commitment or potential of an individual to future actions, whereas the latter set (epistemic and propositional) regards the speaker’s commitment to the truth of the relevant proposition. Here von Fintel (2006). Epistemic modality (Greek episteme, meaning ‘knowledge’) concerns what is possible or necessary given what is known and what the available evidence is. Deontic modality (Greek: deon, meaning ‘duty’) concerns what is possible, necessary, permissible, or obligatory, given a body of law or a set of moral principles or the like.10

It is not uncommon to extend the term “epistemic” to include “evidential” readings (i.e., modality regarding the speaker’s source of evidence), as von Fintel does here, and the term “deontic” is often extended so as to include 10

The “body of law” or “set of moral principles” or the person inducing the obligation or permission is often referred to as “the modal source”, or “the source of modality.”

Modality in Germanic

“dispositional” or “dynamic” readings, i.e., volition and ability. Other works are careful to keep dynamic and dispositional readings separate from deontic modality, which is only taken to denote, e.g., “modality of necessity or possibility of acts performed by morally responsible agents” (Lyons 1977: 823, von Wright 1951). Very often reference to the deontic-epistemic distinction is used as a shorthand for the two domains; subsuming “an individual’s commitment to future actions” and “the speaker’s current commitment to the truth of the relevant proposition,” respectively. Although most works on modality discuss these two domains as both belonging to the very core of modality meanings, it is not obvious that these two notions (commitment to future actions and current truth evaluation) constitute a natural, conceptually coherent field within a domain of linguistic expressions. As we will discuss shortly, modals and semantically similar elements often display exactly these two readings cross-linguistically. The very fact that the same linguistic form may be used for deontic/dynamic and epistemic/evidential modality in language after language studied, naturally entices modality researchers to find the missing link between the two domains, in strongly suggesting that the relation is there, albeit recalcitrant to characterization. Thus many authors have tried to find the right level of abstraction allowing for one semantic-conceptual field encompassing both, or possibly two adjacent fields, allowing elements to leak from one domain into the other (most often in a unidirectional fashion). Hence, Chung and Timberlake (1985: 246) offer the following insight. There is considerable parallelism between the epistemic and deontic modes. Both can be described in terms of alternative worlds . . .. As a morphosyntactic realization of this parallelism, modal auxiliaries in many languages, notably English, often have both epistemic and deontic senses . . .. The crucial difference between the two, then, is that the epistemic mode deals with a set of alternative worlds at a given time [the alternative worlds are those that could exist instead of the given world], while the deontic mode deals with a set of alternative worlds that develop out of a given world and time [alternative futures of a given world].

Others point out that this parallelism in morphosyntactic realization leads to a rather forced quest for a common denominator for the two types of modality. According to Palmer (1986: 96) There are two features that they share: subjectivity . . ., i.e. the involvement of the speaker, and non-factuality. Yet it must be admitted that the chief reason for treating them as a single category lies in the fact that in English, and many other languages, the same forms (e.g. modal verbs) are used for both. There are possibly some deeper reasons.

The phenomenon of two meanings in one form is illustrated by the Norwegian, German, and Faroese examples in (5) below,11 but 11

From Eide (2005), Kratzer (2002), and Thrá insson et al. (2004), respectively.

573

574

KRISTIN MELUM EIDE

corresponding and comparable examples can be found in all Germanic languages, as cross-linguistically, modal verbs and auxiliaries typically display these two types of reading. Note that some utterances need a bit of context to allow for a second reading; e.g., (5b I) becomes more natural in the context of casting a play, whereas (5c I) is felicitous when discussing Mary’s future boyfriend and the specifications he will need to meet. (5)

(a) Jon maº være hjemme. Jon must be home

(Norwegian)

(I) ‘John is obligated to be at home.’ (root/deontic) (II) ‘It must be the case that John is at home.’ (epistemic) (b) Der Kastenjakl kann der Mo¨rder sein. the Kastenjakl can the murderer be

(German)

(I) ‘Kastenjakl is allowed to be the murderer’ (root/deontic) (II) ‘It may be the case that Kastenjakl is the murderer’ (epistemic) (c) Hann skal vera sera evnarı´kur. he shall be very talented

(Faroese)

(I) ‘He needs to be very talented.’ (root/deontic) (II) ‘He is said to be very talented’ (epistemic/evidential) The ability of modals to take on these two different readings is often considered a crucial and, in some works, even defining property of modal auxiliaries (e.g., Thra´insson and Vikner 1995). Others claim that this requirement is too strong, and that the English modal can is no less a modal even though it no longer gives rise to an epistemic reading (although could does). Likewise, the Swedish evidential la¨r ‘seems to’ has no root reading, but is still considered a modal, and the Norwegian faº ‘get’, like the Icelandic þurfa ‘need’ both seem very modal-like but lack a clear epistemic reading. The ability to take on both deontic and epistemic/evidential readings is by no means limited to bona fide modal auxiliaries.12 Cross-linguistically verbal elements denoting directive actions may gradually take on a reading where the complement changes from denoting a future action into denoting a current or past situation and the directive verbal element changes into an evidential or epistemic marker. (6)

12

(a) (b) (c) (d)

He insists that I leave him. He insists that I left him. John is supposed to become an architect. John is supposed to be an architect.

(English)

Authors are often reluctant to count a Germanic modal verb as a proper auxiliary unless it meets the syntactic criterion of taking bare infinitival complements, i.e., lacking an infinitival marker. For others, this is not a valid criterion. Hence, Thrá insson and Vikner (1995) list a range of Icelandic modals not meeting this criterion.

Modality in Germanic

In (6a) and (6c), the complement sentences that I leave him and to become an architect are reported directions, denoting potential, but as yet unrealized situations. The complement sentences in (6b) and (6d), in contrast, are reported claims about a situation perceived as real (by someone other than the speaker). In this case, the reading of the semi-modal be supposed to receives the same interpretation as the German modal sollen, Flemish zou, Norwegian skulle; i.e., an evidential (“hear-say”) reading. Note that the the root/deontic reading in (6a) in fact depends on a (relative) future reading of the complement, and in the literature there is a certain consensus that the temporal and aspectual properties of the complement (i.e., the “proposition”) determine the possible readings of the modal element (e.g., verb or auxiliary) as deontic/root or epistemic/evidential. Specifically, it is observed that root and epistemic modals select for different aspect feature matrices of their complements: root modals allegedly select for dynamic (eventive) predicates only and epistemic/evidential modals for stative predicates only. This claim is often augmented with the assumption that root readings are impossible with progressive and perfect complements (aspectually stative). In (7a) the stative predicate as predicted gives rise to an epistemic reading of the modal, and in (7b) and (7c) the perfect and progressive complements, respectively, seem to yield no reading beside the epistemic one. However, note that when we add a purpose clause or a particular type of temporal adverbial, as in (7d), (7e), and (7f, from Brennan 2004), the root reading becomes the natural one. (7)

(a) Jon maº virkelig like pannekaker. Jon must really like pancakes ‘Jon must really like pancakes.’

(Norwegian)

(b) Jon maº ha spist. Jon must have eaten ‘Jon must have eaten.’ (c) The water must be boiling.

(Norwegian)

(English) (d) Jon maº virkelig like pannekaker (Norwegian) Jon must really like pancakes for at svigermora skal like ham. for that mother-in-lawDEF shall like him. ‘Jon must really like pancakes for his mother-in-law to like him.’ (e) Jon maº ha spist før han kommer. Jon must have eaten before he arrives ‘Jon must have eaten before he arrives.’ (f) The water must be boiling when you pour it over the tomatoes.

(Norwegian)

(English)

Likewise, it has often been assumed in the literature on modals that the root senses denote two-place semantic relations, whereas the epistemic

575

576

KRISTIN MELUM EIDE

(or more generally, nonroot) senses denote one-place relations. Thus, on the root reading of must “there is such a role as someone who must something” (Vikner 1988: 14); there seems to be a predication between the subject referent and the modal, unlike on the epistemic (nonroot) reading, where no such relation between the two exists (cf. for instance the data in (5) above). Dyvik (1999: 4) argues that Every modal can be interpreted either as a one-place epistemic modal or as a two-place root modal. Under the epistemic interpretations the subject referent is not an argument of the modal, which only takes the entire proposition as an argument . . .. Under the root interpretation the subject referent is an argument of the modal.

While this generalization holds to some extent for epistemic/evidential versus dynamic root modals, it does not hold for the deontic root senses. As pointed out by numerous authors (e.g., Brennan 1993 Wurmbrand 1999, Eide 2005), deontic root modals may very well yield so-called proposition-scope readings; thus even root modals seem capable of taking the entire proposition as an argument. Feldman (1986: 179) notes that Sometimes, instead of saying that a certain person ought to do a certain thing, we may say that a certain state of affairs ought to be, or ought to occur . . .. The ought-to-do involves a relation between an agent and a state of affairs. The ought-to-be involves a property of a state of affairs.

In the latter sense, ‘ought to be’, the obligation or permission is clearly not directed towards the subject referent. Thus, it is not the subject referent who has an obligation or permission to do something in (8a), and (8b–8d) show that root modals may perfectly well occur with expletive subjects, although there exist many proposals to the contrary (8c is from von Fintel 2006). The presence of an expletive naturally excludes a reading where the subject has an obligation/ permission. (8) (a) Skilpadden bør være i badekaret. (Norwegian) turtle-DEF ought-to be in bath tub-DEF ‘The turtle should stay in the bath tub.’ (#The turtle has an obligation) (b) Es soll rote Rosen regnen. (German) It shall red roses rain ’It ought to rain red roses.’ (c) There ought to be a law against double parking. (English) (d) Det kan komme ti gjester i fødselsdagen din. (Norwegian) there can come ten guests in birthday-DEF yours ‘There may come ten guests to your birthday (party).’ (# There are ten guests & each has the permission to come)

Modality in Germanic

There have also been recurring claims about strong restrictions on the possible combinations of modals, but the range of possible combinations is in fact wider than predicted by many approaches. Evidential modals can be stacked on top of epistemic modals (two epistemic modals can also be stacked), and root modals can be stacked on top of each other (the Dutch example is from van Gelderen 2004: 165). (9) (a) Butleren skal visstnok maº tte være morderen. (Norwegian) Butler.DEF. shall probably must.INF be killer.DEF. ‘One claims that the butler evidently must be the killer.’ (b) Wer ko¨nnen soll muss wollen du¨rfen. (German) who can.INF shall must want.INF may.INF ‘He who is expected to be able, must be allowed to have a will.’ (c) Je zou hier eigenlijk moeten kunnen mogen kamperen. (Dutch) you should here really must can may camp ‘You should really be able to be allowed to camp here.’ There are also frequent claims about the modal’s ability to interact with tense and aspect, depending on its reading as root or epistemic/evidential. It is true that epistemic and evidential modals are much less susceptible to tense alternations than their root counterparts (since they denote current evaluations about the truth of propositions), but contexts where such alternations are possible do exist, for instance, when referring to past speech acts and their corresponding evaluations. Clearly, though, different types of modals do behave differently with regard to tense alternation. However, the claim (e.g., in Plank 1984) that there exists a finiteness requirement pertaining to epistemic/evidential modals but not to root modals cannot be corroborated. In many Germanic languages even epistemic/evidential modals may occur in nonfinite forms (e.g., the infinitive and the perfect participle); cf. Eide (2005, 2011) for Norwegian, Fagan (2001) for German and Boogart (2007) for Dutch. (10)

(a) Han paº staº s aº skulle være syk. (Norwegian) He claim.PASS to shall.INF be sick ‘It is claimed that he is allegedly sick.’ (b) Er hat krank sein sollen. (German) He has sick be shall.INF ‘They claimed he was sick.’ (c) [. . .] how onzeker Beethoven geweest heeft moeten zijn. (Dutch) how insecure Beethoven been has must be ‘How insecure Beethoven must have been.’

Only a few of the semantic and formal differences often claimed to exist between root and epistemic/evidential modals can be upheld upon closer scrutiny, in inspecting the full range of Germanic modals. Even (11a) does not apply to English or Icelandic, since in neither of these languages

577

578

KRISTIN MELUM EIDE

do modals take nonverbal directional complements. Hence, only (11b) applies to all Germanic modals. The generalization is interesting and calls for an explanation, although no really convincing approach exists (but see Roberts and Roussou 2003 for an attempt). Example (11a) is illustrated in (12a–12b) below, and (11b) is illustrated by the Norwegian data in (12c–12d). (11)

(a) Only root modals take non-verbal directional complements. (b) A nonroot modal will always scope over a root modal, if they occur in the same clause.

(12)

(a) Marit bør hjem. Marit ought-to home ‘Marit should go home.’ (b) Ko¨nnen Sie selber ins Auto? can you self in-the car ‘Are you able to get into the car by yourself ?’ (c) Jon kan ha maº ttet dra tidlig. John may have must-PERF go early ‘John may have had to leave early.’ (d) Jon maº ha kunnet dra tidlig John must have can.PERF go early ‘John must have been able to leave early.’

(Norwegian)

(German)

(Norwegian) (Norwegian)

In a sense, root modality is more fundamental than epistemic modality, since diachronically and cross-linguistically, it has been observed over and over again that modal expressions originating as root modality expressions (which are in turn often derived from experiencer verbs; Diewald 1999) gradually take on an epistemic reading, usually in addition to, but sometimes to the replacement of, the root reading (Bybee et al. 1994: 195, Thra´insson and Vikner 1995, Diewald 1999). This recurring historical development of modal elements is often used as a poster-child in the literature on grammaticalization, where a lexical element gradually takes on inflection-like functions (cf., e.g., Hopper and Traugott 1993, Roberts and Roussou 2003). However, a note of caution is in order since there is no necessity that all deontic modal elements will develop an epistemic reading, and not all epistemic modal elements originate from a deontic element (Nuyts 2006: 16).

24.4 Authority and Controllability Propositional modality (evidential/epistemic), event modality (root; deontic/dynamic), and speech act modality (declarative, interrogative, imperative) very often employ designated syntactic markers as exponents in natural languages. However, as modality involves many different aspects

Modality in Germanic

of language use, it is also evident that certain types of modality are sometimes conveyed even in the absence of any overt marker or specific exponent; cf. the examples in (13) below. (13) (a) Man kommer majsen i kogende vann. (Danish) one take.PRES corn(cob).DEF in boling water ‘Put the corncob in boiling water.’ (b) Man nehme drei Eier. (German) one take.SUBJ1.3sg. three eggs ‘Take three eggs.’ (c) Søknaden sendes til kommunen for behandling. (Norwegian) application send.PASS to municipality for treatment ’Please send the application to the municipality for processing.’ (d) Private Jones will report at 18.00 hrs. (English) In the context of a recipe, as in (13a), what is formally a nonmodal declarative is still perceived as a directive; some German authors would use a present subjunctive here, cf. (13b).13 Likewise, the felicitous interpretation of (13c) is clearly as a mild order, and even though the declarative in (13d) is not formally an imperative, the mere authority of an officer issuing (13d) is enough to ensure that the targeted addressee (here in third person, Private Jones) will perform the described action, which is why the US army often uses declaratives like these for written instructions (cf. Palmer 1986: 30). Authority is a central notion in the modality literature. In order for a speaker to issue an order or to grant a permission to manipulate the world to resemble the situation depicted by the relevant proposition (which is the domain of deontic modality, here interacting with speech act modality), the speaker must possess the necessary amount of authority over the situation and also over the addressee. As seen above, if the authority is strong enough and salient enough, no explicit expression of modality is needed, but there also exist designated syntactic means to issue an order or command. Lyons (1977: 745 ff.) argues that crosslinguistically, the three basic types of utterances, i.e., statements, questions, and commands usually have structural counterparts corresponding to them; declaratives, interrogatives, and imperatives, respectively (cf. also Palmer (1986: 23) and many others). Imperatives are hence specifically marked to primarily convey a command,14 request, or order, whereas interrogatives and especially declaratives are seen as nonmodal sentence types, or unmarked sentence types. Recall from the Palmer (2001: 1) quote 13

This use of the subjunctive is sometimes referred to as a jussive mood; cf. e.g., Palmer (2001: 81): “[F]irst

14

Thieroff (2010: note 3): ”However, there are at least some languages where the imperative form clearly has non-

and third person ‘imperatives’ are often simply called ‘jussives’.” imperative, modal uses, for instance in Dutch . . . ” Thieroff refers to Boogart and Janssen (2010), and the latter give, among others, this example: Had dat mar eerder gezegd; lit. ’Have that but before said’; i.e., ‘You should have said that before.’

579

580

KRISTIN MELUM EIDE

above that it is possible to “make a binary distinction between ‘non-modal’ and ‘modal’ or ‘declarative’ and ‘non-declarative’”. More recent research into Germanic languages, sentence types, and the modality inherent in various types of speech acts offers a new perspective on this question. Truckenbrodt (2006: 264) for instance states that In an imperative, as in [O¨ffne das Fenster ‘open the window’], S[peaker] wants A[ddressee] to do something in the real world. In declaratives and interrogatives, S[peaker] wants to change the world by changing the epistemic states of S[peaker] or A[ddressee] by conveying knowledge to A[ddresee] in an assertion, or by wanting to know something from A[ddressee] in a question; these are the epistemic speech acts [. . .].

In this type of approach, all speech acts (except pure and perhaps involuntary exclamations) hence involve a modal relation, in that they are volitional on the part of the speaker. Volition is one of the notions covered by the term modality, and the speaker always wants something from the addressee.15 In the case of declaratives, the Speaker wants the Addressee to accept the Speaker’s assertion as true; hence admitting this proposition into the common ground (the shared knowledge of Speaker and Addressee; cf. Stalnaker 1978). In the case of interrogatives, “the call is for common ground in regard to the true answer to the question” (Truckenbrodt 2006: 264), normally followed by the Addressee supplying the desired information and thereby adding this proposition into the common ground. In the case of imperatives, the Speaker requires the Addressee to act in a certain way, hence “to do something in the real world.” Another important notion in the modality literature (discussed under very many different headings) is controllability as the Addressee must be able to provide the required response in order for a modal expression to be used felicitously. This is rather easy to prove in the case of imperatives, as be careful is usually a felicitous command whereas #be tall is not (Truckenbrodt 2006: 271). The property of being tall is usually beyond what the Addressee would be able to control, unlike being careful. The notion of controllability is composed of a number of features (see Wa¨rnsby 2009), where being an intentional agent is one. Usually imperatives are directed at interlocutors, i.e., human beings, and if we direct commands at, e.g., inanimate things, we typically need to coerce the context to allow for an animation of the Addressee: (14)

Come on and SNOW, stupid sky!

Moreover, directives and commands are always future-projecting in that one cannot felicitously ask an Addressee to manipulate the world or carry 15

Truckenbrodt dubs this volitional relation deontic, but the more widespread common term for volitional modality and ability is dynamic modality.

Modality in Germanic

out an order in the past. Hence, the situations depicted by the proposition is always posterior to the time of the speech act, like all deontic modality, e.g., permission. Here Lyons (1977: 843): John may have come yesterday construed as a permission-granting utterance is semantically anomalous for the same reason that Come yesterday, John! is anomalous.

According to Truckenbrodt (2006), however, not only deontic speech acts (like imperatives) but even “the epistemic speech acts”, declaratives and interrogatives, require a presupposition of control on behalf of the Addressee. If the Speaker utters It is raining, the Addressee controls whether the proposition (that it is raining) becomes common ground, in accepting or refusing to accept the relevant proposition into the set of shared knowledge. The Speaker is hence tacitly asking for a judgment from the Addresse (is p true or false?). In interrogatives, the Addressee once again controls whether the true answer to a question, like Is it raining? becomes common ground. The Addressee controls this aspect via providing or withholding the answer to the question, hence controlling the update of the bank of shared knowledge. The Addressee’s act of updating common ground is hence also posterior to the the Speaker’s utterance, and the Speaker’s wish for this update to happen always constitutes a deontic (volitional) part of even epistemic speech acts. The proposition constituting the potential object of knowledge, however, can be about a present situation (It is raining), about the past (It was raining) or about the future (It will be raining by then). Whereas the judgment or evaluation of a proposition as true or false, i.e., as corresponding to the perceived and relevant reality always happens in the speech situation, the situation depicted by the proposition can be placed anywhere on a time line. This is different in deontic modality, like directives and imperatives, since these modalities require the Addressee to manipulate the real world in a certain way. Hence all deontic modalities are future-projecting – they operate on propositions depicting a situation that is temporally subsequent to the eventuality-time of the modality, for conceptual reasons. When we require, intend, want, need, or permit a situation to hold, we know that our intentions, wants, or needs cannot possibly influence events that have already taken place.

24.5 Mood and Modality Modality is often used as an overarching category encompassing mood as one of its possible expressions. Other definitions use the term modality as a functional-semantic category, while mood is “but one of the means, namely morphological, of expressing modality”, e.g., Khomutova (2014: 397); or the other way around: “Mood categories express modalities such as

581

582

KRISTIN MELUM EIDE

orders, wishes, (non-)factivity, (non-)reality and the like”; Thieroff (2010: 2). On the other hand, there are also many works where the two terms mood and modality are simply conflated. Hence, for instance Bybee (1985: 28) defines mood as follows. Mood refers to the way the speaker presents the truth of the proposition, whether as probable, possible, or certain.

This clearly amounts to a functional-semantic or conceptual definition of mood, where modal auxiliaries are considered to be an expression of mood, especially on their epistemic reading. The epistemic reading, as discussed in Section 24.3 above, is the interpretation of the modal as a speaker’s evaluation of or “attitude to the truth-value or factual status of the proposition” (Palmer 2001: 8). Huddleston (1984: 164) likewise proposes that modal auxiliaries constitute “an analytic mood system”, and Cinque (1999: 78), although he does distinguish between mood and modality, claims that the difference is in many cases virtually nonsignificant: [T]he same category may be expressed via mood in one language and with a modal in another, thus suggesting a close link between the two. ‘Mood’ is traditionally restricted to modal categories . . . which are expressed in verbal morphology [Cinque’s italics]. Modals instead are typically independent words (verbs, auxiliaries, or particles).

The idea that mood and modal auxiliaries share substantial semantic features and functional domains resonates throughout the literature on mood, modals, and modality. Hence Roberts (1985: 41–42), among others, suggests that the loss of subjunctive mood markings in Middle English paved the way for more frequent nonroot uses of modals, where the modals were construed as clausal operators similar to subjunctive inflection. Hence, modals were “semantic substitutes” for subjunctive mood inflection. Thieroff (2010: 2) states that in some languages (e.g., Slavic languages) there is a drift in the other direction in that modal verbs may transform into particles that at some point contribute to a morphological mood category. This might suggest, perhaps, that modal auxiliary verbs are less likely to be found in languages with inflectional mood or designated morphological mood markers. Moreover, one might be inclined to think that in languages employing inflectional mood or designated mood markers in addition to modal auxiliaries, inflectional mood would render epistemic readings of modal auxiliaries superfluous, so root readings (i.e., deontic and dynamic readings) would be the only readings of modal auxiliaries. However, languages such as German and Icelandic readily defy this expectation since they employ inflectional mood and modal auxiliaries; the modal auxiliaries in these languages express both root and epistemic/ evidential modality, similar to modal auxiliaries in Norwegian and English. For example, in German we even find a subjective inflection on

Modality in Germanic

the modal itself, and the modal may still have a nonroot (i.e., epistemic) and a subjunctive reading simultaneously, as in (15). (15)

a. Das du¨rfte wahr sein. that dareSUBJ2 true be ‘That might be true.’

24.6 Morphosyntactic Markers of Modality The Germanic languages have a comprehensive range of expression types (from lexical via semi-lexical to syntactic) at their disposal in expressing modality. However, the range of designated structural means varies between the languages, and especially within the domain of modality expressed as verbal inflections, i.e., mood. The most widespread mood in European languages is the subjunctive mood, in the German-influenced grammar literature often referred to as der Konjunktiv. It displays an opposition to the unmarked mood, the indicative. As pointed out by Thieroff (2010: 23) the subjunctive is disappearing from the Germanic languages: There is one group of languages which are in the process of losing their subjunctive, thus developing towards moodless languages. These are the Germanic languages Norwegian, Swedish, Danish, Dutch, and English.

We add Faroese to this batch, here Thra´insson et al. (2004: 68): The subjunctive was extensively used in Old Norse (and still is in Modern Icelandic) but there are no remains of that kind of a subjunctive in Modern Faroese. Only a few relic forms still occur in main clauses in relatively fixed expressions (. . .).

Even for the other languages mentioned, the subjunctive is confined to relics and fixed expressions. Icelandic and German constitute the very notable exceptions to this trend, since in these two languages, the subjunctive has a full paradigm of forms as counterparts to the indicative in all tenses (Thieroff 2010: 19). What we refer to as the subjunctive in Germanic is a descendant of the old Indo-European optative (cf. Næs 1972: 267; also Fertig, Chapter 9), and it comes in two paradigms or “tenses”, the present subjunctive (Konjunktiv I / subjunctive 1) and the preterit subjunctive (Konjunktiv II / subjunctive 2). The term “preterit subjunctive” is in some sense a misnomer since there is a general consensus that the difference between the present subjunctive and the preterit subjunctive in Germanic languages is not one of tense, but of mood, the relative perceived distance to reality. In the words of Thieroff (2004: 319): [T]he subjunctive 1 and the subjunctive 2 do not differ with regard to time reference. For example, both er singe and er sa¨nge have non-past time reference and differ only with regard to their modal meaning. In contrast,

583

584

KRISTIN MELUM EIDE

in the indicative the present form er singt has non-past time reference, whereas the preterite form er sang has past time reference. [The] same behaviour holds for the subjunctive forms in all Germanic (and in most other European) languages.

In Old Norse the subjunctive was quite productive (as mentioned in the quote from Thra´insson et al. above) and it occurs frequently in the old saga texts. The two types of subjunctive split the domain of subjunctive meanings roughly in half: The present subjunctive (subjunctive 1) expressed optative meanings (desires, wants, wishes, demands) and the preterit subjunctive (subjunctive 2) expressed more hypothetical, potential, or counterfactual meanings (Haugen 1998: 274 ff.); Eide (in press). (16) (a) Friðr se´ þe´r með o˛llum bo˛rnum þı´num peace be.PRES.SUBJ you with all children yours ‘peace be with you and all your children’

(Old Norse)

(b) Væri nu´ sva´ vel með oss sem þa´ er bezt hefir be.PRET.SUBJ now so well with us as when it best have verit (Old Norse) been ‘Might it be as good between us as when it was at its best’ Many works on mood and modality in Germanic include the imperative as one instantiation of mood.16 This partly follows from the definition of mood as verbal inflection, since Germanic languages mostly have a designated imperative verb form. However, this is not the case for all Germanic languages. In English, the imperative form (the bare form) equals the infinitive, and in Dutch only one verb, the verb zijn ‘be’ has a specialized imperative form; here Boogart and Janssen (2010: 124): It has, therefore, been suggested that the imperative in Dutch is constituted by a specific clause type, characterized not only by the form of the verb, but also by the clause-initial position of the verb and the lack of an explicit subject [. . .].

The tendency to view mood or more generally, modality, as a combination of specific morphological markers and their occurrence in a specific structure or clausal position has had several new adherents recently. For instance, in most Germanic languages the placement of the finite verb in the second or first position of a clause is said to have specific semantic ramifications for the clause. Over several decades there has been a range of proposals trying to shed light on the semantic meaning residing in the phenomena verb second and verb first, and verb second has been claimed to signify the speech act modality “assertion”, whereas verb first signifies 16

Although Thieroff (2010: 6) points out that one might argue against the view of the imperative as a mood, as the imperative form has nonimperative uses, for instance in Dutch.

Modality in Germanic

“interrogative” or “conditional”, i.e., underdetermined truth value. Later these phenomena received a more sophisticated treatment in Lohnstein (2000, 2001). Lohnstein develops a compositional theory of sentence moods making use of Frege’s (1892) three acts linked to form an assertion: thinking (forming the thought), judging (determining the truth of the proposition), and claiming (providing the illocutionary force of claiming). Different speech acts (such as yes-no questions, wh-questions, exclamations, and directives) address only a subset of these three; e.g., of the assertion process. Lohnstein and Bredel (2004) elaborate on the relation between inflectional morphology and sentence mood in German, and put forth that only a subset of finite markers (i.e., indicative present and preterit and the subjunctive preterit) may perform the role usually attributed to finiteness markers; i.e., in serving as morphological cues to the act of asserting. Finite markers such as imperative and subjunctive present do not lead to assertions, cannot be used as ordinary declaratives, and neither allow for question formation; cf. (17) below: (17)

(a) Peter steige auf die Leiter. Peter climb.SUBJ1 on the ladder ‘Peter climb(SUBJ1) the ladder.’ (*Nein, das ist nicht wahr) (‘*No, that isn’t true’.) (b) *Wer steige auf die Leiter? who climb.SUBJ1 on the ladder ‘*Who climb(SUBJ1) the ladder?’ (c) Wer steigt/stieg/stiege auf die Leiter? who climb.IND.PRES/IND.PRET/SUBJ2 on the ladder ‘Who climbs/climbed/would climb the ladder?’

Nordstro¨m (2010) suggests that the verbal inflectional moods have syntactic counterparts in the Germanic complementizers, such that the unmarked mood, the indicative, has as its correspondent the complementizer THAT. The subjunctive likewise has a correspondent in the complementizer IF; (Nordstro¨m 2010: 173):17 In the Germanic languages, THAT introduces declarative complement clauses, whereas IF introduces either conditional protases, embedded polar questions, or both. This is clearly a Realis-Irrealis distinction. Declarative clauses are semantically Realis, whereas conditional protases and embedded polar questions are semantically Irrealis (i.e., hypothetical or counterfactual and uncertain respectively).

Nordstro¨m also explores the notions realis and irrealis in relation to verb first and verb second phenomena, and argues that V1 signifies a type of Irrealis, too (Nordstro¨m 2010: 152): 17

Whereas Kratzer (1986) sees If-clauses instead as “devices for restricting the domains of various operators”, cf. also von Fintel (2006).

585

586

KRISTIN MELUM EIDE

The V1-order in polar questions and conditional protases [. . .] is in complementary distribution with IF [. . .]. The common denominator between these clause types is that they are both notationally Irrealis. Polar questions are Irrealis in the sense that the truth of the proposition is unknown by the speaker [. . .], whereas conditional protases are Irrealis in the sense that the proposition is hypothetical or counterfactual.

If these ideas are on the right track, even in those Germanic languages where inflectional mood no longer exist, the distinction between the modally unmarked indicative and the modally marked subjunctive is expressed through the complementizers THAT and IF, respectively.

24.7 Conclusion Modality is a category encompassing very many different aspects of language use, and the literature concerned with this topic extends from antiquity to the modern age. Hence there is no way one small chapter could do justice to the vast literature existing on this topic for the Germanic languages. The present chapter focused on some recurring topics from the modality literature, such as propositional modality, event modality, and speech act modality. These different types of modality were illustrated by some major classes of syntactic markers, such as conditionals, subjunctive mood, imperatives, and modal auxiliaries. As pointed out by Thieroff (2010: 3), morphological moods are generally restricted to expressing either deontic or epistemic modality, whereas we have seen that modals are typically defined by their ability to express both. Unlike subjunctive mood, modal auxiliaries also exist in all Germanic languages; cross-linguistically displaying both root readings (deontic and dynamic) and epistemic/evidential readings. The relation between mood and modality is another topic discussed, and the potential semantic relation between types of mood and certain complementizers, specifically THAT and IF. We have hardly touched upon the intriguing topic of counterfactuals, or conditionals, for reasons of space. Furthermore, we have excluded from our discussion such important categories as modal adverbs, for which claims have been made that their relative positions reveal the universal hierarchy of syntactic heads (Cinque 1999) and hence point to an underlying universal syntactic structure. Another very interesting field of modality is discourse markers of various kinds, for instance instantiated by discourse particles like the German ja, Swedish ju, Norwegian jo. Although these particles may seem to contribute only modestly to the propositional content of the clause, it seems that their semantics may be quite intricate. The aforementioned particle jo, for instance, is given the following description in Berthelin et al. (2013):

Modality in Germanic The analysis of 150 natural occurrences of jo shows that jo encodes the following evidential restriction: p is mutually manifest i.e. the hearer and speaker both have access to all the evidence required for entertaining p as true.

In an analysis inspired by Truckenbrodt (2006), one might say that jo is employed by the speaker to remind the hearer that p is already part of common ground. We started this chapter claiming that modality lies at the heart of language use as it highlights the human ability to entertain ideas about things that are not real. We will conclude the same way, in reminding the reader that although the study of modality started out a couple of thousand years ago, it is still seen as an exciting and vibrant field of linguistic study. Here we end with a quote from Harrog (2012: 1): Modality has become one of the most vibrant areas of linguistic studies in the past decade, as linguists have come to recognize its importance for language and communication. Human beings constantly imagine things that are not real but still possible or even contrary to facts. They continually evaluate the status of states-of-affairs, or urge or allow interlocutors to bring about states-of-affairs not present in reality. And they constantly put these activities into words. Imagining things that are not real, things that go beyond what is factual, is certainly one of the most interesting and complex features of human cognition. Its representation in language, in turn, is certainly worth investigation, and the fact that this has been done so intensely only recently is not least due to the very complexity of the processes behind it, and of the linguistic expressions themselves.

References Auwera, J. van der and A. Z. Aguilar 2016. “The History of Modality and Mood.” In J. Nuyts and J. van der Auwera (eds.), The Oxford Handbook of Modality and Mood. Oxford University Press: 9–29. Barbiers, S. 2002. “Modality and Polarity.” In S. Barbiers, F. Beukema, and W. van der Wurff (eds.), Modality and Its Interaction with the Verbal System. Amsterdam: Benjamins: 51–73. Barbiers, S. and A. van Dooren 2017. “Modal auxiliaries.” In M. Everaert and H. C. van Riemsdijk (eds.), The Wiley-Blackwell Companion to Syntax, 2nd edn., Vol. V. Oxford: Wiley-Blackwell: 2646–2676. Berthelin, S., K. Borthen, and C. S. Knudsen 2013. “Midtstillet jo–semantiske og pragmatiske aspekter.” Paper presented at MONS 15. University of Oslo, November 21, 2013. Boogart, R. 2007. “The Past and perfect of epistemic modals.” In L. Saussure, J. de Moeschler, and G. Puskas (eds.), Recent Advances in the Syntax and Semantics of Tense, Mood and Aspect. Berlin: Mouton de Gruyter: 47–71.

587

588

KRISTIN MELUM EIDE

Boogart, R. And T. Janssen 2010. “Mood in Dutch.” In B. Rothstein and R. Thieroff (eds.), Mood in the Languages of Europe. Amsterdam: Benjamins: 117–132. Bouchard, D. 2013. The Nature and Origin of Language. Oxford University Press. Brennan, V. 1993. Root and Epistemic Modal Auxiliary Verbs. Ph.D. dissertation, University of Massachusetts, Amherst. Brennan, V. 2004. Modalities. Ms., Vanderbilt University, Nashville. Bybee, J. 1985. Morphology – A Study of the Relation between Meaning and Form. Amsterdam: Benjamins. Bybee, J., R. Perkins, and W. Pagliuca 1994. The Evolution of Grammar – Tense, Aspect, and Modality in the Languages of the World. Chicago: The University of Chicago Press. Carlson, G. N. 1977. References to Kinds in English. Doctoral dissertation, University of Massachusetts, Amherst. Cinque, G. 1999. Adverbs and Functional Heads. Oxford University Press. Chung, S. and A. Timberlake 1985. “Tense, aspect and mood.” In T. Scopen (ed.), Language Typology and Syntactic Description: Grammatical Categories and the Lexicon. Cambridge University Press: 203–258. Comrie, B. 1985. Tense. Cambridge University Press. Diewald, G. 1999. Die Modalverben im Deutschen. Grammatikalisierung und Ployfunktionalita¨t. Tu¨bingen: Niemeyer. Dyvik, H.1999. “The universality of f-stucture: Discovery or stipulation? The case of modals.” Proceedings of the LFG99 Conference, CSLI Publications. Eide, K. M. 2005. Norwegian Modals. Berlin: Mouton de Gruyter. Eide, K. M. 2011. “Modals and the present perfect.” In T. Mortelmans, J. Mortelmans, and W. de Mulder (eds.), In the Mood for Mood. Rodopi: Cahiers Chronos: 1–20. Eide, K. M. 2017. “Counterfactual Present Perfects.” In M. Fryd and P-D. Giancarli (eds.), Aorists and Perfects: Synchronic and diachronic Perspectives. Leiden: Brill Academic Publishers: 245–288. Eide, K. M. in press. “Have-less perfects.” In K. M. Eide and M. Fryd (eds.), Norwegian: An Old Norse Heritage. The Perfect Volume. Papers on the Perfect. Amsterdam: Benjamins. Elsness, J. 2003. “A contrastive look at the present perfect/preterite opposition in English and Norwegian,” Languages in Contrast 3.1: 3–40. Faarlund, J. T., S. Lie, and K. I. Vannebo 1997. Norsk Referansegrammatikk. Oslo: Universitetsforlaget. Fagan, S. 2001. “Epistemic modality and tense in German,” Journal of Germanic Linguistics 13.3: 197–230. Falk, C. 2010. “The Interaction of Temporal and modal Auxiliaries in Counterfactual Contexts in Swedish,” Tampa Papers in Linguistics 1: 1–12. Fintel, K. von 2006. “Modality and Language.” In D. M. Borchert (ed.), Encyclopedia of Philosophy, 2nd edn., Vol. 6. Detroit: Macmillan Reference: 20–27.

Modality in Germanic

Frege, G. 1892. “Der Gedanke. Eine Logische Untersuchungen.” In G. Patzig (ed.), 1986: Logische Untersiehurg (3): 30–53. Go¨ttingen: Vandenhoeck and Ruprecht. Gelderen, E. van 2004. Grammticalization as Economy. Amsterdam: Benjamins. Haugen, E. 1972. “The inferential perfect in Scandinavian: a problem of contrastive Linguistics,” The Canadian Journal of Linguistics 17.2: 132–139. Haugen, O. E. 1998. Grunnbok i norrønt spraº k. Oslo: Ad Notam Gyldendal. Hopper, P. J. and E. C. Traugott 1993. Grammaticalization. Cambridge Textbooks in Linguistics. Cambridge University Press. Huddleston, R. 1984. Introduction to the Grammar of English. Cambridge University Press. Izvorski, R. 1997. “The present perfect as an epistemic modal.” In A. Lawson and E. Cho (eds.), 1999: SALT VII. Ithaca: Cornell University: 99–116. Khomutova, T. N. 2014. “Mood and Modality in Modern English,” Procedia: Social and Behavioral Sciences 154: 395–401. Kinnander, B. 1974. “Perfektum i sekunda¨r anva¨ndning,” Nysvenska studier 53: 127–172. Klein, W. 1992. “The present perfect puzzle,” Language 68: 525–552. Kratzer, A. 1986. “Conditionals,” Chicago Linguistics Society 22.2: 1–15. Kratzer, A. 2002. “The Notional Category of Modality.” In P. Portner and B. H. Partee (eds.), Formal Semantics: The Essential Readings. Oxford: Blackwell: 289–323. Larsson, I. 2014. “Conditional clauses and the shape of HAVE,” Nordic Atlas of Language Structures (NALS) Journal 1: 287–299. Leech, G. N. 1969. Towards a Semantic Description of English. London: Longman. Lohnstein, H. 2000. Satzmodus – kompositionell. Zur Parametrizierung der Modus-phrase im Deutschen. Berlin: Akademie Verlag. Lohnstein, H. 2001. “Sentence mood constitution and indefinite noun phrases,” Theoretical Linguistics 27.2–3: 187–214. Lohnstein, H. and U. Bredel 2004. “Inflectional morphology and sentence mood in German.” In H. Lohnstein and S. Trissler (eds.), Syntax and Semantics of the Left Periphery. Berlin: Mouton de Gruyter: 235–264. Lyons, J. 1977. Semantics. Cambridge University Press. Mezhevich, I. 2008. “A time-relational approach to tense and mood.” In N. Abner and J. Bishop (eds.), Proceedings of the 27th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Proceedings Project: 326–334. Mithun, M. 1999. The Languages of Native North America. Cambridge University Press. Narrog, H. 2012. Modality, Subjectivity, and Semantic change. A Cross-Linguistic Perspective. Oxford University Press.

589

590

KRISTIN MELUM EIDE

Næs, O. 1972. Norsk grammatikk. Elementære strukturer og syntaks. Oslo: Fabritius. Nordstro¨m, J. 2010. Modality and Subordinators. Amsterdam: Benjamins. Nuyts, J. 2006. “Modality: Overview and linguistic issues.” In W. Frawley, W. Klein, and S. Levinson (eds.), Expression of Cognitive Categories: The Expression of Modality. Berlin: Walter de Gruyter: 1–26. Palmer, F. R. 1986. Mood and Modality. Cambridge Textbooks in Linguistics. Cambridge University Press. Palmer, F. R. 2001. Mood and Modality, 2nd edn. Cambridge Textbooks in Linguistics. Cambridge University Press. Patard, A. 2014. “When tense and aspect convey modality. Reflections on the modal uses of past tenses in Romance and Germanic languages,” Journal of Pragmatics 71: 69–97. Plank, F. 1984. “The modals story retold,” Studies in Language 8: 305–364. Portner, P. 2009. Modality. Oxford University Press. Roberts, I. 1985. “Agreement parameters and the development of English modal auxiliaries,” Natural Language and Linguistic Theory 3: 21–58. Roberts, I. and A. Roussou 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge University Press. Sigurðsson, H. 2010. “Mood in Icelandic.” In B. Rothstein and R. Thieroff (eds.), Mood in the Languages of Europe. Amsterdam: John Benjamins: 33–55. Stalnaker, R. 1978. “Assertion.” In P. Cole (ed.), Syntax and Semantics 9: Pragmatics. New York: Academic Press: 315–332. Thieroff, R. 2004. “The subjunctive mood in German and the Germanic languages.” In W. Abraham (ed.), Focus on Germanic Typology. Studia Typlogica 6. Berlin: Akademie Verlag: 315–358. Thieroff, R. 2010. “Moods, moods, moods.” In B. Rothstein and R. Thieroff (eds.), Mood in the Languages of Europe. Amsterdam: John Benjamins: 1–29. Thra´insson, H., H. P. Petersen, J. I´ L. Jacobsen, and Z. S. Hansen. 2004. Faroese. An Overview and Reference Grammar. To´rshavn: Føroya Froðskaparfelag. Thra´insson, H. and S. Vikner 1995. “Modals and double modals in the Scandinavian languages,” Working papers in Scandinavian Syntax 55: 51–88. Truckenbrodt, H. 2006. “On the semantic motivation of verb movement to C in German,” Theoretical Linguistics 32.3: 257–306. Vikner, S. 1988. “Modals in Danish and event expressions,” Working papers in Scandinavian Syntax 39: 1–33. Wa¨rnsby, A. 2009. “Controllability as a contextual variable.” In Studies on English Modality: In Honour of Frank Palmer: 69–98. Wright, G. H. von 1951. “Deontic Logic.” Mind 60: 1–15. Reprinted in G. H. von Wright, A. Tsangalidis, and R. Fachinnetti (eds.), Logical Studies. London: Routledge and Kegan Paul: 58–74. Wurmbrand, S. 1999. “Modal verbs must be raising verbs.” In S. Bird, A. Carnie, J. D. Haugen, and P. Nordquest (eds.), Proceedings of the 18th West Coast Conference on Formal Linguistics (WCCFL18): 599–612.

Chapter 25 Tense and Aspect in Germanic Languages Kristin Melum Eide

25.1 Introduction Many natural languages signal via designated grammatical markers the time when a predicated event takes or took place; in the past, the present or the future. Such designated markers are for instance particular inflections (affixes), particles, or auxiliaries, usually elements with an affinity to the verbal domain. Exponents serving to locate a situation in time, especially relative to the moment of utterance or the speech event, are described as tense markers. Likewise, languages often encode the internal constituency of events and states, distinguishing a dynamic action from a stative one; denoting whether or not an action is bounded or unbounded, completed, or if it repeats itself. These features of an event are usually described as their aspect. Tense and aspect interact in intricate ways cross-linguistically and have been subjected to scrutiny by several generations of grammarians. This chapter examines some seminal assumptions and descriptions of these notions with a particular view to Germanic languages.

25.2 Tense According to a widespread consensus, “tense is grammaticalized expression of location in time”; Comrie (1985a: 9). Tense as a grammatical category is used to locate a given situation (typically encoded by a predicate) relative to some other situation or time, usually the speech event or the moment of speech.1 Hence, it falls under Lyon’s (1977: 637) definition of a deictic category, belonging to the category of deixis. I want to thank the editors for their support, and Michael T. Putnam and Tor Åfarli for valuable comments and suggestions. Remaining errors are my own. 1

Von Fintel (2006: 1): ”Together, modality and temporality are at the heart of the property of ‘displacement’ (one of Charles F. Hockett’s design features of human language) that enables natural language to talk about affairs beyond the actual here and now.”

592

KRISTIN MELUM EIDE

By deixis is meant the location and identification of persons, objects, events, processes and activities talked about, or referred to, in relation to the spatiotemporal context created and sustained by the act of utterance and the participation in it, typically, of a single speaker and at least one addressee.

Thus, the sentences in (1a) and (1b) can be described as identical in all respects, involving the same situations, events, and participants, only their tense features differ. In (1a), the verb has a tense affix encoding (at least) present, locating the smiling event as simultaneous to the speech event; the corresponding verb in (1b) has a preterit tense affix encoding past and locating the smiling event at a point in time prior to the speech event. (1)

a. Marit smiler. Marit smile.PRES ‘Marit smiles.’

(Norwegian)

b. Marit smilte. Marit smile.PRET ‘Marit smiled.’

(Norwegian)

When temporally relating one event or situation to another, three relative orderings exhaust the potential of possible relations. The first event (A) may be past with respect to the second event (B); as in (2a); A can be simultaneous to B, as in (2b), or A can be subsequent to B, as in (2c). Take event A to be the situation depicted by the predicate, whereas event B is the speech situation, i.e., the “act of utterance”, as described by Lyons in the quote above. These three relations amount to past, present, and future time reference, respectively. (2)

A

B

a. A&B b. B

A

c.

Whereas most Germanic languages employ tense affixes (as in (1)), certain languages, for instance many Creole languages, use freestanding tense particles. These particles encode tense, like the tense affixes in (1); they give rise to specific temporal construals and have scope over the proposition depicting the relevant event, as illustrated by the data in (3) from the English-based creole language Hawaiian Creole English; data from Bickerton (1999: 53–54), originally from Roberts (1993). The particles bin, go, and stay are unbound morphemes and encode past, future, and present, respectively. (3)

a. You bin say go up on roof and paint him. you PAST say go up on roof and paint it ‘You told me to go up on the roof and paint it.’

Tense and Aspect in Germanic Languages

b. Bimeby I go look one tenement house in Nuuanu Street. by-and-by I FUT look a tenement house on Nuuanu Street ‘Soon I will look at a tenement house on Nuuanu Street.’ c. This time he stay coming. this time he PRES coming2 ‘This time he is coming.’ Although all languages have means of expressing all three of the temporal relations in (2), either via lexical words and phrases (for instance adverbs) or via designated grammatical markers (as in (1) and (3)), many languages have only a two-way tense split in their primary, basic tense system. Hence, Comrie (1985a: 48 ff) discusses languages where (a) only a future/ nonfuture distinction is encoded in the grammar – for instance the language Hua in New Guinea; (b) languages where only a past/nonpast distinction is present such as German and Finnish; and (c) “languages that lack tense altogether” (Comrie 1985a: 50), like Burmese and probably also the Australian language Dyirbal.3 This does not exclude the possibility of ternary (or n-ary) tense systems, of course, and all the Germanic languages have specific means (like auxiliaries) to single out for instance future time reference when this is specifically intended, e.g., the English future modal will. But as Comrie observes (op.cit. 48), most European languages (Germanic,4 non-Germanic, and non-Indo-European) have only a two-way split in their basic tense systems, past versus nonpast, “with subdivisions within non-past (especially future as opposed to the present) being at best secondary; thus the socalled present tense in such languages is frequently used for future time reference.” (see also Fertig, Chapter 9). In this perspective, only the paradigm of tense affixes (illustrated in (4) with German examples) constitute the primary tense system; the auxiliaries and the categories they denote belong to a secondary tense system. (4)

2

a. Ich lerne fu¨r mich selbst zu denken. I learn.PRES for myself to think ‘I am learning to think for myself.’ b. Ich lernte dann selbst Linsen zu schleifen. I learned then myself lenses to grind ‘I then learned to grind lenses on my own.’

(German)

(German)

The particle stay or ste is sometimes glossed as an aspectual auxiliary encoding the progressive, either by itself or in combination with the affix –in (wat yu (ste) it(-in); ‘what are you eating?’).

3

Comrie (1985a: 50): Given the existence of binary past/nonpast and future/nonfuture splits, the question might arise whether it is possible to have a present/nonpresent split, with a single nonpresent category covering time reference in the past and future to the exclusion of the present. [. . .] I am not aware of any example of a language which has a clearcut system opposing non-present as a single category to present. [. . .] This would suggest a possible universal of tense systems: in a tense system, the reference of each tense is a continuity.

4

König and van der Auwera (1994: xi) describe “the inflectional contrast between only two tenses” as a “typically Germanic feature”, but Comrie (1985a: 50) ascribes this trait to most European languages.

593

594

KRISTIN MELUM EIDE

In the various Germanic languages, these simple tenses have restrictions, such that the English present tense usually cannot express ongoing actions (instead, the progressive is used), but only habitual situations (I work as a teacher), scheduled or repeated events (The train arrives at noon), or general truths (Two plus two equals four). Moreover, the simple present tense in English usually does not denote the future (see 5a), instead the auxiliary will is enrolled. Note however, that in adverbial clauses headed by temporal connectives like when, before, until, and others, only the simple present is possible and the auxiliary will is not only unnecessary, but ungrammatical (see 5b). As shown in (5c and 5e), both German and Norwegian allow for a simple present to express the future (in main clause declaratives), although the future can also be expressed by enrolling an auxiliary verb, for instance werden ‘will’ and skulle ‘shall’, respectively (cf. also Fertig, Chapter 9). (5)

a. He *gives / will give it to you tomorrow. b. She will see him before he arrives/*will arrive. c. Ich schreibe dir morgens wieder. I write.PRES you.DAT tomorrow again ‘I will write to you again tomorrow.’ d. I werde dir gelegentlich schreiben. I will you.DAT occasionally write.INF ‘I will write to you occasionally.’ e. Jeg skriver til deg i morgen. I write.PRES to you tomorrow ‘I will write to you tomorrow.’ f. Jeg skal skrive til deg senere. I shall write.INF to you later ‘I will write to you later.’

(English) (English) (German)

(German)

(Norwegian)

(Norwegian)

Likewise, in (standard) German the preterit is used in writing and especially in narratives, but in spoken German the present perfect is much more frequently used (except with the copula zu sein and with auxiliaries), and in several varieties the simple past is more or less extinct. A parallel situation can be observed for a range of standard and dialectal variants of closely related Germanic languages; here Harbert (2006: 314). [I]n some G[erman] dialects, particularly in the south (e.g. Swiss G[erman], Keller 1961: 67), as well as southern dialects of D[utch] (de Vooys 1957: 42), Y[iddisch] (Jacobs 2005: 217) and A[frikaans] (Ponelis 1993: 383), the simple past tense has disappeared aside from relics, and the present perfect is the only means remaining of talking about past time.

We will return to the present perfect and other temporally complex constructions in more detail below, but let us note for now that all Germanic languages make use of auxiliaries, i.e., modal auxiliaries and perfective auxiliaries (the latter corresponding to HAVE and BE) to

Tense and Aspect in Germanic Languages

construct complex tenses. We have already mentioned the modals will in English, werden in German, and skulle in Norwegian. However, all Germanic languages have an inventory of modal verbs allowing for an array of complex temporal and modal constructions, and a small set of perfective auxiliaries to form complex tenses. On the surface, these constructions seem rather identical semantically and syntactically from one Germanic language to the next, but as we have already suggested, the felicitous use and semantic-pragmatic restrictions on these complex tenses and modal constructions may vary tremendously from one Germanic language (and dialect) to the next.

25.3 The Reference of Tense Elements As already mentioned, Comrie (1985a: 50) suggests that “in a tense system, the reference of each tense is a continuity.” The “reference” of tense elements – represented in (1) and (4) above by the present affixes -er/-e and the preterit -te, and by the particles bin, go, and stay in (3) – has been a topic of much debate. Hence there exist a wide range of proposals on the semantics and nature of tense elements. Prior (1967) and Montague (1974) suggest that tense elements are operators, taking the proposition (or predicated event) in their scope; others have suggested that tense elements are referential entities, resembling pronouns (Enc¸ 1986, Partee 1973), even adverbial elements (Hornstein 1990). Moreover, a range of proposals have advocated that tense elements are best analyzed as predicates, see for instance Zagona (1995), Stowell (1995, 1996), Julien (2000, 2001), and Eide (2005, 2009a, 2009b). If tense elements are predicates, what constitutes the arguments of these predicates? Reichenbach (1947) and Comrie (1985a), inter alia, suggest that tenses amount to a relation between time points. Reichenbach (1947) is the classical analysis of tense in Germanic languages, and tense is seen here as anchoring the clause to the utterance context (hence being deictic). Reichenbach proposed that three time points were relevant: Time of (predicated) event E, time of speech S, and a “mediating” Reference time R. Reichenbach believed that all tenses expressed in natural languages could be stated as a relative ordering of these three time points. A muchquoted approach which is seen by many as a refinement of Reichenbach’s theory is Klein’s theory on tense and finiteness (1994, 1998, 2006, 2009).5 In contrast to the role ascribed to tense in Reichenbachian proposals, Klein advocates instead the role of finiteness in anchoring. In this theory, finiteness is thus not merely a morphological category, but a host for the crucial elements Assertion and Topic Time. Assertion determines the illocutionary

5

Klein´s theories are also advocated in formalist approaches, e.g., Demirdache and Uribe-Extebarria (2000, 2004).

595

596

KRISTIN MELUM EIDE

force of a sentence, and Topic Time amounts to one of the three time spans distinguished, Time of Utterance (TU), Topic Time (TT), and Time of Situation (TSit). Topic Time refers to the time for which a particular utterance makes an assertion, referring to a time span to which the assertion made is constrained. In Klein’s view, tense expresses a temporal ordering between Time of Utterance and Topic Time, whereas Aspect expresses a temporal ordering relation between Topic Time and Time of Situation. Although Klein clearly builds on the insights from Reichenbach, there are some important differences between Klein’s notion of Topic Time and Reichenbach’s Reference Time R. Crucially, in Reichenbach’s theory, R is a mediator of a viewpoint, a relative time point serving as an anchor for the event time. Klein’s Topic Time, on the other hand, is a time for which the truth of the assertion is evaluated. Reichenbach and Comrie hence see tense as a relation between time points, whereas Klein sees tense as a relation between time spans or time intervals (cf. also Bennett and Partee 1978, Demirdache and UribeExtebarria 2000, 2004). Others still claim instead that tense elements encode a relation between two events or situations (Giorgi and Pianesi 1997; Julien 2001; Eide 2005, 2009a, 2009b). As noted by Julien (2001: 127), when taking tense to express the relation between events, “It follows that the precise extension in time of the event or state becomes irrelevant”; and this allows for a certain underspecification of temporal relations which seems useful in characterizing the category tense in Germanic languages. This chapter will hence adhere to the latter view, assuming that tense is a grammatical category semantically consisting of a dyadic predicate, locating a situation on a temporal axis by means of creating a temporal relation between its two arguments, both being events (or situations), where the speech event may be one of these events. In anchoring the predicated event to the speech event, finite tense is the deictic anchoring category par excellence.

25.4 Aspect In the literature on aspect, the distinction between stative (states) and dynamic (events) situations is considered of fundamental importance to conceptual and linguistic organization. For instance, Michaelis (1998: 16) quotes Langacker (1987: 258), for the claim that the distinction between states and events has “a primal character” because it is linked to a basic cognitive capacity – the ability to perceive change (or the lack of change) over time. [E]vents . . . are situations which (a) have salient boundaries (i.e., points of inception and / or termination) and (b) involve change over time; states . . . are situations which do not involve change over time, and which do not have salient endpoints. I will maintain that the event-state distinction, as outlined here, should form the basis of all explanation in aspectology.

Tense and Aspect in Germanic Languages

In this chapter, we will use the term “event” to encompass both bounded events and unbounded states, hence as an overarching term for both types.6 Likewise, many linguists have found it useful to distinguish between aktionsart and (grammatical) aspect.7 The former is usually taken to denote the aspectually relevant inherent semantic properties of the verb, either in isolation or with its arguments. This is in contrast to aspectual properties acquired by the verb in specific syntactic environments.8 The aspectual properties of a verb may change when the “aktionsart” of the verb interacts with other lexical items in the clause such as adverbials. For instance, when a verb with a dynamic, punctual aktionsart (e.g., flash) combines with a durative adverbial (e.g., until dawn) and as a result produces an iterative reading (The light flashed until dawn), one may refer to this as iterative aspect.9 Aspect in its most restricted sense denotes a grammatical category, the “grammaticalization of the relevant semantic distinctions” (Comrie 1976: 7). When describing the differences between aspect and tense, Comrie (1976: 5) claims that the latter relates the time of a situation to some other time point. The relevant features of aspect however are described as follows: Aspect is not concerned with relating the time of the situation to any other time-point, but rather with the internal temporal constituency of the one situation.

In the traditional literature on Germanic languages both the perfect (John has read the book) and the progressive (John is reading a book) have been considered aspectual categories. In both cases, there is an aspectual auxiliary heading the construction and – in combination with some relevant properties of the verbal complement – presumably providing this construction with specific aspectual and temporal properties. For our present purposes, the distinction between aktionsart and aspect is less important; we will use aspect as an overarching term and describe a certain construction simply as encoding some specific aspect although in some cases the term aktionsart might in fact be more accurate.10 At an informal level, aspect refers to how we conceptualize a situation or event with regard to how it begins, continues, and ends (and possibly repeats itself). Figure 25.1, for instance, illustrates a situation conceptualized as an event with a beginning, duration, and end. The figure implies a timeline from left to right, and correspondingly, the potential 6

In coping with literature where event is construed as a cover term for events in the narrow sense (i.e., nonstates / dynamic situations) and states; synonymous with the eventuality of Bach (1981) or the situation of Barwise and Perry (1983).

7

Cf. Smith (1991), among many others.

8

The use of the term aktionsart in the literature on Slavic languages restricts this term to the properties of

9

The sentence is from Jackendoff (1997: 51), who refers to this type of aspectual shifting as aspectual coercion.

morphological affixes (Comrie 1976: 6, footnote 4). 10

Verb semantics is the subject of a large body of linguistic literature, and this chapter cannot possibly do justice to the attested and conceivable verb types represented in this literature. However, I will mention the seminal work of Vendler (1967) and his four verb types: states, activities, accomplishments, and achievements; and the aspect calculus of Dowty (1979).

597

598

KRISTIN MELUM EIDE

S ¬S

¬S

Figure 25.1 Potential boundaries of an event

S

¬S

¬S

Figure 25.2 Ingressive, dynamic, eventive, inchoative11

boundaries of the event are often referred to as “left boundary” (the beginning) and “right boundary” (the end) of the event. Before the event takes place, it does not yet exist (¬S; not-S). Likewise, when the event has ended, it no longer exists (¬S). At the point where the event begins, the state of affairs changes from ¬S to S. This change-of-state is encoded or at least implied by a range of predicate types. There are for instance lexical verbs such as the Norwegian sovne ‘fall asleep’ and gjenkjenne ‘recognize’; there are also (semi-)lexical verbs, such as the Norwegian inchoative copula bli ‘become’; and the Swedish verb bo¨rja ‘begin’, specialized to encode this aspectual information. Moreover, a range of Germanic languages have functional markers conveying exactly this type of information, for instance classes of verbal prefixes in German, cf. Maylor (2002 :140). The prefixes ent-, er-, ver- are found on many verbs that traditional grammars describe as inchoative, and describe not just a change of state, but rather more a coming into being.12

The aspectual information coercing a zooming in on the beginning of an event is often referred to as inchoative aspect, as mentioned above, or as ingressive aspect. Verbs that encode a change from ¬S to S are frequently also referred to as dynamic or eventive verbs.13 11

Needless to say, these terms are not synonymous. They are used in different types of literature and emphasize different features of aspectual information. What they have in common, though, is their focus on the beginning of an event and the implication of change of state.

12

However, note that the prefix er- sometimes also encodes the right boundary of the event, see for instance Comrie (1985b: 343): “[T]he German simple verb kämpfen ‘to fight’ as in Die Partisanen kämpften für ihre Freiheit ‘The partisans fought for their freedom’, gives no indication of the outcome of the struggle, whereas the telic derivative erkämpfen incorporates the successful outcome of the struggle, as in Die Partisanen erkämpften ihre Freiheit ‘The partisans fought for [and won] their freedom’.”

13

Comrie (1976: 49) characterizes the difference between a stative and a dynamic situation as follows: “With a state, unless something happens to change that state, then the state will continue . . .. With a dynamic situation, on the other hand, the situation will only continue if it is continually subject to a new input of energy . . .. To remain in a state requires no effort, whereas to remain in a dynamic situation does require effort, whether from inside . . . or from outside.” Comrie later divides dynamic situations into events and processes.

Tense and Aspect in Germanic Languages

(6)

a. Die Pflanze erblu¨hte. the plant er-bloomed ‘The plant bloomed.’ b. Kain blev vred og dræbte sin bror. Kain became angry and killed his brother. ‘Cain got mad and killed his brother.’ c. Det bo¨rjade regna smaº sten. it began rain.INF pebbles ‘It started to rain little pebbles.’

(German)14

(Danish)

(Swedish)

There is a large class of predicates that by themselves refer to (or imply) neither the beginning nor the end of the situation they describe; instead, they simply refer to the stative situation, S in Figure 25.3. Accordingly, these predicates are referred to as stative. Lexical predicates encoding this aspect are for instance sove ‘sleep’ (Norwegian), veta ‘know’ (Swedish), være intelligent ‘be intelligent’ (Danish / Norwegian), and bleiben ‘stay’ (German). Semi-lexical constructions are blive ved med at ‘keep v-ing’ (Danish), forsa¨tta at ‘continue’ (Swedish), holde paº med aº ‘be in the process of’ (Norwegian), but also pseudo-coordination structures with posture verbs in several languages, like the Dutch Ik zit te eten ‘I am eating (while sitting)’ (cf. also Fertig, Chapter 9). Both the English progressive be v-ing and the Dutch an het-progressive have this aspectual feature as their core interpretation. Hence this aspect is sometimes referred to simply as progressive (ongoing), but also as imperfective, continuative, durative, atelic or simply stative. These different terms emphasize slightly different features of the stative event; for instance, both imperfective and atelic usually emphasize the lack of right boundary of the event, not the lack of left boundary, and the progressive is sometimes described as “seeing the event from within.” What they all have in common, however, is their focus on the stative, ongoing, and unbounded aspect of the situation. Adverbials denoting time spans, like for a year, are easily compatible with such predicates. (7)

a. Die Pflanze blu¨hte. the plant bloomed ‘The plant bloomed.’

S ¬S

¬S

Figure 25.3 Progressive, imperfective, continuative, durative, stative, atelic

14

From Maylor (2002: 140).

(German)

599

600

KRISTIN MELUM EIDE

(Danish) b. Jeg forblev i landsbyen. I stayed in village.DEF ‘I stayed in the village.’ c. Regnmolnen fortsa¨tter att driva runt o¨ver landet. (Swedish) rain.clouds.DEF continue to drift around over country.DEF ‘The rain clouds keep drifting across the country.’ d. Ik ben aan het werken (Dutch) I am at the working ‘I am working.’ e. Jeg sitter og spiser. (Norwegian) I sit.PRES and eat.PRES ‘I am eating.’ Then there are predicates that encode information about the natural endpoint, or “the right boundary” of an event. For instance, predicates such as eat an apple or paint a house (unlike eat apples or paint houses) imply or encode the natural endpoint of the event they describe: The event will end when the apple is eaten or the house is painted.15 This type of aspect is often referred to as telic (egressive is a less common term), and telic predicates are typically compatible with time-frame adverbials such as in an hour, for instance John reached the top in an hour. A grammaticalized exponent of such right endpoints, for instance as an affix, is typically referred to as the perfective aspect, and in Germanic, the prefix ge- has been ascribed exactly this function; cf. for instance van Gelderen (2004: 167) quoting Mustanoja (1960: 446) on Middle English, Salmons (2012: 199) on Modern German,16 Lockwood (1968: 104) on Middle High German. Semi-lexical predicates are verbs denoting quitting, stopping, or finishing, like Norwegian slutte, German aufho¨ren. (8)

a. Die Pflanze verblu¨hte. the plant withered ‘The plant withered.’

(German)

S ¬S

¬S

Figure 25.4 Egressive, perfective, telic

15

The object or theme of such verbs are often referred to as incremental themes.

16

“The ge- prefix still retains some perfective meaning – that is, serving to mark the completion of something.”

Tense and Aspect in Germanic Languages

forlod hovedstaden. b. Lægen doctor.DEF left capitol ‘The doctor left the capitol.’ c. Flickorna slutade idrotta. girls.DEF stopped sporting ‘The girls stopped doing sports.’

(Danish)

(Swedish)

In some languages, like Finnish, case alternations on the object (for instance partitive versus accusative; Kiparsky 1998) has been shown to induce a distinction between atelic and telic predicates in a manner resembling the alternation between bounded and unbounded themes mentioned above (e.g., paint a house versus paint houses, eat an apple versus eat apples). To some extent, a similar effect of morphological case can be observed in Germanic languages in the use of dative and accusative case in adverbial prepositional phrases; see for instance the German data in (9) from Zwarts (2006), see also Zwarts (Chapter 26, especially the discussion of examples 46 and 51). Here the preposition in allows for an alternation between dative case and accusative, and the dative conveys an atelic reading of the predicate, whereas the accusative induces a telic reading.17 (9)

a. Sie tanzten in dem Zimmer. they danced in the.DAT room ‘They danced in the room.’ b. Sie tanzten in das Zimmer. they danced in the.ACC room ‘They danced into the room.’

(atelic)

(telic)

Some predicates imply both the left boundary and the right boundary of an event, and this particular aspect is sometimes called punctual aspect. Marit datt ‘Mary fell (down)’, Den gamle mannen døde ‘the old man died’ are examples of events that are typically construed as punctual or instantaneous.18 Finally, we will mention iterative (habitual, generic) aspect. These terms denote aspects of repeated or recurring events, and although each event by itself may be interpreted as bounded or even punctual, this repetition of events induces a reading where these punctual events are wrapped in a state.

17

Zwarts notes, however, that although this pattern is upheld for the prepositions alternating between accusative and dative complements, it does not generalize to all types of prepositions, i.e., it is not always the case that the dative induces atelicity and the accusative induces telicity. Compare for instance Sie tanzten aus dem.DAT Zimmer ‘they danced out of the room’, which shows dative, but is still telic, and Sie tanzten durch das.ACC Zimmer ‘they danced through the room’, which is accusative, but still ambiguous with respect to telicity.

18

Intriguingly, a study of strong verbs in Norwegian, specifically of all the strong verbs listed in Nynorsk-ordboka, suggests that a puzzlingly large number of contemporary strong verbs in Norwegian denote punctual events. The author suggests firstly that the strong conjugations of verbs historically have been very productive for verbs of this specific semantic type, and secondly that as more and more strong verbs have leaked into classes of weak verbs and hence express past tense via dental suffixes instead of via Ablaut, these punctual verbs have resisted to a larger degree the migration to weak verb classes. She concludes that language users are able to associate certain semantic features to verbs undergoing Ablaut (Brodahl Nilsen 2012).

601

602

KRISTIN MELUM EIDE

S ¬S

Figure 25.5 Iterative, generic, habitual

Hence these predicates may be compatible in principle with punctual adverbials on a singular, punctual reading, but with time-span adverbials on an iterative reading. Moreover, adverbs like often, usually, typically, always, regularly are very much compatible with the iterative reading; in fact, these adverbs are exactly what brings about the iterative reading in many cases. (10)

a. Kongen spiser middag i denne restauranten. (Norwegian) king.DEF eat.PRES dinner in this restaurant.DEF ‘The king eats/is eating his dinner in this restaurant.’ b. Marit spiller tennis. (Norwegian) Marit play.PRES tennis Marit plays/is playing tennis. c. Jon drikker ikke øl (akkurat naº / vanligvis). (Norwegian) Jon drink.PRES not beer (right now / usually) ‘John is not drinking beer right now/ John does not usually drink beer.’ d. The light flashed (once/until dawn). (English)

25.5 Complex and Periphrastic Tenses: Temporal and Aspectual Features Germanic languages use auxiliaries to construct more complex tenses, and the temporal construal of these potentially long chains of auxiliaries modifying a main verb gave rise to Reichenbach’s (1947) system of nine tenses. As mentioned above, Reichenbach proposed the presence of a mediating reference point R to account for the temporal complexity of attested tenses. Vikner (1985) acknowledged that there are several attested temporal construals that even Reichenbach’s R is not able to account for, and suggested that we need at least one more reference time to account for certain temporal construals of Germanic tenses.19

19

This example is from Vikner (1985: 91), quoted here from Julien (2001: 140).

Tense and Aspect in Germanic Languages

(11)

She promised in November that they would have received her paper by the first day of term.

In this example, there are two reference times, Vikner says, in November and the first day of term, neither of which corresponds to the speech time S or the event time E. Hence, Vikner’s system employs four temporal primitives; the speech time S, the event time E, and two reference times, R1 and R2. Julien (2001) suggests instead that the number of reference times is a function of the number of tense elements that are present in the construction in question. This proposal has a predecessor in Janssen (1988: 99), who claims that “a separate time can in principle be assigned to each verb of a verbal group, whether an auxiliary or a main verb.” He illustrates this with Dutch data showing that in general it is the case that the more verbs and auxiliaries, the more temporal interpretation possibilities exist (1988: 121).20 (12) a. Wim zal morgenochtend

om acht uur vertrokken zijn. (Dutch) Wim will tomorrow.morning at eight o’clock left be ‘Wim will have left tomorrow morning at eight.’

b. Wim zal gistermorgen om acht uur

vertrokken zijn. Wim will yesterday at eight o’clock left be ‘Wim will have left yesterday at eight.’

(Dutch)

The examples illustrate how time adverbials may refer to the time of the perfect auxiliary in (12a) and the time of the main verb in (12b).21 The perfect, in particular, but also the progressive, has a long history in the literature on Germanic grammar as a bit hard to categorize as temporal or aspectual. Here Michaelis (1998: 3), discussing “the debate over whether the perfect is best regarded as tense or aspect”: Analysts have long expressed uncertainty as to the appropriate categorization of the perfect within the tense-aspect system of English. Aspectual marking in general evokes the internal temporal constitution of a situation (Comrie 1976). By contrast, the perfect “tells us nothing directly about the situation in itself, but relates some state to a preceding situation” (Comrie 1976: 52). [. . .] This fact has induced a number of analysts [. . .] to regard the perfect as a relative tense.

Whether you analyze the perfect as an aspect or as a periphrastic tense, it is commonly assumed that the perfect encodes some kind of past-time reference; in the words of Dyvik (1999: footnote 1), the perfect is a “non20

As pointed out by Eide (2005: 119), this assumption is supported by evidence from Portuguese, in that adjunction of adverbs seems to be highly dependent on the overt presence of a lexical head (cf. Costa 1998): The more verbs and auxiliaries overtly occur in a given sentence, the higher the number of adverbs possible in this sentence.

21

Note that the modal can also have two different readings in these examples, a deontic reading and an epistemic / evidential reading, but this does not affect the issue here.

603

604

KRISTIN MELUM EIDE

referential relative past.”22 However, this by no means rules out the possibility that the garden variety Germanic perfect simultaneously encodes a (present) state, and in the literature we often find the claim that it indeed does, cf. ter Meulen (1995: 5–6): The perfect Jane has sighed describes th[e] state caused by the end of [Jane’s] sighing. Such perfect states are atemporal in the sense that once they have begun, they never end . . . The difference between the simple past and the perfect is hence aspectual in nature: The former describes events in a context-dependent way, whereas the latter gives only stative information.

What this proposal suggests is that, unlike most states, the perfect denotes the event whose culmination caused the given state. The “relative past” reading is an inference: If there is a state described as commencing at the exact time of the culmination of the very event that caused the state, and the culmination of this event is an essential part of the description of the state, it follows that this event must already have taken place in order for the state to hold at the time of utterance. Therefore, the “past event” reading is an inevitable inference, according to this proposal. Michaelis (1998: 51), following Herweg (1991a, 1991b), refers to the perfect as a stativizing operator, an operator that maps an event predication true at one time (e.g., Madge swallow-the fly) into a stative predication true at a later time. This stative predication is described as the aftermath of the flyswallowing event, just as in ter Meulen’s proposal. The perfect provides a stative construal of what would otherwise be an event predication. Several proposals hence suggest that the perfect encodes a stative situation. Moreover, it has been argued that the perfect is also some kind of present tense; Jespersen (1931: 47) claims that The perfect . . . is itself a kind of present tense, and serves to connect the present time with the past. This is done in two ways: first, the perfect is a retrospective present, which looks upon the present state as a result of what has happened before in the past; and second, the perfect is an inclusive present, which speaks of a state that is continued from the past into the present time.

The progressive is widely taken to encode a stative aspect23 (Langacker 1987, 1991; ter Meulen 1995: 66). Michaelis (1998: 52) follows Herweg (1991a, 1991b) in referring to the progressive as a stativizing operator, an operator that maps an event into a state. For any given event, there is a state that holds prior to the time at which that event has reached its point of culmination. This state is the progressive state. Thus, the event (interval) 22

Dyvik makes this claim for the Norwegian and English perfect only: “The Norwegian perfect is semantically very close to the English perfect, and less close to the French and German perfects, which can be used to refer to specific past times (Ich habe ihn gestern gesehen ‘I saw him yesterday’). The meaning of Norwegian (and English) perfect is neither deictic past tense nor perfective aspect, but rather nonreferential relative past – the category existentially quantifies over times preceding the time indicated by the tense of the finite verb: Jeg har sett ham ‘I have seen him’ = There exists a time in the past such that I saw him then.”

23

But cf. Comrie (1976: 35) for a different view.

Tense and Aspect in Germanic Languages

“selected” by the progressive is not a state in and of itself (e.g., Madge swallow the fly), but the progressive selecting this event denotes a state (Madge is swallowing a fly), as argued in Michaelis (1998: 256). Likewise, most proposals on the subject suggest that the progressive encodes some kind of present tense; Kreidler (1998: 222), for example, states that [The progressive] is also about the present moment but calls attention to the fact that the activity is in process now.

Julien (2001: 138) remarks that perfects, but not the progressive, are traditionally held to be tenses in descriptions of Germanic perfects. This is reflected in works like Reichenbach (1947) and Vikner (1985), who both include a number of perfect constructions in their tense system while leaving out the progressive. The reason for this, one might guess, is that the temporal contribution of the past participle is more easily recognized than that of the progressive: whereas the progressive, being a present tense, never introduces a time that is not referred to by the auxiliary, the past participle introduces an event time which is different from any other time referred to in the same sentence.

Julien recognizes the progressive as a present tense, specifically a nonfinite periphrastic present tense. Like for instance Stowell (1996) she also ascribes a substantial temporal function to the past participle, and describes the perfect as a nonfinite past tense. Moreover, she assumes that since there are future finite tenses, there ought to be a non-finite future tense as well. Hence she proposes that (2001: 132) [T]he progressive, the perfect, and the prospective are in fact tenses and not aspects. More precisely, the progressive is a nonfinite present, the perfect is a nonfinite past, and the prospective is a nonfinite future.

Of course, these tenses can perform in finite functions when they are headed by a finite auxiliary, but the perfect, the progressive and the prospective themselves are nonfinite, according to Julien. Klein (1995) proposes that the difference between what we call tenses and what we call aspects should be analyzed as minimal, and Demirdache and Uribe-Extebarria (2000) pursue this idea of parallelism between tenses and aspects (2000: 161).24 We reduce Tense and Aspect to the same set of semantic primitives by assuming that since Tense has been analysed as relating two times, Aspect should be analysed as relating two times. This is precisely the proposal made by Klein (1995).

The difference between tenses and aspect, on this view, is that aspects establish a relation between a reference time (Reichenbach’s R; Klein’s assertion time) and the predicate event time, whereas tenses relate a reference time to the utterance time, i.e. the time of the speech event. 24

Cf. also Comrie (1976), and Guéron and Hoekstra (1995) for closely related ideas.

605

606

KRISTIN MELUM EIDE

25.6 Finite and Nonfinite Tenses Eide’s (2005, 2009a, 2009b, 2016) approach to finiteness emerges from Comrie’s (1985a: 36) observation that natural languages display examples of absolute and relative tenses. Absolute tenses (e.g., the preterit) typically take the moment of speech as their deictic centre and relative ones (e.g., participles) take some other contextually given time point as their point of reference. While in many languages specific verb forms encode the distinction between absolute and relative tenses, there are also languages, e.g., creoles, where the same verb form is used for both. This lack of formal distinctions between absolute and relative tense forms has traditionally been considered a lack of formal finiteness distinction in the literature on Creoles (cf. Romaine 1993; Muysken & Law 2001). In most Germanic languages, however, the finiteness distinction is morphologically encoded. Eide (2005, 2009a, 2009b, 2016) argues that even in Germanic languages the distinction between finite and nonfinite verb forms corresponds to the distinction between absolute and relative tense forms. As Comrie (1985a: 48) observes, and as discussed above, most European languages have only a two-way split in their tense systems, past versus nonpast, “with subdivisions within non-past (especially future as opposed to the present) being at best secondary; thus the so-called present tense in such languages is frequently used for future time reference.” Pairing this assumption with the claim that the absolute-relative distinction equals the finiteness distinction in the relevant languages, we arrive at the paradigm in (13) for the garden variety of Germanic language. Here each and every verb form encodes a tense element consisting of two pieces of information: [±Past] and [±Finite]. +Finite

−Finite

+Past

Preterit

Participle

−Past

Present

Infinitive

(13)

In this analysis, being tensed is an essential part of being a verb,25 and this ties in nicely with the proposal of Janssen (1988), where each verb and auxiliary in a verbal chain provides some temporal contribution to the construction. However, it differs from Julien’s approach, where some nonfinite verb forms are ascribed tense features (e.g., the past participle) and some are not (e.g., the infinitive), and some auxiliaries are ascribed tense properties (will) and others not (infinitival have).26 In Eide’s approach all verb 25

Present participles are seemingly adjectival, at least in Norwegian, see Faarlund et al. (1997: 119). Imperatives are likewise often described as lacking tense features. However, they are often believed to encode a type of mood (e.g., jussive or deontic mood). It might be that the description of verbness requires either tense or mood for a lexical root to be encoded as a verb.

26

In Julien’s (2001: 136) system, the aspectual have is a proxy element. It does not have any semantic content (except for formal features, p. 137); it is simply a realization of a V, allowing for the projection of the tense head hosting will on top of it.

Tense and Aspect in Germanic Languages

forms, main verbs and auxiliaries alike, are ascribed tense features, and they all contribute temporal points to the construction. The difference between finite and nonfinite tenses is that the latter are unable to relate to the moment of utterance (or the speech event), whereas the former typically do relate to it. Hence, nonfinite tenses will usually be embedded under some other, finite verbal element, whereas finite tenses occur as the highest tense element in a sentence functioning as a complete utterance. Unlike Creole languages most of the Germanic languages productively encode finiteness with all types of verbs; any verb form in for instance Norwegian is inherently specified with respect to finiteness.27 A [+Finite] form like spiser ‘eat.PRES’ or spiste ‘eat.PRET’ has distinct morphology showing that it is a finite (i.e., typically absolute) tense form; no context needs to be provided to determine this. Likewise, spise ‘eat.INF’ and spist ‘eat.PART’ are unambiguously nonfinite, regardless of context. In contrast, an out-of-context main verb in English, e.g., love or loved, could be either a finite (absolute) or a nonfinite (relative) tense form. Loved may be either a preterit or a participle, love either a present or an infinitive (in fact, love might even be a subjunctive or an imperative). This could be an accidental fact of formal homonomy and need not point to anything profound separating English morphology from Norwegian morphology. However, it is certainly possible to compare paradigms of basic verb forms and expect the formal syncretism, e.g., the systematic correspondence of two forms, preterit and participle to one, generalized [+past] form; to involve the across-the-board deletion of one specific morphosyntactic feature. To see the consequences of this particular syncretism, we need to acknowledge that the infinitive and the past participle are genuine, fullfledged participants in the basic tense system, giving rise to a four-way tense paradigm with two finite and two nonfinite forms; see the table in (13), repeated here as (14a). Over time, English (weak) main verbs developed into a system employing the collapsed paradigm in (14b); here the finiteness distinction, the distinction between absolute and relative tense forms, is lost.

(14)

a +Past -Past

+Finite Preterit Present

-Finite Participle Infinitive

b +Past -Past

Preterit/Participle Present/Infinitive

This loss results in a paradigm employing one generalized [+Past] form and one generalized [−Past] form (“the bare form”). Old English, like present day Norwegian, had distinct forms for all four cells in this paradigm for both strong verbs like singan and weak verbs like lufian, see paradigm (15a). However, English is changing and becoming more like the Creole languages mentioned earlier. The finiteness distinction starts to disappear in weak verbs, adhering to the collapsed paradigm in (15b). In 27

Here I am glossing over the fact that in the clearly most productive verb class (the kasta-class) the preterit and the past participle are formally identical. This is also the case for Frisian, see Fertig (Chapter 9).

607

608

KRISTIN MELUM EIDE

(15b), one generalized [+Past] form covers the function of the preterit and the past participle, and one [−Past] form (“the bare form”) covers the present and the infinitive.28

(15)

a +Past Old English Norwegian -Past Old English Norwegian

+Finite Preterit: sang, lufode sang, likte Present: singe, lufie synger, liker

-Finite Participle: gesungen, lufod sunget, likt Infinitive: singan, lufian synge, like

b Weak verbs +Past Modern English -Past Modern English

c Standard strong verbs +Past Modern English -Past Modern English

+Finite

-Finite

Preterit: went,saw, drove

Participle: gone,seen, driven Infinitive: go, see, drive

d Dialectal strong verbs +Past Non-standard English -Past Non-standard English

Present: go, see, drive

Preterit/Participle: liked, killed, smiled Present/Infinitive like, kill, smile

Preterit/Participle gone, seen, drove30 Present/Infinitive go, see, drive

While English regular weak verbs have been inflected according to the collapsed paradigm in (15b) for centuries, irregular strong verbs have maintained a slightly more complex system. Although there is only one generalized [−Past] form, there are distinct forms for the preterit and the past participle; instead of a four-way system, there is a three-way system in Standard English, see (15c). Currently, these remains of the finiteness distinction in main verbs, hitherto maintained in strong irregular verbs, are disappearing from many variants and dialects of English, resulting in alignment with the rest of the system (15d). English modals, on the other hand, only come in finite forms (in the standard varieties) and hence occur in only finite surroundings,29 and the only verbal elements occurring in both finite and nonfinite contexts are the auxiliaries have and be (see Eide 2016 for details).

25.7 The Present Perfect Puzzle: Remote and Immediate Past The “present perfect puzzle” (Klein 1992) is usually characterized as the observation that a punctual past-denoting adverbial such as yesterday cannot co-occur with the present perfect. However, “current cycle” adverbials, i.e., adverbials denoting the present daily, weekly, monthly, or yearly cycle

28

This resembles the situation in Afrikaans, where there is no distinction between the infinitive and the present form, and the preterit has disappeared except for in a handful of verbs. There is hence one generalized [+past] form (for instance Ek het gewerk ‘I worked’) and one generalized [−past] form (Ek werk ‘I work’, om te werk ‘to work’). The [+past] form consists in a finite and a nonfinite part, although the auxiliary het is even ambiguous between the present and the past interpretation (‘I have worked, I had worked’).

29

Note that they also cannot occur in subjunctives, that require verbal heads without finiteness markings.

Tense and Aspect in Germanic Languages

are always licit with the present perfect in Norwegian. Adverbials setting up a cycle, such as hver gang ‘every time’, are also perfectly fine.30 (16) a. *Jeg har sett ham i gaº r. I have seen him yesterday ‘I have seen him yesterday.’ b. Hver gang jeg ringer min mor,

har hun sovet daº rlig natten før. Every time I call my mother, has she slept poorly night.DEF before ‘Every time I call my mother, she has slept poorly the night before.’

Thus, it is possible to describe the present perfect as a tense construction signifying “there currently exists a state consisting of the aftermath of a past event”, where “currently” can be paraphrased ”within some (salient) current cycle.” In Norwegian, the present perfect is hence the proper “past tense for the current cycle”, and it seems legitimate to characterize it as an immediate past. Any event taking place in the previous cycle must be signaled by the preterit, which acts as the remote past. As observed by Bybee et al. (1994: 100), many languages distinguish between immediate and remote pasts, and Norwegian seems to be such a language. This is the reason for the “present relevance” reading of the present perfect – it denotes a state consisting of the aftermath of a past event that took place within the current cycle. One important difference between the uses of the German and the Norwegian present perfect is that Norwegian makes the distinction between immediate and remote past, whereas German does not. The German present perfect can be used to describe a situation in the previous cycle, unlike its Norwegian counterpart. It is thus not necessary to assume that the semantics of the German present perfect is very different from the present perfect in Norwegian. It is more a question of usage; the German present perfect has a wider domain of functions as it is allowed to signify previous cycle, unlike Norwegian present perfect. The distinction between remote past (preterit) and immediate past (present perfect) disappears when the present perfect is embedded under a modal, as in (17a). In this context (unlike in the present perfect, cf. (17b)), the participle may be modified by an adverbial denoting previous cycle. (17)

30

a. Jon maº ha kommet i gaº r. Jon must have arrived yesterday ‘Jon must have arrived yesterday.’

(Norwegian)

b. *Jon har kommet i gaº r. Jon has arrived yesterday ‘Jon has arrived yesterday.’

(Norwegian)

See also the observations made in Comrie (1985a: 78–79) and Giorgi and Pianesi (1997: 111) that habitual adverbials are possible with the present perfect.

