Word-Formation. Volume 1 Word-Formation: An International Handbook of the Languages of Europe 9783110246254, 9783110246247

Table of contents :
Acknowledgements
Introduction
Contents
I. Word-formation as a linguistic discipline
1. The scope of word-formation research
2. Word-formation research from its beginnings to the 19th century
3. Word-formation in historical-comparative grammar
4. Word-formation in structuralism
5. Word-formation in inhaltbezogene Grammatik
6. Word-formation in onomasiology
7. Word-formation in generative grammar
8. Word-formation in categorial grammar
9. Word-formation in natural morphology
10. Word-formation in cognitive grammar
11. Word-formation in optimality theory
12. Word-formation in construction grammar
13. Word-formation in psycholinguistics and neurocognitive research
II. Units and processes in word-formation I: General aspects
14. The delimitation of derivation and inflection
15. Units of word-formation
16. Derivation
17. Conversion
18. Backformation
19. Clipping
20. Composition
21. Blending
22. Incorporation
23. Particle-verb formation
24. Multi-word expressions
25. Reduplication
26. Word-creation
27. Allomorphy
III. Units and processes in word-formation II: Special cases
28. Affective palatalization in Basque
29. Parasynthesis in Romance
30. Affix pleonasm
31. Interfixes in Romance
32. Linking elements in Germanic
33. Synthetic compounds in German
34. Verbal pseudo-compounds in German
35. Particle verbs in Germanic
36. Particle verbs in Romance
37. Particle verbs in Hungarian
38. Noun-noun compounds in French
39. Verb-noun compounds in Romance
40. Co-compounds
41. Multi-word units in French
42. Multi-word expressions and univerbation in Slavic
43. Compounds and multi-word expressions in Slavic
44. Paradigmatically determined allomorphy: the “participial stem” from Latin to Italian

Citation preview

Word-Formation HSK 40.1

Handbücher zur Sprach- und Kommunikationswissenschaft Handbooks of Linguistics and Communication Science Manuels de linguistique et des sciences de communication Mitbegründet von Gerold Ungeheuer Mitherausgegeben (1985−2001) von Hugo Steger

Herausgegeben von / Edited by / Edités par Herbert Ernst Wiegand

Band 40.1

De Gruyter Mouton

Word-Formation An International Handbook of the Languages of Europe Volume 1 Edited by Peter O. Müller Ingeborg Ohnheiser Susan Olsen Franz Rainer

De Gruyter Mouton

ISBN 978-3-11-024624-7 e-ISBN (PDF) 978-3-11-024625-4 e-ISBN (EPUB) 978-3-11-039320-0 ISSN 1861-5090 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2015 Walter de Gruyter GmbH, Berlin/Boston Typesetting: Meta Systems Publishing & Printservices GmbH, Wustermark Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen Cover design: Martin Zech, Bremen 앝 Printed on acid-free paper 앪 Printed in Germany www.degruyter.com

Acknowledgments The beginnings of this handbook date back to the year 2007 as Herbert Ernst Wiegand, during a guest visit in Erlangen, accepted the suggestion of adding a handbook on wordformation to the series Handbooks of Linguistics and Communication Science of which he is the editor. For his support we are extremely grateful. In the spring of 2008, the circle of current editors materialized. In two meetings in July and October in Berlin the structure of the handbook was conceived and suggestions for the choice of authors were worked out. The first articles reached us in 2010, the last arrived in 2014. During the proofreading of the galley proofs in 2014, the authors of the earlier articles were given the opportunity to update their contributions so that all the articles in the handbook could reflect the current state of the discipline. The editors have taken care to interconnect the individual contributions in the handbook by crossreferencing the articles, thus encouraging a systematic evaluation of the topics and information represented therein. First of all, we would like express our gratitude to the authors of the 207 articles for their constructive co-operation and their understanding for our comments and suggestions. A few of them were even prepared to jump in at short notice to take over articles whose topics otherwise would have been left untreated. For this we are especially grateful. The effort on the part of the authors to adhere for the most part to the proposed submission deadlines made it possible for us to keep to the time frame agreed upon by the editors and the publisher for the publication of the handbook that, according to the original plans, was to include two volumes but in actual fact has turned out to encompass five. For their disciplined work within the given time restraints the authors deserve special thanks, also because in many cases the articles were originally written in the native language of the author and then translated into the lingua franca English – sometimes by the authors themselves and other times by a translator. Not all authors have desired to make the translation known and therefore we can only express our gratitude to the few translators whose names are known to us. Alphabetically these are: Iraide Ibarrexte Antuñano (article 29), Tatiana Bogrdanova (191), Pavel S. Dronov (162), Svenja Grabner and Philip Herdina (43, 59, 130), Beate Seidel (129), Dirk Siepmann (131), George Smith (4, 5, 93, 109, 134), Christo Stamenov (167) and Daniel Węgrzyn (107). In addition, we are deeply indebted to Barbara Karlson who has supervised our handbook from within the publishing house. She has accompanied our project from the very start, assisting us in every possible way and has always been open to our suggestions and wishes. Without her contribution the work on the handbook as well as its timely completion would undoubtedly not have come about so smoothly. The handbook ends with a subject index and a map of languages. The subject index was compiled by Sophie Salmen (Berlin) who, in her function as a student of linguistics at the Humboldt-Universität, also managed the homepage of the project that was originally set up by Antonia Fegeler, also a student of linguistics in Berlin. For their excellent und untiring commitment to the project we would like to extend a very cordial thanks to both. The content of the subject index is a product of the individual authors’ suggestions who each contributed a number of key words from their topic that were then brought together to form the index. The map of languages was designed by Hans-Jörg Bibiko (Max Planck Institute for Evolutionary Anthropology, Leipzig) to whom we extend our appreciation. It offers an

vi

Acknowledgments overview of the geographic distribution of the 74 languages that belong to the language portraits treated in articles 134−207 of the handbook. The article number is placed before each language name for ease of reference. With the completion of this handbook a long collaboration of the editors comes to an end. The work was characterized by an extremely creative, harmonic and profitable team spirit that was enjoyable and rewarding to each one of us. Our work on the handbook served to increase our knowledge of word-formation in many ways. But most of all it has brought to our attention that the word-formation of the European languages has been treated in a variety of different ways and that there is still great need for further interactive discussion and cross-fertilization. To bring this out is one of the essential goals of our handbook. January 2015

The Editors

Introduction 1. Why a handbook on word-formation? 2. Aims of the handbook 3. Outline of the handbook

1. Why a handbook on word-formation? Word-formation has been considered a central component of grammar for quite some time. Already in Antonio Nebrija’s Gramática de la lengua castellana (1492), the first grammar of a modern European language, a separate chapter was devoted to wordformation (Libro tercero que es de la etimología y dición). For Justus Georg Schottelius’ Ausführliche Arbeit Von der Teutschen HaubtSprache (1663), word-formation is constitutive as well; it is discussed in two comprehensive volumes of Jacob Grimm’s Deutsche Grammatik (1826, 1831) and forms its own chapters in the influential grammars, also entitled Deutsche Grammatik, by Wilhelm Wilmanns (1899) and Hermann Paul (1920), although Paul observes in his introduction that for “the position of word-formation within the system of grammar […] there is as yet no confirmed habit”. This statement is not obsolete; even today the question of whether word-formation is a self-contained linguistic discipline is still being debated. Often word-formation is subsumed together with inflection under the heading of “morphology” (as, e.g., in the HSK volume 17). There have also been attempts − primarily during the transformational phase of generative grammar (1957−ca. 1975), but also again in recent “minimalist” and “distributed” frameworks − to reduce word-formation to syntax and phonology, which however has not gained wide acceptance among morphologists. At present there is a broad consensus that word-formation operates on the basis of words and most of the recent psycholinguistic research also confirms this. Nevertheless, central features of word-formation do interface on the formal side with phonology, morphology and syntax and on the side of content with semantics, lexicology and pragmatics. Word-formation, therefore, is clearly an interdisciplinary phenomenon situated between lexicon and grammar, even if one shares the view of the editors of this handbook that word-formation, that is the study and description of the processes and regularities that form new words on the basis of the existing vocabulary, constitutes an independent area of scientific study. In the course of the reorientation from a historically oriented to a synchronic-structural field of research, word-formation has, beginning in the 1960s, developed into a favored object of linguistic theorizing. The initial word-formation boom, which continues on today, has led to a number of interesting controversies, for instance about the architecture of grammar or the nature of language-based rules. But, the importance of the area of word-formation results in the end from the large proportion of word-formations in the lexicon. About 80 % of the words in any Romance language are morphologically complex and fall into the sphere of word-formation. For many other languages similar statistics can be adduced. The numerous connections between word-formation and other linguistic subareas are also documented in the other HSK-volumes which often contain articles on particular aspects of word-formation. These include, for instance, the volumes Sprachgeschichte

viii

Introduction [Language History] (HSK 2), Name Studies (HSK 11), Language Typology and Language Universals (HSK 20), Lexicology (HSK 21), The Nordic Languages (HSK 22), Romanische Sprachgeschichte [Romance Language History] (HSK 23), The Slavic Languages (HSK 32) and Morphology (HSK 17). The latter volume, which essentially reflects the state of research at the beginning of the 1990s, includes two chapters on word-formation (XI. Word formation I: Fundamental problems, XII. Word formation II: Processes), but these 14 articles could offer only a very limited account of this field of knowledge, resulting in the omission of many important aspects. From these findings, we believe, emerges the necessity for a HSK-volume which is exclusively dedicated to word-formation and which presents the field from a cross-linguistic perspective and in so doing consolidates different research aspects.

2. Aims of the handbook The handbook has the following main objectives: a) The five-volume handbook intends to provide a comprehensive account of the subject area of word-formation under consideration of the following major aspects: word-formation as a scientific discipline (history of science, theoretical concepts), units and processes of word-formation, rules and restrictions of word-formation, semantics and pragmatics of word-formation, foreign word-formation, language planning and purism, historical word-formation, word-formation in language acquisition and aphasia, word-formation and language use, tools of word-formation research, word-formation in the individual languages of Europe. b) Several chapters contain, in addition to the general overview articles, special articles on specific issues of word-formation. The objective of these articles is to deepen our knowledge of particular, especially intriguing questions usually exemplified by an individual language or a language family. In the case of some of these questions, extensive literature already exists which is then summarized in the handbook for the first time for an international readership. Many of the general questions play a central role in the discussion of theoretical word-formation. For this reason the overview articles will be especially interesting to researchers of other languages or language families. It is often the case that the literature dealing with even the more specific issues is already so extensive that it is hard to keep track of, even by experts of the language in question. This state of affairs has lead to a large amount of redundancy in research as well as to superfluous discussion when attempting to gain acquaintance with the current state of research. The handbook is designed to contravene this tendency as well. c) Last but not least, the handbook is intended to promote the internationalization of our discipline. Discussion still too often tends to proceed along language boundaries. This can already be inferred from the wide variety of competing terminologies to be found in the literature. While many national philologies depend at least partially on literature written in English, there is only very limited exchange in the opposite direction or between national philologies. This is true even within language families like Romance or the Germanic languages (although the expert commission “Word-formation” of the Slavic languages, founded in 1996, serves as an exceptionally good model here in promoting

Introduction exchange via joint publications and projects). The objective of promoting the internationalization of our discipline demands an internationally comprehensible metalanguage (also for the translation of word-formation examples and the annotation of word-formation structures) − a role generally relegated to English. The individual articles are not meant for immediate colleagues but primarily for word-formation researchers of other languages. An account of Kalmyk word-formation written in Kalmyk or even in Russian, for example, would automatically be constrained to a limited sphere. d) At all times there have been complaints about the barely manageable flood of publications. Today, especially, this is doubtlessly a truism. There are, for instance, approximately 1,000 publications on Spanish word-formation alone and other languages like English, German or Russian display considerably higher numbers. Therefore, a synthesis is desirable for which this handbook will provide a framework.

3. Outline of the handbook The handbook consists of five volumes and contains 207 articles. The XVI chapters of the handbook provide the reader, in both general articles and individual studies, with a wide variety of perspectives. The precise structure of the handbook, in more detail, is as follows:

I. Word-formation as a linguistic discipline (Articles 1–13): This introductory chapter opens with a contribution to the subject matter of word-formation research and then gives a, for the most part, chronological overview of the emergence and development of word-formation research and its foundation in theoretical concepts – extending from its beginning to the current neurocognitive approach. Up to now an historically-based systematic account of word-formation research of this kind has existed only in an approximate form in Štekauer and Lieber (Handbook of WordFormation, Dordrecht 2005) and, with respect to compounding, only in a few articles in Lieber and Štekauer (The Oxford Handbook of Compounding, Oxford 2009). In addition, just recently Lieber and Štekauer’s The Oxford Handbook of Derivational Morphology, Oxford 2014, has appeared. The account to be offered in this HSK-volume, on the other hand, is more systematic and more comprehensive and aims primarily at capturing the theoretical progress in the field during the last twenty years. It will, thus, make a specific contribution to the history of word-formation as a science and clarify among other things its role in linguistic paradigm change as well as its relevance for certain neighboring disciplines and interdisciplinary research.

II. Units and processes in word-formation I: General aspects (Articles 14–27) This chapter begins with an article on the delimitation of inflection and derivation. Then, in an overview article, the units of word-formation (types of morphemes) are discussed.

ix

x

Introduction Following this, eleven articles treat different types of word-formation processes. The chapter closes with a discussion of allomorphy.

III. Units and processes in word-formation II: Special cases (Articles 28–44) In addition to the articles on general aspects in chapter II, this chapter focuses on special questions having to do with the units and processes of word-formation. The individual articles make reference partly to language families (e.g., Multi-word expressions and univerbation in Slavic), and partly to particular languages (e.g., Affective palatalization in Basque).

IV. Rules and restrictions in word-formation I: General aspects (Articles 45–48) This short chapter encompasses four general articles on rules, patterns and schemata in word-formation, on analogy, on productivity as well as on the restrictions applying to them.

V. Rules and restrictions in word-formation II: Special cases (Articles 49–55) In addition to the general aspects discussed in chapter IV, this chapter offers seven articles on particular questions pertaining to rules and restrictions in word-formation. Central to this chapter are semantic, syntactic, phonological and argument-structure related restrictions which are discussed with regard to particular morphemes (for instance, the influence of the prosodic structure of the base on the choice of suffix in English, dissimilatory phenomena in French or the systematic changes in the verbal argument structure in different forms of prefixation in German) as well as affix types (negative prefixes, closing suffixes).

VI. Semantics and pragmatics in word-formation I: General aspects (Articles 56–62) This chapter focuses on general semantic-pragmatic aspects of word-formation and contains articles on motivation and idiomatization in word-formations, on the role of folk etymology, on the possibility of the semantic categorization of word-formations, on word-formation schemes and semantic roles, on argument structure and on metonymy. The overview article The pragmatics of word-formation closes the chapter.

VII. Semantics and pragmatics in word-formation II: Special cases (Articles 63–89) In a parallel fashion to chapter V, the articles in this chapter deal with selected questions concerning the semantics of word-formation on the basis of particular word-formation categories and word-formation processes and using examples from various individual

Introduction languages, often applying a contrastive point of view. Morphopragmatic questions are discussed on the basis of the Slavic and Romance languages.

VIII. Foreign word-formation, language planning and purism I: General aspects (Articles 90–92) The three articles in this chapter form a basis for the handling of specific questions in chapter IX and give an overview of Types of foreign word-formation, Word-formation in Neo-Latin and Foreign word-formation, language planning and purism.

IX. Foreign word-formation, language planning and purism II: Special cases (Articles 93–101) In the discussion of special cases, this chapter is concerned with the clarification – from an historical and sociolinguistic perspective – of differently conditioned word-formation processes illustrated using selected examples from individual languages. Four articles discuss foreign word-formation (German, English, Italian, Polish), three focus on purism phenomena (German, French, Croatian) and another two on questions of language planning (Estonian, Russian).

X. Historical word-formation I: General aspects (Articles 102–103) The two articles in this chapter on forms and causes of word-formation change as well as on fluctuations in productivity form the basis for the individual studies in chapters XI and XII.

XI. Historical word-formation II: Special cases (Articles 104–108) Central to this chapter are particular studies of grammaticalization, e.g., the grammaticalization of word-formation in German and in the Slavic languages, the grammaticalization of prepositions in French word-formation, the grammaticalization of -mente in the adverbs of Romance as well as the origin of suffixes in Romance.

XII. Historical word-formation III: Language sketches (Articles 109–119) This chapter contains eleven language sketches on the history of word-formation of particular languages, taking into account Germanic, Romance and the Slavic languages, Irish, Greek, Uralic, Hungarian and Turkish.

XIII. Word-formation in language acquisition and aphasia (Articles 120–122) Here the concentration is on the role of word-formation in language acquisition and language loss. Three overview articles are offered on word-formation in first language acquisition, in second language acquisition and in aphasia.

xi

xii

Introduction

XIV. Word-formation and language use (Articles 123–130) This chapter focuses on specific characteristics of word-formation influenced by language use. The eight articles discuss the areas of word-formation and text, brand names, planned languages, sign languages, technical languages, literature, orthography and visuality.

XV. Tools in word-formation research (Articles 131–133) Here an overview of the tools of word-formation research is provided, making reference in three articles to dictionaries, corpora and the internet.

XVI. Word-formation in the individual European languages (Articles 134–207) The final chapter of the handbook contains 74 portraits of word-formation in the individual languages of Europe and offers an innovative perspective. These portraits afford the first overview of this kind and will prove useful for future typological research. The restriction to European languages with an underlying geographical Europe-concept makes sense insofar as this region of the world is, on the one hand, the best described in terms of word-formation and, on the other, also exhibits a considerable typological variety (Indo-European, Uralic, Basque, Semitic, Turkic, Mongolic, North Caucasian). In addition to the insight into the respective typological distinctive features, the articles also contain a short outline of the research history of each language considered and contain information about its standard reference works. To allow for a good comparison of European word-formation phenomena, the authors have been provided with a uniform structure for their articles: 1. Introduction Obligatorily includes information on research history and standard works. 2. General overview Should briefly address the main characteristics of word-formation in language X, including the place of neoclassical word-formation, where it exists (depending on the situation in language X, neoclassical word-formation can be treated either separately or together with other processes of compounding or derivation). Should address problems of demarcation: Compounding vs. syntax (including multi-word expressions and particle-verb formation) Compounding vs. derivation 3. Composition Nominal compounds should be treated before adjectival compounds before verbal compounds (including incorporation) before other compounds.

Introduction Within these categories, determinative compounds should be treated before copulative ones, appositive compounds before dvandvas, endocentric compounds before exocentric ones. 4. Derivation In all (sub-)sections, prefixes should be treated before suffixes, and suffixes before circumfixes or infixes. 4.1. Nominal derivation 4.1.1. Denominal nouns: personal nouns before status nouns before place nouns before other categories; evaluative categories last: diminutives before augmentatives and other categories 4.1.2. Deadjectival nouns: quality nouns before other categories 4.1.3. Deverbal nouns: action before agent before instrument before place nouns before other categories 4.2. Adjectival derivation 4.2.1. Denominal adjectives: relational before other categories 4.2.2. Deadjectival adjectives: intensive before evaluative and other categories 4.2.3. Deverbal adjectives: active before passive before other categories 4.3. Verbal derivation 4.3.1. Denominal verbs 4.3.2. Deadjectival verbs 4.3.3. Deverbal verbs 4.4. Adverbial derivation 5. Conversion 5.1. 5.2. 5.3. 5.4.

Nominal conversion Adjectival conversion Verbal conversion Adverbial conversion Within these categories denominal conversion should be treated before deadjectival conversion and deadjectival conversion before deverbal conversion.

6. Backformation 7. Reduplication 8. Blending 9. Clipping 10. Word-creation The proposed structure suggests a standardized terminology that, however, in reality simply does not exist. We have seen time after time that the individual philologies are

xiii

xiv

Introduction anchored in different terminological traditions. Even though, for this reason or due to the preferences of certain authors, it was not possible to maintain uniformity in terminology throughout all the studies in the handbook, we have nevertheless attempted to maintain consistency where possible. Achieving a standard descriptive terminology in word-formation presents a challenge to our discipline that yet awaits realization. We will be pleased if our handbook has made a contribution to this task. The 74 language portraits of this handbook represent a major portion of the languages of Europe. That not all the languages of Europe could be treated has several reasons. First, for some languages it was simply not possible to find competent authors (e.g., Saami). In other cases, authors were found who in the end were not able to complete their work. This is the reason why portraits of Occitan, Romani, Chechen and Ingush are missing. Finally, a few languages were intentionally left out of consideration because experts were in agreement that due to their similarity with other languages a large amount of redundancy would result if they were also included, for instance Galician (Portuguese) and Scottish Gaelic (Irish). The order of the word-formation portraits of the individual European languages has the following macrostructure: 1. Indo-European, 2. Uralic, 3. Basque, 4. Semitic, 5. Turkic, 6. Mongolic, 7. North Caucasian. The Indo-European languages have the following internal structure: Germanic, Romance, Celtic, Slavic, Baltic, Albanian, Greek, Indo-Iranian. In the realm of North Caucasian, the Northwest Caucasian languages precede the Northeast Caucasian languages. The enumeration of the language families does not contain a qualitative evaluation and could have been arranged differently. This is also valid for the individual languages grouped under the language families (e.g., Breton – Welsh – Irish for Celtic) with the exception of the cases in which the accepted order of arrangement is followed as in Slavic. Peter O. Müller, Erlangen (Germany) Ingeborg Ohnheiser, Innsbruck (Austria) Susan Olsen, Berlin (Germany) Franz Rainer, Vienna (Austria)

Contents

Volume 1 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

I.

v vii

Word-formation as a linguistic discipline

1. The scope of word-formation research · Hans-Jörg Schmid . . . . . . . . . 2. Word-formation research from its beginnings to the 19 th century · Barbara Kaltz and Odile Leclercq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Word-formation in historical-comparative grammar · Thomas Lindner . . . 4. Word-formation in structuralism · Wolfgang Motsch . . . . . . . . . . . . . 5. Word-formation in inhaltbezogene Grammatik · Johannes Erben . . . . . . 6. Word-formation in onomasiology · Joachim Grzega . . . . . . . . . . . . . 7. Word-formation in generative grammar · Rochelle Lieber . . . . . . . . . . 8. Word-formation in categorial grammar · Ulrich Wandruszka . . . . . . . . . 9. Word-formation in natural morphology · Hans Christian Luschützky . . . . 10. Word-formation in cognitive grammar · John R. Taylor . . . . . . . . . . . 11. Word-formation in optimality theory · Renate Raffelsiefen . . . . . . . . . . 12. Word-formation in construction grammar · Geert Booij . . . . . . . . . . . 13. Word-formation in psycholinguistics and neurocognitive research · Gary Libben . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.

1

. . . . . . . . . . .

22 38 52 66 79 94 112 123 145 158 188

.

203

. . . . . . . . . . . . . .

218 235 301 322 340 352 364 386 413 434 450 467 485 500

II. Units and processes in word-formation I: General aspects 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

The delimitation of derivation and inflection · Pavol Štekauer . . . Units of word-formation · Joachim Mugdan . . . . . . . . . . . . . Derivation · Andrew Spencer . . . . . . . . . . . . . . . . . . . . . . Conversion · Salvador Valera . . . . . . . . . . . . . . . . . . . . . . Backformation · Pavol Štekauer . . . . . . . . . . . . . . . . . . . . Clipping · Anja Steinhauer . . . . . . . . . . . . . . . . . . . . . . . Composition · Susan Olsen . . . . . . . . . . . . . . . . . . . . . . . Blending · Bernhard Fradin . . . . . . . . . . . . . . . . . . . . . . . Incorporation · Jason D. Haugen . . . . . . . . . . . . . . . . . . . . Particle-verb formation · Andrew McIntyre . . . . . . . . . . . . . Multi-word expressions · Matthias Hüning and Barbara Schlücker Reduplication · Thomas Schwaiger . . . . . . . . . . . . . . . . . . Word-creation · Elke Ronneberger-Sibold . . . . . . . . . . . . . . Allomorphy · Wolfgang U. Dressler . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

xvi

Contents

III. Units and processes in word-formation II: Special cases 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

Affective palatalization in Basque · José Ignacio Hualde . . . . . . . . . . . Parasynthesis in Romance · David Serrano-Dolader . . . . . . . . . . . . . . Affix pleonasm · Francesco Gardani . . . . . . . . . . . . . . . . . . . . . . Interfixes in Romance · Michel Roché . . . . . . . . . . . . . . . . . . . . . Linking elements in Germanic · Nanna Fuhrhop and Sebastian Kürschner Synthetic compounds in German · Martin Neef . . . . . . . . . . . . . . . . Verbal pseudo-compounds in German · Christian Fortmann . . . . . . . . . Particle verbs in Germanic · Nicole Dehé . . . . . . . . . . . . . . . . . . . Particle verbs in Romance · Claudio Iacobini . . . . . . . . . . . . . . . . . Particle verbs in Hungarian · Mária Ladányi . . . . . . . . . . . . . . . . . . Noun-noun compounds in French · Pierre J. L. Arnaud . . . . . . . . . . . Verb-noun compounds in Romance · Davide Ricca . . . . . . . . . . . . . . Co-compounds · Bernhard Wälchli . . . . . . . . . . . . . . . . . . . . . . . Multi-word units in French · Salah Mejri . . . . . . . . . . . . . . . . . . . . Multi-word expressions and univerbation in Slavic · Olga Martincová . . . Compounds and multi-word expressions in Slavic · Ingeborg Ohnheiser . . Paradigmatically determined allomorphy: the “participial stem” from Latin to Italian · Anna M. Thornton . . . . . . . . . . . . . . . . . . . . . . . . . .

Volume 2 IV. Rules and restrictions in word-formation I: General aspects 45. 46. 47. 48.

Rules, patterns and schemata in word-formation · Heike Baeskow Word-formation and analogy · Sabine Arndt-Lappe Productivity · Livio Gaeta and Davide Ricca Restrictions in word-formation · Livio Gaeta

V. Rules and restrictions in word-formation II: Special cases 49. 50. 51. 52.

Argument-structural restrictions on word-formation patterns · Holden Härtl Phonological restrictions on English word-formation · Renate Raffelsiefen Morphological restrictions on English word-formation · Lothar Peter Semantic restrictions on word-formation: the English suffix -ee · Heike Baeskow 53. Dissimilatory phenomena in French word-formation · Marc Plénat 54. Closing suffixes · Stela Manova 55. Closing suffix patterns in Russian · Dmitri Sitchinava

. . . . . . . . . . . . . . . .

517 524 537 551 568 582 594 611 627 660 673 688 707 727 742 757

.

780

Contents

VI. Semantics and pragmatics in word-formation I: General aspects 56. 57. 58. 59.

Motivation, compositionality, idiomatization · Daniela Marzo Word-formation and folk etymology · Sascha Michel Categories of word-formation · Volkmar Lehmann Schemata and semantic roles in word-formation · Hanspeter Ortner and Lorelies Ortner 60. Word-formation and argument structure · Manfred Bierwisch 61. Word-formation and metonymy · Manfred Bierwisch 62. The pragmatics of word-formation · Lavinia Merlini Barbaresi

VII. Semantics and pragmatics in word-formation II: Special cases 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89.

Noun-noun compounds · Christina L. Gagné and Thomas L. Spalding Gender marking · Ursula Doleschal Singulatives · Paolo Acquaviva Collectives · Wiltrud Mihatsch Action nouns · Maria Koptjevskaja-Tamm Action nouns in Romance · Livio Gaeta Verbal nouns in Celtic · Paul Russell Nominalization in Hungarian · Tibor Laczkó Result nouns · Chiara Melloni Quality nouns · Franz Rainer Status nouns · Hans Christian Luschützky Agent and instrument nouns · Franz Rainer Patient nouns · Susanne Mühleisen Place nouns · Bogdan Szymanek Intensification · Franz Rainer Negation · Marisa Montero Curiel Negation in the Slavic and Germanic languages · Jozef Pavlovič Spatial and temporal relations in German word-formation · Ludwig M. Eichinger Adverbial categories · Davide Ricca Denominal verbs · Andrew McIntyre Valency-changing word-formation · Dieter Wunderlich Word-formation and lexical aspect: deverbal verbs in Italian · Nicola Grandi Word-formation and aspect in Samoyedic · Beáta Wagner-Nagy Verbal prefixation in Slavic: a minimalist approach · Petr Biskup and Gerhild Zybatow Denumeral categories · Bernhard Fradin The semantics and pragmatics of Romance evaluative suffixes · Martin Hummel Morphopragmatics in Slavic · Alicja Nagórko

xvii

xviii

Contents

Volume 3 VIII. Foreign word-formation, language planning and purism I: General aspects 90. Types of foreign word-formation · Wieland Eins 91. Word-formation in Neo-Latin · Thomas Lindner and Franz Rainer 92. Foreign word-formation, language planning and purism · Wolfgang Pöckl

IX. Foreign word-formation, language planning and purism II: Special cases 93. 94. 95. 96. 97. 98. 99. 100. 101.

Foreign word-formation in German · Peter O. Müller Foreign word-formation in English · Klaus Dietz Foreign word-formation in Italian · Claudio Iacobini Foreign word-formation in Polish · Krystyna Waszakowa Word-formation and purism in German · Mechthild Habermann Word-formation and purism in French · Petra Braselmann Word-formation and purism in Croatian · Branko Tošović Word-formation and language planning in Estonian · Virve Raag Individual initiatives and concepts for expanding the lexicon in Russian · Wolfgang Eismann

X. Historical word-formation I: General aspects 102. Mechanisms and motives of change in word-formation · Franz Rainer 103. Change in productivity · Carmen Scherer

XI. Historical word-formation II: Special cases 104. Grammaticalization in German word-formation · Mechthild Habermann 105. The grammaticalization of prepositions in French word-formation · Dany Amiot 106. The Romance adverbs in -mente: a case study in grammaticalization · Ulrich Detges 107. Grammaticalization in Slavic word-formation · Krystyna Kleszczowa 108. The origin of suffixes in Romance · David Pharies

XII. Historical word-formation III: Language sketches 109. 110. 111. 112.

Historical word-formation in German · Peter O. Müller Historical word-formation in English · Klaus Dietz From Latin to Romance · Éva Buchi and Jean-Paul Chauveau From Latin to Romanian · Marina Rădulescu Sala

Contents 113. 114. 115. 116. 117. 118. 119.

xix

From Old French to Modern French · Franz Rainer and Claude Buridant From Old Irish to Modern Irish · David Stifter Historical word-formation in Slavic · Swetlana Mengel From Ancient Greek to Modern Greek · Io Manolessou and Angela Ralli The history of word-formation in Uralic · Johanna Laakso From Old Hungarian to Modern Hungarian · Tamás Forgács Historical word-formation in Turkish · Claus Schönig

XIII. Word-formation in language acquisition and aphasia 120. Word-formation in first language acquisition · Hilke Elsen and Karin Schlipphak 121. Word-formation in second language acquisition · Cornelia Tschichold and Pius ten Hacken 122. Word-formation in aphasia · Carlo Semenza and Sara Mondini

XIV. Word-formation and language use 123. 124. 125. 126. 127. 128. 129. 130.

Word-formation Word-formation Word-formation Word-formation Word-formation Word-formation Word-formation Word-formation

and and and and and and and and

text · Anja Seiffert brand names · Elke Ronneberger-Sibold planned languages · Klaus Schubert sign languages · Ronnie B. Wilbur technical languages · Ivana Bozděchová literature · Peter Handler orthography · Hannelore Poethe visuality · Lorelies Ortner

XV. Tools in word-formation research 131. Dictionaries · Renate Belentschikow 132. Corpora · Ulrich Heid 133. Internet · Georgette Dal and Fiammetta Namer

Volume 4 XVI. Word-formation in the individual European languages Indo-European Germanic 134. German · Irmhild Barz 135. English · Ingo Plag

xx

Contents 136. 137. 138. 139. 140. 141. 142. 143.

Dutch · Geert Booij Frisian · Jarich F. Hoekstra Yiddish · Simon Neuberg Faroese · Hjalmar P. Petersen Danish · Hans Götzsche Norwegian · John Ole Askedal Swedish · Kristina Kotcheva Icelandic · Þorsteinn G. Indriðason

Romance 144. 145. 146. 147. 148. 149. 150. 151.

Portuguese · Bernhard Pöll Spanish · Franz Rainer Catalan · Maria Teresa Cabré Castellví French · Franck Floricic Ladin · Heidi Siller-Runggaldier Sardinian · Immacolata Pinto Italian · Franz Rainer Romanian · Maria Grossmann

Celtic 152. Breton · Elmar Ternes 153. Welsh · Paul Russell 154. Irish · Brian Ó Curnáin Slavic 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168.

Upper Sorbian · Anja Pohontsch Polish · Alicja Nagórko Kashubian · Edward Breza Czech · Ivana Bozděchová Slovak · Martina Ivanová and Martin Ološtiak Ukrainian · Ievgeniia Karpilovska Belarusian · Alâksandr Lukašanec Russian · Igor’ S. Uluhanov Slovene · Irena Stramljič Breznik Croatian · Mario Grčević Serbian · Božo Ćorić Bosnian · Branko Tošović Bulgarian · Cvetanka Avramova and Julia Baltova Macedonian · Lidija Arizankovska

Map of languages

Contents

Volume 5 Baltic 169. Lithuanian · Bonifacas Stundžia 170. Latvian · Agnė Navickaitė-Klišauskienė

Albanian 171. Albanian · Monica Genesin and Joachim Matzinger

Greek 172. Greek · Angela Ralli

Indo-Iranian 173. Ossetic · David Erschler 174. Tat · Gilles Authier

Uralic 175. 176. 177. 178. 179. 180. 181.

Nenets · Beáta Wagner-Nagy Finnish · Kaarina Pitkänen-Heikkilä Estonian · Krista Kerge Permic · László Fejes Mari · Timothy Riese Mordvinic · Sándor Maticsák Hungarian · Ferenc Kiefer

Basque 182. Basque · Xabier Artiagoitia, José Ignacio Hualde and Jon Ortiz de Urbina

Semitic 183. Maltese · Joseph Brincat and Manwel Mifsud

Turkic 184. Turkish · Jens Wilkens 185. Bashkir · Gulnara Iskandarova 186. Tatar · László Károly

xxi

xxii

Contents 187. 188. 189. 190.

Crimean Tatar · Henryk Jankowski Gagauz · Astrid Menz Karaim · Éva Á. Csató Chuvash · Galina N. Semenova and Alena M. Ivanova

Mongolic 191. Kalmyk · Danara Suseeva

Northwest Caucasian 192. Abkhaz · Viacheslav A. Chirikba 193. Adyghe · Yury Lander 194. Kabardian · Ranko Matasović

Northeast Caucasian 195. 196. 197. 198. 199. 200. 201. 202. 203. 204. 205. 206. 207.

Rutul · Mikhail Alekseyev Budugh · Gilles Authier and Adigözel Haciyev Udi · Wolfgang Schulze Aghul · Timur Maisak and Dmitry Ganenkov Archi · Marina Chumakina Khinalug · Wolfgang Schulze Lak · Wolfgang Schulze Dargwa · Nina Sumbatova Bezhta · Madzhid Khalilov and Zaira Khalilova Botlikh · Mikhail Alekseyev Akhwakh · Denis Creissels Avar · Madzhid Khalilov and Zaira Khalilova Khwarshi · Zaira Khalilova

Subject index Map of languages

I. Word-formation as a linguistic discipline 1. The scope of word-formation research 1. 2. 3. 4. 5. 6. 7. 8. 9.

Introduction Morphological building blocks and the internal structures of complex lexemes Word-formation patterns Approaches to word-formation research Levels of analysis and description in word-formation research Theoretical models of word-formation Modelling dynamic aspects of word-formation: productivity and lexicalization Conclusion References

Abstract The first article of this volume presents an introductory survey of the scope of wordformation research. It defines and demarcates the subject-matter of word-formation and explains the basic notions related to the internal structures of complex lexemes and the cross-linguistically important word-formation patterns. Major approaches, analytical and descriptive levels and models in the field of word-formation research are outlined from a bird’s eye view. The final section deals with productivity and lexicalization.

1. Introduction Word-formation research investigates the patterns and regularities underlying the formation of complex lexemes by means of existing building blocks with the aim of formulating rules and other types of generalizations. Complex lexemes (e.g., E. headteacher or trivialize) are characterized by the fact that they consist of two or more constituents. Unlike most simple lexemes such as head, teach and trivial, complex lexemes are not entirely arbitrary signs, but instead are morphologically motivated by their constituents and by the semantic links shared with other structurally identical formations. A precise understanding of the nature of this motivation forms the main interest of word-formation research. The scope of word-formation research in linguistics can be defined by demarcating word-formation from neighbouring fields. The adjacent domain of inflectional morphology deals with elements and operations which produce word-forms of lexemes (e.g., teaches, teaching, taught) rather than new lexemes (teacher, headteacher, to teamteach, etc.), as is the case in word-formation. Word-formation and inflectional morphology are not separated by a clear boundary, however (Bybee 1985; Scalise 1988; Plank 1994; Booij 2000; Stump 2005; see article 14 on the delimitation of derivation and inflection). A classic example of a delimitation problem is the dispute over whether the English

2

I. Word-formation as a linguistic discipline adverb-forming suffix -ly as in really or elegantly should be treated as a derivational, i.e. lexical, or inflectional and thus grammatical morpheme (cf. Giegerich 2012). Syntax, while having emerged as a prominent source of inspiration for theory-building in wordformation, differs from word-formation in that the output of syntactic operations is phrasal and clausal rather than lexical in nature. Needless to say, boundary issues exist as well, e.g., in the distinction of certain types of nominal compounds from noun phrases (e.g., Benveniste 1967; Bauer 1988a; Olsen 2000; see articles 20 on composition, 38 on noun-noun compounds in French and 135 on English). Demarcation problems are also very common at the porous boundary to phraseology, for instance when it comes to classifying semi-idiomatic phrases such as black market or particle verbs of the type get up and make up for (see articles 23 on particle-verb formation and 24 on multi-word expressions). Many practitioners of word-formation research distinguish word-formation (in a narrow sense) from ways of extending the lexical resources which do not involve changes in the forms of linguistic signs, mainly metaphorical or metonymic transfers and other forms of lexical change resulting in purely semantic extensions or shifts (see article 61 on word-formation and metonymy). Finally, as suggested by the definition given above, word-formation can be, but is not always, distinguished from what is referred to as coinage (Lieber 2010: 51), word-creation or word manufacture (Bauer 1983: 239), which does not rely on existing building blocks (see article 26 on word-creation). Frequently quoted examples include product and brand names such as Kodak or Google. The subject-matter of word-formation research is also demarcated by the definition provided above. Essentially, four aspects define the remit of word-formation research. Firstly, word-formation research analyses and describes the internal structures and constituents of complex lexemes and identifies and classifies the forms and meanings of the lexical and morphological building blocks of a given language (see article 15 on units of word-formation). As a part of this segmentation, identification and classification procedure, the lexical building blocks involved in word-formation must be distinguished from inflectional morphemes (see article 14 on the delimitation of derivation and inflection). The results of analytical and classificatory efforts feed into models of word-formation processes. Secondly, word-formation research identifies, classifies and models the processes underlying the formation of existing and new complex lexemes (see chapters II and III of the handbook). This is typically accomplished by segmenting established complex lexemes and describing their grammatical, morphological, semantic and phonological properties as well as those of their constituents. While most researchers in the field agree on a set of basic types of word-formation processes, there has been considerable controversy over the precise way in which they should be modelled. Thirdly, because of the multifaceted nature of complex lexemes, word-formation research tends to be a multi-level (Lipka 1983) or multi-perspectival endeavour. Traditionally, morphological, syntactic, semantic and phonological aspects have taken centre stage in word-formation research, as these perspectives provide the basis for systematic and parsimonious generalizations regarding word-formation rules and patterns. More recently, sociopragmatic, psycholinguistic, cognitive and textual aspects have been attracting increasing interest (see article 13 on word-formation in psycholinguistics and neurocognitive research and chapter XIV of the handbook). Fourthly, word-formation research tries to provide adequate models of the creative and dynamic aspects of word-formation. On the level of word-formation rules and pat-

1. The scope of word-formation research terns, this relates to the changing productivity of word-formation processes and the elements involved in them (see articles 47 on productivity and 103 on change in productivity). On the level of individual complex lexemes, an explanation must be found for how new creations are motivated, how they find their way into the lexicon of a language and how their forms and meanings change in the course of time (see article 56 on motivation, compositionality, idiomatization). As regards the terminology used to refer to the core interest of word-formation research, the terms word-formation process, word-formation type, word-formation model, word-formation rule and word-formation pattern will be used interchangeably here, even though they highlight different aspects of the phenomena at hand and have been defined in more specific ways by some authors (e.g., Fleischer and Barz 2012: 67–69; Hansen et al. 1990: 28). Section 2 of this article will be devoted to the morphological building blocks and internal structures of complex lexemes. Section 3 will provide a survey of cross-linguistically important word-formation patterns. The next three sections will discuss different approaches to word-formation research (section 4), survey the linguistic perspectives included in word-formation research (section 5) and provide a sketch of the major types of theoretical models attempting to capture the nature of word-formation rules and patterns (section 6). Section 7 will focus on research into the temporal and dynamic aspects of word-formation, i.e. the productivity of word-formation patterns and the types of changes that take place as complex lexemes are coined, spread and become part of the lexicon.

2. Morphological building blocks and the internal structures of complex lexemes Three basic approaches to describing the constituents of complex lexemes can be distinguished: a word-based, a root-based and a morpheme-based approach (see article 15 on units of word-formation). The first type proceeds from the assumption that words constitute the cores of complex lexemes. Word-formation rules combine several words in the case of compounding, and words and affixes in derivation. This assumption is known as the word-based hypothesis (Aronoff 1976: 21; Scalise 1986: 40–42, 71–78). In the second approach, roots or stems, rather than full-fledged words, are considered the key constituents of complex lexemes. Terminology is far from uniform in this area, especially regarding the term stem (see Chelliah and de Reuse 2011: 314 for a survey). This notion is sometimes used to refer to the base of a word-formation process, i.e. the element to which further morphological material is added, and sometimes to the part which remains constant before inflectional endings are added (Bauer 1988b: 11). According to the first interpretation, national would be the stem of (the company was) nationalized, and according to the second, nationalize. To keep these readings apart, some authors (e.g., Hansen et al. 1990: 41) distinguish the word stem of a lexeme, qua word-form minus inflectional affix, from the word-formation stem, i.e. the base of the (final) wordformation process. Whether words or stems are more useful as basic units of wordformation research may well depend on the language being investigated (Bloomfield 1933: 224–226; Kastovsky 1999).

3

4

I. Word-formation as a linguistic discipline The third approach relies on the notion of morpheme (Baudouin de Courtenay 1895: 10), usually defined as the smallest meaning-bearing units of words (Bloomfield 1926: 155). Morphemes are classified with regard to their distributional properties into (potentially) free morphemes and (obligatorily) bound morphemes, and with regard to their function into grammatical morphemes and lexical morphemes. In English word-formation research (see article 15 on units of word-formation for traditions in other languages), free lexical morphemes, also known as root morphemes, are considered to correspond to simple lexemes (e.g., hand, great, eat); bound grammatical morphemes correspond to inflectional endings, e.g., {genitive -s}, {past tense -ed}, marking word-forms for case, number, tense, etc. Bound lexical morphemes are derivational affixes used for the purpose of word-formation (e.g., un-, -ment); free grammatical morphemes are function words such as of, to or the, which are orthographically autonomous, serve grammatical functions and are synsemantic rather than autosemantic. Morphemes are theoretical constructs abstracting commonalities over morphs, which are often regarded as perceptible physical events realizing morphological building blocks in actual speech or writing (see article 15 for a critique of this view). While the notion of morpheme has proved useful for the description of basic constituents of complex lexemes, demarcation problems are rampant, with regard to both the identification of morphemes as such and the classification of morphemes into lexical vs. grammatical and free vs. bound morphemes. The general properties of morphemes are reviewed by Mugdan (1986) and Luschützky (2000), see also article 15 on units of wordformation. Prototypical lexical morphemes, which lie at the heart of word-formation research, have to meet the following morpheme- and lexeme-related criteria: a) Identifiability: It must be possible to describe each of the potential morphemes precisely in terms of their extent and form and the corresponding meanings. b) Exhaustive segmentability of complex lexemes into morphemes: It must be possible to segment a given complex lexeme exhaustively into morphemes and other clearly identifiable non-meaning-bearing elements such as linking elements (as in G. Notenständer ‘music stand’ ← {note} + n + {ständer}). c) Autonomy of at least one constituent: Every complex lexeme must consist of at least one free lexical morpheme. d) Compositionality of complex lexemes: It must be possible to trace the meaning of the complex lexeme back to the meanings of the morphemes. In practical applications problems tend to arise with regard to all four criteria, resulting in the postulation of different types of pseudo-morphemic, quasi-morphemic or submorphemic units (cf. Kubrjakova 2000 and article 15). One notorious challenge are English verbs of Romance origin such as insist, persist and resist or ascribe, prescribe and subscribe which suggest an analysis in terms of a prefix (a-, in-, per-, pre-, re- and sub-) and a stem (-sist or -scribe). Such an analysis runs into difficulties with criteria a), c) and d), because, at least from a synchronic point of view, it is neither possible to ascribe a meaning to the potential stems -sist and -scribe nor are these stems free morphemes; as a result, compositionality is also violated. The term bound root has been introduced to capture such meaning-bearing units. Neoclassical compounds such as biology or bibliography, which are found in most European languages, cause partly similar problems. While their parts – bio-, biblio-, -logy and -graphy – seem to carry identifiable lexical meanings, none of them are free forms, thus violating criterion c). In English and

1. The scope of word-formation research

5

Romance linguistics, the terms combining form and affixoid have become established to describe this phenomenon; in German linguistics the term Konfix (Schmidt 1987: 50) is commonly used. Typical lexical blends also defy descriptions in terms of morphemes, as they cause problems for all four criteria. The constituent units of blends, for example br- and -unch in the case of the classic brunch, are known as splinters (Lehrer 1996). Furthermore, the particles of English phrasal verbs (give up, pass out), German prefix and particle verbs (beraten, anbahnen) and similar multi-word lexical items also fall within the scope of morpheme-like building blocks of complex lexemes. The semantic contribution of these elements to the meaning of the multi-word unit is often opaque, however, so that criteria a) and d) are not met. The term formative has been suggested either to refer to such minimal units which lack an identifiable meaning, also including linking elements (G. Fugenelemente) (Kastovsky 1982: 70; Lipka 2002: 87), or as a superordinate term comprising both morphemes proper and semantically empty morphological building blocks (Bauer 1988b: 24; Bauer, Lieber and Plag 2013: 16). Finally, phonaesthemes (e.g., fl- in flicker, flip, flap, flurry) are non-arbitrary pairings of phoneme clusters and meanings (Firth 1964: 184–185). Their semantic significance is usually restricted to a limited range of words and their meanings do not go beyond soundsymbolic allusions and associations which are shared by a set of words but are quite difficult to pinpoint. The portions of words that remain when the phonaestheme is segmented do not carry meanings; as a result, the words are non-compositional. The internal structures of complex lexemes are usually described not in terms of shallow, chain-like sequences of morphemes and morpheme-like elements but as hierarchical structures which follow the principle of binary branching of immediate constituents (cf., e.g., Booij 1977: 32; Lieber 1990: 80; Scalise 1984: 146–151) also prominent in syntactic theorizing. In addition, the right-hand head rule (Williams 1981) states that the two sister constituents on one level are not equipotent but related to each other in a determinans-determinatum (Marchand 1969) or modifier-head relation (cf. also Lieber 1992: 26–76). The determinatum or head is the semantically and, more importantly, grammatically dominant constituent, which is specified by the determinans or modifier. To conclude this section, Figure 1.1 illustrates the most important concepts introduced here. nationalized

nation al free lexical bound lexical morpheme morpheme root derivational suffix base (word-formation stem) (word) stem modifier modifier head

ize bound lexical morpheme derivational suffix suffix

ed bound grammatical morpheme inflectional suffix/ inflectional ending

head

Fig. 1.1: Terms used for the description of the internal structures of complex lexemes

6

I. Word-formation as a linguistic discipline

3. Word-formation patterns Figure 1.2 represents one of the most common ways of classifying the major types of word-formation patterns found in the languages of the world. While many other classifications are of course possible, depending, among other things, on the type of language and the individual researcher’s aims and convictions, the key categories found in Figure 1.2 have proved adequate for the descriptions of the wide range of different languages provided in articles 134 to 207 of this handbook.

Fig. 1.2: Survey of word-formation patterns

As all these types of patterns are discussed in greater detail and with reference to a large number of languages in parts II, III and XVI of this handbook, it will suffice here to provide a general survey. Composition, or compounding, is defined as the combination of (at least) two lexemes or words (Bauer 1988b: 33; article 20 on composition), or stems (Hansen et al. 1990: 43; Fleischer 2000: 889), or bases (Lieber 2010: 43; see also article 135 on English), or free lexical morphemes (Schmid 2011). Many authors explicitly include bound roots or confixes – in addition to free lexical morphemes proper – as potential bases of compounds (Fleischer and Barz 2012: 84; Plag 2003: 135; article 135 on English). Compounds are sub-classified in terms of their semantic structure into endocentric and exocentric compounds, with the former being further divided into determinative and coordinative ones. As regards their forms, root compounds, consisting of only two free lexical morphemes, are distinguished from synthetic (verbal, verbal nexus) compounds, which include bound lexical morphemes (see articles 20 on composition and 33 on synthetic compounds in German). Rules detailing the formation of compounds have to take the word-classes of their constituents into account, granting a prominent role to the head constituent, so that blackbird would be referred to as a nominal compound of the form Adj + N. Derivation (see article 16 on derivation) is the process of adding an affix to a stem, or a bound lexical morpheme to a free one, in order to create a new lexeme. The main forms of derivation are prefixation or prefix-derivation, where the bound morpheme precedes the free one, and suffixation or suffix-derivation where the order is reversed. Rarer types not rendered in Figure 1.2 are infixation and circumfixation. In certain schools of word-formation research mainly in English studies (Marchand 1969; Kastovsky 1982; Hansen et al. 1990), zero-derivation or derivation by zero-morpheme is postu-

1. The scope of word-formation research lated as a further type of derivation producing new lexemes by the addition of a formally empty zero-morpheme. As indicated by its place in Figure 1.2, the process of backformation (see article 18) straddles the boundary between morphemic and non-morphemic word-formation patterns. In contrast to derivation by means of the addition of an affix, lexemes formed by backformation are the product of the deletion of a bound morpheme or morpheme-like element. Roughly speaking, conversion (see article 17) is a word-formation process which transposes a lexeme to a new word class without the addition of an overtly marked suffix. As Valera (see article 17) puts it, “the form of the converted item does not change, while its inflectional potential, its syntactic function and its meaning do, such that the item displays inflectional, syntactic and semantic properties of a new word-class”. Conversion and zero-derivation – as well as the concept of paradigmatic derivation used in work on word-formation in Polish; see article 156 on Polish – are essentially competing ways of making sense of the same phenomena. Where the converted item does not acquire the full range of inflectional possibilities (e.g., the rich), this is termed partial conversion by proponents of the conversion approach and conversion by those of the zero-derivation approach. Morphemic word-formation processes make up the core of word-formation in the sense that they are to a large extent regular and predictable (in hindsight) and therefore amenable to generalizations couched in the format of rules or schemas (see section 7). Equipped with knowledge of such rules, no competent speaker of English who knows the verb to tweet will be surprised when confronted with a derived verb such as to detweet, as this is clearly a potential or possible word (Aronoff 1976: 17–19). The same kind of hindsight predictability does not apply to the group of non-morphemic processes listed in Figure 1.2, which are also known as minor word-formation types. In addition, these processes are more flexible and more creative, which causes difficulties when it comes to attributing a shared meaning to formations of a similar type. Reduplication (see article 25) is a word-formation process which involves the repetition of a word, word-like element or part of a word either in unchanged form (e.g., hushhush), with a different vowel (e.g., hip-hop) or a different consonant (e.g., boogie-woogie). Blending (see article 21) is a cover term for a range of processes which, like composition, combine (at least) two lexemes, but, unlike composition, also fuse their forms by either shortening one or both of the input lexemes or by telescoping them into each other at portions where their forms overlap. Unlike all other word-formation processes, clipping and acronym-formation (see article 19) both preserve the denotative meanings of the source lexemes. Both are form-shortening processes only. Clipping, the deletion of initial and/or final portions of words, can be applied to single words, while acronyms, with some variation, are created by deleting everything but initial letters of two-word or longer expressions. Some researchers distinguish between acronyms in a narrow sense, which can be pronounced like normal words, and initialisms, pronounced as series of individual letters. Needless to say, the description of different word-formation models and the analysis of individual complex lexemes do not stop at this level of granularity. More specific descriptions from a structural perspective tend to consist of at least six types of information. These are illustrated by English deverbal nominalizations in Table 1.1. Textual, pragmatic and sociolinguistic aspects related to characteristics of the use and frequency of word-formation patterns in different text types and with different functions can complement this information.

7

8

I. Word-formation as a linguistic discipline Tab. 1.1: Parameters for the description of complex lexemes (inspired by Fleischer and Barz 2012: 73–74) parameter

exemplified for English deverbal nominalization in -(a)tion

a) general morphological properties of the constituents

base: – word-class: transitive verbs, e.g., explain, combine, describe – morphological status: free lexical morpheme affix: – morphological status: bound lexical morpheme, suffix – form: -(a)tion

b) order of constituents

– base – suffix

c) formal and semantic characteristics of the process

– formal description: [V + -(a)tion]N – semantic description: ‘process or product of V-ing’

d) various other characteristics and effects of the complex lexemes generated by the process, especially phonological and graphological ones, sometimes depending on the specific characteristics of the input

– stress movement and vowel change, e.g., /ɪks'pleɪn/ > /eksplɘ'neɪʃn/ – possible insertion of /eɪ/ before /ʃn/

e) restrictions on the nature of the input and on the applicability of the process (see articles 48 to 52)

– generally not applicable to English words of Germanic origin, cf. hear > *hearation, eat > *eatation

f) degree of productivity (see article 47)

– fully productive, especially in formal contexts

4. Approaches to word-formation research The study of word-formation has been approached from a variety of angles. Some of these are shared by other linguistic disciplines, while others are specific to word-formation research or manifested in specific forms. As in other fields of linguistics, we can distinguish between historical approaches (see parts X, XI, XII), and those which investigate the present-day language in its current state. Historical investigations can be carried out in a diachronic manner (cf. Kastovsky 2009) or in a synchronic one. What is specific to word-formation research is that there is always a latent diachronic element in synchronic descriptions, because the very idea that complex lexemes are the products of a formation process actually entails a dynamic perspective (see article 16 on derivation). While a synchronic description can in principle rely solely on observing paradigmatic relations between words sharing the same apparently productive elements, complex lexemes, much more obtrusively than unmotivated simple lexemes, evoke the impression that they are the effects of a process (cf. Hansen et al. 1990: 31–32). As in all fields of linguistic inquiry, typological word-formation research compares different languages with the aim of identifying similarities and differences or producing a universally valid description of word-formation in language as such.

1. The scope of word-formation research Word-formation research has been carried out from a semasiological perspective, which has been the dominant one, but also from an onomasiological one (e.g., Štekauer 1998, 2005; see article 6 on word-formation in onomasiology). Semasiological investigations aim to describe the structures and meanings of complex lexemes and wordformation patterns. In doing so, they start out from an examination of existing complex lexemes – e.g., explanation, combination, description – which share the same morphological structure and are therefore hypothesized, at least metaphorically speaking, to share a common formation history (roughly [V + (a)tion]N). Therefore, it is the analytical aspect of a hearer or reader confronted with a complex lexeme – or indeed the linguist trying to discover structural generalizations – that comes to the fore in semasiological approaches. The complementary onomasiological perspective takes into consideration the fact that competent speakers are not only able to segment complex words into their constituents, but also use the results of such analyses for their own generative potential to create new words. The onomasiological perspective reflects speakers’ states of mind while trying to encode a given conceptualization by means of applying a word-formation model. What, for instance, are the speaker’s choices when aiming to encode the result of an action by deriving a noun from a verb: -ation as in transformation ← transform; -ment as in achievement ← achieve; -ence as in existence ← exist; -ing as in painting ← paint; -al as in denial ← deny; -age as in blockage ← block? The onomasiological approach thus corresponds to the synthetic aspect of word-formation. The semasiological and the onomasiological approaches differ in terms of their potential to be applied for descriptive purposes. Due to the virtually unlimited range of meanings that can be expressed by means of compounds, especially root compounds, the semasiological perspective has been dominant in this field. Systematic descriptions of prefixation and suffixation can, in principle, be arranged from a semasiological angle, by providing the meanings associated with the different prefixes and suffixes depending on the types of bases, or from an onomasiological one, by listing, for example, the prefixes that encode negation, location or time, or the suffixes that turn verbs into person-denoting nouns.

5. Levels of analysis and description in word-formation research Due to its position at the interface or crossroads of many different aspects of language, word-formation research potentially includes the full range of linguistic levels, from phonetics and phonology to syntax, semantics, pragmatics and sociolinguistics. Morphological considerations related to the constituents of complex lexemes and the formation rules and schemas naturally form the core of word-formation research, since a solid understanding of the composition of complex lexemes and their internal structures is required for all other levels of investigation (see article 15 on units of word-formation). Semantic aspects are of key importance not only for the internal segmentation and structural analysis of complex lexemes, but also for describing the semantic links between constituents and the semantic characteristics of word-formation processes and patterns such as compounding or derivation (see articles 5 on word-formation in inhaltbezogene Grammatik, 49 on argument-structural restrictions on word-formation patterns and 58 on categories of word-formation). Phonetic, phonological and morphonological aspects are equally important from a descriptive, an analytical and a heuristic perspective. In the

9

10

I. Word-formation as a linguistic discipline field of derivation, for example, it is crucial to understand and model regularities regarding stem allomorphy, e.g., systematic changes in the stress patterns and vowel qualities of bases and derivatives, as, for example, in English explain → explanation or sane → sanity (see articles 16 on derivation and 50 on phonological restrictions on English wordformation). Stress is also considered a diagnostic for compound status in English and other languages (cf. Bauer 1988a; Giegerich 2009). Syntactic aspects of word-formation came into focus with attempts to transfer insights from syntactic structures and rules to the internal grammar of words in early generative grammar (see article 7 on wordformation in generative grammar). They also provide important insights into word-classspecific restrictions on word-formation rules and their productivity (cf. Plag 1999). The remaining levels to be mentioned here hold a less traditional position in wordformation research. This does not mean that they are less important, however. Over the past two or three decades, psycholinguistic and neurolinguistic research has greatly enriched the field of word-formation research by probing the extent to which models of word-formation are realistic and plausible from a psychological and neuronal perspective (see article 13). Questions that have been addressed from this perspective include the way in which novel and lexicalized complex lexemes, especially compounds, derivations and blends, are represented in the mind and the brain, and how they are processed in actual usage (e.g., Libben and Jarema 2006; Schmid 2008). More recently, the cognitivelinguistic perspective focusing on the way in which knowledge of word-formation models and schemas as well as of individual complex lexemes becomes entrenched and is influenced by general cognitive abilities such as categorization and figure-ground segregation has gained in importance (cf., e.g., Ungerer 2002 and 2007; Heyvaerts 2003 and 2009; Onysko and Michel 2010; Schmid 2011). The sociolinguistic perspective looking at word-formation in diverse regional and social varieties (e.g., Biermeier 2008; Braun 2009) and the pragmatic perspective (Downing 1977; Bauer 1979; Clark and Clark 1979; Schmid 2011) highlighting interactional contexts and communicative functions in actual usage-events have also been gaining momentum in the field of word-formation research. This includes the study of lexical creativity in various registers and text-types (Munath 2007) involving a text-related and discourse-related perspective (see chapter XIV).

6. Theoretical models of word-formation The first chapter of this handbook (see articles 3 to 12) presents models of word-formation proposed in different theoretical frameworks, ranging from the historical-comparative tradition and structuralism to categorial grammar, generative grammar, natural morphology, optimality theory, cognitive grammar and construction grammar. As one would expect, in each case word-formation is modelled in line with the aims and assumptions typical of the corresponding approaches to linguistics in general. While categorial grammar and generative grammar focus on formal aspects of word-formation, natural morphology, optimality theory, cognitive grammar and construction grammar take a functional stance. As generative grammar favours a modular architecture of language (see article 7), a key issue has been to clarify whether word-formation belongs to syntax or to the lexicon or is an interface or even a module in its own right. Cognitive grammar

1. The scope of word-formation research (see article 10) and construction grammar (see article 12), on the other hand, favour a holistic conception of word-formation processes which unites morphological, syntactic, semantic, cognitive and even pragmatic aspects. Natural morphology (see article 9) and optimality theory (see article 11) attempt to develop explanatory frameworks which are of cross-linguistic or even universal validity and relevance rather than describing individual processes. Supporters of the various theoretical models advocate very different conceptions of how word-formation processes are to be modelled. A frequently quoted early classification of these conceptions was proposed by Hockett (1954), who distinguished between three types of models: item and arrangement, item and process, and word and paradigm. The first approach aims to describe the patterns of word-formation starting out from listing morphemes and describing “the arrangements in which they occur relative to each other in utterances” (Hockett 1954: 212). Secondly, item and process models, as suggested by the term, highlight the procedural aspects of word-formation; they regard the root, rather than the morpheme, as the basic input to morphological processes. Thirdly, word and paradigm models, which tend to focus on inflectional rather than derivational morphology, consider unsegmented words as the basic unit of word-formation and try to disclose paradigmatic similarities by comparing them. Moving to the present state of the art, the range of current theories which aim to model word-formation patterns and processes can be broadly divided into four types of approaches: – – – –

rule-based models schema-based models exemplar-based models exemplar-cum-schema-based models

The most prominent representatives of rule-based models are the different variants of generative approaches. Much of the research carried out in this framework is concerned with devising general rules applying across different word-formation types as well as type-specific rules or rule schemas, as they are called, in such a way that they can account for the empirical facts largely gleaned from introspection. The precise way in which these rules have been formulated has changed in line with the different stages of generative grammar: transformations and rewriting rules expressed in analogy to the phrase-structure rules of early generative grammar (Chomsky and Halle 1968; Selkirk 1982) were followed by applications of X-bar-theoretical principles (e.g., Scalise 1986) and sets of projection rules compatible with the government-and-binding and the principles-and-parameters approaches (see article 7 on word-formation in generative grammar and Lieber and Mugdan 2000 for more details). What rule-based approaches share is their focus on structural and formal rather than semantic or functional aspects, their commitment to the modularity assumption, giving rise to the need to identify the place where word-formation is situated in the architecture of language and linguistic knowledge, and their goal to formulate maximally generalizable insights and predictions. Well-known general hypotheses that have been postulated for the field of word-formation include, next to the binary-branching hypothesis and the right-hand head rule already mentioned, the unitary base hypothesis (Aronoff 1976: 47–48; Scalise 1984: 137–146) and the unitary output hypothesis (Scalise 1984: 137), both of which have turned out to be problematic, however. The former states that the syntactic and semantic properties of

11

12

I. Word-formation as a linguistic discipline the bases of derivational rules are clearly specified and unique, which implies that affixes cannot be attached to words of different word-classes. The latter states that affixes must be functionally and semantically unitary in the sense that they cannot be attached to different word-classes and bring about different changes in meaning. In both cases, potential empirical counter-evidence has been dealt with by maximizing the homonymy of affixes. This means, for instance, that the form -able in the deverbal adjective manageable and the denominal adjective marriageable (Bauer, Lieber and Plag 2013: 635–636) and the form -ese in the noun Japanese and the adjective Japanese would each have to be considered as two different but homonymic affixes. As a concrete example of the rule-based approach to describing individual wordformation models, the rule schema proposed by Aronoff (1976: 63) for English adjectival prefixation with un- is provided in (1). (1)

“Rule of negative un# a. [X]Adj → [un#[X]Adj]Adj semantics (roughly) un#X = not X b. Forms of the base 1. XVen (where en is the marker for the past participle) 2. XV #ing 3. XVable 4. X+y (worthy) 5. X+ly (seemly) 6. X#ful (mindful) 7. X#like (warlike)” (Aronoff 1976: 63; original italics)

The endeavours of schema-based models are also directed towards reaching generalizations, but these are not couched in the form of rules, but are instead formulated in terms of (constructional) schemas (see article 45 on rules, patterns and schemata in wordformation). These are defined as schematic form-meaning pairings representing lexical and phrasal knowledge and sanctioning concrete uses of complex lexemes. While rules are essentially variable procedural instructions, schemas are unit-like elements containing variable slots (Booij 2010: 41–43). In this way, schemas account for productivity and analyzability as well as creativity in word-formation. Schemas are considered to be connected by formally and semantically motivated hierarchical relations, yielding multidimensional networks of schemas and sub-schemas arranged on several levels of specificity (Ryder 1994; Tuggy 2005: 248–264). Schemas on lower levels inherit information from superordinate schemas. In contrast to rule-based approaches, schema-based models subscribe to a holistic, non-modular conception of linguistic knowledge and therefore unite and integrate formal structural, semantic and functional aspects in their accounts of schemas. While rule-based models tend to keep up the strict separation of grammar and usage, schema-based models are compatible with usage-based approaches, which assume that linguistic knowledge emerges from the experience of concrete usage-events in social situations and is subject to the frequencies of occurrence of certain elements (cf., e.g., Kemmer 2003; Bybee 2010). Figure 1.3 provides an idealized representation of a schema-based network for un-adjective formation inspired by Tuggy (2005) and Booij (2010):

1. The scope of word-formation research

Fig. 1.3: Idealized illustration of a schema-based network for adjectival un-prefixation in English

In Figure 1.3, schemas are represented by boxes detailing formal properties above the dividing horizontal lines and meanings below them. Boxes indicate the status of symbolic units, with bolder boxes marking hypothetically more entrenched schemas. The top-level schema in Figure 1.3 represents the most general prefixation schema which is not specified with regard to the form and the meaning of the base. In line with Schmid (2011: 160–162), the meaning of the pattern of prefixation as such, i.e. across different prefixes and different word-classes of the base, is glossed as expressing a contrast to the base. The sub-schemas represent increasingly specific information, from un-prefixation applicable not only to adjectives but also verbs and nouns to adjectival un-prefixation, different types of un-adjectives and, finally, individual words sanctioned by these schemas. Schema-based and exemplar-based models have more in common than rule-based and schema-based models. Like schema-based models, exemplar-based models (Krott, Schreuder and Baayen 2002; Eddington 2004: 71−98; Bybee 2010: 14−33; 165−193; Bybee and Beckner 2010; Arndt-Lappe 2011) conceive of lexical knowledge and knowledge about word-formation processes in terms of associative networks that are continuously reorganized under the influence of usage and exposure. The main difference between the two approaches is that strong versions of exemplar-based approaches, especially connectionist ones (e.g., Skousen 1992; Skousen, Lonsdale and Parkinson 2002), deny the existence of symbolic representations such as schemas and assume that linguistic knowledge is only available in the form of representations of individual exemplars stored under the impression of specific usage events. In theory at least, the structures of exem-

13

14

I. Word-formation as a linguistic discipline plar-based networks are much more complex than those of schema-based ones, since the nodes are individual exemplars related by means of similarity relations, ideally on all possible levels of description, i.e. meaning, morphological form and phonological form. The structure of networks – in terms of the strengths of nodes and the links between them – is considered to be influenced by probabilistic information extracted from the frequencies of exemplars. Instead of rules or schemas, analogies based on similarities are considered to play the key role of motor and motivation for productive and creative processes. Some network models (e.g., Bybee 2010) try to reconcile characteristics of schemabased and exemplar-based models. They postulate the existence of schemas representing experience that has been abstracted from usage-events without denying the importance of exemplar-based knowledge and thus argue that multiple and redundant representations on different levels of schematicity co-exist, from general schemas all the way down to individual lexemes. This way, they accommodate the potential to form new complex lexemes both on the basis of productive schemas and also by means of analogical formations based on similarities to stored exemplars (Eddington 2006). Like pure exemplarbased models, such exemplar-cum-schema-based models keep track of usage frequencies of exemplars but have a more differentiated view of the effects of frequency of exposure. While token repetition is considered to reinforce the representation of exemplars, type repetition strengthens schemas sanctioning novel uses (Bybee 2010). Figure 1.4 takes up English adjectival un-prefixation once more and provides an exemplary and idealized fragment of an exemplar-cum-schema network indicating form-based clusters of exemplars which are candidates for schemas.

Fig. 1.4: Idealized schema-cum-exemplar-based network representation of English adjectival unprefixation (Clusters: (1) un-Adj; (2) un-X-able (various morphological structures); (3) un-V-able; (4) un-Adj(-y) (various morphological structures); (5) un-V-D2 (various allomorphs of D2); un-V-D2, only /ɪd/ allomorph; (7) un-N(-al); (8) un-X-al (various morphological structures); (9) un-N-ful; (10) un-Adj(monosyllabic); (11) un-Adj(bisyllabic, ′--)

1. The scope of word-formation research The figure renders information on frequency of occurrence of exemplars in terms of the size of lexeme labels and the frequencies of occurrence in the British National Corpus. Clusters are labelled with numbers and glossed in the figure caption. The lexemes not included in clusters are – at least in this illustrative fragment – only related to other exemplars by the very general commonality of un-Adj. It should be noted that since exemplar-based networks are multi-dimensional, the figure mirrors only one of a large number of aspects contributing to the coherence of the network.

7. Modelling dynamic aspects of word-formation: productivity and lexicalization As noted in section 4, word-formation research has an intrinsic dynamic element, since the very notion of word-formation itself evokes the image of a process. Two complementary types of observations reinforce the need to integrate dynamic aspects into wordformation research. The first type is illustrated by words such as E. depthN (← deepAdj) and widthN (← wideAdj) or G. FahrtN ‘act of driving, ride, journey’ (← fahrenV ‘to drive’) and NahtN ‘seam’ (← nähenV ‘to sew’). While these words are morphologically analysable as bases and suffixes – E. -thAdj and G. -tN – it seems unlikely that new words with the help of these suffixes would be formed in present-day English or German, respectively. This suggests that these words were, at some time, formed by means of nominal suffixes for English deadjectival and German deverbal derivation which are no longer productive today. In its most extreme form, the second type of observation concerns words that are known to be the result of a productive word-formation model but are no longer analysable from today’s perspective. English daisy (← OE dæges éage ‘day’s eye’) or lord (← OE hláfweard ‘loaf keeper’) are cases in point. The opacity of such words is usually due to long-term formal, i.e. morphological, phonological and orthographical changes. Semantic developments can bring about the similar, though less extreme effect that “a complex lexeme may be synchronically analysable but no longer motivated” (Lipka 2002: 95), either because the extra-linguistic denotata have changed (e.g., cupboard, blackboard) or because the meaning itself has undergone changes (e.g., holiday, lit. ‘holy day’). The first phenomenon has been investigated under the label of productivity (see article 47), the second under lexicalization (see article 58 on categories of word-formation). Research into productivity involves two major facets: firstly, the measurement and description of degrees of the productivity of different word-formation models and patterns. Various attempts have been made to come up with dictionary-based (e.g., Plag, Dalton-Puffer and Baayen 1999) and corpus-based measures of productivity, taking into consideration the number of lexemes manifesting a pattern, i.e. type-frequency (Baayen 1989), the relation of hapax legomena to type frequency (Baayen and Lieber 1991), and the relation of hapaxes to types, while additionally taking into account the distinction between hapaxes that are just rare and hapaxes which are indeed novel formations relying on a pattern (Plag 2006: 542). Corpus-based measures can also be applied to investigate diachronic changes in the productivity of individual patterns (see article 103; Scherer

15

16

I. Word-formation as a linguistic discipline 2005; Trips 2009; Schröder 2008, 2011) or of several patterns which are in onomasiological competition, e.g., -ity and -ness (cf. Gaeta and Ricca 2006; Baeskow 2012). The second major branch of research into productivity tries to provide maximally detailed descriptions of the general or pattern-specific limits or restrictions on productivity (see article 48). For example, a wide-ranging and in fact somewhat trivial restriction on productivity is that there must be a communicative need for a potential formation. Pattern-specific restrictions can be morphological, semantic, phonological, syntactic and etymological in nature or derive from semasiological or onomasiological competition (blocking) of existing lexemes (Kastovsky 1982: 159–164; Rainer 2005). Recent webbased research (e.g., Schröder and Mühleisen 2010) has indicated that many productivity restrictions are tendencies rather than hard and fast rules. Research in the field of the lexicalization of individual complex lexemes has been characterized by considerable variation in the choice of terms. Among the terms causing confusion are lexicalization, institutionalization, idiomatization, conventionalization, diffusion, spread, propagation, listing and establishment (cf. Brinton and Traugott 2005: 32–61). The term lexicalization itself, for example, has been conceptualized as “the process by which complex words come to have meanings that are not compositional” (Lieber 2010: 201), i.e. more or less synonymous with idiomatization; as “the strength of the lexical representation of a particular lexeme and its forms” (Bauer, Lieber and Plag 2013: 117), i.e. more or less synonymous with entrenchment; and by stating that “[w]hen a possible word has become an established word, we say that it has lexicalized” (Booij 2005: 17), i.e. in sociopragmatic terms. As this terminological diversity is partly caused by lumping together different levels of description, it seems desirable to separate these levels both conceptually and terminologically. Schmid (2011: 71–83) therefore proposes that the term lexicalization be reserved for the structure-oriented description of formal processes (fusion, reduction, erosion) and semantic processes (idiomatization) noticeable in the word itself. The sociopragmatic perspective focusing on how a given word spreads in the speech community is captured by the terms institutionalization and, more generally, conventionalization, while the cognitive perspective describing changes taking place in the mind, such as the strength of a unitary holistic representation or the density of the link to “neighbouring” words, is described in terms of degrees of entrenchment.

8. Conclusion In keeping with the nature of the opening contribution to any handbook, this article can supply no more than a broad-brush survey of the main issues in word-formation research. In doing so, it aims to chart the terrain for the much more fine-grained explorations offered in the subsequent articles. As has been shown, one of the beauties and challenges of word-formation research is that its ramifications branch out onto all levels of linguistic analysis, description and theorizing. Recently, the field has been marked by an increasing awareness that speakers and writers use the productive and creative resources of their language in a much more flexible and variable manner than had been predicted by rigid and overly reductionist models. Doing justice to this flexibility while upholding the aim to produce valid generalizations could be one of the major challenges to be faced in future research into word-formation.

1. The scope of word-formation research

9. References Aronoff, Mark 1976 Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Arndt-Lappe, Sabine 2011 Towards an exemplar-based model of stress in English noun-noun compounds. Journal of Linguistics 11: 549–585. Baayen, R. Harald 1989 A corpus-based approach to morphological productivity: Statistical analysis and psycholinguistic interpretation. Ph.D. dissertation, Vrije Universiteit, Amsterdam. Baayen, R. Harald and Rochelle Lieber 1991 Productivity and English word-formation. A corpus-based study. Linguistics 29: 801– 843. Baeskow, Heike 2012 -Ness and -ity: Phonological exponents of n or meaningful nominalizers of different adjectival domains? Journal of English Linguistics 40: 6–40. Baudouin de Courtenay, Jan N. 1895 Versuch einer Theorie phonetischer Alternationen. Straßburg: Trübner. Bauer, Laurie 1979 On the need for pragmatics in the study of nominal compounding. Journal of Pragmatics 3: 45–50. Bauer, Laurie 1983 English Word-Formation. Cambridge: Cambridge University Press. Bauer, Laurie 1988a When is a sequence of two nouns a compound in English? English Language and Linguistics 2: 65–86. Bauer, Laurie 1988b Introducing Linguistic Morphology. Edinburgh: Edinburgh University Press. Bauer, Laurie, Rochelle Lieber and Ingo Plag 2013 The Oxford Reference Guide to English Morphology. Oxford: Oxford University Press. Benveniste, Émile 1967 Fondements syntaxiques de la composition nominale. Bulletin de la Société de Linguistique de Paris 62: 15–31. Biermeier, Thomas 2008 Word-formation in New Englishes. A corpus-based analysis. Berlin: LIT. Bloomfield, Leonard 1926 A set of postulates for the science of language. Language 2: 153–164. Bloomfield, Leonard 1933 Language. London: Allen & Unwin. Booij, Geert 1977 Dutch Morphology. A Study of Word Formation in Generative Grammar. Dordrecht: Foris. Booij, Geert 2000 Inflection and derivation. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 360–369. Berlin/New York: de Gruyter. Booij, Geert 2005 The Grammar of Words. An Introduction to Morphology. Oxford: Oxford University Press. Booij, Geert 2010 Construction Morphology. Oxford: Oxford University Press.

17

18

I. Word-formation as a linguistic discipline Braun, Maria 2009 Word-Formation and Creolisation. The Case of Early Sranan. Tübingen: Niemeyer. Brinton, Laurel J. and Elizabeth Closs Traugott 2005 Lexicalization and Language Change. Cambridge/New York: Cambridge University Press. Bybee, Joan L. 1985 Morphology. A Study of the Relation between Meaning and Form. Amsterdam/Philadelphia: Benjamins. Bybee, Joan 2010 Language, Usage and Cognition. Cambridge/New York: Cambridge University Press. Bybee, Joan L. and Clay Beckner 2010 Usage-based theory. In: Bernd Heine and Heiko Narrog (eds.), The Oxford Handbook of Linguistic Analysis, 826–855. Oxford: Oxford University Press. Chelliah, Shobhana L. and Willem J. de Reuse 2011 Handbook of Descriptive Linguistic Fieldwork. Dordrecht: Springer. Chomsky, Noam and Morris Halle 1968 The Sound Patterns of English. New York: Harper and Row. Clark, Eve and Herbert H. Clark 1979 When nouns surface as verbs. Language 55: 767–811. Downing, Pamela 1977 On the creation and use of English compound nouns. Language 53: 810–842. Eddington, David 2004 Spanish Phonology and Morphology. Experimental and Quantitative Perspectives. Amsterdam/Philadelphia: Benjamins. Eddington, David 2006 Spanish diminutive formation without rules or constraints. Linguistics 40: 395–419. Firth, John R. 1964 The Tongues of Men & Speech. Oxford: Oxford University Press. Fleischer, Wolfgang 2000 Die Klassifikation von Wortbildungsprozessen. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 886–897. Berlin/New York: de Gruyter. Fleischer, Wolfgang and Irmhild Barz 2012 Wortbildung der deutschen Gegenwartssprache. 4th ed. Berlin/Boston: de Gruyter. Gaeta, Livio and Davide Ricca 2006 Productivity in Italian word formation. A variable-corpus approach. Linguistics 44: 57− 89. Giegerich, Heinz J. 2009 Compounding and lexicalism. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 178−200. Oxford: Oxford University Press. Giegerich, Heinz J. 2012 The morphology of -ly and the categorial status of ‘adverbs’ in English. English Language and Linguistics 16: 341–359. Hansen, Barbara, Klaus Hansen, Albrecht Neubert and Manfred Schentke 1990 Englische Lexikologie. 2nd ed. Leipzig: Enzyklopädie. Heyvaert, Liesbeth 2003 A Cognitive-Functional Approach to Nominalization in English. Berlin/New York: Mouton de Gruyter. Heyvaert, Liesbeth 2009 Compounding in cognitive linguistics. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 233–254. Oxford: Oxford Univerity Press.

1. The scope of word-formation research Hockett, Charles F. 1954 Two models of grammatical description. Word 10: 210–231. Kastovsky, Dieter 1982 Wortbildung und Semantik. Düsseldorf: Bagel & Francke. Kastovsky, Dieter 1999 English and German morphology. A typological comparison. In: Wolfgang Falkner and Hans-Jörg Schmid (eds.), Words, Lexemes, Concepts – Approaches to the Lexicon. Studies in Honour of Leonhard Lipka, 39–52. Tübingen: Narr. Kastovsky, Dieter 2009 Diachronic perspectives. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 323–340. Oxford: Oxford Univerity Press. Kemmer, Suzanne 2003 Schemas and lexical blends. In: Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.), Motivation in Language. Studies in honor of Günter Radden, 69–97. Amsterdam/Philadelphia: Benjamins. Krott, Andrea, Robert Schreuder and R. Harald Baayen 2002 Analogical hierarchy: Exemplar-based modeling of linkers in Dutch noun-noun compounds. In: Royal Skousen, Deryle Lonsdale and Dilworth B. Parkinson (eds.), Analogical Modeling, 181–206. Amsterdam/Philadelphia: Benjamins. Kubrjakova, Elena S. 2000 Submorphemische Einheiten. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 417–426. Berlin/New York: de Gruyter. Lehrer, Adrienne 1996 Identifying and interpreting blends: An experimental approach. Cognitive Linguistics 7: 359−390. Libben, Gary and Gonia Jarema (eds.) 2006 The Representation and Processing of Compound Words. Oxford: Oxford University Press. Lieber, Rochelle 1990 On the Organization of the Lexicon. New York/London: Garland. Lieber, Rochelle 1992 Deconstructing Morphology. Word Formation in Syntactic Theory. Chicago: Chicago University Press. Lieber, Rochelle 2010 Introducing Morphology. Cambridge: Cambridge University Press. Lieber, Rochelle and Joachim Mugdan 2000 Internal structure of words. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 404–416. Berlin/New York: de Gruyter. Lipka, Leonhard 1983 A multi-level approach to word-formation. Complex lexemes and word semantics. In: Shirô Hattori and Kazuko Inoue (eds.), Proceedings of the XIIIth International Congress of Linguists, Tokyo 1982, 926–928. Tokyo: The Committee. Lipka, Leonhard 2002 English Lexicology. Lexical Structure, Word Semantics and Word Formation. Tübingen: Narr. Luschützky, Hans Christian 2000 Morphem, Morph und Allomorph. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 451–462. Berlin/New York: de Gruyter.

19

20

I. Word-formation as a linguistic discipline Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Mugdan, Joachim 1986 Was ist eigentlich ein Morphem? Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 39: 29–43. Munat, Judith (ed.) 2007 Lexical Creativity, Texts and Contexts. Amsterdam/Philadelphia: Benjamins. Olsen, Susan 2000 Compounding and stress in English: A closer look at the boundary between morphology and syntax. Linguistische Berichte 181: 55–69. Onysko, Alexander and Sascha Michel (eds.) 2010 Cognitive Perspectives on Word-Formation. Berlin/New York: Mouton de Gruyter. Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Plag, Ingo 2003 Word-Formation in English. Cambridge: Cambridge University Press. Plag, Ingo 2006 Productivity. In: Bas Aarts and April McMahon (eds.), The Handbook of English Linguistics, 537–556. Oxford: Blackwell. Plag, Ingo, Christiane Dalton-Puffer and Harald Baayen 1999 Morphological productivity across speech and writing. English Language and Linguistics 3: 209–228. Plank, Frans 1994 Inflection and derivation. In: Ronald E. Asher (ed.), Encyclopaedia of Language and Linguistics. Vol. 3, 1671–1678. Oxford: Pergamon. Rainer, Franz 2005 Constraints on productivity. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 335–352. Dordrecht: Springer. Ryder, Mary Ellen 1994 Ordered Chaos. The Interpretation of English Noun-Noun Compounds. Berkeley: University of California Press. Scalise, Sergio 1984 Generative Morphology. Dordrecht: Foris. Scalise, Sergio 1986 Generative Morphology. Berlin/New York: de Gruyter. Scalise, Sergio 1988 Inflection and derivation. Linguistics 26: 561–581. Scherer, Carmen 2005 Wortbildungswandel und Produktivität. Eine empirische Studie zur nominalen -er-Derivation im Deutschen. Tübingen: Niemeyer. Schmid, Hans-Jörg 2008 New words in the mind: Concept-formation and entrenchment of neologisms. Anglia 126: 1−36. Schmid, Hans-Jörg 2011 English Morphology and Word-formation. An Introduction. Berlin: Schmidt. Schmidt, Günter Dietrich 1987 Das Affixoid. Zur Notwendigkeit und Brauchbarkeit eines beliebten Zwischenbegriffes der Wortbildung. In: Gabriele Hoppe, Alan Kirkness, Elisabeth Link, Isolde Nortmeyer, Wolfgang Rettig and Günter Schmidt (eds.), Deutsche Lehnwortbildung. Beiträge zur

1. The scope of word-formation research

21

Erforschung der Wortbildung mit entlehnten WB-Einheiten im Deutschen, 53–101. Tübingen: Niemeyer. Schröder, Anne 2008 Investigating the morphological productivity of verbal prefixation in the history of English. Arbeiten aus Anglistik und Amerikanistik 33: 47–69. Schröder, Anne 2011 On the Productivity of Verbal Prefixation in English. Synchronic and Diachronic Perspectives. Tübingen: Narr. Schröder, Anne and Susanne Mühleisen 2010 New ways of investigating morphological productivity. Arbeiten aus Anglistik und Amerikanistik 35: 43–58. Selkirk, Elisabeth 1982 The Syntax of Words. Cambridge, MA: MIT Press. Skousen, Royal 1992 Analogy and Structure. Dordrecht: Kluwer. Skousen, Royal, Deryle Lonsdale and Dilworth B. Parkinson (eds.) 2002 Analogical Modeling. Amsterdam/Philadelphia: Benjamins. Štekauer, Pavol 1998 An Onomasiological Theory of English Word Formation. Amsterdam/Philadelphia: Benjamins. Štekauer, Pavol 2005 Onomasiological approach to word-formation. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 207–232. Dordrecht: Springer. Stump, Gregory T. 2005 Word-formation and inflectional morphology. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 49–71. Dordrecht: Springer. Trips, Carola 2009 Lexical Semantics and Diachronic Morphology. The development of -hood, -dom and -ship in the history of English. Tübingen: Niemeyer. Tuggy, David 2005 Cognitive approach to word-formation. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 233−265. Dordrecht: Springer. Ungerer, Friedrich 2002 The conceptual function of derivational word-formation in English. Anglia 120: 534– 567. Ungerer, Friedrich 2007 Derivational morphology and word-formation. In: Dirk Geeraerts and Hubert Cuyckens (eds.), The Oxford Handbook of Cognitive Linguistics, 991–1025. Oxford: Oxford University Press. Williams, Edwin 1981 On the notions of ‘lexically related’ and ‘head of a word’. Linguistic Inquiry 12: 245– 274.

Hans-Jörg Schmid, Munich (Germany)

22

I. Word-formation as a linguistic discipline

2. Word-formation research from its beginnings to the 19th century 1. 2. 3. 4. 5.

Introduction Greek and Roman antiquity The German tradition (16th–19th century) The French tradition (16th–19th century) References

Abstract In Greco-Latin grammatical theory as well as in the first grammars of European vernacular languages, word-formation was treated within the presentation of word classes. This article aims to show how specific theories of derivation and composition were progressively developed by grammarians of German and French.

1. Introduction This article starts with a brief survey of research on word-formation in Greek and Roman antiquity (Aristarchus, Dionysius Thrax, Quintilian, Varro). The following sections treat theories developed in modern times by major grammarians of German (Schottelius, Adelung, Becker, Grimm, Paul) and French (Meigret, Arnauld and Lancelot, Beauzée, Butet de la Sarthe, Darmesteter). Our overall focus lies on salient aspects, such as the distinction between derivation and composition and the place of word-formation in grammatical theory.

2. Greek and Roman antiquity The assumption that there are two basic processes of word-formation, derivation and composition, still widely held in contemporary linguistics (Lallot 2008: 63; McLelland 2010: 13), may be traced back to ancient Greek linguistic thought (Vaahtera 1998: 60− 76). Aristarchus’s analysis of complex words in Greek is based upon the distinction between derivatives and compounds. The latter are formed by combining two or more autonomous words, the first by adding derivational suffixes (considered to be non-autonomous) to a word (Matthaios 1999: 254−272; summarized in Matthaios 2004: 16−20). This idea was later developed more fully by Dionysius Thrax (Τέχνη, ed. Lallot 1998). It must be stressed that ancient Greek grammarians did not view word-formation as a separate domain; rather, they treated it within the framework of word class theory, on the basis of categories describing properties (“accidents”, e.g., number and gender) of word classes (Matthaios 2004: 8−9). Of these accidents, two are specific to word-formation, “είδος” [species] and “σχήμα” [figure]. Both were attributed to various word classes; the most detailed description was made for the nominal class (Matthaios 2004: 11−

2. Word-formation research from its beginnings to the 19th century 16). The category species allows for a distinction between “primary” and “derived” lexical items, while figure is the basis for distinguishing “simple”, “composed” and “decompound” words, the latter term designating nominal derivatives from compounds. (Greek grammarians used the term “parasyntheta”, in Latin grammar this latter term was rendered as “decomposita”; cf. article 29 on parasynthesis in Romance.) This view of word-formation as part of word class theory, including the fundamental dichotomy composition – derivation, was then adopted by grammarians of classical Latin, who latinized the Greek terms pertaining to word-formation (e.g., “species primitiva”/“derivativa”, “figura simplex”/“composita”/“decomposita”). A discussion of complex words in classical Latin is to be found in writings by Quintilian, Priscian, and Varro, among others (Fögen 2008; Lindner 2002). Classical and medieval Latin grammaticography (entitled “De partibus orationis”) is characterized by a canonized system of “parts of speech” (word classes), word-formation is considered strictly within this framework, with a focus on derivation. In early modern times, this word class system in turn constituted the theoretical framework for the process of “grammatisation” (Auroux 1992: 11–64) of vernacular languages (Kaltz 2000: 693–698).

3. The German tradition (16th–19th century) The development of word-formation theory in German grammaticography from the 16th to the 19th century may be briefly summarized as follows. In the first stage of the “grammatisation” process (16th–early 17th century), grammarians deal with derivation and composition strictly in terms of the categories “species” and “figura”, defined as accidents of word classes of German (see section 3.1). 17th-century grammarians generally continue to present German word-formation within the section on word classes (“etymologia”, “Wortforschung”) though some (Ratke, Schottelius) begin to conceive of word-formation as being a grammaticographical topic in its own right (see section 3.2). In the 18th century, theories of word-formation in German gradually gain more independence from the Latin tradition, particularly with Mäzke, Heynatz, and Adelung (see section 3.3). 19th-century linguists (Becker, Grimm, Paul, among others) contribute to further development and refinement of word-formation theory as a domain of its own (see section 3.4).

3.1. 16th–early 17th century Early grammars of German, mostly written in Latin (e.g., Albertus, Ölinger, Clajus) closely adhere to the conventions of Latin grammar. The focus is on derivation by “terminationes” (‘endings’) while authors of grammar books pay little attention to composition. Complex words formed by the addition of “praepositiones”, i.e. either separable or inseparable prefixes, are analyzed as compounds, as are “decomposita”, resulting from both prefixation and suffixation: “gerecht” (“figura simplex”), “ungerecht” (“figura composita”), “Ungerechtigkeit” (“figura decomposita”; Clajus 1894 [1578]: 43). Albertus occasionally refers to the notion of “radix” (“root”; 1573 [1895]: 66, 74); originally elaborated by grammarians of Hebrew, unknown to the tradition of Greek and Roman grammar

23

24

I. Word-formation as a linguistic discipline (Kaltz 2005: 107–111), this notion was to become a key concept in Schottelius’s theory (see section 3.2).

3.2. 17th century While the latinizing tradition is still present in early 17th-century German grammaticography (Padley 1985: 98; note 44), more grammarians now make use of the vernacular, for pedagogical reasons (e.g., Brücker 1620: 5; Ratke 1959 [1612–1615], 1959 [1630]) as well as ideological motives (legitimation of the German language; e.g., Schottelius 1641). In his Wortbedeütungslehr [Treatise on the meaning of words], Ratke redefines the traditional categories “Art” [species] and “Gestalt” [figura], arguing that derivation and composition are essential means by which the proper and necessary meaning of discourse is grasped (“wesentliche Mittel [...] darauß die Bedeütung der Rede eigentlich vnd nothwendig erkennet wird”; 1959 [1630]: 277). This work, “the first attempt to provide a semantically based theory of word derivation applied to the vernacular” (Padley 1985: 112), is also the first to deal with word-formation not just in the framework of “Wortforschung”, but rather as a topic in its own right (Kaltz 2005: 113–116). The central figure among 17th-century German grammarians is undoubtedly Schottelius (cf. Gützlaff 1989; McLelland 2010, 2011); he had a major impact on late 17th and early 18th-century grammaticography and lexicography (Kaltz 2005: 117–125). While Schottelius (1663) follows tradition by discussing derivation and composition within the section on word classes (“Wortforschung”), he also addresses the topic of word-formation separately. In three Lobreden [discourses of praise], he praises the German language for its abundance of “Stammwörter” or “Wurzelen” [root words], which are monosyllabic, he claims. This section of his grammar may be seen as the first attempt to describe the constituents of complex words in a somewhat systematic manner. Schottelius analyses both derivatives and compounds as being formed on the basis of root words (1663: 49–103); the first result from the addition of “Haubtendungen”, i.e. derivative suffixes, which are now differentiated from inflectional ones (“zufällige Endungen”). The binary structure of compounds, formed by combining “Stammwörter” with suffixes, prefixes or other root words, is described by the terms “Beygefügtes/vorderstes Glied des Wortes” ([attributive/first element of the word]; i.e. determiner) and “Grund/Haubtglied” ([basis], [main element]; i.e. primary word; 1663: 75). According to Schottelius, there are four “Verdoppelungs=arten” [types of compounds] in German. The first class, formed by nouns only (“welche aus lauter Nennwörteren entstehet”; 1663: 77), includes determinative compounds as well as copulative ones (e.g., Freudenpein; 1663: 79); compounds of the structure “Nennwort + Zeitnennwort” (i.e. a deverbal noun) form the second (e.g., Mordbrenner, Taglöhner; 1663: 84). The third class assembles compounds formed by adding a preposition to verbs or nouns (“wan durch die Vorwörter (Praepositiones) die Verdoppelung geschiehet”; primarily verbs such as erheirathen, einbrokken, vertieffen, but nouns as well, e.g., Gezisch, Gesäusel; 1663: 88 f.) while combinations of one or two Haubtendungen with one or two root words constitute the last (“wan mit einem oder zweyen Stammwörtern eine oder zwo Haubtendungen der abgeleiteten verdoppelt werden”, e.g., weibisch, Jüngling, Hofnung, Gottseligkeit; 1663: 90). Contrary to present-day linguistics, “Ableitung” [derivation] refers to suffixation only here (1663:

2. Word-formation research from its beginnings to the 19th century 318) whereas the term “Verdoppelung” [composition] is used for composition proper and prefixation, and includes the formation of complex verbs with inseparable prefixes.

3.3. 18th century Among early 18th-century grammarians, Longolius is worth mentioning as he maintains the distinction between “Primitiva” [root words], “Derivativa”, and “Composita” (also labelled “zusammengestückte Wörter”; 1715: 63) but rejects the idea of the monosyllabic nature of root words (1715: 617 f.). Following Schottelius, he considers the last word (“das letzte Wort”) to be the main element (“Grund”), which is merely restricted and determined by the [word] added (“welchen das beygefügte nur gewisser Maaßen restringiret und determiniret”; 1715: 623). His analysis of nominal compounds, however, is somewhat more elaborate than that of Schottelius; in particular, Longolius notes that verb stems may be first constituents of nominal compounds: “Ein Teutsches Substantivum, das mit einem Verbo componiret ist/bedeutet eine Sache von einem Vermögen zu derjenigen Verrichtung/so besagtes Verbum anzeigen” [a German noun that is combined with a verb denotes the ability to perform the action referred to by the said verb], e.g., Lockvogel, Heuchelchrist, Hakkebret (1715: 623 f.). The term Decomposita is used here for complex compounds such as Erzpfaffenfreund and stockpechdickfinster (1715: 627). Like his predecessors, Aichinger (1754) differentiates “Stammwörter” [root words] and “abstammende” [derivatives], “einfache” [simple words] and “zusammengesetzte” [compounds] (Kaltz 2005: 130); he also discusses the “Gattung” and “Gestalt” of complex words in the traditional manner as accidents of word classes (1754: 136 f., 157). Aichinger points out an essential difference between derivation and composition in German: “niemand [darff] leicht sich selber deriuativa schmieden, ausser den Dichtern” ([with the exception of poets, no one has the right to form derivatives]; 1754: 139); on the other hand, he notes that speakers of German have the liberty to create new compounds every day (“die Freyheit, alle Tage neue zusammengesetzte Wörter zu machen”; 1754: 104). In the late 18th century, significant changes in the theory of word-formation occur as grammarians progressively turn away from the traditional perspective. Heynatz (1777 [1770]: 121) deserves special mention for his analysis of monosyllabic, non-suffixed deverbal derivatives such as Druck, Schlag, Trieb, while Mäzke (1776: 5 f.) criticizes the use of the traditional term “praepositiones inseparabiles” for inseparable German prefixes such as be, er, ent (Kaltz 2005: 132 f.). Adelung, the most interesting and most influential 18th-century grammarian of German, no longer argues in terms of “figura” and “species”. Just as many linguists nowadays, he considers prefixation to be a type of derivation: “Die Ableitung der Wörter geschiehet entweder durch Vorsylben oder durch Nachsylben, oder durch beide zugleich” ([derivation of words occurs either by prefixes or suffixes, or both]; 1784: 97 f.). Derivatives are described in detail within the context of word class theory while composition is dealt with in a separate lengthy section (Adelung 1782: II, 209– 274). Adelung stresses the special importance of compounds in German, noting that this has been a much neglected topic due to the excessive weight accorded to the Latin tradition (Kaltz 2002; 2005: 131). As some of his predecessors, he insists upon the binary structure of (determinative) compounds, including complex ones such as Fastnachtspiel,

25

26

I. Word-formation as a linguistic discipline which he describes with the terms “Grundwort” [basic/primary word] and “Bestimmungswort” ([determiner]; 1782: II, 215), still commonly used in modern studies of wordformation in German. Copulative compounds (e.g., Fürst=Bischof), on the other hand, are analyzed as a type of apposition (Adelung 1782: II, 229).

3.4. 19th century Throughout the 19th century, word-formation was a much discussed topic in general grammar as well as in pedagogical (e.g., Becker 1831) and historical-comparative grammar (e.g., Grimm 1878 and article 3 on word-formation in historical-comparative grammar). Becker, who represents both general and pedagogical grammar (Jankowsky 2004; Forsgren 2008: 134–135), is the author of the first comprehensive study dealing exclusively with word-formation in German (Deutsche Wortbildung, 1824; Kaltz 2005: 136). The author establishes a distinction between “Begriffswörter” and “Formwörter” ([notional words] vs. [form words]) on the one hand, “Wurzelwort” [root word] and “abgeleitetes Wort” [derivative] on the other. There are two types of derivatives, “Stamm” ([primary form], e.g., Bund) and “Sproßform” ([secondary form], e.g., bündig). Composition, Becker argues, must be seen as “wahrhafter Ableitungsvorgang” ([true process of derivation]; Becker 1824: 369 f.) since compounds are new words for new concepts formed on the basis of existing linguistic material (“Stoff”), the same way as derivatives stricto sensu are formed by “Ablautung” [ablauting] and “Umendung” ([change of ending], [suffixation]). In other words, Becker uses “Ableitung” as the comprehensive term for both derivation and composition. The terms “Bestimmungswort” and “Grundwort”, introduced by Adelung for the constituents of compounds, are reinterpreted by Becker: the first constituent of compounds is the principal one: “Hauptwort” [main word] whereas the second one is defined as “Beziehungswort” [relational word]. He does follow Adelung, however, as far as the principle of the binary structure of compounds is concerned, noting that it applies to complex compounds as well (e.g., Nußbaum-holz, Schneider-handwerk, Herzbeutel-wassersucht; Becker 1831: 45). J. Grimm had a major impact on 19th- as well as early 20th-century grammar; in fact, he was perceived as the initiator of scientific word-formation theory: “Die wissenschaftliche Wortbildungslehre ist [...] eine Schöpfung J. Grimms” (Paul 1896: 17). Grimm starts out by observing that word-formation theory in particular has been unduly neglected in traditional grammar (“zumahl die wortbildungslehre [ist] ungebührlich verabsäumt worden”; 1878: VI). Word-formation, he argues, occurs “durch innere änderung oder durch äußere mehrung der wurzel” [through internal change or external addition to the root]. “Zusammensetzung” and “Ableitung” differ as follows: “Zusammensetzung kann vorne oder hinten an der wurzel eintreten, ableitung nur hinten” ([composition may occur by combining elements preceding or following the root, derivation by elements following the root only]; 1878: 1). In contrast to “innere wortbildung” ([internal word-formation]; 1878: 1), derivation (i.e. suffixation only) and composition are dealt with in great detail (Kaltz 2005: 144–146). Grimm’s classification of compounds is based upon the distinction between “eigentliche” and “uneigentliche composition” ([proper]/[improper composition]; e.g., wein-stock, gras-grün vs. tages-licht; 1878: 386). In spite of his sharp criticism of traditional grammar, he does retain the term “decomposita” for both complex

2. Word-formation research from its beginnings to the 19th century compounds (e.g., schlafkammer-thürhüter) and complex derivatives such as un-absehlich (1878: 383, 912). Similarly to Adelung, Grimm stresses the binary structure of determinative compounds (including “decomposita”; 1878: 912), but does not give much consideration to copulative compounds such as christ-kind, tier-mensch and fürst-bischof, which result from “appositionelle verhältnisse” ([appositional relations]; 1878: 416). Grimm’s theory of word-formation was later refined by Paul and other Neogrammarians (“Junggrammatiker”; Fleischer 1983). As his 18th-century predecessors Mäzke and Adelung, Paul insists that it is impossible to draw a sharp line between inflection, derivation, and composition: “Die Scheidelinie zwischen Kompositionsglied und Suffix kann nur nach dem Sprachgefühl bestimmt werden” ([linguistic intuition is the only way to decide between a compound constituent and a suffix]; 1880: 348); “auf die gleiche Weise wie die Ableitungssuffixe entstehen Flexionssuffixe. Zwischen beiden gibt es ja überhaupt keine scharfe Grenze” ([inflectional suffixes arise the same way as derivational suffixes. In fact, there is no sharp line separating both]; 1880: 349). While the focus is on morphological aspects here, Paul later argues that more consideration should be given to semantics in word-formation theory; when analyzing complex words, “Funktionen” [functions] representing meaning, not “Bildungsweisen” [word-formation types] ought to be the primary criterion (1896: 18). In his Deutsche Grammatik (1920), Paul further distances himself from Grimm by rejecting the categories of “proper” and “improper” composition for nominal compounds as they are motivated by morphological considerations. Instead, referring explicitly to the classification of compounds in Sanskrit grammaticography (Brocquet 2008: 19), he now argues that the logical relation between constituents as well as that between the compound as a whole and its constituents must be decisive when classifying compounds (“nach dem logischen Verhältnis der Glieder zueinander und des Ganzen zu den Gliedern”; 1920: 6). This principle leads him to differentiate three types of nominal compounds (Kaltz 2005: 150): a) “kopulative” or copulative compounds (e.g., Prinzregent, Fürstbischof; = “dvandva” in Indian grammar; 1920: 7), b) “solche, in denen das zweite Glied durch das erste bestimmt wird” [compounds in which the second constituent is determined by the first], such as Senfsauce, Mitbürger (1920: 9, 23), and c) “possessive Zusammensetzungen” or possessive compounds (e.g., Lügenmaul, Rotkehlchen; = the Indian “bahuvrihi”; 1920: 30). With respect to the place of word-formation, Paul notes that 19th-century grammarians have chosen various options: following “Flexionslehre” [theory of inflection] and preceding syntax (e.g., Grimm), or preceding the former. Paul argues that the status of inflectional suffixes may change as a result of processes of isolation and that their syntactic function changes as they become “Wortbildungssuffixe” [word-formation suffixes]. Compounds, on the other hand, result from syntactic structures. Therefore, he suggests that the most appropriate way is to discuss word-formation in a separate section, following syntax (1920: 3 f.). In 20th-century German linguistics, the separate treatment of word-formation became predominant with the gradual shift from the historical perspective to the synchronic, though some grammarians still choose the traditional position by integrating wordformation into the description of the word class system (Kaltz 2005: 152).

