1,647 76 5MB
English Pages [884] Year 2020
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
-Ø
−
+
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
-Ø
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
V°
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
C°
V2 = V°-TO-T°-TO-C° MOVEMENT
TP
FINITE VERB
DP
T'
SUBJECT
T°
VP
V°
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
C°
TP-spec
AdvP
V°
V°
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
C°
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.
Að
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
C°
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'
C°
TP
SUB. CONJ.
DP
V°-TO-T° MOVEMENT
T'
SUBJECT
T°
VP
FINITE VERB
AdvP MEDIAL ADVERBIAL
VP
V°
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
T°
Jo´n
AdvP V°
V°
DP
oft
hafi borðað to´mata
oft
borðað to´mata
(kemur flestum a´ o´vart.)
b.
Að
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.)
<