609

610

KRISTIN MELUM EIDE

In cases like (17a), the domain of the immediate past, the present perfect, is extended to cover the remote past, the preterit, which semantically would be the correct past in this construction. In such cases, there are conflicting requirements from the syntax and the semantics. The semantic component requires the remote past as the adverbial i gaº r denotes previous cycle. However, there is no way to embed a remote past under a modal, due to the modal’s strict syntactic selectional requirements, which specify that it accepts only a complement with the matrix [−PAST, −FIN]. Hence, as observed by Hofmann (1976: 94), the preterit must be replaced by the nonfinite perfect under a modal, and is “of course, the only way to represent past time [under a] modal.” The remote past is replaced by the immediate past, fulfilling the syntactic requirements: The auxiliary ha contributes the right feature specification and the past participle provides the past tense element. Hence B is the answer to the question in (18b) as well as the one in (18a). (18)

a. A:

Tror du Marit har drept ham? believe you Marit has killed him ‘Do you believe that Marit has killed him?’

(Norwegian)

b. A:

Tror du Marit drepte ham? believe you Marit killed him ‘Do you believe that Marit killed him?’

(Norwegian)

Hun maº ha drept ham! she must have killed him ‘She must have killed him.’

(Norwegian)

B:

The nonfinite present shows up not only under modals, but in every case where the semantic component requires the preterit and the selecting element requires a nonfinite complement. The clause in (19b) is the nonfinite version of the one in (19a). It employs the nonfinite perfect to express the past expressed by the preterit in the corresponding finite clause. The infinitival marker aº requires a complement with a matrix [−PAST, −FIN], just like a modal; the way to meet this requirement and preserve the past reading is to employ the immediate past, i.e., the (nonfinite) present perfect, as a substitute for the remote past, the preterit. (19)

a. Hun hevder at hun ankom i gaº r. she claims that she arrived yesterday ‘She claims that she arrived yesterday.’

(Norwegian)

b. Hun hevder aº ha ankommet i gaº r. she claims to have arrived yesterday ‘She claims to have arrived yesterday.’

(Norwegian)

Tense and Aspect in Germanic Languages

25.8 Conclusion The tense and aspect systems of Germanic languages are not extremely complex when compared to certain other language groups, but are constantly subject to fresh scrutiny from new generations of linguists and philosophers nevertheless. The underlying meaning of tense and aspect markers seemingly never cease to attract interest and attention, and felicitous usage of particular tenses and aspects in context always poses a challenge to any learner of a language, even the Germanic ones. The choice between the simple tenses and the complex, periphrastic tenses in particular constructions is governed by different sets of rules from one Germanic language to the next, and the ways of expressing particular aspectual features vary from one language to another. An inventory of aspectual prefixes and the availability of specific Case marking give, e.g., German a wider range of fine-grained formal markers in the aspectual domain, as compared to for instance Mainland Scandinavian languages, where many aspectual features are left to be specified via lexical or semilexical means. Likewise, the morphological expression of finiteness is not evenly distributed across the Germanic languages; English and Afrikaans, for instance, employ productive verbal paradigms with a lot fewer morphological distinctions than German and Icelandic.

References Bach, E. 1981. “On time, tense, and aspect: An essay in English metaphysics.” In P. Cole (ed.), Radical Pragmatics. New York: Academic Press: 63–81. Barwise, J. and J. Perry 1983. Situations and attitudes. Cambridge, MA: MIT Press. Bennett, M. and B. Partee 1978. Toward the Logic of Tense and Aspect in English. Bloomington, IN: Indiana University Linguistics Club. Bickerton, D. 1999. “How to acquire language without positive evidence: What acquisitionists can learn from creoles.” In M. DeGraff (ed.), Language Creation and Language Change – Creolization, Diachrony and Development. Cambridge, MA: MIT Press. Brodahl Nilsen, M. 2012. Sterke verb og semantiske fellestrekk. Om semantisk motivasjon for bøyingsklassetilhørighet ved norske verb. [Strong verbs and common semantic features. On the semantic motivation for conjugational classes of Norwegian verbs.] M.A. thesis, Dept. of Linguistic and Scandinavian Studies, University of Oslo. Bybee, J., R. Perkins, and W. Pagliuca 1994. The Evolution of Grammar – Tense, Aspect, and Modality in the Languages of the World. Chicago: The University of Chicago Press. Comrie, B. 1976. Aspect – An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge Textbooks in Linguistics. Cambridge University Press.

611

612

KRISTIN MELUM EIDE

Comrie, B. 1985a. Tense. Cambridge Textbooks in Linguistics. Cambridge University Press. Comrie, B. 1985b. “Causative verb formation and other verb-deriving morphology.” In T. Shopen (ed.), Language Typology and Syntactic Description, Vol. 3. Grammatical Categories and the Lexicon. Cambridge University Press: 309–348. Costa, J. 1998. Word Order Variation. A Constraint-Based Approach. Doctoral dissertation, Leiden University. Distributed by: Holland Academic Graphics (HIL), The Hague. Demirdache, H. and M. Uribe-Extebarria 2000. “The primitives of temporal relations.” In R. Martin, D. Michael, and J. Uriagereka (eds.), Step by Step. Cambridge, MA: MIT Press: 157–186. Demirdache, H. and M. Uribe-Etxebarria 2004. “The syntax of time adverbs.” In J. Gue´ron and J. Lecarme (eds.), The Syntax of Time: Current Studies in Linguistics 37. Cambridge, MA: MIT Press: 143–179. De Vooys, C. G. N. [1957]. 1967. Nederlandse spraakkunst, 7th edn., rev. by M. Scho¨nfeld. Groningen: Wolters. Dowty, D. R. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel. Dyvik, H. 1999. “The universality of f-stucture: discovery or stipulation? The case of modals.” Proceedings of the LFG99 Conference, CSLI Publications. Eide, K. M. 2005. Norwegian Modals. Studies in Generative Grammar 74. Berlin: Mouton de Gruyter. Eide, K. M. 2009a. “Tense, finiteness and the survive principle: Temporal chains in a crash-proof grammar.” In M. T. Putnam (ed.), Towards a Derivational Syntax: Survive-Minimalism. Amsterdam: John Benjamins: 91–132. Eide, K. M. 2009b. “Finiteness: The haves and the have-nots.” In A. Alexiadou, J. T. Hankamer, J. McFadden, J. Nuger, and F. Scha¨fer (eds.), Advances in Comparative Germanic Syntax. Amsterdam: John Benjamins: 357–390. Eide, K. M. 2016. “Finiteness, inflection, and the syntax your morphology can afford.” In K. M. Eide (ed.), Finiteness Matters: On Finiteness-Related Phenomena in Natural Languages. Amsterdam: John Benjamins: 121–168. Enc¸ M. 1986. “Towards a referential analysis of tense expressions,” Linguistics and Philosophy 9: 405–426. Faarlund, J. T., S. Lie, and K. I. Vannebo 1997. Norsk referansegrammatikk. Oslo: Universitetsforlaget. Fintel, K. von 2006. “Modality and Language.” In D. M. Borchert (ed.), Encyclopedia of Philosophy, 2nd edn. http://mit.edu/fintel/www/modality.pdf. Gelderen, E. van 2004. Grammaticalization as Economy. Amsterdam: John Benjamins. Giorgi, A. and F. Pianesi 1997. Tense and Aspect: From Semantics to Morphosyntax. Oxford University Press. Gue´ron, J. and T. Hoekstra 1995. “The temporal interpretation of predication.” In A. Cardinaletti and M. T. Guasti (eds.), Small Clauses: Syntax and Semantics, Vol. 28. New York: Academic Press: 77–107. Harbert, W. 2006. The Germanic Languages. Cambridge University Press.

Tense and Aspect in Germanic Languages

Herweg, M. 1991a. “A critical account of two classical approaches to aspect,” Journal of Semantics 8: 362–403. Herweg, M. 1991b. “Perfective and imperfective aspect and the theory of events and states,” Linguistics 29: 969–1010. Hofmann, T. R. 1976. “Past tense replacement and the modal system.” In J. McCawley (ed.), Syntax and Semantics, Vol. 7. New York: Academic Press: 86–100. Hornstein, N. 1990. As Time Goes By: Tense and Universal Grammar. Cambridge, MA: MIT Press. Jackendoff, R. 1997. The Architecture of the Language Faculty. Cambridge, MA: MIT Press. Jacobs, N. G. 2005. Yiddish: A Linguistic Introduction. Cambridge University Press. Janssen, T. A. J. M. 1988. “Tense and temporal composition in Dutch: Reichenbach’s ‘point of reference’ reconsidered.” In V. Ehrich and H. Vater (eds.), Temporalsemantik. Beitra¨ge zu¨r Linguistik der Zeitreferenz. Tu¨bingen: Max Niemeyer Verlag: 96–128. Jespersen, O. 1931. A Modern English Grammar on Historical Principles. Book IV. Syntax. London: George Allen and Unwin. Julien, M. 2000. Syntactic Heads and Word Formation – A Study of Verbal Inflection. Doctoral dissertation, University of Tromsø. Published as Julien 2002 by Oxford University Press. Julien, M. 2001. “The syntax of complex tenses,” The Linguistic Review 18:125–167. Keller, R. E. 1961. German Dialects: Phonology and Morphology. Manchester University Press. Kiparsky, P. 1998. “Partitive case and aspect.” In W. Geuder and M. Butt (eds.), The Projection of Arguments: Lexical and Compositional Factors. Stanford, CA: CSLI: 265–307. Langacker, R. W. 1987. Foundations of Cognitive Grammar, Vol.1. Stanford, CA: Stanford University Press. Langacker, R. W. 1991. Foundations of Cognitive Grammar, Vol. 2. Descriptive Application. Stanford, CA: CSLI Publications. Klein, W. 1992. “The Present Perfect puzzle.” Language 68: 525–552. Klein, W. 1994. Time in Language. London: Routledge. Klein, W. 1995. “A time-relational analysis of Russian Aspect,” Language 71: 669–695. Klein, W. 1998. “Assertion and Finiteness.” In N. Dittmar and Z. Penner (eds.), Issues in the Theory of Language Acquisition: Essays in Honor of Ju¨rgen Weissenborn. Bern: Peter Lang Verlag: 225–245. Klein, W. 2006. “On finiteness.” In V. van Geenhoven (ed), Semantics in Acquisition. Heidelberg: Springer: 245–272. Ko¨nig, E. and J. van der Auwera 1994. The Germanic Languages. London: Routledge. Kreidler, C. W. 1998. Introducing English Semantics. London: Routledge.

613

614

KRISTIN MELUM EIDE

Lockwood, W. 1968. Historical German Syntax. Oxford: Clarendon. Lyons, J. 1977. Semantics. Cambridge University Press. Maylor, B. R. 2002. Lexical Template Morphology: Change of State and the Verbal Prefixes in German. Amsterdam: John Benjamins. Meulen, A. G. B. ter 1995. Representing Time in Natural Language: The Dynamic Interpretation of Tense and Aspect. Cambridge, MA: MIT Press. Michaelis, L. A. 1998. Aspectual Grammar and Past-Time Reference. London: Routledge. Montague, R. 1974. Formal Philosophy. New York: Yale University Press. Mustanoja, T. 1960. A Middle English Syntax. Helsinki: Societe Neophilogique. Muysken, P. and P. Law 2001. “Creole Studies: A theoretical linguist’s field guide,” GLOT International 5.2: 47–57. Partee, B. 1973. “Some structural analogies between tenses and pronouns in English,” The Journal of Philosophy 70: 601–609. Ponelis, F. A. 1992. The Development of Afrikaans. Bern: Peter Lang Verlag. Prior, A. 1967. Past, Present and Future. Oxford University Press. Reichenbach, H. 1947. Elements of Symbolic Logic. New York: Macmillan. Roberts, J. 1993. “The transformation of Hawaiian plantation pidgin and the emergence of Hawaii Creole English.” Paper presented at the conference of Society for Pidgin and Creole Linguistics, Amsterdam, June 10–12. Romaine, S. 1993. Pidgin and Creole Languages. London: Longman. Salmons, J. 2012. A History of German: What the Past Reveals about Today’s Language. Oxford University Press. Smith, C. S. 1991. The Parameter of Aspect. Dordrecht: Kluwer. Stowell, T. 1995. “What is the meaning of the present and past tenses?” In P. M. Bertinetto, V. Bianchi, J. Higginbotham, and M. Squartini (eds.), Temporal Reference, Aspect and Actionality, Vol. 1. Semantic and Syntactic Perspectives. Turin: Rosenberg and Sellier: 381–396. Stowell, T. 1996. “The phrase structure of tense.” In J. Rooryck and L. Zaring (eds.), Phrase Structure and the Lexicon. Studies in Natural language & Linguistic Theory, Vol. 33. Dordrecht: Kluwer: 277–291. Vendler, Z.1967. Linguistics in Philosophy. Ithaca, NY: Cornell University Press. Vikner, S. 1985. “Reichenbach revisited: One, two or three temporal relations,” Acta Linguistica Hafniensia 19: 81–98. Zagona, K. 1995. “Temporal argument structure: Nonfigurational elements of construal.” In P. M. Bertinetto, V. Bianchi, J. Higginbotham, and M. Squartini (eds.), Temporal Reference, Aspect and Actionality. Vol. 1. Semantic and Syntactic Perspectives. Turin: Rosenberg and Sellier, 397–410. Zwarts, J. 2006. “Case Marking Direction: The Accusative in German PPs,” Proceedings from the Annual Meeting of the Chicago Linguistic Society 42.2: 93–107.

Chapter 26 Prepositions and Particles Place and Path in English, German, and Dutch Joost Zwarts

26.1 Introduction All Germanic languages have a set of nondeclinable words or expressions with a spatial meaning, like the English elements in (1). A fairly exhaustive list is presented in (1), to give an impression of the size and variety of this class. (1)

abaft, aboard, about, above, abreast, abroad, across, adjacent, adrift, aft, after, against, aground, ahead, aloft, along, alongside, amid(st), among(st), apart, around, ashore, aside, astride, at, atop, away, back, backward(s), before, behind, below, beneath, beside, between, betwixt, beyond, by, chez, close, down, downhill, downstage, downstairs, downstream, downward(s), downwind, east, eastward(s), far away, forth, forward(s), from, heavenward(s), hence, here, hither, home, homeward(s), in, in back, indoors, in front, inside, into, inward(s), left, leftward(s), near(er/est), nearby, north, northward(s), off, on, onward(s), onto, on top, opposite, out, outdoors, outside, outward(s), over, overboard, overhead, overland, overseas, past, right, rightward(s), round, roundabout, seaward(s), sideways, skyward(s), south, southward(s), thence, there, thither, through, throughout, to, together, toward(s), under, underfoot, underground, underneath, up, uphill, upon, upstage, upstairs, upstream, upward(s), upwind, via, west, westward(s), whence, where, whither, with, within, without, yonder

When combining with a noun phrase (from and inside in (2a)) these items are traditionally classified as prepositions and as adverbs when they do not combine with a complement (in, back, inside, upstairs in (2b)–(2d)). Within the latter class, those items that can combine more tightly with a verb are known as particles (in and back in (2b)), as demonstrated by their ability to

616

JOOST ZWARTS

occur before a (nonheavy, nonpronominal) direct object [(2c), as opposed to inside and upstairs, (2e)] (adapted from McIntyre 2015: 235–236): (2)

a. b. c. d. e.

We pushed the box from the shelf/inside the room. We pushed the box in/back. We pushed in/back the box. We pushed the box inside/upstairs. *We pushed inside/upstairs the box.

Some elements in (1) have both prepositional and adverbial uses (like inside), while others are exclusively prepositional (like from) or exclusively adverbial (like back and upstairs). I will call the items in (1) Ps, following a tradition that sees spatial prepositions, adverbs, and particles as instances of the same syntactic category P: Prepositions are transitive Ps and adverbs (including particles) are intransitive Ps. The label P can also be taken to apply to prefixal (bypass, underbelly) and maybe even bound forms (enfold, defriend). Here, the label P also descriptively covers fixed multiword elements like in front and on top, ignoring the syntactic structure that these might have. The primary function of Ps is to help communicate where something is (place) or where it is going (path). The goal of this chapter is to describe (with a minimum of theory, especially on the syntactic side) how Ps do this, focusing on standard, present-day English, German, and Dutch, the three languages that probably have been most intensively studied in this respect. After a section with basic semantic preliminaries (Section 26.2), Section 26.3 is devoted to Ps for place and Section 26.4 deals with Ps for path. It is impossible to do justice to the rich literature on the syntax and semantics of Ps in this chapter see Zwarts (2017) for a review of recent semantic literature.*

26.2 Figures and Grounds, Places and Paths, Functions and Arguments Describing where something is or is going usually involves an asymmetric spatial relation. One entity serves as the ground (landmark, reference object) with respect to which another entity, the figure (trajector, located object) is located or moving (Talmy 1975). If expressed, the ground corresponds to the object (internal argument) of the P and the figure to an argument external to the whole PP (Svenonius 2007): (3)

a. The birdfigure was [PP in the cageground] / was [PP inside]. b. The birdfigure flew [PP into the cageground] / flew [PP in].

* I gratefully acknowledge, with the usual disclaimers, Anja Goldschmidt for her help with the German data and Michael T. Putnam for his helpful comments.

Prepositions and Particles

The figure can be realized as the subject of the sentence, as in (3), but there are other explicit or implicit elements in a sentence outside the PP that can be the figure of a spatial relation. I use the term “PP” (“prepositional phrase”) descriptively here, ignoring the possibility that particles like in in (3b) might not be phrases in a theoretical sense. Spatial relations come in two types. Example (3a) involves a place, the interior of the cage, where the figure was located (4a). In (3b) the figure followed a path that ends in the interior of the cage (4b). It will be convenient to refer to the PPs in (3a) as PlacePPs and to those in (3b) as PathPPs, and to the Ps that head them as PlacePs and PathPs, respectively. (4)

in the cage (a), into the cage (b) x a

b

If we represent the interior of the cage as I N (T H E - C A G E ), with I N standing for the place function that maps an object to its interior place, then the bird can be related to this place with a general location relation L O C , as shown in (5a) (e.g., Jackendoff 1983 and many others). (5)

a. b.

LOC(THE-BIRD,IN(THE-CAGE)) LOC(THE-BIRD,TO(IN(THE-CAGE)))

The path relation can be decomposed as in (5b). The part T O ( I N ( T H E - C A G E )) makes explicit that the “into” path is defined on the basis of the “in” place: T O maps the interior of the cage to the set of paths that have their end point in its interior. In other words, at the end of the path of (4b), we are at the place of (4a). The T O + I N decomposition of into in (5b) exemplifies a semantic layering that is typical for many path relations: A path function (here T O ) applies to the result of a more basic place function (here O N ). In (5b), the predicate L O C represents that the bird is “located” at the path, which in this case boils down to the bird moving along the path (obviously, a simplification of the way paths function in sentential semantics). There are contexts in which the figure of the spatial relation is the event e described by the sentence, which is definitely needed for (6a), and maybe also for (3b) (instead of analysis (5b)), representing that the motion event e stretches out along the path: (6)

a. The bird sang in the cage / inside. ∃e [ S I N G ( T H E - B I R D ,e) & L O C ( e, I N ( T H E - C A G E ) ) ] b. The bird flew into the cage / flew in. ∃e [ F L E W ( T H E - B I R D ,e) & L O C ( e, T O ( I N ( T H E - C A G E ) ) ) ]

In both (6a) and (6b), the bird is then indirectly related to the (interior of the) cage, through its involvement in an event e that has that relation more directly.

617

618

JOOST ZWARTS

I N and T O are both unary functions, taking one argument. There are no spatial functions with more than one ground argument. Between requires minimally two reference objects, but these are always part of a conjunction (between Brussels and Paris) or plural (between the two cities). From . . . to . . . might behave as a discontinuous P requiring two noun phrases, but its semantics is a combination of two unary functions. There might be “nullary” path functions, corresponding to adverbial Ps like backward and up, which require no ground argument (but see Section 26.4.1). It is not obvious that there are also nullary place functions. The only expressions that can refer to places in an absolute way, without the help of a ground, are nouns like place and the pronominal form where (in somewhere, for instance). In addition to this basic type of argument structure functionality, Ps may also depend on the spatial point of view of the speaker (or another contextual participant), most directly in forms like here and there, but also in abroad, home, upstairs and in PPs like beyond the ocean. Note that quite a few Ps in (1) might have a semantic structure that involves the application of a more general spatial function to a more specific ground, as partially and opaquely reflected in the incorporation of a noun in a (possibly affixal or empty) P, like underground, outdoors, downhill, ashore, heavenward, home. However, such structures do not always correspond to a saturated spatial function, as illustrated by the prepositional use of such forms in aboard the ship, beside his son, inside the box, and on top of it. In this chapter, I will henceforth mostly ignore the Ps that involve a very specific nonspatial component (like “shore”) and focus on the Ps that help to express more general spatial meanings. Given (5) and (6), I assume that the place or path meaning of PPs is given by representations like I N (T H E C A G E ) and T O ( I N (T H E - C A G E )), respectively, and I take it for granted that these will be related to a figure as part of the compositional sentential semantics. This allows me to concentrate on the question of how English, German, and Dutch express their place and path meanings through Ps, addressing the following two clusters of questions. Transitivity: Given that the class of Ps involves both transitive (prepositional) and intransitive (adverbial, particle) elements, as well as elements doing double duty, the question is how this grammatical distinction relates to the semantics of place and path functions. Under what conditions does a P allow for an implicit ground argument, like the cage of inside in (3a) and in in (3b)? Is intransitivity always a matter of an implicit ground or are there intransitive Ps that denote “‘ground-less”, zero-place spatial functions? In other words, what is the role of transitivity and intransitivity in the P-domain? Complexity: How are spatial distinctions and relations reflected in the system of Ps, through lexical and grammatical means? This question does not only concern the way the three languages relate places to paths (like English in and into), but also the distinctions among different types of place functions (like in versus inside) and path functions (like into versus from). In general, the question will be what the complexity and

Prepositions and Particles

combinatorics of Ps can reveal about the underlying spatial system of places and paths. These two questions will be addressed separately for places and paths. Since paths are often defined on the basis of places, we start with the latter.

26.3 How Ps Express Places 26.3.1 More About Places In order to explore how English, Dutch, and German express place functions, it is useful to have some sort of shared semantic “grid.” There is a rich set of geometric and nongeometric features that define place functions and they may pattern in different ways across languages. In this section, I will work with a set of about 20 place functions, labeled with unanalyzed markers, like A T and B E H I N D , standing for complex bundles of features, as customary in the semantics of Ps. The only distinction that I will make at this point is between topological and projective functions (Herskovits 1986). Topological relations, like contact, containment, and their negations (A T , O N , I N , O U T S I D E ), are geometrically the most basic spatial relations between objects, definable (in some formal geometric frameworks) on the basis of a primitive notion of connection (Randell et al. 1992), that is insensitive to distance and direction. A T represents a loose and general notion of spatial connection, without a commitment to either inclusion (I N ) or contact (O N ). O U T S I D E is the complement of I N . Much research (see for instance Garrod and Coventry 2004 and Carlson and van der Zee 2005) has demonstrated the influence of nongeometric (functional and force-dynamic) factors like containment and support in the application of I N and O N , respectively. This is one thing that should be kept in mind when we are using such capitalized markers to stand for the bundles of features that define particular place functions. Projective relations involve a direction (axis, reference frame) defined by properties of the environment (absolute frame, (7a)), the ground object (intrinsic frame, (7b)), or by an observer (relative frame, (7c)). These reference frames have received much attention in the literature (e.g., Levinson 1996). Apart from the place relations O V E R , F R O N T , and L E F T in (7) and their inverses U N D E R , B E H I N D , and R I G H T we also find B E S I D E here, the cardinal directions (N O R T H , etc.), and O P P O S I T E . (7)

a. The bird is above the cage. (as determined by gravity) b. The bird is in front of the cage. (where its door is) c. The bird is left of the cage. (from my point of view)

Note that one and the same relation (like L E F T ) can be used with different frames. Left of the car might be intrinsically defined (with respect to the driver’s side in a US context) or relatively (as seen from my point of view).

619

620

JOOST ZWARTS

The status of some place relations with respect to this distinction is unclear. Mathematically speaking, proximal (N E A R ), distal (F A R ), and comitative relations (W I T H , e.g., I’m with John) are neither topological nor projective. N E A R and W I T H share with I N and O N a notion of “closeness” to the ground that correlates with their resistance to certain distal modifiers, like measure phrases (e.g., *a mile in the city, *an inch on the ground, *a yard near the house, *a foot with John), but this notion of closeness is not part of topology in the strict sense. The relation denoted by B E T W E E N could be approached from a projective angle (involving a bounded axis connecting two objects) or a topological angle (the smallest convex region including two or more objects). The latter approach would also cover among(st) and amid(st), that require a plurality of objects as a ground. In the following two sections I describe English, German, and Dutch Ps that express these place relations. Items that are archaic, regional or specialized (like English abaft, abroad, aboard, astride, for instance) are mostly left out, as well as transparent multiword expressions, like instances of the pattern in the N (of) in English (with N = back, midst, vicinity, . . .). Also omitted are together and apart and their German and Dutch counterparts, because of the central role that the nonspatial notion of collectivity plays in their meaning. The Ps for “north” will be taken as representative for all the cardinal direction Ps. With these restrictions, let us now consider the behavior of PlacePs, from the perspectives of transitivity (whether and how they express their ground, Section 26.3.2) and complexity (how spatial distinctions are reflected in their relative phonological, morphological, and syntactic complexity, Section 26.3.3).

26.3.2 The Transitivity of Places PlacePs always require some sort of reference point, typically the ground. However, whether and how this reference point is made explicit or not varies across the classes of Ps in the three languages. Starting with English, we can first make a distinction between transitive and ambitransitive PlacePs (examples (8a) and (8a’) from Svenonius 2010 and (8b) and (8b’) from Griffiths and Sailor 2015, for British English): (8) a. I saw a small house. Beside *(it) was a gas pump. a’. There was a beach. Above (it), the cliffs swarmed with birds. b. A kangaroo with a joey with *(it) just hopped through the park. b’. The notebook of Dali’s with doodles on (it) should sell for millions. Beside and with are transitive PlacePs (and so are among(st), amid(st), at, beside, next, and upon): They always have an object that expresses the ground explicitly. Above and on are ambitransitive PlacePs: They can optionally leave an inanimate ground argument implicit by picking it up from the linguistic context, like the beach in (8a’) and the notebook in (8b’). Most English PlacePs are actually ambitransitive (with some regional and syntactic restrictions).

Prepositions and Particles

In Dutch, ambitransitivity is much more restricted. With most Ps, the counterpart of the covert ground in examples like (8a’) and (8b’) is a neuter pronoun, spelled out as er before the P (van Riemsdijk 1978). (9) a. Er there Er there b. Het the moet must

was een strand. was a beach boven krioelden de kliffen van vogels. above swarmed the cliffs of birds notitieboek van Dali met doedels er op notebook of Dali with doodles there on miljoenen opbrengen. million bring.in

In Dutch, ambitransitivity can be found with compound-like Ps, as illustrated in (10). (10) a. een driehoek, met een tekening (er) binnenin a triangle with a drawing there inside-in ‘a triangle with a drawing inside it’ b. een doos, met een label (er) bovenop a box with a label there above-on ‘a box with a label on top of it’ c. een optocht, met een muziekkorps (er) achteraan a procession with a band there behind-on ‘a procession with a band behind it’ These Ps refer to particular regions of the interior (in) or exterior (op, aan) of the ground through modification by one of the Ps binnen ‘inside’, buiten ‘outside’, boven ‘over’, onder ‘under’, voor ‘in front’, achter ‘behind’. The resulting ambitransitivity is surprising, since the component Ps themselves are not ambitransitive (as we will see below). Also ambitransitive are a few Ps that mark their object with van ‘of’: (11)

Ik zie een kasteel en I see a castle and ‘I see a castle and a small lake . . . ’

een a

meertje . . . lake-D I M . . .

van) a. links (er left (there of) ‘left of it’ b. terzijde (er van) at-side (there of) ‘beside it’ c. ten noorden (er van) at north-E N (there of) ‘north of it’ Most Dutch PlacePs are not ambitransitive, unlike most English PlacePs. However, these nonambitransitive Ps do alternate between a transitive and an intransitive use, but with a strongly restricted implicit argument and sometimes a different meaning. I will call these PlacePs (in)transitive. Their

621

622

JOOST ZWARTS

implicit argument is not simply a covert version of an overt pronominal object which could pick up any object from the linguistic context, but it is a narrowly described contextual parameter. (English also has (in)transitive PlacePs, but they are more difficult to recognize because of their ambitransitivity.) One class of intransitively used PlacePs refers to regions in relation to the house (‘home’, or a suitable generalization of that notion): (12)

Zij is . . . ‘She is . . . ’ a. voor ‘in (the) front of the house’ c. boven ‘upstairs’ e. binnen ‘in the house, inside’

b. d. f.

achter ‘in (the) back of the house’ beneden ‘downstairs’ buiten ‘outside the house, outside’

These intransitive uses are not elliptical versions of transitive counterparts with the complement het huis ‘the house’. (12c), for example, refers to a region inside the house, boven het huis ‘above the house’ only to a region outside the house. (12e) functions as the intransitive counterpart of in het huis ‘in the house’ and not of binnen het huis (which does not have a clear spatial meaning). We could say that these intransitive Ps do not just “spell out” place functions, but place functions in combination with particular grounds: e.g., binnen is the P that spells out I N (H O M E ). Another class involves implicit body parts in a stereotypical relation with clothes or similar items: (13)

Hij heeft . . . ‘He has . . . ’ a. een hoed op b. een a hat on (his head) a c. een bit in d. schaatsen a bit in (his mouth) skates e. een schort voor an apron in.front (of his body)

jas aan coat on (his body) onder under (his feet)

To the limited extent that the ground can be made explicit, it is restricted to a body part of the possessor. The domain of sport provides another illustration of the narrow interpretations that intransitively used PlacePs can get: de bal is naast / in / uit ‘the ball is wide / in / out’, which requires a ball as the figure and a goal or field as the implicit ground. In general, most PlacePs have highly restricted intransitive uses, sometimes only in morphological contexts: For instance, tussen ‘between’, is only used intransitively in nominal compounds, like tussendeur ‘dividing door’ (lit. between-door). If we recognize these intransitive uses, then only very few PlacePs remain that are exclusively transitive in Dutch, namely te ‘at’ (see Section 26.3.3) and the archaic form benoorden ‘to the north of’. The status of comitative met ‘with’ depends on whether we count mede in mede-student ‘fellow student’ as an intransitive comitative allomorph.

Prepositions and Particles

What German adds to the picture is first a small class of PlacePs that are exclusively intransitive (“adverbial”), morphologically related to (in)transitive PlacePs (Eschenbach 2005), and semantically similar to the Dutch forms in (12): (14)

oben (ob, u¨ber, ‘above’) vorne (vor, ‘in front’) draußen (außer, ‘outside’)

unten (unter, ‘below’) hinten (hinter, ‘behind’) drinnen (in, ‘inside’)

We do not find such exclusively intransitive PlacePs in English and Dutch (apart from incorporation structures with more specialized meanings, like abroad and (at) home in English, thuis te-huis ‘home’ in Dutch). The case government distinction that German makes within the class of PlacePs also involves the transitivity dimension: (15) a.

b.

an ‘on’, auf ‘on’, außer ‘outside’, bei ‘near’, hinter ‘behind’, in ‘in’, mit ‘with’, neben ‘beside’, unter ‘under’, u¨ber ‘over’, vor ‘in front of’, zu ‘at’, zwischen ‘between’ innerhalb ‘inside’, außerhalb ‘outside’, unterhalb ‘below’, oberhalb ‘above’, diesseits ‘on this side’, jenseits ‘on the other side’, beiderseits ‘on both sides’, links ‘left’, rechts ‘right’, no¨rdlich ‘north’, o¨stlich ‘east’, su¨dlich ‘south’, westlich ‘west’, eingangs ‘at the start’, ausgangs ‘at the end’, abseits ‘aside, away’, fern, weitab ‘far away’

Class (15a) governs dative case (when referring to places), as illustrated in (16), class (15b) governs genitive case (17) or marks the object with von (18). Only gegenu¨ber ‘opposite’ fits neither class because it alternates between von and dative case. It is also special in being the only PlaceP-element that can be postpositional. (16)

(17)

(18)

a. bei einem Baum b. hinter dem Haus near a.D A T tree behind the.D A T house ‘near a tree’ ‘behind the house’ a. innerhalb des Hauses b. no¨rdlich des Flusses inside the.G E N house-G E N north-ly the.G E N river-G E N ‘inside the house’ ‘north of the river’ a. innerhalb von Berlin b. no¨rdlich von dem Fluss inside of Berlin north-ly of the.D A T river ‘inside Berlin’ ‘north of the river’

The dative Ps are all morphologically simplex (apart from außer ‘outside’ < aus ‘out’) and more common, the genitive/von Ps are all morphologically complex (but fern ‘far’ only historically so) and less common. Crucially, the dative Ps are transitive or (in)transitive, the genitive/von Ps are ambitransitive. The resulting picture for the three languages is that (in)transitivity (restricted alternation, with a deictic implicit ground) is typical for the

623

624

JOOST ZWARTS

core PlacePs. Both transitive and intransitive PlacePs are less common, and ambitransitivity (free alternation, with an anaphoric implicit ground) is a special feature of (varieties of) English and of noncore PlacePs.

26.3.3 The Complexity of Places In order to get an impression of the distribution of formal complexity over the different place relations, I have organized the most important PlacePs on the basis of their formal complexity in Tables (19)–(21). The rows of the tables group PlacePs with roughly similar meanings, ignoring many finer distinctions. For instance, German and Dutch divide up the O N function along force-dynamic lines, by distinguishing (roughly speaking) between support from below (auf G(erman), opD(utch), for a cup on the table) and support from above (anG, aanD, for a picture on the wall), but this is not reflected in the table. The two columns separate simple Ps from complex Ps. Simple Ps are monosyllabic and monomorphemic and do not mark their object with an additional P (like north of Paris, close to London). All the other Ps are complex. In the complex column we find many types of morphological and syntactic complexity, but also a fair number of basic monomorphemic disyllabic forms ending in -ər or -ən, in all three languages. (19)

Complexity of PlacePs in English Simple at

Complex

AT ON

on

on top

IN

in

inside, within

NEAR

by

near(by), close

WITH

with

together

FRONT

before, in front

BESIDE

next, beside, aside

NORTH

north

FAR

far away

LEFT/RIGHT

(to / on the) left / right

OUTSIDE

outside

BEHIND

behind

OVER

above, over

UNDER

below, beneath, under(neath)

BETWEEN

between, among, amid

OPPOSITE

opposite

Prepositions and Particles

(20)

(21)

Complexity of PlacePs in German Simple zu

Complex

AT ON

auf, an

obendrauf

IN

in

innerhalb

NEAR

bei

nahe(bei)

WITH

mit

zusammen

FRONT

vor

vorne

BESIDE

neben, beiseite, abseits

NORTH

no¨rdlich

FAR

fern, weit weg / von

LEFT/RIGHT

links / rechts

OUTSIDE

außer(halb), außen

BEHIND

hinter, hinten

OVER

u¨ber, oben, oberhalb

UNDER

unter(halb), unten

BETWEEN

zwischen

OPPOSITE

gegenu¨ber

Complexity of PlacePs in Dutch Simple te

Complex

AT ON

op, aan

bovenop

IN

in

binnen(in)

NEAR

bij

nabij, dichtbij

WITH

met

samen

FRONT

voor

BESIDE

naast, terzijde

NORTH

benoorden, ten noorden

FAR

ver weg / van

LEFT/RIGHT

links / rechts

OUTSIDE

buiten

BEHIND

achter

OVER

boven

UNDER

onder, beneden

BETWEEN

tussen

OPPOSITE

tegenover

625

626

JOOST ZWARTS

The tables highlight, along the vertical dimension, a distinction between two groups of place functions. At the top we find a handful of functions (A T , O N , I N , N E A R , W I T H , and F R O N T to some extent) that can be expressed by a simple (monosyllabic and monomorphemic) P combining with its object in a direct way, without the help of an additional P, in contrast to the functions further down in the tables. At first sight, this grouping seems to align with the topological/projective distinction. However, N E A R and W I T H are expressed by simple forms, but they are not topological, while O U T S I D E and B E T W E E N are topological but not expressed in a simple way. Instead, I suggest that one of the major semantic factors promoting a simple P-form is closeness to the ground (with contact as maximal closeness). The functions A T , O N , I N , N E A R , W I T H deliver places that are always necessarily close to the ground, which is not generally true for the other functions (although individual functions, like F R O N T and B E T W E E N , require more study). The functions A T , O N , I N , N E A R , and W I T H form a tight semantic field and the rows are not always as neatly separated as in the tables. A T , the most general place function, is lexicalized as at in English in a variety of uses and by zu in German and te in Dutch in archaic and fossilized uses (zu Ko¨lnG, te KeulenD; zu HauseG, thuisD ‘home’). However, other “close” Ps can also be used to express the A T -relation, in ways that can be idiomatic and subject to variation between the languages: (22)

a. at the corner – an der EckeG, at school – in der SchuleG, at the party – auf der PartyG, at the finish line – bei der ZiellinieG b. at the entrance – aan de ingangD, at the hotel – in het hotelD, at the corner – op de hoekD, at the station – bij het stationD

This gives rise to an ambiguity of such PlacePs between a more general (A T ) and a more specific meaning (e.g., I N or O N ). In Dutch, this ambiguity can be demonstrated through pronominalization. Pronominalizing the ground with the pronoun er is possible with an I N /O N interpretation, as shown in (23), but with an A T interpretation the whole place must be pronominalized (24). (23)

(24)

a. Ze zit in de kist / she sits in the box / ‘She is in the box / in it.’ b. Hij zit op de tafel he sits on the table ‘He is on the table / on it.’ a. Ze zit in de she sits in the ‘She is in the room / there.’ op Texel b. Hij zit he sits on Texel ‘He is on Texel / there.’

/ /

er there / /

kamer room er there

in. in

er there

/ / (*op). (*on)

op. on

er there

(*in). (*in)

Prepositions and Particles

Something similar can be seen in the German examples in (25) and (26). The I N / O N use can be strengthened with a postposition (drin in (25a), drauf in (26a)), but not the A T use ((25b), (26b)). At the same time, the “doubling” in (25a) and (26a) makes the topological meaning I N of in and O N of auf more specific or emphatic. (25)

(26)

a. Sie ist in der Kiste she is in the.D A T box ‘She is inside the box.’ b. Luisa ist in Frankfurt (*drin). ‘Luisa is in Franfurt.’

(drin). (Noonan 2010: 164–165) D R -in

a. Hans war auf dem Tisch (dr-auf). Hans was on the.D A T table there-on ‘Hans was on top of the table.’ b. Hans war auf den Hans was on the.D A T . P L ‘Hans was on the Canary islands.’

Kanaren Canaries

(Haselbach 2016)

(*dr-auf). there-on

This is part of a more general phenomenon that is visible in the horizontal dimension of the tables (19)–(21): there are often two forms for a particular function, differing in complexity. This is not only true when we compare the simple and complex columns (e.g., on and on top), but also within the complex column (e.g., under versus below). The more complex P often identifies a more specific place than the simpler P, in a hyponymic fashion. This is illustrated in (27), with ) standing for the entailment from the more specific PlaceP (hyponym) to the more general PlaceP (hypernym). (27)

a. within the city binnen de stadD

() in the city) () in de stad)

b. inside the box in der Kiste drinG

() in the box) () in der Kiste)

c. on top of the roof bovenop het dakD

() on the roof) () op het dak)

d. in between Mom and Dad tussen mama en papa inD

() between Mum and Dad) () tussen mama en papa)

e. below my window beneden mijn raamD

() under my window) () onder mijn raam)

The more complex forms impose additional spatial requirements, concerning boundaries (27ab), vertical orientation (27c), alignment/proximity (27d), or noncontact (27e). A spoon covered by a napkin cannot be below or beneden that napkin, but only under/onder it. Throughout such examples, English, German, and Dutch often use different formal ways (descriptively, at least) to make the same sort of semantic distinction. For instance, English has the

627

628

JOOST ZWARTS

P+N structures in+side and on+top where German has the postpositions dr+in and dr+auf and Dutch has the compounds binnen+in and boven+op. Furthermore, the English compound with+in has the German compound in +er+halb and Dutch derivation be+in+en as its counterparts. What seems to be important, is that the grammar provides a formal asymmetry (in whatever way) that can be used to mirror the semantic asymmetry between a general meaning (simpler form) and a more specific and emphatic submeaning (complex form). Together with the formal differentiation between close (≈ topological) and nonclose (≈ projective) functions and the transitivity variations, this results in classes of PlacePs in English, German, and Dutch that are much richer in various respects than the closed lists of short function words that they are often assumed to be.

26.4 How Ps Express Paths 26.4.1 More About Paths The places that we discussed in Section 26.3 are systematically related to paths (e.g., Kracht 2002; Gehrke 2008). We saw in Section 26.2 how an I N place (example (28a), (29a)) is mapped to an “into” path by means of the path function T O ((28b), (29b)), which is also known as the goal, cofinal, or allative function. The opposite path functions is F R O M (source, coinitial, ablative, (28c), (29c)). Because of the complementariness of the inside and outside of the ground, I assume that F R O M (I N (T H E - C A G E )) = T O (O U T S I D E (T H E - C A G E )). The function V I A (route, transitory, perlative) is illustrated in (28d) and Figure (29d). The R E F L ( E X I V E ) path function is a bit different, because it does not involve distinct figures and grounds, but situations where the path of a figure is described in terms of its initial position. The figure is its own ground, hence the term reflexive (Lindner 1983). As illustrated in (29e), (28e) involves a path where the cage ends up being inside its own earlier place. (28)

(29)

a. The bird is in the cage. IN(THE-CAGE) b. The bird flew into the cage. TO(IN(THE-CAGE)) c. The bird flew out of the cage. FROM(IN(THE-CAGE)) d. The bird flew through the cage. VIA(IN(THE-CAGE)) e. The cage caved in. R E F L ( I N (T H E - C A G E )) Path relations based on I N x a

b

c

d

e

Prepositions and Particles

I ignore a few less important part functions: approximative (towards), recessive (away from), terminative (up to), egressive (‘starting from’), and make no distinction between V I A paths that are bounded (past) and those that are not bounded (along). It is also important to realize that not all paths might be related to places in this way. The paths described by along, across, and around seem to be based on the overall orientation or shape of the path with respect to the ground object: With along the path is parallel to the main axis of the ground, with across it is orthogonal, and with around it is (roughly) circular with the ground in its center. It is also possible to use paths to describe places. In (30a) a figure is positioned at the end point of an imaginary path leading from a perspective point (“here”) across the street to the location of the bird: E N D is the place function that takes a path and gives a place. In (30b) the path gives us an “linear” place providing a distributed location for the shops, as represented through the function E X T (“extension”) here. (30)

a. The bird is across the street (from here). END(ACROSS(THE-STREET)) b. There are shops along the street. EXT(ALONG(THE-STREET))

Like I did in Section 26.3 for places, I will now explore what role transitivity (26.4.2) and complexity (26.4.3) play in the way paths are expressed in English, German, and Dutch.

26.4.2 The Transitivity of Paths We saw in Section 26.3.2 that places are almost always defined with respect to a ground (though not necessarily an explicit one). The same is obviously true for most paths, because they involve a sequence of positions of a figure with respect to a ground. We can generalize the notion of a reflexive path in such a way that all paths involve a ground, but that this ground can sometimes be the figure itself, as illustrated in (31). (31)

Reflexive PathsPs English REFL+IN cave in REFL+OUTSIDE spread out R E F L +O V E R go up REFL+UNDER push down R E F L +F R O N T go forward R E F L +B E H I N D shrink back REFL+NORTH fly north REFL+ROUND

turn around/ over

German ein stu¨rzen aus breiten auf steigen nieder dru¨cken vorwa¨rts gehen zuru¨ck schrecken nach Norden fliegen um drehen

Dutch in storten uit spreiden op stijgen neer drukken vooruit gaan terug deinzen naar het noorden vliegen om draaien

629

630

JOOST ZWARTS

In all of these cases the figure is changing shape (in, out), moving (up, down, forward, backward), or rotating (around, over) with respect to its initial position. If I fly north, for instance, then my later position is to the north of my earlier position. These reflexive PathPs are always intransitive, with a few idiomatic exceptions (like Dutch in elkaar storten literally in each.other collapse ‘cave in’, with the idiomatic reciprocal elkaar). For instance, independent of their shape (monomorphemic (32a), suffixed with -ward (32b) or phrasal (32c)), the expressions in (32) are consistently intransitive because of their exclusively reflexive path semantics (i.e., R E F L (F R O N T )). (32)

a. forth, fortG, voortD b. forward, vorwa¨rtsG, voorwaartsD c. nach vorneG, naar vorenD

Apart from these reflexive cases, there are PathPs of which the intransitivity might be the result of an explicitly “‘incorporated” nominal ground, as board in overboard. The more productive examples of this involve the suffixal P -ward (33a) and the Ps up/auf G/opD and down/abG/ af D (33b): (33)

a. heavenward, heimwa¨rtsG ‘homeward’, zeewaartsD ‘seaward’ b. upwind, flussabG ‘downstream’, bergopD ‘uphill’

The noun in (33a) designates the goal of the path. In (33b) it provides a directed spatial dimension (wind, stream, slope) along which a reflexive path is defined. The forms in (33b) are “reduced” versions of full PathPPs, like those in (34), omitting the referential elements (the determiner and the deictic elements hinG ‘away from speaker, thither’ and herG ‘toward speaker, hither’). (34)

a. up the mountain, den Berg hinauf G, de berg opD b. down the river, den Fluss herabG, de rivier af D

On the other side of the transitivity spectrum we find PathPs that are exclusively transitive, but there are not many. The set includes Ps that correspond more or less to the three path functions F R O M (from, vonG, vanD), G G D V I A (the Latinate P via, in all three languages), and T O (bis , nach , tot , D naar ), as well as some composite forms (toward(s), into, onto, upon in English, vanaf, vanuit, vandaan in Dutch). The simple Ps combine with noun phrases (with an implicit A T (35)) or with a PP (36). (The asterisk in T O * indicates that this is a slightly different type of (terminative) path function, ‘up to’.) (35)

F R O M (A T (N Y )): V I A (A T (N Y )): T O (A T (N Y )): T O * (A T (N Y )):

from/vonG/vanD New York viaE,G,D New York nachG/naarD New York bisG/totD New York

Prepositions and Particles

(36)

F R O M (U N D E R (T A B L E )): D

from under the table, von unter dem TischG,

van onder de tafel G D T O (O U T S I D E (H O M E )): nach außen , naar buiten G D T O * (B E H I N D (H O T E L )): bis hinter dem Hotel , tot achter het hotel Most PathPs alternate between a transitive use (with the ground realized as a complement within the PathPP) and an intransitive use (where this is not the case). In the transitive use the complement can be realized as a noun phrase (off the wall, down the tree), or as a PP (e.g., out of the box, away from the house). There are at least three different types of intransitivity with PathPs. First, with some PathPs, any type of ground object can be anaphorically picked up from the context, as illustrated in the English example (37a) from Huddleston and Pullum (2002: 283) and similar examples in (37b), where off, in, and out behave as ambitransitive PathPs. (37)

a. She climbed onto the wall and immediately jumped off. b. I fell in (the dam), I got out (of the box)

The Dutch translations in (38) require the pronoun er to make the ground explicit, which demonstrates again (like Section 26.3.2 did for PlacePs) that English allows more ambitransitivity in the P domain. (38) a. Ze klom op de muur en sprong she climbed on the wall and jumped b. Ik viel er in, Ik kwam er I fell there in, I came there

er meteen af. there immediately off uit out

Second, the implicit ground can be a more restricted deictic parameter, especially with PathPs that are used as resultative particles, as illustrated with the Dutch examples in (39). (39)

a. de hond uit laten ‘take out the dog’ (out of the house) b. een bezoeker binnen laten ‘let in a visitor’ (into the house) c. in ademen ‘breathe in’ (into the body) d. onder dompelen ‘immerse’ (under the water surface) e. geld weg gooien ‘throw away money’ (away from agent) f. over drijven ‘drift over’ (over point of view) g. langs lopen ‘walk past’ (past point of view)

All of these examples involve a more specific ground than when the object pronoun er would have been used, often functioning as part of a more stereotypical event. (39a), for example, does not just refer to taking the dog outside of the house, but to the activity of walking the dog. Third, the ground of an intransitive PathP that is used as a particle can also be expressed as the direct object of the verb (what is sometimes known as “ground promotion”, because the ground is “promoted” to the object position of the verb):

631

632

JOOST ZWARTS

(40)

a. b. c. d.

een formulier in vullen ‘fill in a form’ (letters into a form) een doek uit knijpen ‘squeeze out a rag’ (liquid out of a rag) een tafel af vegen ‘wipe off a table’ (dirt off a table) een woord door strepen ‘strike through a word’ (line through a word) e. een woord onder strepen ‘underscore a word’ (line under a word) f. een rivier over bruggen ‘bridge a river’ (bridge over a river) g. een kind om armen ‘embrace a child’ (arms around a child)

The alternation between transitive and intransitive uses might sometimes involve allomorphy, with the German in /ein ‘in’, Dutch naar/toe ‘to’, met/ mee ‘with’, and van/af ‘off’: (41)

a. in einem Kreis ‘in a circle’ – ein Wort einkreisen ‘to encircle a word’ b. wuiven naar iemand – iemand toewuiven ‘wave to someone’ c. lopen met iemand ‘walk with someone’ – meelopen ‘walk (with someone)’ d. van de muur breken ‘break off the wall’ – afbreken ‘break off’

These allomorphs might also co-occur, one as a particle or prefix and the other heading the PP that optionally complements it: (42)

a. b. c. d.

in das Zimmer eintreten ‘enter into the room’ toegang naar/tot iets ‘access to something’ met iemand meegaan ‘go with somebody’ van de muur afbreken ‘break off from the wall’

The path domain partially confirms what we saw in the place domain: Transitivity is limited, ambitransitivity is restricted, and core Ps are (in)transitive. However, the application of the reflexive path function gives rise to a relatively large and prominent class of intransitive PathPs in all three languages.

26.4.3 The Complexity of Paths Given that (most) paths are semantically composed of a path function applying to a place, we can ask how the different path+place combinations are expressed by means of Ps in the three languages and what this shows about the structure of the P domain. We can distinguish five different ways in which this is done. I first describe the situation with the path functions F R O M , V I A , and T O . I will ignore combinations of Ps like in down into the darkness or back in the box, which I take to involve intersection of two path categories. The first path, for instance, is a path that goes both “down” and “into the darkness”, but these are two independent characterizations, irrelevant for the kind of

Prepositions and Particles

complexity that I am interested in here. The same is true for German auf das Dach hinunter ‘down onto the roof’, which describes a path as being auf das Dach ‘onto the roof’ and hinunter ‘down’ (van Riemsdijk and Huybregts 2007). Marking expression: The path function can be expressed most transparently through a separate PathP that combines with and “marks” a PlaceP(P). In other words, the path and place function are expressed analytically (separately), although sometimes tied more closely together. (43)

FROM

VIA

TO

from between the clouds, von zwischen den WolkenG, van tussen de wolkenD, zwischen den Wolken hervorG, tussen de wolken vandaanD unter dem Fluß durchG, onder de rivier doorD ‘under the river’ to behind the cage, into the cage, onto the cage, nach hintenG, naar achterenD ‘to the back’

All three languages use this strategy prepositionally for F R O M and German and Dutch also postpositionally. V I A is expressed through a postpositional P in German and Dutch. English has an overt T O in into and onto and marginally with other PPs; in German and Dutch it is restricted to intransitive PlacePs. This formally more complex strategy occurs mainly with the “secondary” type of places of Section 26.3.3. The expression of path confirms here the observation that formal complexity correlates with semantic “markedness.” Zero expression: When the path function is left unexpressed, we have a path expression that has the form of a place expression. This is never possible with F R O M paths in any of the three languages. Only English uses it for V I A , instead of the postpositional option that we just saw for German and Dutch. (44)

a. b. c.

VIA TO TO

(go) under the river (to the other side) (dive) under the table, onder de tafel (duiken)D (breathe) in, ein(atmen)G, in(ademen)D

It occurs fairly often in English and Dutch T O paths, but only if licensed by appropriate verbs. The intransitive Dutch PathP binnen ‘inside’, for instance, can only drop the T O marker naar with certain verbs (45a), but not with others (45b), and definitely not in contexts where it is not governed by a verb (45c, 45d). (45)

a. b. c. d.

(naar) binnen komen/stappen ‘come/walk in’ ?(naar) binnen gaan/dansen ‘go/dance in’ de weg *(naar) binnen ‘the way in’ Gauw, *(naar) binnen! ‘Quick, get in!’

The particle examples in (44c) illustrate that the particle position neutralizes the T O + I N versus I N distinction, at least in English and Dutch, because

633

634

JOOST ZWARTS

it uses in for T O + I N . This holds for German too if one views ein as only a superficial allomorph of in. Case expression: Several core PlacePs in German refer to a T O path by using accusative case on the noun phrase instead of dative case. (46)

a. in dieACC StadtG (gehen) ‘(go) into the city’ b. unter denACC Tisch (tauchen) ‘(dive) under the table’

This alternation occurs with most of the dative PlacePs identified in Section 26.3.2 (an, auf, hinter, neben, in, u¨ber, unter, vor, zwischen) and it is the German counterpart of the English and Dutch zero expression of T O that we saw in (44b). Postpositional expression: It is possible that Dutch expresses the T O versions of I N and O N by putting the corresponding PlacePs in and op postpositionally: (47)

a. de stad in (gaan)D ‘go into the city’ b. de tafel op (klimmen)D ‘climb onto the table’

It is not always easy to distinguish these (transitive) postpositional Ps from (intransitive) particle Ps because postpositions can also incorporate in the verb in the same way as particle Ps, as shown in (48), where the postpositional Ps have become part of the progressive verb form aan het V- I N F . (48)

a. de stad [aan het in gaan] the city on the in go-I N F ‘going into the city’ b. de tafel [aan het op klimmen] the table on the on climb-I N F ‘climbing onto the table’

Suppletive expression: The path and place function can together be expressed as one special form. This happens most clearly with the F R O M / V I A + I N /O N combinations: (49)

FROM+IN VIA+IN FROM+ON VIA+ON

out of the garden, aus dem GartenG, uit de tuinD through the garden, durch den GartenG, door de tuinD break off, abbrechenG, afbrekenD over the bridge, u¨ber die Bru¨ckeG, over de brugD

We see here again a reflection of the “closeness” of the place relations involved (see Section 23.3). Not only places but also paths based on the “closer” place functions (especially A T , O N , I N ) are expressed by shorter, tighter forms than the other place functions. Cumulation: Finally, the path can be expressed in a cumulative way, using more of these mechanisms at the same time.

Prepositions and Particles

(50)

a. vanuit de kamerD b. in dasA C C Zimmer hineinG c. de kamer doorD d. de brug onderdoorD e. durch den Wald hindurchG f. unter denA C C Schrank drunterG g. um den Tisch drum rumG

‘out of the room’ ‘into the room’ ‘through the room’ ‘(via) under the bridge’ ‘through the forest’ ‘(via) under the wardrobe’ ‘around the table’

In (50a) van marks F R O M and uit expresses it again in a suppletive way. In (50b) T O is expressed through accusative case, but also through the complex postpositional hinein form. Example (50c) illustrates a combination of marking the path (V I A ) with the suppletive form door and postpositioning the item. In (50d) there is a combination of marking (by door) with a phrase-final position. Note that door functions as the suppletive expression of V I A + I N in (50c), but as the marking of V I A in (50d). Examples (50e) and (50f) show that the same Ps can be circumpositionally reduplicated, or even retriplicated (50g) (Noonan 2010: 169). Because of their postpositional options, German and Dutch have rich possibilities of “overexpressing” path information, creating room for interpretive asymmetries, such as those in (51): (51)

a. Wij reden door Belgie¨. ‘We drove through Belgium.’ a’. Wij reden door Belgie¨ heen. ‘We drove through Belgium.’ b. Hij kwam op de berg. ‘He came on the mountain.’ b’. Hij kwam de berg op. ‘He came up the mountain.’

Example (51a’), with the postposition heen, has a stricter meaning than (51a): It can only be true if we drove from one end of the country to the other end, e.g., from the Netherlands to France (Claessen and Zwarts 2010: 35–36). Example (51b’) also has a stricter meaning than (51b): It can only be true if we reached the top by climbing up the mountain, not if we arrive on the top by helicopter. This illustrates a specialization of shorter and longer forms that we already observed in the place domain in Section 26.3.3. Also here we can observe that languages differ in how they express a particular distinction (if they make it at all), as illustrated in (52): (52)

a. herumlaufen – umherlaufenG ‘walk round’ b. omlopen – rondlopenD ‘walk round’

Both German and Dutch (but not English) make a distinction between ‘detour round’ and ‘crisscross round’ with a range of verbs. German uses the morphological contrast between herum and umher, while Dutch uses the lexical contrast between om and rond. The reflexive PathPs show such contrasts more extensively. I have tabulated the more common reflexive PathPs in Tables 26.1 and 26.2,

635

636

JOOST ZWARTS

Table 26.1 Simple reflexive PathPs

IN OUTSIDE OVER UNDER FRONT BEHIND

English

German

Dutch

in out up down forth back

ein aus auf nieder fort zurück

in uit op neer voort terug

Table 26.2 Complex reflexive PathPs

IN OUTSIDE OVER UNDER FRONT BEHIND

English

German

Dutch

inward outward upward downward forward backward

hinein hinaus hinauf, nach oben hinunter, nach unten vorwärts, nach vorn rückwärts, nach hinten

naar binnen naar buiten naar boven, omhoog naar beneden, omlaag vooruit, naar voren achteruit, naar achteren

separating simple forms in Table 26.1 from complex forms in Table 26.2, treating zu-ru¨ck and te-rug as synchronically simple. Tables 26.1 and 26.2 show different strategies for encoding the R E F L path function relative to the PlaceP: marking (e.g., R E F L +voor = vooruit), zero (e.g., R E F L +in = in), suppletion (e.g., R E F L +under = down). The simple, but not the complex reflexive PathPs can enter into the close combination with verbs that is characteristic for particles. When we compare a simple form with its complex counterpart, as illustrated in (53) for Dutch, we can see that the simple forms have more stereotypical and idiomatic meanings than the complex forms, irrespective of how those complex forms are built up. (53)

a. inkijken ‘take a look at’ – naar binnen kijken ‘look inside’ b. uitgaan ‘go out at night’ – naar buiten gaan ‘go outside’ c. opkijken ‘look up by lifting the eyes’ – omhoog kijken ‘look upward’ d. neerkomen ‘come down to the ground’ – omlaag komen ‘come downward’ e. voortbewegen ‘move on’ – vooruit bewegen ‘move forward’ f. teruglopen ‘walk back to origin’ – achteruit lopen ‘walk backward’

In this respect, particles are not so special within the wider landscape of Ps, because we have seen that intransitive and simple Ps in general often have such specialized meanings with respect to their transitive and complex counterparts.

Prepositions and Particles

26.5 Conclusion This overview of place and path expressions is necessarily restricted in many ways, because of its descriptive focus on the internal structure of spatial PPs in three Germanic languages. Due to time and space restrictions, I could not pay attention to the sophisticated approaches to the syntax and semantics of PPs that have developed over the past decades, especially for these three languages, to the interaction of PPs with the larger sentential context, in particular with verbal semantics, aspect, and argument structure, and to the position of English, German, and Dutch Ps within a wider typological and grammaticalization setting, including also the other Germanic languages. But most obvious is what remains to be said about the nonspatial part of category P: the way prepositions and particles are also used to talk about time, aspect, quantity, force and other nonspatial domains and the grammatical roles they are often required to play. Hopefully, this overview of the spatial P system of English, German, and Dutch can also help to deepen our understanding of this bigger picture.

References Carlson, L. and E. van der Zee (eds.) 2005. Functional Features in Language and Space: Insights from Perception, Categorization, and Development. Oxford University Press. Claessen, C. and J. Zwarts 2010. “On the directional particle heen.” In J. van Kampen and R. Nouwen (eds.), Linguistics in the Netherlands 2010. Amsterdam: John Benjamins: 31–43. Eschenbach, C. 2005. “Contextual, functional, and geometric components in the semantics of projective terms.” In L. Carlson and E. van der Zee (eds.), Functional Features in Language and Space: Insights from Perception, Categorization, and Development. Oxford University Press: 71–91. Garrod, K. R. and S. A. Coventry 2004. Saying, Seeing, and Acting: The Psychological Semantics of Spatial Prepositions. New York: Psychology Press. Gehrke, B. 2008. Ps in Motion: On the Semantics and Syntax of P Elements and Motion Events. Ph.D. dissertation, Universiteit Utrecht. Griffiths, J. and C. Sailor 2015. “Prepositional object gaps in British English.” In B. Ko¨hnlein and J. Audring (eds.), Linguistics in the Netherlands 2015. Amsterdam: John Benjamins: 63–74. Haselbach, B. 2016. P’s at the Interfaces: On the Syntax, Semantics, and Morphology of Spatial Prepositions In German. Ph.D. dissertation, Universita¨t Stuttgart. Herskovits, A. 1986. Language and Spatial Cognition: An Interdisciplinary Study of The Prepositions in English. Cambridge University Press.

637

638

JOOST ZWARTS

Huddleston, R. and G. K. Pullum (eds.) 2002. The Cambridge Grammar of the English Language. Cambridge University Press. Jackendoff, R. 1983. Semantics and Cognition. Cambridge, MA: MIT Press. Kracht, M. 2002. “On the semantics of locatives,” Linguistics and Philosophy 25.2: 157–232. Levinson, S. C. 1996. “Frames of reference and Molyneux’s question: Crosslinguistic evidence.” In P. Bloom, M. A. Peterson, L. Nadel, and M. F. Garrett (eds.), Language and Space. Cambridge, MA: MIT Press: 109–169. Lindner, S. J. 1983. A Lexico-Semantic Analysis of English Verb Particle Constructions with OUT and UP. Bloomington, IN: Indiana University Linguistics Club. McIntyre, A. 2015. “Particle-verb formation.” In P. Mu¨ller, I. Ohnheiser, S. Olsen, and F. Rainer (eds.), Word-Formation: An International Handbook of the Languages of Europe, Vol. 1. Berlin: Mouton de Gruyter: 434–450. Noonan, M. 2010. “A` to Zu.” In G. Cinque and L. Rizzi (eds.), Mapping Spatial PPs: The Cartography of Syntactic Structures, Vol. 6. Oxford University Press: 161–195. Randell, D. A., Z. Cui, and A. G. Cohn 1992. “A spatial logic based on regions and connection.” In Proceedings of the 3rd International Conference on Knowledge Representation and Reasoning. San Mateo: Morgan Kaufman: 165–176. Riemsdijk, H. van 1978. A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Dordrecht: Foris. Riemsdijk, H. van and R. Huybregts 2007. “Location and locality.” In S. Karimi, V. Samiian, and W. Wilkins (eds.), Clausal and Phrasal Architecture: Syntactic Derivation and Interpretation: A Festschrift for Joseph E. Emonds. Amsterdam: John Benjamins: 339–364. Svenonius, P. 2007. “Adpositions, particles, and the arguments they introduce.” In E. Reuland, T. Bhattacharya, and G. Spathas (eds.), Argument Structure. Amsterdam: John Benjamins: 63–103. Svenonius, P. 2010. “Spatial p in English.” In G. Cinque and L. Rizzi (eds.), Mapping Spatial PPs: The Cartography of Syntactic Structures, Vol. 6. New York: Oxford University Press: 126–160. Talmy, L. 1975. “Figure and ground in complex sentences.” In C. Cogen (ed.), Proceedings of the First Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society: 419–430. Zwarts, J. 2017. “Spatial semantics: modeling the meaning of prepositions,” Language and Linguistics Compass 11.5: 1–20.

Chapter 27 Negative and Positive Polarity Items Doris Penka

27.1 Introduction Like many, if not all, other languages, the Germanic languages have expressions that are sensitive to the polarity of the clause in which they occur. Negative polarity items (NPIs) can only occur in clauses that are negative (in a sense to be made precise below), whereas positive polarity items (PPIs) can only occur in clauses that are affirmative. The prime examples are the determiners any and some in English (as well as their derivatives like somebody/anybody, etc.). As an NPI, any is acceptable under negation, but gives rise to ungrammaticality if it occurs in a plain affirmative clause. (1)

a. I didn’t call anyone. b. *I called anyone.

Some, in contrast is acceptable in affirmative clauses, but cannot be used under negation. While a sentence in which some co-occurs with negation, such as (2b), is grammatical, some crucially cannot be interpreted in the semantic scope of negation. For (2b) only a reading is available where the indefinite takes wide scope over negation (‘There is somebody who I didn’t call’). The sentence cannot mean the same as (1a), i.e., that I called nobody. (2)

a. I called somebody. b. I didn’t call somebody. * not ≫ somebody, ✓ somebody ≫ not

This contrast exhibited by any and some is usually described by saying that negation licenses NPIs like any and antilicenses PPIs like some. At this point it is useful to mention some caveats that one should be aware of when investigation polarity items. The first concerns indefinite NPIs like any in English. It is well known that there are uses of any that do not require a negative licenser. In these so-called Free Choice uses,

640

DORIS PENKA

illustrated in (3), any is not interpreted existentially as in NPI-uses, but receives a quasi-universal reading. (3)

a. You may choose any advanced course. b. Any five-year old could answer this question.

It is a long-standing question whether any is lexically ambiguous or whether both NPI and Free Choice uses can be derived under a unified analysis. This question will be addressed in Section 27.6. For now it is important to be aware that examples like (3) do not undermine the claim that any is an NPI. The second caveat concerns negation, which as said above licenses NPIs and antilicenses PPIs. There is, however, a type of negation that is known to behave differently with respect to polarity items. So-called metalinguistic or echoic negation is not used to deny the truth of the embedded proposition but rather to object to the assertability of the corresponding positive utterance (see Horn 1985). The sentence in (4), for instance, which contains the PPI some, would typically be used to object to the previous utterance John met some old lady on grounds of it being inappropriate to call the person he met some old lady, as the continuation makes clear. (4)

John didn’t meet some old lady. He met the Queen of England.

Since metalinguistic negation does not affect truth-conditions, it does not license NPIs and antilicense PPIs. A final caveat concerns another class of elements that show a certain affinity to negation, so-called negative concord items (also called negwords or n-words). In Afrikaans, for example, negative indefinites like niemand ‘nobody’ have to co-occur with the negative marker nie. (5)

Hier slaap niemand *(nie) (Afrikaans) here sleeps nobody not ‘Nobody sleeps here.’ (from Biberauer and Zeijlstra 2012: 357)

While negative concord items are sometimes discussed on a par with NPIs (e.g., Laka 1990; Giannakidou 1998 and sequel; Chierchia 2013) it is quite clear that the restrictions governing their distribution are different (see Penka 2011 for extensive discussion). Most importantly, negative concord items are self-licensing in the sense that in certain contexts they seem to be able to induce the licensing negation by themselves (Ladusaw 1992). For instance, negative concord items can be used as negative fragmentary answers, whereas NPIs cannot. (6)

a. Wie het my boek gesien? who has my book seen? ‘Who saw my book? b. Who knows the answer?

Niemand. No-one No-one’ *Anybody.

(Afrikaans) (from Biberauer and Zeijlstra 2012: 357)

Negative and Positive Polarity Items

Because of these differences, negative concord items will not be discussed in this chapter. Questions that the phenomenon of polarity sensitivity raises for linguistic research and that will be addressed in the rest of the chapter include the following (see also Ladusaw 1997): (i) (ii) (iii) (iv)

What are the lexical properties of polarity sensitive items? How can the environments in which NPIs are licensed and PPIs antilicensed be characterized as natural class? How is the licensing condition to be formulated? Why are polarity items sensitive to the polarity of the context in which they occur?

Section 27.2 takes a closer look at the lexical semantics of polarity sensitive items. While this section addresses NPIs and PPIs alike, the following sections will focus on NPIs, with which the bulk of work on polarity items is concerned. Section 27.3 takes a closer look at the distribution of NPIs and gives a semantic characterization of the environments in which NPIs are licensed. Section 27.4 considers the fact that not all NPIs behave alike and discusses several subtypes that have been identified in the literature. Section 27.5 is concerned with the licensing condition and addresses issues about the structural relation between an NPI and its licenser. Section 27.6 addresses the question why there should be expressions in language that are polarity sensitive. Section 27.7 returns to PPIs and discusses their distribution as well as analyses that have been proposed in the literature. Section 27.8 is the conclusion.

27.2 Lexical Properties of Polarity Sensitive Items While polarity items are found in many syntactic categories – determiners like any, verbs like bother in English and hoeven ‘need’ in Dutch, VPs like budge an inch, adverbs like ever, and particles like yet and either among others – it has been observed that across languages, expressions that are polarity sensitive tend to have certain properties concerning their lexical semantics. Cross-linguistically, indefinites are often NPIs, like indefinites formed with any in English and irgend- in German. While not all Germanic languages have a full paradigm of NPI indefinites, most have an indefinite temporal adverb that is an NPI: ever in English, ooit in Dutch, jemals in German, naº gonsin in Swedish. A second common type of NPIs across languages comprises idiomatic expression describing minimal amounts or minimal intensities of actions. Examples of such minimizer NPIs from English include lift a finger, budge an inch, give a damn, drink a drop, sleep a wink, give a red cent, and have the faintest

641

642

DORIS PENKA

idea. Similar idiomatic expressions are found in other Germanic languages, e.g., einen Finger ru¨hren ‘move a finger’, ein Auge zu tun ‘close an eye’, and einen blasen Dunst haben ‘have a pale haze’ in German; ett ro¨tt o¨re ‘a red cent’, ett dug ‘a bit’, ett jota ‘a jot’ in Swedish (Brandtler 2012). Another type consists of scalar focus particles like English so much as, German auch nur, Dutch ook maar, Swedish ens (all of which are NPI-versions of English even). While the above-mentioned types of NPIs all denote or combine with low-scalar expressions, other, less common types denote or involve high quantities. These high-scalar NPIs come in two different flavors, depending on the way they are used conversationally. First, there are maximizing NPIs that are used to make an utterance stronger and add emphasis, similarly to the minimizer NPIs discussed above. Examples from English are the following: (7)

a. He would*(n’t) do it for all the tea in China. b. I *(don’t) agree at all. c. I have*(n’t) seen him in weeks/ages.

High-scalar NPIs of the second type are used in the opposite way, i.e., to make a claim weaker. English examples of such NPIs, which Israel (1996) calls understating or attenuating, are given in (8). (8)

a. John is *(not) all that clever. b. I *(don’t) much like coffee.

Turning now to PPIs, we also find emphatic and attenuating uses. But the pattern is now the reverse: High-scalar expressions make emphatic PPIs (i.e., they make an utterance stronger) and low-scalar PPIs are attenuating (Israel 1996). Examples of emphatic PPIs are the following:1 (9)

a. They (*don’t) earn tons of money. b. We are (*not) far better off without him.

Moreover, most intensifiers are PPIs, e.g., utterly, entirely (van der Wouden 1997). Examples of understating or attenuating PPIs include degree modifiers like rather and pretty and quantifiers like a tad and a little bit, as illustrated in (10). (10)

a. John is(*n’t) rather tall. b. I am (*not) a little bit worried.

It should, however, be noted that these generalizations in terms of lexical properties correspond to mere tendencies. There are also instances of polarity sensitive items that do not straightforwardly fit into this 1

All cited examples of PPIs are acceptable under a metalinguistic interpretation of negation (see Section 27.1 above). The ‘*’ applies if negation is read as plain sentential negation.

Negative and Positive Polarity Items

classification. Examples for such NPIs include the temporal preposition until (and its cognate fo¨rran in Swedish) when combined with a time point, and the modal verbs hoeven ‘need’ in Dutch and brauchen in German. Any attempt to reduce polarity sensitivity entirely to lexical semantic properties is moreover called into question by the existence of pairs of expressions that arguably mean the same, where one member is a PPI and the other an NPI, such as the determiners some – any. Other examples are the particles already – yet and too – either. (11)

a. John has arrived {already / *yet} b. John hasn’t arrived {*already / yet}

Having taken a closer look at the lexical semantics of polarity items, we next turn to the environments in which they can occur.

27.3 Distribution of NPIs It is well known that NPIs cannot only be licensed by negation, but can also occur in many other grammatical contexts as well. A nonexhaustive list of environments in which NPIs are licensed is given in (12), along with examples illustrating with the NPI ever. (12)

a. Negative quantifiers and adverbs: nobody, no N, never Nobody has ever seen him again. b. Semi-negative quantifiers and adverbs: few, at most n, only, rarely, seldom, hardly At most five people have ever read this book. c. Negative conjunctions: without, neither . . . nor He left without ever looking back. d. Semi-negative predicates: doubt, be surprised, be unlikely I doubt that she’ll ever come back. e. Comparative constructions: more than, too The situation is worse than we ever imagined. f. First argument (restrictor) of universal quantifiers: Everyone who has ever been to Paris remembers the atmosphere. g. Antecedents of conditionals: If you ever visit China, you should see the Great Wall. h. Questions: Have you ever heard of such a thing?

This rather diverse list raises the question of what property all these contexts have in common and how the environments in which NPIs are licensed can be characterized as a natural class. As observed by Fauconnier (1975, 1979) and Ladusaw (1979), (most of) the contexts that license NPIs can be characterized as giving rise to downward entailing inferences. This means that these contexts allow

643

644

DORIS PENKA

inferences from general to more specific statements, i.e., from sets to subsets. Most other contexts are upward entailing and allow inferences from sets to supersets, i.e., from the specific to the general. Entailment patterns are therefore reversed in downward entailing contexts. For illustration, consider the two arguments of the quantifier every. Given that dogs are pets and that terriers constitute a subkind of dog, (13) shows that the first argument of every allows downward entailing inferences from sets to subsets, but not upward entailing inferences from sets to supersets. (13)

a. Everyone who owns a dog has to pay a fee. b. ==> Everyone who owns a terrier has to pay a fee. c. =/=> Everyone who owns a pet has to pay a fee.

(14)

a. Everyone who wants to join this club has to own a dog. b. =/=> Everyone who wants to join this club has to own a terrier. c. ==> Everyone who wants to join this club has to own a pet.

The second argument of every, in contrast, allows inferences from sets to supersets, rather than inferences from sets to subsets, and thus constitutes an upward entailing context, as shown in (14). Formally, downward entailment (also called downward monotonicity) can be defined as follows (where “)” stands for cross-categorial entailment)(von Fintel 1999: 100): (15)

A function f of type is downward entailing if and only if for all x,y of type σ such that x ) y: f(y) ) f(x).

The hypothesis that NPIs are licensed in downward entailing contexts then accounts for the fact that NPIs can occur in the first argument of every, but not in the second, as illustrated in (16). (16)

a. Everyone who answered any question correctly passed the test. b. *Everyone who passed the test answered any question correctly.

While the characterization of NPI licensing contexts in terms of downward entailment has been widely adopted, it also has some well-known problems. The most obvious is that some of the contexts in which NPIs are licensed are not intuitively downward entailing. Examples are the scope of only and complement clauses of be surprised and be sorry, in which NPIs like any and ever are licensed, as illustrated in (17). Inferences from sets to subsets, however, are not intuitively valid and the inferences in (18) do not go through. (17)

a. Only Sam ate anything. b. Tom is sorry that he ever left.

Negative and Positive Polarity Items

(18)

a. Only Sam ate a vegetable. #Therefore, only Sam ate spinach. b. Tom is sorry that he bought a car. #Therefore, Tom is sorry that he bought a Mercedes.

Von Fintel (1999) argues that what is responsible for the invalidity of the inferences in (18) is a presupposition triggered by the licenser, which interferes with downward entailment. In the case of only for instance, it is usually assumed that while the exclusive meaning component is asserted, the truth of the prejacent (the clause without only) is presupposed. Thus, while the inference from (19) to (20) goes through as far as the truth conditions in (20a) are concerned, the presupposition in (20b) does not follow from anything in (19) (Sam could have eaten carrots, for instance), and thus spoils the overall inference. (19)

Only Sam ate a vegetable. a. Truth conditions: Nobody who is not Sam ate a vegetable. b. Presupposition: Sam ate a vegetable.

(20)

Only Sam ate spinach. a. Truth conditions: Nobody who is not Sam ate spinach. b. Presupposition: Sam ate spinach.

Similarly, factive verbs like be surprised and be sorry trigger the presupposition that the complement clause is true. In case of (18b), the second sentence presupposes that Tom bought a Mercedes, which is not guaranteed by the first sentence, making the overall inference invalid. Thus, while the scope of only and complement clauses of be surprised and be sorry constitute downward entailing contexts as far as truth conditions are concerned, they do not do so anymore once presuppositions are also taken into account. Von Fintel (1999) defines a form of entailment, which he called Strawson entailment, in which presuppositions are factored out, and argues that this is the relevant notion underlying NPI licensing. (21)

A function f of type is Strawson downward entailing if and only if for all x,y of type σ such that x ) y and f(x) is defined: f(y) ) f(x).

But even if Strawson downward entailment is adopted as the unifying property of contexts in which NPIs are licensed, there are still cases that are not covered by this generalization. There are examples of acceptable NPIs that do not occur in a (Strawson) downward entailing context. A wellknown example from Linebarger (1987: 373) is (22a), where any occurs in the scope of the nonmonotone quantifier exactly n (i.e., in an environment that is neither upward nor downward entailing). Similar, even more extreme cases where “negation is everywhere present yet nowhere visible” like (22b) and (22c) are discussed by Horn (2001: 176f) under the label “Flaubert polarity.”

645

646

DORIS PENKA

(22)

a. Exactly four people in the world have ever read that dissertation: Bill, Mary, Tom, and Ed. b. A fat lot of good THAT ever did me. c. Small thanks you get for THAT, either.

To deal with such cases, it has been proposed that there are two ways in which NPIs can be licensed (Baker 1970; Linebarger 1980, 1987). Besides the core case of licensing, in which NPIs are licensed by negation, it is assumed that NPIs can also be licensed pragmatically via a “Negative Implicatum” (which is supposed to subsume entailments as well as presuppositions and implicatures). The NPI ever in (22a), for instance, is argued to be licit because the sentence plausibly gives rise to the inference “Everyone who is not Bill, Mary, Tom, or Ed has not ever read that dissertation”, in which ever is in the scope of negation and thus licensed under the core condition. The appeal of such an account lies in the assumption that the Negative Implicatum has to be pragmatically salient, thus making leeway for context-dependency. This squares well with the observation that the possibility of NPIs to occur in such environments crucially depends on the context. While ever is acceptable under nonmonotone exactly four in (22a), it is not under nonmonotone exactly one hundred in (23). (23)

*Exactly one hundred people in the world have ever read that dissertation.

The problem with this pragmatic approach to NPI licensing is that it is not restrictive enough and severely overgenerates. Because there are many inferences of any given sentence and many different ways to represent them, it will always be possible to come up with a Negative Implicatum, and it is unclear how and under which conditions precisely this Negative Implicatum is salient enough for an NPI to be licensed.

27.4 Varieties of NPIs While identifying (Strawson) downward entailment as the unifying property of the contexts that licenses NPIs has been an important step towards our understanding of NPIs, it cannot be the whole story. It has been observed that not all NPIs behave uniformly and different classes have been identified. The first dimension of variation concerns the locality requirements that NPIs pose on their licenser. While NPIs of the any-kind can also be licensed by a negation in a higher clause, as in (24a), so called strict NPIs such as punctual until require a clausemate licenser (Horn 1978): (24)

a. I didn’t claim that Mary stole anything. b. *I didn’t claim that Mary would arrive until midnight.

Negative and Positive Polarity Items

Another dimension along which NPIs differ has to do with the strength of negativity that an NPI requires in order to be licensed. While it was shown above that NPIs like any and ever are licensed in all (Strawson) downward entailing contexts, there are other NPIs that have a more limited distribution and require a stronger form of negativity from their licensers. The following examples illustrate this for in weeks and yet: (25) a. I have*(n’t) seen Mary in weeks. b. No doctor has seen Mary in weeks. c. *At most five doctors have seen Mary in weeks. (Gajewski 2011: 114) (26) a. b. c. d.

I have *(n’t) been here yet. No one has been here yet. *At most five people have been here yet. *I regret that you have been here yet. (Szabolcsi 2004: 427)

One approach to tackle NPIs that are only licensed in a subset of downward entailing contexts distinguishes different logical strengths of negativity. While downward entailment corresponds to a rather weak notion of negativity, stronger notions can be formally defined (Zwarts 1996). Besides downward entailment, other logical properties that have been argued to play a role in the licensing of polarity items are antiadditivity and antimorphism.2 The formal definitions (in terms of Boolean algebras) are given in (27).3 (27)

a. A function f is anti-additive if and only if for all x,y in its domain: f(x _ y) , f(x) ^ f(y). b. A function f is antimorph if and only if for all x,y in its domain: f(x _ y) , f(x) ^ f(y) and f(x ^ y) , f(x) _ f(y)

These logical properties define a hierarchy of negative contexts: Antimorphism is the strongest form of negativity and downward entailment the weakest. Antimorphic contexts constitute a proper subset of antiadditive contexts, which in turn are a proper subset of downward entailing contexts. 2

Yet other logical properties are based on the notion of veridicality (Zwarts 1995): (i) A function f is nonveridical if and only if f(p) does not entail p, for all propositions p. (ii) A function f is antiveridical if and only if f(p) entails ¬p, for all propositions p. Giannakidou (1998 and sequel) argues nonveridicality and antiveridicality to be relevant for NPI-licensing in Greek and other languages.

3

Formally, these finer-grained notions of negativity are defined via the entailments making up De Morgan’s Laws, which negation in classical logic obeys. Generalized to arbitrary functions, the two Laws of De Morgan are: (i) 1. f(x ˄ y) , f(x) ˅ f(y) 2. f(x ˅ y) , f(x) ˄ f(y) Different logical properties can be defined depending on the entailments that are valid. Functions that validate both of de Morgan’s Laws are antimorph. Antiadditive functions validate only the second Law of De Morgan. For merely downward entailing functions, only half of the second Law of De Morgan is valid, i.e., f(x ˅ y) ) f(x) ˄ f(y).

647

648

DORIS PENKA

Intuitively, antiadditivity can be tested for by checking whether wide scope conjunction is equivalent to narrow scope disjunction. This is the case for negative quantifiers like no student, but not, for instance, for quantifiers formed with at most. As illustrated in (28), the entailment is valid in both directions in case of no student, whereas the entailment from wide scope conjunction to narrow scope disjunction does not go through for at most (example (29b) is true, for instance, in a situation where two students gave a class presentation and two students wrote a term paper, whereas (29a) is false in this situation). Quantifiers with at most are therefore not antiadditive and merely downward entailing. (28)

a. No student gave a class presentation or wrote a term paper. , b. No student gave a class presentation and no student wrote a term paper.

(29)

a. At most three students gave a class presentation or wrote a term paper. ) ⇍ b. At most three students gave a class presentation and at most three students wrote a term paper.

Other operators that are downward entailing, but not antiadditive, are the quantifier few and the adverbs seldom and hardly. Besides negative quantifiers, the conjunctions without and before have been argued to be antiadditive and so have the complements of verbs like refuse or deny. Antimorphic operators are sentential negation and expressions like allerminst ‘least of all, not at all’ in Dutch and keinesfalls ‘in no case’ in German. In line with these different degrees of negativity, Zwarts (1998) and van der Wouden (1997) distinguish different classes of polarity items, depending on which negative strength they require in order to be licensed:4 (30) Classes of NPIs (Zwarts 1998; van der Wouden 1997): Weak NPIs: licensed in all downward entailing contexts Strong NPIs: licensed in antiadditive contexts, but not in merely downward entailing contexts Superstrong NPIs: licensed in antimorphic contexts only The distribution of in weeks and yet observed in (25)–(26) above now follows from their status as strong NPI: While they are licensed by negation (an antimorphic operator) and negative quantifiers 4

It should be noted that the terminology is not used consistently in the literature. In contrast with the terminology of Zwarts (1998), which is most often adopted, van der Wouden (1997) calls NPIs that are licensed in antimorphic contexts only “‘strong” and expressions that require an antiadditive context “NPIs of medium strength.” Krifka (1995) uses the term “strong NPI” in yet another sense, namely for emphatic NPIs like stressed ANY. Sometimes the items that Zwarts (1998) calls “strong NPIs” are also referred to as “strict NPIs”, whereas following Horn (1978) the term “strict NPI” is usually used for an NPI that requires its licenser to be in the same clause.

Negative and Positive Polarity Items

(antiadditive operators) they are not licensed by merely downward entailing operators like at most. Other expressions that van der Wouden (1997) argues to require an antiadditive licenser include Dutch ook maar ‘at all, even’ and med en finger aanroeren ‘touch with a finger.’ NPIs that have been classified as requiring an antimorphic licenser are mostly highly idiomatic, such as Dutch mals ‘tender’, pluis (literally ‘plush’), and voor de poes (lit. ‘for the cat’). Another account of the difference between weak and strong NPIs is offered by Gajewski (2011). He argues that strong NPIs are sensitive to non-truth conditional meaning of their licensers while weak NPIs are not. His proposal starts from the observation (see also Krifka 1995) that typical antiadditive licensers coincide with expressions forming endpoints of scales made up of downward entailing expressions. Negative quantifiers, for example, constitute the strongest member of the scale < no, few, not many, not all >. It is also well-known that weaker, nonendpoint scalar elements trigger scalar implicatures to the effect that statements involving their stronger scale-mates are false. The sentences in (31a) for instance, all give rise to the implicature in (31b). (31)

a. {Few / not many / not all} students failed the exam. b. Implicature: Some students passed the exam (= It is false that no student passed the exam).

Gajewski now argues that these positive implicatures interfere with the licensing of strong NPIs. Endpoint scalar items, in contrast, do not give rise to implicatures, which accounts for the contrast in (32). (32)

a. *{Few / not many / not all} students have failed in years. b. *No student has failed in years.