4. The French tradition (16th–19th century) The history of word-formation theories in France from the 16th to the 19th century may be summarized as follows. In the first grammars of French, written in Latin or French,

27

28

I. Word-formation as a linguistic discipline word-formation is studied using the conceptual framework of Latin grammar. The richest developments on lexical morphology are undoubtedly to be read in the first grammar of French written in French, Le Trȩtté de la grammȩre françoȩze (Meigret 1980 [1550]) (see section 4.1). 17th-century grammarians of French no longer refer to the Latin categories of “espèce” [species] and “figure” [figure]. Even if the contrasts between primitive and derivative or between simple and compound are sometimes mentioned, general grammar assigns the task of dealing with derivation and composition to the dictionary (see section 4.2). Word-formation is then reintroduced into the field of grammar with Beauzée’s theses. The encyclopaedist’s approach to derivation and composition – based on the theory of the sign specific to general grammar – had a long-standing influence (see section 4.3). In the 19th century, word-formation theories are mostly developed outside the field of grammar, first and foremost in Darmesteter’s works (see section 4.4).

4.1. 16th century From Sylvius’s Grammatica latino-gallica (1531) to Ramus’s Grammaire (1572), derivation and composition were regarded as word accidents, in compliance with the GrecoRoman tradition, with the consequence that word-formation was integrated into grammar but had no real autonomy: it was included in the presentation of word classes. The species implies an opposition between “primitifs” and “dérivatifs” and corresponds to suffixal derivation whereas the figure opposes “simples” and “composés” and describes combinations of at least two words considered as autonomous words. Both the species and the figure are discussed along with the other word accidents – essentially morphosyntactic variations. The parts devoted to noun formation are by far the most developed and some elements pertaining to lexical morphology can also be read into the description of the category of “diminutif” [diminutive], traditionally dealt with separately from other suffixed derivatives. 16th-century grammarians pay variable attention to the species and the figure, most of them simply asserting the two dichotomies and exemplifying them. The processes of composition are sometimes studied more precisely. Estienne, for example, distinguishes three types of formations taken up from Donat (Colombat 1999: 243): two “whole words” (“mots entiers”; e.g., malheur), a “whole word” and a “corrupted word” (“mot corrompu”; e.g., ennemi), a “corrupted word” and a “whole word” (e.g., chascun; 1557: 17). Meigret, known for his interest in morphology (Glatigny 1985), added a fourth pattern based on the association of two corrupted words (e.g., benivole; 1980 [1550]: 49). But Le Trȩtté, which uses French as metalanguage for the first time, deals above all with derivation in long and detailed discussions in chapter 8 (“Des noms”) and chapter 12 (“Des dénominatifs”, that is to say proper nouns derived from nouns but nouns derived from verbs, for example, are addressed in the same chapter as well). Meigret attempts to identify different classes built, in particular, on semanticoreferential categories: French derivatives can result from names of “affections et qualités” ([affections and qualities]; e.g., fiévreux), “arts et sciences” ([arts and sciences]; e.g., mathématicien), “chefs de disciplines et sectes” ([heads of disciplines and sects]; e.g., platonique), etc. But the classification is primarily organized around suffixes, which are extensively listed. It appears that suffixal derivation is one of the privileged features enabling the differentiation between Latin and the vernaculars. The linguistic filiation is

2. Word-formation research from its beginnings to the 19th century indeed systematically studied in the chapter “Des noms”: it must also be observed that all the Latin derivatives ending in cus, or Greek derivatives ending in cos become qe or çien (“Il faut aussi entendre que nous tournons en qe ou çien tous les dérivatifs que nous tirons de la langue latine terminés en cus, ou de la grecque en cos”; 1980 [1550]: 27). However, it is also important to highlight the specificities of the French language: the Latin ending -arius becomes -ȩre in French, but in the case of Censorius, it becomes Çȩnsorin and not Çȩnsoȩre: because it does not sound right (“au regard de Censorius, nous le tournons en Çȩnsorin et non pas Çȩnsoȩre: parce qu’il sonne mal à l’oreille”; 1980 [1550]: 28). Meigret tries to shed light on general principles whenever possible (for example, the relation between the grammatical classes of derivative and primitive or between the suffix and the gender of the derivative), but he is also very attentive to restrictions on usage. While the form and the meaning of derivatives are often associated in an attempt at generalisation, the grammarian also remarks that the same suffix does not always have the same meaning and that one meaning can be expressed by different suffixes.

4.2. 17th century After Ramus’s Grammaire, no more references to the categories of species and figure can be found in the grammars of French. The contrast between primitive and derivative or between simple and compound may be mentioned from time to time in relation to the class of the nouns (Chifflet 1659: 8 and Irson 1662 [1656]: 18–19), but a specific section is no longer devoted to them in the presentation of word classes. However, Irson’s Nouvelle méthode (1662 [1656]) contains a final section entitled “Les étymologies”, which is a list of entries arranged alphabetically. These entries correspond to a collection of comments on both the origin of the words and their derivatives in compliance with etymology in its traditional sense (Delesalle and Mazière 2002). A parallel can be drawn between this insertion of a lexical section at the end of a grammar and 17th-century distribution of the roles of grammar and the dictionary in the treatment of word-formation. This distribution is explicitly asserted by Arnauld and Lancelot’s Grammaire générale et raisonnée in which we find the following passage (1660: 105): “On n’a point parlé, dans cette Grammaire, des mots dérivés ni des composés, dont il y aurait encore beaucoup de choses très-curieuses à dire, parce que cela regarde plutôt l’ouvrage d’un Dictionnaire général, que de la Grammaire Générale” [derived and compound words are not mentioned in this grammar even though a great many fascinating things remain to be said on the subject but they belong in a General Dictionary rather than in a General Grammar]. The authors of the Grammaire générale, however, assume a certain form of derivation – between substantives and adjectives – by contrasting “signification” and “manière de signifier” [manner of signifying]. All nouns which appear independently in discourse (“subsistent par eux-mêmes dans le discours”) – which represents their “manner of signifying” – are called nouns, even if they signify “accidents” (e.g., rougeur). Conversely, if the element expressing “connotation” – which implies that an adjective cannot be self-sufficient in discourse – is added to a noun which signifies a substance, the process results in the creation of an adjective (e.g., humain; Delesalle 1990). As shown, morphological regularity depends above all on the mental conceptions of things,

29

30

I. Word-formation as a linguistic discipline which is consistent with the programme of French general grammar: Arnauld and Lancelot show that derivation is a linguistic fact contributing to the expression of thought. Furthermore, only the transformation leading to the creation of an adjective noun from a substantive noun is regarded as enabling generalization. When the link between form and meaning in the lexicon can no longer be systematized, the reader is invited to turn from grammar to the dictionary, or in other words not expect a rule but to consult a list of items. The Port-Royal logic and linguistic theory were taken up in the Dictionnaire de l’Académie française (1694) with a view to applying them to a particular language. This dictionary, which aims at showing the creativity of the French language, is important for lexical morphology for two reasons. Firstly, at a micro-structural level, it enables the emergence of morphosemantic definitions (Mazière 1996; Leclercq 2002). The syntactic forms and semantic categories used at the beginning of the definitions attest to the remarkable stability characterizing the definitions of derivative words. These recurring syntactico-semantic patterns allow an organization of the definition which renders the link between the base form and the derivative explicit. Secondly, at a macro-structural level, not only does the dictionary set up a morphological arrangement of entries but – more importantly – it imposes constraints (explained in the preface) on this classification, which shows that derivation and composition are perceived in synchrony, regardless of etymology (Leclercq 2002). Essentially, the arrangement is used only if a French primitive word is attested: construire and destruire, for example, are regarded as independent since Latin struere has not come down to French (“n’a point passé en français”). To be taken into account in the nomenclature, the formal structure of the derivation must be observable in the French language from a synchronic point of view. It should be noted that the academician Regnier-Desmarais, in his grammar published in the early 18th century, distinguished between the etymon and the base form in synchrony: Simple or primitive nouns […] are those which do not derive their origin from another noun of the same language, but owe their signification to the first institution of this language (“Les noms simples ou primitifs […] sont ceux qui ne tirent point leur origine d’un autre nom de la même langue, mais qui doivent leur signification à la première institution de cette langue”; 1705: 179).

4.3. 18th–early 19th century Régnier-Desmarais is one of the few 18th-century grammarians of French who devoted a substantial discussion to derivatives and compounds (both are studied in the section dealing with the noun). As in the 17th century, word-formation was very rare in particular grammars. However, Beauzée reintroduced it into the grammarian’s field of studies, integrating it into general grammar. He didn’t address the question in his Grammaire générale (1767), but in several articles of the Encyclopédie (1751–1765). Through a series of subdivisions (article “Grammaire”, cf. Beauzée 1751–65b), word-formation was made a part of “etymology”, which was itself a part of “lexicology” (a term which, according to Beauzée, was introduced by Girard), lexicology in its turn being included in grammar. Word-formation groups together composition, derivation and inflection (derivation and inflection are both derivational processes that use “inflexions”, the first

2. Word-formation research from its beginnings to the 19th century being called “philosophical derivation” and the second “grammatical derivation”). But in the article “Formation” (cf. Beauzée 1751–65a; probably written with Douchet; Bourquin 1980a: 25), Beauzée blames the grammarians who preceded him for having preferred inflection at the expense of derivation. According to him, the explanation lies in the widespread position that grammar may legitimately restrict its focus to ready-made words (“les mots tout faits”). The encyclopaedist justifies the reintegration of the lexical aspects of word-formation into grammar using an argument compatible with the principles of general grammar: derivation and composition not only suppose a uniformity in processes (“uniformité de procédés”) within a language, but also some feature common to all languages (article “Formation”). More precisely, the link between form and meaning is interpreted as a regular phenomenon: “Nous disons en premier lieu, que ces terminaisons sont soumises à des lois générales, parce que telle terminaison indique invariablement une même idée accessoire” [First of all we believe that these endings are subject to general laws, because an ending invariably indicates the same accessory idea]. The theory of word-formation is incorporated into the theory of the sign that characterizes general grammar: both derivation and composition allow a “primitive” or “fundamental” idea to be modified due to an accessory idea. However, despite the grammarian’s asserting the principle of generality, it should be noted that most of the examples given by Beauzée are in Latin and the author explicitly points out that French has fewer regularities in this field than ancient languages. In the early 19th century, influenced by the work of etymologists (especially de Brosses and Le Bel) and probably inspired by the intense activity of lexical creation that characterized the revolutionary period (Dougnac 1982; Steuckardt 2008), Butet de la Sarthe developed and above all implemented Beauzée’s rationalist theory. He produced a new “science of words” (1801: 2), as he called it in his Abrégé d’un cours complet de lexicologie. We can consider this book as the first systematic description of formation processes in French (le “premier traitement systématique des procédés de formation en français”; Schlieben-Lange 2000: 31). This book isolates “lexicology” from grammar and gives it autonomy. However this autonomy is relative in fact since suffixal derivation and inflection are considered jointly, as was the case in Beauzée’s work. The systematic approach inaugurated by Beauzée and Douchet is reinforced. Butet, who taught physics, exposes not simply a metaphor but an analogy between the combinations that give rise to words and the molecules of matter assembled into bodies: both obey laws and are reducible to formulae. His theses on word-formation aim at implementing a semantic structuring of the lexicon and are fundamentally based on the compositionality of meaning. He takes up his predecessor’s theory of the sign: “prepositions” and “endings” are “accessory ideas” which change the main idea expressed by the “root”. However, the classification of the resulting words that is proposed is tripartite and no longer bipartite. Indeed Butet distinguishes between “radical constructions” (“constructions radicales”), which only consist of “roots”, that is to say elements used to represent the “main idea” (“primitive” or “simple”), “prepositive constructions” (“constructions prépositives”) and “postpositive constructions” (“constructions postpositives”, which comprise suffixal derivation and inflection). So composition is divided and organized with a view to singling out “compound words”, as they are called nowadays. The effort of completeness and systematicity is otherwise remarkable, even though the classification of suffixes, which is based on meaning rather than form, does not take into account the diachronysynchrony opposition: the formation of French words is put next to that of Latin words

31

32

I. Word-formation as a linguistic discipline and French words resisting analysis are placed on the same footing as French derivatives (words like instant or protéger constitute examples of prepositive constructions).

4.4. 19th century In the 19th century, grammarians dealing with word-formation did not innovate and generally adopted a pedagogical perspective: suffixal derivation, related to etymology, is mainly seen as a way of focusing on the meaning of words and above all their spelling from a didactical point of view (Jullien 1849: 150). Word-formation is still not clearly separated from etymology and priority is always given to meaning, in compliance with the tradition of general grammar. During this period, progress is made above all in books which are outside the field of grammar. One example is the case of synonym dictionaries that offer a systematic description of suffixes through the study of morphologically related synonyms. It is exemplified by the work of Lafaye (1858 [1841]). Already addressed by Roubaud (1785) and taken up by Guizot (1809), this systematic description is taken one step forward by Lafaye, the author of the Dictionnaire des synonymes de la langue française (1858 [1841]). He introduces a distinction between synonyms formed from different stems and grammatical synonyms. With the latter, whose “prefixes” or “endings” constitute the main difference (e.g., renunciation and renoncement), it is possible to extend the difference found in a particular example to all others sharing the same modification (“faire servir la différence trouvée dans un exemple particulier à la distinction de tous les autres qui présentent la même modification”; 1858 [1841]: 26). Thus synonym dictionaries create lists of suffixes that were taken up during the 19th century. But the pivotal work of the century is of course Darmesteter’s. While associated with historical grammar (his Cours de grammaire historique de la langue française was posthumously published from 1891 on), along with Raynouard and Diez, the author breaks new ground by dealing with “today’s language”, “living language” and by defining a field of study centered on the creativity of the French lexicon. His aim is to separate it from the historical study of word-formation. In his book De la création actuelle des mots nouveaux dans la langue française, he proposes to shift the point of view on the language: he concentrates on word-formation seen as a creative process in progress and not on the historical development of French vocabulary (1877: 1–2). Darmesteter takes into consideration three strategies of lexicon enrichment: a) “French formation”, lexical constructions created from French “stems” (the term “popular formation” is used but rejected by Darmesteter), b) “learned formation”, corresponding to borrowings from Greek or Latin and derivatives and compounds derived from Greek and Latin stems and c) borrowing from modern languages. Concern for the observation of the “real language” is also noticeable in the expansion of the notion of demotivation: a derived word can be considered as a single word if it does not express a “double-idea” (1877: 70), that is to say if compositionality of meaning is no longer obvious. Darmesteter’s theories are also characterized by a detailed classification of the different modes of formation. Derivation is split into “dérivation propre” (using a suffix) and “derivation impropre” (covering the processes now called “conversion” and “back formation”). Composition (to which he devotes a separate work entitled Traité de la formation des mots composés dans la langue française, 1894 [1875]), which is defined by its ability to create a conceptual

2. Word-formation research from its beginnings to the 19th century unity which obliterates the specific meanings of the elements, implies a division between compounds formed by juxtaposition (“les composés formés par voie de juxtaposition”; e.g., plafond, arc-en-ciel), compounds formed with particles, viz. adverbs or prepositions (“les composés formés à l’aide de particules”; e.g., surprendre, malaise) and compounds formed by composition as such (“les composés formés par composition proprement dite”; e.g., chou-fleur, timbre-poste). Although isolated, prefixation is therefore still regarded as being part of composition and the term of “parasynthesis”, taken over from the Greek grammatical tradition, is introduced with its meaning of “résultat d’une composition et d’une dérivation agissant ensemble sur un même radical” [result of a composition and a derivation acting together on the same stem] (1894 [1875]: 96). The opposition between the two remaining sets is founded on the nature of the syntactic relationships between the various elements of the compounds, with the use of the notion of “ellipsis”: the composition as such is indeed elliptical, the compound is in this case a “proposition en raccourci” – “timbre-poste ne veut pas dire simplement timbre et poste, mais timbre de la poste, timbre pour la poste” (1890: 72) [postage-stamp doesn’t mean just stamp and postage, but stamp for postage] – while composition by juxtaposition follows the usual syntactic construction. This innovative syntactic theory of composition was taken up widely in the 20th century, especially in a transformational perspective (particulary in Guilbert’s and Dubois’s works). More generally, Darmester’s theses and categories came down with minor changes through the first half of the 20th century and are still used today, despite Saussure’s reflections on synchrony and analogy (Kerleroux 2000).

Acknowledgements We are most grateful to Helmut Puff (University of Michigan, Ann Arbor) and Wilfrid Andrieu (Université de Provence) for their valuable help with the English version.

5. References Adelung, Johann Christoph 1782 Umständliches Lehrgebäude der Deutschen Sprache. Zur Erläuterung der Deutschen Sprachlehre für Schulen. Vol. 2. Leipzig: Breitkopf. [Repr. Hildesheim: Olms, 1971]. Adelung, Johann Christoph 1784 Gebrauch und Mißbrauch der Etymologie. Magazin für die Deutsche Sprache II(2): 96– 121. Aichinger, Carl Friedrich 1754 Versuch einer teutschen Sprachlehre, anfänglich nur zu eignem Gebrauche unternommen, endlich aber, um den Gelehrten zu fernerer Untersuchung Anlaß zu geben, ans Liecht gestellt von C.F.A. Frankfurt/Leipzig: Kraus. [Repr. Hildesheim: Olms, 1972]. Albertus, Laurentius 1573 [1895] Teutsch Grammatick oder Sprach-Kunst. Augsburg: Manger [Die deutsche Grammatik des Laurentius Albertus. Ed. by Carl Müller-Fraureuth, 1895, Straßburg: Trübner]. Arnauld, Antoine and Claude Lancelot 1660 Grammaire générale et raisonnée. Paris: Le Petit.

33

34

I. Word-formation as a linguistic discipline Auroux, Sylvain 1992 Histoire des idées linguistiques. Vol. 2: Le développement de la grammaire occidentale. Liège: Mardaga. Beauzée, Nicolas 1751−65a Formation. In: Denis Diderot and Jean Le Rond d’Alembert (eds.), Encyclopédie ou Dictionnaire raisonné des sciences, des arts et des métiers. Vol. 7, 172−176. Paris/ Neuchâtel: Le Breton. Beauzée, Nicolas 1751−65b Grammaire. In: Denis Diderot and Jean Le Rond d’Alembert (eds.), Encyclopédie ou Dictionnaire raisonné des sciences, des arts et des métiers. Vol. 7, 841−847. Paris/ Neuchâtel: Le Breton. Beauzée, Nicolas 1767 Grammaire générale. Paris: Barbou. Becker, Karl Ferdinand 1824 Die Deutsche Wortbildung oder die organische Entwickelung der deutschen Sprache in der Ableitung. Frankfurt/M.: Hermann’sche Buchhandlung. [Repr. Hildesheim: Olms, 1990]. Becker, Karl Ferdinand 1831 Schulgrammatik der deutschen Sprache. Frankfurt/M.: Hermann’sche Buchhandlung. Bourquin, Jacques 1979 La place et la fonction de la morphologie dérivationnelle dans la grammaire scolaire au XIXe siècle. Langue française 41: 60–76. Bourquin, Jacques 1980a La dérivation suffixale (théorisation et enseignement) au XIXe siècle. Thèse présentée devant l’Université de Besançon (1977). Lille: Université de Lille III. Bourquin, Jacques 1980b La terminologie du lexique construit (dérivation suffixale et préfixation). Langue française 47: 33−47. Brocquet, Sylvain 2008 Les mots non simples dans la tradition indienne paninéenne. In: Barbara Kaltz (ed.), Regards croisés sur les mots non simples, 11–33. Lyon: ENS Editions. Brücker, Jacob 1620 Teutsche Grammatic / das ist / Kurtzer Vnterricht / wie eyner etlicher massen recht reden und schreiben lehrnen solle. Frankfurt/M.: Jennis. Butet de la Sarthe, Pierre Roland François 1801 Abrégé d’un cours complet de lexicologie à l’usage des élèves de la quatrième classe de l’Ecole polymathique. Paris: Renouard. Chifflet, Laurent 1659 Essai d’une parfaite grammaire française. Anvers: van Meurs. Clajus, Johannes 1894 [1578] Die deutsche Grammatik des Johannes Clajus. Nach dem ältesten Druck von 1578 mit den Varianten der übrigen Ausgaben. Ed. by Friedrich Weidling. Freiburg: Trübner. Colombat, Bernard 1999 La grammaire latine en France à la Renaissance et à l’Age classique. Grenoble: ELLUG (Université Stendhal). Corpus de textes linguistiques fondamentaux: http://ctlf.ens-lyon.fr Darmesteter, Arsène 1877 De la création actuelle de mots nouveaux dans la langue française. Paris: Vieweg. Darmesteter, Arsène 1894 [1875] Traité de la formation des mots composés de la langue française. Paris: Vieweg.

2. Word-formation research from its beginnings to the 19th century Darmesteter, Arsène 1895 Cours de grammaire historique de la langue française. Vol. 3: Formation des mots et vie des mots. Paris: Delagrave. Delesalle, Simone 1990 De la définition du nom et du verbe dans la Logique et la Grammaire de Port-Royal. In: Jacques Chaurand and Francine Mazière (eds.), La définition, 72–77. Paris: Larousse. Delesalle, Simone and Francine Mazière 2002 La liste dans le développement des grammaires. Histoire Epistémologie Langage 24(1): 65–92. Dictionnaire de l’Académie française 1694 Paris: Coignard. Dougnac, Françoise 1982 La néologie. Histoire Epistémologie Langage 4(1): 67–72. Estienne, Robert 1557 Traité de la grammaire française. Genève: Estienne. Fleischer, Wolfgang 1983 Zur Geschichte der germanistischen Wortbildungsforschung im 19. Jahrhundert: Jacob Grimm und die Junggrammatiker. In: Wolfgang Fleischer (ed.), Entwicklungen in Wortbildung und Wortschatz der deutschen Gegenwartssprache, 74–100. Berlin: Akademie der Wissenschaften der DDR – Zentralinstitut für Sprachwissenschaft. Fögen, Thorsten 2008 La formation des mots et l’enrichissement de la langue vus par quelques auteurs latins. In: Barbara Kaltz (ed.), Regards croisés sur les mots non simples, 65–84. Lyon: ENS Editions. Forsgren, Kjell-Åke 2008 La conception de la formation des mots selon Karl Ferdinand Becker. In: Barbara Kaltz (ed.), Regards croisés sur les mots non simples, 131–149. Lyon: ENS Editions. Glatigny, Michel 1985 L’exception dans le système morphologique de L. Meigret. Langue française 66: 9–19. Grimm, Jacob 1878 Deutsche Grammatik. Zweiter Theil. Besorgt durch Wilhelm Scherer. 2nd ed. Berlin: Dümmler. [Repr. Hildesheim: Olms, 1989]. Gützlaff, Kathrin 1989 Von der Fügung Teutscher Stammwörter. Die Wortbildung in J. G. Schottelius’ ‘Ausführlicher Arbeit von der Teutschen HaubtSprache’. Hildesheim: Olms. Guizot, François 1809 Nouveau dictionnaire universel des synonymes. Paris: Payen. Heynatz, Johann Friedrich 1777 [1770] Deutsche Sprachlehre zum Gebrauch der Schulen. 3rd ed. Berlin: Mylius. Irson, Claude 1662 [1656] Nouvelle méthode pour apprendre facilement les principes et la pureté de la langue française contenant plusieurs traités. 2nd ed. Paris: Beaudouin. Jankowsky, Kurt R. 2004 Karl Ferdinand Becker’s (1775−1849) concept of word formation within the framework of his general linguistic theory. In: Kjell-Åke Forsgren and Barbara Kaltz (eds.), Studien zur Geschichte der Wortbildungstheorien, 89–106. Münster: Nodus. Jullien, Bernard 1849 Cours supérieur de grammaire. Paris: Hachette. Kaltz, Barbara 2002 Zur Entwicklung der Wortbildungstheorie in der deutschen Grammatikographie 1750– 1800. Beiträge zur Geschichte der Sprachwissenschaft 12(1): 27–47.

35

36

I. Word-formation as a linguistic discipline Kaltz, Barbara 2005 Zur Herausbildung der Wortbildungslehre in der deutschen Grammatikographie: Von den Anfängen bis zum Ende des 19. Jahrhunderts. In: Peter Schmitter (ed.), Sprachtheorien der Neuzeit III/1, 105–162. Tübingen: Narr. Kerleroux, Françoise 2000 France and Switzerland. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 138–145. Berlin/New York: de Gruyter. Lafaye, Pierre Benjamin 1858 [1841] Dictionnaire des synonymes de la langue française. 2nd ed. Paris: Hachette. Lallot, Jean 1998 La grammaire de Denys le Thrace. Traduite et annotée. 2nd ed. Paris: Editions du CNRS. Lallot, Jean 2008 De Platon aux grammairiens: Regards grecs sur la structure des mots non simples. In: Barbara Kaltz (ed.), Regards croisés sur les mots non simples, 51–63. Lyon: ENS Editions. Leclercq, Odile 2002 Aspects grammaticaux d’un dictionnaire de langue: Deux traitements de la morphologie par le Dictionnaire de l’Académie (1694). Histoire Epistémologie Langage 24(1): 107– 118. Lindner, Thomas 2002 Lateinische Komposita. Morphologische, historische und lexikalische Studien. Innsbruck: Institut für Sprachwissenschaft. Longolius, Johann Daniel 1715 Einleitung zu gründlicher Erkäntniß einer ieden / insonderheit aber Der Teutschen Sprache / Welcher man sich Zu accurater Untersuchung jeder Sprache / und Besitzung einer untadelhafften Beredsamkeit in gebundenen und ungebundenen Reden / Wie auch besonders In Teutschen für allerley Condition, Alter und Geschlechte / Zu einem deutlichen und nützlichen Begriff der Mutter=Sprache / bedienen kan. Budissin [Bautzen]: Richter. Mäzke, Abraham Gotthelf 1776 Grammatische Abhandlungen über die Deutsche Sprache. Breslau: Meyer. Matthaios, Stephanos 1999 Untersuchungen zur Grammatik Aristarchs. Texte und Interpretation zur Wortartenlehre. Göttingen: Vandenhoeck & Ruprecht. Matthaios, Stephanos 2004 Die Wortbildungstheorie in der alexandrinischen Grammatik. In: Kjell-Åke Forsgren and Barbara Kaltz (eds.), Studien zur Geschichte der Wortbildungstheorien, 5–22. Münster: Nodus. Mazière, Francine 1996 Un événement linguistique: La définition des noms abstraits dans la première édition du Dictionnaire de l’Académie (1694). In: Nelly Flaux, Michel Glatigny and Didier Samain (eds.), Les noms abstraits. Histoire et théories. Actes du colloque international “Les noms abstraits” (Dunkerque, sept. 1992), 161–174. Lille: Presses Universitaires du Septentrion. McLelland, Nicola 2010 Justus Georgius Schottelius (1612−1676) and European linguistic thought. Historiographia Linguistica 37(1): 1–30. McLelland, Nicola 2011 J. G. Schottelius·s Ausführliche Arbeit von der Teutschen Haubtsprache (1663) and its place in early modern European vernacular language study. Oxford: Blackwell.

2. Word-formation research from its beginnings to the 19th century Meigret, Louis 1980 [1550] Le traité de la grammaire française. Ed. by Franz Josef Hausmann. Tübingen: Narr. Padley, George Arthur 1985 Grammatical Theory in Western Europe 1500–1700. Trends in Vernacular Grammar I. Cambridge: Cambridge University Press. Paul, Hermann 1880 Prinzipien der Sprachgeschichte. Halle/S.: Niemeyer. [Repr. Tübingen: Niemeyer, 10th ed. 1995]. Paul, Hermann 1896 Ueber die Aufgaben der Wortbildungslehre. In: Sitzungsberichte der philosophisch-philologischen und der historischen Classe der k.b. Akademie der Wissenschaften zu München, 692–713. [Repr. in: Leonhard Lipka and Hartmut Günther (eds.), Wortbildung, 17–35. Darmstadt: Wissenschaftliche Buchgesellschaft, 1981]. Paul, Hermann 1920 Deutsche Grammatik. Vol. 5: Wortbildungslehre. Halle/S.: Niemeyer. Ramus, Petrus 1572 Grammaire. Paris: Wechel. Ratke, Wolfgang 1959 [1612–1615] Sprachkunst. In: Erika Ising (ed.), Wolfgang Ratkes Schriften zur deutschen Grammatik. Vol. 2, 7–22. Berlin: Akademie Verlag. Ratke, Wolfgang 1959 [1630] Die WortbedeütungsLehr der Christlichen Schule [...]. In: Erika Ising (ed.), Wolfgang Ratkes Schriften zur deutschen Grammatik. Vol. 2, 269–318. Berlin: Akademie Verlag. Régnier-Desmarais, François Séraphin 1705 Traité de la grammaire française. Paris: Coignard. Rey, Alain 1970 La lexicologie. Paris: Klincksieck. Roubaud, Pierre-Joseph 1785 Nouveaux synonymes françois. Paris: Moutard. Schlieben-Lange, Brigitte 2000 La révolution française. In: Sylvain Auroux (ed.), Histoire des idées linguistiques. Vol. 3, 23–34. Liège: Mardaga. Schottelius, Justus Georg 1641 Teutsche Sprachkunst [...]. Braunschweig: Gruber. Schottelius, Justus Georg 1663 Ausführliche Arbeit Von der Teutschen HaubtSprache. Braunschweig: Zilliger. [Repr. Tübingen: Niemeyer 1967; online]. Steuckardt, Agnès 2008 Présentation. In: Pierre-Nicolas Chantreau, Dictionnaire national et anecdotique. 1790. Reprint ed. by Agnès Steuckardt, 9–91. Limoges: Lambert–Lucas. Sylvius, Jacobus Ambianus 1531 Grammatica Latino-Gallica. Paris: Estienne. Vaahtera, Jaana 1998 Derivation. Greek and Roman Views on Word Formation. Turku: Turun Yliopisto.

Barbara Kaltz, Freiburg (Germany) Odile Leclercq, Aix-en-Provence (France)

37

38

I. Word-formation as a linguistic discipline

3. Word-formation in historical-comparative grammar 1. 2. 3. 4. 5. 6.

Terminological preliminaries Word-formation immediately before the rise of the historical-comparative method Historical-comparative descriptions Descriptions based on semantic criteria Compounding in comparative grammar References

Abstract Word-formation in historical-comparative linguistics emerged on the one hand from Classical and German studies at the turn of the 18 th and 19 th centuries and on the other hand from Indian grammar which had become known in Europe early in 1800. The present article tries to delineate the main developments from the 19 th century to the present.

1. Terminological preliminaries 1.1. Basic terminology Morphological terms such as root, stem or affix as well as the segmentation of words and the consciousness of word-formational processes which they imply represented fundamental linguistic insights that provided concepts of morphological analysis which Indo-European studies, but also other grammatical traditions, could no longer do without. Nevertheless, in comparison with terms relating to syntax, parts of speech or case they are not really old: they do not, as one might be inclined to think, reach back to antiquity, nor were they well-established elements of traditional terminology at the outset of IndoEuropean studies. Morphological analysis was neglected in ancient grammar: the procedures of analyzing words which are now so familiar to linguists and which Bopp (1824− 31) referred to as dissection [Zergliederung] were unknown to the ancient grammarians, concepts such as ‘root’, ‘stem’ or ‘affix’ were also completely unfamiliar to them. In the wake of the Indian grammatical tradition, Franz Bopp was the first to recognize that Indo-European words could generally be broken down into the structure root + derivational affix + inflectional affix. The identification of what was indistinctly called Grundform ‘basic form’, Stammform ‘stem form’, Stamm ‘stem’ or Thema ‘theme’, was so new and groundbreaking at the time that Bopp felt obliged to provide the following clarifications: Die Indischen Grammatiker fassen die Nomina (sowohl Substantive, als Adjektive, Pronomina und Zahlwörter), in ihrem absoluten, von allen Casusverhältnissen unabhängigen, und von allen Casuszeichen entblößten Zustande auf, und nehmen daher eine Grund- oder Stammform an, zu welcher der Nominativ und die obliquen Casus der drei Zahlen sich als

3. Word-formation in historical-comparative grammar abgeleitet verhalten. Diese Grundform kommt häufig in zusammengesetzten Wörtern vor, indem die ersten Glieder eines Compositums aller Casusendungen beraubt, und somit identisch mit der Grundform sind. (Bopp 1827: 23) [The Indian grammarians conceived of nouns (substantives, adjectives, pronouns and numerals) in their absolute state, independent of all relations and markers of case, assuming the existence of a basic or stem form, from which the nominative and the oblique cases of all three numbers were derived. This basic form often appears in compound words, the first members of compounds being deprived of all case endings and therefore identical with the basic form.]

These stem forms of the old Indo-European languages thus have only been recognized in the Western world since Bopp and his disciples (cf. Lindner 2012: 121). However, Jacob Grimm in his Deutsche Grammatik (1819 ff.) did not yet recognize the stem principle: In the analysis of the first members of compounds, where it should have been most obvious, he resorted to the concept of a “composition vowel” (Compositionsvocal) which would haunt Indo-European studies until the early Neo-Grammarians (cf. Lindner 2012: 115). German linguists soon adopted Bopp’s analysis, first and foremost among them was Eberhard Gottlieb Graff (cf. Lindner 2012: 122), but the classical handbooks continued to perpetuate the old view codified over the centuries in school grammars of Latin and Greek. Philipp Buttmann was the first to use and differentiate the concepts of root (Wurzel), stem (Stamm) and theme (Thema), but he did not recognize that the first members of compounds were stems, calling them declensional or nominal endings, or linking vowels (cf. Lindner 2012: 124). Nevertheless, he was the first classical philologist who had recognized, at least theoretically, the role of the stem in the segmentation of Greek nominal inflection, especially of the third declension. However, he did not yet recognize clearly the amalgamation of stem vowels and inflectional desinences, which is why segmentation did not find its way into school grammars for still quite some time. We will have to wait until well into the second half of the 19th century to see Bopp’s insight establish itself in grammars of the classical languages, essentially thanks to the indefatigable endeavors of his disciple Georg Curtius. It was especially the precise conceptual distinctions in Curtius’ Griechische Schulgrammatik (1852) which eventually enhanced general consciousness regarding morphological structure. As we have seen, only comparative linguistics and, hesitantly, classical grammar started differentiating on the one hand root (Germ. Wurzel) and stem (Stamm), and on the other ending (Endung) and termination (Ausgang). The latter distinction has become established in school grammars and has lived on in scientific discourse until the present day. As a consequence Ausgang still refers to the amalgamated combination of a stemforming element and an inflectional morpheme and probably constitutes the ending par excellence for the unsophisticated speaker (cf. Lindner 2012: 138). Alongside the terms root, stem, (inflectional) ending, and termination, which became established as late as around 1870 in scientific morphology, Hermann Perthes introduced Wortstock ‘root or stem, respectively’ exclusively for pedagogical purposes from 1875 onwards (cf. Lindner 2012: 138). As this terminology was becoming popular in school grammars of the classical languages in the last quarter of the 19th century, an umbrella term for morphological elements was introduced around 1880 in entirely different scientific contexts, viz. morpheme. It had been coined by Jan Baudouin de Courtenay in the Kazan school of linguistics towards the end of the decade of 1870, taking as a model the term phoneme which was already in use (cf. Lindner 2012: 140). Thema, originally a synonym of Stamm

39

40

I. Word-formation as a linguistic discipline ‘stem’ in morphology, also shows an interesting conceptual evolution: In the course of the 20th century in Old German studies it has narrowed its meaning and become equivalent to Stammbildungsvokal ‘stem-forming vowel’ or Formans (cf. Lindner 2012: 140− 143).

1.2. The ambiguity of Wortbildung ‘word-formation’ In the second half of the 19th century there was also some confusion concerning the notion Wortbildung ‘word-formation’ itself. The early Neogrammarians used Wortbildung in the sense of ‘inflection’, while what is called Wortbildung today went under the term Stammbildung ‘stem-formation’: Es zerfällt demnach die wortbildungslehre (formen-, flexionslehre) in die lehre von der bildung der nomina und in die lehre von der bildung der verba. Jenen liegen nominal-, diesen verbalstämme zu grunde. Die lehre von der bildung der nomina nennt man declination, die lehre von der bildung der verba conjugation. (Miklosich 1876: 1; vgl. Miklosich 1875) [Word-formation (the study of forms, of inflection) therefore comprises the formation of nouns and verbs. The former are derived from noun stems, the latter from verb stems. The study of the formation of nouns is called declension, the study of the formation of verbs conjugation.] Die Suffixe theilt man ein in Wortbildungssuffixe oder Flexionssuffixe im engern Sinne, wozu einerseits die Casusendungen […], anderseits die Personalendungen […] gehören, und Stammbildungssuffixe […]. Eine scharfe Grenze zwischen beiden Suffixgattungen ist nicht zu ziehen, da manches Element, das ursprünglich nur ableitend (stammbildend) war, mit den wortbildenden Suffixen [= Flexionsendungen, Th. L.] auf gleiche Linie gekommen ist. (Brugmann 1886: 15) [Suffixes are divided into word-formational or inflectional suffixes in the narrow sense, which on the one hand comprise case endings […] and on the other personal endings, and stem-forming suffixes […]. It is impossible to draw a sharp line between the two kinds of suffixes, because some elements that originally were derivational (stem-forming) have turned into word-forming suffixes [i.e. inflectional endings; Th. L.].]

This interpretation as well as the arrangement of word-formation before inflection in grammars goes back to Schleicher’s Compendium, who subsumed both “roots and stems” (stem-formation) as well as word-formation (inflection) under the heading Morphologie ‘morphology’ (cf. Lindner 2012: 94). Brugmann in turn, in the first edition of his Grundriß (1892), subsumed both stem-formation and inflection under word-formation. This arrangement met with criticism on the part of Hermann Paul (1896: 692), which is why it was later on abandoned again (also by Brugmann in his final edition of the Grundriß in 1906). At the latest in the first decades of the 20th century the traditional terminology had again become standard, cf. Debrunner’s preface to his Griechische Wortbildungslehre: Noch ein Wort über die Begrenzung des Stoffs: Das Büchlein soll die Wortbildung, nicht die Formenbildung behandeln. Ausgeschlossen ist also die gesamte Deklination und Konjugation […]. (Debrunner 1917: IX)

3. Word-formation in historical-comparative grammar [Just one more word concerning the demarcation of the subject: the book is intended to treat word-formation, not form-formation. All declensions and conjugations are therefore excluded […].]

The very fact that Debrunner felt obliged to make such a remark, however, shows that the term Wortbildung could still provoke uncertainties, otherwise he would not have made it. We have come full circle, returning to the usage of early historical-comparative grammar (and earlier traditions), when word-formation was used in the sense it has today and was located after inflection (cf. Lindner 2012: 94−95, with references).

2. Word-formation immediately before the rise of the historical-comparative method At the beginning of the 19th century important insights had been gained with respect to the approaches of the ancient grammarians which had been in use in grammars and textbooks until around 1800. Before Buttmann, nobody talked of word-internal structure, only of “endings” (Endungen, Endigungen). Especially first elements of compounds were still treated in the traditional way by regarding them as case-forms or mutilations thereof; Christian August Lobeck, for example, continued this tradition of analysis well into the 19th century (cf. Lindner 2012: 125). The systematization of derivational, especially suffixal, word-formation and composition was still in its infancy. Until the 18th century the poorly developed word-formational analysis of the ancient grammarians was treated in the etymological parts of grammars, i.e. those dealing with words and parts of speech, under the headings of species and figura (etymologia had the meaning of ‘word analysis’, cf. Lindner 2011: 29 and article 2 on word-formation research from its beginnings to the 19th century). Only towards the end of the 18th century could a new orientation be observed that manifested itself in a separate treatment of word-formation, especially composition, by Adelung in his Umständliches Lehrgebäude (1782a: 216−236, 1782b: 209−274). In classical grammar that was considered as a revolution, which is why Trendelenburg could write in the second edition of his Griechische Grammatik: Ueberhaupt, hoffe ich, wird man nicht leicht einen Theil der Sprachlehre an vorhin unbekannten Bemerkungen ganz leer finden […] Als Beispiel nenne ich nur das eilfte Kapitel […], in welchem ich einen Versuch gemacht habe, die Wortbildung genauer, wie bisher, auseinander zu setzen. Wenigstens glaube ich mir den zweiten Abschnitt […] von der Zusammensetzung der Wörter mit Recht zueignen zu können. Denn ich erinnere mich nicht, in irgend einem grammatischen Werke über diesen Theil der Wörter, an welchen das Griechische so außerordentlich reich ist, etwas gelesen zu haben, was die Grundsätze und die Analogie, welche die Sprache in Zusammensetzung der Wörter befolgt, nur einigermaßen auseinandersetzte. Alle Sprachlehrer haben sich bloß auf den mechanischen Theil, auf die Art und Weise, wie zwei Wörter an einander gefügt werden, eingeschränkt […]. Ich erwehne dieses Versuchs, besonders deswegen, um Freunde von dergleichen Untersuchungen zu bitten, so wol die Grundsätze, welche ich hier angenommen habe, zu prüfen und, wo es nöthig ist, zu berichtigen, als auch gelegentlich das Ihrige dazu beizutragen, diese Lehre mit neuen Bemerkungen zu bereichern, welche sich in diesem Theil der Sprache besonders bei der Vergleichung des Griechischen mit dem Deutschen darbieten werden. (Trendelenburg 1790: VI f.)

41

42

I. Word-formation as a linguistic discipline [I hope that one will hardly find any part of this grammar devoid of hitherto unknown observations […] I would just like to mention Chapter XI […], where I have tried to deal with word-formation in more detail than is usual. I think that at least the second section […] about composition is mostly original work. I do not remember having read in any grammatical treatise about this kind of words which abound in Greek anything that would explain the principles and the analogy which the language follows in compounding. All grammarians have limited themselves to the purely mechanical part of the question, the way in which two words are put together […]. I mention this essay particularly because I would like to invite friends of this kind of research to assess and, where necessary, correct the principles I have assumed and to enrich this analysis with new observations that one will not fail to make in this area of language especially by comparing Greek and German.]

One should also point out the progress made in the treatment of Greek word-formation by Buttmann, from the short chapter of the first, second and third editions (1805: 266− 268, Von der Zusammensetzung) of his Griechische Grammatik to the extensive chapter in the fourth and fifth editions (1808, 1810: 399−419, Wortbildung durch Endungen, 420−428, Wortbildung durch Zusammensetzung): Der bedeutendste größere Zusatz […] ist der Abschnitt von der Wortbildung, der mich zwar noch keineswegs befriedigt, von welchem ich aber doch hoffe, daß er auch so schon seinen Zweck in der Hauptsache erreichen wird. (Buttmann 1810: X) [The most important substantial addition […] is the section on word-formation, which is still far from satisfying me completely, but which hopefully will also attain its main aims as it stands.] Die Wortbildung im vollen Verstande des Wortes liegt außerhalb der Grenzen der gewöhnlichen Sprachlehre. Denn da die Analogien in dem älteren Theile des Wortvorrathes […] vielfältig zerrissen und verdunkelt sind […], [wird daher] eine gewisse Masse von Wörtern lexikalisch voraus[gesetzt] […]. Gewisse Arten der Ableitung jedoch, von welchen man eben deswegen annehmen kann, daß sie neuer sind, haben sich so vollständig und innerhalb gewisser Grenzen durchgehend erhalten, daß sie mit Sicherheit zusammen gestellt werden können; und diese Vereinigung derselben unter einem Gesichtspunkt erleichtert und beschleunigt die Kenntnis der Sprache […]. (Buttmann 1810: 399) [Word-formation in the full sense of the word lies outside ordinary grammar. Since in the older part of vocabulary the analogies […] have often been broken and are now opaque […], a certain number of words must be considered as lexically primitive […]. However, some kinds of derivation, which for that very reason may be considered as younger, have been preserved so fully and completely within certain limits, that they can be put together without question; and assembling them from one specific viewpoint makes knowledge of the language easier and more rapid […].]

Formal analogies were also decisive for early comparative grammar which built on these forerunners. At the beginning, derivational word-formation was formally oriented exclusively according to the “endings”, i.e. stem-forming suffixes: Bei der Anhängung der Endungen [scil. Wortbildungssuffixe] walteten zwei Prinzipe vor, das Bestreben gleichartige Bedeutungen durch einerlei Endung auszudrücken, und das Bestreben, die Endung der Form des Stammworts möglichst anzupassen. Allein durch die Kollision dieser Prinzipe entstand zweierlei Verwirrung der Analogie: 1) ist dieselbe Art der Bedeutung häufig unter verschiedene Formen vertheilt; 2) Endungen, die ursprünglich nur

3. Word-formation in historical-comparative grammar von gewissen Formen des Stammworts gebildet wurden […], gingen, wenn eine gewisse Bedeutung bei mehreren Wörtern gleichen Ausgangs fühlbar geworden war, auch auf andere Stammwörter über, deren Form nicht dazu paßte […]. (Buttmann 1810: 400) [Two principles preside over the attachment of endings [i.e. derivational suffixes], the endeavor to express the same meaning with the same ending, and the endeavor to adapt the ending to the form of the stem as far as possible. Through the collision of these two principles analogy became confused in two ways: 1) often the same meaning is distributed among different forms; 2) endings that had originally been attached only to stems of a certain form came to be attached to other stems whose form was inappropriate as soon as more words with the same ending were felt to form a semantically coherent group […].]

It still took quite some time before word-formation would also be analysed according to “sameness of meaning”, i.e. according to semantic derivational categories. At the beginning, all treatments of word-formation were organized according to formal parameters concerning the suffixes. Shortly before the publication of Grimm’s monumental German word-formation (1826), Becker published his Deutsche Wortbildung (1824). In this first monographic treatment of German derivation and composition Becker analyzed the “formation of verbs” (71−261), “derivation by suffixes” (262−368) and “composition” (369−451). Concerning derivation, Becker speaks of Kernformen (‘nuclear forms’, i.e. bases) and Sproßformen (‘offshoot forms’, i.e. derivatives); the process of derivation is called Umendung ‘change of ending’. Ableitung ‘derivation’ for him is the hyperonym, which leads him to distinguish derivation by change of ending from derivation by composition. In the realm of composition he builds on Adelung’s criteria (1782b: 215 ff.) distinguishing Verschmelzung ‘fusion’, the amalgamation of two words into a conceptual unit, and Zusammenfügung ‘putting together’, the syntactic union of two words. These distinctions can also be found in his Organism (1827) and in his Deutsche Grammatik (1829), placed in the context of a wider grammatical system.

3. Historical-comparative descriptions 3.1. 19th century Jacob Grimm (1826) represents the historical-comparative turn in German word-formation: “The scientific study of word-formation is […] a creation of J. Grimm” (cf. Paul 1896: 692). He takes into account older stages of the language and provides abundant comparisons with other Germanic and non-Germanic material. First, he distinguishes derivation in the more restricted sense and composition: ableitung heißt die zwischen wurzel und flexion eingeschaltete, an sich selbst dunkle mehrung des wortes, kraft welcher der begriff der wurzel weiter geleitet und bestimmt wird. […] die ableitung unterscheidet sich von der zusammensetzung […] letztere verbindet zwei lebendige oder doch deutliche wurzeln miteinander […]. (Grimm 1826: 89−90) [Derivation refers to the in itself opaque material that comes between the root and the inflection and which develops and determines the concept expressed by the root. […] Derivation must be distinguished from composition […] the latter combines two roots that are still distinctly felt to be such […].]

43

44

I. Word-formation as a linguistic discipline Grimm then goes on to differentiate formal patterns of “vocalic” and “consonantic” derivations (92−386), the semantics of which is briefly touched upon (395−398). The latter part is constituted by composition (405−985). Here Grimm introduces the concept of Compositionsvocal ‘composition vowel’, which he did not identify as a stem vowel and which would live on for a long time in German school grammars. Only around 1840 does Grimm abandon this doctrine after harsh criticism from comparatists (cf. Lindner 2012: 117). His distinction of eigentliche (echte) and uneigentliche (unechte) Komposition (407 f.; cf. Lindner 2011: 15−20), compounding proper and univerbation, on the other hand, has stood the test of time. Franz Bopp treats Old-Indian word-formation for the first time in his Ausführliches Lehrgebäude, where he introduces the Indian terms for types of compounds which are used up to the present day (1827: 268 ff., 310 ff.; cf. section 5). In a comparative context he does the same in the fifth volume of his Vergleichende Grammatik (1833−52: 1072 ff., 1410 ff.), classifying derivation according to the form of suffixes and composition according to the Indian categories. Already before Bopp, August Friedrich Pott in his Etymologische Forschungen (1836: 351 ff.) had treated word-formation from a comparative perspective, but in a rather casuistic and unsystematic manner. In the meantime Friedrich Diez had published a treatise of Romance word-formation in the second volume of his Romanische Grammatik (1838: 219 ff.). The treatment of derivation follows the formal criterion of the form of suffixes (244 ff.), that of composition is by parts of speech following the example of Grimm (334 ff.). A book that was to become important for the analysis of word-formation, especially of Ancient Greek, by the early Neogrammarians was Curtius’ De nominum Graecorum formatione linguarum cognatarum ratione habita (1842), dedicated to his teacher Bopp. In Neogrammarian times the theoretical manifesto, also for word-formation, was Hermann Paul’s Prinzipien (1880: 161 ff., 51920: 325 ff.). In turn, the model up to the present day for all descriptively oriented treatises of the word-formation of individual IndoEuropean languages is constituted by the monumental volume on word-formation in Brugmann’s Grundriß. The second edition of 1906 set the standard and continues to be indispensable due to the wealth of its comparative material and its comprehensive description. The introduction discussing the structure of Indo-European word-forms as well as the motives and kinds of word-formation processes (1−49, Allgemeines) is followed by a description of composition (49−120) and thereupon the listing of Stammbildungsformantien based on formal criteria (120−582; the relevant literature on word-formation is summarized on pp. 49 ff. and 120 ff.). In the concluding section the material is arranged according to semantic groups (582−685, Bedeutung der Nominalstämme). What is worth highlighting is Brugmann’s rejection of the affix-terminology and his introduction of the term Formans ‘formative’ to refer to a derivational affix (1906: 8 ff.).

3.2. 20th century The reference works on the word-formation of old Indo-European languages published during the 20th century can only be enumerated here (further references can be found in the bibliographies of the works cited as well as in Heidermanns 2005, Lühr and Balles 2008 and Lühr and Matzinger 2008): from a comparative perspective Brugmann (1906)

3. Word-formation in historical-comparative grammar (esp. Delbrück 1900: 139 ff. and Brugmann 1904: 297 ff., 1906: 49 ff.); for Old Indian: Wackernagel (1905) (Wackernagel and Debrunner 1957), Debrunner (1954); for Avestan: Duchesne-Guillemin (1936), Kellens (1974); for Greek: Debrunner (1917), Chantraine (1933), Schwyzer (1939) as well as Risch (1974) for the language of Homer, for Latin: Leumann (1977), Kircher-Durand (2002) and Lindner (1996, 2002a), for Germanic: Paul (1920), Meid (1967) and Carr (1939). I would also like to draw attention to the part dedicated to word-formation in the on-going project of the Indogermanische Grammatik published by Winter in Heidelberg: Lindner (2011 ff.) on compounding, Sadovski (forthc.) on derivation, both with exhaustive bibliographies.

4. Descriptions based on semantic criteria The first scholar to take a semantic category, viz. verbal abstracts, as point of departure was Karl von Bahder (1880), subordinating the formal side to the semantic perspective: Ich hoffe, dass es mir gelungen ist, einige neue gesichtspuncte für die wortbildung aufzudecken […] Abschliessende resultate wird kein billig denkender von einem versuche fordern, der als der erste einer zusammenhängenden betrachtung einer wortkategorie […] wol bezeichnet werden darf. (von Bahder 1880: I f.) [I hope to have been able to discover some new viewpoints for word-formation […] It would be unfair to expect definitive results from an essay that may probably be called the first one to treat a word-category from a coherent perspective.]

In his Nominale Stammbildungslehre (1886) Friedrich Kluge followed this model for all categories. He was the first to abandon the prevailing formal approach in a comprehensive treatise of word-formation and to proceed according to the meanings of the formatives; those belonging together from a semantic point of view were treated together in special chapters (suffixes for personal nouns, diminutive suffixes, collective suffixes, suffixes for inanimate concrete nouns, abstract nouns, etc.). Cf.: Freilich schließe ich mich in der gruppierung des stoffes nicht an linguistische vorbilder an; ich habe nicht die lautform, wie es bisher üblich war, zum ausgangspunkte für die anordnung gemacht. (Kluge 1886: VIII) [In my arrangement of the material I do not follow existing models; I did not take as the point of departure the phonic form, as has been done up to now.]

This way of proceeding was to become widely accepted. Wilhelm Meyer-Lübke, for example, followed Kluge’s arrangement in his Italienische Grammatik (1890): Wenn so von Seiten einer streng historischen Grammatik die italienischen Fachgenossen, die an der Quelle sitzen, mein Buch werden vielfach ergänzen können, so ist das in noch höherm Grade der Fall bei der Wortbildungslehre. Die Darstellung derselben, wie ich sie im III. Kapitel gegeben habe, hat die Unkömmlichkeit, dass sie einmal viele Suffixe nicht bespricht, und sodann, dass manche Erscheinungen, wie die Verknüpfung verschiedener Suffixe, die Präpositionalbildungen, die Suffixvertauschungen nicht zur Sprache kommen. Allein ich habe nicht den historischen Entwicklungsgang geben wollen, mir lag hauptsächlich daran, eine Darstellung zu bieten, wie sie meines Wissens noch für keine romanische

45

46

I. Word-formation as a linguistic discipline Sprache geboten ist. Die begriffliche Seite ist für diesen Theil der sprachlichen Biologie ebenso wichtig wie die formale, und um dies hervorzuheben, habe ich die letztere fast ganz beiseite gelassen, um so eher, als bei der einseitig aufs Formale gerichteten Aufmerksamkeit der romanischen Forschung eine Ergänzung nach dieser Seite hin nicht schwer fällt. (MeyerLübke 1890: ix–x) [Colleagues working on Italian from a strictly historical perspective on the basis of more abundant material will certainly be able to complete the present book, especially regarding word-formation. My description in chapter III has the disadvantage of not treating many suffixes and that some phenomena such as the combination of different suffixes, prepositional formations, or suffix exchange are not mentioned. However, my intention was not to present the historical development, I wanted to provide a description which, to the best of my knowledge, does not yet exist for any Romance language. The conceptual side is as important for this part of linguistic biology as the formal side. In order to highlight this fact, I have almost completely left aside the latter, all the more so as the one-sided attention of research in Romance on the formal side easily allows completing it in that respect.]

Brugmann’s Grundriß (1892, 1906) eventually combined both a formal and functional perspective. Furthermore, Paul (1896) is an important contribution to the semantic aspects of word-formation categories and their diachronic developments.

5. Compounding in comparative grammar The formally-oriented doctrine of compounding of the ancient grammarians (cf. Lindner 2002a: 181) remained unchallenged until the 18th century. But when Sanskrit appeared on the scene of European scholars in the last decades of the 18th century, the Western world became acquainted with the approach and terminology of Indian grammarians, who had been led early on by the high productivity of compounding in Sanskrit to develop a syntactico-semantic, functional typology of compounds which mutatis mutandis is still in use today (cf. Lindner 2011: 20−21). The first interface between the Indian tradition and its transmission in Europe can be found in the first Sanksrit grammars printed in Europe at the end of the 18th and the beginning of the 19th century (Lindner 2012: 148). Bopp eventually coined the latinized versions of the Indian terms that are still in use: 1. Copulative compounds, called dvandva; 2. possessive compounds, called bahuvrīhi (“the compounds of this class are adjectives or common nouns denoting the possessor of what the parts of the compound mean, so that the notion of possessor always has to be supplemented. That is why I call them ‘possessive compounds’”); 3. Determinative compounds, called karmadhāraya (“The last member of this class of compounds is a noun or adjective, which is determined or described more precisely by the first member”); 4. Dependency compounds (Abhängigkeits-Composita), called tatpuruṣa (“this class comprises compounds whose first member depends on or is governed by the second one, always realizing some oblique case relation”); 5. Collective compounds, called dvigu; 6. Adverbial compounds, called avyayībhāva (Bopp 1827: 311 ff., 1833−52: 1427 ff.). This classification remains practically unchanged over the next decades, excepting order and denomination. The insight that the determinative compounds comprised both karmadhāraya and tatpuruṣa was formulated somewhat later, by Theodor Benfey in his Vollständige Grammatik der Sanskritsprache (1852), who did not follow the six-class-

3. Word-formation in historical-comparative grammar system of the Indian grammarian Vopadeva directly (as most Europeans before him), but another tradition going back directly to Pāṇini, who calls karmadhāraya and dvigu subclasses of tatpuruṣa; this system, however, has not always been adopted by later research on compounding (cf. Benfey 1852: 245 ff.: 1st class: Copulative compounds (dvandva); 2nd class: Determinative compounds (tatpuruṣa), subdivided into: 1st species: case-determined compounds (tatpuruṣa stricto sensu), 2nd species: apposition-determined compounds (karmadhāraya), 3rd species: numeral-determined compounds (dvigu), which he eventually sees as a special case of the second species, in which the appositional determination is realized by a numeral; 3rd class: relative compounds (bahuvrīhi). Benfey goes as far as to consider the bahuvrīhi as a special kind of compounding of the second class). At the beginning of the decade of 1860 an important study was published that would set standards for the later research on compounds in general and Indo-European linguistics, Ferdinand Justi’s Über die Zusammensetzung der Nomina in den indogermanischen Sprachen (1861). What is at issue here is genetic and typological comparison based on material also from non-Indo-European languages, with a wealth of examples that would form the basis of Neogrammarian research some fifteen or twenty years afterwards. I would like to draw attention particularly to the first in-depth analysis of the bahuvrīhi type, interpreted as “a higher kind of compounding” condensing a relative clause into a word, a stance that gave rise later on to a controversial discussion about the origin of possessive compounds leading to different theories (cf. Lindner 2002b: 269−273). Chronologically the first one was Justi’s relative-clause theory, which was further developed and expanded by Jacobi (1897): Es gibt nun eine art wortzusammensetzung, welche einen ganzen bezüglichen satz zu éinem wort vereinigt, das aber wie der ganze satz ebenfalls bezügliche, relative bedeutung hat. Statt zu sagen ἐφάνη Ἠὼς ᾗτινι οἱ δάκτυλοι ὥστε ῥόδα εἰσὶν, zieht man den ganzen relativsatz zusammen und bringt ihn in numerale, casuelle und geschlechtliche congruenz mit dem nomen, auf das er sich bezieht, und sagt also ἐφάνη Ἠὼς ῥοδοδάκτυλος, welches aber genau aufgelöst bedeutet ,Eos, welcher finger wie rosen sind‘. – ibid. Der bahuvrîhi (sic!) ist nun die bildung, in welcher die wortzusammensetzung den gipfel ihrer vollendung erreicht hat; sie ist ebenso schön wie kurz und bündig […]. (Justi 1861: 117) [There is one kind of composition that condenses a whole relative clause into one word, which however also has a relative meaning like the whole clause. Instead of saying ἐφάνη Ἠὼς ᾗτινι οἱ δάκτυλοι ὥστε ῥόδα εἰσὶν, the whole relative clause is pulled together and made agree with the noun it is predicated of in number, case and gender, saying ἐφάνη Ἠὼς ῥοδοδάκτυλος, whose explicit meaning however is: ‘Eos, whose fingers are like roses’. […] The bahuvrihi is that kind of formation where compounding has reached its epitome; it is as beautiful as it is concise […].]

The question of the nature, history, and origin of stem composition has also become the subject of controversy (cf. Lindner 2012: 88−121, 2013: 149−154). Furthermore, the general and philosophical aspects of compounding first addressed by Justi have been elaborated in Ludwig Tobler’s book Über die Wortzusammensetzung nebst einem Anhang über die verstärkenden Zusammensetzungen (1868). Moreover, Justi’s book stimulated further research on Greek and Latin compounds published in shorter essays (cf. Lindner and Oniga 2005; Lindner 2011: 12, 2012: 131−134). Apart from such smaller contributions, one has to mention the comprehensive descriptions of syntactic compounds by Francis Meunier (1872) – we usually call them Juxta-

47

48

I. Word-formation as a linguistic discipline posita in Indo-European studies (cf. Lindner 2011: 18) –, as well as Leopold Schröder’s great monograph Über die formelle Unterscheidung der Redetheile im Griechischen und Lateinischen mit besonderer Berücksichtigung der Nominalcomposita (1874). In this first near-exhaustive description of the dispersed material Schröder rightly criticizes that one defect of his predecessors was “that they were far from complete in their empirical coverage, being content with single examples where complete enumerations would be highly desirable” (cf. Schröder 1874: 194). What is more, he coined the terms “synthetic compound” (1874: 206; synthetisches Kompositum) as well as composita immutata and mutata (1874: 208) to replace Justi’s “lower” and “higher” formations. Some fifteen years later the present term “exocentric compounds” was introduced. It is first attested in Aleksandrow (1888: 110), going back probably to Baudouin de Courtenay (Lindner 2009: 190−192): exocentric compounds (“compounds whose semantic center is not located in its members”) and esocentric compounds (“composition whose semantic center corresponds to one of the members”), the latter, nowadays called endocentric, further subdivided into bicentric (= dvandva) and monocentric compounds (with the subtypes primocentralia and alterocentralia, depending on whether the center coincides with the first or the second member). On the history of this terminology cf. Lindner (2009) and (2011: 27−28); on reference works of the 20th and 21st century concerning Indo-European composition as well as further literature see section 3.2 as well as Lindner (2002a: 322−323, 2003: 134), and (2011: 51−53).

6. References Adelung, Johann Christoph 1782a Umständliches Lehrgebäude der Deutschen Sprache, zur Erläuterung der Deutschen Sprachlehre für Schulen. Vol. 1. Leipzig: Breitkopf. Adelung, Johann Christoph 1782b Umständliches Lehrgebäude der Deutschen Sprache, zur Erläuterung der Deutschen Sprachlehre für Schulen. Vol. 2. Leipzig: Breitkopf. Aleksandrow, Aleksander 1888 Litauische Studien I. Nominalzusammensetzungen. Dorpat: Hermann. Bahder, Karl von 1880 Die Verbalabstracta in den germanischen Sprachen ihrer Bildung nach dargestellt. Halle/S.: Niemeyer. Becker, Karl Ferdinand 1824 Die Deutsche Wortbildung oder die organische Entwickelung der deutschen Sprache in der Ableitung. Frankfurt/M.: Hermann. Becker, Karl Ferdinand 1827 Organism der Sprache als Einleitung zur deutschen Grammatik. (Deutsche Sprachlehre, Vol. 1.) Frankfurt/M.: Reinherz. [2nd ed. 1841]. Becker, Karl Ferdinand 1829 Deutsche Grammatik (Deutsche Sprachlehre, Vol. 2). Frankfurt/M.: Hermann/Kettembeil. [2nd ed. 1836, 3rd ed. 1842]. Benfey, Theodor 1852 Vollständige Grammatik der Sanskritsprache (Handbuch der Sanskritsprache, Vol. 1). Leipzig: Brockhaus. Bopp, Franz 1824−31 Vergleichende Zergliederung des Sanskrits und der mit ihm verwandten Sprachen. 5 Abhandlungen. In: Abhandlungen der historisch-philologischen Klasse der Königlichen Akademie der Wissenschaften zu Berlin, Jg. 1824−1831.

3. Word-formation in historical-comparative grammar Bopp, Franz 1827 Ausführliches Lehrgebäude der Sanskrita-Sprache. Berlin: Dümmler. Bopp, Franz 1833–52 Vergleichende Grammatik des Sanskrit, Zend, Griechischen, Lateinischen, Litthauischen, [Altslawischen,] Gothischen und Deutschen. 6 Vol. Berlin: Dümmler. Brugmann, Karl 1886 Grundriß der vergleichenden Grammatik der indogermanischen Sprachen. Vol. 1: Einleitung und Lautlehre. Straßburg: Trübner. Brugmann, Karl 1892 Grundriß der vergleichenden Grammatik der indogermanischen Sprachen. Vol. 2: Wortbildungslehre (Stammbildungs- und Flexionslehre). Straßburg: Trübner. Brugmann, Karl 1904 Kurze vergleichende Grammatik der indogermanischen Sprachen. Straßburg: Trübner. Brugmann, Karl 1906 Vergleichende Laut-, Stammbildungs- und Flexionslehre nebst Lehre vom Gebrauch der Wortformen der indogermanischen Sprachen. Vol. 1: Lehre von den Wortformen und ihrem Gebrauch. Part 1: Allgemeines, Zusammensetzung (Komposita), Nominalstämme. 2nd ed. Straßburg: Trübner. Buttmann, Philipp 3 1805, 51810 Griechische Grammatik. Berlin: Mylius. Carr, Charles T. 1939 Nominal Compounds in Germanic. London: Milford. Chantraine, Pierre 1933 La formation des noms en grec ancien. Paris: Klincksieck [Reprint 1979]. Curtius, Georg 1842 De nominum Graecorum formatione linguarum cognatarum ratione habita. Berlin: Dümmler. Curtius, Georg 1852 Griechische Schulgrammatik. Prag: Tempsky. Debrunner, Albert 1917 Griechische Wortbildungslehre. Heidelberg: Winter. Debrunner, Albert [and Jacob Wackernagel] 1954 Altindische Grammatik. Vol. 2,2: Die Nominalsuffixe. Göttingen: Vandenhoeck & Ruprecht. Delbrück, Berthold 1900 Vergleichende Syntax der indogermanischen Sprache. Vol. 3. Straßburg: Trübner. Diez, Friedrich 1838 Grammatik der romanischen Sprachen. Vol. 2. Bonn: Weber. [5th ed. 1882]. Duchesne-Guillemin, Jacques 1936 Les composés de l’Avesta. Liège/Paris: Droz. Grimm, Jacob 1819 Deutsche Grammatik. Vol. 1. Göttingen: Dieterich. [2nd ed. 1822]. Grimm, Jacob 1826 Deutsche Grammatik. Vol. 2. Göttingen: Dieterich. [= 3. Buch: Von der Wortbildung]. [2nd ed. 1878]. Heidermanns, Frank 2005 Bibliographie zur indogermanischen Wortforschung. Wortbildung, Etymologie, Onomasiologie und Lehnwortschichten der alten und modernen indogermanischen Sprachen in systematischen Publikationen ab 1800. 3 Vol. Tübingen: Niemeyer (also available as CD-ROM).

49

50

I. Word-formation as a linguistic discipline Jacobi, Hermann 1897 Compositum und Nebensatz. Studien über die indogermanische Sprachentwicklung. Bonn: Cohen. Justi, Ferdinand 1861 Ueber die zusammensetzung der nomina in den indogermanischen sprachen. Göttingen: Dieterich. Kellens, Jean 1974 Les noms-racines de l’Avesta. Wiesbaden: Reichert. Kircher-Durand, Chantal (ed.) 2002 Grammaire Fondamentale du latin. Tome IX: Création lexicale: la Formation des noms par dérivation suffixale. Louvain: Peeters. Kluge, Friedrich 1886 Nominale Stammbildungslehre der altgermanischen Dialekte. Halle/S.: Niemeyer. [3rd ed. 1926]. Leumann, Manu 1977 Lateinische Laut- und Formenlehre. Vol. 1 of Manu Leumann, Johann Baptist Hofmann and Anton Szantyr Lateinische Grammatik. München: Beck. [1st ed. 1926–28]. Lindner, Thomas 1996 Lateinische Komposita. Ein Glossar, vornehmlich zum Wortschatz der Dichtersprache. Innsbruck: Institut für Sprachwissenschaft. Lindner, Thomas 2002a Lateinische Komposita. Morphologische, historische und lexikalische Studien. Innsbruck: Institut für Sprachen und Literaturen. Lindner, Thomas 2002b Nominalkomposition und Syntax im Indogermanischen. In: Heinrich Hettrich (ed.), Indogermanische Syntax. Fragen und Perspektiven, 263–279. Wiesbaden: Reichert. Lindner, Thomas 2003 Aspekte der lateinisch-romanischen Kompositaforschung. Moderne Sprachen 47: 115– 141. Lindner, Thomas 2009 A Note on ‘endocentric’. Historiographia Linguistica 36(1): 190–192. Lindner, Thomas 2011 Komposition (Indogermanische Grammatik 4/1). Lieferung 1. Heidelberg: Winter. Lindner, Thomas 2012 Komposition (Indogermanische Grammatik 4/1). Lieferung 2. Heidelberg: Winter. Lindner, Thomas 2013 Komposition (Indogermanische Grammatik 4/1). Lieferung 3. Heidelberg: Winter. Lindner, Thomas and Renato Oniga 2005 Zur Forschungsgeschichte der lateinischen Nominalkomposition. In: Gualtiero Calboli (ed.), Lingua Latina! Proceedings of the Twelfth International Colloquium on Latin Linguistics, Bologna 2003, 149–160. Roma: Herder. Lühr, Rosemarie and Irene Balles 2008 Nominale Wortbildung des Indogermanischen in Grundzügen. Die Wortbildungsmuster ausgewählter indogermanischer Einzelsprachen. Vol. 1: Latein, Altgriechisch. Hamburg: Kovač. Lühr, Rosemarie and Joachim Matzinger 2008 Nominale Wortbildung des Indogermanischen in Grundzügen. Die Wortbildungsmuster ausgewählter indogermanischer Einzelsprachen. Vol. 2: Hethitisch, Altindisch, Altarmenisch. Hamburg: Kovač. Meid, Wolfgang 1967 Wortbildungslehre. Vol. 3 of Hans Krahe and Wolfgang Meid Germanische Sprachwissenschaft. Berlin: de Gruyter.

3. Word-formation in historical-comparative grammar Meunier, L.-Francis 1872 Les composés syntactiques en grec, en latin, en français et subsidiairement en zend et en indien. Paris: Durand et Pedone-Lauriel. Meyer-Lübke, Wilhelm 1890 Italienische Grammatik. Leipzig: Reisland. Miklosich, Franz 1875 Vergleichende Grammatik der slavischen Sprachen. Vol. 2: Stammbildungslehre. Wien: Braumüller. Miklosich, Franz 1876 Vergleichende Grammatik der slavischen Sprachen. Vol. 3: Wortbildungslehre. 2nd ed. Wien: Braumüller [1st ed. 1856 as Vol. 3: Formenlehre]. Paul, Hermann 1880 Prinzipien der Sprachgeschichte. Halle/S.: Niemeyer. [5th ed. 1920]. Paul, Hermann 1896 Über die Aufgaben der Wortbildungslehre. In: Sitzungsberichte der philosophisch-philologischen und historischen Classe der k. b. Akademie der Wissenschaften zu München. Jahrgang 1896, Heft 4, 692–713. München: Verlag der K. Akademie. [Reprinted in: Leonhard Lipka and Hartmut Günther (eds.) 1981: Wortbildung, 17–35. Darmstadt: Wissenschaftliche Buchgesellschaft]. Paul, Hermann 1920 Deutsche Grammatik. Vol. 5: Wortbildungslehre. Tübingen: Niemeyer. Pott, August Friedrich 1836 Etymologische Forschungen auf dem Gebiete der Indo-Germanischen Sprachen, insbesondere des Sanskrit, Griechischen, Lateinischen, Littauischen und Gothischen. Vol. 2: Grammatischer Lautwechsel und Wortbildung. Lemgo: Meyer. Risch, Ernst 1974 Wortbildung der homerischen Sprache. 2nd ed. Berlin/New York: de Gruyter. Sadovski, Velizar forthc. Derivation (Indogermanische Grammatik 4/2). Heidelberg: Winter. Schröder, Leopold [von] 1874 Über die formelle Unterscheidung der Redetheile im Griechischen und Lateinischen mit besonderer Berücksichtigung der Nominalcomposita. Leipzig: Köhler. Schwyzer, Eduard 1939 Griechische Grammatik. Vol. 1: Allgemeiner Teil, Lautlehre, Wortbildung, Flexion. München: Beck. Tobler, Ludwig 1868 Über die Wortzusammensetzung nebst einem Anhang über die verstärkenden Zusammensetzungen. Ein Beitrag zur philosophischen und vergleichenden Sprachwissenschaft. Berlin: Dümmler. Trendelenburg, Johann Georg 1790 Anfangsgründe der griechischen Sprache. 3rd ed. Leipzig: Barth. [1st ed. 1782]. Wackernagel, Jakob 1905 Altindische Grammatik. Vol. 2,1: Einleitung zur Wortlehre. Nominalkomposition. Göttingen: Vandenhoeck & Ruprecht [New impression 21957 with additions by Albert Debrunner]. Wackernagel and Debrunner 1954 see Debrunner 1954. Wackernagel and Debrunner 1957 see Wackernagel 1905.

Thomas Lindner, Salzburg (Austria)

51

52

I. Word-formation as a linguistic discipline

4. Word-formation in structuralism 1. 2. 3. 4. 5.

Overview Ferdinand de Saussure Directions in structuralism Word-formation in structuralist schools References

Abstract The subsumption under the designation structuralism of directions in linguistic research from the first half of the 20 th century which were very different in detail refers to efforts to found a linguistic theory which meets the demands of mathematical requirements. The guiding impulses for this development stem from F. de Saussure with his establishment of the language system as an autonomous subject of linguistics which is independent of and a precondition to all other aspects of the study of language. The language system was seen as a hierarchy of levels, the units of which are connected to one another through specific relations. The scientific ideal demanded a strict formal, inherent differentiation of levels, units and classes of the language system. The analysis was applied primarily to phonemes and morphemes. The suggested theoretical approaches established guiding foundations for phonology, morphology, and for syntactic dependency or constituent structure grammars. Word-formation played a rather marginal role. This can be explained by the strict methodological postulates.

1. Overview At the beginning of an, in a narrow sense, scientific study of natural languages stood language comparison and language history. The systematic comparison of languages lead to insights regarding the family relationships among natural languages. There was also a central interest in cultural similarities and differences. The investigation of the history of individual languages was also originally motivated by ethnological interest. With the neogrammarians, this phase of linguistics reached its zenith. Both language comparison as well as the study of the history of individual languages were, however, only based on individual linguistic forms, in particular in the context of the word. The characteristics which constitute the essence of natural language were not systematically investigated. The historical-comparative linguistics of the 19th century could carry out its program with a rather undifferentiated concept of grammar. Linguistic research was oriented towards models which had been developed for classical languages. As a fundamental criticism of this one-sided direction of research, the idea emerged at the beginning of the 20th century that language comparison and language history are only possible on a scientific basis when a theory is proposed which describes the actual core of natural language. Since human languages are by their nature sign systems, i.e. systems which link sounds with their meanings and thus make possible communication

4. Word-formation in structuralism between people, a linguistic theory must describe fundamental characteristics of linguistic sign systems. All other phenomena linked with language presuppose such a theory. The results of preceding linguistic research were, thus, not rejected on principle, but merely seen as provisional and inadequate. The goal of the directions in linguistics collectively designated as structuralism was to lift linguistic research to the level of natural sciences such as physics, chemistry and biology. The requirements for theoretical languages for these precise empirical sciences were also discussed in contemporaneous language philosophy. The Vienna Circle, in particular Rudolf Carnap, Hans Reichenbach, and Herbert Feigl, played a decisive role in the formulation of a program which rationally reconstructed the theories of empirical sciences with the help of mathematical logic. The model was Gottlob Frege’s attempt to reconstruct mathematics within the framework of logic. Logical empiricism demanded that all meaningful statements be either directly reduced to observation sentences, or that they are at least able to be brought into a logical relation to observation sentences, so that they can be verified and confirmed by accepted observation sentences. Structuralism is the designation for a phase in the development of linguistics which is quite varied in its details. The common features which led to this designation are, above all, of methodological nature. Similar stages of development can also be found in the history of the humanities, in psychology, economics, and in other social sciences. The debates on methodology in linguistics had a strong influence on anthropology, ethnology, psychology and literary studies, cf. Levi-Strauss (1958). The main motivator behind the new direction in linguistics was the Genevese linguist Ferdinand de Saussure (1857−1913). His posthumously published work Cours de linguistique générale decisively influenced the development of all structuralist lines of research. The questions he raised led directly to those of the European schools which have gone down in history as the Geneva school, the Copenhagen school and the Prague school. But also the directions in American linguistics which fall under the designation of structuralism were decisively characterized by Saussure’s ideas and formulations of the problem. The term structure is only rarely used by Saussure. His central term is system. Especially in the Prague school and in American linguistics, the term structure is found more often. Structuralism can be seen as the true onset of scientifically founded investigation of human language if one considers the introduction of strict, i.e. verifiable, methodological postulates as characteristic of scientific activity. In any case, the linguistic research of the following period would be inconceivable without structuralist approaches. The current widely accepted organization of possible topics in the scientific study of human languages was already outlined by Saussure. Even the methodological foundations of linguistic research developed by Noam Chomsky can be seen as an extension and a more precise rendering of Saussure’s ideas, cf. Motsch (2006).

2. Ferdinand de Saussure Ferdinand de Saussure initially started his career with historical linguistic research in neogrammarian circles, but then sharply criticized the research of his time. He proposed the outline of a theory of language which differed substantially in its fundamental as-

53

54

I. Word-formation as a linguistic discipline sumptions from historical-comparative linguistics. In this theory, he demands a fundamental departure from the atomistic study of language details and emphasizes the importance of an examination of the relationships into which the individual units enter with one another within the framework of a system. Only in this way can one arrive at a complex system which consists of units of different levels and their mutual relationships, strictly speaking, a system of subsystems. Language is, according to Saussure, by its very nature a system of values, of units of different levels which mutually condition each other and are only determined through their position in the system. This basic idea was taken up and developed further in, to some extent, very different variations in the linguistic research of the following period. Beginning with a basic model of linguistic signs, Saussure attempts to differentiate various levels of observation which determine separate objects of research. Central to this endeavor is the distinction between langue and parole. The langue corresponds to a sign system which is used by the speakers or writers of a social community to transmit messages to members of this community. The processes which take place during the transmission and the reception of messages compose the parole. They are individual in nature in contrast to the social character of the langue. The langue is the sign system that determines a particular individual language. A further generalization comprises the assumption of a sign system which underlies all human languages. For this, Saussure introduces the term langage, the human language faculty. Although the langue has social and psychological facets – it must be accepted by a language community and it is situated in the human brain – according to Saussure it can and must be described completely independently of these perspectives. Even meanings may only be included from the formal point of view of the system. The thoughts or ideas which stand behind the meanings, the substance of the content, form an amorphous mass which does not belong to the subject matter of the description of the system of a language. This means: the essential properties of the sign system which underlies a language are independent both of analyses of the substance of the content, the psychological existence and the social use of this system, as well as of the processes of the formation and comprehension of utterances. One can of course investigate the relationships between the sign system of a language and questions of language philosophy or its existence in the human brain, just as one can, e.g., investigate the influence of social factors on the structuring of a language into a standard language and regional and social variants, or on the vocabulary of a language. But in any case, such studies presume a description of the sign system which is completely independent of these kinds of questions. Thus an investigation of the processes of the formation and understanding of utterances, i.e. the parole, demands a knowledge of the system of a language and additionally the inclusion of psychological questions and theories as well as the consideration of the given context. According to Saussure, if one sees the sign system as a fundamental prerequisite for communication among the members of a speech community, a separation of the description of this system from an investigation of the history of this system follows automatically. Linguistic signs are valeurs, values in a system in which they coexist with other signs. The coexistence is only possible on the axis of simultaneity (axe des simultanéités). Change in the sign system is irrelevant for the user of the sign system. It arises only from the perspective of the linguist. Saussure views every natural language as very much simultaneously a contemporary institution and a product of the past (Saussure 1916: 24). But studying the relationships is only possible if the inner organism of the

4. Word-formation in structuralism language is initially described without reference to society and history (Saussure 1916: 40, 124). These considerations led Saussure to the dichotomy of synchrony and diachrony. The observations thus outlined strengthen Saussure’s assumption that the langue must be the central subject matter of linguistics. It is the only autonomous topic within the scope of the overall phenomenon of language which is not investigated by any other discipline and which is always assumed in the investigation of other aspects of language. This train of thought decisively influenced the further development of linguistics. It forms the methodological basis for almost all directions in modern linguistic research. Saussure provided stimulating guidelines with respect to the internal structure of linguistic sign systems. Especially his distinction between syntagmatic and paradigmatic relations was taken up and expanded upon by other structuralist schools. Saussure differentiates between relations into which units of the system enter in praesentia, i.e. with other linguistic units at the same level, e.g., in syntactic chains, as well as associative relations (Saussure 1916: 170); these are units which can occupy the same position in a syntagm. This differentiation forms the basis for the assumption of syntactic categories and levels of sentence structure. Syntagm and paradigm are complementary concepts. The language system thus consists of sets of units (phonemes, words, word groups) and a small quantity of very general patterns in which phonemes and words can be inserted under certain conditions. Every linguistic unit can thus also be characterized through the relationships into which it can enter with other units at the same level, cf. Wells (1947) and Jäger (2006).

3. Directions in structuralism The Geneva school: Immediately following Saussure are his students and colleagues Charles Bally, Albert Sechehaye, Henri Frei, Sergej Karcevski and Robert Godel. Their work was discussed in the Genevan Cercle F. de Saussure and was published primarily in the series Cahiers F. de Saussure. In these papers, Saussure’s thoughts and arguments are expanded upon and further developed, and misunderstandings are cleared up. Special attention is due Lucien Tesnière, also considered part of the Geneva school, who in his book “Éléments de syntaxe structurale” (Tesnière 1959), sketched in 1934 and only released posthumously in 1959, published a model of syntactic analysis which was later extended as dependency grammar. An overview of Tesnière and the further development of his ideas can be found in Godel (1961), Kunze (1975), Baumgärtner (1976), Baum (1976), Heringer, Strecker and Wimmer (1980) and Hudson (1980). The Copenhagen school: Head of the Copenhagen school was Louis Hjelmslev, who in 1943 published a programmatic work in Danish Omkring sprogteoriens grundlæggelse. An English translation appeared in 1961 under the title Prolegomena to a Theory of Language. To the Copenhagen school belong further Hans Jørg Uldall, Viggo Brøndal, Berta Siertsema, Henning Spang-Hanssen, Knut Togeby, and Paul Diderichsen.

55

56

I. Word-formation as a linguistic discipline This approach refers to itself as glossematics. The goal of this approach is to create an immanent algebra of language, a general calculus for the description of texts in a natural language. Hjelmslev initially only proposes prolegomena of such a linguistic theory, i.e. methodological principles and a few basic assumptions about fundamental properties of the system which underlies natural languages. He emphasizes that a linguistic theory cannot be a sum of hypotheses; it must rather be a system of premises and definitions which is arbitrary, yet suitable for the description of generally accepted and empirically verified data. When establishing the calculus, direct reference to either phonetic or ontological concepts is not allowed. To the premises which determine the central properties of the theory belong the distinction between expression and content as well as between form and substance. This corresponds to distinctions which had already been made by Saussure: signifié [signified] and signifiant [signifier] as well as forme [form] and substance [substance]. The calculus must take into consideration the character of natural languages as sign systems as well as the distinction between syntagmatic and paradigmatic relations. It must also permit hierarchical relations. The glossematic algebra is, as are calculuses of formal logic, a system of dependencies (functions) between terms, which are only determined through their mutual dependencies. Units must be the result of an analytic procedure, a deduction. Hjelmslev thus does not accept the discovery procedures which were later suggested in American approaches. Hjelmslev’s analysis does not lead to a separation of syntax and morphology. To a degree, the distinction between paradigmatic and syntagmatic relations corresponds to these components in other linguistic theories. Most applications of glossematics relate to areas which belong to phonology or morphology in other theories. Products of wordformation are only used as examples for general morphological questions. Glossematic grammar corresponds rather more to a dependency grammar than to a constituent structure grammar. An overview of glossematics is provided by i.a. Martinet (1946), Siertsema (1955), and Spang-Hanssen (1961). A comparison of glossematics with American linguistic theories can be found in Haugen (1951) and Garvin (1954). The Prague school: In 1926, Vilém Mathesius, Bohuslav Havránek, Jan Mukařovský, Bohumil Trnka and Josef Vachek founded the Cercle linguistique de Prague. At the first International Linguistic Congress, which took place in 1928 in The Hague, Roman Jakobson, Sergej Karcevski and Nikolai S. Trubetzkoy presented theses in the name of this circle, cf. Jakobson, Karcevski and Trubetzkoy (1929). In these theses, theoretical positions on phonology, morphology including word-formation, and on the functional view of linguistic phenomena are presented. In these theses as well, the basis is Saussure’s requirement of investigating language as an autonomous system, i.e. independently of philosophy, psychology, acoustics, sociology and other disciplines. But members of the Prague school also concern themselves with topics that go beyond the language system. In contrast to other structuralist schools, the Prague school stipulates a functional approach to the language system, taking the path from function to form, cf. Mathesius (1929). Typical for the Prague school is a broadly based interest in language phenomena. Besides phonology, morphology and syntax, i.e. the areas which belong to the language

4. Word-formation in structuralism system, problems of literary texts, text analysis, language acquisition, and a structurally oriented historical linguistics are also discussed. American approaches: The actual father of American structuralism is considered to be Leonard Bloomfield, whose groundbreaking work Language was published in 1933. His most important successors are Zellig S. Harris, Bernard Bloch, Rulon S. Wells, Charles F. Hockett and Charles C. Fries. As did the European schools originating with Saussure, the American structuralists tried to find methods which permit the identification of linguistic units and their organization into classes by purely formal means. Bloomfield, who stands at the beginning of this movement, refers principally to Saussure. Through Jakobson and Martinet, ideas of the Prague school reached the Linguistic Circle of New York in the 30s. The focal points were initially phonology and morphology. The distributionalists, especially Harris, also systematically include syntax in their linguistic analysis. Characteristic for American structuralism is the search for discovery procedures, i.e. for procedures which permit the determination of the grammatical structure of sentences after a finite number of steps. With Leonard Bloomfield, theoretical linguistics in the United States reached a new level. All following language researchers are in Bloomfield’s debt. Bloomfield’s methodological premises are influenced by the dominating physicalistic scientific ideal of his time. He only recognizes such data that can be discovered through direct observation and immediate experience. Mentalistic procedures which in principle permit reference to the linguistic knowledge of speakers/hearers of a language are excluded. Of course Bloomfield doesn’t deny the character of natural languages as sign systems, he only excludes the explicit reference to meanings in the definition of grammatical concepts. However, he states that the ability of native speakers to decide whether two forms are the same or not must be permitted (Bloomfield 1933: 77). The meanings associated with linguistic forms are, in Bloomfield’s view, not accessible to scientific description, at least not for the current state of linguistic research. They must be investigated according to principles which the behaviorist psychology of his time applied in the study of the behavior of animals. He gives an example of how one should proceed according to these principles in order to make scientifically founded statements about mental phenomena such as linguistic meaning. With this example, he would like to demonstrate that even the application of behavioristic methods to the analysis of linguistic meanings is practically impossible. We owe a significant further development of Bloomfield’s approaches to Zellig S. Harris. Harris viewed linguistics, similarly to Hjelmslev and later Chomsky, as applied mathematics (Harris 1968: 1). His first great work was published in 1951 under the title Methods in Structural Linguistics. He is the founder of distributionalism. The task of linguistics is, according to his conception, “to describe the distribution or arrangement within the flow of speech of some parts or features relatively to each other”. There are three tasks to be solved: 1. The units of different levels must be segmented, 2. The resulting segments must be classified by means of an observation of their distribution. The appearance of segments in the same context means membership in the same class. 3. The relations between the resulting classes must be determined (Harris 1951: 20). Harris assumes that these tasks can be completed via an effective procedure which after a finite number of steps determines the grammatical structure of linguistic utteran-

57

58

I. Word-formation as a linguistic discipline ces. This procedure must be possible without explicitly taking into account the meaning or the sounds of linguistic units. “In exact descriptive linguistic work […] considerations of meaning can only be used heuristically, as a source of hints, and the determining criteria will always have to be stated in distributional terms” (Harris 1951: 365 fn. 6). In fact Harris must also allow for a native speaker who only answers the questions: Is that the same? and Is that possible? Harris is however of the opinion that this only makes short cuts possible which otherwise abbreviate very elaborate procedures. “It may be presumed that any two morphemes having different meanings also differ somewhere in distribution” (Harris 1951: 365 fn. 4). The assumption of effective discovery procedures generated great interest, but was also seriously questioned. In particular two points are controversial: are the assumed procedures really effective, i.e. do they lead to an acceptable conclusion? And is the justification for the acceptability of short cuts acceptable? The tenor of the criticism is: it is practically impossible to investigate all segments in every possible context. In phonology this is in any case simpler than in morphology or syntax. For the substitution-in-frames-procedures suggested by Harris, it cannot be theoretically proven that it delivers the desired definitions after a finite number of steps (Lees 1957). Problems which arise for the definition of the concept of morpheme are also pointed out by Bierwisch (1961). The background of distributionalistic thinking is immediate constituent grammar. This type of grammar assumes a very clear and manageable schema involving a strictly hierarchical structuring of the sentence. A sentence consists of phonemes. The combination of phonemes yields morphemes. The combination of morphemes yields words and word groups. The combination of word groups yields sentences (Postal 1964). As most structuralists, Harris also follows Saussure’s view that only an immanent linguistic description of language can be the foundation for “historical linguistics, dialect geography, for relations of language to culture and personality, to phonetics and semantics and for the comparison of language structure with systems of logic” (Harris 1951: 3).

4. Word-formation in structuralist schools In a description of the history of research on English word-formation, Valerie Adams notes that American structuralism was only interested in questions of word-formation in the context of morphological and syntactic problems, since its adherents focused their main interest on units which are smaller or larger than words. Furthermore, Saussure’s distinction between synchrony and diachrony discredited the primarily historically oriented research on word-formation of the 19th century (Adams 1973: 5). It also applies to the European schools that typical phenomena of word-formation were not systematically followed up on by the founders of structuralist schools. Only to the extent that means of word-formation are associated with the analysis of morphemes or with syntactic operations were they taken into consideration. Hockett discusses, for example, models according to which words can be segmented into morphemes (Hockett 1958: 393). He prefers an item-and-arrangement model, i.e. the segmentation of expressions in units and the determination of types of arrangements of these units. The German verb form sucht ‘searches’ can, for example, be segmented in

4. Word-formation in structuralism such + t. The unit such also combines with other units: Such+e ‘search’, Ver+such+ung ‘temptation’, such+te ‘searched’. All of these units are morphemes, i.e. the smallest grammatical units which are combined with a meaning. Problems are caused by words such as Engl. cran+berry, which contain units with which no meaning is associated, words with ablaut (Engl. took : take, Ger. zieh : zog ‘pull : pulled’ and other problematic cases in which words can be segmented into grammatical units which do not correspond to a segmentation of the word into morphemes (Engl. worse, bad + comparative, Ger. Lauf ‘walk’, lauf + nominalization). Problems of this type are associated both with inflectional forms of words as well as with products of word-formation, cf. zieh ‘to pull’ : Zug ‘pull, train’, seh ‘to see’ : Sicht ‘sight’. To describe these phenomena, Bloomfield suggests an item-and-process model. Bloomfield also discusses relationships between compounds and syntactic expressions (Bloomfield 1933: 227−32). He differentiates semisyntactic compounds from asyntactic compounds. Syntactic compounds have syntactic expressions as direct parallels, cf. house keeper: keep house, blue-eyed: blue eyes. For asyntactic compounds, no syntactic constructions exist in which the members of the compound are constituents, cf. door knob. This distinction makes clear Bloomfield’s physicalistic scientific ideal. The members of a compound must be able to be immediately adjacent to each other in syntactic constructions. Without syntactic argumentation, he rejects parallels such as knob of a door or door has a knob. Tesnière includes products of word-formation which function as translatives in his syntactic theory. Characteristic of sentence structure in Tesnière’s model is the possibility to transfer words with a specified grammatical category into words with another grammatical category. In the French construction le livre rouge ‘the red book’, rouge ‘red’ is a word categorized as an adjective. But also in le livre de Pierre ‘the book of Peter’, de transfers the noun Pierre to an adjective. De is a translative and Pierre a translate; de Pierre and rouge are syntactically equivalent. The noun Pierre becomes an adjective via the preposition de in de Pierre. Adjectives can be transferred to verbs with the help of être ‘to be’, e.g., La maison est neuve ‘The house is new’. Bound morphemes can also function as translatives. In Latin Venit Romam ‘(He/she) came to Rome’ the accusative morpheme -am causes the transfer of a noun to an adverb. The use of German gut ‘good’ in the syntactic position Er schläft gut ‘he sleeps well’ makes the adjective an adverb without a translative. Verbs are seen as the basic category. But they can also be transferred to other categories. Thus, the subordination of relative clauses is a translation from a structure with verb into an adjective. An object clause makes a structure with verb into a noun. Means of word-formation can also be means of recategorization. The German verb erobern ‘to conquer’ becomes a noun via the suffix -ung, as does the adjective frei ‘free’ via the suffix -heit. The reserved attitude toward a systematic search for rules for new words is, strictly speaking, an expected consequence of the methodological principles assumed by the individual schools: 1. 2. 3. 4.

Strict synchrony; Mathematical regularity of the combination of linguistic expressions; Meanings may only be taken into consideration via purely formal methods; A mentalistic capturing of data, i.e. reference to the linguistic knowledge of native speakers, is not permitted, with the exception of strictly regulated exceptions; 5. Language use does not influence statements about the language system;

59

60

I. Word-formation as a linguistic discipline 6. These requirements cause concepts such as affixoid, unproductive rule, degree of productivity of a rule, lexicalized word, ad-hoc formation, or acceptability of a new formation to be excluded from the analysis of the system. They require a reference to historical contexts or to language use. Saussure already distinguishes suffixes from other morphemes. He assumes semantic and grammatical functions for suffixes. This is permissible, because the question of whether a morpheme is connected with a meaning or not is acceptable, just as is the question of the function of a form within a syntagm. The suffixes of word-formation (derivational suffixes) are associated with meanings, other suffixes (inflectional suffixes) are markers for grammatical relations. But problems are caused by products of wordformation which entered the lexicon of a language in historic times, cf. Ger. Brom+beere ‘blackberry’, Dick-icht ‘thicket’, Fahr-t ‘ride’, Klang ‘sound’ (nominalization of kling via ablaut). In this case, neither the question of whether the suffixes have their own meaning, nor the question of why there are no comparable new formations can be decided on the basis of the principles considered to be permissible. It should generally be noted that the syntactic aspect preferred in the structuralist theories only concerns a portion of word-formation patterns. Except for the syntactic categorization of complex words, recategorization with the help of patterns of word-formation, and the few cases of syntactic restructuring, the principles for the formation of new words are of a semantic nature, cf. Fanselow (1987) and Motsch (2000). A systematic investigation of the regularities of word-formation within the framework of the methodological principles of most structuralist schools would have led to relatively uninteresting, strongly overgeneralized statements. For example: all words occurring in texts can be divided into derivatives and compounds. Derivatives are combinations of words and affixes. Affixes are suffixes or prefixes. In contrast to words, which can occur alone, affixes are bound to words. Affixoid is strictly speaking not a permissible concept, since it assumes historical or semantic concepts. Affixes are signs, and thus carry meaning. Some suffixes have an innergrammatical function; they indicate the change in category of a word, cf. Ger. prüf ‘to test’ : Prüf+ung ‘test’, frei ‘free’ : Frei+heit ‘freedom’. Suffixes and prefixes are bound to word categories which are to be syntactically defined. We must however note that, for example, a rule “N+ig yields a possible adjective in German” is much too undifferentiated. Although there are many words for which this rule applies, cf. sandig ‘sandy’, eckig ‘angular; lit. edgy’, steinig ‘stony’, staubig ‘dusty’, wolkig ‘cloudy’, for many other words, this statement is problematic, cf. *tischig ‘tabley’, *stuhlig ‘chairy’, *zimmerig ‘roomy’, *tagig ‘day-y’, *hausig ‘housey’, *dorfig ‘villagey’, *knopfig ‘buttony’. Even if more specific statements about the meanings of affixes were permitted and subclassifications of base words were possible, the problem of overgeneralized statements would remain, cf. Motsch (2011). Compounds are combinations of words which belong to specific word categories. Whether formations with syntactic constructions as a first member – cf. Ger. warm-herz+ig ‘warm hearted; lit. warm hearty’, kurz-atm+ig ‘short winded; lit. short breathy’, Zwei-zu-eins-Sieg ‘two-toone-win’ – can be satisfactorily described in the framework of the principles of structuralist theories would need to be clarified, see article 33 on synthetic compounds in German. Compounds can be differentiated through the category membership of their constituents. Beyond that, only parallels to syntactic constructions can be noted. Probably even the classic, semantically based differentiation between determinative, copulative and exocentric compounds could not be reproduced.

4. Word-formation in structuralism It is not surprising that word-formation in European languages was primarily investigated by scholars who placed more moderate conditions on methodological principles. This includes in particular the work of Marchand (1969), Dokulil (1968), and Fleischer (1969). Synchronic analysis is the focus of this work as well. Many suggestions of structuralist approaches are adopted, although in a weakened form. The greatest impact was from Saussure’s requirement of separating synchrony and diachrony. Apart from a few historical references, one doesn’t find systematic historical investigations in the large body of works containing descriptions of word-formation. Thus, for example, an investigation of derivational affixes following the assumptions of Jacob Grimm that they trace back to independent words was almost completely lost. Grimm’s distinction between proper compounds (echte/eigentliche Komposita) and improper compounds (unechte/uneigentliche Komposita) no longer plays a systematic role either, see article 3 on word-formation in historical-comparative grammar. Particularly worth mentioning in this context is also Coseriu’s (1970) attempt to determine the structure of the vocabulary of a language. He assumes thereby ideas of the classical structuralists, in particular the postulate that all linguistic units, and therefore also the meanings of linguistic expressions, enter into syntagmatic and paradigmatic relations. He follows up on ideas of Hjelmslev, who, parallel to the segmentation of morphemes into phonemes, suggested a decomposition of the meanings of the content plane in figurae. According to Coseriu, the phenomena of the expression plane correspond to semes, with the help of which semantic relations between lexemes in the lexicon can be described. Word-formation patterns are seen as special means for the establishment of semantic relations in the lexicon of a language. Generative grammar also directed attention to structural connections which appear to point out new directions for research on word-formation. Thus Lees (1960) uses transformations in the first version of a generative grammar for the analysis of compounds formed by two nouns. He establishes a connection between syntactic constructions and compounds which makes it possible to highlight differences in meaning. The problem nevertheless remains whether all prominent phenomena of word-formation can be described in the context of a stricter theory of grammar, cf. Motsch (2011). The mentalistic perspective taken by Chomsky and the idea of a modular organization of systems of knowledge adopted from cognitive psychology freed the way for the systematic integration of semantic and pragmatic aspects in the analysis of word-formation. However, for central questions of the theory of grammar, word-formation still remains as a whole of marginal interest. Chomsky (1982: 96) took the view in a discussion with Henk van Riemsdijk that “where we have options to get an infinite vocabulary, it appears to be through pretty trivial mechanisms”. In any case, it is fair to say that structuralist schools have made great contributions to the analysis of the elementary structure of words. Fundamental concepts were worked out for analyzing the internal structure of words and their role in the combination of words to phrases and sentences. We thank structuralist approaches for concepts such as stem morpheme, derivational and inflectional morpheme, discontinuous morpheme, and free and bound morpheme. Bloomfield, Harris and Hockett contributed greatly to these results. It would be impossible to imagine modern research in morphology and wordformation without the preliminary work of the structuralists. As an example for the influence of classical structuralist schools on research in wordformation, the influential theory of Hans Marchand will be briefly illustrated here.

61

62

I. Word-formation as a linguistic discipline Marchand’s The Categories and Types of Present-Day English Word-Formation was first published in 1960 and appeared in a second edition in 1969. As one of his most well-known students, Dieter Kastovsky, writes, this book is even today “the most comprehensive synchronic description of Modern English word-formation” (Kastovsky 1999: 29). Marchand had already developed the theoretical basis of his work in the 1940s of the last century. He bases his important points on European structuralist thought, especially on the basic ideas of Saussure and extensions of these ideas by Charles Bally. But he also takes American structuralists into account. The confrontation with the wordformation theory of Robert Lees led to an extensive reworking of his book, which appeared in a second edition in 1969. In contrast to the largely neogrammarian orientation of the linguistics of his time, Marchand emphasizes the priority of a synchronic description. References to historical contexts do not have a systematic character. To the foundation of his theory of wordformation belong the concepts syntagm and motivation, which can be traced back to Saussure and Bally, cf. Bally (1944). The subject matter of word-formation are regular formations which can be analyzed as a syntagm which consists of two morphemes. Morphemes are seen as linguistic units with the character of a sign. According to Saussure, the bond between the two components of the sign, the form and the meaning, is fundamentally arbitrary and unmotivated. Complex linguistic expressions, syntagms, are on the other hand principally motivated, since they can be semantically interpreted on the basis of their constituents and a grammatical pattern. Marchand draws from this the conclusion that meaning is just as important as form, since the smallest linguistic signs, the morphemes, must be interpreted as signs (Marchand 1969: 1). He thus does not accept the trend to a purely distributional definition of linguistic units in American linguistics and makes, as did later Chomsky, the mentalistic assumption that the speakers of a language know the meanings of linguistic expressions and that this aspect of linguistic competence must be taken into consideration in an adequate theory, even though there were at the time no fully-developed proposals as to how linguistic meanings should be described in an appropriate theoretical form. This approach results in important consequences for determining the subject matter of a theory of word-formation. There are words that form a syntagm, i.e. which are formed from morphemes. Among these are, according to Marchand, compounds, prefixations, suffixations, zero-derivatives and backformations. Other analyzable words cannot be reduced to syntagms. This includes onomatopoeia (Ger. Schnick-Schnack ‘tittle-tattle’), blends (Ger. Kurlaub ‘medical treatment holiday’, from Kur ‘cure’ and Urlaub ‘vacation’), words with ablaut (Ger. Zug ‘train’ : zieh(en) ‘to pull’). Words with parts which do not have a meaning or for which there are no recognizable patterns also do not belong to synchronic descriptions of word-formation (Engl. cran+berry). In contrast to the distributionalists, Marchand rejects analyses such as Engl. re-ceive, de-ceive, conceive, etc., since neither the prefixes nor the bases have a recognizable meaning. He regards such words as monomorphemic. Word-formation is based on bimorphemic syntagms, which must be productive to a certain degree, i.e. are the basis for new formations. In contrast to Harris, who completely excludes productivity from grammatical analysis, Marchand uses this phenomenon for the definition of word-formation patterns. Harris (1951: 255) writes with theoretical consequence from his methodological premises, “The methods of descriptive linguistics cannot treat the degree of productivity of elements”. The term, to be defined by language use, offers, together with the concept of

4. Word-formation in structuralism syntagm, the possibility of excluding certain lexicalized formations, but it remains out of place in a, in principle, syntactically grounded theory. Problematic is also the assumption of a zero-morpheme. Verbs such as to dirty, to clean, to tidy have a meaning which is composed of the meaning of the homonymous adjective and the meaning of make. They thus stand in relation to the pattern ‘adj. + -ize’, which underlies the forms legal+ize, national+ize, steril+ize. According to Marchand, one can on this basis assume morphemes which do not have a phonological form. The strict concept of syntagm is noticeably softened here, as signs are assumed which do not have a phonological form. The justification based on parallel patterns with suffixes is also not convincing. In German, we have denominal verbs such as geigen ‘to (play the) fiddle’, trompeten ‘to (play the) trumpet’, kämmen ‘to comb’, angeln ‘to fish; lit. to fishing-rod’, hämmern ‘to hammer’, the meanings of which can be rendered with ‘to do something typical with N’. But there are no parallel patterns with an affix. Phenomena of this type demand a semantically based concept of word-formation pattern. Characteristic for Marchand’s theory of word-formation is the relationship of wordformation patterns to syntactic syntagms. Bally and Bloomfield have also pointed out such connections. He thus sees the relationship of compounds such as house-keeping to syntactic constructions such as keep the house. He notes, that “Cpds [compounds] are chiefly based on a ‘predicate/object’ relation, but as such cpds are, on principle, nominalised sentences” (Marchand 1969: 29). He highlights further connections to syntactic relations in his analysis of zero-nominalizations and denominal verbs. He rejects, however, the description of such relations through special restructuring rules, or transformations. The assumption of zero-morphemes makes clear that Marchand is still deeply bound to the item-and-arrangement model. Since units at the lowest syntactic level are segmented in morphemes, and morphemes are the smallest units which have meaning, in problematic cases an identity in meaning of word components is assumed as sufficient justification for the existence of signs without phonetic marking. Only the assumption of semantic patterns which are linked in different ways with phonological indices is a theoretically satisfying alternative. According to Marchand, composites, words which have been assembled, are syntagms which consist of a determinant and a determinatum. This is the case for compounds and derivatives. In head-ache, father-hood, and un-do, the first member determines in each case the second. It is typical of prefixations and suffixations that the affix also occurs with other stems: father-hood, mother-hood, boy-hood; un-do, un-fasten, unroll. This aspect was later discussed under the perspective of head of a construction. Backformations such as the verb peddle, which is later attested than the noun peddler, are only peculiar from a historical perspective; from a synchronic point of view they must be dealt with like the noun writer, which is derived from the verb write on the basis of a pattern.

5. References Adams, Valerie 1973 An Introduction to Modern English Word-Formation. London: Longman. Bally, Charles 1944 Linguistique générale et linguistique française. 2nd ed. Bern: Francke.

63

64

I. Word-formation as a linguistic discipline Baum, Richard 1976 Dependenzgrammatik. Tesnières Modell der Sprachbeschreibung in wissenschaftsgeschichtlicher und kritischer Sicht. Tübingen: Niemeyer. Baumgärtner, Klaus 1976 Konstituenz und Dependenz: Zur Integration der beiden grammatischen Prinzipien. In: Hugo Steger (ed.), Vorschläge für eine strukturale Grammatik des Deutschen, 52−77. Darmstadt: Wissenschaftliche Buchgesellschaft. Bierwisch, Manfred 1961 Über den theoretischen Status des Morphems. Studia grammatica 1: 51−89. Berlin: Akademie Verlag. Bloomfield, Leonard 1933 Language. New York: Holt. Chomsky, Noam 1982 On the Generative Enterprise. A Discussion with Riny Huybregts and Henk van Riemsdijk. Dordrecht: Foris. Coseriu, Eugenio 1970 Einführung in die strukturelle Betrachtung des Wortschatzes. Tübingen: Niemeyer. Dokulil, Miloš 1968 Zur Theorie der Wortbildungslehre. Wissenschaftliche Zeitschrift der Karl Marx Universität Leipzig. Gesellschafts- und Sprachwissenschaftliche Reihe 17: 203−211. Fanselow, Gisbert 1987 Gemeinsame Prinzipien von Wort- und Phrasensemantik. In: Brigitte Ansbach-Schnitker, Herbert E. Brekle and Johannes Roggenhofer (eds.), Neuere Forschungen zur Wortbildung und Historiographie der Linguistik, 177−194. Tübingen: Niemeyer. Fleischer, Wolfgang 1969 Wortbildung der deutschen Gegenwartssprache. Tübingen: Niemeyer. Garvin, Paul L. 1954 Review of Prolegomena to a Theory of Language by Louis Hjelmslev. Language 30(1): 69−96. Godel, Robert 1961 L’École saussurienne de Genève. In: Christine Mohrmann, Alf Sommerfelt and Joshua Whatmough (eds.), Trends in European and American Linguistics 1930−1960, 294− 299. Utrecht/Antwerpen: Spectrum. Harris, Zellig S. 1951 Methods in Structural Linguistics. Chicago: Chicago University Press. Harris, Zellig S. 1968 Mathematical Structures of Language. New York: Wiley Interscience. Haugen, Einar 1951 Directions in modern linguistics. Language 27: 211−222. Heringer, Hans Jürgen, Bruno Strecker and Rainer Wimmer 1980 Syntax: Fragen − Lösungen − Alternativen. München: Fink. Hjelmslev, Louis 1943 Omkring sprogteoriens grundlæggelse. Copenhagen: Munksgaard. Hjelmslev, Louis 1961 Prolegommena to a Theory of Language. Madison: University of Wisconsin Press. Hockett, Charles F. 1958 A Course in Modern Linguistics. New York: Macmillan. Hudson, R.A. 1980 Constituency and Dependency. Linguistics 18: 179−198. Jakobson, Roman, Sergej Karcevski and Nikolai S. Trubetzkoy 1929 Proposition au Premier Congrès International de Linguistes à La Haye, 10−15 avril 1928, 33−36. Leiden: Sijthoff.

4. Word-formation in structuralism Jäger, Ludwig 2006 Ferdinand de Saussure zur Einführung. Hamburg: Junius. Jespersen, Otto 1942 A Modern English Grammar on Historical Principles. Vol. 6: Morphology. London/ Copenhagen: George Allen & Unwin. Kastovsky, Dieter 1999 Hans Marchand’s theory of word-formation: Genesis and development. In: Uwe Clark and Peter Lucko (eds.), Form, Function and Variation in English. Studies in Honor of Klaus Hensen, 19−39. Frankfurt/M.: Lang. Kunze, Jürgen 1975 Abhängigkeitsgrammatik für das Deutsche. Berlin: Akademie Verlag. Lees, Robert 1957 Review of Syntactic Structures. Language 33: 375−408. Lees, Robert 1960 The Grammar of English Nominalizations. The Hague: Mouton. Lévi-Strauss, Claude 1958 Anthropologie Structurale. Paris: Plon. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. München: Beck. Martinet, André 1946 Au sujet des fondaments de la théorie linguistique de L. Hjelmslev. Bulletin de la Société Linguistique de Paris 124: 19−43. Mathesius,Vilém 1929 Zur Satzperspektive im modernen Englisch. Archiv für das Studium der neueren Sprachen und Literaturen 155: 200−210. Motsch, Wolfgang 2000 Syntaktische Konsequenzen von Wortbildungsmustern. In: Josef Bayer and Christine Römer (eds.), Von der Philologie zur Grammatiktheorie: Peter Suchsland zum 65. Geburtstag, 289−302. Tübingen: Niemeyer. Motsch, Wolfgang 2006 Methodologische Aspekte der neueren Sprachforschung. In: Kristel Proost and Edeltraud Winkler (eds.), Von der Intentionalität zur Bedeutung konventionalisierter Zeichen. Festschrift für Gisela Harras zum 65. Geburtstag, 327−336. Tübingen: Narr. Motsch,Wolfgang 2011 Grammatische und sprachpsychologische Aspekte der Wortbildung. In: Hilke Elsen and Sascha Michel (eds.), Wortbildung im Deutschen zwischen Sprachsystem und Sprachgebrauch. Perspektiven − Analysen − Anwendungen, 43−72. Stuttgart: ibidem. Postal, Paul 1964 Constituent Structure. The Hague: Mouton. Saussure, Ferdinand de 1916 Cours de linguistique générale. Ed. by Charles Bally and Albert Séchehay. Paris/Lausanne: Payot. Siertsema, Berta 1955 A Study of Glossematics. The Hague: Mouton. Spang-Hansen, Henning 1961 Glossematics. In: Christine Mohrmann, Alf Sommerfelt and Josua Whatmough (eds.), Trends in European and American Linguistics 1930−1960, 128−164. Utrecht/Antwerpen: Spectrum. Tesnière, Lucien 1959 Élements de syntaxe structurale. Paris: Klincksieck.

65

66

I. Word-formation as a linguistic discipline Trubetzkoy, Nikolai S. 1939 Grundzüge der Phonologie. Prag: [Harrassowitz]. Wells, Rulon S. 1947 De Saussure’s system of linguistics. Word 3: 1−3.

Wolfgang Motsch, Berlin (Germany)

5. Word-formation in inhaltbezogene Grammatik 1. 2. 3. 4.

Preliminary remarks on principles On the specifics of an inhaltbezogene theory of word-formation Conclusion and prospects References

Abstract Word-formation in inhaltbezogene Grammatik [content-based grammar] directs its interest at the activity of the word-formation faculty and thus at the creative potential developed in a language community, to make an organized reality accessible through the formation of words and concepts.

1. Preliminary remarks on principles Every standard language (written, standardized language of culture, further developed through modernization) is the manifestation of the creativity of many generations, groups and creative personalities. In confrontation with their respectively experienced reality, they have expanded their linguistic talents and – in extending the language’s means of access – created a many-membered, much usable instrument of mind that goes far further than the purposes of elementary understanding. On the one hand, through its vocabulary, it makes a network of semantic categories available, with the help of which classes of entities can be differentiated and named in a nuanced manner from various points of view, by means of partially synonymous words from lexical paradigms (word fields). On the other hand, during language acquisition, during the process of becoming familiar with the possibilities of symbolic action (of communicative practice) in a community of language and interaction, grammatical forms for the organization of speech are also conveyed and thus syntactic categories for the construction of complex forms which can categorize situations and intentions. How many and which categories are made distinct in a language and how they are applied depends obviously on the cognitive and communicative expressive needs of the speakers, the former and present participants in a lan-

5. Word-formation in inhaltbezogene Grammatik guage community. In order to gain new means of expression, they especially use the efficacy of their word-forming faculty (Humboldt 1973 [1836]: 99), their ability “sich über Zeichenkonstruktionen eine gegliederte Wirklichkeit zugänglich zu machen” [to make an organized reality accessible for themselves via symbolic constructions] (Schlögl 2004: 14). Thereby the orientation necessary for any communal life becomes possible – within the scope of the respective socio-cultural circumstances – agreement on “Weltwissen” [world knowledge] and the “Welt des gesellschaftlichen Handelns” [world of social action] (Soeffner 2004: 47). Considering the creativity of the individual language communities which have become selectively effective under specific historical conditions, their different character and often also formal as well as functional alteration of categories, it is not surprising that the discipline of comparative linguistics has ascertained “unterschiedlichste Verteilungen [...] über die Welt hin” [the most variable distributions across the world] (Hartmann 1958: 44) and come to the finding “dass sich eine Parallelität zwischen Kategorien und Realität kaum von nur einer Sprache aus bündig entscheiden lässt” [that a parallelity between categories and reality can hardly be conclusively decided from the perspective of only one language] (Hartmann 1958: 45). For example, the color spectrum is not subdivided into the same number and value of color terms in all languages, also, not all individual languages correspond in their rendering of space and time reference, for example in the temporal gradation of a verbal tense system, that is, the development of the human language faculty has not generally led to the development of language systems which completely correspond with respect to the available semantic and syntactic categories: “The empirical fact is that different cultural groups of the world have conventionalized very different sets of linguistic conventions and constructions, in some cases based on very different grammatical principles” (Tomasello 2008: 309). This is not only shown by the examination of non-European languages, for instance the “American aboriginal languages” (Boas 1948: 199 and 207). Even the genetically related languages of the Indo-European language family do not show “in allen Sprachen gleichermaßen vorhandene Kategorien ohne Defektivität und mit eindeutigen, unterschiedenen Signifikanten” [in all languages equally available categories without defectiveness and with unequivocal signifiers] (Adrados 1985: 9). From this follows the task for linguistics to also work out what goes beyond those universals assumed as probable: the specifics of individual languages with regard to their “obligatory aspects” (Deutscher 2010: 151). Making conscious points of view (Sichtweisen) with a partially differing categorial network for grasping reality is a special concern of the “inhaltbezogene Grammatik” [content-based grammar] (Weisgerber 1962) which is able to resume ideas of Wilhelm von Humboldt “über die Verschiedenheit des menschlichen Sprachbaues und ihren Einfluss auf die geistige Entwickelung des Menschengeschlechts” [on the diversity of human language construction and its influence on the mental development of the human species] (1830−35, first printing 1836). “Die Vorstellung, daß Denken und Sprache miteinander korrelieren, ist charakteristisch für den Rationalismus der Aufklärungszeit” [The notion that thinking and language correlate with one another is characteristic for the rationalism of the Age of Enlightenment] (Gardt 2000: 177). Gottfried Wilhelm Leibniz has already ascertained that language is “kein bloßer Spiegel des Verstandes mehr, der nur passiv das Denken abbildet, sondern zugleich dessen Instrument und Stimulans” [no longer a simple mirror of the intellect, which just passively reflects thinking, but also at the same time its instrument and stimulant] (Gardt 2000: 179). Thus language has a cognitive function alongside its

67

68

I. Word-formation as a linguistic discipline communicative function. Leibniz still had the hope of being able to overcome the perspectivity of perception. “Für Humboldt dagegen scheint die sprachbedingte Perspektivität fast zur conditio humana zu gehören” [For Humboldt, on the other hand, perspectivity conditioned by language appears to almost belong to the human condition] (Gardt 1999: 23). “Die Zuspitzung des Problems zur These vom sprachlichen Relativitätsprinzip geht auf den Amerikanisten B. L. Whorf zurück” [The extreme version of the problem of the thesis of the principle of linguistic relativity dates back to the Americanist B. L. Whorf] (Gipper 1972: 5; see also Gumperz and Levinson 1996 as well as Wildgen 2008: 12−14 and 42−44). “However, a weaker version of the Sapir-Whorf hypothesis is generally accepted” (Crystal 1991: 15). For all constructivist theories it is common knowledge “dass sie der Sprache ein erkenntnistheoretisches Apriori zuerkennen, Sprache als ganz entscheidend für die Wahrnehmung und intellektuelle Verarbeitung der Wirklichkeit betrachten” [that they judge language to possess an epistemological apriority, regard language as fully decisive for the perception and intellectual processing of reality] (Gardt 2009: 1199). Incorporating modern insights into the nature and role of linguistic signs (Ferdinand de Saussure, Walter Porzig, Jost Trier, Ernst Cassirer), inhaltbezogene Grammatik was designed in its objectives and its systematics above all by Leo Weisgerber (from 1942 to 1967 Professor of General Linguistics and Celtic Studies in Bonn) and founded as a linguistic line of thought which endeavors to go beyond the sound-based view of language. According to this conception, it is essential to convert sound-based observations into content-based insights (Weisgerber 1962: 119) and to make the content side, which structures reality, the point of reference of the description. This holds for the description of patterns of sentential semantics which are available to conceptualize situations (as actions or events, as existence or essence). This is also true for the examination of the vocabulary, which is not only sound-based, entered in individual entries of an alphabetically ordered dictionary – with occasional glances at semantically related words – but rather studied and described as a semantic network. In this way, the conceptual grasping of reality may become clear, the respectively reached particularity of potential for differentiation which a linguistic convention offers and suggests to the participants in a language community. With the attempt to place the mental side of language and its effects for the individual and the respective society in the focus of research, linguistics increases its value as a science of culture. It would become a “Basiswissenschaft innerhalb der Kulturwissenschaften” [basic discipline within the sciences of culture] (Kämper and Eichinger 2008: 370), and could contribute to a “Kulturanthropologie in der heutigen Konstellation einer […] zunehmend multikulturell begriffenen Welt” [cultural anthropology in today’s constellation of an […] increasingly multiculturally understood world] (Fuchs 2001: 21). “The lexicon of a language is perhaps the most direct link with cultural conceptualizations in the sense that lexical items largely act as labels, and hence ‘memory banks’ for conceptualizations that are culturally constructed […]. In short, the lexical items of human languages need to be viewed as capturing and storing cultural conceptualizations such as cultural schemas and categories” (Sharifian 2009: 168). That is why the cognitive anthropology developed in the USA sees as necessary an “ethnosemantischen Zugriff” [ethnosemantic access] (Maeder and Brosziewski 2007: 269) to “den in einer Kultur angelegten Sprachkategorien” [the language categories inherent in a culture] and points out “dass Dinge, Ereignisse und Verhaltensweisen Einheiten sind […], die überhaupt erst

5. Word-formation in inhaltbezogene Grammatik im Netz sprachlicher Differenzierungen, Abstraktionen und Verweisungen konstituiert werden” [that things, events and behaviors are entities […] which are anyway only constituted in a network of linguistic differentiations, abstractions and references] (Maeder and Brosziewski 2007: 273). “Dass Weisgerber in der (heute wieder sehr aktuellen) Debatte um sprachlichen Universalismus oder Relativismus eine berücksichtigungswerte Position eingenommen hatte, die es verdienen würde, rezipiert und diskutiert zu werden” [That Weisgerber has taken a position worthy of consideration in the (today once again very current) debate about linguistic universalism or relativism which would deserve being received and discussed] (Sylla 2008: 702) is the result of a detailed critical examination of the theses of research on language content. It is thus not surprising “dass die Sprachinhaltsforschung […] in der ehemaligen UdSSR, in Japan und in Südkorea auf Interesse stößt” [that research on language content […] meets with interest in the former USSR, in Japan and in South Korea] (Kaltz 2006: 2174). It is without doubt that it – although modified and more strongly operationalized – has become especially important for the further development of semantics and likewise for research on word-formation. For this, Hermann Paul already demanded a “Ergänzung durch die Bedeutungslehre” [supplementation through the study of meaning] (Paul 1981 [1896]: 18 and 19). One must also begin with functions and investigate “welche Ausdrucksformen dafür neben einander zur Verfügung stehen” [which forms of expression are simultaneously available in order] to get a clear view of the “Konkurrenz gleichbedeutender Ausdrucksformen” [competition of synonymous expressions] (Paul 1981 [1896]: 19). Inhaltbezogene Grammatik then seeks to fulfill the demand of reaching “eine zusammenfassende Behandlung der verschiedenen gleicher Funktion dienenden Bildungsweisen” [a summarizing treatment of the various kinds of formations which serve the same functions] and paying attention to functional “Schattierungen” [shadings] (19).

2. On the specifics of an inhaltbezogene theory of word-formation 2.1. General remarks The creativity which was mentioned initially, the productive process of the development of linguistic energeia in a language community is especially obvious in the area of wordformation. Here, the objective is to gain insight how, i.e. with which methods, objects (that which is experienced, felt, thought) which have become relevant for the text are made nameable and communicable. It is a question of preferred methods of operation for conceptualization and nomination, also of clarifying how patterns of word-formation – in the service of particular speaker strategies – “sich auf die Erfassung und Darstellung sprachlich vermittelter Wirklichkeit auswirken” [affect the capturing and depiction of reality conveyed by language] (Erben 2003: 2537). It is thus necessary to not be content with taking inventory of individual means of word-formation and with working out possible patterns of construction (structural schemata), that is with the form-oriented descriptions of a morphological theory of word-formation. In such a theory, individual forms and combinations are indeed assigned functional values, but the systematic, paradigmatically graded interaction of means of word-formation, which differ in form but

69

70

I. Word-formation as a linguistic discipline are similar in function, in the realization of patterns typical in language are not worked out. As a result, the task does not become apparent to also investigate the contentual directions of expansion of the vocabulary (Weisgerber 1971: 114) especially the development of cognitively and communicatively important derivational categories, for each of which a number of word-forming morphemes have become common. If one would like to avoid an all too simple conception of word-formation as a mechanism for the creation of words and to better perceive the creativity of the word-formation faculty in the development of speech (texts) and in the expansion of language (of the lexicon), then the perspectives and findings of the inhaltbezogene theory of word-formation must additionally be gained. Since the German language possesses a particular wealth of options in word-formation and extensive descriptions of this are available, especially the multivolume inventory of the Institut für deutsche Sprache in Mannheim: “Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache” (1973−1992, cf. Ortner and Ortner 1984; Ortner, Ortner and Wellmann 2007), the opportunity presents itself to examine the available results from the perspective of the theory of content-based word-formation and to point out much of fundamental significance. It is important “dass die in ihrer Leistung sich nahe kommenden Ableitungsmittel (wortbildenden Morpheme, Affixe) auch in ihrer inhaltlichen Nähe und Zusammengehörigkeit erkennbar gemacht [werden]” [that the means of derivation (morphemes of word-formation, affixes) which are similar to one another in terms of performance also [be] made distinguishable in terms of their contentual nearness and coherence] (Erben 1979: 158) as members of a word class (Wortstand) or a functional class (Funktionsstand). A simple example is for instance the regular interaction of -chen and -lein in the diminution of base nouns, in order to express a perspective of the speaker regarding subjective assessment, sympathy (Wellmann 1975: 123) with respect to the denoted entities. In a content-based theory of word-formation, the word-formation paradigms (word or functional classes – Wort- oder Funktionsstände) are thus to be described, which are available to the participants in a language community for particular contentual directions of expansion of the vocabulary. This availability relates not only to the possibilities of semantic modification of base words (base morphemes) such as in the case of diminution (Hünd-chen ‘doggy’) or of gender transposition, gender differentiation (Hünd-in ‘bitch’). Since word-formation occupies a central position between the lexicon and syntax, and the word is almost to be seen as a “Vermittlungseinheit zwischen semantischer und syntaktischer Struktur” [unit of mediation between semantic and syntactic structure] (Heidolph, Flämig and Motsch 1981: 159), the fact is still of fundamental importance that word-formation can make areas important to communication accessible in various semantic-syntactic subdivisions (Aufgliederungen). The word-formation faculty helps the expansion of the open classes of lexematic words, i.e. especially the nouns, adjectives (adverbs) and verbs, which in regular interaction take particular roles in the sentence as well as in the development of speech. They specify through their categorial meaning at the same time “[das] Wie der Erfassung der außersprachlichen Welt” [the how of the comprehension of the world outside of language] (Coseriu 1972: 82). This is variable, for example in the case of the words of the series warm – Wärme – erwärmen ‘warm – warmth – to warm’, which grasp and portray a phenomenon as a property, material (indeed measurable) substance, or as an action. Word-formation can therefore make possible a change in word class (transposition) and thereby in categorial

5. Word-formation in inhaltbezogene Grammatik meaning. The methods of operation of derivation can create transitions and thus increase the flexibility of statements via the remodelling of base words. Thus a base adjective such as warm ‘warm’ can be verbalized or nominalized: er-wärm-en ‘to warm’, Wärm-e ‘warmth’. In this way, various representations of a state of affairs are possible: as something that is there/thus (Das Wasser ist warm, hat eine Wärme von 38 Grad ‘The water is warm, has a warmth of 38 degrees’) or rather happens/causes/acts (Die Sonne erwärmt das Wasser ‘The sun warms the water’), so that a situation can be presented to the hearer (reader) as static or dynamic, for instance as in the “Denkmodell der Handlung” [paradigm of action] (Bühler 1934: 251).

2.2. The diversity of the formation of nouns For the formation of nouns, a particularly large number of means and patterns of wordformation for compounding and derivation have been developed in German so that constantly new or newly conceived entities can be seen and named as bearer, goal or circumstance of an occurrence (being). This occurs via semantic modification of base nouns or via transposition of verbs or adjectives to the word class of nouns (reclassification). What is captured verbally as occurrence (doing) can be detached from the individual event, e.g., x warms y, and at the same time grasped on a level of abstraction, hypostasizing and typifying, as a discussible entity via the derivation of action nouns (nomina actionis) such as Erwärm-ung ‘warming’. The necessity of often being obligated to speak about actions and actors has caused the derivational categories action noun and agent noun to be particularly markedly developed, whereby even a series of foreign suffixes has been taken into service. Thus, beside the particularly frequent -ung, the suffix -(at)ion has also become available, beside -er, we also find -ant/-ent, -(at)or, -eur and -ist. Equally important is the development of quality nouns (nomina qualitatis), which make it possible to speak about human or animal qualities or the nature of envisaged things. Here, the adjective is used – on the level of simple existential predication – but when one wishes to make general statements about qualities, then abstract nouns (Nominalabstrakta) must be activated or formed, e.g., Wärm-e ‘warmth’, Streng-e ‘strictness’, Schlau-heit ‘cleverness’, Naiv-ität ‘naivety’. For the development of important derivational categories, a large number of affixes are available which correspond in their functional direction, that is, they belong to the same functional class, and, where applicable, can be deployed in a paradigmatically graded manner, depending among other factors on the phonological-morphological nature of the base word: Bäll-chen ‘small ball’, Büch-lein ‘small book’. The range of choices and the conditions of use of converging or competing affixes must each be described, also with respect to their frequency of occurrence and the share that they each have as “Indikatoren für semantische Muster” [indicators for semantic patterns] (Motsch 2004: 12). Thus in contemporary German, it is for example the primary function of -schaft (-schaft1) to indicate collective terms for persons (Arbeiter-schaft ‘employees’, Ärzteschaft ‘medical profession’). Collective formations with -tum, the most frequent function of which (-tum1) is the assignment-of-roles type Außenseitertum ‘outsiderdom’ (‘the fact of being an outsider’), are less frequent. They only follow from a secondary function of

71

72

I. Word-formation as a linguistic discipline -tum (-tum2: Bauern-tum ‘peasantry; lit. farmer-dom’, Bürger-tum ‘bourgeoisie; lit. citizen-dom’) and are derivatives which not only designate a multiplicity grasped as a unity (of farmers or citizens) but rather “immer noch die Komponente der geistigen Richtung und des Verhaltens enthalten” [still contain the components of mental direction and of behavior] (Wellmann 1975: 178). For the functional class of collective formations as well as for other paradigms of word-formation as contentual directions of expansion of the vocabulary, “das Zusammenwirken inhaltlich benachbarter Wortnischen aus verschiedenen Ableitungstypen” [the interaction of contentually neighboring word niches from various derivational types] (Weisgerber 1981 [1964]: 42) is of particular importance, whereby in principle the possibility of semantic differentiation holds, which accentuates particular points of view. In this way, Bürger-schaft ‘the entirety of the citizens of a community’ can be differentiated from Bürger-tum ‘social class of citizens’. In the present case, especially the lemma-rich semantic niches (Baldinger 1950: 117) of the formations with -schaft1 and -tum2 serve as a model for the formation of additional collective nouns from personal nouns. The more semantic niches an affix helps to develop, the less explicit its signal value appears to be, but it is not least the number and usualness of the formations which belong to a semantic niche which are decisive for its degree of explicitness and systemic importance. For this reason statistical findings are just as indispensable as analyses substantiated by corpus linguistics. They show, for example, that the agent type Lehr-er ‘teacher’ (-er1) accounts for approximately two thirds of the total inventory of formations with -er, whereas formations such as Versprech-er ‘slip of the tongue; lit. misspeaker’ (-er5), only constitute a very small niche in the paradigm of action nouns, at most 1 % of all deverbal nouns. In the case of designations for personal roles, a combination with the gender differentiating suffix -in is possible, that is the coupling -er-in. As a result, the unmarked form which abstracts from natural gender (cf. the occupational title Lehr-er ‘someone who teaches’) becomes a designation which is unambiguously gender determined, which refers exclusively to female actors (Lehr-er-in ‘female teacher; lit. teacheress’). In contrast, according to a feminist approach, the generic, that is the common noun used for both genders, grammatically masculine forms (e.g., jeder/alle Lehr-er ‘every/all teacher/s’, Benutz-er ‘user’, Les-er ‘reader’, Mitarbeit-er ‘co-worker’) should be avoided, because “l’oscillation entre masculin spécifique et masculin générique” [the oscillation between the specific masculine and the generic masculine] suggests a specific masculine reading (Elmiger 2008: 112) and thereby “une vision masculine du monde” [a masculine view of the world] (Elmiger 2008: 35). This linguistic effect cannot be completely excluded, but is probably not generally intended. “Mit der inhaltbezogenen Wortbildungslehre rücken auch die Wortfamilien in ein neues Licht […], alles das, was aus demselben Wortstamm durch Ableitung und Zusammensetzung an Wortgut gewonnen ist” [With the content-based theory of word-formation, the word families are also shown in a new light […], all that which is won from the same word stem in terms of vocabulary through derivation and compounding] (Weisgerber 1962: 213), the “Ganzheit der dem ‘Sprachgefühl’ gegenwärtigen Bildungen gleichen Stammes” [entirety of the formations of the same stem which are available to ‘linguistic intuition’] (Weisgerber 1962: 233). Which degree is respectively reached by the diversification of the derivationally and compositionally active base morphemes depends on their relevance for the discourse of an age, and from their semantic appropriateness as an

5. Word-formation in inhaltbezogene Grammatik onomasiological base for the conceptual or metaphorical classification of phenomena. This has an effect not only in derivation, but also in compounding. In the case of the particularly frequent determinative compounds, the first element appended for differentiation states “unter welchem Gesichtspunkt eine Erscheinung dieses Feldes gesehen und gesondert ist” [under which point of view a phenomenon of this field is seen and separated] (Brinkmann 1981 [1956]: 196), cf. Wach-hund ‘watch dog’, Hüte-hund ‘herding dog’, Haus-hund ‘house dog’, Jagd-hund ‘hunting dog’, Schäfer-hund ‘shepherd dog’, Schoß-hund ‘lap dog’, Wild-hund ‘wild dog’. The conventionally preferred means of naming, which for instance highlight use or affiliation, correspond in the norm for designation characteristic of individual languages to a series of relational or constructional meanings (Erben 2006: 73), which facilitate the formation and understanding of additional compounds. In principle it is of course correct, that, e.g., a compound such as Papier-korb ‘wastepaper basket; lit. paper basket’ only says in a very abstract manner “Korb, der etwas mit Papier zu tun hat” [basket that has something to do with paper] (Coseriu 1977: 50). The semantic relation between the elements of a compound is not always more precisely characterized. But the existence in the mental lexicon of compounds which have long since become established, such as Brot-korb ‘bread box’, Flaschen-korb ‘bottle basket’, Salat-korb ‘salad basket’, Wäsche-korb ‘laundry basket’, together with the corresponding specialized and world knowledge (not acquired and formed without communicative, that is, linguistic influence) as well as the contexts typical in language (wirf die Zeitung in den Papierkorb ‘throw the newspaper in the wastepaper basket’) leave little doubt that Papier-korb cannot be a ‘basket made out of paper’, but rather a ‘basket which is used to collect paper’, a ‘basket for paper’, that is an expression of purpose. Beyond that, it can be established that for formations with this semantic pattern it is necessary “in vielen Beispielen […] das Prädikat AUFBEWAHREN zu interpolieren” [in many cases […] to interpolate the predicate AUFBEWAHREN ‘keep, store’] (Motsch 2004: 415), when the second element designates a vessel or container. According to its constructional meaning, Papier-korb is in accordance with a considerable series of additional formations such as Akten-schrank ‘filing cabinet; lit. file cabinet’, Blumen-vase ‘flower vase’, Milch-kanne ‘milk can’, Sand-kasten ‘sand box’. In the domain of compounding as well the task presents itself to go beyond the level of a theory of form-based word-formation, to recognize formations related to constructions and to grasp them as realizations of semantic patterns, and thus also the particularity of the concept and mental-linguistic classification of that which is designated. A characteristic example for this are the dialectal designations of the potato, which as a bulbshaped fruit is associated with its place of occurrence: Erd-apfel lit. ‘earth-apple’ or Grund-birne lit. ‘ground-pear’. In the lesser used type of the coordinative compounds “wird das Relatum unter zwei verschiedenen Gesichtspunkten im Hinblick auf zwei verschiedene Seins-/Existenzweisen mit zwei kohyponymen Lexemen angesprochen” [the relatum is referred to with two co-hyponymic lexemes from two different perspectives with respect to two different modes of being/existing] (Ortner and Ortner 1984: 53): Gott-mensch ‘God-man; lit. godhuman being’ or Hass-liebe ‘love-hate; lit. hate-love’. Similar bundlings of different aspects are shown by adjective compounds such as nass-kalt ‘wet and cold; lit. wetcold’ or schaurig-schön ‘scary and beautiful; lit. scary-beautiful’.

73

74

I. Word-formation as a linguistic discipline

2.3. On the formation of verbs and adjectives For the formation of verbs, above all a multitude of prefixes and prefix-like particles are available to semantically or syntactically modify base verbs or to verbalize nouns with a particular marking. Here as well many semantic niches are marked respectively, mostly with ver- and er-, the fewest with ent- and zer-, which are thus more explicit. And there are functional classes – “mit stärkerer Haltekraft und größerer Dichte, aber […] auch äußerst lose” [with stronger cohesion and greater density, but […] also very loose] (Henzen 1981 [1958]: 80) – which more or less indicate a graduated interaction of different prefixes. Thus mainly formations with an- and be- are found when the concern is to indicate the spatial orientation of a convergence on an object or its grasping through the process of the action of the verb. But an- often signals more “einen punktuellen Kontakt, behingegen die volle Erfassung des Objekts” [a punctual contact, be- on the other hand the full grasping of the object] (Kühnhold and Wellmann 1973: 202), cf. etwas an- und be-strahlen ‘to beam at and illuminate something’. If the objective is to show that the action of the verb has led to a conclusion, then formations with er- typically show the successful conclusion, with ver- on the other hand often the missing of a target, or the impairment or the elimination of an object, cf. etwas er-wirtschaften ‘to gain something’ and ver-wirtschaften ‘to squander away something’. The semantic modification of a base verb can be linked to a change in syntactic valency. In many cases, a transitivizing of intransitive verbs is brought about, that is the establishing of a relation to an object: weinen ‘to cry’ → jemanden/etwas be-weinen ‘to cry about something/someone; lit. to cry something/someone’. Even more important is the possibility of reversing common relations to objects, that is of changing the syntactic valency, which at the same time changes the perspective of the representation of a state of affairs: jemanden um etwas bitten ‘to ask someone for something’ → etwas von jemandem er-bitten ‘to ask something of someone’. Means of word-formation can extend the manner of construction of base verbs; they make possible via introduction of arguments or alteration of argument structure a presentation of states of affairs which is adequate to the situation and effective on the level of sentential semantics. Here as well it becomes clear that word-formation offers different modalities of presentation, and thus contributes significantly to the constitution of texts. Important thereby are also forms of reclassification of base verbs to deverbal adjectives. This concerns especially the formation of adjectives which evaluate verb processes as behavior (rühr-ig ‘active’) or as suitability (ess-bar ‘edible’). Formations with -bar can be formed from every transitive verb with which an action can be described. They have then a passive-modal meaning, i.e. ess-bare Früchte ‘edible fruits’ are those which can be eaten (according to human experience without harm). The sorting of phenomena, whether they are suitable for human purposes, finds a clear formulation in this type of word-formation, which spares laborious syntactic construction.

2.4. The diversity of verbal or nominal structures of comparison If one asks what kinds of contentual relations appear particularly important, which are therefore expressed in the expansion of the vocabulary in all three open word classes,

5. Word-formation in inhaltbezogene Grammatik then one encounters the endeavor to grasp new perceptions through a comparing referral back to known classes, that is to express relationships of comparison: pendel-artig ‘like a pendulum’, pendel-n ‘to move back and forth like a pendulum’, Pendl-er ‘commuter’. Often the nomination forms of compounding and derivation compete: Gold-kind lit. ‘gold-child’, gold-ig-es Kind lit. ‘golden child’ for the meaning ‘welcome, cherished like gold’. It is not surprising that especially the adjective, which also features comparing forms in its inflectional morphology (comparative and superlative), exhibits compounds with second elements such as -ähnlich ‘similar’, -artig ‘in the manner of; lit. mannerly’, -förmig ‘in the form of; lit. formly’, and -gleich ‘same’. Further, there are a multitude of suffixes with similar functional direction (cf. Kühnhold, Putzer and Wellmann 1978: 330), not only -haft, which is above all used to join something likened to another or an equivalent (comparable) thing. Beside tölpel-haft ‘booby-like’ can be found tölp-isch, with a shortened base before a vowel-initial suffix, beside balladen-haft ‘ballade-like’ also ballad-esk ‘balladesque’ (after a foreign base). In the case of specialized meanings, adjectival and nominal comparative formations can also be found with homo-, homöo-, iso- and -oid. Since judgments of comparison evidently belong to the essential mental-linguistic performances of communication, it is not surprising that such a “Vielfalt von Vergleichssignalen und Vergleichsstrukturen entwickelt [worden ist]” [diversity of signals and structures of comparison [has been] developed] (Erben 1988: 329), whereby the interaction of syntax, semantics and word-formation deserves particular interest. The central position of word-formation between lexicon and syntax becomes particularly apparent here and likewise the necessity of also making its contentual relations recognizable.

3. Conclusion and prospects The task of the theory of inhaltbezogene word-formation is to present the word-formation paradigms of a language and to give insight into mental-linguistic methods of operation: how that which is experienced, felt or thought is made nameable and communicable by means of particular patterns of word-formation. In derivation, it is necessary to describe the functional classes of affixes, which differ in form but are partially similar in function (converging or competing with each other), and which can serve as indicators of particular categories (word-formation meanings). In compounding, it is the semantic patterns (constructional meanings) which characterize the general structural schema of compounding on the level of a norm for designation of an individual language and constitute conventionally preferred means of naming, highlighting particular aspects and features of the denotatum. Ultimately the question arises how the availability of particular patterns of word-formation and derivational categories, which make possible and suggest a particular mental-linguistic grasping of reality, affect the participants in a language community. Attempts to find a “wirkungbezogene Wortbildungslehre” [effect-based theory of word-formation] (Weisgerber 1981: 51−54) have not thus far been convincing. As to the controversy surrounding the accusativization of man, cf. Helbig (1974: 156−159) and Werlen (2002: 291−293). But it is probably uncontroversial “that memory and perception

75

76

I. Word-formation as a linguistic discipline are affected by the availability of appropriate words and expressions” (Lyons 1981: 305), just as that neologisms can serve to influence opinions and to steer the behavior of particular groups. Specialized terms also open “längst nicht ausnahmslos die Möglichkeit, wertneutral über fachliche Denotate zu sprechen, die bereits vor jeder sprachlichen Gliederung klar umrissen sind. Oft klassifizieren sie die Wirklichkeit, indem sie unterschiedliche Inhaltsmerkmale eines Begriffs verbalisieren und so die Aufmerksamkeit auf jeweils spezifische Aspekte lenken” [by far not without exception the possibility to neutrally speak about specialized denotata which are already clearly outlined before any linguistic classification. They often classify reality in that they verbalize different contentual features of a concept and thus draw attention to specific respective aspects] (Reinart 2009: 498−499). Regarding laboratory studies, cf. Knorr Cetina (2007: 330−331). A convincing answer to the fundamental question, “Was machen wir mit der Sprache und was macht sie mit uns?” [What do we do with language and what does language do with us?] (Bieri 2010: 215) is evidently only to be gained in an interdisciplinary manner, in interaction between linguistics, psychology, cognitive science and translation studies. At present it must already be noted on the basis of available studies that “the evidence from research indicates that language does influence thought and perception of reality, but language does not govern thought or reality” (Tohidian 2009: 7; further cf. Deutscher 2010: 233−236).

4. References Adrados, Francisco 1985 Der Ursprung der grammatischen Kategorien des Indoeuropäischen. In: Bernfried Schlerath (ed.), Grammatische Kategorien. Funktion und Geschichte, 1−46. Wiesbaden: Harrassowitz. Baldinger, Kurt 1950 Kollektivsuffixe und Kollektivbegriff. Ein Beitrag zur Bedeutungslehre im Französischen mit Berücksichtigung der Mundarten. Berlin: Akademie Verlag. Bieri, Peter 2010 Die Vielfalt des Verstehens. In: Christian Starck (ed.), Jahrbuch der Akademie der Wissenschaften zu Göttingen 2009, 215−231. Berlin/New York: de Gruyter. Boas, Franz 1948 Race, Language and Culture. New York: MacMillan. Brinkmann, Hennig 1981 [1956−1957] Die Zusammensetzung im Deutschen. In: Leonhard Lipka and Hartmut Günther (eds.), Wortbildung, 187−199. Darmstadt: Wissenschaftliche Buchgesellschaft. Bühler, Karl 1934 Sprachtheorie. Die Darstellungsfunktion der Sprache. Jena: Fischer. Coseriu, Eugenio 1972 Semantik und Grammatik. In: Hugo Moser (ed.), Neue Grammatiktheorien und ihre Anwendung auf das heutige Deutsch, 77−89. Düsseldorf: Schwann. Coseriu, Eugenio 1977 Inhaltliche Wortbildungslehre. In: Herbert E. Brekle and Dieter Kastovsky (eds.), Perspektiven der Wortbildungsforschung. Beiträge zum Wuppertaler Wortbildungskolloquium vom 9.−10. Juli 1976. Anläßlich des 70. Geburtstags von Hans Marchand am 1. Oktober 1977, 48−61. Bonn: Bouvier. Crystal, David 1991 The Cambridge Encyclopedia of Language. Cambridge: Cambridge University Press.

5. Word-formation in inhaltbezogene Grammatik Deutscher, Guy 2010 Through the Language Glass. Why the World Looks Different in Other Languages. London: Heinemann. Elmiger, Daniel 2008 La féminisation de la langue en français et en allemand. Paris: Champion. Erben, Johannes 1979 Bemerkungen zur “inhaltbezogenen” Wortbildungslehre. Wirkendes Wort 29: 158−164. Erben, Johannes 1988 Vergleichsurteile und Vergleichsstrukturen im Deutschen. Sprachwissenschaft 13: 309− 329. Erben, Johannes 2003 Hauptaspekte der Entwicklung der Wortbildung in der Geschichte der deutschen Sprache. In: Werner Besch, Anne Betten, Oskar Reichmann and Stefan Sonderegger (eds.), Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. 2nd ed. Vol. 3, 2525−2539. Berlin/New York: de Gruyter. Erben, Johannes 2006 Einführung in die deutsche Wortbildungslehre. 5th ed. Berlin: Schmidt. Fuchs, Martin 2001 Der Verlust der Totalität. In: Heide Appelsmeyer and Elfriede Billmann-Mahecha (eds.), Kulturwissenschaft. Felder einer prozessorientierten wissenschaftlichen Praxis, 18−53. Weilerswist: Velbrück Wissenschaft. Gardt, Andreas 1999 Geschichte der Sprachwissenschaft in Deutschland. Vom Mittelalter bis ins 20. Jahrhundert. Berlin/New York: de Gruyter. Gardt, Andreas 2000 Nation und Sprache in der Zeit der Aufklärung. In: Andreas Gardt (ed.), Nation und Sprache. Die Diskussion ihres Verhältnisses in Geschichte und Gegenwart, 169−198. Berlin/New York: de Gruyter. Gardt, Andreas 2009 Stil und Bedeutung. In: Ulla Fix, Andreas Gardt and Joachim Knape (eds.), Rhetorik und Stilistik. Ein internationales Handbuch historischer und systematischer Forschung. Vol. 2, 1196−1210. Berlin/New York: de Gruyter. Gipper, Helmut 1972 Gibt es ein sprachliches Relativitätsprinzip? Untersuchungen zur Sapir-Whorf-Hypothese. Frankfurt/M.: Fischer. Gumperz, John and Stephen Levinson (eds.) 1996 Rethinking Linguistic Relativity. Cambridge: Cambridge University Press. Hartmann, Peter 1958 Sprache und Erkenntnis. Zur Konstitution des exemplarischen Bestimmens. Heidelberg: Winter. Heidolph, Karl Erich, Walter Flämig and Wolfgang Motsch 1981 Grundzüge einer deutschen Grammatik. Berlin: Akademie Verlag. Helbig, Gerhard 1974 Geschichte der neueren Sprachwissenschaft. Unter dem besonderen Aspekt der Grammatik-Theorie. Reinbek: Rowohlt. Henzen, Walter 1981 [1958] Inhaltbezogene Wortbildung. In: Leonhard Lipka and Hartmut Günther (eds.), Wortbildung, 55−81. Darmstadt: Wissenschaftliche Buchgesellschaft. Humboldt, Wilhelm von 1973 [1836] Über die Verschiedenheit des menschlichen Sprachbaues und ihren Einfluss auf die geistige Entwicklung des Menschengeschlechts. In: Michael Böhler (ed.), Schriften zur Sprache, 30−207. Stuttgart: Reclam.

77

78

I. Word-formation as a linguistic discipline Kaltz, Barbara 2006 Die Entwicklung der inhaltbezogenen Grammatik in Deutschland. In: Sylvain Auroux, Ernst Frideryk Konrad Koerner, Hans-Josef Niederehe and Kees Versteegh (eds.), Geschichte der Sprachwissenschaften. Ein internationales Handbuch zur Entwicklung der Sprachforschung von den Anfängen bis zur Gegenwart. Vol. 3, 2166−2178. Berlin/New York: de Gruyter. Kämper, Heidrun and Ludwig M. Eichinger (eds.) 2008 Sprache – Kognition – Kultur. Sprache zwischen mentaler Struktur und kultureller Prägung. Berlin/New York: de Gruyter. Knorr Cetina, Karin 2007 Neue Ansätze zur Wissenschafts- und Techniksoziologie. In: Rainer Schützeichel (ed.), Handbuch Wissenssoziologie und Wissensforschung, 328−342. Konstanz: UVK Verlagsgesellschaft. Kühnhold, Ingeburg and Hans Wellmann 1973 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Eine Bestandsaufnahme des Instituts für deutsche Sprache. Forschungsstelle Innsbruck. Erster Hauptteil: Das Verb. Mit einer Einführung von Johannes Erben. Düsseldorf: Schwann. Kühnhold, Ingeburg, Oskar Putzer and Hans Wellmann 1978 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Eine Bestandsaufnahme des Instituts für deutsche Sprache. Forschungsstelle Innsbruck. Drittter Hauptteil: Das Adjektiv. Düsseldorf: Schwann. Lyons, John 1981 Language and Linguistics. An Introduction. Cambridge: Cambridge University Press. Maeder, Christoph and Achim Brosziewski 2007 Kognitive Anthropologie. In: Rainer Schützeichel (ed.), Handbuch Wissenssoziologie und Wissensforschung, 268−275. Konstanz: UVK Verlagsgesellschaft. Motsch, Wolfgang 2004 Deutsche Wortbildung in Grundzügen. 2nd ed. Berlin/New York: de Gruyter. Ortner, Hanspeter and Lorelies Ortner 1984 Zur Theorie und Praxis der Kompositaforschung. Mit einer ausführlichen Bibliographie. Tübingen: Narr. Ortner, Lorelies, Hanspeter Ortner and Hans Wellmann 2007 Das Projekt “Deutsche Wortbildung”. In: Heidrun Kämper and Ludwig M. Eichinger (eds.), Sprach-Perspektiven. Germanistische Linguistik und das Institut für Deutsche Sprache, 91−131. Tübingen: Narr. Paul, Hermann 1981 [1896] Über die Aufgabe der Wortbildungslehre. In: Leonhart Lipka and Hartmut Günther (eds.), Wortbildung,17−35. Darmstadt: Wissenschaftliche Buchgesellschaft. Reinart, Sylvia 2009 Kulturspezifik in der Fachübersetzung. Die Bedeutung der Kulturkompetenz bei der Translation fachsprachlicher und fachbezogener Texte. Berlin: Frank & Timme. Schlögl, Rudolf 2004 Symbole in der Kommunikation. In: Rudolf Schlögl, Bernhard Giesen and Jürgen Osterhammel (eds.), Die Wirklichkeit der Symbole. Grundlagen der Kommunikation in historischen und gegenwärtigen Gesellschaften, 9−38. Konstanz: UVK Verlagsgesellschaft. Sharifian, Farzad 2009 On collective cognition and language. In: Hanna Pishwa (ed.), Language and Social Cognition. Expression of the Social Mind, 163−180. Berlin/New York: de Gruyter. Soeffner, Hans-Georg 2004 Protosoziologische Überlegungen zur Soziologie des Symbols und Rituals. In: Rudolf Schlögl (ed.), Die Wirklichkeit der Symbole. Grundlagen der Kommunikation in historischen und gegenwärtigen Gesellschaften, 41−72. Konstanz: UVK Verlagsgesellschaft.

6. Word-formation in onomasiology

79

Sylla, Bernhard Josef 2008 Leo Weisgerbers Sprachinhaltsforschung, ihre philosophischen Implikationen und ihr Bezug zu Heidegger. Ph.D. dissertation, Universidade do Minho, Braga. Tohidian, Iman 2009 Examining linguistic relativity hypothesis as one of the main views on the relationship between language and thought. Journal of Psycholinguistic Research 38: 65−74. Tomasello, Michael 2008 Origins of Human Communication. Cambridge, MA: MIT Press. Weisgerber, Leo 1962 Grundzüge der inhaltbezogenen Grammatik. 3rd ed. Düsseldorf: Schwann. Weisgerber, Leo 1971 Die geistige Seite der Sprache und ihre Erforschung. Düsseldorf: Schwann. Weisgerber, Leo 1981 [1964] Vierstufige Wortbildungslehre. In: Leonhard Lipka and Hartmut Günther (eds.), Wortbildung, 36−54. Darmstadt: Wissenschaftliche Buchgesellschaft. Wellmann, Hans 1975 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Eine Bestandsaufnahme des Instituts für deutsche Sprache. Forschungsstelle Innsbruck. Zweiter Hauptteil: Das Substantiv. Düsseldorf: Schwann. Werlen, Iwar 2002 Sprachliche Relativität. Eine problemorientierte Einführung. Tübingen/Basel: Francke. Wildgen, Wolfgang 2008 Kognitive Grammatik. Klassische Paradigmen und neue Perspektiven. Berlin/New York: de Gruyter.

Johannes Erben, Bonn (Germany)

6. Word-formation in onomasiology 1. 2. 3. 4.

Introduction Onomasiologists and their models Outlook: European word-formation issues from an onomasiological perspective References

Abstract This article is structured in a historical order. Each section is dedicated to the onomasiological model of one linguist: Miloš Dokulil, Ján Horecký, Pavol Štekauer, Andreas Blank, Peter Koch, Joachim Grzega. Some models refer particularly to word-formation, some are general models of the name-giving process. The article illustrates the various theoretical works with examples from the original works and additional examples. Finally, the article will briefly discuss a number of onomasiological word-formation issues with respect to European languages.

80

I. Word-formation as a linguistic discipline

1. Introduction In word-formation studies, approaches have always been rather analytical. An onomasiological perspective, in contrast, takes as its starting point an extralinguistic entity (a concept) and the need of a speaker to denote an extralinguistic entity and looks for forms that denote or may denote this concept. The term onomasiology was coined by Zauner in 1902, but the onomasiological approach was already carried out earlier, notably by Diez (1875). The onomasiological perspective may be purely synchronic, or static, or it may include diachronic, procedural aspects. If procedural aspects are included, the coinages of designations can be grouped into a) already existing names used in a new way (i.e. semantic change), b) borrowings, c) new coinages based on indigenous linguistic material. This latter is referred to as word-formation.

2. Onomasiologists and their models 2.1. Miloš Dokulil Dokulil (1962, 1968, 1997) elaborated the first comprehensive onomasiological theory of word-formation. Its influence reached mainly linguists from Central and Eastern European countries: Horecký (e.g., 1983, 1994, 1999), Buzássyová (e.g., 1974), Furdík (e.g., 1993), Štekauer (e.g., 1996, 1997, 1998, 2001, 2005a, 2005b), Puzynina (1969), Grzegorczykowa (1979), Szymanek (1988), Waszakowa (1994), Neščimenko (1963, 1968), Kubrjakova (e.g., 1977a, 1977b), Fleischer (1969), Polenz (1973) and Huke (1977). Dokulil’s essential idea is the “onomasiological category”, that is the different ways of structuring the concept with respect to its expression in a language, in other words: the basic conceptual structures that initiate the naming process in humans’ mind. These structures consist of two elements: As in traditional logic, the concept to be named is first classified as a member of a certain conceptual group in an “onomasiological base” or “onomasiological basis” (Cz. onomaziologická báze), and this basis is then specified by an “onomasiological mark”, or “onomasiological feature” (Cz. onomaziologický příznak); the terms basis and feature are the translations used by Horecký (cf. section 2.2), base and mark are the translations used in the works by Štekauer and Grzega (cf. sections 2.3 and 2.6); Dokulil himself uses basis and mark in his English summary. For example, in the words Cz. černice, Cz. černozem, Engl. black earth the onomasiological bases are -ice ‘something’, -zem ‘earth’ and earth respectively. The onomasiological marks are čern- ‘black’ and black. The base is always simple, the mark may be simple or complex. A simple mark in the conceptual category of SUBSTANCE is quality (e.g., Cz. černice ‘black earth’ or Engl. black·berry) or action without regard to its object (e.g., Cz. soudce ‘judge’ or Engl. singer). An example with a complex mark is Cz. knihvazač ‘book binder’, where the object of action (books) is specified, too. The two components of the mark, i.e. the determining and the determined one, may or may not be visible in the expression. In Dokulil’s system, the basic types of onomasiological structure can be determined according to the categorial nature (SUBSTANCE, ACTION, QUALITY, CIRCUMSTANCE) of the base and the determining part of the mark, called motive. For example,

6. Word-formation in onomasiology a concept of the category SUBSTANCE is determined by its relation to a concept of the category SUBSTANCE (e.g., Cz. straník ‘member of a political party’ ← strana ‘political party’ + -(n)ík ‘thing or person related to X’), QUALITY (e.g., Cz. černice ‘black earth’), ACTION (e.g., Cz. soudce ‘judge’ ← soud[it] ‘[to] judge’ + -ce ‘person having to do with X’), or CONCOMITANT CIRCUMSTANCE (e.g., Cz. večerník ‘evening paper’ ← večer ‘evening’ + -(n)ík ‘thing or person related to X’). Other onomasiological structure types are characterized in a parallel way. These types can mirror a broad range of semantic relations, for which Dokulil (1962) offers an elaborate classification. A selected structure may then be verbalized by a variety of naming units, highlighting its various aspects. After choosing an onomasiological basis and mark on the semantic level of the word-formation process, the speaker selects a word-formation base and a formative element from a set of word-formation categories, classes and subtypes on the formal level.

2.2. Ján Horecký Highlighting Dokulil’s concept of onomasiological structure, Horecký (1983) splits the name-giving process into six levels: 1. an object of extralinguistic reality or an ideal object. 2. the pre-semantic level of concepts (logical predicates, noemes). 3. the level of semantic features: Some of the logical predicates of the pre-semantic level are verbalized as semantic markers (Horecký presents 39 semantic distinctive features for Slovak, their relations, and their hierarchical organization). 4. the level of onomasiological structures: The base and the mark are selected. The base also integrates relevant grammatical categories, including a word-class. 5. the level of onomatological structures: This is the level of the inventory of morphemes for expressing the base and the mark. 6. the phonological level: The name-giver determines the specific form of morphemes and other phonological features. Due to the structure of the Slovak language, Horecký focuses on affixations. However, he occasionally uses the terms compound words and derived compound words (e.g., Horecký 1994: 44 f., 49), by which he apparently means words that are made of a compound and a derivational affix. For instance, his analysis of the Slovak word delostrelec ‘artilleryman’ (i.e. delo-strel-ec lit. ‘cannon-shoot-er’) yields the suffix -ec as the base because this suffix puts the word in the large set of nouns denoting the actor of action; the onomasiological mark of such compounds is complex.

2.3. Pavol Štekauer Notwithstanding the contributions by Dokulil and Horecký, it was Pavol Štekauer’s reaction to generativism in his 1998 book that brought the cognitive and onomasiological approach closer to the center of lexicologists’ attraction (cf. also the preliminary works by Štekauer 1996, 1997 and the concise illustrations of his onomasiological theory in Štekauer 2005a; 2001; 2000: 1−28). In Štekauer’s model, the word-finding process is

81

82

I. Word-formation as a linguistic discipline triggered by an individual speaker’s concrete intent for creating a new naming unit. The process includes five levels: 1. the conceptual level: The concept to be named is analyzed and conceptually categorized in the most general way (i.e. “SUBSTANCE, ACTION (with internal subdivision into ACTION PROPER, PROCESS, and STATE), QUALITY, and CONCOMITANT CIRCUMSTANCE (for example, that of Place, Time, Manner, etc.)” (Štekauer 2001: 11)). 2. the semantic level: The semantic components are structured. 3. the onomasiological level: This is the level of naming in a more abstract sense. One prominent semantic component is selected as the onomasiological base (representing a class like agent, object, instrument, etc.) and another as the onomasiological mark of this base (the mark, in turn, is divisible into a determining constituent – sometimes showing a specifying and a specified element – and a determined constituent). 4. the onomatological level: This is the level of naming in a more concrete sense, featured by the morpheme-to-seme-assignment principle (MSAP). Concrete morphemes are selected, allowing the speaker creativity within productivity constraints (cf. Štekauer et al. 2005). 5. the phonological level: The forms are combined according to morphological, phonotactic and suprasegmental rules. Štekauer defines the following onomasiological types: a)

b) c) d) e)

Onomasiological Type 1, where all three constituents are present in the form – onomasiological base, determining constituent, determined constituent –, e.g., [[ piano] [ play]]-[er]. Onomasiological Type 2, where the determining constituent is not present on the formal level, e.g., [lock] [ pin], [ play]-[er]. Onomasiological Type 3, where the determined (actional) constituent is not present on the formal level, e.g., [ pian(o)][ist ], [ piano][man]. Onomasiological Type 4, where the onomasiological mark cannot be split into a determining and a determined part, e.g., [un][happy] and [extra][large]. Onomasiological Type 5, traditionally known as “conversion”.

In Štekauer’s model, the classical juxtaposition of compounds and derivations is considered dispensable. Speakers simply transform the salient aspects and relations selected on the onomasiological level into words by combining morphemes (either free or bound) on the onomatological level. The relation of agent of an action, for instance, can be verbalized in English by the morphemes -man, -er, -ist, -ant and others.

2.4. Andreas Blank Blank obtained fame for his cognitive approach to historical semantics (cf. Blank 1997a), but he also applied his theory to word-formation (Blank 1997b). Although this extension did not draw very much attention, its basic elements shall be mentioned for the sake of completeness. In Blank’s model, too, speakers’ first step is the perception of salient subconcepts of the concept to be named. One subconcept for which there already is a

6. Word-formation in onomasiology name is then taken as the semantic basis for a coinage. The semantic difference between the basic concept and the concept to be named is bridged by a “co-basis” in the form of an affix or a second subconcept. These relations between basis, co-basis and the new concept rest on the same associative principles as are semantic changes, namely contiguity (i.e. the cooccurrence of things or concepts in a specific context), contrast, and similarity – principles first formulated by Roudet (1921). Blank distinguishes various types of compounding and affixation. In section 5 of his paper he mentions, among others, the type “similarity/contrast within a category + conceptual contiguity”, a type which is “characterized by the similarity between a prototype and a peripheral member as well as by conceptual contiguity”, such as in Fr. wagon-lit ‘sleeping car; lit. car-bed’, It. autostrada ‘freeway; lit. auto-street’, Sp. máquina de escribir ‘type-writer; lit. machine of writing’, and the type “similarity/contrast within a category plus metaphorical similarity”, similar to the preceding one, but in which the determinant is based on a metaphor, such as in ModE frogman.

2.5. Peter Koch Koch has proposed an overall model of word-finding processes based on an interplay of three dimensions, or axes (cf. Koch 2001: 19, 2002: 1159 ff.): a) the morphological axis (including “zero” [= semantic change], conversion, suffixation, prefixation, composition and others), b) the cognitive-associative, or semantic, axis (including various forms of contiguity, similarity and contrast), and c) the stratification axis (differentiating between foreign models and native models). Koch’s definitions of the word-formation types are apparently traditional ones. Like semantic shifts, word-formations, too, can be activated by the same cognitive-associative relations, e.g., cab driver only shows contiguity, screwdriver also shows similarity (the instrument is seen like a person who does something). Such an overall model avoids the neglect of compounds triggered by foreign influences: loan translations, loan renditions and pseudo-loans (including “neoclassical compounds”). A loan translation is created on the onomatological level: the single elements of a foreign coinage are translated, e.g., Ger. Über·mensch > Engl. super·man; Engl. sky·scraper > Fr. gratte-ciel. Loan renditions are created on the onomasiological level: the name-giver seems to pass the semantic level, looking at a foreign language on the way to the onomatological level, and returns to the native language on the onomatological level, where the image, or motive, of a foreign coinage is now rendered with native material, e.g., Engl. sky·scraper > Ger. Wolken·kratzer lit. ‘cloud-scraper’. The third crossbreed of word-formation and borrowing are lexical pseudo-loans, i.e. words that are made of foreign material, but never existed as such in the assumed donor language. So-called neoclassical compounds, such as Engl. microphone, are one form of such pseudo-loans, which have become an essential process in the creation of technical terms all over Europe and the Americas. Today, the prestige of English attracts many Europeans to form pseudo-Anglicisms. Here, it seems as if the name-giver person reaches the onomatological level and there takes material from a foreign language, which then undergoes the usual adaptation changes on the morphonological level. However, with calques it is not always easy to decide whether the coinage was really modelled on a foreign term or whether it represents an independent, albeit parallel construction.

83

84

I. Word-formation as a linguistic discipline

2.6. Joachim Grzega In Grzega (2004), an overall model of both processes and forces of lexemic change was established (cf. also the preliminary study in Grzega 2002a): the Cognitive and Social Model for Onomasiological Studies (CoSMOS) (cf. also Grzega 2007, 2009; 2012b). CoSMOS incorporates and further elaborates the models of Štekauer, Blank and Koch. Moreover, it takes into account psycholinguistic observations, too (e.g., Mangold-Allwinn et al. 1995; Dietrich 2002). It is presented as a linguistic sign model (a shorter textbook version of the model is given in Grzega and Schöner 2007: 53 ff.).

Fig. 6.1: Visualization of CoSMOS

The following paragraphs describe how to read the model, illustrating the single phases with the help of terms for the concept BENGAL TIGER. 1. The starting-point is either a thought, or the concept, or a specific referent in context, or a type of referent. The term context is to cover the speaker-hearer situation, the type of discourse, the communicative goal, the syntactical context (this comes close to what Nerlich and Clarke 1992 call “meta-semantic expert system”). Example: I have an animal in front of me and I need to give it a neutral-sounding, but catchy name. 2. The speaker classifies the referent through processing its more basic, “global” and its more specific, “local” traits. The speaker then categorizes the referent by using some kind of mental checklist for the absence and presence of specific features and by comparing the overall image of the referent with other images already in the mind. This is the perceptual level. Example: The animal is a subspecies of the tiger family. It is particularly large and some exemplars show a completely white background of their coat. In France, they call it tigre royal. 3. If the (concrete) referent can be identified as the member of a familiar (abstract) concept, the speaker may resort to an already existing designation or create a new designation, sometimes more, sometimes less consciously. The decision will be based

6. Word-formation in onomasiology on a cost-benefit-analysis in which the speaker weighs the goals of the designation and the utterance it is embedded in: do I want to sound like the hearer, do I want to sound different from others, do I want to sound primitive, elaborate, polite, impolite, do I want to simply take the first word that comes to my mind? The cost-benefit analysis can be described as “linguistic economy”. If the speaker does not want to coin an innovation consciously, but doesn’t have the best word at hand, a word whose semantics is close might even be used on the spur of the moment, in the hope that the hearer may infer the suitable meaning from the context of the interaction. In the case of intentional, conscious innovation the speaker subsequently goes through several levels of a word-finding, or name-giving, process. Example: What can I call this animal (to achieve my communicative goals)? Actually, there seems to be a supra-force, or super-force, that could be termed “pressure of acceptance” or rather a supra-goal, or super-goal, of “Get accepted”, either because the group you want to be accepted by recognizes your style as similar to theirs and different from others or because the group you want to be accepted by admires your being different from them. The less informal and the less private a communicative situation is, the more important becomes this supra-goal. Maxims of differentiation may lead to systemic language change. Enduring linguistic change can either be planned (cf., e.g., the contributions in Coulmas 1989), or it simply occurs, as a by-product. The coinage of a new designation can be incited by various, potentially concurrent, forces. Early attempts of lists of forces were unsystematic and sometimes did not neatly separate processes from forces (cf. the overview in Grzega 2004: 163−165). Blank was the first to establish a comprehensive empirically based, cognitive-linguistic catalog for forces triggering semantic change (1997a: 345 ff., 1999: 70 ff.). Grzega’s works (cf. 2002a and 2004) are the first that strive for an empirically based, comprehensive and systematic catalog of forces for all types of lexical change. These factors are positioned on a conscious-subconscious continuum using the “word death” metaphor, where the gradual subconscious loss of a word is compared to “natural (designation) death” and where the conscious avoidance of a word is compared to “(designation) murder” (these two extremes embrace several intermediate degrees) (cf. Grzega 2004: 272, 2007: 22). Among the forces of the CoSMOS catalog are cognitive, sociocultural and linguistic reasons that can also be linked with Grice’s conversational maxims (cf. Grzega 2004: 268, 2007: 21 f.). 4. The next phase in the name-giving process is again an analysis of the specific traits of the concept (= feature analysis) – with a focus on the local traits. This step is skipped if the speaker simply borrows a word from a foreign language or variety that is used for the concept in question; it can also be skipped if the speaker simply picks an already existing designation and reduces its form. Example: This species of the tiger family typically lives in India, Bengal, and neighboring countries. In contrast to other species of the wider area, it is particularly large and some exemplars showing a completely white background of their coat. In France they call it tigre royal. 5. The speaker will then highlight one or two features as a basis for the designation. Štekauer refers to this as “naming in a more abstract sense”. The designation motives (or iconemes or, as Alinei 1995 said, iconyms) are commonly based on similarity, contrast, partiality and contiguity/contact relations. These relations can affect the linguistic side as well as the extralinguistic side and the abstract as well as the concrete

85

86

I. Word-formation as a linguistic discipline side. In Štekauer’s model this is called “onomasiological level”. Here again, the speaker keeps in mind the extralinguistic context. The concrete associations can or cannot be incited by a model, which may be of a speaker’s own idiom or a foreign idiom. Example: Variant 1: This tiger typically lives in Bengal. Variant 2: This tiger typically lives in India. Variant 3: This member of the tiger family is comparatively large that it seems like the “king” of the tiger family. Variant 4: Exemplars of this tiger can be white. Variant 5: The corresponding French term tigre royal literally means ‘royal tiger’. 6. The next step is what Štekauer calls “onomatological level” (“naming in a more concrete sense”). Here, concrete morphemes are selected. The speaker either reduces the form of an already existing lexeme for the concept or has decided to create a new one, which can be achieved through several types of processes, founded either on the speaker’s own idiom or on a model from a foreign idiom or on no model at all. Due to the already existing words in a language, the perceptual level, feature analysis and onomasiological level may lead to the preference of a certain process on the onomatological level (e.g., a preference of using -er with agent nouns in English), but such model-like concept-to-process matching is not a general, let alone universal or diachronically rigid rule. Example: Variant 1 (English): {Bengal} + {tiger}. Variant 2 (Spanish): {tigre} + {India} + {relatedness}. Variant 3 (German): {König ‘king’} + {Tiger}. Variant 4 (Russian): {bel- ‘white’} + {quality morpheme/adjectival ending} + {tigr}. Variant 5 (English): {royal} + {tiger} The catalog of processes embraces these: 1.

adoption of an already existing lexeme a) of the speaker’s own idiom (semantic change) (N.B.: This also includes the phenomenon traditionally known as “semantic loan”, the copy of a certain polysemy found in a donor language.) (i) (ii) (iii) (iv) (v) (vi) (vii)

metaphor (“similar-to” relation) metonymy (“neighbor-of” relation) synecdoche (“part-of” relation) generalization and specialization (“kind-of” relation) cohyponymic transfer (“sibling-of” relation) antiphrasis and auto-antonymy (“contrast-to” relation) conceptual recategorization

b) from a foreign idiom, including dialects and jargons (loanword) (i) (ii) (iii)

“true loan” “incomplete loan” (traditionally called “morphological pseudo-loan”) “mis-loan” (i.e. folk-etymological formal change of a loan, folk-etymological semantic extension due to an only phonetically similar loan) (iv) “creative loan” (traditionally called “lexical pseudo-loan”) 2. syntactical recategorization (traditionally called “zero-derivation” and “conversion”) 3. composition (lato sensu, i.e. the combination of existing morphemes) (N.B.: This includes what is traditionally called “compounds” and “derivations”. This also includes what is traditionally called “loan translations” and “loan renditions”.)

6. Word-formation in onomasiology (i)

4.

5. 6. 7. 8. 9. 10. 11. 12.

13. 14.

“complete complex structure” (complex composites, i.e. complete determinative composites with a base and a mark, consisting of a determining component and a determined component; e.g., Engl. pianoplayer, skyscraper) (ii) “incomplete complex structure 1” (composites with absence of determining component of the mark; e.g., Engl. pianist, pianoman) (iii) “incomplete complex structure 2” (composites with absence of determined component of the mark; e.g., Engl. player) (iv) “incomplete complex structure B” (composites with absence of the base; e.g., Engl. redbreast) (v) “simplex structure” (simplex composites, no action-determinated relationship between the elements; e.g., Engl. blackberry, vice chair) (vi) “copulative structure” (copulatives composites; e.g., Engl. actor-singer, deaf-mute) (N.B.: The variety of associations and relations that can be expressed by just combining two word(stem)s was already underlined by Whitney 1875: 121 – but in a way that rather reminds one of a generative approach.) blendings (i.e. overlapping of already existing lexemes, sometimes folk-etymologically; e.g., Engl. smog ← smoke + fog ‘fog consisting of smoke’, Ger. blass erstaunt ‘highly astonished; lit. pale-astonished’ ← blass ‘pale’ + bass erstaunt ‘highly astonished; lit. well astonished’, with an otherwise obsolete word for ‘good’) back-derivation (e.g., Engl. to edit ← editor) reduplication (full or partial) (e.g., Engl. bye-bye, wishy-washy) morphological alteration (e.g., number change as in people sg ‘several human beings’ < peoplepl ‘nation’, gender change as in It. melo ‘apple-tree’ < mela ‘apple’) wordplaying phonetic-prosodic alteration (e.g., stress shift in Engl. ímport vs. impórt) graphic alteration (e.g., Engl. discrete vs. discreet) phraseologism (e.g., Engl. the red thread ‘the recurrent theme’) root creation (including onomatopoetic words, i.e. words that verbalize the sounds the concept makes, e.g., Engl. cuckoo = Ger. Kuckuck = Sp. cuco = Hung. kakukk, etc., and expressive words, i.e. words that verbalize the sounds that a concept is metaphorically related with, e.g., Engl. to peck = Ger. picken = Sp. picar = Hung. csipegetni) clarifying compounds/composites (i.e. tautological compounds/composites; e.g., Engl. hound dog instead of just hound) formal shortening of already existing designations

a) morpheme deletion (ellipsis; e.g., Engl. turkey ← turkey-hen, Fr. dinde ← geline d’Inde ‘hen from India’) b) morpheme shortening (clipping; e.g., Engl. Ger. Du. Fr. It. prof ‘professor’) c) morpheme symbolization (acronyms, incl. alphabetisms, and short-forms; e.g., Engl. laser ← light amplification by stimulated emission of radiation) Processes may be combined. 7. Then, the word is provided with a fixed form-content relation and certain grammatical traits – the sign is completed.

87

88

I. Word-formation as a linguistic discipline Example: Variant 1 (e.g., English): /beŋ'gɔːl 'taɪgər/ ‘Pantheris tigris tigris L.’, noun, regular. Variant 2 (e.g., Spanish): /tigre indio/, masculine noun. Variant 3 (e.g., German): /'kønɪkstiːgɐ/, masculine noun, declension pattern like other nouns ending in the suffix -er. Variant 4 (e.g., Russian): /'bjelʉj 'tjigr/, masculine noun, regular. 8. Eventually, the sign is phonetically realized in a concrete context. This may possibly be influenced by a foreign sound model. Example: Variant 1a (e.g., British): [beŋ'gɔːl 'taɪgə]. Variant 1b (e.g., American): [beŋ'gɔːl taɪgɚ]. Like Koch’s model, CoSMOS enables the inclusion of composition types effected by foreign influences: loan translation, loan renditions, and pseudo-loans (including the combination of Latin morphemes in “neoclassical compounds”, as Bauer, e.g., 1998 called them). In addition, the model includes the phenomenon of folk-etymologies like sparrow-grass (from Lat. asparagus): The levels of the word-finding process do not seem to be relevant; what the speaker does is assume a wrong selection on the onomatological and onomasiological level with the consequence that even the elements on the semantic level (connotation and some of the semantic markers) are re-ordered, or reinterpreted. An onomasiological approach as suggested by CoSMOS does not only allow for an abdication of the compounding-versus-derivation distinction, but also for an abdication of the compounding-versus-phrase distinction. Onomasiologically, Sp. tigre de Bengala (with a morpheme originally meaning ‘from, of’), It. tigre del Bengala (with a morpheme originally including ‘from, of’ and ‘the’), Swed. Bengalisk tiger (with an adjectival component in first place, the suffix -isk indicating relatedness), Pol. tygrys bengalski (with an adjectival component in second place, the suffix -ski indicating relatedness, again), Hung. bengáli tigris (with a derivational component in first place, the suffix -i indicating relatedness), Engl. Bengal tiger (written with a space), Ger. Bengaltiger (written as one word) and the synonymous Ger. Königstiger lit. ‘king-tiger’ (with a submorphemic linking -s-) all underwent the same process and can all be termed “incomplete complex structures type 1”, or composites with absence of the determining component of the mark. Apart from this, the model also includes, unlike Štekauer’s, cases like reindeer or hound dog either. These are remarkable, as the meaning of the second element is already encompassed in the first, which is especially visible in the compound hound dog. Here the name-giver has not gone through the word-finding process in the usual way. At the beginning of the process is an unmotivated word, e.g., rein and hound. On the onomasiological level, the name-giver selected a base, but not a mark, since the mark was already represented by the unmotivated word. Therefore, on the onomatological level, only the morpheme for the base had to be selected. On the phonological level, the original word is then treated like a mark, which is why it appears in first position in English, for example (hound dog, not *dog hound).

2.7. Contrastive summary of the onomasiological models To sum up, there are two groups of cognitive-onomasiological approaches to the namefinding process, Dokulil/Horecký/Štekauer on the one hand and Blank/Koch on the other,

6. Word-formation in onomasiology that are merged in the model presented in Grzega (2004). In the first approach the main point is the differentiation between a step in the name-finding process that can be termed naming in a more abstract sense, where an onomasiological base and mark are selected, and a step that can be termed naming in a more concrete sense, where the concrete morphemes are selected. In the second approach the associations related to extralinguistic, formal and mental types of similarity, contrast and contiguity are focussed on. Dokulil, Horecký, Blank and Koch understand “compounding” – a bit unfortunately – in its traditional formal definition and not in a definition that follows from the overall model; therefore, they view this process differently from affixation. Štekauer’s wordformation model abstains from making this distinction at all (the model only knows “onomasiological types”, excluding any sort of non-regular name-giving processes, such as blends, semantic changes and loans). The overall model in Grzega (2004) does not distinguish between the traditional forms of compounding, affixation and phrases either, but it uses the term “composites” in order to distinguish the concatenation of morphemes from other types of name-giving processes (such as blending, clipping, semantic change, borrowing). In comparison to analytical models, all onomasiological models offer the advantage of carrying out contrastive studies not only on the level of the sign, but also on previous cognitive steps.

3. Outlook: European word-formation issues from an onomasiological perspective As already demonstrated with the designations for the Bengal tiger in several European languages, an onomasiological approach enables one to overcome a purely form-driven word/phrase-distinction, which in many languages would exclude many words that functionally work not as two units, but as one (especially in Romance languages, which favor a construction “base + specified element of the mark in the form of a preposition + specifying element of the mark”, and in Slavic languages, which favor a pattern in the order “base-morpheme + specifying element of the mark + specified element of the mark in the form of an adjectival suffix” or in reverse order). Second, comparisons before the level of the sign, particularly the onomasiological and the onomatological levels, are possible as well. Only some examples can be given here for illustration. A culturally interesting topic may be the study of the frequency of suppletive vs. morphologically consociated patterns. Such opposites were first studied thoroughly by Osthoff (1899). Among other conceptual fields, he analyzed kinship terms and pointed out the replacement of the pair frate ‘brother’ and sora ‘sister’ in Northern Italy by frate and frata. Likewise, we can mention the Standard Spanish pair hermano ‘brother’ and hermana ‘sister’. Similar cases are It. figlio ‘son’ and figlia ‘daughter’, Sp. hijo ‘son’ and hija ‘daughter’, It. zio, Sp. tío ‘uncle’ and It. zia, Sp. tía ‘aunt’. A contrastive analysis of the word-pairs for ‘father : mother’, ‘brother : sister’, ‘son : daughter’, ‘uncle : aunt’ and ‘nephew : niece’ in a selection of European languages shows that Spanish has 5 and Italian and Greek 4 consociative pairs, while Polish and Rumanian have 2, English, Swedish, French, Finnish and Russian 1 and German, Dutch and Czech 0 consociative pairs. With such comparisons a consociation index may be established.

89

90

I. Word-formation as a linguistic discipline Another topic of contrastive word-formation research from an onomasiological perspective can be the degree of technical terms formed with Neo-Latin morphemes in comparison to formation with indigenous material. (Neo-)Latin morphemes that have adapted equivalents all over Europe including Russia are inter-, intra-, trans-, multi-, maxi-, mini-, auto-, -ismus, -logia, -graphia, etc. Equivalents of Engl. international are found in all European languages but Icelandic, Irish and Greek. In some languages, the internationalism has to share the stylistic ranges with a synonym formed with indigenous components, e.g., Ru. meždunarodnyj, Dan. mellemfolkelig, Cz. mezinárodní, Lith. starptautisk, and Hung. nemzetközi. Usually, the internationalism is the stylistically more marked item. The prominence of internationalisms in contrast to formations with indigenous morphemes varies from country to country. Internationalisms seem most prominent in France. Despite manifold diastratic variation, it seems as if the trend to form words with indigenous material is, on the whole, weakest in the Romance and in the English-speaking countries and strongest in Iceland, Finland, Eastern Central Europe and South Eastern Europe (cf. Grzega 2012a: 90 f.). Onomasiological word-formation studies may be especially interesting for hierarchically structured conceptual fields. Contrastive word-formation onomasiology need not be restricted to nominal categories. Comrie (2006) illustrates that in causative/anti-causative pairs (e.g., ‘to bring to an end’/‘to come to an end’) the anti-causative is often derived from the causative in European languages, while globally it is mostly the other way around. A prominent exception is English with its many labile pairs (i.e. forms that serve both as causative and anticausative). A final topic could be the investigation of less frequent word-formation processes such as acronyms, blends, and clippings. What is their real prominence in specific registers (such as expert language and youth language)? Some minor word-formation processes seem to be restricted to a certain substandard – interestingly, in several languages (e.g., back slang, known as verlan in France and šatrovački in the South Slavic languages, e.g., ‘head’: Engl. deeache for head, Fr. teuté for tête and Croat. vugla for glava). The forces influencing the decision when a new word is coined and what type of process is used are manifold. A complete catalog of such forces has been suggested by Grzega (2002b, 2004). An experiment on the preference of word-formation patterns with new words was carried out for English and Slovak by Štekauer et al. (2005) and Körtvélyessy (2009).

4. References Alinei, Mario 1995 Theoretical aspects of lexical motivation. Svenska Landsmål och Svenskt Folkliv 118(321): 1–10. Bauer, Laurie 1998 Is there a class of neoclassical compounds, and if so is it productive? Linguistics 36: 403–422. Blank, Andreas 1997a Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. Tübingen: Niemeyer.

6. Word-formation in onomasiology Blank, Andreas 1997b Outlines of a cognitive approach to word-formation. In: Proceedings of the 16 th International Congress of Linguists, Paper No. 0291 [s.pag.]. Oxford: Pergamon. Blank, Andreas 1999 Why do new meanings occur? A cognitive typology of the motivations for lexical semantic change. In: Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition, 61–90. Berlin/New York: Mouton de Gruyter. Buzássyová, Klára 1974 Sémantická štruktúra slovenských deverbatív. Bratislava: Veda. Comrie, Bernard 2006 Transitivity pairs, markedness, and diachronic stability. Linguistics 44: 302–318. Coulmas, Florian (ed.) 1989 Language Adaptation. Cambridge: Cambridge University Press. Dietrich, Rainer 2002 Psycholinguistik. Stuttgart/Weimar: Metzler. Diez, Friedrich 1875 Romanische Wortschöpfung. Bonn: Weber. Dokulil, Miloš 1962 Tvoření slov v češtině. Vol. 1: Teorie odvozování slov. Prague: Nakladatelství Československé Akademie Věd. Dokulil, Miloš 1968 Zur Theorie der Wortbildungslehre. Wissenschaftliche Zeitschrift der Karl Marx-Universität Leipzig. Gesellschafts- und sprachwissenschaftliche Reihe 17: 203–211. Dokulil, Miloš 1997 The Prague School’s theoretical and methodological contribution to “word-formation” (Derivology). In: Jarmila Panevová and Zdena Skoumalová (eds.), Obsah – výraz – význam. Miloši Dokulilovi k 85. narozeninám, 179–210. Praha: FF UK. Fleischer, Wolfgang 1969 Wortbildung der deutschen Gegenwartssprache. Leipzig: Bibliographisches Institut. Furdík, Juráj 1993 Slovotvorná motivácia a jej jazykové funkcie. Levoča: Modrý Peter. Grzega, Joachim 2002a Some thoughts on a cognitive onomasiological approach to word-formation with special reference to English. Onomasiology Online 3.4. http://www.onomasiology.de [last access 9 Sept 2014]. Grzega, Joachim 2002b Some aspects of modern diachronic onomasiology. Linguistics 40: 1021–1045. Grzega, Joachim 2004 Bezeichnungswandel: Wie, Warum, Wozu? Ein Beitrag zur englischen und allgemeinen Onomasiologie. Heidelberg: Winter. Grzega, Joachim 2007 Summary, supplement and index for Grzega, Bezeichnungswandel, 2004. Onomasiology Online 8: 18–196. http://www.onomasiology.de [last access 9 Sept 2014]. Grzega, Joachim 2009 Compounding from an onomasiological perspective. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 217–232. Oxford: Oxford University Press. Grzega, Joachim 2012a Europas Sprachen und Kulturen im Wandel der Zeit. Eine Entdeckungsreise. Tübingen: Stauffenburg.

91

92

I. Word-formation as a linguistic discipline Grzega, Joachim 2012b Lexical semantic variables. In: Juan M. Hernández-Campoy and J. Camilo Conde-Silvestre (eds.), Handbook of Historical Sociolinguistics, 271–292. Oxford: Wiley-Blackwell. Grzega, Joachim and Marion Schöner 2007 English and General Historical Lexicology. Materials for Onomasiology Seminars. Eichstätt: Katholische Universität. http://www.onomasiology.de [last access 9 Sept 2014]. Grzegorczykowa, Renata 1979 Zarys słowotwórstwa polskiego. Słowotwórstwo opisowe. Warszawa: PWN. Horecký, Ján 1983 Vývin a teória jazyka. Bratislava: SPN. Horecký, Ján 1994 Semantics of Derived Words. Prešov: Acta Facultatis Philosophicae Universitatis Šafarikanae. Horecký, Ján 1999 Onomaziologická interpretácia tvorenia slov. Slovo a slovesnost 60: 6–12. Huke, Ivana 1977 Die Wortbildungstheorie von Miloš Dokulil. Ph.D. dissertation, Universität Gießen. Koch, Peter 2001 Bedeutungswandel und Bezeichnungswandel: Von der kognitiven Semasiologie zur kognitiven Onomasiologie. Zeitschrift für Literaturwissenschaft und Linguistik 121: 7–36. Koch, Peter 2002 Lexical typology from a cognitive and linguistic point of view. In: D. Alan Cruse, Franz Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen. Vol. 1, 1142– 1178. Berlin/New York: de Gruyter. Körtvélyessy, Lívia 2009 Productivity and creativity in word-formation: A sociolinguistic perspective. Onomasiology Online 10: 1–22. http://www.onomasiology.de [last access 9 Sept 2014]. Kubrjakova, Elena S. 1977a Slovoobrazovatel’naja nominacija. In: Boris A. Serebrennikov and Anna A. Ufimceva (eds.), Jazykovaja nominacija: Obščie voprosy, 55–98. Moskva: Nauka. Kubrjakova, Elena S. 1977b Teorija nominacii i slovoobrazovanie. In: Boris A. Serebrennikov and Anna A. Ufimceva (eds.), Jazykovaja nominacija: Vidy naimenovanij, 222–303. Moskva: Nauka. Mangold-Allwinn, Roland, Stefan Baratelli, Markus Kiefer and Hans-Gerhard Koelbing 1995 Wörter für Dinge. Von flexiblen Konzepten zu Benennungen. Opladen: Westdeutscher Verlag. Nerlich, Brigitte and David D. Clarke 1992 Outline of a model for semantic change. In: Günter Kellermann and Michael D. Morrissey (eds.), Diachrony with Synchrony, 125–141. Frankfurt/M.: Lang. Neščimenko, Galina 1963 Zakonomernosti slovoobrazovanija, semantiki i upotreblenija suščestviteľnyx s suffiksami sub”ektivnoj ocenki v sovremennom českom jazyke. In: Aleksandra G. Širokova (ed.), Issledovanija po češskomu jazyku. Voprosy slovoobrazovanija i grammatiki, 105– 158. Moskva: Nauka. Neščimenko, Galina 1968 Istorija imennogo slovoobrazovanija v češskom literaturnom jazyke konca XVIII–XX vekov. Moskva: Nauka. Osthoff, Hermann 1899 Vom Suppletivwesen der indogermanischen Sprachen. Heidelberg: Hörning.

6. Word-formation in onomasiology Polenz, Peter von 1973 Synpleremik I. Wortbildung. In: Hans P. Althaus, Helmut Henne and Herbert E. Wiegand (eds.), Lexikon der germanistischen Linguistik, 145–163. Tübingen: Niemeyer. Puzynina, Jadwiga 1969 Nazwy czynności we współczesnym języku polskim (słowotwórstwo, semantyka, składnia). Warszawa: Wydawnictwo UW. Roudet, Léonce 1921 Sur la classification psychologique des changements sémantiques. Journal de Psychologie 18: 676–692. Štekauer, Pavol 1996 A Theory of Conversion in English. Frankfurt/M.: Lang. Štekauer, Pavol 1997 On some issues of blending in English word formation. Linguistica Pragensia 7: 26– 35. Štekauer, Pavol 1998 An Onomasiological Theory of Word-Formation in English. Amsterdam/Philadelphia: Benjamins. Štekauer, Pavol 2000 English Word-Formation. A History of Research (1960–1995). Tübingen: Narr. Štekauer, Pavol 2001 Fundamental principles of an onomasiological theory of English word-formation. Onomasiology Online 2.4. http://www.onomasiology.de [last access 9 Sept 2014]. Štekauer, Pavol 2005a Onomasiological approach to word-formation. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 207–232. Dordrecht: Springer. Štekauer, Pavol 2005b Compounding and affixation: Any difference? In: Wolfgang U. Dressler, Dieter Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.), Morphology and Its Demarcations. Selected Papers from the 11 th Morphology Meeting, 151–159. Amsterdam/Philadelphia: Benjamins. Štekauer, Pavol, Don Chapman, Slávka Tomaščíková and Štefan Franko 2005 Word-Formation as creativity within productivity constraints: Sociolinguistic evidence. Onomasiology Online 6: 1–55. http://www.onomasiology.de [last access 9 Sept 2014]. Szymanek, Bogdan 1988 Categories and Categorization in Morphology. Lublin: Katolicki Uniwersytet Lubelski. Waszakowa, Krystyna 1994 Słowotwórstwo współczesnego języka polskiego – rzeczowniki sufiksalne obce. Warszawa: Wydawnictwa Uniwersytetu Warszawskiego. Whitney, William D. 1875 The Life and Growth of Language. 6th ed. London: Kegan Paul, Trench and Co. Zauner, Adolf 1902 Die romanischen Namen der Körperteile. Eine onomasiologische Studie. Erlangen: Junge.

Joachim Grzega, Eichstätt and Pappenheim (Germany)

93

94

I. Word-formation as a linguistic discipline

7. Word-formation in generative grammar 1. 2. 3. 4. 5. 6. 7.

Introduction Early generative grammar The beginnings of generative morphology as a separate discipline The heyday Reactions to classic generative morphology Conclusion References

Abstract This article traces the treatment of word-formation in generative grammar from the early days of the theory in which word-formation, if discussed at all, is treated either as a matter of syntax or phonology to the present. We look at the beginnings of the theory in the work of Halle (1973) and Jackendoff (1975) and its heyday in the 1970’s and 1980’s with the work of Aronoff (1976), Lieber (1980), Williams (1981), and Selkirk (1982). Of special interest is the placement of rules of word-formation with respect to other components of the grammar as well as the issue of lexical integrity. Later developments, such as the return to a syntactic theory of word-formation and the fragmenting of generative morphology into a variety of different approaches are covered as well.

1. Introduction We use the term generative grammar to refer broadly to the various theoretical frameworks that have arisen from the work of Noam Chomsky, Morris Halle, and their students in the 1950’s and 1960’s (Chomsky 1957, 1968, 1981, 1995; Chomsky and Halle 1968). Generative grammar has gone through many phases in its more than six decades of existence. Here we will not provide a general overview of the framework, as this can be found elsewhere (e.g., Newmeyer 1986). What is critical for the purposes of this article are the philosophical and psychological foundations of generative grammar which have remained constant over the history of the theory. Those foundational assumptions are: first, that linguistics is a fundamentally mentalist enterprise in the sense that the linguist must attempt to model the native speaker’s mental representation of the grammar; second, that the grammar must be able to generate all and only acceptable utterances in the language under study; and third, that all humans are born with some set of linguistic universals, by which we mean mental constructs that allow them to acquire language, and indeed that determine the form of the grammars that they acquire. The formalization of rules, the organization of the grammar, and the assumed nature of linguistic universals have all changed in various ways over the years, but these foundational assumptions have not. In its early years, generative grammar was largely concerned with syntax and phonology, which can to some extent be interpreted as a natural reaction to the American Structuralist tradition that preceded it. The latter tradition was at least partly built on the

7. Word-formation in generative grammar attempt in the early part of the twentieth century to document what we would now call understudied and endangered languages, and was quite heavily concerned with the description of morphology and phonology, and much less so with syntax. In contrast, in its early years generative grammar had very little to say directly about morphology, and especially about word-formation, by which we mean here derivation, compounding, and conversion. Here we will focus on the treatment of word-formation in generative grammar. We will approach our subject from two directions. Our overall treatment of generative morphology will be historical, looking at the treatment of word-formation in the earliest works on generative grammar, at the rise of generative morphology as a discipline in its own right, and at the many ways in which generative morphology has subsequently developed. But we will also treat several themes that have been prominent since the inception of generative grammar, among them the status of the morpheme as a linguistic construct, the nature of rules and the organization of the mental lexicon.

2. Early generative grammar Although early work in generative grammar did not recognize word-formation as a topic of study per se, word-formation did figure obliquely in both syntactic and phonological theory. We will illustrate this here by looking briefly at two representative works that deal with different aspects of derivation and compounding. On the syntactic side, we take up Lees’s (1960) The grammar of English nominalizations and on the phonological side Chomsky and Halle’s (1968) Sound Pattern of English (henceforth, SPE).

2.1. Syntactic approaches Lees’s The grammar of English nominalizations is typical of the treatment of derivation and compounding in early generative grammar. In the earliest versions of generative theory, the grammar consisted of a set of phrase-structure rules that determine basic constituent structure and a set of transformational rules that modify initial phrase structure trees in various ways, including the addition, deletion, and reordering of morphemes and indeed whole constituents. Lees used this basic framework to explore the relationship between nominal compounds like pronghorn (a kind of sheep) and the sentential underlying structure from which they were presumed to derive, in this case the sentences The sheep has a horn. The horn is like a prong. In his derivation, a generalized transformation first combines the two base sentences and then a series of simple transformations reduces that structure to the compound, as follows (Lees 1960: 156): (1)

a. b. c. d. e. f.

The sheep has a horn. The horn is like a prong. → The sheep has a horn which is like a prong. → The sheep has a horn like a prong. → … sheep with a horn like a prong … → … sheep with a pronghorn … → … pronghorn…

95

96

I. Word-formation as a linguistic discipline The actual compound formation transformation which Lees calls NPN effects the change from (1d) to (1e), deleting a central preposition or verb, reordering the nouns on either side, and inducing the distinctive left stress pattern typical of English compounds; NPN in its simplest form is stated as in (2): (2)

x1 + x2 + x3 → x´3 + x̀1

A further process of ellipsis takes us from (1e) to the compound in (1f). Lees’s analysis of compounds obviously uses the theoretical devices at the disposal of early generative theory, but it also reflects the major question with which generative theory was preoccupied at that moment, specifically the derivational relationship between semantically “equivalent” sentences or complex words. It is important to keep in mind that in the first blush of theoretical exuberance, the power of transformational rules was not yet an issue for generative grammar.

2.2. Phonological approaches Chomsky and Halle’s SPE is a comprehensive effort to apply the principles of generative grammar to phonology and has little explicit to say about morphology. But as a great deal of English phonology is specific to particular derivational patterns, SPE of necessity requires some implicit assumptions about complex words. Several sorts of derivational complexity are recognized in SPE and are coded in lexical representations with three different sorts of boundaries. The # symbol is used to represent word boundaries, but it is also deployed to represent the boundary between a root and a so-called “neutral” affix, by which Chomsky and Halle mean an affix that does not affect the stress placement of its base (-like, -able, -ish, -ness, etc.). Non-neutral affixes, that is, those which affect stress like -ity, -ic, -al, and -ive are signalled by a + boundary. Finally, Latinate formatives like ceive, pel, mit, scribe, and so on are separated from the prefixes with which they occur (as in, for example, deceive, repel, transmit, prescribe) by a special boundary represented as =. Chomsky and Halle are not generally concerned with how affixes and bases are put together, but passing remarks give an idea of how complex words might be constructed: neutral affixes are “syntactically distinguished” (1968: 86), being “assigned to a word by a grammatical transformation […]”. Non-neutral affixes, on the other hand, are “largely, internal to the lexicon.”

3. The beginnings of generative morphology as a separate discipline The treatment of word-formation as a special subdiscipline of generative morphology is generally said to have its origins in Chomsky’s (1970) Remarks on nominalization. In this seminal article, Chomsky notes that whereas a gerundive form exists for any given verb (3), derived nominals are far more idiosyncratic in terms both of the form of the derived noun and the productivity of the process with which the noun is formed (4).

7. Word-formation in generative grammar (3)

a. John’s refusing the offer b. John’s criticizing the book

(4)

a. John’s refusal of the offer b. John’s criticism of the book

97

Given the formal and semantic regularity of the gerundive construction, a transformational derivation is conceivable, but the derived nominals, Chomsky argues, are a matter for the lexicon. The subject of what exactly the lexicon looks like is not broached in that article, but Chomsky’s proposal that complex words might be a matter for the lexicon clearly suggested to researchers of the time that the internal structure of the lexicon merited further thought. Halle (1973) takes up the structure of the lexicon in his Prolegomena to a theory of word-formation. In that article, Halle sketches a separate morphological component of the grammar which consists of a list of bases and affixes, a set of word-formation rules, and a dictionary which contains idiosyncratic information concerning semantics and phonology as well as a feature [−lexical insertion] that applies to words produced by the word-formation rules that are not “actual” words. The dictionary thus serves not only to store item-familiar words, but to filter out items that do not exist. The actual form of word-formation rules is left vague, although Halle mentions the possibility of stating them in the form of templates (1973: 10). Jackendoff’s (1975) Morphological and semantic regularities in the lexicon is another early effort to flesh out the internal structure of the lexicon. Jackendoff proposes that the lexicon should consist of lexical entries fully specified for syntactic, phonological, and semantic information about words. Lexical entries are related to one another by “lexical redundancy rules”, essentially formal devices that evaluate the degree to which those entries overlap: for example, the lexical entry for the derived word decision will be “cheaper” to the grammar to the extent that it overlaps with the entry for decide (1975: 642−643): (5)

/decīd/ +V +[NP1 __ on NP2] NP1 DECIDE ON NP2

/decīd ī + ion/ +N +[NP1’s __ on NP2] ABSTRACT RESULT OF ACT OF NP1’s DECIDING ON NP2

Lexical redundancy rules are conceived as part of the evaluation metric of a grammar, that is, the mechanism by which the simplicity of the grammar is assessed. The more redundancy, the simpler the grammar, simplicity, of course, being highly valued in a generative grammar. The creative aspect of word-formation is not a central concern in Jackendoff’s article, although he does note that to the extent that new forms may be derived, the lexical redundancy rules are permitted to function as generative devices. Both proposals are very much products of their times. Both display an emphasis on formalism that has characterized generative grammar throughout its existence. And both also exhibit a somewhat prescriptive stance towards the lexicon, starting from the assumption that there is something we might characterize as “the existing words of a

98

I. Word-formation as a linguistic discipline language”. Perhaps the most important point that they have in common, however, is simply the fact that they identify morphology as a component of the grammar distinct from both phonology and syntax, and worthy of study in its own right.

4. The heyday The mid-1970’s saw the publication of the first monograph on generative morphology, Aronoff’s (1976) Word-formation in generative grammar, and this was quickly followed by a number of developments that set the stage for discussions of word-formation within generative grammar for the next two decades. We have in mind here the work of Siegel (1974, 1977), Allen (1978), Lieber (1980), Williams (1981), Kiparsky (1982), Selkirk (1982), Scalise (1984) and Di Sciullo and Williams (1987), among others. Rather than look chronologically at key works of generative morphology within this period, we will approach our review thematically in this section, looking at the overall architecture of the theories, including the status of the morpheme as a theoretical construct, the formal nature of rules of word-formation, and the relationship among components of the grammar, and then at a number of specific proposals that form some of the foundational claims of generative morphology.