In a similar vein, strong NPIs are argued not to be licensed in the scope of only and in the complement of factive verbs like regret, because these introduce positive presuppositions. Summarizing the approach of Gajewski (2011), the licensing conditions of strong NPIs require them to be in a downward entailing context at all levels of meaning – be it truth conditions, presuppositions or implicatures. Weak NPIs, in contrast, require downward entailment only at the level of the asserted content, and disregard non-truth conditional meaning. Further distinctions among different subtypes of NPIs have been proposed based on the interpretational effects they give rise to. Minimizer NPIs and other NPIs like stressed ANY add emphasis to the utterance they occur in. They moreover contrast with any and ever in their behavior in questions. While any and ever can be used in information seeking questions, minimizer NPIs in questions give rise to a strong negative bias. In contrast to (33a), (33b) feels like a rhetorical question in which the speaker suggests that Bill did not help (Guerzoni 2004).

649

650

DORIS PENKA

(33)

a. Did Bill help you with anything? b. Did Bill lift a finger to help you?

We will return to this contrast between different subtypes of NPIs in Section 27.6, when possible analyses to account for it are discussed. But for now we ignore the more fine-grained distinctions among different types of NPIs and stick to the simplifying assumption that NPIs are licensed in downward entailing environments.

27.5 The Nature of the Licensing Relation Having identified logical properties by which the contexts in which NPIs are licensed can be characterized, the next question to address is how precisely the licensing requirements of NPIs should be formulated. Assuming that downward entailment is the relevant property, the licensing condition on NPIs can be stated in two different ways: (34)

Licensing condition on NPIs: a. An NPI is only grammatical if it is in a downward entailing environment. b. An NPI is only grammatical if it is in the scope of a downward entailing operator.

The two conditions might prima facie seem equivalent – after all, a downward entailing environment is induced by a downward entailing operator. There are, however, important differences, as pointed out by Homer (in press), who provides arguments in favor of formulating the constraint in terms of environments rather than operators. The view that is prevalent in the literature, however, is operator-based, and this raises issues about the structural relation between an NPI and its licenser. Several structural constraints for NPI licensing have been proposed. One was motivated by the observation that in the Germanic languages NPIs cannot occur in subject position of negated clauses, see (35), even though it is in principle possible for a subject to be interpreted in the scope of a clausemate negation. (35)

*Anyone didn’t come.

This observation has led to the claim that the licensing requirement cannot only be formulated with respect to the level of interpretation (in terms of semantic scope), but must also refer to surface syntax, and that an NPI has to be c-commanded by its licenser in the surface syntax. There are, however, systematic exceptions to this restriction. NPIs can be embedded in preverbal subjects, as in the following example from Linebarger (1980). (36)

A doctor who knew anything about acupuncture was not available.

Negative and Positive Polarity Items

De Swart (1998) argues that the restrictions on the surface position of NPIs reduce to pragmatic constraints on inverse scope and shows that when a subject can be interpreted in the scope of clausemate negation, it can also contain an NPI. Another structural constraint that has been proposed to be effective in the licensing of NPIs is the Immediate scope constraint of Linebarger (1987), according to which another semantic operator cannot intervene between an NPI and its licenser. This explains why the following sentence does not have a reading where the universal quantifier every party takes scope in between negation and the indefinite object if it involves any, even though the reading “It was not to every party that she wore earrings” is available if the indefinite object is realized as the bare plural earrings. (37)

She didn’t wear any earrings to every party.

* not ≫ every ≫ any, ✓not ≫ any ≫ every

While Linebarger’s Immediate scope constraint assumes a structural relationship between an NPI and its licenser, other accounts of intervention effects are also compatible with an environment-based approach. Chierchia (2013) argues that scalar implicatures triggered by intervening operators interfere with the licensing of NPIs. As it is the common characteristic of NPI licensers that they induce scale reversal, it is now scalar endpoint expressions like every that give rise to scalar implicatures in their scope. The nonavailable reading “It was not to every party that she wore earrings” for (37), for instance, would yield the positive implicature that she wore earrings to some parties, which interrupts downward entailment.

27.6 The Source of Polarity Sensitivity So far the discussion in this chapter has been mainly descriptive, providing semantic characterizations of different types of NPIs and the environments in which they are licensed as well as formulations of the licensing conditions. But that does not yet answer the question of what it is that makes an expression polarity sensitive. Syntactic accounts of NPI licensing essentially stipulate that NPIs are endowed with a feature that forces them to enter a certain configuration with a licenser (e.g., Klima 1964, Progovac 1994, den Dikken 2002). Szabolcsi (2004) and Collins and Postal (2014) view NPI features as negations in the lexical semantics of NPIs themselves. But such syntactic approaches still do not explain why there are elements in natural language that are sensitive to logical properties of the context in which they appear. Moreover, it is puzzling why it is precisely downward entailment, rather than any other logical property, that is decisive for NPI licensing. These and related questions have gotten into the focus of the study of polarity items since the mid-1990s.

651

652

DORIS PENKA

Focusing on one particular common type of polarity item, low-scalar NPIs, there is an intuitive reason for why it makes sense to use them in downward entailing contexts: In downward entailing contexts, which allow inferences from smaller to larger quantities, the use of an expression denoting a minimal quantity or action results in a strong statement. In (38), for instance, the use of the minimizer NPI lift a finger excludes any kind of help from Bill, even the most minimal activity, and thus results in the strongest possible statement a speaker could make about Bill’s contribution. (38)

Bill didn’t lift a finger to help.

In upward entailing contexts, in contrast, low-scalar NPIs make for such weak statements that they are effectively unusable. It seems then that NPIs can be used precisely when they strengthen an utterance. To formally spell out this intuition, however, has proven challenging. There are two different issues that need to be addressed: First, in what way do NPIs lead to stronger statements? While the answer to this question seems rather straightforward for minimizer NPIs, it is less obvious in the case of other NPIs like any. Second, how does the grammar enforce strengthening, such that NPIs are excluded from occurring in environments in which they do not strengthen the utterance? Different analyses have been proposed to deal with these questions. In a seminal paper, Kadmon and Landman (1993) propose that any induces widening of the quantificational domain. This can be illustrated by way of the following example: (39)

a. I don’t want presents for my birthday. b. I don’t want any presents for my birthday.

In (39a) the quantificational domain of the bare indefinite presents is restricted contextually to include only things that are usually regarded as presents. These contextual restrictions are lifted when any is used in (39b), such that things that might not always count as presents, e.g., flowers and chocolate are also part of the quantificational domain. In this way, the use of any reduces tolerance of exceptions – provided that it occurs in a downward entailing environment. Because more restricted quantificational domains are subsets of the widened domain, and downward entailing contexts support inferences from sets to subsets, an utterance with any in a downward entailing environment entails all possible alternative statements with a more restricted domain of quantification. Domain widening is therefore useful in downward entailing contexts, where it leads to stronger statements. Domain widening has also been argued to be key to understanding why across languages, many indefinite expressions that are NPIs can also be used as Free Choice items (see Hapelmath 1997 for a cross-linguistic survey). Besides any-indefinites in English, German indefinites formed with irgend- also lead this double life.

Negative and Positive Polarity Items

(40)

a. Du kannst irgendeine Farbe nehmen. Free Choice (German) you can any color pick ‘You can pick any color.’ b. Niemand hat irgendetwas no one has anything ‘No one said anything.’

gesagt. said

NPI

It has been a long-standing question whether a unified analysis of NPI and Free Choice uses of any is possible. Early unified accounts (Reichenbach 1947: section 21; Quine 1960: section 29) treat any as a universal quantifier that obligatorily takes wide scope. Because of the semantic equivalence of a universal quantifier taking scope above negation and an existential quantifier in the scope of negation (∀¬ , ¬∃) this seems a viable option when NPI any occurs in a strict negative (i.e., antiadditive or antimorphic) context. But for NPI any in other contexts, the equivalence does not hold, as shown in example (41) from Fauconnier (1979: 297–298). (41)

I wonder if Susan married anybody. I wonder if there is an x such that Susan married x. NOT: For every x: I wonder if Susan married x.

It is clear that in such cases any is interpreted as an existential in the scope of its licenser. Thus NPI any cannot generally be analyzed as wide scope universal. Instead, an existential analysis seems warranted. Other attempts have tried to reduce Free Choice any to generic uses of indefinites (e.g., Kadmon and Landman 1993, Lahiri 1998), but have proven equally problematic. Under the domain widening approach, being an NPI and being a Free Choice item represent two sides of the same coin, corresponding to two different strategies in which domain widening can be exploited for communicative purposes. While the wish to make a stronger statement is the reason for a speaker to use an NPI in a downward entailing context, strengthening is not the only possible use of a domain widening indefinite. Kratzer and Shimoyama (2002) argue that another reason is the wish to prevent the hearer from drawing unintended exhaustivity inferences. By instructing the hearer to consider a wide domain of quantification, the speaker could signal that he does not want to rule out any conceivable option. In the German example with irgendein in (40a) above, for instance, the speaker conveys that there are no restrictions regarding color and any color is a conceivable option. Since the exhaustivity inferences that nondomain widening indefinites would give rise to (namely that an interpretation with a wider domain would not be a possible option) arise as implicatures and are limited to upward entailing contexts, the antiexhaustivity effect of domain widening indefinites does not arise in downward entailing contexts. This explains the flip-flop behavior of any- and irgendindefinites: In upward entailing contexts (with a modal component)

653

654

DORIS PENKA

domain widening is used to convey antiexhaustivity, while in downward entailing contexts domain widening strengthens the utterance. See Chierchia (2013) for a detailed proposal for a domain widening analysis of NPI and Free Choice uses of indefinites, which also accounts for different flavors of Free Choice indefinites as well as for the fact that crosslinguistically not all indefinite NPIs can also serve as Free Choice items. While domain widening provides a widely adopted answer to the question in what sense indefinite NPIs make a statement stronger, the second question – why strengthening is obligatory for NPIs – remains open. (Kadmon and Landman essentially stipulate a strengthening requirement, which has been criticized as being noncompositional.) One type of approach assumes that NPIs obligatorily associate with the focus sensitive particle even, which is covert in the Germanic languages but overt in other languages such as Hindi (Lahiri 1998). The presupposition of even requires the proposition to which it attaches to be less likely than the focus alternatives. Since NPIs denote elements located at the bottom of their scale, i.e., very general properties, this presupposition is satisfied in downward entailing contexts, but cannot be fulfilled if the NPI occurs in an upward entailing environment. In nonmonotone contexts, whether the presupposition is fulfilled depends on the context, which accounts for the contextdependency of NPIs in nonmonotone contexts illustrated in (22) and (23) above (Crnicˇ 2014). Some theories of NPI licensing assume that there are two different types of NPIs: emphatic NPIs – minimizers such as lift a finger and stressed ANY – which involve (a covert operator like) even, and others like unstressed any and ever, which do not associate with even (Heim 1984, Krifka 1995, Guerzoni 2004, Chierchia 2013). The presence of covert even is argued to account for the rhetorical effect of the first type of NPIs in questions, which was illustrated in (33) above: The presupposition triggered by even reduces the set of possible answers to the question to the singleton set containing the negative answer (Guerzoni 2004). Another type of approach relates NPI licensing to the mechanism underlying scalar implicatures (Krifka 1995, Chierchia 2013). Similar to the evenbased account, this approach also assumes that NPIs obligatorily activate alternatives that are more informative. These alternatives are factored into meaning via an exhaustification operator, whose effect is to deny all more informative alternative propositions. For utterances with NPIs in upward entailing contexts, this results in a contradiction: It is simply impossible for a more general statement to be true while at the same time all possible more specific statements are false. For NPIs in downward entailing contexts, in contrast, the alternative propositions are all logically weaker and exhaustification does not have an effect. A challenge that implicaturebased approaches face is the question why unlicensed NPIs lead to ungrammaticality, rather than just pragmatic infelicity. This has partly motivated the recently popular assumption that scalar implicatures are generated by

Negative and Positive Polarity Items

a mechanism that is part of the core grammar system rather than being derived by pragmatic reasoning (Chierchia 2013). Finally, it should be mentioned that there are licensing contexts as well as types of NPIs for which strengthening-based approaches cannot account. This concerns the licensing of NPIs in questions as well as understating NPIs. As in unifying NPI and Free Choice uses of any, the key to accounting for these NPI uses seems to be realizing that there are potentially different reasons for why a certain utterance would be more useful and thus preferred over alternatives. The most basic reason is that the utterance is more informative because it logically entails its alternatives. This entailment-based notion of strength is what most accounts of NPIs build on. But another way in which an utterance is more useful than an alternative is found in questions: Using an NPI in a question has the effect of making possible answers more informative and thus maximizes the utility of the question (Krifka 1995, van Rooy 2003). To account for understating NPIs, it has been argued that they can be used by a speaker to avoid making a strong claim in a situation where it might be disadvantageous for the speaker to commit to a strong proposition (Israel 2001, van Rooy 2003). For instance, by uttering (42a) rather than (42b) a speaker makes a weaker claim about the insufficiency of John’s intelligence and thus appears more polite. (42)

a. John is not all that clever. b. John is not clever.

In sum, recent investigations of NPIs have identified two key ingredients for an explanatory account of NPI licensing: First, alternatives play a crucial role for polarity sensitivity. NPIs trigger alternatives to which an utterance containing an NPI is compared. The second ingredient is a mechanism ensuring that these alternatives are factored into the meaning computation and give rise to ungrammaticality whenever an NPI does not occur in a licensing context.

27.7 Positive Polarity Items While there is a rich body of work on NPIs, PPIs have only recently started to gain attention (recent discussions include Nicolae 2012, Spector 2014, Zeijlstra 2017 and Homer in press). The null hypothesis would be that PPIs are the exact opposites of NPIs in terms of distribution and analysis. In this vein, van der Wouden (1997) proposes the following classification of PPIs, which is the mirror image of the classification of NPIs in (30) above. (43)

Classification of PPIs (van der Wouden 1997): Strong PPIs: anti-licensed in all downward entailing contexts

655

656

DORIS PENKA

PPIs of medium strength: compatible with downward entailing contexts, but banned from anti-additive contexts Weak PPIs: anti-licensed in antimorphic contexts only (scope of clausemate negation) For illustration, consider English some, which is banned from the scope of (clausemate) negation and negative quantifiers, but acceptable in the scope of merely downward entailing operators such as at most five, and is thus classified as a medium strength PPI: (44) a. I didn’t call somebody. *not ≫ somebody, ✓somebody ≫ not b. No one called somebody. *no one ≫ somebody, ✓somebody ≫ no one c. At most five girls called somebody. ✓at most five ≫ somebody As further examples of PPIs of medium strength van der Wouden lists the adjectival modifier rather, and Dutch een beetje ‘a bit’, nogal ‘rather’, maar ‘but’, and somes ‘sometimes.’ Examples of weak PPIs in English are already and still and their Dutch counterparts al and nog. As examples of strong PPIs, finally, van der Wouden cites Dutch allerminst ‘not all all’, inderdaad ‘indeed, actually’ and verre van ‘far from.’ There are, however, empirical facts that blur this nice parallel between NPIs and PPIs. First, there are expressions that appear to be NPIs and PPIs simultaneously. An example for such a bipolar item (in the terminology of van der Wouden 1997) is the German temporal adverb jemals ‘ever’, which is restricted to downward entailing contexts and thus a weak NPI, see (45a). At the same time, jemals cannot occur with a clausmate sentential negation, see (45b), and would thus be classified as a weak PPI. (45)

a. *(Ho¨chstens) fu¨nf Personen haben jemals dieses At most five people have ever this Buch gelesen. (German) book read ‘At most five people have ever read this book.’ b. *Peter hat nicht jemals dieses Peter has not ever this ‘Peter hasn’t ever read this book.’

Buch book

gelesen. read

But it seems that there is an independent reason why jemals is banned from occurring in the scope of negation. In German, indefinite expressions cannot be immediately adjacent to sentential negation and have to be realized as a negative indefinite instead, as illustrated for the indefinite determiner ein ‘a’ in (46). (46)

a. Ich habe { *nicht ein/ I have not a ‘I didn’t buy a book.’



kein} no

Buch gekauft. book bought.

Negative and Positive Polarity Items

This is, however, a surface syntactic constraint, since indefinites can be interpreted in the scope of negation as long as they are not adjacent to negation in the surface syntax, e.g., when they occur in topic position as in (47). With the intonation of the so-called topic-focus accent, which forces the topicalized indefinite to reconstruct into the scope of negation (Bu¨ring 1997), (47) means something like “As for books, I didn’t buy any.” (47)

Ein a

Buch book

habe have

ich I

nicht not

gekauft. bought

This shows that indefinites in German can be interpreted in the scope of an antimorphic operator. The fact that indefinites in German are banned from occurring with clausmate negation can thus not be due to polarity sensitivity. Instead, the distribution of general and NPI indefinites is best explained in terms of morphological blocking: Because there is a more specific morphological form for this use, i.e., negative indefinites, the less specific forms, i.e., general and NPI indefinites, are blocked from occurring in this context (see Hoeksema 1999 for Dutch). But even if attention is restricted to the prime example of a PPI, English some, it turns out that PPIs and NPIs are not in complementary distribution. There are environments where both PPIs and NPIs can occur. In (48), for example, the polarity item is separated from the (anti)licensing negation by a clause boundary and both some and any can be interpreted within the embedded clause (Ladusaw 1979). (48)

I don’t think that Sam called {someone / anyone}.



not ≫ someone

In certain configurations, PPIs can even be interpreted in the scope of a clausemate antilicenser. This is the case if both the PPI and the antilicenser are in the scope of another operator, which is said to rescue the PPI (Baker 1970). (49)

I am surprised that Sam didn’t call someone.



surprise ≫ not ≫ someone

Szabolcsi (2004) argues that PPIs of the some-type are rescued in those contexts where standard NPIs such as any and ever are licensed. In line with this, she proposes that the antilicenser and the PPI in its scope compose to form a derived NPI, which is assumed to be subject to the same licensing conditions as lexical NPIs. Taking a different perspective on the distribution of some-PPIs and their relation to NPIs, Nicolae (2012) argues that PPIs can be integrated into the alternative-based account of polarity sensitivity developed by Chierchia (2013). As in the case of NPIs, PPIs are assumed to activate alternatives that are obligatorily factored into the meaning via exhaustification. The crucial difference is argued to be in the nature of the alternatives: While any-NPIs widen the quantificational domain by activating subdomain alternatives,

657

658

DORIS PENKA

which lead to strengthening in downward entailing contexts, some-PPIs activate superdomain alternatives, which result in strengthening in upward entailing environments. Similar analyses in terms of weaker, i.e., more general, alternatives have been proposed for maximizer PPIs like tons of by Krifka (1995). For the adjectival modifiers rather and pretty, Krifka proposes that the alternatives are interpretations of the adjective at weaker precision standards. To conclude the discussion of PPIs, I would like to remark that the expressions that have been subsumed under the label “positive polarity item” constitute a much more heterogeneous class than NPIs. The factors governing the distribution of various different types of expressions that somehow resist being in the scope of negation include blocking by more specific expressions and obligatory widest scope for speech-act-related reasons. Recent alternative-based accounts of PPIs, however, analyze certain PPIs as true counterparts of NPIs.

27.8 Conclusion This chapter reviewed some lines of inquiry on polarity sensitive items that have been developed within the framework of generative grammar over the past 50 years. Because there is an enormous body of work on this topic, only the approaches and ideas that I take to be most influential could be considered. While a lot of progress has been made over the past decades, a number of questions remain open and there is still no unified analysis that can account for all types of NPIs and PPIs. As the discussion in this chapter showed, syntax, semantics, and pragmatics closely interact in the constraints that govern the distribution of polarity items. The investigation of polarity items can thus serve as a window into the interaction between these components of grammar and into the architecture of grammar more generally.

References Baker, C. L 1970. “Double Negatives,” Linguistic Inquiry 1: 169–186. Biberauer, T. and H. Zeijlstra 2012. “Negative concord in Afrikaans: Filling a typological gap,” Journal of Semantics 29: 345–371. Brandtler, J. 2012. The Evaluability Hypothesis: The Syntax and Semantics of Polarity Item Licensing in Swedish. Amsterdam: John Benjamins. Bu¨ring, D. 1997. The Meaning of Topic and Focus – The 59th Street Bridge Accent. London: Routledge. Chierchia, G. 2013. Logic in Grammar. Oxford University Press. Collins, C. and P. M. Postal. 2014. Classical NEG Raising: An Essay on the Syntax of Negation. Cambridge, MA: MIT Press.

Negative and Positive Polarity Items

Crnicˇ, L. 2014. “Non-monotonicity in NPI Licensing,” Natural Language Semantics 22:169–217. de Swart, H. 1998. “Licensing of negative polarity items under inverse scope,” Lingua 105: 175–200. den Dikken, M. 2002. “Direct and Indirect Polarity Item Licensing,” Journal of Comparative Germanic Linguistics 5: 35–66. Fauconnier, G. 1975. “Pragmatic Scales and Logical Structure,” Linguistic Inquiry 6: 353–376. Fauconnier, G. 1979. “Implication reversal in a natural language.” In F. Guenther and S. J. Schmidt (eds.), Formal Semantics and Pragmatics for Natural Languages. Dordrecht: Reidel: 289–301. Gajewski, J. R. 2011. “Licensing Strong NPIs,” Natural Language Semantics 19.2: 109–148. Giannakidou, A. 1998. Polarity Sensitivity as (Non)Veridical Dependency. Amsterdam: John Benjamins. Guerzoni, E. 2004. “Even-NPIs in Yes/No Questions,” Natural Language Semantics 12.4: 319–343. Haspelmath, M. 1997. Indefinite Pronouns. Oxford University Press. Heim, I. 1984. “A Note on Negative Polarity and Downward Entailingness.” In C. Jones and P. Sells (eds.), Proceedings of the 14th North Eastern Linguistic Society meeting. Amherst: GLSA: 98–107. Hoeksema, J. 1999. “Blocking Effects in the Expression of Negation,” Leuvense Bijdragen 80: 403–423. Homer, V. in press. “Domains of Polarity Items,” Journal of Semantics. Horn, L. R. 1978. “Remarks on Neg-raising.” In P. Cole (ed.), Syntax and Semantics 9: Pragmatics. New York: Academic Press: 129–220. Horn, L. R. 1985. “Metalinguistic Negation and Pragmatic Ambiguity,” Language 61.1: 121–174. Horn, L. R. 2001. “Flaubert triggers, squatitive negation, and other quirks of grammar.” In J. Hoeksema, H. Rullmann, V. Sanchez-Valencia, and T. van der Wouden (eds.), Perspectives on Negation and Polarity Items. Amsterdam: John Benjamins: 173–200. Israel, M. 1996. “Polarity sensitivity as lexical semantics,” Linguistics and Philosophy 19: 619–666. Israel, M. 2001. “Minimizers, maximizers and the rhetoric of scalar reasoning,” Journal of Semantics 18: 297–331. Kadmon, N. and F. Landman 1993. “Any,” Linguistics and Philosophy 16.4: 353–422. Klima, E. 1964. “Negation in English.” In J. A. Fodor and J. J. Katz (eds.), The Structure of Language. Englewood Cliffs, NJ: Prentice-Hall: 246–323. Kratzer, A. and J. Shimoyama 2002. “Indeterminate pronouns: The view from Japanese.” In Y. Otso (ed.), Proceedings of the Third Tokyo Conference on Psycholinguistics. Tokyo: Hituzi Syobo. Krifka, M. 1995. “The semantics and pragmatics of polarity items,” Linguistic Analysis 25: 209–258.

659

660

DORIS PENKA

Ladusaw, W. A. 1979. Polarity Sensitivity as Inherent Scope Relations. Doctoral dissertation, University of Texas, Austin. Published 1980 by Garland, New York. Ladusaw, W. A. 1992. “Expressing negation.” In C. Barker and D. Dowty (eds.), Proceedings of SALT II, Ohio State Working Papers in Linguistics 40. Columbus: Ohio State University: 237–259. Ladusaw, W. A. 1997. “Negation and polarity items.” In S. Lappin (ed.), The Handbook of Contemporary Semantic Theory. Oxford: Blackwell: 321–341. Lahiri, U. 1998. “Focus and negative polarity in Hindi,” Natural Language Semantics 6: 57–123. Laka, I. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. Doctoral dissertation, MIT. Linebarger, M. C. 1980. The Grammar of Negative Polarity. Doctoral dissertation, MIT. Linebarger, M. C. 1987. “Negative polarity and grammatical representation,” Linguistics and Philosophy 10: 325–387. Nicolae, A. 2012. “Positive polarity items: An alternative-based account.” In A. Aguilar Guevara, A. Chernilovskaya, and R. Nouwen (eds.), Proceedings of Sinn und Bedeutung 16: 475–488. Penka, D. 2011. Negative Indefinites. Oxford University Press. Quine, W. van Orman 1960. Word and Object. Cambridge, MA: MIT Press. Reichenbach, H. 1947. Elements of Symbolic Logic. New York: The Free Press. Spector, B. 2014. “Global Positive Polarity Items and Obligatory Exhaustivity,” Semantics and Pragmatics 7.11: 1–61. Szabolcsi, A. 2004. “Positive polarity – negative polarity,” Natural Language and Linguistic Theory 22.2: 409–452. van der Wouden, T. 1997. Negative Contexts: Collocation, Polarity and Multiple Negation. London and New York: Routledge. van Rooy, R. 2003. “Negative polarity items in questions: Strength as relevance,” Journal of Semantics 20: 239–273. von Fintel, K. 1999. “NPI Licensing, Strawson-entailment, and Context Dependency,” Journal of Semantics 16: 97–148. Zeijlstra, H. 2017. Universal quantifier PPIs. Glossa: A Journal of General Linguistics 2.1: 91.1–25. DOI:https://doi.org/10.5334/gjgl.220. Zwarts, F. 1995. “Nonveridical contexts,” Linguistic Analysis 25: 286–312. Zwarts, F. 1996. “A hierarchy of negative expressions.” In H. Wansing (ed.), Negation: A Notion in Focus. Berlin: Mouton de Gruyter: 169–194. Zwarts, F. 1998. “Three Types of Polarity.” In F. Hamm and E. Hinrichs (eds.), Plurality and Quantification. Dordrecht: Kluwer: 177–238.

Chapter 28 Grammatical Reflexes of Information Structure in Germanic Languages Caroline Féry

28.1 Introduction The effect of information structure (IS) on prosody and syntax in Germanic languages cannot be overestimated. Focus, topic, and givenness, which are related to the way information is stored in human memory and organized in communication, have a profound influence on the prosodic and syntactic structure of sentences. They can add and delete pitch accents or words, use focus and discourse particles, and change the word order of sentences in significant ways. These properties have numerous consequences for the architecture of grammar. In this handbook, IS is relevant in the discussions of nearly half of the chapters, and this chapter is intended to review its effects on grammar. We will see that different Germanic languages use similar strategies in implementing IS in grammar – especially as compared to other families of languages –but that they also differ from each other in the detail of IS’s grammatical effects. Before turning to these strategies, Section 28.2 presents a summary of the notions of information structure as they are used in this chapter. Section 28.3 discusses the effects of IS grammar on prosodic structure (Section 28.3.1), on syntax (Section 28.3.2), and on focus particles (Section 28.3.3). Section 28.4 offers a conclusion.

28.2 Notions of Information Structure Three pairs of information structural notions are present in every sentence: focus-background, given-new, topic-comment. For each of these pairs, one Thanks to Mary O’Brien for reading the chapter, and making significant corrections.

662

CAROLINE FÉRY

of them is considered the default and the other one the complementary part. Part of the sentence is focused, and the remainder is backgrounded; part of it is given and the remainder is new; part of it is the topic and the remainder is the comment.1 It will be shown that these pairs are orthogonal to each other, and that all three are necessary, thereby falsifying attempts to collapse givenness with background, or focus with newness.

28.2.1 Focus/Background Consider first the pair, focus-background. Either the entire sentence is in focus, or only part of it is. The parts of the sentence that are not focused form the background. The definition of focus used in this chapter is based on the theory of alternative semantics of focus proposed by Rooth (1985, 1992, 2016): Focus assigned to a linguistic expression α always indicates that there are alternatives to α relevant in the current discourse. Putting it differently, anything that does not indicate any alternatives to α should be called background. Krifka (2008: 247) defines focus as in (1).2 (1)

Focus Focus indicates the presence of alternatives that are relevant for the interpretation of linguistic expressions.

The semantic use of focus affects the truth-conditional aspects of the discourse in the sense that the position of the focus changes the meaning of the sentence. The sentences in (2) have different presupposition structures: the first reading can be paraphrased in the following way: in (2a), when she makes a call, Justine calls somebody other than her father; and the second reading, in (2b), means: Whatever relationship Justine has with her father, it does not include calling him. Subscripted F indicates focus, and capital letters a pitch accent. (2)

a. Justine never calls [her F A T H E R ]F. b. Justine never [C A L L S ]F her father.

Sentence (3a) can be uttered by speaker A as an answer to a question like “What happened?” Speaker B, who knows better, can correct speaker A’s utterance by changing the person who ordered pizza, the result being a narrow focus on Alicia. The alternatives induced by focus are contextually restricted, as for instance the individuals in (3c), to the persons in an office who have worked late and are now hungry. Thus the alternative set 1

A sentence can also be entirely focused, in which case there is no given part. Or the whole sentence can be new (or given). But there is no sentence consisting entirely of a topic. A sentence consisting of only backgrounded material is exceptional.

2

Focus can also be defined on the basis of what information it adds to the so-called Common Ground (CG). The CG shapes the background to which new information is added to the information that is mutually known (or believed) to be shared by a speaker and an addressee, and that is continuously modified in communication. We do not elaborate the CG in this article, but see Stalnaker (1974) and Krifka (2008) for theoretical approaches to the CG.

Information Structure in Germanic Languages

consists of the individuals who may have ordered pizza in the particular context of this utterance. (3b) presupposes that somebody ordered pizza. Ordered pizza is the noncontrastive part of the sentence, the background, which is accepted by speaker B and which is complementary to the focused part, i.e., the person who did the ordering. Rooth’s intuition is that the focus contrasts with all possible right and wrong answers to a contextually given question – its ‘Hamblin denotation’ (Hamblin 1973). (3)

a. A: [Gregory ordered P I Z Z A ]F b. B: No, [A L I C I A ]F ordered pizza. c. Alternative set: {G R E G O R Y ordered pizza, A L I C I A ordered pizza, F A T I M A ordered pizza}

In the Roothian version of the theory, focus in English has a uniform semantic import related to “contrast” within a set of alternatives: One of the alternatives is the correct one, the others are wrong. In this survey, contrast is not a special information structural notion, because every focus – and every topic – introduces some contrast, relative to the other members of the alternative set. Thus, contrast is an inherent feature of focus.3 The classic use of focus is to highlight the part of an answer that corresponds to the wh-part of a constituent question, as illustrated in (4a)–(4b), usually by a pitch accent. The wh-question does not need to be overt: It can be covert, and structured by implicit questions. Focus is generally sub-sentential, that is, it only affects one part of the sentence, the other parts being backgrounded (except in so-called ‘fragment sentences’ in which the backgrounded material is elided). But a sentence can also contain more than one focus; (4c) is a dual focus sentence (see Wang and Fe´ry 2018, for the prosodic correlates of such sentences in German). (4)

a. A: What did Mary order? B: She ordered [P I Z Z A ]F

b. A: What did Mary do? B: She [ordered P I Z Z A ]F

c. A: A: Who ordered what? B: [M A R Y ]F ordered [P I Z Z A ]F

If the accent is not on the item requested via a wh-question, the result is infelicitous, as illustrated in the dialogue (5), where Alicia is accented and ordered cake is deaccented. In this case, a backgrounded constituent (here Alicia, a given constituent) is assigned a pitch accent, and the focus (cake) is deaccented. (5)

3

A: What did Alicia order? B: #A L I C I A ordered [cake]F.

Other authors assume that contrast is an IS notion on its own (see, for example, Vallduvi and Engdahl 1996), including contrasting focus and contrasting topic. Here it is proposed that both of them are inherently contrastive in the sense that both trigger an alternative set, but they do so in different ways.

663

664

CAROLINE FÉRY

In a spoken sentence, the hearer generally knows which element is focused because the context specifies what element is new in the sentence. The reason for adding a pitch accent on the focus may be related to attentional factors. The mechanism consisting of accenting the focus and deaccenting the background facilitates processing of new information. Furthermore, pitch accents are present for purely syntactic purposes, and since they are part of the prosodic structure of Germanic languages anyway, it may be simply more economical to use them for signalling the focus structure as well. Finally, notice that focus cannot be conflated with new, and background is not the same as given. Selection and correction focus in (6) and (7) respectively, show that a given constituent can be focused rather than backgrounded. (6)

Selection focus A: Did Louise’s father or mother pick her up? B: It was [her M O T H E R ]F.

(7)

Confirmation focus A: Did Alicia order a pizza? B: Yes, everybody thought Gregory did it, but it was indeed [A L I C I A ]F.

28.2.2 Given/New The second pair of information structural notions is given/new. If a referent is new, it means that it has been inactive in the conscience of the listener (or at least the speaker supposes it to be inactive) at the point of its introduction into the discourse. If it is given, it is already active in the consciousness of the interlocutors when it is mentioned.4 Givenness is divided into text-givenness (previously mentioned in the discourse) and context-givenness (contextually salient). See (8) for a definition. (8)

Givenness A referent or part of a sentence is given if it is anaphoric to a constituent mentioned previously in the discourse, if it is entailed by the previous discourse, or it is salient in the context.

For Schwarzschild (1999) and for Halliday (1967/1968), givenness is entailed by the preceding discourse. See also Rochemont (2016) for a presuppositional account of salience-based givenness based on entailment and coreference. For these authors, givenness is not a gradient property of referents. By contrast, Prince (1981) and Baumann & Riester (2013) have proposed hierarchies of givenness. A constituent can be more or less given. The hierarchy in (9) comes from Prince (1981): Going from left to right, givenness increases. 4

According to Clark and Haviland (1977: 3), given information is “information [the speaker] believes the listener already knows and accepts as true,” and new is “information [the speaker] believes the listener does not yet know.”

Information Structure in Germanic Languages

(9)

Assumed familiarity: Unanchored brand-new < Anchored brand-new < Non-containing inferable < Containing inferable < Textually evoked < Situationally evoked

According to this hierarchy, a given constituent needs not to have been mentioned in the discourse, or to be situationally available. It can also be accessible or predictable. Examples of accessibility (or inferability, or bridging) appear in (10), adapted from Prince. Many concerts involve a singer, and by mentioning concert in (10a), singer may become cognitively available (inferable or accessible). In the second part of (10b), the singer has been mentioned, and is given (textually evoked). But things may be more complicated. In (10a), singer is not necessarily deaccented. Deaccenting is a device that is used when the referent is available. But this availability is a matter of how we share knowledge of the world. If the interlocutors were discussing symphonies just before, singer is definitely not available in (10a). The subscripted G indicates givenness. (10)

a. Max went to a concert and the singer was hoarse. b. Max listened to the singer and noticed that the singerG was hoarse.

Givenness can also involve co-referent expressions, as illustrated in (11). (11)

Leila listened to what the new president had to say and found the idiot laughable.

Rochemont (1986, 2016) cites examples of entailments like the following one illustrating that a hyponym can trigger givenness of its hypernym but the contrary is not true. (12)

a. John traps gorillas and he also T R A I N S animals. b. #John traps animals and he also T R A I N S gorillas.

28.2.3 Topic/Comment The last pair of information structural notions is topic-comment. A “topic” is analysed here as a referent that the remainder of the sentence is about, possibly contrasting with other referents under discussion, and necessarily followed by a comment, which itself contains a focus constituent.5 A sentence can have a topic or not. Sentences divided into a topic and a comment are called “categorical”, and sentences consisting of a comment only are called “thetic”. We return to this distinction at the end of this subsection. The topic has often, but not necessarily, been previously introduced into the discourse. It can be new or given, again showing that the three IS pairs

5

Steedman (2000) speaks of “theme” which is largely equivalent to “topic.”

665

666

CAROLINE FÉRY

of notions cannot be conflated. A topic constituent is characterized as in (13); see also Reinhart (1981), Jacobs (2001) and Krifka (2008: 265) for alternative definitions. (13)

Topic A topic is a denotation of a referential expression about which the remainder of the sentence expresses a proposition.

Consider the following examples containing aboutness topics. In (14) B, the topic is the beans and in (14)C, it is the meat. Top stands for topic, and Com for comment. In both (14)B and (14)C, the entire VP is new and focused. (14) A: What a delicious smell! How did you prepare the food? B: [The C: [The

B E A N S ]T O P M E A T ]T O P

[were cooked together with the [was cooked together with the

M E A T ]C O M B E A N S ]C O M

Besides aboutness topics, it is useful to further distinguish between contrastive topics and frame-setting topics. Contrastive topics resemble focus. Consider the example in (15)B, where my younger daughter and my older daughter are topics and contrast with each other. In each clause, the entire subject is the contrastive topic. Additionally, within each of them, younger and older are contrastively focused, demonstrating that a topic can have an embedded focus. (15) A: What do your children do? B:

[My [[Y O U N G E R ]F D A U G H T E R ]]T O P [studies L A W ]C O M , and [my [O L D E R ]F daughter]T O P [wants to travel to B R A Z I L ]C O M .

According to Bu¨ring (2003, 2016), Tomioka (2010), and Constant (2014) among others, a contrastive topic elicits a set of alternative propositions, which are explicitly not used for an exhaustive answer. A contrastive topic thus deliberately leaves aside other referents that are relevant in the context. The actual utterance does not provide all the information that is expected, as is illustrated in (16). This is why we often find contrastive topics using a strategy of incremental answering, as exemplified, in which an issue is split into sub-issues, as in (15). See Roberts (1996) for this proposal. In (16), there may be other relevant information pertaining to what happened. Such a partial answer to a question may be the best the speaker can come up with in a particular context. “CTop” stands for “contrastive topic.” (16) A: What happened while I was away? B:

[The C A T ]C T O P [[lost its T A I L ]F]C O M

As for the last kind of topic mentioned here, the frame-setting topic is used “to limit the applicability of the main predication to a certain restricted domain” (Chafe 1976). The use of such a topic implies that there are other aspects for which other predications might hold. In this respect they are

Information Structure in Germanic Languages

similar to contrastive topics, as they too split up a complex issue into subissues. “FrTop” stands for “frame-setting topic.” (17) A: How is Pamela? B: [Healthwise/As for her health]FrTop, she is [F I N E ]F

Turning now to thetic sentences, these are sentences lacking a topic. The most remarkable examples of thetic sentences are discourse-new intransitive ones, denoting a whole event. The nuclear accent is falling on the unique argument which happens to be the subject, so that the accent pattern of these sentences is unremarkable (see Fe´ry 2011 for German); see the examples in (18). (18)

a. [Your H A I R ’s on fire]C O M b. [My H U S B A N D is gone]C O M c. [The C A R broke down]C O M

After this rather cursory introduction to the information-structural notions, the grammatical means of expressing them are elaborated in Section 28.3

28.3 Grammatical Reflexes of IS In the remainder of the chapter, the grammatical reflexes of IS in Germanic languages are summed up. All Germanic languages rely on prosodic and syntactic reflexes for the expression of IS, and all have focus particles. For instance, all of them use pitch accents, but pre- and post-nuclear pitch accents can be more or less compressed or even deleted. Germanic languages differ even more with respect to their syntactic reflexes. Presence versus absence of word order flexibility, verb-second versus verb-final properties, scrambling, A- and A-bar movement are all factors that can greatly influence the role of syntax in the expression of IS. Finally, focus particles play a role in all languages, but they have different properties across the different languages.

28.3.1 Prosodic Correlates of Information Structure The main reflex of focus in all Germanic languages is a prominent pitch accent on the so-called focus exponent, at least in their standard varieties.6 This strategy goes hand in hand with the reduction of prominence of the given parts of the sentence, especially in the post-nuclear position; i.e., there is deaccenting or compression of the fundamental frequency (F0) in 6

Nonstandard varieties of English as well as English-based creoles may have different prosodic properties from British, American, or Australian varieties: Deaccenting and the form of pitch accents is especially affected by surrounding or source languages.

667

668

CAROLINE FÉRY

this region (see O’Brien, Chapter 8). Accentuation and deaccentuation are ways to intensify the difference between focus and given. As for the topic, it is phrased individually, generally at the beginning of the sentence and it is bounded by a boundary tone. As a result, it is prosodically separated from the comment. Let us first take a look at the prosodic structure on which phrasing, accenting, and deaccenting take place before we examine the prosodic correlates of IS in more detail.

28.3.1.1 Prosodic Structure The representation of focus-related prosodic prominence goes through the elaboration of a prosodic structure erected by mapping the syntactic structure of a sentence onto a corresponding prosodic structure, as illustrated in (19). The higher prosodic constituents, prosodic words (ωwords), prosodic phrases (Φ-phrases), and intonation phrases (ι-phrases), are organized hierarchically. They roughly correspond to syntactic constituents of the size of a grammatical word, a syntactic phrase and a clause respectively. These constituents are derived from the syntactic structure, but rhythmic factors can reshape them, as can IS. It is assumed here that prosodic constituents can be recursive. In (19), there are two layers of ω-words and two layers of Φ-phrases. (19)

Prosodic hierarchy ( ) ( )( ( )) ( ( )) (Elena) (watched) ((Breaking) (Bad)) (last) (Monday)

ι-phrases Φ-phrases ω-words

All prosodic constituents have a head in the form of a metrical prominence, as shown in (20). Metrical prominence reflects the hierarchical relationship between the prosodic constituents at each level, represented by the relative heights of the grid columns. (20)

Metrical prominence ( (× ) ( (× ) ( ( (× ) (× ) ( (× ) (× ) (× ) Elena watched Breaking

× ) × ) × ) (×) Bad

( ( (× ) (× ) last

× ) × ) ( × ) ( × ) ( × ) Monday

ι-phrases Φ-phrases Φ-phrases ω-words ω-words

Each prosodic constituent obeys the principle of Culminativity, formulated in (21) which posits that there is exactly one metrical head in each higherlevel prosodic constituent.7

7

Hyman (2006: 231) restricts Culminativity, formulated as in (i) to the lexical domain. (i) Culminativity: Every lexical word has at most one syllable marked for the highest degree of metrical prominence.

Information Structure in Germanic Languages

(21)

Culminativity A prosodic domain (ω-word, Φ-phrase, ι-phrase) has a unique head reflected as a metrical prominence. This head attracts the main pitch accent in its domain.

Besides a syntactically motivated prosodic pattern and Culminativity, the metrical grid also reflects rhythm, where rhythm tends to create wellbalanced prosodic patterns, that is, maximal alternation between weak (less prominent) and strong (more prominent) constituents. A further factor influencing the metrical grid is IS, which tends to create prosodic patterns, thereby helping the speakers to process IS. In line with all these principles, but especially in line with Culminativity, there is only one maximally prominent locus of stress, called the nuclear accent (Chomsky and Halle 1968). The highest prominence in such a sentence is assigned by default to the most prominent position of the final Φ-phrase in the ι-phrase. The pitch accent associated with the highest prominence can also project to any category dominating the accented word or phrase, provided that it sits on the projecting branch in the syntactic structure (see Arregi 2016 for a summary of focus projection). In (19), it is on Monday.

28.3.2 The Role of Focus and Givenness Changing the focus-background structure of a sentence may imply changing the prominence structure. For illustration, let us take a look at a typical Germanic pattern with an Afrikaans sentence in (22). Afrikaans exhibits prototypical Germanic prosody. Assuming a recursive prosodic structure, the direct object is phrased alone but it is also part of the larger Φ-phrase containing the VP. It is thus phrased together with both parts of the verb, the auxiliary and the participle, and the nuclear pitch accent is on the embedded Φ-phrase containing the direct object. In the Φ-phrase consisting of the VP, the direct object carries the main accent, and within the direct object, the embedded Φ-phrase contains the prepositional phrase (see Fe´ry 2011 for a detailed analysis for German, another verbfinal language). Afrikaans is a verb-final language, and due to this recursive prosodic structure, the final verb is unaccented. L*HΦ (22)

LH*

H* LΦ



voe¨ls)Φ)Φ gekoop)Φ]ι (Jy)Φ ( het (’n boek (oor You have a book about birds bought ‘You bought a book about birds.’

The melodic pattern of this sentence is illustrated in Figure 28.1.8 It has three pitch accents, one on jy ‘you’, one on boek ‘book’ and one on voe¨ls ‘birds’.9 The second pitch accent is downstepped relatively to the first one. 8

Thanks to Andries Coetzee for providing these sentences, and to Wilhelm Söhnge for confirming them.

9

The pitch accent on the pronominal subject is not necessary. In this realization, it is especially prominent, and may be a topic.

669

CAROLINE FÉRY

200 150 Pitch (Hz)

670

100 40 Jy you L∗H

het ’n boek

oor

voëls

gekoop

a book

on

birds

bought

have

LH∗

0

H∗ Time (s)



Lι 2.089

Figure 28.1 Afrikaans sentence with wide focus

It is realized on a lower F0 level than the first one. By contrast, the last one, the nuclear accent, is higher than the preceding one, although it is slightly lower than the first one. The upstep in F0 on the nuclear accent renders it more prominent than the preceding ones. Figure 28.1 illustrates a typical tonal scaling pattern also found in Dutch, English, and German. In an Afrikaans all-new sentence, every adjunct and every argument is mapped to a Φ-phrase, and the last one also includes the predicate. As seen in (20), this holds of English as well, but Afrikaans, being verb-final, illustrates the integration of the predicate in the adjacent Φ-phrase better. Because the verb is post-nuclear, it is unaccented, and as a result, its F0 (fundamental frequency contour) is low and flat. The first argument, the pronoun jy ‘you’ has a rising contour analysed here as a pitch accent L* associated with the ω-word, and a high tone HΦ. The last pitch accent, on voe¨ls, is the nuclear one and it has a falling pattern in a declarative sentence, annotated with H* LΦ. The last tone of the sentence is a low boundary tone Lι. Between the two final low tones, the F0 on the participle remains low and flat. The lower-ranked boundary tones LΦ and HΦ are aligned with the syllables immediately following the pitch accents, and the final Lι delimits the intonation phrase. The Focus Prominence Principle in (23) is adapted from the Focus Rule of Jackendoff (1972: 247), which expresses F0 raising in terms of “highest stress”, and from Truckenbrodt’s (1995) notion of focus prominence. It posits that a focus attracts the highest prominence of an ι-phrase. The consequence of this principle is that a focus may change the prominence pattern of the entire ι-phrase. (23)

Focus Prominence Principle

The highest prominence of an ι-phrase is assigned to the focus. However, the effect of focus is not limited to assigning highest stress on the focus. It also changes the relationship between the different parts of a sentence. Accenting the focused constituent goes together with

Information Structure in Germanic Languages

deaccenting the given or backgrounded postnuclear parts of the sentence, and this is why we examine both effects together in this section. It must nevertheless be kept in mind that the two effects, focus prominence principle and deaccenting as a consequence of givenness are separate processes, not necessarily linked. In West Germanic languages, givenness has a double effect: in the postfocal position, deaccenting or at least extreme compression of the register is the rule. In the prefocal position, reduction of pitch accents is often achieved, but no deaccenting results from givenness. Consider sentence (24), uttered as an answer to a question like Who baked the cake?, illustrated in Figure 28.2.

(24)

H* LΦ Lι (Andries)Φ ( het (die koek )Φ gebak)Φ]ι Andries has the cake baked ‘Andries baked the cake.’

Some words are unaccented due to their syntactic position rather than to their information structural status, as is the case for the participle gekoop ‘bought’ in (22) and gebak ‘baked’ in (24). This word would be unaccented in a wide focus context as well. The unaccented status of the participles is thus not due to givenness, but rather to its position in the sentence. The only instance in which it could be accented is when it carries a narrow focus. Prosodically, deaccenting because of givenness and postfocal compression cannot be distinguished. In both cases, it may even be the case that accents are not completely deleted, but rather extremely compressed (see Katz and Selkirk 2011 and Ku¨gler and Fe´ry 2016). In (25A), the default nuclear pitch accent placement is on cookies. As a result, it is not the best carrier of pitch accent for speaker B. The other accentable element in the Φ-phrase formed on the VP is the verb, which gets the nuclear accent. Notice that, for the sake of nuclear accent assignment, it does not 200

Pitch (Hz)

150 100

40 Andries

het die

koek

gebak

Andries

has the

cake

baked

H∗ 0



Lι Time (s)

Figure 28.2 Afrikaans sentence with a narrow focus

1.603

671

672

CAROLINE FÉRY

matter whether cookies is discourse-given, context-given, or accessible. As far as accent assignment is concerned, it is unstressable for speaker B. (25) A: Hou Andries van koekies? ‘Does Andries like cookies?’ B:

Andries [[H O U nie]F [van koekies]G nie]Φ.

‘Andries doesn’t like cookies.’ From these remarks, it can be concluded that, if an element is deaccented that would be accented if the sentence were all-new, it is interpreted as given. The same is true if it is deleted. The reverse does not hold: It is not the case that, if an element is given or accessible, its prosody and accentuation can be predicted: Discourse givenness has no invariant prosody. A given referent can still be a topic or a focus, as in example (26), adapted from Schwarzschild (1999). Even though Mary is already mentioned in the preceding question in (26), it is still the focus in the answer, because it answers the wh-question. Here the focus wins and the constituent is accented. (26) A: Who did Mary’s grandmother greet first? B:

She greeted [M A R Y ]G/F first.

In other situations, as for instance in Second Occurrence Focus (SOF), where an element is also both focused and given, givenness wins in certain circumstances, as in (27), where to Sue is text-given, but focused by being associated with the focus operator only. When the SOF is in the postnuclear position, as in (27), it is deaccented. (27) A: Yesterday Bill only introduced his friends [T O S U E ]F B:

Even [T O D A Y ]F Bill only introduced his friends [to Sue]SOF.

The asymmetry in the deaccenting strategy of languages such as German is accounted for by the alignment effect of focus. Metrical heads are preferably aligned with the right edge of an ι-phrase (see Selkirk 2000, Fe´ry 2011, 2013). If they are not aligned, for example if there is additional material after a narrow focus, the post-nuclear material is deaccented. The prefocal material is left untouched or only slightly compressed because additional accents in this position do not affect right alignment. To sum up, the role of accenting for expressing focus and the role of deaccentuation of the postfocal material for expressing givenness are related but crucially independent of each other. Northern Germanic languages present some differences from the West Germanic that are well documented in the literature (see Bruce 1977, Myrberg and Riad 2016 for Swedish). Similar to the West Germanic languages, Swedish and Norwegian have a nuclear accent in every ι-phrase, the so-called “focal accent”, that has the same properties of Culminativity as noted above. The main difference between the two groups of languages comes from a binary lexical accent

Information Structure in Germanic Languages

distinction between Accent 1 and Accent 2. Individual words are realized differently according to this lexical accent. The choice between Accent 1 and Accent 2 is determined by the morphemic and the phonological structure of the words (Riad 2014). The nuclear accent is a combination of the word accent and an extra tone. Following Bruce (1977), Myrberg and Riad (2016) distinguish between word accents, which appear on every ω-word and focal accent (or big accent), which is assigned at a higher prosodic level. In Stockholm Swedish, the realization is as follows: Word accent Focal accent Lexical accent 1 HL* L*H Lexical accent 2 H*L H*LH (in compounds: H*L*H) Post-nuclear givenness is related to the absence of focal accent rather than to deaccentuation at the lexical level. The contrastive word accents are maintained in the post-nuclear part of the sentence.

28.3.3 The Role of Topic Topics usually occupy the first position in a clause. Germanic languages with V2 properties allow not only subjects, but any constituent in this position. In all Germanic languages, a subject can have properties that are more or less typical for topics. In prosody, a topicalized constituent has the effect of creating a separate phrase, as shown in (28) and Figure 28.3 for Afrikaans. This sentence shows reordering of the constituents relative to (22), with the PP that was embedded into the direct object now topicalized at the beginning of the sentence. Afrikaans is a V2 language, as reflected in the auxiliary-subject inversion. As before, nuclear stress is on the direct object, but this time on boek.

(28)

H* Lι H* LΦ Lι [ Oor voe¨ls)Φ ( het jy (’n boek)Φ gekoop)Φ]ι about birds have you a book bought ‘You bought a book about birds.’

The pitch track shows a falling contour on the topic, and downstep of the pitch accent on boek. The pronoun has lost its status as topic that it had in Figure 28.1, and is now unaccented. In sum, both focus and topic are singled out by prosody: The focus attracts the prominence of the sentence and the topic exhibits particular phrasing properties. The given and backgrounded parts do not show the same tendency, and the result is that given parts of a sentence (in the information-structural sense) may intervene between topic and focus or be

673

CAROLINE FÉRY

250 200 150 Pitch (Hz)

674

100 40 Oor

voëls

het

jy

’n boek

gekoop

on

birds

have

you

a book

bought

L

H∗

H∗







0

2.005 Time (s)

Figure 28.3 Afrikaans sentence with a topic

located at the beginning or end of the sentence. When this happens, they have different prosodic properties. Word order plays an important role, and we therefore now examine syntax.

28.4 Syntactic Correlates of Information Structure Since prosodic correlates are predominant in Germanic languages, it is sometimes assumed that the syntactic changes merely accompany the prosodic changes, which are themselves the reflexes of information structure. For instance, since the nuclear accent is obligatorily on the focused constituent, and since the nuclear accent is preferably aligned with the end of the sentence, there is a tendency to realize backgrounded constituents before focused ones. In this view, syntax is flexible and delivers all possible word order options, and prosody decides which of the structures is most appropriate in a particular context. The prosodic effect of placing a nuclear accent as far to the end of the sentence as possible has been related to the psycholinguistic principle Given-before-New in (29). See Clark and Haviland (1977) for this principle. (29)

Given-before-New: the given material precedes the new material.

However, it will be shown in this section that most syntactic changes are not exclusively motivated by IS, but by other factors as well. In examining syntactic correlates, it is important to make a distinction between the correlates of the different notions of IS on the one hand, and the syntactic strategies like word order changes and constituent deletions or ellipsis on the other hand. In Icelandic, for example, the subject can precede or follow an adverb, like still in (30). However, the choice of the subject position is only partly related to IS considerations (Svenonius 2002). Scope considerations can be another factor.

Information Structure in Germanic Languages

(30) a. Þess vegan o¨gra ennÞa´ mo¨r g leikrit a´horfendum nu´timans this cause provoke still many plays audiences today’s ‘Because of this many plays still provoke today’s audiences’ b. Þess vegan o¨gra mo¨r g leikrit ennÞa´ a´horfendum nu´timans. Word order changes take different forms: scrambling, topicalization, focus preposing, stylistic fronting, object shift, heavy NP shift, extraposition, and verb inversion are some of them, and they have different effects. Some syntactic strategies are specialized for information structural notions: topicalization affects topics, word order changes affect focus, and pronominalization, deletion, or ellipsis only affect given constituents. Moreover, not all Germanic languages behave alike even though the syntactic reflexes of information structure are similar. SOV languages like German, Afrikaans, and Dutch allow constituent scrambling, but this option is not available, or it is heavily restricted, in languages like Swedish and English. As a result, SOV languages have greater freedom in their constituent order than SVO languages. In this section, we first review word order changes, and we end the section with pronominalization and ellipsis. See Fanselow (2016) for a more detailed review of the correlates of IS in Germanic syntax.

28.4.1 Scrambling Scrambling is used in SOV languages as a standard means to change the order of the constituents. It is generally analyzed in generative syntax as A¯-movement, where adverbials are involved, and A-movement, where only arguments are reordered. However, some analysts consider the different word orders as base generated. The question arises whether changes in constituent order are primarily the consequence of prosodic preferences that are themselves triggered by IS, as proposed by different authors (see for instance Zubizarreta 1998), or whether syntactic reordering does not need such a motivation. In this short overview, only the effects of word order on IS can be addressed; a formal analysis of word order variation goes beyond the scope of this chapter. It has been shown a number of times that IS is only one of the factors affecting scrambling, others being an optimization of the LF interface, as for example to mark scope (Frey 1993), to enable binding relations, or to avoid superiority and weak crossover violations (Haider 1981, Fanselow 2001). In German, scrambling can affect arguments, and adjuncts alike and it can take place both in main and in embedded clauses. In a German sentence like (31)a, the focus is naturally located on Paket, while it is preferably on gestern in (31)b. In other words, the object is normally in the immediately preverbal position, as in (31)a, but in (31)b, the given object has been

675

676

CAROLINE FÉRY

scrambled in front of the focused adjunct. See Neeleman and van de Koot (2016) for similar examples in Dutch, which demonstrate that whether scrambling applies in a specific circumstance depends on the kind of adverb the object has to cross. (31) a. Peter denkt, dass Beata gestern

das P A K E T

Peter thinks that Beata yesterday the package

geschickt hat. sent

has

‘Peter thinks that Beata has sent the package yesterday.’ b. Peter denkt, dass Beata [das Paket] G G E S T E R N geschickt hat.

To sum up, whether scrambling is obligatory for the sake of IS or not is subject to discussion. There is a difference between A- and A¯-movement, and there are also important differences among languages. Comparing Dutch and German, scrambling is more readily applied to German, probably because case-marking disambiguates constituent roles.

28.4.2

Leftward-Movements: Topicalization, Passivization, Dative Construction, Object-Shift In topicalization, or focus preposing, a constituent is placed in a preverbal position. See the examples in (32). In the German tradition, there is only one such position called Vorfeld or prefield. The verb or the auxiliary occupies then the second position of the sentence, and the element preceding it can be the subject – the unmarked option – as in (32b) and (32d), or any constituent of the sentence, with some exceptions. For a more thorough discussion see Chapter 15 by Haider. If a constituent is a true topic, it occupies the highest syntactic position, that is before sentence adverbials.10 (32) a. In Hamburg wurde

endlich die

In Hamburg became at.last

Elbphilharmonie fertiggebaut.

the Elbphilharmony

ready.built.

‘At last they finished building the Elbphilharmony in Hamburg.’ b. Die Elbphilharmonie wurde in Hamburg endlich fertiggebaut. c. Endlich wurde die Elbphilharmonie in Hamburg fertiggebaut. d. Die Stadt Hamburg baute die Elbphilharmonie endlich fertig.

V2 characteristics are crucial for word order in general. In German, the prefield position is obligatorily occupied, and the verb is in the second position. Compare the passive sentences in (33). In (33)a, the prefield position is occupied by the adverb and there is no subject. In (33)b, when the adverb is postverbal, the prefield position must be filled by an expletive pronoun.

10

It was mentioned in Section 28.4.1 that English, Scandinavian, and Icelandic do not allow scrambling of an object across a subject. As a result, in these languages, only subjects can be topics.

Information Structure in Germanic Languages

(33)

a. Morgen wird gestreikt. tomorrow be.3sg striked ‘Tomorrow there will be a strike.’ b. Es wird morgen gestreikt.

Because of the operation consisting in fronting the object, passivization is directly related to IS, but it also suppresses the agent or expresses it in a prepositional phrase, both processes that do not necessarily relate to IS. Thus, a sentence like (32a) and (32b) can easily be interpreted as containing a topic. This process can even be considered as focus preposing, but it does not have to. Fanselow and Lenertova´ (2010) show that in many languages, the highest argument of the verb can occupy the prefield in a wide focus reading, in which case the fronted argument is accented, like in example (34). Notice that die Zeitung ‘the newspaper’ is not a topic in this sentence, as it does not need an actual referent. The focus is the entire VP would like to read the newspaper. Such “formal fronting” is only possible if there is no other accentable element in the sentence. The whole sentence has to form one Φ-phrase to be interpreted as wide focus. (34)

{What do you like to do tomorrow?} Die Z E I T U N G wu¨rde ich gerne the newspaper would I with.pleasure ‘I would like to read the newspaper.’

lesen. read

In German, Dutch, and Afrikaans, a contrastive focus can be preposed, as in (35). (35)

A: Who prepared the desserts? B: [Die S U U R lemoenkoek]F het Tom gemaak en The lemon pie has Tom made and [die M A L vapoeding]F het Alain gemaak. the malva pudding has Alain made. ‘Tom made the lemon pie and Alain the tiramisu.’

It has been demonstrated by Givo´n (1984) and Bresnan and Nikitina (2008) among others that in a dative construction like the one in (36), the dative is more prone to be given than in the corresponding prepositional phrase construction, although the relationship between construction and IS is again not systematic or obligatory. (36)

a. Annalisa gave the magpie her grandmother’s ring. b. Annalisa gave her grandmother’s ring to the magpie.

In Swedish and Icelandic, there is a clause-internal leftward movement called Object Shift that places the object before certain adverbials and the negation (Holmberg 1986). Again, several authors have shown a relationship between object shift and IS. Object Shift is the subject of

677

678

CAROLINE FÉRY

a chapter in this handbook (see Broekhuis Chapter 18), and for this reason, it is not commented upon here.

28.4.3

Rightward Movements: Extraposition, Right-Dislocation, Heavy NP-Shift, Afterthought Until now, we have seen how given constituents, topics, and sometimes even focus can be located further to the beginning of the sentence than in their neutral position because of IS. Let us next turn to XPs that move to the right of their canonical position, in particular to clause-final positions. Extraposition, right-dislocation, heavy NP-shift and afterthought can also be IS sensitive. Extraposition is the only rightward movement that is intonationally integrated into the main clause. All others are dislocated. Rochemont’s example of PP extraposition in English in (37)a requires a presentational predicate, as does heavy-NP-shift in (37)b, to a lesser extent. Moreover, it has been proposed that focusing the direct object or part of it licenses its displacement to the right. See Rochemont and Culicover (1990) among many others. (37)

a. There a man came in with blond hair. b. On the next day, Mary introduced to her mother the man she intended to marry next week.

In Scandinavian languages, extraposition is possible, but it is heavily constrained. Compare the following sentence from Norwegian. The clausal complement of the main verb appears after the temporal adjuncts, resulting in a marked word order.11 (38)

Jeg hadde fortalt henne alt i forgaº rs om I had told her already in foreyester about hva som hadde skjedd on market.D E F what that had happened paº markedet ‘I had already told her the day before yesterday what had happened at the market.’

In Dutch and German, both PPs and clauses can be extraposed. This operation correlates with length of the constituents, definiteness/indefiniteness of the PP, and IS. Given constituents are more readily extraposed than focused ones. However, the preference is not strong. Compare the following sentences from Dutch.12 In (39), the extraposed constituent is given, but in (40), it is the narrow focus, and it is equally good.

11

Thanks to Kjell Johan Sæbø for this sentence, and to Siri Gjersøe for confirmation.

12

Thanks to Beata Moskal for the Dutch sentences and to Ruben van de Vijver for confirmation.

Information Structure in Germanic Languages

(39)

A: Kun je iets over je vader vertellen? ‘Can you tell something about your father?’ B: Ik heb afgelopen week over de financie¨le crisis gesproken I have last week about the finance crisis spoken met mijn vader. with my father ‘Last week I spoke about the finance crisis with my father.’

(40)

A: What did you speak with your father about? B: Ik heb met hem gesproken over de financie¨le crisis. I have with him spoken about the finance crisis

In both Dutch and German, extraposition of complement clauses is nearly obligatory. In the preverbal positions, such complements are barely acceptable if at all. (41)

We hebben in de vergadering gisteren besloten, dat We have in the meeting yesterday decided that we nieuwe bureaus gaan kopen. we new desks will buy Yesterday at the meeting we decided that we will buy new desks.

Finally, discontinuous nominal phrases may arise as a consequence of the need to separate two IS roles in one constituent. In (42), the fronted head noun is a topic and the remnant, a quantifier, is the focus. The roles of the two parts may vary, but the very fact that both parts have different roles legitimates discontinuity. (42)

Seeadler hat Derk schon sehr White-tailed eagles has Derk already very ‘Derk already saw a lot of white-tailed eagles.’

viele gesehen. many seen

28.4.4 Pronominalization and Ellipsis It does not come as a surprise that givenness correlates with pronominalization and deletion of constituents. Pronominalization and deletion in syntax correspond to compression and deletion of pitch accents in the prosody, thus to absence of the highlighting correlates of focus. Specific anaphoric expressions, like personal pronouns, clitics, demonstratives, and definite articles are the usual anaphoric elements of given referents; see (43) with she and him. (43)

Leilai saw the blond clownj, shei even listened to himj, but in the end, shei didn’t trust himj.

In the same way as established previously for word order, the use of such devices is not limited to the expression of IS, but they have other functions

679

680

CAROLINE FÉRY

as well. In the present case, they help to establish coherence in discourse. A main pitch accent on a pronoun is of course possible, but it implies an unusual anaphoric relation between pronoun and antecedent, as in Halliday’s (1967) oft-cited example in (44). If the pronouns he and him are accented in (44), a hearer assumes that the anaphoric relationship between the pronouns and antecedents is reversed.13 In other words, the referents are given since previously mentioned, but the pitch accents on the pronouns indicate newness of their roles. (44)

John called Bill a Republican and then H E insulted H I M .

Deletion or ellipsis of given material correlates with contrastive remnants (see Merchant 2001, 2004, Vicente 2006, Winkler 2016 among others). Examples of ellipsis appear in (45). The following constraints are crucial for deletion: (1) Two constituents are coordinated, (2) one of the constituents contains the antecedent of the deleted constituent in the other constituent (they have to be semantically identical), (3) at least part of the remnant is in a constrasting or parallel relationship with the corresponding constituent. (45)

Ellipsis a. Gapping: Beata reads a book and Peter _ the newspaper. b. Right-Node-Raising: Beata writes _ and Peter corrects their joint paper. c. Sluicing: Someone’s writing a paper, but I don’t know who _.

28.5 Particles When focus is associated with focus-sensitive particles like exclusive only, additive also and scalar even, we talk of “association with focus”, following Jackendoff (1972). Such focus particles may have an effect on the truthconditions of the sentence altogether, but this is not necessarily the case. Rooth’s (1985) sentence (46) has different truth-conditions depending on the location of focus, which is indicated by small capital letters. In a context where John introduced Bill to Sue (and did not introduce anyone else to Sue), the sentence is true if the pitch accent is on Bill, as in (46)a but false if the pitch accent is on Sue, as in (46)b. Particles like these are semantic operators whose interpretational effects depend on focus. (46)

13

a. John only introduced B I L L to Sue. b. John only introduced Bill to S U E .

The fact that the verb insulted is deaccented has a different function. It indicates that calling somebody a Republican is an insult.

Information Structure in Germanic Languages

By contrast, the sentences (47) and (48) have the same truth values with or without the focus particles. In (48), for instance, the second part of the sentence means that Bernard had a car crash. The particle too suggests that he was not the only one who had a car crash. In fact, the word too in such a sentence can serve as a pitch accent carrier, since the remainder of the comment is given. In other words, in (47) and (48), the association with focus does not contribute to the (communicated) truth conditions of the sentence but adds a “side condition” (conventional implicature). (47)

a. The magpies also [flew A W A Y ]F (not only the ravens did). b. The elk even broke a [F E N C E ]F (and it also destroyed the flower bed).

(48)

{What did you see yesterday?} [David]T O P [had a C A R crash]F, and [Bernard]T O P [did T O O ]F (have a car crash).

Some discourse particles have a larger role to play in IS, and can be themselves the bearer of IS. In German, a double role for focus particles is not so rare and has been analyzed a number of times in the literature. This double function is illustrated here with wieder ‘again’ and schon ‘already’ that change their role depending whether they are straightforward focus particles, and unstressed, or they carry the focus themselves, and are stressed. Examples are provided in (49) and (50). In the first versions, pitch accent is on the associated constituent of the focus particle. Example (49a) implies that the garbage bin is usually or at least often full, and Alicia has restored this state. Sentence (50)a with main stress on gin and tonic may suggest either that Albert had already a gin and tonic before or that the time of the day is early for having this drink, or both. In the second versions, the focus is on the particle itself. Sentence (49) b suggests that Alicia or someone else already filled the garbage bin before and that this action has been repeated. Example (50b) highlights the fact that Albert already had this drink, that the sentence is true. It has an additional component of possibly refuting incredulity. (49)

(50)

a. Alicia hat die Mu¨lltonne wieder V O L L gemacht. b. Alicia hat die Mu¨lltonne W I E D E R voll Alicia has the garbage bin again full ‘Alicia filled the garbage bin again.’ a. Albert hat schon gin and T O N I C getrunken. gin and tonic b. Albert hat S C H O N Albert has already gin and tonic drunk. ‘Albert already drank gin and tonic.’

gemacht. made.

getrunken.

The use of focus particles in the Germanic languages, although not rare, remains marginal, as compared to other languages. Prosody and syntax are the primary strategies used in the marking of focus and topic.

681

682

CAROLINE FÉRY

28.6 Conclusion This contribution investigated elements of information structure, and its interaction with other domains of grammar (e.g., prosody, morphology, syntax, semantics) across Germanic languages. In doing so, it has followed a pattern that is often found in the literature: Each information structural notion is investigated independently and the same is true for each grammatical strategy used as a reflex of IS. But in fact, this is a simplification. If focus is expressed by prosodic means, by assigning a higher pitch accent than would be the case if the constituent were part of an all-new sentence, the probability is high that the postfocal given material is realized with deaccenting or compression. If the focus is realized by scrambling, and results in right-alignment, the given material finds itself in prenuclear position, and there is no reason to deaccent it. In short, how the information structural blocks are “packaged” is best studied in comparing the blocks with each other.

References Bauman, S. and A. Riester 2013. “Coreference, lexical givenness and prosody in German,” Lingua 136: 16–37. Bresnan, J. and T. Nikitina 2008. “Gradience and the dative alternation.” In L. Uyechi and L-H. Wee (eds.), Reality Exploration and Discovery: Pattern Interaction in Language and Life. Stanford: CSLI Publications: 1–23. Bruce, G. 1977. “Swedish word accent in sentence perspective,” Travaux de l’Institut de Linguistique de Lund 12. Gleerup: Lund. Bu¨ring, D. 2003. “On D-Trees, Beans, and B-Accents,” Linguistics & Philosophy 26.5: 511–545. Chafe, W. 1976. “Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View.” In C. Li (ed.), Subject and Topic. New York: Academic Press: 25–55. Chomsky, N. and M. Halle 1968. The Sound Pattern of English. New York: Harper and Row. Clark, H. H. and S. E. Haviland. 1977. “Comprehension and the given-new contract.” In R. O. Freedle (ed.), Discourse Processes: Advances in Research and Theory, Vol. 1: Discourse Production and Comprehension. Norwood, NJ: Ablex: 1–40. Constant, N. 2014. Contrastive Topic: Meanings and Realizations. Unpublished Ph.D. thesis, University of Massachusetts.

Information Structure in Germanic Languages

Fanselow, G. 2001. “Features, θ-roles, and free constituent order,” Linguistic Inquiry 32: 405–437. Fanselow, G. 2016. “Syntactic and prosodic reflexes of information structure in Germanic.” In C. Fe´ry and S. Ishihara (eds.), The Oxford Handbook of Information Structure. Oxford University Press: 621–641. Fanselow, G. and D. Lenertova´ 2010. “Left peripheral focus: Mismatches between syntax and information structure,” Natural Language and Linguistic Theory 29:169–209. Fe´ry, C. 2011. “German sentence accents and embedded prosodic phrases,” Lingua 121: 1906–1922. Fe´ry, C. 2013. “Focus as prosodic alignment,” Natural Language and Linguistic Theory 31.3: 683–734. Frey, W. 1993. Syntaktische Bedingungen fu¨r die semantische Interpretation. Berlin: Akademie Verlag. Haider, H. 1981. “Empty categories: On some Differences between German and English,” Wiener Linguistische Gazette 25:13–36. Halliday, M. A. K. 1967 /1968. “Transitivity and theme in English (part II),” Journal of Linguistics 3: 199–244. Hamblin, C. L. 1973. “Questions in Montague Grammar,” Foundations of Language 10: 41–53. Holmberg, A. 1986. Word Order and Syntactic Features in the Scandinavian Languages and English. Ph.D. dissertation, University of Stockholm. Hyman, L. 2006. “Word-prosodic typology,” Phonology 23: 225–257. Jackendoff, R. 1972. Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jacobs, J. 2001. “The dimensions of topic-comment,” Linguistics 39: 641–681. Katz, J. & E. O. Selkirk 2011. “Contrastive focus versus discourse-new: Evidence from phonetic prominence,” Language 87: 771–816. Krifka, M. 2008. “Basic notions of information structure,” Acta Linguistica Hungarica 55: 243–276. Ku¨gler, F. and C. Fe´ry 2016. “Postfocal downstep in German,” Language and Speech. DOI:10.1177/0023830916647204. Merchant, J. 2001. The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford University Press. Merchant, J. 2004. “Fragments and ellipsis,” Linguistics and Philosophy 27.6: 661–738. Myrberg, S. and T. Riad 2016. “On the expression of focus in the metrical grid and in the prosodic hierarchy.” In C. Fe´ry and S. Ishihara (eds.), The Oxford Handbook of Information Structure. Oxford University Press: 441–462. Neeleman, A. and H. van de Koot 2016. “Word order and information structure.” In C. Fe´ry and S. Ishihara (eds.), The Oxford Handbook of Information Structure. Oxford University Press: 383–401.

683

684

CAROLINE FÉRY

Prince, E. 1981. “Toward a taxonomy of given-new information.” In P. Cole (ed.), Radical Pragmatics. New York: Academic Press: 223–256. Reinhart, T. 1981. “Pragmatics and linguistics: An analysis of sentence topics,” Philosophica 27: 63–94. Riad, T. 2014. The Phonology of Swedish. Oxford University Press. Roberts, C. 1996. “Information structure in discourse: Towards an integrated formal theory of pragmatics.” In J. H. Yoon and A. Kathol (eds.), OSU Working Papers in Linguistics 49: Papers in Semantics. Columbus: Ohio State University Press: 91–136. Rochemont, M. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins. Rochemont, M. 2016. “Givenness.” In C. Fe´ry and S. Ishihara (eds), The Oxford Handbook of Information Structure. Oxford University Press: 41–63. Rochemont, M. and P. Culicover 1990. English Focus Constructions and the Theory of Grammar. Cambridge Studies in Linguistics 52. Cambridge University Press. Rooth, M. 1985. Associations with Focus. Unpublished Ph.D. thesis, University of Massachusetts. Rooth, M. 1992. “A Theory of Focus Interpretation,” Natural Language Semantics 1: 75–116. Rooth, M. 2016. “Alternative semantics.” In C. Fe´ry and S. Ishihara (eds.), The Oxford Handbook of Information Structure. Oxford University Press: 19–40. Schwarzschild, R. 1999. “Givenness, AvoidF and other constraints on the placement of accent,” Natural Language Semantics 7: 141–177. Selkirk, E. O. 2000. “The interaction of constraints on prosodic phrasing.” In M. Horne (ed.), Prosody: Theory and Experiment. Dordrecht: Kluwer: 231–261. Stalnaker, R. 1974. “Pragmatic presuppositions.” In M. Munitz and P. Unger (eds), Semantics and philosophy. New York University Press: 197–213. Steedman, M. 2000. “Information structure and the syntax-phonology interface,” Linguistic Inquiry 31:649–689. Svenonius, P. 2002. “Subject positions and the placement of adverbials.” In P. Svenonius (ed.), Subjects, Expletives, and the EPP. New York: Oxford University Press: 199–240. Tomioka, S. 2010. “Contrastive topics operate on speech acts.” In M. Zimmermann and C. Fe´ry (eds.), Information Structure. Oxford University Press: 115–138. Truckenbrodt, H. 1995. Phonological Phrases: The Relation to Syntax, Focus, and Prominence. Unpublished Ph.D. dissertation, MIT. Vallduvı´, E. and E. Engdahl 1996. “The linguistic realization of information packaging,” Linguistics 34: 459–519.

Information Structure in Germanic Languages

Vicente, L. 2006. “Negative short replies in Spanish.” In L. Bettelou and J. van de Weijer (eds.), Linguistics in the Netherlands 23. Amsterdam: John Benjamins: 199–210. Wang B. and C. Fe´ry 2018. Prosody of Dual-focus in German: Interaction between Focus and Phrasing. Language and Speech 61.2: 303–333. Winkler, S. 2016. “Ellipsis and information structure.” In C. Fe´ry & S. Ishihara (eds), Oxford Handbook of Information Structure. Oxford University Press: 359–382. Zubizaretta, M-L. 1998. Prosody, Focus, and Word Order. Cambridge, MA: MIT Press.

685

Part V

Language Contact and Nonstandard Varieties

Chapter 29 Second Language Acquisition of Germanic Languages Carrie Jackson

29.1 Introduction The study of second language acquisition focuses on how people acquire and use a second language (L2), and the linguistic and cognitive mechanisms that support these processes (see Gass 2013, Mitchell et al. 2013 for overviews). Key issues include (i) the nature of learners’ L2 linguistic representations, (ii) how such representations develop and how they are shaped by the L2 input to which learners are exposed, and (iii) the ways in which features from learners’ two languages interact to influence L2 acquisition and use. These issues address larger debates concerning the extent to which L2 learners – and especially adult learners – can acquire nativelike L2 proficiency and whether L1 and L2 acquisition rely on the same underlying mechanisms, or whether fundamentally different mechanisms and strategies are involved, especially among adult L2 learners who already possess a mature first language (L1) linguistic system (e.g., Bley-Vroman 1990, Birdsong 1992, White 2003). There is also growing consensus that L2 learners are “NOT the sum of two complete or incomplete monolinguals” (Grosjean 1989: 3, emphasis in original), leading researchers to focus on the unique characteristics of L2 learners’ linguistic systems rather than comparing L2 learners’ linguistic systems to some set of monolingual L1 speaker norms (see also Klein and Perdue 1997, Dimroth 2013). For more than 40 years, Germanic languages, especially German and Dutch, have played a central role in L2 acquisition research. The goal of the present chapter is to review selected findings from this research and discuss how these findings contribute to our understanding of the mechanisms that drive L2 acquisition and use. The scope of this chapter is necessarily limited, focusing primarily on the acquisition and use of L2 morphosyntax among adult L2 learners who begin acquiring their L2 after

690

CARRIE JACKSON

puberty (for work on L2 phonological acquisition in Germanic languages, see Bongaerts et al. 2000, Moyer 2004, O’Brien 2004, Gut 2009). The simultaneous and sequential acquisition of a Germanic L2 among children will not be discussed (see Gawlitzek-Maiwald and Tracy 1996, Haberzettl 2005, Mu¨ller et al. 2009, Blom 2010, Chilla et al. 2010, Unsworth, 2016, among others). Similarly, only limited reference will be made to research on how different classroom instructional methods support L2 learners’ developing linguistic systems (see, e.g., Diehl et al. 2000, Eckerth et al. 2009, Jackson et al. 2018).

29.2 Learner Varieties Starting in the 1970s, researchers conducted several seminal projects on the acquisition of various European languages by immigrants, including the Heidelberger Projekt Pidgin-Deutsch (Klein and Dittmar 1979), the ZISA cross-sectional and longitudinal studies (Meisel et al. 1981, Clahsen et al. 1983), which focused on the acquisition of German by Italian and Spanish foreign workers, and the ESF project (Klein and Perdue 1992), which involved the cross-linguistic comparison of L2 development among adult immigrants from a variety of L1s (Punjabi, Italian, Turkish, Arabic, Spanish, and Finnish) who were in the process of acquiring English, German, Dutch, French, or Swedish. Central to these studies was the emphasis on investigating L2 learners’ linguistic systems as interlanguage systems in their own right, rather than perceiving them as deviant versions of the target language grammar. Participants were primarily untutored L2 learners, whose L2 exposure came from naturalistic input by virtue of living and working in the target language country, thus providing insight into how L2 linguistic systems develop in the absence of formal instruction and – in many cases – on the basis of relatively limited L2 input, as contact with the target language for many participants was confined to everyday interactions in the workplace. The primary goal of these studies was to identify the different developmental stages through which L2 learners progress, what principles guide the internal structure of utterances at any given stage, and the transition from one stage to the next, as well as whether L2 development is influenced by linguistic features from learners’ L1. These studies focused on free production data that the L2 learners produced through personal narratives, interviews, and film retellings, in which participants watched part of a silent film and then reconstructed the events in the film for the interviewer.

29.2.1 Word Order in German and Dutch The general clausal structure is similar in German and Dutch, although differences exist, especially with regard to the robustness of case marking

Second Language Acquisition of Germanic Languages

(see McFadden, Chapter 13). The underlying word order in German and Dutch is SOV. Thus, the finite verb appears clause-finally in embedded clauses, as in (1a) and (1b) (see Haider,Chapter 15; Vikner, Chapter 16). The finite verb raises to verb-second position in main clauses, while verb particles and nonfinite verbs (e.g., infinitive forms and past participles) remain in clause-final position, leading to so-called discontinuous word order or the verbal bracket (Verbklammer in German), as in (2a) and (2b). German and Dutch are verb-second languages, meaning that precisely one constituent appears before the finite verb in main clauses, leading to both preverbal and postverbal placement of the subject, as in (3a) and (3b). (1)

a. Der Vater sagt, dass der Junge am Abend seine Hausaufgabe machen muss. the father says that the boy in.the evening his homework do must. b. De vader zegt, dat de jongen ‘s avonds zijn huiswerk moet doen. the father says that the boy in the evening his homework must do. “The father says that the boy must do his homework in the evening.”

(2)

a. Der Junge muss am Abend seine Hausaufgaben machen. the boy must in.the evening his homework do. b. De jongen moet ‘s avonds zijn huiswerk doen. the boy must in the evening his homework do. “The boy must do his homework in the evening.”

(3)

a. Am Abend muss der Junge seine Hausaufgaben machen. in.the evening must the boy his homework do. b. ‘S avonds moet de jongen zijn huiswerk doen. in the evening must the boy his homework do. “In the evening the boy must do his homework.”

Due to this combination of features and the fact that subject-initial V2 main clauses have SVO surface word order, L2 learners of German and Dutch are faced with inconsistent input with regard to verb placement. Hence, the acquisition of word order requires learning multiple constraints that, from a learner’s perspective, often compete with one another.

29.2.2 The Basic Variety The ESF project consisted of longitudinal data collected from 40 different participants over the course of 30 months (Klein and Perdue 1992). This research culminated in a theoretical framework to account for the initial stages in L2 development, referred to as the Basic Variety. While many ESF respondents eventually progressed beyond the Basic Variety, close to one

691

692

CARRIE JACKSON

third of the participants’ language development remained at this level even at the end of the study (Klein and Perdue 1992, Klein and Perdue 1997, Klein and Dimroth 2009). With regard to the lexicon, the Basic Variety consists primarily of nouns and verbs, with a limited set of adverbs and adjectives, a few quantifiers, a single word for negation, a limited set of prepositions, a highly reduced pronominal system, and no complementizers. The Basic Variety contains no inflectional morphology, meaning that all lexical items appear in their bare forms with no markers for tense, aspect, agreement or case. Here, I will focus on the overall structure of utterances in the Basic Variety (see Dietrich et al. 1995, Becker and Carroll 1997, Dimroth and Starren 2003, Hendricks 2005 for discussion of negation, spatiality, and temporality in the Basic Variety). In the Basic Variety, learners’ utterances are organized according to three sets of structural, semantic, and pragmatic constraints. First, phrasal constraints limit utterances to three basic word order patterns: NP1-V, NP1-V-(NP2)-(NP2), NP1-Copula-{Adj, NP2, PP}, and {V, Copula}-NP2 (Klein and Perdue 1992). Importantly, NP1 is not necessarily the subject of an utterance nor is NP2 necessarily the object, as additional semantic and pragmatic constraints are also at play. All of these phrasal patterns can be preceded by an adverbial element or the coordinating conjunction and (or und or en, ‘and’ in German and Dutch, respectively). In some cases, Turkish learners of German and Dutch – but not Moroccan learners of Dutch or Italian learners of German – also produced a small number of NP-NP-V utterances, providing limited evidence of L1 influence, as Turkish (but not Arabic or Italian) is a verb-final language. However, Klein and Perdue (1997) point out that even Turkish learners progressed to producing NP-V-NP utterances, leading them to conclude that from a longitudinal perspective, L1 influence is more limited at this stage of L2 development. The second set of constraints in the Basic Variety are semantic constraints. The primary constraint is that the controller of an utterance, specifically which noun exerts control over other arguments in the utterance, should occur first. Thus, agentive referents usually appear in utterance-initial position. When utterances involve three arguments (e.g., actions involving giving something to another person), the controller of the source state outweighs the controller of the target state, leading to utterances like Charlie give present for young children (Klein and Perdue 1997: 315). The third set of constraints consists of pragmatic constraints related to the information status of the elements in an utterance. Learners follow the constraint that the focus expression of an utterance should come last. Consider the Dutch sequence, De meneer valt van water. Met valt drie ‘That mister fall from water. With fall three’ (i.e., Charlie fell into the water. There were three of them who fell; Klein and Perdue 1997: 316). Here, the main referent, de meneer, appears in initial position in the first sentence but the main referent of the second utterance, drie, appears at the end of

Second Language Acquisition of Germanic Languages

the utterance, because the focus of the second utterance is that an additional three people fell into the water. The second pragmatic constraint is that given information appears before new information. In situations where constraints conflict, for instance when the controller of an utterance is simultaneously the focus of the utterance, learners can choose to prioritize one constraint over the other (e.g., maintain the semantic constraint of controller first, thereby overriding the pragmatic constraint of focus last), or employ an additional device, such as prosody, to mark the infocus element (Klein and Dimroth 2009).