4.1. Morphology and the architecture of the grammar As the first full-length monograph treating morphology in the generative tradition, Aronoff’s work set the stage for the set of issues that dominated generative morphology in its first decade. Prominent among these was the status of the morpheme and its role in the theory of word-formation. The morpheme is typically defined as the minimal meaningful unit of which words are constituted. Aronoff (1976) begins by pointing out the major issue with this classic definition, namely that there are bits of words that can be formally isolated but are not meaningful (so-called cran-morphs, Latinate formatives like pel, mit, and ceive). He therefore rejects the traditional definition of morpheme, and this move in turn leads him to the proposal that, “All regular word-formation processes are word-based. A new word is formed by applying a regular rule to a single already existing word” (1976: 21). Word-formation rules (WFRs) are conceived of formally as phonological operations on a designated set of bases. In addition to changing the phonological form of the base, WFRs specify a syntactic category and subcategorization, a semantic representation for their output, and a set of appropriate bases. (6) is Aronoff’s WFR for negative un- (1976: 63): (6)

Rule of negative un# a) [X]Adj → [un#[X]Adj]Adj semantics (roughly) un#X = not X b) Forms of the base 1. XVen (where en is the marker for the past participle) 2. XV #ing 3. XVable

7. Word-formation in generative grammar 4. 5. 6. 7.

X+y X+ly X#ful X#like

(worthy) (seemly) (mindful) (warlike)

For cases in which the base to which an affix attaches is not apparently a surface word (for example, nomin in nominee or proscrip in proscription), Aronoff proposes rules of truncation or readjustment. The form nominee, for example, starts out as nominate + ee, with the -ate affix being removed by truncation. As we will see shortly it is important that WFRs in Aronoff’s theory look formally very much like phonological rules, albeit with a wider range of restrictions. Subsequent generative models of this classic period can be divided into those that embrace Aronoff’s word-based framework and those that do not. Scalise (1984), for example, falls into the former camp, although in a slightly different form: he rightly points out that the notion of ‘word’ needs to be modified to ‘stem’ for highly inflected languages like Latin where the bases to which WFRs apply are never themselves free forms. On the other hand, the theories of Lieber (1980), Williams (1981) and Selkirk (1982) explicitly reject the word-based model and assume the traditional morpheme as a legitimate unit of analysis. In all three theories affixes are treated as lexical items on a par with bases as units having lexical entries. Affixes and bases are assembled by rules formally analogous to the phrase-structure rules of generative grammar; the three proposals differ in the technical characterization of the word-structure rules that create complex words; Selkirk’s and Williams’s rules, for example, look like the phrase-structure rules in the standard theory of generative grammar (7a). Lieber’s rules are more like the rules of the later principles and parameters model, with lexical entries like (7b) for morphemes that are then inserted into unlabeled binary-branching word trees to which percolation conventions (see below) apply. (7)

a. A → N AAf (Selkirk 1982: 66) b. -ize (phonological representation) semantic representation: causative category/subcategorization: ]N __ ]V insertion frame: NP __ (NP) diacritics: Level II

These theories make slightly different predictions, but in many ways they are comparable. A second issue that forms part of the landscape of generative morphology concerns the relationship among the components of the grammar. The vast majority of generative morphologists professed some form of what has come to be called the strong lexicalist hypothesis or the lexical integrity hypothesis. Lapointe (1980: 8), for example, proposes what he calls the generalized lexicalist hypothesis, which states that “[n]o syntactic rule can refer to elements of morphological structure”. This version of the hypothesis in effect rules out early generative analyses like Lees’s (1960) as well as Levi’s (1978) transformational analysis of complex nominals. Selkirk’s version, the word structure

99

100

I. Word-formation as a linguistic discipline autonomy condition (1982: 70), is similar. Di Sciullo and Williams’ version is the atomicity thesis, which is a bit more explicit: “Words are ‘atomic’ at the level of phrasal syntax and phrasal semantics. The words have ‘features’, or properties, but these features have no structure, and the relation of these features to the internal composition of the word cannot be relevant in syntax […]” (1987: 49). A corollary of the strong lexicalist hypothesis is Botha’s (1983) no phrase constraint, which prevents syntactic phrases from being the input to rules of derivation and compounding. An important consequence of the lexicalist hypothesis is that generative morphologists like Lieber, Selkirk, and Williams treat both inflection and word-formation as part of a single morphological component of the grammar. Lieber’s argument in favor of this arrangement of the grammar hinges on the behavior of inflectional allomorphy, specifically that in highly inflected languages stem allomorphs that figure in inflectional paradigms also form the input to rules of derivation. Anderson (1982), in contrast, advocates a division between inflection and word-formation; arguing on the basis of case marking and agreement in Breton, for example, he suggests that inflectional features must be generated as part of the syntactic component, with features being associated with phonological forms only after syntactic operations have taken place. This separation between word-formation, which takes place in the morphological component, and inflection, which is part of the syntax, has come to be known as the weak lexicalist hypothesis. Classical generative morphology also saw the formulation of a number of theoretical restrictions on word-formation or word-structure rules, among them the righthand head rule, feature percolation conventions, the unitary base hypothesis, the unitary output hypothesis, the atom condition, and the adjacency condition. One of the most influential proposals to come out of classical generative morphology is the righthand head rule. Borrowing the concept of “head” from syntactic theory, the head of a word was defined as that morpheme that determines the category and morphosyntactic features of the word as a whole. According to Williams (1981), the rightmost morpheme in a complex word is the head, a principle that is known as the righthand head rule (RHHR). While the RHHR is largely correct for English word-formation, it is problematic in at least two ways. Lieber (1980) points out that is not necessarily true in other languages: for example, Vietnamese is a language that has left-headed compounds. Moreover, there are clear exceptions in English as well, where, for example, the prefix de- can be category-changing and therefore are at least apparently heads, and where several affixes like -dom, -hood, and -ship do not change category, and therefore are righthand elements that cannot be considered heads. An alternative to the RHHR is what Lieber (1980) refers to as percolation conventions. In Lieber’s version of generative morphology, features belonging to an affix percolate (in other words are attributed to) the derived word as a whole whether the affix is a prefix or a suffix. Most, but not all, prefixes in English lack categorial and other morphosyntactic features, which accounts for why English is mostly but not entirely right headed. While feature percolation is clearly a weaker proposal than the RHHR, it is cross-linguistically more supportable. Another important theoretical proposal is Aronoff’s (1976) unitary base hypothesis which require that WFRs be stated to apply to a single sort of base. According to the unitary base hypothesis, WFRs cannot be formulated, for example, to apply to both nouns and verbs. Clear counterexamples can be found in English and in many other languages, however. For example, the suffix -er in English can attach to verbs to form

7. Word-formation in generative grammar agent or instrument nouns like writer or computer, but it also attaches to nouns to form personal or instrument nouns like villager or freighter. To the extent that the semantic contribution of -er seems to be the same in both instances, this looks like evidence against the unitary base hypothesis. Cases like this can be circumvented only by claiming that what appears superficially to be a single affix is actually two homophonous affixes. Aronoff’s example is the suffix -able, which he argues is in fact two suffixes, one that attaches to nouns (marriageable) and another to verbs (washable). Scalise (1984) points out an obvious problem with this fix: since nearly every prefix in English attaches to multiple syntactic catgories, we would therefore be forced into the untenable conclusion that we have as two or more homophonous affixes for the vast majority of prefixes. Scalise also proposes what he calls the unitary output hypothesis, which states that WFRs must have as their output words of a single morphosyntactic category. Like the unitary base hypothesis, the unitary output hypothesis has potential counterexamples; for example, the prefix un- in English attaches to adjectives (unhappy) to form adjectives and to verbs (unwind) to form verbs, but has nevertheless been argued to be a single negative prefix (Horn 1989; Lieber 2004). Several proposals in generative morphology are concerned with the locality of wordformation rules. The atom condition proposed by Williams (1981) and the adjacency condition proposed by both Siegel (1977) and Allen (1978) all prohibit the statement of morphological rules from seeing any but the outermost layer of structure. In other words, in a structure like [[x]y], a rule can refer to information contained in y but not x. As in the case of the other restrictions we have reviewed here, there are possible counterexamples. Aronoff (1976: 52−53) mentions the case of the suffix -ment. Nouns in -ment can generally only be made into adjectives by suffixing -al if -ment is itself attached to a bound base (ornamental, regimental). If -ment is attached to a free verb base, the adjective in -al is generally not possible (*discernmental, *containmental). If we were to state a restriction to this effect on suffixation of -al, we would need to refer to x in the configuration [[x]y], which is ruled out by the atom condition and the adjacency condition. Although it is true that the majority of theorists that contributed to the early period of generative morphology were based in North America, we should not forget that there were theorists in Europe, South Africa, and Asia that early adopted a generative stance, among them Booij (1977), Scalise (1984), Botha (1983), and Kageyama (1982).

4.2. Morphology and phonology At the same time that classical generative morphologists were interrogating the relationship between morphology and syntax, the relationship between morphology and phonology was also of great interest, giving rise to both the framework of lexical phonology and morphology and to prosodic morphology. Lexical phonology and morphology, developed in work by Siegel (1974, 1978), Allen (1978), Kiparsky (1982), Halle and Mohanan (1985), Mohanan (1986), among many others, begins from the premise that morphology and phonology are not distinct components, with word-formation strictly preceding the operation of phonological rules. Instead, rules of morphology and phonology are organized into a series of strictly ordered levels with rules of word-formation and phonological rules interspersed within each

101

102

I. Word-formation as a linguistic discipline level. Affixes are introduced by rules that specify the phonological form of the affix and the environment into which they are inserted. They do not themselves have lexical entries. In Kiparsky’s (1982) classic version of the theory, English morphology, for example, consists of three levels. Level 1 includes morphological rules involving what SPE called “non-neutral” derivational affixes, as well as rules for irregular inflectional forms. Level 2 contains rules that add # boundary affixes (what SPE calls “neutral” affixes) and rules that create compound words. Level 3 includes the rules for regular inflection. Each level is associated with its own set of phonological rules. These three levels are distinct from the syntax, which is subject to its own set of phonological rules referred to as postlexical rules. Rules that apply to an earlier level cannot apply to the output of a later level. It is this strict ordering of levels that accounts for the generalizations that non-neutral affixes cannot occur outside neutral ones, that phonological rules that apply to non-neutral affixes do not affect words formed with neutral affixes, that inflection is outside of both types of derivational affix, and so on. Kiparsky’s illustration of the model (1982: 5) is given in (8): (8) underived lexical entries

“+ boundary” inflection and derivation

stress, shortening

“# boundary” derivation and compounding

compound stress

“# boundary” inflection

laxing

Syntax

postlexical phonology

level 1

level 2

level 3

Lexical phonology and morphology was an enormously influential theory, but it was beset by problems from the start, prominent among them the need to introduce loops allowing words formed on later levels to be cycled back to earlier levels; for example, Giegerich (1999: 3) cites examples like systems analyst, where a regularly inflected form appears as the first element in a compound.

7. Word-formation in generative grammar

103

Prosodic morphology, also known as autosegmental morphology, was another theory that developed from generative roots. Originating in the work of McCarthy (1979), prosodic morphology essentially sought to apply the principles of autosegmental phonology to the treatment of non-concatenative processes of word-formation. Autosegmental phonology postulates that phonological segments are not unstructured lists of distinctive features, but rather are organized into a number of tiers that are themselves composed into hierarchically organized prosodic structures (syllables, feet, prosodic words). Prosodic morphology capitalizes on this hierarchical organization by postulating that morphemes can consist of segmental tiers or prosodic constituents, and not just of sequences of whole segments as is typical of generative morphology. This sort of organization makes it possible to treat non-concatenative word-formation processes more on a par with conventional affixation and compounding. McCarthy’s seminal work was on the root and pattern morphology of the Semitic languages, but the theory was soon applied to reduplication (Marantz 1982) and processes of vowel and consonant mutation (Lieber 1984, 1987). McCarthy’s now-classic analysis of the Arabic conjugation partitions each verb into three tiers. One tier consists of a pattern of organization of consonants and vowels that corresponds to what Hebrew grammarians call “binyanim”. A second tier contains the consonantal root, which carries the central lexical meaning of the complex word. Finally a third tier contains vowels that typically signal aspect and voice: (9)

k t b C

C V C V C a

‘write’ binyan 9 perfective active

A major result of the work of this period on prosodic morphology was a substantial deepening of our knowledge of the scope and limits of non-concatenative word-formation cross-linguistically.

5. Reactions to classic generative morphology By the mid 1980’s the time was ripe for reconsidering the assumptions of classic generative morphology. We begin with a consideration of the anti-lexicalist reaction and then consider a number of post-generative frameworks that can be said to have their roots in the generative movement.

5.1. The resyntacticization of morphology By the mid 1980’s morphology was clearly established in the generative framework as a legitimate area of study. This newly established legitimacy allowed theorists of the period to reconsider the organization of the grammar and revive the possibility that some or all of morphology could be treated as a matter of syntax; that is, one could claim morphology as a legitimate area of research interest without committing oneself to a separate

104

I. Word-formation as a linguistic discipline component of morphology in the grammar. The rejection of morphology as a separate component did not, however, mean a return to the sort of analysis proposed by Lees in the early days of generative grammar. By the 1980s the theoretical landscape of generative syntax was very different from Chomsky’s early theory. The move back to syntactic morphology took place in the context of the theory that has come to be known alternatively as government-binding theory or the principles and parameters model (Chomsky 1981). It is in the context of this theory that Fabb (1984), Sproat (1985), Roeper (1988), Baker (1988), and Lieber (1992) adduce several sorts of evidence to call even the weak form of the lexicalist hypothesis into question. By the early 1980’s the nature of argument structure had become a topic of great interest in generative syntax, giving rise within the principles and parameters model to case theory on the one hand and theta theory on the other. Since both derivational affixes and synthetic compounds exhibit argument structure effects – for example, the propensity for a derived word in -er to denote the external argument of its base verb or the interpretation of the first element of a synthetic compound as the internal argument of the second element’s base − it was natural that these types of word-formation come under special scrutiny and that attempts be made to put them in line with the syntactic theory of the time. Fabb (1984), for example, splits derivational affixes into two sorts, what he calls syntactic affixes and lexical affixes. The former are productive, regular, and subject to case theory. They generally (but not always) attach to verbs and either allow the projection of verbal arguments (for example, gerundive -ing) or affect the verbal projection in some regular way (for example, the suffix -er which is associated with the external theta role of the verb in the sense that derived -er words denote whatever the thematic role of the subject of a given verb would be). The lexical affixes are less productive and the forms derived with them are associated with idiosyncratic properties. What is problematic about Fabb’s proposal is that there are lexicalized forms derived from most affixes, which leads to the conclusion that most affixes must have both syntactic and lexical versions. Sproat’s (1985) proposal is much stronger than Fabb’s, abandoning the lexicalist hypothesis entirely. He uses the existence of so-called bracketing paradoxes like unhappier to argue for postulating two levels of analysis for derived words, a syntactic level and a phonological level; there is no separate morphological component. He argues that bracketing paradoxes can be resolved by assuming that on the syntactic level, -er is attached to the negative adjective unhappy, from which we can derive the appropriate semantic representation. On the phonological level, negative un- is attached to the comparative form happier; with the latter bracketing, we are able to maintain the phonological restriction that comparative -er attaches to words of no more than two syllables. The two levels are associated to each other via mapping principles. Baker (1985, 1988) adduces a different sort of evidence for the syntactic treatment of word-formation. He formulates what he calls the the mirror principle, which says that “[m]orphological derivations must directly reflect syntactic derivations (and vice versa)” (1985: 375). This principle allows him to account for facts of languages like Chichewa in which it is possible either to apply a process of applicativization to a passive sentence, or alternatively to passivize an applicative sentence. In the former case, the passive morpheme appears inside the applicative and in the latter the applicative appears inside the passive, the order of the morphology reflecting the ordering in which the operations are performed (example from Baker 1988: 14; abbreviations used in examples (10) and

7. Word-formation in generative grammar

105

(11) are as follows: SP = subject prefix, PASS = passive, APPL = applicative, ASP = aspect, 3M = third person masculine, 3N = third person neuter, DET = determiner, SUF = nominal inflection suffix): (10) a. Mpiringidzo un-na-perek-edw-a

kwa mtsikana ndi mbidzi

SP-PAST-hand-PASS-APPL-ASP to girl Crowbar ‘The crowbar was handed to the girl by the zebras.’

by zebras

b. Mtsikana a-na-perek-er-edw-a mpiringidzo ndi mbidzi SP-PAST-hand-APPL-PASS-ASP crowbar by zebras girl ‘The girl was handed the crowbar by the zebras.’ Baker (1988) also brings to the discussion the process of noun-incorporation in which it appears that a syntactic object can become part of a complex verb. Examples like (11) from Mohawk are prima facie persuasive, as they appear to illustrate that a noun stem can be extracted from an NP leaving behind its possessor (1988: 20): (11) a. Ka-rakv ne sawatis hrao-nuhs-aʔ 3N-be.white DET John 3M-house-SUF ‘John’s house is white.’ b. Hrao-nuhs-rakv ne sawatis 3M-house-be.white DET John ‘John’s house is white.’ Both the mirror principle and the syntactic analysis of noun incorporation have not been uncontroversial, however. Grimshaw (1986), for example, argues that the facts accounted for by the mirror principle can be better explained without assuming a syntactic analysis. Similarly, Rosen (1989) argues against a syntactic treatment of noun-incorporation. In the context of this discussion, synthetic compounds also were taken to suggest the syntactic treatment of morphology. Arguing on the basis of facts concerning control and aspect, Roeper (1988) proposes that compounds like rock-throwing in English should be analyzed as the output of a syntactic operation of head movement that starts out with a sentential structure. Roeper’s derivation of the compound rock-throwing is illustrated in (12). The N rock is incorporated into the verb and the -ing suffix is lowered onto the complex verb from an Infl node. As part of the derivation Roeper assumes that the IP undergoes a category-change to an NP, with a PRO subject under SPEC, thus explaining why in a sentence like John enjoyed rock-throwing for hours, the rock-thrower is typically interpreted as John. The Infl node in the underlying structure of the same sentence accounts for the progressive interpretation of the synthetic compound. Lieber (1992) continues in the effort to apply principles of syntax to word-formation. Lieber starts from the premise that the existence of phrasal compounding and derivation suggests that morphology and syntax cannot be separated into distinct components. She then attempts, on the grounds of simplicity, to develop a theory in which the principles of morphology and syntax are in fact the same. As part of this program she tries to derive the ordering of affixes from the normal settings of the X-bar parameters, an

106

I. Word-formation as a linguistic discipline (12)

IP SPEC

I Infl

VP

ing

N

V

NP

V

N

throw

rock

(S)

attempt that is in the end undermined by the difficulty of applying terms like complement and specifier to morphemes in complex words. All of these proposals have come under criticism and the relative merits of counting word-formation as a part of syntax or as a distinct component are still being debated (see, for example, Lieber and Scalise 2006). Di Sciullo (2005) is notable in this regard: although she seeks to apply the principles of the minimalist program to word-formation, Di Sciullo nevertheless continues to maintain that the mechanisms that apply to wordformation are distinct in key ways from syntactic principles.

5.2. The fragmentation of generative morphology Up to the early 1990’s it is possible to think of the theory of generative morphology as a single framework united not so much by specific principles or a set of formal mechanisms espoused by any specific theorist, but by the set of foundational assumptions mentioned at the outset. What has characterized the study of morphology in generative grammar over the last two decades has been an increasing specialization of theory. In this section we look at some of the outgrowths of generative morphology, giving most attention to those that have focused on word-formation as opposed to inflection.

5.2.1. A digression on inflection Although the subject of this article is word-formation in generative grammar and not inflection, it is difficult to continue a discussion of more recent developments without at least passing mention of frameworks whose primary focus is inflection.

7. Word-formation in generative grammar Anderson’s (1992) a-morphous morphology and Stump’s (2001) paradigm function morphology both have their roots in Matthews’ (1972) word-and-paradigm framework which begins from the observation that the relationship between meaning/morphosyntactic features and phonological representation is rarely one to one. When looking at complex inflectional systems, we often find on the one hand that a single morphosyntactic feature is expressed by more than one morpheme (so-called multiple exponence), and on the other that a single morpheme can express more than one morphosyntactic feature. Anderson (1992) takes this observation as a point of departure for a theory consisting of rules that associate bundles of features (sometimes with internal structure) with phonological forms; like Aronoff’s classic theory, Anderson’s framework is word-based, but his rules do not introduce inflectional features, as Aronovian-style rules would, but rather associate phonological form with sets of features. Stump’s (2001) paradigm function framework differs in formal details from Anderson’s, but is based on a similar premise. What is important for our purposes is that both Anderson’s and Stump’s frameworks are conspicuous in that they have almost nothing to say about word-formation (each contains only a brief chapter or part of a chapter devoted to the subject). Another highly influential framework that was proposed in the early 1990’s is distributed morphology (DM) (Halle and Marantz 1993; Harley and Noyer 1999). Like amorphous morphology and paradigm function morphology, the main emphasis in this framework has been on the analysis of inflection, with relatively little attention devoted to word-formation. DM is conceived broadly as a part of the larger minimalist program (Chomsky 1995). Morphology proper is part of the syntax. Nodes in a syntactic tree are associated with bundles of features, and bundles of features can undergo movement rules from one structural position to another. Only after the syntactic derivation is complete are bundles of features associated with phonological forms and subjected to certain sorts of readjustment rules, an arrangement that is referred to as “late insertion”; DM thus embraces Beard’s separation hypothesis. Roots – the lexical bases to which affixes attach – are inherently categoryless, and are associated with a syntactic category as part of the syntactic derivation; versions of the theory differ on whether roots as well as affixes are subject to late insertion. Attempts have been made within DM to extend the discussion at least to compounding (Harley 2009), but it remains to be seen how the theory might be adapted to account for derivational word-formation.

5.2.2. Lexeme morpheme base morphology Most of the syntactic models discussed in section 5.1 take for granted that the morpheme should be taken as the standard unit of analysis. Beard (1995), however, takes as a point of departure the same lack of a one to one correspondence between meaning and form that motivates Anderson’s and Stump’s theories and explores the consequences of this observation for derivation. The result is what he calls lexeme morpheme base morphology. One of Beard’s most important achievements is to bring the issue of semantics into the discussion of word-formation by observing that derivation, like inflection, often involves mismatches between form and meaning. Beard’s solution to these mismatches is to treat the semantics of derivation on a par with that of inflection. For Beard, wordformation consists in the addition of derivational features like [Agent] to a lexemic base. The agentive nouns corresponding to the verbs write and account would both be the

107

108

I. Word-formation as a linguistic discipline same at this stage of derivation. However, these two representations are spelled out with different phonological forms, resulting in the one case in the form writer and in the other the form accountant. The disjunction between the semantic representation of a word and its ultimate phonological form is what Beard calls the separation hypothesis. A major question that arises from Beard’s theory, however, is whether the semantics of derivation can really be accounted for with a featural system analogous to that needed for inflection.

5.2.3. The lexical semantics of word-formation Beard’s solution to the mismatch between morphology and semantics is essentially an architectural one: by designing a theory that separates the semantic part of word-formation from its phonological realization he is able to address the important issue of meaning-form mismatches. Lieber (2004) agrees that such mismatches must be accounted for, but argues that the success of any treatment depends on having an explicit theory of lexical semantics that accounts for derivation, compounding, and conversion, as well as the lexical semantics of free morphemes. Beginning from the well known observation that the affix -er not only forms agent nouns like writer, but also instrument nouns (printer), location nouns (diner), and even patient nouns (loaner), she proposes a framework of lexical semantic representation that allows her to explain why derivational affixes are so often wildly polysemous. Beard relies on inflectional categories like [agent], [means] (the feature Beard uses for instruments), or [location] to characterize the meanings of derivational morphemes, thus making it seem accidental that a suffix like -er can express all these meanings. Lieber’s lexical semantic representations consist of hierarchical arrangements of features and their arguments. Features cover broad aspects of lexical meaning, such as for example the characteristic of being concrete or abstract, which Lieber encodes in the feature [+/–material] and eventive or non-eventive, encoded in the feature [+/–dynamic]; features are deployed across syntactic categories, so that nouns may be characterized as eventive or non-eventive, just as verbs are. Both affixes and bases in her theory have lexical semantic representations, indeed the same sorts of representations. When these representations are combined, arguments of the affix may be linked to arguments of the base via what Lieber calls a principle of co-indexation: (13) a. write b. -er

[+dynamic [( ), ( )]] [+material, dynamic ([ ], )]

c. writer [+material, dynamic ([i ], [+dynamic([i ], [ ])])] The polysemy of an affix like -er falls out in her theory not so much from the architecture of the grammar but from the extent to which the semantic representation of the affix is underspecified.

6. Conclusion Active study of word-formation continues in the generative tradition. However, the emphasis in recent years has to some extent changed. For many years, generative morpholo-

7. Word-formation in generative grammar gy focused on a broad-brush characterization of the architecture of the theory and relied on formalism and principles modeled on the syntactic or phonological theory of the moment. In recent years we have seen a gradual movement towards a line of thinking that does not essentially piggyback on syntactic or phonological theory. One reason for this change in emphasis is the welcome convergence of theoretical approaches to morphology in the generative tradition with an independent thread of psycholinguistic study of word-formation that has gone on almost in parallel. Represented in the work of Taft (1988), Marslen-Wilson et al. (1994), Baayen, Dijkstra and Schreuder (1997), De Vaan, Schreuder and Baayen (2008), Kuperman, Bertram and Baayen (2010), among many others, this line of research has attempted on the basis of psycholinguistic experiments to model the extent to which complex words are stored as wholes in the mental lexicon as opposed to generated on line. This line of research as well as the increased attention to the issue of productivity in word-formation (for example, Plag 1999; Bauer 2001), has brought home to many current morphologists that words are different from sentences. Words persist in the mental lexicon in a way that sentences typically don’t. Further, the increasing availability of searchable corpora like the British National Corpus or the Corpus of Contemporary American English gives theorists today the means to track productivity and the semantics of complex words in context, and this will certainly lead to new insight into the organization of the mental lexicon.

7. References Allen, Margaret 1978 Morphological investigations. Ph.D. dissertation, University of Connecticut, Storrs. Anderson, Stephen 1982 Where’s morphology? Linguistic Inquiry 13(4): 571−612. Anderson, Stephen 1992 A-morphous Morphology. Cambridge: Cambridge University Press. Aronoff, Mark 1976 Word-formation in Generative Grammar. Cambridge, MA: MIT Press. Baayen, Harald, Ton Dijkstra and Robert Schreuder 1997 Singulars and plurals in Dutch: Evidence for a parallel dual-route model. Journal of Memory and Language 17(1): 94−117. Baker, Mark 1985 The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16(3): 373− 415. Baker, Mark 1988 Incorporation. Chicago: University of Chicago Press. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press. Beard, Robert 1995 Lexeme Morpheme Base Morphology. Albany: State University of New York Press. Booij, Geert 1977 Dutch Morphology. Lisse: Peter De Ridder Press. Botha, Rudolph 1983 Morphological Mechanisms. Oxford: Pergamon Press. Chomsky, Noam 1957 Syntactic Structures. The Hague: Mouton.

109

110

I. Word-formation as a linguistic discipline Chomsky, 1968 Chomsky, 1970

Noam Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Noam Remarks on nominalization. In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 184−221. Waltham, MA: Ginn. Chomsky, Noam 1981 Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam and Morris Halle 1968 The Sound Pattern of English. New York: Harper and Row. De Vaan, Laura, Robert Schreuder and Harald Baayen 2008 Regular morphologically complex neologisms leave detectable traces in the mental lexicon. The Mental Lexicon 2(1): 1−24. Di Sciullo, Anna Maria 2005 Asymmetry in Morphology. Cambridge, MA: MIT Press. Di Sciullo, Anna Maria and Edwin Williams 1987 On the Definition of Word. Cambridge, MA: MIT Press. Fabb, Nigel 1984 Syntactic affixation. Ph.D. dissertation, MIT Cambridge, MA. Giegerich, Heinz 1999 Lexical Strata in English. Cambridge: Cambridge University Press. Grimshaw, Jane 1986 A morphosyntactic explanation for the mirror principle. Linguistic Inquiry 17(4): 745− 749. Halle, Morris 1973 Prolegomena to a theory of word formation. Linguistic Inquiry 4(1): 3−16. Halle, Morris and K. P. Mohanan 1985 Segmental phonology of Modern English. Linguistic Inquiry 16(1): 57−116. Halle, Morris and Alec Marantz 1993 Distributed morphology and the pieces of inflection. In: Keneth Hale and Samuel Jay Keyser (eds.), The View from Building 20, 111−176. Cambridge, MA: MIT Press. Harley, Heidi 2009 Compounding in distributed morphology. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 129−144. Oxford: Oxford University Press. Harley, Heidi and Rolf Noyer 1999 Distributed morphology. Glot International 4(4): 3−9. Horn, Laurence 1989 A Natural History of Negation. Chicago: University of Chicago Press. Jackendoff, Ray 1975 Morphological and semantic regularities in the lexicon. Language 51: 639−671. Kageyama, Taro 1982 Word formation in Japanese. Lingua 57(2−4): 215−258. Kiparsky, Paul 1982 Lexical phonology and morphology. In: In-Seok Yang (ed.), Linguistics in the Morning Calm, 3−91. Seoul: Hanshin Publishing Company. Kuperman, Victor, Raymond Bertram and Harald Baayen 2010 Processing trade-offs in the reading of Dutch derived words. Journal of Memory and Language 62(2): 83−97. Lapointe, Steven 1980 A theory of grammatical agreement. Ph.D. dissertation, University of Massachusetts, Amherst.

7. Word-formation in generative grammar Lees, Robert 1960 The Grammar of English Nominalizations. Bloomington, IN: Indiana University Press. Levi, Judith 1978 The Syntax and Semantics of Complex Nominals. New York: Academic Press. Lieber, Rochelle 1980 On the organization of the lexicon. Ph.D. dissertation, MIT Cambridge, MA. [Published 1981 Bloomington, IN: Indiana University Linguistics Club and 1990 New York: Garland Publications.] Lieber, Rochelle 1984 Consonant gradation in Fula: An autosegmental approach. In: Mark Aronoff and Richard Oehrle (eds.), Language Sound Structure, 329−345. Cambridge, MA: MIT Press. Lieber, Rochelle 1987 An Integrated Theory of Autosegmental Processes. Albany: State University of New York Press. Lieber, Rochelle 1992 Deconstructing Morphology. Chicago: University of Chicago Press. Lieber, Rochelle 2004 Morphology and Lexical Semantics. Cambridge: Cambridge University Press. Lieber, Rochelle and Sergio Scalise 2006 The lexical integrity hypothesis in a new theoretical universe. Lingue e Linguaggio 5(1): 7−32. Marantz, Alec 1982 Re reduplication. Linguistic Inquiry 13(3): 435−482. Marslen-Wilson, William, Lorraine Tyler, Rachelle Waksler and Liane Olde 1994 Morphology and meaning in the English mental lexicon. Psychological Review 101(1): 3−33. Matthews, Peter H. 1972 Inflectional Morphology. Cambridge: Cambridge University Press. McCarthy, John 1979 Formal problems in Semitic phonology and morphology. Ph.D. dissertation, MIT Cambridge, MA. Mohanan, K.P. 1986 The Theory of Lexical Phonology. Dordrecht: Reidel. Newmeyer, Frederick 1986 Linguistic Theory in America. New York: Academic Press. Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin: Mouton de Gruyter. Roeper, Thomas 1988 Compound syntax and head movement. In: Geert Booij and Jaap Van Marle (eds.), Yearbook of Morphology 1988, 187−228. Dordrecht: Foris. Rosen, Sara 1989 Two types of noun incorporation: A lexical analysis. Language 65(2): 294−317. Scalise, Sergio 1984 Generative Morphology. Dordrecht: Foris. Selkirk, Elisabeth 1982 The Syntax of Words. Cambridge, MA: MIT Press. Siegel, Dorothy 1974 Topics in English morphology. Ph.D. dissertation, MIT Cambridge, MA. Siegel, Dorothy 1977 The adjacency condition and the theory of morphology. In: Mark J. Stein (ed.), Proceedings of the Eighth Annual Meeting of the North East Linguistic Society, 189−197. Amherst, MA: GSLA.

111

112

I. Word-formation as a linguistic discipline Sproat, Richard 1985 On deriving the lexicon. Ph.D. dissertation, MIT Cambridge, MA. Stump, Gregory 2001 Inflectional Morphology. Cambridge: Cambridge University Press. Taft, Marcus 1988 A morphological-decomposition model of lexical representation. Linguistics 26: 657− 67. Williams, Edwin 1981 On the notions ‘lexically related’ and ‘head of a word’. Linguistic Inquiry 12(2): 245− 274.

Rochelle Lieber, Durham, NH (USA)

8. Word-formation in categorial grammar 1. 2. 3. 4. 5.

Introduction Categorial grammar Derivation Compounding References

Abstract So far, morphology has rather been neglected in categorial grammar which was primarily developed for syntax. Categorial grammar presupposes that there are basically two types of linguistic categories, namely functors/operators and arguments/complements. This notation can be applied to morphology, covering both inflection and word-formation. Word-formation can be perceived of as an interface between lexicon and syntax, or as an integration of syntax into vocabulary. The descriptive power of categorial grammar becomes evident when applied to the field of word-formation and morphology in general; it manifests itself in discovering and handling the most detailed morphosyntactic and morphosemantic differentiations.

1. Introduction Up to now, morphology has been rather neglected in categorial grammar, which was primarily developed for syntax. This is true as far as inflection is concerned but also for word-formation as well. In fact, morphology and word-formation are as good as missing in the most comprehensive descriptive categorial grammar so far by Zifonun, Hoffmann and Strecker (1997, 3 volumes, approx. 2500 pages). Categorial Grammars by Mary

8. Word-formation in categorial grammar McGee Wood (1993) comprises a short overview in which several theoretical approaches are treated only briefly and in addition some problematic cases are sketched. Moreover, she points out that (obviously) both inflectional and derivational suffixes are considered by most authors as functors to be applied to lexical bases. Karl Reichl’s work Categorial Grammar and Word-Formation. The De-adjectival Abstract Noun in English (1982) is a more comprehensive and detailed monograph, quoted frequently as such. The main interest here lies in the linguistic material treated in detail but less in general theoretic questions. A rather theoretic focus is given in Jack Hoeksema’s Categorial Morphology (1985) in which categorial approaches are compared to others (cf. also Hoeksema and Janda 1988). This is why it does not offer either a comprehensive presentation of categorial grammar as a theoretical framework nor a discussion of the different patterns of word-formation within this framework. Inflection and word-formation is discussed more extensively in Ulrich Wandruszka (1976, 1997 and 2007). In these contributions, an effort is made, on the one hand, to consider in a formally consistent way the fundamental difference between syntax and morphology, and, on the other, to model the interfaces and transitions between morphology and syntax amongst others by means of the operation of functional composition. In this context, the author also tackles the much discussed phenomenon of government inheritance from a derivational base to the derived word in cases of the type German die Sehnsucht nach Liebe ‘the longing for love’ or die Empörung über den Preis ‘the indignation about the price’ and so on (cf. Moortgat 1988). Furthermore, the question as to whether forms like Korkenzieher ‘corkscrew; lit. cork puller’ should be considered as compounds or derivations is discussed, and how constructions in which the first part is an adjective or a numeral, as, e.g., Dickhäuter ‘pachyderm; lit. thick skin-er’ or Dreimaster ‘three-master’, Tausendfüßler ‘millipede; lit. thousand feet-er’, etc., should be analyzed. As far as inflection is concerned, constructions of the type ((Wein trinkende) N/N MännerN) N ‘wine drinking men’, in which a complex participial attribute is generated by means of functional composition, are treated in principle analogously to Korkenzieher. Also considered in this theoretical framework is the question of whether such bound morphemes are mentally stored as isolated units or always as parts of complex words or word forms.

2. Categorial grammar Categorial grammar presupposes that there are basically two types of linguistic categories, namely functors/operators and arguments/complements. Functors are so-to-speak the active elements, taking an active part in building up a complex expression. This activity arises from the valence of functors, resulting negatively from a blank, i.e. the lack of a particular structural partner and the necessity of closing or filling this blank. In this context, there are two kinds of functors: governing or head functors, operating as the governing head of a combination and dependent functors, constituting the adjunct or attribute of a combination. Together with a complement, head functors form a complex expression whose category they define and represent. Thus, a transitive verb such as drinks takes a noun phrase, such as a beer, forming with it the predicate drinks a beer, which now together with a subject NP like Brigitte forms the sentence Brigitte drinks a beer. Such functors are represented in categorial grammar by a fraction of the type Y/

113

114

I. Word-formation as a linguistic discipline X, i.e. an expression that takes an X and returns a Y. Our predicate drinks a beer thus represents the type S/NP and the verb form drinks the type (S/NP)/NP in that it forms together with an NP the expression S/NP, which then together with an NP forms a sentence. Such a verb form is thus bivalent, as opposed to the monovalent sleeps of the type S/NP, as in (the catNP sleepsS/NP)S. Dependent functors, on the other hand, attach to a head X, forming with it an (extended) expression of the category X. Since dependent functors are dependent attributes they cannot define or change the category of the resulting expression. Because they maintain the respective category, they belong to the formal type X/X. Thus, e.g., an adjective is a functor that combines with a noun and forms a (complex) noun, such as (littleN/N light N) N. Accordingly, an adverb is an expression that combines with a verb and forms an (extended) verb, cf. German ( falschV/V singen V) V ‘to sing out of tune’. The other large group of linguistic expressions embraces arguments or complements that by virtue of their role as complements to head functors do not have empty slots and thus are zero-valent, such as the NP a beer which, when it combines with a transitive verb, is transformed into a direct object as in (the phrase) to drink a beer. NPs can also be the complement of a preposition forming a prepositional phrase with an NP, as in French (dansPP/NP (la ville) NP) PP, ‘in the city’ which is of the type PP/NP. Complements of attributes on the other hand can be of any valence, since the latter do not have any influence on the category of the resulting expression. This is why, e.g., an adverb together with a monovalent verb can produce again the same type: (sleepsS/NP long(S/NP)/(S/ NP))S/NP. Now this notation can be applied to morphology, covering both inflection and wordformation which is of particular interest in the present context. It is generally suitable for all complex linguistic expressions. Thus, the German suffix -er in the noun Maler ‘painter’ is of the type N/St v, in that it forms together with a verb stem (St v) a noun: (Mal-​St v -er​N/St v ) N. Semantically speaking, the function of -er is to produce with the help of an action verb the denomination of the respective doer. In a form like German kirchlich ‘churchly’, the suffix -lich forms with a noun or a noun stem (St n) an adjective (Adj), thus being of the type Adj/St n: (kirch-​St n -lich​Adj/St n ) Adj. The suffix/flexive -en in turn belongs to the category V/St n , generating with a noun or noun stem a verb: (haus-​ St n -en​V/St n ) V ‘(to) live’. The French suffix -ier, as in poirier ‘pear-tree’ forms nouns with the help of noun stems, which can be analyzed as follows: (poir-​St n -ier​N/St n ) N. As we will discuss more in detail, defining derivational suffixes as functors, i.e. as instruments for forming new words, is quite suitable for depicting linguistic reality. The applicability of this model to inflectional morphology, the central grammatical part of morphosyntax, can be touched upon only briefly in the present context. Here again, in the realm of conjugation and declension, the focus lies on generating morphologically complex lexical units by means of bound morphemes. This is the case, e.g., in forms like German (sie) bell-t ‘(she) bark-s’ which result from a functional element, hence a functor, combining with a categorially as yet “unformatted” verb stem to form with it, e.g., a predicate (S/NP). The ending -t thus belongs to the type (S/NP)/St v generating by means of a bound morpheme a morphologically closed form of a certain syntactic category: (bell-​Stv -t​(S/NP)/Stv )S/NP. The fact that symbols of the form (S/NP)/Stv show a morphologically bound element in the denominator and a syntactic category in the numerator makes it clear that this is the domain of morphosyntax, the overlap between lexicon and syntax, connected by the respective inflectional ending. As for declension, one might consider the Latin genitive ending -is, which generates together with a

8. Word-formation in categorial grammar noun stem an attribute, as in (domusN (patr-​Stn -is​(N/N)/Stn )N/N) N ‘father’s house’, i.e. patris is a noun functioning as an attribute.

3. Derivation 3.1. Suffixation Word-formation as well can be perceived of as an interface between lexicon and syntax, or as the integration of syntax into vocabulary − a process also known as lexicalization or morphologization (cf. Morrill 2011). As opposed to inflection, which is about generating different forms of the same word in the sense of a lexeme, here the task is the formation of new words from bound morphemes and/or (free) words, thus extending the lexicon of a language with its own means and own material. The central morphologically bound functor in this field is, along with the prefix, the derivational suffix, a head functor which is different from an inflectional element in that it does not generate word forms but new words and thus − at least in inflectional languages − forms stems. To these products of word-formation, then, the several inflectional elements with syntactic functions and other grammatical features can be applied, as, e.g., in the deverbal Italian noun lava-tor-e/ lava-tor-i ‘washer/washers’ or the adjective lava-bil-e/lava-bil-i ‘washable’, and so on (cf. Werner 1990). Whereas in the realm of inflection the head status of the ending is precarious, in the case of derivation this designation can generally be taken unequivocally, since the suffix always represents the part of speech and also the type of meaning or denotation of the new word. Thus, suffixes like -tor(-e) and -bil(-e) are morphologically and semantically the head of their respective derivations: lava-tore denotes a person or an instrument and lava-bile a property, whereas the nouns lava-ggio/lava-tura ‘washing’ describe an activity, as does the verbal derivational base lav-are ‘(to) wash’. Moreover, even if the suffix does not change the part of speech of the derivational base, as in automobil-ista ‘motorist’ or automobil-ismo ‘motoring’, the morphosemantic dependency structure remains clear (formations like German Ismen ‘-isms’ exemplify that the suffix is in fact perceived of as a head constituent). Inflectional elements, on the other hand, form together with the stem of a given word class a special form of this very word, such as, e.g., parl-ate ‘you talk’, and thus belong to the type word x/stem x, or Wx/Stx, where W stands for a special function of the word. In the case of suffixation however there are generally two stems involved: St y /St x, i.e. stem y /stem x, which can belong to the same word class (St​ny /St​nx ), as in automobil-ist(a). In other words, inflection happens within the borders of a word, whereas suffixation as a type of word-formation generally transcends these borders, relating at least two different words as a matter of principle. The head status of suffixes is thus even more pronounced than that of inflectional elements, in that the former, representing the word class and the respective denotation type, defines the basis for word class specific determination and functional inflectional elements. This is true, e.g., in the above mentioned adjective ((lava-​St v -bil-​Sta​ / St v )​Sta -e ​ dj/St a )Adj, in which the adjective ending -e​Adj/St a together with an adjective stem forms A an adjective, or in a complex noun like Italian formalizzazione ‘formalization’:

115

116

I. Word-formation as a linguistic discipline (((( form-​St n -al-​Sta​ / St n )​St a -izza-​St v​ / Sta )​St v -zion-​Stn​ /St v)​St n -e​N/St n ) N N Stn Stv Sta

N/Stn Stn/Stv

Stv/Sta

Stn

Sta/Stn

form-

-al-

-izza-

-zion-

-e

On the left in the tree diagram, it is quite visible how adding a suffix to the nominal stem form- converts it into an adjective stem, then into a verb stem and finally into a noun stem again. Moreover, the difference between various stem forming suffixes and the word form building the inflectional element -e of the type Wx /St x also becomes apparent. As opposed to the left hand side of the tree, where the derivational bases, i.e. the complements of the suffixes ascend level by level, on the right hand side the suffixes themselves and the ending, i.e. the functors of the right-headed constructions are aligned. From a formal point of view, there is another peculiarity regarding derivation: although we are dealing with head functors and complements, similar to the case of attribution the resulting structure is iterative because applying further suffixes each time will produce new words or word stems. Since suffixes form more complex words or word stems with words or stems, the “products” along with the suffix always contain a word or rather a word stem as an immediate constituent. This is a form of recursiveness that does not occur in syntax, where, e.g., a simple sentence is never an immediate part of a complex sentence, but commonly of an adverbial, as in (before ( you came)S)Adv, or of a verb phrase, as in (hopeV [that] (Elvira comes)S) VP. However, as opposed to attributes, suffixes, as heads, are not optional structurally and thus cannot be simply deleted in a syntactic structure. This means, that on the level of lexical categories suffixes are head functors of the type X/X, whereas head functors in “free” syntax are of type Y/X. This is due to the fact that in syntax the common category word is missing and words instead are combined to a variety of syntactic constructions (syntagmas). Processes of word-formation are so-called relations in, i.e. relations within the set of words of a language. This is why the products of word-formation show the structure of an onion: deleting the head of a complex word again reveals a word, such as in German Nation-al-ität ‘nationality’, whereas Italian represents the generation of complex stems out of simpler ones: nazion-al-ità. Thus, word-formation simply entails producing a new word with a given word or a new stem with a given stem. In the case of category change, that is to say by applying inflectional elements of the type word x/stem x, or Wx/St x, i.e. inflectional elements forming together with a stem a morphologically closed word, no further iteration is possible, cf. (formalizzazion-​Stn -e​Wn​ / St n )​Wn. Likewise, in derivations

8. Word-formation in categorial grammar like German Kind-heit ‘childhood’, Kind ‘child’ does not represent the nominative singular (das Kind), but might be perceived of as a stem with a free slot for a suffix. This constancy of categories in word-formation is true only for the value word, but not for a special part of speech which can be changed by means of word-formation processes. In this domain, suffixes behave like normal head functors of the type Y/X, making with a word of category X one of category Y, allowing also the change of syntactic functions. For example, the German suffix -lich combines with nouns to form adjectives like freund-lich ‘friendly’, abend-lich ‘vespertine’ or väter-lich ‘fatherly’, thus belonging to the functional category adjective/noun, or adj/n, if morphological categories, i.e. parts of words, are represented by lower case letters. Hence, together with an expression of the syntactic category noun (N), a new type of the category nominal attribute is created, i.e. N/N, as in (väterlichesN/N HausN) N (see above for the analogous function of the genitive ending in Lat. ( patr-​Stn -is​(N/N)/St n )N/N). Conversely, the suffix -heit of the type n/adj combines with an adjective to form a noun as in (dieseNP/N Frechheit N) NP ‘this impertinence’. A suffix like -er, in turn, combines with verb stems to give nouns, hence N/St v, as in (der NP/N Fahrer N) NP ‘the driver’, whereas the suffix -ier(en) combines with a noun to produce a verb or a verbal stem, St v /n, as in lackier- ‘lacquer’, which together with an inflectional element becomes a verb: (lackier-​St v -en​V/Stv ) V ‘to lacquer’ or a specific verb form as in (Maria NP ((lackier-​St v -te​((S/NP)/NP)/St v)(S/NP)/NP das Auto NP) S/NP) S In this manner, the syntactic-functional implications of word-formation become quite evident. It is, by the way, an essential feature of natural language that the potential syntactic functions of a sign are predetermined by its part of speech to be used only in special structural contexts. This is why even in functionalistic categorial grammar not only the function but also the part of speech of a syntactic unit has to be named. In cases where the part of speech does not change, as in the aforementioned Automobil-ist ‘motorist’, the argument and the value of the functions have to be differentiated by other means in order to label the suffix formally as the head functor, as, e.g., (Automobil-​n x -ist​n y​ /n x ) N. It is essential that the linguistic expression here, as opposed to attribution (see below) is assigned a new head. Semantically, this suffix could be defined as follows: -ist is a functor that − amongst others − combines with the name of an instrument to form the term for the user of this instrument, for example, as in Maschinist ‘machinist’, especially with musical instruments as in Flötist ‘flutist’, Saxophonist ‘saxophonist’, Hornist ‘hornist’, etc. Categorial grammar is particularly suitable for representing the semantic structure and, in general, the semiotic status of bound morphemes such as suffixes and flexives. The loss of semiotic autonomy influences the meaning and the reference of such signs and, thus, their symbolicalness. The bound morphemes of a complex sign are both morphologically and semantically interdependent, which has interesting consequences for the relation between form and content. The meaning of functional morphemes must always be formulated with respect to the respective stem complement; it is impossible to assign them a meaning or denotation in isolation. For example, the German suffix -er which forms agent nouns, as in Fahrer ‘driver’ or Lehrer ‘teacher’, does not simply mean ‘(male) person’ or functionally speaking ‘doer’ or ‘agent’, but rather ‘(male) per-

117

118

I. Word-formation as a linguistic discipline son, carrying out the activity denoted by the stem’. This means that a suffix only takes on a denotation together with a stem and only together with a stem does it function as a sign. Hence, suffixes can be viewed as synsemantica. As opposed to the nominal head of the German compound Apfelbaum ‘apple tree’ the suffix in Fr. pomm-ier does not mean simply ‘tree’, but ‘tree carrying the fruit denoted by the stem’ based on the general morphosemantic pattern ‘object which has something to do with the referent of the derivational base’. Representing a suffix or an ending by means of a functor that together with a stem forms a morphologically closed term of a certain category avoids the need for assigning meaning to isolated morphemes like the suffix -ier or German -ung in Wohnung ‘flat’ or Italian -oso in sabbioso ‘sandy’. Semantically, these suffixes are functors that together with the meaning of a stem produce the complex meaning of a noun. The meaning of a functor is generally defined with respect to its function, just like its morphosyntactic category. In the compound Apfelbaum, however, the head constituent -baum does not have the same function as the suffix -ier. Rather, it is a complement of the attributive functor Apfel- of the type N/N (see section 4); hence, the “active” operator in the case of the compound is the attributive determinans and in the case of suffixational derivation the head, i.e. the determinatum. The discussion of a possible formal representation for the semantic structure of morphemes is not part of the present article.

3.2. Prefixation Whereas suffixes are normally unequivocal cases of synsemantica that only function as a sign in their morphological context, prefixes in comparison show a more complex behaviour. Above all, prefixes are not heads but attributes, in fact attributes of entire words, as morphological analysis makes clear. As opposed to suffixes and inflectional elements, prefixes do not combine with stems but attach to morphologically complete words (forms), as in German (ein-V/V ( fahr-​St v -en​V/St v) V) V ‘to drive in’. In other words, prefixations behave like compounds. Thus, in contradistinction to suffixes, prefixes are not morphologically necessary constituents of the whole expression. They do not represent the expression, but constitute, just like syntactic attributes, an omissible category that extends the expression and, hence is of the type X/X. This analysis implies that a prefix like ein- would have to be classified as an adverb, having the same function as the adverb hinein ‘into’ in (hineinV/V fahren V) V or even as schnell ‘fast’ in (schnellV/V fahrenV)V ‘to drive fast’. In fact, the prefix ein-, like most of these elements, originates from an adverbial attribute of this verb which, however, has developed into a quasi bound morpheme, since ein- in German is hardly ever used as an adverb. Furthermore, especially in combination with -fahren, ein- has undergone a clear semantic lexicalization. If, by means of a historical process, ein- has lost its status as a free morpheme, it can in principle no longer combine with a free morpheme or a word; ein- combines only with a bound stem and forms with it an extended stem. In this respect, it is a stem-forming expression like a suffix, with the difference that it does not appear as a head functor, but as a dependent functor of the stem, representing type St x /St x, leaving open the slot to the right of the stem to be filled with suffixes or inflectional elements.

8. Word-formation in categorial grammar

119

This is why the following morphological reanalysis takes place when the new stem einfahr- combines with a flexive like the infinitive ending: ((ein-​St v​ / St v fahr-​St v)​St v -en​V/St v) V. The attributive functor ein-​St v​ / St v remains optional morphologically , but is semantically bound, since the prefixless word form fahr-en is in fact morphologically correct. The structural partner of the inflectional ending however is the complex stem einfahr-. What comes to light here is partial loss of isomorphy between the morphosyntactic and the semantic structure. The fact that the prefix is an inherently optional morphological category goes well with the loose association of the prefix to the stem and the possibility of being separated from its verbal base in constructions like Seppi fährt sein neues Auto ein ‘Joey drives his new car in’. Furthermore, it remains true that on the morphological surface the structural partner of a prefix is a word, or rather that a prefixed word morphologically always comprises a prefix and a word, thus showing the typical attributive structure (prefix+word) word − and allowing the iterated structure (prefix (prefix+word) word) word , etc., as in the German verb (vor-prefix (an- prefix meldenword) word) word ‘to preregister’. This is also true for the much discussed so-called parasynthetic derivations whose prefix is lexically obligatory, like French em-bouteill-er ‘to bottle’, a-plat-ir ‘to flatten’, or German ein-tüt-en ‘to bag (sth.)’; they do not occur without a prefix: *bouteiller, *platir, *tüten. Examples like French deadjectival verbs a-grand-ir and grand-ir ‘to aggrandize’, Italian a-moll-are and moll-are ‘to macerate, soak’ or German ver-schmäler-n ‘to narrow’ and schmäler-n ‘to reduce, cut’ as opposed to ver-breiter-n/*breiter-n ‘to broaden, widen’ prove that these are accidental gaps. Such derivations thus do not consist of three parts (prefix−stem−suffix), as the term “parasynthetic” suggests, rather they have a morphologically regular binary structure of the type (prefix (stem−suffix) word) word . Prefixes at least have to be considered as stem-forming elements where they are part of the base of suffixal derivation, as in Italian deformalizza-zione ‘deformalization’, which is a nominalization of the verb deformalizzare ‘to deformalize’, and not a prefixation of the noun formalizzazione, since de- in general only combines with verbs. (((de-​St v​ / St v (( form-​St n -al-​St a​ / St n)​St a -izza-​St v​ / St a )​St v)​St v -zion-​St n/ ​St v)​St n -e​N /St n) N N Stn Stn/Stv

Stv Stv/Stv

Stv Sta

de-

N/Stn

Stv/Sta

Stn

Sta/Stn

form-

-al-

-izza-

-zion-

-e

120

I. Word-formation as a linguistic discipline Here again complements can be found on the left side of the tree and the functors on the right, with the exception of the attributive functor de-. Derivations, both suffixations and prefixations, are thus generally predetermining and right-headed. The tree diagram shows that prefixations can always be analyzed as prefix+word, de+formalizzazione, even if the prefix in the present case, as already mentioned, is not an immediate constituent of the whole derivation morphosemantically but a structural partner of the verb stem formalizza- or the verb formalizzare. Prefixes are, like suffixes, sensitive to parts of speech. For the most part, they only determine representatives of certain parts of speech. The Romance privative prefix in-/ im-, for instance, combines with adjectives, but − for semantic reasons − not with verbs. This is why the derivational base of Italian in-dipendente ‘independent’ is not a verb, hence not an element of the paradigm of dipendere ‘(to) depend’, but the adjective dipendente. Likewise, the German adjective abhängig [compare English dependent] has its privative counterpart unabhängig [independent], whereas the verbal form abhängend [depending] cannot form *unabhängend [*independing], since there is no verb *unabhängen. The form passend [compare fitting] on the other hand has mutated into an adjective and thus allows the derivation of unpassend [unfitting]. As opposed to attributes in free syntax, prefixes can change the valency of their base, as in the following case (in simplified notation): (dringt VP/PP (durchPP/NP das HolzNP) PP) VP → (durchdringt VP/NP das Holz NP) VP in phrases like Wasserdampf dringt durch das Holz → Wasserdampf durchdringt das Holz ‘Water vapour penetrates the wood’. This morphologization can be represented by a so-called functional composition in which the functor dringt VP/PP is applied directly to the immediately following functor durch PP/NP, resulting in the function value durchdringtVP/NP, which then together with a NP forms a verb phrase, formally: VP/PP · PP/ NP = VP/NP. Restructuring a sequence a) into a sequence b) as in: a) b)

(head functor1 − (head functor2 − complement) compl) → ((head functor1 + head functor2) funct − complement)

in which both head functors merge into one, which now makes the original complement of the headfunctor2 its own, can be analyzed and represented as functional composition. In this article, however, this important operation in the field of categorial grammar cannot be discussed in further detail (cf. Dowty 1988). Similar examples are, e.g., fährt um den Baum → umfährt den Baum ‘drives around the tree’ or fliegt über das Haus → überfliegt das Haus ‘flies over the house’. Such an impact of an attributive functor on its head is only possible because the result is not a syntactic structure, but a new word or a new stem with a special meaning which can include a change in syntactic valency.

4. Compounding On the background of the affinity between prefixation and compounding, the question arises of how to distinguish these two patterns of word-formation in our formal model.

8. Word-formation in categorial grammar

121

The first constituent of a German compound noun like Gartenzaun ‘garden fence’ is without doubt an attributive functor and, thus, could on first glance be assigned the same category as the adjective in kleiner Zaun ‘little fence’, namely N/N. However, this again would not make it clear that the value of the adjective functor is a nominal syntagma, whereas the value of the adnominal noun is a compound, hence a word. In order to clarify that we are dealing with innerlexical processes within the scope of a word, the symbol Wx (Word x) for morphologically closed forms of the part of speech x can be used especially for parts of compounds in analogy to the minuscules in St v for verbal or St n for nominal stems. For the sake of convenience, this meaning has to be assigned to the minuscule x itself, so that the determinans of our compound of the type Wn /Wn is categorized n/n. If the part of speech of this nominal attribute should be expressed as well, it has to be assigned the categorial symbol nn/n , representing a noun making a (complex) noun together with a noun − in general notation: xx/x . Thus, the compound has to be assigned the following morpholexical structure: (Garten​n n/n Zaun n) n ‘garden fence’, whereby all expressions of the category n in the sense of a subset are also elements of the category N. Moreover, nouns behaving as word forming functors can also combine with representatives of other parts of speech, in which case they then belong to the type yx/x . This way, we have in German messerscharf ‘razor-sharp’, aalglatt ‘slippery as an eel’ or (schnee​n adj/adj weißadj ) adj ‘snow-white’ a noun forming together with an adjective an adjective (as distinguished from the nominal compound (Schnee​n n/n Flocken)n ‘snowflake’ in which Schnee specifies a noun), and in (rücken​n v/v schwimmen v) v ‘to backstroke’ a noun combines with a verb to form a verb. The combination of nouns with adjectives with a specific inherent valency, as, e.g., arm/reich (an) ‘poor/rich (in)’ or frei/voll (von) ‘free/ full (from/of)’ in compounds like schneearm ‘snow-poor, i.e. snowless’ or bleifrei ‘lead free’ has to be distinguished from the pattern of schneeweiß. In these cases the adjective has to be defined as a suffix-like head functor taking a noun as a complement in order to form a complex adjective and is thus of the type ( y/x) y: (schneen arm​(adj/n) adj ) adj . In this case we put the specification of the function in the first position, followed by a subscript that specifies the part of speech, because an adjective like arm or frei is − in principle and not only in this context − a head functor of the type adj/n or in free syntax Adj/PP (compare also Lat. patris​N/N N as a nominal noun attribute). Whereas in these compounds the relation between noun and adjective is determined by the valency of the adjective, in the case of schneeweiß it is the nominal attribute in its function as an object of comparison which establishes the syntactico-semantic relation between the constituents. Semantically speaking, this is equivalent to the fact that schneeweiß describes a kind of weiß, while schneearm however is not a kind of arm nor is bleifrei a kind of frei. In schneearm, schnee- does not describe a property of arm, in schneeweiß however it does describe a property of weiß. Graphically, the two compound adjective types behave as follows: adj

vs.

adj

nadj/adj

adj

n

schnee-

-weiß

blut-

(adj/n)adj

-arm

122

I. Word-formation as a linguistic discipline Finally, the much discussed type French compte-gouttes, Italian conta-gocce or Spanish cuentagotas ‘dropper, drop counter’ based on a verb-object-relation, remains to be mentioned. The combination of a morphologically bound head functor with a word instead of another morphologically bound element is rather particular. Such a combination is common at best with attributes in the form of a prefix, as, e.g., in Fr. re-voir ‘to see again’, It. s-comodo ‘incovenient, uncomfortable’ or Germ. ent-lassen ‘to dismiss’. A similar morphological pattern can be found in German forms like Schlag-arm ‘punch arm’, Wett-büro ‘betting office’, Kauf-haus ‘department store; lit. buy house’ or Lachtaube ‘laughing dove’ and Prügel-knabe ‘whipping boy’, in which the verbal base functions as a predetermining attribute. In any case, all these constructions do not meet the classic definition of the compound as a complex word consisting of several words. In the case of Fr. compte-gouttes or compte-tours ‘revolution counter’, although parallel in their surface structure, the verb stem, if this constituent is to be interpreted as a stem, is a functor head governing the direct object gouttes just like garde- in gardemagasin ‘quartermaster’, garde-malade ‘home nurse’, etc. − and the structure of the word should be represented accordingly. Now the noun gouttes ‘drops’ (used without the article) at the level of word-formation can be interpreted as a noun phrase in the function of a direct object. We thus are dealing with a kind of morphological NP which according to our convention has to be represented by minuscules: -gouttesnp; the whole construction then is as follows: (compte-​St v -gouttesnp ) n . In this perspective, the interpretation as a stem is preferable to that of a finite verb form as, e.g., 3rd ps. sg., since the latter is not a typical element of word-formation but combines at the syntax level with an entire NP les/des gouttes. Especially in the field of word-formation and generally in morphology, the descriptive and analytical power of categorial grammar becomes evident in its ability to discover and to handle the most detailed morphosyntactic and morphosemantic differentiations (cf. Wandruszka 1999). n/np

5. References Dowty, David 1988 Type raising, functional composition, and non-constituent conjunction. In: Richard T. Oehrle, Emmon Bach and Deirdre Wheeler (eds.), Categorial Grammars and Natural Language Structures, 153−197. Dordrecht: Reidel. Hoeksema, Jack 1985 Categorial Morphology. New York, NY: Garland. Hoeksema, Jack and Richard D. Janda 1988 Implications of process-morphology for categorial grammar. In: Richard T. Oehrle, Emmon Bach and Deirdre Wheeler (eds.), Categorial Grammars and Natural Language Structures, 199−247. Dordrecht: Reidel. Moortgat, Michael 1988 Mixed-composition and discontinuous dependencies. In: Richard T. Oehrle, Emmon Bach and Deirdre Wheeler (eds.), Categorial Grammars and Natural Language Structures, 319−348. Dordrecht: Reidel. Morrill, Glyn V. 2011 Categorial Grammar. Logical Syntax, Semantics, and Processing. Oxford: Oxford University Press.

9. Word-formation in natural morphology

123

Reichl, Karl 1982 Categorial Grammar and Word-Formation. The De-adjectival Abstract Noun in English. Tübingen: Niemeyer. Wandruszka, Ulrich 1976 Probleme der neufranzösischen Wortbildung. Tübingen: Niemeyer. Wandruszka, Ulrich 1997 Syntax und Morphosyntax. Eine kategorialgrammatische Darstellung anhand romanischer und deutscher Fakten. Tübingen: Narr. Wandruszka, Ulrich 1999 Über den Nutzen formaler Modelle der natürlichen Sprache. Papiere zur Linguistik 60(1): 31−50. Wandruszka, Ulrich 2007 Grammatik: Form − Funktion − Darstellung. Tübingen: Narr. Werner, Heinz 1990 Kann die Montague-Grammatik einen Beitrag zur diachronischen Linguistik leisten? Folia Linguistica Historica 11(1−2): 147−182. Wood, Mary McGee 1993 Categorial Grammars. London/New York: Routledge. Zifonun, Gisela, Ludger Hoffmann and Bruno Strecker 1997 Grammatik der deutschen Sprache. 3 Vol. Berlin/New York: de Gruyter.

Ulrich Wandruszka, Klagenfurt (Austria)

9. Word-formation in natural morphology 1. 2. 3. 4. 5. 6. 7.

Introduction From natural phonology to natural morphology Functionalism and natural morphology Naturalness and the lexicon Parameters of naturalness in word-formation Concluding remarks References

Abstract Natural morphology is not a descriptive model of word structure but an explanatory framework designed to evaluate morphological techniques from a functionalist point of view along principles deduced from Peircean semiotics. Word-formation is seen as largely determined by extralinguistic factors such as communicative needs, cognitive exigencies and the conditions on speech performance and processing. Universal preferences are formulated that are reflected in cross-linguistic distribution, order of acquisition, diachronic change and other fields of evidence, but predictions are relativized by typological and language-specific criteria.

124

I. Word-formation as a linguistic discipline

1. Introduction Natural morphology has developed in the late 1970’s and early 1980’s from mainly two different sources. The first of these was a transposition of basic assumptions from natural phonology onto morphology, whence the epithet “natural”. The second is the functionalist approach established by the Prague school, with its concept of markedness, which can be seen as the converse of naturalness, although with the important difference that markedness is conceived as absolute and binary, while naturalness is seen as a relative and gradual quality by the proponents of naturalness theory. For the sake of characterizing the theoretical stance of natural morphology properly, paying regard to its initial motivation, it is necessary first to very briefly recall the basic tenets of natural phonology, inasmuch as they embrace a good deal of the idea of naturalness as a technical notion in linguistics (section 2). In a second step, the functionalist foundations of natural morphology will be sketched (section 3), leading to considerations on lexical enrichment as the primary function of word-formation (section 4). From this outset it will be possible to delineate the specific implementation of the principles of naturalness as applied to the phenomena of word-formation (section 5).

2. From natural phonology to natural morphology In natural phonology, as first conceived by Stampe (1969 and 1979 [1973]), the claim was made that phonology can be reduced to a set of universal processes reflecting the phonetic capacities of language users. In spite of their phonetic motivation, these processes are in themselves not physiological, but mental operations. Still, they are not to be confused with the so-called production rules in formal models of grammar and in artificial intelligence, nor with spellout rules generating target forms. Natural phonological processes are not generative, but are conceived as mental substitutions relating the intention of a speaker to the “actual rumble” of his speech (to use a Sapirean form of wording). The substitutions work both ways, relating not only the intention to the perceived output but also the incoming signal to its perception in terms of distinctive units. These processes are abstract in the sense that they operate at the level of representations specifiable in terms of distinctive features. An aspect of this particular notion of process that has always been emphasized by the proponents of natural phonology is their role as constraints in terms of production, as well as perception, of utterances: “These automatic processes not only govern automatic alternations or add allophonic features. Together with the prosody, they account for the native ‘accents’ of speakers, they constrain the perception and pronunciation of second languages, and they are the basis of systematic variation and change” (Donegan and Stampe 2009: 6; see also Donegan 2001). What makes phonological processes “natural” (in spite of them being called “phonological”) is that their motivation is entirely phonetic, which means that they are based on extralinguistic factors and settings (i.e. human nature, tout court). This makes them universal at the same time, on the premise that the anatomical and (neuro)physiological endowment of humans is roughly the same for all exemplars of the species, at least as far as the handling of what is usually encoded in terms of phonological features is concerned. Notice that for phonology (and for morphology, as will be seen) the theory

9. Word-formation in natural morphology of naturalness circumvents the innateness problem by outsourcing what is universal from the abstract system, i.e. from the grammar of a language, to the apparatus of its execution. No innate explicit system of abstract entities is assumed except the natural processes, but these are unordered at first, there being no such thing as universal grammar in a Chomskian spirit that would keep them in a definite configuration. All the structuredness and patterning making up the phonological system of a language is the result of language acquisition, triggered and directed by communicative experience. What makes phonological systems differ across languages is not a difference in the processes as such but a different pattern of selection from the universal pool of processes, with weighting mechanisms ranging from complete suppression over restriction to the subtleties of ordering of particular processes (for an overview of process application see Donegan and Stampe 1979: 145−158). Since the set of universal processes has never been explicitly defined by the proponents of natural phonology, nor have they ever supplied a technical apparatus that would make the theory applicable as a descriptive model, they can easily be criticized for trying to build a house from the roof downwards instead of having erected firm walls at first. It is hard to overlook, however, that much of the technical affordances of optimality theory could count as an operationalization of the principles of natural phonology, although the proponents of both approaches are far from lying in each other’s arms and the differences with respect to the foundation of constraints are considerable (for a comparison see Tobin 2009). Inasmuch as natural phonology emphasized those aspects of the “sound shape of language” (to use a Jakobsonian form of wording) that are ultimately rooted in phonetics, i.e. outside Saussure’s chessboard, the contrast to the more arbitrary properties of phonological systems was made very explicit in terms of the dichotomy of processes versus rules (cf. Dressler 1985b: 11−24; Donegan 2002). Rules, as opposed to natural processes, were characterized as conventional, language-specific, extrinsic operations that have to be specifically learned (Dressler 1984a: 30; cf. the similar distinction between automatic and nonautomatic rules made in elaborations of generative phonology, e.g., Kenstowicz and Kisseberth 1979). In short, the parameter of rule naturalness could have been assigned the labels “phonetically motivated” on the positive side and “morphological/lexical” on the negative side. Diachronic change might then make a process slide away from the “natural” end of the scale towards the “unnatural” domain, potentially ending up as a morpho(no)logical rule or an ossified alternation in the lexicon. The parallel to lexical phonology (and morphology, cf. Kiparsky 1983) with its layers of cyclic rules and postlexical component is very close in this respect, but, as in the case of optimality theory, the main difference lies in the foundation of the principles rather than in the principles themselves. From what has been briefly sketched in this section it could be expected that morphology should remain outside the scope of naturalness theory, being concerned with everything about the shape of words that was not in some way stimulated and at the same time constrained by the circumstances of linguistic performance. In fact, the founders of natural phonology had always concentrated on core issues in phonology and on the phonetics-phonology interface; they never considered extending their approach to other components of language (cf. Hurch and Rhodes 1996: XI). In the European version of natural phonology, however, there has been a strong tendency to account for interfaces and transition phenomena, in particular morphonology, and to apply the model to dia-

125

126

I. Word-formation as a linguistic discipline chronic change (cf. Dressler 1982). The interaction between phonology and morphology was thus something to be taken up in a more general approach, built on principles that should hold for all components of the language system. At this point the second line of thought joins in that stems from functionalist approaches to language structure (most notably Jakobson, but also Benveniste and Seiler, cf. Dressler et al. 1987: 3), according to which all material components of the language system are determined by performance factors, even though to a disputable degree, under the rule of the principle “form follows function”. Whether this principle is a “sound premise” or a “problematic” one in the sense of Friedrich (1975: 199−200) is still likely to remain a matter of dispute for some time to come.

3. Functionalism and natural morphology Performance factors, if that notion is taken in its widest sense, comprise all kinds of variables, from the neurophysiologically-based mechanisms of signal processing at the inward-directed parts of the communication network to the socially and situationally governed realm of pragmatics at its outward-directed branches. Naturalness theory can be seen as a theoretical approach that tries to unite these very different dimensions under a common heading, with “naturalness” as an evaluation metric for assigning benchmark values to the structural properties of languages according to their contribution to the overall functioning of the communication process. For such a theory, in order to be noncircular on the one hand and avoid over-determinism on the other, it is essential to be non-reductive and to allow for multiple interactions of particular forces. The subtitle of one of the earliest monographs in this tradition (given here in translation from the German original) is illustrative of the breadth of the approach, with its enumeration of phenomena to be accounted for by what was at this time still natural morphology in statu nascendi: “Reduplication, echo words, morphological naturalness, haplology, productivity, rule telescoping, paradigmatic levelling” (Mayerthaler 1977). While Mayerthaler joined the naturalness enterprise with a pronounced consideration for universal markedness theory and a tendency towards axiomaticity and formalization (cf. Mayerthaler 1981: 40−41, 60−64; in English: Mayerthaler 1988), Dressler had been developing what he at that time called a “polycentristic” theory of language, by which he meant “a lexicon and an item-and-process grammar which contains several semi-autonomous components” (Dressler 1977a: 13), one of them being word-formation. “For each corresponding component of language man is endowed with universal, ‘natural tendencies’” (Dressler 1977a: 13), from which preferences directing linguistic behaviour can be deduced. Before reviewing the tendencies affecting word-formation in the following sections, some further remarks on the general character of the “naturalist program” seem appropriate. Since natural morphology is an explicitly non-formalist framework (pace Mayerthaler’s aspirations), its claims and proposals can also be seen as a counterbalance with respect to the treatment of morphology in the generative tradition. After a period of relative neglect, word-structure had become an issue in generative theory in the form of the lexicalist hypothesis (Chomsky 1970; Halle 1973), and since then there have been constant endeavours to accommodate derivation and compounding in the various models

9. Word-formation in natural morphology of universal grammar over the last decades, more or less by mapping principles of syntax onto word structure, but in any case maintaining the fundamental assumptions that grammar is an autonomous mental technique for handling verbal signs and that the potential of this technique as well as the constraints on it are of a strictly formal nature that can be accounted for in terms of computational algorithms (see article 7 on wordformation in generative grammar). Natural morphology not only denies the autonomy of the language system as such but also conceives the system itself as epiphenomenal, being the result of a compromise between a multitude of partly antagonistic functions that cannot all be fulfilled at the same time (cf. Dressler 2002). Moreover, as mentioned already, these functions are assumed to be motivated or determined to a large extent by extra-linguistic factors, but the range of possible morphological phenomena is underdetermined by the entirety of these factors. For instance, the processing of utterances is subject to neurophysiological capacity, from which it follows that for the chunks to be processed in real-time verbal communication there are certain preferred states in terms of size and complexity. All other things being equal, a word form should be long enough as to be perceivable as a separate unit of speech distinguishable from others, and short enough as to allow for concatenation with other word forms, at least up to the phrase-level, within a single breath-group or intonational phrase (cf. Dressler 1985c: 53). Since in reality no other things are ever equal, though, variation within these margins is driven by factors that may challenge the limits set by one particular factor. For example, in the cases of multiple compounding and freezing, which usually result in constructions exceeding the preferred length of a word form (see article 24 on multi-word expressions), it is the function of conceptualization which overrides the processing-based preference for well-portioned chunks, while in the opposite case it is the function of phonetic parsimony which may reduce a word form to the size of a single phoneme or reduce a marker to zero, thus exceeding the limits of preferred size of a signans at the other extreme. Notice that both of these factors are non-morphological, yet morphology is strongly affected by them. By taking the multitude of communicative functions as its basis of reasoning, natural morphology is not a theory that can make strong predictions about the possible structure of languages but rather a theory that explains why the range of variation among language structures is not as perspicuous as the proponents of an autonomous innate universal grammar would like to see, and why unstoppable diachronic change keeps transforming these structures all the time and everywhere. The key aspect of natural morphology as compared to other approaches (except optimality theory, but with important reservations, see article 11 on word-formation in optimality theory) is that it aims at providing an evaluation scheme for morphological processes according to which asymmetries in their distribution can be accounted for. However, a significant restriction on the liability of predictions based on the principles of natural morphology consists in the overall architecture of the approach. Universal naturalness is the overarching layer, based on systemindependent and extra-linguistic factors that can be captured in terms of markedness. The preferences directed at this level can be thwarted, though, at the level of typological adequacy (cf. Dressler 1985c), where the intercomponential interactions of preferences are situated. Finally, at the level of individual languages, system-dependent naturalness may prevail (cf. Wurzel 1985), which is directed mainly by such factors as congruence in patterning, implicational relationships within paradigms and the overall frequency of

127

128

I. Word-formation as a linguistic discipline types and tokens (and is therefore more relevant to inflectional morphology than to word-formation, cf. Dressler 1987: 121).

4. Naturalness and the lexicon According to its explicitly functionalist approach in terms of a means-end model of language, natural morphology embarks on the analysis of word structure by first asking for the function of the phenomena under scrutiny. In the case of word-formation this is lexical enrichment in the first place, i.e. the supply of lexical items expressing either new content (e.g., in English, the coining of the compound dreadlocks upon the introduction of rasta hairstyle in popular culture) or differentiating existing content (e.g., the coining of deathhawk, denoting a particular type of mohawk hairstyle). It follows from the modularity of the language system that the basic function of expressing content can be fulfilled by more than one of its components. For example, instead of by means of a new word (in this case a compound), the content expressed by such items as dreadlocks or deathhawk could in principle as well be expressed by syntax in terms of circumscription with existing lexemes (as in a dictionary-type explication), or by the lexicon itself, without intervening morphology, through items borrowed from another language. However, both alternatives have serious shortcomings (longwindedness in the first case, possible unavailability in the second, above all), so that the preference for fulfilling the onomasiological function by the creation of a naming unit (a term taken up from Mathesius and established by Štekauer 1998; see article 6 on word-formation in onomasiology) with the help of morphology turns out to be the most natural solution. However, for a serious theoretical assessment of such a conclusion it would be necessary to explicitly specify its premises, i.e. to assure that what here has been sloppily referred to as longwindedness is indeed a property of linguistic expressions that can be gauged in terms of enhancement or impediment with respect to the success of a speech act in communication. The factors of frequency and entrenchment are of main importance here, and natural morphology is well disposed to link in with usage-based models (e.g., Bybee 2013) and quantitative approaches to complexity and economy across components (e.g., Fenk and Fenk-Oczlon 2008). As far as methodological rigour is concerned, it is important to mention that natural morphology does not have the tendency to exclude the occurrence of types of constructions that are considered unnatural or dysfunctional by virtue of its own principles. By postulating preferences instead of absolute universals it avoids the problem of immediate empirical fallibility that would arise once an instance of something that has been excluded on theoretical grounds turns up (cf. Dressler 1988). Thus natural morphology does not exclude the use of circumscription as an onomasiological technique, but just maintains that this is not likely to be the main solution for the task of naming concepts in a given language system (cf. Seiler’s 1975 dichotomy of “description” versus “labeling”). Examples of descriptive expressions can easily be found in English, also in the semantic sphere of hairstyles, e.g., high and tight or short back and sides, but this is not the prevailing pattern, apart from the fact that such descriptions are never exhaustive. The same holds for borrowing: it is obvious that borrowing occurs, especially in semantic fields like the one adduced here for illustration (consider, e.g., odango, from Japanese,

9. Word-formation in natural morphology or chignon, from French), but it is highly improbable that a language community will resort actively to borrowing as the main source for new expressions (there may be exceptions to this for technical terminology of languages for special purposes, though, see article 92 on foreign word-formation, language planning and purism and article 127 on word-formation and technical languages, and probably also for language death, cf. Dressler 1977b). The example of lexical enrichment in the semantic field of hairstyles may appear simplistic, but it has been chosen deliberately in order to make the typical line of reasoning in natural morphology as clear as possible. The argumentation is strictly deductive, departing from the definition of a goal in communication and then trying to evaluate the various possibilities of its attainment along the parameters of naturalness. Once a solution is assessed as more natural with respect to others, the reasoning can move on stepwise by further differentiating the spectrum of possibilities for attaining the immediate goal, once the alternatives to it have been discarded. So far it has become clear that circumscription does not enrich the lexicon (by definition, since it is syntactic; but notice that our examples above are not full phrases but rather a sort of complex labels, ready for becoming lexicalized), while borrowing does enrich the lexicon, but not in an optimal way, since it introduces unmotivated items that may require adaptation (and it overthrows the identity of a language when it prevails). For the creation of a naming unit that does not exceed the limits of wordhood and is not borrowed from another language, there are basically two possibilities: either a new formal expression is coined by means of morphology, as in the case of dreadlocks or deathhawk, or a new meaning is attached to an existing lexeme by means of metaphor or metonymy (see article 61 on word-formation and metonymy). Also for these solutions examples can be found in the coiffure-related vocabulary of English: ponytail is a metaphoric transposition on the basis of shape, while ivy league is a metonymic transposition based on the social group of people displaying a preference for the respective haircut (also eponymous hairstyles like pompadour or rachel belong to this type). As in the case of circumscription and borrowing, it can be said that metaphor and metonymy, although pervasive in the lexicons of all languages (and essential to cognition, cf. Danesi 1995), are not natural techniques for lexical enrichment, because they produce polysemy and thus violate the preference for biuniqueness based on the very general semiotic principle of “one meaning one form” (but see Marzo 2013: 60−73 for a refined account). Moreover, the notion “lexical” in the term lexical enrichment cannot be understood to refer to meaning only, otherwise it would be more adequate to speak of “semantic enrichment”. If a language would resort to nothing other than metaphor and metonymy for naming new content, it would become more and more cumbersome for speakers to encode meaning properly and recover intended meaning from form in real-time communication, although “language users seem to have much fewer problems with polysemy than many lexicographers and linguists” (Ungerer 1999: 309). An implicit principle underlying this line of argumentation in natural morphology is akin to Kant’s categorical imperative: for judging the relative naturalness of a linguistic operation, imagine what would happen if it were made to apply across the board. This monopoly principle, as it may be called, is also useful as an antidote against precipitate typological generalizations, as has been shown by Dressler (1985c: 53). As a matter of fact, the monopoly principle was already envisaged by Saussure (2011 [1916]: 133) in the chapter on the “mechanism of lan-

129

130

I. Word-formation as a linguistic discipline guage”, where it is stated that unrestricted arbitrariness “would lead to the worst sort of complication”. Semantic transposition, if applied across the board, would inflate the lexicon with polysemous items instead of enriching it. However, a moderate degree of application of this onomasiological technique seems to be common, and there are types of regular polysemy that can be found with considerable cross-linguistic frequency and uniformity, like the relation between designations for plants and their respective fruits, or between containers and their contents (Murphy 2010: 89−90). This kind of regularity seems to be more typical for metonymic transpositions than for metaphorical ones (Cruse 2006: 151), but the boundary between lexical meaning and pragmatic inference is fluid in such cases (cf. Nunberg 1979), so that nothing conclusive can be said about how much of the task of word-formation can be taken care of by semantics alone. The remaining candidates to be judged for their aptness to supply a language with items for fulfilling the most basic of its functions, viz., the referential function, are wordcreation, composition and derivation. The first of these falls outside the scope of wordformation, if this is defined as the creation of new words from existing ones. Still, leaving aside most obvious presuppositions, one could ask why lexical items are hardly ever created from scratch, like the English word blurb, which is reported to have been invented by the American humorist Gelett Burgess (cf. Bloomfield 1933: 424) and eventually gained general acceptance. It could be claimed that a new form for a new concept would represent a state of affairs that should not be considered dispreferred in terms of isomorphism and biuniqueness. However, this would lead a long way back, one step behind Saussure, who had called attention to the importance of lexical motivation (cf. the quotation above). Furthermore, new concepts are hardly ever new in all respects: there are relations of hyponymy, meronymy, etc., accommodating any new content in an existing network of semantic specifications. Unmotivated coinage can be at work, of course, in well-defined functional niches like the creation of brand names, but a closer look at such items usually reveals that they are hardly ever unmotivated in all respects (cf. Piller 1999; see article 26 on word-creation and article 124 on word-formation and brand names). Special standards hold, of course, for poetic language (cf. Dressler 1981, 1994). In spite of Marchand’s extensive use of the notion “motivation” (cf. Kastovsky 2005), this term has not gained much currency in English linguistic metalanguage. Definitions are lacking in Bauer’s (2004) and Cruse’s (2006) glossaries; also Haspelmath and Sims (2010) do not include it in their otherwise quite comprehensive list of definitions (but see Nänny and Fischer 2001; Radden and Panther 2004; Panther 2008 and Lehmann 2007 for recent appraisals; see also article 10 on word-formation in cognitive grammar and article 56 on motivation, compositionality, idiomatization). Haiman (1985: 2) uses the term in a different sense (“ways in which the linguistic form is a diagram of conceptual structure, and homologous with it”). In natural morphology, motivation has the prominent role of a main function of word-formation, viz., to create complex items whose meaning can be recovered from the meanings of their respective parts, according to the principle of compositionality (on which see Cruse 2011: 44; cf. also Klos 2011: 35−59). By establishing regular formal relationships between semantically related items, motivation also facilitates storage and retrieval in the declarative compartment of longterm memory (on the distinction of declarative versus procedural memory see Ullman 2004: 244−248). From this follows a general preference for transparency as opposed to

9. Word-formation in natural morphology opacity of complex word forms that is further differentiated in natural morphology in terms of semiotically-based naturalness parameters, which will be discussed in the following section.

5. Parameters of naturalness in word-formation Natural morphology assumes that each (sub-)component of the language system harbours system-specific tendencies that are not only potentially in conflict with the tendencies of other (sub-)components but also partly antagonistic within each (sub-)component itself. For the lexicon, a number of tendencies have been identified that cannot be all fulfilled at the same time (cf. Waugh and Newfield 1995): motivation, which warrants the semantic interpretability of complex words, stands against a tendency towards autonomy of complex words. This tendency, which manifests itself in idiomatization and lexicalization, can in turn not be fulfilled without detriment to another tendency, that of enhancing the solidarity between semantically related words in lexical fields. The multiple antagonisms inherent to the (sub-)components of the system is a main source of variation and makes language resemble a perpetuum mobile, with the impulse for diachronic change already built into the mechanism. This is in accordance with the conclusions to be drawn from the historical study of languages, which boil down to the insight that it is not change as such which has to be explained but rather its absence (cf. Shapiro 1991: 11−12). Since semiotics is at the basis of all evaluative considerations in natural morphology, the parameters or scales that have been set up in this approach are all directly or indirectly derived from semiotic principles, of which iconicity is the dominant and most general one. In the following subsections only those parameters will be discussed that are of immediate relevance to word-formation. In the interest of space, the use of examples will be kept to a minimum.

5.1. Iconicity Our look at the semantic field of hairstyles would be incomplete if we neglected formations that do not fit to any of the types of expression considered thus far. For example, perm or afro are clippings (of permanent wave and afro-american, respectively) and as such represent the process-type of subtractive derivation (see article 19 on clipping), as opposed to incremental derivation. Providing that no change in meaning is expressed by the clipping, as in the case of perm, this kind of process can be considered “extragrammatical”, as has in fact been proposed (Dressler and Merlini Barbaresi 1994: 36−41; cf. also Dressler 2000b and Dressler 2005: 269−270). On the other hand, clipped forms easily become lexicalized and acquire a meaning that differs from that of the full form. This seems to be the case for afro, which has a specific meaning as a hairstyle, while the full form may refer to anything being afro-american. Furthermore, afro is a noun, while afro-american is an adjective (potentially convertible to a noun), so the clipping here involves a change in word-class too, which is quite typical for grammatical but not for extragrammatical derivation. These examples show that formations that belong to the same type in one respect can behave very differently in other respects, which means that

131

132

I. Word-formation as a linguistic discipline the criteria for defining what is extragrammatical need to be refined beyond what has been brought forward thus far in the literature (cf. Ronneberger-Sibold 2010; for a summary of the problems and proposals see Mattiello 2013: 1−6). The fact remains, however, that clipping, as popular as it may be in the creation of all kinds of expressions enriching the lexicon, is hardly ever used regularly and productively as an exponent of a particular semantic category (cf. Dressler 1984b; 2000a). From the fact that lexical enrichment is the primary function of word-formation it follows that this function is best fulfilled if procedures are made available that can be applied productively and multiply, and it is fulfilled even better if the productive and multiple application of these procedures does not impair redundancy. Subtraction is definitely not a good candidate for such a job description, since it diminishes redundancy in direct proportion to the multiplicity of its application. If the monopoly principle mentioned in section 4 is applied to subtraction, the result is a language whose lexicon implodes instead of expanding. Only in view of the criterion of optimal word-size may subtractive derivation be rated as relatively natural (cf. Ronneberger-Sibold 1995), responding to the preference for bi- and trisyllabicity that seems to hold cross-linguistically: according to Dressler (1985c: 53), the optimal size of word forms, i.e. with bound morphemes already included (but excluding compounds), is about two to three syllables (for general considerations and a comparative survey of word length cf. Popescu et al. 2013). However, while subtraction is preferred from the point of view of economy of expression, at least as long as the need for redundancy in the lexicon is not compromised, other semiotically-based preferences stand against it, with the preference for constructional iconicity at the top.

5.1.1. Constructional iconicity In section 4, isomorphism was invoked as a most natural state of the relation between meaning and form, but without paying regard to the relations between the meanings of particular items. This missing aspect is brought in by the parameter of constructional iconicity (or diagrammaticity, in Peircean semiotic terms, cf. Johansen 1993: 98), which affords that an addition to the semantic specification of an item be accompanied by an addition of phonic substance. In the case of word-formation this leads to the apparent paradox that even diminutive meaning is most commonly expressed incrementally, i.e. by means of affixes (also termed “additive coding”, cf. Carstairs-McCarthy 1992: 219). This is because the isomorphy of meaning and form as conceived in the notion of constructional iconicity does not refer to the semantic content as such, which is to be dealt with under the Peircean rubric of “image” (see section 5.1.2), but to the abstract relation between what is basic and what is derived. All this rests, of course, on the prerequisite that meanings like ‘smallness’ or ‘lessness’ are cognitively represented in such a way as to be encoded in terms of a semantic category; otherwise the lexicon of a language would be full of unmotivated words like dwarf or puppy, where smallness is semantically implied but not encoded categorially the way it is in derivational diminutives like booklet, starlet or statelet (English is, however, exceptionally parsimonious with productive diminutive formations). Formations like diminutives and augmentatives have received much attention in natural morphology (cf. Dressler and Merlini Barbaresi 1994 for a

9. Word-formation in natural morphology thorough treatment), because they are related to a wider array of functional dimensions (especially pragmatics, see article 62 on the pragmatics of word-formation) than category-changing derivation, like the formation of deverbal action nouns or deadjectival quality nouns, so that the interplay and trade-off between a variety of partially conflicting preferences can be studied (cf. Zacarías 2006).

5.1.2. Imaging iconicity In the functional evaluation of word-formation processes as envisaged by natural morphology, the parameter of iconicity is compartmentalized according to Peircean semiotics into three types of iconic relations, diagrammaticity being agreed upon to be the most important for word-formation (cf. Dressler 2005: 269). A second type of relation is constituted by the most immediate kind of similarity between form and content, manifesting itself most obviously in onomatopoeia, but also, to a lesser degree, in phonaesthetic features of morphological markers, e.g., the preference for palatality in expressing diminutivity or attenuativity (cf. Dressler and Merlini Barbaresi 2001, 1994: 93; see also article 28 on affective palatalization in Basque). The parameter of imaging iconicity is of course only applicable to a highly restricted set of semantic features. It works well in the formation of hypocoristics from personal names and pet names, in forms of endearment and the like, whereas it is highly improbable for a category like agentivity or instrumentality to be ever expressed by means of phonic substance that bears a resemblance to its semantic content, since no speaker is likely to come up with an idea of what agentivity or instrumentality “sound like”. The imaging type of iconicity is also very limited in its applicability because of intervening preferences of a different kind. For example, the most iconic way of expressing duality would be full reduplication (see article 25 on reduplication). However, duality ranks far below plurality in terms of semantic relevance and generality, so reduplication is much more often used for expressing plurality than duality cross-linguistically (cf. Rubino 2005; Kouwenberg 2003). In a study of reduplication in Bikol (Central Philippine), Mattes (2006) further elevates the semantics of the reduplication process to a higher level of abstraction, arguing that “change of quantity” is the semantic value covering all the different senses of fully reduplicated forms in that language (cf. also Kouwenberg and LaCharité 2001). Furthermore, from the point of view of economy of expression, full reduplication should be dispreferred, since it produces forms that easily exceed the optimal size of word forms. Apart from the fact that the category of number is not a prototypical derivational category (cf. Dressler 1989: 9), this shows that the image-type, although the most iconic of all types of iconic signs, can only be employed as a supplementary device in word-formation and is not an optimal candidate for the task of lexical enrichment.

5.1.3. Metaphoric iconicity The third type of iconic relation between content and form is more abstract and at the same time more problematic than the other two. Under the assumption that derivation is semantically additive, in that it introduces an extra specification to the meaning of a

133

134

I. Word-formation as a linguistic discipline base form, affixation is iconic, mirroring the semantic increment through the attachment of phonic substance, while subtraction is clearly anti-iconic. This leaves the morphological process of modification, as in verbal believe versus nominal belief, somewhere in the middle of the scale of diagrammaticity, because a change from a voiced obstruent to a voiceless one is hardly interpretable as an addition of phonic substance (but rather as a loss of the feature [+ voiced], which might even suggest a classification of this type of modification as “mildly subtractive”). While it can be said that the modification of the sound shape of a lexeme, as in apophony and metaphony (also termed “modulatory coding”, cf. Carstairs-McCarthy 1992: 219), is still iconic to the degree of signaling a change in meaning (e.g., sing : song, full : fill), it is obvious that the way of signaling this change inflicts a loss of formal identity on the respective lexeme. A morphological operation that, in order to fulfill the task of producing semantically closely related items from one another, resorts to the very technique by which semantically unrelated lexemes are kept apart (e.g., string : strong, pull : pill) cannot be ranked high in terms of semiotic convenience.

5.1.4. Scales of iconicity The parameterization approach in natural morphology consists in setting up scales along which formal operations on lexical items are ordered in proportion to their relative naturalness with respect to the parameters set up according to the Peircean model of semiotics. The gamut of possible formal operations on lexical items comprises the following: they can be invented, substituted, combined, mixed, multiplied, enlarged, reduced or reshaped. In terms of word-formation this corresponds to coinage, suppletion, compounding, blending, reduplication, affixation, subtraction and modification, respectively. Signs can also be deleted, of course, but this is an operation of a different order and certainly not of interest to the study of processes serving the function of lexical enrichment. Even without recourse to any of the possible formal operations, the lexicon can still be enriched by means of an abstract operation. This is the case of conversion, as in verbal to host from nominal host (see article 17 on conversion). Since the absence of any change in the shape of a word cannot seriously be interpreted as the signaling of an addition or change of meaning, conversion should be ranked lowest on the parameter of diagrammaticity. Instead, it usually appears somewhere in the middle of the scales that have been proposed in the literature. For example, it has been argued by Crocco Galèas (1990) that conversion is a kind of metaphor, although at a first glance this is not in accordance with the Peircean definition of metaphor, which affords some sort of parallelism in representation between the content and the expression of the respective signs. However, in conversion this parallelism is only invisible when word forms are analyzed in isolation. Crocco Galèas proposes to extend the scope of formal properties of a word form to its (morpho)syntactic behaviour, which leads to the conclusion that the semantic difference between nominal and verbal host is paralleled by a difference in syntagmatic collocation (cf. Crocco Galèas 1998: 84−86). Consequently, in her scale of diagrammaticity (Crocco Galèas 1998: 32), conversion (under the name of “metaphoricity”) is placed below modification and above total suppletion, while in the equally explicit and

9. Word-formation in natural morphology even more detailed scale of diagrammaticity proposed by Kilani-Schoch (1988: 166− 117) conversion is ranked below strong suppletion, but together with total suppletion under the heading of “adiagrammaticité”. An ex-aequo solution for the arrangement of the scale with respect to this type of process is envisaged also by Dressler (2005: 269), who ranks modification as more diagrammatic than conversion, but does not exclude at the same time that both of them could occupy an equal rank on the scale, referring to the fact that conversion is very productive in English whereas modification is unproductive. If this means that universal scales of naturalness can be rearranged according to what is frequent and productive in particular languages, the often-repeated claim that the framework is deductive and that the preference degrees on the parameters, which are claimed to be universal, are based on extralinguistic factors (Dressler 2005: 268) loses much of its credibility (but cf. Dressler and Manova 2005 for a more deliberate account).

5.2. Problems with scaling It comes as no surprise that a theory that explicitly rejects formalism runs into problems as soon as it starts to pin down any of its components in a formal guise, even in as simple a form as a one-dimensional scale. Problems like the ranking of process-types are indicative of the fact that semiotics alone is perhaps not sufficient for the purpose of setting up generalizations concerning particular types of construction. The general impression is that the naturalness scales that can be found in the literature appear so plausible to the authors who contrive them that not enough care is taken in the arguments leading to the desired result. For example, it is not clear why the modification of the stem vowel in a derived agent noun like German Läufer ‘runner’ (from laufen ‘to run’) should downgrade the diagrammaticity of this derivational type. One could as well assume that diagrammaticity is even enhanced by the modification, since the addition of the agentive suffix -er could be seen as mirroring the “addition of meaning” (whatever that may look like in terms of cognition), while the modification of the diphthong in the stem would mirror the overall semantic change brought about by this addition. Instead, the modification is qualified as indexical (see section 5.3 below), but it is not made clear in what way the presence of an indexical element subtracts from the diagrammaticity of a construction. By founding its main principles on semiotics, which is basically sensible, natural morphology does not always succeed in keeping the adopted semiotic concepts apart and in specifying what counts as what in a given type of construction. This is a conundrum for which Anttila (1975: 10) has set the stage with the precept that linguistic signs “are not simple, but intricate blends of symbolic, indexical, and iconic elements” (emphasis in the original; cf. also Nöth 2008: 84). Furthermore, natural morphology relies heavily on semantics but has never made explicit what kind of semantics it is that underlies its assumptions. In the case of word-formation this is crucial, since derivational morphemes, unlike inflectional ones, are semantically complex and especially prone to interact with lexical meaning. Another problem consists in the fact (which is never stated explicitly by proponents of natural morphology), that the naturalness scales regard only a subset of the formal operations mentioned above (section 5.1.1): suppletion, affixation, subtraction, modification and conversion. For example, Crocco Galèas (1998: 32) ranks agglutinative affixa-

135

136

I. Word-formation as a linguistic discipline tion highest on her scale of diagrammaticity. However, if the definition of diagrammaticity rests on compositionality (“diagrammaticity entails a relation of biuniqueness between segmentability of signans and compositionality of signatum”, Crocco Galèas 1998: 30, italics in the original), one may ask why the top position on this scale is not assigned to compounding (cf. Stolz 1990, who asks the same question − and many more − in his review of Dressler et al. 1987). From the design of the naturalness scales it follows that compounding and derivation are implicitly kept apart as different compartments of wordformation, but this neglects the gradualness of the passage from one to the other, with intermediate phenomena like affixoids and combining forms. Reduplication, which would also deserve a place in the upper ranks of diagrammaticity, is likewise missing for no obvious reason. The same holds for the scale in KilaniSchoch (1988: 116−117), who assigns the highest rank of diagrammaticity to suffixation, with other kinds of affixes unaccounted for. Dressler (2005: 269−269), who discusses the diagrammaticity of different types of compounds but does not explicitly place them on a scale, also omits reduplication, while Thornton (2005: 164) places reduplication as the maximally iconic type of construction at the top of a scale of constructional iconicity (which is restricted to plural formations, however). As far as subtypes are concerned, Crocco Galèas (1998: 32) considers only one type of synaffixation, viz., the combination of an affix with modification of the base, while other variants of synaffixation, like circumfixation, are left out, and so forth. The main problem with naturalness parameters, for which divergent assignments like the one concerning conversion are only symptomatic, lies in fact that the extra-linguistic foundation of the tendencies or preferences postulated in natural morphology lacks a solid empirical basis. This is in the nature of things, of course. While natural phonology, dealing with meaningless form, could walk its way safely on the relatively firm ground of physiology and phonetics, natural morphology has to cope with meaningful units, which affords explicit reference to semantics, the extralinguistic basis of which is cognition. It seems that by relying directly on semiotics, the founders of natural morphology have tried to avoid a commitment to any explicit theory of semantics, which is understandable in view of the labyrinthic nature of this branch of linguistics and its extensions into similarly entangled fields like cognitive science and philosophy of language. However, that shortcut takes its toll when it comes to being concrete about relations between meaning and form. Only if the way in which meaning is represented is made explicit can judgements be made on how iconically it is expressed by form. Mayerthaler seems to have been aware of the problem and provided what he called “semantic markedness relations” (Mayerthaler 1987: 39−40; cf. also Mayerthaler 1980: 27−29), but these are rather conceived for inflection than for word-formation, and no further steps have followed this first one (cf. the criticism in Marzo 2008); cf. also Andersen’s (2008: 14) critical assessment: “N[atural]ness theory’s odd idea that all linguistic relations are scales makes it impossible for the linguist who applies this theory to attain descriptive adequacy in the face of the actual variety of, say, semantic relations.” There is nothing that could compensate for the lack of a semantic component in a theory dealing with the expression of meaning, nor for the lack of a theory of the lexicon in a theory dealing with lexical enrichment. The wealth of external evidence adduced by natural morphology in support of its claims is impressive, but cannot be used as an explanans for naturalness. Early acquisition, cross-linguistic frequency, diachronic stability and the like can only be explananda in a deductive framework. It is indicative for

9. Word-formation in natural morphology this dilemma that proponents of natural morphology often skip semantics and resort directly to cognition as a justification for their assertions, as if cognition were something obvious or at least well-defined. For example, Dressler (1987: 99), delineating the functions of word-formation, states that “the cognitive function is best served by labelling the concepts needed as precisely as possible”, without adding any hint on what evidence from cognitive science in support of this claim should look like (nor making explicit what it means for labeling to be “precise”). In spite of great advances in the recent past, very little is known as yet about the nature of cognition, so any assumption about what may be acting in its service remains intuitive.

5.3. Other parameters The parameter of diagrammaticity has been discussed in some detail here since the proponents of natural morphology consider it to be of central relevance to word-formation (Crocco Galèas 1998: 125; Dressler 2005: 268−269), and because it best reveals the basic merits and problems of this approach. A less perspicuous parameter derived from Peircean semiotics is indexicality. “The relation between sign and object is indexical when it is defined by a spatio-temporal (factual or existential) contiguity between them” (Shapiro 1983: 39). In natural morphology this is interpreted as a matter of distance of an operator from its operand, from which it follows that an affix is maximally indexical if it is adjacent to the lexeme it modifies. This is said to explain why inflectional morphemes, in that they refer to external constituents, occupy the peripheral positions in a complex word, while derivational morphemes are placed closer to the lexical base (Dressler 1987: 111). The fact that proponents of natural morphology are not always at ease with their own claims can be illustrated with the interpretation of indexicality in terms of word structure: While Dressler (1987: 111) has insisted that multiple modification, as in introflecting languages, constitutes the highest degree of indexicality (which is plausible, there being no way of being closer to each other than being interwoven), the same author later holds that “affixation is more natural than modification, because the indexical relation between an affix and its base is clearer” (Dressler 2005: 270). Carstairs-McCarthy (1992: 226), relying on the earlier conception, has applied the monopoly principle (without naming it so) to indexicality, arriving at the conclusion that “perfect indexicality would presumably require in every word form the total fusion of the lexical stem with the realisations of all its associated derivational processes and morphosyntactic properties, yielding inflectional paradigms consisting entirely of unsegmentable suppletive ‘stems’”. This provides a more reasonable solution to the problem than Dressler’s appeal to “clarity”, since “even with only the few inflectional categories of English, such a language would impose a totally implausible burden on the memory” (Carstairs-McCarthy 1992: 226). Thus, in spite of the fact that “all of morphology is indexical insofar a morphological marker refers to the base of the rule that introduces it” (Dressler 2005: 270), unlike other parameters, indexicality does not lend itself easily to operationalization, nor does it seem to be understood as a major factor by proponents of natural morphology. If indexicality were a driving force in the structuring of word forms, one should expect modification to be preferred for derivational morphology and dispreferred for inflection, in view of

137

138

I. Word-formation as a linguistic discipline the explanation for the relative order of derivational and inflectional morphemes with respect to their base mentioned above, but this does not seem to be borne out by the facts. Kilani-Schoch (1988: 126), who explicitly affirms that the contiguity between the grammatical morpheme and the lexical base is particularly strong in derivations like Arabic katab-a ‘he wrote’ → kātib ‘writer’, makes no claim in this direction. Crocco Galèas (1998: 22) is right in pointing out that “within natural morphology, there is no general agreement on the number of the universal parameters, nor is there a fixed list of semiotic principles (or factors) from which parameters can be deduced”. The present survey is not intended as a complete overview of all the proposals that have been made, repeated and modified in the literature. The parameters based on iconicity have been reviewed more thoroughly in the preceding subsections in order to expose the ambitious and at the same time tentative nature of the approach. Other important topics, like transparency (in terms of compositionality and segmentability, thus closely related to diagrammaticity and often treated under the same heading), biuniqueness and optimal shape of units have been mentioned only in passing. This imbalance is justifiable, however, in view of the fact that the universal preferences postulated in natural morphology hold for word structure as such and not for word-formation in particular. Therefore, and with regard to the explicitly functionalist design of the framework, it has been considered appropriate to highlight those aspects that are of most direct concern to the principal function of word-formation, viz., lexical enrichment.