29.3 The Development of Word Order in L2 German Around the same time as research on the Basic Variety, a related line of research focused on the development of word order rules among immigrant populations acquiring German. This research also addresses questions regarding developmental stages in L2 acquisition, but with the primary goal of understanding what role, if any, the principles of Universal Grammar and the transfer of L1 linguistic features play in adult L2 acquisition (see Bley-Vroman 1990, Jordens 2001, White 2003). Clahsen and colleagues (Meisel et al. 1981, Clahsen and Muysken 1986, Clahsen 1990), developed a multistage model to explain the developmental sequence of word order among L2 German learners. In stage 1, Clahsen and Muysken (1986) hypothesized that L2 German learners rely on SVO word order, as in (4). In stage 2, L2 learners optionally move adverbs into sentence-initial position, but maintain SV order, as in (5). In stage 3, L2 learners acquire discontinuous word order, correctly placing nonfinite verbs and verb particles in sentence-final position, as in (6). In stage 4, L2 learners acquire inversion and the verb-second rule, placing the subject immediately after the finite verb when an adverbial phrase appears in sentence-initial position, as in (7). In stage 5, L2 learners realize that adverbial phrases can appear between the verb and the object, as in (8). In stage 6, learners correctly place the finite verb in sentence-final position in embedded clauses (all examples from Clahsen 1990, Clahsen and Muysken 1986). (4)

ich studieren in Porto (Stage 1: SVO) I study.INF in Porto

(5)

vielleicht andere kollege sagen (Stage 2: ADV-PREP) perhaps other colleague say.INF

(6)

ich muss jetzt putzen (Stage 3: Particle) I must.SG now clean.INF

(7)

und dann arbeiten die zwei in zuhause (Stage 4: INV) and then work.INF/PL the two in at home

(8)

ich soll aus Italien eine apparat bringen (Stage 5: ADV-VP) I should.SG from Italy a machine bring.INF

693

694

CARRIE JACKSON

As with the Basic Variety, not all L2 learners progress to stage 6. For some learners, language development stabilizes at an earlier phase. Clahsen and Muysken (1986) argue that all L2 learners progress through these phases, including the initial SVO stage, even when their L1 is verbfinal, thus positing no significant role for L1 transfer in the early stages of adult L2 acquisition. They claim that any early utterances containing a verb in final position are best interpreted as containing some sort of preposed element, as in meine bruder er helfen ‘my brother [is who] he helps’ (Clahsen and Muysken 1986: 108), rather than as verb-final per se. Clahsen and Muysken also contrast these developmental stages of word order acquisition among adult L2 learners to the developmental stages in child L1 acquisition of German. Specifically, L1 German speaking children acquire verb-final word order early on. Further, in child L1 acquisition there is a tight coupling between the acquisition of subject-verb agreement and the placement of finite verbs in verb-second position, whereas these two linguistic features are acquired separately among adult L2 learners (but see Parodi 2000 for counterevidence). Based on the lack of significant L1 transfer and these differences between child L1 and adult L2 acquisition, Clahsen and colleagues conclude that adult L2 acquisition relies on fundamentally different cognitive processes from child L1 acquisition, with adult L2 acquisition relying on domain-general cognitive processes and child L1 acquisition driven by the principles of Universal Grammar (see also Clahsen 1990, but see Meisel 1997 for the possibility that adult L2 acquisition can involve both domain-general and UG-constrained processes). Building on this research, Pienemann (1998) developed Processability Theory to explain the sequence of these developmental stages in L2 German in terms of processing constraints. Pienemann bases Processability Theory on psycholinguistic models of L1 production (Kempen and Hoenkamp 1987, Levelt 1989), framed within LexicalFunctional Grammar (Kaplan and Bresnan 1982). Such models emphasize the incremental nature of language processing and that the processing procedures involved in L1 production first require the activation of individual words, followed by lexical procedures, then phrasal procedures, then sentence-level procedures (i.e., word order rules), and finally subordinate clause procedures (if needed). Applied to L2 acquisition, L2 learners must successfully acquire the L2-specific processing routines necessary to process L2 input at any given level before they can proceed to the next level. For example, L2 learners must be able to successfully produce utterances containing fronted adverbial phrases in the absence of inversion (i.e., XSVX word order), a phrasal-level procedure, before acquiring inversion, a sentence-level procedure. Such implicational sequences are identical for all L2 learners, regardless of the L1. L1 transfer occurs only when the L2-specific processing routines that are necessary for processing such grammatical features are already in place, at which point L1 transfer may facilitate accurate production of target L2 features within a given

Second Language Acquisition of Germanic Languages

developmental stage (see Haº kasson et al. 2002, Bohnacker 2006, Pienemann and Haº kasson 2007 for further discussion). Although developed to explain the acquisition of word order in L2 German (see also Jansen 2008), Processability Theory has since been applied to the acquisition of case marking among Dutch L2 learners of German (Baten 2011), word order sequences in L2 Swedish (Glahn et al. 2001, Haº kasson and Norrby 2010), and additional linguistic phenomena in non-Germanic languages (see Pienemann 2005 for review). Contra Clahsen and Muysken (1986) and Pienemann (1998), Schwartz and Sprouse (1994, 1996) proposed the Full Transfer / Full Access hypothesis, arguing that the starting point of adult L2 acquisition is the end state of L1 acquisition and that adult L2 learners have full access to the principles and parameters of Universal Grammar throughout all stages of L2 acquisition. Their hypothesis rests largely on a reanalysis of data from a Turkish L2 learner of German from the ESF project (Klein and Perdue 1992). Schwartz and Sprouse (1994) characterize this learner as going through three developmental stages which are all UG-compatible in that they are instantiated in other natural languages, even if they are not instantiated in German. Stage 1 contains both SV and SOV word order, with verb particles and nonfinite verbs appearing in verb-final position, as in (9), while finite verbs occur earlier in the sentence, as in (10). At this stage there are no instances of a nonfinite verb preceding an object NP, leading Schwartz and Sprouse to conclude that this learner’s underlying word order is SOV, which he has transferred from his L1 Turkish. The high proportion of SV utterances, even at this early stage, reflects the frequency and saliency of this word order for main clauses in German. Stage 2 is characterized by a high proportion of inversion and verb-second word order with pronominal subjects, as in (11), but not nonpronominal subjects, as in (12). The learner also exhibits verbfinal word order in embedded clauses, as in (13). To explain the divergence in inversion with pronominal versus nonpronominal subjects, Schwartz and Sprouse appeal to generative accounts of subject-verb inversion in French, as French similarly allows inversion with pronominal but not nonpronominal subjects (Rizzi and Roberts 1989). In stage 3, this learner now produces sentences containing inversion with nonpronominal subjects, as in (14), although some instances of ungrammatical XSV utterances with nonpronominal subjects remain, leading Schwartz and Sprouse to conclude that he has still not fully acquired the verb-second constraint in German (all examples from Schwartz and Sprouse 1994). (9)

der Mann seine Frau geku¨sst (Stage 1: SOV) the man his wife kissed.PST “The man kissed his wife.”

(10)

jetzt er hat Gesicht [das ist falsches Wagen] (Stage 1: SV) now he has.SG face that is false wagon “Now he makes a face that [that] is the wrong car.”

695

696

CARRIE JACKSON

(11)

dann trinken wir bis neun Uhr (Stage 2: XPronV) then drink.PL/INF we until nine o’clock “Then we will drink until nine o’clock.”

(12)

in der Tu¨rkei der Lehrer kann der Schu¨ler schlagen (Stage 2: XSV) In the Turkey the teacher can.SG the pupil beat.INF “In Turkey the teacher can beat the student.”

(13)

dass ich mit Brot war (Stage 2: SOV, embedded) that I with bread was.SG “That I had some bread.”

(14)

draußen hatte die Polizei eine Wagen brauchen sollten (Stage 3) outside had.SG the police a wagon bring.INF should.PL/INF “Outside the police should have used a vehicle.”

Schwartz and Sprouse (1994) emphasize that the critical test for whether adult L2 acquisition is supported by Universal Grammar is not whether key developmental stages in adult L2 acquisition directly map onto developmental stages in child L1 acquisition (cf. Clahsen and Muysken 1986), or whether L2 learners’ linguistic systems perfectly match the target language grammar, as constructed by mature native speakers of that language. Rather, the crucial test is whether the language produced by an L2 learner at any given developmental stage can be adequately explained by Universal Grammar, exemplified here by their analysis of inversion with pronominal versus nonpronominal subjects in stage 2. Schwartz and Sprouse highlight that adult L2 learners – similar to child L1 speakers – must construct and continually update their linguistic systems to account for the dominant patterns they encounter in the L2 input, and that the mechanisms by which they do so are constrained by the principles of Universal Grammar. In essence, even if the initial and end states of child L1 and adult L2 acquisition differ, the cognitive and linguistic processes that drive acquisition are fundamentally similar. Each stage in adult L2 development represents how L2 learners construct a grammar based on the interaction of UG-based principles, the architecture of the L1 grammar, the L2 input to which they are exposed, and general constraints in language learning procedures. Based on data from Turkish, Korean, Italian, and Spanish L2 learners of German, Vainikka and Young-Scholten (1996a, 1996b, 2011) proposed the Minimal Trees Hypothesis, or more recently, Organic Grammar, as yet another model to explain adult L2 acquisition within a generative framework. In contrast to the Full Transfer / Full Access Hypothesis (Schwartz and Sprouse 1994, 1996), Vainikka and Young-Scholten propose that the full array of functional projections (i.e., CP, AgrP, TP/FP) are not available in the initial stages of adult L2 acquisition, but rather such projections, and their corresponding syntactic structures, develop incrementally over time. As outlined in Vainikka and Young-Scholten (1996b), a central argument in this model is that the initial state of L1 acquisition consists of only a bare

Second Language Acquisition of Germanic Languages

verb phrase (VP), which is either head-initial or head-final based on the parameter settings from a learner’s L1, leading to different word order preferences in L2 German (i.e., SVX versus SXV) based on the L1 grammar, as in (15) and (16). The absence of modal and auxiliary verbs, as well as tense and agreement features, indicates that at this initial state, L2 learners lack any functional projections beyond the VP. In stage 2, the acquisition of modal verbs serves to trigger the development of a single functional projection above the VP, leading to verb raising and discontinuous word order, as in (17). Based on the preponderance of SVX input in German main clauses, this functional projection (FP; finite phrase) is head-initial, regardless of learners’ L1. Importantly, learners still raise nonfinite forms, as in (18), meaning they have yet to acquire subject-verb agreement, suggesting that the acquisition of bound morphemes does not serve as a trigger for syntactic development in adult L2 acquisition in the same way that it does in child L1 acquisition (Clahsen and Muysken 1986, see also Vainikka and Young-Scholten, 2011). In stage 3, L2 learners begin to consistently produce correct agreement and verb raising becomes obligatory, leading to the projection of an agreement phrase (AgrP). Importantly, this AgrP differs from the AgrP in German, as it is head-initial instead of head-final. Agreement features are also acquired similarly regardless of a learner’s L1 grammar, providing further evidence against a strong role for L1 transfer at this stage (Vainikka and Young-Scholten 2011). In a final stage, L2 learners acquire a complementizer phrase (CP). However, as long as they continue to classify the AgrP as head-initial, learners continue to place the finite verb in second position in embedded clauses (all examples from Vainikka and Young-Scholten 1996b). (15)

Oya Zigarette trinken (Stage 1: Turkish L2 learner) Oya cigarette drink.INF “Oya smokes cigarettes.”

(16)

ich sprechen in meine Firma (Stage 1: Italian L2 learner) I speak.INF in my firm “I speak in/at my firm.”

(17)

jetzt brau Wohnungsamt fragen (Stage 2) now need.0/1SG housing authority ask.INF “Now [I] need to ask [the] housing authority”

(18)

ich sehen Schleier (Stage 2) I see.INF veil “I see the veil.”

Vainikka and Young-Scholten (2011) reported that English-speaking high school students exhibit a similar pattern of development over the course of an exchange year in Germany, although they pass through the FP-stage (stage 2) more quickly than the immigrant populations investigated in earlier research, perhaps due to increased L2 input as a result of

697

698

CARRIE JACKSON

living with German host families and attending German-speaking high schools. From this, they conclude that these developmental stages may not be limited to immigrant populations with less education and less overall L2 exposure – a critical step in evaluating the extent to which the extensive research on learner varieties extends to adult L2 acquisition more generally.

29.4 The Acquisition of Finiteness The L2 acquisition of finiteness (i.e., subject-verb agreement) is often investigated in tandem with the acquisition of word order in Germanic languages, because the acquisition of morphological finiteness marking (i.e., the production of bound morphemes to mark subject-verb agreement on thematic verbs) and syntactic finiteness (i.e., verb raising), are hypothesized to be linked in child L1 acquisition (Clahsen and Muysken 1986, Clahsen and Penke 1992). There is evidence from adult L2 Dutch and adult L2 German acquisition that auxiliary verbs play a critical role in the acquisition of subject-verb agreement and verb movement among untutored L2 learners, like the participants in the earlier ESF project (Klein and Perdue 1992, Parodi 2000, Vainikka and Young-Scholten 2011), and in more recent data from tutored learners, i.e., L2 learners who complete formal L2 language courses after emigrating to The Netherlands or Germany (van de Craats 2009, van de Craats and van Hoet 2010, Schimke 2011, Verhagen 2011). The precise acquisitional trajectory is influenced by L1 features, including the choice of auxiliary verb (e.g., is ‘is’ versus ga(at) ‘go(es)’ for Turkish versus Moroccan L2 Dutch learners), the preferred location of accompanying thematic verbs (i.e., sentence-final position versus earlier in the sentence), and the preferred forms for these thematic forms (i.e., finite versus nonfinite forms) (van de Craats 2009, van de Craats and van Hoet 2010, Schimke 2011, Verhagen 2011). However, commonalities across studies and different L1-L2 pairings in the general developmental pattern are very pronounced. L2 learners produce finite forms for modal, auxiliary and copula verbs prior to producing thematic verbs that are marked for number and person. L2 learners are more likely to place these inflected auxiliary verbs earlier in the sentence, prior to negators and other constituents, compared to nonfinite thematic verbs (but see Vainikka and Young-Scholten 2011). Further, L2 learners exhibit sensitivity to these distributional differences in elicited imitation and sentence matching tasks before they produce such patterns in their own productions (Schimke 2011, Verhagen 2011), although even here performance – especially when imitating correct verb placement with finite versus nonfinite thematic verbs – is linked to L2 learners’ ability to produce auxiliary verbs

Second Language Acquisition of Germanic Languages

during free production (Verhagen 2011) and the correct marking of third person singular agreement on thematic verbs (Schimke 2011). Such findings have led researchers to conclude that L2 learners use the production and placement of finite auxiliary verbs as an intermediate step in acquiring verb raising and subject-verb agreement. This research highlights that verb placement is not random even at earlier stages of acquisition and that L2 functional categories may be built incrementally over the course of L2 acquisition (Vainikka and YoungScholten 1996a, 1996b, 2011, see also Schimke 2011).

29.5 L2 Psycholinguistic Research L2 researchers also increasingly use psycholinguistic methods to investigate how L2 learners build the grammatical structure and meaning of an utterance during L2 comprehension and production (see MorganShort 2014, Juffs and Rodriguez 2015, for review). Here, research focuses less on the final outcome of such processes (e.g., whether an L2 learner comprehends an utterance), and more on how lexical items and grammatical structures are activated and retrieved, and how the structure and meaning of a sentence is built up incrementally, as one comprehends or produces an utterance. This research seeks to understand the underlying cognitive mechanisms that support the processing of L2 input and the production of L2 utterances, and the extent to which these mechanisms are similar to the mechanisms used by native speakers. When differences between L1 and L2 processing emerge, the question arises whether such differences reflect qualitative differences in how grammatical information is stored, retrieved, and used by L2 learners versus native speakers (Ullman 2005, Clahsen and Felser 2006), or whether such differences reflect quantitative differences between L2 speakers and native speakers in overall processing capacity and lexical retrieval speed, combined with potential interference from the co-activation of L1 structures, rather than fundamental differences in L2 versus native processing mechanisms per se (Dekydtspotter et al. 2006, Hopp 2010, Cunnings 2017). As with research on learner varieties, studies involving L2 learners of German and Dutch have played a central role in advancing our understanding of real-time L2 processing. Here I will focus on research on the processing of case marking in L2 German and of grammatical gender in L2 German and L2 Dutch, to showcase how research in German and Dutch has addressed key issues in L2 psycholinguistic research more generally. Readers are directed to Schimke and Hopp (2018), and the chapters therein, for a broader overview of L2 psycholinguistic research, and L2 psycholinguistic research on Germanic languages in particular.

699

700

CARRIE JACKSON

29.5.1

L2 Processing of Case Marking and Grammatical Role Assignment As outlined in Section 29.2.1, beyond strict V2-SOV rules for verb placement in German and Dutch, the ordering of constituents in a sentence is fairly flexible. In German, case marking on determiners in noun phrases (NPs) are thus the most reliable cue to assign grammatical roles (e.g., subject, object, indirect object). For instance, the subject and object in sentences (19) and (20) are identical even though the word order differs. (19)

Der Mann sieht den Bruder. the.NOM man sees the.ACC brother

(20)

Den Bruder sieht der Mann. the.ACC brother sees the.NOM man “The man sees the brother.”

This reliance on case marking differs from other languages, like English, where word order is the most reliable cue to identify grammatical roles. According to the Competition Model (MacWhinney 2008), such crosslinguistic differences in cue strength and reliability can lead to difficulties during L2 acquisition, as learners often rely on the strongest cues from their L1 to assign agent/patient roles when comprehending the L2 input (McDonald 1987, Kempe and MacWhinney 1998, Jackson 2007, Henry et al. 2009). Although case marking is the most reliable cue to identify the subject and object of a German sentence, even German native speakers exhibit a subject-first bias during real-time sentence processing. In the absence of unambiguous case marking, German native speakers interpret the first NP in a sentence as the subject, and exhibit longer reading times and associated processing difficulties upon encountering disambiguating morphosyntactic information (e.g., case marking on the second NP or number marking on the finite verb) that requires speakers to reanalyze this initial interpretation towards a less-preferred object-first structure (Bader and Meng 1999, Schlesewsky et al. 2000). L2 learners of German exhibit similar processing difficulties when reading sentences that disambiguate to a less-preferred object-first structure (Hopp 2006, Jackson 2008). However, Hopp (2006) and Jackson (2008) found that only highly-proficient L2 learners and/or learners with an analogous case-marking system in the L1 (e.g., Russian) exhibited processing difficulties immediately upon encountering disambiguating case-marking information. Less-proficient L2 learners’ processing difficulties emerged only at the end of the sentence. This suggests that only highly proficient L2 learners exhibit nativelike processing routines that include incremental reanalysis. In contrast, less proficient L2 learners use casemarking information to assign grammatical roles, but they may have difficulty doing so immediately upon encountering such information. Additional research finds that reanalyzing an initial subject-first interpretation may be more difficult for L2 learners than “native speakers” (Gerth

Second Language Acquisition of Germanic Languages

et al. 2017). The strength and immediacy of such reanalysis is also influenced by task-based factors, including the extent to which accompanying comprehension or grammaticality judgment tasks draw explicit attention to case-marking information (Jackson and Bobb 2009, Jackson and Dussias 2009), and whether target sentences are presented in isolation or as part of a larger discourse context that supports the information structural constraints modulating the use of object-first structures in German discourse (Hopp 2009). In an eye-tracking study, Jackson et al. (2012) also found that intermediate-level English L2 learners of German spent longer processing accusative-marked than nominative-marked determiners, even when they appeared in a subject-first sentence (e.g., Welche Ingenieurin traf den Chemiker gestern Nachmittag im Cafe´? ‘Which.NOM engineer met the.ACC chemist yesterday afternoon at.the cafe´?’). This points towards an intermediate stage in L2 acquisition where learners have identified the importance of case marking cues for determining grammatical roles, even if they do not yet exhibit clear signs of incremental structure building during real-time language comprehension (see also Henry 2015). Research with German L2 learners of Dutch reveals similar difficulties with the incremental reanalysis of subject-object ambiguities, even among L2 learners whose L1 and L2 are maximally similar. Havik et al. (2009) found no reading time differences at the disambiguating numbermarked verb for subject- versus object-relative clauses, like (21), among advanced German L2 learners of Dutch, even though both German and Dutch require verb-final word order with relative clauses. This contrasts with Dutch native speakers’ performance in the same experiment and previous research investigating the online processing of similar objectand subject-relative clauses among German native speakers (Schriefers et al. 1995). (21)

Daar is de machinist die de conducteurs heeft/hebben bevrijd uit het brandende treinstel. There is the train-driver.SG who the conductors.PL has.SG/have.PL freed from the burning train-carriage. “There is the train driver who has freed the conductors/who the conductors have freed from the burning train carriage.”

Havik et al. concluded that L2 learners may not incrementally assign grammatical roles and commit to a particular analysis during L2 processing, especially in the absence of disambiguating case-marking information prior to the thematic verb, regardless of cross-language syntactic similarity. However, in a follow-up study, Jackson and Roberts (2010) found that manipulating the animacy of the critical NPs led similarly proficient German L2 learners of Dutch to incrementally assign grammatical roles, resulting in reading time patterns at the disambiguating region that paralleled those of Dutch native speakers. Comparing their results to

701

702

CARRIE JACKSON

Havik et al., Jackson and Roberts concluded that, in the presence of sufficient lexical-semantic information, L2 learners can build argument structure in real time, even without the aid of disambiguating case-marking information and number marking on the thematic verb – further highlighting that at least under certain circumstances, L2 learners can apply processing routines that do not differ from those used by native speakers.

29.5.2 Grammatical Gender in L2 German and L2 Dutch In both German and Dutch, NPs are marked for grammatical gender. In German, all nouns are classified as either masculine, feminine, or neuter, whereas in Dutch, nouns are classified as common or neuter. These lexical categories have consequences for morphosyntactic agreement, as other words in a sentence, like determiners and adjectives, must be marked based on the gender classification of the noun they modify (see also Ku¨rschner, Chapter 12). Gender assignment in both languages is largely arbitrary, and the morphophonological cues to gender assignment that exist are probabilistic in nature (Ko¨pcke and Zubin 1984). Dominant L1 psycholinguistic models (Levelt et al. 1999) propose that gender information is stored at the lemma level, with each noun entry connected to an abstract gender node for that particular gender (i.e., masculine, feminine, or neuter in German; common or neuter in Dutch). When a noun’s lemma is activated during production or comprehension, its corresponding grammatical information stored at the lemma level – here, grammatical gender – is also activated, to then be used when producing or comprehending agreement features associated with that particular gender. Thus, the acquisition of L2 grammatical gender involves acquiring both lexical and syntactic knowledge, as learners must learn (i) the correct gender for each individual noun and (ii) which constituents in a sentence are marked for gender agreement, and what the corresponding morphological markers are for each gender category. Grammatical gender is notoriously difficult for L2 learners to master, with even highly proficient L2 learners exhibiting residual variability in both gender assignment and agreement (see Bordag and Pechmann 2018, Hopp 2018 for review). Paralleling child L1 acquisition (Szagun et al. 2007), English L2 learners of German are sensitive to morphophonological cues to gender assignment in German, with greater gender assignment accuracy and faster determiner+noun naming times for nouns containing a frequent and reliable cue to gender assignment (e.g., 90 percent of German words that end in –e are feminine, Wegener 2000), and lower accuracy and longer naming times for nouns with atypical gender assignment (e.g., das Auge ‘the.NEUT eye.NEUT’; whereby –e is usually associated with feminine, not neuter, nouns) (Bordag et al. 2006, Bobb et al. 2015). For L2 learners whose L1 also contains grammatical gender, accuracy and naming times are faster

Second Language Acquisition of Germanic Languages

for nouns that are assigned the same gender in both languages (e.g., Teller and talı´ȓ ‘plate’ are masculine in both German and Czech) than nouns with conflicting gender assignment (e.g., Haus ‘house’ is neuter in German but duº m ‘house’ is masculine in Czech), suggesting that gender nodes are shared across languages and both languages are co-activated during language production (Bordag and Pechmann 2007, Salamoura and Williams 2007, Klassen 2016, Fowler 2017, but see Lemho¨fer et al. 2008 for an alternative explanation). Sabourin et al. (2006) further showed that the extent to which L2 learners exhibit nativelike accuracy in gender agreement depends on the degree of cross-linguistic similarity between an L2 learner’s two languages. In a grammaticality judgment task involving nouns and their corresponding gender-marked relative pronoun, German L2 learners of Dutch outperformed Romance L2 learners of Dutch (Spanish, French, and Italian L1), although their performance was still below that of Dutch native speakers. The German and Romance learners, in turn, outperformed English L2 learners of Dutch. Sabourin et al. concluded that the English learners exhibited the lowest accuracy because they did not possess grammatical gender in their L1. Romance learners were more accurate because Romance languages contain the abstract feature of grammatical gender, even if the system – in terms of the grammatical gender assigned to individual nouns and how gender agreement is realized – exhibits little overlap with Dutch. Sabourin et al. argued that the German learners exhibited the highest accuracy because grammatical gender in German and Dutch largely overlap, both in terms of gender assignment and how gender agreement is marked (see also Sabourin et al. 2008, Meulman et al. 2014, 2015, and Hopp and Lemmerth 2016, for related findings using eyetracking or event-related potentials). Using event-related potentials (ERPs), Lemho¨fer et al. (2014) reported that – in contrast to other studies claiming persistent difficulty in the realtime processing of L2 grammatical gender violations (Meulman et al. 2014, 2015) – even intermediate German L2 learners of Dutch can exhibit nativelike neurophysiological responses to grammatical gender violations in L2 Dutch (e.g., Het volk verlangt naar *de einde van de dictatuur ‘The people longed for the.COM end.NEUT of the dictatorship’). However, such effects emerged only once responses were coded according to participants’ “subjective” gender assignment (i.e., which gender they assigned to the target Dutch nouns in an offline gender naming questionnaire), rather than the “objective” gender assignment for each target noun (i.e., the gender assigned to each noun according to Dutch native speakers). Together, these studies underscore how L2 learners struggle to acquire the lexical and syntactic knowledge necessary for producing and comprehending L2 grammatical gender. However, research involving learners whose L1 also possesses grammatical gender reveals that learners’ persistent difficulties with this grammatical feature lie not just with acquiring a novel L2 feature, but also

703

704

CARRIE JACKSON

stem from cross-language interaction and competing L1 information during real-time language processing.

29.5.3 Predictive Processing in L2 Learners Researchers have also begun to investigate predictive processes during real-time language comprehension, specifically the processes whereby native speakers and L2 learners use linguistic cues to anticipate upcoming words and morphosyntactic structures. Prediction facilitates language comprehension and aids communication in dialogue (see Huettig 2015 for review), and may even play a role in L1 and L2 acquisition (Dell and Chang 2014, Phillips and Ehrenhofer 2015, Hopp 2016). However, L2 learners often have difficulty using predictive cues in the L2. In light of research showing that even native speakers may not always predict upcoming input during real-time language processing, the central goal of this research is to identify the factors that modulate L2 prediction (see Kaan 2014 for review). Scherag et al. (2004) found that German native speakers and English L2 learners of German used semantic information on adjectives to facilitate recognition of a subsequent noun, in that participants were quicker to decide that the word Gesicht ‘face.NEUT’ was a real German word when preceded by the word faltiges ‘wrinkled’, than when the word faltiges preceded a semantically unrelated word, like Geru¨cht ‘rumor’. However, only German native speakers similarly used gender marking on the preceding adjective to facilitate subsequent word recognition (e.g., faltiges ‘wrinkled.NEUT’ facilitated recognition of Gesicht ‘face.NEUT’ more than Haut ‘skin.FEM’). Hopp (2013) reported that overall gender assignment accuracy was a critical factor in whether highly proficient English L2 learners of German used gender to predict upcoming words in an experiment using the visual world paradigm. In such experiments, participants’ eye movements toward different images are tracked while they listen to a target sentence. In Hopp’s study, participants heard sentences like Wo ist die gelbe Karte? ‘Where is the.FEM yellow.FEM card.FEM?’. For half of the trials, three of the four target images contained yellow nouns with the same grammatical gender (e.g., a card, Karte.FEM, a cup, Tasse.FEM, a flower, Blume.FEM) but for half of the trials, the three yellow target images bore different gender (e.g., a card, Karte.FEM, a dress, Kleid.NEUT, and a button, Knopf.MASC). If participants use grammatical gender to predict upcoming nouns, they should look more quickly towards the target noun (i.e., the card) on the trials in which competitor nouns bear different gender than the trials in which target and competitor nouns all bear the same gender. Indeed, German native speakers used gender to predict upcoming target nouns. However, only those L2 learners with near perfect gender assignment accuracy, as measured by their accuracy in naming the target nouns and their gender

Second Language Acquisition of Germanic Languages

prior to the onset of each trial, used gender information predictively. In a follow-up study, Hopp (2016) showed that after training L2 learners on grammatical gender assignment, those L2 learners who exhibited high gender recall accuracy on a production task after the training unit then used gender predictively on the visual world paradigm post test. These studies underscore that having grammatical gender in the L1 is not a necessary prerequisite for using this grammatical feature during realtime L2 comprehension, although gender assignment accuracy can influence the strength of such predictive processes. Turning to case marking, Hopp (2015) found that English L2 learners of German did not use case marking on the first noun in unambiguous OVS and SVO sentences, like (22) and (23), to predict whether the second noun would be a logical agent or patient of the utterance. (22)

Der Hase frisst gleich die Blume. (SVO) the.NOM rabbit eats soon the.ACC flower “The rabbit soon eats the flower.”

(23)

Den Hasen frisst gleich der Fuchs. (OVS) The.ACC rabbit eats soon the.NOM fox “The fox soon eats the rabbit.”

L2 learners, regardless of proficiency, exhibited a subject-first preference, looking towards the picture of a flower independent of whether the first NP (i.e., Hase ‘rabbit’) had been introduced as a nominative-marked subject or an accusative-marked object. Hopp concluded that while the L2 learners used lexical-semantic information on the verb to anticipate upcoming input – they heard the noun rabbit and the verb eat, and then looked towards the likely patient of that action – they did not use case marking, assuming instead that the rabbit was always the subject of the utterance. This begs the question of whether L2 learners always use lexical-semantic information predictively, or whether even the use of lexical-semantic cues may vary according to how such information is instantiated in the L1. Van Bergen and Flecken (2017) used the visual world paradigm to measure whether Dutch native speakers and German, French, and English L2 learners of Dutch used Dutch placement verbs zetten ‘set’ and leggen ‘lay’ to anticipate the placement of objects on the computer screen, as zetten is used to describe the placement of objects on their natural base (e.g., a bottle standing up) whereas leggen is used to describe the placement of objects in a position other than on their natural base (e.g., a bottle lying down). Similar to Dutch native speakers, German L2 learners of Dutch used the placement information encoded by verb choice to quickly look towards the target object (e.g., a bottle standing up after hearing the verb zetten), which van Bergen and Flecken attribute to the fact that German similarly encodes placement information via the choice between stellen ‘set’ and legen ‘lay’. Equally proficient English and French L2 learners of Dutch did not exhibit significant predictive

705

706

CARRIE JACKSON

eye movements, presumably because they lack this distinction in their L2 (i.e., a single verb, like put, encodes both placement events). This work on L2 predictive processing highlights the difficulties L2 learners have in using linguistic cues to anticipate upcoming input, including the accuracy and stability of lexical and syntactic representations (Hopp 2013) and competing information from the L1 (Hopp and Lemmerth 2016, van Bergen and Flecken 2017). At the same time, some of the same issues that hinder predictive processing in L2 learners also modulate predictive processing among native speakers (see Kaan 2014).

29.6 Conclusion As demonstrated by the review of L2 research presented here, we have made much progress over the last 40 years in understanding the linguistic and cognitive mechanisms that support L2 acquisition, and research involving the L2 acquisition of German and Dutch has been critical to the success of this endeavor. However, many debates remain unresolved and many questions remain unanswered, especially regarding the interaction between the development of L2 linguistic representations and the realtime processing of L2 input, and particularly how different instructional methods, as well as age of acquisition, may modulate such interactions. Especially with the recent influx of migrants to Germany and The Netherlands, many projects are underway to investigate precisely these questions. Thus, without a doubt, research on the L2 acquisition of Germanic languages will continue to make significant contributions to the field of L2 acquisition into the future.

References Bader, M. and M. Meng 1999. “Subject-object ambiguities in German embedded clauses: An across-the-board comparison,” Journal of Psycholinguistic Research 28: 121–143. Baten, K. 2011. “Processability Theory and German case acquisition,” Language Learning 61: 405–455. Becker, A. and M. Carroll 1997. The Acquisition of Spatial Relations in a Second Language. Amsterdam and Philadelphia: John Benjamins. Bergen, G. van and M. Flecken 2017. “Putting things in new places: Linguistic experience modulates the predictive power of placement verb semantics,” Journal of Memory and Language 92: 26–42. Birdsong, D. 1992. “Ultimate attainment in second language acquisition,” Language 68: 706–755. Bley-Vroman, R. 1990. “The logical problem of foreign language learning,” Linguistic Analysis 20: 3–49.

Second Language Acquisition of Germanic Languages

Blom, E. 2010. “Effects of input on the early grammatical development of bilingual children,” International Journal of Bilingualism 14: 442–446. Bobb, S., J. Kroll, and C. Jackson 2015. “Lexical constraints in second language learning: Evidence on grammatical gender in German,” Bilingualism: Language and Cognition 18: 502–523. Bohnacker, U. 2006. “When Swedes begin to learn German: From V2 to V2,” Second Language Research 22: 443–486. Bongaerts, T., S. Mennen, and F. van der Slik 2000. “Authenticity of pronunciation in naturalistic second language acquisition: The case of very advanced learners of Dutch as a second language,” Studia Linguistica 545: 298–308. Bordag, D., A. Opitz and T. Pechmann 2006. “Gender processing in L1 and L2: The role of noun termination,” Journal of Experimental Psychology: Learning, Memory, and Cognition 32: 1090–1101. Bordag, D. and T. Pechmann 2007. “Factors influencing L2 gender processing,” Bilingualism: Language and Cognition 10: 299–314. Bordag, D. and T. Pechmann 2018. “Genuserwerb und -verarbeitung in einer L2.” In S. Schimke and H. Hopp (eds.), Sprachverarbeitung im Zweitspracherwerb. Berlin: Mouton de Gruyter: 75–92. Clahsen, H. 1990. “The comparative study of first and second language development,” Studies in Second Language Acquisition 12: 135–153. Clahsen, H. and C. Felser 2006. “Grammatical processing in language learners,” Applied Psycholinguistics 27: 3–42. Clahsen, H., J. Meisel, and M. Pienemann 1983. Deutsch als Zweitsprache. Der Spracherwerb ausla¨ndischer Arbeiter. Tu¨bingen: Narr. Clahsen, H. and P. Muysken 1986. “The availability of universal grammar to adult and child learners – a study of the acquisition of German word order,” Second Language Research 2: 93–118. Clahsen, H. and M. Penke 1992. “The acquisition of agreement morphology and its syntactic consequences: New evidence on German child language from the Simone-corpus.” In J. Meisel (ed.), The Acquisition of Verb Placement. Dordrecht: Kluwer: 181–223. Chilla, S., M. Rothweiler, and E. Babur 2010. Kindliche Mehrsprachigkeit. Grundlagen, Sto¨rungen, Diagnostik. Munich: Reinhardt. Craats, I. van de 2009. “The role of is in the acquisition of finiteness by adult Turkish learners of Dutch,” Studies in Second Language Acquisition 31: 59–92. Craats, I. van de and R. van Hoet 2010. “Dummy auxiliaries in the second language acquisition of Moroccan learners of Dutch: Form and function,” Second Language Research 26: 473–500. Cunnings, I. 2017. “Parsing and working memory in bilingual sentence processing,” Bilingualism: Language and Cognition 20: 659–678. Dekydtspotter, L., B. Schwartz, and R. Sprouse 2006. “The comparative fallacy in L2 processing research.” In M. O’Brien, C. Shea, and J. Archibald (eds.), Proceedings of the 8th Generative Approaches to Second

707

708

CARRIE JACKSON

Language Acquisition Conference (GASLA 2006). Somerville, MA: Cascadilla Proceedings Project: 33–40. Dell, G. and F. Chang 2014. “The P-chain: relating sentence production and its disorders to comprehension and acquisition,” Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 369: 20120394. http:// doi.org/10.1098/rstb.2012.0394. Diehl, E., H. Christen, and S. Leuenberger 2000. Grammatikunterricht: Alles fu¨r der Katz? Untersuchungen zum Zweitsprachenerwerb Deutsch. Tu¨bingen: Niemeyer. Dietrich, R., W. Klein, and C. Noyau 1995. The Acquisition of Temporality in a Second Language. Amsterdam and Philadelphia: John Benjamins. Dimroth, C. 2013. “Learner varieties.” In A. Carol (ed.), The Encyclopedia of Applied Linguistics. DOI:10.1002/9781405198431.wbeal0673. Dimroth, C. and M. Starren (eds.) 2003. Information Structure and the Dynamics of Language Acquisition. Amsterdam and Philadelphia: John Benjamins. Eckerth, J., K. Schramm, and E. Tschirner 2009. “Review of recent research (2002–2008) on applied linguistics and language teaching with specific reference to L2 German (part 1),” Language Teaching 42: 41–66. Fowler, C. 2017. Cross-language Interaction During Language Production among L1 German-L2 Italian Bilinguals. Ph.D. dissertation, The Pennsylvania State University. Gass, S. 2013. Second Language Acquisition: An Introductory Course. New York: Routledge. Gawlitzek-Maiwald, I. and R. Tracy 1996. “Bilingual bootstrapping,” Linguistics 34: 901–926. Gerth, S., C. Otto, C. Felser, and Y. Nam 2017. “Strength of garden-path effects in native and non-native speakers’ processing of object-subject ambiguities,” International Journal of Bilingualism 21: 125–144. Glahn, E., G. Haº kansson, B. Hammarberg, A. Holmen, A. Hvenekilde, and K. Lund 2001. “Processability in Scandinavian second language acquisition,” Studies in Second Language Acquisition 23: 389–416. Grosjean, F. 1989. “Neurolinguists beware! The bilingual is not two monolinguals in one person,” Brain and Language 36: 3–15. Gut, U. 2009. Non-Native Speech: A Corpus-Based Analysis of the Phonetic and Phonological Properties of L2 English and L2 German. New York and Frankfurt: Peter Lang Verlag. Haberzettl, S. 2005. Der Erwerb der Verbstellungsregeln in der Zweitsprache Deutsch durch Kinder mit russischer und tu¨rkischer Muttersprache. Tu¨bingen: Niemeyer. Haº kansson, G. and Norrby, C. 2010. “Environmental influence on language acquisition: Comparing second and foreign language acquisition of Swedish,” Language Learning 60: 628–650. Haº kansson, G., M. Pienemann, and S. Sayehli 2002. “Transfer and typological proximity in the context of second language processing,” Second Language Research 18: 250–273.

Second Language Acquisition of Germanic Languages

Havik, E., L. Roberts, R. van Hout, R. Schreuder, and M. Haverkort 2009. “Processing subject-object ambiguities in the L2: A self-paced reading study with German L2 learners of Dutch,” Language Learning 59: 73–112. Hendricks, H. (ed.) 2005. The Structure of Learner Varieties. Berlin: Mouton de Gruyter. Henry, N. 2015. Morphosyntactic Processing, Cue Interaction, and the Effects of Instruction: An Investigation of Processing Instruction and the Acquisition of Case Marking in L2 German. Ph.D. dissertation, The Pennsylvania State University. Henry, N., H. Culman, and B. van Patten 2009. “More on the effects of explicit information in instructed SLA,” Studies in Second Language Acquisition 31: 559–575. Hopp, H. 2006. “Syntactic features and reanalysis in near-native processing,” Second Language Research 22: 369–397. Hopp, H. 2009. “The syntax-discourse interface in near-native L2 acquisition: Off-line and online performance,” Bilingualism: Language and Cognition 12: 463–483. Hopp, H. 2010. “Ultimate attainment in L2 inflection: Performance similarities between non-native and native speakers,” Lingua 120: 901–931. Hopp, H. 2013. “Grammatical gender in adult L2 acquisition: Relations between lexical and syntactic variability,” Second Language Research 29: 33–56. Hopp, H. 2015. “Semantics and morphosyntax in L2 predictive sentence processing,” International Review of Applied Linguistics 53: 277–306. Hopp, H. 2016. “Learning (not) to predict: Grammatical gender processing in second language acquisition,” Second Language Research 32: 277–307. Hopp, H. 2018. “Kasus- und Genusverarbeitung des Deutschen im Satzkontext.” In S. Schimke and H. Hopp (eds.), Sprachverarbeitung im Zweitspracherwerb. Berlin: Mouton de Gruyter: 141–168. Hopp, H. and N. Lemmerth 2016. “Lexical and syntactic congruency in L2 predictive gender processing,” Studies in Second Language Acquisition. Advance online publication. DOI:10.1017/S0272263116000437. Huettig, F. 2015. “Four central questions about prediction in language processing,” Brain Research 1626: 118–135. Jackson, C. 2007. “The use and non-use of semantic information, word order, and case markings during comprehension by L2 learners of German,” Modern Language Journal 91: 418–432. Jackson, C. 2008. “Proficiency level and the interaction of lexical and morphosyntactic information during L2 sentence processing,” Language Learning 58: 875–909. Jackson, C. and S. Bobb 2009. “The processing and comprehension of wh-questions among second language speakers of German,” Applied Psycholinguistics 30: 630–636. Jackson, C. and P. Dussias 2009. “Cross-linguistic differences and their impact on L2 sentence processing,” Bilingualism: Language and Cognition 12: 65–82.

709

710

CARRIE JACKSON

Jackson, C., P. Dussias, and A. Hristova 2012. “Using eye-tracking to study on-line processing among intermediate L2 learners of German,” International Review of Applied Linguistics 50: 101–33. Jackson, C., C. Johnson-Fowler, B. Gavin, and N. Henry 2018. “Zusammenha¨nge zwischen der Sprachverarbeitung und dem Erwerb neuer Strukturen bei erwachsenen L2-Lernern.” In S. Schimke and H. Hopp (eds.), Sprachverarbeitung im Zweitspracherwerb. Berlin: Mouton de Gruyter: 299–324. Jackson, C. and L. Roberts 2010. “Animacy affects the processing of subject-object ambiguities in the second language: Evidence from self-paced reading with German second language learners of Dutch,” Applied Psycholinguistics 31: 671–691. Jansen, L. 2008. “Acquisition of German word order in tutored learners: A cross-sectional study in a wider theoretical context,” Language Learning 58: 185–231. Jordens, P. 2001. “Constraints on the shape of second language learner varieties,” International Review of Applied Linguistics 39: 51–74. Juffs, A. and G. Rodriguez 2015. Second Language Sentence Processing. New York: Routledge. Kaan, E. 2014. “Predictive sentence processing in L2 and L1: What is different?” Linguistic Approaches to Bilingualism 4: 257–282. Kaplan, R. and J. Bresnan 1982. “Lexical-functional grammar: A formal system for grammatical representation.” In J. Bresnan (ed.), The Mental Representation of Grammatical Relations. Cambridge, MA: MIT Press: 173–281. Kempe, V. and B. MacWhinney 1998. “The acquisition of case marking by adult learners of Russian and German,” Studies in Second Language Acquisition 20: 543–587. Kempen, G. and E. Hoenkamp 1987. “An incremental procedural grammar for sentence formulation,” Cognitive Science 11: 201–258. Klassen, R. 2016. “The representation of asymmetric grammatical gender systems in the bilingual mental lexicon,” Probus 28: 9–28. Klein, W. and N. Dittmar 1979. Developing Grammars: The Acquisition of German Syntax by Foreign Workers. Berlin: Springer. Klein, W. and C. Dimroth 2009. “Untutored second language acquisition.” In T. Bhatia and W. Ritchie (eds.), The New Handbook of Second Language Acquisition. Bingley: Emerald: 503–522. Klein, W. and C. Perdue 1992. Utterance Structure. Amsterdam and Philadelphia: John Benjamins. Klein, W. and C. Perdue 1997. “The Basic Variety (or couldn’t natural languages be simpler?)” Second Language Research 13: 301–347. Ko¨pcke, K-M. and D. Zubin 1984. “Sechs Prinzipien fu¨r die Genuszuweisung im Deutschen: Ein Beitrag zur natu¨rlichen Klassifikation,” Linguistische Berichte 93: 26–50.

Second Language Acquisition of Germanic Languages

Lemho¨fer, K., H. Schriefers, and P. Indefrey 2014. “Idiosyncratic genders: Syntactic processing in second language comprehension uses subjective feature representations,” Journal of Cognitive Neuroscience 26: 1428–1444. Lemho¨fer, K., K. Spalek, and H. Schriefers 2008. “Cross-language effects on grammatical gender in bilingual word recognition and production,” Journal of Memory and Language 59: 312–330. Levelt, W. 1989. Speaking: From Intention to Articulation. Cambridge, MA: MIT Press. Levelt, W., A. Roelofs, and A. Meyer 1999. “A theory of lexical access in speech production,” Behavioral and Brain Sciences 22: 1–75. MacWhinney, B. 2008. “A unified model.” In P. Robinson and N. C. Ellis (eds.), Handbook of Cognitive Linguistics and Second Language Acquisition. New York: Routledge: 341–371. McDonald, J. 1987. “Sentence interpretation in bilingual speakers of English and Dutch,” Applied Psycholinguistics 4: 379–413. Meisel, J. 1997. “The L2 Basic Variety as an I language,” Second Language Research 13: 374–385. Meisel, J., H. Clahsen, and M. Pienemann 1981. “On determining developmental stages in natural second language acquisition,” Studies in Second Language Acquisition 3: 109–135. Meulman, N., L. Stowe, S. Sprenger, M. Bresser, and M. Schmid 2014. “An ERP study on L2 syntax processing: When do learners fail?” Frontiers in Psychology 55: 1072. Meulman, N., M. Wieling, S. Sprenger, L. Stowe, and M. Schmid 2015. “Age effects in L2 grammar processing as revealed by ERPs and how (not) to study them,” PlosONE 10: e0143328. Mitchell, R., F. Myles, and E. Marsden 2013. Second Language Learning Theories. New York: Routledge. Morgan-Short, K. 2014. “Electrophysiological approaches to studying second language acquisition: A field reaching its potential,” Annual Review of Applied Linguistics 34: 15–36. Moyer, A. 2004. Age, Accent and Experience in Second Language Acquisition. Bristol: Multilingual Matters. Mu¨ller, N., T. Kupisch, K. Schmitz, and K. Cantone 2009. Einfu¨hrung in die Mehrsprachigkeitsforschung: Deutsch, Franzo¨sisch, Italienisch. Tu¨bingen: Narr. O’Brien, M. G. 2004. “Pronunciation Matters,” Die Unterrichtspraxis 37: 1–9. Parodi, T. 2000. “Finiteness and verb placement in second language acquisition,” Second Language Research 16: 355–381. Phillips, C. and E. Ehrenhofer 2015. “The role of language processing in language acquisition,” Linguistic Approaches to Bilingualism 5: 409–453. Pienemann, M. 1998. Language Processing and Second Language Development: Processability Theory. Amsterdam andPhiladelphia: John Benjamins. Pienemann, M. (ed.) 2005. Cross-Linguistic Aspects of Processability Theory. Amsterdam and Philadelphia: John Benjamins.

711

712

CARRIE JACKSON

Pienemann, M. and G. Haº kansson 2007. “Full transfer vs. developmentally moderated transfer: A reply to Bohnacker,” Second Language Research 23: 485–493. Rizzi, L. and I. Roberts 1989. “Complex inversion in French,” Probus 1: 1–30. Sabourin, L. and L. Stowe 2008. “Second language processing: When are first and second languages processed similarly?” Second Language Research 24: 397–430. Sabourin, L., L. Stowe, and G. de Haan 2006. “Transfer effects in learning a second language grammatical gender system,” Second Language Research 22: 1–29. Salamoura, A. and J. Williams 2007. “The representation of grammatical gender in the bilingual lexicon: Evidence from Greek and German,” Bilingualism: Language and Cognition 10: 257–275. Scherag, A., L. Demuth, F. Ro¨sler, H. Neville, and B. Ro¨der 2004. “The effects of late acquisition of L2 and the consequences of immigration on L1 for semantic and morpho-syntactic language aspects,” Cognition 93: B97–B108. Schimke, S. 2011. “Variable verb-placement in second language German and French: Evidence from production and elicited imitation of finite and nonfinite negated sentences,” Applied Psycholinguistics 32: 635–685. Schimke, S. and H. Hopp (eds.) 2018. Sprachverarbeitung im Zweitspracherwerb. Berlin: Mouton de Gruyter. Schlesewsky, M., G. Fanselow, R. Kliegl, and J. Krems 2000. “The subject preference in the processing of locally ambiguous wh-questions in German.” In B. Hemforth and L. Konieczny (eds.), German Sentence Processing. Dordrecht: Kluwer: 65–93. Schriefers, H., A. Friederici and K. Ku¨hn 1995. “The processing of locally ambiguous relative clauses in German,” Journal of Memory and Language 34: 499–520. Schwartz, B. and R. Sprouse 1994. “Word order and nominative case in non-native language acquisition: A longitudinal study of (L1 Turkish) German interlanguage.” In T. Hoekstra and B. Schwartz (eds.), Language Acquisition Studies in Honor of Kenneth Wexler from the 1991 GLOW Workshops. Amsterdam and Philadelphia: John Benjamins: 317–368. Schwartz, B. and R. Sprouse 1996. “L2 cognitive states and the Full Transfer/Full Access model,” Second Language Research 12: 40–72. Szagun, G., B. Stumper, N. Sondag, and M. Franik 2007. “The acquisition of gender marking by young German-speaking children: Evidence for learning guided by phonological regularities,” Journal of Child Language 34: 445–471. Ullmann, M. 2005. “A cognitive neuroscience perspective on second language acquisition: The declarative/procedural model.” In C. Sanz (ed.), Mind and Context in Adult Second Language Acquisition: Methods, Theory, and Practice. Washington, DC: Georgetown University Press: 141–178. Unsworth, S. 2016. “Early child L2 acquisition: Age or input effects? Neither, or both?” Journal of Child Language 43: 603–634.

Second Language Acquisition of Germanic Languages

Vainikka, A. and M. Young-Scholten 1996a. “The early stages in adult L2 syntax: Additional evidence from Romance speakers,” Second Language Research 12: 140–176. Vainikka, A. and M. Young-Scholten 1996b. “Gradual development of L2 phrase structure,” Second Language Research 12: 7–39. Vainikka, A. and M. Young-Scholten 2011. The Acquisition of German: Introducing Organic Grammar. Berlin: Mouton de Gruyter. Verhagen, J. 2011. “Verb placement in second language acquisition: Experimental evidence for the different behavior of auxiliary and lexical verbs,” Applied Psycholinguistics 32: 821–858. Wegener, H. 2000. “German gender in children’s second language acquisition.” In B. Unterbeck, M. Rissanen, T. Nevalainen, and M. Saari (eds.), Gender in Grammar and Cognition. Berlin: Mouton de Gruyter: 511–544. White, L. 2003. Second Language Acquisition and Universal Grammar. Cambridge University Press.

713

Chapter 30 Urban Speech Styles of Germanic Languages Pia Quist and Bente A. Svendsen

30.1 Introduction This article provides an overview of the last two decades of sociolinguistic research on contact induced speech styles in Northern Europe. Referencing parallel findings in Germany, the UK, and in the Netherlands, our focus will be on studies that have been conducted in Scandinavian cities. Across Germanic languages, sociolinguists have documented the emergence of urban speech styles in multilingual neighborhoods among young people with and without migration backgrounds (cf., e.g., the edited volumes Ka¨llstro¨m and Lindberg 2011, Kern and Selting 2011, Nortier and Svendsen 2015, Quist and Svendsen 2010). These styles have been denoted ethnolects or multiethnolects (Clyne 2000, Quist 2000, Sollid 2013, Cheshire et al. 2011) or they have been given more emic labels, such as Rinkeby Swedish in Stockholm, Straattaal in Amsterdam, Kiezdeutsch in Berlin, and Kebab Norwegian in Oslo (Aasheim 1995, Appel and Schoonen 2005, Kotsinas 1988, Wiese 2011, 2012). These speech styles are seen as connected-but-distinct from the migrant languages in a given locality (e.g., Turkish and Arabic), national standard language(s), traditional nonstandard dialect and adult second language speaker styles or learner language (e.g., such as “Gastarbeiterdeutsch”, Hinnenkamp 2003: 33–34), as well as from “prestigious counter-standard styles” (such as American Vernacular Black English) (Rampton 2015: 39). Contemporary urban speech styles are noted by speakers beyond their localities of origin and they are represented in media (Androutsopoulos 2010, Quist 2016, Svendsen and Marzo 2015), in commercials (Milani et al. 2015) and popular culture (Cutler and

Bente A. Svendsen’s part of this work is partly supported by the Research Council of Norway through its Centres of Excellence funding scheme, project number 223265 and the project Family language policy in multilingual transcultural families in Oslo (MultiFam), project number 240725.

Urban Speech Styles of Germanic Languages

Røyneland 2015), as well as in the informal speech of people outside of the locality in question (Rampton 2015). Within sociolinguistics, there is, however, no consensus on how to label these speech styles and the particular label is often anchored in the researcher’s theoretical and methodological stance and research interest (Quist 2008), such as “sociolect” (Kotsinas 1988), “multiethnolect” (Clyne 2000; Quist 2000), “ethnolect” (Appel and Muysken 1987), “multiethnic urban heteroglossia” (Blackledge and Creese 2015), “late modern urban youth style” (Madsen 2008), or “contemporary urban vernacular” (Rampton 2015). In this article, we use as a general operational term “contemporary urban speech styles” coined to include Rampton’s (2015) description of “contemporary urban vernacular”, as well as the criticism raised against the term “vernacular” (Cornips, Jaspers and De Rooij 2015) suggesting to replace “vernacular” with “style” (Eckert 2004, Svendsen 2015). The word “contemporary” indicates that the emergence of these speech styles is of a recent phenomenon. This is, however, by no means our opinion. There are older contact induced speech styles such as Tsotsitaal in South Africa, Sheng in Kenya (Kiessling and Mous 2004), Cite´taal in Flanders (Marzo and Ceuleers 2011) and “ethnolects” in language-shift contexts from Sa´mi/Kven to Norwegian in Northern Norway (Sollid 2013). In this article, however, our focus will be on the recent and ongoing sociolinguistic developments as we will restrict ourselves to the last two decades of sociolinguistic research conducted in multilingual urban spaces. That said, there is still a need to conduct comparative studies across time of the parallels and differences between “older” and more recent sociolinguistic development in social and linguistic contact zones. We will, moreover, narrow our overview to studies conducted in urban spaces in Europe where a Germanic language is included with an emphasis on Scandinavian studies (cf., Kiesling 2005, Kiessling and Mous 2004, and Dorleijn et al. 2015 for studies in African and Australian contexts and where a Germanic language is involved). Notwithstanding our broad scope, we have not managed to include each and every study conducted in this field; we have concentrated on the linguistic parallels that have been documented (see Section 30.2 below), and on the ways contemporary urban speech styles are used as a marker of identification as it is manifested in informal talk, perceived by others and the ways these speech styles and their alleged speakers are portrayed and valorised (Agha 2007) in the media (see Section 30.3 below). Our appraisals are driven by a plea for synthesized knowledge on the semiotics of contemporary urban speech styles and the ways we experience and relate to the sociolinguistic complexity in the wake of globalization in Europe.

715

716

PIA QUIST AND BENTE A. SVENDSEN

30.2 Contemporary Urban Speech Styles of Germanic Languages Scandinavian linguists have contributed significantly to this field of research, pioneered by the work of Ulla-Britt Kotsinas in the 1980s. Compared to other European countries, linguists in Scandinavia were relatively early in carrying out studies of linguistic practices in multilingual urban spaces. The work by Kotsinas (1988) of so-called rinkebysvenska (‘Rinkeby Swedish’) paved the way for sociolinguistic studies of multilingual urban speech in Scandinavia; she contributed to moving the study of multilingual youth from a second language learning approach by applying a sociolinguistic perspective and describing the language of the young bilingual people in its own right, i.e., as a fully valid variety of Swedish and not as an incomplete learner language.1 Considering the demographic developments of the post-World War II period in Scandinavia, it is not surprising it was in Sweden that the first sociolinguistic study of linguistic practices in multilingual urban spaces was carried out since Sweden has a long history of immigration. Approximately 16 percent of Sweden’s 10 million residents are born abroad, and a total of 21 percent including children born in Sweden have “foreign-born” parents (Statistics Sweden 2017). This is considerably more than in Denmark and Norway. In these two countries the socalled “immigrant populations” are of similar proportions, i.e., 12.33 percent in Denmark and 13.8 percent in Norway, including children born in Norway and Denmark respectively to “foreign-born” parents (Statistics Denmark 2017, Statistics Norway 2017). The parallel sociodemographic developments of the three Scandinavian welfare states display on the one hand relatively similar sociopolitical backgrounds for young people in multilingual urban places. On the other hand, the apparent socioeconomic similarities in the mainland Scandinavian countries may fluctuate in times of higher unemployment rates; for example, in May 2013 adolescents in segregated and culturally diverse neighborhoods in Stockholm participated in violent acts. Although it is beyond the scope of this article, a comparison across mainland Scandinavian research in multilingual urban spaces may shed light on sociopolitical and language ideological differences between the three countries that may have an impact on the conditional contexts for contemporary urban speech styles (see Madsen and Svendsen 2015, Quist 2018).

1

The sociolinguistic perspective has since been applied and developed in larger research projects all over Scandinavia, especially during the first ten years of the twenty-first century such as in the Swedish SUF-project (Language and language use among youth in multilingual urban settings, see Boyd 2010) and the Norwegian UPUS/Oslo-project (Developmental processes in urban linguistic settings, see Svendsen 2010).

Urban Speech Styles of Germanic Languages

30.3 Structure In the following three sections, we present and compare results from the Scandinavian and Northern European studies, with an emphasis on Scandinavia, in terms of the described structures of contemporary urban speech styles, and in terms of perceptions and media coverages of the styles. The Scandinavian languages are quite similar in grammar and vocabulary. Historically, Danish, Norwegian, and Swedish belong to the Old Norse branch of the Indo-European languages; now and then they are described as a dialect continuum with no internal linguistic borders. The three languages are still mutually intelligible, at least with some effort (see Gooskens, Chapter 32). Hence, a comparison across the mainland Scandinavian studies reveals parallel linguistic developments that, we would argue, have emerged as a result of the relatively similar Scandinavian linguistic contexts.

30.3.1 Variation of the Verb-Second Constraint One of the structural features characteristic of contemporary urban speech styles which has been documented in various studies across Germanic languages is the variation in the syntactic verb-second constraint (V2) (Freywald et al. 2015, Ganuza 2008, Kotsinas 1988, Quist 2000, Nortier 2001, Svendsen and Røyneland 2008, Opsahl and Nistov 2010). According to Vikner (Chapter 16), most Germanic languages (not including Middle English and Modern English) adhere to the V2 constraint: “This means that the finite verb occupies the second position in the clause, irrespective of which constituent occupies the first position.” In a clause where another element has been topicalised or fronted (X), such as an adverbial or an embedded clause, and in interrogatives, the canonical subject position is after the finite verb, and before any of the verb complements (XVS), as exemplified in Norwegian in (1b), marked in bold: (1a)

han gikk paº kino i gaº r ‘he went to the cinema yesterday’

(1b)

i gaº r gikk han paº kino yesterday went he to the cinema ‘yesterday he went to the cinema’

In the spoken Scandinavian (standard) languages the V2-rule, however, might be abolished when the fronted element is extraposed, that is when there is a pause after the fronted element and the following clause. In six Scandinavian cities – Oslo, Copenhagen, Aarhus, Stockholm, Malmo¨, and Gothenburg – researchers have documented variation in the V2-rule and in grammatical gender marking. These studies have all pointed out that “straight” word order or V2 deviation cannot be assigned to learner language but is a stylistic feature that the young people in question alternate according to context. In Sweden, Ganuza (2008, 2010) has found variation in the use of

717

718

PIA QUIST AND BENTE A. SVENDSEN

the V2-rule in topicalized declarative main clauses among young people in multilingual areas in the three largest cities Stockholm, Gothenburg, and Malmo¨ (aº sen jag har lunch ‘and then I have lunch’ versus standard Swedish: aº sen har jag lunch ‘and then have I lunch’, Ganuza 2010: 38). A similar tendency is found in Norwegian in Oslo (Opsahl and Nistov 2010, Svendsen and Røyneland 2008) and in Danish in Aarhus (Christensen 2012) and Copenhagen (Quist 2000, 2008), for example, in the Norwegian data: hvis noen staº r og breaker alle stopper opp (‘if someone is breakdancing everybody stops up’, Opsahl and Nistov 2010: 57), and the Danish data: naº r man er i puberteten man tænker mere ‘when you are in puberty you think more’, Quist 2000: 152). The fronted elements in the above-mentioned examples are not extraposed in the young people’s utterances. Hence, in these contexts according to the V2-rule the canonical subject position is after the finite verb, and before the verb complements (XVS), but the subject and verb are not inverted in these examples. According to Opsahl and Nistov (2010) there is a strong correlation between the situational context and the use of noninversion. The number of noninverted declarative main clauses is more than three times higher in informal peer conversations than in semi-formal interviews with researchers, despite the fact that the number of possible contexts for non-inversions is higher in the interviews (Opsahl and Nistov 2010: 54). In Sweden, Ganuza (2008, 2010) demonstrates that the number of non-inversions is higher when the speakers get involved or are more enthusiastic in the conversations, i.e., when they demonstrate a so-called high involvement style (Ganuza 2010: 43). Likewise, in Copenhagen, Quist found a higher proportion of noninverted declarative main clauses in peer conversations compared to the research interviews (Quist 2000: 162). In other words, variation in the V2-rule seems to be characteristic for informal language use in peer groups and less characteristic outside of the peer groups in more formal situational contexts. In this research on V2-variation, especially in Ganuza’s (2008, 2010) work, not all young people in multilingual urban spaces show a high degree of variability in particular contexts. No one consequently uses non-inversion in all declarative main clauses in all conversational contexts. Variation in the V2rule seems, however, to be an established style feature that indexes a young person from linguistically and culturally diverse urban spaces. Ka¨llstro¨m (2010) examined how V2-variation is used as a part of the literary formation of themes and characters in recent Swedish literature. In the novel by Jonas Hassen Khemiri Ett o¨ga ro¨tt (‘An eye red’) the protagonist Halim, an ArabicSwedish boy in his mid-teens, uses non-inversion in all contexts where inversion is canonically used in standard Swedish (Ka¨llstro¨m 2010: 146). Noninversion is used as a performative literary tool which is not as frequent – as demonstrated by Ka¨llstro¨m’s (2010) comparison with Ganuza’s (2008) work – outside of the fiction, but it is consciously used in the novel to draw a picture of a young person who is struggling to reconcile himself in Stockholm. It is plausible to argue that when a linguistic feature is used as

Urban Speech Styles of Germanic Languages

a performative literary tool, there are reasons to believe that it is a conventionalized, well-established feature and therefore possible for a broader audience to recognize. This argument is supported by the fact that non-inversion has been known and also described by researchers for over 20 years (Kotsinas 1988, Quist 2000). The Scandinavian literature on V2variation among young people in multilingual urban spaces was expanded by Freywald et al. (2015) through including Dutch and German data, two other Germanic languages that adhere to the V2-rule. The researchers found striking parallels in their Norwegian, Swedish, and German (although not the Dutch) data, and conclude that the motivation for the occurrence of XSV order lies in discourse pragmatics, and demonstrates that the informationalstructural preferences are realized in novel ways. Against this background Freywald et al. (2015) pose the question of whether we are in fact witnessing a process of language change, reminding us of the loss of V2 in English during the Middle English period.

30.3.2 Variation of Grammatical Gender Variation in the marking of grammatical gender is another aspect of grammar documented in Sweden, Denmark, and Norway (Kotsinas 1988, Quist 2000, Opsahl 2009a, Opsahl and Nistov 2010). In Norway, according to Opsahl and Nistov (2010: 60), the indefinite feminine article ei (‘a’) is left out, a common linguistic change documented all over Oslo: “Not a single one of our informants uses the feminine indefinite article ei [‘a’ (f)].” The feminine definite article, the suffix -a (døra ‘the door’) is, however, widespread in use, both among the respondents in the multilingual urban spaces and among people in general in Oslo East, the traditional working class area (Stjernholm 2012). In Oslo West, the traditional middle and upper class area, the masculine definite article -en (døren ‘the door’) is preferred (Stjernholm 2012), a typical feature of the so-called vestkantmaº l (‘West End vernacular’), a linguistic compromise between written Danish (the former colonial language) and spoken Norwegian. Variation in gender marking is thus indexing socioeconomic variation. In multilingual urban spaces, however, other types of gender variation are found, such as absence of gender congruence (er dette min glass? ‘is this my (m) glass (n)?’) (Opsahl and Nistov 2010: 61). In Denmark, there is a tendency among young people in multilingual urban spaces to use common gender, especially in demonstratives in contexts where standard Danish marks neuter, for instance den blad (‘that (common gender) magazine’) for det blad (‘that (n) magazine’) (Quist 2000: 153). As for the variation of the V2-rule, it seems that variation in gender is more frequent in informal conversations among friends. Common for V2-variation and grammatical gender is that the variation involves “marked features”, i.e., features that may be considered “special” for the language in question or rare among the world’s languages (Ellis 1994). Marked forms are sensitive for change in language contact areas and

719

720

PIA QUIST AND BENTE A. SVENDSEN

they are considered to be more difficult for second language learners to acquire (Ellis 1994, Lund 1997). It might thus not come as a surprise that it is the marked forms that vary among adolescents. These features are amenable to stylistic rules which can be exploited socially: You can “increase” or “decrease” the extent to which you sound like a“majority” or “minority” speaker by using the standard form or an alternative form. Hence, the above-mentioned Scandinavian literature on variation in the V2 rule and in grammatical gender demonstrates that for the young people in question variation is a possibility. They can choose to use the alternative form or they can use the “standard” form according to context. In other words, the variation is not a symptom of lack of competence in Norwegian, Swedish, or Danish syntax or gender marking; you can shift from one situation to another. Young people’s ability to vary their own language use in different contexts was precisely Kotsinas’ (1988) main argument for classifying rinkebysvenska (“Rinkeby Swedish”) as a Swedish dialect or variety and not as a learner language. Hence, the latest Scandinavian research on the topic supports Kotsinas’ argument. Similar arguments have been proposed by Wiese and colleagues in connection with their studies of “Kiezdeutsch” in Berlin, Germany (Wiese 2012).

30.3.3 Phonetic Variation Phonological variation is found in Danish, Norwegian and Swedish studies of language use in culturally and linguistically diverse urban spaces. The pronunciation is clearly recognizable by adolescents (Bijvoet and Fraurud 2010; Bode´n 2010; Christensen 2010, 2012; Hansen and Pharao 2010). Researchers, however, have difficulty extracting the specific traits that are different from the standard forms since both segmental and suprasegmental features, such as tone, stress, and intonation are involved. Hansen and Pharao (2010) show, for instance, that there is a difference in the duration of short and long vowels when they compare Copenhagen “multiethnolect” with “Copenhagen-based speech in general”, and that the speakers of “multiethnolect” occasionally employ a different stress group pattern. They suggest that it is these differences in vowel durations that “might contribute to the impression of an alternate rhythm in the spoken multiethnolect” (Hansen and Pharao 2010: 94). A juxtaposition of parallel phonological features across Scandinavia does not make sense in the same way as a grammatical comparison does since the phonological systems are quite different in the Scandinavian languages. There is, however, one exception – a parallel feature which we will briefly touch upon here. It has been argued by some phoneticians that there is a historical link between the Danish stød and the Swedish and Norwegian tonal system (Grønnum 1998: 181). Therefore, variation in stød and tones could be seen as parallel phenomena. In Denmark, there is a tendency to omit stød in contexts where standard Danish has stød (Christensen 2012, Quist 2000). There are

Urban Speech Styles of Germanic Languages

examples where words such as tusind (‘thousand’), sammen (‘together’) and grim (‘ugly’) are pronounced without stød by the young people in culturally and linguistically diverse Nørrebro (Quist 2000: 154). These words would traditionally have been pronounced with a stød on the vowel [u] [tuˀsen], and the consonant [m] [samˀən] and [gʁɛmˀ]. In Norway, there is a tendency to level out the accent (tone 1 and 2) in minimal pairs (Svendsen and Røyneland 2008). Tone 2 seems to be replaced with tone 1. Both instances in the minimal pair bønder [1bøn:ər] (‘farmers’) and bønner [2bøn:ər] (‘beans’) are pronounced with tone 1 among young people in multilingual urban spaces. It would be reasonable to expect that the distinction in pitch accent in Swedish would also vary in the Stockholm, Gothenburg, and Malmo¨ data, but there is so far no evidence of such a leveling (see Bode´n 2010). In Sweden, Bode´n (2010) reports on a new “foreign” way of speaking Swedish. She has studied pronunciation among young people in diverse language areas in Stockholm, Gothenburg, and Malmo¨ as part of the SUFproject (see Boyd 2010). Bode´n (2010) describes a new accent which is not only colored by various heritage languages such as Turkish and Arabic, but influenced by numerous languages. She has compared Swedish pronunciation inter alia with Turkish and Arabic, and cannot find phonological transfer from these languages. Hence, there is no evidence to state that these young people have an Arabic or Turkish accent, but rather a Swedish accent colored by the multilingual environment it has emerged in. Bode´n argues in line with Kotsinas (1988) that the young people’s speech style is to be considered as a local Swedish variety, a “multiethnolect”. She qualifies her statement by documenting that the speakers’ linguistic background does not determine a specific pronunciation. There are for instance speakers with parents born outside of Sweden who use a standard Swedish pronunciation, and speakers with parents born in Sweden who use the “foreign” sounding accent. Hitherto, there are unfortunately no similar studies in Denmark and Norway that systematically compare the pronunciation of the local contemporary urban speech style with the pronunciation of the young people’s various heritage languages, but there is reason to believe that there is the same trend in Copenhagen and Oslo as Bode´n found in Sweden, that is, a local pronunciation with a “foreign” colored accent without direct transfer from the speakers’ heritage languages (see Quist, Svendsen and Røyenland 2008). Nonetheless, the results from all three Scandinavian countries are consistent: Parental background is not a determiner for the use of contemporary urban speech styles, or pronunciation, since they are used by young people regardless of their descent (Quist 2012; Quist and Svendsen 2010).

30.3.4 Lexical Features A series of words, often with slang and discourse regulating functions, are associated with contemporary urban speech styles. These words are often loans from the largest heritage languages such as Arabic, e.g., habibi

721

722

PIA QUIST AND BENTE A. SVENDSEN

(‘friend’), and Turkish, such as kız (‘girl’) and para (‘money’), but also from languages such as Urdu, Berber, Serbian, Kurdish, and Greek depending on the local context. Some words are only registered locally and not found elsewhere. Other words, however, are found to be used across Northern European cities; for instance jalla (Arabic: ‘hurry up, come on, move away’), habibi (Arabic: ‘friend, darling’), wallah (Arabic/Turkish: ‘by Allah, to swear by Allah, I swear’), lan (Turkish: ‘man’ or ‘young man’) (there are no regulated spelling conventions) (e.g., Quist 2000, Cornips 2002, Auer and Dirim 2003, Madsen 2008, Svendsen and Røyneland 2008, Opsahl 2009b, Aarsæther 2010, Wiese 2012, Dorleijn et al. 2015). Here, we will restrict ourselves to commenting on a couple of the most frequently documented words. However, the lexical items of contemporary urban speech styles change relatively fast (e.g., Dorleijn et al. 2015) and there is a need for further (comparative) research on the lexical traits of these speech styles. The discourse marker wollah is in widespread use among young people in urban language areas in Denmark, Norway, and Sweden (or vollah or wolla, Opsahl 2009b). Wollah is originally Arabic and denotes ‘by Allah’. It is used in Danish, Norwegian, and Swedish to emphasize the correctness or the significance of what is said. Wollah is thus placed in front of or after the utterance, for instance wollah, jeg snakka med han (‘wollah, I spoke to him’), denoting ‘it is true that I spoke to him’, or han irriterer meg, wollah (‘he annoys me, wollah’), denoting ‘he annoys me, I swear’ (Opsahl 2009b: 228, 235). In Oslo, the word wollah has become emblematic for the young people’s ways of speaking in multilingual urban areas. Aarsæther (2010), Ims (2013), and Svendsen and Røyneland (2008) report that several of the young people in Oslo use the word as a label of the speech style in itself, by referring to it as ‘to speak wollah.’ Another widespread word in Denmark and Sweden is the Turkish word lan (‘man’ or ‘young man’) (Kotsinas 2000, Quist 2000, Madsen 2008, Møller 2010), and among the young people lan is used as an emphasizer, for instance det er sejt, lan (‘it is cool, man’). Lan is also used as a label or proper noun for a person, friend, or stranger, for example hey, lan (‘hey, man’). Møller (2010) demonstrates in a detailed analysis of a group of Turkish-Danish young people from Køge that lan is used with Danish pronunciation with stød [lanˀ]. Lan is thus integrated in the boys’ Danish language, and according to Møller it is no longer meaningful in its Danish context to conceive of the word as a Turkish word. In Malmo¨, Ekberg (2010) has explored the use of the particle saº n where the original meaning can be translated into English as ‘such a’ / ‘such’ (see also Ekberg et al. 2015). Saº n is perceived as being widespread in use among young people in Sweden, and Ekberg (2010) finds that the young people in Malmo¨ use the word for a series of other functions, for instance as a determiner, that is, in positions where it seemingly acts as a substitute for the indefinite article (jag var paº saº n badplats, ‘I was at such bathing-place’). In a cross-linguistic analysis of ‘such a’ / ‘such’ as it is represented in German

Urban Speech Styles of Germanic Languages

(so), Norwegian (saº nn), and Swedish (saº n), Ekberg et al. (2015) reveal that the particles undergo strikingly parallel development, pointing to general patterns of functionalization. They demonstrate ways in which semantic and syntactic restrictions are loosened in favor of a pragmatic expansion of the lexical items. Hence, their comparative account reveals findings that might be proved relevant to other European linguistic settings as well – as emphasized by the authors – given the general cross-linguistic similarities that have emerged from different national studies. This general pattern of functionalization is found among all adolescents in Sweden, Norway, and in Germany, but – what is interesting in the context of this chapter – it appears that the young people in linguistically diverse urban spaces spearhead this language development. Hitherto, no one has documented this general pattern of functionalization among young people in multilingual urban spaces in Denmark. Loans with a slang function such as lan, para, wollah, and kız are common in other European multilingual urban spaces (Cornips 2002, Auer and Dirim 2003, Wiese 2012), but so far no one has provided working explanations for why exactly these words coincide among the countries’ adolescents. It is perhaps not surprising that the loanwords originate from major heritage languages such as Arabic and Turkish. A word such as wollah is used in Arabic as well as in Turkish and it is thus plausible that many European citizens of migrant descent know this word from their heritage language. However, this is not the case for Turkish words used by young people without competence in Turkish. The rationale behind such loans might be traced in the conciseness of the words and their function as discourse particles, a function that makes them “easy” to combine with Danish, Norwegian, and Swedish sentences. The use of these words offers, moreover, resources for identity work. The ways adolescents perceive and relate to ways of speaking in multilingual urban spaces in Scandinavia is the topic for the next section.

30.4 Perception Studies and Ascribed Identities As shown in the sections above, a series of studies have focused on the linguistic description of features belonging to recently emerged new speech styles used by young people in multilingual urban spaces in Scandinavia and beyond. These studies have raised new questions and called for further investigations of the status and ideological impact of such linguistic developments in the broader speech communities, locally as well as nationally. In recent years, therefore, a handful of studies have been designed specifically with the aim to shed light on how contemporary urban speech styles are perceived by in- and out-group speakers in society. Studies have demonstrated that the use of features associated with these speech styles index in-group-solidarity and prestige, familiarity to a certain place or neighborhood, as well as representing a social contestation against

723

724

PIA QUIST AND BENTE A. SVENDSEN

mainstream society (Quist and Svendsen 2010, Nortier and Svendsen 2015). The contemporary Scandinavian urban speech styles seem to be ideologically positioned in contrast to the national standard languages (Stroud 2004, Milani 2010, Madsen 2013; Pharao et al. 2014) and analyses show that media discourses tend to homogenize the linguistic practices in question and to construct the adolescents as a homogeneous group (see below Androutsopoulos 2010; Milani 2010; Rampton 2006; Svendsen and Marzo 2015; Quist 2015). In Sweden, Bijvoet and Fraurud (2008, 2010; Bijvoet 2003) designed a listening task inspired by the matched guise techniques used in social psychology (Lambert et al. 1960; Giles et al. 1983). A young man born and raised in Rinkeby was told to tell the same story in two different speech styles. He produced his first speech sample (referred to as “speaker Leo”) when asked (just like all the other speakers) to imagine talking to a close friend. The second sample (referred to as “speaker Sam”) was produced when he was asked to speak the way he would when accommodating a “typical (monolingual) Swede.” A group of young people listened to the two samples and were asked to fill in a questionnaire and to point out on a map of the Stockholm subway system where they believed the speaker came from. It turned out that Leo’s voice was judged to be less organised, “tougher” and more humorous than Sam’s voice, while the opposite was the case for Sam’s voice, in particular with regard to his alleged toughness and organizing ability. Bijvoet and Fraurud demonstrate that listeners recognize specific ways of speaking Swedish in Stockholm. Listeners are able to point out on a map where they believe speakers of the speech style associated with youth of minority backgrounds live, and, moreover, they connect very clear stereotyped values to the two speech styles in the test. Some similar results are found in Denmark where Møller (2009) studied a group of 20 linguistic minority Danes in an evaluation test. The group of listeners were asked to categorize and rate four different voices – belonging to people who spoke different types of Danish – including a young man from Copenhagen speaking so-called “late modern urban youth style.” The young man was characterized by the listeners as speaking ghetto-Danish or street-language, and the respondents quite homogeneously described the young man and the other people in the group talking like that as gangsters and immigrants (or wannabe-immigrants) and evaluated this type of speech negatively (e.g., bandit, ghetto kid, hoodlum, I don’t like him, this is not my type, Møller 2009: 242). On a scale measuring “social status” the young man’s talk was rated lowest. Interestingly, Møller (2009) found that the respondents clearly distinguished between “late modern urban speech style” and “accented speech” (i.e., Turkish accented), and that they treated the young man’s voice as a marker of identity. Maegaard (2010: 205) examined whether social meanings attached to different speech styles are the same in different communities in Copenhagen. She concluded that the connections between category membership, linguistic practice, and linguistic variation are not locally bound to one part of

Urban Speech Styles of Germanic Languages

Copenhagen, but can be found in the larger Copenhagen speech community as well, although there are local variations. Maegaard (2010: 204) documents interesting gender differences in her study, where the girls with a “foreigner” background and a “foreigner” stylistic practice talk ”super nice”, that is with a higher frequency of the same variants typical of the “nice Danish girls.” In Aarhus, Christensen (2012) conducted a language recognition study with a view to explore how various groups of adolescents in Aarhus and Copenhagen respectively perceive and relate to different speech styles. Whereas “ethnicity” is pivotal in the recognition of the young people’s voices within Aarhus, the young Copenhageners give salience to regionality when they listen to the “ethnically marked” speech samples from Aarhus. Hence, Christensen’s study is significant in documenting that there exists no such phenomenon as a pan-Danish way of speaking Danish in multilingual urban spaces, although some of the linguistic features between what she labels “multiethnolect” in Aarhus and Copenhagen (Quist 2000) are of a similar kind, such as variation in the V2-rule. In Norway, there are only a few studies on how people perceive ways of speaking in multilingual urban spaces (Ims 2013, 2014, Kulbrandstad 2004, Svendsen 2014). Ims’ (2013) focus is on how language ideology is expressed through people’s emic labeling of ways of speaking in multilingual urban spaces in Oslo. Based on an internet-based questionnaire which has triggered an enormous number of answers (ca. 100,000), Ims’ findings reveal that the respondents categorize “this way of speaking” in a myriad of ways, with kebabnorsk being the most frequent label. She emphasizes that the speech style is not clearly stigmatized by the respondents; the degree of loaded labels (negative or positive) vary according to respondents’ individual, geographical, and social background. Hence, Ims’ study clearly indicates that there is a need for large-scale studies to nuance the picture drawn by many of the more qualitative oriented studies that have been conducted so far on the ways contemporary urban speech styles are conceived of on a macrosocietal level. In sum, the perception studies reveal that the recent urban speech styles are recognized by in- and out-groups in Scandinavia. There is evidence that these styles are generally perceived negatively in terms of social status, although there also seems to be an appealing “covert” prestige connected to the speech styles. However, as Ims’ (2013) study indicates, there is clearly a need for more research on the value ascription of these speech styles. The perception studies, furthermore, point to the local anchoring of these speech styles. Listeners link the different urban voices to specific places in the cities. These results suggest that contemporary urban speech styles constitute integrated parts of the Scandinavian speech communities in terms of value attachment and language attitudes – something that preconditions the possibilities of using these styles as meaningful resources for identity construction and negotiation (see also Quist 2009, Kristiansen et al. 2005, Madsen and Svendsen 2015).

725

726

PIA QUIST AND BENTE A. SVENDSEN

Perception studies that apply a social psychological approach (Giles et al. 1983) have a strong tradition in Scandinavia especially in studies of dialect attitudes and ideologies (Kristiansen 1997, Ladegaard 2001, Kristiansen 2010; see also Jørgensen and Quist 2001, Kristiansen 2006). However, ethnographic and interactional approaches have arguably an even stronger tradition among Scandinavian sociolinguists which has resulted in a series of studies based on ethnographic fieldwork and interaction studies in multilingual urban communities of practice (Ganuza 2008, Madsen 2008, Aarsæther 2010, Haglund 2010, Jørgensen 2010, Møller 2010, Quist 2012, Madsen and Svendsen 2015, Magnusson 2015). A common insight from these studies is that the young people appear to conceive contemporary urban speech styles as a typical peergroup phenomenon, used for in-group activities and topics (Svendsen and Røyneland 2008, Madsen 2011). In Norway, Aarsæther (2010) shows how young girls with Norwegian-born parents in the inner city use “multiethnic youth language” to position themselves in contrast to the traditional posh West-End and the alleged “posh” way of speaking. Svendsen and Røyneland (2008: 79) argue that the “multiethnolectal style” found in Oslo is used deliberately to express distance from an alleged “majority” or from the “society” at large, as a sort of resistance, and as a consequence of never having been accepted as “hundred percent Norwegian,” as one of their respondents puts it. Contrasting identities are also in focus for Madsen’s (2011, 2013) ethnographic study in and around a Copenhagen school where she explores how social power differences are refracted in the metalinguistic activities of a group of adolescents with a minority background. Her participants introduced labels for two ways of speaking that differed from what they referred to as “normal”: integreret (‘integrated’) and gadesprog (‘street language’) where “integrated” is associated with upscale culture, sophistication, authority, emotional control, academic skills, politeness, respect, and aversion to rudeness. The stereotypical indexical values of “street language” are contrasted to those values of “integrated” and are associated with toughness, masculinity, youth, pan-ethnic minority street culture, and academic non-prestige. In her study, Madsen (2011: 21) concludes that the social status relations brought into play by the adolescents is better understood as an awareness of “high” and “low” societal stratification signified by certain cultural and linguistic practices, rather than as “social class in a classical sense of inequality between clearly bounded groups” (see also Madsen and Svendsen 2015).

30.5 Media and Popular Culture – Identity Work The Scandinavian adolescents in multilingual urban spaces – as every other adolescent – draw on a multitude of resources, including

Urban Speech Styles of Germanic Languages

popular culture, peer groups, and young people outside of the Scandinavian context, in their identifications and allegiances (Haglund 2010, Magnusson 2015). Popular cultural expressions such as hip-hop and rap seem to be one of the key factors in the enregisterment process (Agha 2007) and of the ideological formation of contemporary urban speech styles. The connection between hip-hop culture, rap lyrics, and the development of “glocal” language practices has gained interest within sociolinguistics (e.g., Androutsopoulos 2009, Ibrahim et al. 2009, Knudsen 2010, Stæhr 2010, Cutler and Røyneland 2015). Knudsen (2010), for instance, documents that the use of street language or “multiethnolectal language features” is perceived as an act of social and cultural resistance for the young hip-hop crew in Oslo, Minoritet1 (‘Minority1’), and Cutler and Røyneland (2015) compare hip-hop lyrics and self-presentations in Norwegian and American hip-hop videos on YouTube. They show how the hiphop affiliated adolescents’ linguistic practices function as a means to differentiate themselves from other immigrant minority groups, their own ethnic group, and/or mainstream youth, and to signal social and discursive stances, that is solidarity with or distance from an interlocutor; a street-orientation; status as a hip-hop insider/artist; and as a way to resist and transform traditional social and ethnic categories. In the Swedish context, Milani (2010) has demonstrated through an analysis of print-mediated discussions about linguistic practices in Sweden that are perceived by many commentators to be “deviant” from “standard Swedish,” that language is used as a proxy for other social differentiation processes. He reveals, as has also been found in Norway (Ims 2014, Svendsen 2014), that contemporary urban speech style is stigmatized in projections of a speech style or a language which inevitably leads to unemployment (Milani 2010, Ims 2014, Svendsen 2014). Milani et al. (2015) study how rinkebysvenska is removed from its local origin and strategically used in media texts such as advertisements. Milani et al. (2015) show that stereotypically odd and comic characters get to count as “the ethnic other” in these Swedish advertisements. By comparing how rinkebysvenska in Sweden and Tsotsitaal in South Africa have been strategically recontextualized in mainstream advertising for consumerist purposes, Milani et al. (2015) demonstrate how the recontexualization of Tsotsitaal has led to changes in the values ascribed to it. After Soweto, Tsotsitaal has become an index of political emancipation. Rinkebysvenska has in contrast not undergone a similar process of re-evaluation, which leads the authors to conclude that the recontextualization of Tsotsitaal might function as a future scenario for rinkebysvenska. Further comparisons across time and space such as the one conducted by Milani et al. (2015) as well as by Rampton (2015) are important to explore the future of these contemporary urban speech styles,

727

728

PIA QUIST AND BENTE A. SVENDSEN

whether or not they become permanent registers within the sociolinguistic ecology of late modern urban agglomerations.