6. Concluding remarks Natural morphology can be characterized as an ambitious research program designed to explain asymmetries in the occurrence and distribution of structural aspects of complex word forms. The epistemological design of the program is strictly deductive, without recourse to ad-hoc conventions and circular argumentation, which guarantees that predictions are empirically testable (cf. Dressler 1985a), according to the Popperian criterion of fallibility. It has stimulated research in many areas previously neglected and even opened up new sub-fields of morphology, like morphopragmatics (Dressler and Merlini Barbaresi 1994; see article 62 on the pragmatics of word-formation). As far as word-formation is concerned, an impressive amount of empirical research has been inspired by natural morphology over the last three decades, especially in the field of child language studies (cf. Korecky-Kröll 2011 for a recent account). The success of the framework in this respect is not paralleled in other domains, however. The predictive potential of naturalness theory concerning the directionality of morphological change has been overshadowed by the enormous amount of attention devoted to grammaticalization theory (cf. Giacalone Ramat 1995: 134), and the proponents of both sides have refrained from integrating their respective approaches (cf. also Dahl 2004: 115−117 for criticism in another vein). Such drawbacks notwithstanding, the idea of multidimensional scaling underlying the approach of natural morphology remains attractive as a model for a multifunctional phenomenon like word-formation, especially in view of recent increase of typological research: “In our view, the universal principles of Natural Morphology and its three sub-theories (universal preferences, typological adequacy and system-dependent naturalness), are at the heart of typological explanations” (Štekauer, Valera and Körtvély-

9. Word-formation in natural morphology essy 2012: 9). On the other hand, Bauer’s (2003: 265) request for “greater detail in some of the many areas which are currently rather obscure” has relinquished none of its topicality in the decade since it was made.

7. References Andersen, Henning 2008 Naturalness and markedness. In: Klaas Willems and Ludovic De Cuypere (eds.), Naturalness and Iconicity in Language, 101−119. Amsterdam/Philadelphia: Benjamins. Anttila, Raimo 1975 The Indexical Element in Morphology. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Bauer, Laurie 2003 Introducing Linguistic Morphology. 2nd ed. Edinburgh: Edinburgh University Press. Bauer, Laurie 2004 A Glossary of Morphology. Edinburgh: Edinburgh University Press. Bloomfield, Leonard 1933 Language. New York: Holt. Bybee, Joan L. 2013 Usage-based theory and exemplar representations of constructions. In: Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford Handbook of Construction Grammar, 49−69. Oxford: Oxford University Press. Carstairs-McCarthy, Andrew 1992 Current Morphology. London/New York: Routledge. Chomsky, Noam 1970 Remarks on nominalization. In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 184−221. Waltham, MA: Ginn. [Reprinted 1972 in: Noam Chomsky, Studies on Semantics in Generative Grammar, 11−61. The Hague: Mouton]. Crocco Galèas, Grazia 1990 Conversion as morphological metaphor. In: Julián Méndez Dosuna and Carmen Pensado (eds.), Naturalists at Krems. Papers from the Workshop on Natural Phonology and Natural Morphology (Krems, 1−7 July 1988), 23−32. Salamanca: Ediciones Universidad de Salamanca. Crocco Galèas, Grazia 1998 The Parameters of Natural Morphology. Padova: Unipress. Cruse, Alan D. 2006 A Glossary of Semantics and Pragmatics. Edinburgh: Edinburgh University Press. Cruse, Alan D. 2011 Meaning in Language. An Introduction to Semantics and Pragmatics. 3rd ed. Oxford/ New York: Oxford University Press. Dahl, Östen 2004 The Growth and Maintenance of Linguistic Complexity. Amsterdam/Philadelphia: Benjamins. Danesi, Marcel 1995 The iconicity of metaphor. In: Marge E. Landsberg (ed.), Syntactic Iconicity and Linguistic Freezes. The Human Dimension, 265−284. Berlin/New York: Mouton de Gruyter. Donegan, Patricia 2001 Constraints and processes in phonological perception. In: Katarzyna Dziubalska-Kołaczyk (ed.), Constraints and Preferences, 43−68. Berlin/New York: Mouton de Gruyter.

139

140

I. Word-formation as a linguistic discipline Donegan, Patricia 2002 Phonological processes and phonetic rules. In: Katarzyna Dziubalska-Kołaczyk and Jarosław Weckwerth (eds.), Future Challenges for Natural Linguistics, 57−81. München: LINCOM Europa. Donegan, Patricia and David Stampe 1979 The study of natural phonology. In: Daniel A. Dinnsen (ed.), Current Approaches to Phonological Theory, 126−173. Bloomington: Indiana University Press. Donegan, Patricia and David Stampe 2009 Hypotheses of natural phonology. Poznań Studies in Contemporary Linguistics 45(1): 1−31. Dressler, Wolfgang U. 1977a Elements of a polycentristic theory of word formation. Wiener Linguistische Gazette 15: 13−32. Dressler, Wolfgang U. 1977b Wortbildung bei Sprachverfall. In: Herbert Brekle and Dieter Kastovsky (eds.), Perspektiven der Wortbildungsforschung. Beiträge zum Wuppertaler Wortbildungskolloquium vom 9.−10. Juli 1976 anläßlich des 70. Geburtstags von Hans Marchand am 1. Oktober 1977, 62−69. Bonn: Bouvier. Dressler, Wolfgang U. 1981 General principles of poetic license in word formation. In: Harald Weydt (ed.), Logos semantikos. Studia linguistica in honorem Eugenio Coseriu; 1921−1981. Vol. 2: Sprachtheorie und Sprachphilosophie, 423−431. Berlin: de Gruyter. Dressler, Wolfgang U. 1982 A semiotic model of diachronic process phonology. In: Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 93−131. Amsterdam/Philadelphia: Benjamins. Dressler, Wolfgang U. 1984a Explaining natural phonology. Phonology Yearbook 1: 29−51. Dressler, Wolfgang U. 1984b Subtraction in word formation and its place within a theory of natural morphology. Quaderni di Semantica 5(1): 78−85. Dressler, Wolfgang U. 1985a On the predictiveness of natural morphology. Journal of Linguistics 21: 321−337. Dressler, Wolfgang U. 1985b Morphonology. The Dynamics of Derivation. Ann Arbor: Karoma Press. Dressler, Wolfgang U. 1985c Typological aspects of natural morphology. Acta Linguistica Academiae Scientiarum Hungaricae 35: 51−70. Dressler, Wolfgang U. 1987 Word formation as part of natural morphology. In: Wolfgang U. Dressler, Willi Mayerthaler, Oswald Panagl and Wolfgang U. Wurzel (eds.), Leitmotifs in Natural Morphology, 99−126. Amsterdam/Philadelphia: Benjamins. Dressler, Wolfgang U. 1988 Preferences vs. strict universals in morphology: Word-based rules. In: Michael Hammond and Michael Noonan (eds.), Theoretical Morphology. Approaches in Modern Linguistics, 143−154. San Diego: Academic Press. Dressler, Wolfgang U. 1989 Prototypical differences between inflection and derivation. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42(1): 3−10. Dressler, Wolfgang U. 1994 Word-formation: Poetic license. In: Robert E. Asher and J. M. Y. Simpson (eds.), The Encyclopedia of Language and Linguistics, 5028−5029. Oxford: Pergamon Press.

9. Word-formation in natural morphology Dressler, Wolfgang U. 2000a Subtraction. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 581−587. Berlin/New York: de Gruyter. Dressler, Wolfgang U. 2000b Extragrammatical vs. marginal morphology. In: Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal Morphology, 1−10. München: LINCOM Europa. Dressler, Wolfgang U. 2002 Naturalness and functionalism. In: Katarzyna Dziubalska-Kołaczyk and Jarosław Weckwerth (eds.), Future Challenges for Natural Linguistics, 83−102. München: LINCOM Europa. Dressler, Wolfgang U. 2005 Word-formation in natural morphology. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 267−284. Dordrecht: Springer. Dressler, Wolfgang U. and Stela Manova 2005 The morphological technique of conversion in the inflecting-fusional type. In: Laurie Bauer and Salvador Valera (eds.), Approaches to Conversion/Zero-Derivation, 67−101. Münster: Waxmann. Dressler, Wolfgang U., Willi Mayerthaler, Oswald Panagl and Wolfgang U. Wurzel 1987 Introduction. In: Wolfgang U. Dressler, Willi Mayerthaler, Oswald Panagl and Wolfgang U. Wurzel (eds.), Leitmotifs in Natural Morphology, 3−22. Amsterdam/Philadelphia: Benjamins. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 1994 Morphopragmatics. Diminutives and Intensifiers in Italian, German and other Languages. Berlin/New York: Mouton de Gruyter. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 2001 Morphopragmatics of diminutives and augmentatives: On the priority of pragmatics over semantics. In: István Kenesei and Robert Harnish (eds.), Perspectives on Semantics, Pragmatics, and Discourse. A Festschrift for Ferenc Kiefer, 43−58. Amsterdam/Philadelphia: Benjamins. Fenk, August and Gertraud Fenk-Oczlon 2008 Complexity trade-offs between the subsystems of language. In: Matti Miestamo, Kaius Sinnemäki and Fred Karlsson (eds.), Language Complexity: Typology, Contact, Change, 43−65. Amsterdam/Philadelphia: Benjamins. Friedrich, Paul 1975 The lexical symbol and its relative non-arbitrariness. In: Dale M. Kinkade, Kenneth L. Hale and Otmar Werner (eds.), Linguistics and Anthropology. In Honor of C. F. Voegelin, 199−247. Lisse: Peter de Ridder Press. Giacalone Ramat, Anna 1995 Iconicity in grammaticalization processes. In: Raffaele Simone (ed.), Iconicity in Language, 119−139. Amsterdam/Philadelphia: Benjamins. Haiman, John 1985 Introduction. In: John Haiman (ed.), Iconicity in Syntax, 1−7. Amsterdam/Philadelphia: Benjamins. Halle, Morris 1973 Prolegomena to a theory of word formation. Linguistic Inquiry 4(1): 3−16. Haspelmath, Martin and Andrea D. Sims 2010 Understanding Morphology. London: Hodder Education. Hiraga, Masako K. 1994 Diagrams and metaphors: Iconic aspects in language. Journal of Pragmatics 22: 5−21.

141

142

I. Word-formation as a linguistic discipline Hurch, Bernhard and Richard Rhodes 1996 Introduction. In: Bernhard Hurch and Richard Rhodes (eds.), Natural Phonology. The State of the Art, VII−XII. Berlin/New York: Mouton de Gruyter. Johansen, Jørgen Dines 1993 Dialogic Semiosis. An Essay on Sign and Meaning. Bloomington: Indiana University Press. Kastovsky, Dieter 2005 Hans Marchand and the Marchandeans. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 99−124. Dordrecht: Springer. Kenstowicz, Michael J. and Charles W. Kisseberth 1979 Generative Phonology. Description and Theory. New York: Academic Press. Kilani-Schoch, Marianne 1988 Introduction à la morphologie naturelle. Berne: Lang. Kiparsky, Paul 1983 Word-formation and the lexicon. In: Frances A. Ingemann (ed.), Proceedings of the 1982 Mid-America Linguistics Conference, 3−29. Lawrence (Kansas): Department of Linguistics, University of Kansas. Klos, Verena 2011 Komposition und Kompositionalität. Möglichkeiten und Grenzen der semantischen Dekodierung von Substantivkomposita. Berlin/New York: de Gruyter. Korecky-Kröll, Katharina 2011 Der Erwerb der Nominalmorphologie bei zwei Wiener Kindern: Eine Untersuchung im Rahmen der Natürlichkeitstheorie. Ph.D. dissertation, University of Vienna. Kouwenberg, Silvia 2003 Twice as Meaningful. Reduplication in Pidgins, Creoles and other Contact Languages. London: Battlebridge Publications. Kouwenberg, Silvia and Darlene LaCharité 2001 The iconic interpretations of reduplication: Issues in the study of reduplication in Caribbean creole languages. European Journal of English Studies 5(1): 59−80. Lehmann, Christian 2007 Motivation in language: Attempt at a systematization. In: Peter Gallmann, Christian Lehmann and Rosemarie Lühr (eds.), Sprachliche Variation. Zur Interdependenz von Inhalt und Ausdruck, 100−135. Tübingen: Narr. Marzo, Daniela 2008 What is iconic about polysemy? A contribution to research on diagrammatic transparency. In: Klaas Willems and Ludovic De Cuypere (eds.), Naturalness and Iconicity in Language, 101−119. Amsterdam/Philadelphia: Benjamins. Marzo, Daniela 2013 Polysemie als Verfahren lexikalischer Motivation. Theorie und Empirie am Beispiel von Metonymie und Metapher im Französischen und Italienischen. Tübingen: Narr Francke Attempto. Mattes, Veronika 2006 One form − opposite meanings? Diminutive and augmentative interpretation of full reduplication in Bikol. In: Tenth International Conference on Austronesian Linguistics, 17−20 January, Palawan, Philippines. Linguistic Society of the Philippines / SIL International. Available online at: http://www.sil.org/asia/philippines/ical/papers.htm [last access 2 Feb 2013]. Mattiello, Elisa 2013 Extra-grammatical Morphology in English. Abbreviations, blends, reduplicatives, and related phenomena. Berlin/New York: Mouton de Gruyter.

9. Word-formation in natural morphology Mayerthaler, Willi 1977 Studien zur theoretischen und zur französischen Morphologie. Reduplikation, Echowörter, morphologische Natürlichkeit, Haplologie, Produktivität, Regeltelescoping, paradigmatischer Ausgleich. Tübingen: Niemeyer. Mayerthaler, Willi 1980 Ikonismus in der Morphologie. Zeitschrift für Semiotik 2: 19−37. Mayerthaler, Willi 1981 Morphologische Natürlichkeit. Wiesbaden: Athenaion. Mayerthaler, Willi 1987 System-independent morphological naturalness. In: Wolfgang U. Dressler, Willi Mayerthaler, Oswald Panagl and Wolfgang U. Wurzel (eds.), Leitmotifs in Natural Morphology, 25−58. Amsterdam/Philadelphia: Benjamins. Mayerthaler, Willi 1988 Morphological Naturalness. With a preface by Rupert Riedl. Ann Arbor: Karoma Press. Murphy, M. Lynne 2010 Lexical Meaning. Cambridge: Cambridge University Press. Nänny, Max and Olga Fischer 2001 Introduction: Veni vidi vici. In: Olga Fischer and Max Nänny (eds.), The Motivated Sign, 1−14. Amsterdam/Philadelphia: Benjamins. Nöth, Winfried 2008 Semiotic foundations of natural linguistics and diagrammatic iconicity. In: Klaas Willems and Ludovic De Cuypere (eds.), Naturalness and Iconicity in Language, 73−100. Amsterdam/Philadelphia: Benjamins. Nunberg, Geoffrey 1979 The non-uniqueness of semantic solutions: Polysemy. Linguistics and Philosophy 3: 143−184. Panther, Klaus-Uwe 2008 Conceptual and pragmatic motivation as an explanatory concept in linguistics. Journal of Foreign Languages 31(5): 2−19. Piller, Ingrid 1999 Iconicity in brand names. In: Max Nänny and Olga Fischer (eds.), Form Miming Meaning, 325−341. Amsterdam/Philadelphia: Benjamins. Popescu, Ioan-Iovitz, Sven Naumann, Emmerich Kelih, Andrij Rovenchak, Haruko Sanada, Anja Overbeck, Reginald Smith, Radek Čech, Panchanan Mohanty, Andrew Wilson and Gabriel Altmann 2013 Word length: Aspects and languages. In: Reinhard Köhler and Gabriel Altmann (eds.), Issues in Quantitative Linguistics 3. Dedicated to Karl-Heinz Best on the occasion of his 70 th birthday, 224−281. Lüdenscheid: RAM Verlag. Radden, Günter and Klaus-Uwe Panther 2004 Introduction: Reflections on motivation. In: Günter Radden and Klaus-Uwe Panther (eds.), Studies in Linguistic Motivation, 1−46. Berlin/New York: Mouton de Gruyter. Ronneberger-Sibold, Elke 1995 Different ways of optimizing the sound shape of words. In: Henning Andersen (ed.), Historical Linguistics 1993. Selected Papers from the 11th International Conference on Historical Linguistics, Los Angeles, 16−20 August 1993, 421−432. Amsterdam/Philadelphia: Benjamins. Ronneberger-Sibold, Elke 2010 Word creation: Definition − function − typology. In: Franz Rainer, Wolfgang U. Dressler, Dieter Kastovsky and Hans Christian Luschützky (eds.), Variation and Change in Morphology. Selected Papers from the 13th International Morphology Meeting, Vienna, February 2008, 201−216. Amsterdam/Philadelphia: Benjamins.

143

144

I. Word-formation as a linguistic discipline Rubino, Carl 2005 Reduplication: Form, function and distribution. In: Bernhard Hurch (ed.), Studies on Reduplication, 11−29. Berlin/New York: Mouton de Gruyter. Saussure, Ferdinand de 2011 [1916] Course in General Linguistics. Translated by Wade Baskin. Edited by Perry Meisel and Haun Saussy. New York/Chichester: Columbia University Press. Seiler, Hansjakob 1975 Die Prinzipien der deskriptiven und der etikettierenden Benennung. In: Hansjakob Seiler (ed.), Linguistic workshop III. Arbeiten des Kölner Universalienprojekts 1974, 2−57. Mu¨nchen: Fink. Shapiro, Michael D. 1983 The Sense of Grammar. Language as Semeiotic. Bloomington: Indiana University Press. Shapiro, Michael D. 1991 The Sense of Change. Language as History. Bloomington/ Indianapolis: Indiana University Press. Stampe, David 1969 The acquisition of phonetic representation. In: Anthony Bruck, Robert A. Fox and Michael W. La Galy (eds.), Papers from the Parasession on Natural Phonology, 443−453. Chicago: Chicago Linguistic Society. Stampe, David 1979 [1973] A Dissertation on Natural Phonology. New York: Garland. Štekauer, Pavol 1998 An Onomasiological Theory of English Word-Formation. Amsterdam/Philadelphia: Benjamins. Štekauer, Pavol, Salvador Valera and Lívia Körtvélyessy 2012 Word-Formation in the World’s Languages. A Typological Survey. Cambridge: Cambridge University Press. Stolz, Thomas 1990 Review of Dressler, Mayerthaler, Panagl and Wurzel (1987). Linguistics 28(1): 162− 167. Thornton, Anna M. 2005 Morfologia. Roma: Carocci. Tobin, Yishai 2009 Comparing and contrasting natural phonology, optimality theory and the theory of phonology as human behavior. Poznań Studies in Contemporary Linguistics 45(1): 169− 189. Ullman, Michael T. 2004 Contributions of memory circuits to language: The declarative/procedural model. Cognition 92: 231−270. Ungerer, Friedrich 1999 Diagrammatic iconicity in word-formation. In: Max Nänny and Olga Fischer (eds.), Form Miming Meaning, 307−324. Amsterdam/Philadelphia: Benjamins. Waugh, Linda and Madeleine Newfield 1995 Iconicity in the lexicon and its relevance for a theory of morphology. In: Marge E. Landsberg (ed.), Syntactic Iconicity and Linguistic Freezes. The Human Dimension, 189−221. Berlin/New York: Mouton de Gruyter. Wurzel, Wolfgang Ullrich 1985 On morphological naturalness. Nordic Journal of Linguistics 7: 165−183. Zacarías, Ramón 2006 Formación de deminutivos con el sufijo /-ít-/: Una propuesta desde la morfología natural. Anuario de Letras 44: 77−103.

Hans Christian Luschützky, Vienna (Austria)

10. Word-formation in cognitive grammar

10. Word-formation in cognitive grammar 1. 2. 3. 4. 5. 6. 7.

Basic principles of cognitive grammar Word structure: analyzability Entrenchment and drift; countering analyzability Characterizing the parts Composition Conclusion References

Abstract This article summarizes some of the basic assumptions of cognitive grammar with an eye on their relevance to the study of word-formation. Issues addressed include analyzability, entrenchment, and productivity; the different processes of word-formation (affixation, compounding, and blending); the rationale for recognizing words as distinctive units of linguistic structure; and the similarities and differences between word structure and phrase structure.

1. Basic principles of cognitive grammar Cognitive grammar is a theory of language which Ronald Langacker began working on in the late 1970s. The standard exposition of the theory is the two-volume Foundations of Cognitive Grammar (Langacker 1987, 1992); an updated account is Langacker (2008). Taylor (2002) is a text-book introduction. (For the position of cognitive grammar within a more broadly characterized cognitive linguistics movement, see Taylor 2002, Ch. 1.) Each of these publications has extended discussions of words and their structure. Additional resources − both those which explicitly adopt Langacker’s theory and those which are broadly compatible with its goals − are cited in this article. The starting point of cognitive grammar is uncontroversial: a language is a means for relating sound and meaning. More specifically, a language enables speakers to represent their thoughts and intentions by making available to them an inventory of symbolic associations between units of form (phonological structures) and units of meaning (semantic structures). Hearers familiar with the symbolic associations are able to recover, or to intimate, the speaker’s semantic intentions. A notable feature of the theory, in contrast with competing accounts, is its minimalist approach to the form-meaning relation. There are, namely, only three objects of study in cognitive grammar. The first comprises language in its perceptible form, typically as sound, but also as writing, or (in the case of signed languages) as gesture. Using the term rather broadly, this aspect of language is referred to as phonological structure. The second is the symbolized content, referred to as semantic structure. Semantic structure comprises a person’s conceptualizations, in a very general understanding of the term; it includes not only referential intent and propositional content, but also affect, evaluations,

145

146

I. Word-formation as a linguistic discipline and attitudes towards possible hearers. The third object of study are symbolic associations between phonological and semantic structures. The latter are taken, following Saussure (1964 [1916]), to be conventional in character; their properties and indeed their very existence need to be specifically learned by members of a speech community. Cognitive grammar proposes a direct relation between phonological and semantic structure. In many alternative theories the relation, on the contrary, is mediated by intervening levels of organization. These are autonomous in the sense that they are composed of elements and relations which are unique to these levels and which cannot be reduced to matters of phonology or semantics. One such intervening level (or, in some theories, cluster of intervening levels) is the syntax. Cognitive grammar does not deny the existence of syntactic phenomena. The claim, rather, is that syntactic elements − notions of lexical and phrasal categories (noun, noun phrase, etc.); categories such as complement and adjunct; syntactic relations (clausal subject, phrasal head, etc.), and even the very notions of word and phrase − can themselves be fully described in terms of the minimalist ontology of the theory. An aim of this article is to indicate the validity of this approach to the description of words and their internal structure.

1.1. A structured inventory of units In cognitive grammar, knowledge of a language is equated with knowledge of an inventory of units, a unit being defined as any phonological, semantic, or symbolic structure that has been established, or entrenched, in the speaker’s mind through frequency of previous use. Importantly, the units do not constitute a random, unordered list. Rather, the inventory is structured by virtue of the ways in which the units are related to each other. Three kinds of relation are of special importance. These are the part-whole relation, the schema-instance relation, and the relation of similarity (or, in the limiting case, identity). a) The part-whole relation. This is the relation whereby a linguistic structure may be analysed into its component parts. The part-whole relation applies to phonological, semantic, and symbolic structures alike. Thus, the phonological form [sɪŋ] breaks down into its component sound units, namely, the sounds [s], [ɪ], and [ŋ], while the symbolic unit [singer] can be analysed into its constituent symbolic units [sing] and [-er]. The counterpart to analysis is composition. This is the process whereby speakers recruit already established units in order to create structures which are not already entrenched in their mental grammar. b) The schema-instance relation. A linguistic structure − whether phonological, semantic, or symbolic − may count as an instance of a more abstractly characterized unit. The phonological structure [sɪŋ] may be seen as an instance of a schematic phonological unit [Syllable], which in turn can be analysed into its schematically characterized parts [Onset] and [Rhyme]. While speakers encounter only instances, the sanctioning schemas are immanent in the instances to the extent that the instances may be recognized as matching the specifications of the schema. These two relations are recursive. [A] may be a part of [B], which in turn is a part of [C]. Likewise, [X] may be an instance of [Y], which in turn is an instance of [Z]. The

10. Word-formation in cognitive grammar two relations interact in numerous ways. For example, the very possibility of analysing a structure into its parts is constrained by the existence of a schema which sanctions the analysis; at the same time, the analysis strengthens the sanctioning schema. The interplay of the two relations is also evident in the process of composition. The possibility of creatively combining already established units rests not only on the availability of the combining units themselves, but also on the entrenchment of a schema which sanctions the combination. c) The similarity relation. Although difficult to define formally, similarity − or more accurately, speakers’ subjective perception of similarity − plays an important role in the emergence of schemas. The fact that [A] and [B] come to be regarded as instances of [C] rests on the prior recognition that [A] and [B] are similar in some respect(s). Moreover, the similarity between [A] and [B] may be perceived to be similar to the way in which [D] and [E] are similar. The commonality between the two cases may give rise to a higher order schema (Nesset 2008). For example, the perceived similarity between word pairs such as urbane/urbanity, insane/insanity, profane/profanity may give rise to a schema representing the alternating phonological forms.

1.2. The autonomy of phonological and semantic structures In view of the role of language as a means for giving overt expression to a speaker’s conceptualizations, it is only to be expected that the study of symbolic relations will constitute the heart of any cognitive grammar investigation. However, the theory also allows for the possibility that phonological and semantic structures exhibit a degree of autonomy vis-à-vis symbolic relations, in the sense that components of these structures need not participate in symbolic associations. Thus, with respect to phonological structure, units such as vowels and consonants, syllables and their parts, do not of themselves have symbolic value. Hickory-dickory-dock, of the nursery rhyme, is perfectly wellformed phonologically, in that the expression is composed of established units and is sanctioned by well-entrenched phonological schemas; the expression, however, corresponds to nothing at all at the semantic level. Similarly, there are many elements of semantic structure which are not overtly symbolized by any phonological material. This is very obviously the case with noun-noun compounds (Benczes 2007). There is nothing in the phonological structures of the expressions to indicate that while a cheese shop sells cheese, a barber shop does not sell barbers (rather, it sells the service which barbers provide) (Jackendoff 2010: 446).

1.3. A usage-based grammar Cognitive grammar is a strictly bottom-up, or usage-based model of linguistic knowledge and its acquisition (Tomasello 2003). Abstractions (as encapsulated in schemas) are abducted on the basis of acquaintance with their instances. Repeated exposure leads to the progressive entrenchment of instances and of the sanctioning schemas. Entrenchment is, to be sure, a matter of degree; its extent also varies from language user to language user.

147

148

I. Word-formation as a linguistic discipline It is therefore taken as rather evident that different speakers of the “same” language may have acquired rather different mental representations. The usage-basis of cognitive grammar entails that a speaker’s mental grammar may contain a good deal of redundancy. Expressions which are fully consistent with a sanctioning schema and which could, in principle, be created compositionally, may nevertheless have unit status, provided that they have been entrenched through previous linguistic experience. For example, high frequency plurals may well be stored in the mental grammar alongside the singular forms.

1.4. Motivation Symbolic relations are conventional, in the sense that they have to be specifically learned (on the basis of experience). This wording is noncommittal with respect to the Saussurian notions of arbitrariness and motivation. There is, to be sure, much in a language which is arbitrary, the fact that the phonological form [sɪŋ] is associated with the semantic concept ‘sing’ being but one example. Equally, a very great deal in a language is motivated. Motivation is often taken to refer to the role of semantic structure, or conceptualization more generally, in the shaping of phonological and symbolic structures (Panther and Radden 2011). While not discounting this aspect, the proposal that linguistic knowledge constitutes a structured inventory of units offers an alternative perspective, where each unit stands at the hub of a network of relations (part-whole, schema-instance, and similarity) to other units which are already entrenched (to a greater or lesser degree) in the mental grammar (Taylor 2004). Even a phonological form such as [sɪŋ] − whose association with the concept ‘sing’ may indeed be arbitrary − is motivated by the fact that it conforms with a schema for syllable structure and is made up of units which are well entrenched elsewhere in the language. Language-internal motivation undoubtedly facilitates the learning, storage, and accessing of linguistic units, and testifies, again, to the redundancy inherent in the language system.

2. Word structure: analyzability Knowledge of a language consists of a network of units, structured by the part-whole, schema-instance, and similarity relations. The focus in this section is on the part-whole relation, with specific reference to words. On what basis it is possible to identify a part of a word? Observe that the question, as posed, refers to words, that is, to symbolic associations of form and meaning. The fact that [s] is part of the phonological pole of the symbolic unit [sing] is not strictly speaking relevant to word structure. In speaking of word parts we are primarily interested in parts which themselves have symbolic status, whereby a part of the phonological structure of a word can be associated with a semantic value which contributes to the word’s semantic structure. In order for a word-part to be recognized, a number of conditions need to be satisfied, at least in the canonical cases (some non-canonical cases are discussed below). First, we need to be able to establish an identity relation (or, at least, a relation of similarity)

10. Word-formation in cognitive grammar between a piece of a word’s phonology and (a piece of ) the phonology of at least one other symbolic unit. Second, we need to be able to associate the phonological segments with an identical (or similar) semantics. Third, the postulated semantic value of the word-part needs to feature in the semantic structure of the word itself. Grounds for identifying the word-part will be strengthened in proportion to the number of symbolic units exhibiting the postulated word-part; in the limiting case, the correspondence may be perceived to be accidental or fortuitous, or indeed not noticed at all. Below, I discuss a range of examples, selected so as to illustrate some of the issues which arise when considering the internal structure of words.

2.1. farmer The analyzability of this word into its parts [farm] and [-er] is rather evident. It is instructive to reflect on why this intuition should be so clear. Essentially, we accept this analysis because it fits in with so many other facts about the English language; the analysis, in other words, is very strongly motivated by language-internal relations. First, farmer is obviously related to the well-entrenched verb form farm, both phonologically (the phonological structure of the verb is fully incorporated into the derived form) and semantically (the derived word invokes the same kind of activity as the base verb). Second, there exists a large number of words for which a comparable analysis is indicated: singer, walker, driver, and countless more. These words characterize a person in terms of what they do, the activity in question being supplied by the verbal element while the agent notion can plausibly be attributed to the suffix. The existence of this cohort of examples, with comparable semantics, enables the abstraction of a schema which captures their commonality, viz. the agentive noun schema [ NV-er]. The schema is supported by its compatibility with a very general pattern of word-formation in English, namely suffixation, whereby the suffix determines the kind of entity that a word refers to while the word-specific semantic content is supplied by the stem. In this sense, the agentive noun schema is itself an instance of a higher order schema for derived words.

2.2. philosopher and barber In the canonical case, a word can be exhaustively divided up into constituent symbolic units. The “building block” metaphor (Langacker 1992: 186) does not always apply. Given the high degree of entrenchment of the agentive noun schema (this being a function of the large number of examples on which it is based), we are likely to detect the [-er] suffix in words such as philosopher and grocer, and perhaps even in barber. These words correspond only partially with the agentive noun schema. While the words undoubtedly characterize a person in terms of what they do, English lacks the putative base verbs (to) philosoph [fɪ'lɒsəf ], (to) groce, and (to) barb. In the case of philosopher we can appeal to the phonological similarity (and even orthographic identity) with the underlined portions of philosophy and philosophical. Not only this, but the phonological and semantic relatedness of philosophy and philosopher matches the phonological and se-

149

150

I. Word-formation as a linguistic discipline mantic relatedness of other word pairs, such as geography and geographer, biography and biographer. For barber, the situation is less clear, not only because there is no verb (to) barb, but also because there are no semantically related words containing the phonological sequence [bɑːb], or phonologically similar sequences.

2.3. Thursday The case of barber is not unusual. We often encounter words, a part of which can be identified as a symbolic unit, the remainder of the word, however, lacks this status. Take, as a simple case, the names of the days of the week. These all terminate in [-day], a unit whose semantic contribution to the names is evident. Yet no meaning at all can be attributed to [thurs-], except in the trivial sense that [thurs-] contrasts with [wednes-] and the dangling parts of the other weekday names.

2.4. perform Sometimes the case for analysis may be quite compelling, even in the absence of semantic relatedness. Take the case of (mostly) bisyllabic words of usually abstract meaning. Initial segments include ad-, con-, in-, per-, pro-, trans-; final segments include -ceive, -fer, -form, -late, -tain. To be sure, not all combinations are possible, though a good many are. The semantic contribution of the constituent units is likely to be opaque to English speakers, even to those who have an inkling as to their etymology. One would be hard pressed to explain the meaning of perform on the basis of the putative meanings of per- and -form.

2.5. meat In spite of the evident semantic relation between meat and eat − meat is something that you eat − we would probably reject outright an analysis of meat in terms of prefixation of the verb eat with [m-]. This is because the analysis fails to fit in with other facts about the language. Other words terminating in -eat have nothing to do with eating: heat, bleat, pleat, etc. Neither are there any other examples of nouns being formed by prefixing a verb with m-. There is therefore no basis for proposing the word-formation schema [ N m-V]. Moreover, such a schema would conflict with general patterns of English word-formation. While English is certainly not lacking in derivational prefixes (un-, over-, out-, etc.), prefixation by a single consonant is rare. One of the few examples is the negative import of initial n-, as in one/none, either/neither, or/nor, the now archaic aye/nay, and dialectal owt (‘something’)/nowt (‘nothing’). Even in the case of meat, however, we may need to tread with caution. In light of the vagaries of English spelling, English speakers sometimes have recourse of mnemonic aids for distinguishing the different spellings of homophones. Readers may be familiar with the school-room rule for distinguishing the spellings and :

10. Word-formation in cognitive grammar , with an , is an adjective. Likewise, one can imagine poor spellers appealing to the semantic relation in order to distinguish from . These observations serve as a reminder, not only of the speaker-specific nature of the mental grammar, but also of the role of orthography in the mental representation of words.

2.6. more Even though we would probably not want to recognize m- as a meaningful part of meat, there are numerous cases where a phonological constituent may be felt to be associated with a semantic value (Rhodes and Lawler 1981). Several words commencing in m- are associated with a large number or quantity: many, more, most, multiple, magnitude, million, myriad, as well as, perhaps, magnificence, magnanimous, mighty. There are, of course, many words in m- which do not share this semantic value, and some, even, which convey the notion of smallness: minuscule, minimal, microscopic. Nevertheless, the association may contribute to the feeling that the phonological form of seemingly unanalysable words is not entirely arbitrary. Take the case of droop, discussed next.

2.7. droop This word would probably be regarded as internally simple, yet correspondences and resonances with other words can be detected. Consider words which terminate in -oop, such as hoop, loop, coop, swoop, stoop, and scoop. These words invoke the notion of curvature. (Imagine your posture if you are ‘cooped up’ in an enclosed space.) Now consider words commencing in dr-. These include dry, drip, drop (of water), draft (of beer), drain, dredge, drench, drought, draw (water from a well), drool, and dribble. These have to do with water (or its absence). Remarkably, the meaning of droop is almost compositional: curvature (of a plant) through (lack of) moisture. Often, the proposed semantic value appears to be based in facts of articulation or perception, or in some more vaguely characterized synaesthesic association between sound and meaning (Firth 1930). High front vowels − where there is a small aperture between tongue and palate − tend to be associated with ideas of smallness, while consonant clusters with voiceless plosives may be associated with sudden or jerky movement (split, spike, strip, etc.). The above discussed examples, however, lack this synaesthesic component. By no stretch of the imagination can [dr] be considered a “watery”, or even as a “dry” sound. Whatever their status, these conventionalized sound-symbolic relations do, however, set up resonances within the language network and no doubt contribute to speakers’ intuitions that words somehow match their meanings.

2.8. motel This word is a blend of motor and hotel. The initial part of the one word is dovetailed with the final part of the other: [məʊ[t]ɛl], the blend having a semantic value akin to

151

152

I. Word-formation as a linguistic discipline that of a compound (motor hotel). Blending is very different from canonical cases of analyzability, since the sequences [məʊt] and [tɛl] do not in themselves have symbolic status (with respect to the blend in question). Nevertheless, the viability of a blend depends on the possibility of hearers being able to recover the input words, a crucial factor in this regard being the cohort of words containing the phonological strings in question (Gries 2006). It would be inconceivable to propose [lɒt] as a blend of land and yacht, since initial [l] is hardly likely to activate the word land.

3. Entrenchment and drift; countering analyzability Wheeler and Schumsky (1980) report a remarkable finding. When speakers of English were invited to divide words up into their component parts, about half failed to mark any internal division in the word baker. We can imagine speakers being reluctant to mark a boundary in barber. But baker? A likely explanation lies in the entrenchment of the word (Hay 2001; Taylor 2012: 131). A feature of entrenchment is that a unit does not need to be assembled (compositionally) from its parts on each occasion of its use, nor do language users need to refer to its parts in order to understand it (Langacker 1987: 59). It already exists as a preformed unit and is stored and accessed as such. As a consequence, the unit’s internal structure tends to be obscured. Monitoring tasks show that speakers are slower to recognize a word (such as of ) when it occurs as part of a frequently occurring string (such as sort of ) than when it is part of a less entrenched sequence (such as example of ) (Sosa and MacFarlane 2002). One symptom of the entrenchment of (potentially analysable) units is the phenomenon of drift, whereby an expression takes on phonological and semantic properties in addition to (or even at variance with) its compositional value; the expression, as it were, acquires a life of its own and drifts away from its source. Inflected forms of nouns and verbs typically have their own characteristic distributions. Plural eyes distributes in the language quite differently from singular eye: the singular tends to be used in idiomatic locutions (keep an eye on, etc.), whereas reference to the organ of sight dominates for the plural form (Sinclair 2004). Derived forms often acquire semantic nuances additional to their compositional values. Bybee (1985) cites the example of dirty, an adjective whose use need not invoke the notion of ‘dirt’. Similarly, baker, as an entrenched unit, is not restricted to referring to ‘one who bakes’; more commonly, the word is used as the name of a kind of retailer, or retail outlet, which sells certain kinds of baked goods, typically breads, pastries, and cakes. If you buy something ‘from the baker’, you are not necessarily buying from ‘the person who baked’ the stuff. Phonological drift − evidenced by such processes as assimilation, elision, palatalization, consonant lenition, and vowel reduction, sometimes operating cumulatively over lengthy time periods − is liable to obscure the internal make-up of words. Although still evident from the more conservative spelling, the relation of cupboards to cups and boards is probably lost on most English speakers, as is the relation between state and station. Informal enquiries suggest that many speakers fail to perceive the relatedness of preside and president, and even of horizon and horizontal.

10. Word-formation in cognitive grammar

4. Characterizing the parts Derived words − the example of farmer is typical − exhibit a marked asymmetry with respect to the properties of their parts (Taylor 2002; Tuggy 1992). Some of the parameters are applicable to the analysis of phrases; others are specific to the domain of word structure. a) Autonomy vs. dependence. A unit is autonomous to the extent that it can be conceptualized in and of itself, without reference to other units. Although initially proposed with reference to semantic units, in particular to verbs and their complements (Langacker 1992: 286), the notion also applies to phonology. Stop consonants are phonologically dependent in that we cannot conceive of a stop consonant without invoking some notion of an adjacent vowel-like segment. Suffixes, whether derivational or inflectional, are overwhelmingly phonologically dependent. The suffix [-er], being unstressed, absolutely requires as a host a stressed syllable. Prefixes, on the other hand, tend to be phonologically autonomous. Forms such as over- (as in overeat), out- (as in outperform), and anti- (as in anticommunist) in fact exist as phonologically autonomous words. The same, of course, goes for parts of a compound; both air and port − parts of airport − are phonologically autonomous. b) Schematic vs. contentful. Some conceptual units, such as those symbolized by do, be, and it, have a highly schematic semantics; others, such as those symbolized by basic level terms such as walk, sing, and table, are more contentful. Prefixes and suffixes are notable for their semantic schematicity while the stems to which they attach supply the bulk of the semantic content. Prefixed out- merely refers to some schematically characterized process (whose content is supplied by the stem to which it attaches) which exceeds some expected norm. Some affixes are phonologically schematic. The (regular) plural morpheme of English is an alveolar fricative. Whether the sound emerges as voiced or voiceless, and, if voiced, whether preceded by an epenthetic schwa, is determined by properties of the host. c) Valence. The specification of a unit often requires reference to the kinds of items with which it can combine. (The phenomenon is studied in syntax under the rubric of subcategorization.) A transitive verb (by definition) is one which requires a direct object complement. Another way of putting this is to say that the characterization of a unit may make reference to the (schematically characterized) construction in which it needs to occur, in the case in point, the transitive clause construction. Noun-forming -ness needs to attach to an adjective, while agentive -er attaches to verb stems. The case of -er is, to be sure, more complex. Like many symbolic units, the suffix is polysemous (Panther and Thornburg 2001; Ungerer 2007). Thus, on one of its meanings, the suffix can characterize a person in terms of their location, as in villager, islander, Londoner, and back-bencher. d) Selection (or choosiness). Syntacticians distinguish between subcategorization and selectional restrictions. While drink subcategorizes for a nominal object, it semantically selects the name of a liquid. The “choosiness” of word-parts goes well beyond matters of semantic plausibility. There is no semantic reason why apt and correct should prefer the periphrastic comparatives more apt and more correct over the in-

153

154

I. Word-formation as a linguistic discipline flected forms apter and correcter, while poor and simple exhibit the opposite tendency, preferring the inflected forms poorer and simpler. One factor is the phonological properties of the adjectives − those terminating in a consonant cluster, more specifically an obstruent cluster, tend to shun the inflection (Hilpert 2008; Mondorf 2003). Amongst suffixes with roughly comparable semantic values we likewise find large differences in their range of uses. While noun-forming -ness is fairly promiscuous, in that it can attach to a very large and possibly open-ended set of adjectival stems, noun-forming -th, -ity, and -al are much more choosy, being virtually restricted to occurring with a closed set of hosts. Factors which motivate these differences are an important site of research (Baayen and Renouf 1996; Bauer 2001). e) Profile determinant. This term refers to that constituent of an expression which determines the kind of entity that the expression refers to (Langacker 1987: 235). (The notion thus corresponds to the traditional notion of head.) Suffixes are profile determinants. The fact that farmer refers to a person, or that farmed refers to a past-time event, is due to the affix. The case of prefixes is more diverse. Un-, n-, and anti- do not impinge of the semantic type of the referent. Out-, on the other hand, imposes a verbal conceptualization; (to) out-Nixon (Nixon). f) Coercion. One unit may influence the phonological shape of a neighbouring unit. The shape of the (regular) plural morpheme in English is determined by the phonological properties of (the final segment of) its host. More complex are the effects of an affix on its host. For example, a feature of adjective-forming -ic is that stress is placed on the immediately preceding syllable; this, in combination with other phonological schemas, may trigger changes in vowel quality. Compare photograph and photographic. The above parameters allow us to characterize some of the basic categories in morphological theory (Taylor 2002; Tuggy 2005; Ungerer 2007; van Huyssteen 2010). Suffixes (in English) are phonologically dependent, are semantically schematic, are profile determinants, they can be quite choosy with respect to the hosts to which they attach, and can coerce the phonological structure of their hosts. Clitics, such as possessive s in English, are phonologically dependent, are non-coercive, and can attach to practically any kind of host (think of examples like The person I met’s new car, where the clitic attaches to a verb). Freedom of combination is also a defining feature of words, along with such features as phonological autonomy and (in most cases) semantic contentfulness. The approach also accounts for − indeed, leads us to expect − that the boundaries of these categories will be fuzzy, and that some items will be able to “migrate” from one category to another. When unstressed, prepositions have clitic-like properties, while prefixes such as anti- and ex- can function as independent words, as befits their phonological autonomy. Parts of a blend can also achieve symbolic status. Kemmer (2003) discusses the case of forms in -erati (chatterati, culturati, and, more recently, twitterati), based, according to her, on the blend glitterati (glitter + literati).

5. Composition The counterpart of analyzability is composition, the process of creating new forms from existing resources. One approach − we might call it the Lego model − views composition

10. Word-formation in cognitive grammar in terms of the “fit” between the parts. The model works tolerably well for syntax (at least, for basic phrase structure configurations). Tea matches the subcategorization and selectional requirements of drink; hence drink tea is an acceptable (grammatical) formation. The Lego model does not always work, even for syntax. By all accounts, lie ‘to speak untruthfully’ is an intransitive verb. Yet in He lied his way through the interview the verb takes a direct object. The transitive use of lie is imposed by the constructional schema in which the verb occurs, whereby the specifications of the schema override the properties of a constituent unit (Jackendoff 2010). The Lego model is even less appropriate for word-formation. For one thing, some word-formation processes, such as blending, do not consist at all in the fitting together of already entrenched parts. But even when parts can be identified, their combination is associated with all manner of restrictions, idiosyncrasies, and overrides. One cannot, willy-nilly, attach -al to a verb stem to form a noun designating an instance of the event. Cooker does not refer to a person who cooks (the concept is pre-empted by the noun cook). A person who practises geology is not a geologer (geologist is the term). Nor, for that matter, can one be sure that attaching /s/ to a singular noun will always generate the acceptable plural form.

5.1. Productivity It is common in the literature to speak of the productivity of an affix. By this is meant, not so much the number of forms which exhibit the affix, but the potential of the affix for the creation of new forms. By this measure, -ness turns out to be a highly productive noun-forming affix, -ity, -th, and -al much less so (Baayen and Lieber 1991). It would be more accurate, perhaps, to speak of the productivity of the schema which sanctions the use of the affix. Productivity is a function of entrenchment, which itself is a function of frequency of occurrence. The relation to frequency is not a simple one, however (Bybee 1995, 2001; Dąbrowska 2004). At issue is not the frequency with which the schema per se is instantiated, but the number of different instances on which it is instantiated (in comparison with the number of counter-indications). Word-formation processes which are not particularly frequent in the language at large may nevertheless be productive, provided that they are instantiated on a sufficiently wide range of examples. Nominals in -ee are not all that frequent, yet the word-formation process appears to be quite productive, as a consequence of the large number of different types. Equally, processes which are frequently attested may not be productive. The pattern of past tense formation exhibited by the high frequency verbs do, say, and have is not able to be extended to other verbs.

5.2. The power of the schema In view of its evident inadequacies, for both syntax and word-formation, the Lego model is being abandoned by an increasing number of theorists in favour of an approach which focuses on the sanctioning schemas (or constructions) (cf. article 12 on word-formation in construction grammar; Taylor 2012; Tomasello 2003). As an illustration of the power of schemas in word-formation, consider the example of hamburger (Taylor 2002: 293). According to the Oxford English Dictionary, the word was used in the late 19th century

155

156

I. Word-formation as a linguistic discipline to refer to a minced meat patty, a so-called ‘Hamburger steak’, supposedly a speciality of the city of Hamburg. At this time, the word was presumably analysed as [Hamburg] + [-er]. At some point, however, the word was re-analysed as [ham] + [burger]. We can be confident of this because of the existence of forms such as cheeseburger, and even the emergence of burger as an autonomous unit. What could have triggered the reanalysis? Hamburgers, after all, are made of beef, not ham. The crucial factor is not the semantics, but the phonology, more precisely, a phonological schema. The stress pattern of hamburger ['σ ˌσ σ] matches the stress pattern of compound nouns; compare dog-whistle, clothes cupboard, soup-kitchen, and countless more. When applied to hamburger, the schema indicates a comparable analysis, with [burger] emerging as the profile determinant and [ham-] as a kind of nominal modifier. The way was then open for burger to emerge as an independent word and for it to be modified by other words in a compound structure.

6. Conclusion The minimalist ontology of cognitive grammar, in association with its rigid adherence to a usage-based account of language acquisition, offers the possibility of a rigorous and constrained description of the internal structure of words and of the processes which sanction creative word-formation. As will be apparent from the brief overview presented in this article, the cognitive grammar approach relies heavily on the notions of entrenchment and of schematic representations; the latter arise through familiarity with instances and are able to sanction both the analyzability of encountered expressions and the creative construction of new expressions. Moreover, the issues raised in this article are by no means unique to the analysis of words and the creation of new words, being applicable both to “pure” phonology and to syntactic organization alike. To this extent, wordformation takes its place in a broad, unified conception of the symbolic system which constitutes a human language.

7. References Baayen, Harald and Rochelle Lieber 1991 Productivity and English derivation: A corpus-based study. Linguistics 29: 801−843. Baayen, Harald and Antoinette Renouf 1996 Chronicling the Times: Productive lexical innovations in an English newspaper. Language 72: 69−96. Bauer, Laurie 2001 Morphological Productivity. Cambridge: Cambridge University Press. Benczes, Réka 2007 Creative Compounding in English. The Semantics of Metaphorical and Metonymical Noun-Noun Combinations. Amsterdam/Philadelphia: Benjamins. Bybee, Joan 1985 Morphology. A Study of the Relation between Meaning and Form. Amsterdam/Philadelphia: Benjamins. Bybee, Joan 1995 Regular morphology and the lexicon. Language and Cognitive Processes 10: 425−455.

10. Word-formation in cognitive grammar Bybee, Joan 2001 Phonology and Language Use. Cambridge: Cambridge University Press. Dąbrowska, Ewa 2004 Rules or schemas? Evidence from Polish. Language and Cognitive Processes 19: 225− 271. Firth, John 1930 Speech. London: Benn’s Sixpenny Library. Gries, Stefan Thomas 2006 Cognitive determinants of subtractive word formation: A corpus-based perspective. Cognitive Linguistics 17: 535−558. Hay, Jennifer 2001 Lexical frequency in morphology: Is everything relative? Linguistics 39: 1041−1070. Hilpert, Martin 2008 The English comparative − language structure and language use. English Language and Linguistics 12: 395−417. Jackendoff, Ray 2010 Meaning and the Lexicon. The Parallel Architecture 1975−2010. Oxford: Oxford University Press. Kemmer, Suzanne 2003 Schemas and lexical blends. In: Hubert Cuyckens, Thomas Berg, Rene Dirven and Klaus-Uwe Panther (eds.), Motivation in Language, 69−97. Amsterdam/Philadelphia: Benjamins. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol. 1: Theoretical Prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Vol. 2: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 2008 Cognitive Grammar. A Basic Introduction. Oxford: Oxford University Press. Mondorf, Britta 2003 Support for more-support. In: Günter Rohdenburg and Britta Mondorf (eds.), Determinants of Grammatical Variation in English, 251−304. Berlin: Mouton de Gruyter. Nesset, Tore 2008 Abstract Phonology in a Concrete Model. Cognitive Linguistics and the MorphologyPhonology Interface. Berlin: Mouton de Gruyter. Panther, Klaus-Uwe and Günter Radden (eds.) 2011 Motivation in Grammar and the Lexicon. Amsterdam/Philadelphia: Benjamins. Panther, Klaus-Uwe and Linda Thornburg 2001 A conceptual analysis of English -er nominals. In: Martin Pütz, Suzanne Niemeier and René Dirven (eds.), Applied Cognitive Linguistics. Vol. 2, 149−200. Berlin: Mouton de Gruyter. Rhodes, Richard and John Lawler 1981 Athematic metaphors. In: Roberta Hendrick, Carrie Masek and Mary Frances Miller (eds.), Papers from the Seventeenth Regional Meeting of the Chicago Linguistic Society, 318−342. Berkeley, CA: Chicago Linguistic Society. Saussure, Ferdinand de 1964 [1916] Cours de linguistique générale. Paris: Payot. Sinclair, John 2004 Trust the Text. Language, Corpus and Discourse. London: Routledge.

157

158

I. Word-formation as a linguistic discipline Sosa, Anna Vogel and James MacFarlane 2002 Evidence for frequency-based constituents in the mental lexicon: Collocations involving of. Brain and Language 83: 227−236. Taylor, John R. 2002 Cognitive Grammar. Oxford: Oxford University Press. Taylor, John R. 2004 The ecology of constructions. In: Günter Radden and Klaus-Uwe Panther (eds.), Studies in Linguistic Motivation, 49−73. Berlin: Mouton de Gruyter. Taylor, John R. 2012 The Mental Corpus. How Language is Represented in the Mind. Oxford: Oxford University Press. Tomasello, Michael 2003 Constructing a Language. A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Tuggy, David 1992 The affix-stem distinction: A Cognitive Grammar analysis of Data from Orizaba Nahuatl. Cognitive Linguistics 3: 237−300. Tuggy, David 2005 Cognitive approach to word formation. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 233−265. Dordrecht: Springer. Ungerer, Friederich 2007 Word-formation. In: Dirk Geeraerts and Hubert Cuyckens (eds.), The Oxford Handbook of Cognitive Linguistics, 650−675. Oxford: Oxford University Press. van Huyssteen, Gerhard B. 2010 (Re)defining component structures in morphological constructions: A Cognitive Grammar perspective. In: Alexander Onysko and Sascha Michel (eds.), Cognitive Perspectives on Word Formation, 97−126. Berlin/New York: de Gruyter. Wheeler, Cathy J. and Donald A. Schumsky 1980 The morpheme boundaries of some English derivational suffixes. Glossa 14: 3−34.

John R. Taylor, Christchurch (New Zealand)

11. Word-formation in optimality theory 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

Introduction Markedness vs. faithfulness in phonology Constraint interaction and the prosodic organization of affixes Affix placement Neutralization in affixes Faithfulness in base-derivative relations Unmarkedness in hypocoristics Allomorphy and gaps Anti-faithfulness or anti-homophony constraints Criticism Summary References

11. Word-formation in optimality theory

Abstract Optimality theory (henceforth OT) models natural language competence in terms of interactions of universal constraints, notably markedness and faithfulness constraints. This article illustrates some of the major advances in the understanding of word-formation phenomena originating from this theory, including the prosodic organization of morphologically complex words, neutralization patterns in derivational affixes, allomorphy, and infixation.

1. Introduction The goal of OT, shared with all models of generative grammar, is to account for linguistic well-formedness by teasing apart the rule-governed from the idiosyncratic and the universal from the language-specific. This task is approached in traditional generative grammar by mapping underlying forms, which are designed to encode the idiosyncratic, to observable surface forms through a successive application of structure-changing rules. OT addresses this task by mapping (hypothetical) inputs to observable surface forms by selecting optimal forms among output candidates. The selection process involves the evaluation of the candidates with respect to ordered, often inherently conflicting constraints. While the constraints themselves are assumed to be universal, the order among them is largely language-specific (Prince and Smolensky 2004). To clarify some of the major new ideas developed within OT, the model will first be introduced with phonological data. This is because the pioneering work in OT concerns phonology and OT analyses of word-formation phenomena have been largely focused on the morphology-phonology interface (cf. McCarthy 2006). Articles with other foci typically concern inflectional rather than derivational morphology (cf. Ackema and Neeleman 2005; Wunderlich 2006; Xu 2011). OT-based approaches to word-formation focusing on semantics are somewhat tentative at this point (Lieber 2010). An emphasis on the phonology-morphology interface is motivated not only by the work hitherto carried out in OT but also by the significance of this interface to one of the central controversies in word-formation theory. This controversy concerns the question of whether word-formation phenomena can be adequately captured in a simple IA (item-and-arrangement) model. Such a model presupposes inputs consisting of linearly ordered morphemes, including both stems and affixes, where output forms may be subject to phonological modification. Noting that such modifications can be drastic and “unnatural”, some have found such a model inadequate. Instead, it has been argued that both inputs and outputs in word-formation consist of words to be related by a series of processes, including both affixation and so-called “allomorphy rules” (Aronoff 1976). This view raises the questions of how to capture shared properties between affixes and stems and whether or not such IP (item-and-process) models are subject to any restrictions. OT is most relevant to this discussion as it allows for analyses which are consistent with a simple IA model without assuming phonological modifications of input forms.

159

160

I. Word-formation as a linguistic discipline

2. Markedness vs. faithfulness in phonology To illustrate the main differences between OT and previous generative grammar consider the patterns of loanword adaptation in (1) (the transcriptions of the German words are adopted from Duden 2005). (1)

a. English [bɛt] [pɛp] [stɛp] [wɛb]

> > > >

German [bɛt] [pɛp] [stɛp] [vɛp]

b.

English [bæɡ] [spæm] [fæn] [dæd]



> > > >

German [bɛk] [spɛm] [fɛn] [dɛt]

The data show that both English [ɛ] and [æ] are uniformly adapted as [ɛ] in German. Also English voiced and voiceless obstruents in syllable-final position are uniformly adapted as voiceless obstruents in German. In traditional generative grammar the mapping of input to output forms involves a series of rules, which successively modify structure. By contrast, in OT this mapping rests on the comparison of potential input and output forms. Given for instance the form /bæɡ/ as an input, a device called GEN(erator) supplies output candidates, including the form identical to the input but also a potentially infinite number of forms deviating from the input in various ways. A few example candidates are given in the lefthand column in (2). (It is assumed here that the constraints of a phonological grammar concern the phonemic level of representation.) bæɡ

(2)

*OBS[+VOICE]COD

*[–back]/[+low] *

a.

/bæɡ/

*

b.

/bɛɡ/

*

c.

/bæk/

d.

☞ /bɛk/

e.

/pɛk/

FAITH

* *

* ** ***

The core component of the grammar, called EVAL(uation metric), is responsible for ensuring that the actually attested form, in this case the form /bɛk/, emerges as the winner. (The symbol ☞ indicates the winner). This task is solved by ordering constraints such that all rival candidates are eliminated from the competition and only the winner remains. The relevant procedure involves the successive evaluation of the candidates with respect to the ordered constraints, starting with the highest constraint, and eliminating candidates whenever they fare worse than at least one of the remaining competitors. The task of the linguist is then to identify and order the constraints relevant for eliminating non-winning competitors. Returning to the tableau in (2), a key role for selecting the winner is played by the constraint FAITH (“faithfulness”), a novel type of constraint introduced in OT which militates against deviations between output and input structures. The assumption of such a constraint implies that all candidates which are more similar to the input than the winner must be eliminated due to intrinsic flaws (cf. (2a, b, c)). Such flaws concern

11. Word-formation in optimality theory violations of markedness constraints, which express universal preferences pertaining to the structural well-formedness of output forms. The two markedness constraints referred to in (2) are stated in (3). The winner in (2d) differs from the more faithful candidates in (2a, b, c) in that it satisfies those two constraints. (3)

a. *[+low]/[−back] b. *OBS[+VOICE] COD

Low front vowels are prohibited Voiced obstruents in coda position are prohibited

Whereas all candidates exhibiting greater similarity to the input than the winner are eliminated due to violations of specific markedness constraints, those exhibiting less similarity can be eliminated as a result of violating FAITH. The tableau in (2) illustrates the elimination of the candidate /pɛk/, which, unlike the winner, does not contain any segment with an unaltered correspondent in the input. Consideration of additional candidates such as /bak/, which like the winner also differs in only two segments, indicates the need to split FAITH into several subconstraints referring to individual features. Ordering FAITH(±back) above FAITH(+low) would account for the preference of the candidate /bɛk/ over /bak/. Additional possibilities for differentiating between faithfulness constraints are listed in (4). An adequate ordering of these constraints ensures the elimination of other promising candidates such as /bæɡi/, /bæ/, or /ɡæp/ (for an overview over constraints commonly used in OT, cf. Kager 1999): (4)

a. FAITH(±back) b. FAITH(+low) c. MAXIMALITY d. DEPENDENCE e. LINEARITY f.

CONTIGUITY

Correspondent segments have identical values for feature [±back] (No fronting or backing!) Correspondent segments have identical values for feature [+low] (No raising!) Every segment in the input has a correspondent in the output (No deletion!) Every segment in the output has a correspondent in the input (No epenthesis!) The input string reflects the precedence structure of the output string and vice versa (No reversal!) Correspondent segments in inputs and outputs form contiguous strings (No morpheme-internal insertion or deletion!)

The description of the data in (1) is complete when all markedness and faithfulness constraints have been ordered such that, for any given input, the relevant winner is selected as the output. More generally, an OT grammar is complete, when the optimal output is determined for any hypothetical input. This ban on any restrictions on inputs is known as “richness of the base” and, together with the constraint-based parallel evaluation of output forms, constitute major innovations originating with OT. Conceiving of grammar in terms of interacting markedness and faithfulness constraints sheds new light on the notion of function in language. For instance, the constraint ranking in (2) indicates an advantage of German grammar compared to English in that markedness constraints prevail, which facilitate the pronunciation, perception, and perhaps also the learning and storage of words. English, a language where FAITH dominates both markedness constraints, is on the other hand “better” in more fully exploiting the contrastive potential by allowing more feature combinations in more positions. Hence,

161

162

I. Word-formation as a linguistic discipline the phonology of English, but not that of German, allows for the meaning differences in (5) to be formally distinguished as follows: (5)

/bæɡ/ vs. /bɛɡ/ vs. /bæk/ vs. /bɛk/

This, in a nutshell, is the basic conflict to be resolved by a phonological grammar: the need to optimize form by favoring maximally unmarked structures versus the need to optimize the contrastive potential by allowing maximal freedom of choice for combining features or phonemes. The observation that the resolution of this conflict can differ for various positions can be captured by linking constraints to specific contexts. One approach is known as “positional faithfulness” (Beckman 1999), the other as “positional markedness” (Zoll 2004). On the latter approach, illustrated in the tableau in (2), the absence of voiced obstruents in coda position is captured by ordering the restricted markedness constraint as specified in (3b) above a general faithfulness constraint FAITH(±VOICE) (Corresponding segments have identical values for the feature [±voice]), cf. (6a)). By contrast, on the “positional faithfulness” approach, the absence of voiced obstruents in coda position is captured by ordering a faithfulness constraint FAITH(±VOICE)ONS linked to the syllable onset above the constraint *OBS[+VOICE], a general markedness constraint against voiced obstruents (cf. (6b)), which then in turn dominates FAITH(±VOICE). As a result, voiced and voiceless obstruents contrast in the syllable onset, but not in the coda. The notion of “positional faithfulness” hence captures the well-known generalization that prominent positions tend to exhibit contrasts which are neutralized in nonprominent positions (Trubetzkoy 1958: 212). (6)

a. *OBS[+VOICE] COD >> FAITH(±VOICE) b. FAITH(±VOICE) ONS >> *OBS[+VOICE] >> FAITH(±VOICE)

The notion of (positional) markedness has a close analogue in formal constructs such as “surface structure constraints” or “filters” introduced in generative grammar (cf. Shibatani 1973). However, such devices have an awkward status in that model, in contrast to the key role played by markedness constraints in OT. The innovation in OT without any formal analogue in previous frameworks concerns the concept of faithfulness, that is, constraints which actively favor sameness of structure between input and output. Given the premises of OT, all permutations of universal markedness and faithfulness constraints constitute possible grammars and all grammars consist of specific rankings of universal markedness and faithfulness constraints. There are no constraints on permissible inputs.

3. Constraint interaction and the prosodic organization of affixes Prosodic organization plays an important role in word-formation, to be demonstrated throughout this article. Before addressing the conditions concerning the prosodic organization of morphemes in complex words, I will briefly present some constraints determining the basic organization of segments in syllables. Typological studies of syllable structure have revealed a universal preference for simple CV syllables, consisting of a consonant in the onset followed by a vowel in the

11. Word-formation in optimality theory

163

nucleus. In OT this preference is captured by various markedness constraints, including ONSET (Every syllable needs an onset), NOCODA (Codas are prohibited), and *COMPONS (Complex onsets are prohibited). Other constraints limiting the association of specific segments with specific syllable positions concern the position of a segment within the sonority scale. The order of segments on this scale roughly correlates with the openness of the vocal tract during the articulation, ranging from low vowels, the most open sounds, to plosives, the maximally closed sounds. In OT, the relevant generalizations are expressed in terms of so-called anti-association constraints, which prohibit the association of segments with specific syllable positions. The orders for onsets and nuclei are sketched in (7a, b). The top ranking of “*ONS /ɑ” (“/ɑ/ is prohibited in the onset”) in (7a) means that /ɑ/, due to its maximal sonority, is the worst occupant of the syllable onset while the low ranking of “*ONS/ PLS” means that plosives, due to their minimal sonority, are the most preferred segments in onset position. The opposite preferences obtain for the nucleus position (cf. (7b)) (cf. Prince and Smolensky 1993; for the relevant descriptive generalizations see also Vennemann 1988). (7)

a. *ONS/ɑ >> *ONS/i >> *ONS/ʀ >> *ONS/l >> *ONS/NAS >> *ONS/ FRC >> *ONS/ PLS b. *NUC/ PLS >> *NUC/ FRC >> *NUC/NAS >> *NUC/l >> *NUC/ʀ >> *NUC/i >> *NUC/ɑ

The ranking of faithfulness constraints relative to these anti-association constraints accounts then for the thresholds observed in individual languages: in German only the least sonorous vowel /i/ is allowed in the onset (cf. /iɑ/ ‘yes’) whereas in English, both /i/ and /u/ are allowed (cf. /iɛs/ , /ui/ ). These examples also illustrate the effect that syllabic organization has on the phonetic realization of a phoneme. In both /ui/ and /iu/ the respective first vowel is pronounced as a short glide followed by a longer and more steady sound (i.e. [wi:] , [ju:] ). This is because in both words the first vowel is organized as an onset while the second vowel is organized as a nucleus. Turning now to complex words in English, there are cases where the phonetics indicates the association of a stem-final consonant with the syllable coda, even when a stressed vowel follows (cf. Kahn 1980). In (8a) /t/ is flapped before a stressed vowel, which clearly signals that the vowel belongs to a separate stem. Within simplexes /t/ is always phonetically aspirated in that context (cf. (8b)). Aspiration also applies when the plosive occurs in stem-final position before a suffix as in the word hittee shown in (8c). (8)

Phonetic surface forms: ´ ɾɑ̀ɪ] a. [khæ b. [lɑ´:thɛ̀ɪ] c. [hɪt̀ hí:]

Morphological structure: [kæt]STM[ɑi]STM [lɑtɛi]STM

[hɪt]STM[i]SFX

The /t/ allophony is readily explained with reference to prosodic structure. The trees in (9) exhibit a hierarchical structure, ranging from composite groups (CG), over phonological words (ω), feet (Σ), syllables (σ), down to the subsyllabic constituents onset (O), nucleus (N), and coda (C). In branching nodes, one of the daughters is strong, marked by the subscript “s” in (9), the others are weak, marked by the subscript “w”. Signifi-

164

I. Word-formation as a linguistic discipline cantly, in (9b, c) the /t/ occurs exclusively in the onset, in accordance with general constraints on the syllabification of consonants before stressed vowels. Aspiration can be considered a consequence of this prosodic structure. The apparent deviation from this pattern in (9a) indicates the presence of an internal domain boundary, which prevents the regular syllabification of /t/ as the onset of the stressed syllable. As a consequence, /t/ associates with the coda, resulting in lenition (flapping) rather than fortition (aspiration). (It is assumed here that syllabic tense, high vowels in English associate with the nucleus and coda.) (9)

The representations in (9) raise the question of how to capture the associations between boundaries of prosodic domains and those of specific morphological categories. In OT such cases of systematic coincidences are described in terms of so-called “alignment” constraints, a special type of markedness constraints (for a similar idea in pre-OT work see Selkirk 1986). The general schema of such constraints, referred to as “generalized alignment”, is given in (10) (Cat = category, PCat = prosodic category (e.g., foot, syllable), GCat = grammatical category (e.g., stem, affix), cf. McCarthy and Prince 1993). (10) Align (Cat1, Edge1, Cat2, Edge2) = def c Cat1 d Cat2 such that Edge1 of Cat1 and Edge2 of Cat2 coincide. Where Cat1, Cat2 2 PCat g GCat Edge1, Edge2 2{R (right), L (left)} The alignment constraints relevant for mapping the morphological structures in (8) to the prosodic domains (kæt)ω(ɑi)ω, (lɑtɛi)ω, and (hɪti)ω concern stem and phonological word boundaries. The “fused” prosodic domain of the complex word (hɪti)ω, compared to the two domains of (kæt)ω(ɑi)ω, could then be captured by ranking the constraint ONSET between the two alignment constraints in (11). (“ALIGN (STEM, L, ω, L )” = “align the left boundary of every stem with the left boundary of a phonological word”; “ALIGN (STEM, R, ω, R )” = “align the right boundary of every stem with the right boundary of a phonological word”.) (11) ALIGN (STEM, L, ω, L ) >> ONSET >> ALIGN (STEM, R, ω, R )

11. Word-formation in optimality theory

165

The constraint order in (11) accounts for the generalization that the vowel-initial suffix -ee is integrated into the phonological word of the stem, to satisfy ONSET, while compound members such as eye form separate phonological words, regardless of their segmental structure. In functional terms, the grammar in (11) expresses the generalization that in English, it is less important for syllables to have an onset than for boundaries of individual compound members to be prosodified separately. At the same time, it is more important for syllables to have an onset than to mark the boundary between a stem and a suffix. As a result, stem-final /t/ occurs in the onset position in (9c) but in the coda position in (9a). This effect is reached without resyllabification rules employed in rule based frameworks, where the /t/ in the stem hit would disassociate from the coda to reassociate with the onset position, once the suffix has been attached. The ranking in (11) captures not only the difference in the prosodic organization between compounds and suffixed words but also predicts potential differences between vowel-initial and consonant-initial suffixes, which satisfy ONSET independently. The ranking thus expresses a connection between the phonological shape of a suffix and its prosodic organization, thereby accounting for the fundamental difference between socalled “cohering” (integrated) and “non-cohering” (non-integrated) affixes (cf. Dixon 1977). At the phonetic level, this difference is manifest in the absence of perfect rhymes between words with consonant-initial suffixes and simplexes in English. This is because the pronunciation of phonemes is necessarily affected by their position within higher prosodic constituents. In English, vowels are potentially lengthened more extensively when occurring in foot-final position as in (12c), compared to foot-internal position, as in (12a, b). (12) a.

b.

[vinəs]STM

c.

[vin]STM[əs]SFX

[fɹi]STM[nəs]SFX

As a result, words with vowel-initial suffixes such as venous can be homophonous with simplexes such as Venus, whereas a word with a consonant-initial suffix such as freeness will exhibit variable but measurable − often noticeable − phonetic differences. The wellaligned prosodic structures in (12c) further account for the systematic “stress-neutrality” of consonant-initial suffixes in English. As a result of being outside of the phonological word of the stem, they are outside of the domain of foot formation as well, unable to influence the stress patterns of the stem. The same generalizations obtain in German, where again vowel-initial suffixes cohere, whereas consonant-initial suffixes or compound members form separate prosodic domains. (The glottal stop in stressed suffixes

166

I. Word-formation as a linguistic discipline such as German [ ʔɑ´:l] or [ ʔø´:s] is non-phonemic and consequently does not influence the formation of prosodic domains. Reference to the phonemic level (i.e. /ɑl/, /øz/) results in the correct analysis of such suffixes as cohering ((fʀɔntɑ´l)ω ‘frontal’, (poʀø´z)ω ‘porous’)). Further studies of the prosodic organizations of complex words in English and German show that all suffixes without an onset integrate into the phonological word of the stem, even when the stem fails to provide one. In addition, English prefixes can also be integrated as is shown by the contrast between enable, where the morpheme-final /n/ forms the onset of the following syllable (/ɪ.nɛ´i.bəl/), versus unable, where the morpheme-final /n/ closes the syllable in careful speech and both morphemes are stressed (/ɐ̀n.ɛ´i.bəl/) (Jones 2006). This prosodic contrast shows that alignment constraints are sensitive to the difference between regular category-changing versus strictly modifying prefixes, such that the latter form separate phonological words ([ɪn] PFX [ɛibəl]STM → (ɪnɛibəl)ω, [ɐn] MD-PFX [ɛibəl]STM → (ɐn)ω(ɛibəl)ω). These observations, while calling for a modification of the specific grammar in (11), are in line with the approach to describing prosodic domains in terms of interacting alignment and markedness constraints. The important pattern concerns the observation that prosodic integration is tied to a context where a markedness constraint is violated (i.e. syllables lacking an onset). It follows from the basic premises of OT that while there may be languages where all suffixes are integrated into the prosodic domain of the stem − to satisfy a ban on non-integrated suffixes − there cannot be a language where only consonant-initial suffixes are integrated while vowel-initial suffixes form separate domains. This is because there are no markedness constraints whose ranking would produce such a result. Reference to properly aligned prosodic domains in output structures is crucial not only for capturing phenomena relating to the prosodic organization of phonemes and their phonetic realization, but also for the description of various morphophonological phenomena, including neutralization patterns and allomorphy rules (cf. sections 5 and 8). Ways in which interacting constraints can determine morpheme position are illustrated in the next section.

4. Affix placement Dominant markedness constraints can condition not only the prosodic integration of affixes but can also affect their placement. The much-discussed case of Tagalog umaffixation is a case in point. The actor-focus affix um- occurs as a regular prefix in combination with vowel-initial stems as in (13a), but appears as an infix before the first vowel in consonant-initial stems (cf. (13b); Schachter and Otanes 1972: 292). (13) a. abot ‘to reach for’ → umabot ibig ‘to love’ → umibig

b. tawag ‘to call’ → tumawag ɡradwet ‘to graduate’ → ɡrumadwet

The distinct locations of the affix can be explained with reference to the markedness constraint NOCODA, which prohibits closed syllables (cf. Prince and Smolensky 2004: 40−42). Specifically, ranking NOCODA above the faithfulness constraint LINEARITYM, which requires all linear precedence relations among morphemes given in the input to

11. Word-formation in optimality theory

167

be preserved in the output, ensures the avoidance of C + C junctures in affixation (Horwood 2002). This point is illustrated by comparing the first candidate in (14), where um properly precedes the consonant-initial stem thereby incurring an additional violation of NOCODA, with the second candidate, where this problem is circumvented by inserting um before the stem-internal vowel. The intrusion of the affix within the stem indicates the low ranking of CONTIGUITY (cf. (4f)). Its precise location appears to be determined by the linear ordering between the morphemes: due to its leftmost position in the input, the morpheme um appears in the leftmost position which satisfies NOCODA. (It is assumed here that the grammar ensures the presence of the stem-final consonants and forces their association with the coda.) (14)

um < tawag /um.ta.wag/ ☞ /tu.ma.wag/

NOCODA

LINEARITYM

** *

*

The observation that infixation of the morpheme -um- has the effect of enhancing the phonological wellformedness of the output structure is the key insight expressed in the analysis in (14). If both morphemes are vowel-initial and nothing is gained by infixation, LINEARITYM will be satisfied and the affix appears as a regular prefix (cf. (15)): (15)

um < abot

NOCODA

☞ /u.ma.bot/

*

/a.bu.mot/

*

LINEARITYM *

The analysis illustrates the “P >> M” ranking schema, which challenges traditional modular views of grammar in that strictly phonological constraints (P) take precedence over morphological constraints (M). To adherents of such views, infixation as in (13b) is analysed by subcategorizing the affix for phonological “pivots” such as the “leftmost vowel” (cf. Yu 2007: 8). This alternative is less compelling in that it fails to express the link between the inherent shapes of the respective morphemes and the universal markedness constraints on syllable structure. Proponents of the analysis based on subcategorization indeed deny the existence of such a link, dismissing the effects concerning markedness in the output forms as incidental (cf. Yu 2007; Paster 2009). Empirically this debate boils down to the question of whether or not infixation can always be shown to be motivated by the satisfaction of phonological markedness in output forms. A notorious case of allegedly non-optimizing infixation, ni-affixation in Leti, is briefly discussed in section 10.

5. Neutralization in affixes Affixes like English -ness or -ee are like stems in that their phonological structure consists of prosodically organized phonemes arranged in a linear order. They differ with

168

I. Word-formation as a linguistic discipline respect to the potential for contrast they exhibit. A grammar consisting of interacting faithfulness and markedness constraints is well-suited to account for both the shared and the separate properties, as will be illustrated below. Comparing English stem forms such as /bɹɐuz/ , /ɡlɪmps/ , /smuð/ with the forms of English non-cohering suffixes shows significant differences with respect to the distribution of phonemes (see article 50 on phonological restrictions on English word-formation). Regarding the distribution of vowels, the relevant suffixes exhibit a clear preference for schwa (/fəl/ , /dəm/ , /səm/ , /mən/ , /iəɹ/ , , /uəɹd/ , /mənt/ ). Full vowels are limited to high vowels and occur only when schwa is not allowed, be it because a high consonant or non-coronal obstruent follows (/lɪŋ/ , /nɪk/ , /ʃɪp/ ), or because no consonant follows (/li/ ). Schwa is ruled out in /hʊd/ , because that suffix forms a separate foot as a consequence of the necessary alignment of /h/ with a left foot-boundary in English grammar. The fact that the non-cohering suffixes contain no low or mid full vowels, along with the association between /ʊ/ and the presence of a foot, indicate the need to differentiate anti-association constraints referring to nuclei with respect to their associations with higher prosodic structure. Specifically, it is necessary to link the ranking of anti-association constraints based on a preference for maximal sonority presented in (7a) to syllables dominated by a phonological word, marked by the subscript “ω” in (16a); all other syllables are associated with a ban on non-high full vowels. Instead such syllables prefer a rounded high vowel in the nucleus, when dominated by a foot (/hʊd/ ), (cf. (16b)) and an unrounded high vowel when not dominated by a foot (cf. (16c)). In the latter case full vowels occur only when schwa is not allowed. The relevant restrictions can be expressed by the ranking of faithfulness constraints earmarked for prosodic constituents as is shown in (16). (16) a. FAITH( V) >> *NUC/i(ω) >> *NUC/u(ω) >> *NUC/ɛ (ω) >> *NUC/ɑ (ω) b. *NUC/[-high](Σ) >> *NUC/i(Σ) >> FAITH( V) >> *NUC/u(Σ) c. *NUC/[-high](σ) >> *NUC/u(σ) >> FAITH( V) >> *NUC/i(σ) >> *NUC/ə(σ) A rather drastically reduced potential for contrast is also seen in German inseparable verbal prefixes, which are non-cohering and exhibit vocalic neutralizations similar to those observed in the English suffixes. Monosyllabic prefixes contain schwa when the syllable is open or ends in /n/, /l/, or /ʀ/ (cf. 17a). Other monosyllabic prefixes contain /ɪ/ or, when a labial follows, /ʊ/ (cf. 17b). Disyllabic prefixes differ in that they form feet and require a high, preferably rounded vowel in the strong syllable. Unrounded vowels are associated with significantly lower productivity, indicated with the symbol “(†)” in (17c). (The productive prefix /dʊʀç/ counts as disyllabic due to its consonant cluster, whose final member is best represented as the onset of a syllable with an empty nucleus (cf. Harris and Gussman 2002). The evidence for this syllabic organization concerns the foot structure in longer words ending in /ʀç/: stable final main stress as in Monárch ‘monarch’ indicates the presence of a final trochaic foot (cf. (mo.(náʀ.ç0̸) Σ) ω), whereas words ending in coronals often have initial stress ((hʊ´ndəʀt) Σ) ω ‘hundred’). The differences to English suffixes indicate then only slight changes in the order of FAITH(V) relative to the anti-association constraints in (16b, c).

11. Word-formation in optimality theory (17) a. /bə/ , /ɡə/ , /əʀ/ , /fəʀ/ , /tsəʀ/ , /ənt/ b. /mɪs/ , /ʊm/ c. /ybəʀ/ , /ʊntəʀ/ , /dʊʀç/ , (†)/vidəʀ/ , (†)/hɪntəʀ/ Additional severe restrictions concern the distribution of consonants. German prefixes which contain a single voiced obstruent are structured such that this obstruent appears in initial position followed by schwa (e.g., /bə/, /ɡə/). This generalization is captured by formulating a grammar which for any of the inputs containing no consonant other than for instance /b/ as in (18a) selects /bə/ as the optimal output. A fragment of such a grammar is sketched in (18b). (18) a. {/bə/, /əb/, /b/, /ɪb/, /bi/, /ubu/, /əbə/} → /bə/ b. *NUC/ PLS, *OBS[+VOICE] COD, FAITH(C) >> FAITH( V), DEPENDENCE, MAXIMALITY The need to faithfully preserve the consonant (FAITH(C)) and organize it prosodically, along with high-ranking markedness constraints against associating /b/ with the nucleus (due to its low sonority, cf. (7b)), or with the coda (due to its voicedness, cf (3b)) force this segment into the syllable onset. A following vowel is needed to satisfy a constraint requiring every syllable to have a nucleus. Schwa is the optimal choice here as it exhibits the minimal structure needed to provide a syllable nucleus, without gratuitously violating either faithfulness or markedness constraints prohibiting various sorts of structure. Full vowels appear only when necessary to satisfy higher-ranking constraints, including a constraint prohibiting schwa in the strong syllable of a foot or in a syllable closed by a consonant other than {/n/, /l/, /ʀ/}. While /b/ is forced into the onset, sonorants seem to avoid this position. This avoidance indicates that high-ranking anti-association constraints militating against sonorants in the onset (cf. 7a) dominate the faithfulness constraint LINEARITYP, which requires all precedence relations among phonemes in the input to be preserved in the output. The prefixes /ʊm/ or /əʀ/ show that the relevant anti-association constraints also dominate the constraint ONSET, which prohibits onsetless syllables. (19) a. {/m/, /mu/, /mə/, /əm/, /ʊm/} → /ʊm/ b. {/ʀəf/, /fʀə/, /fʀ/, /fəʀ/, /ʀf/, /əʀəf/, /ʀɪf/, /ʀaf} → /fəʀ/ c. *ONS/ʀ >> *ONS/ l >> *ONS/ NAS >> LINEARITYP >> *ONS/ FRC >> *ONS/ PLS Apparent exceptions indicate the impact of higher-ranking markedness constraints. Hence the linear order of the segments in the prefix /mɪs/ is due to the undominated constraint against /s/ in word-initial position in German, along with the constraint FAITH(C), which ensures that input /s/ surfaces as /s/, rather than /z/ or /ts/. The necessary organization of /s/ in the coda forces /m/ into the onset. Arranging /t/ after /n/ in the prefix /ənt/ allows for both consonants to share a place node ([+alveolar]), thereby reducing markedness. The preference for /ənt/ over /tən/ or /nət/ is further evidence for the low ranking of ONSET in the grammar of German verbal prefixes.

169

170

I. Word-formation as a linguistic discipline An investigation of the disyllabic prefixes in (17c) shows that here, too, the sequence of all segments appears to be determined by the markedness constraints mentioned above. Again, complex onsets are prohibited (e.g., */dʀʊç/), all voiced obstruents are restricted to onset positions and /h/ is moreover restricted to foot-initial position. The organization of /t/ in unter- and hinter- as the onset of the second syllable serves to satisfy the constraint against sonorants in onset position. The consonant /n/, rather than /ʀ/, precedes /t/ to allow both segments to share a place node. Historically, inseparable prefixes which do not conform to these constraints have ceased to be productive and appear only in a few fossilized words (e.g., /ɔp/ as in obsiegen ‘to win’ /fɔl/ as in vollbringen ‘to accomplish’). Loan prefixes which happen to satisfy the constraints in question may be more prone than others to join the class of inseparable verbal prefixes and become productive in German (e.g., /zʊp/ as in suboptimieren, /hypəʀ/ as in hyperoptimieren, /dɪs/ as in disoptimieren). In OT, the highly reduced potential for contrast observed in these affixes vis-à-vis stems can be captured as in (20), where specific markedness constraints (e.g., the constraint against complex onsets, the constraint against sonorants in onset position) are sandwiched between faithfulness constraints linked to stems versus affixes. (Other markedness constraints, including *OBS[+VOICE] COD, are undominated in German.) As a result, contrasts concerning the linear order of phonemes such as /ɡaʀtən/ vs. /ɡantəʀ/ vs. /ɡʀantə/ vs. /ɡnaʀtə/ vs. /tʀaɡən/ are possible in stems, but not in affixes. (20) LINEARITYP-STM >> MARKEDNESS >> LINEARITYP-AFF The order between the two faithfulness constraints illustrates the metaconstraint ranking FAITH ROOT >> FAITHAFF proposed by McCarthy and Prince (1995). While the observation that affixes tend to exhibit less phonemic contrast than stems is not new, an analysis in terms of interacting faithfulness and markedness constraints yields novel insights into the extent of the neutralizations. The generalization that neither the order among the consonants nor the choice among the vowels play (much of) a role in the distinction among German prefixes can be expressed concisely in an OT analysis, where various aspects of faithfulness such as FAITH(C), FAITH( V), and LINEARITY can be referred to. A generalization not yet expressed by the ranking FAITH STM >> FAITHAFF concerns the restriction of the neutralization patterns illustrated above to non-cohering affixes. Reference to prosodic constituent structure is already explicit in the account of vocalic neutralizations described in (16) but there is evidence that all aspects of affixal neutralizations are determined by their prosodic organization. For instance, English cohering suffixes differ from non-cohering suffixes in that they exhibit the full range of vowels (/ɛt/ -ette, /ɛit/ -ate, /ɪfai/ -ify, /ɔɹi/ -ory), rather than only high vowels. Cohering suffixes also exhibit the full range of contrast among consonant phonemes, including affricates (/ɪdʒ/ -age), voiced fricatives (/ɪv/ -ive, /iz/ -ese, /aiz/ -ize, /ɪzəm/ -ism), non-coronal voiced plosives (/əbəl/ -able), as well as clusters with non-coronal obstruents in peripheral position (/ɛsk/ -esque), none of which occur in non-cohering suffixes. These generalizations can be captured by replacing FAITH STM with FAITH ω, where the subscript “ω” refers to the entire phonological word, including the cohering affixes. Indeed there is evidence that cohering affixes may exhibit faithfulness effects not seen in stems. In English, the suffix /i/ -ee is associated with stable main stress (cf. (21a)), thereby differ-

11. Word-formation in optimality theory

171

ing from stem-final /i/ (cf. (21b)), which shows the affinity between low sonority nuclei and stresslessness mentioned above. (21) a. (dɹæf(ti)Σ)ω (tɹɛi(ni)Σ)ω (tɹɐst(ti)Σ)ω (pɛi(i)Σ)ω (flɹ(ti)Σ)ω (ɹɛn(ti)Σ)ω





b.

((kɔ´fi)Σ)ω ((tɹə´uki)Σ)ω ((típi)Σ)ω (təu(pi)Σ)ω ~ ((tə´upi)Σ)ω (bæn(ʃi)Σ)ω ~ ((bæ´nʃi)Σ)ω (ɹu(pi)Σ)ω ~ ((ɹúpi)Σ)ω





Additional English suffixes associated with stable main stress include -eer (cf. suffixed mòuntainéer versus simplex càvalíer ~ cávalìer), -ese (Jàpanése versus díocèse) -ette (kìtchenétte versus sìlhouétte ~ sílhouètte) and -esque (pìcturésque versus óbelìsk). Similar cases abound in other languages. In German, final /i/ is associated with stable main stress when representing the feminine suffix -ie as in (22a), but appears in the weak syllable of the foot otherwise (22b). (mɑ(ɡi) Σ)ω (22) a. [mɑɡ]ST [í]SF-FEM (teʀɑ(pi) Σ)ω [teʀɑp]ST [í]SF-FEM [melankol]ST [í]SF-FEM (melanko(li) Σ)ω b. [tsɔmbi]N-MSC [ ʃpɑɡɛti]N-PL [hɑʀɑkiʀi]N-NEU

((tsɔmbi) Σ)ω ( ʃpɑ(ɡɛti) Σ)ω (hɑʀɑ(kiʀi)Σ)ω



‘magic’ ‘therapy’ ‘melancholy’



‘zombie’ ‘spaghetti’ ‘harakiri’

Similar effects can be observed for syllable structure. In French, final /i/ is regularly organized in the syllable margin when a syllabic vowel precedes (cf. /bui/ [buj] ‘face’, /œi/ [œj] ‘eye’)), but consistently forms a nucleus when representing a suffix. As a result, the simplex /abɛi/ ‘bee’ ends in a glide ([a.bɛj]) while the suffixed word [abe]STM [i]SF-FEM ‘abbey’ ends in a syllabic vowel ([a.be.i]) (cf. the feminine suffix /i/ in boulangerie ‘bakery’, charcuterie ‘butcher shop’). The presence and stability of the marked structures in question could be captured by positing high-ranking faithfulness constraints which target the prosodic structure of affixes, specifically the association of certain phonemes with the syllable peak (e.g., French [i]SF-FEM) or the stress peak (e.g., German [i]SF-FEM). (23) FAITH( PROS)AFF/ω >> MARKEDNESS >> FAITHω >> MARKEDNESS >> FAITHAFF The ranking in (23) captures the restriction of “deviant” prosodic structure as in (21a) and (22a) to cohering affixes. The prosodic organization of other affixes is determined entirely by the respective phonemes, which themselves are severely restricted due to the low ranking of FAITHAFF in (23). To the extent that the relevant generalizations require reference to fully specified prosodic output structures they could hardly be captured in traditional generative frameworks based on rules for combining and modifying individual morphemes.

172

I. Word-formation as a linguistic discipline

6. Faithfulness in base-derivative relations Up to now, reference to faithfulness constraints has been aimed at describing potential contrastiveness, with a focus on relevant differences between stems and affixes. There is evidence for a second type of faithfulness in morphology, which concerns the relation between a base, that is, a word forming the input to word-formation, and the derived form. To illustrate the need to distinguish the relevant constraints consider first the potential contrastiveness between front and back low vowels in English before /m/ in (24a), compared to the restriction to the back low vowel before syllable-final /ɹ/ illustrated in (24b). Reference to the syllable-final position is important as /æ/ does appear before heterosyllabic /ɹ/ (cf. /kæɹi/ (Kahn 1980)). (24) a. /kɑm/ − /kæm/

b.

/kɑɹ/ − */kæɹ/

The distribution illustrated in (24) is a straightforward case of neutralization, captured in OT by ranking a specific markedness constraint “*æɹ)σ” above a faithfulness constraint FAITH(V), which requires feature values in output vowels to match those given in the input. (25)

/kæɹ/

*æɹ)σ

/kæɹ/

FAITH(V)

*

☞ /kɑɹ/

*

The data motivating the second type of faithfulness are given in (26), where clipped forms such as /sæɹ/ retain the vowel quality of their base despite deviating from the regular patterns. Note that the base Sarah conforms to regular phonotactics as /æ/ is not followed by tautosyllabic /ɹ/. (26) /sæɹə/ → /sæɹ/ /læɹi/ → /læɹ/ The generalization that /sæɹ/ is not a possible word in English, except when related to a base in which /æ/ is followed by (heterosyllabic) /ɹ/, is captured by a constraint FAITH BD, which requires structural identity in the relation between a derived form and its base. Ranking this constraint above the relevant markedness constraint and faithfulness to the input (henceforth referred to as FAITH IO) introduced earlier captures the conditions under which /sæɹ/ is grammatical in English (cf. McCarthy and Prince 1995; Benua 1995). (27)

/sæɹə/

FAITH BD

☞ /sæɹ/ /sɑɹ/

*æɹ)σ *

*

FAITH IO

11. Word-formation in optimality theory

173

The concept of faithfulness in the relation between a given base word and its output accounts for all so-called cyclic effects, including the observation, first noted by Brame (1974), that cyclicity crucially involves reference to (surface) words rather than smaller units such as roots. The observation that preservation of structure does not necessarily concern the entire base word, but can target specific aspects thereof, can be captured through positional faithfulness. For instance, the retention of the first vowel of /sæɹə/ in (26), as opposed to the last, is achieved by linking the relevant faithfulness constraint to a prominent position such as the stressed syllable or the first syllable in the phonological word. Recognizing both FAITH IO and FAITH BD allows for intricate constraint interactions. Empirical motivation for such interactions comes from putative “rule ordering paradoxes” in reduplication (cf. McCarthy and Prince 1995, 1999; Russell 1997).

7. Unmarkedness in hypocoristics The cases of word-formation illustrated in (28) differ from typical affixations in that the strings preceding the vocalic endings are not necessarily stems. (This is assuming a notion of stem consisting of a word with recognizable affixes removed.) In addition, the formations are atypical in that the endings exhibit few syntactic restrictions, at times with no apparent effect on the category of the derived word (cf. Schneider 2003: 87). (28) a. weird]A nuts]A sick]A derelict]N aggressive]A ammunition]N

→ → → → → →

weirdo]N nutso]A sicko]A,N dero]N aggro]A ammo]N

b.

cook]V good]A short]A comfortable]A vegetable]N Australian]A,N

→ → → → → →

cookie]N goody]Intj. shorty]N comfy]A veggie]N Aussie]A,N

The final vowels in the formations in (28) appear to have primarily a pragmatic function, acting as “familiarity markers” characterized by an informal tone and social closeness (Quirk et al. 1985: 1584). Indeed, it is not just the vowels that serve this function but the entire shape of the outputs, including the trochaic foot structure. Formations characterized by a fixed phonological shape are known as “templatic” (McCarthy and Prince 1986), seemingly involving the extraction of some specific part of the structure of the base, which may then combine with additional markers. This procedure is known as “positive prosodic circumscription” and is also illustrated by the formations in (26). From an OT perspective, the formations in (28) can be described without assuming these separate steps but instead indicate dominance of various markedness over faithfulness constraints (cf. McCarthy and Prince 1993; Benua 1995; Lappe 2007). Ranking the constraint FOOTBIN, which requires binary feet, over FAITHBD captures the fact that each output in (28) consists of a single trochaic foot. Reference to the foot also accounts for the choice of the “familiarity markers” as high front vowels and schwa are the only phonemes which can occupy the nucleus of the weak syllable in a foot in English. This restriction is revealed by the potential of flapping of intervocalic /t/ before just these vowels in American English (cf. /sɪ´[ɾ]i/ , /dɛ´i[ɾ]ə/ , /ɔ´[ɾ]əu/ ), as

174

I. Word-formation as a linguistic discipline opposed to necessary aspiration before all other vowels (cf. /lɑ´[th]ɛ̀i/ , /kɐ´u[th]ɐ̀u/ , /tú[th]ù/ , /iú[th]ɑ̀ / ). This difference is indicated in the structures in (29), where /t/ occurs foot-internally, versus the structure depicted in (9b) above, where /t/ occurs foot-initially. (29)

a.

b.



Non-occurring “familiarity markers” such as /u/ or /ɑ/ can accordingly be ruled out by independently motivated markedness constraints prohibiting their association with the nucleus of the weak syllable within a foot. In that position, tense /i/, the least sonorous vowel, will emerge as the least marked vowel, as is reflected in the high productivity of /i/-formations in English. The choice of /əu/ as in (28a) presupposes some input specification, which also accounts for the specific (derogatory) meaning associated with those formations. Only when /u/ is present in the input independently, as part of the stem, will the relevant negative connotations be absent (cf. amm/u/nition > amm/əu/ ). Focusing now on the formation of English nicknames it appears that markedness constraints determine not only their overall shape but also the “extraction sites”. The formations in (30a, b) indicate satisfaction of FAITH(ω(σ)BD, which requires correspondence between the respective first syllables, regardless of stress. If, however, the wordinitial syllable is both unstressed and begins with a vowel, there is a tendency for the respective stressed syllable to be “extracted” instead (cf. (30c)): (30) a. (Ábra)Σham (Álfred)Σ (Ésther)Σ (Ágnes)Σ (Édward)Σ (Élspeth)Σ c. Au(gústus)Σ An(tónius)Σ E(mmánuel)Σ E(lísa)Σbeth E(zéki)Σel Antoi(nétte)Σ

→ → → → → →

(Ábe)Σ (Álfie)Σ (Éssie)Σ (Ággie)Σ (Éddy)Σ (Élspie)Σ

→ → → → → →

(Gússie)Σ (Tóny)Σ (Mánny)Σ (Lízzy)Σ ( Zéke)Σ (Nétty)Σ

b.

So(phía)Σ Ma(ría)Σ Ber(nárd)Σ Vic(tória)Σ To(bías)Σ Chris(tína)Σ

→ → → → → →

(Sóphie)Σ (Máry)Σ (Bárney)Σ (Vícky)Σ (Tóby)Σ (Chríssie)Σ

11. Word-formation in optimality theory

175

From an OT perspective the data in (30a, b) show that FAITH(ω(σ)BD is more important than FAITH(Σ(σ)BD, another positional faithfulness constraint which requires that footinitial syllables in the derived form and the base correspond. The data in (30c) indicate that the simultaneous violations of both ONSET and FAITH( Σ)BD would be worse than violating FAITH(ω(σ)BD. This condition can be captured by allowing constraints to conjoin, such that the conjunction is violated whenever each of its components is violated locally, that is, by a single piece of structure (Smolensky 1993, 2006). Constraint conjunction captures the circumstance that a given structure may be tolerated despite certain flaws, unless it is flawed in too many regards. The conjunction of ONSET and FAITH(Σ(σ)BD has the desired effect of eliminating the candidate /ɑ´ɡi/ for /ɑɡʌ´stəs/, allowing its rival /ɡʌ´si/ to win (cf. (31a)). In (31b, c) the optimal nickname starts with the same syllable as its base since the high-ranking constraint conjunction is satisfied. This is because conjoined constraints are violated only when each individual constraint is violated. (31) a.

ɔ(ɡʌ´stəs) Σ (ɔ´ɡi) Σ ☞ (ɡʌ´si) Σ

b.

{ONSET, FAITH( Σ) BD}

FAITH (ω(σ)BD

*ONSET

* *

kɹɪs(tínə) Σ (tíni) Σ

*

☞ (kɹɪ´si) Σ

c.

FAITH(Σ(σ) BD

*

(ɛ´lɪ) Σ (nɔ̀ɹ) Σ ☞ (ɛ´lɪ) Σ (nɔ´ɹi) Σ

* *

The analysis in (31) does not account for the fact that several of the non-winning candidates are attested as well. Such variation, to the extent that it is associated with distinct groups of speakers, indicates differences in the individual grammars, expressed in terms of different constraint rankings in OT. The important prediction here concerns the limits imposed by the theory on possible grammars. For instance, it is predicted that “extraction” of non-initial syllables in consonant-initial words such as /tíni/ from /kɹɪstínə/ presupposes extraction in vowel-initial words, but not vice versa. It is further predicted that extraction of vowel-initial strings such as /ɔ´ɡi/ entails the extraction of consonantinitial strings, but not vice versa. These predictions follow from the existence of the markedness constraint ONSET, which can have the effect illustrated in (31), while no ranking of known constraints would produce the opposite effect. It is this sort of explanation and predictive power which distinguishes OT from approaches based on prosodic circumscription rules (“Pick the word-initial syllable, unless …”) and can hardly be mimicked by those. A grammar where markedness constraints freely interact with FAITH BD constraints can capture intricate conditions on word-formation rules which appear outside of the realm of rule-based approaches. Consider the modifications in the nicknames based on female names in (32), which indicate satisfaction of the anti-association constraint *ONS/

176

I. Word-formation as a linguistic discipline ɹ. In some cases /ɹ/ appears to be replaced by /l/, in others by various obstruents, which have a correspondent in the portion following /ɹ/ in the respective base words. (32) a. Sarah Mary Caroline Laura

→ → → →

Sally Molly Callie Lolly

b.

Teresa Harriet Florence Charity

→ → → →

Tessy Hatty Flossie Chattie

The consistent preservation of word-initial /ɹ/ as in Rachie from Rachel shows that *ONS/ɹ is dominated by FAITH(ω(σ)BD (cf. 33a). The choice of the specific underlined obstruents in the nicknames in (32b) indicates the ranking of FAITH(C)BD, which requires complete identity of a consonant in the base and the derived form, above CONTIGUITY (see (4f)). (33) a.

ɹɛ´itʃəl

FAITH(ω(σ) BD

*ONS /ɹ

☞ (ɹɛi.tʃi) Σ *

sæ´ɹə ´ .ɹi) Σ (sæ

*

´ .li) Σ ☞ (sæ

c.

CONTIG BD

*

(lɛi.tʃi) Σ

b.

FAITH(C)BD

*

hæ´ɹiət (hæ´.ɹi) Σ

*

(hæ´.li) Σ

*

´ .ti) Σ ☞ (hæ

*

The choice of Sally as the optimal nickname based on Sarah, rather than *Sanny or *Satty, which have less sonorous and hence better onsets, indicates faithfulness in basederivative relations. Ranking the constraint FAITH(±APP)BD, which requires identical specifications for the feature [±approximant], among the relevant anti-association constraints accounts for the selection of the candidate containing /l/. (34)

´ ɹə sæ

*ONS /ɹ

´ .ɹi) Σ (sæ

*

´ .ni) Σ (sæ ☞ (sæ´.li) Σ

FAITH(±APP) BD

*ONS/l

*

*ONS/NAS

* *

Again, the grammars in (33) and (34) express limits on what is possible. The top ranking of ONS/ɹ among the relevant anti-association constraints in (7a) accounts for the fact that the “replacements” affect precisely /ɹ/, rather than one of the less sonorous consonants. While faithfulness constraints such as FAITH(±APP) BD can have the effect of limiting distinctness among correspondents (e.g., the replacement of /ɹ/ by /l/, which is also

11. Word-formation in optimality theory

177

[+approximant]) there are no constraints which would have the effect of maximizing distinctness (e.g., the replacement of /ɹ/ by /p/). While the high ranking of positional faithfulness as in (33) can ensure the preservation of /ɹ/ in word-initial position there is no corresponding constraint which could ensure its preservation only in non-prominent positions.

8. Allomorphy and gaps The term allomorphy refers to the occurrence of phonemically distinct morphemes, whose distribution is determined by the linguistic context alone and which consequently do not differ in meaning. Consider the alternation between the English adjectival suffixes /əl/ and /əɹ/ in (35), the latter occurring only in combination with stems ending in /l/. (35) a. /tɹɑib-əl/

/ænɪkdəut-əl/ /kəust-əl/

b.

/pəul-əɹ/

/pɹəutəkɑl-əɹ/ /kɑnsəl-əɹ/

Phonologically conditioned allomorphy as in (35) is analysed by ranking a markedness constraint (i.e. *CiVCi “The flanking of nuclei by identical phonemes is prohibited”) above faithfulness constraints. The constraint FAITH(±APP) BD again has the effect of choosing the minimally different form /əɹ/ in (36b). Locating the “repair” to the affix, rather than the stem, can be expressed by ranking faithfulness to the segmental structure of the stem above that of the affix. As a result, violations of *CiVCi are tolerated within stems (cf. parallel, lull), but not in stem-affix combinations. (36) a.

[tɹɑib] STM+[əl] AFF

FAITH(STM) BD

*CiVCi

FAITH(±APP) BD FAITH(AFF) BD

☞ /tɹɑib-əl/ /tɹɑib-əɹ/

b.

*

[pəul] STM+[əl] AFF /pəul-əl/ /pəur-əl/

* *

/pəul-ən/

*

☞ /pəul-əɹ/

c.

*

[pæɹəlɛl] STM ☞ /pæɹəlɛl/ /pæɹəlɛɹ/

* *

The assumption of FAITHBD constraints brings to light the close relatedness between allomorphy and gaps. Hence the avoidance of the candidate /pəul-əl/ in adjectival -alsuffixation due to the violation of *CiVCi is mirrored in nominal -al-suffixation in (37).

178

I. Word-formation as a linguistic discipline The cases differ in their response to the problem: in one case the suffix is “repaired” (cf. (35b)), in the other suffixation is avoided altogether (cf. (37b)). (37) a. /uɪθdɹɔ-əl/ /ɐphiv-əl/ /bɪtɹɛi-əl/



b.

/fəlfɪl/ + /əl/ → 0̸ /bɪɡɐil/ + /əl/ → 0̸ /ɪnɹəul/ + /əl/ → 0̸

An OT grammar concisely captures the similarities and differences between the two suffixes. Both are associated with the undominated markedness constraint *CiVCi and undominated FAITH(STM) BD. They differ in that faithfulness to affixes is violable for adjectival, but not for nominal, -al-suffixation (for more discussion of gaps see Raffelsiefen 2004; Rice 2007). Returning to the account of allomorphy in (36), the analysis entails neither the assumption of several allomorphs in the input nor a process whereby underlying /əl/ in the suffix is changed to /əɹ/ in the output. Rather it expresses static constraints concerning the structural identity in the relation between input and output forms. Deviations from identity in this relation are conditioned by higher-ranking phonological markedness constraints and are always minimal. From this perspective many cases of allomorphy which have been labelled “suppletive” in rule-based frameworks could be reanalysed in terms of constraint interaction. Consider the complementary distribution between the third singular masculine pronominal clitics /u/ and /h/ in Moroccan Arabic, where /u/ appears after consonants and /h/ appears after vowels (Harrell 1962:136): (38) a. /menn-u/ /ktab-u/ /ʃaf-u/

‘from − him’ ‘book − his ’ ‘he saw − him’

b.

/mʕa-h/ /xtʕa-h/ /ʃafu-h/

‘with − him’ ‘error − his’ ‘they saw − him’

The data are consistent with the assumption of /u/ in the input, which does not surface after vowels because the hiatus would violate ONSET. Assuming that epenthesis or deletion are not allowed, ONSET can be satisfied only when /u/ corresponds to a non-vocalic segment. /h/ is a promising candidate as it agrees with vowels in being [+approximant] (Clements 1992) and does not introduce place or manner features at odds with those of /u/. (Historically, the /u/-/h/-allomorphy goes back to the single morpheme /hu/.) (39) a.

[menn] STM+[u]CLT

FAITH (STM) BD

ONSET

FAITH(±APP) BD

FAITH(AFF) BD

☞ /menn-u/ /menn-h/

b.

*

[mʕa] STM+[u] CLT /mʕa-u/ /mʕa-k/ ☞ /mʕa-h/

* * *

11. Word-formation in optimality theory

179

Assuming the possible representation of input morphemes as sets of allomorphs the correct distribution could also be captured with reference to additional markedness constraints. That is, given the listed allomorphs {/u/, /h/} in the input, the candidate /mennh/ would no longer violate faithfulness, but could be eliminated as a result of violating *COMP-CODA (Complex codas are prohibited). An analysis of this sort accounts for the allomorphy in cases like the passive marking in Turkish, where /n/ after vowel-final stems corresponds to /il/ after consonant-final stems (Underhill 1976: 332). (Additional alternations are due to vowel harmony constraints and will be ignored here.) (40) a. /oku-n/ /de-n/

‘to be read’ ‘to be said’

b.

/kaibed-il/ /ior-ul/

‘to be lost’ ‘to be tired’

In this case it is presumably not possible to account for the allomorphy by positing a single underlying morpheme. However, the phonological conditions determining the distribution of the listed allomorphs {[n], [il]} can be captured by ranking the markedness constraints ONSET and *COMP-CODA above faithfulness constraints (cf. tableau (41)). (41)

[oku] STM+{[n], [il]} /oku-il/

ONSET

*COMP-CODA

FAITH(AFF) BD

*

/oku-l/

*

☞ /oku-n/ [kaibed] STM+{[n], [il]} ☞ /kaibed-il/ /kaibed-n/

*

Not all cases of allomorphy lend themselves to an analysis based on listed allomorphs. For instance, positing both allomorphs /əl/ and /əɹ/ in the input for English adjectival -al-suffixation would cause a problem concerning the elimination of the faithful and phonologically wellformed, yet ungrammatical, candidate /tɹɑib-əɹ/. To express the default status of the allomorph /əl/ it would be necessary to specify priority relations among the listed allomorphs (i.e. {/əl/ < /əɹ/}). This necessity may support the alternative analysis in (36), which is based on a single underlying morpheme and requires no ordering. In general, decisions concerning underlying forms are based on a principle called “lexicon optimization” in OT (cf. McCarthy 2002: 76−78). This principle essentially favors maximal similarity between surface and underlying forms, but disfavors listed allomorphs. The questions of whether or not allomorphs must be listed and possibly ordered are decided based on possible constraint permutations: variation which cannot be captured in terms of constraint interaction by assuming a single underlying form requires listed allomorphs. The evidence for listed and ordered allomorphs presented in the OT literature appears to be limited to inflectional morphology (cf. Mascaró 2007). Similarly the evidence for assuming candidates consisting of entire paradigms in OT, rather than individual word forms, concerns only inflectional morphology (Raffelsiefen 1995; McCarthy 2005). Such theory-internally motivated formal differences, be it multi-

180

I. Word-formation as a linguistic discipline ple allomorphs in underlying forms or multiple forms within single candidates, may shed light on the distinction between word-formation proper and inflectional morphology.

9. Anti-faithfulness or anti-homophony constraints In section 7 it was demonstrated how reference to markedness constraints such as FOOTBIN can account for the invariance in the shape of output forms in formations like English comfy (← comfortable). A more recalcitrant problem concerns cases known as truncation, where a fixed part of the base appears to be removed in derivation, while the remaining part may vary in shape. Gender distinctions in a subclass of French adjectives, manifest in the presence versus absence of final consonants, may illustrate this (cf. Bloomfield 1933: 217). The variance in the relevant consonants seen in (42a) rules out the assumption of a regular suffix marking the feminine gender. The data in (42b) illustrate the variance in the shape of the “truncated” masculine forms. (42) a. [ ʃod] F − [ ʃo] M [blɑ˜ʃ ] F − [blɑ˜] M [ɡʀɑs] F − [ɡʀɑ] M [sul] F − [su] M b. [ot]F − [o] M [fɔʀt] F − [fɔʀ] M [pətit] F − [pəti]M [elɛɡɑ˜t] F − [elɛɡɑ˜] M



‘warm’ ‘white’ ‘fat’ ‘drunk’



‘high’ ‘strong’ ‘small’ ‘elegant’

It has been recognized that the sort of modifications seen in truncated forms cannot in general be captured with reference to markedness constraints (Horwood 1999). A possible solution lies in the recognition of “anti-faithfulness” constraints, which militate against identity in the relation between a derived form and its base (cf. Alderete 2001; Horwood 1999). The idea is that for every faithfulness constraint there is a corresponding “inverted” anti-faithfulness constraint, which penalizes the occurrence of identity in the relevant respect (cf. (43a vs. b)). (43) a. MAXIMALITY(C) BD b. ANTI-MAXIMALITY(C) BD

Every consonant in the base has a correspondent in the derived form. (No deletion of consonants!) Not every consonant in the base has a correspondent in the derived form. (Deletion of at least one consonant!)

Ranking ANTI-MAXIMALITY(C) BD above other faithfulness constraints has the desired effect of “deleting” exactly one consonant. Properly localizing truncation to the wordfinal position is achieved by ranking positional faithfulness constraints preserving the prominent word-initial position together with CONTIGUITYBD above ANTI-MAXIMALITY(C) BD.

11. Word-formation in optimality theory (44)

[pətit] STM

FAITH(ω(σ) BD

181

CONTIGUITYBD

ANTI-MAX(C) BD

/pətit/ /ətit/

FAITH (STM) BD

* *

/pəit/

*

☞ /pəti/

*

/pət/

**

Reference to anti-faithfulness constraints concerning specific features could account for “devoicing” of English past tense forms in the small class of verbs illustrated in (45): (45)

[bɪld] STM

FAITH(ω(σ)BD

ANTI-FAITH ([+VOICE]) BD

/bɪld/ /pɪld/

FAITH (STM) BD

* *

☞ /bɪlt/

*

The strongest evidence for anti-faithfulness concerns so-called “polarity” rules, which require a switch of feature values. The basic idea can be illustrated with plural formation in a subclass of nouns in Luo, a Western Nilotic language (Gregersen 1974). Unlike English, where derived forms differ from their bases only in one direction (e.g., [bɪld]PRES − [bɪlt] PAST, but not [mɛlt] PRES − *[mɛld] PAST), the derivations in Luo involve change in both directions. Final voiceless obstruents in singular forms correspond to voiced obstruents in plural forms (e.g., [ɡɔt] SG ‘mountain’ − [ɡɔd-ɛ] PL ‘mountains’), and vice versa (e.g., [hiɡ] SG ‘year’ − [hik-e] PL ‘years’). These patterns are concisely expressed in the grammar in (46). (46)

[ɡɔt] STM+[ɛ]AFF /ɡɔt-ɛ/

FAITH(ω(σ) BD

ANTI-FAITH ([±VOICE]) BD

FAITH(STM) BD

*

☞ /ɡɔd-ɛ/

*

[hiɡ] STM+[e]AFF /hiɡ-e/ ☞ /hik-e/

* *

Problematic aspects of the notion of anti-faithfulness concern its restriction to basederivative relations. (Anti-faithfulness in input-output relations has no empirical motivation and would explode the space of possible grammars.) Another question concerns the observable restriction to maximally one anti-faithfulness effect in a word-formation rule (Horwood 1999: 20). A solution might be to adopt a general constraint DISTINCT(STM), which militates against homophonous stems in morphologically distinct word forms (cf. Urbanczyk 1998). The restriction to specific segments or specific features would then

182

I. Word-formation as a linguistic discipline be accommodated by ranking the relevant faithfulness constraints above DISTINCT(STM). Objections to this option concern cases where different “anti-faithfulness effects” are attested in a single language (Horwood 1999). However, the possibility that distinct word-formation rules may associate with distinct constraint rankings must be granted independently, as is shown for instance by the patterns associated with nominal versus adjectival -al-suffixation in English (cf. (35), (37)). The possible existence of different “anti-faithfulness effects” is expected then, as long as different cases of word-formation are concerned.

10. Criticism OT grammars consist of ordered markedness and faithfulness constraints. There is consequently a prediction that all violations of faithfulness, be it allomorphy or infixation, are caused by higher-ranking markedness constraints and must accordingly be reflected in some sort of improvement of phonological output structure. This prediction has been claimed to be refuted by cases where phonologically conditioned allomorphy or infixation are said to be clearly non-optimizing. Since one of the main characteristics distinguishing OT from other theories of word-formation is its predictive power, the validity of such counter-examples has to be examined in detail for each individual case. The investigation of nominalization in the Austronesian language Leti, which has been heralded as a prime example of allegedly non-optimizing allomorphy- and infixation patterns (Blevins 1999; Yu 2007), shows that at least some of the criticism is unfounded. The regular phonologically conditioned variation in Leti nominalization is illustrated in (47), where the surface affixal material appears in brackets. Before vowel-initial stems, a prefix [nj] occurs as in (47a) while [nja] precedes stems starting with a consonant cluster as in (47b). The data in (47c, d, e and f) illustrate the generalization that affixal material does not precede verb stems starting with a single consonant. Instead the allomorph [nj] is infixed after initial obstruents (47c) and [j] after initial sonorants (47e). The glide does not occur before high vowels, leaving [n] as the only marker in cases like (47d). Nominalization is accordingly not marked at all in verbs with an initial sonorant before a high vowel (47f). (47) a. b. c. d. e. f.

[(n)j]atu [nja]mnèsa k[nj]asi t[n]utu r[j]esi ruru

‘knowledge’ ‘equality’ ‘digging’ ‘support’ ‘victory’, d[j]avra ‘trembling’

‘cut’

The glide in the affixes represents phonemic /i/ and indicates syllabification in onset position when another vowel follows. The allomorphy can then be shown to be consistent with the assumption of a single underlying morpheme /ni/, such that all variants can be captured in terms of markedness constraints dominating FAITH(AFF) BD. The insertion of [a] in (47b) is optimizing in that this vowel appears when necessary to avoid the syllabification of /i/ in nucleus position. This condition is captured by ranking the anti-associa-

11. Word-formation in optimality theory

183

tion constraint *NUC /i(ω) >> above DEPENDENCE(AFF) BD, which prohibits the insertion of segments in affixes (cf. (4d)). The segment [a] emerges as it is the most sonorous and hence optimal vowel to fill the nucleus position in the relevant prosodic context (cf. (16a)). Similarly, the “deletion” of affixal material seen in (47d, e, f) indicates the ranking of MAXIMALITY(AFF) BD below markedness constraints. The absence of /i/ as in (47d, f) satisfies a constraint against adjacent high vowels while the absence of /n/ after sonorants as in (47e, f) satisfies a constraint against onset clusters with even or falling sonority. Reference to sonority may also capture the absence of /n/ in cases like d[j]avra ‘cut’. This is because the segment represented with the symbol [d] does not precede other consonants and hence patterns with liquids rather than plosives in Leti (cf. van Engelenhoven 1995: 53). (Words like /rpa:ri/ ‘they pay’ are not counter-examples as /r/ ‘they’ is a clitic.) In sum, the allomorphy shown in (47) is entirely optimizing, mostly indicative of markedness constraints concerning sonority and syllable structure. Turning now to the position of the affixes, it has been claimed that the sound patterns resulting from infixation in Leti are “exactly opposite of those predicted by OT” (Blevins 1999: 384). Whereas simple prefixation in cases like /ni + kasi/ would result in a maximally unmarked CV pattern, infixation as in /kniasi/ has been argued to produce marked consonant and vowel clusters “for no obvious prosodic or phonotactic gains” (Yu 2007: 5). However, infixation does serve a purpose as can be demonstrated by comparing the two wellformed cases of prefixation in (48a, b) with the ill-formed case in (48c). The generalization is that prefixation is grammatical only for cohering affixation, where the respective initial syllable spans both affix and stem phonemes (cf. (48a, b)), not for noncohering affixation (48c). (The fact that the affix-initial nasal optionally deletes in cases like [(n)j]atu, but not in [nja]mnèsa, may be due to its occurrence in foot-initial position.) (48) a.

b.

/ni+atu/

c.

/ni+mnèsa/

/ni+kasi/

Whether or not an affix is cohering is determined by the ranking between markedness and alignment constraints in (49) (“*ONS /CC” = “No two consonants in an onset”). (49) ONSET, *ONS /CC >> ALIGN (STEM, L, ω, L) Assuming the distinction in the prosodic organizations in (48) the ungrammaticality of (48c) can be captured by ranking a constraint *CG, which prohibits non-cohering affixation, above the faithfulness constraint LINEARITY M. The observation that *CG is satisfied not by simple integration of the prefixes, but by infixation, indicates the ranking of LINEARITY M below ALIGN (STEM, L, ω, L). The precise site of the infixation could follow

184

I. Word-formation as a linguistic discipline from a constraint on foot structure, as infixation in well-formed k[nj]asi, as opposed to *kanisi, satisfies foot-binarity. While more challenging cases of phonologically conditioned yet allegedly non-optimizing allomorphy or infixation patterns may exist, it appears that much of the criticism is due to the less than full exploration of the analytical possibilities provided by the theory. This raises the question of what exactly those possibilities are, in particular, what the full set of markedness constraints is and on the basis of which criteria they are established. Another question concerns possible limits on constraint orderings, including limits on conjunctions (cf. Smolensky 1993, 2006). Finally the theory must tackle phenomena such as gradability and exceptions. For instance, Leti includes a few nominalizations which deviate from the regular patterns in (47), including cases where the prefix nia precedes stems not starting with a consonant cluster. Also the grammar in (36), as it stands, accounts for the preference of polar to polal in English, but says nothing about the unacceptability of both goalal and goalar as adjectives based on the noun goal. While differences concerning productivity may be amenable to modeling in OT to some extent (cf. “stochastic OT”) it appears that the phenomenon that semantic coherence and stability in base-derivative relations may go hand in hand with entirely unproductive types of exponence (e.g., English hot − heat, strong − strength, hate − hatred) is beyond all generative theories focused on accounting for grammatical output forms. For an approach in which the modeling of morphological analyses to account for relatedness among existing words is separated from a model for morphological synthesis, which then accounts only for productive rules, see Raffelsiefen (2010).

11. Summary OT is a highly restrictive linguistic theory in that it imposes strict limits on how grammars can be modeled. Restrictions on input forms are ruled out and so is reference to intermittent steps in the relation between input and output forms. Instead, grammars assign the optimal output form to any (hypothetical) input form by evaluating candidates in parallel with respect to ordered constraints. There are only two types of constraints, markedness constraints concerning well-formedness in output forms and faithfulness constraints, which require identity (possibly also non-identity) in the relation between inputs and outputs. Regarding word-formation, such a grammar can be shown to account for seemingly drastic modifications of structure, manifest in allomorphy or in varying position of affixes, by assuming a simple IA model. The idea is that all such modifications reflect the domination of faithfulness by markedness constraints. Whether or not such an approach is empirically sound or whether or not the theory can be amended to accommodate all counter-evidence remains to be seen. For now it holds that OT-based analysis have brought to light numerous connections between observable aspects of morpheme shape and universal markedness constraints which appear to yield insight into the fundamental causes of surface deviations from simple IA-structure. Assuming that the relevant connections are not all coincidental it can be said that OT surpasses competing theories with respect to its explanatory potential.

11. Word-formation in optimality theory

12. References Ackema, Peter and Ad Neeleman 2005 Word-formation in optimality theory. In: Pavol Štekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 285−313. Dordrecht: Springer. Alderete, John D. 2001 Dominance effects as transderivational anti-faithfulness. Phonology 18(2): 201−253. Aronoff, Mark 1976 Word-formation in Generative Grammar. Cambridge, MA: MIT Press. Beckman, Jill N. 1999 Positional Faithfulness. An Optionality Theoretic Treatment of Phonological Asymmetries. New York: Garland. Benua, Laura 1995 Identity effects in morphological truncation. In: Jill N. Beckman, Suzanne Urbanczyk and Dickey Walsh (eds.), University of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory, 77−136. University of Massachusetts, Amherst. Blevins, Juliette 1999 Untangling Leti infixation. Oceanic Linguistics 38(2): 383−403. Bloomfield, Leonard 1933 Language. New York: Holt Rinehart and Winston. Brame, Michael 1974 The cycle in phonology: Stress in Palestinian, Maltese, and Spanish. Linguistic Inquiry 5: 39−60. Clements, Geoge N. 1992 Phonological primes: Features or gestures? Phonetica 49: 181−193. Dixon, Robert M. W. 1977 Some phonological rules in Yidiny. Linguistic Inquiry 8: 1−34. Duden 2005 Duden. Das Aussprachewörterbuch. Ed. by Max Mangold. Mannheim: Dudenverlag. Gregersen, Edgar A. 1974 Consonant polarity in Nilotic. In: Edgar Voeltz (ed.), Third Annual Conference on African Linguistics, 105−109. Bloomington, IN: Indiana University Press. Harrell, Richard S. 1962 A Short Reference Grammar of Moroccan Arabic. Washington: Georgetown University Press. Harris, John and Edmund Gussmann 2002 Word-final onsets. UCL Working Papers in Linguistics 14: 1−42. Horwood, Graham 1999 Anti-faithfulness and subtractive morphology. ROA-466. New Brunswick, NJ: Rutgers University. Horwood, Graham 2002 Precedence faithfulness governs morpheme position. In: Line Mikkelsen (ed.), Proceedings of the 21st West Coast Conference on Formal Linguistics, 166−179. Somerville, MA: Cascadilla Press. Jones, Daniel 2006 Cambridge English Pronouncing Dictionary. Ed. by Peter Roach, James Hartman and Jane Setter. Cambridge: Cambridge University Press. Kager, René 1999 Optimality Theory. Cambridge: Cambridge University Press. Kahn, Daniel 1980 Syllable-based Generalizations in English Phonology. New York: Garland.

185

186

I. Word-formation as a linguistic discipline Lappe, Sabine 2007 English Prosodic Morphology. Dordrecht: Springer. Lieber, Rochelle 2010 Towards an OT morphosemantics. In: Susan Olsen (ed.), New Impulses in Word-Formation, 173−232. Hamburg: Buske. Mascaró, Joan 2007 External allomorphy and lexical representation. Linguistic Inquiry 38(7): 715−735. McCarthy, John 2002 A Thematic Guide to Optimality Theory. Cambridge: Cambridge University Press. McCarthy, John 2005 Optimal paradigms. In: Laura J. Downing, T. Alan Hall and Renate Raffelsiefen (eds.), Paradigms in Phonological Theory, 170−210. Oxford: Oxford University Press. McCarthy, John 2006 Morphology: Optimality theory. In: Keith Brown (ed.), Encyclopedia of Language & Linguistics, 308−316. 2nd ed. Amsterdam: Elsevier. McCarthy, John and Alan S. Prince 1986 Prosodic morphology. Ms. University of Massachusetts at Amherst and Brandeis University, Waltham, MA. McCarthy, John and Alan S. Prince 1993 Prosodic morphology. Constraint interaction and satisfaction. Technical Report 3. New Brunswick, NJ: Rutgers Center for Cognitive Science. McCarthy, John and Alan Prince 1995 Faithfulness and reduplicative identity. In: Jill N. Beckman, Suzanne Urbanczyk and Dickey Walsh (eds.), University of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory, 249−384. Amherst: University of Massachusetts. McCarthy, John and Alan Prince 1999 Faithfulness and identity in prosodic morphology. In: René Kager, Harry van der Hulst and Wim Zonneveld (eds.), The Prosody-Morphology Interface, 218−309. Cambridge: Cambridge University Press. Paster, Mary 2009 Explaining phonological conditions on affixation: Evidence from suppletive allomorphy and affix ordering. Word Structure 2(1): 18−37. Prince, Alan and Paul Smolensky 1993 Optimality theory. Constraint interaction in generative grammar. Technical Report CUCS-696−95. RuCCS-TR-2 [2002, as ROA-537]. Prince, Alan and Paul Smolensky 2004 Optimality Theory. Constraint Interaction in Generative Grammar. Malden, MA: Blackwell. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Raffelsiefen, Renate 1995 Conditions for stability. The case of schwa in German. Arbeiten des Sonderforschungsbereichs 282. Nr. 69. Heinrich Heine Universität Düsseldorf. Raffelsiefen, Renate 2004 Absolute ill-formedness and other morphophonological effects. Phonology 21(1): 91− 142. Raffelsiefen, Renate 2010 Idiosyncrasy, regularity, and synonymy in derivational morphology: Evidence for default word interpretation strategies. In: Susan Olsen (ed.), New Impulses in Word-Formation, 173−232. Hamburg: Buske.

11. Word-formation in optimality theory

187

Rice, Curt 2007 Gaps and repairs at the phonology-morphology interface. Journal of Linguistics 43(1): 197−221. Russell, Kevin 1997 Optimality theory and morphology. In: Diana B. Archangeli and D. Terence Langendoen (eds.), Optimality Theory. An Overview, 102−133. Malden, MA: Blackwell. Schachter, Paul and Fe T. Otanes 1972 Tagalog Reference Grammar. Berkeley: University of California Press. Schneider, Klaus P. 2003 Diminutives in English. Tübingen: Niemeyer. Selkirk, Elisabeth O. 1986 Phonology and Syntax. The Relation between Sound and Structure. Cambridge, MA: MIT Press. Shibatani, Masayoshi 1973 The role of surface phonetic constraints in generative phonology. Language 49(1): 87− 106. Smolensky, Paul 1993 Harmony, markedness and phonological activity. Rutgers Optimality Workshop − 1, October 23. Ms. ROA-87. Smolensky, Paul 2006 Optimality in phonology II: Harmonic completeness, local constraint conjunction, and feature domain markedness. In: Paul Smolensky and Géraldine Legendre (eds.), The Harmonic Mind. From Neural Computation to Optimality-Theoretic Grammar. Vol. 2, 453−535. Cambridge, MA: MIT Press. Trubetzkoy, Nikolaj S. 1958 Grundzüge der Phonologie. Göttingen: Vandenhoeck & Ruprecht. Underhill, Robert 1976 Turkish Grammar. Cambridge, MA: MIT Press. Urbanczyk, Suzanne 1998 Avoidance of the marked. UBC Ms. ROA-286. van Engelenhoven, Aone 1995 A description of the Leti language (as spoken in Tutukei). Ph.D. dissertation, University of Leiden. Vennemann, Theo 1988 Preference Laws for Syllable Structure and the Explanation of Sound Change. With special reference to German, Germanic, Italian, and Latin. Berlin: Mouton de Gruyter. Wunderlich, D. 2006 Syntax: Optimality theory. In: Keith Brown (ed.), Encyclopedia of Language & Linguistics, 408−418. 2nd ed. Oxford: Elsevier. Xu, Zheng 2011 Optimality theory and morphology. Language and Linguistics Compass 5(7): 466−484. Yu, Alan C. L. 2007 The phonology-morphology interface from the perspective of infixation. In: Walter Bisang, Hans H. Hock, Werner Winter, Matti Miestamo and Bernhard Wälchli (eds.), New Challenges in Typology, 35−54. Berlin: Mouton de Gruyter. Zoll, Cheryl 2004 Positional asymmetries and licensing. In: John McCarthy (ed.), Optimality Theory in Phonology, 365−378. Oxford: Blackwell.

Renate Raffelsiefen, Mannheim (Germany)

188

I. Word-formation as a linguistic discipline

12. Word-formation in construction grammar 1. 2. 3. 4. 5. 6. 7.

Introduction The hierarchical lexicon Holistic properties of word structure Phrasal and phrase-based lexical items Form-meaning asymmetries Conclusions References

Abstract The notion “construction” that plays a central role in construction grammar, is an indispensable notion for the analysis of word-formation patterns. In the study of wordformation, we investigate the systematic correspondences between form and meaning at the word level. Constructional schemas provide an adequate format for expressing these systematic correspondences. Moreover, they are part of a hierarchical lexicon in which both complex words and morphological patterns of various levels of abstraction can be specified. An important advantage of the construction morphology approach is that it can express the relevant similarities between morphological and phrasal lexical expressions, and the paradigmatic relations between morphological and phrasal schemas. Thus, lexical knowledge is characterized as a complicated network between words and phrasal expressions on a range of levels of abstractions, varying between individual words and completely abstract patterns.

1. Introduction Word-formation is the domain of linguistics in which systematic correspondences between the form and meaning of complex words are studied. Consider the following sets of English adjectives: (1)

steady social suitable stressed sympathetic

unsteady unsocial unsuitable unstressed unsympathetic

The meaning of the adjectives in the column on the right can be circumscribed as ‘not A’, where A is the meaning of the corresponding adjective in the column on the left. That is, there is a systematic correspondence between the presence of un- and the meaning component ‘not’. Therefore, we consider the adjectives in the rightmost column as complex words, with the word structure [un[x]A]A, where x is a variable for a phonological string. Thus, we assign morphological structure to words on the basis of systematic paradigmatic relations with other words.

12. Word-formation in construction grammar The structure [un[x]A]A is part of the following constructional schema in which the correspondence between form and meaning of a morphologically defined class of words is specified: (2)

The angled brackets demarcate a constructional schema. The correlation between form and meaning is expressed by the double-arrowed symbol ↔. The meaning contribution of the base word on the right of the arrow is co-indexed with the relevant part of the formal structure on the left of the arrow. The schema thus represents the meaning of these un-adjectives as a compositional function of that of their base words. The format of these schemas derives from the parallel architecture framework, as developed in work by Ray Jackendoff (Jackendoff 2002; 2009; 2010; 2013). The meaning (SEM) of the base words is specified independently in the lexicon, whereas the meaning contribution of affixes is specified in constructional schemas, since their meaning is not accessible outside of the morphological structure in which they occur. Note, moreover, that the meaning of un- depends on the kind of morphological structure it occurs in. In the structure [un[x] N] V for instance, the meaning of un- is ‘reversative action’, as in the denominal verbs un-cork and un-root. Hence, the meaning of the prefix un- cannot be specified in isolation of the morphological structure of which it forms a part. Word-formation patterns can thus be considered constructions at the word level, and the individual complex words that instantiate these patterns are (morphological) constructs. The constructional schema in (2) differs from the format of the word-formation rule as used in traditional generative morphology (Aronoff 1976) in that it is neutral as to production or perception. This schema is a declarative statement that characterizes a set of existing English complex adjectives, and at the same time indicates how new adjectives of this type can be formed. This type of morphological knowledge can be used both in language perception and in language production, and therefore, it is quite appropriate for morphological regularities to be expressed in declarative form. The notion “construction” has been shown to be essential for a proper characterization of the syntax of natural languages, and this theoretical stance is referred to as construction grammar (Goldberg 2006). The use of the notion “construction” in the domain of morphology has been argued for in my monograph Construction Morphology (Booij 2010). In this article I will present a number of arguments in favour of a construction morphology approach to word-formation. In section 2 the concept of the hierarchical lexicon is introduced. In section 3 holistic properties of complex words are shown to be an argument for constructional schemas at the word level. Section 4 argues that there is no sharp boundary between lexicon and grammar, a basic idea of construction grammar, and that constructional schemas can be used in order to express the parallelisms between morphological and syntactic constructs. In section 5 we will see that we need the concept of paradigmatic relationships between constructional schemas in order to account for form-meaning asymmetries in complex words. Section 6 presents a summary of the argumentation for a construction morphology approach to word-formation.

189

190

I. Word-formation as a linguistic discipline

2. The hierarchical lexicon The acquisition of word-formation patterns is based on knowledge of memorized complex words. Once a pattern has been discovered, this can be expressed in a constructional schema that dominates the individual instantiations of this schema. The schema and its instantiations co-exist, as the discovery of a schema will not lead to removal from one’s memory of the set of complex words that formed the basis of the constructional schema. This insight can be expressed in a hierarchical lexicon, in which abstract schemas dominate their instantiations. For instance, the following substructure can be assumed for the English lexicon: (3)

< [un[suitable]A]A ↔ [ not suitable]>

The information about these three adjectives is almost completely redundant, as their properties are inherited from the general schema by which these adjectives are dominated, and the meaning of their independently specified base words. The only non-redundant information about these complex adjectives is that they exist, that is, that they are conventionalized complex words of present-day English. The partial structure of the English lexicon specified in (3) has only one level of abstraction above the level of the individual complex words. However, it might be necessary to have more levels of abstraction. A clear case for this necessity can be found in the domain of compounding in Germanic languages. The most general statement about Germanic compounds is that they are right-headed. This can be expressed at the topmost level of the substructure of the lexicon for compounds. However, the various classes of compounds may also have specific properties of their own. Examples are that verbal compounding is unproductive in languages like Dutch, English, and German, and that in Dutch AN compounds, unlike NN compounds, the non-head tends to be simplex. For instance, whereas the AN compound hoog-bouw ‘high-building, high rise building’ is a correct compound of Dutch, a compound like huizenhoog-bouw ‘houses-high building, very high rise building’, with the complex adjective huizen-hoog lit. ‘houses high, very high’ as modifier, is ungrammatical. Hence, the following partial hierarchy is relevant for Dutch AN-compounds: (4)



Condition: Ai = simplex

The lower schema is an instantiation of the upper schema, with the category variables specified as A and N, and the nature of the semantic relation R specified as ‘having property’; moreover, a restriction on the complexity of the non-head constituent applies. Another argument for the use of intermediate levels of abstraction for the description of regularities in compounds is that constituents within compounds may have lexicalized,

12. Word-formation in construction grammar yet productive meanings. That is, they may have bound meanings dependent on their occurrence in compounds. An example from Dutch is the use of the noun dood ‘death’ as a modifier in NA compounds, with the meaning ‘very, to a high degree’ as in: (5)

dood-gewoon dood-moe dood-simpel dood-stil

‘very ‘very ‘very ‘very

normal’ tired’ simple’ quiet’

This productive use of the bound meaning ‘very’ of the noun dood can be expressed in a subschema for NA compounds that is dominated by the NA compound schema that in its turn is dominated by the general schema for adjectival compounds: (6)

In this schema one of the slots is lexically specified, and hence this is a constructional idiom (Jackendoff 2002). The semantic specification for dood in (6) overrules the literal meaning ‘death’ of dood when used as an independent word. Words with bound meanings are often referred to as affixoids, as they are similar to affixes in having bound meanings. By making use of the concept of ‘constructional idiom’ we can avoid introducing a new morphological category for word constituents besides words and affixes. A constructional idiom is a constructional schema in which at least one slot is lexically fixed, and at least one slot is open. We need this type of constructional schema for another form of boundedness as well. Consider the German word Macher ‘maker’, discussed in detail in Joeres (1995). Joeres’ observation is that -macher with the regular meaning ‘maker’ is a very productive constituent of compounds, whereas it has a lexicalized meaning when used as an autonomous lexeme, namely ‘strong personality who achieves a lot’. Hence, Joeres (1995: 151) concluded that -macher can be qualified as a “Halbsuffix”, that is, an affixoid. Note, however, that macher is not one morpheme, as was the case for the affixoids discussed above, but consists of two, the verbal stem mach- ‘to make’ and the agentive suffix -er. Examples of this type of compounding in German are: (7)

with A as first constituent: Fit-macher ‘fit-maker’ Krank-macher ‘ill-maker’ Wach-macher ‘awake-maker’ with N as first constituent: Baby-macher ‘baby-maker’ Eis-macher ‘ice-maker’ Programm-macher ‘program-maker’

Different from what is at stake with the noun dood, the meaning of macher in these words is completely regular: it has the meaning ‘entity that causes or creates something’. This meaning, however, is only productive within compounds, and not available for the word Macher in isolation, which only has the lexicalized meaning mentioned above. In

191

192

I. Word-formation as a linguistic discipline order to account for its bounded productivity, we may assume the following constructional idioms for this class of compounds: (8)

Old Fr. eissir, It. uscire, Rom. a ieşi, Sard. bessiri, bessire) and the simultaneous formation and spread in the use of other strategies that are typical of the verb-framed type (e.g., the coinage of verbs expressing the directional component in the root: Fr. hausser, It. alzare, Sard. artziai, arziare ‘to lift, raise’ < Lat. *altiare ← altus ‘high’). Late Latin and the initial stages of Romance languages, on the other hand, are also characterized by the expansion of PVs (e.g., Late Lat. Quisquis servus sine dominico iussu foras exierit ‘Any slave going abroad without the master’s permission’ Petron. Satyr. 28, 7; Old Fr. Lors saut avant Girflez et dist a la reïne: … ‘Then Girflet comes forward (lit. jumps forward) and says to the queen: …’), a phenomenon that is not congruent with the emergence of the verbframed type. The subsequent history of the Romance languages sees for most of them (Portuguese, Spanish, Catalan, French, and Romanian) a progressive loss of productivity and marginalization of PVs starting from the second half of the 15th century circa, a recovery in recent times, and a current stage of spreading in use for standard Italian, and a constant vitality in other languages (Rhaeto-Romance, Northern dialects of Italy). Since Italian distinguishes itself from the other major Romance languages both by the number and use of PVs and by the number of studies on the topic, I believe it is useful first to provide (cf. sections 2 to 4) a description of the morpho-syntactic, semantic and aspectual properties of PVs on the basis of contemporary Italian. This is true also because the main characteristics observed in Italian are shared by the Romance varieties that make a more extensive use of PVs (the few characteristics differentiating these varieties from Italian are also highlighted). Section 5 is devoted to the origins of PVs and to their main features during the Middle Ages. Section 6 deals with variation in the contemporary Romance languages. The characteristics and the use of PVs in the contemporary Romance languages are described in section 7.

2. Morpho-syntactic properties PVs used in contemporary Italian display a strong morpho-syntactic and semantic cohesion. Their characteristics allow them to be set apart from other constructions (cf. Iacobini and Masini 2006). The minimal PV configuration found in contemporary Italian, which is by far the less controversial in classification in both diachronic and implicational terms, consists of a simple (non-pronominal, non-reflexive) verbal base and a postverbal modifying spatial particle that is not followed by an argument expressing ground. The argumental structure of PVs usually corresponds to that of their verbal base. PVs may be both intransitive (1) and transitive (2); in the latter case, the direct object normally occurs to the immediate right of the particle. (1)

L’ascensore va su. ‘The elevator goes up.’

(2)

Il vento spinge via le nuvole. The wind pushes away the clouds ‘The wind blows the clouds away.’

36. Particle verbs in Romance Giovanni mise giù il telefono. John put down the telephone ‘John hung up the phone.’ A quite regular minor change in argument structure is illustrated in (3). Here the particle su in (3b) absorbs the indirect ground argument (sul fuoco) of the verbal base (cf. 3a), whereas the direct argument (il caffè) is not affected. (3)

a. Metti il caffè sul fuoco. ‘Put the coffee on the stove.’ b. Metti su il caffè. ‘Put on the coffee.’

An increase of argumental roles is less frequent than their reduction, but it is possible, cf. (4). (4)

a. Il cane corre. ‘The dog runs.’ b. Il cane corre dietro ai gatti. The dog runs after to_the cats ‘The dog runs after the cats.’

A more important and frequent argumental change is the formation of an unaccusative PV from a verbal base admitting both transitive and/or unergative constructions. This shift in valency is signaled in Italian by the change of the auxiliary verb: the auxiliary essere ‘to be’ is used in the expression of telic events (5a), the auxiliary avere ‘to have’ in atelic ones (5b). Auxiliary shift in the same verb distinguishes the Italian language from the current stage of the other major Romance languages, while in their initial stages all of them share these characteristics, cf. sections 5 and 6. (5)

a. Il piccione è volato via. ‘The pigeon flew away.’ b. Il piccione ha volato per tre ore. ‘The pigeon flew for three hours.’

Romance PVs are characterized by morphosyntactic cohesion and fixity. Normally, only clitics (6a) and other light constituents, e.g., adverbs or focal elements (6b), can occupy an intermediate position between the verb and the particle. (6)

a. Devi buttarla fuori. ‘You must throw her out.’ b. Devi andare sempre avanti. You must go always forward ‘You must always go forward.’

629

630

III. Units and processes in word-formation II: Special cases c. Non devi andare mai indietro. You must go never back ‘You must never go back.’ The default position of argumental lexical elements is to the right of the particle (7). Yet examples like the ones in (8) are not ungrammatical: they are more frequent in spoken language, especially in contexts that focus on the course of the event (cf. Masini 2008 for an analysis of the phenomenon and quantitative data; prosodic restrictions on the insertion of lexical elements between the verb and the particle are investigated also by Schwarze 2008: 220−221). (7)

Irene buttò via il libro vecchio. ‘Irene threw away the old book.’ Irene portò giù le bottiglie con l’ascensore. ‘Irene took down the bottles with the elevator.’

(8)

mettere la palla dentro to put the ball in ‘to make a goal’ mettere le mani avanti to put the hands forward ‘to play it safe; to cover yourself/your back’

The topicalization of the particle (9b) and its left-dislocation (9c) are normally unacceptable. (9)

a. Irene è saltata fuori all’improvviso. Irene jumped out all of a sudden ‘Irene suddenly popped up.’ b. *Fuori Irene è saltata all’improvviso. c. *È fuori che Irene è saltata all’improvviso.

Cohesion between the verb and the particle is highlighted not only by the semantic unity of the construction, but also by the structural characteristics of the sentences. For example, comparing the argument structure of two apparently similar sentences, note that the one in (10a) is formed by a PV and one argument, while the verb of sentence (10b) has two arguments. (10) a. [Metti dentro] [la borsa e i guanti.] [put in] [the bag and the gloves] ‘Put in the bag and the gloves.’ b. [Metti] [dentro la borsa] [i guanti.] [put] [in the bag] [the gloves] ‘Put the gloves in the bag.’

36. Particle verbs in Romance Cohesion between verb and particle is particularly evident in coordinating structures, where PVs (11a, b) behave differently from apparently similar prepositional phrases (11c, d). (11)

a. Irene porta su il tavolo e Anita ___ le sedie. ‘Irene brings up the table and Anita the chairs.’ b. *Irene porta su il tavolo e Anita su le sedie. Irene bring.3SG up the table and Anita up the chairs. c. Irene mangia sul tavolo tondo e Anita su quello quadrato. ‘Irene eats on the round table and Anita on the square one.’ d. *Irene mangia sul tavolo tondo e Anita ___ quello quadrato. Irene eat.3SG on.the table round and Anita ___ that square.

The frequency of use of PVs in final clause position and the fact that they form a unitary prosodic constituent, in which the particle plays a prominent role (cf. Simone 1997; Iacobini 2008), may be considered further support for the strong bond between verb and particle. Besides the configurations in (1) and (2), Italian PVs display a number of other possibilities, such as reflexive transitive constructions (12a), pronominal unaccusative constructions (12b), and constructions with an indirect argument that may be expressed by a locative (12c) or a pronominal clitic (12d). The pronominal clitic construction is almost absent in Spanish and Catalan (cf. Calvo Rigual 2008), but it is quite common in French (cf. Pourquier 2001). (12) a. tirar=si fuori pull=REFL out ‘to get out’ b. precipitar=si fuori rush=PRON out ‘to rush out’ c. ber=ci sopra drink=LOC.PRT up ‘to drink to forget something’ dar=ci dentro give=LOC.PRT inside ‘to go hard at, crank it up’ d. saltar=gli addosso jump=PRON on ‘to jump on, pounce on’ When the referent of sentences like the one in (12d) is not expressed by a clitic, but by a nominal autonomous form (noun or pronoun), it must be preceded by a preposition (usually the preposition a), cf. (13). (13) saltare addosso al nemico jump on to_the enemy ‘to jump on the enemy’

631

632

III. Units and processes in word-formation II: Special cases correre dietro alle ragazze run behind to_the girls ‘to chase after the girls’ passare sopra a qualcosa pass on to something ‘to pass, forgive, let something pass’ It is not always easy or possible to decide how to interpret these sequences of particles whether as a PV (saltare addosso) that governs a prepositional phrase (al nemico) or as a PV with a complex P (saltare addosso a). Quaglia (2012) argues that examples like the ones in (12d) and (13) are not to be considered as instances of PVs, but as a particular sub-class of verb+preposition constructions. One may also consider as PVs other constructions in which the particle position is filled by elements that are not spatial, i.e. temporal (14a) or manner (14b) adverbs. Nonetheless, these constructions are beyond the scope of this paper. (14) a. fare presto do early ‘to hurry up’ b. finire male finish badly ‘to come to a bad end’

3. Semantic properties PVs express a range of spatial meanings that are comparable to those expressed by prefixation, and also richer with respect to the ones that can be expressed by the (more frequently used) path verbs (e.g., salire ‘to rise, go up’, montare ‘to climb on/in, get on/ in’). However, while preverbal prefixation has lost most of its productivity, post-verbal particles can be freely combined with a wide array of verbal bases. PVs mainly convey directional meanings. The particle may function as a direction marker with manner verbs (15a), with generic and deictic verbs of motion (15b), and with verbs of putting and removal (15c). (15) a. saltare dentro/fuori ‘to jump in/out’ b. andare dentro/fuori ‘to go in/out’ c. portare dentro/fuori ‘to bring in/out’ PVs may also result from the combination of a path verb and a particle; in this case the particle strengthens the directional information already lexicalized in the verbal base (16).

36. Particle verbs in Romance (16) entrare dentro enter in ‘to enter’ uscire fuori exit out ‘to exit’ PVs formed with stative verbs express locational meanings. This kind of construction is often interpreted metaphorically (17). (17) stare accanto stay next to ‘to stand by, support’ stare dietro stay behind ‘to keep after, keep up, court’ covare dentro brood inside ‘to harbour’ PVs may display idiomatic meanings independently of the semantics of the verb and of the particle. Nevertheless, polysemous PVs tend to maintain their spatial meanings along with the new idiomatic meanings. (18) illustrates some examples of PVs that are used most often with a non-compositional meaning, rather than a spatial meaning, and (19) shows how they are used in contexts. (18) portare avanti carry forward ‘to carry forward/on/out; to hold, handle, manage, conduct’ tirare avanti pull forward ‘to get by, draw forth; to survive, resist’ tirare su pull up ‘to pull up; to raise, bring up, cheer up, hike up’ venire fuori to come out ‘to come out; to turn out, result, emerge’ (19) questo modo di portare avanti la discussione ‘this way of handling the discussion’ un po’ di soldi per poter tirare avanti ‘a little money in order to survive’

633

634

III. Units and processes in word-formation II: Special cases Ho avuto due figli, li ho tirati su da sola. ‘I have two kids, I brought them up by myself.’ Alla fine del conteggio è venuto fuori che … ‘At the end of the tally it turned out that …’ Another sign of semantic lexicalization and of the entrenchment of the construction is the coinage of PVs from non-motion verbs. The meaning of such constructions often results from metaphoric or metonymic interpretation, cf. (20). (20) fare fuori do out ‘to wipe out; to kill; to eat up’ parlare dietro speak behind ‘to tittle-tattle, buzz about’ tenere sotto hold down ‘to keep down, subjugate, submit, put under pressure’ tenere su hold up ‘to keep up, cheer up’ The verbs used in PVs that are attested more frequently and combine with a higher number of particles are generic motion verbs, deictic verbs, and verbs of putting or removal (e.g., andare via ‘to go away’, mettere dentro ‘to put in/away’, tirare fuori ‘to pull out’, venire su ‘to come up/out’). Path verbs permit only particles that express the same orientation that is already encoded in the verb root (e.g., salire su lit. ‘ascend up’, scendere giù lit. ‘descend down’). Manner verbs are less frequently used, even if they are much less constrained by semantic compatibility with the particle (e.g., correre via/ fuori/dentro/su/giù ‘to run away/out/in(to)/up/down’). Table 36.1 shows the range of spatial meanings that may be conveyed through particles in semantically transparent PVs in contemporary Italian. The data have been extracted from LIP (Lessico di frequenza dell’italiano parlato; cf. De Mauro et al. 1993), which is a spoken corpus amounting to approximately 500,000 tokens balanced for diaphasic and regional varieties. The left column of the table lists the particles grouped by semantic type. The second column shows their meanings. The third column reports the percentage of use of PVs (tokens). And the rightmost column allows a comparison between postverbal particles and verbal prefixes. Prefixes that cannot be productively used with spatial meaning are underlined; question marks (?) indicate prefixes whose productivity is doubtful. The most frequent particle meanings result as: the opposites ‘up/down’ on a vertical axis, and the expression of distancing, separation, or removal, rendered with the particles fuori and via. Fuori and via, considered by themselves, represent the most frequently used particles, but the corresponding opposite meaning ‘in, into’ is much less frequent. The expression of forward movement is well represented, while the corresponding back-

36. Particle verbs in Romance

635

Table 36.1: Spatial meanings expressed by particles in semantically transparent PVs in LIP and comparable prefixes Italian post-verbal particles

Spatial meanings

% in PVs

sopra, su

up, over

14

sopra-/sovra-, sor-

giù, sotto

down, under

15

sotto-

dentro

in, into

fuori, via

away, off, out

37

ab-, de- (?), dis-, e-/es-, estro(?), s-, se-

avanti

forward

15

ad-, pre-, pro-

dietro, indietro

back

2

re-/ri- (?), retro-

accanto, appresso, vicino

near, by

4

giusta-

oltre

across, beyond

1

per-, trans-

intorno

around, about

1

circum-

lontano

far

1

—

addosso

against

1

contro-

—

between

—

9

Italian prefixes

in-, intro-

fra-, inter-

ward movement particles occur only a few times. Another important feature of Italian (as well as of other Romance languages) is the lack of particles and prepositions expressing goal attainment (e.g., Engl. into), and the almost total absence of particles expressing inherently directional meaning (e.g., Engl. to). Comparison of the spatial meaning expressed by PVs with those expressed by verbal prefixes shows that the two ways of expressing direction of motion largely overlap. However, only a small number of the prefixes can be productively used to express the direction of motion. Moreover, only one (intro-) of the few productive prefixes listed in Table 36.1 expresses mainly spatial meaning when added to verbs. Whereas the other prefixes that formerly had a spatial value are now productively employed with mostly other meanings (the same can be said for the verbal prefixation in the other Romance languages, cf. Lüdtke 1996). The most frequent PVs in LIP expressing transparent directional meanings are listed in (21) in alphabetical order. (21) andare avanti ‘to go ahead/forward/on’ andare fuori to go out/outside’ andare giù ‘to go down’

636

III. Units and processes in word-formation II: Special cases andare su ‘to go up/on’ buttare via ‘to throw away’ entrare dentro enter into ‘to enter’ mandare via ‘to send away’ mettere sopra ‘to put on/over’ portare su to carry up ‘to carry up/on/over to/over onto, bring up’ portare via ‘to carry/take away’ scendere giù descend down ‘to descend, to go/come/climb/walk/step down, get down/off/out’ tirare fuori ‘to pull out’ uscire fuori exit out ‘to exit’ venire via ‘to come away’ Evidence of the role of particles for the semantics of PVs is provided by synonymous PVs formed by using the same particle with different base verbs (e.g., buttare/portare/ sbattere/spedire dentro il ladro ‘to imprison the thief; lit. throw/bring/haul/send inside the thief’, cf. Guglielmo 2013: 216−243).

4. Aspectual properties The use of post-verbal particles as aspectual markers, besides their spatial function, shows that PVs are a structure entrenched in the grammatical system of a given language. This phenomenon, which is very frequent in satellite-framed languages, was common in all the Romance languages during the medieval period, and it is currently present in Italian, Rhaeto-Romance, and northern dialects of Italy, but absent in the other major Romance languages. Iacobini and Masini (2006) describe the emergence of PVs with

36. Particle verbs in Romance

637

Tab. 36.2: Proportion of telic vs. atelic PVs (tokens) in LIP formed with the most typically telic and atelic particles Particle

% PV +Tel 100

via

% PV −Tel —

giù

92.7

7.3

su

92.1

7.9

fuori

89.7

9.2

dentro

83.8

16.2

dietro

40.0

60.0

accanto

12.5

87.5

1.8

98.2

avanti

—

intorno

100

via ‘away’ expressing telicity. When added to verbs, the particle via may stress the completion of the process and emphasize the result of the action expressed by the verb (cf. strofinare le macchie ‘to rub the stains’ and strofinare via le macchie ‘to rub off, to wipe off the stains’). Even though Italian does not present a developed system of aspectual particles, there are nonetheless some traces of regularity. Particles that can be used to express movement oriented towards a goal or originating from a source may come to imply attainment of the goal (telic situations) (22a), whereas particles that express stasis, location, or motion without a specific endpoint may contribute to indicate atelic situations (22b). (22) a. saltare ‘to jump’ –TEL volare ‘to fly’ –TEL b. tirare ‘to throw, pull’ ±TEL girare ‘to turn’ ±TEL

saltare dentro ‘to jump in’ +TEL volare via ‘to fly away’ +TEL tirare avanti ‘to get by’ –TEL girare intorno ‘to go round and round; to roam, wander around’ –TEL

The results of an analysis performed on the LIP corpus (cf. Table 36.2, taken from Iacobini 2008: 113) confirm the high correlation between the meanings expressed by particles and the aspectual characteristics of the PVs they contribute to. Perhaps the most significant difference between standard Italian and the Romance varieties in which PVs are extensively employed is the more frequent and systematic use of aspectual post-verbal particles. In Rhaeto-Romance and in the dialects spoken in the North-East of Italy PVs are currently used to encode telicity (23a) − beside spatial

638

III. Units and processes in word-formation II: Special cases information −, and they can also express durative (23b), progressive (23c), or inchoative values (23d), with both motion and non-motion verbs (cf. section 7.3). (23) a. comedàr fòra repair out ‘to repair’

Trentino dialects

ligàr su ‘to tie up’ b. vardàr le carte ‘to look at the documents’

Trentino dialects

vs. vardàr fora le carte lit. look out ‘to examine the documents’ (Cordin 2008a: 184) quàśe dré ca la seca. c. L’ è 3SG.F be.3SG almost behind COMPL 3SG.F dry3SG ‘[The plant] is dying.’ d. I s’ è metìt dré a löcià. 3PL.M REFL be.3SG put.PART.PAST.PL behind to cry.INF ‘[The people] started to cry.’ (Bernini 2008: 152–153)

Bergamasque dialect

Bergamasque dialect

5. Particle verbs in Medieval Romance languages The current uneven distribution of PVs tilted toward Rhaeto-Romance, Northern dialects of Italy, and standard Italian does not reflect the situation in the oldest stages of the Romance languages. There was a period in which PVs were largely used in all the Romance languages. The transition from Late Latin to Romance languages does not coincide with an abrupt shift from the satellite- to the verb-framed type. Both the univerbation of prefixed verbs and the emergence of PVs are evident in the initial stage of Romance languages. The formation of new path verbs (resulting from the blurring of the boundary between stem and prefix of Latin verbs (24a), or from nominal or adjectival bases referring to entities that have an intrinsic direction reference (24b), or from directional prepositions (24c)) and PVs in early stages of Romance languages (e.g., Old Catalan pujar sus ‘to ascend’, Old Provençal venir foras ‘to come out’) may be interpreted as two different strategies emerging at the same time and sharing the same purpose: the encoding of motion events. (24) a. Old Sp. exir, Cat. eixir, Old Fr. eissir, It. uscire, Rom. a ieşi, Sard. bessiri, bessire < Lat. ex-ire ‘to exit’

36. Particle verbs in Romance b. Old Fr. monter, amonter, It. montare, Cat. muntar, Occ. montar, Rhaeto-R. muntar ‘to mount, ascend’ < Lat. *montare ← mons, montis ‘mountain, mount’ Fr. hausser, It. alzare, Sard. artziai, arziare ‘to lift, raise’ < Lat. *altiare ← altus ‘high’ c. Port., Cat. avançar, Sp. avanzar, Fr. avancer, It. avanzare ‘to advance, go forward’ < Lat. *abantiare ← abante ‘forward, ahead’; Rom. înainta < Lat. in abante (cf. Stolova 2008) In all the major Romance languages, up until at least the mid 15th century, PVs played an important role in the expression of spatial events. In French and Spanish a progressive marginalization of PVs has been witnessed from the end of 15th century on. They were the first languages to define a standard following a deliberate process of unification and regularization (cf. Schøsler 2008; Burnett and Tremblay 2012). A similar decrease in the use of PVs also happened in the other major Romance languages. Italian distinguishes itself by the re-emergence of PVs in the course of 20th century. The large number and the extensive use of PVs in the Romance languages during the Middle Ages suggest that there were already traces of this construction in Latin. In fact Late Latin shows signs of the emergence of PVs. Some examples can already be found in texts of less formal register from the pre-classical period (25). They become more and more numerous from the 4th century on (26) in Latin texts of varying style and degree of formality, cf. Iacobini (2009a). (25) inde effugi foras ‘thence did I betake myself off’ (Plaut., Most., 315) i intro ‘go in’ (Plaut., Most., 807) abii foras depart off ‘I went off’ (Plaut., Most., 879) (26) Et quod sursum est, deorsum faciunt. what up is they put down ‘They turn everything upside down.’ (Petron., Satyr., 63, 10) Cecidit de tertio cenaculo deorsum. ‘And fell down from the third loft.’ (Act. apost., 20, 9) Noli foras ire, in te ipsum redi. ‘Do not wish to go outside, return into yourself.’ (Aug., De vera relig., 39, 72)

639

640

III. Units and processes in word-formation II: Special cases Besides PVs, during the Middle Ages Romance languages developed the whole array of strategies which conform to Talmy’s satellite-framed typology. These strategies include productive directional and aspectual prefixes and particles, a large use of manner of motion verbs, directed motion constructions, and adjectival resultatives (cf. Dufresne, Dupuis and Tremblay 2003; Herslund 2005; Burnett and Tremblay 2012). For example, specialists in Old-French are unanimous in classifying it as a satellite-framed language. Medieval Romance manner of motion verbs allowed both an unaccusative and an unergative interpretation signaled by an auxiliary shift: ‘to be’ for unaccusative constructions expressing perfective meaning and source/goal oriented displacement motion (27a), ‘to have’ for unergative imperfective meaning expressing motion at a location (27b). por Old French (27) a. Mais tot li chevalier ensamble i sont coru but all the knights together there AUX run.PST.PTCP for lui rescorre. him rescue ‘But together the knights quickly ran there in order to rescue him.’ (Vengeance Raguidel, 1200, 33) coru et porchacié, b. Tant a so.much AUX run.PST.PTCP and chase.PST.PTCP ‘So much did he run and chase,’ (Saint-Cloud, Roman de Renart Branche 7, c.1175, quoted from Troberg and Burnett 2011)

Old French

The inventory of posture verbs was quite rich (cf. Old Fr. ester ‘to stand’, seoir ‘to sit’, gesir ‘to lie’ / Fr. être debout, être assis, être couché). There was a distinction between directional (cf. Old Fr. amont ‘upwards’; aval ‘downwards’; contremont ‘upwards’; contreval ‘downwards’) and stative particles; prefixes and post-verbal particles were productively used to express both direction (28, 29) and aspectual and argumental properties (30, 31) (cf. Buridant 2000; Dufresne, Dupuis and Longtin 2001; Burnett and Tremblay 2009). Old French (28) A une part se sunt retrait. at one part REFL were withdrawn ‘At one point, they drew back.’ (TFA: Wace, p. 168, quoted from Burnett and Tremblay 2012: 339) (29) et pou ce se trait il ariere Old French and for that REFL drew he back ‘and, because of this, he drew back’ (TFA: Trisper, p. 263, quoted from Burnett and Tremblay 2012: 339) (30) devancer / adevancer ‘to be in front’ / ‘to arrive before’ durer / adurer ‘to last; to go on’ / ‘to extend sth.’ grocier / agrocier ‘to grunt’ / ‘to insult’

Old French

36. Particle verbs in Romance parler / aparler ‘to speak; to talk’ / ‘to address’ (cf. Martin 2001: 311) (31) Il s’entrecommencent a regarder et semont li uns Old French to look and ask the one they CL.ENTRE.begin l’autre de parle avant. the other to talk forward ‘They begin to look at each other and ask each other to start talking.’ (Artu, p. 13, quoted from Burnett and Tremblay 2009: 32) According to Troberg and Burnett (2011), in Middle French texts there is clear evidence of strong adjectival resultatives, cf. (32). (32) que tricherie abat jus plate Middle French ‘that deception beats down flat’ (Christine de Pisan, Livre de la Mutacion de Fortune, t. 2, 29; DMF 2009)

6. Variation among contemporary Romance particle verbs In addition to PVs, contemporary French, together with the majority of Romance languages, has strongly reduced or completely lost the above mentioned satellite-framed type features (cf. Old Fr. Des que cil furent fort issu − Brut, Wace, 12th century ‘As soon as these people had come out; lit. had exited outside’ vs. Fr. dès qu’ils furent sortis). Nevertheless, a lot of variation can still be observed among the Romance languages. Currently, among the major Romance languages, auxiliary shift signaling the distinction between transitive and unergative from unaccusative constructions is allowed only in Italian. In French, auxiliary selection responds to the same basic requirements as in Italian, but auxiliary shift in the same verb is very limited. In contemporary Portuguese, Spanish, Catalan, and Romanian the use of the auxiliary ‘to have’ is generalized to all predicates, replacing the old contrast in auxiliary selection (cf. Loporcaro 2007, which, looking at nonstandard (Italo-)Romance varieties, describes different patterns of auxiliary verb variation that are unknown in standard languages). One of the major obstacles to the use of PVs is the fact that nowadays major Romance languages do not have pairs of prepositions or particles distinguishing locative from directional meanings (cf. English in/into, on/onto). Inherently directional prepositions and particles are very few in number (e.g., Sp. hacia, Fr. vers, It. verso ‘towards’), neither of them can unambiguously express the attainment of a goal. A result location interpretation can be univocally expressed only by complex prepositions (e.g., Fr. jusque dans, It. fin dentro ‘up to the inside of’), but, among Romance languages, this strategy seems to be commonly used in post-verbal position only in Rhaeto-Romance varieties. Since the Romance languages have lost most satellites (e.g., directional prefixes) or other morphosyntactic devices (e.g., Lat. in + accusative case), which clearly express goal-oriented motion, they tend to express path in the main verb. If manner cannot be omitted, it is typically expressed in an adjunct, as in (33), that represents a “heavier” construction if compared with Eng. to run into or Latin incurro.

641

642

III. Units and processes in word-formation II: Special cases (33) entrar corriendo entrer en courant entrare di corsa intră în fugă enter running

Spanish French Italian Romanian

The preferential encoding of path in the verb root rather than in particles associated with manner of motion verbs has consequences in the structure of discourse. Cross-linguistic studies have shown that in the narration of a motion event, speakers of Romance languages privilege descriptions of the locations of figures compared to the ground and of end states of motion (using few manner-specifying verbs). At the same time, speakers of satellite-framed languages prefer to express manner through verbs, and they provide more detailed and dynamic path descriptions (cf. Berman and Slobin 1994; Slobin 1996). The restrictions on the use of manner verbs as the main verb in the representation of boundary-crossing events limit, but do not prevent, the expression of displacement motion via PVs. In the lack of inherently directional prepositions or verb satellites, the boundary-crossing reading is made possible through the interaction of the semantic and aspectual features of the verb, co-textual information, and pragmatic inferences that are principally drawn from the meaning of the preposition and/or particle and from the physical characteristics of locations designated by the noun expressing the ground (cf. Nikitina 2008; Kopecka 2009a, 2009b; Levin, Beavers and Tham 2009). Unlike Germanic languages, the most frequently employed verbs in Romance PVs are generic verbs of motion, verbs of putting and removing, and path verbs. The manner of motion verbs that preferentially occur in Romance PVs expressing displacement or boundary-crossing events are those “that are not readily conceived of as activities, but, rather, as ‘instantaneous’ acts” Slobin (2004: 226), i.e. verbs that either encode a rapid, often sudden movement (e.g., ‘to jump’), or verbs which express an orientation: e.g., removal from a reference point (e.g., It. sbucare ‘to come out suddenly, pop out’, scappare ‘to escape’) or movement toward a goal (e.g., It. irrompere ‘to burst into’, scagliarsi ‘to lunge’, tuffarsi ‘to dive’). (34) Acquista un pacco su internet e salta fuori un topo. ‘(S/he) Buys a packet on the internet and a mouse pops out.’ Ragazza rapina il supermarket, salta sul bancone e scappa via. ‘Girl robs the supermarket, jumps on the counter and runs away.’

Italian Italian

The least compatible verbs with a displacement or a boundary-crossing reading are durative verbs expressing movement of a rather aimless sort (e.g., ‘to stroll’) or the verbs focusing on the coordination of specific movements (e.g., ‘to dance’). Italian differs from the other major Romance languages in the use of manner verbs for the expression of displacement or boundary-crossing events. Even if verbs expressing orientation and punctual high force dynamics are preferred, the formation of PVs with verbs describing slow, quiet or careful movement is also possible. (35) Le onde sorprendono Ilary, il costume scivola via. ‘The waves surprise Ilary, the swimsuit slips away.’ Provo a inserire la prima pallottola. Scivola dentro senza attrito. ‘I try to insert the first bullet. It slides in without friction.’

Italian Italian

36. Particle verbs in Romance

643

Tab. 36.3: Implicational scale of PVs in languages lacking means for encoding the attainment of a goal 1

2

uscire fuori ‘exit out’

andare fuori ‘go out’

Path

Deixis

3

4

5

6

7

8

spingere fuori ‘push out’

precipitarsi fuori ‘bolt out’

correre fuori ‘run out’

camminare fuori ‘walk out’

scivolare fuori ‘slide out’

??ballare fuori ‘dance out’

Caused Motion

Manner

Manner

Manner

Manner

Manner

±Force Dyn. −Specific

−Force Dyn. −Specific

−Force Dyn. +Specific

+Orientation −Orien+Force tation Dynamics +Force Dyn.

La ragazza si inginocchiò a fatica e strisciò fuori. ‘The girl knelt down with difficulty and crawled out.’

Italian

Quando finalmente la libreria apre i battenti la gente sciama dentro. ‘When the library finally opens its doors the people stream in.’

Italian

Verbs that are more likely to be used in PVs in languages lacking particles and other means inherently expressing direction or attainment of a goal can be ordered from left to right in an implicational scale like the one in Table 36.3 with examples of Italian verbs. It is a distribution that speaks in favor of a scalar ordering of PV features.

7. Contemporary distribution of particle verbs in Romance Although the attention on Romance PVs is relatively recent, their presence was already reconized in the comparative grammars at the end of 19th century (cf. Diez 1882: 992− 993; Meyer-Lübke 1899: 517 “Die Verbindung eines beliebigen Verbums mit einem Ortsadverbium zur Bildung einer festen Formel ist […] allen romanischen Sprachen eigen” [The combination of any verb with a spatial adverb to form a fixed expression is characteristic of all Romance languages]). Despite the current preferred use by the Romance languages of path verbs for the encoding of motion events, PVs are still used in all Romance varieties, albeit with a rather uneven distribution. In major contemporary Romance languages, PVs are preferred in less formal registers and in the spoken modality (these differences are more pronounced in the varieties in which the PVs are used less). The preference for oral use suggests that PVs are present to a greater degree in less standardized varieties, although research on the diffusion of PVs in Romance dialects is still scarce. Comparative studies are limited in number (among the most recent, cf. Mateu and Rigau 2010; Cordin 2011; Iacobini and Fagard 2011; Iacobini 2012), and corpus-based research is even rarer (cf. Schøsler 2008; Spreafico 2009; Hijazo-Gascón and Ibarretxe-Antuñano 2013). The following sections provide some indications and bibliographical references on the main characteristics and the current use of PVs in both standard and non-standard varieties.

644

III. Units and processes in word-formation II: Special cases

7.1. Spanish, Catalan and Portuguese The most detailed description of PVs in contemporary Spanish and Catalan is Calvo Rigual (2008) (see also Calvo Rigual 2010), based on a survey of grammars and monoand bilingual dictionaries. Corpus-based analyses on Spanish data are presented in Hijazo-Gascón and Ibarretxe-Antuñano (2013). Some data and bibliographical references concerning the different strategies used in Spanish for the encoding of motion events can be drawn from the works of Slobin and his followers (cf. Slobin 1996, 2004; Guo et al. 2009), as well as from studies comparing Spanish with satellite-framed languages, especially English (cf. Cadierno and Robinson 2009; Cifuentes-Férez 2009). According to Calvo Rigual (2008), PVs are much less numerous in Spanish and Catalan than in Italian: they amount to about fifty for Spanish and fewer for Catalan (whereas Calvo Rigual’s open list of Italian PVs contains more than three hundred items). The base verbs are mostly motion verbs (e.g., Sp. andar ‘to go, walk’, caer ‘to fall’, dejar(se) ‘to leave’, echar(se) ‘to chase out’, ir(se) ‘to go’, llegar ‘to arrive, reach’, llevar ‘to carry, bring’, pasar ‘to go, proceed’, seguir ‘to follow’, venir(se) ‘to arrive’, volar ‘to fly’, volver(rse) ‘to come back’; Cat. anar ‘to go’, caure ‘to fall’, continuar ‘to continue’, deixar ‘to leave’, dur ‘to carry’, sortir ‘to go out’, tirar ‘to throw’, tornar ‘to come back’, volar ‘to fly’), but also a smaller number of verbs with a highly generic meaning are used (e.g., Sp. estar ‘to be’, tener ‘to have’; Cat. dir ‘to say’, fer ‘to do’, mirar ‘to watch’, restar ‘to remain’). The number of motion particles is also limited; the most frequently used are: Sp. abajo ‘down’, adelante ‘forward’, alto ‘high’, aparte ‘aside’, arriba ‘up’, atrás ‘back’, bajo ‘down’, delante ‘forward’, derecho ‘straight’, detrás ‘behind’, encima ‘above’, fuera ‘outside’, lejos ‘far away’; Cat. alt ‘high’, amunt ‘up’, baix ‘down’, damunt ‘on, above’, darrere ‘behind’, endarrere ‘back(ward)’, endavant ‘forward’, enrere ‘back(ward)’, fora ‘out’, lluny ‘far away’. Besides the expression of motion in both transitive and intransitive constructions (Sp. irse abajo ‘to go down’, salir adelante ‘to go on/forwards’, volver atrás ‘to turn back’; Cat. anar amunt ‘to go up’, anar avall ‘to go down’, anar lluny ‘to go away’, tornar enrere ‘to turn back’), a fair number of PVs are used with an idiomatic meaning (Sp. echarse atrás ‘to back out’, sacar adelante ‘to get ahead’, venirse abajo ‘to break down’; Cat. fer fora ‘to keep out, to throw out’, tirar endavant ‘to get by, resist’). Manner verbs used in PVs are very few. Constructions in which a path verb is followed by a particle doubling the expression of direction are quite common in the spoken modality, although they are not reported in descriptive grammars (Sp. salir fuera lit. ‘to exit out’, entrar dentro lit. ‘to enter into’, subir arriba lit. ‘to ascend up’; Cat. eixir/sortir fora lit. ‘to exit out’, entrar dins lit. ‘to enter into’, pujar dalt lit. ‘to ascend up’); a diachronic survey on this subject is provided by González Fernández (1997). The expression of space in European Portuguese is investigated in Batoréo Jakubowicz (2000); a work based on original experimental data, corpus-analysis, and the thorough investigation of grammars and dictionaries, to which I refer for the few bibliographic references on Portuguese PVs. According to Batoréo Jakubowicz (2000), cf. especially ch. 4, Portuguese PVs share the same characteristics of Spanish PVs, and their diffusion in use is also comparable. Some of the most common Portuguese PVs are: andar para tràs ‘to go backwards’, carregar para baixo ‘to load down’, deitar abaixo ‘to shoot down, demolish, break down’, deitar para fora ‘to empty, excrete’, ir à frente ‘to guide, conduct’, ir-se embora ‘to escape, run away’, partir/sair embora ‘to leave,

36. Particle verbs in Romance emigrate’, pôr antes ‘to advance’, pôr contra ‘to oppose’, pôr fora ‘to put off, exclude’, ter dentro ‘to include, contain’ (cf. Batoréo Jakubowicz 2000: 789−791). Among the few studies on diatopic variation in the lexicalization of motion events there are: Rojo (1974) on Galician dialects, and the careful survey of Aragonese varieties by Hijazo-Gascón and Ibarretxe-Antuñano (2010).

7.2. French It is generally agreed that French at its current stage is the Romance language that is farthest away from the satellite-framed type, and in particular the Romance language that uses PVs least. In publications dealing with the typological shift from the satellite-framed to the verbframed pattern and with the actual typological intricacy of path expression in French, Kopecka (2009a, forthc.) focuses her attention on prefixation rather than on PVs. The almost complete absence of PVs in contemporary French is also confirmed by Herslund (2005) and Hijazo-Gascón and Ibarretxe-Antuñano (2013). We may distinguish two kinds of PVs used in contemporary French. In the first one, the path verb is followed by a directional particle that reinforces the expression of direction (e.g., descendre en bas lit. ‘to descend down’, entrer dedans lit. ‘to enter into’, monter en haut lit. ‘to mount up’, sortir dehors lit. ‘to exit out’). The other, formed from a short list of manner verbs and directional particles, is characterized by the presence of a clitic pronoun (which can only refer to animate referents) and by the dominant metaphorical meaning next to the spatial one (je lui cours après I him/her run after ‘I run after/chase/court him/her’; il m’a sauté dessus ‘he jumped on me’; je vais lui rentrer dedans ‘I am going into him/her’). PVs of the last kind are common especially in nonformal use, although they are confined to the spoken modality; cf. Pourquier (2001, 2003) for an analysis of their characteristics and a list of verbs and particles employed. In areas where the contact between French and the Germanic languages is more direct, we can identify some PVs that are not present in standard French, cf. Kramer (1981) about French spoken in Brussels (e.g., tomber bas, sauter bas) and King (2000) on French spoken in Canada. Both authors argue that the spread of such constructions is not due to the calque of individual Flemish or English expressions, rather to the Germanic languages that acted as an enhancing factor favoring the use of a strategy present at earlier stages of French. Treffers-Daller (2012) comes to different conclusions, stressing the importance of individual calques of Brussels French (e.g., recevoir dehors ‘to get out’) derived from the regional variety of Dutch.

7.3. Rhaeto-Romance varieties and dialects of Northern Italy Rhaeto-Romance varieties spoken in Italy and in Switzerland, and the dialects of the North-East of Italy (Venetian, Lombard, and Emilian dialects) are the Romance languages that make a wider use of PVs. In northern dialects such a mode of expression is even more used that in Tuscany, cf. Milanese dà föra to give out ‘to spend’, dà giò to give down ‘to go down’, di sü to say up

645

646

III. Units and processes in word-formation II: Special cases ‘to recite the lessons’, tirà sü to pull up ‘to wind the clock’, Lombard levà sü ‘to get up’ Venetian contar su ‘to tell’, often without any need, cf. si mangian su cogli occhi they each other eat up with the eyes ‘they look at each other with burning desire’ (Fogazzaro, “Malombra”, 159); in Poschiavo taśé gió to stop talking down ‘to stop talking’, i ligan sü Tell ‘they tie up Tell’. The idea to cut down a tree is expressed in Lombard with tajà giò lit. ‘to cut down’, in Venetian with tajar śo lit. ‘to cut up’ (see AIS, 532). In Upper Valtellina they say séntet ó ‘sit down’, cf. Italian siediti. (Rohlfs 1969: § 918; my transl.)

PVs are the main way of expressing verbal spatial meanings also in the dialects spoken in the Franco-Provençal and Occitan area in Piedmont and Valle d’Aosta, cf. examples in (36) taken from Cini (2002). (36) allà əndré giü sut sü via

lit. go back ‘down’ ‘below’ ‘up’ ‘away’

‘to ‘to ‘to ‘to ‘to

draw back’ go down’ sink’ go up’ go away’

‘to throw down’ ‘to throw up’ ‘to throw away’

bütà giü sü via tiría avanti fora

lit. pull ahead ‘out’

‘to get by, manage’ ‘to take out’

sü via

‘up’ ‘away’

‘to pick up, bring up’ ‘to take away’

A number of studies have been dedicated to the lexicalization of motion in RhaetoRomance and in Romance varieties spoken in the north-east of Italy. Gsell (1982) provides an overall study on PVs, Kramer (1987) gives an overview and rich bibliographical references on Ladin, and the monograph of Vicario (1997) focuses on Friulian from the earliest documents in the fourteenth century to the present day. Some interesting contributions and updated bibliographical references may also be found in Berthele (2006), Cini (2008), Cordin (2011), and Bernini (2012). Rhaeto-Romance varieties make an almost exclusive use of PVs in the encoding of motion (cf. Mair 1984; Kramer 1987). According to Vicario (1997), contemporary Friulian uses PVs to express the basic motion events listed in (37). (37) lâ su ‘to go up’ lâ jù ‘to go down’ lâ fûr ‘to go out’ lâ dentri ‘to go into’

36. Particle verbs in Romance

647

Friulian does not have a synthetic form for the expression of upward/downward direction, while the verbs jessi ‘to exit’ and jentrâ ‘to enter’ are used very sparcely and have a defective paradigm limited to the participle and the infinitive forms. Likewise, according to Bernini (2008: 143−144), the Bergamasque dialect uses PVs exclusively for the first three meanings, cf. (38), while ‘to go into’ can be expressed both by a PV (indà de dét) and a path verb (entrà). (38) indà de ’n sö ‘to go up’ dà de ’n śó ‘to go down’ indà de fò ‘to go out’ In all these varieties PVs represent the most natural and frequent way to express spatial meanings. The base verbs are mainly generic and causative motion verbs, but also a number of manner and path verbs are used (39). (39) δanociarse δo ‘to kneel down’

Trentino dialects

nar via ‘to go away’ rampegarse su ‘to climb up’ sentarse zó ‘to sit down’ culeté ju ‘to tumble down’

Ladin (Gardenese variety)

ficë/tucë ite ‘to stuff into’ scané ju ‘to climb down’ PVs can also be formed from non-motion verbs (e.g., Trent. star su ‘to stand upright’). A consequence of the spread of the construction is the high number of PVs with metaphorical meanings from both motion and non-motion verbs (40). (40) entrà sö enter up ‘to understand’ butàrse via throw oneself away ‘to despair’

Bergamasque dialect

Trentino dialects

648

III. Units and processes in word-formation II: Special cases meterse zó lay one-self down ‘to worsen’ serarse su tighten up ‘to cover oneself, shelter from the cold’ dir fora speak out ‘to blurt, reveal’

Venetian

A typical feature of this area is the wide use of PVs with aspectual values (41). (41) basarse su kiss each other on ‘to kiss each other repeatedly’

Trentino dialects

far su make up ‘to build up’ (cf. Cordin 2006: 221–222) dé su give up ‘to resign; cede’

Ladin (Gardenese variety)

fé ju make down ‘to undo’ fé via make away ‘to finish’ lascé do leave after ‘to reduce’ unì sëura to come over ‘to come to know’ ciapar su catch up ‘to collect everything’ magnar fora eat out ‘to squander’

Venetian

36. Particle verbs in Romance

649 Western Emiliano dialects

dèr so ‘to give up’ dèr vî ‘to give away’ dèr zå give down ‘to brush; to beat’ dîr so tell up ‘to call up’ fèr so do up ‘to tidy up’ mètter vî put away ‘to lock’ (cf. Begioni 2003)

A typical characteristic of Rhaeto-Romance varieties (cf. among others Meyer-Lübke 1899 Vol. 3: § 482; Siller-Runggaldier 2009) is the use of compound particles (see examples in Table 36.4) that allow a very detailed spatial encoding. These elements can also be used in post-verbal position, thus favoring the formation of PVs. Tab. 36.4: Examples of Ladin (Gardenese) compound particles ite ‘into’ do ‘back’

doite

sot ‘under’

sotite

sëura ‘over’

sëuraite

su ‘on’ dossù

ju ‘down’

sëura ‘over’

dojù

dovia sotsëura

sëurajù

via ‘away’

sotvia sëuravia

A peculiarity of some Venetian and Ladin varieties signaled by Vigolo (2007) is the selectional restriction among synonymic particles depending on the nature of the referent, cf. (42). (42) netà δu clean down (for glasses, walls, clothing) netà fora clean out (for rooms, stables)

650

III. Units and processes in word-formation II: Special cases netà su clean up (for the floor) netà via clean away (for tables, nose) The traditional view according to which the presence of PVs in the Romance languages depends on calques or loans from German (Rohlfs 1969: § 916, 918) has been criticized and rejected especially by specialists of Rhaeto-Romance (cf. among others Gsell 1982; Mair 1984; Kramer 1987). In these varieties there are of course singular PVs that can be explained as a calque from the corresponding German constructions (e.g., Ladin Badiot variety morì fòra from German aussterben), and contact with the German language has certainly facilitated the spread of such structures. However, the characteristics of PVs in the system of verbal spatial lexicalization lean towards the hypothesis that we are in the presence of a construction that developed on its own, and took its origin from Vulgar Latin. Corpus analyses recently conducted on various Northern dialects of Italy (cf. Bernini 2008; Cordin 2008a, b) have shown important differences with respect to German regarding the expression of path components (i.e. vector, conformation, deictic) encoded in the adverbal locus. Rather than the constructions present in the Germanic OV languages, Romance PVs more closely resemble the constructions employed in Germanic VO languages (see article 35 on particle verbs in Germanic, section 2). Constructions similar to the Surmiran example in (43) are very common in German (cf. er springt zum Fenster hinaus), while they are exceptional in all Romance languages, where particle stranding is ruled out and the insertion of heavy lexical elements between the verb and the particle is very constrained and limited in use (moreover, Siller-Runggaldier 2009 suggests that, contra Berthele 2006, the sentence in 43 is not to be interpreted as an instance of PV construction). (43) El saglia da fenestra or he jumps from the window out ‘He jumps out of the window.’

Surmiran

Currently, scholars agree that the influence of German did not so much determine the birth of PVs as extend the usage of a pre-existing construction consistent with the language’s system. The widespread use of PVs in all the Romance languages during the medieval period is a further element in support of this position.

7.4. Standard Italian and other dialects of Italy The recent interest in Italian PVs has been stimulated especially by Schwarze (1985) and Simone (1997). In the wake of these seminal articles a number of studies have emerged. Among the most recent I would like to mention Iacobini and Masini (2006), the papers collected by Cini (2008), Iacobini (2009b), Bernini (2012) and all the bibliographic references contained in these works.

36. Particle verbs in Romance

651

The characteristics of Italian PVs have already been described in sections 2 to 4. Lists of contemporary Italian PVs can be found in Simone (1997), Calvo Rigual (2008), Antelmi (2002), Iacobini (2008), Mosca (2010). The first two articles are mainly based on dictionary sources, the others on corpus analysis. Diachronic data may be found in Iacobini and Masini (2009), where a first quantitative account on the use of PVs in Italian from its origins to the present day is provided. An example of the hybrid (and to a certain extent redundant) system of motion expression in current Italian is given in Table 36.5 (taken from Iacobini and Masini 2006: 161) through the comparison of English PVs with the verb to go and their translation into Italian with both a synthetic form and a PV (cf. Talmy 2000: 145). It is important to observe that while some monolexemic verbs are more frequently used compared with their synonymous PVs, the two ways of expression display comparable colloquiality in the representation of the same type of motion event. Tab. 36.5: Italian analytic and synthetic counterparts of English PVs with to go English

Italian

Verb Root+Satellite

Verb Root

Verb Root+Satellite

to go after

inseguire

andare/correre dietro

to go ahead

procedere/continuare

andare avanti

to go away

andarsene

andare via

to go back

(ri)tornare

andare/tornare indietro

to go down

scendere

andare giù

to go for

avventarsi

andare/lanciarsi contro

to go in

entrare

andare dentro

to go on

continuare

andare avanti

to go out

uscire

andare fuori

to go (a)round

girare

andare attorno

to go up

salire

andare su

The more widespread use of PVs in contemporary standard Italian as compared to the other major Romance languages can be explained in terms of an internal change due to the interaction between dialects and standard language (cf. Jansen 2004; Iacobini 2009a). Italian was mainly confined to written usage for many centuries, and it was not until the second half of the 20th century that it became the language spoken by the majority of Italians. The most significant characteristic of 20th century Italian lies in the re-emergence of sub-standard features belonging to the Umgangssprache, which were present in the origins of Italian. They are features that have been kept and developed in dialects and regional spoken varieties, but have long been excluded from, or marginalized in standard Italian. Two independent studies (Benincà and Poletto 2006 on Venetian, and Amenta 2008 on Sicilian dialects) maintain that speakers of local dialects and regional Italian (with no competence of standard Italian) were those originally producing the

652

III. Units and processes in word-formation II: Special cases interference between local dialects and colloquial regional Italian varieties, but these speakers refrained from using PVs in more formal contexts because these were considered to be dialectally overconnotated. Italian speakers with less or no competence in a dialect and fewer socio-cultural inhibitions acquired the PVs circulating in the regional communities and extended their usage to more formal contexts. The strong presence of PVs in regional varieties of Northern Italian has played a positive role in the diffusion of these constructions in Italian, thanks to the migration (and the subsequent linguistic integration) of millions of people from Southern Italy to the industrial towns of Northern Italy after World War II, and to the fact that Northern varieties of Italian have gained in prestige. Nowadays, PVs have become a lexical resource of standard Italian that is almost independent of a regional distribution. Contemporary Italian PVs do not originate only from the Northern dialects. Although they are less frequent than in Northern dialects, they have been quite common in Tuscan dialects since the 13th century at least (cf. Meyer-Lübke 1899: § 370; Masini 2006). PVs have also been documented from the first texts in the other dialects of the central regions, and (although less used) in the insular and Southern dialects, cf. Rohlfs (1969: § 916, 918), Tekavčić (1972: 634−694), Loi Corvetto (1982), Cini (2002: 147−148), Amenta (2008), Simone (2008: 27−28), Iacobini (2009a). The presence of PVs in medieval Tuscan texts (as well as in other central and insular dialects) shifts back to in the Early and High-Middle Ages the presumed Germanic influence on Italian. It is an influence that can be regarded as a supporting effect to the diffusion and entrenchment of a construction resulting from an autonomous development of Romance varieties. This reveals itself as a catalyzer of processes that were already in use in Late Latin.

7.5. Romanian The two reference works on Romanian PVs are Vicario (1995) and Papahagi (2009), on the expression of motion in Romanian, see also Drăgan (2012). The first contains bibliographical indications and an examination of dictionaries and grammars, the latter adopts a typological cross-linguistic perspective and is enriched by data extracted from corpus analysis. Currently, after Italian Romanian is the standard language in which PVs are most numerous and most commonly used. However, according to Papahagi (2009), the proportion of path verbs used to encode the trajectory confirms the classification of Romanian as a verb-framed language. The list of Romanian post-verbal particles is quite rich: înǎuntru ‘inside’, afarǎ ‘outside’, înainte ‘forth, ahead’, înapoi, îndǎrǎt ‘back’, dincolo ‘beyond’, sus (în sus) ‘up’, jos (în jos) ‘down’, drept ‘straight’, la vale ‘downstream’, la deal ‘upstream’, la dreapta ‘to the right’, la stânga ‘to the left’. Romanian seems to be the only Romance language to have deictic particles (încoace ‘hither’, încolo ‘thither’) and an interrogative directional post-verbal particle (încotro ‘whereto’). The two denominal particles (aƫǎ thread ‘straight’, roatǎ wheel ‘around’) are probably due to a calque from Hungarian, a fact confirmed by their limited geographical distribution. PVs may result from both transitive (44a) and intransitive motion verbs (44b).

36. Particle verbs in Romance (44) a. un jucǎtor care a fost trimis afarǎ de pe terenul de joc ‘a player who has been sent out of the playing field’ înainte și era gata să treacă b. Sultana, ca o somnambulă, păși Sultana, as a sleepwalker, stepped forward and was about to pass.SBJ pragul biroului. threshold office ‘As a sleepwalker, Sultana stepped forward and was about to enter the office.’ (George Călinescu, Bietul Ioanide (1953). Chișinău: Litera, 2001, p. 43) The verb a da ‘to give’ is the one that combines with the highest number of particles (e.g., a da afară ‘to cast out’, înapoi ‘to retract, draw back’, înainte ‘to continue, progress’, încolo ‘to put aside’, jos ‘to get down’). The most used particle is înapoi, which can express both iteration (marfa vândutǎ nu se mai primeşte înapoi ‘sale goods cannot be exchanged’) and motion from a place (a trimite înapoi ‘to send back’). Romanian also has a verb that encodes pure motion a se duce ‘to go’ and a small number of constructions using support verbs (expressing nothing more than the fact of moving) and an element expressing either manner or direction (45). (45) a o lua la picior / la vale / la goanǎ to take it at foot / at valley / at run ‘to walk away’/‘to descend’/‘to start running’ a se da înapoi / jos / la vale to give oneself back / down / downstream ‘to go back’/‘to descend (vertical)’/‘to descend (a slope)’ Path verbs may combine with a redundant satellite (46a), sometimes specifying deixis (46b). (46) a. Ieşi afarǎ! exit out! ‘Get out!’ b. Vino încoace! come hither! ‘Come here!’ Manner is rarely encoded by the main verb, and when it is expressed, it tends to be lexicalized in general gaits (e.g., a alerga ‘to run, to escape’, a sǎri ‘to jump’). Manner verbs, however, are sometimes used in PVs also to express boundary-crossing events, cf. (47). (47) Petru a fugit în casǎ / înǎuntru. Peter ran in home / into ‘Peter ran home / into the house.’ Despite the fact that PVs are part of the system of Romanian, they are not very frequent in use. Many PVs correspond to synonymous monolexemic verbs (e.g., a se da jos / a

653

654

III. Units and processes in word-formation II: Special cases coborî ‘to descend’) that constitute the standard or the more formal variant, while the PVs are primarily distributed in spontaneous or informal communicative situations. According to Papahagi (2009), the satellite-framed characteristics displayed by Romanian can be interpreted as conservative features that characterize Romanian in comparison to the other Romance languages. Furthermore, contact with satellite-framed languages (German, Hungarian and Slavic languages) may play a positive role in the maintenance of PVs.

8. References Amenta, Luisa 2008 Esistono i verbi sintagmatici nel dialetto e nell’italiano regionale di Sicilia? In: Monica Cini (ed.), I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, 159−174. Frankfurt/M.: Lang. Antelmi, Donatella 2002 Il verbo senza significato: Possibilità di slittamento del contenuto lessicale su elementi di tipo nominale. Rivista italiana di linguistica e di dialettologia 4: 97−117. Batoréo Jakubowicz, Hanna 2000 Expressão do espaço no portugues europeu. Contributo psicolinguístico para o estudo da linguagem e cognição. Lisboa: Fundação Calouste Gulbenkian. Begioni, Louis 2003 Le costruzioni verbali V + indicatore spaziale nell’area dialettale dell’Appennino parmense. In: Mathée Giacomo-Marcellesi and Alvaro Rocchetti (eds.), Il verbo italiano. Studi diacronici, sincronici, contrastivi, didattici, 327−342. Roma: Bulzoni. Benincà, Paola and Cecilia Poletto 2006 Phrasal verbs in Venetan and Regional Italian. In: Frans Hinskens (ed.), Language Variation − European Perspectives, 9−22. Amsterdam/Philadelphia: Benjamins. Berman, Ruth and Dan I. Slobin 1994 Relating Events in Narrative. A Cross-Linguistic Developmental Study. Hillsdale, NJ: LEA. Bernini, Giuliano 2008 Per una definizione di verbi sintagmatici: La prospettiva dialettale. In: Monica Cini (ed.), I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, 121−138. Frankfurt/M.: Lang. Bernini, Giuliano 2012 Lexicalization and grammaticalization in the area of multi-word verbs: A case-study form Italo-Romance. In: Valentina Bambini, Irene Ricci and Pier Marco Bertinetto (eds.), Language and the brain − Semantics, 131−154. Roma: Bulzoni. Berthele, Raphael 2006 Ort und Weg. Eine vergleichende Untersuchung der sprachlichen Raumreferenz in Varietäten des Deutschen, Rätoromanischen und Französichen. Berlin/New York: de Gruyter. Buridant, Claude 2000 Grammaire nouvelle de l’ancien française. Paris: Sedes. Burnett, Heather and Mireille Tremblay 2009 Variable-Behaviour Ps and the Location of PATH in Old French. In: Enoch O. Aboh, Elisabeth van der Linden, Josep Quer and Petra Sleeman (eds.), Romance Languages and Linguistic Theory 2007, 25−50. Amsterdam/Philadelphia: Benjamins. Burnett, Heather and Mireille Tremblay 2012 The evolution of the encoding of direction in the history of French: A quantitative approach to argument structure change. In: Ans M. C. van Kemenade and Nynke de Haas (eds.), Historical Linguistics 2009, 333−354. Amsterdam/Philadelphia: Benjamins.

36. Particle verbs in Romance Cadierno, Teresa and Peter Robinson 2009 Language typology, task complexity and the development of L2 lexicalization patterns for describing motion events. Annual Review of Cognitive Linguistics 7: 245−276. Calvo Rigual, Cesáreo 2008 I verbi sintagmatici italiani: Appunti contrastivi con lo spagnolo e il catalano. In: Carmen González Royo and Pedro Mogorrón (eds.), Estudios y análisis de fraseología contrastiva. Lexicografía y traducción, 47−66. Alicante: Universidad de Alicante. Calvo Rigual, Cesáreo 2010 Trattamento nella lessicografia monolingue (italiana) e bilingue (italiano−spagnolo e catalano) dei verbi sintagmatici: Panorama attuale e proposte di futuro. In: Maria Iliescu, Heidi Siller-Runggaldier and Paul Danler (eds.), Actes du XXVe Congrès International de Linguistique et Philologie Romanes. Vol. 7, 375−383. Berlin/New York: de Gruyter. Cifuentes-Férez, Paula 2009 A Crosslinguistic Study on the Semantics of Motion Verbs in English and Spanish. München: LINCOM Europa. Cini, Monica 2002 I verbi sintagmatici negli etnotesti dell’ALEPO. In: Giovanna Marcato (ed.), La dialettologia oltre il 2001, 143−150. Padua: Unipress. Cini, Monica (ed.) 2008 I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca. Frankfurt/M.: Lang. Cordin, Patrizia 2006 Su e giù modificatori del verbo in alcune varietà dell’italiano. In: Nicola Grandi and Gabriele Iannaccaro (eds.), Zhì. Scritti in onore di Emanuele Banfi in occasione del suo 60° compleanno, 215−225. Cesena/Roma: Caissa Italia. Cordin, Patrizia 2008a L’espressione di tratti aspettuali nei verbi analitici dei dialetti trentini. In: Monica Cini (ed.), I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, 175−192. Frankfurt/M.: Lang. Cordin, Patrizia 2008b Spazio fisico e spazio figurato nelle collocazioni verbo più locativo in italiano e in alcune sue varietà. In: Gerhard Bernhard and Heidi Siller-Runggaldier (eds.), Sprache im Raum − Raum in der Sprache, 3−20. Frankfurt/M.: Lang. Cordin, Patrizia 2011 Le costruzioni verbo-locativo in area romanza. Dallo spazio all’aspetto. Berlin/New York: de Gruyter. De Mauro, Tullio, Federico Mancini, Massimo Vedovelli and Mirian Voghera (eds.) 1993 Lessico di frequenza dell’italiano parlato. Milano: Etas Libri. Diez, Friedrich 1882 Grammatik der romanischen Sprachen. 5th ed. Bonn: Weber. Drăgan, Ruxandra 2012 Aspects of Lexical Structure. Verbs in Locative Constructions in English and Romanian. Bucureşti: Editura Universităţii din Bucureşti. Dufresne, Monique, Fernande Dupuis and Marie-Catherine Longtin 2001 Un changement dans la diachronie du français: La perte de la préfixation aspectuelle en a-. Revue québecoise de linguistique 29: 33−54. Dufresne, Monique, Fernande Dupuis and Mireille Tremblay 2003 Preverbs and particles in Old French. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2003, 33−60. Dordrecht: Kluwer. González Fernández, María Jesús 1997 Sobre la motivación semántica de las expresiones pleonásticas de movimiento: subir arriba, bajar abajo, entrar adentro y salir afuera. In: Concepción Company Company

655

656

III. Units and processes in word-formation II: Special cases (ed.), Cambios diacrónicos en el español, 123−141. México: Universidad Nacional Autónoma de México. Gsell, Otto 1982 Las rosas dattan ora − les röses da fora − le rose danno fuori: Verbalperiphrasen mit Ortsadverb im Rätoromanischen und im Italienischen. In: Sieglinde Heinz and Ulrich Wandruszka (eds.), Fakten und Theorien. Beiträge zur romanischen und allgemeinen Sprachwissenschaft. Festschrift für Helmut Stimm zum 65. Geburtstag, 71−85. Tübingen: Narr. Guglielmo, Daniela 2013 A particle-centred approach on Italian verb-particle constructions. Ph.D. dissertation, University of Salerno. Guo, Jiansheng, Elena Lieven, Nancy Budwig, Susan Ervin-Tripp, Keiko Nakamura and Şeyda Özçalişkan (eds.) 2009 Crosslinguistic Approaches to the Psychology of Language. Research in the Tradition of Dan Isaac Slobin. New York: Psychology Press. Herslund, Michael 2005 Lingue endocentriche e lingue esocentriche: aspetti storici del lessico. In: Iørn Korzen and Carla Marello (eds.), Tipologia linguistica e società, 19−30. Firenze: Franco Cesati. Hijazo-Gascón, Alberto and Iraide Ibarretxe-Antuñano 2010 Tipología, lexicalización y dialectología aragonesa. Archivo de Filología Aragonesa 66: 245−280. Hijazo-Gascón, Alberto and Iraide Ibarretxe-Antuñano 2013 Las lenguas románicas y la tipología de los eventos de movimiento. Romanische Forschungen 125: 467−494. Iacobini, Claudio 2008 Presenza e uso dei verbi sintagmatici nel parlato dell’italiano. In: Monica Cini (ed.), I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, 103−119. Frankfurt/M.: Lang. Iacobini, Claudio 2009a The role of dialects in the emergence of Italian phrasal verbs. Morphology 19: 15−44. Iacobini, Claudio 2009b Phrasal verbs between syntax and lexicon. Italian Journal of Linguistics 21: 97−117. Iacobini, Claudio 2012 Grammaticalization and innovation in the encoding of motion events. Folia Linguistica 46: 359−385. Iacobini, Claudio and Benjamin Fagard 2011 A diachronic approach to variation and change in the typology of motion event expression: A case study: From Latin to Romance. Faits de Langues − Les Cahiers 3: 152− 171. Iacobini, Claudio and Francesca Masini 2006 The emergence of verb-particle constructions in Italian: Locative and actional meanings. Morphology 16: 155−188. Iacobini, Claudio and Francesca Masini 2009 I verbi sintagmatici dell’italiano fra innovazione e persistenza: Il ruolo dei dialetti. In: Anna Cardinaletti and Nicola Munaro (eds.), Italiano, italiano regionali e dialetti, 115− 135. Milan: FrancoAngeli. Jansen, Hanne 2004 La “particella spaziale” e il suo combinarsi con verbi di movimento nell’italiano contemporaneo. In: Paolo D’Achille (ed.), Generi, architetture e forme testuali, 129−144. Firenze: Cesati. King, Robert 2000 The Lexical Bases of Grammatical Borrowing. Amsterdam/Philadelphia: Benjamins.

36. Particle verbs in Romance Kopecka, Anetta 2009a Continuity and change in the representation of motion events in French. In: Jiansheng Guo, Elena Lieven, Nancy Budwig, Susan Ervin-Tripp, Keiko Nakamura and Şeyda Özçalişkan (eds.), Crosslinguistic Approaches to the Psychology of Language. Research in the Tradition of Dan Isaac Slobin, 415−426. New York: Psychology Press. Kopecka, Anetta 2009b L’expression du déplacement en français: L’interaction des facteurs sémantiques, aspectuels et pragmatiques dans la construction du sens spatial. Langages 173: 54−75. Kopecka, Anetta forthc. From a satellite- to a verb-framed pattern: A typological shift in French. In: Hubert Cuyckens, Walter De Mulder, Michèle Goyens and Tanja Mortelmans (eds.), Variation and Change in Adpositions of Movement. Amsterdam/Philadelphia: Benjamins. Kramer, Johannes 1981 Die Übernahme der deutschen und der niederländischen Konstruktion Verb + Verbzusatz durch die Nachbarsprachen. In: Wolfgang Meid and Karin Heller (eds.), Sprachkontakt als Ursache von Veränderungen der Sprach- und Bewusstseinsstruktur, 129−140. Innsbruck: Institut für Sprachwissenschaft. Kramer, Johannes 1987 Tedeschismi e pseudo-tedeschismi nel ladino e altrove. Quaderni Patavini di Linguistica 6: 9−30. Levin, Beth, John Beavers and Shiao Wei Tham 2009 Manner of motion roots across languages: Same or different? Conference held at the workshop: Roots Word formation from the perspective of “core lexical elements”, Stuttgart, June 10−12, 2009. http://www.stanford.edu/~bclevin/stutt09mot.pdf [last access 1 Oct 2014]. Loi Corvetto, Ines 1982 L’italiano regionale di Sardegna. Bologna: Zanichelli. Loporcaro, Michele 2007 On triple auxiliation in Romance. Linguistics 45: 173−222. Lüdtke, Jens 1996 Gemeinromanische Tendenzen IV. Wortbildungslehre. In: Günter Holtus, Michael Metzeltin and Christian Schmitt (eds.), Lexikon der Romanistischen Linguistik. Vol. 2, 235− 272. Tübingen: Niemeyer. Mair, Wolfgang 1984 Transferenz oder autonome Bildung? Zeitschrift für romanische Philologie 100: 408− 432. Martin, Robert 2001 Le préfixe a-/ad- en moyen français. Romania 119: 289−322. Masini, Francesca 2006 Diacronia dei verbi sintagmatici in italiano. Archivio Glottologico Italiano 91: 67−105. Masini, Francesca 2008 Verbi sintagmatici e ordine delle parole. In: Monica Cini (ed.), I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, 83−102, Frankfurt/M.: Lang. Mateu, Jaume and Gemma Rigau 2010 Verb-particle constructions in Romance: A lexical-syntactic account. Probus 22: 241− 269. Meyer-Lübke, Wilhelm 1899 Grammatik der Romanischen Sprachen. Vol. 3: Romanische Syntax. Leipzig: Reisland. Mosca, Monica 2010 L’evento di moto in italiano tra sintassi e semantica. Pisa: Pisa University Press.