30.6 Conclusion The main purpose of this article has been to provide a general, although not complete, overview of the parallel results from the Scandinavian research in multilingual urban spaces with a glance at parallel findings in Germany, the UK, and in the Netherlands. We do not believe nor have found evidence for any kind of supraterritorial way of speaking in these urban spaces (for instance a “pan-Scandinavian multiethnolect”). On the contrary, one of the themes to explore further is the local and regional differences in linguistic practices across languages and research sites with a view to understanding the sociolinguistic complexity of 21stcentury (young) people’s lives. At the same time, the Scandinavian languages seem to undergo some parallel developments in multilingual urban spaces: clustering of syntactic (variation in the V2-rule), morphological (gender variation), phonological (intonation and tone variation), and lexical features. The grammatical traits analyzed involve marked features and are perhaps therefore more prone to change in language contact areas. The perception studies, moreover, reveal that young people recognize such clustering of features as belonging to certain speech styles and position them rather negatively in terms of supposed personal characteristics. They also assume that they indicate low social status. However, there is a need for large-scale studies to nuance the social status of these speech styles (see Ims 2013) and their (covert?) function as a marker of prestige, and to explore whether there are differences in the value ascriptions to these styles across Scandinavia (see, e.g., Madsen and Svendsen 2015). The Scandinavian literature on language practices in multilingual urban spaces demonstrates that contemporary urban speech styles are exactly styles in which the young language users are able to step in and out from; they are often considered and used as in-group phenomena in more informal settings; exposed and used in processes of identity negotiations and stance taking in different genres such as hip-hop, rap lyrics and literature inter alia as a means to broaden and transform traditional social and ethnic categories. Hitherto, the Scandinavian research (as elsewhere) has focused on the emergence and use of contemporary urban speech styles as a youth phenomenon. As these language users grow older, there is a need to explore whether their speech styles are age-graded; for example, the extent to which the language users continue to employ these styles during their lifespan, or whether they leave them aside as they leave other semiotic age-graded signs behind. Rampton (2015) has, for

Urban Speech Styles of Germanic Languages

instance, documented the use of a “contemporary urban vernacular” by an English middle-aged business man in informal conversations with a friend. The respondents in the international literature on linguistic practices in multilingual urban spaces are predominantly male. There is a need to explore to which extent these speech styles index gender, for instance in light of the intriguing above-mentioned contrasting results of Aarsæther (2010) and Maegaard (2010). The Scandinavian research on linguistic practices in multilingual urban spaces has often focused either on the linguistic forms or on the forms’ functions in situated discourse (Svendsen and Quist 2010). To explore the semiotics of these contemporary urban speech styles there is a need to integrate the study of linguistic forms, usages, and ideological value ascriptions to these speech styles in different situated discourses (Silverstein 1985; cf. Quist and Svendsen 2015, Rampton 2015, Svendsen 2015). A narrow focus on one of the three angles might obscure the extent to which we are conducting research on the same phenomena or simply grouping respondents based on stereotypes of an ostensible group of people, their speech, their physical appearance or residential address.

References Aarsæther, F. 2010. “The use of multiethnic youth language in Oslo.” In Quist and Svendsen (eds.): 111–126. Aasheim, S. 1995. Kebab-norsk: fremmedspraº klig paº virkning paº ungdomsspraº ket i Oslo. [Kebab Norwegian: Influence of foreign languages on the youth language in Oslo.] M.A.thesis, University of Oslo. Agha, A. 2007. Language and Social Relations. Cambridge University Press. Androutsopoulos, J. 2009. “Language and the three spheres of hip-hop discourse.” In A. Ibrahim, H. S. Alim, and A. Pennycook (eds.), Global Linguistic Flows: Hip Hop Cultures, Identities and the Politics of Language. Mahwah, NJ: Lawrence Erlbaum Associates: 43–62. Androutsopoulos, J. 2010. “Ideologizing ethnolectal German.” In S. Johnson and T. M. Milani (eds.), Language Ideologies and Media Discourse: Texts, Practices and Politics. New York: Continuum: 182–202. Appel, R. and P. Muysken 1987. Language Contact and Bilingualism. New York: Edward Arnold. Appel, R. and R. Schoonen 2005. “Street language: A multicultural youth register in the Netherlands,” Journal of multilingual and multicultural development 26.2: 85–117. Auer, P. and I˙. Dirim 2003. “Socio-cultural orientation, urban youth styles and the spontaneous acquisition of Turkish by non-Turkish adolescents in Germany.” In J. Androutsopoulos and A. Georgakopoulou (eds.),

729

730

PIA QUIST AND BENTE A. SVENDSEN

Discourse Constructions of Youth Identities. Amsterdam and Philadelphia: John Benjamins: 223–246. Bijvoet, E. 2003. “Attitudes towards ‘Rinkeby Swedish’, a group variety among adolescents in multilingual suburbs.” In K. Fraurud and K. Hyltenstam (eds), Multilingualism in Global and Local Perspectives. Stockholm: Centre for Research on Bilingualism, Stockholm University, and Rinkeby Institute of Multilingual Research: 307–318. Bijvoet, E. and K. Fraurud 2008. “Svenskan i dagens flerspraº kiga storstadsmiljo¨er: en explorativ studie av unga stockholmares perceptioner av variation och varieteter” [Swedish spoken in today’s multilingual cities: An explorative study of young Stockholmers’ perceptions of variation and varieties], Nordand 3.2: 7–38. Bijvoet, E. and K. Fraurud 2010. “‘Rinkeby Swedish’ in the mind of the beholder. Studying listener perceptions of language variation in multilingual Stockholm.” In Quist and Svendsen (eds.): 170–188. Blackledge, A. and A. Creese 2015. “Emblems of identities in four European urban settings.” In Nortier and Svendsen (eds.): 167–182. Bode´n, P. 2010. “Pronunciation in Swedish multiethnolect”, in Quist and Svendsen (eds.): 65–78. Boyd, S. 2010. “Research on Language in Multilingual Urban Settings in Sweden.” In Quist and Svendsen (eds.): 1–5. Cheshire, J., P. Kerswill, S. Fox, and E. Torgersen 2011. “Contact, the feature pool and the speech community: The emergence of Multicultural London English,” Journal of Sociolinguistics 15.2: 151–196. Christensen, M. V. 2010. “One of My Kind? Language and Ethnicity among Danish Adolescents.” In Quist and Svendsen (eds.): 207–224. Christensen, M. V. 2012. 8220, 8210. Sproglig variation blandt unge i multietniske omraº der i Aarhus [8220, 8210. Linguistic Variation among Youth in Multiethnic Areas in Aarhus]. Ph.D. thesis, Aarhus University. Clyne, M. 2000. “Lingua franca and ethnolects in Europe and beyond.” Sociolinguistica, 14: 83–89. Cornips, L. 2002. “Etnisch Nederlands.” In H. Bennis, G. Extra, P. Muysken, and J. Nortier (eds.), Een buurt in beweging. Talen en Culturen in het Utrechtse Lombok en Transvaal [A Neighbourhood on the Move. Languages and Cultures in Lombok and Transvaal, Utrecht]. Amsterdam: Aksant: 285–302. Cornips, L., J. Jaspers, and V. de Rooij 2015. “The Politics of Labelling Youth Vernaculars in the Netherlands and Belgium.” In Nortier and Svendsen (eds.): 45–69. Cutler, C. and U. Røyneland 2015. “Where the fuck am I from? Hip-Hop youth and the (re)negotiation of language and identity in Norway and the US.” In Nortier and Svendsen (eds.): 139–163. Dorleijn, M., M. Mous, and J. Nortier 2015. “Urban youth speech styles in Kenya and the Netherlands.” In Nortier and Svendsen (eds.): 271–289.

Urban Speech Styles of Germanic Languages

Eckert, P. 2004. ”The meaning of style”. www.stanford.edu/eckert/PDF/ salsa2003.pdf. Ekberg, L. 2010. “Extended uses of saº n ‘such’ among adolescents in a multilingual context.” In Quist and Svendsen (eds.): 17–30. Ekberg, L., T. Opsahl, and H. Wiese 2015. “Functional gains: A crosslinguistic case study on three particles in Swedish, Norwegian and German.” In Nortier and Svendsen (eds.): 93–115. Ellis, R. 1994. The Study of Second Language Acquisition. Oxford University Press. Freywald, U., N. Ganuza, T. Opsahl, and L. Cornips 2015. “Beyond verb second – A matter of novel information structural effects? Evidence from Norwegian, Swedish, German and Dutch.” In Nortier and Svendsen (eds.): 73–92. Ganuza, N. 2008. Syntactic Variation in the Swedish of Adolescents in Multilingual Urban Settings: Subject-Verb Order in Declaratives, Questions and Subordinate Clauses. Ph.D. thesis, University of Stockholm. Ganuza, N. 2010. “Subject-verb order variation in the Swedish of young people in multilingual urban areas.” In Quist and Svendsen (eds.): 31–48. Giles, H., M. Hewstone, and P. Ball 1983. “Language attitudes in multilingual settings: Prologue with priorities,” Journal of Multilingual and Multicultural Development 4.2–3: 81–100. Grønnum, N. 1998. Fonetik og Fonologi. Almen og Dansk. Copenhagen: Akademisk Forlag. Haglund, C. 2010. “Transnational identifications among adolescents in suburban Sweden.” In Quist and Svendsen (eds.): 96–110. Hansen, G. F. and N. Pharao 2010. “Prosody in Copenhagen Multiethnolect.” In Quist and Svendsen (eds.): 79–95. Hinnenkamp, V. 2003. “Mixed language varieties of migrant adolescents and the discourse of hybridity.” In J. N. Jørgensen (ed.), Bilingualism and Social Relations: Turkish Speakers in North-West Europe. Bristol: Multilingual Matters: 12–41. Ibrahim, A., H. S. Alim, and A. Pennycook (eds.) 2009. Global Linguistic Flows: Hip Hop Cultures, Identities and the Politics of Language. Mahwah, NJ: L. Erlbaum Associates. Ims, I. 2013. ”Spraº klig registerdanning og verditilskriving: Betegnelser paº nye maº ter aº snakke norsk paº i Oslo” [Enregisterment and valorisation: Labeling new ways of speaking Norwegian] NOA: Tidsskrift for Norsk som andrespraº k, 1. 37–71. Ims, I. 2014. “Alle snakker norsk. Spraº kiidedogiog spraº klig differensiering i mediene” [Everybody speaks Norwegian. Language idedogy and language differentiation in the media] NOA: Tidsskrift for Norsk som andrespraº k, 30.1: 5–40. Jørgensen, J. N. 2010. Languaging: Nine Years of Poly-Lingual Development of Young Turkish-Danish Grade School Students. Copenhagen: Københavnerstudier i tosprogethed.

731

732

PIA QUIST AND BENTE A. SVENDSEN

Jørgensen, J. N. and P. Quist 2001. ”Native speakers’ judgements of second language Danish,” Language Awareness 10.1: 41–56. Ka¨llstro¨m, R. 2010. “Literary use of multiethnic youth language. Noninversion in Swedish fiction.” In Quist and Svendsen (eds.): 142–155. Ka¨llstro¨m, R. and I. Lindberg (eds.) 2011. Young Urban Swedish. Variation and Change in Multilingual Settings. University of Gothenburg: Go¨teborgsstudier i Nordisk spraº kvetenskap 14. Website: http://hdl.handle.net/2077/26570 (last accessed October 26, 2017). Kern, F. and M. Selting 2011. Ethnic Styles of Speaking in European Metropolitan Areas. Amsterdam: John Benjamins. Kiesling, S. F. 2005. “Variation, style, and stance: Word-final -er and ethnicity in Australian English,” English World Wide 26: 1–42. Kiessling, R. and M. Mous 2004. “Urban youth languages in Africa.” Anthropological Linguistics 46.3: 303–341. Knudsen, J. S. 2010. “‘Playing with words as if it was a rap game’: Hip-hop street language in Oslo.” In Quist and Svendsen (eds.): 156–169. Kotsinas, U.-B. 1988. “Immigrant children’s Swedish – A new variety?,” Journal of Multilingual and Multicultural Development 9.2: 129–140. Kotsinas, U.-B. 2000. “Spraº kkontakt og slangspraº k i Stockholm.” [Language contact and slang in Stockholm] In A.-B. Stenstro¨m, U.-B. Kotsinas, and E.-M. Drange (eds.), Ungdommers spraº kmøter. Copenhagen: Nordisk Ministerraº d: 19–60. Kristiansen, T. 1997. “Language attitudes in a Danish cinema.” In Sociolinguistics: A Reader and Coursebook. London: Macmillan Press Ltd: 291–305. Kristiansen, T. (ed.) 2006. Nordiske sprogholdninger: En masketest [Nordic language attitudes: A matched guise test]. Oslo: Novus forlag. Moderne importord i spraº ka i Norden, Vol. 5. Kristiansen, T. 2010. “Attitudes, ideology and awareness.” In R. Wodak, B. Johnston, and P. Kerswill (eds.), The SAGE Handbook of Sociolinguistics. Thousand Oaks: Sage: 265–278. Kristiansen, T., N. Coupland, and P. Garrett 2005. “Introducing subjectivities in language variation and change,” Acta Linguistica Hafniensia 37: 9–35. Kulbrandstad, L. A. 2004. “‘Kebabnorsk’, ‘perkerdansk’ og ‘gebrokken’ – ord om innvandreres maº te aº snakke majoritetsspraº ket paº ” [‘KebabNorwegian’, ‘perker Danish’ and ‘broken’ – labels on ways of speaking the majority language]. In H. Sandøy (ed.), Den fleirspraº klege utfordringa. Oslo: Novus: 108–130. Ladegaard, H. J. 2001. “Popular perceptions of standard language: attitudes to ‘regional standards’ in Denmark,” Language Awareness 10.1: 25–40.

Urban Speech Styles of Germanic Languages

Lambert, W. E., R. C. Hudgson, R. C. Gardner, and S. Fillenbaum 1960. “Evaluational Reactions to Spoken Languages,” Journal of Abnormal and Social Psychology 60.1: 44–51. Lund, K. 1997. Lærer alle dansk paº samme maº de?En længdeundersøgelse af voksnes tilegnelse af dansk som andetsprog. Herning: Special-pædagogisk forlag. Madsen, L. M. 2008. Fighters and Outsiders: Linguistic Practices, Social Identities, and Social Relationships among Urban Youth in a Martial Arts Club. Ph.D. thesis, University of Copenhagen. Madsen, L. M. 2011. “Social status relations and enregisterment in contemporary Copenhagen,” Working Papers in Urban Language and Literacies 72. King’s College London. Madsen, L. M. 2013. “ ‘High’ and ‘low’ in urban Danish speech styles,” Language in Society 42: 115–138. Madsen, L. M. and B. A. Svendsen 2015. ”Stylized voices of ethnicity and social division.” In Nortier and Svendsen (eds.): 207–229. Maegaard, M. 2010. “Linguistic practice and stereotypes among Copenhagen adolescents.” In Quist and Svendsen (eds.): 189–206. Magnusson, J. E. 2015. “‘Sja¨lvklart ka¨nns det mer tryggt att vara da¨r inne i huset’ – om den sociokulturella kontextens betydelse fo¨r spraº kliga repertoarer och identiteter” [The impact of the sociocultural context on linguistic repertoires and identities], Nordand – Nordisk tidsskrift for andrespraº ksforskning 10.1:7–28. Marzo, S. and E. Ceuleers 2011. “The use of Cite´taal among adolescents in Limburg. The role of space appropriation in language variation and change,” Journal of Multilingual and Multicultural Development 32: 451–60. Milani, T. 2010. “What’s in a name? Language ideology and social differentiation in a Swedish print-mediated debate,” Journal of Sociolinguistics 14: 116–142. Milani, T., R. Jonsson, and I. J. Mhlambi (2015). “Shooting the subversive: When non-normative linguistic practices go mainstream in the media.” In Nortier and Svendsen (eds.): 119–138. Møller, J. S. 2009. ”Stereotyping categorisations of speech styles among linguistic minority Danes in Køge.” In M. Maegaard, F. Gregersen, P. Quist, and J. N. Jørgensen (eds.), Language Attitudes, Standardization and Language Change. Oslo: Novus: 231–254. Møller, J. S. 2010. “Polylingualism in the steak house: Exploring linguistic practices in late modern Copenhagen.” In Quist and Svendsen (eds.): 127–141. Nortier, J. 2001. Murks en Straataal [Murks and street Language]. Amsterdam: Prometheus. Nortier, J. and B. A. Svendsen 2015. Language, Youth and Identity in the 21st Century: Linguistic Practices across Urban Spaces. Cambridge University Press.

733

734

PIA QUIST AND BENTE A. SVENDSEN

Opsahl, T. 2009a. “Egentlig alle kan bidra!” – en samling sosiolingvistiske studier av strukturelle trekk ved norsk i multietniske ungdomsmiljøer i Oslo. [Actually everybody may contribute. A collection of sociolinguistic studies of structural features associated with Norwegian as spoken among adolescents in multiethnic areas in Oslo.] Ph.D. thesis, University of Oslo. Opsahl, T. 2009b. “Wolla I swear this is typical for the conversational style of adolescents in multiethnic areas in Oslo.” In F. Gregersen and U. Røyneland (eds.), Special issue of Nordic Journal of Linguistics. Sociolinguistics 32.2: 221–244. Opsahl, T. and I. Nistov 2010. “On some structural aspects of Norwegian spoken among adolescents in multilingual settings in Oslo.” In Quist and Svendsen (eds.): 49–64. Pharao, N., M. Maegaard, J. S. Møller, and T. Kristiansen 2014. “Indexical meanings of [s+] among Copenhagen youth: Social perception of a phonetic variant in different prosodic contexts,” Language in Society 43.1: 1–32. Quist, P. 2000. “Ny københavnsk ‘multietnolekt’. Om sprogbrug blandt unge i sprogligt og kulturelt heterogene miljøer” [New Copenhagen ‘multiethnolect’. Language use among young people in linguistically and culturally heterogeneous settings], Danske Talesprog 1: 143–212. Quist, P. 2008. “Sociolinguistic approaches to multiethnolect: Language variety and stylistic practice,” International Journal of Bilingualism 12: 43–61. Quist, P. 2009. “Multietnolekt – del af stilistiske praksisser i storbyen [Multiethnolect – part of stylistic practices in the cities].” In NyS. Nydanske Studier og Almen Kommunikationsteori 37: 10–33. Quist, P. 2012. Stilistisk Praksis [Stylistic Practice]. København: Museum Tusculanum. Quist, P. 2015. “Konstruktion af ‘perkerdansk’ og danskhed i online debatter om multietnisk ungdomsstil,” Danske Talesprog 15: 173–192. Quist, P. 2016. “Representations of multi-ethnic youth styles in Danish broadcast media.” In N. Coupland, N. J. Thøgersen, and J. Mortensen (eds.), Style, Media and Language Ideologies. Oslo: Novus: 217–234. Quist, P. 2018. “Alternative place naming in the diverse margins of an ideologically mono-lingual society.” In L. Cornips and V. de Rooij (eds.), The Sociolinguistics of Place and Belonging: Perspectives from the Margins. Amsterdam: John Benjamins: 239–259. Quist, P. and B. A. Svendsen (eds.) 2010. Multilingual Urban Scandinavia: New Linguistic Practices. Bristol: Multilingual Matters. Quist, P. and B. A. Svendsen 2015. “MultiNord 2007–2015: Et nettverk for forskning om spraº k i heterogene byrom – empiriske og teoretiske hovedlinjer” [MultiNorth 2007–2015: A network on research on language in

Urban Speech Styles of Germanic Languages

heterogeneous urban spaces – empirical and theoretical main approaches]. NOA – Tidsskrift for norsk andrespraº ksforskning, 1–2: 151–194. Rampton, B. 2006. Language in Late Modernity: Interaction in an Urban School. Cambridge University Press. Rampton, B. 2015. “Contemporary urban vernaculars.” In J. Nortier and B. A. Svendsen (eds.): 24–45. Silverstein, M. 1985. “Language and the culture of gender.” In E. Mertz and R. J. Parmentier (eds.), Semiotic Mediation. New York: Academic Press: 219–259. Sollid, H. 2013. “Ethnolects in Northern Norway: From national negligence to local linguistic pride.” In P. Siemund, I. Gogolin, M. E. Schulz, and J. Davydova (eds.), Multilingualism and Language Diversity in Urban Areas. Amsterdam: John Benjamins: 63–94. Statistics Sweden, Norway and Denmark 2017. http://www.scb.se/en/, https:// www.ssb.no/en/, http://www.dst.dk/en/Statistik, last accessed October 21, 2017. Stjernholm, K. 2012. Stedet velger ikke lenger deg, du velger et sted. Tre artikler om spraº k i Oslo [You are no longer Chosen by a Place, you Choose a Place. Three Articles on Language in Oslo]. Ph.D. thesis, University of Oslo. Stroud, C. 2004. “Rinkeby Swedish and semilingualism in language ideology debates: A Bourdieuan perspective,” Journal of Sociolinguistics 8.2: 196–214. Stæhr, A. 2010. Rappen reddede os. Et studie af senmoderne storbydrenges identitetsarbejde i fritids- og skolemiljøer [Rap saved us. A study of late modern urban boys’ identity work in leisure and school environments]. University of Copenhagen: Copenhagen Studies in Bilingualism, Vol. 54. Svendsen, B. A. 2010. “Linguistic Practices in Multilingual Urban Contexts in Norway: An Overview.” In Quist and Svendsen (eds.): 12–16. Svendsen, B. A. 2014. “Kebabnorskdebatten. En spraº kideologisk kamp om sosial identitet og verdier” [Social identities and values in the debate on ‘Kebab-Norwegian’ – an ideological battlefield], Norsk tidsskrift for ungdomsforskning 1: 33–62. Svendsen, B. A. 2015. “Language, youth and identity in the 21st century: Content and continuations.” In Nortier and Svendsen (eds.): 1–23. Svendsen, B. A. and U. Røyneland 2008. “Multiethnolectal facts and functions in Oslo, Norway,” International Journal of Bilingualism 12: 63–83. Svendsen, B. A. and P. Quist 2010. “Introduction.” In Quist and Svendsen (eds.): xiii–xxiii. Svendsen, B. A. and S. Marzo 2015. “A ‘new’ speech style is born: The omnipresence of structure and agency in the life of semiotic registers in heterogeneous urban spaces,” European Journal of Applied Linguistics 3.1: 47–85. Wiese, H. 2011. “The role of information structure in linguistic variation: evidence from a German multiethnolect.” In F. Gregersen, J. Parrott, and P. Quist (eds.), Language Variation – European Perspectives III. Amsterdam: John Benjamins: 83–95. Wiese, H. 2012. Kiezdeutsch: Ein Neuer Dialekt Entsteht. Mu¨nchen: C. H. Beck.

735

Chapter 31 The West Germanic Dialect Continuum William D. Keel

31.1 Introduction The linguistic variation within the area of north central Europe traditionally labeled West Germanic can be considerable. While everyone is familiar with “the standard varieties of” Dutch and German, the terms cover a rich array of different spoken forms of the language. Peter Wiesinger (1980) devised a four-way classification of these spoken varieties that may be applied generally throughout the areal called Continental West Germanic: (i) the high or standard variety (Standardsprache); (ii) a broader regional colloquial variety (Umgangssprache); (iii) a narrow regional colloquial variety (Verkehrsdialekt); (iv) a localized base dialect (Basisdialekt). In this chapter, we will explore the varieties within the traditional base dialects that make up the Continental West Germanic dialect continuum. This continuum involves all aspects of the varieties in question: the sound system, its morphosyntactics, its lexicon. Typically, only neighboring base dialects are mutually intelligible, while base dialects a great distance from one another are not. Low German varieties, Upper German varieties, and even some Middle German varieties when spoken as a base dialect may not be intelligible to people who know only Standard German. All West Germanic base dialects belong to the dialect continuum that includes High German (Middle and Upper German) and Low German. This West Germanic dialect continuum also includes the base varieties spoken in the area where Netherlandic serves as the high variety. There are also pockets of Frisian varieties belonging to West Germanic along the North Sea Coast (see Frisian at the end of this chapter). The terms “Low” and “High” German derive from the topography where these base dialects are spoken rather than from social conditions. Low German base dialects are thus found in the north in the lowlands (e.g., The Netherlands) along the seacoasts of the North and Baltic Seas; High

The West Germanic Dialect Continuum

German base varieties include Middle German base dialects (in the Central Highlands) and Upper German base dialects in the pre-Alpine and Alpine region in the south. This chapter will discuss the major characteristics defining Low, Middle and Upper German base dialects in West Central Europe, to include varieties associated with the continental West Germanic languages bordering on the core German-speaking countries of Germany, Switzerland, and Austria such as Dutch, Luxembourgish, or Alsatian. For a more in-depth overview of these dialects, the reader is referred to Keller (1961), Russ (1989), and Schirmunski (2010). The continental West Germanic region roughly encompasses the territory of modern-day Germany, Austria, the German-speaking part of Switzerland, the Netherlands, the northern half of Belgium (Flanders), Liechtenstein, Luxembourg, and South Tyrol in northern Italy. The continuum also includes areas of neighboring countries where a variety of German is spoken: the southern border region of Denmark (North Schleswig), as well as enclaves of German in Bohemia (western Czech Republic), Slovenia, Hungary, and Romania among others. One should also note the presence of non-German/Low Franconian/Frisian enclaves within this West Germanic areal, e.g., Sorbian in eastern Germany, which is a West Slavic variety. The principal dialects are outlined on the map in Figure 31.1. The labeling of the dialects according to the extent of the spread of the characteristics of the second consonant shift led to the division into Low and High German and the division of the High German dialects into Middle and Upper German. The labels for the individual dialects are often based on the names of the early medieval Germanic tribes. These dialects can also be grouped and classified differently, which may be illustrated at the very least by the transitional dialects and areas of overlap and convergence existing between all dialects. An example of an area of dialect convergence is the dialect known as Berlinerisch, which is widely used in the region of Brandenburg and which has both Low- and Middle-German linguistic characteristics. The area of our focus is largely contained within the area that uses Modern Standard German as its high variety (Germany, Austria, Liechtenstein, South Tyrol, Switzerland, Luxembourg, and a few cantons in eastern Belgium). In the Netherlands and northern cantons of Belgium, that standard variety is Standard Dutch (Algemeen Beschaafd Nederlands). Standard Luxembourgish (Le¨tzebuergesch) is also a high variety in the Grand Duchy of Luxembourg plus the region of Arlon in southeastern Belgium. Standard German itself is nowadays considered to have at least five regional standard varieties, largely divided along the boundaries for the dialectal varieties of West Germanic: (i) North German corresponding to the Low German dialect region; (ii) Middle German encompassing the central section

737

738

WILLIAM D. KEEL

Map 31.1 Westgermanic Dialect Continuum: Low German varieties 1–11; Low Franconian (Dutch) varieties 12–20; Middle German varieties 21–28; Upper German varieties 29–32; Frisian varieties 33–35. https://commons.wikimedia.org/wiki/File:Deutsch-Niederl%C3% A4ndischer_Sprachraum_(nach_Werner_K%C3%B6nig).png.

The West Germanic Dialect Continuum

739

740

WILLIAM D. KEEL

of Germany from Luxembourg to Saxony; (iii) South German with the major states of Baden-Wu¨rttemberg and Bavaria; (iv) Austrian German; and (v) Swiss German (cf. Knipf-Komlo´si and Berend 2001). From the beginning of recorded Central European history in the early Middle Ages, there was a realization that numerous varieties of “German” were spoken and written. The medieval German author Hugo von Trimberg (ca. 1230–1313) describes the linguistic variety in the German world (ignoring those in northern Germany) in his literary work Der Renner (ca. 1300, verses 22265–22276, www.staff.uni-marburg.de/~naeser/renner.htm) as follows (modern German translation lejastierras.blogspot.com/2011/08/ von-mancherlei-sprache.html; English by the author): Swaˆben ir wo¨rter spaltent, Die Franken ein teil si valtent, Die Beier si zezerrent, Die Du¨ringe si uˆf sperrent, Die Sahsen si bezu¨ckent, Die Rıˆnliute si verdru¨ckent, Die Mıˆsener si vol schu¨rgent, Egerlant si swenkent, Oesterrıˆche si schrenkent, Stıˆrlant si baz lenkent, Kernde ein teil si senkent . . .

Die Schwaben spalten ihre Wo¨rter, Die Franken falten sie teilweise, Die Bayern zerren sie auseinander, Die Thu¨ringer sperren sie auf, Die Sachsen u¨berlisten sie, Die Leute vom Rhein unterdru¨cken sie, Die Wetterauer wu¨rgen sie, Die Meißener stoßen sie vo¨llig, Egerla¨nder schleudern sie, ¨ sterreicher stellen sie O schra¨g, Steierla¨nder biegen sie besser, Ka¨rntner senken sie teilweise.

The Swabians split their words, The Franks merge them partially, The Bavarians tear them apart, The Thuringians open them up, The Saxons circumvent them, The Rhinelanders suppress them, The Wetterauers choke them, The Meißeners shove them totally out, Egerlanders hurl them, Austrians skew them, Styrians turn them better, Carinthians lower them partially. . .

Some might read into this description of the Middle and Upper German dialects an attempt to describe the various vowel sounds in these varieties: some varieties diphthongizing vowels, others monophthongizing, lowering or raising vowels. As we shall see, some of those characterizations may have been close to reality and mirror the variation found today within the continuum. The following discussion will reflect the varieties of West Germanic depicted on the map at Figure 31.1, which includes base varieties / dialects such as Low Prussian and Silesian, which, due to the displacement of the German-speaking population following the Second World War, are no longer part of that continuum.

31.2 Netherlandic and Low German The northernmost group of dialects in Germany is termed Low German (often called Platt, Plattdeutsch or Low Saxon) and has the status of a regional

The West Germanic Dialect Continuum

minority language in Europe. There is also an Institute for Low German Language in the city of Bremen, which promotes and helps maintain the use of Low German. The Low German varieties along the lower Rhine River, based on their distinctive characteristics, are classified as Low Franconian dialects. Closely related to these Low Franconian varieties in northwestern Germany are those dialects with Dutch (Algemeen Beschaafd Nederlands) as their standard language. The Dutch-speaking area of Central Europe includes the Netherlands and the northern cantons of Belgium, where the language is called Flemish, which extends historically into northeastern France (Dunkirk). The Flemish area in Belgium and the varieties in the southern, central and western areas of the Netherlands are part of the Low Franconian dialect area. The northeastern part of the Netherlands, however, is part of the Low Saxon/Low German dialect area. On the other hand, West Frisian in the northwest of the Netherlands as well as the North Frisian on the North Sea Coast of Schleswig-Holstein and the remnant of East Frisian (Saterland Frisian) in northwestern Germany are part of a separate West Germanic language group (see Frisian in Section 31.8). “Low” varieties of West Germanic are found in the lowlands along the coasts of the North Sea and the Baltic Sea as noted earlier. The defining factor for all “Low” varieties of West Germanic is that they do not reflect the consonant changes of the High German consonant shift (see below). Internally, the “Low” West Germanic varieties are further divided into (northwestern) Dutch Low Franconian, West Low German and East Low German based on a feature of their verb morphology. A common characteristic of the verb in all of these varieties is the uniform ending for all three persons in the present indicative plural, as in (1) (see also the discussion of the morphological continuum later in this chapter). (1)

Plural verbal inflection in Dutch and varieties of Low German Dutch (Low Franconian) West Low East Low German German wij maken ‘we make’ wi maakt wi make(n) jullie maken ‘you make’ ji maakt ji make(n) zij maken ‘they make’ se maakt se make(n)

31.2.1 Varieties of Low Franconian As described above, the Low Franconian dialect continuum is located in the western, central, and southern Netherlands, the northern half of Belgium (Flanders), an area in northeastern France (French Flanders) and northwestern North Rhine-Westphalia along the Lower Rhine. This Low Franconian continuum includes such dialects as Hollandic, Limburgish and Brabantian as well as East and West Flemish in Belgium and Kleverla¨ndisch and Bergisch in northwestern Germany (see Figure 31.1). South Africa and Namibia also

741

742

WILLIAM D. KEEL

exhibit the dialects of Afrikaans, which were possibly derived via creolization of the Low Franconian dialects spoken by the early European settlers in South Africa (see Markey 1982; Roberge, Chapter 35). Some Low Franconian varieties such as Limburgish are viewed today as regiolects (informed by Althaus et al. 1980: 458–464; Donaldson 1983; Brachin 1985; “Dialects” at en .wikipedia.org/wiki/Dutch_language; “Low German” at en.wikipedia.org /wiki/German_dialects); “Niederfra¨nkisch” at de.wikipedia.org/wiki/ Deutsche_Dialekte.

31.2.2 Low German Dialects Low German dialects have been spoken and written in northern Germany until the present day, especially in rural areas. From the late Middle Ages until the Early Modern period, varieties of Low German were also widely employed in written documents in northern Germany for legal and commercial purposes, e.g., the variety common in Lu¨beck was in essence the official language of the Hanseatic League throughout the Baltic and North Sea region. After the sixteenth century with the decline of the Hanseatic League and the massive influence of the High German Lutheran Bible in the aftermath of the Protestant Reformation, there was a near complete loss of written forms of Low German by the middle of the seventeenth century. Beginning in the mid-nineteenth century, a rebirth of literary writing in Low German came about through such authors as Fritz Reuter and Klaus Groth. Despite efforts by some radio and television stations to broadcast in Low German (such as Norddeutscher Rundfunk with its popular shows “Talk op Platt” and “Die Welt op Platt”) and some newspapers publishing articles in Low German, the prognosis for the continuance of the Low German varieties is not good. Since 1800, mandatory schooling and the use of Standard German (High German) in schools have had a massive impact on the use of Low German varieties in everyday life. Often, only the older population is able to speak and understand the Low German varieties. The dialects of Low German are subdivided as described above into West Low German and East Low German. Extending eastward from just west of the German-Dutch border, West Low German includes the dialects of the northeastern provinces Overijssel, Drenthe, Gelderland, and Groningen in the Netherlands. Within Germany, West Low Saxon is traditionally divided into Westphalian, Eastphalian, and North Low Saxon dialects with further subdivisions reflecting the specific usage in the northern German states of Schleswig-Holstein (e.g., Holsteinish and Schleswigish) and Lower Saxony (e.g., East Frisian and North Hannoverian). The East Low German dialects developed from the varieties brought by medieval settlers from the West Low German region as they migrated across the northern German plain parallel to the Baltic Sea coast to establish new communities extending to the easternmost part of the Baltic Sea. East Low German includes Brandenburgish (Ma¨rkisch) and Mecklenburg-

The West Germanic Dialect Continuum

Vorpommersch. Historically, East Pomeranian and Lower Prussian belonged to East Low German. Linguistically, the city dialect of Berlin, Berlinerisch, is a Middle-German/Low-German mixed dialect as noted earlier, and is associated along with Su¨dma¨rkisch to both East Low German and East Middle German (see Althaus et al. 1980; Russ 1989; “Low German” at en . w i k i p e d i a . o r g / w i k i / L o w _ G e r m a n a nd e n . w i k i p e d i a . o r g / w i k i / German_dialects; “Niederdeutsche Mundarten” at de.wikipedia.org/wiki/ Deutsche_Dialekte).

31.3 High German Dialects The dialects to the south of the Low German group are defined by the impact of the Second or High German Consonant Shift. The High German Consonant shift did not occur at all in the varieties of Low German and Low Franconian, with isolated exceptions (e.g., ich instead of ik for the pronoun ‘I’ in Limburgish). In the Middle German dialects the shift took place to a limited extent, in Upper German dialects to a greater extent. This sound shift is generally believed to have occurred in the early Middle Ages (prior to the first written sources in the eighth century CE) in the southernmost regions of the Germanic (German) language area, spread gradually to the north and influenced the dialects to varying degrees. The precise origin and manner of spread of this consonant shift is not without controversy. In addition to the traditional view of a southern origin (see Sturtevant 1917; Salmons 2012), some see the origin in the border area between Low and High German with a more generalized spread into the southern varieties (see Prokosch 1917; Becker 1967). Still others view the consonant shift as originating in a more autochthonous pattern with many points of origin (see Schu¨tzeichel 1976). This High German consonant shift involves changes in consonants which are used to demarcate the linguistic boundary between Low and High German dialects, including the shifts in words such as German machen ‘make’ (maken/machen; the so-called Benrath Line) and ich ‘I’ (ik/ich, the so-called U¨rdinger Line) (see Section 31.4).

31.3.1 Central or Middle German Dialects Central or Middle German dialects are located in a broad area delimited on the north by the maken/machen isogloss (Benrath Line) and to the south by the Appel/Apfel ‘apple’ isogloss (Speyer Line). On the west, these varieties are limited by the boundary with French in northern Alsace and Lorraine as well as in southern Belgium and with Low Franconian in northeastern Belgium and the southern part of the Dutch province of Limburg. On the east, they border on Polish as well as Czech with the enclave of Sorbian in the extreme eastern part of Saxony. The Middle German dialects are traditionally divided into a western and eastern half. A bundle of isoglosses

743

744

WILLIAM D. KEEL

separating West from East Central German runs through the area near the rivers Werra and Fulda in the border region of the states of Hesse and Thuringia. The isogloss separating western Pund from eastern Fund (for standard German Pfund ‘pound’) is considered the principal dividing line between East and West Middle German. In Germany, West Middle German, the dialects are divided into Rhine Franconian with Hessian and Palatine (Mainz), Moselle Franconian (Trier) and Ripuarian (Cologne), to some extent reflecting the dominance of the medieval bishoprics associated with those cities. The variant of Moselle Franconian found in Luxembourg (Luxembourgish) has been developed into a written language and functions with German and French as an official language. West Middle German extends into the Saarland as well as large parts of Rhineland-Palatinate, Hesse, northwestern Baden-Wu¨rttemberg, and southwestern North Rhine-Westphalia. Outside of Germany, there are West Middle German varieties in France (Lorraine and northern Alsace), Belgium (Eupen and Malmedy), and the southeastern tip of the Netherlands (Limburg). East Middle German has several groups of closely related dialects including Thuringian and Upper Saxon, each with a number of sub-dialects, as well as Lusatian and South Ma¨rkisch varieties to the south of Berlin along the border with Poland. Historically, the dialects of the former province Silesia (now in Poland) and the so-called High Prussian (also formerly located in today’s northeastern Poland) were classified as East Middle German. The East Middle German varieties bordering on Czech in the south (Erzgebirgisch) exhibit the affricate pf in Apfel and Pfund as an exception to the general rule of Appel and Fund for East Middle German (informed by Althaus et al. 1980: 468–478; Russ 1989; “Mitteldeutsche Mundarten” at de.wikipedia.org/wiki/Deutsche_Dialekte).

31.3.2 Upper German Dialects To the south of the Speyer Line (Appel/Apfel isogloss) all dialects are classified as Upper German, which in turn is differentiated into North, West and East Upper German. North Upper German is subdivided into East Franconian and South Franconian, and some would also include North Bavarian. West Upper German encompasses the Alemannic dialects and East Upper German the three divisions of the Bavarian dialects, North, Middle and South Bavarian. The Upper German dialects, including the High and Highest Alemannic in Switzerland as well as the South Bavarian dialects in Austria and Tyrol (northern Italy) are characterized by the most extensive realization of the products of the second sound shift (see Section 31.4). South Franconian is located primarily in northern Baden-Wu¨rttemberg, but also is located in small parts of the southern Palatinate and northern Alsace in France. East Franconian extends over a larger area in northern Bavaria, extends into southwestern Thuringia (to the Thuringian Forest) and into northeast Baden-Wu¨rttemberg.

The West Germanic Dialect Continuum

Alemannic (West Upper German) dialects include the German varieties of Switzerland (Schwyzertu¨u¨tsch), where some two-thirds of the Swiss speak an Alemannic variety. In neighboring Liechtenstein and in the Austrian province of Vorarlberg, dialects similar to Swiss German are also spoken. In Germany, Alemannic dialects can be found in most of the administrative district of Swabia in western Bavaria, southern Baden-Wu¨rttemberg, and most of Alsace in eastern France (including Strasbourg). Alemannic is divided into Swabian, Low Alemannic varieties along the Rhine River and near Lake Constance as well as High and Highest Alemannic within Switzerland, reflecting the Alpine terrain in the extreme south of that country. Swabian varieties of Alemannic are separated from Bavarian by a bundle of isoglosses in the region known as the Lechrain (see Sprechender Sprachatlas von Bayrisch-Schwaben). South Bavarian dialects are found primarily in the Austrian states of Ka¨rnten, Tyrol and the southern part of Steiermark, as well as in the region of South Tyrol in Italy. Transitional dialects to Middle Bavarian are spoken in the rest of Steiermark as well as in southern parts of Salzburg and Burgenland. Middle Bavarian is spoken in the states of Upper and Lower Austria and in the northern parts of Salzburg and Burgenland. In the German federal state of Bavaria we find Middle Bavarian varieties in the administrative districts of Upper Bavaria and Lower Bavaria, but in the district of the Upper Palatinate the distinctive North Bavarian variety is spoken. The city dialects of Vienna, Austria, and Munich in Germany may be viewed as special urban types of Middle Bavarian dialects. The transition to East Franconian in the northern area of Upper German is gradual with the city dialect of Nu¨rnberg playing a prominent role. A characteristic marker (shibboleth) of the Bavarian dialects is the second person plural pronoun that distinguishes them from both East Franconian to the north and Alemannic to the west: instead of variants of ihr/euch (nominative and oblique forms of you-plural), the corresponding Bavarian forms are es/enk (see map in Ko¨nig 2007: 156), with many internal variations (reflecting a retention of the dual second person pronouns reconstructed for the prehistoric Indoeuropean ancestor language) (informed by Althaus et al. 1980: 479–491; Russ 1989; “Oberdeutsche Dialekte” at de.wikipedia.org/wiki/Deutsche_Dialekte).

31.4 Consonant Continuum: The High German Consonant Shift While vowel distinctions play a significant role in differentiating the dialects of the West Germanic continuum (see Section 31.5), the changes in the historic Germanic plosive consonants */p, t, k/ have served as the primary basis for classifying these dialects. The High German consonant shift or second Germanic consonant shift was a sound change that occurred in the southern region of the West Germanic dialect continuum.

745

746

WILLIAM D. KEEL

As noted in Section 31.3, it is believed to have begun between the third and fifth centuries CE and was largely complete prior to the earliest written records produced in the late eighth and early ninth centuries (for instance, Attila the Hun [fifth century] appears in medieval written sources as Etzel). The historical language that emerges at this time in this southern region is traditionally called Old High German. The written forms of Old High German, with much regional variety, can be contrasted with the other continental West Germanic varieties such as Old Saxon, which was spoken and written in today’s Low German region. Old Saxon generally did not exhibit the shift in the consonants * /p, t, and k/. (Old) English also retained the older Germanic consonants as a rule. This consonant shift led to the regional differentiation of the numerous varieties in the West Germanic continuum where it occurred. However, as noted earlier, the precise manner in which these sound changes evolved and spread remains undetermined (see Sonderegger 1979; Speyer 2010; Salmons 2012). The most widespread of these changes altered the Germanic voiceless stops */p, t, k/ in postvocalic position into the voiceless fricatives /f, s, x/ as in (2). This part of the consonant shift occurred throughout the Middle and Upper German region. The examples in (2) were selected to illustrate the consonant shift in cognate forms without regard to semantic differentiation, e.g., English dapper is cognate with German tapfer, but the meaning of the two words is quite different. The consonants remain unshifted in the English and Dutch examples while the corresponding consonants have shifted in the cognate German forms. (2)

High German consonant shift in postvocalic position */p/ > /f/

*/t/ > /s/

*/k/ > /x/

English up vs. German auf open vs. offen, sleep vs. schlafen, ship vs. Schiff, ripe vs. reif, sheep vs. Schaf, deep vs. tief, pepper vs. Pfeffer English water vs. German Wasser; better vs. besser, nit vs. Nisse, foot vs. Fuss, nut vs. Nuss, eat vs. essen, hot vs. heiss, goat vs. Geiss English make vs. German machen; token vs. Zeichen, break vs. brechen, book vs. Buch, like vs. gleichen, rich/Dutch rijk vs. reich

The bundle of isoglosses delineating the northernmost extent of this shift is used to distinguish Low German / Low Franconian varieties to the north and High (Middle and Upper) German varieties to the south. This major linguistic boundary is frequently referred to as the Benrath Line (Benrather Linie) due its passage through the town of Benrath south of Du¨sseldorf where it crosses the Rhine River. Specifically, the Benrath Line is the boundary between unshifted */k/ in maken ‘make’ to the north and shifted machen to the south. Dialectal varieties to the north of this line are defined

The West Germanic Dialect Continuum

as Low German or Low Franconian (Dutch/Flemish). These include the dialects noted earlier such as Westphalian, Eastphalian, Mecklenburgish, Pomeranian, Brandenburgish, Holsteinish, Schleswigish, among others. In the northwest of the High German area, primarily in the Middle Franconian dialects, one also finds exceptional unshifted forms such as dat ‘that’, wat ‘what’, so¨ken ‘seek’, reflecting the modern German das, was, suchen. In addition, Germanic */p, t, k/ became the affricates /pf, ts, kx/, respectively, in non-postvocalic environments defined as follows: in word-initial position /p-, t-, k-/; when lengthened /-pp-, –tt-, -kk-/; and following a resonant consonant (a liquid /lC, rC/ or a nasal /mC, nC/). After /l/ and /r/ the affricate frequently simplified to a voiceless fricative. The English and Dutch examples in (3) remain unshifted while the corresponding consonants in their German cognates have undergone the High German consonant shift. (3)

High German consonant shift in non-postvocalic position English pipe vs. German Pfeife; help vs. helfen */p/ > /p͡ f (historically helpfan); warp vs. werfen; path vs. Pfad; sharp vs. scharf; slip vs. schlu¨pfen; camp vs. Kampf; -thorpe vs. Dorf; dapper vs. tapfer. */t/ > /t͡ s/ (written English tide vs. German Zeit; smelt vs. schmelzen; ⟨z⟩ or ⟨tz⟩) sit vs. sitzen; holt vs. Holz; too/to vs. zu; slit vs. Schlitz; heart vs. Herz; two vs. zwei; twelve vs. zwo¨lf; timber vs. Zimmer; tooth vs. Zahn; town vs. Zaun; toll vs. Zoll */k/ > /k͡ x/ English child (Dutch kind) vs. Kchind; drink vs. (written ⟨(k)ch⟩) trinkche; clean vs. chlini; cold vs. chalt; cow vs. Chue; cup vs. Chopf (ch = kx; all shifted forms only in extreme south Upper German)

However, only the shift of */t/ to the corresponding affricate /ts/ spread throughout the Middle and Upper German dialects. The shift of */p/ to /pf/ is in turn only found in the Upper German dialects. The primary isogloss between Middle and Upper German, the Appel/Apfel line (‘apple’), is called the Speyer line (Speyerer Linie), since it crosses the Rhine River near the city of Speyer, which in West Middle German is paralleled by the line demarcating Pund to the north from Pfund to the south (‘pound’). However, this distinction becomes blurred in the East Middle German region, where one typically finds Fund and Appel, but in Erzgebirgisch in the southern part of the region Pfund and Apfel, more characteristic of Upper German. The third group of the Germanic voiceless stops, */k/, is only realized as the affricate /kx/ in the extreme southern group of Upper German varieties (High and Highest Alemannic with South Bavarian). The isogloss bundle known as the Sundgau-Lake Constance Barrier, or Kind/Kchind line (‘child’), marks the division between the northern Alemannic and Middle Bavarian dialects without /kx/ and those exhibiting the velar affricate, at least in some forms, in Swiss German and partially in South Bavarian.

747

748

WILLIAM D. KEEL

In summary, the High German consonant shift has been used by linguists since the nineteenth century to classify the West Germanic dialects. The major isoglosses resulting from this historical consonant shift are used to delineate the major dialect regions such as Low German vs. High German or Middle German vs. Upper German as well as the varieties within the larger regions. Along the Rhine River, within West Middle German, the numerous isoglosses based on the consonant shift have produced the intriguing figure of an open fan when displayed on a map and given rise to the term known as the “Rhenish Fan” (Rheinischer Fa¨cher). The shift of the voiceless Germanic stops to fricatives was realized in both Upper and Middle German, together with that of the dental stop to the affricate (*/t/ > /ts/). The shifts of */p, k/ to /pf, kx/ were only realized in Upper German, but in a graduated fashion with the shift of the velar stop limited to the extreme southern dialects, largely to Switzerland and southern Austria. The geographical distribution of each change can be quite complex. Not only do the individual consonants follow their own geographical distribution, but also individual words, at times, seem to follow their own patterns. As Jakob Grimm stated in 1819, “each word has its own history.” For example, the ik/ich line runs further north and west than the maken/machen line from the Lower Rhine of Germany into Belgium, creating a major division within Low Franconian: the dialects with unshifted /k/ are classified as North Low Franconian while those with shifted /k/ are classified South Low Franconian. Moving east, the two lines coincide across central Germany, and then the ik/ich line runs further south of the maken/machen line prior to reaching their eastern terminal points on the linguistic border with Poland. Berlinerisch exhibits ik and machen–reversing the situation in South Low Franconian. Yet both lines reflect the shift of the very same voiceless stop */k/ > /x/ in postvocalic position (see Ko¨nig 2007: 230–231).

31.4.1 High German Consonant Shift in West Middle German South of the Benrath Line, one encounters a number of isoglosses, all based on the consonant shift described above (see Althaus et al. 1980: 469). The Dorp/Dorf line delineates the Ripuarian dialect centered on the city of Cologne from the Moselle Franconian, including Luxembourgish, which in turn is divided into North Moselle Franconian and South Moselle Franconian by the op/auf line. The next dialect region is often labelled Rhine Franconian (Rheinfra¨nkisch), separated from Moselle Franconian by the isogloss dat/das with several subdivisions: The main division into Palatine (Pfa¨lzisch) in the southwest and Hessian in the northeast is marked by the isogloss fest/fescht, which is not related to the consonant shift. Both Palatine and Hessian have a number of subregions within their areas. For instance, to the southwest we find Lorrainese

The West Germanic Dialect Continuum

(Lothringisch), primarily in eastern France, and to the northeast Low Hessian – both of these sub-varieties are marked by the absence of the socalled New High German diphthongs ai/ei, au, eu/a¨u in place of the Middle High German long, high monophthongs ˆı /i:/, uˆ /u:/, iu /y:/ (see below). Also of interest is the division of Palatine into a western and eastern subvariety on the basis of the isogloss for the loss of the full ending in the past participle of strong verbs as in gebroch/gebroche ‘broken.’ Western Palatine dialects as well as western Moselle Franconian, including Luxembourgish, exhibit such past participles with no ending, while Eastern Palatine dialects have the final schwa. The major boundary separating West Middle German from Upper German varieties to the south is marked by the Appel/Apfel isogloss (together with the Pund/Pfund isogloss). This major isogloss, as noted above, is called the Speyer Line (Speyerer Linie).

31.4.2 High German Consonant Shift in East Middle German Lying to the east of the West Middle German area, the eastern counterpart to the numerous varieties of West Middle German is traditionally split into the dialects of Thuringian in the western part and Upper Saxon in the east. Bordering Poland, one also finds Lusatian (closely related to the nearly extinct Silesian) and South Ma¨rkisch bordering on the Low German variety of Brandenburgish to the north. On the Czech border in the southeast, one can also distinguish Erzgebirgisch. With respect to the consonant shift discussed above, the East Middle German dialects generally follow the pattern of West Middle German group. The central and northern part of the East Middle German area, however, while retaining the unshifted reflex of */pp/ in Appel, exhibits a fricative reflex for an initial */p/ with a simple /f/ as in Fund ‘pound.’ This contrasts with both the retained stop characteristic of West Middle German and the affricate /pf/ characteristic of the Upper German dialects to the south. The Pund/ Fund line offers a distinctive boundary between East and West Middle German all the way north to the Benrath Line, the boundary with the Low German dialects in the north. In addition, southern East Middle German varieties, such as Erzgebirgisch, exhibit the shift of initial */p/ and geminate */pp/ to the affricate /pf/ (Pfund instead of Fund/Pund ‘pound’; Apfel instead of Appel ‘apple’), thus sharing this feature with the Upper German dialects.

31.4. 3 High German Consonant Shift in Upper German Consistency in the reflection of the affricate /pf/ in place of the Germanic initial and geminate */p/ is characteristic of all the Upper German dialects. This includes the Alemannic group in the southwest (Swabian, Badish, Alsatian, Swiss German), the Bavarian dialects in the southeast, including

749

750

WILLIAM D. KEEL

Austria and South Tyrol, and the Upper Franconian dialects in the north of the Upper German region. The Swiss region as well as the southern Bavarian group of dialects also exhibit the shift of initial */k/ (and some post-consonantal */k/) to the affricate /kx/ (written kch or ch) as in the wellknown isoglosses Kchind versus Kind ‘child’ or trinkchan versus trinken ‘to drink,’ characteristic of Swiss German and some Tyrolean dialects. While not part of the High German Consonant shift, a significant consonantal marker of the southwestern German dialects should be mentioned here: the palatalization of /s/ to /ʃ/ (), especially when followed by /p/ or /t/. Palatalization of Germanic */s/ is common in initial position except in the northwestern area of the West Germanic continuum, e.g., Schwester ‘sister’ versus swester/su¨ster. However, in the extreme southwest of the continuum (Alemannic and Rhine Franconian) */s/ palatalizes in medial environments as well: e.g., fescht versus fest ‘firm’; geschtern versus gestern ‘yesterday’; Weschpe versus Wespe ‘wasp’ and Schweschter versus Schwester ‘sister’ (see Ko¨nig 2007: 150–151).

31.5 Vowel Continuum Peter Wiesinger (1970) utilized the development in the vowels of medieval German varieties to discern the dialect boundaries of the High German dialects, in essence describing the continuum of vocalic distinctions in these varieties of High German. The same vowels also have reflexes in the Low German and Low Franconian varieties as can be seen on the series of maps at Figures 31.2–31.6. The vowel systems of the West Germanic dialects constitute an important part of the distinctions between the dialects. A few of those vocalic differences are discussed below in Sections 31.5.1–31.5.3.

31.5.1

Diphthongization of Medieval High Long Vowels /i:, u:, y:/ (written î, û, iu) Beginning in the twelfth century, there is evidence of new diphthongs in place of the historical long, high vowels /i:, u:, y:/ represented in the traditional orthography as < ıˆ >, , , respectively (see maps in Ko¨nig 2007: 146–147). These new diphthongs may be represented as (with many orthographical variants). Thus medieval huˆs ‘house’ becomes Haus; hiuser ‘houses’ becomes Ha¨user; wıˆn ‘wine’ becomes Wein. The new diphthongs begin to be realized throughout the southern half of the West Germanic area. Areas retaining the old monophthongs include the Low German varieties in the north, some bordering Middle German varieties just to the south of Low German such as Ripuarian (Cologne) and the southwest varieties of Upper German (Swiss German, Alsatian, Low Alemannic).

The West Germanic Dialect Continuum

31.5.2

Monophthongization of Medieval Diphthongs /iə, uə, yə/ (written ) In a similar fashion, a widespread monophthongization of the medieval diphthongs /iə, uə, yə/ (written ) occurred in the Middle German varieties of the High German dialects (see maps in Ko¨nig 2007: 146, 148). Thus, medieval bruoder ‘brother’ became Bruder; gu¨ete ‘goodness’ became Gu¨te; and lieber [liɘb-] ‘dear’ became lieber [li:b-]. Retained diphthongs are evident in the southern High German dialects such as Bavarian and Swiss German. In some dialects (North Bavarian, Central Hessian, Moselle Franconian) we also encounter forms that have been characterized as “toppled diphthongs” (gestu¨rzte Diphthonge) or reversed diphthongs.

31.5.3 Unrounding of Front Rounded Vowels A phenomenon widespread in the High German dialects is the unrounding of front rounded vowels in the transition from medieval to modern German varieties. Only the dialects of East Franconia (east of Frankfurt), the extreme southwest (High Alemannic / Swiss), and northwest (Ripuarian/Cologne, here showing an affinity with the neighboring Low German to the north) retain front rounded vowels in reflexes of medieval mu¨rede ‘tired’. Again, several varieties in the middle of the region exhibit “reversed diphthongs” (meid/moid) as reflexes of the old diphthongs. Interestingly, the Low German and Dutch dialects generally retain the front rounded vowels except in the easternmost varieties (located in today’s Poland), such as Mennonite Plautdietsch, which were displaced following the Second World War (see map in Ko¨ng 2007: 148).

31.6 Morphological Continuum A number of other phenomena in the realm of morphology also serve to divide the continuum.

31.6.1 Apocope – Loss of Final Schwa A significant vocalic phenomenon that serves to divide the continuum of West Germanic varieties is the loss of final, unstressed vowels, typically schwa [ə]. As is illustrated by the map (at Figure 31.2), only dialects in a band from the Dutch border area in the northwest across northern Germany to the southeast part of the Middle German dialects retain unstressed schwa in final position as in forms such as heute ‘today’ (see also the map in Ko¨nig 2007: 159). While this changes the prosodic shape of many words, it also has a major impact on verb and noun morphology. A number of verb endings and case/plural forms in the noun are impacted.

751

752

WILLIAM D. KEEL

Rostock Hamburg

Berlin

Hannover Munster

Köln Dresden Erfurt Würzburg Nürnberg Wien Mainz Stuttgart

Bern

Munich Innsbruck

Heut(e) um vier(e) (Treffen wir uns doch ...) Zweitmeldungen kleiner

Heute um vier Heut um vier Heut um viere Hüt em vieri Heute um viere

Map 31.2 Apocope of final schwa Atlas zur deutschen Alltagssprache (Elspaß and Möller 2003ff.)

For instance, some have attributed the loss of the preterite tense of verbs in part to the loss of final schwa, e.g., arbeitete versus arbeitet ‘worked’ with the apocopated past tense form now identical with the present tense form for the third person singular. The loss of final schwa has also triggered changes in the morphology of noun case and number. For instance, the loss of -e in the plural of Tag/Tage ‘day/days’ has led to adjustments in marking of the plural often by Umlaut: Tag/Ta¨g.

The West Germanic Dialect Continuum

Other changes in the morphological components of the continuum have affected, for instance, both the grammar of nouns and pronouns (expression of case) and verb morphology (expression of number and tense).

31.6.2 Noun/Pronoun Case An interesting distinction in the West Germanic dialect continuum is the treatment of the morphology of noun and pronoun case distinctions in the dialects. While Standard German retains a four-case distinction for masculine gender nouns (nominative – accusative – dative – genitive), the dialects have, in general, eliminated the genitive case while reducing the distinctions among the other three for definite and indefinite articles as well as personal pronouns. Shrier (1965) demonstrated that the forms for masculine articles and pronouns pattern quite distinctly. Varieties in the north, east, and southeast reflect the morphological distinction of a nominative versus a common oblique case for articles and pronouns in the masculine gender (merging the accusative/dative distinction, not unlike Modern English and Dutch in pronominal forms he/him). On the other hand, the varieties of the southwest exhibit a combined nominative/accusative form versus a distinct dative form. In some instances, dialects in this region exhibit the historical accusative as the common form (Luxembourgish); in others the nominative serves as the common case (Ripuarian). The two maps illustrate this phenomenon. The map at Figure 31.3 with arrows pointing to the north and east demarcates common accusative/dative forms in masculine noun phrases for (a) the definite article, (b) indefinite article, (c) personal pronoun, (d) adjective ending, and (e) first person personal pronoun. The areas with the common oblique forms for the masculine gender include the Bavarian, East Franconian, Thuringian, and Upper Saxon regions. The map at Figure 31.4 illustrates the second pattern in which there is a common nominative/accusative form for the items in question and a distinct dative case for the masculine gender forms. The area demarcated is to the west of the lines on Figure 31.4, encompassing the western parts of the Rhineland, Luxembourg, Switzerland, and BadenWu¨rttemberg.

31.6.3 Expression of Past Time The “Upper German Pra¨teritumsschwund (loss of simple preterite)” is one of the most important morphosyntactic changes in German dialects, in which the synthetic past tense (ich ging ‘I went’) is replaced by the analytic perfect tense (ich bin gegangen ‘I have gone’). Although this phenomenon has repeatedly been the subject of linguistic-historical, dialectological, and contrastive investigations, no precise documentation of the

753

754

WILLIAM D. KEEL

Map 31.3 N-AD case systems to the east (Shrier 1965: 434). Solid line, definite article; short dashes, personal pronoun; dotted line, indefinite article; line of diagonals, adjective; long dashes, 1 S G pronouns

“preterital borders” in the West Germanic continuum has yet been made, nor has a satisfactory explanation for this process been found. In general, however, we may characterize the Upper German varieties as not exhibiting the synthetic preterite forms at all. Middle German varieties exhibit sporadic instances of the synthetic preterite, especially in high frequency verbs such as modal auxiliaries, while the synthetic preterite has largely been retained in the Low German varieties (see map in Ko¨nig 2007: 163).

31.6.4 Plural Verb Endings As noted above in the discussion of Low German verb forms, another aspect of verb morphology which creates distinctions in the continuum

The West Germanic Dialect Continuum

Map 31.4 NA-D case systems to the west (Shrier 1965: 435). Solid line, definite article; short dashes, personal pronoun; dotted line, indefinite article; line of diagonals, adjective

of the West Germanic dialects is the present tense ending of plural verb forms across the three persons (first, second, third) (see Ko¨nig 2007: 158). Many of the Middle and Upper German dialects pattern like Standard German with a two-way distinction: first and third person ending in -en (or some variety of that) and the second person ending in -t or the like. This two-part system is also evidenced in the West Swiss dialects. Other dialects, nearly all of the Low German and the northern Low Franconian dialects as well as most of the southwestern High German dialects, exhibit the so-called “common plural” with one ending on all persons within the present tense plural conjugation. Distinctions are exhibited between West Low German and East Low German (-et versus -en) with the northern Low Franconian forms more typically ending in -en rather than like the neighboring West Low German forms. In the southwest of the Upper and some Middle German dialects we find a number of variants of the common

755

756

WILLIAM D. KEEL

plural ending with some Swiss varieties exhibiting a longer -ent ending throughout the plural. A few High(est) Alemannic varieties even exhibit three distinct plural verb endings in the present tense, e.g., the dialect of Bosco Gurin in the canton of Ticino in Switzerland (Russ 1989).

31.7 Lexical Continuum The West Germanic dialect continuum also displays a great variety of vocabulary differences. This can encompass a variety of different forms for a particular vocabulary item such as “bread roll” with forms such as Rundstu¨ck in the extreme north, Bro¨tchen across the northern area, Schrippe concentrated in the region of Berlin, Semmel throughout the Bavarian/ Austrian continuum and a form of Wecken in the southwest (Figure 31.5). It also includes different forms for the “tag question,” as in the English “It’s a beautiful day, isn’t it?” The variety of forms for the “tag” across the West Germanic continuum range from southern gell (plus its own variants) to nicht, wa, stimmt’s, woll and several others (Figure 31.6). Some vocabulary variants are restricted to a particular dialect such as the pronominal forms es/enk for ‘you plural nominative/oblique’ in Bavarian. Others extend across dialect boundaries and involve more of a regional vocabulary. The variety is almost limitless. For some examples, see the maps in Ko¨nig (2007:166–229, 232–242) and online in the excellent Atlas zur deutschen Alltagssprache (Elspaß and Mo¨ller 2003ff).

31.8 Frisian Varieties Frisian is classified into three major divisions consisting of many subdialects. They belong to the North Sea branch of the West Germanic languages and are quite distinct from the neighboring members of the West Germanic continuum. Frisian was historically along the North Sea coast between the mouths of the Rhine and Elbe rivers and later also extending north from the mouth of the Eider River into southwestern Denmark. Today its varieties are still spoken by slightly more than 400,000 persons, especially in the Dutch province of Friesland, where the variety called West Frisian has been designated, together with Standard Dutch, as an official language. West Frisian is spoken in the Netherlands in the province of Friesland, on two of the offshore West Frisian Islands, as well as in four villages in the western part of the neighboring province of Groningen. The second Frisian variety, Saterland Frisian or Saterlandic, is only spoken in four villages in the Saterland in a very small area of the administrative district of Cloppenburg in the north German state of Lower Saxony (Niedersachsen). The villages lie just outside the borders of the

The West Germanic Dialect Continuum

Rostock Hamburg

Berlin

Hannover Munster

Köln Dresden Erfurt Würzburg Nürnberg Wien

Mainz

Stuttgart

Bern

Munich Innsbruck

Brötchen (länglich) Zweitmeldungen kleiner

Brötchen Brötli Brötla Semme(r)l Weck Weck(er)le

Weckerl Weggi Schrippe Laabla Mütschli Kipf Gibt es am Ort so nicht

Map 31.5 Brötchen ‘bread roll’ (Elspaß and Möller 2003ff.)

administrative district of East Frisia, where, in addition to German, East Frisian Low Saxon, a variant of Low German influenced by East Frisian, is spoken. The varieties which are commonly called North Frisian, the third division of Frisian, are spoken in the northernmost administrative district of Nordfriesland in the state of Schleswig-Holstein along the North Sea just south of the border with Denmark. These distinctive dialects are also spoken on the nearby North Frisian Islands of Sylt, Fo¨hr, Amrum, and

757

758

WILLIAM D. KEEL

Rostock

Hamburg

Berlin

Hannover Munster

Köln Dresden Erfurt Würzburg Nürnberg Wien Mainz Stuttgart

Munich Bern

Innsbruck

Gell (Das ist ein herrliches Wetter heute, ...?) Zweitmeldungen kleiner

Gell Oder Na Nich(t) Nicht wahr Wa

Was Stimmt’s He Woll [Nichts]

Map 31.6 Tag question gell? (Elspaß and Möller 2003ff.)

the Halligen islands as well as on the island of Heligoland in the North Sea. Each North Frisian variety is typically only spoken in its limited location. Communication between speakers of neighboring North Frisian varieties is typically in Low German, High German or even Danish. Experts consider North Frisian – not unlike the situation with Romansh in Switzerland – to be a collection of a number of distinct dialects rather than a language, with the number of dialects hovering around seven (Russ 1989; Munske 2001; see also en.wikipedia.org/wiki/Frisian_languages).

The West Germanic Dialect Continuum

31.9 Conclusion Each base variety within the areal of the West Germanic dialect continuum participates or does not participate to various degrees in each of the linguistic phenomena discussed above and many more. Some evidence hardly any of the major consonantal changes but do exhibit vocalic or morphosyntactic changes. Others exhibit nearly all of the consonantal changes, but retain historical vowel forms and morphology. Lexical variation among the dialects seems to find no limits. In sum, the study of the linguistic variation in the West Germanic base dialects is truly a “neverending story.”

References Althaus, H. P., H. Henne, and H. E. Wiegand (eds.) 1980. Lexikon der germanistischen Linguistik, 2nd edn. (Studienausgabe). Tu¨bingen: Max Niemeyer Verlag. Becker, D. A. 1967. Generative Phonology and Dialect Study: An Investigation of Three Modern German Dialects. Ph.D. dissertation, University of Texas at Austin. Besch, W. and U. Knoop 1982. Dialektologie: Ein Handbuch zur deutschen und allgemeinen Dialektforschung. Berlin: Walter de Gruyter. Brachin, P. The Dutch Language: A Survey. Paul Vincent (trans.). Leiden: E. J. Brill. Donaldson, B. C. 1983. Dutch: A Linguistic History of Holland and Belgium. Leiden: Marinus Nijhoff. Elspaß, S. and R. Mo¨ller 2003ff. Atlas zur deutschen Alltagssprache (AdA). OpenAccess Publication. www.atlas-alltagssprache.de. Grimm, J. 1819. Deutsche Grammatik. 1. edn. Go¨ttingen: Dieter’sche Buchhandlung. Keller, R. E. 1961. German Dialects: Phonology and Morphology, with Selected Texts. Manchester: Manchester University Press. Knipf-Komlo´si, E. and N. Berend (eds.) 2001. Regionale Standards: SprachVariationen in den deutschsprachigen La¨ndern. Budapest: Dialo´g Campus Kiado´. Ko¨nig, W. 2007. Dtv-Atlas: Deutsche Sprache, 16. edn. Mu¨nchen: Deutscher Taschenbuch Verlag. Markey, T. L. 1982. “Afrikaans: Creole or Non-Creole?” Zeitschrift fu¨r Dialektologie und Linguistik 49(2): 169–207. Munske, H. H. (ed.) 2001. Handbuch des Friesischen / Handbook of Frisian Studies. Tu¨bingen: Max Niemeyer. Prokosch, E. 1917. “Die deutsche Lautverschiebung und die Vo¨lkerwanderung,” Journal of English and Germanic Philology 16: 1–21.

759

760

WILLIAM D. KEEL

Rijpma, E. and F. G. Schuringa 1972. Nederlandse sprachkunst. Groningen: Wolters-Noordhoff. Russ, C. V. J. (ed.) 1989. The Dialects of Modern German: A Linguistic Survey. Stanford University Press. Salmons, J. 2012. A History of German: What the Past Reveals about Today’s Language. Oxford University Press. Schirmunski, V. M. 2010. Deutsche Mundartkunde: Vergleichende Laut- und Formenlehre der deutschen Mundarten. L. Naiditch (ed.), W. Fleischer, (trans.). Frankfurt am Main: Peter Lang. Schu¨tzeichel, R. 1976. Die Grundlagen des westlichen Mitteldeutschen: Studien zur historischen Sprachgeographie. Tu¨bingen: Niemeyer. Shrier, M. 1965. “Case systems in German dialects,” Language 41: 420–38. Sonderegger, S. 1979. Grundzu¨ge deutscher Sprachgeschichte: Diachronie des Sprachsystems. Band 1: Einfu¨hrung—Genealogie—Konstanten. Berlin: Walter de Gruyter. Speyer, A. 2010. Deutsche Sprachgeschichte. Go¨ttingen: Vandenhoeck & Ruprecht. Sprechender Sprachatlas von Bayern. www.bayerische-landesbibliothekonline.de/sprachatlas Sprechender Sprachatlas von Bayrisch-Schwaben. www.bayerischelandesbibliothek-online.de/sprachatlas-schwaben Sturtevant, E. H. 1917. Linguistic Change: An Introduction to the Historical Study of Language. University of Chicago Press. Wiesinger, P. 1970. Phonetisch-phonologische Untersuchungen zur Vokalentwicklung in den deutschen Dialekten. Berlin: Mouton de Gruyter. Wiesinger, P. 1980. “‘Sprache’, ‘Dialekt’, und ‘Mundart’ als sachliches und terminologisches Problem.” In J. Go¨schel, P. Ivic, and K. Kehr (eds.), Dialekt und Dialektologie: Ergebnisse des internationalen Symposions “Zur Theorie des Dialekts.” Wiesbaden: Franz Steiner Verlag: 177–194. Wiesinger, P. and E. Raffin 1982. Bibliographie zur Grammatik der deutschen Dialekte. Laut-, Formen-, Wortbildungs- und Satzlehre. 1800–1980. Bern and Frankfurt am Main. Supplemental volume, 1987.

Chapter 32 The North Germanic Dialect Continuum Charlotte Gooskens

32.1 Introduction The North Germanic languages belong to the Germanic language group along with the West Germanic and the extinct East Germanic languages. The North Germanic languages are also often referred to as the Nordic languages, a translation of the term mostly used by the speakers themselves and it refers to the closely related Germanic languages spoken in Denmark, Norway, Sweden, Finland, Iceland, and the Faroe Islands. Traditionally, the North Germanic language family is further subdivided genetically into East Germanic (Danish and Swedish) and West Germanic (Norwegian, Icelandic, and Faroese). Another subdivision is based on the present relationship between the languages and divides the languages into a continental group (Danish, Norwegian,1 and Swedish) and an insular group (Icelandic and Faroese). Sometimes the term Scandinavian is used in a narrow sense to refer to the three mutually intelligible continental Scandinavian languages, while Nordic is used in a wider sense to include Icelandic and Faroese. In this chapter we will follow this terminology. Very diverse local dialects are spoken in the Nordic language area, especially in rural communities. Boundaries between dialect areas are gradual and form a dialect continuum that does not always coincide with national borders. In this chapter we will focus on the Scandinavian language continuum, but we will also discuss the position of Faroese and Icelandic within the Nordic language family.

32.2 The Origin and Development of the North Germanic Dialect Continuum We can distinguish three periods in the history of the Nordic languages (Faarlund 1994: 38, Torp 2002: 19). The first period (Ancient Nordic) ends 1

Norwegian has two official forms, Dano-Norwegian (Bokmål) and New-Norwegian (Nynorsk).

762

C H A R LOT T E GO O S K E N S

around the seventh century. Our knowledge about this period stems from contemporary Roman and Greek authors and from ancient runic inscriptions. The language of the ancient Nordic people was extended over a considerable geographic area but still it appears to have been fairly homogeneous with no known or significant dialect differences. During the second period, from the seventh to the fifteenth century (Old Nordic), the dialects still belonged to one dialect continuum and their speakers all lived in Scandinavia or had recently emigrated to the islands in the Atlantic Ocean, Great Britain, Normandy, Russia etc. The dialect differences were small enough for all Scandinavians to communicate without difficulty and the Nordic language was regarded as one language, referred to as do˛nsk tunga (Danish tongue) (Ottosson 2002: 789). From the end of the eighth century when Iceland and the Faroe Islands were colonized (the Viking Age), dialect differences started to show up that were large enough to justify a division into West Nordic (the varieties spoken on Iceland, the Faroe Islands, and Norway) and East Nordic (Sweden and Denmark). The colonization of Iceland and the Faroe Islands took place mainly from western Norway and the dialectal origin of the language of the emigrants can still be traced in modern Icelandic and Faroese. Until around 1200 the differences between West and East Nordic were small, but the southern part of the language area started to change due to the economic influence from the Hanseatic League and the resulting extensive contact with the German language. This period is often referred to as “middle” (Venaº s 2002: 34–35), but Faarlund (1994: 39) notes that it is a period of transition and many changes at different times in different areas of Scandinavia so that it is not possible to talk about a uniform stage of Middle Scandinavian. However, Torp (1998: 34) suggests distinguishing a separate period in the history of the North Germanic languages between the years 1200 and 1500. In this period the innovations from the south and the large geographical distances and resulting diminishing amount of contact between Scandinavia and the islands in the Atlantic Ocean resulted in a gradual division of the language area into a northern part consisting of Icelandic, Faroese, Norwegian, and Swedish and a southern part (Danish). At this stage the differences between Norwegian, Faroese, and Icelandic were still very minimal. The changes usually took place from the south starting with Danish, followed by Swedish and East Norwegian, then West Norwegian and ending with Icelandic. The Danish pronunciation has undergone an exceptionally fast development during the last century (Brink and Lund 1975, Grønnum 1998) and generally speaking, Danish has been the most innovative of the Nordic languages. On the other hand Icelandic is considered the most conservative of the languages. Both insular Nordic languages have retained a complex inflectional morphology (Thra´insson et al. 2017). Torp (1982: 92) also suggested that there are two centers of innovation in Scandinavia: one in the

The North Germanic Dialect Continuum

Nordic

West Icelandic

Faroese

East Norwegian

Danish

Swedish

Figure 32.1 The Nordic language tree

south where most innovations took place and one in the north that resulted in a division between inner (innovative) and outer (conservative) North Scandinavian. In the modern Nordic languages we make a distinction between insular Nordic (Icelandic and Faroese) and Scandinavian (Danish, Norwegian, and Swedish). The spoken languages of these two groups are not mutually intelligible and the same goes for the two languages within the insular Nordic group. The Scandinavian languages on the other hand are still mutually intelligible with some effort. Within the Scandinavian group Torp (1998: 34) makes a distinction between North Scandinavian (Norwegian and Swedish) and South Scandinavian (Danish). We will return to investigations concerned with mutual intelligibility in Section 32.4. In Figure 32.1 the standard family tree model for the Nordic languages is presented. Such a language tree (also Stammbaum, genetic, or cladistic model) is intended to show the genetic linguistic relationships between languages. The description of the periods in the history of the Nordic languages above makes clear that in the case of the Nordic languages this model is not a good reflection of the linguistic similarities and differences between the modern languages. The tree suggests that Norwegian is more closely related to Icelandic and Faroese than to Swedish and Danish. Inspired by Schmidt (1872), who acknowledged the large role played by patterns of language contact following from political and economic changes at different places in a dialect area, Torp (1982) proposed an alternative model of the development of the Nordic languages, the socalled wave model, where successive linguistic innovations are compared to waves on a water surface. The resulting model of the modern Nordic languages is shown in Figure 32.2, where the number of vertical lines illustrates how different the languages are, from one line (small dialectal differences) to four lines (mutually unintelligible).

32.3 Classification of the North Germanic Dialects The many local Germanic dialects that exist today in the Nordic countries developed in the Middle Ages. As a result of the historical developments described above, the dialects are distributed over three different dialect

763

764

C H A R LOT T E GO O S K E N S

Insular Nordic Icelandic

Scandinavian

Faroese

North Norwegian

South Swedish

Danish

Figure 32.2 The model of the modern Nordic languages (from Torp 2002: 19)

continua: an Icelandic, a Faroese, and a mainland Scandinavian continuum. The dialects in geographical dialect continua form chains of dialects with cumulative differences. This means that dialects spoken in adjacent villages are mutually intelligible. However, due to the cumulative effect of linguistic differences some dialects in a dialect continuum may be so different that they are no longer mutually intelligible. The traditional Scandinavian dialects spoken in most of Denmark, Norway, Sweden, and the southwestern parts of Finland form a classical example of a dialect continuum that crosses geographical borders. The dialects differ in terms of how far they moved away from Old Scandinavian in various parts of the linguistic system (Faarlund 1994: 39). Since the boundaries between the dialect areas are gradual and do not always coincide with national borders, the traditional dialect division is sometimes somewhat arbitrary. Some of the dialects, such as those on the border between Norway and Sweden, especially parts of Bohusla¨n, Dalsland, western Va¨rmland, western Dalarna, Ha¨rjedalen, Ja¨mtland, and Scania, could be described as intermediate dialects of the national standard languages. The differences between the traditional dialects within the countries of Norway, Sweden, and Denmark are often larger than the differences across the borders, but the political independency of the countries and the strong influence of the standard languages, particularly in Denmark and Sweden, leads continental Scandinavian to be classified into Norwegian, Swedish, and Danish by most linguists. The concept heteronomy is used to refer to such a situation where dialects spoken in a certain part of a country are considered a dialect of the standard language of the country even though they may be more similar to dialects spoken in the neighboring country. For example, the present southern Sweden was part of Denmark until 1658 and the dialects spoken in this area were considered Danish dialects. After southern Sweden became part of Sweden the dialects spoken there were also classified as Swedish dialects even though they had not changed linguistically (Chambers and Trudgill 1998: 9–11). Like most European languages, the Nordic languages have recently gone through processes of dialect leveling due to industrialization, urbanization, and migration and today, many traditional Scandinavian dialects have disappeared. Especially in Denmark many dialects are mostly extinct, with only the oldest generations and people living in the geographical periphery still speaking them, probably because of the increased mobility and negative stereotypes connected to dialects (Nyborg 2004). However,

The North Germanic Dialect Continuum

there is still a lot of regional, social, and ethnic variation in the Danish language (Pedersen 2000). Most Danes speak a regional variety of Standard Danish and dialect speakers are mostly bilinguals and only use their dialect at home. In Sweden too, many traditional dialects have been under pressure, even though there is an increasing tolerance towards dialect speakers (Wentzel 2004). Rural, local dialects are disappearing and beeing replaced by varieties that can be situated on a scale between regional dialects and regional standard language. Only in some peripheral areas (especially Upper Dalarna, Norrbotten, and Gotland) local dialects are still spoken. The position of the Norwegian dialects is much stronger than in Sweden and Denmark. Dialects find significant support at all levels of Norwegian society from casual speech to teaching, broadcasting, and parliament. Students freely speak their local dialects in school. Even the Education Act states that they have the right to do so. It states that “in their choice of words and mode of expression, the teaching staff and the school leaders shall as far as possible take the spoken form used by the pupils into account” (Education Act 2014: chapter 2.5). In comparison to the other Nordic languages the dialectal differences in Modern Icelandic are very minor in spite of the large geographical area that covers the Icelandic language area and the isolated living conditions of the inhabitants of Iceland during the centuries (Thra´insson 1994: 142). Most differences are phonetic (Kvaran 2005: 1743). Two major dialect surveys have been carried out in the twentieth century on Iceland with an interval of 40 years. Thra´insson and A´rnason (1992) compared the results of these two investigations and concluded that the dialects that were spoken by the smallest numbers of people around 1940 had disappeared 40 years later and only the majority dialects such as the Reykjavik dialect had survived and spread to larger areas. The homogeneity of the Icelandic language has been explained by the strong literary traditions and the linguistic conservatism on the island (A´rnason 2004). The Faroese language, on the other hand, existed only as a spoken language until the middle of the nineteenth century. This is probably the explanation for the large degree of dialectal variation in Faroese. The differences are primarily phonetic and despite the small number of speakers and restricted geographical spread of the language it is easy to hear where a speaker is from. Still, the differences do not lead to any problems with mutual intelligibility (Barnes and Weyhe 1994: 190). Each of the Scandinavian countries has a strong tradition of dialectological work that has resulted in large numbers of dialect maps showing the geographical distribution of particular phenomena (see Edlund 2010 for an overview of the geolinguistic mapping of Nordic languages). There are also dialect maps showing the classification and geographical distribution of the dialect areas in each of the countries e.g., Nielsen and Pedersen (1991) and dialect.dk (2016) for Danish, Sandøy (1996) and Skjekkeland (2005) for Norwegian, Wesse´n (1969), Elert (1994) and Dahl and Edlund

765

766

C H A R LOT T E GO O S K E N S

(2010) for Swedish, Spraº kinstitutet (2016) for Finland Swedish dialects, A´rnason (2005) for Icelandic, and Thra´insson et al. (2004) for Faroese. However, as noted by for example Ringgaard (2002: 279) and Edlund (2010: 219–220), Scandinavian dialectologists have tended to stay within their own national borders and avoided a pan-Scandinavian or pan-Nordic approach. As a result, few dialect maps have been made showing the whole Scandinavian dialect continuum. Traditional dialectologists present the geographical spread of dialectal features, for instance a particular word form or pronunciation, by drawing isoglosses (lines on a map between locations where speakers use different variants of a feature). A dialect division is said to be major if several isoglosses coincide (isogloss bundles). However, isoglosses usually only coincide approximately, resulting in different dialect areas with a transition zone in between reflecting the fact that the dialects belong to a dialect continuum. In Bandle (1973) a collection of dialect maps is presented, showing a large number of isoglosses and distributions of phonological and other linguistic characteristics of the dialects spoken in the Scandinavian language area. Bandle’s map 22 (see Figure 32.3) shows the characterization of the language area that he made on the basis of the isogloss bundles presented in the maps. He divides the language area into three major dialect areas, West Scandinavian, South Scandinavian, and North Scandinavian. The last one is subdivided into four areas. Inspection of the map makes clear that, in general, the dialect borders do not coincide with national borders. Dialect maps such as the one presented by Bandle are based on a limited number of single linguistic features. Such isogloss maps are verifiable, but the motivation for the selection of the isoglosses remains unclear. The isogloss method cannot be applied without making subjective choices (Goossens 1977) and there are no principles to determine which isoglosses or which bundles should outrank others (Chambers and Trudgill 1998: 96–97). Dialectometric research has sought more satisfying characterizations by aggregating over a large number of linguistic features to express linguistic distances between dialects. The aggregating step has allowed the introduction of powerful quantitative techniques into dialectology in many language areas (for an overview see Nerbonne and Heeringa 2010) and dialectometric methods can find relationships in data sets that are too large and complex for manual analysis on an objective basis. These statistical techniques make it possible to draw dialect maps presenting dialect areas as continua rather than as areas separated by sharp borders. To carry out dialectometric measurements a large quantity of digital data is needed. In Scandinavia various dialect projects have been set up to collect corpora of dialect data, for example the Scandiasyn project (Johannessen et al. 2009), SweDia 2000 (Eriksson 2004), and Talko (So¨dergaº rd and Leinonen 2017). However, these collections of data have only sporadically been used for creating dialect maps.

The North Germanic Dialect Continuum

ICELAND

Icelandic

Faroese

FÆROE ISLANDS West Scandinavian South Scandinavian

SC A N DI N

A

V

IA

N

North Scandinavian

TH R

O

E a s t

s h d i S w e

N

WEST SCAND INAVIA N

North Scandinavian narrowly defined

Göthic Gutnish

SOUTH SCANDINAVIAN

Map 32.1 Characterization of the Nordic dialects (from Bandle 1973, map 22)

An example of a dialectometric investigation is the analysis carried out by Leinonen (2010) of the vowel pronunciations of over one thousand Swedish speakers from the SweDia 2000 project. She used aggregate differences in vowel quality (formant measurements on nearly 20,000 vowel tokens) and subjected the result to factor analysis and multidimensional scaling. She then projected the results onto maps of Sweden showing the geographical distribution of vowel pronunciations. A comparison of the pronunciations of older and younger speakers showed that a large amount of variation is being lost in the ongoing process of dialect leveling. Gooskens and Heeringa (2004) measured

767

768

C H A R LOT T E GO O S K E N S

phonetic distances between 15 Norwegian dialects by means of the Levenshtein algorithm, a string edit distance measure that has proven to be a successful method for measuring phonetic distances between dialects in various language areas (Nerbonne and Heeringa 2010). They projected the distances onto a map of Norway. They also asked speakers of the 15 dialects to judge the distances between their own dialect and each of the other 14 dialects. The results showed a high correlation between the distances as perceived by the speakers themselves and objective dialectometric distance measurements. The classification of the Norwegian dialect area obtained from the two measures is rather similar. In both, a north-south division is found. Trudgill (1986) suggests that dialect distances reflect mobility and social contact and that this explains the fact that dialects form continua with geographically remote areas being linguistically less similar than geographically close areas. If a place is geographically close, people are likely to go to this place more often and therefore language varieties spoken in villages that are geographically close have a greater chance of influencing each other than if there is a larger geographical distance between them. Nerbonne et al. (1996) found a high correlation (r=.67) between straight line geographic distances and linguistic distances between 350 Dutch dialects by means of the Levenshtein distance method. Gooskens and Heeringa (2004) found a considerably lower correlation (r=.22) between linguistic distance and geographic distance in the case of 52 Norwegian dialects. Differences in geography are likely to explain the difference between correlations in the Dutch and the Norwegian language areas. In the Netherlands it has always been rather easy to travel from place to place since there are hardly any natural obstacles such as mountains. In Norway on the other hand, is has been difficult to travel between villages due to the many mountains and in the past most of the traveling took place by boat along the coast. When correlating linguistic distances between Norwegian dialects with old traveling distances Gooskens (2005) got a higher correlation (r=.53). This is an indication that varieties spoken in places which are easily reached show a greater linguistic similarity to each other than varieties spoken in more isolated places and that this effect is still reflected in the distances between modern Norwegian dialects.