657

658

III. Units and processes in word-formation II: Special cases Nikitina, Tatiana 2008 Pragmatic Factors and Variation in the Expression of Spatial Goals: The Case of into vs. in. In: Anna Asbury, Jakub Dotlačil, Berit Gehrke and Rick Nouwen (eds.), Syntax and Semantics of Spatial P, 175−209. Amsterdam/Philadelphia: Benjamins. Papahagi, Cristiana 2009 La place du roumain dans la typologie motion event: Esquisse de description. In: Gavril Neamţu, Ştefan Gencărău and Adrian Chircu (eds.), Limba română. Abordări tradiţionale şi moderne, 397−408. Cluj-Napoca: Presa Universitară Clujeană. Porquier, Rémy 2001 ‘Il m’a sauté dessus’, ‘je lui ai couru après’: Un cas de postposition en français. Journal of French Language Studies 11: 123−134. Porquier, Rémy 2003 ‘Gli corro dietro/Je lui cours après’: A propos d’une construction verbale spécifique en italien et en français. In: Mathée Giacomo-Marcellesi and Alvaro Rocchetti (eds.), Il verbo italiano. Studi diacronici, sincronici, contrastivi, didattici, 491−500. Roma: Bulzoni. Quaglia, Stefano 2012 On the Syntax of some Apparent Spatial Particles in Italian. In: Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG12 Conference. Standford: CSLI. http:// web.stanford.edu/group/cslipublications/cslipublications/LFG/17/papers/ lfg12quaglia.pdf [last access 1 Oct 2014]. Rohlfs, Gerhard 1969 Grammatica storica della lingua italiana e dei suoi dialetti. Turin: Einaudi [= revised version of Historische Grammatik der italienischen Sprache und ihre Mundarten. Bern: Francke 1949−1954]. Rojo, Guillermo 1974 Perífrasis verbales en el gallego actual. Vigo: Universidad de Santiago de Compostela. Schøsler, Lene 2008 L’expression des traits manière et direction des verbes de mouvement: Perspectives diachroniques et typologiques. In: Elisabeth Stark, Roland Schmidt-Riese and Eva Stoll (eds.), Romanische Syntax im Wandel, 113−132. Tübingen: Narr. Schwarze, Christoph 1985 “Uscire” e “andare fuori”: Struttura sintattica e semantica lessicale. In: Annalisa Franchi de Bellis and Leonardo Maria Savoia (eds.), Sintassi e morfologia della lingua italiana d’uso. Teorie e applicazioni descrittive, 355−371. Roma: Bulzoni. Schwarze, Christoph 2008 I verbi sintagmatici: Prospettive di ricerca. In: Monica Cini (ed.), I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, 209−224. Frankfurt/M.: Lang. Siller-Runggaldier, Heidi 2009 Review of: Raphael Bertele Ort und Weg. Berlin/New York: de Gruyter, 2006. Zeitschrift für romanische Philologie 125: 145−153. Simone, Raffaele 1997 Esistono verbi sintagmatici in italiano? In: Tullio De Mauro and Vincenzo Lo Cascio (eds.), Lessico e grammatica. Teorie linguistiche e applicazioni lessicografiche, 155− 170. Roma: Bulzoni. Simone, Raffaele 2008 I verbi sintagmatici come costruzione e come categoria. In: Monica Cini (ed.), I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, 13−30. Frankfurt/M.: Lang.

36. Particle verbs in Romance

659

Slobin, Dan Isaac 1996 Two ways to travel: Verbs of motion in English and Spanish. In: Masayoshi Shibatani and Sandra Thompson (eds.), Grammatical Constructions, 195−219. Oxford: Clarendon Press. Slobin, Dan Isaac 2004 The many ways to search for a frog. In: Sven Strömqvist and Ludo Verhoeven (eds.), Relating Events in Narrative. Vol. 2: Typological and Contextual Perspectives, 219− 257. Mahwah, NJ: LEA. Spreafico, Lorenzo 2009 Problemi di tipologia lessicale. Roma: Bulzoni. Stolova, Natalya 2008 From satellite-framed Latin to verb-framed Romance: Late Latin as an intermediate stage. In: Roger Wright (ed.), Latin vulgaire − latin tardif VIII, 253−262. Hildesheim: Olms. Talmy, Leonard 1985 Lexicalization patterns: Semantic structure in lexical forms. In: Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. 3: Grammatical Categories and the Lexicon, 57−149. Cambridge: Cambridge University Press. Talmy, Leonard 2000 Toward a Cognitive Semantics. Typology and Process in Concept Structuring. Vol. 2. Cambridge, MA: MIT Press. Tekavčić, Pavao 1972 Grammatica storica dell’italiano. Vol. 2: Morfosintassi. Bologna: Il Mulino. Treffers-Daller, Jeanine 2012 Grammatical collocations and verb-particle constructions in Brussels French: A corpuslinguistic approach to transfer. International Journal of Bilingualism 16: 53−82. Troberg, Michelle and Heather Burnett 2011 Resultative secondary predication in medieval French and the satellite-framed/verbframed distinction. In: Workshop on verbal elasticity. Framing the verb/satellite distinction from a biolinguistic perspective. October 3−5, 2011, Centre de Lingüística Teòrica, Universitat Autònoma de Barcelona. https://focus.library.utoronto.ca/attachments/0000/ 1172/Verbal_Elasticity_Troberg_Burnett2011.pdf [last access 1 Oct 2014]. Vicario, Federico 1995 Sul tipo a da afară, a veni înapoi: Verbi con avverbio in rumeno. Revue de Linguistique Roumaine 40: 149−164. Vicario, Federico 1997 I verbi analitici in friulano. Milano: FrancoAngeli. Vigolo, Maria Teresa 2007 Indicatori spaziali in dialetti alto-veneti e in cadorino. In: Rosanna Maschi, Nicoletta Pennello and Piera Rizzolati (eds.), Miscellanea di studi offerta a Laura Vanelli, 353− 363. Udine: Forum.

Claudio Iacobini, Salerno (Italy)

660

III. Units and processes in word-formation II: Special cases

37. Particle verbs in Hungarian 1. 2. 3. 4. 5. 6. 7. 8. 9.

The status of verbal particles and particle verbs Defining the verbal particle, and the verbal particle stock Becoming a verbal particle: grammaticalization processes The meaning and functions of verbal particles Presence or absence of verbal particles in the sentence Patterns of particle verbs The syntactic behavior of particle verbs Outlook for the future References

Abstract This article surveys the most important problems concerning Hungarian verbal particles and particle verbs. Particle verbs, on the one hand, form a lexical unit, which is established via a morphological process, but on the other hand, in the sentence they behave as syntactic structures, whose parts can be divided. The determination of the verbal particle stock as well as that of the verbal particle function and presence in the sentence must be considered preliminary. Due to both the many facets of the verbal particles and the different theoretical frameworks in which the research is carried out, the problems of the particle verbs will and must partly remain conjectural.

1. The status of verbal particles and particle verbs In modern Hungarian morphology, the formation of particle verbs is considered a morphological operation and is dealt with as a special means of derivation in the realm of word-formation (Kiefer and Ladányi 2000: 453; Kiefer 2006: 120). The verbal particle (just like the verbal prefixes of other languages) is canonically placed directly before the verb (e.g., be-megy in-go ‘to go in’, ki-jön out-come ‘to come out’, fel-repül up-fly ‘to fly up’, meg-ír (valamit) megPERF-write (sth.) ‘to write down’, el-olvas (valamit) elPERF-read (sth.) ‘to read (sth.) through’). However, the elements of the particle verbs can be separated in the sentence (see section 6), consequently the verbal particle cannot be regarded as a verbal prefix, and the particle verb in Hungarian, in spite of being a lexical unit, belongs not only in the scope of morphology, i.e. word-formation, but also in that of syntax. In syntax, verbal particles are classified among the verbal modifiers, because they cannot usually be distinguished from the latter on the basis of their syntactic behavior. In Hungarian, aside from verbal particles, bare nominal complements are also typical verbal modifiers, e.g., szén-né éget (valamit) coal-to burn-CAUS (sth.) ‘to carbonize (sth.)’, újság-ot olvas newspaper-ACC read ‘to read a newspaper’, piros-ra fest (valamit) red-onto paint (sth.) ‘to paint (sth.) red’, okos-nak tart (valakit) clever-to consider (sb.) ‘to consider sb. clever’ (cf. Komlósy 1992: 500). Preverbal verbal modifiers, such as

37. Particle verbs in Hungarian verbal particles, form a phonological and semantic unit with their verb. However, the verbal particle differs from the other verbal modifiers in that, unlike them, a) it cannot satisfy any of the basic verb’s arguments, i.e. it cannot be governed by the verb, and b) the particle verb can be the input into a morphological rule, whereas complexes formed by the other verbal modifiers cannot take part in secondary word-formation (cf. Kiefer 2003: 177−178, 181−185). Hungarian particle verbs are referred to as both verbal prefixes and preverbs in the English linguistic literature. With the term verbal prefix it is the morphological character of the particle verb’s formation and the lexical unity of the verbal particle with the verb that is stressed, while the term preverb emphasizes the canonical preverbal position of the verbal particle.

2. Defining the verbal particle, and the verbal particle stock There is a partial overlap between the category of verbal particles and the category of adverbs, because the bulk of the verbal particles have grammaticalized from directional adverbs and have kept their original directional meanings as polysemantic units. The same lexical unit after the verb, i.e. in a non-canonical position, can often be understood as both verbal particle and adverb; in the case of some preverbal adverbial elements syntactically behaving like verbal modifiers the question arises as to whether or not they should be classified as new verbal particles. This is the chief reason why the transitional character of the category of verbal particle was emphasized in 20th-century Hungarian linguistics. It remains to be seen on what grounds an element can be grouped under the category of verbal particles. Beöthy and Altmann (1985) evidently start from a wider syntactic category of verbal modifier, assuming that Hungarian has around 100 verbal particles (in their terminology: verbal prefixes), whereas Jakab (1982: 65−66), on the basis of lative-type directional meaning and lexical/textual density, excludes 39 out of 75 verbalparticle-like elements chosen from special literature. Kiefer and Ladányi (2000: 463− 465, 481) define the so-called productive verbal particle stock considering the following three criteria: a) non-argument status; b) marking of aspectuality/perfectivity (see section 4.1) and c) occurrence in productive patterns of particle verbs (this requirement was mainly drawn into the research in connection with defining the newer, controversial elements as verbal particles on the basis of Ladányi 2000, see section 3). On the strength of the above three guidelines, according to Kiefer and Ladányi (2000: 482), and also D. Mátai (2007: 86), the stock of Hungarian verbal particles consists of the following: a) verbal particles originating from adverbs − tovább ‘further, on(ward)’, újra ‘re-’, végig ‘to the (very) end, throughout’, vissza ‘back’, át ‘across, through’, keresztül ‘across, through’, túl ‘over’; b) verbal particles also used with possessive endings, which originated in parallel with postpositions − alá ‘under(neath)’, elé ‘before, in front of’, fölé ‘above, over’, mellé ‘next to, quite near’, mögé ‘behind’, utána ‘after’ as well as verbal particles that are closely related to case-endings: hozzá ‘to(wards)’, neki ‘to, at’, rá ‘on(to)’;

661

662

III. Units and processes in word-formation II: Special cases c) verbal particles with directional meanings: be ‘in’, bele ‘into, inside’, el ‘away’, fel ‘up’, félre ‘aside’, hátra ‘back’, ide ‘here’, ki ‘out (of)’, körül ‘around’, le ‘down’, oda ‘there’, össze ‘together’, szét ‘apart, asunder’; d) verbal particles that cannot be classified among any of the above-mentioned groups: agyon ‘to death, to excess’, megPERF and tönkre ‘(becoming) totally ruined’. The primary meaning of the bulk of the verbal particles in points a) to c) is some kind of directionality, which equals Jakab’s (1982) criterion relating to the lative meaning. Kiefer (2003: 185−186) also states that locative adverbs (e.g., fent ‘up’, kint ‘outside’, itt ‘here’, ott ‘there’), despite their syntactic behavior, cannot be looked upon as verbal particles, nor as verbal modifiers, because in their combination both the adverb and the verb can be stressed independently of one another, thus indicating that the adverb is unable to form a semantic unit with the verb. These elements have other qualities that distinguish them from verbal particles as well. At the same time, É. Kiss (2005, 2006) takes into consideration locative verbal particles as well (see section 5). Presuming prototype categories instead of classical ones, prototypical verbal particles can be defined by the mechanic application of the three criteria stated above. Verbal particle being a category “perpetually widening during language history” (D. Mátai 2007: 83), there exist non-prototypical verbal particles only partially satisfying the above guidelines. In case productivity is defined not as a yes-no property, but a gradual one, the elements excluded in Kiefer and Ladányi (2000) because of lack of productivity will be found on the category’s periphery if compared to the ones enumerated in the stock of productive verbal particles.

3. Becoming a verbal particle: grammaticalization processes Verbal particles have been grammaticalized partly from local adverbs, partly from suffixed nouns. However, in the case of the so-called primary verbal particles (megPERF, el ‘away, from’, be ‘into’, ki ‘out (of)’, fel ‘up’, le ‘down’; cf. J. Soltész 1959) this grammaticalization process has not yet even arrived at the phase of univerbation/morphologization, therefore in this instance we cannot speak of verbal prefixes. In Pais’ (1959) view, the predecessors of the most ancient verbal particles are appositive-like adverbs without meaning independent of a situation that, together with contentful adverbial noun phrases, could also be found in preverbal position in the sentence (az erdő-ből, ki | jön DET forest-out, out come ‘out of the forest, (s)he comes’ → az erdőből ki-jön DET forest-out out-come ‘to come out of the forest’). According to Forgács (2004: 55), the occurrence of these antecedents to the verbal prefix (“protopreverbs” in his terminology) may also have been supported by communicative and pragmatic factors manifested in accentual conditions; their preverbal position enabled their lexicalization with the verb, i.e. their grammaticalization into verbal particles. As another source of verbal particles, Forgács (2004: 57) names the postpositional elements in possessive constructions. Such a construction is for example N-DAT/GEN + alá ‘under’ (originally ‘to the bottom of sth.’), e.g., az ágy-nak al-á | búj-ik DET bed-DAT/GEN bottom-to hide-3SG ‘under the bed, (s)he creeps’ → alá-búj-ik (az ágy-nak) under-hide-3SG (DET bed-DAT/GEN) ‘to creep under (the bed)’. (In D. Mátai’s refinement − cf. 2007: 86 − both the VPs and the postpositions may have concurrently developed from suffixed-

37. Particle verbs in Hungarian noun and adverbial antecedents in genitive constructions.) By contrast, the newer verbal particles, being enriched with new semantic content in lexicalized verbal complexes, have their source in suffixed nouns (cf. Ladányi 2000; Forgács 2004: 60; D. Mátai 2007: 87). For example, the original meaning of agy-on üt brain-on hit ‘to kill’ was ‘to hit on the head’. Later on, it acquired the new meaning ‘to kill’, and as a consequence agyon ‘to death, to excess’ could also be used along with other verbs expressing the semantic content ‘to kill’, e.g., agyon-lő ‘to shoot sb. dead’. Finally, through hyperbolic usages (e.g., agyon-gyötör ‘to torture sb. to death’) the meaning of agyon attains the aktionsart meanings intensity (e.g., agyon-díszít ‘to overdecorate’) and exhaustivity (agyon-dolgozza magát ‘to work oneself to death’); cf. Ladányi (2000). Forgács (2004: 60−64) calls this type verbal particles derived from idiom chunks. The semantic change connected with grammaticalization is closely related not only to the acquisition of the status of a verbal particle, that is the category shift of the elements, but also to their connectivity. In the case of relatively new verbal particles this is an easily graspable phenomenon. The formation of new, grammaticalized meaning(s) considerably extends the element’s sphere of usage, therefore among the disputed properties of the productivity of verbal-particle-like elements, according to Ladányi (2000), can be the index of category shift, i.e. the verbal-particle (in her terminology: “verbalprefix”) status. According to Kiefer and Honti (2003), in primary Hungarian verbal particles it was the aspectual meanings that first grammaticalized out of the local meanings, and to them were later added aktionsart meanings, whereas in the case of newer verbal particles the formation of the aspectual meaning was preceded by that of some aktionsart meanings (cf. Ladányi 2000). The grammaticalization process is a continuum in which the stratification of meanings plays an important role (cf., e.g., Hopper and Traugott 2003: 78). Since the grammaticalized unit often keeps its older, more specific meaning or a remainder thereof, its usages can be arranged as parts of a chain connected together; these are the stages of the grammaticalization path that appear as polysemantic meanings in synchrony (Bybee, Perkins and Pagliuca 1994: 17−18). In connection with Hungarian agyon ‘to death, to excess’ and tönkre ‘(becoming) totally ruined’, Ladányi (2000) sketches a possible grammaticalization path, which is consistent with the accepted position on the formation of verbal particles in the field of Hungarian historical linguistics (cf. D. Mátai 2007: 87).

4. The meaning and functions of verbal particles Hungarian verbal particles (like verbal prefixes and particles of other languages) are polysemantic elements. Their meanings, which come about by way of semantic changes, are interconnected and create a system. According to Kiefer (2006: 48, 120), the meaning of verbal particles can contribute in a variety of ways to the compound meaning of particle verbs with compositional meaning. The author assigns three functions to the verbal particles: a) a purely perfectivizing function, b) an aktionsart-forming function accompanied by perfectivization, and c) perfectivization and a compound-like function − this latter function is connected with the verbal particles’ lexical (directional or direction-like) meanings.

663

664

III. Units and processes in word-formation II: Special cases

4.1. Aspectual meaning In Hungarian, unlike, for example, in Russian, aspect is not merely a morphologically determined category, though we can talk of verbal aspect in Hungarian as well (see, e.g., Kiefer 2006; Péter 2008) − thus Hungarian is situated between the so-called aspectual and non-aspectual languages (cf. Kiefer 2006: 14−15). In the case of productive wordformation, the particle verb belongs to the perfective aspect on the lexical-semantic level. It may also happen that the verbal particle is attached to a perfective verb, or the particle verb is aspectually continuous, yet these are lexicalized, atypical cases (cf. Kiefer 2006: 43−44). Rarely do the particle verb and its particleless counterpart form a purely aspectual pair; the verbal particle usually denotes the perfective aspect along with its local or aktionsart meaning. Péter (2008: 3) identifies the perfective aspect with viewing the action or process in its totality (which corresponds to the so-called Slavic aspect concept), in the expression of which, owing to the immaturity of the Hungarian aspect category (cf. also Wacha 2001), different lexical-semantic, syntactic and pragmatic factors cooperate. By verbal aspect Kiefer (2006: 26) understands the sum of those qualities whereby the verb contributes to the aspect of the sentence; according to his definition, sentence aspect expresses the inner temporal structure of the sentence. The aspect of the Hungarian sentence is largely compositional; it is also affected by factors other than verbal aspect like “the type of the object, temporal modifier, numeral construction, adjectival or nominal expression in resultative meaning and local adverb in terminal role” (Gyuris and Kiefer 2008: 240). The aspect of the sentence coincides with the verbal aspect only when other factors do not override it. Since productively formed particle verbs as lexical units express perfective aspect, particle verbs in the canonical order verbal particle + verb occur in perfective sentences, and only they may be used with a non-durative modifier (cf. Kiefer 2006: 22): (1)

a. A könyv-et három nap-on belül el-olvas-t-am. DET book-ACC three day-on inside elPERF-read-PAST-1SG ‘I read the book within three days.’ vs. b. *A könyve-t három nap-on át el-olvas-om. DET book-ACC three day-on during elPERF-read-1SG(DEF) ‘I read the book during three days.’

4.2. Aktionsart meaning Based on Isačenko (1962: 385−418), Kiefer (2006: 144) defines aktionsart as “the morphologically compound verb’s additional semantic quality, productively introduced by suffixation or verbal particle”. In Hungarian, aktionsart meanings are chiefly conveyed by verbal particles, with the exception of the iterative (expressing regular or irregular repetition) and diminutive (expressing decreased intensity) aktionsart, which are represented by the suffix -gat/-get (e.g., sétál-gat walk-FREQ ‘to walk about’, int-eget waveFREQ ‘to wave one’s hand to’).

37. Particle verbs in Hungarian The verbal particle in its aktionsart-forming function not only introduces an aktionsart quality, but also perfectivizes the verb (cf. Kiefer and Ladányi 2000: 464) and vice versa: the perfectivizing function is usually accompanied by some type of resultativity. This semantic element is secondarily present in several more specific aktionsart meanings such as total, saturative, terminative, exhaustive and intensive aktionsarten as well (see Kiefer 2006: 148, 172). The following enumeration, relying on Kiefer and Ladányi (2000: 475−480) and Kiefer (2006: 149−181), lists the different aktionsart meanings that can be expressed by verbal particles: a) r e s u l t a t i v e (expressing that the action is completed and results in sth.): meg, ki, be, el, fel, le − e.g., fel-söpri az udvart ‘to sweep up the yard’, le-szedi az asztalt ‘to clear the table’, meg-írja a levelet ‘to write a letter’ b) t e r m i n a t i v e ( expressing that the action is completed): el − e.g., el-énekel ‘to sing sth. (to the end)’ c) i n c h o a t i v e / i n g r e s s i v e ( expressing the beginning of the action): verbal particle el and pronoun magát ‘oneself’ − e.g., el-sírja magát ‘to break out crying’, el-neveti magát ‘to burst out laughing’ d) s e m e l f a c t i v e (expressing single occurrence): meg − e.g., meg-kondul ‘to (begin to) toll’, meg-kavar ‘to stir’ e) d e l i m i t a t i v e (expressing temporal limitation): el − e.g., el-ábrándozik ‘to be lost in reveries’, el-beszélget (valakivel) ‘to chat (with sb.)’ f) t o t a l (expressing that the action/process affects (almost) the whole surface of sth.): be − e.g., be-ken ‘to spread sth. over sth.’, be-jár ‘to walk all over’, be-porosodik ‘to get dusty’ g) s a t u r a t i v e (expressing that the action/process is brought to full satisfaction): be; verbal particle ki and pronoun magát ‘oneself’ − e.g., be-eszik ‘to eat one’s fill’, kialussza magát ‘to have one’s sleep out, to sleep one’s fill’ h) i n t e n s i v e (expressing that the action/process is excessive): agyon, tönkre − e.g., agyon-díszít ‘to overdecorate’, tönkre-szárad ‘to dry excessively’ i) e x h a u s t i v e (expressing that the agent performs the action until tired/exhausted): verbal particles agyon or tönkre and pronoun magát ‘oneself’ − e.g., agyon-tanulja magát ‘to learn too much’, tönkre-dolgozza magát ‘to overwork oneself’ j) s u b m e r s i v e (expressing getting into an intensive state and submerging in it): be − e.g., be-szerelmesedik ‘to fall in love (with)’, be-szomorkodik ‘to become distressed’, be-hisztizik ‘to throw a fit’ (all the examples are neologisms). The frequentative aktionsart refers to irregular repetition. There is no particular verbal particle for this meaning. It is expressed by the reduplication of the verbal particle contained by the particle verb − e.g., el-el-olvas ‘to read through (sometimes)’, be-benéz ‘to look in on sb. (sometimes)’. Kiefer (2006: 176) defines terminative aktionsart as a special subtype of resultative aktionsart, which is expressed by the verbal particle el (productively with verbs of saying and singing). Submersive is a new aktionsart suggested by Nádasdy (2003: 287); in a corpus-based analysis of neologisms with the verbal particle be, Ladányi (2004) confirmed the existence of this aktionsart, and also defined its characteristic features. In his detailed overview of Hungarian aktionsarten, Kiefer (2006: 149−181) describes the barri-

665

666

III. Units and processes in word-formation II: Special cases ers to, and consequences of, aktionsart formation. Kiefer (2010) points out the areal relationships of the formation of aktionsart meanings. The aktionsart meanings expressed by the particle verbs may be complex as well, especially in neologisms (cf. the observations made by Nádasdy 2003; Pátrovics 2003 and Ladányi 2004). For example, according to Ladányi (2004), the submersive aktionsart is linked to the aktionsart meanings of inchoativity and intensity. The formation of aktionsart meanings is a result of grammaticalization processes, accompanied by further processes of meaning change. Because of this − though a polysemantic verbal particle may keep its directional meaning besides its aktionsart meaning(s) − the directional meaning and the aktionsart meaning(s) cannot occur together. The aktionsart meaning can change the argument structure of the word (e.g., be-jár a város-ba in-go DET town-into ‘to visit the city’ − directional meaning, be-jár-ja a várost in-go-3SG(DEF) DET town-ACC ‘to go round the town’ − totality: aktionsart meaning).

4.3. Lexical meaning − compound-like function About the verbal particles enumerated among the productive verbal particle stock under section 2 we can say that most of them have a lexical (directional or direction-like) meaning, too, though there is a difference as to how specific this directional meaning is. For example, directionally the verbal particle el ‘away’ (denoting departure from a given point to an unspecified direction) is less specific than other verbal particles denoting direction, like be ‘in’, ki ‘out’, le ‘down’, fel ‘up’, etc., in whose case the direction is exactly set. Verbal particles having directional meanings, too, are attached to verbs expressing motion as well as to verbs denoting a motion quality or containing the semantic moment of direction, and in these cases the directional meaning of verbal particles compositionally contributes to the meaning of the particle verbs. It is in this regard that Kiefer (2006: 48) mentions the compound-like function of verbal particles having additional directional meanings. At the same time, directional verbal particles also often denote the perfective aspect: here the function of expressing direction is accompanied by that of expressing aspect. Some verbal particles do not express directional meaning at all. In the case of the early-grammaticalized megPERF the directional meaning (< mög ‘the back of sth.’) of the verbal particle has almost completely disappeared in the course of grammaticalization. Nor do the recently-grammaticalized agyon ‘to death, to excess’ and tönkre ‘(becoming) totally ruined’ have any directional meaning.

5. Presence or absence of verbal particles in the sentence On the lexical-semantic level, the verbal particle usually perfectivizes the verb (excluding verbs of perception, e.g., át-lát through-see ‘to see through’, be-lát in-see ‘to see in’, and cases when the sequence verbal particle + verb denotes a spatial configuration, e.g., le-lóg down-hang ‘to hang down’, ki-hajlik out-lean ‘to lean out’, cf. Kiefer and Németh

37. Particle verbs in Hungarian 2012), but its presence or absence in the sentence is not directly connected to the aspectual value of the sentence, because it is influenced by other factors as well (see section 4.1). However, it is only in perfective sentences that productively formed particle verbs with a canonically placed (preverbal) verbal particle can occur. An exception could be provided by the so-called locative verbal particles mentioned in É. Kiss (2005, 2006), but the elements belonging to this group are not considered as verbal particles in other Hungarian sources, because − while their syntactic behavior corresponds to that of the verbal particles − they do not form a semantic unit with the verb, cannot be repeated unlike the verbal particles and play no role in aspect-forming, either − cf. Kiefer (2003: 185−186); see also section 2. É. Kiss (2005, 2006) does not regard verbal particles as aspectual operators, and ascribes their syntactic function chiefly to the fact that they contribute to the event structure of complex verbal predicates as secondary predicates referring a) to the state of the thing or person that experiences the change as a result of the verbal activity − resultative verbal particles, e.g., meg: (2)

A vaj meg-olvad-t. DET butter megPERF-melt-PAST ‘The butter melted.’

b) to the destination of the shifting thing’s or person’s movement − terminative verbal particles, e.g., be ‘into’: (3)

János be-gurít-ott-a a labdá-t a kapu-ba. John roll-PAST-3SG(DEF) DET ball-ACC DET gate-into ‘John rolled the ball into the net.’

or c) relates to the existence/spatial place of the thing or person in question − locative verbal particles, e.g., fent ‘up’: (4)

János fent alsz-ik az emelet-en. John up sleep-3SG DET storey-on ‘John sleeps upstairs.’

According to É. Kiss (2005: 82), the presence or absence of a verbal particle in a sentence is mostly predictable on the basis of whether the verbal particle functions as a secondary predicate: a verbal particle usually occurs in the sentence in case the event structure of the predicate is complex. The verbal particle can only be omitted if the canonical preverbal slot is occupied by a resultative or terminative phrase functioning, like the verbal particles do, as a secondary predicate, yet contrary to them, having a descriptive content (cf. É. Kiss 2005: 61): (5)

Mari zöld-re fest-ett-e [be] a kerítés-t. Mary green-on paint-PAST-3SG(DEF) [bePERF] DET fence-ACC ‘Mary painted the fence green.’

667

668

III. Units and processes in word-formation II: Special cases

6. Patterns of particle verbs Since the combination of verbal particles and verbs is not restricted by any kind of phonological or morphological condition (cf. Kiefer and Ladányi 2000: 455), it is worth talking about the patterns of Hungarian particle verbs from a semantic point of view. These patterns can only be defined as to certain meanings of verbal particles, because the connectivity of one and the same verbal particle in its various meanings may differ. The establishment of the patterns is closely connected, on the one hand, to the determination of the meanings of the verbal particles, and on the other hand, to the establishment of the semantic properties of verbal groups the verbal particle in a given meaning can be attached to. There are partial results in this field, but comprehensive works are still lacking. When establishing the patterns of particle verbs with the verbal particles be ‘in’, szét ‘apart, asunder’, agyon ‘to death, to excess’ and tönkre ‘(becoming) totally ruined’, Kiefer and Ladányi (2000: 483−518) − based on Ladányi (1999−2001) and Ladányi (2000), among others − aimed to determine both certain lexical and aktionsart meanings of the verbal particles examined, and the productive particle-verb patterns belonging to them. Principally, Kiefer and Ladányi (2000) set up patterns referring to groups of morphologically complex primary words, which are performed by productive derivational rules, assuming that the particle-verb patterns built on them have to be productive as well (Kiefer and Ladányi 2000: 483). In connection with the compatibility of the various aktionsart meanings of the verbal particles with the event types, i.e. the verbal groups defined on the basis of lexical meaning, several restrictions can be formulated. Hence, for example, “aktionsarts are constructed only with special meanings: resultative aktionsart with resultative verbs, semelfactive with lexically iterative verbs, inchoative with verbs of sound-giving, total aktionsart with resultative verbs denoting the covering of an area or surface, intensive with resultative verbs implying a change of state” (Kiefer 2003: 179). In a cognitive-functional framework, particle verbs as complex structures are derived from the semantics of verbal particles and verbs, based on Langacker’s (1987) reciprocal elaboration concept and Fauconnier and Turner’s (1996) blending concept (cf. Tolcsvai Nagy 2001, 2005). This is quite a promising direction, but there are as of yet only scant results on the basis of Hungarian material. A constructional framework could also be very useful for describing the patterns of particle verbs to overcome the contradiction between the particle verb being both a lexical unit and a syntactic structure at the same time (cf. Booij 2010: 130). However, the various verbal particle + verb constructional schemas and the constructional meanings belonging to them have not yet been elaborated for Hungarian. Szili, in several of her works (e.g., 2003, 2005), attempts to sketch the path of the verbal particles from directional meanings to abstract meanings and their occurences in the patterns of particle verbs, invoking some results of cognitive linguistics as well. She analyzes the material of the entries of the Explanatory dictionary of the Hungarian language (MNÉS 1959−62), while Ladányi’s work (2004) rests upon the examination of neologisms with the verbal particle be taken from the Hungarian National Corpus in an attempt to grasp how the new meanings of the verbal particle are built on the old ones, and how the new meanings lead to the verbal particle’s greater connectivity.

37. Particle verbs in Hungarian

7. The syntactic behavior of particle verbs Verbal particles behave syntactically like other verbal modifiers − cf., for example, Komlósy (1992: 494−505) and É. Kiss (2002: 55−71); her term is “particle-like verbal prefix”. In neutral (evenly stressed) sentences, the verbal particle is placed directly before the verb, bearing an accent (ˈ) and forming a phonological unit with the verb, e.g.: (6)

ˈEl-olvas-sa a ˈkönyv-et. elPERF-read-SG3(DEF) DET book-ACC ‘(S)he reads the book.’

In the same position, it can also fulfil the role of the focus (ˈˈ) of a non-neutral sentence, e.g.: (7)

A ˈfiú ˈtegnap ˈˈbe-men-t a ˈszobá-ba. DET boy yesterday in-go-PAST DET room-into ‘Yesterday the boy did go into the room.’

According to the behavior of the verbal modifiers, the verbal particle can be separated from the verb, if some word, for example the augmentative particle is ‘too’, is inserted between the verbal particle and the verb (e.g., ˈbe is ˈmegy in too go ‘(s)he does go in’), or if the verbal particle is placed after the verb. This is the situation when, compared to the neutral sentence, some stressed element, for example, an adverb or an object, appears in the preverbal (focus) position: (8)

A ˈˈkönyv-et olvas-sa el. DET letter-ACC read-SG3(DEF) elPERF ‘It is the book (s)he reads through.’

Or in the case of negation as well: (9)

ˈNem olvas-sa ˈel a ˈkönyv-et. no read-SG3(DEF) elPERF DET book-ACC ‘(S)he does not read the book’.

In continuous/progressive sentences/clauses, perfective particle verbs can only occur in syntactic structures showing the sequence verb + verbal particle and special accentual conditions (often completed with the time-modifying adverb éppen ‘just’ expressing reference time): (10) ˈPisti (éppen) ˈmász-ott ˈfel a ˈfá-ra (ˈ amikor Steve (just) climb-PAST up DET tree-onto ( when ˈmeg-lát-t-am). megPERF-see-PAST-1SG) ‘Steve was climbing the tree (when I noticed him).’

669

670

III. Units and processes in word-formation II: Special cases As Kiefer (2006: 118−127) points out, this characterizes only particle verbs with transparent meanings. This is the case with particle verbs formed by directional verbal particles. In a functional-cognitive framework, Imrényi (2011: 97−102) looks upon the verbal particle + verb structure as a protostatement forming the kernel of the sentence. By default interpretation, the kernel sentence built on the protostatement is declarative, affirmative and neutral. As compared with the basic word order (verbal particle + verb), Imrényi interprets the word order change (verb + verbal particle) as a restriction of the protostatement’s default interpretation, i.e. as a change of the pragmatic type of the sentence (Imrényi 2009: 364−370, 2011: 84−90): (11) default interpretation, declarative, affirmative and neutral sentence Zsuzsi tegnap fel-hív-t-a Mari-t. Sue yesterday up-call-PAST-3SG(DEF) Mary-ACC ‘Sue called Mary yesterday.’ (12) interrogative sentence (instead of declarative) Ki-t hív-ott fel tegnap Zsuzsi? who-ACC call-PAST up yesterday Sue ‘Whom did Sue call yesterday?’ (13) negative sentence (instead of affirmative) Zsuzsi nem hív-ja fel Mari-t Sue not call-3SG(DEF) Mary-ACC ‘Sue doesn’t call Mary.’ (14) non-neutral sentence ’’Mari-t hív-ja fel Zsuzsi. Mary-ACC call-3SGF(DEF) up Sue ‘It’s Mary Sue is calling.’

8. Outlook for the future According to Komlósy (1992: 498), “the circle of problems of the particle verbs and similar verbal structures has until now been one of the greatest puzzles of Hungarian grammar”. During the last twenty years, this picture has become clearer in several respects: there have been partial results achieved in numerous fields, including the most problematic semantic questions as well. However, not all problems connected with verbal particles and particle verbs have been solved so far. Further empirical research will probably clarify new details, yet there will remain disputed questions regarding the stock of verbal particles and the function and connectability of the verbal particles, as well as regarding the explanation of the meaning and syntactic behavior of the particle verbs − partly because of the many facets of the verbal particles, partly because the research is often carried out in different theoretical frameworks, and the different approaches lead to different results.

37. Particle verbs in Hungarian

9. References Beöthy, Erzsébet and Gabriel Altmann 1985 The diversification of meaning of Hungarian verbal prefixes I. meg. Nyelvtudományi Közlemények 87: 187−196. Booij, Geert 2010 Construction Morphology. Oxford: Oxford University Press. Bybee, Joan L., Revere Perkins and William Pagliuca 1994 The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: The University of Chicago Press. D. Mátai, Mária 2007 A magyar szófajtörténet általános kérdései. Budapest: Akadémiai Kiadó. É. Kiss, Katalin 2002 The Syntax of Hungarian. Cambridge: Cambridge University Press. É. Kiss, Katalin 2005 First steps toward a theory of the verbal particle. In: Christopher Pinon and Péter Siptár (eds.), Approaches to Hungarian. Vol. 9: Papers from the Düsseldorf Conference, 59− 88. Budapest: Akadémia Kiadó. É. Kiss, Katalin 2006 The function and the syntax of the verbal particle. In: Katalin É. Kiss (ed.), Event Structure and the Left Periphery. Studies on Hungarian, 17−55. Dordrecht: Springer. Fauconnier, Gilles and Mark Turner 1996 Blending as central process of grammar. In: Adele E. Goldberg (ed.), Conceptual Structure, Discourse and Language, 113−129. Cambridge: Cambridge University Press. Forgács, Tamás 2004 Grammaticalisation and preverbs. Acta Linguistica Hungarica 51: 45−84. Gyuris, Beáta and Ferenc Kiefer 2008 Az igék lexikai ábrázolása és az eseményszerkezet. In: Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 4. A szótár szerkezete, 229−267. Budapest: Akadémiai Kiadó. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Grammaticalization. 2nd ed. Cambridge: Cambridge University Press. Imrényi, András 2009 Toward a unified functional account of structural focus and negation in Hungarian. Acta Linguistica Hungarica 56: 342−374. Imrényi, András 2011 A magyar mondat viszonyítási modellje. Ph.D. dissertation, ELTE University, Budapest. Isačenko, Alexander V. 1962 Die russische Sprache der Gegenwart. Part 1: Formenlehre. Halle/S.: Niemeyer. Jakab, István 1982 A magyar igekötő szófajtani útja. Budapest: Akadémiai Kiadó. J. Soltész, Katalin 1959 Az ősi magyar igekötők (meg, el, ki, be, fel, le). Budapest: Akadémiai Kiadó. Kiefer, Ferenc 2003 A kétféle igemódosítóról. Nyelvtudományi Közlemények 100: 177−186. Kiefer, Ferenc 2006 Aspektus és akcióminőség különös tekintettel a magyar nyelvre. Budapest: Akadémiai Kiadó. Kiefer, Ferenc 2010 Areal-typological aspects of word-formation: The case of aktionsart-formation in German, Hungarian, Slavic, Baltic, Romani and Yiddish. In: Franz Rainer, Wolfgang U. Dressler, Dieter Kastovsky and Hans Christian Luschützky (eds.), Variation and Change in Morphology. Selected papers from the 13th International Morphology Meeting, Vienna, February 2008, 130−147. Amsterdam/Philadelphia: Benjamins.

671

672

III. Units and processes in word-formation II: Special cases Kiefer, Ferenc and László Honti 2003 Verbal ‘prefixation’ in the Uralic languages. Acta Linguistica Hungarica 50: 137−153. Kiefer, Ferenc and Mária Ladányi 2000 Az igekötők. In: Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 3. Morfológia, 453− 518. Budapest: Akadémiai Kiadó. Kiefer, Ferenc and Boglárka Németh 2012 When the preverb does not perfectivize. In: Morphology and Meaning. 15th International Morphology Meeting, Book of Abstracts, 73. Vienna, February 9−12, 2012. Komlósy, András 1992 Régensek és vonzatok. In: Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 1. Mondattan, 299−527. Budapest: Akadémiai Kiadó. Ladányi, Mária 2000 Productivity as a sign of category change: The case of Hungarian verbal prefixes. In: Wolfgang U. Dressler, Oskar E. Pfeiffer, Markus Pöchtrager and John R. Rennison (eds.), Morphological Analysis in Comparison, 113−141. Amsterdam/Philadelphia: Benjamins. Ladányi, Mária 1999−2001 Synonymy in Hungarian verbal prefixes. In: István Szathmári (ed.), Annales Universitatis Scientarium Budapestinensis de Rolando Eötvös nominatae. Sectio Linguistica 24, 67−84. Budapest: Eötvös Loránd Tudományegyetem. Ladányi, Mária 2004 Rendszer − norma − nyelvhasználat: Igekötős neologizmusok „helyi értéke”. In: László Büky (ed.), Nyelvleírás és nyelvművelés, nyelvhasználat, stilisztika, 97−118. Szeged: Magyar Nyelvészeti Tanszék, Általános Nyelvészeti Tanszék. Langacker, Ronald 1987 Foundations of Cognitive Grammar. Vol. 1: Theoretical Prerequisites. Stanford: Stanford University Press. MNÉS 1959−62 A magyar nyelv értelmező szótára. Vol. 1−7. Budapest: Akadémiai Kiadó. Magyar Nemzeti Szövegtár. http://corpus.nytud.hu/mnsz. Nádasdy, Ádám 2003 Besír, beröhög. In: Ízlések és szabályok, 285−288. Budapest: Magvető. Pais, Dezső 1959 Az igekötők mivoltához és keletkezéséhez. Magyar Nyelv 55: 181−184. Pátrovics, Péter 2003 Megjegyzések igekötőink használatának újabb tendenciáiról. In: Mihály Hajdú and Borbála Keszler (eds.), Köszöntő könyv Kiss Jenő 60. születésnapjára, 202−206. Budapest: ELTE Magyar Nyelvtudományi és Finnugor Intézet, Magyar Nyelvtudományi Társaság. Péter, Mihály 2008 A magyar aspektusról − más aspektusból. Magyar Nyelv 104: 1−11. Szili, Katalin 2003 A ki igekötő jelentésváltozásai. Magyar Nyelv 99: 163−188. Szili, Katalin 2005 A be igekötő jelentésváltozásai I−II. Magyar Nyelvőr 129: 151−164, 282−299. Tolcsvai Nagy, Gábor 2001 Conceptual metaphors and blends of “understanding” and “knowledge” in Hungarian. Acta Linguistica Hungarica 48: 79−100. Tolcsvai Nagy, Gábor 2005 Kognitív jelentéstani vázlat az igekötőről. Magyar Nyelv 101: 27−43. Wacha, Balázs 2001 Időbeliség és aspektualitás a magyarban. Budapest: Akadémiai Kiadó.

Mária Ladányi, Budapest (Hungary)

38. Noun-noun compounds in French

38. Noun-noun compounds in French 1. 2. 3. 4. 5. 6.

Introduction Definitions and restrictions Classification Grammatical characteristics The status of French [NN]N compounds References

Abstract Although the [NN]N pattern is not predominant for the naming of combinatory concepts, the corresponding class of compounds is expanding in contemporary French. The article presents a semantically-based taxonomy of [NN]N units, followed by a survey of their grammatical characteristics. The last section examines theoretical positions with respect to their grammatical status.

1. Introduction Languages differ in the position they accord to compounding among their lexical extension processes. In a comparison of 13 languages with respect to lexical items corresponding to 30 culture-independent concepts like ‘eyelash’ and ‘rainbow’, German had a “score” of 18 out of 30, English 12 and French, with 4, was close to the lowest score of the language sample (Arnaud 2004). A random sample of 100 English [NN]N compounds corresponds to only 8 similar (but left-headed) units in French. These differences are due to the fact that [NN]N is in competition with other patterns for the naming of combinatory concepts in French. The most important of these alternative patterns are listed below (compound glosses are literal, so in most cases the order of components is the opposite of the English one; English equivalents or explanations are given in brackets when useful): (1)

[N prep N]N units, mainly with the vague prepositions de ‘of, from’ and à ‘to, at’ frein à main ‘brake to hand (handbrake)’

(2)

N + relational adjective forêt pluviale ‘forest pluvial (rainforest)’

(3)

lexicalized N + prepositional phrase sequence marché aux puces ‘market to-the fleas (flea market)’

Obviously, French does not produce [NN]N lexical units as freely as the Germanic languages, but the pattern is available in its grammar. After a section on preliminary notions

673

674

III. Units and processes in word-formation II: Special cases and precautions, the article presents a classification of [NN]N units based on semantic criteria, followed by a survey of the characteristics and behaviour of the different categories. The last section will be devoted to a review of theoretical positions on their grammatical status. Until then the term compound will be used in a theory-neutral way.

2. Definitions and restrictions An important consideration in the analysis of NN sequences is that nominal lexical units are the names of categories of entities. Also important is the fact that the word-formation resources of a language produce items that can be lexicalized and become units of the lexicon or of a particular terminology. Word-formation processes also produce items that refer to categories that are nameworthy (Downing 1977) only in a particular communicative situation. For this reason, hapaxes or types with few occurrences will be used as examples of productive categories along with established compounds. French [NN]N units are formally similar to syntactic NN sequences. Some of these constructions, with non-heads such as (4)

charnière, choc, ‘hinge, shock,

clé, key,

modèle, model,

phare, lighthouse,

record, record,

surprise surprise’

are semantically close to subordinative compounds (see section 3), but they can be distinguished from them. French nouns may undergo different degrees of conversion to adjectives and conversion entails a semantic reduction (marron ‘chestnut’ as an adjective only denotes a colour). As a result, modification by a converted N2 does not create a subclass of N1 but only qualifies it: année clé ‘year key’ does not denote a type of year as année bissextile ‘year bisectile (leap year)’ does. Conversion also goes together with the acquisition of adjectival syntax and another reason to reject an NN sequence like annéeclé as a compound is that clé is found in typically adjectival positions as in the following example: (5)

Nous sommes dans une période assez clé de l’année. ‘We are in a rather key period of the year.’

Another precaution is necessary when collecting French NN sequences. “Telegraphic style” can result in NN sequences that are not necessarily lexical items. This is the case when N+prep+(art)+N sequences are reduced, as in the following text seen on a roadside sign: (6)

parking cars 200m

This is a telegraphic version of (7)

parking des cars à 200 m ‘park of-the coaches in 200 m (coach park in 200m)’

38. Noun-noun compounds in French For this reason, NN sequences from signs, catalogues, etc. are not safe to use in a discussion of lexical units and, therefore, only examples found in normal syntactic environments or in dictionaries will be quoted here.

3. Classification 3.1. Headedness In a discussion of [NN]N units it is necessary to distinguish different types of heads. Therefore, I will differentiate between semantic headedness and syntactic headedness, as there are a few cases where the corresponding heads do not coincide, as will be seen in section 3.4. An endocentric compound has (only) one semantic head, which is the name of the generic class of its denotatum: (8)

Un oiseau-mouche est un oiseau. ‘A bird fly (hummingbird) is a bird.’

Other compounds may have two or more semantically equal constituents, in which case pluricentric is a more adequate term than the usual exocentric. In other cases yet, due to a metaphoric or metonymic construal of an endocentric compound, the denotation shifts and the name of the denotatum’s category is no longer present in the compound, as exemplified by (9): (9)

poids plume ‘weight feather (a person with a light weight)’

As plume initially modifies poids the term exocentric is somewhat misleading (see Bauer 2008) and secondary exocentric is preferable for units of this type which, due to their secondary nature, need not figure in the classification given here (Benczes 2006). Syntax-inspired terms with wide currency are subordinative compounds for endocentrics, and coordinatives for pluricentrics. These will be used below. The top level of the classification is binary. The reason for this is that semantic headedness is an essential parameter and if we go beyond the case of French, there are significant phonological and grammatical differences between subordinative and coordinative compounds in a number of languages, like stress position and sandhi phenomena (see, for instance, Kim 2001).

3.2. Subordinative units The classification below is supported by a number of semantic tests inspired by Cruse (1986) and Riegel (1988), and it is justified by the difference in pluralization patterns which will be examined in section 4. The first tests T1 and T2 are tests of semantic headedness: units that test positive on T1 and negative on T2 are endocentric (subordinative):

675

676

III. Units and processes in word-formation II: Special cases (10) (T1) Un nœud-papillon est un nœud. ‘A knot butterfly (bow tie) is a knot.’ (T2) *Un nœud-papillon est un papillon. ‘A knot butterfly is a butterfly.’ Units that satisfy neither test are secondary exocentric compounds, and units that satisfy both tests are coordinative − but see below for apparently ambiguous cases.

3.2.1. Attributive units Test T3 tests the attribution of features from the nonhead’s semantic representation to the head: (11) (T3) Une tente-igloo est une tente qui est (comme) un igloo. ‘A tent igloo is a tent that is (like) an igloo.’ The analogy manifested in comme rests on a subset of the semantic features in the representation of the denotatum of N2, namely the function and the hemispherical shape in the above example, while other features are suppressed. Analogy, however, is less perceptible in cases like voiture-bélier ‘car-ram (a car used by thieves to shatter a shop window)’ where N2 is independently metaphorized in the lexicon, and it is absent from a unit like date-limite ‘date limit (deadline)’: (12) (T3) Une date-limite est une date qui est (0̸) une limite. ‘A date limit is a date which is (0̸) a limit.’ In such a case, both T1 and T2 give positive results, so we are apparently close to the category of coordinative compounds. It is possible, however, using various tests which cannot be detailed here, to ascertain that a unit like date-limite is semantically leftheaded in the same way as tente-igloo is. The attributive category is productive, and hapax occurrences can be found in the media, like: (13) des moutons-tondeuses embauchés au fort de Dardilly. ‘sheep lawnmowers (lawnmower sheep) hired at the fort of Dardilly’ (with analogical attribution) A particular subcategory of attribution is to be found in a limited set of units like thonalbacore ‘tuna albacore’, mésange nonette ‘tit little-nun (marsh tit)’. N2 is the name of a species while N1 is that of a genus, and N1N2 has the same denotation as N2: (14) (T4) Le thon albacore et l’albacore, c’est la même chose. ‘Tuna albacore and albacore, it is the same thing.’

38. Noun-noun compounds in French

677

3.2.2. Relational units Other endocentric compounds test negative on T3: (15) (T3) *Une assurance vie est (comme) une vie. ‘An insurance life is (like) a life.’ Semantically, the modification relation belongs to diverse types. Some cases correspond to different implicit predications which link the constituents of the compound. The following is a very limited sample: (16) N1 consists of N2 N1 includes N2 N1 originates in N2 N1 belongs to N2

code barres toile émeri point jury flotte client

‘code bars (barcode)’ ‘cloth emery’ ‘point jury (an extra point awarded by an examination jury)’ ‘fleet customer (customer’s vehicle fleet)’

The relation may also be of the verb-complement type if N1 is deverbal. Examples in (17) are followed by a logical notation and a translation: (17) descente dames placement produit

descend (ladies) place (x, product)

(women’s downhill race) (product placement)

In some cases, the modification is semantically ambiguous and different relations coexist, as in local barre ‘room helm (steering gear compartment)’, N1 place of N2 and N1 for N2. As in this example, the telic relation present in compounds naming artefacts often coexists with some other relation. In other cases yet, the compound is underpinned by a complex scenario: (18) banane dollar (a banana produced in a non-euro zone) amphi garnison ‘lecture-theatre garrison (a meeting during which officer school graduates choose their first post)’ Listing and classifying these relations with sufficient granularity is a daunting task, which has rarely been undertaken (Arnaud 2003: 61−87). A comparison with English compounds shows that some relations may be absent from French. This is apparently the case of the “habitat” relation present in sea bream or marshmallow, but it may be an effect of the smaller number of French units. Analogy of construction results in the emergence of series of compounds with identical heads or non-heads: (19) compte client, fiche client, relation client, suivi client, flotte client, etc. ‘account customer, card customer (customer knowledge sheet), relation customer, follow-up customer, fleet customer’

678

III. Units and processes in word-formation II: Special cases

3.3. Coordinative compounds In coordinative compounds N1 and N2 are co-hyponyms. As noted above, these units test positive on both T1 and T2. Tests of anaphora show that they do not include a single semantic head, and this is confirmed by the fact that, contrary to subordinatives, sentences with the constituents in reverse order are acceptable: (20) Un enseignant-chercheur pourrait être appelé chercheur-enseignant. ‘A teacher researcher (an academic) could be called researcher-teacher.’ The two versions may be actually attested, like propriétaire-dirigeant ‘owner-manager’ and dirigeant-propriétaire, but the order is generally lexicalized and results from a combination of factors of length and pragmatic salience (Renner 2008). Three semantic subclasses can be distinguished, as was suggested by Hatcher (1951: 32) for English: (21) Multifunctional coordinatives denote entities that hold several functions or purposes boucher-charcutier ‘butcher pork-butcher’ chasseur-bombardier ‘fighter bomber’ (22) Hybrid coordinatives denote entities that are perceived to be a mix of other entities baryton-basse ‘barytone bass’ tram-train (a tramcar that can be used on a railway line) (23) Additive coordinatives (dvandvas), which are rare, denote entities that are the sum of separate entities bains-douches ‘baths showers (public baths)’ batterie-fanfare ‘drum-corps fanfare (a band)’ Individual examples may be categorially ambiguous like pli-faille ‘fold fault’, which could be a hybrid or a dvandva.

3.4. Intermediate units Units like femme médecin ‘woman doctor’, expert-comptable ‘expert accountant’, candidat médicament ‘candidate medicine (a new medicine undergoing tests)’, which generally include one constituent denoting a status relative to the other constituent’s denotatum, are not easy to fit into a taxonomy. Tests of semantic headedness and anaphora do not

38. Noun-noun compounds in French return identical results on different units and show a notable amount of disagreement among informants. It seems wiser to conclude that these units occupy various positions along a scale between attributives and coordinatives. In some cases, however, like that of bébé (as in bébé-phoque ‘baby seal’), the N1 has acquired some of the characteristics of a prefix (Amiot and van Goethem 2010).

3.5. Exceptions “In many languages there are exceptions or fuzzy transitions to non-compounding” (Dressler 2006: 24). The most numerous exceptions are semantically and syntactically right-headed units like auto-école ‘auto school (driving school)’ or radio-balise ‘radio beacon’, where N1 results from the reanalysis of a neo-classical component and may be considered an “affixoid” (Booij 2009). In bidonvillemasc ‘can town (shanty town)’, the syntactic head is bidonmasc and the semantic head villefem. Other exceptions, like the graphically fused secondary exocentrics chiendent ‘dog tooth (couch grass)’ and chèvrefeuille ‘goat leaf (honeysuckle)’, are opacified remnants of earlier patterns. The next section is devoted to non-exceptional units only.

4. Grammatical characteristics This section presents a survey of the grammatical characteristics of French [NN]N units. Two characteristics of French complicate the analysis of compounding: French does not have lexical stress, and noun plurals are phonetically identical to the singulars, except for a few “irregular” nouns rarely present in compounds anyway, so it is often necessary to rely on spellings and then caution is necessary with Web occurrences. When possible, examples with audible plurals have been selected (these are in bold type). a) Syntactic headedness French [NN]N units are syntactically left-headed with N1 communicating its gender to the compound, as in the following examples from different categories described in section 3: (24) [chapeaumasc-clochefem]masc ‘hat bell (cloche hat)’ [stylomasc billefem]masc ‘pen ball (ballpoint pen)’ [femmefem-commissairemasc]fem ‘woman police-commissioner’ [pointmasc-virgulefem]masc ‘point comma (semicolon)’

679

680

III. Units and processes in word-formation II: Special cases b) Plural forms Two pluralization patterns exist, but the head (N1) is always in the plural, so French units have internal plurals. Note however that, due to the non-audible nature of most plurals, there is considerable variation in the spelling of occurrences. (25) subordinatives: relationals: plural on N1 chevaux vapeur ‘horses steam (horsepower)’ œufs cocotte ‘eggs pot (a recipe)’ Occurrences of some of these units are occasionally found with double plurals, which may be a sign of regularization toward a unique rule of double plurals. (26) attributives: plural on both N1 and N2 animaux machines ‘animals machines (a term of Descartes’s mechanist philosophy)’ tuiles canaux ‘tiles canals (Roman tiles)’ (27) coordinatives: plural on all the constituents couvents-hôpitaux ‘convents hospitals’ auteurs-compositeurs(-interprètes) ‘songwriters composers (singers)’ c) Nonhead pluralization The nonhead (N2) of a relational unit is generally in the plural if it denotes multiple entities. (28) train travaux ‘train works’ marché actions ‘market shares’ d) Internal modification Modification of the head is impossible. (29) *des pneus chers neige ‘tyres expensive snow’ *un poisson noir chat ‘a fish black cat’ This, however, is not in itself a criterion for compounding but results from lexicalization and the categorizing effect of the modification by N2, as lexicalized phrases have the same characteristic.

38. Noun-noun compounds in French

681

e) Phrasal inclusions Syntactic sequences may be included as modifiers, as in the following examples: (30) laitue feuille de chêne ‘lettuce leaf of oak (oakleaf lettuce)’ kit mains libres ‘kit hands free (hands-free phone kit)’ responsable news, tendances et tablettes ‘person-in-charge news, trends and e-pads’ f) Derivation Derivation from compounds is a phenomenon with limited productivity, consequently very few examples based on [NN]N units are currently available. Examples where the suffix attaches to N1 like chéquier services ‘cheque-book services (a book of vouchers for paying service workers)’, épargnant retraite ‘saver retirement (a person saving up toward his/her pension)’, assureur vie ‘insurer life (life insurance company)’ are equivocal as compounding may have occurred after the derivation of N1 and so there is no firm evidence of compound-internal derivation. That external derivation is possible appears in recent isolated units with implicit noun-to-verb conversion of N2 like capital-risqueur ‘capital risk-er (venture capital lender)’, mot-valisage ‘word portmanteau-age (lexical blending)’. g) Recursivity That coordinative units do not have a unique semantic head is also apparent from the fact that longer coordinatives are formed by stringing more than two nouns together in a semantically non-hierarchical fashion: (31) moissonneuse-batteuse → ‘harvester thresher (combine harvester)’

moissonneuse-batteuse-lieuse ‘harvester thresher binder’

The following example is a hapax: (32) il y a aussi la technique de Laurent le prêtre-chanteur-écrivain ‘there is also the technique of Laurent the priest singer writer’ Triple subordinatives, which are rare in the first place, may have structures that are ambiguous between [[NN]N] and [N[NN]], like officier sécurité incendie ‘officer safety fire (fire safety officer)’, plan épargne logement ‘plan savings housing (housing savings scheme)’, as officier sécurité and sécurité incendie, plan épargne and épargne logement are separately lexicalized and metalinguistic reflection does not produce a preferred interpretation. An example like contrat initiative emploi ‘contract initiative employment (employment initiative contract)’ appears to be structurally opaque and is not very different from a telegraphic style sequence. In contrast, very few English triple subordinatives cause hesitation about their structure. However, although the pattern’s productivity is low, there exist a few examples of clearly left-branching [[NN]NN]N units, such as poisson-feuille cacatoès ‘fish leaf cockatoo (cockatoo leaf-fish)’, oiseau-mouche abeille ‘bird fly bee (bee hummingbird)’, which are attributive, and crédit impôt recherche ‘credit tax

682

III. Units and processes in word-formation II: Special cases research (R&D tax credit)’, a relational compound. As for chèque emploi service ‘cheque employment service (a voucher for paying service workers)’, it is right-branching. h) Origin and expansion Relational [NN]N units were rare before 1850, and the first author known to have mentioned them, Darmesteter (1875), provided a very limited list of examples. Although underrepresented in dictionaries, the category is undergoing expansion in Contemporary French. Jenkins (1972) noted the “sharp increase” and Noailly (1990: 6) the “proliferation” of NN constructions, and Picone (1996: 175) considers this growth to be one of the most important recent trends of the language. Some units are apparently calques of English compounds: (33) station service ‘service station’ analyste marché ‘analyst market’ empreinte carbone ‘footprint carbon’ Even if [NN]N compounds were rare before 1850, the existence of units like papier formule ‘paper formula (sheets with formulas)’, laurier cerise ‘laurel cherry’ (Furetière 1690) at a time when English had no influence on French proves that the pattern did not originate in another language. As Clas (1987) has remarked, recent relational units that are obvious calques are as acceptable as attributive units, which shows that the corresponding rule must be indigenous to the grammar of French. i) Prepositional variants From an onomasiological perspective, naming a combinatory concept attributively can only be done with the [NN]N pattern. In contrast, when the relation is of another type, the [NN]N pattern is, as noted in section 1, in stiff competition with others and [NN]N units are outnumbered. Another interesting fact is that some [NN]N units also exist in a prepositional version: (34) stylo-bille / stylo à bille ‘pen ball / pen to ball (ballpoint pen)’ groupe contrôle / groupe de contrôle ‘group control / group of control (control group)’ In some cases, however, the prepositional variant is attested in negligible numbers in comparison with the [NN]N form, and, more conclusively, many units are without prepositional equivalents, like: (35) portrait robot (identikit portrait) dose carrière ‘dose career (the dose of radioactivity received during one’s career)’

38. Noun-noun compounds in French

5. The status of French [NN]N compounds Not all definitions of compounding used in French linguistics are restricted to combinations of open-class units. For example, Guilbert (1971: LXIX), Mathieu-Colas (1996), Gross (1996: 33, 49, 53), Apothéloz (2002: 18−19), Di Sciullo (2005), Riegel, Pellat and Rioul (2009: 912) consider N+prep+N units as compounds, a position criticized by Fradin (2009) on the grounds that it confuses compounding with idiomaticity. Another problem is that definitions that include binary structure as a criterion, like Benveniste’s (1966), should lead to the exclusion of coordinative units if we take them to refer to semantic interpretation, since, as we saw in section 4, these can be expanded to longer sequences while keeping a “flat” interpretation. The pioneer of French compounding studies, Darmesteter, suggested that [NN]N units represented a “clause in short form” (Darmesteter 1875: 4). The idea that compounds are syntactic was taken up by, for instance, Benveniste (1967), without formalization and outside the framework of generative grammar where the formation of compounds by transformations of underlying sentences had been introduced earlier by Lees (1960). The generative approach had an influence on French linguistics and various authors presented transformational accounts of compounding (Barbaud 1971; Guilbert 1971; Wandruszka 1972; Giurescu 1975; Rohrer 1977; Lamy 1978), some of which only mentioned presumed underlying sentences in non-formalized accounts. This approach was criticized on account of the implausibility of some of the proposed underlying sentences (Lifetree-Majumdar 1974; Noailly 1990: 34 n. 24), or on the grounds that compounds do not include clausal relations that can be asserted or modalized like those in a sentence (Riegel 1988). Because transformations were later abandoned by generative grammar and theoretical works suggesting a radical separation of morphology and syntax appeared in the 1980s (Selkirk 1982; Di Sciullo and Williams 1987), the place of compounding in the grammar became the object of attention. The characteristics of Romance compounds set them apart from Germanic ones. In particular, the left-headedness of French units led Di Sciullo and Williams (1987: 81, 83), applying William’s (1981) right-hand head rule, to classify them as listed “phrasal idioms”, not as compounds. Explicitly following Di Sciullo and Williams, Zwanenburg (1992) wrote that “in French, real, righthand headed compounding is very restricted”. Such approaches, which were fundamentally inspired by the structure of English, were later criticized: Ten Hacken (1999), for instance, concludes that the right-hand head rule should not be part of a cross-linguistic definition of compounds, and Guevara and Scalise (2009) criticize the restriction of the class of compounds to right-headed units as a “theory-internal construct”. Outside the generative community, Noailly (1990) distinguished four functions of syntactic NN constructions with attributive nouns, three of which, if lexicalized, correspond to compounds: qualification corresponds to attributive compounds, complementation to relationals and coordination, obviously, to coordinatives. A recent example of the syntactic view is the model presented by Barbaud (2009). While derivation belongs to morphology, compounding is a process that expands the lexicon by combining lexemes, in which speakers interpret syntactically generated sequences on the basis of their identity with a polylexemic “dicteme” memorized in the dictionary of the language (p. 31). The definition of the dicteme associates a phonetic signifiant, a symbolic signifiant (with an ontologic domain and formal features), and a prototypical signifié comprising a set of semes (p. 15).

683

684

III. Units and processes in word-formation II: Special cases The morphologist viewpoint is represented by Corbin (1992), who notes that timbreposte ‘stamp post (postage stamp)’ exhibits a “syntactic rupture”, i.e. the lack of a determinant of its N2, which differentiates it from a lexicalized phrase. In Corbin (1997), she also criticizes Zwanenburg (1992), who compared timbre-poste and homme-grenouille ‘man frog (frogman)’ with the phrases le projet Delors ‘the project Delors’ and un avocat ami ‘a lawyer friend’ and concluded that the former also had syntactic structure but, in having been used for the naming of entities, they have become lexicalized. Corbin notes that poste and grenouille have a categorizing effect resulting in N1 being the hyperonym of N1N2, while modifiers like Delors and ami respectively have an identifying and a qualifying effect on N1. The fact that the categorizing effect is not found in the syntactic constructions leads Corbin to the conclusion that [NN]N units like timbre-poste and homme-grenouille result from different rules, i.e. word-formation rules. Fradin in turn applies the principle that “compounds may not be built by syntax” (Fradin 2009). Like Corbin, he considers attributives like poisson-lune ‘fish moon’, requin-marteau ‘shark hammer (hammerhead shark)’ as morphologically formed, but unlike her he claims that examples like langage auteur ‘language author’, impôt sécheresse ‘tax drought (a farm relief tax)’ are formed syntactically (Fradin 2003: 195). He later (Fradin 2009) reiterates this distinction with more explicit arguments: according to him, the perceptible properties of the compound’s denotatum do not cogently produce a subcategory of the head, “N2 never introduces a semantic predicate that N1 would be an argument of”, and these units exist with prepositional variants. The characteristics described above in section 4 weaken these arguments: for instance, as we have seen, many relationals are without a safely attested prepositional variant. It has become progressively apparent that compounds are problematic in models of grammar with segregated morphology and syntax. In particular, English “phrasal” compounds show that morphology may take in syntactic constructions (Lieber 1992: 14). Montermini (2008), examining Italian units like raccolta rifiuti ingombranti ‘collection refuseplur cumbersomeplur (cumbersome refuse collection)’, sees in them objects that include syntactic sequences, but are in other respects morphological. This kind of consideration has led to the construction of models where the separation is less strict, for instance in the framework proposed by Di Sciullo (2005) where morphology and syntax can exchange word sequences: English compounds are derived in the morphological space, but French compounds are derived in the syntactic space before being transferred to the morphology. All categories of compounds are derived by the application of identical rules. As the derivation of French compounds is syntactic, it involves a functional projection realized as an operator that provides the semantic relation between the two nouns. In the case of coordinatives, this operator is AND/OR and in the case of relationals, it is the semantically vague SORT, which may surface in prepositional units (épargne de précaution ‘savings of precaution’) or not be realized phonologically (épargne logement ‘savings housing’) [my examples]. In construction grammar, morphology and syntax are on a continuum, which solves the disputes over their delineation. Research on compounds in cognitive linguistics has been focused mainly on semantics (Heyvaert 2009), however, and to the best of my knowledge, no construction grammar study of French [NN]N units has been published at the time of writing, but we can refer to an article by Masini (2009), who examines Italian N+prep+N and V+and+V constructions. In this study [NN]N units like effetto serra ‘effect greenhouse’ figure among phrasal nouns, an intermediate category, together

38. Noun-noun compounds in French with the above constructions and [NAdj]N, and this would presumably apply to French [NN]N units. From the characteristics presented in section 4, it is obvious that French [NN]N units are less prototypical compounds than the corresponding Germanic units, and they have been the object of debate. Although descriptive knowledge about them has accumulated since 1875, no definitive answer as to their status has been provided so far because opinions have been as varied as the theoretical frameworks they were formed in.

6. References Amiot, Dany and Kristel van Goethem 2010 Le statut de -clé et de sleutel- dans mot-clé / sleutelwoord: Une analyse unifiée? In: Franck Neveu, Valelia Muni Toke, Jacques Durand, Thomas Klingler, Lorenza Mondada and Sophie Prévost (eds.), Congrès mondial de linguistique française CMLF 2010, 847− 859. Paris: Institut de Linguistique Française. http://www.linguistiquefrancaise.org [last access 18 Apr 2011]. Apothéloz, Denis 2002 La Construction du lexique français. Principes de morphologie dérivationnelle. Gap: Ophrys. Arnaud, Pierre J. L. 2003 Les Composés timbre-poste. Lyon: Presses Universitaires de Lyon. Arnaud, Pierre J. L. 2004 Problématique du nom composé. In: Pierre J. L. Arnaud (ed.), Le Nom composé. Données sur seize langues, 329−353. Lyon: Presses Universitaires de Lyon. Barbaud, Philippe 1971 L’ambiguïté structurale du composé binominal. Cahiers de Linguistique [Montréal] 1: 71−116. Barbaud, Philippe 2009 Syntaxe référentielle de la composition lexicale. Un profil de l’homme grammatical. Paris: L’Harmattan. Bauer, Laurie 2008 Exocentric compounds. Morphology 18: 51−74. Benczes, Réka 2006 Creative Compounding in English. The Semantics of Metaphorical and Metonymical Noun-Noun Combinations. Amsterdam/Philadelphia: Benjamins. Benveniste, Émile 1966 Formes nouvelles de la composition nominale. Bulletin de la Société de Linguistique de Paris 61: 82−95. Benveniste, Émile 1967 Fondements syntaxiques de la composition nominale. Bulletin de la Société de Linguistique de Paris 62: 15−31. Booij, Geert 2009 Compounding and construction morphology. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 201−216. Oxford: Oxford University Press. Clas, André 1987 Sur les binominaux juxtaposés. Lebende Sprachen 87(3): 120−121. Corbin, Danielle 1992 Hypothèses sur les frontières de la composition nominale. Cahiers de Grammaire 17: 25−55.

685

686

III. Units and processes in word-formation II: Special cases Corbin, Danielle 1997 Locutions, composés, unités polylexématiques: Lexicalisation et mode de construction. In: Michel Martins-Baltard (ed.), La Locution entre langue et usages, 53−101. FontenaySt Cloud: ENS Editions. Cruse, D. Alan 1986 Lexical Semantics. Cambridge: Cambridge University Press. Darmesteter, Arsène 1875 Traité de la formation des noms composés dans la langue française comparée aux autres langues romanes et au latin. Paris: Franck. Di Sciullo, Anna Maria 2005 Decomposing compounds. SKASE Journal of Theoretical Linguistics 2: 14−33. Di Sciullo, Anna Maria and Edwin Williams 1987 On the Definition of Word. Cambridge, MA: MIT Press. Downing, Pamela 1977 On the creation and use of English compound nouns. Language 53: 810−842. Dressler, Wolfgang U. 2006 Compound types. In: Gary Libben and Gonia Jarema (eds.), The Representation and Processing of Compound Words, 23−44. Oxford: Oxford University Press. Fradin, Bernard 2003 Nouvelles approches en morphologie. Paris: Presses Universitaires de France. Fradin, Bernard 2009 IE, Romance: French. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 417−435. Oxford: Oxford University Press. Furetière, Antoine 1690 Dictionnaire universel contenant généralement tous les mots françois tant vieux que modernes. The Hague: Leers. Giurescu, Anca 1975 Les Mots composés dans les langues romanes. The Hague: Mouton. Gross, Gaston 1996 Les Expressions figées en français. Noms composés et autres locutions. Gap: Ophrys. Guevara, Emiliano and Sergio Scalise 2009 Searching for universals in compounding. In: Sergio Scalise, Elisabetta Magni and Antonietta Bisetto (eds.), Universals of Language Today, 101−128. Dordrecht: Springer. Guilbert, Louis 1971 De la formation des unités lexicales. In: Louis Guilbert, René Lagane and Georges Niobey (eds.), Grand Larousse de la Langue Française. Vol. 1, IX−LXXI. Paris: Larousse. Hatcher, Anna Granville 1951 Modern English Word-Formation and Neo-Latin. A Study of the Origins of English (French, Italian, German) Copulative Compounds. Baltimore, MD: The Johns Hopkins Press. Heyvaert, Liesbet 2009 Compounding in Cognitive Linguistics. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 233−254. Oxford: Oxford University Press. Jenkins, Fred M. 1972 Double-noun compounds in contemporary French. French Review 46: 67−73. Kim, Gyung-Ran 2001 Obstruent alternations in sub-compounds. Studies in Phonetics, Phonology and Morphology 7: 315−331. Lamy, Marie-Noëlle 1978 Neological noun-noun compounds in contemporary French. Semasia 5: 125−147.

38. Noun-noun compounds in French

687

Lees, Robert B. 1960 The Grammar of English Nominalizations. Bloomington, IN: Indiana University. Lieber, Rochelle 1992 Deconstructing Morphology. Word Formation in Syntactic Theory. Chicago: Chicago University Press. Lifetree-Majumdar, Margaret J. 1974 Contribution à l’analyse des modes de composition nominale en français écrit contemporain. Cahiers de Lexicologie 24: 63−84. Masini, Francesca 2009 Phrasal lexemes, compounds and phrases: A constructionist perspective. Word Structure 2: 254−271. Mathieu-Colas, Michel 1996 Essai de typologie des noms composés français. Cahiers de Lexicologie 69: 71−125. Montermini, Fabio 2008 La composition en italien dans un cadre de morphologie lexématique. In: Dany Amiot (ed.), La Composition dans une perspective typologique, 161−187. Arras: Artois Presses Université. Noailly, Michèle. 1990 Le Substantif épithète. Paris: Presses Universitaires de France. Picone, Michael D. 1996 Anglicisms, Neologisms, and Dynamic French. Amsterdam/Philadelphia: Benjamins. Renner, Vincent 2008 On the semantics of English coordinate compounds. English Studies 89: 606−613. Riegel, Martin 1988 Vrais et faux noms composés: Les séquences binomiales en français moderne. In: Actes du Troisième Colloque Régional de Linguistique, Strasbourg, 28−29 avril 1988, 371− 394. Strasbourg: Université des Sciences Humaines. Riegel, Martin, Jean-Christophe Pellat and René Rioul 2009 Grammaire méthodique du français. 4th ed. Paris: Presses Universitaires de France. Rohrer, Christian 1977 Die Wortzusammensetzung im modernen Französisch. Tübingen: Narr. Scalise, Sergio and Antonietta Bisetto 2009 The classification of compounds. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 34−53. Oxford: Oxford University Press. Selkirk, Elizabeth O. 1982 The Syntax of Words. Cambridge, MA: MIT Press. ten Hacken, Pius 1999 Motivated tests for compounding. Acta Linguistica Hafnensia 31: 27−58. Wandruszka, Ulrich 1972 Französische Nominalsyntagmen. München: Fink. Williams, Edwin 1981 On the notions ‘lexically related’ and ‘head of a word’. Linguistic Inquiry 12: 245−274. Zwanenburg, Wiecher 1992 Compounding in French. Rivista di Linguistica 4: 221−240.

Pierre J. L. Arnaud, Lyon (France)

688

III. Units and processes in word-formation II: Special cases

39. Verb-noun compounds in Romance 1. 2. 3. 4. 5. 6. 7.

Introduction: a productive exocentric strategy The semantic spectrum of nominal verb-noun compounds Adjectival verb-noun compounds and the unitary output hypothesis Non-lexical verb-noun compounds and the border with syntax Constraints on V and N and “peripheral” formations Formal details: interlinguistic differences within Romance References

Abstract Verb-noun compounds are a very characteristic device in Romance word-formation. The present article tries to deal briefly with most of the facets which have made verb-noun compounds the object of intense debate from the 19 th century onward. Emphasis is placed on: (i) the semantic and categorial polyfunctionality of the output; (ii) its borderline position on the syntax-morphology interface, although it is argued that verb-noun compounds are definitely the output of a morphological procedure; (iii) some interlinguistic discrepancies in the build-up of forms, including the long-term issue of the formal status of the verb-component.

1. Introduction: a productive exocentric strategy The verb-noun compounds − hereafter also VNCs − in Romance (e.g., Cat. llevataps, Fr. tire-bouchon, It. cavatappi, Port. saca-rolhas, Sp. sacacorchos, all meaning ‘corkscrew’, to quote a beautiful example from Gather 2001: 1, with wide lexical diversity in both V and N) occupy a very peculiar place within Romance word-formation, and unsurprisingly they have been the object of a great amount of interest and scholarly work from the 19th century up to present times: it goes without saying that such an amount of work can be referred to only minimally in this article. At least the following characteristic issues have been widely discussed: − they are a Romance innovation with respect to Latin (cf. Lloyd 1966, followed by most authors; for a contrary view see the few lines on Bork 1990 below), but also sufficiently old to display a basically unitary character at least across all Western Romania (they are not absent in Romanian either, but apparently much less productive, see Schapira 1985, Grossmann 2012: 155); − they offer one of the most convincing instances of a productive exocentric compounding strategy, at least in European languages; − they display a very wide semantic spectrum in their output, both as nouns and adjectives, and at the same time interesting restrictions on the features of both V and N; − they raise long-debated questions regarding the formal description of the V component, both synchronically and diachronically;

39. Verb-noun compounds in Romance − they may challenge the received idea that a morphologically defined procedure is necessarily a lexeme-forming procedure as well. Outside Romance, VNCs are attested in other Indo-European languages, but much less pervasively. Apart from the few English instances like killjoy, scarecrow, which may be dependent on French models, the type is also present at least in Ancient Greek (e.g., pheréoikos carry-home ‘snail’, proper names like Arkhélaos, lit. ‘lead-people’) and in Slavic languages (e.g., Progovac 2006 gives many examples for Serbian, like ispičutura drink up-flask ‘drunkard’, vadičep take out-cork ‘corkscrew’; as for Czech, Štichauer 2009: 198 states that the process is not particularly productive and typically displays a jocular connotation, e.g., kazisvět ruin-world ‘vandal’). From a diachronic point of view, Bork (1990) has challenged a long established tradition by arguing that VNCs were not unknown to Latin either (ultimately from Greek models), and therefore Romance languages should have inherited the type from Vulgar Latin. However, as remarked by Gather (2001: 202−203), Bork’s investigation has yielded only a handful (precisely 16) of sure cases, mostly rare words: moreover, most of them are not formed according to the Romance pattern, but rather by a parasynthetic template which adds a derivational suffix -i(us) to the V and N stems, e.g., Verticordia turn-heart ‘who changes the hearts’, epithet of Venus, poscinummius request-coin ‘eager for money’. At any rate, it seems unlikely that the relevance of the VNCs in Romance languages could be matched by their putative Latin antecedents: as said in the heading above, VN compounding is a productive strategy in all Western Romance languages, and has been so since medieval times. Much more disputable, and most probably untrue, is the lieu commun repeatedly found in the literature without empirical support, that it should be the most productive compounding strategy in Romance overall. Leaving aside the wellknown difficulties in assessing a quantitative notion of productivity, especially when different domains are compared, Gather (2001: 8) correctly remarks that in the rare instances in which frequency counts on comparable corpora have been made, N-N compounds have turned out to be by far more numerous. Clearly, type frequency does not coincide with productivity (cf. article 47 on productivity). However, Ricca (2010: 240) adds some evidence from a very large Italian newspaper corpus (around 330 million tokens), concerning both types and hapax legomena: the latter − i.e. the words occurring only once in a corpus − give an idea of a word-formation process “in real time”. Comparing the figures for VNCs with those found for A-A compounds in the same corpus by Grossmann and Rainer (2009), the ratio is only about 1: 6 for both types and hapaxes, which looks quite indicative, especially because N-N compounds are expected to be even more widespread than A-A compounds in Italian. Coming to the second qualification in the heading, exocentricity is normally viewed as the exception rather than the rule in compounding (for an attempt at typologization see Bauer 2010), so it is remarkable that the VN compounding rule is productive and has been so for a long time. Exocentricity in a compound like It. portalettere carryletters ‘postman’ seems scarcely questionable, since the head cannot be the first element, a verb, nor the second one, which is indeed a noun, but does not transfer any of its semantic and morphosyntactic features to the whole compound: lettere is [−animate], feminine plural, while portalettere is [+animate] and may be masculine or feminine, singular or plural; moreover, obviously a postman is not a kind of letter. However,

689

690

III. Units and processes in word-formation II: Special cases perhaps just because exocentric compounding is felt as somehow anomalous, there has been much theoretical effort in trying to analyze Romance VNCs as endocentric (not just a recent suggestion anyway, as it goes back to Osthoff 1878). The reader is referred to Gather (2001: 109−136) for a survey. A possibility which has often been put forward consists in interpreting the first element not as a V, but as a deverbal (chiefly agentive) noun derived from the base verb by a conversion or zero-suffix procedure (Coseriu 1977; Grossmann 1986 for Catalan; Varela 1990 for Spanish; Zuffi 1981 and Bisetto 1999 for Italian, among others). In this way, porta- in portalettere would be on a par with portatore ‘bearer’ (some have even suggested it is a truncation of the latter), and could be treated as the head of the compound. These kinds of proposals are probably too ad hoc to be convincing. However, opting for the exocentric analysis, one must at the same time recognize the hierarchical dependency between the two elements of the compound: in a sense, V behaves as an “internal” head in its relationship with N, but has no impact on the “external” syntax of the whole compound. In the recent classification proposal by Scalise and Bisetto (2009), this double level is satisfactorily captured by assigning VNCs like portalettere to the “subordinate exocentric” type. The label “subordinate” takes into account the governor-argument relationship which the VNCs share with the so-called “synthetic compounds” like truck driver, while the exo-/endocentric distinction keeps them apart.

2. The semantic spectrum of nominal verb-noun compounds VNCs display a very wide semantic spectrum, basically the same in the different Romance languages (except Romanian), summarized in (1): (1)

a. Agent N: Fr. porte-drapeau, It. portabandiera, Port./Sp. porta(-)estandarte carry-flag ‘standard-bearer’ b. Instrument N: Fr. brise-glace, It. rompighiaccio, Port. quebra-gelo(s), Sp. rompehielos, Cat. trencaglaç break-ice(s) ‘icebreaker’ c. Location N: Fr. coupe-gorge cut-throat ‘ill-famed place’, It. spartiacque separate-waters ‘watershed’, Port. corrimão run-hand ‘handrail’ d. Event N: Cat. besamà, Fr. baise-main, It. baciamano, Port. beija-mão, Sp. besamanos kiss-hand(s) ‘hand-kissing’ e. Relational A: Fr. (porte) coupe-feu, It. (porta) tagliafuoco, Port. (porta) cortafogo, Sp. (puerta) cortafuego cut-fire ‘fire (door)’ f. Qualifying A: Fr. casse-gueule break-face ‘dangerous, risky’, It. mozzafiato cut-breath ‘breathtaking’, Sp. rompepiernas break-legs ‘very tiring (esp. in cycling)’

Among the four main nominal subclasses (1a–d), the most productive (at least in presentday Romance languages) are undoubtedly the first two, namely agent and instrument nouns, and particularly the latter. A common morphological meaning for the two subclasses may be formulated as ‘entity that (usually/typically) performs the action V on N’. This general meaning incorporates the agentive (or at least [+dynamic]) character of V and the patient argument role of N (see section 5 for some exceptions), leaving

39. Verb-noun compounds in Romance unspecified the [±animate] feature. This section of the semantic spectrum is just another instance of systematic agent-instrument polysemy, not rarely found in derivation. Such a polysemy has intermediate steps, as is shown by the [−human, +animate] VNCs (denoting animals identified by a sterotypical property, like Cat. pica-soques peck-logs ‘nuthatch’, It. beccafico peck-fig ‘figpecker, warbler’, Port. papa-formigas eat-ants ‘anteater’, Sp. chupaflor suck-flower ‘hummingbird’), and by those denoting plants/flowers, which are living things but [−animate], e.g., Fr. attrape-mouches catch-flies ‘dionaea’, It. bucaneve pierce-snow ‘snowdrop’, Rom. suge-pin suck-pine ‘pinesap (a parasitic plant)’, Sp. detienebuey stop-ox ‘restharrow’. Further examples in Gather (2001: 64−66). Agentive, [+human] nouns seem to group around two main lexical fields, namely those denoting regular, job-like activities, often of lowly prestige (Cat. guardabosc watch-wood ‘forester’, Fr. porte-parole carry-word ‘spokesperson’, It. cantastorie singstories ‘street singer’, Port. limpa-chaminés clean-chimneys ‘chimney sweep’, Sp. limpiabotas clean-boots ‘shoeshine boy’) and those identifying a kind of human character, often by a derogatory and/or jocular expression (e.g., Cat. somiatruites dream-omelettes ‘daydreamer’, Fr. rabat-joie reduce-joy ‘spoilsport’, It. rompiballe break-balls ‘bore’, ficcanaso put-nose ‘meddler’, Sp. papanatas eat-cream:PL ‘dupe’, Rom. zgârie-brânză scratch-cheese ‘stingy person’, It. mangiapreti = Sp. comecuras eat-priests ‘radical anticlerical’). In the latter category the creative component is very prominent. This is apparently the only meaning which has been (weakly) productive in Romanian as well (Schapira 1985: 26), where it may still give rise to some independent new formations: an instance from Grossmann (2012: 156) is fură-becuri steal-lightbulbs ‘tall person’. On the other hand, [−animate] nouns do not always denote instruments proper. For instance, a possible meaning only vaguely relatable to the prototypical instrument is ‘(chemical) product performing the action V on N’, as in It. levamacchie = Port. tiranódoas remove-stains ‘stain remover’. The same could be said for Fr. pousse-café pushcoffee = It. ammazzacaffè kill-coffee, both ‘liqueur taken after the coffee’. Another unprototypical instrument subclass is given by items in which the V is [−dynamic], like Cat. cobrellit = Fr. couvre-lit = It. copriletto = Sp. cubrecama cover-bed ‘bedcover’. The agent/instrument readings are by no means mutually exclusive for a given compound. Cf. It. portabagagli carry-luggage:PL, both [+animate], ‘porter’, and [−animate], ‘luggage container’; Fr. garde-côte protect-coast ‘coastguard/coastguard vessel’ (Villoing 2009: 191); Cat. rentaplats wash-dishes ‘dish-washer/dish-washing machine’ (Gràcia 2002: 797), etc. As for gender assignment, [+human] VNCs quite naturally select the gender according to the sex of the referent: It. il/la portabandiera = Fr. le/la porte-drapeau, etc. For all other kinds of VNCs, the default gender is masculine, although counterexamples are found in all languages, usually when the most common hyperonym(s) is/are feminine. However, this is surely not a sufficient condition: cf. It. la lavastoviglie f. vs. Fr. le lavevaisselle m., Port. o lava-louça m. (but often f. in Brazil), all three meaning ‘dishwasher’ and relatable to the feminine hyperonym macchina/machine/máquina. The two further kinds of nominal outputs (1c) and (1d) deserve some more comment. Location nouns appear to be very rare in new formations (see, e.g., Villoing 2009 for French; Ricca 2010 has found pratically none in the whole 16-year corpus of the newspaper La Repubblica). Moreover, locative semantics may be disputable even in many apparently clear-cut cases. For instance, It. spartiacque ‘watershed’ has been taken as a good illustration of a location noun in (1c), as it denotes a portion of space (admittedly

691

692

III. Units and processes in word-formation II: Special cases a linear one), but a residual instrumental flavour cannot be denied. Conversely, I would view It. appendiabiti hang-clothes ‘coat-hanger’ rather as a limiting case of instrument noun, but its Catalan synonym penja-robes is considered a location noun by Gràcia (2002: 797). By the same token, a very productive subclass in all Romance languages denotes containers: in Italian the usual verbal base is porta- ‘carry’, often with no motion implied (e.g., portacenere carry-ash ‘ashtray’, portapenne carry-pen ‘pen-case’, etc.). These items, like It. appendiabiti above, can be viewed as borderline between instrumental and locative semantics (cf. Rainer 1993: 274), although they properly denote an object and not a place. Perhaps the same can be said of the often cited Fr. gratte-ciel = It. grattacielo = Port. arranha-céus = Sp. rascacielos, calqued on Eng. skyscraper (also Rom. zgârie-nori lit. ‘scratch-clouds’ like Ger. Wolkenkratzer), which is, probably, a bit more place-like than the instances above. As already noted by Varela (1990: 67), a VNC like Sp. lavacoches lit. ‘wash-cars’ may be freely used with all three main meanings mentioned above: ‘person who washes cars’, ‘instrument to wash cars’ and ‘place where cars are washed’. This is unsurprising, since the instrument-location polysemy is apparently quite common among derivational morphemes cross-linguistically. However, it is difficult to decide if in these cases the location interpretation has to be considered as a property of the VN-formation rule, or rather results from a common metonymic shift generally applicable to both agent and action nouns coming to denote the place where the activity is performed. Of course, this potentially available polysemy may vary in diachrony and across languages. Fr. coupe-gorge lit. ‘cut-throat’ had developed all the three meanings ‘murderer’, ‘large knife (as a weapon)’ and ‘ill-famed place’ in the history of French, but apparently it is now available only with locative meaning. The Italian parallel tagliagole ‘murderer’, however, is agentive only. Contrary to their rarity among new formations, exclusively locative VNCs are relatively common as toponyms, which are among the best attested VNCs in medieval times, and have been traditionally viewed as one of the original sources for the whole wordformation procedure. Note that in several cases the N is not a patient, but the performer of the action denoted by V: Fr. Chanteloup = It. Cantalupo/a lit. ‘sing-he/she wolf’, Fr. Hurlevent lit. ‘cry-wind’ (Villoing 2009: 189), It. Cantarana = Sp. Cantarranas lit. ‘sing-frog(s)’, all (originally) interpretable as ‘place where N performs the action V’ (although toponyms with a patient N are also attested, e.g., It. Bagnacavallo lit. ‘wethorse’, Serravalle lit. ‘close-valley’). The above pattern is very rare in “normal” VNCs − some cases are mentioned in section 5. Therefore, VNC toponyms do not seem to provide the main model for the contemporary formation rule. Event VNCs belonging to type (1d) stand perhaps even more apart, because they cannot be accomodated along an equally well attested polysemic path in word-formation (although occasionally some may cross the border: Villoing 2009: 192 cites Marc Plénat’s instance of Fr. vide-bouteille lit. ‘void-bottle’ as potentially having all four main meanings: ‘drinker’, ‘instrument to void bottles’, ‘drinking place’ and ‘good drinking’). Some well established examples of event VNCs are Cat. correbou run-ox ‘bull run’, Fr. remue-ménage move-household ‘mess’, It. voltafaccia turn-face ‘U-turn (esp. in figurative sense)’, Sp. cumpleaños accomplish-years ‘birthday’. They are undoubtedly much rarer than agent or instrument ones, and this has led several authors to consider them as plainly unproductive (cf. Corbin 2004: 1296 for French; Bisetto 1999: 509 for Italian;

39. Verb-noun compounds in Romance for Catalan, Gràcia 2002 just mentions event VNCs in a footnote). However, their productivity cannot be ruled out if one looks to sufficiently extended databases (see Villoing 2009 for French; Ricca 2010 for Italian). Ricca (2010: 251) remarks that, apart from strict analogical models, new formations among Italian event VNCs tend to concentrate in limited semantic niches (especially names of games, or activities involving body parts), but there are exceptions. It may also be the case that several instances of event VNCs, e.g., faire du ramassemiettes ‘to do garbage collecting (in computer jargon)’ cited in Villoing (2009: 192), are only licensed − or strongly favoured − by particular syntactic contexts (here, the faire du … construction). However, this cannot surely be extended to all occurrences. A clear case of syntactic conditioning which operates across different Romance languages is the construction “a/à + VN” in instances like It. correre a perdifiato (losebreath) ‘to run at breakneck speed’, Sp. a regañadientes grind-teeth ‘unwillingly’, Fr. crier/chanter à tue-tête lit. ‘kill-head’ = It. gridare/cantare a squarciagola rip up-throat ‘to shout/sing as loud as possible’, It. sparare a bruciapelo lit. ‘burn-hair’, Sp. disparar a quemarropa = Port. disparar à queima-roupa burn-dress ‘to shoot at a very short distance’. The VNCs involved normally do not exist as independent lexemes, but they are best interpretable as event nouns, even if licensed by the specific construction only. At any rate, the productivity of the pattern is minimal, and in most cases these “a/à + VN” sequences collocate quite rigidly with just one or very few verbs, giving rise to highly idiomatic, nearly lexicalized VPs (same ideas in Gather 2001: 155−156).

3. Adjectival verb-noun compounds and the unitary output hypothesis The instances of adjectival VNCs as given in (1e–f) raise a different issue. A single morphological rule appears to form productively items belonging to two different syntactic categories, and thus provides a clear challenge to the unitary output hypothesis (henceforth UOH) found, e.g., in Scalise (1984: 137). Even outside the lexicalist framework, this constraint has been considered empirically very solid, or even “inherent in an output-oriented approach” like the one of Plag (1999: 243). Probably for this reason, there is no general consensus about Romance VNCs really constituting a counterexample to the UOH. To be sure, there can be little doubt that they can be used extensively as noun modifiers: for some further data, see, e.g., Gather (2001: 155−159). However, for many authors this does not automatically imply a substantial violation of the UOH. Several ways out are found in theoretically-oriented descriptions (especially in the Italian tradition; less unanimously, but prevalently in Spain as well, as discussed for instance in Gather 2001: 155−165). It can be argued that a modifier VNC is first formed as a noun by the compounding rule, and then transformed into an adjective by virtue of a further N → A conversion rule. Alternatively, the VNC may be considered as a noun in an appositional relationship with its head; or even as the modifier N in a N-N (loose) compound. The latter solution, proposed by Zuffi (1981), seems, however, hardly applicable to the instances in which the VNC occurs separated from its head (like It. scatola d’argento portafiammiferi box of-silver hold-matches ‘silver matchbox’).

693

694

III. Units and processes in word-formation II: Special cases It is not by chance that the advocates of the all-nominal approach normally refer to instances like It. nave portacontainer = Fr. navire porte-conteneurs ship carry-containers ‘container carrier’. In such cases, the VNC is also widely employed autonomously as a noun, often much more widely than as a modifier. Therefore, it is easy to consider the modifier use as derived from the nominal one. However, in many other instances the modifier function is the only one attested, or at least is largely prevailing in use. For these cases, a reductionist approach complying with the UOH becomes much less convincing (similarly Rainer 1993: 273). Perhaps it may still be feasible for the relational VNC adjectives, as the one in (1e) (many more examples in Gather 2001: 155), because − even when a corresponding nominal use is scarcely attested − they keep some nominal features: they have often an instrumental reading and they are not gradable. Moreover, the tight syntagmatic links with their head, with which they often constitute a single referential unit, may suggest a N-N compoundlike interpretation of the whole sequence. But cases like It. mozzafiato ‘breathtaking’ in (1f ) cannot be treated this way. This item has all the features of a full-fledged qualifying adjective, including: − occurring in predicate position: quel panorama è veramente mozzafiato ‘that panorama is really breathtaking’; − occurring in coordination with other clearly qualifying adjectives: panorama stupendo e mozzafiato ‘marvellous and breathtaking panorama’; − being gradable: un panorama ancora più mozzafiato ‘a still more breathtaking panorama’. Full-fledged qualifying VNC adjectives are present in all Romance languages and should suffice to rule out a rigid version of the all-nominal approach. On the other hand, they are undoubtedly a small minority among well established VNCs, as is reflected also in dictionary-based counts (cf. Gather 2001: 155). Therefore, from lexicographical data it could be argued that the UOH for VNCs still holds as a strong tendency. However, Ricca (2010: 249−254) has shown that corpus data draw quite a different picture, at least for Italian. While they confirm the prevalence of nouns for high frequency items, the low frequency ones, and especially the hapax legomena, turn out to be quite balanced between nominal and modifier uses. Since, as said above, the hapax legomena in a very large corpus are the best approximation to the behaviour of a productive word-formation rule “in action”, it seems fair to argue that the VN-compounding process in itself is better left unspecified regarding its output category. The clear preference for nouns among the firmly entrenched items can then hardly be ascribed to the formation pattern itself, and has probably to do with the different lexicalization potential between nouns (referential entities) and adjectives (property concepts) in general.

4. Non-lexical verb-noun compounds and the border with syntax The investigations of large newspaper corpora by Ricca (2005, 2010) have put in evidence a further interesting feature of many hapax or nearly hapax VNCs, namely their markedly non-lexical features, especially when employed as modifiers.

39. Verb-noun compounds in Romance Typical instances are constructions like the following (Ricca 2010: 252), which are quite often met in Italian newspapers (it would be interesting to verify if similar structures are common in other Romance languages as well): (2)

a. con il colpo di testa fissa-risultato di Fonseca ‘with Fonseca’s result-securing header’ b. alludendo alla ventilata astensione salva-Prodi ‘alluding to the proposed Prodi-saving abstention’

From a communicative point of view, such items “work” more or less like a reduced relative clause. However, they undoubtedly belong to the same compounding pattern seen till now, and therefore in our view they are still the output of a morphological procedure, not a syntactic one. But taking the view that the items in (2a–b) are compounds does not straightforwardly imply that they are lexemes: not just because they are nonce formations, but because they are also extremely poor candidates for lexical storage. A good argument for the inherently non-lexical nature of the VNCs in (2) is that they often keep the referential autonomy of the noun intact, which is patent when a proper noun is involved, as in (2b), but holds for (2a) as well. It is also significant that in this use the V element does not convey at all the habituality feature which is nearly always present in “lexical” VNCs, nonce formations included (and is usually taken as a general property of VNCs, see, e.g., Varela 1990: 65). From these − and several other − instances of compounds, Gaeta and Ricca (2009) argue for the necessity of keeping well separated the issues of lexicalizability (i.e storage/ storability as a unit in the mental lexicon) and “morphologicalness” (i.e. being the result of a non-syntactic procedure), an issue neatly raised already by Corbin (1992: 50). By the suggested two-feature analysis, instances like the VNCs in (2) would be [+morphological], [−lexical] entities, contrary to the often tacit assumption that noninflectional morphology is always a lexeme-forming device, at least potentially. More generally, the assignment of the whole VN compounding procedure − not just the non-lexical cases in (2) − to syntax rather than morphology has been often proposed. We cannot deal with the theoretical debate here (apart from the few hints below, the reader is referred to the survey in Gather 2001: 28−63 and 115−126). Apparently, an argument in favour of the syntactic option could be given by the well-known instances of V[NP] formations, like It. porta-carta igienica = Fr. porte-papier hygiénique = Port. porta papel higiênico ‘toilet paper holder’. Gather (2001: 153) already reports a fairly long list for French, but innumerable instances with bases like porta-/porte- can now be found most easily on the Internet for all Western Romance languages, for any conceivable kind of container/holder on sale on the Web. The NP involved may well have a complex structure itself, and in particular govern a PP, as in It. copriborsa dell’acqua calda cover-hot water bag ‘hot water bag cover’ (Ricca 2005: 479), Fr. nettoie-chaînes de vélo ‘bike chain cleaner’ (quoted in Gather 2001: 153), despite the contrary statement for Spanish in NGLE (2009: 771). In many cases, as the ones above, the NP (and consequently the whole compound) can be considered a unit in the mental lexicon, but not necessarily so: instances of V[NP] formations with features like those in (2) occur as well.

695

696

III. Units and processes in word-formation II: Special cases Nevertheless, the morphological option for VN formations still seems us to fit better the data. Perhaps the strongest argument against the syntactic option, put forward by Corbin (1992: 48), Rainer (1993: 275) and many others, is the virtual absence of V[DP] compounds (i.e. compounds including the determiner). If VNCs were to be synchronically derived from VPs structures, they should freely allow for, or indeed prefer, a V[DP] pattern (*porta-la-carta-igienica and the like), which is definitely not the case. Instances of lexicalized V-Det-N formations are in fact attested (e.g., Fr. trompe-l’oeil id.; lit. ‘deceive-the-eye’, It. battiloro beat-the-gold ‘gold-beater’, apart from family names like Fr. Boileau = It. Bevilacqua lit. ‘drink-the-water’, It. Cantalamessa lit. ‘sing-the-mass’, etc.), but they are exceedingly rare and are best treated, as Gather (2001: 22−23) argues, as a separate process, i.e. as idiosyncratic instances of univerbation from frozen clausal chunks, not qualitatively different from the types It. nontiscordardimé ‘forget-me-not’, Sp. hazmerreír make-me-laugh ‘laughing stock’. A further strong argument in favour of the essentially morphological nature of VN formations (the same point is raised, e.g., by Rainer 1993: 75) is given by the interesting syntactic and semantic restrictions involving both V and N, which will be discussed below in section 5. They should not be expected if the VN formations were the output of the same syntactic rule which forms VPs.