32.4 Linguistic Distances between North Germanic Languages To categorize closely related language varieties Kloss (1967) introduced the terms Ausbausprache (language by development) and Abstandsprache (language by distance). An Ausbausprache is mostly an official national language with its own standardized form. It is often taught in schools and used as a written language for various social and political functions. In an Ausbau

The North Germanic Dialect Continuum

definition, “languages” and “dialects” are defined in terms of their sociopolitical and cultural status rather than by independently identifiable structural entities. The official standard Scandinavian languages, Danish, Norwegian (Nynorsk and Bokmaº l), and Swedish, are an often mentioned example of Ausbau languages that are very closely related but still regarded as different languages because they are spoken in three different countries and have distinct, standardized forms, with their own orthographies, dictionaries, grammars, and literatures. One language variety is called an Abstand language with respect to another language variety if the two are so different from each other that they can be considered different languages. Kloss (1967) did not specify exactly how to measure the differences between two language varieties objectively, probably because the necessary tools were not yet available. As discussed in Section 32.3, methods to measure linguistic distances objectively have now been developed by dialectometrists. In this section we will summarize research on linguistic differences between the Nordic languages. Due to phonological innovations in the Danish language, the largest differences between Norwegian and Swedish on the one hand and Danish on the other hand are found at the phonetic level (Torp 1998: 69–70). Lexical, morphological, and syntactic differences between the Scandinavian languages are generally assumed to be small (Torp 1998: 105). Dialectometric measurement techniques make it possible to quantify linguistic distance between the languages and such measurements have confirmed the general impression expressed by Scandinavian linguists. Torp (1998) found that only six percent of the 343 most frequent nouns in a corpus of Swedish newspapers are likely to cause problems to a Norwegian reader. Gooskens and Heeringa (2014) measured lexical distances (expressed as percentages of noncognates, i.e., words without a common etymological origin, in the fable The North Wind and the Sun) and phonetic distances (measured by means of the Levenshtein algorithm, see Section 32.3) between standard Danish and Norwegian (represented by the Oslo variety) and 18 Nordic language varieties (including standard Swedish, various Scandinavian dialects, and the variety of Torshavn as a representative for Faroese). The results confirmed the small lexical distances between the Scandinavian languages (all lower than 2.2 percent). The phonetic distances between Danish and Norwegian/Swedish are larger (21.6 and 23.5 percent) than between Norwegian and Swedish (18.5 percent). Both the lexical and the phonetic distances to Faroese are largest (lexical distance to Norwegian 13.8 percent and to Danish 13.7 percent, phonetic distances to Norwegian 25.2 percent and to Danish 27.8 percent), but a number of Swedish and Norwegian dialects are almost as phonetically deviant from standard Danish as Faroese. Unfortunately, the investigation did not include Icelandic. Dialectometric measurements within the insular Nordic language group have hardly been carried out so far, but

769

770

C H A R LOT T E GO O S K E N S

Gooskens and Heeringa (2005) showed that the distance between the Icelandic and the Faroese pronunciation is large and almost as large as the distances between these two languages and Swedish. As mentioned, morphological and syntactic distances between the Scandinavian languages are generally assumed to be very small (Torp 1998: 105). Only recently, methods for measuring syntactic distances dialectometrically have been developed. Heeringa et al. (2017) measured syntactic distances between five Germanic languages (Danish, Dutch, English, German, and Swedish) by means of three different dialectometric methods. As expected, the results show that the syntactic distances between Swedish and Danish are smaller than between any of the other nine language pairs in the investigation. Gooskens and Heeringa (in preparation) measured linguistic distances between 35 closely related European language pairs (national standard languages from the Germanic, Romance, and Slavic language families), including Swedish-Danish. Distances were only measured between language pairs within the same language family. The comparison of the Swedish-Danish distances to the distances between the other languages pairs can help to place the measurements in a larger perspective. The lexical and syntactic distances between Swedish and Danish are the lowest among the 35 language pairs in the investigation. However, compared to the other language pairs the phonetic distances are rather high. Twentythree language pairs have smaller phonetic distances, including all Slavic, most of the Romance language pairs, and German-Dutch.

32.5 Mutual Intelligibility of North Germanic Languages and Dialects As explained in Section 32.4, linguistic distance is one of the criteria for distinguishing between dialects and languages. The problem remains, however, that languages may differ to different extents in their lexicon, phonetics and phonology, morphology, and syntax. It is difficult to decide how much weight should be given to each of these linguistic dimensions when determining overall distance. Maybe to circumvent this problem, Trudgill (2000) introduced the intelligibility criterion and this has become the primary criterion among many linguists. According to this criterion, dialects are mutually intelligible varieties, whereas languages are so linguistically different that their speakers are unable to understand each other. From this it follows that a language is a collection of mutually intelligible dialects. As became clear from the discussion of the Scandinavian dialect continuum in Section 32.2 and the linguistic differences in Section 32.4, the standard Scandinavian languages are so similar that with some effort speakers of Danish, Swedish, and Norwegian can communicate sufficiently well to sustain a meaningful exchange of information each using their own native

The North Germanic Dialect Continuum

languages. Haugen (1966) introduced the term semi-communication for this manner of communicating. In later research other terms were coined, e.g., receptive multilingualism (Braunmu¨ller 2007), plurilingual communication (Lu¨di 2007), intercomprehension (Grin 2008), and lingua receptiva (Rehbein et al. 2012), see Ten Thije (2017) for a discussion of terminology. We will use the term receptive multilingualism in the remainder of this chapter. It is an alternative to other kinds of cross-border communication such as English (or another language) used as a lingua franca and offers many advantages, especially on the production side. People can express themselves more easily and more precisely in their mother tongue than in a later acquired language. There is evidence that receptive multilingualism was a common means of communication between speakers of Germanic languages across the Nordic/non-Nordic border during the era of the Hanseatic League from about 1200 to 1550 (Braunmu¨ller 1996). Today, receptive multilingualism is a common way of communicating in Scandinavia among speakers of Danish, Norwegian, and Swedish. Furthermore, it is supported by the Scandinavian governments as a way of uniting the countries (Nordic Language Convention 1987) and relatively much research on the mutual intelligibility among the mainland Scandinavian languages has been carried out. In Table 32.1 the results of the most important investigations including all the three Scandinavian languages are presented. Haugen (1953, 1966) was the first to investigate how well Scandinavians are able to understand each other’s languages. He distributed a questionnaire among 252 Danes, Norwegians and Swedes (88 percent male) whose names Table 32.1 Results of five investigations on the mutual intelligibility between spoken and written Danish, Norwegian, and Swedish Target language Danish

Native lang.

Spoken

Written

Norwegian Spoken 72 69 63 92 39

Da

Haugen Maurud Bø Börestam INS

No

Haugen Maurud Bø Börestam INS

50 73 62 80 46

93 75 81

Sw

Haugen Maurud Bø Börestam INS

44 23 32 48 23

69 41 60

61 48 74 91 61

Written 89 48 73

86 56 58

Swedish Spoken

Written

40 43 74 80 29

69 56 65

70 88 80 98 75

89 76 80

771

772

C H A R LOT T E GO O S K E N S

were selected at random from the national telephone directories. The questionnaire contained several questions with regard to cross-language intelligibility, such as how well the participants could understand their Scandinavian neighbors when they met them for the first time, how well they understood the neighboring languages at the time of the investigation, and how well they thought their Scandinavian neighbors could understand them. The results showed that communication involving Norwegians is most successful (most scores were above 60 percent). Danish is a difficult language for both Norwegians (50 percent) and Swedes (44 percent). The lowest mutual intelligibility is found between Danes and Swedes (40 percent for Danes and 44 percent for Swedes, Haugen 1966: 289). Haugen’s results were based on self-reports and he suggested that his data should be checked against functional tests. Maurud (1976) presented two spoken dialogues and three monologues to Danish, Norwegian, and Swedish participants and asked them six questions about the content of each text. Afterwards they were asked to translate twelve words in each of six written texts. The results confirm the high intelligibility of Norwegian among Danes. Swedes have difficulties with both Norwegian and Danish resulting in asymmetric intelligibility scores: 48 versus 88 percent correct answers to content questions in the case of Norwegian-Swedish mutual intelligibility and even lower (23 versus 43 percent) in the case of SwedishDanish mutual intelligibility. The scores for written intelligibility are all higher than the scores for spoken intelligibility and show no asymmetries. Again, intelligibility involving Norwegian is especially high (all scores above 86 percent) and the Swedish-Danish mutual intelligibility is lowest (both 69 percent). A weakness of Maurud’s (1976) study is that the Danish participants came from Copenhagen and the Swedish participants came from Stockholm. Copenhagen is closer to the Swedish border (30 km) than Stockholm is to the Danish border (570 km), so the Danes were likely to have had more exposure to Swedish than vice versa because they could more easily travel to the neighboring country and watch television in the neighboring language. When the investigation was carried out only inhabitants in the border areas had access to television programs from the other countries. Bø (1978) tested participants from the border area as well as from places outside the border area. He let secondary school pupils read and listen to a text and translate certain key words. In general, Bø’s (1978) results are similar to those of Maurud (1976). They confirm the asymmetry found between spoken Danish and Swedish (32 versus 74 percent). The mutual intelligibility between spoken Norwegian and Swedish is highest (74 and 80 percent). As expected the results of Bø’s investigation showed that participants who lived in a border region had fewer problems understanding the neighboring language than participants who lived outside a border region. Bo¨restam Uhlmann (1991) collected self-reports on the mutual intelligibility of Danish, Norwegian, and Swedish. Just as Haugen (1953, 1966), she

The North Germanic Dialect Continuum

asked her participants (members of several Nordic language meetings) to estimate how well they understood the neighboring languages, to which the participants could answer “very well,” “fairly well,” and “very badly.” The self-reports elicited by Bo¨restam Uhlmann (1991) confirmed the asymmetry between Danish and Swedish found by Maurud (1976) and Bø (48 versus 80 percent). A comparison of Haugen’s (1966) and Bo¨restam Uhlmann’s results shows that Scandinavians seem to have become more positive about their ability to understand their neighboring languages. Several factors could explain this difference. Firstly, there is a time difference of about 30 years between the two investigations. Since the Scandinavian governments have promoted the use of receptive multilingualism, Bo¨restam Uhlmann’s participants might have been more familiar with this way of communicating than Haugen’s participants. Secondly, the participant groups were different. Haugen (1966) selected random participants from telephone directories, while Bo¨restam Uhlmann (1991) visited Nordic language meetings, where the participants were used to interScandinavian communication. The most recent large investigation of the mutual intelligibility of spoken and written Danish, Norwegian, and Swedish was carried out with support from the Nordic Culture Fund and is referred to as the INSinvestigation. The results are reported in Delsing and Lundin Åkesson (2005). In contrast with earlier investigations the INS-investigation included the testing of the intelligibility of the three Scandinavian languages in all Nordic countries. It tested reading and listening comprehension of both adolescents and adults with different levels of education and with different language backgrounds. To measure intelligibility of spoken texts the researchers had secondary school students and their parents answer a number of open questions about a video fragment and a news item. To test intelligibility of written texts they had the participants read a newspaper article and translate key words in the text. The results in Table 32.1 are the results of the young native speakers in the three Scandinavian countries.2 In general, the recent Scandinavian results are similar to results found in previous investigations. The INS results replicate the asymmetry between spoken Danish and Swedish found in most earlier investigations. Mutual intelligibility is highest between Norwegians and Swedes (61 and 58 percent correct answers for the Swedes and higher, 75 percent and 80 percent, for the Norwegians); Danish is hard to understand, especially for Swedish-speaking listeners (23 percent correct). Contrary to most previous research, there was also an asymmetric intelligibility between written Danish and written Swedish. To examine the influence of previous exposure to the test language, participants from two places in each country were scored. In general, 2

The videos used for the different test languages were different and not equally difficult, and therefore the results are not presented here.

773

774

C H A R LOT T E GO O S K E N S

participants from places close to the border of the country where the test language was spoken performed better that participants living further away. This confirms the importance of exposure for the intelligibility of a closely related language as found by Bø (1978), even in the present-day situation where access to television from the neighboring countries is not restricted to the border areas. The parents of the adolescents also participated in the investigation and got higher scores on all tests than their children, especially among the Swedes and the Danes. It is possible that this can be explained by the fact that adults have had more exposure to the neighboring languages, but Delsing and Lundin Åkesson (2005: 143–144) also assume that political and cultural developments such as membership of the European Union, globalization, migration, increased mobility, and internationalization of the mass media explain the lower scores of the Swedish and Danish adolescents. The investigations that have been carried out on the mutual intelligibility between Danish, Norwegian, and Swedish since the pioneering research by Haugen (1966) show similar overall trends. In Figure 32.4 we summarize the results by presenting the mean scores of the five investigations. In general Norwegians are better at understanding the neighboring languages than Danes and Swedes. Danish is the most difficult language to understand. Danes understand spoken Swedish better than Swedes understand Danish, but this asymmetry in not found in the written mode. The five investigations cover a time frame of 40 years, but since the methods and the participants involved in the five studies are different it is not possible to compare the results directly and we do not know whether the level of mutual intelligibility in Scandinavia has changed. This means that there is no general way to express how well

100 90 80 70 60 50 40 30 20 10 0

Danish

Norwegian spoken

Swedish

Danes

Danish

Norwegians

Norwegian written

Swedish

Swedes

Figure 32.3 The mean results of the five investigations on mutual intelligibility in Scandinavia in Table 32.1

The North Germanic Dialect Continuum

Scandinavians understand each other’s languages. The results of Haugens’s (1966), Bø’s (1978), and Delsing and Lunding Åkesson’s (2005) investigations make clear that the amount of exposure that speakers of related languages have had to the languages spoken in the neighboring countries play an important role. This shows that it is important to pay attention to the geographical background when selecting participants for an investigation. As far as the intelligibility of the neighboring languages among Norwegians is concerned, it has often been mentioned that the Norwegian dialects command a strong position and that Norwegians are therefore used to listening to varieties that are different from their own variety (Torp 1998: 118). It has also been suggested that there is a relationship between intelligibility and attitudes towards the neighboring language, culture, and people (Haugen 1966; Delsing and Lundin Åkesson 2005; Gooskens 2006). However, it has not been possible, so far, to draw conclusions about the causality of the relationship between attitude and intelligibility. It is possible that people are positive toward a language if they find it easy to understand, but it is also possible that people make a greater effort to understand a language if they have a positive attitude toward the language, its speakers, and the country where it is spoken. It is generally assumed that Danes are more positive towards Swedes than the other way round because Sweden has for a long time been the most prosperous country with the largest population and this is part of the explanation for the asymmetric intelligibility found in most investigations on Swedish-Danish mutual intelligibility. Schu¨ppert et al. (2016) showed that the large number of reduction processes in Danish may also be part of the explanation for the asymmetry. These reduction processes are hardly reflected in the Danish orthography and Schu¨ppert (2011) showed that Danes can use their orthographic knowledge when decoding spoken Swedish cognates while this is much less the case for Swedes decoding Danish cognates. Returning to the discussion of how to define a language in Section 32.4, it is difficult to decide whether the mutual intelligibility between Danish and Swedish is now so low that the two languages should be defined not only as Ausbau languages but also as Abstand languages. Predicting mutual intelligibility between two related languages from structural linguistic differences can only be done as long as the interlocutors have had no prior exposure to the other language (inherent intelligibility). Gooskens et al. (2018) measured the mutual intelligibility between 35 European language pairs (national standard languages from the Germanic, Romance, and Slavic language families), including Swedish-Danish (but not Norwegian). To analyze the results from participants who had not learned the languages at school and had had minimal exposure to the test language (inherited intelligibility) 11 language combinations had to be excluded from the data set. By comparing the results of mutual intelligibility of Swedish and Danish to the results of the remaining 23 language pairs the results can be

775

776

C H A R LOT T E GO O S K E N S

put into perspective. The results show that Swedish-Danish mutual intelligibility is higher than that of most other language pairs. Their mean scores on a spoken cloze test was 49.9 percent. The Czech participants had considerably higher scores when tested in their intelligibility of Slovak (87.5 percent)3 and the mean score for Slovenian-Croatian mutual intelligibility was also higher (53.4 percent). The Italian-Spanish mutual intelligibility was slightly lower (47.9 percent) and next came the Spanish-Portuguese mutual intelligibility (45.2 percent). These results still do not allow us to draw conclusions about the status of Danish and Swedish as Abstand languages, but it provides us with some indications of the level of mutual intelligibility compared to other language pairs in Europe. As mentioned above, extra-linguistic factors seem to play an important role in explaining the level of mutual intelligibility in Scandinavia. However, linguistic distances also show a strong relation with intelligibility. Gooskens (2007) correlated spoken intelligibility data collected by Delsing and Lundin Åkesson (2005) with phonetic and lexical distances (see Section 32.4) and found a significant correlation with phonetic distances (r = .80, p < .001) but not with lexical distances (r = –.42, p = .111). Gooskens et al. (2008) investigated the role of phonetic and lexical distances in the intelligibility of 17 Scandinavian language varieties among young Danes from Copenhagen. They found significant correlations with both lexical (r = .86, p < .01) and phonetic distances (r = .64, p < .01), but a multiple regression analysis included only phonetic distances, which suggests that phonetic distance is a better predictor of intelligibility than lexical distances in the Scandinavian language area. To get a better impression of particular word characteristics that play a role in intelligibility Ku¨rschner et al. (2008) tested the intelligibility of 384 frequent Swedish words among Danish subjects and correlated the results with 11 linguistic factors. In addition to phonetic distances, word length, different number of syllables than in Danish, foreign sounds not present in Danish, neighborhood density, word frequency, orthography, and the absence of the prosodic phenomenon of ‘stød’4 in Swedish had a significant influence on the level of intelligibility.

32.6 Conclusions and Desiderata for Future Research In this chapter we have seen that the Nordic languages have developed from a common origin into three different dialect continua where six official standard languages are spoken, so-called Ausbau languages. Icelandic and Faroese are two Abstand languages and the Scandinavian 3

The results of the Slovak participants listening to Czech had to be excluded since almost all Slovak participants had had prior exposure to the Czech language.

4

Phonetically a kind of glottalization of either a vowel or a sonorant, sometimes produced as creaky voice, sometimes as a glottal stop.

The North Germanic Dialect Continuum

languages are mutually intelligible to such an extent that they are usually considered one Abstand language. However, the Danish pronunciation has changed considerably during the past century and is sometimes hard to understand for other Scandinavians, especially to Swedes with no prior exposure to the language. Since the first investigations of mutual intelligibility in Scandinavia there have been many changes in Scandinavian society and English often takes over as a lingua franca in the communication between people in the Nordic countries. The distance and intelligibility criterion introduced by Kloss for defining languages as opposed to dialects is problematic in cases like Danish and Swedish because it is not clear exactly how little intelligibility and precisely how large the distances on different linguistic levels should be for two languages to be considered Abstand languages. It would be of great value for language planning and policies in different countries to develop a standardized method for determining how to distinguish between dialects and languages, for example a standard intelligibility test or objective distance measurements with a standardized threshold. Traditionally, Scandinavians communicate by means of receptive multilingualism. It is important for collaboration in Scandinavia that communication can take place in the Scandinavian languages rather than in a lingua franca such as English. Scandinavian authorities therefore strongly encourage cross-border communication in the native languages of the inhabitants. Also to the individual citizen there are many advantages of being able to express themselves in their own languages. Previous research (e.g., Hedquist 1985, Golubovic´ 2016) has shown that in the case of closely related languages only very little exposure or a short course that makes speakers conscious of the most important differences and similarities between their native language and the language of the speaker can improve mutual intelligibility considerably. It is therefore of great importance to make people conscious about the advantages and possibilities of communicating with receptive multilingualism and to create opportunities for people to be exposed to the neighboring language and learn strategies for improving communication, for example through school exchange programs. To make such programs as efficient and successful as possible, more research is needed into what linguistic differences between the languages and what strategies the speakers should be taught to communicate efficiently in the receptive multilingualism mode. Even though dialect leveling has taken place in many parts of Scandinavia, many dialects are still spoken and initiatives have been taken to collect and digitalize dialect corpora. With the many dialectometric methods that have recently been developed the following challenge will be to use the digital resources to characterize the dialects and develop maps covering the whole Scandinavian or Nordic area. Such panScandinavian mappings across language boundaries may provide a safer

777

778

C H A R LOT T E GO O S K E N S

basis for interpretation and new insights into Nordic language history and the forces determining the spreading and vitality of the dialects spoken in the language area today. Dialectometrical studies can establish the geographical distribution of particular linguistic phenomena at various linguistic levels as well as at an aggregate level and help explaining the history, context, and dynamics of dialectal variation in the Nordic countries.

References A´rnason, K. 2004. “Island – landet uden dialekter.” In I. S. Sletten (ed.), Nordens sprog med rødder og fødder. Nord, Nordic Council of Ministers, Copenhagen K, https://doi.org/10.6027/Nord2004-010. A´rnason, K. 2005. I´slensk tunga, Vol. 1: Hljo´ð. Handbo´k um hljoðfræði og hljoðkerfisfræði [Icelandic Language, Vol. 1: Sounds. Handbook of Phonetics and Phonology]. Reykjavı´k: Almenna bo´kafe´lagið. Bandle, O. 1973. Die Gliederung des Nordgermanischen (Beitra¨ge zur nordischen Philologie 1). Basel and Stuttgart: Helbing & Lichtenhahn. Barnes, M. and E. Weyhe 1994. “Faroese.” In E. Ko¨nig and J. van der Auwera (eds.), The Germanic Languages. London and New York: Routledge: 190–218. Bø, I. 1978. Ungdom og naboland. En undersøkelse av skolens og fjernsynets betydning for nabospraº ksforstaº elsen. [Youth and neighbouring country. An investigation of the influence of school and TV on inter-Scandinavian comprehension.] Stavanger: Rogalandsforskning. Bo¨restam Uhlmann, U. 1991. Spraº kmo¨ten och mo¨tesspraº k i Norden. [Language meetings and meeting languages in the Nordic countries.] Nordisk spraº ksekretariats rapporter 16. Oslo: Nordisk spraº ksekretariat. Braunmu¨ller, K. 1996. “Forms of language contact in the area of the Hanseatic League: dialect contact phenomena and semicommunication,” Nordic Journal of Linguistics 19: 141–154. Braunmu¨ller, K. 2007. “Receptive multilingualism in Northern Europe in the Middle Ages: A description of a scenario.” In J. D. ten Thije and L. Zeevaert (eds.), Receptive Multilingualism. Amsterdam: John Benjamins: 25–47. Brink, L. and J. Lund 1975. Dansk rigsmaº l: lydudviklingen siden 1840 med særligt henblik paº sociolekterne i København. Copenhagen: Gyldendal. Chambers, J. K. and P. Trudgill 1998. Dialectology, 2nd edn. Cambridge University Press. ¨ . and L.-E. Edlund (eds.) 2010. Spraº ken i Sverige. [The languages of Dahl, O Sweden]. 1. utg. Sveriges nationalatlas. Stockholm: Norstedts. Delsing, L.-O. and K. Lundin Åkesson 2005. Haº ller spraº ket ihop Norden? En forskningsrapport om ungdomars fo¨rstaº else av danska, svenska och norska. [Does the language keep together the Nordic countries? A research report of

The North Germanic Dialect Continuum

mutual comprehension between young speakers of Danish, Swedish, and Norwegian.] Copenhagen: Nordiska ministerraº det. Dialekt.dk 2016. dialekt.ku.dk/dialekter/dialektgraenser/. Edlund, L.-E. 2010. “Mapping the North Germanic languages.” In A. Lameli, R. Kehrein, and S. Rabanus (eds.), Language and Space: An International Handbook of Linguistic Variation. Berlin: Mouton de Gruyter: 203–243. Education Act 2014. www.regjeringen.no/en/dokumenter/education-act /id213315/. Elert, C.-C. 1994. “Indelning och gra¨nser inom omraº det fo¨r den talade svenskan: en aktuell dialektgeografi.” In L.-E. Edlund (ed.), Kulturgra¨nser – myt eller verklighet? DIABAS, Vol. 4. Institutionen fo¨r nordiska spraº k, University of Umeaº : 215–228. Eriksson, A. 2004. “SweDia 2000: A Swedish dialect database.” In P. J. Henrichsen (ed.), Babylonian Confusion Resolved. Proc. Nordic Symposium on the Comparison of Spoken Languages. Copenhagen Working Papers in LSP: 33–48. Faarlund, J. T. 1994. “Old and Middle Scandinavian.” In E. Ko¨nig and J. van der Auwera (eds.), The Germanic Languages. London and New York: Routledge: 38–71. Golubovic´, J. 2016. Mutual Intelligibility in the Slavic Language Area. Groningen: Center for Language and Cognition. Gooskens, C. 2005. “Traveling time as a predictor of linguistic distance,” Dialectologia et Geolinguistica 13: 38–62. Gooskens, C. 2006. “Linguistic and extra-linguistic predictors of InterScandinavian intelligibility.” In J. van de Weijer and B. Los (eds.), Linguistics in the Netherlands, Vol. 23. Amsterdam: John Benjamins: 101–113. Gooskens, C. 2007. “The contribution of linguistic factors to the intelligibility of closely related languages,” Journal of Multilingual and Multicultural Development 28.6: 445–467. Gooskens, C. and W. Heeringa 2004. “Perceptive evaluation of Levenshtein dialect distance measurements using Norwegian dialect data,” Language Variation and Change 16.3: 189–207. Gooskens, C. and W. Heeringa 2005. “De moderne norske dialekters stilling indenfor den nordiske sprogfamilie.” In S. Lie, G. Nedrelid, and H. Omdal (eds.), MONS 10. Utvalde artiklar fraº det tiande Møte om norsk spraº k i Kristiansand 2003. Kristiansand: Høyskoleforlaget: 133–146. Gooskens, C. and W. Heeringa 2014. “The role of dialect exposure in receptive multilingualism,” Applied Linguistics Review 5.1: 247–271. Gooskens, C. and W. Heeringa in preparation. Linguistic distances between Germanic, Romance and Slavic languages. Gooskens, C., W. Heeringa, and K. Beijering 2008. “Phonetic and lexical predictors of intelligibility,” International Journal of Humanities and Arts Computing 2.1–2: 63–81. Gooskens, C., V. J. van Heuven, J. Golubovic´, A. Schu¨ppert, F. Swarte, and S. Voigt 2018. “Mutual intelligibility between closely related languages

779

780

C H A R LOT T E GO O S K E N S

in Europe,” International Journal of Multilingual Communication, 15.2: 169–193. Goossens, J. 1977. Inleiding tot de Nederlandse dialectologie [Introduction to Dutch dialectology]. Groningen: Wolters-Noordhoff. Grin, F. 2008. “L’intercompre´hension, efficience et e´quite´.” In V. Conti and F. Grin (eds.), S’entendre entre langues voisines: vers l’intercompre´hension. Cheˆne-Bourg: Georg: 79–109. Grønnum, N. 1998. Fonetik og fonologi. Almen og dansk. Copenhagen: Akademisk forlag. Haugen, E. 1953. “Nordiske spraº kproblemer – en opinionsunderso¨kelse. [Nordic language problems – a poll],” Nordisk tidskrift fo¨r vetenskap, konst och industry 29: 225–249. Haugen, E. 1966. “Semicommunication – The Language Gap in Scandinavia,” Sociological Inquiry 36: 280–297. Hedquist, R. 1985. Nederla¨ndares fo¨rstaº else av danska och svenska. En spraº kpedagogisk underso¨kning med utnyttjande av likheterna mellan spraº ken. [Dutchmen’s understanding of Danish and Swedish. A language pedagogical investigation with the use of similarities between the languages.] Umeaº : Umeaº universitet. Institutionerna fo¨r fonetik och nordiska spraº k. Heeringa, W., F. Swarte, A. Schu¨ppert, and C. Gooskens 2018. “Measuring syntactical variation in Germanic texts,” Digital Scholarship in the Humanities, 33.2: 279–296. Johannessen, J. B., J. Priestley, K. Hagen, T. A. Åfarli, and Ø. A. Vangsnes 2009. “The Nordic Dialect Corpus – an advanced research tool.” In K. Jokinen and B. Eckhard (eds.), Proceedings of the 17th Nordic Conference of Computational Linguistics NODALIDA 2009. NEALT Proceedings Series, Vol. 4: 73–78. Kloss, H. 1967. “‘Abstand languages’ and ‘ausbau languages’,” Anthropological Linguistics 9.7: 29–41. Ku¨rschner S., C. Gooskens, and R. van Bezooijen 2008. “Linguistic determinants of the intelligibility of Swedish words among Danes.” International Journal of Humanities and Arts Computing 2.1–2: 83–100. Kvaran, G. 2005. Orð: handbo´k um beygingar- og orðmyndunarfræði. Reykjavı´k: Almenna bo´kafe´lagið. Leinonen, T. 2010. An Acoustic Analysis of Vowel Pronunciation in Swedish Dialects. Ph.D. thesis, University of Groningen. Lu¨di, G. 2007. “The Swiss model of plurilingual communication.” In K. Bu¨hrig and J. D. ten Thije (eds.), Beyond Misunderstanding: Linguistic Analyses of Intercultural Communication. Amsterdam and Philadelphia: John Benjamins: 159–178. Maurud, Ø. 1976. “Reciprocal comprehension of neighbour languages in Scandinavia: An investigation of how well people in Denmark, Norway and Sweden understand each other’s languages,” Scandinavian Journal of Educational Research 20: 49–71.

The North Germanic Dialect Continuum

Nerbonne, J., W.Heeringa, E. van den Hout, P. van de Kooi, S. Otten, and W. van de Vis 1996. “Phonetic distance between Dutch Dialects.” In G. Durieux, W. Daelemans, and S. Gillis (eds.), Proceedings of Computer Linguistics in the Netherlands ’95. Antwerpen: Centre for Dutch Language and Speech (UIA): 185–202. Nerbonne, J. and W. Heeringa 2010. “Measuring dialect differences.” In J. E. Schmidt and P. Auer (eds.), Language and Space: Theories and Methods. Berlin: Mouton de Gruyter: 550–567. Nielsen, B. J. and K. M. Pedersen 1991. Danske talesprog. Copenhagen: Gyldendal. Nordic Language Convention 1987. https://www.norden.org/en/avtal/nordiclanguage-convention. Nyborg, L. 2004. “Farvel til dialekterne i Danmark.” In I. S. Sletten (ed.), Nordens sprog med rødder og fødder. http://eplads.norden.org/nordens sprak/kap5/5c.asp. Ottosson, K. 2002. “Old Nordic: A definition and delimitation of the period.” In O. Bandle, K. Braunmu¨ller, E. H. Jahr, A. Karker, H-P. Naumann, and U. Teleman (eds.), The Nordic Languages: An International Handbook of the History of the North Germanic Languages. Berlin and New York: Walter de Gruyter: 787–793. Pedersen, I. L. 2000. “Fra dialekt til multietnolekt.” In T. Kristiansen and F. Køster (ed.), Danske Talesprog, Vol. 1: 5–29. Rehbein, J., J. D. ten Thije, and A. Verschik 2012. “Lingua receptiva (LaRa) – remarks on the quintessence of receptive multilingualism,” International Journal of Bilingualism 16.3: 248–264. Ringgaard, K. 2002. “General history of Nordic dislectology.” In O. Bandle, K. Braunmu¨ller, E. H. Jahr, A. Karker, H.-P. Naumann, and U. Teleman (eds.), The Nordic Languages: An International Handbook of the History of the North Germanic Languages. Berlin and New York: Walter de Gruyter: 272–282. Sandøy, H. 1996. Talemaº l, 2nd edn. Oslo: Novus Forlag. Schmidt, J. 1872. Die Verwandtschaftsverha¨ltnisse der indogermanischen Sprachen. Weimar: H. Bo¨hlau. Schu¨ppert, A. 2011. Origin of Asymmetry: Mutual Intelligibility of Spoken Danish and Swedish. Groningen: Center for Language and Cognition. Schu¨ppert, A., N. H. Hilton, and C. Gooskens 2016. “Why is Danish so difficult to understand for fellow Scandinavians?” Speech Communication 79: 47–60. Skjekkeland, M. 2005. Dialektar i Noreg: Tradisjon og fornying [Dialects in Norway: Tradition and Modernization]. Kristiansand: Høyskoleforlaget, So¨dergaº rd, L. and T. Leinonen 2017. “Talko – korpus o¨ver den talade ¨ stman, svenskan i Finland: Korpusbygge i teori och praktik.” In J.-O. O º C. Sandstro¨m, P. Gustavsson, and L. So¨dergard (eds), Ideologi, identitet, intervention: Nordisk dialektologi 10: 331–340. Helsingfors: Finska, finskugriska och nordiska institutionen vid Helsingfors universitet.

781

782

C H A R LOT T E GO O S K E N S

Spraº kinstitutet 2016. www.sprakinstitutet.fi/sv/ordbocker/dialektordbocker/ ordbok_over_finlands_svenska_folkmal/karta_over_dialektomraden. Ten Thije, J. D. 2017. “Receptive multilingualism.” In D. Singleton (ed.), Twelve Lectures on Multilingualism. Bristol: Multilingual Matters. Thra´insson, H. 1994. “Icelandic.” In E. Ko¨nig and J. van der Auwera (eds.), The Germanic Languages. London and New York: Routledge: 142–189. Thra´insson, H. and K. A´rnason (1992). “Phonological variation in 20th century Icelandic,” I´slenskt ma´l 14: 89–128. Thra´insson, H., C. B. Heycock, P. Petersen, and Z. S. Hansen (eds.) 2017. Syntactic Variation in Insular Scandinavian. Amsterdam University Press. Thra´insson, H., H. P. Petersen, J. ı´ L. Jacobsen, and Z. Svabo Hansen 2004. Faeroese: An Overview and Reference Grammar. To´rshavn: Føroya Fro´ðskaparfelag. Torp, A. 1982. Norsk og nordisk før og naº [Norwegian and North Germanic past and present]. Oslo: Universitetsforlaget. Torp, A. 1998. Nordiske spraº k i nordisk og germansk perspektiv. Oslo: Novus. Torp, A. 2002. “The Nordic languages in a Germanic perspective.” In O. Bandle, K. Braunmu¨ller, E. H. Jahr, A. Karker, H.-P. Naumann, and U. Teleman (eds.), The Nordic Languages: An International Handbook of the History of the North Germanic Languages. Berlin and New York: Walter de Gruyter: 13–24. Trudgill, P. 1986. Dialects in Contact. Oxford: Blackwell. Trudgill, P. 2000. Sociolinguistics: An Introduction to Language and Society. London: Penguin. Venaº s, K. 2002. “Previous attempts at establishing periods in Nordic language history.” In O. Bandle, K. Braunmu¨ller, E. H. Jahr, A. Karker, H-P. Naumann, and U. Teleman (eds.), The Nordic Languages: An International Handbook of the History of the North Germanic Languages. Berlin and New York: Walter de Gruyter: 31–38. Wentzel, M-L. 2004. “Faº r du tala ditt eget spraº k?” In I. S. Sletten (ed.), Nordens sprog med rødder og fødder. Nord, Nordic Council of Ministers, Copenhagen K, http://doi.org/10.6027/Nord2004-010. Wesse´n, E. 1969. Vaº ra folkmaº l, 9th edn. Stockholm: Fritzes [first published in 1935].

Chapter 33 Heritage Germanic Languages in North America Janne Bondi Johannessen and Michael T. Putnam

33.1 Introduction In this chapter, we focus on varieties of Germanic languages, including Scandinavian varieties, acquired as a first language (L1), yet spoken outside of the region of origin. Such languages are commonly referred to as heritage languages, as explained by Rothman (2009: 156): A language qualifies as a heritage language if it is a language spoken at home or otherwise readily available to young children, and crucially this language is not a dominant language of the larger (national) society. Like the acquisition of a primary language in monolingual situations and the acquisition of two or more languages in situations of societal bilingualism / multilingualism, the heritage language is acquired on the basis of an interaction with naturalistic input and whatever inborn linguistic mechanisms are at play in any instance of child language acquisition. Differently, however, there is the possibility that quantitative and qualitative differences in heritage language input and the introduction and influence of the societal majority language, as well as differences in literacy and formal education can result in what on the surface seems to be arrested development of the heritage language or attrition in adult bilingual knowledge.

America, and in particular, the Midwest, is home to heritage languages for a majority of the Germanic languages. Many are documented and discussed in Putnam (2011), Johannessen and Salmons (2015), Page and Putnam (2015), Johannessen (2018), as well as in numerous individual monographs and papers. Here we focus on structural aspects from varieties of heritage Norwegian and German spoken in the Midwestern US and

Many thanks to Bill Keel, Mark Louden, and Robert Klosinski for comments and suggestions that improved this piece. Thanks to Lara Schwarz for both substantive comments as well as editorial assistance.

784

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

beyond. On occasion, when applicable, we will also refer to other Germanic heritage grammars to illustrate a particular point. The heritage Germanic languages in America all belong to the traditional immigrant language category, as opposed to modern immigrant languages and indigenous heritage languages (Johannessen in progress), as well as minority languages (see Louden Chapter 34). Unlike some languages of Germanic ancestry, e.g., Mennonite Low German, Hutterite German, Pennsylvania Dutch, and Yiddish, that continue to be acquired and spoken as an L1 for future generations, most of today’s heritage languages are almost exclusively moribund. The eldest generation make up the vast majority of these varieties and is the final one with members who possess significant proficiency in the grammar. Additionally, we recognize that there are global varieties of heritage Germanic languages that extend well beyond the United States; however, due to time and space considerations, we focus here on varieties spoken in the Midwestern US. Although we acknowledge that there are some important differences between the American varieties of heritage Germanic compared with those in other countries – especially with respect to the particular sociolinguistic conditions and the typological similarity/distance between the two languages in contact – the similarities between these groups outweigh their differences. Studying traditional immigrant heritage languages often takes place from a comparative perspective: The mother language (in our case European Germanic languages) is used as a baseline against which one can measure variation and change. There are noted advantages and disadvantages to this approach; although it is advantageous to establish a means of comparison with a baseline language and/or dialect of origin, it comes at the expense of establishing a comparison fallacy between monolinguals and bilinguals (Cook 1997; Judy et al. 2018). This caveat notwithstanding, the comparative method represents common practice in this field of linguistic inquiry. The study of structural aspects of heritage grammars is both rewarding and fascinating. Studies focusing on heritage grammars also force linguists to learn and investigate diachronic developments in synchronic forms, from periods where we know more about the language situation and the speakers, in instances where recordings from previous generations exist. The kinds of comparative study that heritage languages allow us to do, from comparing with the European baseline language to comparing with the language of the first immigrant generation or two, to comparing with genetically similar and different languages, can give us much knowledge of how languages change and what factors cause it. Comparisons with child language acquisition and those classified as moribund provide insight into factors like incomplete acquisition, attrition, and instances of restructuring. In addition to these studies whose focus primarily concerns the structure of these grammars, there exists additional sociolinguistic research on these communities. In this chapter, we offer a survey of

Heritage Germanic Languages in North America

structural developments in the phonological, morphosyntactic, syntactic, and semantic/pragmatic developments of heritage German, heritage Norwegian, and examples from other Germanic languages. In the final section of this chapter we discuss theoretical challenges and controversies associated with research on heritage languages in these communities.

33.2 Phonology We begin our overview by taking a closer look at phonological developments in these Germanic heritage grammars. The most comprehensive work on the phonology of Norwegian in America is Hjelde (1992), who compares the phonological system of the American heritage Norwegian dialect of Trøndsk with European Trøndsk. The Trøndsk heritage dialect has lost the feature [rounding] in the realization of /y/, yielding forms like [bɪː] for by ‘town’ and [mɪː] for mye ‘much’. However, importantly, there is still a slight phonetic difference between the realization of the delabialization product and the old /i/, resulting in minimal pairs such as [mɪː] mye ‘much’ and [miː] mi ‘my’ (Hjelde 1992: 42). Hjelde discusses the reasons for this change. Since American English does not have [rounding] as a distinctive feature, it is perhaps interference (transfer) from English that has caused this change. However, the new distinction in heritage Norwegian Trøndsk is now [spread lips], which is not a distinctive feature in American English. Hjelde (1992: 43) suggests that this new feature is just temporary, while the language is on its way to an American system, which has five degrees of distinctive mouth opening, in contrast to the Norwegian four. The American English /ɪ/ has the same phonetic features as heritage Norwegian Trøndsk /e/, except that the latter has an additional feature, viz. [spread lips] (Hjelde 1992: 8–9). The /ɪ/ is never retained as such in loanwords into Trøndsk, but is replaced with allophones of existing phonemes. Some examples of loanwords illustrate this (the pronunciation of the loanword in Trøndsk is presented with phonetic transcription in square brackets): [i] as in gipsy [d͡ ʒipsi], [iː] as in beer [biːr], [e] as in brick [bɹeks], [æ] as in brickstone [brækːstein], and even [ø] as in crib [krøb:a]. These findings are especially interesting seen next to the development of [y] to [ɪ] above, and show that the system as such remains intact, even with the addition of new words that have phonemes whose allophones are already in use. The vibrant /r/ has also gone through a process of change, from an alveolar vibrant to an approximant /ɻ/, though not necessarily for all words and phonological contexts (Hjelde 1992: 44). Hjelde characterizes both phenomena as interference from English, though he stresses that the /r/ phoneme changes for other reasons, too. In European Norwegian this phoneme has undergone a transformation from alveolar to uvular in many dialects, partly for sociolinguistic reasons, but also because this

785

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

VOT Change in real time 80 VOT in ms

786

Std. English

60 Std. German

40 20

WI English

0 1946

1970

2001

Year

Dane Cty Kölsch

Figure 33.1 VOT-times in Wisconsin Kölsch and varieties of English (from Geiger and Salmons 2006)

sound is speech-motorically difficult to produce and is thus one of the last to be mastered by children, making it vulnerable to interference. Hjelde also reports some reanalyses of the old phoneme flap /ɺ/ as an allophone of /r/. Though there are changes in the realization of the phonemes, it seems that the phonological system remains. Hjelde, then, suggests three factors for variation and change: speech motorics, sociolinguistics, and interference (transfer). In other domains, such as Voice Onset Time (VOT) in Germanic dialects, not only do we see evidence of the retention of native-like contrast, but also lasting effects that the heritage L1 may potentially have on the dominant L2. Geiger and Salmons’s (2006) study of VOT-times in the Ko¨lsch dialect spoken in Dane County, Wisconsin, reveals that the English spoken in this region by both monolinguals and bilinguals has had a long-term effect on the English spoken here. In summary, research into the phonological aspects of heritage Germanic grammars demonstrates that these systems are highly resistant to (significant) change and attrition over the course of the lifespan.

33.3 Morphology and Morphosyntax The inflection of loanwords has been the focus of study for many scholars (Haugen 1953, Hjelde 1992, Åfarli 2012, Grimstad et al. 2014, Riksem in press). Riksem (2018, section 6) uses the modern speech corpus CANS, Corpus of American Norwegian Speech (Johannessen 2015a), and finds nearly 1,000 English nouns used in heritage Norwegian conversations. Of these, 10 percent have (an American English) plural -s, while the rest are integrated into the Norwegian morphological system. The number of words with the English morphological suffix is so high that it could be argued that -s is a plural morpheme in heritage Norwegian (at least among those who use it).

Heritage Germanic Languages in North America

Hjelde (2015) studies changes in a Norwegian heritage dialect in a little town in the American Midwest, based on data from speakers born as far back as 1849 and as recently as 1961. There is some paradigmatic morphological leveling in the past tense suffix and instances of dative case. These changes suggest that a koine´ization has been taking place, without having been completed (Hjelde 2015: 295). Eide and Hjelde (2015: 85) find some loss of finiteness morphology. Johannessen and Larsson (2015) and Lohndal and Westergaard (2016) both discuss gender morphology with empirical evidence from the CANS corpus. Johannessen and Larsson (2015) look at noun phrase internal agreement, but also discuss noun declension, and show that there are many more deviations (relative to a European Norwegian baseline) in agreement in preposed determiners and adjectives than in the definiteness suffix, something that they attribute to complexity. Lohndal and Westergaard (2016) find that there is an overgeneralization of masculine gender in indefinite articles, and that the masculine targets more neuter than feminine nouns. Like Johannessen and Larsson, they find that the definiteness suffix is target-like. Thus, while nontarget gender forms can be found on preposed determiners and adjectives, the definite form is expressed as a suffix on the noun itself, and is rarely or never found with nontarget gender. Example (1) exemplifies a nontarget demonstrative determiner and a target noun: The speaker has used a masculine determiner (denna) with a neuter noun (skolehus). The definiteness on the noun has the correct neuter suffix (-et). (1)

denna andre skolehuset this.M . S G . D E F other. D E F school.building. N . S G . D E F this other school building’ (target: detta ‘this’. N . S G . D E F ) (westby_WI_06gm) (from Johannessen and Larsson 2015: (7c))

While Johannessen and Larsson argue that the definiteness suffix also exhibits gender, Lohndal and Westergaard argue that the different morphological suffixes are nothing more than declension class markers (in addition to the definiteness), and conclude that gender is vulnerable to change. Both of these aforementioned studies have good arguments for their views, and new research is needed to ascertain how the data should finally be interpreted. Rødvand (2017) studied gender based not just on noun phrase-internal elements, but also on pronouns, in a group of speakers using elicitation techniques. She found that the speakers must be divided into different categories depending on their individual gender system. While half of the speakers have a three-gender system that is target-like, the rest have various versions of a less full system, for example one in which there are no longer feminine pronouns. She finds, like Lohndal and Westergaard, that there is massive overgeneralization of the masculine form. She also sees a connection between the definite suffix and

787

788

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

the gender system, again similar to the claims advanced by Johannessen and Larsson. Research on the morphosyntactic case systems of heritage German varieties abounds, with a traditional focus on the paradigmatic loss of case markings first with articles, and eventually pronouns (e.g., Minronow 1957, Shrier 1965, Keel 1994, Boas 2009, among others). As an illustrative example, consider this history of Pennsylvania Dutch. In the literature on Pennsylvania Dutch, there are two primary dialects, a sectarian variant spoken by members of the Old Order Amish church, and nonsectarian variant that has largely become moribund over the course of the last century. The literature on case systems in Pennsylvania Dutch has observed an important distinction between the morphosyntactic case systems of these two groups of Pennsylvania Dutch-speakers. Over the course of the twentieth century nonsectarians, most of whom no longer speak the language on a regular basis, can be classified as heritage speakers. This situation contrasts with that of sectarian speakers, who still pass on the language to their children and are diglossic bilinguals. These sectarian speakers (i.e., the Old Order Amish) display a reduction in their case system from a three-tier distinction (nominative, accusative, and dative) to a two-tier distinction similar to English (nominative versus nonnominative/oblique). This pattern stands in opposition to what is found in the grammar of nonsectarian Pennsylvania Dutch, which has largely retained dative case marking, and thus, a three-tier paradigmatic distinction (Louden 1988, Huffines 1989). In addition to these patterns observed in the morphosyntax of Old Order Amish, the so-called Schwartzentruber Amish in Ohio, who broke away from the Old Order Amish in 1913, still actively use the dative case (and thus a three-tier distinction in their speech) (see, e.g., Louden 2016: 324, 352; p.c.). According to Louden (2016: 324), the Schwartzentrubers “in many areas have very little contact with non-Schwartzentrubers, both Amish and non-Amish”, thus “the use of English is therefore much less widespread [. . .] hence it is not surprising that their Pennsylvania Dutch shows less structural influence from English.” This discussion of the case system in varieties of Pennsylvania Dutch leads to a question that commonly resurfaces in studies focusing on morphosyntactic case in heritage varieties of German; namely, whether or not this perceived loss of morphosyntactic case markings is primarily attributed to internal or external language change (Keel 1994). The prevailing view is that this change is most likely driven by internal factors, given that similar instances of multi-generational case loss have taken place in the former Soviet Union, where Russian was the societal dominant second language (L2) (for an overview of this research, see Berend and Jedig 1991, Rosenberg 2003). If external factors, such as transfer, were mostly at play here, one would anticipate paradigmatic expansion rather than reduction, since Russian has more case distinctions than (dialectal varieties of) German.

Heritage Germanic Languages in North America

In addition to the oft-documented collapse of case paradigms in these heritage German varieties, there also exist studies that argue for expansion and restructuring. Keel (1994, 2015) calls to attention the creation of an ablative case in Volga German varieties spoken on the Great Plains of Kansas. Consider the following examples taken from the Volga German dialect spoken in Victoria (Herzog) in Ellis County, Kansas (data from Keel 2015: 135): Standard German: (2)

(a) Der Mann ist ins Wasser gefallen. the man.N O M is in+the.A C C water fallen ‘The man fell into the water.’ (b) Der Mann hat im Wasser gestanden. the man.N O M has in+the.D A T water stood ‘The man stood in the water.’

Volga German dialect of Victoria / Ellis County, Kansas: (3)

(a) Der Mann is in+den Wasser gefall. (target: ins) the man.N O M is in+the.A B L water fallen ‘The man fell into the water.’ (b) Der Mann hot in+den Wasser gestand. (target: im) the man.N O M has in+the.A B L water stood ‘The man stood in the water.’

In standard German – as well in most other dialect varieties – there is a distinction between the singular neuter accusative definite article das and its dative equivalent dem. In prepositional phrases that can be interpreted as either involving motion towards a goal or being stative, German marks the object of a preposition (i.e., the Ground in a Figure-Ground configuration) with accusative case in the former and dative in the latter. Compare this with the forms from Volga German; here the declined form for singular masculine accusative nouns (i.e., den) appears before a neuter noun, Wasser ‘water.’ Based on the fact the noun Wasser is a neuter form when it does not appear as an object of a preposition in this dialect of Volga German (cf. (2a) with (3a) and (2b) with (3b), respectively), we can conclude that the form den is a new case form (ablative). It is interesting to note, however, that emergence of this ablative class does not appear in all Volga German dialects in this isolated area; for example, whereas Johnson (1994) reports similar findings in the community of Schoenchen (also in Ellis County, Kansas), Khramova’s (2011) field work on the Volga German dialect spoken in neighboring Milberger (in Russell County, Kansas) documents the maintenance of dative forms. The creation of an ablative case in the Kansas Volga German dialects found in the communities of Schoenchen and Victoria (Herzog) are canonical examples of internal restructuring (see, e.g., Polinsky’s 2011 work on relative clauses in heritage

789

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

Russian). It is difficult to classify this shift as attrition or an expansion of the already existing paradigm; however, what is clear from examples such as these is that the common narrative of attrition and loss may be more complex than often assumed. Recent research by Yager et al. (2015) and Yager (2016) has uncovered a somewhat surprising finding in connection with reported morphosyntactic case loss. In fact, rather than losing paradigmatic case distinctions, these studies show that multiple varieties of heritage German have developed innovative restructuring mappings between form and semantic and pragmatic meaning. In particular, the dative case adopts a more indexical discourse function along the lines of what is commonly observed in Differential Object Marking (DOM) in Romance and other language families (Bossong 1985, 1991; Aissen 2003). Yager et al. (2015) incorporates data from five distinct heritage German-speaking communities; i.e., data from three different Wisconsin heritage German varieties, Texas German, and Misionero German (a Palatinate variety spoken in the Misiones Province of Northern Argentina). Figure 33.2 from Yager et al. (2015: 6) shows that 32.2 percent of oblique definite noun phrases are marked with some form of morphosyntactic case, while third person singular pronouns are marked on 41.4 of all tokens, with the difference between these two proportions being statistically significant. Similar distributions were observed in the other two varieties of heritage German under investigation. As Yager et al. (2015: 8) report, “The patterns observed in our data, though, cannot be explained simply in terms of direct influence from sociolinguistically dominant L2 grammars, i.e., English, rural vernacular Portuguese, and Spanish. [. . .] Instead, we see a new, divergent grammatical property, the rise of DOM. As is often the case with DOM, its occurrence is tendential rather

Differences in Wisconsin German case marking between NPs and pronouns 100% Proportion of object tokens

790

80% 60% 40% 20% 0%

marked unmarked

definite NPs 273 573

3 sg. Pronouns 60 85

Figure 33.2 Differences in Wisconsin Heritage German case marking between NPs and pronouns

Heritage Germanic Languages in North America

Table 33.1 Proportions of case marking on elements by definiteness Degree of definiteness Most definite Least definite

Category

Frequency of case marking

Proportion of casemarked tokens

Personal pronouns Proper nouns Definite NPs Indefinite NPs

81.8% 62.5% 64.7% 3.2%

54/66 20/32 176/271 1/31

than categorical.” Focusing exclusively on the varieties of Wisconsin heritage German in a more expansive study, Yager (2016) analyzed the case marking on nominal and pronominal tokens (n=5,191) of 21 speakers, coding for gender, number, case, article type, animacy, etc., elicited from semistructured interviews. If DOM-effects are present, we would anticipate a higher frequency of case marking on pronouns when compared with full nouns since the former show a higher degree of definiteness and animacy. Indeed, this prediction is borne out in Yager’s (2016) findings, where elements classified as most definite did display a case marking versus those that were not (i.e., indefinite NPs). Yager’s findings are summarized in Table 33.1 below (adopted from Yager 2016: 251, Table 66). In conclusion, Yager et al. (2015) and Yager (2016) make the case that appealing to incomplete acquisition for the rise of DOM-effects in these varieties of heritage German is unsatisfactory based on the fact that L1 German-speaking children acquire command of structural case marking by age three. Case is thus acquired so early in childhood that one must assume it is completely acquired. The treatment of morphosyntactic case highlights a larger debate in research on heritage languages and language attrition more generally; namely, whether or not instances such as these can correctly be considered as evidence for incomplete acquisition or some form of attrition. We return to this point in the final section of this paper (Section 33.6).

33.4 Syntax 33.4.1 Word Order Variation While linguists studying the Scandinavian heritage languages in America during the twentieth century mainly looked at phonology and the lexicon, the empirical object of study has expanded, and word order is now a popular topic among linguists. European Norwegian, like most Germanic languages (other than English), adheres to verb-second (V2) pattern in main clauses, i.e., the verb will always be the second constituent (see Vikner 1995, Chapter 16, for a detailed overview). This is independent of whether the first constituent is a subject or a topicalized constituent, and whether the verb is an auxiliary verb or a lexical verb, see (4).

791

792

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

(4)

a. Jon kjøpte litt mer land i gaº r Jon bought little more land in.yesterday ‘Jon bought some more land yesterday.’ b. I gaº r kjøpte Jon litt mer land in.yesterday bought Jon little more land ‘Yesterday Jon bought some more land.’ (Eide and Hjelde 2015: 68)

Eide and Hjelde (2015) looked at old recordings of heritage Norwegian (recorded by Einar Haugen in the 1940s), and find no instances of nontarget word order. They compare this with the word order of a speaker in the 2010s, and find that she has target word order in subject-initial clauses, but an instability in nonsubject-initial clauses. While short and frequent clause-initial adverbs occur with V2, heavier fronted topics trigger V2 violations. Johannessen (2015b) also investigated the language production of one speaker. Unlike Eide and Hjelde she finds several examples of V2 violations in subject-initial sentences, see (5), but, like Eide and Hjelde, she finds that light adverbs trigger V2 more often than other sentence-initial adverbials, see (5)–(6): 92 percent of the clauses that have fronted light adverbs have target V2 word-order, but none of the heavier sentence-initial clausal adverbials, and few of the PP-initial ones. (5)

(6)

Vi bestandig hadde fisk we always had fish ‘We always had fish.’ (Johannessen 2015b: 60)

(target: hadde bestandig)

a. Da fløtte dem til Flekke (target) then moved they to Flekke ‘Then they moved to Flekke.’ b. I skolen vi snakte bare engelsk (target: snakte vi) in school we talked only English ‘At school we only spoke English.’ c. Naº r jeg kom tilbake, jeg kunne ikke gaº bak (target: kunne jeg) when I came back I could not walk behind ‘When I came back, I couldn’t walk behind.’ (Johannessen 2015b: 60–62)

Both Eide and Hjelde (2015) and Johannessen (2015b) find that the speech of their informants shows a robust V2-pattern, and that topicalized sentences are more likely to have V2 if the constituent is small and frequent. Their speakers differ, however, in subject-initial clauses, where Eide and Hjelde’s speaker only has target-like word order, while Johannessen’s speaker has a nontarget V3-order if the sentence contains an adverb. These two speakers are interesting because their nontarget speech is systematic. However, it should be mentioned that Johannessen

Heritage Germanic Languages in North America

chose her speaker precisely because this informant showed deviation from the V2-pattern in her language production. Johannessen’s goal was to test the Regression Hypothesis (Jakobsen 1941; Keijzer 2007, 2010). There are probably many heritage speakers that have a fully target-like V2-system, but this remains to be investigated. Taranrød (2011) and Larsson and Johannessen (2015a, 2015b) investigate subordinate clause word order, particularly the order of relative clauses. They show that heritage Norwegian exhibits both V–Adv and Adv–V order, even from the same speaker, (7a, 7b), while in European Norwegian only the latter is found.1 (7)

a. . . . var mye melk der [som ble ikke brukt] was much milk there that was not used ‘Was a lot of milk there that wasn’t used.’ b. . . . [som ikke tar mer penger] who not take more money ‘Who don’t take out more money.’ (Westby_WI_02gm, from Taranrød 2011: 51)

In European Norwegian, then, main clause word order is V2 and subordinate clause word order V3 (also known as non-V2). This is generally explained as verb movement to the C position in main clauses, and blocking of verb movement by the presence of an overt complementizer in C in subordinate clauses. Larsson and Johannessen (2015a, 2015b) discuss the facts of the subordinate word order in heritage Norwegian, and after discarding explanations of transfer from English, and overgeneralizations of V2, conclude that they must be due to incomplete acquisition. These authors find that other facts support their conclusion in that it shows that the subordinate clause word order is acquired late in childhood, and that the relevant generation of heritage language speakers was disrupted in their learning of Norwegian by the massive and dominant influx of English input during the early formative years of linguistic development when they should have acquired the subordinate clause word order. Interestingly, these results contrast with Haº kansson’s (1995) work on heritage Swedish affected by prolonged contact with English. German also displays a difference between main clauses and subordinate clauses. It is known as “asymmetric V2”: While the finite verb is found in the second position in declarative matrix clauses, it is found as the rightmost element in subordinate clauses (V(erb)-last). Perhaps surprisingly, the asymmetric word order commonly found in continental-European varieties of German is largely maintained in both the spoken and written registers of these heritage varieties of German. In particular, 1

In some Northern Norwegian dialects the facts are different, as are clauses embedded under the complementizer at ‘that’ (see Larsson and Johannessen 2015a, 2015b).

793

794

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

Pennsylvania Dutch, which has co-existed with English for three centuries, has retained this distinction (Louden 1988, 2006; Fuller 1997; Gross 2005; Fitch 2011; Stolberg 2014). The same trend holds in most other intragenerational language contact situations; for example, Clyne (2003) and Clyne and Cain (2000) report that their data set, comprised of 39 secondgeneration Dutch-German-English trilinguals in Australia, only produced an overgeneralized SVO-ordering in subordinate clauses in approximately 5% of all observed cases (45/846) (Clyne 2003: 133). In a recent study, Hopp and Putnam (2015) investigated the status of asymmetric V2 in a moribund variety of heritage German spoken in South Central Kansas, Moundridge Schweitzer German (MSG). MSG shares many affinities with Pennsylvania Dutch, since both have Palatinate heritage. Hopp and Putnam analyzed natural discourse and carried out an acceptability judgment task (AJT) to test the robustness of the V-last ordering in subordinate clauses. As one might expect, there is great variation regarding the word order of subordinate clauses in MSG (see Figure 33.3 below). As reported by Hopp and Putnam (2015: 194), “Of the 101 subordinate clauses, 67 showed ambiguous or verb-final word order, which is compatible with (standard) German word order. [. . .] In particular for syntagms such as [Aux/Modal + V[fin]] with dass-clauses, there existed a substantial amount of V2-ordering and a relatively minor amount of SVO.”2 Hopp and Putnam broke down the subordinate clauses further according to which complementizer headed the clauses; i.e., finite declarative

90% 88% 86% 84% V2

82%

V-last 80% 78% 76% 74% dass (n = 17)

weil (n = 9)

wenn (n = 25)

wo (n = 25)

Figure 33.3 Frequency of the two word orders in the production of subordinate clauses in MSG according to clause type

2

As a point of clarification, Hopp and Putnam (2015) tested the difference between V2 and SVO with the inclusion of a direct object in order to determine the placement of the past participle in relation to the rightmost edge of the subordinate clause.

Heritage Germanic Languages in North America

clauses introduced by dass ‘that’ (n=17), clauses introduced by weil ‘because’ (n=9), clauses introduced by temporal complementizers such as wenn ‘when’ (n=25), relative clauses (n=25), and other types of subordinate clauses (n=25). In their production data, the MSG grammar prefers V2-orderings in subordinate clauses in those beginning with weil ‘because’ (8/9 = 89%) or dass ‘that’ (15/17 = 88%); however, the opposite preference (i.e., for V-last ordering) holds for relative clauses (19/25 = 79%) and those that are headed by the temporal complementizer wenn ‘when’ (22/25 = 88%). The examples in (8) below represent examples from natural speech (from Hopp and Putnam 2015: 195–6): (8) a. dass-clause: . . . dass da lieber Gotthot uns auch net alles

genomm wie dat

that the dear God has us also not everything taken

like there

‘that the dear God hasn’t taken everything away from us like in Oklahoma. in Oklahoma in Oklahoma” (dass-Subj-Aux-Neg-IO-DO-PPart-PP / V2)

(MSG Participant 102)

(target: dass . . .. alles genomm hat)

b.

c.

d.

weil-clause: . . . weil ich duh nett Hochdeutsch redde. because I do/can not High German talk ‘because I can’t speak standard German.’ (weil-Subj-Aux-DO-V / V2) (target: weil . . . redde duh)

(MSG Participant 103)

wenn-clause: . . . wenn mir erscht geheirat henn. when we first married have ‘when we first got married.’ (C-Subj-PPart-Aux / V-last) (MSG Participant 102) (target) relative clause (headed by wo): . . . die wo in die Schul jetzt sin(d). those there in the school now are ‘those that are in school currently.’ (RP-PP-Aux / V-last) (MSG Participant 103) (target)

The finding that dass-clauses co-occurred predominantly with V2-ordering is quite surprising, as it is virtually unattested in other heritage varieties of

795

796

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

German.3 Hopp and Putnam’s AJT focused on the acceptability of various permutations of the order of the nonfinite verb in connection with subordinate clauses headed by dass ‘that.’ Each informant (n=48) heard 48 tokens and was asked to evaluate the well-formedeness/acceptability of each token. The scale, which ranged from 1 (totally unacceptable) to 6 (completely acceptable), revealed V2 was the preferred order, echoing the dominant pattern in the free narration. As discussed in Section 33.3, recent studies by Yager et al. (2015) and Yager (2016) reveal that it is difficult to label the reduction in case morphology exclusively as a loss of structural options. Rather, these works point towards nuanced restructurings at the syntax-semantics/ pragmatics interface. From this perspective, this restructuring process in these aforementioned communities regarding morphosyntactic case as well as the word order preferences in subordinate clauses in MSG can be viewed as a complexification effect in some ways; i.e., the preference of the nonfinite verb is determined by complementizer type rather than adopting the V2-ordering or simply retaining the V-last option.

33.4.2 Possessive Pronouns in Norwegian In European Norwegian, possessive pronouns can be either prenominal or postnominal. Both are used in all dialects, but there are dialectal differences in their distribution, and there are pragmatic factors involved. Westergaard and Anderssen (2015) compare the use of possessives in bilingual children in Norway with heritage Norwegian speakers in the USA. A first hypothesis would be that heritage Norwegian speakers use the preposed possessive more, as a transfer phenomenon, given that English possessives are preposed. It turns out, however, that the heritage speakers use the postnominal possessive construction more than in Norway. This is unexpected given that it is more complex than the prenominal possessive, and most different from English. It is also different from the speech of young children. Westergaaard and Anderssen (2015: 26, 29–30) show that both monolingual Norwegian children and bilingual Norwegian-English children in Norway prefer the (syntactically simpler) preposed possessives. There is therefore no doubt that the heritage speakers have acquired the target grammar. They have looked at various factors, and suggest that the overall frequency of the postnominal posessive protects it from change. Examples from heritage Norwegian of both types are given in (9).

3

It should be noted, however, that in the Scandinavian (North Germanic) languages, the equivalent of dass, i.e. at(t) ‘that’, introduces subordinate clauses with main clause word order almost as often as subordinate word order (Larsson & Johannessen 2015b: 250).

Heritage Germanic Languages in North America

(9)

a. schoolhouse’n din school.house.D E F your ‘your schoolhouse.’ b. min bestmor my grandmother ‘my grandmother’ (Westergaard and Anderssen 2015: 35–36)

In Anderssen and Westergaard (2017) possessive structures are compared with the noun phrase structure known as double or composite definiteness, which is the Norwegian phenomenon of “double definiteness” requiring an extraposed determiner whenever a definite noun is modified, also discussed in Johannessen (2015b). In European Norwegian, while definiteness is expressed as a suffix on noun phrases consisting of a nonmodified noun, (10a), an additional preposed definite determiner is required when the noun is modified by, for example, an adjective. (10)

a. flagg-et flag-N . S G . D E F ‘the flag’ b. det norske the.D E F . S G . N Norwegian.D E F ‘the Norwegian flag’

flagget flag-N . S G . D E F

A lot of variation is found with respect to this construction in heritage Norwegian; (11a) illustrates a nontarget phrase where the preposed determiner is missing, while (11b) illustrates a missing definiteness suffix on the noun. (11)

a. *norske flagget (target: det norske flagget) Norwegian.D E F flag.S G . D E F ‘The Norwegian flag’ b. *den store building (target: den store buildingen) the.D E F . big. building. SG.M DEF SG ‘The big building’ (Johannessen 2015b: 58)

However, it is a fact that heritage speakers differ with respect to their language production. Anderssen and Westergaard (2017) investigate whether speakers can be divided into three different groups, following Kupisch (2014), using the position of the possessive as a diagnostic: (i) one group that has the target-like distribution; (ii) one that shows crosslinguistic influence, i.e., has a predominantly English-like distribution (with preposed possessives) and (iii) one that has cross-linguistic hypercorrection, i.e. shows a predominantly Norwegian distribution, with

797

798

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

postposed possessives. The distribution of double definiteness constructions is expressed differently among three groups. Regarding the group with POSS–N as an English group, and the N–POSS as a Norwegian group, they find that the English group indeed produces significantly more structures in which the suffix is dropped, in contrast to the Norwegian and the Middle groups. They take this to indicate that the English group has crosslinguistic influence and the others have cross-linguistic hypercorrection.

33.5 Semantics and Pragmatics An obvious place in which to find semantic change is in the lexicon. Changes in the lexicon from the baseline language to the heritage language are very noticeable to lay speakers as well as to linguists, and have been commented upon and studied by a wide range of people, from educated nonlinguists to linguists, for more than a 150 years, from the pastor Duus (1855–1858 [1958]), to Flom (1901, 1926), Flaten (1901), Haugen (1953), Hjelde (1992), to the twenty-first century: Johannessen and Laake (2012a, 2017), Åfarli et al. (2013), Grimstad et al. (2014), Annear and Speth (2015), Eide and Hjelde (2015), Hjelde and Johannessen (2017), Riksem (2018). It is interesting that many of the loanwords mentioned in these works have been stable for more than a hundred years. For example, the loanwords in (12) (replacing Norwegian words) are all documented in Haugen (1953), and thus existed in his recordings from the 1930s and 1940s. (12)

Heritage Norwegian farm figgera ut filda grævel grandchild job kidsa krikken

Modern Norwegian gaº rd funnet ut jordet grus barnebarn jobb/arbeid barna bekken

English farm figured out the field gravel grandchild job the kids the creek

Johannessen and Laake 2017: 14)

An additional domain at the syntax-semantics/pragmatics interface that may be susceptible to attrition and restructuring is anaphoric and logophoric binding. Previous research has demonstrated that binding relations in heritage L1 grammars can be altered (see, e.g., Cole et al. 1990; Gu¨rel 2002, 2004, 2007; Kim et al. 2009). Putnam and Arnbjo¨rnsdo´ttir (2015) show that heritage Icelandic as spoken in Eastern Manitoba, Canada, no longer licenses long-distance binding. Long-distance binding is possible in European Icelandic under certain conditions, but is never possible in

Heritage Germanic Languages in North America

English. Compare (13a) with (13b) (data from Putnam and Arnbjo¨rnsdo´ttir 2015: 208 and Thra´insson 2007: 466): (13)

a. *Greg1 told Marsha [that she saw himself1 in the mirror]. b. Jo´n1 heldur [að e´g hafi logið að se´r1]. John thinks that I have.S B J T lied to ANPH.DAT ‘John thinks that I lied to him.’

Example (13a) confirms that anaphoric binding in English cannot cross a clausal boundary. In contrast, long-distance binding is possible in European Icelandic as long as an additional condition is met; namely, the verb in the subordinate clause must exhibit subjunctive mood marking, otherwise long-distance binding is not possible. In addition to the loss of long-distance binding in heritage Icelandic, Arnbjo¨rnsdo´ttir (2006) and Putnam and Arnbjo¨rnsdo´ttir (2015) noted the additional loss of verbal inflections for subjunctive mood. The remaining heritage Icelandic speakers often include the verb mundi ‘would’ to mark subjunctive mood in place of the various allomorphs (Note: the subjunctive mood from standard Icelandic appears below in [brackets]): (14) a. Hann skrifaða þeim að hann mundi [yrði] ekki vera he wrote them that he would not be kallaður ı´ herinn. called into the army ‘He wrote to them that he would not be called into the army.’ b. E´g mundi [vildi] ekki vanta að vera. I would not want to be ‘I would not want to be.’ According to Putnam and Arnbjo¨rnsdo´ttir (2015), the loss of long-distance anaphoric binding is accompanied by the loss of inflectional categories to mark subjunctive mood in heritage Icelandic. Here we witness a reduction of grammatical complexity in multiple domains of grammar (i.e., morphosyntax, syntax, and semantics/pragmatics) that eliminates the possibility of long-distance binding relations.

33.6 Challenges and Controversies This overview of the acquisition and further development of Germanic heritage languages and their effect on the grammar systems of these languages is the focus of ongoing research from multiple perspectives. Here we note some of the challenges and controversies surrounding the general research program of these heritage grammars. A particular challenge unique to these communities of speakers is that they are almost exclusively moribund (see Johannessen and Salmons 2015, and Page and Putnam 2015 for an overview of the sociolinguistic make up

799

800

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

of these communities). This contrasts with the situation that many minority language communities currently find themselves in (see, e.g., Louden, Chapter 34), where other factors such as religious affiliation and agrarian occupations have continued to foster the need of transmitting the socially less-dominant L1 to future generations. Heritage Germanic languages are spoken by the final generation of speakers of the community who are elderly and exhibit various degrees of proficiency in the grammar with respect to both comprehension and production capabilities. Any form of literacy of the heritage language is markedly rare, and possessed usually by clergy or those in the community who were educated beyond high school. These factors often shape and dictate the nature of the designed experiments beyond discourse and targeted narration tasks. In this vein, we see that the classification of grammars with the label “heritage” is heterodox, including research on multi-generational communities in which the “heritage speakers” consist of the youngest generation of speakers of the heritage L1 who still have contact with the eldest generation of speakers (see e.g., Bayram et al. 2019 for an overview of this state of affairs). The significant contrast in these “heritage language communities” poses difficulties for direct comparisons with some groups within this intra-heritage distinction. Future research in this area must take this into consideration. The fact that the vast majority of remaining heritage Germanic speakers consists of elderly and moribund speakers means they can make important contributions to research on language, in particular bilinguals, in aging populations. Research directly focusing on cognitive aging and bilingualism has become an important and relevant topic, especially given the improvement of life expectancy (see, e.g., Keijzer 2007, 2010; Bialystok 2016; Blumenfeld et al. 2016; Ivanova et al. 2016; Keijzer and Schmid 2016; Rossi and Diaz 2016). Data from these communities will figure prominently in this growing body of research. Finally, we come to the intensely debated nature of the acquisition and maintenance of heritage grammars across the lifespan. Seminal proposals by Polinsky (1997, 2006) and Montrul (2008) advanced the proposal that those acquiring a heritage grammar receiving limited or insufficient input in the period of their formative linguistic development may not have ultimately attained full proficiency in the heritage grammar, resulting in an incompletely acquired grammar. This contrasts with language attrition, which is viewed in this system as the difficulty in retrieving aspects of a fully-acquired grammar. This interpretation of heritage language acquisition and its further development has been challenged by some scholars who argue that heritage language acquisition is on equal footing with its monolingual counterpart but under different pressures due to conflicting source inputs (see, e.g., Rothman 2007, Pires and Rothman 2009, Pascual y Cabo and Rothman 2012, Putnam and Sa´nchez 2013, Perez-Cortes 2016,

Heritage Germanic Languages in North America

Putnam et al. 2018). Even the term incomplete acquisition has come under fire (e.g., Kupisch and Rothman 2018) in this ongoing debate. Moving forward, one of the daunting challenges that faces researchers involved in the study of these heritage grammars is arriving at a more nuanced and detailed understanding of the differences between loss and attrition versus the restructuring of grammatical systems (if one can be ascertained at all). A second equally arduous task will be to determine whether or not one can reliably measure ultimate attainment of a grammar system; for example, see Montrul’s (2015) discussion of incomplete mastery of grammatical forms. Lastly, this research program should expand to include passive/recipient bilingual populations from these communities to come to a more comprehensive understanding of the possible asymmetries of the atrophy of production and comprehension of these heritage grammars under investigation (e.g., Peres-Cortez 2016).

References Åfarli, T. A. 2012. “Hybride verbformer i amerikanorsk og analyse av tempus,” Norsk Lingvistisk Tidsskrift: 381–398. Aissen, J. 2003. “Differential object marking: iconicity versus economy,” Natural Language and Linguistic Theory 21: 438–485. Anderssen, M. and M. Westergaard 2017. Crosslinguistic Similarities and Differences in Bilingual Acquisition and Attrition: Possessives and Double Definiteness in Norwegian Heritage Language. Ms., University of Tromsø. Annear, L. and K. Speth 2015. “Maintaining a multilingual repertoire: Lexical change in American Norwegian.” In J. B. Johannessen and J. Salmons (eds.), Germanic Heritage Languages in North America: Acquisition, Attrition and Change. Amsterdam: John Benjamins: 201–216. Arnbjo¨rnsdo´ttir, B. 2006. North American Icelandic: The Life of a Language. Winnipeg: University of Manitoba Press. Bayram, F., D. Pascual y Cabo, and J. Rothman 2019. “Cross-generational attrition contributions to heritage speaker competence.” In B. Ko¨pke and M. S. Schmid (eds.), The Oxford Handbook of Language Attrition. Oxford University Press: 446–457. Berend, N. and H. Jedig (eds.) 1991. Deutsche Mundarten in der Sowjetunion: Geschichte der Forschung und Bibliographie. Marburg: N. G. Elwert Verlag. Bialystok, E. 2016. “Aging and bilingualism: Why does it matter?” Linguistic Approaches to Bilingualism 6.1/2: 1–8. Boas, H. 2009. The Death of Texas German. Durham, NC: Duke University Press. Bossong, G. 1985. Empirische Universalienforschung: Differentielle Objektmarkierung in der Neuiranischen Sprachen. Tu¨bingen: Narr. Bossong, G. 1991. “Differential object marking in Romance and beyond.” In D. Wanner and D. Kibbee (eds.), New Analyses in Romance Linguistics. Amsterdam: John Benjamins: 143–167.

801

802

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

Blumenfeld, H. K., S. R. Schroeder, S. C. Bobb, M. R. Freeman, and V. Marian 2016. “Auditory word recognition across the lifespan: Links between linguistic and nonlinguistic inhibitory control in bilinguals and monolinguals,” Linguistic Approaches to Bilingualism 6.1/2: 119–146. Clyne, M. 2003. Dynamics of Language Contact: English and Immigrant Languages. Cambridge University Press. Clyne, M. and H. Cain 2000. “Trilingualism in related languages – DutchGerman-English near and far.” In G. Hirschfelder, D. Schnell, and A. Schrutka-Rechtenstamm (eds.), Kulturen – Sprachen – U¨berga¨nge. Festschrift fu¨r H.L. Cox zum 65. Geburtstag. Cologne: Boehlau: 135–152. Cole, P., G. Hermon, and L.-M. Sung 1990. “Principles and parameters of long-distance reflexives,” Linguistic Inquiry 21: 1–22. Cook, V. 1997. “Monolingual bias in second language acquisition research,” Revista Canaria de Estudios Ingleses 34: 35–50. Duus, O. F. 1855–1858 [1958]. “Frontier Parsonage: The letters of Olaus Fredrik Duus. Norwegian Pastor in Wisconsin, 1855–1858. NAHA, Northfield 1947.” In I. Blegen and C. Theodore (eds.), Amerikabrev. Oslo: H. Aschehoug & Co. Eide, K. M. and H. Arnstein 2015. “Verb second and finiteness morphology in Norwegian heritage language of the American Midwest.” In B. R. Page and M.T. Putnam (eds.), Leiden: Brill: 64–101. Fitch, G. W. 2011. “Changes in frequency as a measure of language change: Extraposition in Pennsylvania German.” In M. T. Putnam (ed.), Studies on German-Language Islands. Amsterdam: John Benjamins: 371–384. Flaten, Ns. 1901. “Notes on American-Norwegian with a Vocabulary.” Dialect Notes, Vol. 2. Flom, G. T. 1901. “English Elements in Norse Dialects of Utica, Wisconsin.” Dialect Notes, Vol. 2, 1900–1904. Flom, G. T. 1926. “English loanwords in American Norwegian. As spoken in the Koshkonong settlement, Wisconsin,” American Speech 1: 541–558. Fuller, J. 1997. Pennsylvania Dutch with a Southern Touch: A Theoretical Model of Language Contact and Change. Ph.D. dissertation, University of South Carolina-Columbia. Geiger, S. and J. Salmons 2006. “Reconstructing variation at shallow time depths: The historical phonetics of 19th century German dialects in the U.S.” In T. Cravens (ed.), Variation and Reconstruction. Amsterdam: John Benjamins: 37–58. Grimstad, M. B., T. Lohndal, and T. A. Åfarli 2014. “Language mixing and exoskeletal theory: A case study of word-internal mixing in American Norwegian,” Nordlyd 41.2: 213–237. Gross, S. 2005. “Attrition and typological change.” Presentation, 2nd International Conference on First Language Attrition. Vrije Universitet Amsterdam. Gu¨rel, A. 2002. Linguistic Characteristics of Second Language Acquisition and First Language Attrition: Turkish overt versus null pronouns. Ph.D. dissertation, McGill University.

Heritage Germanic Languages in North America

Gu¨rel, A. 2004. “Selectivity in L2-induced attrition: A psycholinguistic account,” Journal of Neurolinguistics 17: 53–78. Gu¨rel, A. 2007. “(Psycho)linguistic determinants of L1 attrition.” In B. Ko¨pke, M. Schmid, M. Keijzer, and S. Dosterst (eds.), Language Attrition: Theoretical Perspectives. Amsterdam: John Benjamins: 99–120. Haugen, E. 1953. The Norwegian Language in America: A Study in Bilingual Behavior. Philadelphia: University of Pennsylvania Press. º Hakansson, G. 1995. “Syntax and morphology in language attrition: A study of five bilingual expatriate Swedes,” International Journal of Applied Linguistics 5: 153–171. Hjelde, A. 1992. Trøndsk talemaº l i Amerika. Trondheim: Tapir. Hjelde, A. 2015. “Changes in a Norwegian dialect in America.” In Johannessen and Salmons (eds.): 283–298. Hjelde, A. and J. B. Johannessen 2017. “Amerikanorsk: Orda vitner om kontakt mellom folk.” In T. M. Hasle Joranger and H. T. Cleven (eds.), Norwegian-American Essays 2017. Oslo: Novus: 257–282. Hopp, H. and M. T. Putnam 2015. “Restructuring in heritage grammar: Word order variation in heritage German,” Linguistic Approaches to Bilingualism 5.2: 180–203. Huffines, M. L. 1989. “Case usages among the Pennsylvania German sectarians and nonsectarians.” In N. C. Dorian (ed.), Investigating Obsolescence: Studies in Language Contraction and Death. Cambridge University Press: 211–226. Ivanova, I., M. Murillo, R. I. Montoya, and T. H. Gollan 2016. “Does bilingual language control decline in older age?” Linguistic Approaches to Bilingualism 6.1/2: 86–118. Jakobson, R. 1941. Kindersprache, Aphasie, und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell. Johannessen, J. B. 2015a. “The Corpus of American Norwegian Speech (CANS).” In B. Megyesi (ed.), Proceedings of the 20th Nordic Conference of Computational Linguistics, NODALIDA 2015, May 11–13, 2015, Vilnius, Lithuania. NEALT Proceedings Series 23. Johannessen, J. B. 2015b. “Attrition in an American Norwegian heritage language speaker.” In J. B. Johannessen and J. C. Salmons (eds.), Germanic Heritage Languages in North America: Acquisition, Attrition and Change. Amsterdam: John Benjamins: 46–71. Johannessen, J. B. 2018. “Factors of variation, maintenance and change in heritage languages,” International Journal of Bilingualism 22.4: 447–465. Johannessen, L. J. B. and I. Larsson 2015. “Complexity matters: On gender agreement in heritage Scandinavian,” Frontiers in Psychology, https://doi .org/10.3389/fpsyg.2015.01842. Johannessen, J. B. and J. C. Salmons (eds.) 2015. Germanic Heritage Languages in North America: Acquisition, Attrition and Change. Amsterdam: John Benjamins. Johnson, C. D. 1994. The Volga-German Dialect of Schoenchen, Kansas. Ph. D. dissertation, University of Kansas.

803

804

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

Judy, T., M. T. Putnam, and J. Rothman 2018. “When bilingualism is the common factor: Switch reference at the junction of competence and performance in both Second Language and Heritage Language performance,” Journal of Language Contact 11.3: 592–618. Keel, W. D. 1994. “Reduction and loss of case marking in the noun phrase in German-American speech islands: Internal development or external influence?” In N. Berend and K. Mattheier (eds.), Sprachinselforschung: eine Gedenkschrift fu¨r Hugo Jedig. Frankfurt am Main: Peter Lang Verlag: 93—104. Keel, W. D. 2015. “Noun phrase case shift in Volga German varieties on the Great Plains of Kansas.” In B. R. Page and M. T. Putnam (eds.), Moribund Germanic Heritage Languages in North America: Theoretical Perspectives and Empirical Findings. Leiden: Brill: 133–153. Keijzer, M. C. J. 2007. First In Last Out? An Investigation of the Regression Hypothesis in Dutch Emigrants in Anglophone Canada. Ph.D. dissertation, University of Utrecht. Keijzer, M. C. J. 2010. First Language Acquisition and First Language Attrition: Parallels and Divergences. Saarbrucken: Lambert Academic Publishers. Keijzer, M.C. J. and M. S. Schmid 2016. “Individual differences in cognitive control advantages of late Dutch-English bilinguals,” Linguistic Approaches to Bilingualism 6.1/2: 64–85. Khramova, M. 2011. The Volga German Dialect of Milberger, Kansas. Ph. D. dissertation, University of Kansas. Kim, J-H., S. Montrul, and J. Yoon 2009. “Binding interpretation of anaphors by Korean heritage speakers,” Language Acquisition 16: 3–35. Kupisch, T. 2014. “Adjective placement in simultaneous bilinguals (German-Italian) and the concept of cross-linguistic overcorrection,” Bilingualism: Language and Cognition 17: 222–233. Kupisch, T. and J. Rothman 2018. “Terminology matters! Why difference is not incompleteness and how early child bilinguals are heritage speakers,” International Journal of Bilingualism 22.5: 564–582. Larsson, I. and J. B. Johannessen 2015a. “Incomplete acquisition and verb placement in heritage Scandinavian.” In B. R. Page and M. T. Putnam (eds.), Moribund Germanic Heritage Languages in North America: Theoretical Perspectives and Empirical Findings. Leiden: Brill: 153–189. Larsson, I. and J. B. Johannessen 2015b. “Embedded word order in Heritage ¨ stman, C. Mertzlufft, and M. Rießler Scandinavian.” In M. Hilpert, J-O. O (eds.), New Trends in Nordic and General Linguistics. Berlin: Mouton de Gruyter: 239–266. Lohndal, T. and M. Westergaard 2016. “Grammatical gender in American Norwegian Heritage Language: Stability or attrition?” Frontiers in Psycholology, 7: 344. Louden, M. L. 1988. Bilingualism and Syntactic Change in Pennsylvania German. Ph.D. dissertation, Cornell University.

Heritage Germanic Languages in North America

Louden, M. L. 2006. “Pennsylvania German in the twenty-first century.” In N. Berend and E. Knipf-Komlo´si (eds.), Sprachinsel weltweit – The world of language islands. Berlin: Peter Lang Verlag: 89–107. Louden, M. L. 2016. Pennsylvania Dutch: The Story of an American Language. Baltimore: Johns Hopkins University Press. Mironow, S. A. 1957. “Zur vergleichenden Formenlehre der deutschen Mundarten,” Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 79 (supplemental volume): 388–414. Montrul, S. 2008. Incomplete Acquisition in Bilingualism: Re-examining the Age Factor. Amsterdam: John Benjamins. Montrul, S. 2015. The Acquisition of Heritage Languages. Cambridge University Press. Øverland, O. and S. Kjærheim (eds.) 1992. Fra Amerika til Norge. I norske utvandrerbrev 1838–57. Bind 1. Oslo: Solum. Page, B. R. and M. T. Putnam (eds.) 2015. Moribund Germanic Heritage Languages in North America: Theoretical Perspectives and Empirical Findings. Leiden: Brill. Pascual y Cabo, D. and J. Rothman. 2012. “The (il)logical problem of heritage speaker bilingualism and incomplete acquisition,” Applied Linguistics 33: 450–455. Perez-Cortes, S. 2016. Acquiring Obligatory and Variable Mood Selection: Spanish Heritage Speakers and L2 Learners’ Performance in Desideratives and Reported Speech Contexts. Ph.D. dissertation, Rutgers University. Pires, A. and J. Rothman 2009. “Disentangling sources of incomplete acquisition: An explanation for competence divergence across heritage grammars,” International Journal of Bilingualism 13: 211–238. Polinsky, M. 2006. “Incomplete acquisition: American Russian,” Journal of Slavic Linguistics 14: 191–262. Polinsky, M. 2011. “Reanalysis in adult heritage language,” Studies in Second Language Acquisition 33: 305–328. Putnam, M. T. (ed.). 2011. Studies on German-Language Islands. Amsterdam: John Benjamins. Putnam, M. T. and L. Sa´nchez 2013. “What’s so incomplete about incomplete acquisition? A prolegomenon to modeling heritage language grammars,” Linguistic Approaches to Bilingualism 3.4: 478–508. Putnam, M. T. and B. Arnbjo¨rnsdo´ttir 2015. “Minimizing (interface) domains: The loss of long-distance binding in North American Icelandic.” In B. R. Page and M. T. Putnam (eds.), Moribund Germanic Heritage Languages in North America: Theoretical Perspectives and Empirical Findings. Leiden: Brill: 203–223. Putnam, M. T., T. Kupisch, and D. Pascual y Cabo 2018. “Different situations, similar outcomes: Heritage grammars across the lifespan.” In F. Bayram, N. Denhovska, D. Miller, J. Rothman, and L. Serratrice (eds.), Bilingualism: The State of the Science Across Its Subfields. Amsterdam: John Benjamins: 251–279. Riksem, B. R. 2018. “Language mixing in American Norwegian noun phrases,” Journal of Language Contact 11.3: 481–524.

805

806

J A N N E B O N D I J O H A N N E S S E N A N D M I C H A E L T. P U T N A M

Rødvand, L-I. S. 2017. Empirical Investigations of Grammatical Gender in American Heritage Norwegian. M.A. thesis, University of Oslo: Department of Linguistics and Scandinavian Studies. Rosenberg, P. 2003. “Comparative speech island research: Some results from studies in Russia and Brazil.” In W. D. Keel and K. J. Mattheier (eds.), Deutsche Sprachinseln weltweit. Interne und externe Perspektiven/German language varieties worldwide. Internal and external perspectives. Frankfurt am Main: Peter Lang Verlag: 199–238. Rossi, E. and M. Diaz 2016. “How aging and bilingualism influence language processing,” Linguistic Approaches to Bilingualism 6.1/2: 9–42. Rothman, J. 2007. “Heritage speaker competence differences, language change and input type: Inflected infinitives in heritage Brazilian Portuguese,” International Journal of Bilingualism 11: 359–389. Shirer, M. 1965. “Case systems in German dialects,” Language 413: 420–438. Stolberg, D. 2014. Changes Between the Lines: Diachronic Contact Phenomena in Written Pennsylvania German (Studia Linguistica Germanica). Berlin: Mouton de Gruyter. Taranrød, B. 2011. Leddstillingen i relativsetninger i amerikansknorsk. M.A. thesis. University of Oslo: Department of Linguistics and Scandinavian Studies. Thra´insson, H. 2007. The Syntax of Icelandic. Cambridge University Press. Vikner, S. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford University Press. Westergaard, M. and M. Anderssen 2015. “Word order variation in Norwegian possessive constructions: Bilingual acquisition and attrition.” In J. B. Johannessen and J. C. Salmons (eds.): 21–45. Yager, L. 2016. Morphosyntactic Variation and Change in Wisconsin Heritage German. Ph.D. dissertation, University of Wisconsin-Madison. Yager, L., N. Hellmold, H-A. Joo, M. T. Putnam, E. Rossi, C. Stafford, and J. Salmons 2015. “New structural patterns in moribund grammar: Case marking in heritage German,” Frontiers in Psychology 6: 1716.

Chapter 34 Minority Germanic Languages Mark L. Louden

34.1 Introduction 34.1.1 Identifying Minority Germanic Languages This chapter examines the sociolinguistic situation of Germanic languages that are spoken by a minority of residents of a given nation. While the definitions of the descriptors “minority” and “Germanic” are uncontroversial, the problem of distinguishing between “languages” and “dialects” is a familiar one. It is well-known that there are no absolute scientific criteria according to which linguistic systems may be labeled languages or dialects, rather it is the external, often political, situation of speaker groups that determines whether varieties sharing a common linguistic ancestor are sufficiently autonomous from one another as to be viewed as distinct languages or not. The language-dialect question is a relevant one in this chapter. The standard reference work on the documentation of linguistic diversity worldwide, Ethnologue (Simons and Fennig 2017), lists 47 languages as belonging to the Germanic branch of the Indo-European family, 6 of which are part of the North subgroup; the remaining 41 are in the West subgroup. These are listed below. Germanic Languages Listed in Ethnologue Underscored varieties are regarded as autonomous languages for the purpose of this chapter. The numbers in parentheses, which are given by Ethnologue, indicate the number of languages in a particular group. (https://www.ethnologue.com/subgroups/germanic) North (6) East Scandinavian (4) ¨ vdalian (Elfdalian, Dalecarlian) O Danish-Swedish (3)

808

MAR K L. LOU DEN

Danish-Bokmaº l (1) Norwegian Danish-Riksmaº l (1) Danish (1) Danish Swedish (1) Swedish West Scandinavian (2) Faroese Icelandic West (41) English (2) English Scots Frisian (3) Frisian Frisian, Northern Saterfriesisch High German (20) German (18) Hunsrik Middle German (8) East Middle German (4) German, Standard Saxon, Upper Silesian, Lower Wymysorys West Middle German (4) German, Pennsylvania Palatinate Franconian Ripuarian Moselle Franconian (1) Luxembourgish

Minority Germanic Languages

Upper German (9) Eastern Franconian Alemannic (4) German, Colonia Tovar German, Swiss Swabian Walser Bavarian-Austrian (4) Bavarian Cimbrian Hutterisch Mo´cheno Yiddish (2) Yiddish, Eastern Yiddish, Western Low Saxon-Low Franconian (16) Low Franconian (5) Afrikaans Dutch Limburgish Vlaams Zeeuws Low Saxon (11) Achterhoeks Drents Gronings Plautdietsch Sallands Saxon, East Frisian Low Saxon, Low Stellingwerfs Twents Veluws Westphali[a]n There appears to be little if any disagreement that Danish, O¨vdalian (also Elfdalian, an endangered language spoken in Sweden), Faroese, Icelandic, Norwegian, and Swedish should be regarded as languages. However, the status of most of the 41 West Germanic languages identified as such by Ethnologue is less clear. Just seven of these varieties are recognized as languages without debate either administratively or by

809

810

MAR K L. LOU DEN

linguists: Afrikaans, Dutch, English, Frisian, (standard) German, Luxembourgish, and Yiddish. Three more varieties, Limburgish, Low Saxon (Low German), and Scots, are less readily identified as languages in their own right, but for reasons discussed below are included in this chapter. There are an additional four extraterritorial West Germanic varieties descended historically from German regional dialects that have enjoyed little official recognition but on structural and sociolinguistic grounds could well be argued to be as autonomous from German as, say, Yiddish is, namely Hutterisch, Pennsylvania German (henceforth, Pennsylvania Dutch), Plautdietsch (Mennonite Low German), and Wymysorys. As is the case with Yiddish, the speakers of these Germanrelated varieties have for generations felt little or no sociolinguistic connection to German-speaking Central Europe. Three of these languages, Hutterisch, Pennsylvania Dutch, and Plautdietsch, are spoken by rapidly growing conservative Anabaptist Christian groups and are thus in a robust state of health. The fourth, Wymysorys, is a nearly extinct German-related language spoken in a single town in southern Poland. A few remarks about Frisian and Yiddish are in order. Ethnologue divides each into sublanguages, namely Frisian, North Frisian, and Saterfriesisch; and Eastern and Western Yiddish, respectively. These subdivisions could be justified on linguistic, historical, and geographic grounds, as the varieties in question are structurally distinct from one another and their speakers have occupied noncontiguous regions for centuries. Nonetheless, I will follow the practice of linguists with special expertise in these languages and subsume the three and two varieties under Frisian and Yiddish, respectively. Four additional German-related varieties, Cimbrian, Colonia Tovar German, Hunsrik, and Mo´cheno, are spoken in non–German-speaking countries, namely in Italy (Cimbrian and Mo´cheno), Venezuela (Colonia Tovar German), and Brazil (Hunsrik). All are endangered to varying degrees. As extraterritorial varieties spoken by people whose ancestors left German-speaking Europe generations ago (in the case of Cimbrian, over a millennium ago), one could argue that they are as autonomous from German as Hutterisch, Pennsylvania Dutch, and Plautdietsch, hence they are included in this chapter. However, their recognition in this chapter as minority Germanic languages raises the valid question whether other socalled colonial or heritage dialects of German – and also of Dutch, Frisian, and Scandinavian languages – spoken in speech islands around the globe by the descendants of immigrants should not also be included. If Cimbrian is treated here as a minority Germanic language, why not also Transylvania Saxon (spoken in Romania) or Pella Dutch (from the US state of Iowa)? The sheer numbers of such Germanic varieties, which can be found on every continent save Antarctica, preclude their all being treated in a chapter of this size. We will therefore limit the discussion here to varieties selected from the Ethnologue list (including Lower Silesian)

Minority Germanic Languages

supplemented by two others that are not on the list for reasons that will be indicated below. The editors of Ethnologue recognize the complexity of according certain linguistic varieties the status of languages. They follow three major criteria for language identification under what is known as the ISO 693–3 standard (www.ethnologue.com/about/problem-language-identification). (a) Two related varieties are normally considered varieties of the same language if speakers of each variety have inherent understanding of the other variety at a functional level (that is, can understand based on knowledge of their own variety without needing to learn the other variety). (b) Where spoken intelligibility between varieties is marginal, the existence of a common literature or of a common ethnolinguistic identity with a central variety that both understand can be a strong indicator that they should nevertheless be considered varieties of the same language. (c) Where there is enough intelligibility between varieties to enable communication, the existence of well-established distinct ethnolinguistic identities can be a strong indicator that they should nevertheless be considered to be different languages. With regard to these three criteria, it seems reasonable to reconsider the language status accorded by Ethnologue to 19 West Germanic varieties, which are spoken in Germany, the Netherlands, Belgium, and Switzerland. These are: Bavarian, East Frisian Saxon, Eastern Franconian, Palatinate Franconian, Ripuarian, Swabian, Upper Saxon, and Westphalian (Germany); Achterhoeks, Drents, Gronings, Sallands, Stellingwerfs, Twents, Veluws, and Zeeuws (Netherlands); Vlaams (Belgium); and Swiss German and Walser (Switzerland). With regard to Vlaams (Flemish) and Swiss German, I am guided by the consensus of the speakers of these varieties themselves, who identify what they speak as dialects of Dutch and German, respectively, despite the considerable structural differences between Vlaams and standard Dutch and Swiss German and standard German, not to mention the sociolinguistic autonomy the speakers of these vernacular varieties enjoy as citizens of the countries of Belgium and France (for Vlaams) and Switzerland, as opposed to the Netherlands and Germany (and Austria). I mentioned above that three West Germanic varieties identified by Ethnologue as languages, Low Saxon (henceforth referred to as Low Saxon / Low German), Limburgish, and Scots, are not universally regarded as such. Referring to the Ethnologue criteria for language identification above, it is an open question whether majorities of the speakers of Low Saxon / Low German, Limburgish, and Scots have “ethnolinguistic identities” that are strongly “distinct” from those of their immediate neighbors who do not speak Low Saxon / Low German, Limburgish, and Scots. Each

811

812

MAR K L. LOU DEN

variety is endangered, with the practical consequence that the communities in which each is spoken include speakers and nonspeakers who at least share a common ethnolinguistic heritage if not a contemporary ethnolinguistic identity. It is safe to say that every active speaker of Low Saxon / Low German, Limburgish, and Scots has close relatives who speak German, Dutch, or English only. However, these three varieties are recognized as languages under the European Charter for Regional or Minority Languages (ECRML), a treaty to which most of the 47 current members of the Council of Europe (which includes all but three states in Europe, namely Belarus, Kazakhstan, and Vatican City) are signatories (https://www.coe.int/t/dg4/education/minlang/). Mainly for this reason I include Low Saxon / Low German, Limburgish, and Scots in this chapter. To summarize, of the 47 Germanic varieties identified by Ethnologue as languages, I have decided to recognize just 25 as such for the purpose of this chapter. Four of these languages, Faroese, Icelandic, Luxembourgish, and Norwegian, will not figure into the discussion below as they were spoken as minority languages only in immigrant enclaves (e.g., in North America) that are today nearly nonexistent, the descendants of the original immigrants now having almost completely shifted to a majority language (in the US and Canada, English). I also do not discuss Dutch, which aside from its use in immigrant enclaves, could perhaps be considered a minority language in northern France (French Flemish). English is also excluded from this chapter due to the focus of this volume, despite the fact that it is a minority language in multiple nations across the globe (cf. World Englishes). To the 19 remaining languages (Afrikaans, Cimbrian, Colonia Tovar German, Danish, Frisian, German, Hunsrik, Hutterisch, Limburgish, Low Saxon / Low German, Lower Silesian, Mo´cheno, ¨ vdalian, Pennsylvania Dutch, Plautdietsch, Scots, Swedish, Wymysorys, O and Yiddish) I am adding two small but sociolinguistically healthy extraterritorial West Germanic languages that are not included in Ethnologue, both of which are spoken by subgroups of the Old Order Amish in the US, namely Amish Alsatian German and Amish Swiss German.

34.1.2 Sociolinguistic Typology of Minority Languages We turn now to the question of how the 21 minority Germanic languages under discussion may be grouped together. John Roberts, a Canadian social psychologist with a special interest in multilingual societies, has advanced a descriptive typology for the classification of minority languages according to external criteria (e.g., Edwards 1992, 2004) that derives from the work of a British geographer, Paul White (White 1987). According to this typology, minority languages may be classified according to three sets of geographic criteria. The first set of criteria has to do a language’s minority status across one or more political states. If a minority language is spoken in a single state, it

Minority Germanic Languages

¨ vdalian, which is is unique. An example of a unique minority language is O spoken exclusively in Sweden. If a language is spoken in multiple states but is in all of them a minority language, it is nonunique, as, for example, Frisian vis-a`-vis the Netherlands and Germany. Finally, if a language is a minority language in one country but spoken by a majority population in another, it is classified as local-only. German is an example of a local-only language, as it is a minority language in Belgium and Denmark, but a majority language elsewhere (e.g., Germany). The second set of criteria in the White/Edwards typology classifies nonunique or local-only minority languages as either adjoining or nonadjoining if the regions of the states in which they are spoken are adjacent or not. Limburgish, which extends over the southeastern Netherlands, northeastern Belgium, and western Germany, is an adjoining minority language. Afrikaans, which is spoken across noncontiguous parts of South Africa and Namibia, is nonadjoining. The final category refers to the “spatial cohesion” of the speakers of a given language. If a minority language is spoken by people who occupy a geographic space that is cohesive it is designated as such. Spatial cohesion should not be understood as strictly binary but as a matter of degree. For example, Wymysorys, the Germanic language of a single village in Poland, ¨ vdalian is spoken in is a highly spatially cohesive minority language. O multiple discrete communities in central Sweden; however, all are located ¨ vdalian to the in a single county, Dalarna, hence my decision to assign O category of cohesive minority languages. A clearly noncohesive minority language is Scots, which is spoken today in noncontiguous regions of Great Britain (Scotland and Northern Ireland). Since the adjoining/nonadjoining dichotomy is by definition not applicable to unique minority languages, the total number of possible sociolinguistic types is ten. A further distinction between indigenous (autochthonous) and immigrant varieties doubles the number of logically possible minority language types. The 21 minority Germanic languages under discussion here fall into 10 of the 20 possible types under the White/ Edwards model, as shown in Table 34.1. Examples for the ten other types not represented by minority Germanic languages, some of which are taken from Edwards 2004, are given in italics. The countries where various languages are spoken are indicated in parentheses by their respective ISO Alpha-2 codes. Not all of the assignments given here are clear-cut. For example, Edwards treats Romani as a language indigenous to Europe, even though it traces its origins to northwest India. Other ambiguous classifications will be discussed below. In the remainder of this chapter, we present brief overviews of the 21 minority Germanic languages with an emphasis on basic sociolinguistic information justifying their assignment into the ten White/Edwards categories above. We will also consider the relative health of each minority language in the regions where they are spoken, that is, the degree to which

813

814

MAR K L. LOU DEN

Table 34.1 Classification of minority Germanic languages minority language type 1. unique, cohesive 2. unique, noncohesive

3. nonunique, adjoining, cohesive 4. nonunique, adjoining, noncohesive 5. nonunique, nonadjoining, cohesive 6. nonunique, nonadjoining, noncohesive

7. local-only, adjoining, cohesive 8. local-only, adjoining, noncohesive

9. local-only, nonadjoining, cohesive 10. local-only, nonadjoining, noncohesive

indigenous

immigrant

Lower Silesian (PL) Övdalian (SE) Cimbrian (IT) Mócheno (IT) Scots (GB) Limburgish (NE, BE, DE) Low Saxon / Low German (NE, DE) Sami (FI, NO, SE, RU)

Colonia Tovar German (VE) Wymysorys (PL) Amish Alsatian German (US) Amish Swiss German (US)

Frisian (NE, DE)

Basque (ES, FR)

Hunsrik (BR, AR, PY) Hutterisch (CA, US) Pennsylvania Dutch (US, CA) Afrikaans (ZA, NA)

Romani (throughout Europe) Plautdietsch (KZ, CA, BO, MX, PY, DE, US, BR) Yiddish (IL, throughout Europe and North and South America) Danish (DE) Spanish in the US Southwest German (BE, DK) (vis-à-vis Mexico) Swedish (FI) German speakers in Italy living outside of enclaves (vis-à-vis Switzerland and Austria) French in the Apulia region of German (NA) Italy (vis-à-vis France) Albanian in the Mezzogiorno German speakers in North region of Italy (vis-à-vis America living outside of Albania) enclaves (vis-à-vis Germany)

they are endangered or not. Where possible, we will refer to the Atlas of the World’s Languages in Danger maintained by the United Nations Educational, Scientific, and Cultural Organization (UNESCO). The UNESCO Atlas identifies five levels of endangerment: vulnerable, definitely endangered, severely endangered, critically endangered, and extinct (www.unesco.org/languagesatlas/index.php). In the concluding section of this chapter, we will consider to what extent factors promoting minority language maintenance versus shift to a majority language are associated with the ten minority language types discussed.

34.2 Minority Germanic Languages According to Type 34.2.1

Unique, Cohesive Indigenous Languages (Lower Silesian, Övdalian) Lower Silesian refers to the indigenous varieties of German spoken historically in Lower Silesia, the western part of a region that is located today

Minority Germanic Languages

mostly in southwestern Poland, extending into the Upper Lusatia (Oberlausitz) region of eastern Germany and a small part of the northeastern Czech Republic. Prior to World War II, Silesia was politically part of Germany, thus what is considered by Ethnologue as a distinct language was until 1945 a contiguous part of the east-central dialects of German. The justification for the designation of Lower Silesian as a language rather than a German dialect group is based mainly on the fact that its speakers have for the past two to three generations been citizens of a country, Poland, which did not officially recognize German as a minority language. Speakers of Lower Silesian and other German varieties in Poland experienced considerable discrimination in the postwar era, leading to a dramatic decline in their active use today. In 1991, Germans were officially designated a national minority under the Polish constitution, however most Polish ethnic Germans reside in historical Upper Silesia and speak not a form of German but Silesian, a West Slavic language related to but still distinct from Polish (www.ethnologue.com/16/show_language/sli/). The UNESCO Atlas identifies eight endangered minority languages in Poland that include three Germanic varieties, though Lower Silesian is not one of them. The three Germanic languages are Low Saxon (Low German), Vilamovian (Wymysorys), and Yiddish, the first two of which are designated as critically endangered. (Yiddish is considered definitely endangered.) Were Lower Silesian to be included in the Atlas, it would almost certainly be identified as at least severely endangered, a direct result of the repression of ethnic Germans by the Polish state in the decades after World War II and its ongoing lack of official recognition (Lasatowicz and Weger 2008, Ammon 2015: 320–328). A second unique and cohesive indigenous Germanic minority language ¨ vdalian (Elfdalian, Dalecarlian). O ¨ vdalian is spoken in the central is O Swedish province of Dalarna and in 2007 was estimated to have approximately 2,400 speakers, of whom 1,700 were concentrated in one commu¨ vdalian a distinct North nity, A¨lvdalen (Zach 2013: 9). Linguists consider O Germanic language as it is not mutually intelligible with Swedish; however, this view is not shared by the Swedish government, which under the European Charter for Regional or Minority Languages recognizes five other minority languages, Sami, Finnish, Mea¨nkeli (a group of Finnish dialects), Romani, and Yiddish. ¨ vdalian is definitely endangered, according to UNESCO, since very few O children speak it. Its decline is not due to the outright persecution of its speakers, as was the case for Lower Silesian. Rather, the shift from ¨ vdalian to Swedish is largely a consequence of the increasing integration O ¨ vdalian speakers into the larger society. In the early twentieth century, of O most speakers were farmer folk living in relative isolation from their fellow Swedes. Between 1920 and 1950, the Dalarna region became increasingly industrialized and urbanized, leading to both out-migration ¨ vdalians and in-migration of Swedish speakers. The use of O ¨ vdalian of O

815

816

MAR K L. LOU DEN

was stigmatized, and speakers increasingly began to promote the use of ¨ vdalian enjoys a measure of overt Swedish among their children. Today, O prestige; however, since only a small fraction of children in Dalarna speak it, the long-term health of the language is in question (Zach 2013: 103–107). Also, numerous efforts in recent years to have the Swedish ¨ vdalian as a regional or minority language parliament formally recognize O have been unsuccessful. The Swedish government continues to insist that ¨ vdalian is a dialect of Swedish and not a language in its own right (Zach O 2013: 100–103).

34.2.2

Unique, Cohesive Immigrant Languages (Colonia Tovar German, Wymysorys) The first of the two unique, cohesive immigrant Germanic languages under discussion here is the product of the immigration of German dialect speakers to South America in the nineteenth century. Colonia Tovar German is descended from varieties of Lower Alemannic (southern Baden) German spoken by people who settled in the community of Colonia Tovar in the northern Venezuelan state of Aragua. Structurally, the language still strongly resembles Lower Alemannic German. Ethnologue deems the status of the language as “shifting”; it is estimated that there are approximately 1,500 speakers today (www.ethnologue.com/language/gct). Although Colonia Tovar German is not formally recognized as a minority language by the Venezuelan government and Spanish is the sole language used in Tovar schools, there are concerted efforts on the part of Tovar Germans to maintain their language, and Germanthemed tourism is the main industry in their community. Nevertheless, the shift among youth toward dominance in Spanish is clear (Da Rin 1997). Wymysorys (Vilamovian) is the most severely endangered of the minority languages discussed in this chapter. It is a Germanic language spoken today by fewer than 40 people in the town of Wilamowice in southern Poland in the far southeastern part of historic Silesia (Upper Silesia). The origins of the language are obscure. Wilamowice is believed to have been settled in the thirteenth century by Germanic speakers from Holland, Friesland, Germany, and Scotland. The language itself bears the greatest resemblance to East Central German dialects, which include varieties of Silesian German. Until World War II, the language was the dominant idiom of Wilamowice; however, the repressions of ethnic Germans in postwar Poland referred to above in the discussion of Lower Silesian advanced the shift among Wymysorys speakers, most of whom were bilingual, to using Polish only. The language has been moribund since the 1950s and has now become nearly extinct, though recent efforts have been undertaken to revive the language (Wicherkiewicz 2003).

Minority Germanic Languages

34.2.3

Unique, Noncohesive Indigenous Languages (Cimbrian, Mócheno, Scots) Cimbrian and Mo´cheno are Germanic linguistic isolates in northeastern Italy that are both classified by UNESCO as severely endangered, though Ethnologue identifies their status as “vigorous” (“used for face-to -face communication by all generations and the situation is sustainable”; www.ethnologue.com/about/language-status). Cimbrian is spoken today by around 2,300 people. In only one community in the Autonomous Province of Trento (Trentino), Lusern, a village of fewer than 300 people, is Cimbrian actively spoken (www.ethnologue.com /language/cim). Mo´cheno (known in German as Fersentalerisch) is also spoken in Trentino, mainly in the Bernstol valley, which is just north of Lusern. Arguably just as endangered as Cimbrian, the Mo´chenospeaking population numbers around 1,900 (www.ethnologue.com/lan guage/mhn). Both Cimbrian and Mo´cheno share structural similarities with Bavarian German dialects, which suggest the languages have their origins in immigration to the region from farther north. More speculative work posits that the languages are the modern descendants of Lombardic, the ancient Germanic language spoken in what is today northern Italy. For the purpose of this chapter, I choose to regard Cimbrian and Mo´cheno as indigenous languages since they have been spoken in this area for at least a millennium (Rowley 1996, Bidese 2004). The sociolinguistic situation of Scots is considerably better than that of Cimbrian and Mo´cheno, yet it is still designated vulnerable by UNESCO. Spoken today across Lowland Scotland and parts of Ulster, Scots is considered by many of its speakers a dialect of English due in large measure to the fact that it has been in a diglossic relationship with English for hundreds of years. The closer genetic relationship between Scots and English versus that between both these languages and Scots Gaelic, an indigenous Celtic language, reinforces the popular view of Scots as being subordinate to English. The consensus among most linguists is that Scots is a distinct Germanic language that developed from Early Middle English varieties spoken in Scotland and diverged from the English spoken in England in the twelfth and thirteenth centuries. Today, Ethnologue gives a figure of 1,589,200 users of Scots, for most of whom (1,500,000) the language is not a mother tongue (www.ethnologue.com/language/sco). A 2010 study conducted by the Scottish government on popular attitudes toward the Scots language revealed a mixed picture. In general, Scots regard the language (which most in fact consider a dialect of English) as an important part of the cultural heritage of Scotland but with limited practical value in the present day (Scottish Government Social Research 2010).

817

818

MAR K L. LOU DEN

34.2.4

Unique, Noncohesive Immigrant Languages (Amish Alsatian German, Amish Swiss German) The highly traditional Anabaptist group known as the Old Order Amish is comprised mainly of speakers of Pennsylvania Dutch. There are subgroups of Amish, however, known collectively as “Swiss Amish,” who speak two other languages, Amish Alsatian German and Amish Swiss German. Living today mainly in the US state of Indiana but also in multiple other states, Swiss Amish are the descendants of immigrants to the American Midwest from Alsace and Switzerland in the first half of the nineteenth century. The larger of the two subgroups, whose largest settlement is near the town of Berne, in Adams County, Indiana, speak Amish Swiss German, which is descended from Bernese Swiss German. Their ancestors migrated from the Bern and Jura regions of Switzerland. The second group, who live mainly in an Indiana county adjacent to Adams, namely Allen County, are descended from Amish immigrants from Alsace and still speak a form of Alsatian Low Alemannic known as Amish Alsatian German. As of 2017, the Adams and Allen County communities were the fifth and tenth largest Amish settlements in North America, with estimated populations of 8,595 and 3,190, respectively (https://groups.etown.edu /amishstudies/statistics/largest-settlements/). Considering that there are a number of other smaller settlements across the United States in which Swiss Amish live, one could conservatively estimate the number of Amish Swiss German speakers to be around at least 10,000 and Amish Alsatian German speakers to approach 4,000, or roughly 3 percent and 1 percent of the total Amish population in North America. Despite their small numbers relative to the Pennsylvania Dutch–speaking majority of Old Order Amish, all evidence points to a stable maintenance situation. Swiss Amish children continue to acquire the respective German-related language of their community, Amish Swiss or Amish Alsatian, and the overall growth rate is on par with that of the larger Amish community, which is doubling in size every 20 years due to an unusually high birth rate and low attrition. As such, Amish Alsatian German and Amish Swiss German are not endangered. (See Humpa 1996 and Fleischer and Louden 2011 for Amish Swiss German, Thompson 1994 for Amish Alsatian German.)

34.2.5

Nonunique, Adjoining, Cohesive Indigenous Languages (Limburgish, Low Saxon / Low German) Limburgish is a Germanic language descended from Low Franconian and thus related historically to both Dutch and German. It is spoken in a cohesive region where the Netherlands, Germany, and Belgium meet by an estimated 1,300,000 people, according to Ethnologue, 700,000 of whom live in the Netherlands. Limburgish is still widely spoken by

Minority Germanic Languages

children, though there is evidence to suggest some attrition due to generational differences in its use (www.ethnologue.com/language/lim). Recognized as a regional language under the European Charter by the Dutch government, Limburgish does not enjoy similar status in Germany and Belgium. UNESCO classifies it as a vulnerable language. North of the region in which Limburgish is spoken is the Germanic minority language alternately called Low Saxon or Low German. Once used across a wide area that includes most of modern northern Germany and the eastern Netherlands, the number of native speakers is estimated at around 300,000, though with proportionally many fewer children than adults, an ominous sign that supports its classification by UNESCO as, like Limburgish, vulnerable (“shifting,” according to Ethnologue; www.ethnologue.com/lan guage/nds). Although Low Saxon / Low German is historically distinct from both Dutch and standard German, these languages function as Dachsprachen for Low Saxon / Low German varieties in the Netherlands and Germany. The governments of both countries have formally recognized Low Saxon / Low German as a regional language under the European Charter. The medieval ancestor of the language, Middle Low German, was the main vehicle of communication of the Hanseatic League, whose decline in the fifteenth century contributed to the decline of the language. The popularity of an emerging East Central German–based standard variety through the northward spread of the Reformation in the sixteenth century likewise promoted attrition from Low Saxon / Low German in Germany. The decline of the language accelerated in the nineteenth century in the face of industrialization, urbanization, and an expanding educational system in which standard German was the sole medium. Large-scale emigration of Low Saxon / Low German speakers in the nineteenth century mainly to North and South America, but also to Australia and southern Africa, has made Low Saxon / Low German a truly global language, as documented in a popular television show produced by Norddeutscher Rundfunk in the 2000s, Die Welt op Platt (The World in Low Saxon / Low German).

34.2.6

Nonunique, Adjoining, Noncohesive Immigrant Languages (Hunsrik, Hutterisch, Pennsylvania Dutch) The language known as Hunsrik is spoken predominantly in southern Brazil (especially the state of Rio Grande do Sul) and parts of Argentina and Paraguay by approximately 3,000,000 people, according to Ethnologue, which describes its maintenance status as “shifting” (www.ethnologue.com /language/hrx). As its name implies, Hunsrik has its origins in the Hunsru¨ck region of central-western Germany, located within the West Central German dialect area, from which emigrants began settling in Brazil in 1824. Modern Hunsrik shows the effects of considerable contact with other European German dialects, notably Eastern Pomeranian and Westphalian varieties of Low German, as well other European immigrant

819

820

MAR K L. LOU DEN

and languages and, of course, Portuguese. All German-related varieties spoken in Brazil have experienced attrition as younger speakers shift to Portuguese, yet Hunsrik has remained the strongest demographically, due in part to its use across multiple geographically separate rural communities in what Steffen and Altenhofen (2014) term a “language archipelago” that they compare to that of Plautdietsch-speaking Mennonites in the same region. Whereas a shared religious faith underlies the ties that bind Plautdietsch speakers to one another, the Hunsrik network is based primarily on ethnic and commercial ties. In both situations, an important factor that promotes language maintenance is endogamy (Altenhofen 1996, Ammon 2015: 369–380, Rosenberg 1998). Hutterisch is the in-group language of a highly traditional Anabaptist group known as the Hutterites, who live in the United States and Canada. Although they share most of the same basic Christian beliefs as Amish and Mennonites, the Hutterites, unlike other Anabaptist groups, live in an archipelago of communal settlements called Bruderho¨fe on both sides of the US-Canadian border. The origins of the Hutterite movement lie in early sixteenth-century Tyrol; however, the Hutterisch language is most closely related to the German dialects spoken in Austrian Carinthia (Rein 1977: 216–267). The migration history of Hutterites is a complicated one, similar to that of the Plautdietsch-speaking Mennonites, which is largely due to a legacy of religious persecution. In the 1870s, all Hutterites left Russia, into which they had moved a century prior, for the Dakota Territory in the United States. In the wake of severe repressions at the hands of the US government during World War I, most Hutterites relocated to Canada, where approximately three-quarters of their members live today. The most recent estimate of the total Hutterite population is 45,000. Their birth rate is comparable to that of the Amish, Old Order Mennonites, and Old Colony Mennonites, that is, very high, which, coupled with low attrition from the group, ensures the rapid growth of the number of speakers of Hutterisch (www.hutterites.org/the-leut/distribution/). The demographic health of the Hutterites is an important part of the overall sociolinguistic stability of Hutterisch. The factors of endogamy and ruralness similarly promote heritage language maintenance. Hutterites are basically trilingual. Lorenz-Andreasch (2004), building on earlier work by Kurt Rein (1977), describes a stable diglossic relationship between Vernacular Hutterisch (Alltagshutterisch) and Church Hutterisch (Kirchenhutterisch), which differ from one another structurally and sociolinguistically, the latter used as the main vehicle for religious worship. In addition, all Hutterites are fluent in English, which is the dominant medium of instruction in their schools. LorenzAndreasch also notes the increasing importance of European Standard German among Hutterites, mainly through occasional contacts with German speakers from Europe.

Minority Germanic Languages

Like Hutterisch, Pennsylvania Dutch (Pennsylvania German, as it is often referred to in scholarship, including Ethnologue) is a minority Germanic language spoken in the United States and Canada (plus a very small number in Belize) that is not endangered. It emerged in the eighteenth century through the immigration of German speakers from southwestern Germany, Alsace, and Switzerland to the colony of Pennsylvania, of whom a critical mass came from the Palatinate (Pfalz). Pennsylvania German is lexically and structurally very similar to modern Palatine German dialects, though not identical to any one of them. The bilingualism of Pennsylvania Dutch speakers in English has promoted the divergence from the language’s Palatine German cousins, especially at the lexical and semantic levels. Historically, most speakers were members of Protestant churches, especially Lutheran and (German) Reformed; only a small fraction of the Pennsylvania Dutch founding population was part of Anabaptist and Pietist groups such as Mennonites and Amish. Today, nearly all active speakers are affiliated with the most traditional Amish and Mennonites, known as the Old Orders (Louden 2016). In 2017, the Amish population exceeded 318,000, distributed in yet another archipelago of communities across 31 US states and three Canadian provinces (https://groups.etown.edu/amishstudies/statistics/ population-2017/). Horse-and-buggy–driving Old Order (“Team”) Mennonites live in 13 US states and the Canadian provinces of Ontario and Manitoba, as well as in Belize. Their total population was estimated at just under 40,000 in 2015 (Kraybill et al. 2017: 126–127). There are a number of sociolinguistic parallels between Pennsylvania Dutch-speaking Amish and Mennonites and the Hutterites. Old Order communities are concentrated in rural areas and marriage within the faith is the norm as are exceptionally large family sizes and low attrition rates. Indeed, with populations doubling every 20 years, the Amish (including Swiss Amish), Old Order Mennonites, and Hutterites are the fastestgrowing groups on the planet, meaning that their languages, though small, are growing exponentially. Also, as is the case among Hutterites, trilingualism is the norm for Amish and Old Order Mennonite adults. Pennsylvania Dutch is in a diglossic relationship with another Germanrelated variety, which in the case of the Old Orders is an archaic form of standard German (Hochdeitsch), and all speak English. Old Order sectarians, like their Hutterite counterparts, value the continued use of forms of German as an essential symbol of their sociospiritual identity, a strong factor promoting the long-term health of Pennsylvania Dutch.

34.2.7

Nonunique, Nonadjoining, Cohesive Indigenous Languages (Frisian) The Frisian group of West Germanic languages has three varieties, according to Ethnologue: Frisian (also known as West Frisian), which is spoken in

821

822

MAR K L. LOU DEN

the Netherlands, mainly in the province of Friesland; and North(ern) Frisian and Saterland Frisian (Saterfriesisch in Ethnologue), which are both spoken in Germany. The linguistic distance between the three varieties is such that many people, including a number of Frisian speakers themselves, view them as distinct languages. Of the three varieties, (West) Frisian is the healthiest in terms of maintenance and shift, though it is still deemed vulnerable by UNESCO. In the Dutch province of Friesland it shares official status with Dutch, where it is spoken by approximately three-quarters of the population; most of the rest of Friesland’s nearly 650,000 residents have receptive knowledge of Frisian. The official recognition of Frisian, combined with widespread positive attitudes toward the language and its use among Frieslanders, promotes its continued maintenance. However, the economic base of Friesland is agricultural, meaning that the province has experienced steady out-migration, which raises concerns about the long-term health of the language (Gorter et al. 2001). In Germany, both North Frisian and Saterland Frisian are severely endangered. There are an estimated 8,000 to 10,000 speakers of North Frisian in the state of Schleswig-Holstein, with perhaps twice that number of people who have receptive knowledge of the language. Despite its recognition as a regional language by the German government under the European Charter and support for the language in education, North Frisian remains severely endangered (www.schleswig-holstein.de/DE/Fachinhalte/M/minder heiten/minderheiten_friesen.html). The situation of Saterland Frisian is grimmer. It is spoken by an estimated 2,000 people, mostly elderly, or about one-quarter of the residents of the municipality of Saterland in the state of Lower Saxony (https://www.ethnologue.com/language/stq).

34.2.8

Nonunique, Nonadjoining, Cohesive Immigrant Languages (Afrikaans) Afrikaans is the only (non-English) Germanic language with official status that is spoken exclusively outside of Europe. It is not endangered. It developed through the migration of Dutch dialect speakers during the eighteenth century to the Cape Colony at the southern tip of Africa in what today is the Republic of South Africa. Though quite similar in many respects to European Dutch dialects, Afrikaans diverged historically from European Dutch in large part due to contact with speakers of indigenous languages as well as Malay and Portuguese, which has led some to suggest that creolization was involved in its development (Roberge 2002, Deumert 2004). It is one of 11 official languages in South Africa and the third most commonly spoken as a first language by approximately 14 percent of the population, after Zulu (23 percent) and Xhosa (18 percent) (https://apps.statssa.gov.za/census01/Census96/ HTML/CIB/Population/28.htm). A 2011 survey reported that 6,900,000 South Africans spoke Afrikaans at home, of whom 2,700,000 (nearly 40 percent) were white. The majority of Afrikaans speakers are Coloureds (of multiracial

Minority Germanic Languages

background) and other nonwhites (https://www.news24.com/SouthAfrica/ News/Minority-of-Afrikaans-speakers-white-20130422). Although there is a well-established standard written variety of Afrikaans subject to prescriptive norms, there is considerable variation in vernacular varieties of the language that points to the role of creolization in its historical development. This aspect of the language’s history has been embraced by some Coloured Afrikaans speakers in the postapartheid era (van der Waal 2012). Afrikaans is also widely spoken as a second and third language in South Africa and enjoys considerable support in print and broadcast media. In neighboring Namibia, which as South West Africa was occupied by South Africa from 1915 until 1988, Afrikaans is not an official language, though it is also the third most common language of the country, used as a main language by approximately 10 percent of Namibians (Namibia 2011 Population and Housing Census Main Report). Along with English, the sole official language of Namibia, Afrikaans is an important lingua franca for speakers of diverse indigenous languages.

34.2.9

Nonunique, Nonadjoining, Noncohesive Immigrant Languages (Plautdietsch, Yiddish) Plautdietsch is a language most closely related to East Low German dialects (part of the Low Saxon / Low German group of varieties) that is spoken by Mennonites who are ethnically distinct from Pennsylvania Dutch–speaking Mennonites. Most Plautdietsch speakers are descended from Netherlandicand Frisian-speaking Anabaptists who migrated eastward from northwestern Europe beginning in the sixteenth century, eventually ending up in the Russian Empire (Moelleken 1987, 1992; Siemens 2012). At around the same time that the Hutterites left Russia, groups of Plautdietsch-speaking Mennonites also began coming to North America, mostly Canada, though unlike the Hutterites, many remained in Russia. Today, there are still Plautdietsch speakers in the Russian Federation and other nations of the former Soviet Union, including Kazakhstan, which according to Ethnologue may have as many as 100,000 speakers, over one-quarter of the total estimated Plautdietsch-speaking population. In the 1920s, Plautdietsch speakers began migrating to northern Mexico, and then into South America, where the largest communities are in Bolivia and Paraguay (Moelleken 1987, Krahn et al. 1989, www.ethnologue.com/language/pdt, http://www .mhsc.ca/index.php?content= http://www.mhsc.ca/mennos/hcanada.html). The migration history of Plautdietsch-speaking Mennonites is more complex than that of the Anabaptist groups who use Hutterisch and Pennsylvania Dutch today. Consequently, the contemporary sociolinguistic situation of Plautdietsch speakers is more differentiated than what we find among Amish and Old Order Mennonites and Hutterites. Among these latter groups, heritage language maintenance correlates with active membership in Amish, Old Order Mennonite, and Hutterite churches. The most traditional

823

824

MAR K L. LOU DEN

Anabaptist groups to maintain Plautdietsch (alongside a form of standard German known as Hu¨agdietsch ‘High German’) are the Old Colony Mennonites, who, like their Old Order and Hutterite counterparts, live at a distance, physically and spiritually, from the social mainstream. However, there are more progressive Mennonite churches, many of which are actively engaged in mission work, whose members continue to use Plautdietsch (see the Ekj Ran [‘I Run’] ministry, www.squareoneworldmedia.com/ministries/ low-german/ekj-ran). This is a different situation than among Pennsylvania Dutch and Hutterisch speakers who choose not to join (or leave) Old Order or Hutterite churches and who then usually shift to speaking English predominantly or exclusively. Overall, Plautdietsch is in much the same robust state of sociolinguistic health as the languages of the Amish, Old Order Mennonites, and Hutterites due to rapid growth, low attrition, endogamy, and a measure of social separation. And as with other traditional Anabaptist groups, minority language maintenance has a symbolic importance as the conservation of a precious spiritual heritage. There are a number of outward parallels between Yiddish and the languages of traditional Anabaptist groups. Structurally, it is descended from German dialects spoken by Ashkenazic Jews going back as far as the tenth century. Like their Anabaptist counterparts, Yiddish speakers, as members of a distinctive religious minority, experienced persecution that in the case of the Jews reached the level of genocide in the twentieth century. Although the Shoah dealt a critical blow to Yiddish, the language has not only endured but is thriving today as a vital vernacular in many so-called Haredi groups, highly traditional orthodox Jewish communities that include the Hasidim and are often compared in the US to the Amish. As was true of Pennsylvania Dutch historically, Yiddish was once widely spoken by Ashkenazic Jews who lived at less of a distance from the social mainstream. Active use of Yiddish among secular Jews is exceptional today – hence its UNESCO designation as “definitely endangered” (in Israel) – however, the growth rate among groups like the Hasidim is comparable to that of Amish, traditional Mennonites, and Hutterites, that is, exponential, thereby securing the future of the language. Endogamy is also a crucial factor in promoting the maintenance of Yiddish. Though most Haredi communities are urban, unlike the rural-dwelling traditional Anabaptists who maintain minority heritage languages, active Yiddish speakers still typically live in distinct enclaves, physically and spiritually, that support language maintenance (Isaacs 1999, Jacobs 2005, Katz 2007).

34.2.10

Local-only, Adjoining, Cohesive Indigenous Languages (Danish in Germany, German in Denmark and Belgium) The Duchy of Schleswig was the political entity that for centuries existed in the territory where today Denmark and Germany meet. Historically, there were three main languages spoken in the duchy, Danish, North Frisian, and Low German. Until 1864, Schleswig was part of Denmark, at

Minority Germanic Languages

which point it became Prussian. After World War I, two plebiscites were held on the political future of the territory, with North Schleswig joining Denmark and South Schleswig remaining German, eventually becoming part of the modern state of Schleswig-Holstein. North and South Schleswig were never linguistically homogeneous, thus the partition resulted in (Low) German- and Danish-speaking minorities in Denmark and Germany, respectively. (North Frisian speakers were concentrated in South Schleswig.) As signatories to the European Charter for Regional or Minority Languages, both Germany and Denmark are committed to supporting the minority communities in their respective countries, including in schools and media. The Danish minority community in Germany is estimated at around 50,000, though Ethnologue estimates there were 20,800 speakers of Danish, likely South Jutish, the dialect of Danish indigenous to the Danish-German border region (www.schleswig-holstein.de /DE/Fachinhalte/M/minderheiten/minderheiten_daenen.html; https:// www.ethnologue.com/language/dan). One study (Pedersen 2002) suggests that the percentage of active Danish speakers among German citizens who identify as ethnic Danes is quite small, which supports the UNESCO assessment that South Jutish is definitely endangered. The language maintenance situation among members of the ethnic German minority of the Danish-German border, in South Jutland, is considerably better. German is the only regional or minority language recognized by the Danish government under the European Charter. There are German-medium schools serving the German minority community, which numbers between 10,000 and 20,000, of whom one-third are estimated to have German as their mother tongue (Ammon 2015: 306). Despite their small numbers, the maintenance of German is supported institutionally by high-quality schools and through media and religious institutions. The importance of German as an international language relative to Danish likely makes the maintenance of the former additionally attractive (Pedersen 1996, 2000; Ammon 2015: 305–311). As in southern Denmark, eastern Belgium is home to an autochthonous group of German speakers. The German-speaking Community of Belgium (Deutschsprachige Gemeinschaft Belgiens) or East Belgium (Ostbelgien, www.ostbelgienlive.be) has a population of approximately 77,000. Although there are also small numbers of French and Flemish speakers in the community, German is its official language. Residents of East Belgium account for less than one percent of the total population of Belgium, nevertheless the maintenance of German there is secure. Bilingualism among German Belgians in French is widespread, as East Belgium is part of the region of Wallonia, though the majority Frenchspeaking Walloons are less likely to speak German. The economic health of the region, a strong infrastructure, including in education, as well as its proximity to Germany, are all factors promoting the maintenance of German there (Ammon 2015: 232–240).

825

826

MAR K L. LOU DEN

34.2.11

Local-only, Adjoining, Noncohesive Immigrant Languages (Swedish in Finland) Until 1809, what is today the nation of Finland was part of the kingdom of Sweden. After another century as part of the Russian Empire, Finland finally gained its independence in 1917. Although Swedish speakers have never comprised more than a small minority of the Finnish population, since independence both Finnish and Swedish are officially national languages. Recent figures the number of Swedish speakers at 325,000, or 5.4 percent of the Finnish population (www.stat.fi/tup/suoluk/suoluk_vaesto_en.html). The Swedish presence in the country is an old one, due to migration from Sweden going back several centuries. Svenskfinland is non-cohesive, including four areas on the western coast of Finland. Owing largely to its official support, Swedish may be considered a safe language, despite a considerable number of bilingual families due to marriage across language communities ¨ stman and Mattfolk 2011). The close genetic relationship of (Østern 2001, O Swedish with other Scandinavian languages also makes its continued maintenance by Finnish citizens a communicative asset for them.

34.2.12

Local-only, Nonadjoining, Cohesive Immigrant Languages (German in Namibia) The German presence in Namibia is a legacy of its status as the sole official language of the German imperial colony German Southwest Africa (DeutschSu¨dwestafrika) between 1884 and 1915, at which point it fell under the administration of neighboring South Africa. German gained official recognition by South African authorities in 1984, but lost it again when Namibia became independent in 1990. Today, German speakers account for less than one percent (20,000–25,000) of the total Namibian population (Ammon 2015: 362); however, many thousands more speak it as a second language. Competition from English, the sole national language of Namibia, and Afrikaans do not strengthen the position of German in the country. Factors favorable to the long-term health of German in Namibia include economic and cultural ties to Germany and the support among German Namibians for German-medium schools, media, and cultural institutions. German’s risk of endangerment in Namibia has not been formally assessed by UNESCO, however Ulrich Ammon has expressed skepticism about its future, hence one could deem it vulnerable (Ammon 2015: 359–369).

34.3 Minority Language Maintenance and Shift By way of concluding this chapter, it is worth considering whether there is a correlation between the White/Edwards typology for the classification of minority languages and language endangerment. Is a language’s health affected by whether it exists as a minority variety solely in one or more

Minority Germanic Languages

states (i.e., whether it is unique or nonunique), or whether it has minority status in one state but is a majority language elsewhere (local-only)? For nonunique and nonlocal languages, does it matter whether the states in which they are spoken are adjoining or not? And does the spatial cohesion of a minority language region correlate in any way with the maintenance of that language? Finally, does a minority language’s status as an indigenous or immigrant variety promote or hinder maintenance or shift? Let us begin by reminding ourselves which of the languages discussed above are not endangered. Of the 21 languages we examined, nine are in a good state of sociolinguistic health: Afrikaans, Amish Alsatian German, Amish Swiss German, German (in Denmark and Belgium), Hutterisch, Pennsylvania Dutch, Plautdietsch, Swedish (in Finland), and Yiddish. Interestingly, just one of these languages, German, is indigenous to the regions in which it is a minority language (except for in Namibia); the remaining eight languages are all the products of immigration at different points in world history, with Swedish and Yiddish being the oldest, followed by Afrikaans, Hutterisch, Plautdietsch, Pennsylvania Dutch, and finally, Amish Alsatian German and Amish Swiss German. The table of minority language types from Section 34.1.2 is reproduced in simplified form in Table 34.2, with the nine healthy languages in boldface. Languages

Table 34.2 Classification of minority Germanic languages and language endangerment minority language type

indigenous

1. unique, cohesive

Lower Silesian Övdalian 2. unique, noncohesive Cimbrian Mócheno Scots 3. nonunique, adjoining, cohesive Limburgish Low Saxon / Low German 4. nonunique, adjoining, noncohesive 5. non-unique, nonadjoining, cohesive 6. nonunique, nonadjoining, noncohesive 7. local-only, adjoining, cohesive

8. local-only, adjoining, noncohesive 9. local-only, nonadjoining, cohesive 10. local-only, nonadjoining, noncohesive

Frisian

immigrant Colonia Tovar German Wymysorys Amish Alsatian German Amish Swiss German

Hunsrik Hutterisch Pennsylvania Dutch Afrikaans Plautdietsch Yiddish

Danish in Germany German in Denmark, Belgium Swedish in Finland German in Namibia

827

828

MAR K L. LOU DEN

deemed vulnerable to loss are in normal typeface and endangered languages are italicized. Of the nine minority language types with Germanic examples, six are represented by healthy languages. There is no obvious correlation between minority language type and whether a given language is more or less likely to be endangered or not. Two languages (Amish Alsatian German and Amish Swiss German) are unique, while the remaining seven are not; three out of the nine (German, Hutterisch, and Pennsylvania Dutch) are adjoining; and four out of nine (German, Hutterisch, Swedish, and Pennsylvania Dutch) are spatially cohesive. The sociolinguistic correlates of minority language health must be sought elsewhere. It is notable that six of the nine healthy minority languages (Amish Alsatian German, Amish Swiss German, Hutterisch, Pennsylvania Dutch, Plautdietsch, and Yiddish) are all associated with highly traditional religious groups who for spiritual reasons maintain clear boundaries between themselves and their neighbors through endogamy and a number of outward symbols, including distinctive dress and limitations on the use of technology. The languages they speak also have symbolic value as they are entirely their own, that is, unique to their communities. Although Amish, conservative Mennonites, Hutterites, and Haredim do not live in complete isolation from the larger world, they do very intentionally delineate physical and spiritual spaces within which they use languages that belong solely to them. In this way, the speakers of these six languages create a form of protective isolation around themselves and their languages that the endangered or vulnerable languages under discussion in this chapter either have lost or are likely to lose. It is no coincidence that these other languages in earlier times enjoyed stability when their speakers lived in relative geographic isolation in rural areas (like Friesland or the Brazilian countryside) or in small ethnolinguistically homogeneous communities (like Wilamowice or isolated villages in northern Italy) at a physical distance from others. How is it that three other languages, Afrikaans, German in Denmark and Belgium, and Swedish in Finland, where social-religious isolation is not relevant, are not endangered? A combination of factors are at play, none of which is unique to each language. All are officially recognized by the governments of the nations in which they are located and have deep historic roots there: Afrikaans began emerging in southern Africa three hundred years ago, Scandinavian speakers have been in Finland for perhaps as long as a millennium, and German is effectively indigenous to the regions of Denmark and Belgium where it is spoken. Each of the three languages has a well-developed and -supported written standard variety, and each is used in a wide range of domains ranging from informal to formal, by both rural- and urban-dwellers. German in Denmark and Swedish in Finland have the added advantage of being languages that have wider international currency than their coterritorial majority

Minority Germanic Languages

languages, Danish and Finnish. The economic health of the communities in question doubtless also plays a role, limiting the incentive or need for younger community members to migrate. The Afrikaans situation is an interesting one. It shares all the positive factors relevant to German and Swedish as minority languages, with one exception: the scope of Afrikaans is limited to southern Africa. It is difficult to imagine that Afrikaans speakers in South Africa and Namibia would be inclined to shift to the most widely spoken languages in their respective countries, Zulu and Ovambo. A shift to English is more plausible; however, in South Africa, at least, Afrikaans is holding its own, due in part to a legacy of distance between English-speaking South Africans and Afrikaners going back at least as far as the Boer Wars at the turn of the twentieth century. The fact that nonwhites comprise the majority of Afrikaans speakers is important for the future language, as well: it is telling that the current directorship of the Afrikaans Language Council (Afrikaanse Taalraad, afrikaansetaalrad.co.za) is made up of three ethnic Afrikaners and three nonwhites. In her monograph on language maintenance and shift, Anne Pauwels (2016) underscores the diversity of circumstances that render some minority language situations more stable than others and explores the important question whether the shift of speakers to a majority language can or even should be addressed. Do the Germanic languages situations above offer any prescriptions to those who would attempt to promote minority language revitalization? To be sure, the successes of the traditional Anabaptist and orthodox Jewish communities in maintaining their heritage languages without special effort (institutional support, use in schools and media, etc.) are not practical for, say, Frisians in the Netherlands or Germany, who are unlikely to choose intentionally to marry exclusively within their ethnic group, live in rural areas only, and have six or seven children whom they would raise to do likewise. However, what the Amish, Mennonite, Hutterite, and Haredi situations do show is that successful minority language maintenance “from below” is possible, even in the heart of industrialized societies like the United States whose majority populations are not inclined to promote linguistic diversity.

References Altenhofen, C. V. 1996. Hunsru¨ckisch im Rio Grande do Sul. Ein Beitrag zur Beschreibung einer deutschbrasilianischen Dialektvarieta¨t im Kontakt mit dem Portugiesischen. Stuttgart: Franz Steiner. Ammon, U. 2015. Die Stellung der deutschen Sprache in der Welt. Berlin: Walter de Gruyter. Bidese, E. 2004. “Die Zimbern und ihre Sprache: geographische, historische und sprachwissenschaftlich relevante Aspekte.” In T. Stolz (ed.),

829

830

MAR K L. LOU DEN

Kolloquium u¨ber Alte Sprachen und Sprachstufen. Beitra¨ge zum Bremer Kolloquium u¨ber “Alte Sprachen und Sprachstufen” (Diversitas Linguarum 8). Bochum: Brockmeyer: 3–42. Da Rin, D. 1997. Das Deutsche in der Colonia Tovar (Venezuela). Munich: LINCOM Europa. Deumert, A. 2004. Language Standardization and Language Change: The Dynamics of Cape Dutch. Amsterdam: John Benjamins. Edwards, J. 1992. “Sociopolitical aspects of language maintenance and loss: towards a typology of ethnic minority language situations.” In W. Fase, K. Jaspaert, and S. Kroon (eds.), Maintenance and Loss of Ethnic Minority Languages. Amsterdam: John Benjamins: 37–54. Edwards, J. 2004. “Language minorities.” In A. Davies and C. Elder (eds.), The Handbook of Applied Linguistics. Malden, MA: Blackwell: 451–475. Extra, G., and D. Gorter (eds.), 2001. The Other Languages of Europe: Demographic, Sociolinguistic, and Educational Perspectives. Bristol: Multilingual Matters. Fleischer, J., and M. L. Louden. 2011. “Das Amish Swiss German im nordo¨stlichen Indiana: eine alemannisch-pfa¨lzische Mischmundart.” In H. Christen, S. Germann, W. Haas, N. Montefiori, and H. Ruef (eds.), Alemannische Dialektologie: Wege in die Zukunft (ZDL-Beiheft 141). Stuttgart: Franz Steiner: 231–244. Gorter, D., A. Riemersma, and J. Ytsma 2001. “The Frisian language in the Netherlands.” In Extra and Gorter (eds.), pp. 103–118. Hinderling, R., and L. M. Eichinger (eds.). 1996. Handbuch der mitteleuropa¨ischen Sprachminderheiten. Tu¨bingen: Gunter Narr. Humpa, G. J. 1996. Retention and Loss of Bernese Alemannic Traits in an Indiana Amish Dialect: A Comparative Historical Study. Ph.D. dissertation, Purdue University. Isaacs, M. 1999. “Haredi, haymish, and frim: Yiddish vitality and language choice in a transnational multilingual community.” International Journal of the Sociology of Language 138: 9–30. Jacobs, N. G. 2005. Yiddish: A Linguistic Introduction. Cambridge University Press. Katz, D. 2007. Words on Fire: The Unfinished Story of Yiddish. New York: Basic Books. Krahn, C., H. S. Bender, and J. J Friesen 1989. “Migrations.” In Global Anabaptist Mennonite Encylopedia Online. Retrieved 16 January 2017. http://gameo.org/index.php?title=Migrations&oldid=143668. Kraybill, D. B., S. M. Nolt, and E. Burge Jr. 2017. “Language use among Anabaptist groups.” In S. J. Bronner and J. R. Brown (eds.), Pennsylvania Germans: An Interpretive Encyclopedia. Baltimore: Johns Hopkins University Press: 108–130. Lasatowicz, M. and T. Weger. 2008. “Polen.” In L. Eichinger, A. Plewnia, and C. M. Riehl (eds.), Handbuch der deutschen Sprachminderheiten in Mittel- und Osteuropa. Tu¨bingen: Gunter Narr: 145–169.

Minority Germanic Languages

Lorenz-Andreasch, H. 2004. “Mir sein jaº kolla Teitschverderber” Die Sprache der Schmiedeleut-Hutterer in Manitoba/Kanada. Wien: Edition Praesens. Louden, M. L. 2016. Pennsylvania Dutch: The Story of an American Language. Baltimore: Johns Hopkins University Press. Moelleken, W. W. 1987. “Die rußlanddeutschen Mennoniten in Kanada und Mexiko: Sprachliche Entwicklung und diglossische Situation,” Zeitschrift fu¨r Dialektologie und Linguistik 54: 145–183. Moelleken, W. W. 1992. “The development of the linguistic repertoire of the Mennonites from Russia.” In K. Burridge and W. Enninger (eds.), Diachronic Studies on the Languages of the Anabaptists. Bochum: Universita¨tsverlag D. N. Brockmeyer: 64–93. Namibia 2011 Population and Housing Census Main Report. https://web .archive.org/web/20131002064316/ http://www.nsa.org.na/files/down loads/Namibia%202011%20Population%20and%20Housing%20Census% 20Main%20Report.pdf. Østern, A. 2001. “Swedish in Finland.” In Extra and Gorter (eds.): 159–173. ¨ stman, J.-O., and L. Mattfolk 2011. “Ideologies of standardisation: Finland O Swedish and Swedish-language Finland.” In T. Kristiansen and N. Coupland (eds.), Standard Languages and Language Standards in a Changing Europe. Oslo: Novus: 75–82. Pauwels, A. 2016. Language Maintenance and Shift. Cambridge University Press. Pedersen, K. M. 1996. “Die deutsche Minderheit in Da¨nemark.” In Hinderling and Eichinger (eds.): 31–59. Pedersen, K. M. 2000. “A national minority with a transethnic identity – the German minority in Denmark.” In S. Wolff (ed.), German Minorities in Europe: Ethnic Identity and Cultural Belonging. New York: Berghahn: 15–28. Pedersen, K. M. 2002. Dansk sprog i Sydslesvig. Aabenraa: Institut for Grænseregionsforskning. Rein, K. 1977. Religio¨se Minderheiten als Sprachgemeinschaftsmodelle. Deutsche Sprachinseln ta¨uferischen Ursprungs in den Vereinigten Staaten von Amerika. ZDL Beihefte, Neue Folge, Nr. 15. Wiesbaden: Franz Steiner. Roberge, P. T. 2002. “Afrikaans: considering origins.” In R. Mesthrie (ed.), Language in South Africa. Cambridge University Press: 79–103. Rosenberg, P. 1998. “Deutsche Minderheiten in Lateinamerika.” In T. Harden and E. Hentschel (eds.), Particulae particularum. Festschrift zum 60. Geburtstag von Harald Weydt. Tu¨bingen: Stauffenburg: 261–291. Rowley, A. 1996. “Die Sprachinseln der Fersentaler und Zimbern.” In Hinderling and Eichinger (eds.): 263–285. Scottish Government Social Research 2010. Public attitudes towards the Scots language. Accessible at http://www.gov.scot/Resource/Doc/298037/ 0092859.pdf. Siemens, H. 2012. Plautdietsch. Grammatik, Geschichte, Perspektiven. Bonn: tweeback verlag.

831

832

MAR K L. LOU DEN

Simons, G. F., and C. D. Fennig (eds.) 2017. Ethnologue: Languages of the World, 20th edn. Dallas: SIL International. www.ethnologue.com Søndergaard, B. 1981. “The fight for survival: Danish as a living minority language south of the German border.” In E. Haugen, J. D. McClure, and D. Thomson (eds.), Minority Languages Today. Edinburgh University Press: 138–143. Steffen, J., and C. V. Altenhofen 2014. “Spracharchipele des Deutschen in Lateinamerika: Dynamik der Sprachvernetzungen im mehrsprachigen Raum,” Zeitschrift fu¨r Dialektologie und Linguistik 81: 34–60. Thompson, C. 1994. “The languages of the Amish of Allen County, Indiana: multilingualism and convergence,” Anthropological Linguistics 36: 69–91. Waal, C. S. van der 2012. “Creolisation and purity: Afrikaans language politics in post-apartheid times,” African Studies 71: 446–463. White, P. 1987. “Geographical aspects of minority language situations in Italy.” Unpublished paper presented at the International Seminar on Geolinguistics, Stoke-on-Trent. Wicherkiewicz, T. 2003. The Making of a Language: The Case of the Idiom of Wilamowice, Southern Poland. Berlin and New York: Walter de Gruyter. Zach, K. 2013. Das A¨lvdalische – Sprache oder Dialekt? M.A. thesis, University of Vienna.

Chapter 35 Germanic Contact Languages Paul T. Roberge

35.1 Contact Languages We can distinguish two types of metropolitan Germanic languages. On the one hand, there are languages that have been transmitted from generation to generation with gradual internal change and with varying degrees of lexical borrowing as outcomes of endogenous agentivity. Icelandic comes immediately to mind. On the other hand, there are languages the history of which includes one or several periods of exogenous agentivity on the part of in-migrating populations. Groups of limited-proficiency nonnative speakers impose their own features (L1-transfer [first language], restructuring) on a co-territorial recipient or target language. The obvious example here is English as compared with the other Germanic languages on the European continent and in the North Atlantic. Danish and Norwegian invaders settled in the north and east of England between 865 and 955. Between 1066 and 1070, French-speaking Normans conquered England. Although their languages have left indelible imprints on the recipient language, by any conventional view, “there would still be English without the Danes and the Normans, only a different English” (Kihm 2012: 659). Colonial varieties of English in the “neo-Europes” (United States, Canada, Australia, New Zealand) developed among settlers from the British Isles, with only limited influence from indigenous languages and adstratal European languages that English displaced in the course of the nineteenth and twentieth centuries. Ethnic varieties of English persist only in communities that are not yet fully integrated into the dominant population (Mufwene 2008:53). So-called “new Englishes” are second-language (L2) varieties that developed under British and American colonialism, specifically in multilingual territories (e.g., Nigeria, Kenya, India, Singapore, the Philippines) where English was introduced in the official sphere and through formal education rather than by a large settler population for whom it was a first language. Dutch is the sole official language of

834

PA U L T. R O B E R G E

postcolonial Suriname and is used alongside of some 20 ethnic home languages. Suriname Dutch spans a continuum, the poles of which are Standard Dutch and a locally colored vernacular form. Such varieties have been “indigenized” through the adoption of lexis, phonological patterns, and grammatical features from local languages, along with some restructuring as the product of L2-acquisition processes. Nonetheless, they continue to belong to the same stock as their metropolitan congeners. Contact language has served as a cover term for languages that show extensive effects of contact-induced change (e.g., Schreier and Hundt 2013) and for languages that are used for wider communication (corresponding, more or less, to the French term langue ve´hiculaire). In this article a different definition is used. Thomason defines contact language as “any new language that arises in a contact situation. Linguistically, a contact language is identifiable by the fact that its lexicon and grammatical structures cannot all be traced back primarily to the same source language; they are therefore mixed languages in the technical historical linguistic sense; they did not arise primarily through descent with modification from a single earlier language” (Thomason 2001: 158, my emphasis). The prototypical contact languages are pidgins, creoles, and (bilingual) mixed languages (Thomason 2001: chapters 6 and 7).1 Pidgins, creoles, and bilingual mixed languages have a long history of stigmatization. They have often been seen as corrupted or parasitic varieties rather than languages in their own right. It is usual and customary for scholars to classify pidgins and creoles according to their principal lexifier language. If lexical affinities are indicia of group membership, then English-, Dutch-, and German-lexified (or -based) pidgins and creoles are in that sense “Germanic”; for convenience, I follow Romaine’s practice (1994) of grouping them under that label even though it is not generally used. Whether creoles are truly new languages that have no genetic affiliation (Thomason and Kaufman 1988) or are to be situated along the phylogenetic branches of their lexifiers (Mufwene 2008, 2015; Aboh and DeGraff 2016) is a contentious issue.2

35.2 Pidgins By pidgin language I understand the linguistic creation of a contact community that has need for a common means of communication but does not 1

Bakker and Matras (2013) add so-called multi-ethnolects to the rubric of contact languages. The former are linguistic styles that are part of the repertoires of speakers of different ethnolinguistic backgrounds. These varieties have their base in the dominant language of the society but draw from the other languages in contact. Exempla are new varieties that have emerged among second and third generations of immigrants, usually in urban environments, in Germany, the Low Countries, Scandinavia, the United Kingdom, and the United States (see Nortier and Dorleijn 2013). Tsotsitaal ‘hoodlum language’ (also known as Fly Taal, fly ‘streetwise’) arose in the twentieth century among Africans in Johannesburg, Pretoria, and other cities in South Africa. It originally had Afrikaans as its base, but there are now many Tsotsitaal varieties based on South Africa’s indigenous languages.

2

Anyone who writes about these languages has to cope with a terminological minefield. I have for the most part followed the naming conventions of the Atlas of Pidgin and Creole Language Structures (Michaelis et al. 2013a).

Germanic Contact Languages

share a preexisting language that fulfills this function. Pidgins are social (group) solutions to the problem of interethnic communication. A pidgin is a restricted linguistic system that is used in limited domains (such as trade, labor, military service) by people who retain their spoken languages. It is native to no one. Generally, pidgins rely on free rather than bound forms. They lack case, gender, and agreement. Nominal plurals and articles are not usually found. It is common for pidgins to lack overt copulas in predicative constructions. Pidgins may show one or two prepositions and bimorphemic question words. Syntactically, pidgins are characterized by derivational shallowness (little or no embedding), a want of movement rules, no passivization, and negation by means of an unbound preverbal operator. Pidgins express tense, mood, and aspect (TMA) by means of context or through adverbials in positions that are external to the propositions that they qualify. Pidgin phonology includes a smaller inventory of contrasts relative to their input languages, simpler phonotactic structure, and lack of tone. The lexicon is minimal. Pidgin speakers make use of polysemy, multifunctionality, compounding, and circumlocution to generate new word forms and express concepts that in full languages might be lexicalized. Reduplication, which is almost universal in creoles, is somewhat rare in pidgins. Pidgins are said to have stabilized with the establishment of communal norms and a sharp enough separation from the languages in contact so as not to be mutually intelligible with any of them. Stabilization implies a gradual transition from a pragmatic mode of speaking to a more regular grammatical one; it develops gradually and is achieved at different times in different parts of grammar (Mu¨hlha¨usler 1997: 138). Grammatical stability is what distinguishes pidgins from jargons and early interlanguages (secondary hybrids in the model of Whinnom 1971), which are unstable, individual solutions to the problem of interethnic communication (cf. Mu¨hlha¨usler 1997:128). Stability is, of course, a matter of degree. Pidgins are likely to show considerable variability in the mouths of users of different linguistic backgrounds. As dynamic systems, pidgins may also develop from lesser to greater complexity as communicative requirements become more demanding (Mu¨hlha¨usler 1997: 6, 119).

35.2.1

Mediums of Interethnic communication (MICs) in the European Metropole With the gradual spread of German eastward from the ninth century until its abrupt retreat after the Second World War, transitory contact varieties of German came into use for interethnic communication. Halbdeutsch refers to L2-versions of German spoken by Estonians and Latvians from the late Middle Ages through the nineteenth century (Stammler 1922, Mitzka 1923, Lehiste 1965). Mitrović (1972) provides a brief, anecdotal description of a pidginized German that existed in Bosnia-Herzegovina ca. 1878–1918 (as

835

836

PA U L T. R O B E R G E

per Mu¨hlha¨usler 1984: 28) among settlers from all parts of the AustroHungarian empire and between these settlers and the multilingual local population. Reports of pidgins and jargons in the North Atlantic are catalogued in Hancock (1996). These include an eighteenth-century Swedish-Sami trade jargon called Borgarmaº let (Broch and Jahr 1984b: 51–52, Velupillai 2015: 31–33) and the ironically named Fa´skru´ðsfjarðarfranska (‘Fa´skru´ðsfjord French’) used in Iceland by North European mariners in the nineteenth century and which may have contained some lexical contributions from French fishermen (cf. Bakker 1989). Broch (1996) calls our attention to an English-Russian trade jargon at Archangel on the White Sea during the second half of the eighteenth century. Russenorsk is an extinct seasonal pidgin that was once spoken by Norwegians and Russians involved in barter trade (most importantly fish and grain) from the late eighteenth century (first attested in 1785) up until the Russian revolution of 1917. Its maximal range subsumed the Arctic coastal region between Tromsø, in Norway, in the west, and Kola, in Russia in the east (Jahr 1996: 108). Its lexicon is drawn in roughly equal measure from the two stock languages, though there are also lexical items from Sami, Low German / Dutch, French, Swedish, and English (Jahr 1996: 110). Jahr (1996) shows that Russenorsk was a stable focused variety that did not form according to Whinnom’s (1971) tertiary hybridization model; that is, it arose among social equals with limited direct contact and without the involvement of a lexifying superstratum.3 Trudgill (1996: 9–10) suggests that we need to distinguish such a pidgin typologically from other pidgins by naming it a dual-source pidgin. The source material for Russenorsk is collected and analyzed in Broch and Jahr (1984a, 1984b). The nation-states of Europe and the British Isles have a vast array of languages spoken within their borders. Of particular interest to linguists are the verbal repertoires of immigrants. A widely studied case is that of Foreign Worker German, which is spoken by laborers who migrated to what was then the Federal Republic of Germany from southern Europe and Turkey during the late 1950s, 1960s, and early 1970s. Clyne (1968) described their German (also known as Gastarbeiterdeutsch) as a pidgin on the basis of shared features with contact languages generally recognized as pidgins. The Heidelberger Forschungsprojekt “Pidgin Deutsch” (1975) questioned the stability of Foreign Worker German while continuing to call this variety a pidgin. Gilbert and Pavlou (1994) outlined a functional typology of pidgins in which they classified Foreign Worker German as an industrial pidgin on the basis of its structural properties and barriers to targeted L2-acquisition due to the ghettoization of its speakers. However, the communis opinio is that the 3

According to Whinnom (1971), a socially more powerful (superstrate) language provides the bulk of vocabulary for speakers of two or more substrate languages, who create the grammar. None of the full languages in the mix serves as a target language.

Germanic Contact Languages

German of the immigrant generation is constitutive of a continuum of jargons and interlanguage varieties, with German and ancestral-language ambilingualism in generations born in the host country. In her study of children in heavily foreign neighborhoods of Berlin, Pfaff (e.g., 1981) found little linguistic evidence of creolized German.

35.2.2

Trade and Maritime Pidgins on the Mainland Coast of West Africa The Portuguese were the first Europeans to explore the coast of Guinea and establish fortresses and lodges there (from 1482). Their main incentive was the potentially lucrative exchange of European goods for gold, ivory, slaves, and local produce. Though attestations are sparse, the establishment of a Portuguese-lexified pidgin in Lower Guinea from the late fifteenth century onwards is beyond doubt; it survived as a trade language on the Gold Coast for at least a century and a half after the Portuguese had surrendered their outposts to the Dutch in 1637–1642. The Dutch, whose objective was likewise mercantile rather than expansionist, adopted the Portuguese pidgin and possibly relexified it. When the British set up trading stations of their own (from 1631), they did so in places where there had been no established Portuguese presence. An early West African pidgin English took shape by the end of the seventeenth century (first attestation 1686, as per Huber 2013: 168) in what are today The Gambia, Sierra Leone, and Ghana; it was used exclusively for communication between Africans and Europeans (Holm 1989: 411, Huber 1999b:119ff., 2013: 167). As the British became involved in the slave trade during the eighteenth century, their ships worked the Grain Coast (on the western coast of the Gulf of Guinea, extending approximately from Cape Mesurado to Cape Palmas — in what is today Liberia) and recruited people to man their crews and serve as middlemen to local traffickers. The use of pidgin English by these so-called Krumen (from the ethnic name of the Kru) probably dates from roughly 1700. As their activity extended eastward along the Guinea coast, the Krumen probably played a role in the spread of the pidgin (Holm 1989: 421–422). Today, forms of pidgin and creole English are spoken along the West African coast, mainly in multilingual countries where English is an official language – The Gambia, Sierra Leone, Liberia, Ghana, Nigeria, and Cameroon – and in Equatorial Guinea on the island of Bioko (formerly Fernando Po´). Present-day West African Pidgin English developed in the nineteenth century under British colonial rule, a subject to which we shall return in Section 35.4 on expanded pidgins. 35.2.3

English-Lexified Trade and Maritime Pidgins in Asia and Pacific The Atlas of Languages of Intercultural Communication in the Pacific, Asia, and the Americas (Wurm et al. 1996) shows the distribution of (inter alia) pidgin

837

838

PA U L T. R O B E R G E

Englishes spoken currently or formerly in the Pacific and along the China coast, alongside descriptive articles relating to the maps.

35.2.3.1 Chinese Pidgin English (China Coast Pidgin) The Portuguese established a permanent trading settlement at Macau in 1557, on the Pearl (Zhujiang) River estuary and downriver from Canton (Guangzhou), which was the main port for foreign commerce in the China trade until the opening of Hong Kong as an entrepoˆt in 1842. Portugal managed to retain its toehold in China despite the loss of most of its Asian empire to the Dutch in the 1640s and 1650s. It was during this period that the English sought to open up trade with China. Eventually, an Englishlexified pidgin developed around the Canton area as a medium for conducting pidgin ‘business’, whence the metonymic general term for restricted systems that serve this purpose (although other etymologies have been suggested; see Mu¨hlha¨usler 1997: 1–2). According to Baker (1987: 164), Chinese Pidgin English (CPE) is known to have existed since the first half of the eighteenth century, possibly as early as 1715 and certainly no later than 1743, the date to which the first published sample relates. However, the Chinese authorities closely regulated mercantile activity and restricted interaction between foreigners and local populations. There is reason to think that the initial impetus to pidgin formation obtained from simplified registers on the part of Anglophone mariners and merchants in informal exchanges. (Formal exchanges were most often mediated by interpreters.) Most Europeans would have had little opportunity to acquire Cantonese (Ansaldo et al. 2010: 91, who grant that direct contact might have been more common). Moreover, the historical and philological records suggest that pidgin Portuguese must have been used as a lingua franca in Macau and Canton throughout the eighteenth century and may have left some imprint on CPE (see especially Li and Matthews 2016). The upshot is that CPE may not have crystallized until at least mid-century (Ansaldo et al. 2010: 91, Li 2016: 299). The First Opium War (1839–1842) changed the situation dramatically. The British were able to force China to cede Hong Kong and to open up Canton and four city-ports for foreign residence and unrestricted trade – Amoy (Xiamen), Foochow (Fuzhou), Ningbo, and Shanghai. With further concessions resulting from the Second Opium War (1856–1860), Western trade expanded further along the China coast as well as inland; and with that expansion the pidgin spread to the Pacific and eventually (with immigration) as far as Australia and the West Coast of the United States (Ansaldo et al. 2010: 1). The pidgin was reportedly used not only in ChineseEuropean communication but also among Chinese of different linguistic backgrounds (Whinnom 1971:102–104). As a class of Chinese educated in Standard English began to emerge in the latter half of the nineteenth century, CPE became associated with service to foreigners (Holm 1989:

Germanic Contact Languages

519). During the twentieth century its use declined until it was spoken only by a few Chinese in Hong Kong as of the late 1960s (Whinnom 1971).4

35.2.3.2 Pacific Pidgin English (PPE) PPE is an umbrella term that refers to varieties of pidgin English in any part of the Pacific other than the China coast (Baker 1987: 165). Connections to CPE are debated, but Baker believes that its influence was modest. The British established their first colony in Australia at Botany Bay in 1788, and it was around this time, too, that American whalers had rounded Cape Horn to hunt in the eastern Pacific. From the beginning of the nineteenth century jargonized forms of English were spoken by polyglot crews on British and American vessels, all the more so as islanders were recruited to replace mariners who had perished or deserted. The lucrative market in China for sandalwood brought about sustained contact between Westerners and Pacific islanders, initially in Fiji, then Hawai‘i, the Marquesas, and finally southern Melanesia (Clark 1979: 6, 36). Western traders – frequently the same people who had carried on the sandalwood trade – dealt in a second commodity for Chinese consumption, viz. the trepang, an edible sea slug called the beach-la-mar (Portuguese bicho do mar ‘animal of the sea’, French beˆche-de-mer), the habitat for which is the coral reefs of Melanesia, the Torres Strait, and Micronesia. South Seas Jargon (after Clark 1979) or Pacific Jargon English (after Romaine 1994) is the aggregate of highly variable English-lexified pre-pidgins in the Pacific of the nineteenth century; in Melanesia it took on the names of the goods in trade, whence Sandalwood English and subsequently Beach-la-Mar. Lexical and grammatical features diffused across the region, among them anterior/ past tense been, future by and by, piccaninny ‘child’ (< Portuguese pequenino), plenty ‘much, many’, sabi / savvy ‘know, understand’ (< Portuguese saber), suppose ‘if’, too much ‘very, very much’ (Clark 1979). These features and others are also to be found in Atlantic varieties of Creole English (see further Holm 1986: 274–275), which suggests that knowledge of a maritime jargon could have been present in both the Atlantic and Pacific (cf. Goodman 1985). Baker and Huber (2001: 192), however, look not to an international nautical English but rather to Wanderwo¨rter and the recurring use of foreigner talk registers on the part of Anglophone mariners seeking to achieve communication with local populations at ports of call. Whatever techniques worked on one occasion were likely to be tried again elsewhere. Baker (1987) has found evidence that leads him to claim that New South Wales Pidgin English (NSWPE) was a far more important influence than South Seas Jargon on the way in which varieties of PE developed on the islands prior to the start of a migrant labor system. In contrast to China and 4

Philip Baker has assembled all known attestations of CPE from English language-sources, but the corpus (referenced in Baker 1987:164, Ansaldo et al. 2010: 72, Velupillai 2015: 162) remains unpublished. The nineteenth century saw the appearance of booklets written by Chinese in Chinese characters for instructional purposes, two of which are known to us today (see Ansaldo et al. 2010: 73–79, Li et al. 2005).

839

840

PA U L T. R O B E R G E

the Pacific islands, the Anglophone presence in Australia was continuous from the beginning. As there was a community of interaction with native peoples in the Sydney area of New South Wales, Baker expects a stable PE to have emerged there earlier than anywhere else. From its establishment Sydney was the important center for shipping in the Southwest Pacific, and there was thus the potential for NSWPE to provide something of a model for whites who subsequently visited Pacific islands. Baker does not wish to imply that the island varieties of PPE are continuations of NSWPE, but merely that the attempted means of communication on the part of whites was likely to have been NSWPE more often than any other. Between the 1860s and the early 1900s Melanesians were trafficked (through a practice known as “blackbirding”) and recruited as indentured laborers (typically on three-year contracts) to work the cotton and sugarcane plantations in Queensland and Fiji. Others were taken to German-run copra plantations in Samoa. Labor procurement began in the New Hebrides and Loyalty Islands in the 1860s, then shifted north to the Solomon Islands in the 1870s, and reached New Guinea after 1880 (Baker 1987, Holm 1989: 528). Melanesian PE varieties began to take their present form in the multilingual settings of the plantations. Lacking continuity of interaction, in contrast to NSWPE, PPE of the Southwest Pacific would, for the most part, have remained highly restricted until substantial numbers of islanders returned with a fluent knowledge of Queensland “Kanaka Pidgin English,” or, particularly in parts of Papua New Guinea, Samoan Pidgin English. The repatriation of most laborers favored the further development of PE in their multilingual homeland islands. Today, the three principal dialects of Melanesian PE are extended pidgins in Papua New Guinea, the Solomons, and Vanuatu (the former New Hebrides). Hawai‘i’s encounter with Westerners began with the landing of James Cook and his party in 1778 and widened with the sandalwood trade (through the 1820s), whaling operations in the Pacific (ca. 1820–1860), layovers by Anglo-American ships, Hawaiians signing up as crewmen, and missionary activity. Commercial sugarcane cultivation started in 1835.5 Initially, the workforce was indigenous, but large-scale production would require imported labor. The ravages of foreign disease reduced the Hawaiian population to a fraction of its pre-1778 size. Chinese indentured laborers were brought in to work on American plantations between 1852 and 1897. The majority spoke Cantonese, Yue, and Hakka. Laborers were also imported from other Pacific islands (Micronesia in particular) and from Europe. The resident Portuguese population in Hawai‘i — originally former crewmen on whalers — burgeoned as thousands of workers from Madeira and the Azores came to work on the plantations between 1878 and 1913. Steady migration from Japan began in 1884, from the Philippines in 1907, and to much lesser extent from Korea 5

Other important crops were coffee (from ca. 1828–1829), rice (1858), and later pineapples (from ca. 1885).

Germanic Contact Languages

(1903–1905) and other regions. See Siegel (2000: 199–200) for demographic data. Reinecke (1969: 35, 89–93) believed that a mixed jargon known as hapa haole ‘half white’ came into being as a means of communication between foreign mariners and Hawaiians ca. 1830–1840. Clark (1979) considers it part of the South Seas Jargon. By the start of Chinese immigration, hapa haole had become the workplace lingua franca on Hawaiian plantations. Hawaiians, Chinese, and Portuguese between them had stabilized the form of the pidgin by the mid 1880s, prior to Japanese immigration (Reinecke 1969: 93). In Reinecke’s view, post-1888 immigrants acquired a stable English-lexified pidgin. Goodman (1985) and Holm (1986, 1989: 517–522) expand this view substantially. Roberts (1995, 1998), however, has presented extensive documentary evidence that points to Pidgin Hawaiian as an important form of communication between Hawaiians and Europeans of different language backgrounds and as the principal plantation language until 1876. Workers were segregated by ethnic groups and maintained their ancestral languages. To be sure, a variety of contact English did exist much earlier in the ports, which drew material from English-lexified jargons and perhaps CPE (e.g., kaukau < CPE chow-chow ‘food’). In the last quarter of the nineteenth century immigrants acquired Pidgin Hawaiian, which gradually mixed with pidginized English and was displaced by the latter only in the 1880s and 1890s (Roberts 1995: 46). The prominence of Pidgin Hawaiian may explain the slow development of Hawaiian PE and its late creolization.

35.2.3.3 German in Namibia, China, and New Guinea Germany’s bid for an overseas empire was both late and short-lived, coming as it did only after unification in 1871 and ending with the First World War.6 As Germany was a late entrant in European colonial expansion, lingue franche were already in place in the vast majority of its possessions: Togo, Cameroon, German East Africa (Tanganyika, present-day Burundi and Rwanda), German Micronesia (Caroline, Mariana [except Guam], and Marshall Islands), and German Samoa. Communicative pressures that might lead to the construction of German-based mediums of interethnic communication were low; see Mu¨hlha¨usler (1984) for a useful survey. When Germany took control of South West Africa (present-day Namibia) in 1884, Afrikaans had already become established as a major lingua franca in the southern parts of the territory. But because of the large number of German settlers and their concentration in the central and central-northern areas, a “Kiche Duits” (Ku¨chendeutsch) emerged among Africans around 1900 in the context of the workforce; see Deumert (2003, 2009) for documentation and analyses of restructuring, substrate 6

A bibliography of work on the German language in former German colonies is at www.philhist.uni-augsburg.de /lehrstuehle/germanistik/sprachwissenschaft/Unserdeutsch/publikationen/.

841

842

PA U L T. R O B E R G E

influences, and how speakers reshaped their linguistic repertoires in response to colonial imposition and retreat. Today, the German-speaking population in Namibia is just under 20,000. Vernacular forms show extensive borrowing from English and Afrikaans. What Deumert recorded at the close of the twentieth century “was the end of a long decline of Kiche Duits in Namibia” (2009: 409). The Kiautschou (Jiaozhou) Bay Pachtgebiet was a German-leased territory (1898–1914) in which the northern Chinese city of Tsingtau (Qingdao) served as naval base and port of entry for German goods. Pidgin German appears to have developed mainly through relexification of an earlier English-based pidgin (presumably some form of CPE) but possibly also independently. Of these concurrent strains, only a few fragments are extant (Mu¨hlha¨usler 1980: 167, 175–176, 1984: 32). According to Mu¨hlha¨usler (1984: 34), German New Guinea would seem to have been the best locale for the development of pidginized varieties of colonial German. Large areas of the country had not been contacted by Europeans when the Germans took control (1884); most missions were in German hands; there was no dominant indigenous lingua franca; and there was a plantation and contract labor system in place. In the islands of the Bismarck Archipelago a rudimentary form of Melanesian PE (early Tok Pisin) was in use mainly among repatriated laborers, whereas on the mainland (Kaiser-Wilhelmsland) relatively little PE was spoken. Relexification of PE appears to have occurred in some areas; elsewhere, pidginized and/or jargonized forms of German arose through independent development (Mu¨hlha¨usler 1980: 175). German-based hybrids cropped up around Rabaul (the administrative center), on other government outposts, and on a number of mission stations — particularly geographically isolated ones such as Ali Island — but apparently not on the plantations. The few examples that have been preserved for posterity are given in Mu¨hlha¨usler (1984: 35–36, 2012: 87–95). In spite of favorable conditions, German New Guinea’s principal language of wider communication was Tok Pisin by the time German control had come to an end (1914). Tok Pisin was not restricted to the plantations but was used in a number of domains that created conduits for lexical borrowing from German. Mu¨hlha¨usler (1980: 176–184) lists more than 150 words of German origin in the source material, of which a few have survived.

35.3 Creoles It has been commonly assumed that a creole language “has a jargon or pidgin in its ancestry” and is “spoken natively by an entire speech community, often one whose ancestors were displaced geographically so that their ties

Germanic Contact Languages

with their original language and sociocultural identity were partly broken,” typically under the conditions of slavery in European colonies (Holm 1988: 6). The importance of plantations is often stressed. In addition to providing the requisite setting for intense language contact, they performed the critical social function of giving group identity to people from geographically different areas and of multiple ethnolinguistic backgrounds. Traditionally, creolization has referred to nativization, that is, a process through which a pidgin acquires native speakers (Holm 1988: 7). Yet, the field has distanced itself from the view (e.g., Bickerton 1981) that children are the primary agents in creole formation (cf. Mufwene 2008: 34, 78–83). It is the mechanisms of untutored adult L2 acquisition that are an essential ingredient in creole genesis (Plag 2008: 307). All creoles are grounded in a “crucial episode, however short its duration” (Kihm 2012: 659) that brought together “conventionalized interlanguages of an early developmental stage” (Plag 2008). The role of children is essentially one of regularization, not elaboration (though see Singler 2006 for an insightful essay on this topic). Some researchers have argued that there is no clear evidence that Caribbean creoles — indeed creoles generally — developed out of prior pidgins (e.g., Alleyne 1980, Mufwene 2008: 75–78, Aboh and DeGraff 2016). They posit a largely uninterrupted transmission of language in the early phases of colonial societies and explicitly deny any cycle of reduction and expansion. While the special circumstances of these societies may have accelerated the pace of language change and facilitated significant influences from substrate languages, the difference between creole and noncreole languages lies not in the linguistic processes that produced them but in the outcomes of the same processes. One such model (Mufwene 2008, 2015) proceeds from the idea of a feature pool. Creoles formed by the same competition-and-selection sequence as other colonial vernaculars that also diverge from their metropolitan congeners, such as Que´be´cois French. It would, on this view, be reasonable to suppose that creoles are socially disfranchised dialects of their lexifiers rather than new languages in any meaningful sense. Proponents of the pidgin-to-creole cycle counter that the Pacific and Australia show direct evidence for pidgins and/or reduced varieties that later creolized. Despite the lack of written evidence, the linguistic facts strongly suggest that Atlantic creoles arose as structurally reduced varieties. Linguistic features that are characteristic of a restricted but dynamic system (Section 35.2 above) can be expected to carry over into a developing creole (Baker 2001; McWhorter 2005: 38; 2011: 31; Bakker 2014a: 187–188). McWhorter (2005: 10, 38; 2011: 6) takes this idea a step further. Prototypical creole languages uniquely combine three traits — little or no inflectional morphology, little or no use of tone to distinguish monosyllabic lexical items or grammatical categories, little or no noncompositional

843

844

PA U L T. R O B E R G E

derivation — that are predicable from their history in pidginization. McWhorter’s trifecta and his claim that creole grammars are “the world’s simplest grammars” (2005: 38) are provocative and warmly disputed (e.g., Aboh and DeGraff 2016). The extent to which creoles differ in fundamental ways from other languages (“creole exceptionalism”) constitutes a special problem in the field. Creolists will agree that there are similarities that require explanation, most saliently in word order, TMA systems, and a tendency toward analycity. But a growing body of descriptive and systematic comparative work (cf. Holm and Patrick 2007, Michaelis et al. 2013a, Velupillai 2015) reveals that cross-creole structural similarities are “neither so numerous nor so profound nor so unexpected as to require extraordinary theories” (Singler 2008: 346). For Blasi et al. (2017), the number of features that are passed along from ancestral languages calls into question the notion of a “transmission bottleneck” in the history of creoles that would explain commonalities among them. Language transmission is resilient even under extreme sociolinguistic conditions. Basic SVO order in virtually all attested creoles may be due to the fact that it is overwhelmingly present in their European lexifiers (Portuguese, Spanish, French, English, and Dutch [with V2]). The expression of TMA by invariant preverbal markers may be historically linked to similar structures in the substrate languages, rather than reflective of some kind of typology in general (cf. Holm 1988: 147; Mufwene 2008: 147, 2015: 354). Nonetheless, testing the hypothesis that creoles have sufficient structural affinities to merit classification as a distinct group is an ongoing project. Using recently developed computational tools to establish phylogenetic networks, Bakker et al. (2011) have argued that creoles cluster separately from noncreoles on the basis of their structure. But their approach has been challenged on both empirical and methodological grounds (cf. Mufwene 2015: 353, Aboh and DeGraff 2016: 424–430, and response in Bakker 2014a). For present purpose, we shall, following Mufwene (2008: 33, 2015: 349), regard creole languages as a group of vernaculars that emerged from the seventeenth through the nineteenth centuries under similar geographic, demographic, and economic conditions, viz. in (usually) tropical colonies settled by Europeans who typically spoke nonstandard varieties of their metropolitan languages and who put in place economies that utilized nonindigenous slave (in some cases indentured) labor. We shall assume further that the ensuing linguistic encounters required people to create a basic medium for interethnic communication (MIC) that was appropriate for the circumstances. The exercise was not one of targeted second language acquisition but rather language construction. To paraphrase Thomason and Kaufman (1988: 178), reduction and simplification of the lexical source language by people who did not know it could play no role at all, because one cannot reduce or simplify what one does not know. Speakers created

Germanic Contact Languages

and elaborated the MIC they needed not only by drawing on the range of existing resources available in all of the languages known to the participants, but also by innovating. In time the speech of locally born speakers may become a more advanced form of the MIC and from there what Baker calls a medium for community solidarity (Baker 2000: 54). In this conceptualization the “classical” distinction between pidgins and creoles based on whether they are the native language of some of their speakers serves no useful purpose (Baker 2000 footnote 48, Singler 2006: 159).

35.3.1 English-Lexified Atlantic Creoles Of the Atlantic creoles, English-lexified varieties represent the largest group. They are surveyed exhaustively in Holm (1989: 432–485) and selectively in Michaelis et al. (2013b). They are structurally similar to one another and to Atlantic creoles based on other European lexifiers, due at least in some measure to the typology of the Niger-Congo languages (primarily the Kwa and West Benue-Congo families) that constituted their substrata. It is reasonable to suppose that what is shared among creoles in a specific region and with a common lexifier might be readily explained as a consequence of diffusion rather than parallel independent development. Hancock (e.g, 1986) sought to show that Krio, spoken today in Sierra Leone, and English-lexified creoles of the New World are descendants of a pidgin that formed along the Upper Guinea coast in the early seventeenth century. Domestic unions between Anglophone males and African females led to the emergence of creole communities. Enslaved persons awaiting shipment overseas acquired some of what Hancock called Guinea Coast Creole English (at least in pidginized form) from multilingual AfroEuropeans and grumettoes (Africans in the employ of resident whites and creoles). Once arrived at their destinations, the captives found themselves in environments that brought them in contact with speakers of regional dialects of British English and of other African languages, in different proportions in different places. However, there has been considerable resistance (e.g., Huber 1999a) to the idea that creolized varieties of English arose in West Africa before they emerged in the Caribbean region.7 35.3.1.1 English Creoles in the Americas British West Indian colonization began with St. Kitts in 1623 and Barbados in 1627 and then progressed throughout the seventeenth and eighteenth 7

According to McWhorter’s Afrogenesis hypothesis (2000), English- and French-lexified New World creole languages were born as pidgins in West African trade forts and only imported to, rather than emerging on, plantations. McWhorter’s position might seem an extension of Hancock’s “domestic hypothesis,” but there are important differences, e.g., the central role of fort slaves (rather than sale slaves), rejection of Afro-Europeans as a “minor aristocracy” in West Africa, and endorsement of the mainstream view on the formation of Krio (McWhorter 2000: 86–87, 117).

845

846

PA U L T. R O B E R G E

centuries, in some cases by hostile takeover from other European powers (e.g., Jamaica captured by Spain in 1655) and culminating in the acquisition of formal control over Guiana in 1814. Sugarcane was the main crop produced on plantations throughout the Caribbean, although other crops such as coffee, indigo, and rice were grown. Following an experiment with an indentured workforce from England and Ireland in the 1650s, the main source of labor was African slaves until the abolition of the oceanic slave trade in the British empire in 1808 and the institution itself in 1834. Diffusion scenarios that proceed from localized points in the New World would seem more promising than Afrogenesis because they correlate with the actual dispersal of people. Barbados is of primary significance in the spread of English to other parts of the Caribbean, Suriname, and North America. By virtue of its geographic location, Barbados became an important entrepoˆt for the British slave trade, although there is disagreement over whether a creolized variety of English developed on the island during the seventeenth century that was dispersed along with the slaves themselves. Baker (1998) has shifted the focus to St. Kitts, site of the first British settlement in the Caribbean and one starting point for creoles in the region. While there can be no doubt that diffusion was a factor in creole formation in the New World (as it was in the Pacific), “some among the individual creoles vary so much one from the other that limits must be recognized to the power of diffusion as an explanatory device” (Singler 2008: 350.)

35.3.1.2 Caribbean English Creole (CEC) Winford (1993) uses this term to refer to the English-lexifier vernaculars spoken in the region except Suriname. Holm (1989: 444–485) subdivides CEC into two geographical groups. The Eastern group includes Barbados, the Leeward Islands, the Dutch Windward islands, the Virgin Islands, the Commonwealth Windward Islands, Trinidad and Tobago, and Guyana. The Western group comprises Providencia and San Andre´s, Jamaica, the Cayman Islands, Belize, and enclaves along the eastern coasts of Honduras and Nicaragua (and offshore islands), Costa Rica, and Panama. While there is some sociohistorical and linguistic justification for this division, it belies the complexity of the historical settlement patterns and linguistic outcomes. Within CEC a further division is indicated between “conservative” varieties, which are structurally far removed from metropolitan English, and “intermediate” varieties, which more closely approximate it (and might be thought of as extraterritorial dialects of English rather than different languages). The latter include Barbados, Trinidad, Dominica, and St. Lucia in the Eastern group and the Caymans and the Bay Islands (Honduras) in the Western group. In principle the intermediate varieties in at least some locales (notably Barbados and Trinidad, but this is debated) could be the result of decreolization, or they emerged ab ovo due to social conditions that favored a close contact with metropolitan English.

Germanic Contact Languages

Complicating matters further still is the continuum effect that is characteristic of many English-lexified creoles in the Caribbean, Jamaica and Guyana being the paradigm cases. The long-term coexistence of creole and localized varieties of English has given rise to a spectrum of lects ranging from acrolect (the variety closest to the metropolitan language) to basilect (the variety furthest removed), with one or more lects in between (mesolects). Speakers are polylectal, the breadth of their range depending on their social activities and networks. At the same time, variation may be horizontal as well as vertical. Le Page and Tabouret-Keller (1985) stress that creole space is in fact multidimensional. Speakers exhibit the effects of communal and individual acts of identity; they regulate their linguistic output so as to resemble or distance themselves from the various groups with which they interact.

35.3.1.3 Suriname The historical profiles of the Suriname creoles are different from those of the varieties that constitute CEC. The territory was under British control for barely sixteen years (1651–1667). Because Suriname was colonized from Barbados by people who already had experience with large-scale sugar production, the transition from homestead society (socie´te´ d’habitation) to plantation society (socie´te´ de plantation) was brief; and the immigrants (voluntary and involuntary) may have brought with them a developing English-based contact vernacular. The departure of a substantial portion of the Anglophone population (including a part of their slaves) after 1667, when the Dutch took over the colony, had the effect of redirecting the language construction process (Arends 2017: 54–55). The 1680s saw an influx of Dutch settlers along with Portuguese-speaking Sephardic Jews from Brazil and Europe. The cohort of slaves who had remained behind in Suriname passed along the local workforce vernacular to slaves brought in later by the Dutch. For their part, the Dutch adopted their slaves’ vernacular as a command language and contributed lexical items. The early withdrawal of English and the massive importation of slaves from the late seventeenth century ensured a high degree of L1-retention within an African substratum representing three main language clusters: Gbe, Akan, and Kikongo (Arends 2017: 11–13). Marronage — the creation of communities of fugitive slaves in the interior — was a fact of life in the colony since the introduction of slavery (Arends 2017: 76). Maroons would not have shared a common preexisting language that could serve as a lingua franca and did not remain in slavery long enough to acquire more than a limited set of lexical items from the superstrate language. Sranan (also known as Sranan Tongo, Taki Taki) is the direct continuation of this early Surinamese plantation creole that formed ca. 1680–1720 (Migge 2003: 56–57). Six maroon creole dialects split off during this period: Saramaccan and its secondary offshoot Matawai and a group sometimes

847

848

PA U L T. R O B E R G E

referred to as Nengee (Ndyuka, Aluku, Paramaccan, and Kwinti). Due to their relative isolation from external influences, these maroon creoles are said to be more radical than (former) plantation creoles, where radicalness is understood as typological distance from the lexifier language. Modern Sranan is spoken both as an L1 or L2 by some 400,000 speakers in Suriname and western French Guiana (Arends 2017: 10). Cultural autonomy and a strong sense of identity may enable the survival of Saramaccan (25,000 speakers) and Ndyuka (20,000 speakers), but the other creole dialects (collectively ca. 7,500 speakers) are endangered (Arends 2017). The Surinamese creoles have been of special interest to linguists for several reasons. Sranan is the first creole for which there is textual material, starting in 1667 and extending through the eighteenth and nineteenth centuries. The corpus includes Bible translations and other religious texts, judicial records, official documents, descriptions of the linguistic scene by European travelers and expatriates, glossaries, and language primers. Missionaries of the Evangelische Bru¨dergemeine (Moravian Brethren), who had been sent to convert the slaves on St. Thomas (1732) and in Suriname (1735), produced a large body of texts as well as invaluable descriptive work.8 Thanks to a robust philological and demographic record, Sranan has provided an excellent opportunity for linguists to gain some understanding of the longitudinal development of a creole, even if there is disagreement over what that record tells us. While Smith (e.g., 2009) dates the creolization of Sranan at 1660–1665, the work of Jacques Arends (culminating in his posthumously published 2017 book) has emphasized the (relatively) gradual nature of its formation over a period of several generations. On this view, creolization is not only gradual but also a differential affair, taking place at different rates for different domains of syntax. Finally, Saramaccan is sometimes considered the most radical creole. It stands out for its having been lexified by two metropolitan languages — English and Portuguese (plus African, Dutch and Native American lexis) — and for the fact that it makes use of tone contrasts (as does Ndyuka, Holm 1989: 141). Maroon creoles like Saramaccan are assumed to be structurally closer to the varieties they were when they first formed, some 300 years ago than (former) plantation creoles like Sranan (Arends 2017: 18).

35.3.1.4 North America Historically, Gullah (also known as Geechee) has been spoken along the southeastern coast of the United States from the mouth of the St. John’s River in northern Florida to Cape Fear in North Carolina, but in particular on the Sea Islands of South Carolina and Georgia and in 8

Arends (2017: 28–29) inventories published editions of early Sranan and Saramaccan texts alongside the contents of a planned online corpus. One also finds an annotated chrestomathy of orally transmitted songs and folk tales and of written texts (Arends 2017: chapters 6 and 7). The online Suriname Creole Archive (www.suca.ruhosting .nl/) remains under construction.

Germanic Contact Languages

the mainland low country.9 Charleston was settled in 1670 from Barbados by whites and enslaved blacks; the town would serve as a major port of entry in the slave trade. The coastal lowlands saw the emergence of a plantation economy that was based on rice, indigo, and later cotton. In the early years the settlement imported slaves from Barbados and other Caribbean islands; but from 1695 it turned to Africa for its source of labor. Newly arrived Africans acquired English from fellow slaves, not whites, and their interactions ensured an African imprint on phonology, syntax, and lexicon. By the end of the eighteenth century blacks significantly outnumbered whites in South Carolina. In tidewater Georgia (which became a British colony in 1733) the white and black populations reached numerical parity by 1776 (Klein 2013: 139). As we should expect, the resulting creole shows structural affinities with other English-lexified Atlantic creoles. At the same time, the speech community of which it was emblematic would long preserve African lexis, names, and oral culture. In our time in-migration on the part of mainland Americans, commercial development, and depopulation have altered the linguistic ecology in which Gullah formed. Contact with vernacular and standard forms of American English have occasioned both language shift and multidialectal proficiency. Estimates of monolingual speakers range from as few as 350 (Ethnologue) to under 10,000 at the high end (Klein 2013: 139). The Ethnologue fixes the “ethnic population” at 250,000, which happens to be the same figure that Holm (1989: 491) reported for the number of speakers living in the core areas. Lectal stratification — whether the outcome of decreolization or inherent variation dating back to the formative period (cf. Mufwene 2008: chapter 13) — tends to blur linguistic boundaries. For example, people might consider Gullah their heritage language even if they do not actually speak the basilect or have only a passive knowledge. Congressional designation of a Gullah-Geechee Heritage Corridor (2006), translation of the New Testament into Gullah, and cultural pride should contribute to the maintenance of the language. English in the Bahamas forms a continuum between an intermediate creole (Winford 1993: 4) to local noncreole varieties spoken by blacks and whites.10 Longstanding historical ties to the North American mainland were strengthened in 1783 with an influx of British loyalists following the American Revolution, many of them with slaves in tow. Hackert and Huber (2007) show that southern Bahamian Creole English is to be considered a diaspora variety of Gullah.

9

According to Klein (2013: 139), the autoglossonym varies by location. One also encounters the term “Sea Islands Creole” in the professional literature.

10

The Turks and Caicos Islands form part of the archipelago but are a politically and probably linguistically separate group.

849

850

PA U L T. R O B E R G E

35.3.1.5 West Africa Krio is spoken as a first language by some 350,000 people in Sierra Leone and as a lingua franca by some 4 million.11 Although its origin is debated, the prevailing view is that the seeds of Krio lie in the creole Englishes — including Gullah — of resettled former slaves in 1787 and 1792 (the latter group by way of Nova Scotia), who had won their freedom by fighting on the British side during the American Revolution (1775–1783), and by deported Jamaican maroon rebels who arrived in 1800 (also by way of Nova Scotia). Between 1808, when Sierra Leone became a Crown Colony, and 1840, an estimated 60,000 “recaptives”or “liberated Africans” were taken at sea from illegal slave ships by the British navy, landed at Freetown, and “apprenticed” to the creolephone Nova Scotians and maroons, and also to whites. The confluence of vernaculars introduced from across the Atlantic became the vehicular language among the recaptives and the first language of their descendants. Offshoots of Krio are Aku in The Gambia and Pichi, the latter of which is today spoken by some 100,000 people on the island of Bioko (Equatorial Guinea). The Republic of Liberia began as a settlement of the American Colonization Society, which acquired land at the present site of Monrovia, with a view toward relocating manumitted slaves to an African homeland. Between 1822 and the end of the American Civil War in 1865, more than 15,000 freed and free-born blacks emigrated, joined over the years by recaptives and some Afro-Caribbeans. The emigre´s brought with them their creolized vernaculars, the legacy of which is a creole that is the first language of their descendants, who live in and around the capital. Standard and vernacular English, the creole, and pidgin varieties of English spoken in Liberia have all influenced one another and appear to form a continuum (Holm 1989: 421).

35.3.2

English-Lexified Pacific Creoles

The Pacific group of English-lexified contact languages took shape during the nineteenth and early twentieth centuries and consists largely of pidgins and extended pidgins, with creoles only in Australia and Hawai‘i.

35.3.2.1 Australia In Baker’s account (1987), white penetration into Queensland from 1839 led to the development of inland (basically transplanted NSWPE, possibly with new Aboriginal lexical items) and coastal varieties of Queensland Aboriginal Pidgin English (QAPE), with the former overtaking the latter by the 1870s. Aborigines did not work on the plantations, but whites in the Brisbane area, at least, were speakers of QAPE and were likely to have used it when addressing Melanesian contract laborers in their employ. The Queensland 11

Unless otherwise noted, population figures in this article are from Michaelis (2013b).

Germanic Contact Languages

plantations would have brought together speakers of at least two varieties of PE from 1863 on, viz. QAPE and an ill-attested Loyalty Islands PE, the outcome of which was a Queensland “Kanaka English.” Today, Torres Strait Creole (also known as Torres Strait Broken) is spoken on several islands in the Strait, northern Cape York, and along the southwest coast of Papua. According to the Ethnologue, it has approximately 6,400 first-language speakers and is used by some 19,000 people as a second or third language. Kriol is the first language of some 20,000 Aboriginal people in communities across Australia’s Northern Territory and in adjacent areas of Western Australia and Queensland. In its background are varieties of PE that came to be used throughout these regions during the nineteenth century. NSWPE moved with white expansion into inland Queensland; pastoralism brought it from there into the Northern Territory and by the 1870s into Western Australia. Chinese immigrants were a significant presence in the nonindigenous population of the Northern Territory in the late 1880s and are likely to have introduced CPE. There is also the probability that pidginized forms of English developed independently among Aborigines living around coastal settlements prior to the arrival of the pastoralists and Chinese immigrants (see Holm 1989: 541–544 for details). Around 1900 these strands converged into a broad lingua franca, which would become the primary language of the remnants of ethnolinguistically diverse Aboriginal groups.

35.3.2.2 Hawai‘i Hawai‘i Creole (HC) has held a prominent place in creolistics. Its entire history is documented in written records, and it emerged recently enough for linguists to have interviewed speakers whose verbal behavior is potentially representative of its early stages (Roberts 1998: 1–2). Bickerton (e.g., 1981) promoted HC as a paradigm case of abrupt creolization under the agency of children, who, he claimed, drew on an innate module of rules and categories embedded in Universal Grammar when confronted with impoverished, highly variable primary linguistic data. Research by Roberts (1998, 2000, 2005) shows that the development of HC was rather more gradual. Substratal explanations for the origins of HC features are pursued in Siegel (2000) and Roberts (2005). The first generation of children born to immigrant laborers acquired ancestral languages but used Hawaiian PE more widely from the mid-1890s because of the large influx of Asians speaking other languages. After leaving the plantations, many workers settled in multiethnic urban areas, which provided the context for the pidgin’s wider use and elaboration. By the early 1900s, many speakers shifted to Hawaiian PE as their primary language, and their offspring acquired it as their first language. Thus, it was the second generation of locally born children who were the original speakers of the creole, which reached its apotheosis in the 1920s. A postcreole dialect (the autoglossonym for which, like the creole, is “Pidgin”) arose with formal education and contact with speakers of vernacular and Standard English.

851

852

PA U L T. R O B E R G E

35.3.3 Dutch-Lexified Creoles in the New World It is striking that the linguistic footprint of a far-flung Dutch empire — with possessions and outposts in the New World, Africa, and Asia at its apotheosis in the mid-seventeenth century— is relatively small. The Dutch language took root only in those territories where the Dutch were the first European settlers (New Netherland, Guyana, Cape of Good Hope) or where they predominated over other Europeans (the Danish West Indies, since 1917 the US Virgin Islands). The Dutch were neither the first nor the only Europeans to arrive in other areas that they colonized (Suriname, Netherlands Antilles); elsewhere (Brazil) their presence was brief. In Ceylon and the Dutch East Indies dialects of Creole Portuguese — alongside an indigenous pidgin, Bazaar Malay, in the latter venue — were in place to mediate interethnic communication. In the East Indies the policies and attitudes adopted by Dutch colonizers prevented the language from becoming dominant (further on this point in Mu¨hlha¨usler 1996). 35.3.3.1 Danish West Indies When Denmark claimed St. Thomas in 1672, there were large numbers of settlers from neighboring islands already there, mainly Dutch planters and slaves. St. John was settled from St. Thomas in 1717. Negerhollands (Virgin Islands Creole Dutch) became established on both islands. St. Croix, which was settled in 1735, appears to have been mainly Anglophone, although Negerhollands was carried there, too. In the nineteenth century language shift in the direction of English was under way. Negerhollands went into inexorable decline and has since died out completely, its last speaker having passed in 1987. Yet, the history of Negerhollands is among the better documented of the world’s creole languages. Moravian and Danish Lutheran missionaries produced translations of scripture, religious texts, and descriptive work (including the first published monographic grammar of a creole language, Grammatica over det Creolske sprog, by J. M. Magens, 1770). What the missionaries have preserved for posterity appears to be an acrolectal “high creole” that was characteristic of locally born Euro-Caribbeans and provided the basis for a more Netherlandicized version for evangelical and educational purposes.12 Representative of the coexistent Afro-Caribbean basilect are the twentieth-century texts collected by Josselin de Jong (1926) and fieldwork among the very last speakers in the 1970s and 1980s; see Sabino (2012) and Bakker (2014b) on the varietal disjunction in our source material.

12

Sabino (2012, chapter 4) surveys the Negerhollands source material and scholarship harking back to the turn of the twentieth century. Linguistically relevant documents are collected in Van Rossem and Van der Voort (1996) and online at https://corpus1.mpi.nl/ (under NEHOL) and https://diecreoltaal.com/.

Germanic Contact Languages

35.3.3.2

Guiana

There are reports of “creole Dutch” in Guiana (present-day Guyana) in the writings of nineteenth-century resident Europeans. But it was not until 1973 when Ian Robertson confirmed the existence of not one but two creole languages lexically related to Dutch (Robertson 1979). Dutch settlements along the Essequibo (1618) and Berbice (1627) rivers saw the development of discrete creole vernaculars during the seventeenth century. The establishment of a colony on the neighboring Demerara River in 1740 by English planters from Barbados and the Leeward Islands marked the beginning of the spread of English creole. The 1763 slave rebellion in Berbice forced the relocation of upriver populations and estates to the coast, where there was increased contact with English and the developing Creolese (Guyana Creole English). The Dutch formally ceded Essequibo, Berbice, and Demerara to Great Britain in 1814. These territories were consolidated into a single colony, British Guiana, in 1831. The population of mixed ancestry that remained in the interior continued to speak Dutch creole, but language shift in favor of English creole was in motion from the mid nineteenth century. Kouwenberg based her comprehensive grammar of Berbice Creole Dutch (1994) on data elicited from eleven competent speakers born between 1895 and 1923. The creole diverges radically from other Atlantic creoles in that it bears the imprint of a single West African language, with Eastern Ijo words in the core vocabulary and inflectional affixes in the TMA system and nominal pluralization. Berbice Dutch became extinct with the death of its last speaker in 2005. Skepi Dutch, once widely spoken along the Essequibo River (whence the name) and a wholly separate development, was only sketchily documented before its last rememberers died in the 1990s. The published corpus consists of about a dozen sentence fragments (collected in Bakker 2014b: 216–218) and 200 lexical items (Robertson 1989).

35.3.4 Unserdeutsch Unserdeutsch was first reported by Volker (1982) and is the only known creole with German as its lexifier. It is unusual in that it arose during the 1890s in the hostels of a mission school and orphanage for children of mixed ethnic backgrounds in Vunapope, near Rabaul in German New Guinea. Standard German was the language of instruction and interaction with mission staff. Many of the older pupils spoke Tok Pisin along with their ancestral languages, to which they often had lost contact. Among themselves, the pupils devised a pidginized German. As adults, they tended to marry one another and settle in the area. The in-group code became the home language of a small community within a single generation, though not necessarily its only language. The position of Standard

853

854

PA U L T. R O B E R G E

German weakened after the Australian occupation of New Guinea in 1914, even though it did hang on as a teaching medium at the school alongside English until World War II. Geographic dispersion — most Unserdeutsch speakers relocated to Australia after Papua New Guinea received independence in 1975 — and exogamy accelerated a shift to English and/or Tok Pisin as home languages. With perhaps a hundred speakers and not all fluent, Unserdeutsch is severely endangered (see Maitz and Volker 2017: 384–392 for a cogent assessment). Unserdeutsch shows structural similarities to Tok Pisin, which suggests that relexification was involved (Mu¨hlha¨usler 1984: 36–41). It also has grammatical features that are common with many L2-varieties of German, including Foreign Worker German (Maitz and Volker 2017: 377). In spite of its stability there is considerable variation between speakers and within individual speakers (on which see Maitz and Volker 2017: 383–384).13

35.4 Extended Pidgins Extended (expanded, full) pidgins — called pidgincreoles by Bakker (2008) — constitute a class in between pidgins and creoles. An extended pidgin has become functional in domains beyond its original range of use. It serves as the primary means of communication in a multilingual speech community, particularly in the context of urbanization; in time it may also become the first language for some of its speakers. Structurally, extended pidgins become elaborated commensurately with their functionality. They are more like creoles insofar as they typically display SVO order, preverbal TMA markers, some innovated morphology, and reduplication. Clusters of extended pidgins are located in the highly heterogeneous regions of West Africa and Melanesia. The West African Pidgin English continuum includes varieties spoken in Ghana (5 million speakers), Nigeria (75 million), and Cameroon (maybe 2 million users of varying degrees of proficiency, as per the Ethnologue), all of which are descendants of Krio (cf. Huber 1999b: 75–134). The region through which Melanesian PE (Section 35.2.3.2) has spread is one of the most linguistically diverse on the planet. Tok Pisin, spoken by some 3 to 5 million people, is today one of three official languages (alongside English and Hiri Motu) of Papua New Guinea. It is used in the print and broadcast media, government, and primary education.14 Bislama (the autoglossonym obtains from Beach-laMar, supra) is official in Vanuatu (alongside English and French) and is 13

The scholarly literature can be accessed from the University of Augsburg Unserdeutsch Research Project website: www .philhist.uni-augsburg.de/lehrstuehle/germanistik/sprachwissenschaft/Unserdeutsch/.

14

On the expansion of Tok Pisin see Mühlhäusler (1997: 163–186), Mühlhäusler and Wurm (1985) remains a valuable handbook. A collection of texts from the earliest attestations (1840s) to the twentieth century can be found in Mühlhäusler, Dutton, and Romaine (2003).

Germanic Contact Languages

widely spoken by the islands’ 218,000 inhabitants. Pijin claims some 341,400 L1- and L2-users in the Solomon Islands but does not have official status (Ethnologue).

35.5 Bilingual Mixed Languages So-called mixed languages are languages with split ancestry that have emerged in situations of community bilingualism. The mechanisms underlying their formation are debated (wholesale relexification, intertwining of grammar of one source language with the lexicon of the other, matrix language turnover); so too is their status as new, autonomous languages vis-a`-vis language maintenance with adoption of contact material or language shift with retention of heritage language material. Mixed languages usually serve as salient markers of new or retained group identity. Some mixed languages show a lexicon-grammar split, whereby the core lexicon is derived from one source language and the grammatical elements from the other. Angloromani, spoken by traveling Roma in the United Kingdom and other Anglophone countries, is of the first type, with Romani lexis and an English grammatical frame. It is not used in sustained conversation or narration but is restricted to individual utterances and used in group-internal interactions (Matras 2010: 148). Other Germanic languages that mix with Romani include German and Swedish.15 Shelta (Irish lexis, English structure) of the Irish travelers in the British Isles and other Anglophone nations has also been identified as a mixed language (Grant 1994, Velupillai 2015: 380–386). During the Dutch East Indies era unions between Dutch men and local women produced Eurasian offspring (“Indos”) who took on new identities marked by new languages. These are Javindo (Dutch lexis, Javanese grammar), which became extinct in the mid twentieth century (De Gruiter 1994), and Petjo (Dutch lexis, Malay grammar), whose numbers had dwindled to a few elderly rememberers in the Netherlands by the mid 1990s (Van Rheeden 1994). Also of the grammar-lexicon type is Old Helsinki Slang (Swedish lexis, Finnish morphosyntax) created by young male street gang members ca. 1890–1950 (Jarva 2008, but see also De Smit 2010 for a different perspective). Other mixed languages show a grammatical split such that the nominal and verbal systems obtain from different source languages. In this category is Gurindji Kriol spoken by the Gurindji people in northern Australia. It derives its lexicon from both source languages (Meakins 2012).

15

Matras (2010) is a book-length treatment of Angloromani. See also the Romani Project website at http://romani .humanities.manchester.ac.uk/.

855

856

PA U L T. R O B E R G E

35.6 Noncanonical Contact Languages In view of the gradient nature of language contact in its social and linguistic aspects, some linguists have discerned the need for an intermediate category between creole and noncreole. There are languages that superficially resemble creoles in some way but upon closer examination have entirely different sociohistorical profiles. In Mu¨hlha¨usler’s view (1998: 304–305) this observation comes close to characterizing Pitcairn Island English and its offshoot on Norfolk Island, which should not be called a creole (pace Holm 1989: 546–551). Platt (e.g., 1975: 372) has characterized the basilectal form of Singapore English commonly known as “Singlish” as a creoloid, which is a cover term for varieties that exhibit linguistic properties that are similar to creoles (e.g., morphological simplification relative to a source language, substrate influences, stratification) but have not passed through a prior stage of reduction; they may also serve as MICs (similarly Trudgill 1996: 8–9). Semicreoles are partially restructured contact languages that combine features of creoles with those of noncreoles but were not necessarily basilectal creoles, since a language can become a semicreole by borrowing from substrate languages (Holm 1988: 9–10, 2004: 5–6; McWhorter 2011: 14–15). Putative examples are Afrikaans (Holm 1989: 338–350, 633, 2004; McWhorter 2011: 15), African American Vernacular English (AAVE) (Holm 1989: 498–503, 633; 2004), and even English itself (McWhorter 2011: 205). If one draws a distinction between these named classes, a prototypical creoloid would be an L2 with significant restructuring due to exogenous agentivity (Trudgill 1996, Mesthrie 2008:270–271). Mesthrie (2008) applies the term language shift varieties to former creoloids whose speakers subsequently undergo language shift with limited access to the target language. Clear-cut examples are South African Indian English and Hiberno-English.16 A semicreole would be an L1 with creolelike contact effects. English is not a creole by any conventional definition of the term; see Mufwene (2008: 45–49) and Roberge (2010: 419–422) for references and discussion of proposals that invoke creolization in some sense. McWhorter (2011: chapter 9) seeks to make a case that in addition to linguistic impositions occasioned by the Norse (especially) and Norman invasions of England, “English grammar is deeply impacted by Celtic in a way that no other Germanic language is, such that depending on where one draws the proverbial line, it could be seen as both moderately simplified and moderately mixed, and therefore . . . a semicreole” (p. 205). It is not certain that proponents of the Celtic hypothesis would be prepared to go that far. In the 1960s and 1970s there was considerable interest in the question of whether AAVE is traceable to creole English formerly spoken widely in the 16

Arguments for heavy influence from Irish on Hiberno-English – but not creolization – are put forward in a detailed study found in Hickey (1997).

Germanic Contact Languages

antebellum American South and now almost completely decreolized, save for some vestigial features such as copula absence (He sick). Although the evidence has not established the existence of a once widespread plantation creole, this does not rule out the possibility of substrate effects due to contact with creole speakers during periods of internment of African slaves in the Caribbean prior to transfer to North America. Rickford (1999) has identified AAVE features that are widespread in Gullah and CEC but rare or nonexistent in British English dialects. Afrikaans is akin to a chimera. One’s assessment depends on which part of the organism one examines. Afrikaans exhibits deflexion and features that are traceable to Khoekhoe, Malay, and Creole Portuguese alongside traits that are prototypically Germanic (basic SOV order with V2phenomena). The establishment of a permanent Dutch East India Company base on the southern coast of Africa in 1652 saw the development of an inchoate fort colony, with complex economic and social relationships between Europeans and the indigenous Khoekhoe. But the disrupted linguistic ecology took a different trajectory with the formal introduction of nonindigenous slave labor in 1658. The Company’s area of operation was in and around the Indian Ocean, and it turned east for most of its slaves — the Indonesian archipelago, the Indian subcontinent, the Mascarenes, Madagascar, and Mozambique. The nearly autonomous slave community that resided in or passed through Cape Town adopted and transformed a Dutch-Khoekhoe jargon (on which see the collected essays in Den Besten 2012) into a MIC within the labor caste, which included not only enslaved but also wage and indentured indigenous workers. As the Cape slave society changed from the urban/household type to a more rurally based plantation system from ca. 1770, an elaborated medium of caste socialization had taken root in the port and spread to the arable regions inland. It was an extended pidgin that was emblematic of a labor caste, within which multiple group identities coexisted. Adaptation on the part of the labor caste to European culture and linguistic norms varied considerably in degree and quality, depending on local conditions. Mutual accommodation at the interface between the expanded pidgin and a coterritorial settler variety of Dutch produced a Cape Dutch Vernacular, which itself showed a high degree of variation. Standard Afrikaans is a focused variety that took shape ca. 1875–1925 as a project of language entrepreneurs in the service of political empowerment. In the absence of agreed-upon linguistic criteria for determining degrees of creoleness (if there is such a thing), classification of contact languages as creoloids and/or semicreoles does not mean much (cf. Thomason 2001: 195, Mu¨hlha¨usler 1998, Mufwene 2008: 40–43). Static synchronic comparison with languages traditionally called creoles tells us nothing about longitudinal development.

857

858

PA U L T. R O B E R G E

It has been suggested that the linguistic processes that regulated the formation of Pennsylvania German and Yiddish might be profitably compared to those that have given rise to contact languages in the strict sense of the term. Enninger (1986) has pondered whether aspects of pidginization might be a feature of Pennsylvania German, but his Fragestellung is essentially a thought experiment. The sociohistorical conditions underlying the formation of Yiddish are incommensurable with those observed in the case of pidgins and creoles (cf. Fishman 1987, Holm 1989: 611, Prince 2001). Wexler’s (e.g., 2002) hypothesis that Eastern Yiddish arose as a kind of mixed language via the relexification of Judeo-Slavic by German-speaking Ashkenazic Jews is radical and controversial.

35.7 Concluding Remark “Es gibt keine vo¨llig ungemischte Sprache,” opined Schuchardt (1884: 5) long before creolistics came into its own as a subdiscipline of linguistics. In his sense all languages are “mixed” to the extent that they contain elements of diverse historical origin. Ultimately, the boundaries between “ordinary” and contact languages, and between pidgins, extended pidgins, and creoles, will prove hard to fix.

References Aboh, E. and M. DeGraff 2016. “A null theory of creole formation based on Universal Grammar.” In I. Roberts (ed.), The Oxford Handbook of Universal Grammar. Oxford University Press: 401–458. Alleyne, M. C. 1980. Comparative Afro-American: An Historical-Comparative Study of English-Based Afro-American Dialects of the New World. Ann Arbor: Karoma. Ansaldo, U., S. Matthews, and G. Smith 2010. “China Coast Pidgin: Texts and contexts,” Journal of Pidgin and Creole Languages 25: 63–94. Arends, J. 2017. Language and Slavery: A Social and Linguistic History of the Suriname Creoles. Amsterdam: John Benjamins. Baker, P. 1987. “Historical developments in Chinese Pidgin English and the nature of the relationships between the various pidgin Englishes of the Pacific region,” Journal of Pidgin and Creole Languages 2:163–207. Baker P. 1998. “Investigating the origin and diffusion of shared features among the Atlantic English creoles.” In P. Baker and A. Bruyn (eds.), St. Kitts and the Atlantic Creoles: The Texts of Samuel Augustus Mathews in Perspective. London: University of Westminster Press: 315–364. Baker P. 2000. “Theories of creolization and the degree and nature of restructuring.” In I. Neumann-Holzschuh and E. W. Schneider (eds.),

Germanic Contact Languages

Degrees of Restructuring in Creole Languages. Amsterdam: John Benjamins: 41–63. Baker, P. 2001. “No creolization without prior pidginization?” Te Reo 44: 31–50. Baker, P. and M. Huber 2001. “Atlantic, Pacific, and world-wide features in English-lexicon contact languages,” English World-Wide 22: 157–208. Bakker, P. 1989. “A French-Icelandic nautical pidgin,” Journal of Pidgin and Creole Languages 4: 129–132. Bakker, P. 2008. “Pidgins versus creoles and pidgincreoles.” In Kouwenberg and Singler (eds.): 130–157. Bakker, P. 2014a. “Creolistics: Back to square one?” Journal of Pidgin and Creole Languages 29: 177–194. Bakker, P. 2014b. “Three Dutch creoles in comparison,” Journal of Germanic Linguistics 26: 191–222. Bakker, P., A. Daval-Markussen, M. Parkvall, and I. Plag 2011. “Creoles are typologically distinct from non-creoles,” Journal of Pidgin and Creole Languages 26: 5–42. Bakker, P. and M. Mous (eds.) 1994. Mixed Languages: 15 Case Studies in Language Intertwining. Amsterdam: IFOTT. Bakker, P. and Y. Matras (eds.) 2013. Contact Languages: A Comprehensive Guide. Berlin: De Gruyter Mouton. Besten, H. den 2012. Roots of Afrikaans. Amsterdam: John Benjamins. Bickerton, D. 1981. Roots of Language. Ann Arbor: Karoma. Blasi, D. E., S. M. Michaelis, and M. Haspelmath 2017. “Grammars are robustly transmitted even during the emergence of creole languages,” Nature Human Behaviour 1: 723–729. Broch, I. 1996. “Solombala-English in Archangel.” In Jahr and Broch (eds.): 93–98. Broch, I. and E. H. Jahr 1984a. Russenorsk: Et Pidginspraº k i Norge, 2nd edn. Oslo: Novus. Broch I. and E. H. Jahr 1984b. “Russenorsk: A new look at the Russo-Norwegian pidgin in northern Norway.” In P. Sture Ureland and I. Clarkson (eds.), Scandinavian Language Contacts. Cambridge University Press: 21–65. Clark, R. 1979. “In search of Beach-la-Mar: Towards a history of Pacific Pidgin English,” Te Reo 22: 3–64. Clyne, M. G. 1968. “Zum Pidgin-Deutsch der Gastarbeiter,” Zeitschrift fu¨r Mundartforschung 35: 130–139. Deumert, A. 2003. “Markedness and salience in language contact and second-language acquisition: Evidence from a non-canonical contact language,” Language Sciences 25: 561–613. Deumert, A. 2009. “Namibian Kiche Duits: The making (and decline) of a Neo-African language,” Journal of Germanic Linguistics 21: 349–417. Enninger, W. 1986. “Pennsylvania German: A pidgin(ized) variety?” In N. Boretzky, W. Enninger, and T. Stolz (eds.), Beitra¨ge zum 2. Essener

859

860

PA U L T. R O B E R G E

Kolloquium u¨ber “Kreolsprachen und Sprachkontakte.” Bochum: N. Brockmeyer: 41–82. Fishman, J. A. 1987. “Post-exilic Jewish languages and pidgins/creoles: Two mutually clarifying perspectives,” Multilingua 6: 7–24. Gilbert, G. G. and P. Pavlou 1994. “Gastarbeiterdeutsch ‘Foreign Workers’ German’: An industrial pidgin.” In C. A. Blackshire-Belay (ed.), The Germanic Mosaic: Cultural and Linguistic Diversity in Society. Westport, CT: Greenwood Press: 147–154. Goodman, M. 1985. Review of Bickerton 1981, International Journal of American Linguistics 51: 109–137. Grant, A. 1994. “Shelta: The secret language of Irish travellers viewed as a mixed language.” In Bakker and Mous (eds.): 123–150. Gruiter, M. de 1994. “Javindo, a contact language in pre-war Semarang,” in Bakker and Mous, (eds.): 151–159. Hackert, S. and M. Huber 2007. “Gullah in the diaspora: Historical and linguistic evidence from the Bahamas,” Diachronica 24: 279–325. Hancock, I. 1986. “The domestic hypothesis, diffusion, and componentiality: An account of Atlantic Anglophone creole origins.” In Muysken and Smith (eds.): 71–102. Hancock, I. 1996. “The special case of Arctic pidgins.” In Broch and Jahr (eds.): 15–29. Heidelberger Forschungsprojekt “Pidgin Deutsch” 1975. Sprache und Kommunikation ausla¨ndischer Arbeiter, Analysen, Berichte, Materialien. Kronberg/Ts.: Scriptor Verlag GmbH & Co. KG. Hickey, R. 1997. “Arguments for creolisation in Irish English.” In R. Hickey and S. Puppel (eds.), Language History and Linguistic Modelling: A Festchrift for Jacek Fisiak on his 60th Birthday. Berlin: Mouton de Gruyter: 969–1038. Holm, J. 1986. “Substrate diffusion.” In Muysken and Smith (eds.): 259–278. Holm, J. 1988–1989. Pidgins and Creoles, 2 Vols. Cambridge University Press. Holm, J. 2004. Languages in Contact: The Partial Restructuring of Vernaculars. Cambridge University Press. Holm J. and P. L. Patrick (eds.) 2007. Comparative Creole Syntax: Parallel Outlines of 18 Creole Grammars. United Kingdom and Sri Lanka: Battlebridge. Huber, M. 1999a. “Atlantic English Creoles and the Lower Guinea coast: A case against Afrogenesis.” In M. Huber and M. Parkvall (eds.), Spreading the Word: The Issue of Diffusion among the Atlantic Creoles. London: University of Westminster Press: 81–110. Huber, M. 1999b. Ghanaian Pidgin English in its West African Context: A Sociohistorical and Structural Analysis. Amsterdam: John Benjamins. Huber, M. 2013. “Ghanaian Pidgin English.” In Michaelis et al. (eds.): 167–175. Jahr, E. H. 1996. “On the pidgin status of Russenorsk.” In Jahr and Broch (eds.): 107–122. Jahr, E. H., and I. Broch (eds.) 1996. Language Contact in the Arctic: Northern Pidgins and Contact Languages. Berlin: Mouton de Gruyter.

Germanic Contact Languages

Jarva, V. 2008. “Old Helsinki Slang and language mixing,” Journal of Language Contact 1: 52–80. Josselin de Jong, J. P. B. de 1926. Het huidige Negerhollandsch. Amsterdam: Koninklijke Akademie van Wetenschappen. Kihm, A. 2012. Review of McWhorter 2011, Language 88: 657–660. Klein, T. B. 2013. “Gullah.” In Michaelis et al. (eds.): 139–147. Kouwenberg, S. 1994. A Grammar of Berbice Dutch Creole. Berlin: Mouton de Gruyter. Kouwenberg, S. and J. V. Singler (eds.) 2008. The Handbook of Pidgin and Creole Studies. Oxford: Wiley-Blackwell. Lehiste, I. 1965. “A poem in Halbdeutsch and some questions concerning substratum,” Word 21: 55–69. Le Page, R. B. and A. Tabouret-Keller 1985. Acts of Identity. Cambridge University Press. Li, M. 2016. “Trade pidgins in China: Historical and grammatical relationships,” Transactions of the Philological Society 114: 298–314. Li, M. and S. Matthews 2016. “An outline of Macau Pidgin Portuguese,” Journal of Pidgin and Creole Languages 31: 141–183. Li, M., S. Matthews, and G. P. Smith 2005. “Pidgin English texts from the Chinese-English Instructor,” Hong Kong Journal of Applied Linguistics 10: 79–168. Maitz, P. and C. A. Volker 2017. “Documenting Unserdeutsch: Reversing colonial amnesia,” Journal of Pidgin and Creole Languages 32: 365–397. Matras, Y. 2010. Romani in Britain: The Afterlife of a Language. Edinburgh University Press. McWhorter, J. H. 2000. The Missing Spanish Creoles: Recovering the Birth of Plantation Contact Languages. Berkeley and Los Angeles: University of California Press. McWhorter, J. H. 2005. Defining Creole. Oxford University Press. McWhorter, J. H. 2011. Linguistic Simplicity and Complexity: Why Do Languages Undress? Berlin: De Gruyter Mouton. Meakins, F. 2012. “Which mix — code-switching or a mixed language? — Gurindji Kriol,” Journal of Pidgin and Creole Languages 27: 105–140. Mesthrie, R. 2008. “Pidgins/creoles and contact languages: An overview.” In Kouwenberg and Singler (eds.): 263–286. Michaelis, S. M., P. Maurer, M. Haspelmath, and M. Huber (eds.) 2013a. The Atlas of Pidgin and Creole Language Structures. Oxford University Press. Michaelis, S. M., P. Maurer, M. Haspelmath, and M. Huber (eds.) 2013b. The Survey of Pidgin and Creole Languages, Vol. 1: English-Based and DutchBased Languages. Oxford University Press. Migge, B. 2003. Creole Formation as Language Contact: The Case of the Suriname Creoles. Amsterdam: John Benjamins.

861

862

PA U L T. R O B E R G E

Mitrović, P. 1972. “Deux sabirs balkaniques,” Linguistique 8: 137–140. Mitzka, W. 1923. Studien zum baltischen Deutsch. Marburg: Elwert. Mufwene, S. S. 2008. Language Evolution: Contact, Competition and Change. London: Continuum. Mufwene, S. S. 2015. “The emergence of creoles and language change.” In N. Bonvillain (ed.), The Routledge Handbook of Linguistic Anthropology. London: Routledge: 348–365. Mu¨hlha¨usler, P. 1980. “German as a contact language in the Pacific (with special reference to the influence of German on Tok Pisin),” Michigan Germanic Studies 6: 163–189. Mu¨hlha¨usler, P. 1984. “Tracing the roots of Pidgin German,” Language and Communication 4: 27–57. Mu¨hlha¨usler, P. 1996. “Dutch in the Pacific area.” In Wurm et al. (eds.), 2: 339–343. Mu¨hlha¨usler, P. 1997. Pidgin and Creole Linguistics, 2nd edn. London: University of Westminster Press. Mu¨hlha¨usler, P. 1998. “How creoloid can you get?” Journal of Pidgin and Creole Languages 13: 355–371. Mu¨hlha¨usler, P. 2012. “Sprachliche Kontakte in den Missionen auf Deutsch-Neuguinea und die Entstehung eines Pidgin-Deutsch.” In S. Engelberg and D. Stolberg (eds.), Sprachwissenschaft und kolonialzeitlicher Sprachkontakt: Sprachliche Begegnungen und Auseinandersetzungen. Berlin: Akademie Verlag: 71–100. Mu¨hlha¨usler, P., T. E. Dutton, and S. Romaine (eds.) 2003. Tok Pisin Texts: From the Beginning to the Present. Amsterdam: John Benjamins. Muysken P. and N. Smith (eds.) 1986. Substrata versus Universals in Creole Genesis. Amsterdam: John Benjamins. Nortier, J. and M. Dorleijn 2013. “Multi-ethnolects: Kebabnorsk, Perkerdansk, Verlan, Knakensprache, Straattaal, etc.” In Bakker and Matras (eds.): 229–271. Pfaff, C. W. 1981. “Incipient creolization in Gastarbeiterdeutsch? An experimental sociolinguistic study,” Studies in Second Language Acquisition 3: 165–178. Plag, I. 2008. “Creoles as interlanguages: Syntactic structures,” Journal of Pidgin and Creole Languages 23: 307–328. Platt, J. T. 1975. “The Singapore English speech continuum and its basilect ‘Singlish’ as a ‘creoloid’,” Anthropological Linguistics 17: 363–374. Prince, E. 2001. “Yiddish as a contact language.” In N. Smith and T. Veenstra (eds.), Creolization and Contact. Amsterdam: John Benjamins: 263–289. Reinecke, J. E. 1969. Language and Dialect in Hawai‘i: A Sociolinguistic History to 1935. Honolulu: University of Hawai‘i Press. Rheeden, H. van 1994. “Petjo: The mixed language of the Indos in Batavia.” In Bakker and Mous (eds.): 223–237.

Germanic Contact Languages

Rickford, J. R. 1999. African American Vernacular English: Features, Evolution, Educational Implications. Malden, MA: Blackwell. Roberge, P. T. 2010. “Contact and the history of Germanic languages.” In R. Hickey (ed.), The Handbook of Language Contact. Oxford: Wiley-Blackwell: 406–431. Roberts, J. M. 1995. “Pidgin Hawaiian: A sociohistorical study,” Journal of Pidgin and Creole Languages 10: 1–56. Roberts, S. J. 1998. “The role of diffusion in the genesis of Hawaiian Creole,” Language 74: 1–39. Roberts, S. J. 2000. “Nativization and the genesis of Hawaiian Creole.” In J. McWhorter (ed.), Language Change and Language Contact in Pidgins and Creoles. Amsterdam: John Benjamins: 257–300. Roberts, S. J. 2005. The Emergence of Hawai‘i Creole English in the Early 20th Century. Doctoral dissertation, Stanford University. Robertson, I. 1979. Berbice Dutch – A Description. Doctoral dissertation, University of the West Indies, St. Augustine. Robertson, I. 1989. “Berbice and Skepi Dutch: A lexical comparison,” TNTL 105: 3–21. Romaine, S. 1994. “Germanic creoles.” In E. Ko¨nig and J. van der Auwera (eds.), The Germanic Languages. London and New York: Routledge: 566–603. Rossem, C. van and H. van der Voort (eds.) 1996. Die Creol Taal: 250 Years of Negerhollands Texts. Amsterdam University Press. Sabino, R. 2012. Language Contact in the Danish West Indies: Giving Jack his Jacket. Leiden: Brill. Schreier, D. and M. Hundt (eds.) 2013. English as a Contact Language. Cambridge University Press. Schuchardt, H. 1884. Slawo-Deutsches und Slawo-Italienisches. Graz: Leuschner & Lubensky. Siegel, J. 2000. “Substrate influence in Hawai‘i Creole English,” Language in Society 29: 197–236. Singler, J. V. 2006. “Children and creole genesis,” Journal of Pidgin and Creole Languages 21: 157–173. Singler, J. V. 2008. “The sociohistorical context of creole genesis.” In Kouwenberg and Singler (eds.): 332–358. Smit, M. de 2010. “Modelling mixed languages: Some remarks on the case of Old Helsinki Slang,” Journal of Language Contact 3: 1–19. Smith, N. S. H. 2009. “English-speaking in early Surinam?” In R. Selbach, H. C. Cardoso, and M. van den Berg (eds.), Gradual Creolization: Studies Celebrating Jacques Arends. Amsterdam: John Benjamins: 305–326. Stammler, W. 1922. “Das Halbdeutsch der Esten,” Zeitschrift fu¨r deutsche Mundarten 17: 160–172. Thomason, S. G. 2001. Language Contact: An Introduction. Washington, DC: Georgetown University Press.

863

864

PA U L T. R O B E R G E

Thomason, S. G and T. Kaufman 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley and Los Angeles: University of California Press. Trudgill, P. 1996. “Dual-source pidgins and reverse creoloids: Northern perspectives on language contact.” In Jahr and Broch (eds.): 5–14. Velupillai, V. 2015. Pidgins, Creoles, and Mixed Languages: An Introduction. Amsterdam: John Benjamins. Volker, C. A. 1982. An Introduction to Rabaul Creole German (Unserdeutsch). Master of Literary Studies dissertation, University of Queensland. Wexler, P. 2002. Two-Tiered Relexification in Yiddish: Jews, Sorbs, Khazars and the Kiev-Polessian Dialect. Berlin: Mouton de Gruyter. Whinnom, K. 1971. “Linguistic hybridization and the ‘special case’ of pidgins and creoles.” In D. Hymes (ed.), Pidginization and Creolization of Languages. Cambridge University Press: 91–115. Winford, D. 1993. Predication in Caribbean English Creoles. Amsterdam: John Benjamins. Wurm, S. A. and P. Mu¨hlha¨usler (eds.) 1985. Handbook of Tok Pisin (New Guinea Pidgin). Canberra: Department of Linguistics, The Australian National University. Wurm, S. A., P. Mu¨hlha¨usler, and D. T. Tryon (eds.) 1996. Atlas of Languages of Intercultural Communication in the Pacific, Asia, and the Americas, 2 Vols. Berlin: Mouton de Gruyter.

Index

A’-movement, 436, 437, 441, 442, 452 abbreviation, 254, 255 ablaut, 193, 198, 201, 205, 206, 207, 210, 216, 226, 238, 248, 251, 478 adjacency, 301, 314, 325, 326, 327, 329, 342, 350, 360, 399 adverb, 241, 245, 327, 328, 346, 349, 349, 354, 360, 368, 380, 428, 430, 442, 471, 482, 520, 521, 524, 586 affixation, 75, 76, 89, 108, 198, 853 case, 288 complementizer agreement, 314, 316, 318–24, 329, 333 contact languages, 855 contact material or language, 855 gender, 276–7 inflectional morphology of nouns, 215, 221 inflectional morphology of verbs, 199–201, 212, 591–2, 600 prosodic effects, 53, 156–7 voice, 475, 480, 484 word formation, 238, 245–50, 254 affixoids, 244, 245 Afrikaans, 2, 16, 25, 26, 34, 39, 44, 195, 198, 199, 201, 203, 204, 205, 206, 214, 241, 253, 255, 267, 268, 274, 282, 288, 300, 339, 438, 446, 464, 471, 476, 482, 822 Amish Alsatian German, 818 Amish Swiss German, 818 A-movement, 304, 426, 427, 429, 430, 431, 466, 467, 469, 470, 474, 675 Anabaptist groups, 810, 818, 820, 821, 823, 824, 829 anaphoric elements, 259, 297, 462, 477, 493, 494 SELF versus SE, 502 anticausatives, 478, 479, 480, 481, 482, 483, 484, 487 apocope, 67, 226, 751, 752 Arabic, 690, 714, 722, 723 Aramaic, 239 argument structure, 6, 292, 293, 296, 358, 461, 499, 509, 521, 527, 528, 532, 539, 618, 637, 702 Armenian (Eastern), 128

articles, 217, 234, 259, 260, 265, 272, 288, 341, 351, 353, 538, 550, 554, 788, 835 definite articles, 162, 265, 550, 679, 753 indefinite articles, 547, 550, 559, 753, 787 spurious indefinite articles, 554, 558 aspect, 6, 199, 200, 204, 391, 569, 571, 575, 577, 591, 596, 611, 637, 692, 835 aspiration, 120, 121, 125, 128, 129, 131, 132, 133, 135, 136, 137 assimilation, 11–16, 22, 23, 39, 56, 105, 121, 126, 127, 128, 129, 131, 133, 134, 136, 137, 138, 201, 232, 320 authority, 578 Barbiers et al., 313 Bavarian Gmunden Bavarian, 494 Egerlandish Bavarian, 495 Bayer, 330 Berber, 722 bilingual mixed languages, 834, 855 binding domain, 497, 500 Binding Theory, 303, 493, 496–502, 508 blending, 238, 254, 255 Boef, 332 borrowing, 248, 249, 250, 833, 842, 856 boundary tone, 145, 146, 154, 158, 171, 174, 176, 177, 178, 181, 184, 668, 670 breaking, 219, 226, 232 Cantonese, 124 Caribbean English Creole, 846 Carstens, 333 case, 5, 263, 282, 342, 351, 353, 389, 390, 391, 611, 676 binding and, 494, 505, 512 dialectal variation in West Germanic, 751, 753 gender and, 265–7, 276–7 genitive case and linking elements, 243 in connection with object shift and scrambling, 426, 429–32 in connection with particles, 516 inflectional morphology of nouns, 214–21, 223–4, 226, 228, 231, 233, 234 in heritage Germanic languages, 787–91, 796

866

INDEX

case (cont.) in relationship to noun and determiner phrase, 539–41, 548–9, 552, 556 in relationship to second language acquisition, 690, 692, 695, 699–700, 705 in relationship to voice, 462, 465, 467, 471, 474, 476–7, 481, 486 lack of case in pidgins, 835 with prepositions, 601, 623, 634, 635 Chinese Pidgin English, 838 Cimbrian, 817 clause-boundedness, 437 clitics, 314, 316, 321, 324, 333, 679 coalescence, 11, 27 Colonia Tovar German, 816 complementary quantity, 100, 101, 104, 114 complementizer agreement, 6, 313 complementizer phrase (CP), 365, 367, 370, 371, 377, 384, 390, 396, 408, 421, 422, 445, 552, 696, 697 compounding, 5, 91, 157, 238, 239, 540, 835 consonant clusters, 33, 34, 42, 57, 59, 86, 106, 107 contact language, 2, 3, 133, 833–858 controllability, 578 conversion, 211, 212, 238, 250, 254 coronals, 12, 14, 18, 19 counterfactuals, 204, 568, 569, 570, 584, 585, 586 Cremers & Van Koppen, 331 creoles, 2, 3, 36, 40, 592, 606, 607, 834, 842 Culminativity, 668, 672 Danish, 3, 4, 34, 37, 75, 76, 78, 86, 89, 92, 99, 109, 132, 143, 153, 154, 173, 174, 175, 178, 180, 185, 195, 198, 203, 205, 214, 220, 221, 238, 240, 246, 252, 255, 259, 263, 264, 266, 272, 289, 300, 301, 344, 371, 375, 379, 416, 417, 418, 421, 424, 427, 440, 444, 453, 454, 455, 463, 469, 475, 487, 508, 558, 718, 719, 720, 722, 723, 724, 725, 761, 768, 770, 776 Colloquial Copenhagen Danish, 439 Danish dialects, 265, 267, 763 Danish in Germany, 824–825 Middle Danish, 439 Western Jutish, 274 De Haan, 320, 332 De Haan & Weerman, 332 declension class, 214, 262, 275, 276, 277, 278, 787 definiteness, 218, 220, 259, 264, 266, 284, 468, 473, 532, 550, 556, 558, 560, 678, 787, 791 double definiteness, 220, 554, 797, 798 degemination, 25, 39, 98, 99, 109, 201 deixis, 551, 591 deletion, 20, 24, 34, 38, 49, 56, 59, 60, 66, 76, 128, 137, 160 deletion of constituents, 674, 679 deletion of pitch accents, 679 deontic modality, 572, 579, 581, 586 Determiner Phrase (DP), 537 discontinuous DPs, 557 doubly filled DP, 558 DP-movement, 555, 558 typological differences in the DP, 559 diphthongization, 750 directed motion, 533 dissimilation, 16, 130 doubly filled Comp, 439

downward entailment, 644, 645, 646, 647, 649, 650, 651 Dutch, 2, 3, 12, 13, 19, 20, 22, 23, 24, 25, 26, 27, 34, 39, 41, 42, 43, 44, 49, 50, 52, 53, 54, 60, 61, 64, 68, 76, 78, 79, 80–84, 86, 99, 109, 112, 114, 119, 124, 125, 126, 127, 128, 129, 137, 172, 173, 174, 177, 180, 181, 184, 195, 196, 197, 201, 203, 204, 205, 207, 208, 209, 210, 211, 214, 215, 228, 240, 241, 242, 243, 245, 249, 250, 251, 253, 255, 262, 265, 273, 282, 288, 300, 305, 306, 313, 339, 346, 347, 348, 351, 353, 353, 354, 357, 382, 397, 413, 417, 419, 422, 424, 425, 427, 428–429, 431, 432, 439, 440, 446, 455, 464, 467, 468, 471, 473, 476, 477, 479, 482, 486, 503–508, 615, 690–691, 702 Brabantic, 319–320, 321 De Panne Dutch, 328 Groningen Dutch, 275 Hellendoorn Dutch, 319, 321, 332 Katwijk Dutch, 318 Lapscheure Dutch, 326, 327, 329 Low Franconian dialects, 741 Maastricht Dutch, 148 Middle Dutch, 196 Nieuwkerken-Waas Dutch, 323 Straattaal, 714 Tegelen Dutch, 326, 330 Tilburg Dutch, 113, 114 effect-on-output, 431, 432 embedded V2, 377, 378, 379, 381, 444 English, 2, 3, 16, 19, 22, 23, 27, 34, 35, 36, 37, 38, 39, 42, 43, 44, 76, 78, 79, 80–84, 99, 109, 119, 120, 121, 123, 125–126, 168, 172, 174, 177, 180, 183, 198, 199, 200, 201, 202, 203, 204, 205, 207, 208, 209, 210, 211, 215, 229, 238, 239, 241, 243, 245, 249, 251, 254, 256, 267, 274, 282, 288, 290, 299, 301, 304, 306, 350, 367, 368, 371, 439, 441, 448, 453, 454, 455, 464, 465, 467, 469, 476, 477, 482, 493, 497–500, 615 African American English, 438 American English, 28, 37, 58, 111, 128, 253 American Vernacular Black English, 714 Appalachian English, 438 Belfast English, 439 British English, 34 East Anglian English, 38 Irish English, 438 Middle English, 230, 267, 306, 367, 374, 439 Old English, 49, 196, 201, 208, 230, 267, 268, 285, 306 South African English, 35 Southern US English, 127 Upper Midwestern US English, 128 Welsh English, 38 epenthesis, 16, 18, 29, 41, 42, 160, 161, 162, 201 epistemic modality, 401, 568, 571, 572, 578, 580, 581, 582, 586 ethnolect, 715 events, 591, 592, 594, 596, 597, 601, 604 exceptional case marking (ECM), 389, 462, 486, 508 expletive, 299, 342, 343, 356, 357, 438, 445, 455, 468, 469, 471, 474, 483, 484, 486, 540, 576, 676

Index exponence, 201, 539 extraction from verb-second clauses, 443, 444 extrametricality, 49, 73, 74, 75, 77, 78, 79, 80, 82 Faroese, 3, 4, 22, 24, 34, 37, 44, 51, 75, 77, 91, 104, 195, 196, 197, 198, 203, 205, 209, 214, 215, 256, 262, 272, 287, 300, 344, 351, 375, 444, 455, 462, 464, 466, 467, 468, 776 Faroese dialects, 763 Flemish, 439 West Flemish, 315, 325, 380, 468 Focus Prominence Principle, 670 focus-sensitive particles, 654, 680 foot, 49, 73, 98, 149, 159, 178 formal fronting, 442, 677 fortis versus lenis consonants, 59, 102, 107, 132, 135, 151 French, 44, 77, 126, 239, 240, 248, 250, 264, 378, 456, 690, 695, 703, 705, 825, 833, 836, 844, 854 Frisian, 2, 6, 13, 16, 17, 20, 21, 22, 23, 25, 26, 34, 38, 39, 41, 50, 69, 129, 132, 144, 196, 200, 203, 204, 205, 206, 208, 209, 210, 211, 214, 215, 238, 241, 243, 245, 251, 255, 256, 313, 314, 316, 320, 321, 322, 325, 331, 332, 339, 346, 347, 355, 356, 380, 440, 446, 479, 482, 736, 756, 810, 812, 813, 821, 823 East Frisian Saterlandic, 262 North Frisian, 262, 264, 267, 824 Old Frisian, 51, 66 Wangeroog Frisian, 67 Weser Frisian, 67 West Frisian, 134, 137, 226, 264, 318 Wursten Frisian, 67 Fuß, 321, 324, 328, 331 geminate consonants, 45, 97, 98, 100, 101, 102, 104, 107, 114, 135 derived, 108 word-initial, 35, 102 geminates, 19, 25 gender, grammatical, 214, 259 gender, onymic, 275 German, 2, 3, 12, 13, 14, 18, 19, 21, 25, 28, 34, 35, 36, 39, 42, 43, 45, 49, 50, 52, 53, 54, 60, 61, 67, 68, 76, 78, 79, 80–84, 86, 98, 99, 109, 119, 120, 122, 123, 124, 125, 128, 172, 173, 174, 175, 180, 181–183, 195, 196, 197, 201, 202, 203, 204, 205, 206, 208, 209, 210, 211, 214, 215, 220, 223, 238, 240, 241, 242, 243, 246, 247, 249, 252, 254, 256, 259, 262, 263, 264, 269, 271, 272, 276, 282, 286, 287, 289, 290, 292, 293, 299, 301, 303, 304, 306, 313, 339, 340, 347, 348, 351, 352, 353, 357, 371, 382, 402, 425, 431, 432, 437, 438, 440, 442, 442–444, 446, 447–448, 451, 452, 453, 455, 463, 465, 467, 468, 471, 473, 476, 477, 479, 482, 483, 484, 485, 493, 500–502, 615, 690–691, 693, 702 Arzbach German, 146, 147, 152 Austrian German, 78 Bavarian German, 14, 16, 17, 28, 98, 288, 316, 317, 328, 330, 331 Central Bavarian, 99, 104, 107, 114 colloquial, 288

Cologne German, 40, 146, 148–150 Early New High German, 277 East Low German, 63 Eastphalian, 63 German in Denmark and Belgium, 824–825 German in Nambia, 826 Hessian, 57 Hessian substractive plurals, 57 High German dialects, 20 Highest Alemannic, 20 Kiel German, 17, 29 Kiezdeutsch, 714 Low Alemannic, 64 Low German, 2, 15, 266, 818–819 Low German dialects, 742 Lower Bavarian, 321 Lower Silesian, 814 Mayen German, 144 Middle German (Central German), 54, 135, 736, 737, 743, 750, 751, 754, 755, 816, 819 Middle High German, 45 Middle Low German, 220 Moselle Franconian, 15 North Low Saxon, 144 North Saxon, 56, 64 Northern Bavarian, 56 Northern German, 40, 172 Odenwald German, 59 Old High German, 225, 277 Southern Bavarian, 17, 29 Southern German, 36, 172, 198, 203 Swabian German, 380 Swiss German (Alemannic), 28, 35, 98, 114, 129, 135, 137, 380, 440, 450 Swiss German (Thurgovian), 102, 104 Swiss German (Zurich), 102, 104, 329 Tyrolean German, 38 Upper German, 744 Upper Saxon, 63 Western German, 270 Westerwald Hessian, 57 Westphalian, 15, 63 givenness, 661, 664, 665, 668, 669, 673, 674, 678, 679, 680, 682 Goeman, 330 Gothic, 2, 34, 39, 40, 45, 49, 129, 196, 197, 198, 201, 202, 206, 207, 210, 298, 440 grammatical functions, 283, 291, 301, 302, 304, 351, 353, 528 grammaticalization, 204, 217, 220, 225, 246, 464, 597, 637 Greek, 16, 36, 77, 239, 242, 248, 249, 391, 722 Gruber, 324, 328 Haeberli, 329 Haegeman, 315, 329 Haegeman & Van Koppen, 326, 329 Hebrew, 38, 239 Hoekstra & Mara´cz, 324 Hoekstra & Smits, 314, 322 Holmberg’s generalization, 418, 419, 422, 423, 431 Hungarian, 124 Hunsrik, 819–821 Hutterisch, 819–821

867

868

INDEX

iamb, 21, 73, 74 Icelandic, 2, 3, 4, 18, 24, 34, 37, 51, 75, 77, 91, 104, 123, 125, 129, 131–132, 144, 169, 173, 174, 177, 183, 195, 197, 198, 199, 201, 203, 204, 205, 206, 209, 214, 215, 218, 262, 272, 282, 290, 298, 299, 301, 303, 304, 306, 344, 349, 351, 353, 358, 374, 375, 378, 413, 416, 418, 423, 426, 430, 438, 444, 455, 462, 464, 465, 467, 470, 472, 473–474, 477, 480, 484, 512, 776 Icelandic dialects, 765 implicit argument, 464, 473, 478, 481, 621 incomplete acquisition, 784, 791, 793, 801 individuation hierarchy, 274, 275 infinitive splitting, 397 infinitives, 200, 244, 251, 303, 339, 389 inflectional morphology adjectival inflection, 43, 543, 547, 556 loss of inflection, 799 nouns, 214 verbs, 193 zero inflection, 215, 217, 221, 223, 225, 226, 229, 230, 234 information structure, 547, 661–682 interlanguage, 690, 835, 837, 843 intonation, 143, 148, 149, 152, 161, 167, 173, 424, 428, 544, 657, 678, 720, 728 declaratives, 174 interrogatives, 145, 146, 148, 150, 151, 155, 172, 179 intonation phrase, 158 intonational phrase, 154, 158, 167, 668 questions, 179 inversion, 179, 321, 322, 438, 673, 675, 693, 694, 695, 696, 718 noninversion, 718 IPP (Infinitivus pro participio = infinitive for a participle), 347, 348, 356 Italian, 77, 690 Japanese, 123 jargons, 835, 839, 841, 842, 857 Kashaya, 128 Kashmiri, 128 Kurdish, 722 language and aging, 800 language contact, 833–858 language endangerment and vitality, 807 language identity, 723, 724, 726, 728, 811, 812, 821, 843, 847, 848, 855 language maintenance, 807 language shift, 133, 807, 849, 852, 853, 855, 856 languages and dialect (distinguishing between) 807 laryngeal features, 11, 119 Latin, 77, 239, 248, 249, 305, 391 learner varieties, 698, 699 left periphery, 321, 323, 330, 438, 441 lenis consonants, 103, 107, 132, 150 fortis-lenis distinction, 102 lenition (weakening), 27, 28, 50, 58 lexical accent, 672 lexical distribution, 156 lexical tone, 148, 158, 173 Limburgish, 818–819 linguistic distances, 766, 768, 776, 822

loanwords, 36, 37, 38, 39, 42, 44, 75, 77, 91, 103, 113, 135, 148, 210, 243, 785, 786, 798 locality, 424, 437, 448, 449, 451, 646, 714 logophor, 493, 497, 498, 500, 501, 503, 506, 514, 798 long-distance reflexive, 504 Luxembourgish, 2, 4, 214, 215, 223, 262, 737, 744, 753, 810, 812 Malay, 253 Maxim of Quantity, 431 Mayr, 332 media, 714, 715, 717, 724, 726, 742, 774, 823, 825, 826, 829, 854 Mennonite Low German (Plautdietsch), 2 middles, 481–487 minority languages, 2, 4, 7, 34, 784, 800, 807 Moche´no, 817 monophthongization, 751 mood, 193, 196, 197, 513, 568, 581, 583, 586, 799, 835 moribund languages, 4, 784, 788, 794, 799, 816 multiethnolect, 714, 715, 720, 721, 725, 726, 727, 728 mutual intelligibility, 4, 717, 736, 761, 763, 764, 765, 770, 776, 811, 815, 835 negation, 199, 374, 418, 432, 482, 639, 640, 643, 646, 677, 692, 835 negative polarity item, 639 neutralization, 5, 27, 103, 104, 119, 128, 130, 132, 134, 136, 137, 138, 146, 154 nonce word experiment, 80 Nonfinality, 73, 74, 75, 77, 78, 82, 83, 84, 85, 87, 92, 178 Norwegian, 3, 4, 12, 13, 26, 27, 29, 34, 36, 40, 41, 42, 43, 44, 76, 78, 79, 84–87, 104, 120, 129, 131, 143, 153, 155, 156, 174, 179, 181, 185, 195, 203, 205, 208, 238, 241, 288, 289, 300, 344, 405–406, 439, 440, 444, 455, 463, 469, 472, 475, 477, 485, 508, 768, 770, 776 Bokmaº l, 4, 198, 220, 263, 264, 273 Hallingdalen Norwegian, 374, 375 Heritage Norwegian, 796 Kebab Norwegian, 714 Mid Gudbrandsdal, 101 Nord-Gudbrandsdalen, 68 North Gudbrandsdal, 101 Norwegian dialects, 763 Nynorsk, 4, 198, 203, 262 Old East Norwegian, 66 Urban East Norwegian, 157 Vaº gaº Norwegian, 68 numerals, 239, 245, 255, 284, 287, 288, 538, 539, 547, 548 object shift, 6, 306, 349, 354, 399, 413, 675, 677 Old Norse, 37, 195, 196, 220, 222, 268 Old Saxon, 45, 196 open syllable lengthening, 98, 99, 104, 109 order preservation, 422, 423 ¨ vdalian, 814 O Pacific Pidgin English, 839 parasitic morphology, 403, 405, 407 participles in connection with word order, 342, 346, 354, 403, 404

Index inflectional morphology of, 199, 201–3, 205–6, 209, 211, 216 in relationship to tense aspect, 605, 607–10 in relationship to voice, 461–3, 469, 473, 476–7, 485 in sentence prosody, 669 in second language acquisition, 691 in word formation, 244, 248, 251 particles, 199, 418, 440, 516, 615, 661, 667, 680, 691, 693, 695, 722, 723 Formation/decomposition of, 241–248 In relation to VO- / OV- ordering, 342–359 In relation with modality, 571–586 In relation with aspect, 591, 595 In relations with polarity items, 641–643 passive, 198, 203, 205, 296, 299, 358, 383, 430, 461–477, 676 adjectival, 463, 465, 480 eventive, 463, 477 get-passives, 297, 475 long passive, 402, 405–406, 408 s-passive, 199, 475 tenses, 464 path, 615 Pennsylvania Dutch, 198 Pennsylvania German, 2, 195, 203, 819–821 periphrasis, 194, 202 phonological activity, 123, 125, 127, 136 phonotaxis, 25, 34, 38, 41, 46, 835 pidgins, 834, 842, 843 pitch accent, 143, 152, 155, 169, 171, 174, 175, 176, 177, 178, 184, 185, 661, 662, 663, 667, 669, 670, 671, 673, 680, 681, 682, 721 pitch contour, 169, 171 place, 615 Plautdietsch, 823–824 plural formation, 49, 53, 65, 68, 226, 228, 228, 232, 233 polarity sensitivity, 641, 651 positive polarity item, 639, 655 possessive reflexive, 493, 496, 508, 510, 512 postposition, 623, 627, 628, 633, 634, 635 Pra¨teritumsschwund (loss of simple preterite), 753 pre-determiners, 552 prepositions, 13, 199, 205, 290, 343, 351, 352, 353, 450, 464, 483, 497, 498, 499, 500, 507, 544, 601, 615, 643, 692, 789, 835 in relation to verbal particles, 516–532 morphological formation and decomposition of, 238–252 privativity, 124, 136 pro-drop, 314, 321, 324, 329, 333 Prokosch’s Law, 84 prolepsis, 450–451, 452 pronouns, 28, 134, 196, 197, 344, 357, 394, 402, 465, 520, 595, 621, 626, 631, 703, 745, 753 case properties of, 283–306 grammatical gender of, 259–275 in connection with Scrambling and Object Shift, 415–431 in connection with the internal structure of DPs, 543–558 in connection with unbounded dependencies, 439–456 in heritage grammar contexts, 787–796 in relation to complementizer agreement, 315–330

in relation to information structure requirements, 670–680 in relation to nominal inflectional morphology, 217, 220 in relation to VO- / OV-base ordering, 340, 357 reflexive pronouns, 198, 474, 525, 529, 531 reflexive versus nonreflexive, 493–514 SE-reflexive pronouns, 479, 480, 483, 487 prosodic morphology, 253, 254 prosodic structure, 75, 80, 168, 277, 661, 664, 668, 669 prosodic template, 49, 53, 57, 60, 61, 254 prosody, 5, 6, 49, 114, 167, 228, 661–682, 693 Proto-Germanic, 194, 206, 210, 211, 214, 277, 282 Proto-Indo-European, 197, 200, 206, 260, 283 Proto-Norse, 219 psycholinguistics, 674, 694, 699 Punjabi, 690 quantity, 35, 43, 49, 65, 74, 91, 97–114, 144, 151 quantity sensitivity, 51, 73, 74, 75, 76, 79, 81, 82, 90, 91, 92 quasi-incorporation, 241, 242 questions, 167, 168, 169, 178, 179, 181, 183, 184 quirky case, 283, 301, 302, 304 quirky subjects, 342, 356, 426, 466, 467 receptive multilingualism, 268, 771, 773, 777 reduplication, 53, 206, 207, 238, 252, 835, 854 reflexivity, 493, 508 relative clauses, 439, 440, 441, 445, 447, 450, 453, 454, 542, 543, 555, 701, 789, 793, 795 restructuring, 6, 389, 396 resultatives, 6, 342, 345, 359, 522, 529, 530, 531, 533 resumption, 448–450 Rich Agreement Hypothesis (RAH), 378, 379, 383 Romance, 38, 42, 46, 75, 77, 92, 124, 135, 230, 239, 240, 264, 379, 417, 475, 703, 770, 775, 790 Sanskrit, 16 Schiepek, 317 Scots, 126, 817 scrambling, 6, 306, 342, 350, 359, 398, 399, 403, 413, 438, 667, 675, 682 second language acquisition, 7, 185, 689–706, 844 sentence processing, 699 Serbian, 722 Slavic, 46, 124, 133, 134, 136, 230, 239, 475, 742, 770, 775, 815 Somali, 124 sonority, 18, 27, 36, 38, 41, 42, 43, 45, 46, 105 space, 615 Spanish, 690 special contours, 181 speech act modality, 578, 579, 584, 586 states, 591, 596, 597, 604 strengthening, 11, 27 stress, 101, 215, 654, 670, 720 aspiration rule, 125 compound stress, 157 contrastive stress, 548 emphatic stress, 506 foot structure and, 49, 51–3, 56, 61, 65–7, 69 in connection with intonation, 170–1, 173, 178

869

870

INDEX

stress (cont.) in connection with tone accent, 143, 150–1, 154, 156–7, 161 level stress, 66, 67, 68 main stress, 73, 74, 75, 76, 91, 160, 201, 214, 246, 249 quantity and, 98, 101, 106–7, 112–3 role in phonological processes, 21, 23, 25–8 role in syllable structure, 41 secondary stress, 87, 160 sentence stress (nuclear stress), 169, 177, 179, 669, 673, 681 stress clash, 160 word stress, 73–92 strong islands, 453, 454 subject-verb agreement, 325, 694, 697, 698, 699 subjunctive, 193, 195, 196, 197, 201, 202, 203, 204, 211, 513, 569, 582, 583, 584, 585, 586, 607, 799 superiority, 675 Swedish, 2, 3, 4, 21, 23, 28, 30, 34, 36, 76, 78, 86, 99, 104, 105, 114, 129, 130–131, 143, 153, 155, 156, 170, 172, 174, 179, 181, 195, 198, 201, 203, 204, 205, 208, 209, 211, 214, 220, 222, 242, 256, 263, 272, 288, 300, 349, 403, 420–421, 439, 440, 449, 454, 455, 462, 468, 469, 475, 485, 487, 508, 768, 770, 776 Central Swedish, 153, 154, 156, 158 Malmo¨ Swedish, 157 Old Swedish, 66 ¨ vdalian, 374 O Rinkeby Swedish, 714, 716 Sollero¨n Swedish, 144 Swedish dialects, 763 Swedish in Finland, 826 Varieties of Swedish, 99 syllable weight, 52, 60, 62, 78, 79, 80, 81, 86, 97, 98 syncope, 160 syncretism, 195, 211, 216, 217, 285, 286, 287, 607 syntactic domains, 271 tense, 125, 126, 128, 390, 391, 404, 405, 418, 419, 420, 461, 464, 591, 595, 611, 753, 755, 787, 835 ablaut, 248 absolute, 606, 607 acquisition of, 692, 697 complex, 595, 602, 611 finite and nonfinite, 606 in relation to modality, 567–583 morphological aspects of, 193–206 periphrastic, 602, 611 present perfect, 608 relative, 606, 607 simple, 594, 611 subjunctive, 195 tense affixes, 19, 119, 206, 322 that-trace effects, 455 thematic roles, 283, 295, 464, 476, 477, 499 tone accent, 5, 143, 168, 179 tone accents synchronic typology, 155 topic, 340, 370, 428, 429, 437, 444, 455, 657, 661, 665, 668, 672, 675, 681 topic islands, 444 topicalization, 345, 346, 358, 420, 424, 553, 557, 675, 676–678, 717, 792

in unbounded dependency constructions, 437–9, 441–2, 450, 453, 456 Topicalization Phrase (TP), 365, 368, 398, 405, 408, 441 transitivity, 618, 620, 629 trochee, 19, 21, 25, 49, 73, 74, 75, 77, 83, 91, 92, 98, 149, 224, 225, 227, 229, 276 Turkish, 690, 714, 722, 723 umlaut, 62, 110, 202, 215, 219, 220, 223, 226, 228, 229, 235, 277, 283, 752 unrounding of front rounded vowels, 751 Unserdeutsch, 2, 853 urban speech styles, 714–729 Urdu, 722 V°-to-T° movement, 366, 373–384 Van Craenenbroeck & Van Koppen, 319, 322 Van der Meer, 332 Van Koppen, 318, 319 Vanacker, 330 verb clusters, 242, 343, 348, 360, 396, 398, 399, 403 verb (finite), 365 verb movement, 306, 354, 378, 418, 421, 698, 793 verb phrase (VP), 339–361, 365, 398, 420, 424, 470, 666, 669, 671, 677, 697 verbal clusters, 355 voice, 193, 198, 407, 461–487 voicing, 11, 35, 39, 60, 119, 151, 201, 229, 230 vowel balance, 50, 65, 69 vowel reduction, 11, 27, 50, 54, 60, 266 vowel tenseness, 109, 110, 113 Weber, 329 Weise, 317 Weiß, 313, 324 wh-copying, 446 wh-islands, 448, 449, 453 wh-movement, 427, 437, 438, 439, 441, 442, 445, 450, 456 word order, 306, 307, 339–361, 398, 418, 419, 462, 463, 465, 506, 523, 531, 661, 667, 674, 675, 678, 679, 692, 791, 844 DP-internal word order, 542–544 head-final, 339–361 head-initial, 339–361 in relation to prosodic intonation, 177, 179, 181 In second language acquisition (SLA), 690, 693, 698, 700–1, 717 object-verb, 339–361 Placement of finite verbs, 377–380 V2, 368, 676 verb-object, 339–361 Wymysorys, 816 Yiddish, 12, 13, 16, 22, 23, 25, 34, 36, 38, 39, 40, 129, 133–134, 137, 195, 196, 198, 200, 201, 203, 204, 205, 206, 209, 239, 241, 253, 264, 287, 339, 340, 374, 375, 378, 419, 425, 431, 438, 450, 472, 558, 559, 823–824 Central Yiddish, 20, 41 West Yiddish, 41 Zwart, 330, 332 Zwicky, 324 Zwicky & Pullum, 324