5. Constraints on V and N and “peripheral” formations Concerning phonology, no categorical restriction appears to hold for VN compounding, although in all Romance languages a tendency may be detected which favours “short” V bases. Being “short” translates as bisyllabic in Italian and Spanish, and monosyllabic in French, due to the different phonotactics of the languages in question. So Rainer (1993: 269) reports an overwhelming majority (around 80 %) of compounds with bisyllabic V bases in his corpus, and a similar percentage (around 75 %) is found by Ricca (2010: 244) among over 2,200 Italian VNCs occurring in a large newspaper corpus. As for French, Villoing (2009: 194) obtains very similar percentages (80 %) for monosyllabic bases. Percentages become still higher if one considers that many trisyllabic bases (bisyllabic for French) begin with a vowel, which often undergoes resyllabification with the preceding word − e.g., an article − and therefore may be considered “extra-metrical”. Methodologically, this can be a tricky issue, since a prevailing number of short bases might be just a side effect of a more general tendency to avoid words that are too long. Ricca (2010: 244−245) checked out this possibility, showing that no similar preference is displayed among deverbal derivatives of equal length in the same corpus. Thus the preference for two-syllable verbal bases appears to be specific to VN compounding, at least for Italian. It is obviously just a tendency, which does not rule out perfectly legitimate (and attested) V bases with up to five syllables at least, like It. macchina distribuisci-biglietti ‘ticket-dispensing machine’. A similar tendency concerning the length of N seems harder to prove, despite some proposals along these lines (convincingly dismissed by Gather 2001: 16−17). As another general limiting factor on VNC formation, it has been often observed that only a very small set of potential verb bases really gives rise to a significant number of VNCs. Gather (2001: 11−12) quotes some lists of the most frequent bases for French,

39. Verb-noun compounds in Romance Spanish, and Portuguese, mostly taken from lexicographic sources, which seriously limits their ability to reflect the formation process proper. For Italian, however, Ricca (2005, 2010) made counts both on lexicographical sources and on large corpora, confirming that the base distribution is very skewed. These facts are not astonishing at all, because a high skewedness is commonplace in every kind of word frequency distribution when productive processes are involved (cf. Baayen 2001). Nevertheless, it may be significant that some semantic types recur almost universally across languages and sources: the top ranked concepts include ‘carry’, ‘protect/save’, ‘cover’, ‘cut’, ‘kill’, ‘break’, ‘pull’, ‘count’, and the first two are probably the leading items. For some of the most frequent bases, there may be some potential for grammaticalization toward prefixal status (cf. Gather 2001: 11−12). This is clearly the case for the French series with pare-, because the corresponding verb with the meaning ‘stop/protect’ is no longer synchronically available, but Gather suggests a similar analysis for Sp. porta- as well, given the limited role played by the free verb portar when compared to the very high productivity and semantic autonomy of the compound base porta- in the sense of ‘container/holder’. Even It. portacould be considered from the same perspective (Ricca 2010: 247), despite the high frequency of the corresponding free verb portare ‘to carry’, because, as in Spanish, in most of its occurrences in compounds there is no motion capability implied. The real restrictions on VNC formation, however, concern the syntactic/semantic features of both its lexical components. The overwhelming majority of VNCs, including the new formations, is formed from transitive verbs which allow for an agentive subject, and the N nearly always takes the role of the internal argument of such verb, and specifically the object (as remarked already, e.g., by Varela 1990: 69−71 for Spanish). This is true also for plant words and the huge class of instrument compounds: although the implied subject is [−animate] in these cases, the verbal base in itself is compatible with [+animate] subjects as well. There are obviously exceptions to the two very general constraints above, but on the whole they are really limited in number, drawing a sort of basically unproductive “periphery” around the core of the VN compounding procedure. This periphery will be briefly illustrated below (a longer discussion may be found in Gather 2001: 76−86, and, for French, in Villoing 2009). The restriction on the agentivity of V rules out in particular those verbs which take an experiencer as their first argument. So formations like It. *odiaparenti (hate-relatives), Sp. *quiereanimales (love-animals, Varela 1990: 70) or Fr. *craint-étrangers (fear-foreigners) are normally excluded, despite their semantic plausibility. Gather (2001: 86) cites as a historically relevant exception a series of French formations with aimer ‘to love’, common among literary coinings in the 16th century only. As already said above, [−dynamic] verbs occur as long as they may allow − in a different context − for a controlling first argument, i.e. they are not stative proper. This is the case for many verbal bases like It. porta- ‘carry’, reggi- ‘hold’, copri- ‘cover’, which are in fact among the most productive of all and give rise to a host of names for containers, holders, covers. But the impossibility of formations like Sp. *tienefiebre lit. ‘have-fever’ (Varela 1990: 70) or It. *pesachili lit. ‘weigh-kilos’ seems to hold across all Romance languages. Concerning the argument role of N, the exceptions are somehow more relevant, and may be divided in two classes: the instances where N is the external argument (subject)

697

698

III. Units and processes in word-formation II: Special cases of V, which normally implies that V is intransitive, and those in which it is a non-object internal argument of V. Well-known and often cited items in the first category include: It. marciapiedi walkfeet ‘pavement’, batticuore beat-heart ‘heartthrob’, Fr. trotte-bébé trot-baby ‘baby walker’, pense-bête think-idiot ‘reminder’, Sp. saltacaballo jump-horse ‘arch springer’, reposacabezas rest-head ‘headrest’ (together with several others with reposa-, cf. Rainer 1993: 270). Even from this shortlist, the peripheral nature of these items may be appreciated also looking at their unpredictable semantic output: the otherwise uncommon location and event nouns are well attested, as well as quite idiosyncratic instances of patient VN (Villoing 2009: 190), like croque-monsieur crunch-Mister ‘toasted sandwich’ (i.e. ‘food to be crunched by men’). As said above, a coherent group of VNCs with subject N is found among toponyms. As for the second type, N as an oblique argument appears in very few scattered items where N may fulfil different semantic roles, e.g., source (It. scendiletto go down-bed ‘bedside rug’), path (the uncommon Sp. andarríos go-brooks ‘wagtail’), direction (the name for ‘sunflower’ shared among most Romance languages: It. girasole, Sp./Cat. girasol, together with the now opaque Fr. tournesol). However, within this periphery an interesting semantic niche appears to be crosslinguistically productive. From verbal bases meaning ‘protect’, two open lists of compounds may be formed, one where N is the object of V (i.e. meaning ‘object protecting N’, often but not exclusively a body part) and another where N is the oblique argument: ‘object protecting from N’. This fact has been often observed for Spanish: Rainer (1993: 270) gives examples from the two bases guarda- (e.g., guardabarros ‘mudguard’) and protege- (the new formation protege-esquinas ‘protector from edges’). Gather (2001: 83), although adding similar French examples with garde-, underestimates the relevance of the pattern, doubting its productivity outside strict analogy. However, there are plenty of similar formations also from Fr. protège- (like protège-vent, protège-soleil ‘wind, sun shelter’ very common on the Internet) and Italian offers a host of further examples with four different bases, not only para- and proteggi- ‘protect’, but also ripara- ‘shelter’ and above all the very productive salva- ‘save’. A list of Italian items of the type ‘protect from’ taken from Internet data (e.g., proteggi-/para-/salva-/ripara-vento ‘wind protecting (device)’), with more than 20 different Ns, has been presented in Ricca (2008) and is found also − with some expansion − in Magni (2010: 19), who discusses extensively the pattern for the three languages. The same behaviour in different languages, without full overlap in the choice of the lexical bases, clearly suggests something more than strict analogy: in this case the saliency of the “oblique” argument − which often may occur alone, without overt specification of the object, as noted, e.g., by Val Álvaro (1999: 4797) − overcomes the syntactic constraint requiring the object role for N. A different kind of peripheral formations involves recursivity. It is well known that Romance languages allow for recursive compounding in general only to a limited extent, at least in comparison with Germanic languages. The VN formations are no exception: recursive V[VN] compounds are neither ungrammatical nor unattested, but they are surely rather uncommon. Not unexpectedly, a high level of lexical entrenchment of the internal VN seems to be required, which does not mean necessarily a high level of opacity. Formations of frequent use include several compounds with the very productive base porta-/porte- denoting containers/holders, like It. portastuzzicadenti = Cat. portaescu-

39. Verb-noun compounds in Romance radents carry-[pick-teeth] ‘toothpick holder’, It. portasciugamani = Fr. porte-essuiemains carry-[dry-hands] ‘towel holder’, It. portacontachilometri = Sp. portacuentakilómetros carry-[count-kilometres] ‘odometer holder’. Other instances often display a higher degree of opacity of the inner VNC, as It. svuotaportafogli empty-[carry-sheets] ‘wallet-emptying (said of taxes, holidays, etc.)’, Sp. limpiaparabrisas wipe-[protect-breezes] ‘windscreen wiper’. Notice that there are VVN sequences which cannot be analysed as recursive V[VN] compounds. These are items like It. tergilavafari = Fr. lave-essuie-phares = Sp. limpialavafaros ‘headlight washer/wiper’, It. lavasciugabiancheria = Sp. lavasecarropas ‘clothes washer/dryer’, in which the two verbs are in a coordinative relationship and both govern the noun. Therefore, the compound must be described according to a [[VV]N] structure. Interestingly, this template is licensed only by the whole constructional pattern, since coordinative VV compounds are exceedingly rare in Romance languages, and there are no compound verbs like It. *lavasciugare, Sp. *lavasecar ‘to wash and dry’ or It. *tergilavare, Sp. *limpialavar ‘to wipe and wash’. Obviously only semantics, or perhaps just world knowledge, may distinguish the two types. For instance, It. lavasciugamani might be interpreted both ways, although it is currently attested with the [[VV]N] meaning, namely ‘hand washer/dryer’ and not ‘towel washing machine’. Both templates combined together, to give a [[VV][VN]] structure, occur at least in one Spanish well established compound, limpia-lavaparabrisas ‘windscreen washer/ wiper’. Finally, brief mention may be made here of exocentric VAdv and VAdj compounds, because, although quite limited in number and scarcely productive at all, they show a great semantic affinity with the VNCs. A short list for Italian, Spanish and French is found in Gather (2001: 18). The adverb is not always a modifier, but often plays an argumental role in these compounds. Both main [+animate] types are represented, the job/role terms (It. buttafuori throw-out ‘bouncer’, Sp. mandamás command-more ‘big boss’) and the qualifying/derogatory terms (It. cacasotto shit-under ‘coward’, Fr. lèvetard wake up-late ‘late riser’), as well as the instrument nouns (Fr. passe-partout passeverywhere ‘master key’, Sp. tirafuera pull-out ‘sort of fishing net with a long pole’) and also the event nouns: It. tagliafuori cut-out ‘boxing out in basketball’, Port. botaabaixo throw-down ‘destructive criticism’ (Río Torto and Ribeiro 2012: 136). Among the perhaps even less numerous VAdj compounds, one could distinguish those where the adjective has in fact a nominal function (e.g., It. menagramo bring-bad ‘jinx’, Fr. gagne-petit earn-little ‘low wage earner’), and those where the Adj has an adverbial use (It. aprifacile open-easy ‘easy to open’, Port. fala-barato speak-cheap ‘chatterbox’, Sp. cantaclaro sing-clear ‘very frank person’). Neither subtype seems to add anything to the general pattern of VNCs.

6. Formal details: interlinguistic differences within Romance VNCs have been treated so far without making any distinction within Romance, and illustrated by examples taken more or less at random from the different Romance languages. The very feasibility of this procedure shows the extent of the uniformity shared by VNCs within the whole (Western) Romance area. However, cross-linguistic discrepan-

699

700

III. Units and processes in word-formation II: Special cases cies do exist, which are difficult to deal with in a short survey. They seem to concern essentially specific points in the build-up of forms, rather than semantic features and general restrictions. In the following, such points of detail will be briefly illustrated in a contrastive perspective between Italian and Spanish, which happen to differ extensively in these respects. French would be harder to evaluate at any rate, due to the great difference in relevance of inflectional markers between the graphematic and the spoken code (the former should not be fully disregarded in this particular domain, given that several new formations plausibly appear first in the written medium; for some discussion see Villoing 2009).

6.1. The V stem The first and most hotly debated issue concerns the precise character of the V element. In this domain care should be taken in keeping the synchronic and diachronic perspectives well separated. The three positions on the nature of V, as surveyed for instance in Rainer (1993: 265−268) and Gather (2001: 87−92), are: (i) V is an imperative 2nd person form; (ii) V is an indicative 3rd person form; (iii) V is a bound verbal stem. However, if formulated as above the three alternatives make sense chiefly as a diachronic issue. Clearly, this was also the main focus of the debate until the first half of 20th century. The imperative proposal was most energetically defended by Darmesteter (1894) with respect to French and, according, e.g., to Rainer (1993: 265), essentially won the contest until the revival of the indicative alternative, prompted especially by Tollemache’s (1945) study of Italian. Not only historical, but also semantic arguments were put forward in defending the different options. For instance Tollemache’s plea for the indicative, despite its formal implausibilty for Italian (see below), rested on the paraphrasability of most VNCs as ‘entity who does the action V on N’. From a contemporary and synchronic perspective however, the option that the V occurring in VNCs is a kind of verbal stem seems hardly questionable, as stressed, e.g., by Villoing (2009: 176−179) for French. Whatever form V takes in these compounds, it carries just its lexical content, not the meaning associated with any specific verbal inflectional marker: neither the imperative nor the indicative, and much less so a specific person value. Consequently, whatever the ultimate origin of the model(s) which triggered the morphological template (quite plausibly the imperative: for a partial re-evaluation of its role even in synchronic word-formation, see Floričić 2008), the present-day productive formation rule should inevitably be formulated in terms of a verbal stem (a morphome in the sense of Aronoff 1994), and not of verbal inflectional forms. The question remains, however, how such a morphome is formed, and some cross-linguistic divergence is found in this respect. In Italian, the V stem in VNCs formally coincides with the 2nd person imperative for all verbs. There is no possibility of relating it to 3rd person indicative, because the two forms differ for all verbs in 2nd and 3rd conjugations (indicative regge ‘holds’, apre ‘opens’ vs. imperative reggi, apri). In Spanish, the opposite pattern holds: while there is a nearly complete overlap between the forms of 2nd imperative and 3rd indicative for all three conjugations, in the rare cases where they contrast (e.g., with tenir ‘keep’ and derivatives) the few instances of lexicalized compounds show the 3rd singular form, and speakers agree that possible new formations would do the same (Val Álvaro 1999: 4789).

39. Verb-noun compounds in Romance In both languages, the stem occurring in VNCs cannot be identified with the derivational stem. In Spanish, for 3rd conjugation verbs the VNC stem ends in -e, but the derivational stem ends in -i: abrelatas open-cans ‘can opener’ vs. abridor ‘opener’. Moreover, many verbs show stem alternations correlated with stress, and the VNC stem keeps the stressed variant, with very few exceptions: e.g., cuentagotas count-drop ‘dropper’ vs. contador ‘counter’ (Rainer 2001: 391). In Italian, the parallel vowel-diphthong alternation involves few verbs (e.g., trattenere ‘to keep’: dispositivo trattieni-odore ‘(bad) smell-keeping device’ vs. trattenimento ‘keeping’), but the same contrast occurs for the open class of verbs with the so-called -isc- augment, as pulire ‘to clean’: they usually keep it in VNCs, as in imperative (pulisciorecchie ‘ear cleaner’), but never in derivatives (pulitore ‘cleaner’). Catalan displays an even more complex picture, since for some bases two different stems coexist, e.g., cobrellit ‘bed cover’ vs. cobriespatlla cover-shoulder ‘shawl’ from cobrir ‘cover’, and both differ from the 3rd indicative/2nd imperative form, which is cobreix in today’s Catalan (Gather 2001: 101). To reduce the impact of the issue, it must be said that the problematic cases of Vstem formation concern essentially the basically unproductive verb classes outside the first conjugation. It is probably not irrelevant that in the productive first conjugation there is full coincidence between 3rd indicative and 2nd imperative in all Romance languages except Sardinian, which may have lead to reanalyses in the different directions.

6.2. The internal N plural Many VNCs in all Romance languages display the N in plural form also when used in the singular, as seen in the examples above. Contrary to the V pseudo-inflectional marker, such plural marker is generally not semantically void. The clearest proof is the − admittedly rare − occurrence of two different VNCs with singular and plural N, e.g., It. portauovo carry-egg ‘egg cup’ vs. portauova carry-eggs ‘egg container’; Sp. matarrata kill-rat (a game) vs. matarratas kill-rats (a poison). The N plural marking occurring in Romance VNCs has been pointed out, at least since Booij (1994: 37), as a good instance of “inflection feeding word formation”, and therefore as an argument for nominal Number belonging to inherent inflection. However, the details vary across the languages. In Italian, its meaningful, non-morphomic character seems quite stable also beyond the cases of contrast like portauovo/a. Ns are always plural in pluralia tantum (e.g., portaocchiali bear-glasses ‘glasses case’), and, more relevantly, when each single state of affairs associated with the compound (typically) involves a plurality of Ns (as in contapassi count-steps ‘step counter’). The contrast between schiaccianoci crush-nut:PL ‘nutcracker’ and cacciavite thrust-screw:SG ‘screwdriver’ shows the area of unpredictable variation, which regards, quite naturally, VNCs whose referent acts on a single N in a given state of affairs, but on a plurality of Ns on different occasions: they could be labelled “variable-N compounds”. In this subclass, both alternatives may often occur in the same item: fermacravatta/e hold-tie(s) ‘tiepin’. Conversely, the singular N occurs quite regularly for “stable-N compounds”, where the VNC typically acts on a single referent throughout (e.g., copriletto cover-bed:SG ‘bedcover’), and categorically when N is a unique referent (giramondo ‘globetrotter’) or a mass noun (spazzaneve ‘snow plough’, rompighiaccio ‘ice-breaker’).

701

702

III. Units and processes in word-formation II: Special cases The pattern in Spanish is different. Singular VNCs display the -s plural marker on N in the great majority of cases. The -s has extended even to unique referents (cf. trotamundos ‘globetrotter’ vs. It. giramondo above) and to mass nouns (cf. Sp. quitanieves ‘snow plough’, rompehielos ‘ice-breaker’ vs. It. spazzaneve, rompighiaccio), although exceptions remain (a list in Val Álvaro 1999: 4798), not only within the “stableN” type, but sometimes even among “variable-N” compounds (Am. Sp. chupaflor suckflower ‘hummingbird’). Such an extension implies a substantial change in the role of the -s marker. As already noted by Rainer and Varela (1992: 130), the -s in Spanish VNCs has mainly taken the role of marking the composition process itself, with no consistent semantic motivation. That is, it has come a long way to becoming a morphome, similar to the linking elements familiar from the compositional processes in German and other languages (cf. article 32 on linking elements in Germanic) with the interesting difference that the -s marker occurs at the word’s end and not at the border between the lexical morphemes. However, the generalization of -s in VNCs is surely not a concluded process, and probably has different relevance across the Spanish-speaking world: for a variety of Argentinian Spanish, Rainer (1993: 272) even reports the opposite tendency of omitting the -s in the singular only. Catalan seems to take an intermediate position between Italian and Spanish: according to Gràcia (2002: 798−99), the plural N is the normal choice for countable Ns, but does not extend to unique referents (rodamón vs. Sp. trotamundos) or mass nouns (llevaneu vs. Sp. quitanieves).

6.3. The plural of the whole compound The above variation in the marking of the internal N plural has obvious consequences for the possibility of marking external plurals, i.e. the plural of the whole compound. Nowhere in Romance languages is there an independent slot available for this function. Consequently, all VNCs with internal plural Ns are necessarily invariable. Very marginal exceptions may at most occur if the compound is so deeply entrenched in the mental lexicon that it may be perceived as non-analyzed unit. For instance, the Italian plural i paracaduti ‘the:M:PL parachutes’ is not found in dictionaries, but on the Internet it is about as frequent as the expected i paracadute. This deviant form necessarily implies a reanalysis of the (F):PL -e ending, occurring in the singular il paracadute stop-fall:PL ‘parachute’, as the (M):SG ending of the -e/-i inflectional class (as in can-e ‘dog’), which in turn cannot happen without a substantial opacization of the compound. A different state of affairs occurs with VNCs displaying a singular N in the singular. Since these are a minority in Spanish, Italian is more interesting in this respect. As a general rule, it appears that the possibility of marking the whole VNC as plural by pluralizing the internal N obeys plausible semantic distinctions. Both “variable-N” and “stable-N” compounds may display a plural N when they have plural reference, but there is great oscillation and in most cases both forms are possible and largely attested: i cacciavite/i thrust-screw(s) ‘screwdrivers’, i copriletto/i cover-bed(s) ‘bed covers’. This makes sense semantically, since for both subclasses a plurality of VNC referents normally implies a plurality of Ns as well. Therefore, there is no need to analyse the plural -i in spazzacamini sweep-chimney:PL ‘chimney sweeps’ and the like − as done by Scalise (1994: 139) − as an instance of external plural implying that the compound is treated as

39. Verb-noun compounds in Romance an unanalysable whole, like paracaduti above. The -i is still an internal plural compatible with the inflectional class of N, and quite naturally licenses the plural interpretation of the whole compound. The latter analysis is supported by the fact that when there is a gender clash between the N and the VNC, the compound is mostly invariable: i portabandiera carry-flag (f.) ‘standard-bearers (m.)’. The clash may be overcome (a plural i portabandiere ‘the:M.PL carry-flags’ is possible, as witnessed by the many occurrences on the Internet), but always in favour of the internal plural (here the (F):PL ending -e of bandiera): a form like *i [portabandier]-i, with the -i modelled on the inflectional class of masculines in -a (like poeta/poeti) is unacceptable.

6.4. Interaction with other morphological domains Traditional lexicalist models usually order derivation before composition (cf. Scalise 1984: 115−122), which would imply that VNCs may freely take derived Ns as input, but would rule out the inverse case, apart from idiosyncratic instances of very opaque, essentially unanalyzed compounds. The first hypothesis is easily confirmed by the data, as stated also by Gather (2001: 142). To give just a single example, cf. It. copri[teiera] = Fr. couvre[thé-ière] = Sp. cubre[te-teras] cover-teapot ‘tea cosy’. Maybe the high frequency items are not very numerous, but in corpora many more can be found, although for Spanish the possibility of internal derivations is rejected in NGLE (2009: 747). The further derivability of VNCs is indeed less straightforward. At least for Italian, data suggest a neat distinction between prefixal and suffixal derivation, which is probably to be expected, given the recognized higher autonomy of prefixes with respect to their bases. Obvious examples of prefixes freely combining with VNCs are anti- ‘anti-, against’, which attaches without problems to the wide subclass of [+animate] derogatory VNCs (e.g., anti-ficcanaso ‘anti-meddler’), and ex- ‘former’, which similarly freely applies to the equally wide domain of job nouns (e.g., ex-cantastorie ‘former storyteller’). Both types have very solid attestations on the Web. As for suffixation, the picture seems different. Surely in every language some well established examples can be found, but often they involve partially opaque VNCs (this is the case for It. paracadut-ista = Sp. paracaid-ista ‘paratrooper’, It. guardaroba guardthing ‘wardrobe’ → guardarob-iera ‘wardrobe maid’, Sp. paraguas protect-water ‘umbrella’ → paragü-ero ‘umbrella maker’, as also Gather 2001: 143−144 suggests). Similar considerations hold for N → V conversions: among the few Italian instances one could list paracadutare ‘to parachute’, ficcanasare ‘to nose around’, rendicontare ‘to report’ from the highly lexicalized paracadute, ficcanaso ‘meddler’ and rendiconto give(ac)count ‘report (N)’. Such a restriction may be weakened, however, for some specific suffixes which more generally display practically no structural limitation on their input: the best example in Italian is -ismo ‘-ism’. So the model of quality nouns from [+animate] derogatory VNCs (like rompiball-ismo, from rompiballe break-balls ‘bore’) is presumably productive, since it is widely attested, at least on the Web, with many VNC bases. The situation is probably similar in Spanish: the only example of a derived VNC cited in the NGLE (2009: 771) is chupamedismo, from chupamedias suck-socks ‘flatterer’, and several

703

704

III. Units and processes in word-formation II: Special cases quality nouns parallel to the Italian ones are abundantly present on the Web, with some even attested in dictionaries, e.g., papanatismo ‘dumbness’. Finally, as for the interaction of VNCs with evaluative morphology, we do not expect to find significant restrictions against diminutive/augmentative Ns serving as input for VN compounding, neither semantically nor theoretically, as long as the Ns in question have some referential autonomy, and this is certainly the case in Italian (cf. porta[telefonino] carry-telephone:DIM ‘mobile holder’, guarda[port-one] guard-door:AUGM ‘doorkeeper’; examples could be multiplied). Such internal diminutivization is however ruled out in NGLE (2009: 747) for Spanish, similarly to derivation in general. The possibility of applying evaluative morphology to the whole compound is more interesting and again displays some degree of contrast between Italian and Spanish. While the option seems rather restricted in both languages, its formal realizations differ. Spanish prefers the diminutive − when acceptable − to attach at the N, although the semantics involves the whole compound, e.g., cortauñas → cortauñitas cut-nails:DIM ‘[little [nail clipper]]’ (NGLE 2009: 770). However, the picture is made complex by the possibility of interpreting the above formations as infixed diminutives according to the pattern Carlos → Carl-it-os (Gather 2001: 150). In Italian, on the contrary, diminutives referring to the whole compound are not as marginal as Gather (2001:150) would suggest, but they overwhelmingly apply externally: so paracadute → [paracadut]-ino ‘little parachute (e.g., on hang gliders)’ vs. *paracadutine, portafogli carry-sheets ‘wallet’ → portafoglino/portafoglietto ‘little wallet’ (not *portafoglini), guardaroba → guardarobino ‘child wardrobe’. This holds surely for the denotative uses of evaluatives, while for the pragmatic use, the “landing site” may vary: so for the meaning ‘little bore m./f’, rompiball-ino/a ‘[break-ball(s)]-DIM:M/F:SG’ alternates on the Web with rompiballine ‘break-[ball:DIM:(F):PL]’.

7. References Aronoff, Mark 1994 Morphology by Itself. Stems and Inflectional Classes. Cambridge, MA: MIT Press. Baayen, Harald 2001 Word-Frequency Distributions. Dordrecht: Kluwer. Bauer, Laurie 2010 The typology of exocentric compounding. In: Sergio Scalise and Irene Vogel (eds.), Cross-Disciplinary Issues in Compounding, 167−175. Amsterdam/Philadelphia: Benjamins. Booij, Geert 1994 Against split morphology. In: Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 1993, 27−49. Dordrecht: Kluwer. Bisetto, Antonietta 1999 Note sui composti VN dell’italiano. In: Paola Benincà, Alberto Mioni and Laura Vanelli (eds.), Fonologia e morfologia dell’italiano e dei dialetti d’Italia. Atti del XXXI Congresso internazionale di studi della Società di Linguistica Italiana, 505−538. Roma: Bulzoni. Bork, Hans Dieter 1990 Die lateinisch-romanischen Zusammensetzungen Nomen + Verb und der Ursprung der romanischen Verb-Ergänzung-Komposita. Bonn: Romanistischer Verlag.

39. Verb-noun compounds in Romance Corbin, Danielle 1992 Hypothèses sur les frontières de la composition nominale. Cahiers de Grammaire 17: 25−55. Corbin, Danielle 2004 Français (Indoeuropéen: Roman). In: Geert Booij, Christian Lehmann, Joachim Mugdan and Stavros Skopeteas (eds.), Morphology. An International Handbook on Inflection and Word-formation. Vol. 2, 1285−1299. Berlin/New York: de Gruyter. Coseriu, Ernesto 1977 Inhaltliche Wortbildungslehre (am Beispiel des Typs “coupe-papier”). In: Herbert E. Brekle and Dieter Kastovsky (eds.), Perspektiven der Wortbildungsforschung, 48−61. Bonn: Bouvier. Darmesteter, Arsène 1894 Traité de la formation des mots composés dans la langue française comparée aux autres langues romanes et au latin. 2nd ed. Paris: Champion. Floričić, Franck 2008 The Italian verb-noun anthroponymic compounds at the syntax/morphology interface. Morphology 18(2): 167−193. Gaeta, Livio and Davide Ricca 2009 Composita solvantur: Compounds as lexical units or morphological objects? Italian Journal of Linguistics 21(1): 35−70. Gather, Andreas 2001 Romanische Verb-Nomen-Komposita. Wortbildung zwischen Lexikon, Morphologie und Syntax. Tübingen: Narr. Gràcia, Lluïsa 2002 Formació de mots: composició. In: Joán Solà, Maria Rosa Lloret, Joan Mascaró and Manuel Pérez Saldanya (eds.), Gramàtica del català contemporani. Vol. 1, 777−829. Barcelona: Empúries. Grossmann, Maria 1986 Anàlisi dels compostos catalans del tipus somiatruites. Estudis de Llengua i de Literatura catalanes 12: 155−169. Grossmann, Maria 2012 Romanian compounds. Probus 24: 147−173. Grossmann, Maria and Franz Rainer 2009 Italian adjective-adjective compounds: Between morphology and syntax. Italian Journal of Linguistics 21(1): 71−96. Lloyd, Paul 1966 A possible structural factor in the development of Verb-Noun Compounds in the Romance languages. Studia Neophilologica 38: 257−262. Magni, Elisabetta 2010 From the periphery to the core of Romance [VN] compounds. Lingue e linguaggio 9: 3−39. NGLE 2009 Nueva gramática de la lengua española. Madrid: Asociación de Academias de la Lengua Española. Osthoff, Hermann 1878 Das Verbum in der Nominalcomposition im Deutschen, Griechischen, Slawischen und Romanischen. Jena: Costenoble. Plag, Ingo 1999 Morphological Productivity. Structural Constraints in English Derivation. Berlin: Mouton de Gruyter. Progovac, Ljiljana 2006 Fossilized imperative in compounds and other expressions: Possible implications for historical and evolutionary studies. In: Online Proceedings of the First Meeting of Slavic

705

706

III. Units and processes in word-formation II: Special cases Linguistics Society. Bloomington, IN. http://www.docin.com/p-442356629.html and http://www.researchgate.net/publication/228732520_Fossilized_Imperative_in_ Compounds_and_Other_Expressions_Possible_Implications_for_Historical_and_ Evolutionary_Studies [last access 30 Oct 2014]. Rainer, Franz 1993 Spanische Wortbildungslehre. Tübingen: Niemeyer. Rainer, Franz 2001 Compositionality and paradigmatically determined allomorphy in Italian word-formation. In: Chris Schaner-Wolles, John Rennison and Friedrich Neubarth (eds.), Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60 th birthday, 383−392. Torino: Rosenberg & Sellier. Rainer, Franz and Soledad Varela 1992 Compounding in Spanish. Rivista di Linguistica 4(1): 117−142. Ricca, Davide 2005 Al limite tra sintassi e morfologia: I composti aggettivali V-N nell’italiano contemporaneo. In: Maria Grossmann and Anna M. Thornton (eds.), La formazione delle parole. Atti del XXXVII congresso della Società di Linguistica Italiana, 465−486. Roma: Bulzoni. Ricca, Davide 2008 VN compounds in Italian: Data from corpora and theoretical issues. Paper presented at the CompoNet Congress on Compounding, Bologna, 6−7/6/2008. Ricca, Davide 2010 Corpus data and theoretical implications: With special reference to Italian V-N compounds. In: Sergio Scalise and Irene Vogel (eds.), Cross-Disciplinary Issues in Compounding, 167−175. Amsterdam/Philadelphia: Benjamins. Río Torto, Graça and Sílvia Ribeiro 2012 Portuguese compounds. Probus 24: 119−145. Scalise, Sergio 1984 Generative Morphology. Dordrecht: Foris. Scalise, Sergio 1994 Morfologia. Bologna: Il Mulino. Scalise, Sergio and Antonietta Bisetto 2009 The classification of compounds. In: Rochelle Lieber and Pavol Štekauer (eds.), The Oxford Handbook of Compounding, 34−53. Oxford: Oxford University Press. Schapira, Charlotte 1985 Les composés roumains à thème verbal et leur place dans l’ensemble des langues romanes. Revue de Linguistique Romane 49: 15−26. Štichauer, Pavel 2009 Compounds in Czech. Lingue e linguaggio 8(2): 188−209. Tollemache, Federico 1945 Le parole composte nella lingua italiana. Roma: Rores. Val Álvaro, José Francisco 1999 La composición. In: Ignacio Bosque and Violeta Demonte (eds.), Gramática descriptiva de la lengua española. Vol. 3, 4757−4841. Madrid: Espasa. Varela, Soledad 1990 Composición nominal y estructura temática. Revista española de lingüística 20: 55−81. Villoing, Florence 2009 Les mots composés VN. In: Bernard Fradin, Françoise Kerleroux and Marc Plénat (eds.), Aperçus de morphologie du français, 175−198. Saint-Denis: Presses Universitaires de Vincennes.

40. Co-compounds

707

Zuffi, Stefano 1981 The nominal composition in Italian: Topics in generative morphology. Journal of Italian Linguistics 2: 1−54.

Davide Ricca, Turin (Italy)

40. Co-compounds 1. 2. 3. 4. 5. 6.

Introduction The meaning, use and form of co-compounds Co-compounds in Europe and elsewhere Co-compounds as context-dependent and context-renewing units Co-compounds and construction morphology References

Abstract Co-compounds are word-like tight units mostly consisting of two parts which express natural coordination and superordinate-level concepts in contrast to sub-compounds which mostly express subordinate-level concepts. In this article it is argued that cocompounds should be considered in their natural environment in texts, since they do not only have characteristic formal and semantic properties, but most importantly characteristic patterns of use. In Europe co-compounds occur particularly in Eastern languages, but also in Basque. However, cross-linguistically co-compounding forms a discrete cline rather than a parametric feature that languages have or lack.

1. Introduction Co-compounds (also known as dvandva, copulative compounds, or pair words [Russian parnye slova], cf. article 20 on composition) are word-like tight units consisting of two parts (or more rarely, more than two parts) which express natural coordination, viz. coordination of things or events that often occur together with characteristic lexical domains including pairs of relatives (Rural Tok Pisin [Mühlhäusler 1979: 377] papamama ‘father-mother’ > ‘parents’, brata-susa ‘brother-sister’ > ‘siblings’), body parts (han-lek ‘hand-foot’ > ‘limbs’), clothes (su-soken ‘shoe-sock’ > ‘footwear’), and collectives and abstract notions (rit-rait ‘reading-writing’ > ‘skills learned at school’). They are not characteristic of European English or other Standard Average European languages, but can be exemplified, for instance, with Indian English:

708

III. Units and processes in word-formation II: Special cases (1)

Indian English: reported speech in an English novel a. ‘Are you maybe married already, captain? Got wife-children waiting somewhere?’ b. ‘However we can help our father-mother that is what it is for us to do.’ (Rushdie 1981: 403, 228)

Co-compounds tend to cluster areally. They typically occur with moderate or high frequency in languages of Asia and Eastern Europe, New Guinea, and Mesoamerica. There is nothing in the structure of Standard Average European languages that prevents them from having co-compounds, as their presence in non-European varieties of English demonstrates. Following the Sanskrit grammarians, traditional morphology considers any compound whose syntactic paraphrase is a coordination (ca-arthé ‘and-denoting’) to be a dvandvacompound, thus disregarding that coordination can be manifested in very different ways in compounds. Here it is argued that it is indispensable to consider the meaning of the whole compound in order to define compound types. While subordinate compounds (sub-compounds) typically denote subordinate-level concepts (Mari [Uralic] kid-tup ‘hand-back, back of the hand’), co-compounds typically express superordinate-level concepts (Mari kid=jol ‘hand=foot > hand and feet, limbs’). This was first noted in the description of coordinate compounds in American Sign Language (Klima and Bellugi 1979). Wälchli (2005) restricts the term “co-compound” to coordinative compounds denoting superordinate-level concepts. However, the superordinate-level meaning component of co-compounds is not equally strong in all examples, which is associated with the fact that the meanings of co-compounds exhibit family resemblance. There is no semantic feature that applies equally well to all examples. “Superordinate-level concept” is the closest one can come with a single criterion. However, compounds with a coordinate and appositional relationship between the parts which denote subordinate-level compounds, such as intermediate-denoting compounds (southwest), appositional compounds (French wagon-restaurant, see article 38 on noun-noun compounds in French and section 3.3) and complex numerals (twenty-three) are excluded. Formally, co-compounds are tight or “word-like” units (they typically, but not universally, lack overt markers of conjunction); only in few languages are co-compounds phonological words as, for example, in Modern Greek and Classical Sanskrit. Usually cocompounds are intermediate between words and phrases, which is represented graphically in many orthographies by hyphenation. In this article hyphens are rendered as equal signs to avoid confusion with the hyphens used in the glosses for the delimitation of morphemes. Co-compounds are dealt with from a typological point of view in Wälchli (2005, 2007a), Bauer (2009), and Arcodia, Grandi and Wälchli (2010). This article focuses on co-compounds in Europe and on the fundamental context-dependent nature of co-compounds − which is notoriously underestimated in most approaches to wordformation. Co-compounds are properties of particular texts belonging to particular registers and cannot easily be detached from the context in which they occur. Consider, for instance, the Russian example travka=muravka in (2). Many texts in Russian, especially in Literary Standard Russian, do not contain any co-compounds at all. (2) is from a Bylina, an epic poem, for which the abundant use of several word-formation devices, such as diminutives and co-compounds is characteristic.

40. Co-compounds (2)

Pod nim trav-ka=murav-ka ne topč-et-sja Russian (Bylina) Under he:INST grass-DIM=ant-DIM not crush-PRS3SG-RFL ‘Under his feet the grass and ants are not crushed’ (Propp and Putilov 1958: 258)

The Byliny constitute part of spoken rural Russian. However, it would be wrong to say in general that travka=muravka ‘grass=ant’ is a co-compound of rural Russian, it can only be understood in the particular context from which (2) is taken. The example is about the hero Čurila Plenkovič who is so noble that he does not even crush what is below his feet when he walks. Travka=muravka is an excellent example for illustrating the function of co-compounds to denote concepts above the basic level of conceptualization. The general notion of ‘animals and plants you can crush when you walk’ is instantiated by the two prototypes ‘grass’ and ‘ant’. In this context, grass and ants exemplify natural coordination. However, it is difficult to imagine any other context where ‘grassant’ would be a suitable instance of a natural pair. A further important factor in (2) favoring the use of a co-compound is the generalizing and non-referential context. We are not talking about concrete animals and plants but rather about eventualities for which collective terms are highly suitable. Given the importance of the meaning of the whole, it is not particularly useful to classify co-compounds according to the semantic relationship between the parts. Rather, the semantic relationship between parts and whole must be considered. Wälchli (2005: 138) proposes a convenient sub-classification into ten types: additive (the whole C is A and B, as in (1, 5, 15)), generalizing (C is everything/everybody/everywhere/always with A and B as extreme components as in (6, 8)), collective (the whole C is represented by the prototypical members A and B, as in (2)), synonymic (C is in the context the same as A and as B as in (3, 4, 6, 9)), ornamental (B or A is misleading and does not contribute to C, as in (22)), imitative (B or A is phonologically similar to the other part and does not mean anything), figurative (C belongs to a domain that differs from the domain to which A and B belong as in (24)), alternative (C is A or B), approximative (C is A or B or something close to A or B), scalar (C is a scale defined by the poles A and B as in (23)). Examples for all types are given below in passing. The contextual character of co-compounds is particularly notable in synonymic cocompounds where the parts A and B are not general synonyms, but only happen to mean the same thing in a concrete context of usage. This is illustrated in (3) from Hungarian: (3)

mi ügy-e=baj-a van Hungarian what thing/matter-POSS3SG=trouble-POSS3SG be:PRS3SG nek-i valahol. DAT-POSS3SG somewhere ‘[Because why shouldn’t I ask you,] what’s wrong with him. [But I said nothing.]’ (Füst 2000: 25)

The two words ügy ‘thing/matter’ and baj ‘trouble’ are far from meaning the same thing in general. However in the given context Mi ügye van? ‘What’s the matter?’ and Mi baja van? ‘What’s wrong?’ are nearly synonymous. This example shows that co-compounds are not formed by a combination of lexemes in abstraction of their contextual meanings. Rather co-compounds are formed by a combination of word-form tokens in a

709

710

III. Units and processes in word-formation II: Special cases concrete speech situation. However, co-compounds are not simply a coordination reduction of two sentences. This can be seen in (3) if case marking is considered. The dative (nek) is idiomatic with baj-a ‘wrong-its’. However, with ügy-e ‘matter-its’ in isolation the comitative (vel) would be used. Example (3) may further serve to illustrate the ways in which co-compounds denote concepts above the basic level of conceptualization. The meaning of the whole co-compound is particularly relevant. The level of the whole compound need not be more general than the meaning of the parts. If the parts denote general concepts already (and there are few nouns with more general meaning than ügy ‘thing’) there is no need for the co-compound to be more general than the simplex would be on its own. Co-compounds can be formed from words from a large variety of non-grammaticalized word classes: nouns, verbs, adjectives, adverbs, demonstratives, and numerals. Numerals are typically used to form alternative co-compounds such as Mari ik=kok ‘one= two > one or two’ or approximative co-compounds such as Chuvash pӗr=ikӗ ‘one=two > some’ (Paasonen 1941: 197). Some languages including Mari and many Turkic languages seem to disprefer co-compounds consisting of finite verbs which may be associated with their dispreference for balanced coordination. However, occasionally verbal cocompounds with a converb in the first part may be encountered (cf. the discussion of (12) below). It is not self-evident how co-compounds should be delimited from other phenomena. This follows quite naturally from the family resemblance character of the category and the absence of explicit formal marking in many languages (see section 2 below). The alternative type, for instance, is quite common even in Standard Average European (in English less so than in German and Italian) which otherwise avoid co-compounds. I include alternative co-compounds in particular because they are used in similar ways as other co-compounds in distributive contexts (see example (16) below). There is an old tradition of viewing co-compounds as a kind of reduplication, recently rearticulated in Inkelas and Zoll (2005). Wälchli (2007b) argues that co-compounds are clearly different from reduplication. Unlike co-compounds, reduplication has only one free lexical slot. In co-compounds the form of one part is not predictable from the form of the other part. For further delimitation problems, such as parallelism and serial verbs, see Wälchli (2005).

2. The meaning, use and form of co-compounds Above I have argued that it is a major characteristic of co-compounds that they express natural coordination (as opposed to accidental coordination). However, natural coordination is not restricted to co-compounds. Most languages have a general tendency to express natural coordination in tighter formal units than accidental coordination, for instance, in English to be able to read and write (natural) vs. to be able to read and to swim (accidental). Put differently, tight coordination is iconic for natural coordination and loose coordination is iconic for accidental coordination. The iconic relationship between meaning and form is summarized in Table 40.1 (see also Wälchli 2005, ch. 2−3 for further discussion).

40. Co-compounds

711

Tab. 40.1: The iconic relationship between different kinds of coordination Semantic dimension

Formal dimension

Close relationship between the coordinands

Natural coordination

Tight coordination

Distant relationship between the coordinands

Accidental coordination

Loose coordination

A typical phrase-like tight coordination pattern are the so-called bare binomials in Germanic languages (Lambrecht 1984), such as brother and sister, law and order, German Pfeil und Bogen ‘bow and arrow; lit. arrow and bow’ whose essential formal property is the lack of articles (thus ‘bare’). Whether bare or not, binomials tend to abound in legalese texts, e.g., we decree and grant that all other cities, boroughs, towns, and ports shall have all their liberties and free customs; safe and secure by land and water (Magna Charta 13; 42). Binomials differ from co-compounds in that the coordinands can consist of more than one word (e.g., free custom) which is not characteristic of co-compounds (except for Northwest Caucasian, see (7) below). Binomials often tend to be formulaic, e.g., French au fur et à mesure ‘gradually as’, German an Ort und Stelle ‘in place’, and this is paralleled by co-compounds. The Kazan Tatar synonymic co-compound kѳç=xəl ‘power=power’, for instance, tends to occur especially in the collocation kѳç=xəl belən ‘power=power with > just barely, with pain and misery’ (cf. German mit Mühe und Not): (4)

Ğabbar kѳç=xəl belən burıç-tan kot-ıl-a. Kazan Tatar Ğabbar power=power with debt-ABL save-PASS/MIDD-PRS3SG ‘Ğabbar gets away from debt only with pain and misery.’ (Fəjsi 1993: 108)

Binomials also share the property with co-compounds that they can contain a component without any meaning of its own. English kith in with kith and kin or fro in to and fro only occur in these combinations very much like Turkish çoluk is restricted to the imitative co-compound çoluk çocuk ‘IMI child > wife and family’. While some co-compounds occur in contexts where other languages can have phrasal coordination or a phrase-like tight coordination pattern such as bare binomials as translational equivalents, there are many other co-compounds which correspond to single words in languages with few or no co-compounds. This is shown in Wälchli (2007: 163) who considers the distribution of co-compounds and coordination in 18 passages in translations of the Gospel according to Mark in a genealogically diverse sample of 41 languages from all continents. The use of co-compounds extends from typical coordinative contexts (‘night and day’, ‘parents, father and mother’, ‘to come and go’) to typical one-word contexts, such as ‘children’, ‘clothes’, ‘fields’, and ‘people’. That co-compounds often correspond to words rather than to coordinate phrases can be illustrated also with their use in proverbs. In (5) from Komi, ‘to sow’ is expressed with the co-compound ‘to plow=sow’. Of course, sowing usually presupposes that the field has been plowed first and ‘plow’ and ‘sow’ thus are a natural pair. However, for the proverb ‘plowing’ is completely irrelevant and it would be rather strange in English to say ‘You reap what you plow and sow’.

712

III. Units and processes in word-formation II: Special cases (5)

Myj ge̮r-an=ke̮ďź-an, sija=j pet-al-as. what plow-PRS2SG=sow-PRS2SG, that=also result-FUT-3SG ‘You reap what you sow.’ (Timušev 1971: 75)

Komi

As far as form is concerned, co-compounds are often not distinctively marked in any particular way (see, e.g., article 182 on Basque). They simply consist of a juxtaposition of unmarked parts as in (1) and (5). However, if they are marked, it is very common for them to have a double marking strategy where both parts of the compound bear the same marker. Double marking is formally symmetric and hence iconic for the symmetric construction type of coordination (Haiman 1985). Double marking often derives from the same syntactic function of the two parts of the compound as in (3) with the repeated Hungarian possessive suffix; this might be called “inflectional harmony”. However, double marking can also involve derivational morphology as with the repeated diminutive suffix in the Russian example (2). The tendency toward “inflectional harmony” can be so strong as to require a repeated affix even if the affix would not occur if a simplex word was used instead of a co-compound. In Mordvin, for instance, most co-compounds in the nominative singular indefinite, which is unmarked, will usually take the plural suffix -t/-ť in both parts of the co-compound, even if only the whole co-compound is to be interpreted as a plural, not the individual parts: Erzya Mordvin ťeťa-t=ava-t ‘fatherPL=mother-PL > parents’ (lit. ‘fathers-mothers’). Such co-compound-specific double marking is particularly common for generalizing co-compounds. In Mari (6), it is usually the possessive suffix of the third person singular or an additive focus clitic, in Mordvin it is the otherwise non-productive comitative marker -n’ek (pokš-n’ek=viški-n’ek ‘big-COM=small-COM’): (6)

Tid-lan verčin Orina-n šüm=čon-žo Meadow Mari this-DAT because Orina-GEN heart=soul-POSS3 jüd-žö=kečy-že tulšol gaj night-POSS3=day-POSS3 embers like ‘Because of that Orina’s heart hurt night and day like glowing embers.’ (Šketan 1991: 16)

Example (6) from Mari further illustrates that co-compounds also can be single marked. The synonymic co-compound šüm=čon ‘heart=soul’ in (6) bears a single possessive suffix. Single marking is iconic for minimal distance which is typical for tight coordination, but languages differ in their preference for double and single marking strategies. While in Mordvin the majority of co-compounds are double marked, Mari and Tatar have a large proportion of single marked co-compounds. Tab. 40.2: Iconicity in marking strategies for natural coordination Symmetry

Minimal distance

(+)

+

Single marking strategy

−

+

Double marking strategy

+

−

Zero marking strategy

40. Co-compounds Whether there is a co-compound-specific double marker does not depend only on the semantic subtype of the co-compound. Not all generalizing co-compounds in Mordvin are marked with -n’ek/nek (Wälchli 2005: 140). What is important to note here is that for many languages there is no single general rule that can produce the form of all cocompounds and not even of one semantic subtype of co-compounds. The morphologically most complex co-compounds in Europe are probably found in Northwest Caucasian languages which are known for their high degree of polysynthesis. In Abkhaz it is no problem to have co-compounds with parts consisting of more than two words. However, there is usually one part repeated and one part varied as in (7). The co-compound 'ɥ-ɑχɑ='ɥə-mʂ ‘two-day=two-night’ is even interesting from a semantic point of view; while ‘night and day’ is natural coordination, be it generalizing as in (6) or additive as in Latvian vien-u dien-nakt-i ‘one-ACC:SG day-night-ACC:SG > one day (period of 24 hours)’, (7) from Abkhaz is a rather unusual kind of additive co-compound. It would be expected that ‘two’ should have scope over the whole co-compound, but it is part of each member. (7)

'ɥ-ɑχɑ='ɥə-mʂ 'ɑ-ʈʂ’ɑqʷ’a-'rɑ j-ɑ-'ʈʂ’ə-n Abkhaz two-night=two-day its-nick-MASDAR it-it-be.in-FIN ‘[Since it (the woodpecker) knew that, if he (the hero Abrskj'yl) went to sleep, he would actually not so quickly wake up, it perched calmly beside his staff and spent] 2 days and nights chipping away at it.’ (Hewitt 2005: 242)

In (8) the prefixes nɑ- ‘thither’ and ɑ:- ‘hither’ yield a typical generalizing co-compound (‘> in all directions’). Similar generalizing co-compounds with alternating prefixes can also be found in Kartvelian languages (see Wälchli 2005: 201). (8)

Abkhaz ɑ-'χʷ(.)ɑdɑ-'kʷɑ nɑ-p'ʂə=ɑ:-p'ʂə-rtɑ-n the-hillock-PL thither-look=hither-look-place-FIN ‘[And it is known why Abrskj'yl would choose hillocks as hiding-place:] hillocks were places with an all-round view; [as soon as his enemies emerged, they at once fell under his sights.]’ (Hewitt 2005: 222)

The morphologically complex co-compounds in Northwestern Caucasian are a morphological variant of discontinuous co-compounds which are not characteristic of European languages, but occur in many languages of Asia (e.g., Hmong, Khasi, Karen), Mesoamerica (e.g., Mixe, Chinantec) and New Guinea (e.g., Toaripi). In discontinuous cocompounds, the parts of the compound A and B are interrupted by a repeated element C, often a word, according to the formula CACB or ACBC. An example from Hmong Daw where discontinuous co-compounds are very frequent is the synonymic co-compound teb chaws ‘land land > land’ as in kuw lub teb lub chaw ‘I CL land CL land, my land’ or Yawm Pus teb Yawm Pus chaw ‘Yau Pu’s land’; Bisang 1988: 36, 56, 37; Wälchli 2005: 102). A further peculiar feature of co-compounds in some Caucasian languages is that they can contain affixes for ‘and’, which can be observed in the two synonymic co-com-

713

714

III. Units and processes in word-formation II: Special cases pounds in Abkhaz in (8). Suffixes for ‘and’ in co-compounds can also be found in some Nakh-Dagestanian languages. (9)

χʲdzə-j='pʂɑ-j Abkhaz pʂɑ-rɑ-j='sɑχʲɑ-j look-ABST-and=face-and name-and=reputation-and 0̸-z-'gə-mə-z 'pħʷəz-bɑ ssəjr-k’ they-whom-be.lacking-not-NFIN woman-young lovely-a ‘[Very long ago there was in Abkhazia] a wonderful maiden who lacked neither looks nor reputation.’ (Hewitt 2005: 209−210)

The examples from Abkhaz show again that the form of co-compounds cannot be generated by a single rule. Some co-compounds have overt markers of coordination, some do not. While prefixes are usually repeated in co-compounds, there is a more varied picture as far as suffixes are concerned. In some verbal co-compounds suffixes occur on both parts, in others only on the second part as in (8).

3. Co-compounds in Europe and elsewhere Co-compounds are characteristic of languages of three macro-areas: Eurasia, New Guinea and Mesoamerica. In Eurasia the highest density of co-compounds is found in the easternmost continental languages (e.g., Mandarin, Vietnamese, Hmong) and the frequency of co-compounding decreases towards the west. Europe is the westernmost part of this continuum, there is no special European kind of co-compounding. Co-compounds in the Caucasus, for instance, have about the same frequency level in European and Asian Caucasian languages at least as far as non-Indo-European languages are concerned. European co-compounds are simply Eurasian co-compounds in languages with moderate, low or very low levels of co-compounding. A first question to address then is whether there are any characteristic ways in which Eurasian co-compounds differ from co-compounds in New Guinea and Mesoamerica. For at least some languages of Mesoamerica, metaphorical co-compounds are highly characteristic. In Classical Nahuatl there are almost only metaphorical co-compounds, e.g., yn j-petla-tzin in i-cpla-tzin ‘DEF his-mat-HON DEF his-seat-HON > throne, government’ and they are characteristic devices of elaborate style (Lehmann and Kutscher 1949). Metaphorical co-compounds do occur in European languages (see, e.g., example (24)), but they are never as dominant as they can be in a Mesoamerican language. Languages in New Guinea with co-compounds including Tok Pisin typically have a co-compound for ‘people’ consisting of the parts ‘man-woman’ and this form tends to be highly frequent in texts, e.g., Tok Pisin man-meri ‘man-woman > people’, Abau uwrsa ‘man-woman > people’, East Kewa oná-aa ‘woman-man > people’, Kobon nɨbi bɨ ‘woman man > people’. The same also occurs in some Australian languages with cocompounds: Wik Mungkan pam wanch ‘man woman > people’. In all the languages mentioned here this co-compound occurs on average in about every seventh verse or more of the translation of the New Testament. Therefore, a useful simple strategy for finding co-compounds in New Guinea is to consider the expression for ‘people’. This test does not work at all in Eurasia.

40. Co-compounds

715

It is not easy to decide which European languages have and which ones lack cocompounds. Co-compounding is a continuous, not a discrete feature. Co-compounds are not evenly distributed across different registers and styles in a language. There are, for instance, huge differences between newspaper, fiction and traditional folklore in Erzya Mordvin (Wälchli 2005: 220). In Russian, most texts in the literary language lack cocompounds almost completely, while they are quite frequent in the epic Byliny (see example (2) above) and are common even in colloquial style in Modern Russian. Tkačenko (1979) observes an areal continuum across East Slavic varieties in the frequency of the co-compound žili=byli ‘lived=were > (once upon a time there) were’ in the beginning of fairy tales (even though žili=byli ‘lived=were’ originally might derive from a pluperfect, see Kiparsky 1967: 230 for discussion). For cross-linguistic comparison it is most convenient to consider parallel texts which largely represent the same register and style (Wälchli 2005, ch. 6). The most reliable single word indicator for co-compounds in Eurasia is ‘parents’. Nominal additive co-compounds are a central group of co-compounds and within them kinship terms tend to be salient and the domain ‘parents’ is particularly important for natural coordination of kinship terms. However, not even the ‘parent’-test divides Eurasian languages into two neat groups. There are some languages such as Georgian and Turkish where ‘parents’ can, but need not, be expressed by a co-compound and in Latvian there is usually a co-compound for ‘parents’ only in folk songs (in the dainas) but not elsewhere. This is opposed to languages such as Avar, Mordvin and Chuvash where the co-compound is the only lexicalized option for ‘parents’ (see Map 6.2 in Wälchli 2005: 217). The areal east-west distribution of co-compounds is salient for instance in FinnoUgric and Turkic languages. Pitkanen and Heikkilä (see article 176 on Finnish) claim that dvandva compounds are the oldest type of compounds in Finnish, but the Finnic languages and Hungarian have a considerably lower frequency of co-compounds than Mordvin and Komi. Uotila (1980) observes parallels in the extent of co-compounding in Finnic and Baltic languages. In Turkic languages across Eurasia co-compounds form an areal cline. As a tendency they are more frequent in Eastern than in Western Turkic languages irrespective of the genealogic affiliation within Turkic. It is no coincidence that co-compounds are not mentioned for Karaim in Lithuania which is strongly influenced by Slavic languages. Example (10) shows that Karaim prefers a Slavic loanword for ‘parents’. (10) bu=mo dz’ed-łar jox juv’-d’a. DEM=Q parent-PL NEG.EX house-LOC ‘Are the parents not at home?’ (Kowalski 1929: 112)

Karaim

An extremely simplistic and therefore inaccurate approximation to describing the occurrence of co-compounds in Europe would be to say that co-compounds are lacking in Indo-European and Semitic and are present in all other languages of Europe. In fact, if you browse through the articles in the final two volumes of this handbook you will find co-compounds described for most non-Indo European languages but not for most IndoEuropean languages of Europe. Notable exceptions for Indo-European include Tat and Modern Greek (and rural Russian for which there is no article).

716

III. Units and processes in word-formation II: Special cases Co-compounds in Greek are an innovation. According to Browning (1983: 99−100) dvandva compounds which first appear in late Hellenistic Greek become particularly frequent in the Turkish period. However, this does not mean that Greek co-compounds are generally due to language contacts with Turkish. Especially their form is very different. Greek co-compounds are exceptional in representing phonological words. This may be because at least some co-compounds have developed from exocentric compounds. In Classical Greek andrógunos ‘hermaphroditic’ is an adjective with a nominalized masculine form andrógunos ‘coward, hermaphrodite’. According to Debrunner (1917) there is only one attested case in Classical Greek where, in an epigram, the adjective can be interpreted as a co-compound: andróguna (PL:N) loutrá ‘baths for men and women’ and there it is an adjective. It may therefore be assumed that Modern Greek andró-guno ‘man-woman > couple’ originates from an exocentric compound and not from coordination diachronically. Turkish co-compounds, however, often look very much like sequences of two words, especially in verbal co-compounds where the first part is a converb. Yet, there are some close lexical parallels between Modern Greek and Turkish, consider, e.g., (11) and (12) from the New Testament. (11)

óso pérn-i na anigo-klisi to as.much take-PRS3SG in.order.to open-close:SUBJ3SG DEF:SG:N máti eye ‘in the twinkling of an eye’ (1st Corinthians 15:52)

(12) … göz aç-ıp kapa-yan-a dek … eye open-CNV close-PTCP-DAT as.far.as ‘in the twinkling of an eye’ (1st Corinthians 15:52)

Modern Greek

Turkish

In translations into Turkic languages it is common to express the passage with two verbs ‘to open’ and ‘to close’ where the first is a converb (e.g., Kazan Tatar, Uzbek). In Kirgiz and Turkmen the collocation is hyphenated, marking in orthography that it is a cocompound. While most other translations use different strategies, “opening and closing of an eye” is not completely restricted to Turkic and Greek. In the Balkans we can add Albanian hap e mbyll sytë ‘open and close eye:PL’ and outside areal reach Portuguese num abrir e fechar de olhos, Spanish en un abrir y cerrar de ojos (only in the modern translation en Lenguage Sencillo) and in Africa Wolof ci xef=ak=xippi ‘PREP blink= with=close.eyes’. The example shows that it is difficult to delimit co-compounds from other looser coordination strategies expressing natural coordination and that it is difficult to track down language contact effects beyond any doubt. The most notable exception to the areal east-west co-compound cline in Europe is Basque which has far too many co-compounds for its western location. But if we now turn to Indo-European western languages, not even these lack co-compounds altogether as can be seen in examples (13) and (14) from French and Spanish. Additive co-compounds of adjacent periods of times such as days of the week are not unusual in Western Europe as, for instance, in (13) from French.

40. Co-compounds (13)

French Le week-end est généralement le vendredi et le samedi dans les pays arabes, parfois jeudi et vendredi, exceptionnellement samedi dimanche (avec un temps laissé pour accomplir la prière le vendredi midi). ‘The week-end is generally Friday and Saturday in the Arab countries, sometimes Thursday and Friday, in exceptional cases Saturday Sunday (with time left to perform the Friday noon prayer).’ (Wikipedia)

The Spanish example in (14) is more intricate: (14) … el bilingüismo es un vaivén y no sólo Spanish the bilingualism is a go:PRS3SG:and:come:IMP2SG and not only un ven … a come:IMP2SG ‘Bilingualism is a give and take and not only a take’ (http://elpais.com/diario/1997/06/18/cultura/866584802_850215.html) The masculine noun vaivén ‘swaying, swinging, up and down, hither and thither’ looks synchronically as if it was the present third singular form of ‘go’ (va) plus ‘and’ (y) plus the second person imperative of ‘come’ (vén). Since it does not make much sense if the category of the two parts is so different, the form might be a loan from Catalan vaivé. What is particularly interesting about (14) is that the second part of this entirely lexicalized noun becomes independent, but not in its original meaning of ‘come!’ A general description of the meaning of vaivén would be ‘repeated alternative movement in more than one direction’. By “decomposing” it, the innovated part vén acquires the meaning of unidirectional movement. Example (14) is taken from a speech of the director of the Real Academia Española about the current situation of Spanish. It is a rhetorical device and it is convenient for the speaker in the context that the exact meaning of vaivén is difficult to pin down exactly. The example is, of course, a pun and hence of limited significance to the plain use of language. However, it demonstrates that co-compounds interfere with style and idiomatic language use.

4. Co-compounds as context-dependent and context-renewing units Handbooks tend to have a bias on subject matters that can easily be generalized and abstracted from their concrete context of use. I have to break against this handbook convention here to emphasize the contextual embedding of co-compounds. The most fundamental misunderstanding many morphological approaches to compounding suffer from is the belief that categories of word-formation can be considered in abstraction of their contexts of use. An exception is Seiffert (see article 123 on word-formation and text) who argues that word-formation does not occur isolated as a rule. This is certainly true of co-compounds. Even though there are few co-compounds which can only be interpreted if more than one sentence is taken into consideration, co-compounds are discourse units in the sense

717

718

III. Units and processes in word-formation II: Special cases that they are context-dependent and context-renewing, to use a common slogan from conversation analysis (Atkinson and Heritage 1984). In order to illustrate this point, I will discuss selected co-compounds from one particular text in one European language Erzya Mordvin, the story Vit’a Kil’d’az’even’ er’amo ‘The life of Vitya Kil’djazev’ (first part) by Vasilij Kudaškin published in the Erzya literary journal S’atko (Kudaškin 1996). It is a story of fourteen short chapters which is representative of Literary Erzya Mordvin. Vitya is a naughty schoolboy in a Mordvin village during the time of Stalinism before World War II. The most obvious discourse structuring aspect of co-compounds is their use with proper nouns. A co-compound consisting of two names expresses that there is a close relationship between two persons; most often this is used for siblings, spouses or other pairs of relatives. In Kudaškin (1996) there are co-compounds of Vitya together with as many as four different other schoolboys and these work to structure the text into episodes of different adventures with different companions. Typically the other boy is first introduced as a simple name before the additive co-compound of proper names occurs. In (15) Mikita is Vitya’s desk neighbor with whom he quarrels which is why the teacher makes the two of them stand in the corner. The co-compound illustrates the double plural marking mentioned in section 2 above. pr’a-n’ komavto-z’ Erzya Mordvin (15) Vit’a-t=Mikita-t Vit’a-NOM:PL=Mikita-NOM:PL head-GEN bow-PTC sirga-s-t’ ugol-s set.off-PST-3PL corner-ILL ‘Vitya and Mikita went to the corner bowing their heads.’ (Kudaškin 1996: 7) The co-compounds of proper names depend on the previous context and create groups which are locally valid in episodes, not in the language as a whole, not even in one text as a whole. Another context-dependent use of co-compounds is distributivity. Example (16) describes how Vitya as an older boy walks a long way to the school in town under difficult circumstances “through snow and dirt, through rain and storm” which is expressed by two co-compounds. While rain and storm often co-occur, snow and dirt can occasionally be combined, but more often they occur on different occasions. There is snow in winter and dirt in autumn and spring. The co-compound thus summarizes different difficult road conditions which apply on different occasions. The expression “wade through snow and dirt” does not make much sense if considered in isolation of this distributive context, marked here with the adverb s’ejet’ste ‘often’. Erzya Mordvin (16) … s’ejet’ste kel’e-ms lov-ga=rudaz-ga, often wade-INF snow-PROL=dirt-PROL piz’eme-n’=davol-on’ pačk … rain-GEN=storm-GEN through ‘[He studied in the second shift. He had to walk home in the dark and] often to wade through snow and dirt, through rain and storm.’ (Kudaškin 1996: 46)

40. Co-compounds

719

The distributive use is particularly common with alternative and approximative co-compounds in examples such as “Each family had five-six horses”. Co-compounds often become good superordinate level concepts only in particular contexts. This can be illustrated with (17): (17)

Erzya Mordvin Paro azor-tnen’ rozor-iz’=pan-s’-iz’ good lord-PL:DEF:GEN destroy/disown-PST3PL >3PL=chase-FREQ-PST3PL>3PL ‘The good peasants were disowned and expelled’ (Kudaškin 1996: 23)

Example (17) occurs in an episode where Vitya sings a song on collectivization he has learned in school. His mother tells him to stop this, he asks why since they themselves are not kulaks (affluent farmers) and his mother replies that the others aren’t either, that Stalin just made this up and that all the good peasants were disowned and expelled. It is difficult to imagine another context of use for the additive co-compound ‘disownexpel’ than collectivization and the formation of kolkhozes. The co-compound moreover comes with a specific evaluative attitude. Very often co-compounds, and especially synonymic co-compounds, are emphatic. They mark important passages in the narrative structure of the text, where unexpected information is given (unexpected at least from the point of view of one or several actors in the story). In the context before (18) a rumor has circulated that the archangel Michael has sent thunder and lightning over the school because the teacher has put an atheistic play on stage, but Vitya reports that this is all nonsense. In order to emphasize that there is nothing wrong with the teacher, counter to the expectation created by the rumor, he uses the co-compound šumbra=paro ‘healthy=well’ with two contextually synonymous parts. (18) Men’-gak pur’gin’e aras’-el’, učitel’-es’ šumbra=paro we-also thunder NEG.EX-2ndPST, teacher-DEF healthy=good ‘We hadn’t any thunder, the teacher is well’ (Kudaškin 1996: 19)

Erzya Mordvin

In another episode where Vitya confesses to his mother that he made a certain observation, the mother thinks, takes a deep breath and then quietly says: “Don’t tell this to anybody, my son. Keep it to yourself.” Emphasis is given to “think” by framing it as a co-compound: ava-zo ars’e-s’=c’ota-s’ ‘mother-POSS3SG think-PST3SG=calculatePST3SG > his mother thought (for a while)’. The emphatic function is context-dependent and context-renewing. Even if lexicalized uses of co-compounds make up the majority of tokens in the text, many occasional co-compounds can be found. Lexicalization is a cline rather than a discrete distinction, ars’ems=c’otams ‘think-calculate > think’, for instance, has an intermediate status on this cline. It is not listed in the dictionary, but can be regularly encountered in Literary Erzya Mordvin texts. Sometimes it is not really the co-compound in itself that is the lexicalized unit, but rather the co-compound functions as part of an idiom. This holds in particular, but not exclusively, for figurative co-compounds. The figurative co-compound tarka=ez’em ‘place=bench’ which must have originated as a

720

III. Units and processes in word-formation II: Special cases contextually synonymic co-compound in the context of a sitting accommodation, is hardly ever used other than in combination with mujems ‘to find’ in negation. In other texts in Literary Erzya Mordvin it sometimes has the metaphorical use of “not having found yet the right place in life”. This co-compound is not listed in the dictionary however. Kudaškin (1996) contains two rather specific uses of it which demonstrate that the meaning of the idiom is much less fixed than the formal collocation of the co-compound with a lexical verb and negation. In the first passage it is the first day in school for Vitya, but he is so excited about the apples on the apple trees outside that he cannot concentrate on anything in the first lesson. He does not find “place-bench” (tarka=ez’em e-z’ muje ‘place-bench NEG-PST3SG find’) which is meant entirely figuratively. It does not mean that he cannot find a seat. In the second passage the teacher is putting on stage an atheistic play at the end of which an actor puts a horse collar on the neck of an orthodox priest. The mother of the boy who is supposed to act the communist’s role takes her son home saying that it is not good to mock people of God. This is now where the teacher does not find “place-bench”: tarka=ez’em e-z’ mu-kšno ‘place-bench NEG-PST3SG findFREQ’. He is put off his stride for a moment and does not know what to do; until he sees Vitya and so Vitya unexpectedly becomes an artist. What is interesting about this co-compound from a formal point of view is that it is never used with a suffix, there are no plural markers as is characteristic for most otherwise unmarked co-compounds and there is never a possessive suffix. This testifies to the idiomatic character of the cocompound. However, the semantics of the idiom is much less fixed than its form. The only thing that is clearly fixed is that the meaning must be metaphorical; it can mean anything ranging from ‘not being concentrated’ to ‘being put off stride’ to ‘not having found one’s vocation in life yet’. All these uses have a family resemblance, but the actual meaning of the co-compound can only be determined in a given context. Preferred collocations are not restricted to figurative co-compounds, however. The common verbal co-compound lovnoms=s’ormadoms ‘read=write’ hardly ever occurs in any other context than that of learning or being able to read and write. In Kudaškin (1996: 39) it is used to state that Vitya’s father learned to read and write himself without going to school. On the opposite pole of least context-dependent lexical units we have the fully lexicalized example t’et’a-t=ava-t ‘father-PL=mother-PL > parents’ which is not only used for pairs of parents, but even in t’et’a-n’=ava-n’ promks-so ‘father-GEN=mother-GEN meeting-INE > in parents’ meeting’ (Kudaškin 1996: 36). However, some co-compounds of kinship terms are quite contextual. The girl Masha which replaces Mikita as Vitya’s bench neighbor after the event in (15) is very quiet; she goes to school to relax from housework and her eight brothers and sisters: kavkso sazor-tne-d’e=jalaks-tne-d’e ‘eight younger.sister-PL:DEF-ABL=younger.brother-PL:DEF-ABL’. The co-compound tells us not only that Masha is older than all her brothers and sisters, but also that the younger sisters are somehow more prominent than the younger brothers (perhaps there are more sisters than brothers or Masha has to take care more of her sisters than of her brothers). In conventionalized co-compounds of kinship terms in Erzya the order is male-female. Masha’s quiet character has the effect that her name never enters a co-compound together with Vitya, she is not a companion in any of his adventures.

40. Co-compounds

5. Co-compounds and construction morphology Here we will consider to what extent construction morphology (see article 12 on wordformation in construction grammar) might be useful to deal with co-compounds. At least at first glance, construction grammar seems very useful for co-compounds, given the relevance of both semantic properties (natural coordination, superordinate-level concepts) and formal properties (juxtaposition of two words) for co-compounds. Construction morphology allows for holistic properties of word structure which is appealing for co-compounds. We might therefore try to formulate a general schema for co-compounds such as (19): (19) One difficulty is to determine whether the category labels for the parts should all be of the same category or of three different categories . In a majority of cases the word class is the same, thus (19) is at least appropriate for prototypical co-compounds. However, it seems nearly impossible to reduce co-compounds and sub-compounds (cf. example (4) in article 12) to one constructional schema which is well-in-line with Wälchli’s (2005, ch. 4) claim that co-compounds are a construction type of their own and not a sub-type of compounds. A highly problematic aspect of construction morphology, however, is that word-formation processes are viewed as parts of a hierarchical lexicon. This presupposes that word-formation is always a relation between parts and wholes which are types rather than tokens. Above I have argued that many co-compounds are not listed in the lexicon and that the parts of productive co-compounds are rather tokens than types, which means that many co-compounds are not formed in the lexicon. In order to become a useful framework for the description of co-compounds therefore, construction morphology must countenance the composition of tokens in particular given contexts. Booij (in article 12, section 2) points out that in sub-compounds parts “may have lexicalized, yet productive meanings”. This phenomenon is much more restricted with co-compounds. There are some schemas with productive parts. However, it is not clear whether they should be considered true co-compounds. In Erzya Mordvin there are nominal compounds of the type N-t=mezt’ ‘N-PL=what:PL > N and the like’ and verbal compounds of the type V-ms=t’eje-ms ‘V-INF=do-INF > V for a short while’. The former have a function similar to echo words in South Asia and m-doublets in Turkic and contact languages (see Wälchli 2005: 167−169 for a survey and further literature). Mordvin compounds with mez’e ‘what’ share a number of features with co-compounds. Generalizing examples and examples in comitative function have the suffix -n’ek on both parts which is characteristic for generalizing co-compounds in Mordvin: pokš kudo-st sadnek=mez-n’ek ‘big house-POSS3PL garden-COM=what-COM > their big house with garden and everything’ (Kudaškin 1996: 22). Like co-compounds, N-t=mezt’ ‘N-PL=what:PL > N and the like’ typically has a nominative indefinite plural marker on both parts if the form would be in the unmarked nominative indefinite singular otherwise. However, what seems to be peculiar to this type is that there can be wide scope semantically over a noun phrase with an adjective as in (20):

721

722

III. Units and processes in word-formation II: Special cases (20) Raj-ev ki-s’ avol’ paro jarsamka-so=mez’e-se Erzya Mordvin paradise-LAT way-DEF NEG good food-INE=what-INE panžo-v-i open-MIDD-PRS3SG ‘The way to paradise does not open itself with good food and the like’ (Doronin 1995: 52) However, =mez’e ‘=what’ is even used in a different pattern with single marking (unmarked first part) where the first part is a name and the whole has the function of an associative plural: Vit’a=mez’e-n’ šabra-st ‘Vitja=what-GEN neigbor-POSS3PL > the neighbors of Vitja and his people/his family’ (Kudaškin 1996: 7). Whether or not the construction with =mez’e ‘=what’ is a subtype of co-compounds, it consists in fact not of a single schema, but rather of a group of highly similar constructions with family resemblance. This is difficult to account for with a hierarchical schema-centered version of construction morphology. The verbal compounds with =t’ejems ‘=do’ has a function similar to the Russian delimitative prefix po-: jarsa-s’=t’ej-s’ ‘eat-PST3SG=do-PST3SG > he ate for a little while’ (=Russian po-el). What is important for this pattern is that it cannot be abstracted from its context. It is hardly ever used at the end of a sentence; the activity is always interrupted because of a subsequent action that will take place: ‘he ate (a bit) and hurried to school’ (Kudaškin 1996: 39). While the construction might have originated from cocompounds and has the same double marking as verbal co-compounds, it has a quite different function from co-compounds semantically which is why it is probably best considered a construction of its own. However, it seems that this construction is not completely unrelated to co-compounds. The two examples with lexicalized parts discussed above show that the best chance for co-compounds to undergo a development towards affixoids is with a kind of pronominal, semantically empty part that can stand in a natural coordinative relationship to any noun or any verb. A similar, more restricted pattern are alternative co-compounds of the type N=kavto ‘N=two > one or two N’: ije-s=kavto-s ‘year-ILL=two-ILL > for a year or two’ (Kudaškin 1996: 38). The noun denotes most typically a period of time. The construction is actually a variant of the alternative co-compound vejke=kavto ‘one=two’ where the numeral for ‘one’ remains unexpressed. While in sub-compounds the meaning of recurrent compound parts tends to lexicalize in the same way, recurrent parts of co-compounds actually tend to have different meanings in different co-compounds depending on the different kinds of natural coordination they enter in with their other part. However, in texts with a high frequency of cocompounds, in particular in the epic register, there may be some playful variation with parallelism of two co-compounds sharing one part (see also Wälchli 2007b for co-compounds and parallelism). Example (21) contains two synonymic co-compounds in a parallel setting which share one compound member (jalga ‘comrade, friend’), while the second part is varied (oj ‘friend’, duga ‘friend, brother, sister’). Here the syntagmatic aspect of repetition in a text is more important than the paradigmatic aspect of a common schema in the lexicon.

40. Co-compounds (21) Vaj, jalg-in’e-m=oj-in’e-m/ oh friend-DIM-POSS1SG=friend-DIM-POSS1SG/ Vaj, jalg-in’e-m=dug-in’e-m oh friend-DIM-POSS1SG=sibling-DIM-POSS1SG ‘Oh my friend, oh my friend’ (Šaronov 1994: 232)

723 Erzya Mordvin, epic poem

The same poem contains another instance of jalg-in’e-m=oj-in’e-m and a co-compound with inverted order ojam=jalgin’em, this time with the diminutive only on the second part. The example shows that the study of recurrent patterns in texts is as important for construction morphology as the study of recurrent patterns in the lexicon. Co-compounds tend to exhibit sub-patterns in texts at least as much as sub-patterns in the lexicon. A construction morphology approach which is suitable for co-compounds should take into account that certain patterns are characteristic of certain texts rather than of a language as a whole. In Russian there is a distinction in some motion events between ‘(to go/lead) on foot’ and ‘(to go/lead) by vehicle or horse’. In the Bylina Knjaz’ Roman i Mar’ja Jur’jevna (Propp and Putilov 1958: 106−111) Mar’ja Jur’jevna is led away by the Lithuanian Voz’jak Kotobrul’evič. This is expressed with the co-compound u-vez= u-vel ‘away-lead.by.vehicle/horse:PV:PST:SG:M=away-lead.on.foot:PV:PST:SG:M > (he) led (her) away’. It is not specified whether Mar’ja really walks and rides while being kidnapped or whether it is a synonymic co-compound where the subtle semantic difference is a pure ornament in the epic text. However, there is another co-compound following the same pattern in the same text 28 lines later: s-xod-i=s”-jezd-i ‘PREV-go/walk:IPVIMP2SG=PREV-ride/go.by.vehicle:IPV-IMP2SG > go (you to the Lithuanian land)!’ The two co-compounds obviously form a pattern with holistic properties of word structure; they are both co-compounds of motion verbs with a manner alternation (on foot vs. by horse/ vehicle). However, it is difficult to generalize them into a construction grammar schema since none of the formal elements is repeated (there is no part of a compound turned into an affixoid). Furthermore, the order of manner elements is not constant, in the first case it is ‘by.horse=on.foot’, in the second case ‘on.foot=by.horse’. The pattern cannot be said to be a construction of Russian in general, it is a construction in one particular text with two tokens in that text, and it has a clear syntagmatic, and not merely a paradigmatic component. The second co-compound of the pattern is obviously primed by the first one. A very good kind of evidence for constructions is priming where one token increases the probability of another token to occur nearby. Co-compounds co-occurring in the same text close to each other are thus good evidence for syntagmatic construction effects. Ornamental co-compounds, such as Erzya Mordvin vel’e=s’ado ‘village=hundred > village’ where one part is misleading and does not contribute to the meaning of the whole are characteristic only of some registers of Mordvin. They occur mainly in folk poetry and in folktales. Example (22) from a fairy tale contains two ornamental co-compounds: ‘forest=maple > forest’ and ‘to belly=to back > have a belly, be pregnant’. The cooccurrence suggests that they prime each other.

724

III. Units and processes in word-formation II: Special cases (22) Pek-ija-s’=lang-ija-s’ Erzya Mordvin: fairytale belly-DENOM-PST3SG=back-DENOM-PST3SG s’ej-ine-s’, tu-s’ vir’-ev=ukštor-ov goat-DIM-DEF depart-PST3SG forest-LAT=maple-LAT l’evksija-mo. have.young-INF ‘A goat was pregnant, (it) went into the forest to have young.’ (Kemajkina 1993: 24) Due to the interconnection of groups and the family resemblance of co-compounds it is sometimes rather unclear how a subschema of co-compounds should best be delimited. In Wälchli (2005: 153) I discussed scalar co-compounds (C is a scale defined by the poles A and B) only for East and South East Asian languages (e.g., Khalkha Mongolian urt bogino ‘long short > length’) and obviously overlooked the fact that Basque has them as well: eta lodi=mehe-an alde Basque (23) … eta luze=labur-rean and long=short-INE:DEF and thick=thin-INE:DEF side/difference handi-a dute. great-DEF be:PRS3PL ‘([Orb-weaver spiders/Araneidae] They have eight legs) and there are great differences in length and thickness (from one species to another).’ (http://eu.wikipedia.org/wiki/Araneido, 17. 12. 2012) The scalar co-compound luze=labur ‘long=short > length, distance, size’ is lexicalized and frequently mentioned in dictionaries and grammars. In (23) it primes another cocompound of the same kind − lodi-mehe ‘thick=thin > thickness’ which is absent in most dictionaries and has only very few hits on Google. It is, however, not obvious that the only relevant sub-schema is that of scalar co-compounds. Artiagoitia, Ortiz de Urbina and Hualde (cf. article 182 on Basque and section 3.1.2) group luze=labur ‘long=short > length’ together with on=gaitz-ak ‘good=bad-PL > pros and cons’ in a group of [Adj= Adj]N copulative compounds, despite the different form (plural) and the figurative meaning. Actually it is difficult to predict the meaning of the co-compound consisting of on ‘good’ and gaitz ‘bad’, interestingly it does not mean ‘quality’. However, even on= gaitz-ak ‘good=bad-PL > pros and cons’ has some closer relatives among Adj=Adj cocompounds, such as the co-compound in (24): (24) Hor ere argi=ilun handi-ak daude Basque there even light=dark big-PL are ‘Although there are large uncertainties’ (http://goiena.net/albisteak/joxe-rojas-ezin-dugu-itxoin-euskarazkohedabideak-desagertzera-haien-garrantzia-aitortzeko/) Hualde and Ortiz de Urbina (2003: 352−353), however, identify an even larger group of co-compounds with an antonymy relationship where even luze=zabal ‘long-wide > extension’ is included. The dominant use of this co-compound is to denote the extension of two-dimensional surfaces in examples such as Mexiko 1.964.375 km2 da ‘Mexico

40. Co-compounds

725

covers an area of 1.964.375 km2’ (Wikipedia). Thus, luze=zabal ‘long-wide’ denotes two scales rather than one. Knowing that there is a [Adj=Adj]N pattern in Basque is only of limited use to predict the meanings of their exponents and this is not only due to several more specific sub-patterns in the lexicon, but also to idiosyncrasies in concrete use. We may conclude that construction morphology might be highly useful for co-compounds if it can be modified in such a way that it can take concrete patterns of use into account.

Acknowledgements I would like to thank Fernando Zúñiga for many useful comments.

Abbreviations 1 2 2ndPST 3 ABL ABSTR CL CNV COM DAT DEF DEM DENOM DIM EX FIN FREQ FUT GEN HON ILL IMI IMP INE

1st person 2nd person 2nd past 3rd person ablative abstract classifier converb comitative dative definite, definite article demonstrative denominal verb diminutive existental finite verb form frequentative future genitive honorific illative imitative part of imitative co-compound imperative inessive

INF INST IPV LAT LOC M MIDD N NEG NFIN NOM PASS PL POSS PREP PREV PROL PRS PST PV Q RFL SUBJ

X>X

infinitive instrumental imperfective lative locative masculine middle neuter negation non-finite nominative passive plural possessive affix omni-purpose preposition verbal prefix prolative present past perfective polar interogative reflexive subjunctive subject form > object form

6. References Arcodia, Giorgio F., Nicola Grandi and Bernhard Wälchli 2010 Coordination in compounding. In: Sergio Scalise and Irene Vogel (eds.), Cross-Disciplinary Issues in Compounding, 177−197. Amsterdam/Philadelphia: Benjamins.

726

III. Units and processes in word-formation II: Special cases Atkinson, J. Maxwell and John Heritage (eds.) 1984 Structures of Social Action. Studies in Conversation Analysis. Cambridge: Cambridge University Press. Bauer, Laurie 2009 Dvandva. Word Structure 1: 1−20. Bisang, Walter 1988 Hmong Texte. Eine Auswahl mit Interlinearübersetzung aus Jean Mottin, Contes et légendes hmong blanc (Bangkok: Don Bosco 1980). Zürich: Seminar für Allgemeine Sprachwissenschaft der Universität Zürich. Browning, Robert 1983 Medieval and Modern Greek. London: Hutchinson University Library. Debrunner, Albert 1917 Griechische Wortbildungslehre. Heidelberg: Winter. Doronin, Aleksandr 1995 Bajagan’ suľejť: Omboc’e kn’igas. S’atko 1995(7): 3−63. Fəjsi, Əxmət 1993 Keçkenə apuş. Kazan: Kazan tatarstan kitap nəşprijaty. Füst, Milan 2000 A feleségem története. Störr kapitány feljegyzései. Budapest: Fekete Sas Kiadó. Haiman, John 1985 Natural Syntax. Cambridge: Cambridge University Press. Hewitt, George 2005 Abkhazian Folktales (with grammatical introduction, translation, notes, and vocabulary). München: LINCOM Europa. Hualde, José I. and Jon Ortiz de Urbina (eds.) 2003 A Grammar of Basque. Berlin: Mouton de Gruyter. Inkelas, Sharon and Cheryl Zoll 2005 Reduplication. Doubling in Morphology. Cambridge: Cambridge University Press. Kemajkina, Raisa S. (ed.) 1993 Jovkson’ kužo. Erz’an’ jovkst. Saransk: Mordovskoj kn’ižnoj izdaťeľstvas’. Kiparsky, Valentin 1967 Russische historische Grammatik. Vol. 2. Heidelberg: Winter. Klima, Edward S. and Ursula Bellugi 1979. The Signs of Language. Cambridge, MA: Harvard University Press. Kowalski, Tadeusz 1929 Karaimische Texte im Dialekt von Troki. Kraków: Nakładem Polskiej Akademji Umiejętności. Kudaškin, Vasilij 1996 Viťa Kiľďaz’even’ er’amozo: Vas’enc’e peľks. S’atko 1996(3): 3−51. Lambrecht, Knud 1984 Formulaicity, frame semantics and pragmatics in German binomial expressions. Language 60(4): 753−796. Lehmann, Walter and Gerdt Kutscher 1949 Sterbende Götter und christliche Heilsbotschaft. Wechselreden indianischer Vornehmer und spanischer Glaubensapostel in Mexiko 1524. Stuttgart: Kohlhammer. Mühlhäusler, Peter 1979 Growth and Structure of the Lexicon of New Guinea. Pidgin Camberra: Australian National University. Paasonen, Heikki 1949 Gebräuche und Volksdichtung der Tschuwassen. Ed. by Eino Karahka and Martti Räsänen. Helsinki: Suomalais-ugrilaisen seura.

41. Multi-word units in French

727

Propp, Vladimir Jakovlevič and Boris Nikolaevič Putilov 1958 Byliny. Vol. 2. Moskva: Gosudarstvennoe izdatel’stvo xudožestvennoj literatury. Rushdie, Salman 1981 Midnight’s Children. London: Vintage. Šaronov, Aleksandr Markovič 1994 Mastorava. Saransk: Mordovijan’ kn’igan’ izdaťeľstvas’. Šketan, M. 1991 Čumyren lukmo ojporo. Kokymšo tom. Ojlymaš, myskara, novella, očerk, staťja, korrespondencij, feľeton-vlak. Joškar-Ola: Marij kn’iga izdateľstvo. Timušev, Dmitrij A. (ed.) 1971 Obrazcy komi-zyrjanskoj reči. Syktyvkar: Akademija Nauk SSSR. Tkačenko, Orest B. 1979 Sopostavitel’no-istoričeskaja frazeologija slavjanskich i finno-ugrorskich jazykov. Kiev: Naukova dumka. Uotila, Eeva 1980 Asyndeton in the baltic and finnic languages: An archaic construction in its typological periphery. Journal of Baltic Studies 11(1): 86−91. Wälchli, Bernhard 2005 Co-Compounds and Natural Coordination. Oxford: Oxford University Press. Wälchli, Bernhard 2007a Lexical classes: A functional approach to “word-formation”. In: Matti Miestamo and Bernhard Wälchli (eds.), New Challenges in Typology. Broadening the Horizons and Redefining the Foundations, 153−175. Berlin: Mouton de Gruyter. Wälchli, Bernhard 2007b Ko-Komposita (im Vergleich mit Parallelismus und Reduplikation). In: Andreas Ammann and Aina Urdze (eds.), Wiederholung, Parallelismus, Reduplikation. Strategien der multiplen Strukturanwendung, 81−107. Bochum: Brockmeyer.

Bernhard Wälchli, Stockholm (Sweden)

41. Multi-word units in French 1. 2. 3. 4. 5.

Introduction Functions of multi-word units and domains of description Classes of multi-word units Multi-word sequences at the crossroads of linguistic preoccupations References

Abstract After pointing out the emergence of the concept of multi-word units in French linguistics, we focus on the principal functions assigned by language to multi-word sequences (referentiation, predication, “argumentalisation”, actualisation, etc.) and on the domains of description. Then we provide a sample of the descriptions devoted to these units classified by parts of speech.

728

III. Units and processes in word-formation II: Special cases

1. Introduction It may be useful to start with the following terminological remark: the concept of multiword units (polylexicalité) is recent in French. Among the first attested uses of the term are those that appear in the works of Gréciano (1983) concerning idiomatic expressions in German. Since then, the term has been used in its adjectival (polylexical) as well as in its nominal form (polylexicality), always with regard to fixedness. Twenty years later, the nominal form appears − already − in the title of the fifth volume of the journal Syntax and Semantics, “Polysémie et polylexicalité” (Mejri 2003). One can currently consider it as an operational concept in French linguistics applying to the description of fixed sequences. Its entry in the Dictionnaire des sciences du langage by Neveu (2004) is proof of its validity. It refers to the plural character of the signifier (several words) of fixed sequences. It is opposed to single-word units of simple monomorphemic or polymorphemic lexemes. This concept is also used to complete the table of the configurations of the word in French: (1)

a. single-word units − monomorphemic (simple words formed by a single morpheme): sac ‘bag’. − polymorphemic (words constructed with at least two non-autonomous morphemes): anti-constitut-ion(n)-elle-ment ‘anticonstitutionnally’. b. multi-word units − units formed by several words, mono- or polymorphemic: rendre l’âme to give back the soul ‘to give up the ghost’.

This new classification in effect replaces others that have ignored multi-word units. The doxa in the field of lexical morphology, still in use in the overwhelming majority of dictionaries and grammar handbooks, is only interested in polymorphemic constructions (non-autonomous morphemes), either affixes (prefixes, infixes, suffixes, confixes) or lexical bases that participate in the formation of several affixed words (lav-age ‘washing’, lav-eur ‘washer’, lav-ure ‘washings’). The single concession made to multi-words is what is commonly referred to as French composition. This concerns essentially nominal sequences (noun compounds) − all other parts of speech not being concerned − with an elliptic structure, put differently compounds missing one element as in the following example: repose-pieds rest-feet ‘footrest’, where the element le(s) ‘the’ is deleted. It is this ungrammaticality that is used as a criterion of distinction between noun compounds and fixed sequences, a criterion introduced by Darmesteter at the end of the 19th century in order to contrast composition and juxtaposition, the latter being “the more or less intimate welding of elements joined together without ellipsis, simply united one besides the other according to ordinary rules of syntax” (1964 [1887]: 72). It is only after the second half of the 20th century that an interest in phrasal formations was shown, within the context of works carried out on the basis of corpora, like those of Dubois (1962) and of Guilbert (1965, 1967). It is especially for specialised subject fields, like those of aviation and astronautics, that Guilbert introduced the category of phrasal formations (formations syntagmatiques). The study of the constitution of lexical fields in specific domains enabled him to isolate a very significant number of multi-word units. This would lead him to integrate what he calls syntagmatic derivation into his famous

41. Multi-word units in French introduction to the Grand Larousse de la Langue Française (Guilbert 1975). In spite of this significant advance, his description remains confined to the noun. As he is interested in specialised discourse, it is natural that the referential dimension, closely related to nouns, orients the description mainly towards the noun and only secondarily to the other parts of speech. Even if his perspective is lexicological, he is in agreement on this point with the terminological work that shows very clearly the importance of multi-word units. If the terminological work has not really taken the step of proposing descriptive tools of a linguistic nature, this is because it privileges terminological aspects such as the adequacy of the term for the concept or conceptual analysis. Thus the study of multi-word units that cover the entire spectrum of parts of speech is excluded from the scope of this approach. It is thanks to lexicographic and data-processing studies that there has been a significant advance in the domain of multi-word units: the focus has shifted from specialised languages (LSP) to the general language and from the category of nouns to all other categories. Rey is among the lexicographers who best grasped the importance of fixedness in the structuring of the French lexicon (Mejri 2010). Indeed the Dictionnaire des expressions et locutions is the work of Rey and Chantreau (1989). It was thinking over this phenomenon which led Rey to propose lexicographical descriptions that he judges adequate to represent the phenomenon in dictionaries. Maurice Gross (1982) and after him Gaston Gross (1996), finding themselves constrained by the requirements of data processing, were among those that systematised the description of multi-word units in French, in particular their syntactic dimension. For a general discussion of the topic of the present article, see also article 24 on multi-word expressions.

2. Functions of multi-word units and domains of description Two angles of attack can be used in order to give an account of the nature of work regarding multi-word units: the functions that are assigned by language to these units and the centres of interest on which linguists have focused.

2.1. Functions of multi-word units Multi-word sequences have continued to suffer from the marginalisation that has long held back the study of fixedness. Even after having rehabilitated multi-word units as an object of study worthy of interest, linguists at first did not integrate multi-word sequences into the general description of the linguistic system. It is only rather belatedly that much of the research has been directed toward demonstrating that multi-word units share with single-word units the same functions, such as referentiation, argumentalisation, predication, grammaticalisation, “pragmatisation”, utterance structuring and stylistic marking (rhetoric dimension), etc. Referentiation is the only function that really held the attention of linguists. As multi-word units have a plural signifier, they have always been used as a common means of designation for minerals, instruments and flora. This has given rise

729

730

III. Units and processes in word-formation II: Special cases to a series of common names which exist alongside the learned ones: tête-de-chat head of cat ‘pieces of rocks’, œil-de-chat eye of cat ‘variety of quartz’, etc. What has interested linguists most in referentiation is primarily the link established between the new terms and the domains of the lexicon from which material is drawn for the creation of a multi-word unit, as all quoted examples illustrate that use parts of the body and animals. Behind this motivation one can raise questions about the mechanisms of abstraction, conceptualisation and categorisation (Mejri 2005 and Petit 2008). These questions find a great deal of material in multi-word units to exploit in order to better grasp not only the operations by which complex designations are constructed but also the patterns (nominal, verbal, etc). Take for instance the French verbal sequence il a un poil dans la main lit. ‘he has a hair in his hand’ used for naming the character of someone who is very lazy. Such a designation reveals the stereotypes shared by the members of the linguistic community and anchored in the lexicon. In this way linguists have gradually allowed multi-word units to fulfil the three primary functions that single-word units do, namely, playing the role of predicates (action of predicating, predication), of predicate arguments (“argumentalisation”, relative to argument) and actualisers (all the elements necessary to ensure the grammatical character of an utterance: time, aspect, gender, number, person …). A sentence as common as La plupart des directeurs adjoints ont fait l’objet d’une contre-enquête avant-hier ‘Most deputy directors were the object of a counter-enquiry the day before yesterday’ nicely illustrates both the omnipresence of multi-word sequences and the coverage they ensure at the level of the three primary functions. It can be analysed as follows: (2)

a. nominal predicate contre-enquête ‘counter-enquiry’, b. argument directeurs adjoints ‘deputy directors’, c. actualisers α. of the predicate: ont fait l’objet de ‘were the object of’ (light verb expressing both perfect and passive), avant-hier ‘day before yesterday’ (adverb expressing the tense: past), β. of the argument: la plupart de ‘most of’ (complex degree modifier/proportional modifier expressing the category of number and of indefiniteness).

As we can see, all components of this sentence, except for the indefinite une ‘a’ which actualizes contre-enquête, are multi-word units. One can continue the exercise and add for example an agent: La plupart des directeurs adjoints ont fait l’objet d’une contreenquête de la part de la cour des comptes ‘Most deputy directors were the object of a counter-enquiry on behalf of the Court of Auditors’. This sentence contains cour des comptes ‘Court of Auditors’ as an additional argument and de la part de ‘on behalf of’, an agentive argument specific to the passive construction. While the functions of referentiation and predication, with what they entail, such as “argumentalisation” and actualisation, are already part of the landscape of research on

41. Multi-word units in French fixedness, the remaining three functions, “pragmatisation”, utterance structuring and stylistic marking fall into the category of exploratory studies. Thanks to the work of Mel’čuk (1998), Polguère (2008) and Blanco (2010), the pragmatic dimension of certain utterances has been admitted as a criterion of analysis. Thus, any sequence implying usage constraints in its semantism would be considered as a pragmateme. The notion concerns lexical units of compositional meaning whose usage is conditioned by considerations of a pragmatic nature. The example that is most usually cited is that of Stationnement interdit parking prohibited ‘No parking’, but as it turns out the extension of the domain of pragmatemes is vast. It covers inter alia all the formulaic utterances that are very frequent in verbal exchanges and whose enunciative dimension generally escapes non-native speakers. Greeting formulas as simple as bonjour ‘good morning’, bonsoir ‘good evening’, and bonne nuit ‘good night’ are likely to pose a problem for non-natives whose mother tongue does not differentiate on the level of constraints relative to the time during which one can use them, namely “to greet somebody one meets during the day” for bonjour, “to greet somebody one meets or leaves in the evening” and “to greet somebody one leaves in the evening before going to sleep” for bonsoir and bonne nuit respectively. The pragmatic elements focused on here are greeting somebody (situation of interlocution and performative dimension) which is common to all three, time of the day, also common to the three sequences, with an opposition between the day for bonjour and the evening for bonsoir and bonne nuit, opposition between meeting, shared by bonjour and bonsoir, and leaving, used for bonsoir and bonne nuit, this last sequence being also distinguished by the time before sleeping. Other types of syntactically and pragmatically autonomous sequences are to be considered as pragmatemes: they are either specific to certain languages, as dou’a in Arabic (formula of injunction, wish, curse, etc.; cf. Mejri forthc.), or shared by a very large number of languages, like formulas of popular wisdom. Proverbs, dictums and sayings have been recently considered as multi-word units that have the double particularity of being of sentential nature, characteristic shared with fixed sequences not relating to popular wisdom such as Un ange passe (an angel passes ‘there is an awkward silence’), and of having a predicative value, a characteristic common with phrasal pragmatemes of the type marché conclu! deal finalized ‘it’s a deal’, non-sententious phrases and sententious discursive sequences such as maxims and apothegms (Kleiber 2000); a proverb like Quand on parle du loup, on en voit la queue (when you speak of the wolf, you can see its tail ‘Speak of the devil’), well illustrates the double predicative and pragmatic dimension. It is a kind of formula, considered by Rey and Chantreau (1989) as a dictum, that predicates “speaking of somebody”, and that is used only when they arrive at the moment that someone is talking about them. Multi-word sequences also ensure sentence and discourse structuring. Most prepositional and conjunctive phrases are used as second-order predicates in non-elementary sentences (cf. Gross and Prandi 2004). As the possibilities of reformulation go beyond the borders of the sentence, these same units ensure an intersentential bond. In (3a), thanks to is a predicate of second order used in a complex sentence (with several predicates); in (3b), moreover, the same predicate ensures, the bond between the first sentence and the second: (3)

a. Pierre a bien réussi à l’examen grâce à sa persévérance. ‘Pierre succeeded well in the exam thanks to his perseverance.’

731

732

III. Units and processes in word-formation II: Special cases b. Pierre a de la persévérance (est persévérant). Grâce à quoi il a bien réussi. ‘Pierre has perseverance (is persevering). Thanks to which he succeeded well.’ At the level of intersentential relations multi-word units also occupy a place of choice concerning the structuring of discourse; this includes all the connectors of the type d’abord ‘initially’, to which one can add all multi-word sequences playing the role of anaphora, including proverbial utterances. To finish with the functions assigned by language to multi-word units, we must mention an area which has been studied very little, but which has begun to interest linguists, that of stylistic marking, which can find its expression inter alia in the recognition of specialised discourses and in the stylistic characteristics of authors. It has been demonstrated, thanks to the frequency of complex denominations in specialised discourses, that the frequency of multi-word units is so great that it can be used as a reliable indication for automatic domain text classification. Other studies have managed to demonstrate how certain authors use multi-word units as stylistic markers. Balzac for example commonly resorts to the de-fixing or diversion of proverbs (Navarro Dominguez 2000). It is a very promising field in which linguists are investing more and more effort.

2.2. Domains of description This stylistic dimension is for the moment rather marginal compared to the linguistic domains that make multi-words a privileged object of study, namely syntax and semantics. Multi-word units have primarily been studied from a syntactic point of view. Faced with the rather widespread idea that multi-word units are supposed to be completely fixed, linguists were initially interested in verifying this working hypothesis. The abundance of works in keeping with this position have made it possible to posit the following characteristics: (4)

a. with the exception of certain productive syntactic patterns especially for noun compounds and archaic constructions (of the type advienne que pourra ‘come what may’), all multi-word units are grammatically well formed; b. multi-word sequences that manifest absolute fixedness, i.e. do not accept any variation of whatever nature, are by no means common: although precise numbers are not available, their volume is estimated at around 10 %; c. the majority of fixed sequences show variations which place them in a continuum that reaches from the most constrained to the least constrained; d. these variations can correspond to the transformations allowed by the internal combinatory of units (cf. section 3) or to the existence of variants of the same sequence as is the case of emporter (enlever) le morceau to carry away the piece ‘to carry the day’.

The challenge is obviously the measurement of the degree of fixedness of multi-word sequences according to the number of variations (transformations) allowed by each se-

41. Multi-word units in French quence. This can only be done by relying on the syntax that is specific to the internal structure of the sequence (cf. section 3). On the semantic level, the following topics need to be investigated: (5)

a. the compositionality/non-compositionality of the meaning of polylexical units: multi-word units are, in general, ambiguous between a compositional and noncompositional meaning. There is a tendency, in particular for the non-native speaker, to understand the expression in its compositional (literal) meaning. This has prompted linguists to contrast the compositional meaning of multiword expressions with their non-compositional meaning. In the first case, although there is a global meaning, one is able to make sense of the components of the sequence. Consider the following examples: Examiner quelque chose à la loupe ‘to examine something with the magnifying glass’. Although figurative, this expression does not pose particular problems of interpretation as long as the implications of the image are obvious. Mettre quelque chose en sûreté ‘to put something in safety’ is even more transparent. Consider now, however, en boucher un coin lit. ‘to plug a corner’: nothing in this sequence suggests the meaning ‘fill with wonder’; b. semantic transparency/opacity: A corollary of the first opposition, this dichotomy focuses on the decoding of sequences. Opacity, like non-compositionality, is of scalar nature: the more opacifying mechanisms are used in the sequence, the greater the opacity is. In the sequence, (faire, jouer) la mouche du coche (to do, to play) the fly of the coach ‘backseat driver’, two mechanisms come into play, figuration (metaphor) and exocentricity (nothing in that sequence indicates that it refers to a person). This is the reason why the meaning of this expression needs to be memorized as it is, “a person who is overactive while claiming to give invaluable help” (Rey and Chantreau 1989). On the other hand, one can find examples of transparent sequences like Beaucoup de bruit pour rien A lot of noise for nothing ‘Much ado about nothing’; c. the role of context in “removing” opacity (Gross and Massoussi 2011): Since the meaning of a unit can only be determined within the sentence frame, taking context into account can de-opacify such sequences. Speaking about tête de Maure (Moorish head) out of a sentence context does not help to access the meaning of the sequence. But to say on nous a servi entre le plat principal et le dessert une tête de Maure (‘we were served a Moorish head between the main dish and the desert’) makes it possible to infer that tête de Maure is something edible, and that it probably refers to a variety of cheese, considering the order in which it is served during a meal; d. space precludes the explicit demonstration of a number of other relevant semantic aspects, as for instance stereotypy, semantic unfolding, referentiation, semantic globalisation, conceptualisation, etc.

733

734

III. Units and processes in word-formation II: Special cases

3. Classes of multi-word units Let us now present in summary an inventory of multi-word units classified according to their surface pattern and more precisely according to the parts of speech to which they belong. Emphasis will be placed on the special characteristics of such patterns discussed in previous research.

3.1. Nominal sequences Nominal sequences are an extensively discussed topic in the linguistic literature on wordformation due to the privileged status that grammar accords them. We have already mentioned the distinction made by Darmesteter between elliptic (compounds according to grammatical tradition) and juxtaposed formations (fixed sequences). This distinction has been accepted in current French textbooks for schools. By recognizing fixedness as a valid domain of study, linguistic descriptions relocate nominal sequences from the domain of composition into their own domain of study. However, besides the degree of fixedness, mainstream research also addresses other topics such as the following: (6)

a. the issue of spelling: The orthographical variation of the components of multiword sequences is relatively random. This is what explains the vivid interest of linguists, and particularly of lexicographers, in this aspect. Natural language processing has made this question a priority for the development of lemmatisers (cf. Mathieu-Colas 1996); b. the internal structure of sequences: Where traditional descriptions limit the number of patterns of nominal compounds to only a few types, recent studies on fixedness extend the notion of nominal compound to include a larger number of syntactic patterns. G. Gross (1996: 48) notes on this subject: “The most explicit typology by far is that established by Mathieu-Colas (1996). It comprises more than 700 types and gives a good idea of the extent of the phenomenon.” In his typology Mathieu-Colas includes 17 elementary classes (abbreviations, VN, Adj N, NN, N de N, N à N, phrases, etc.) to which he adds eight classes of complex compounds (bon vieux temps ‘good old days’, offre publique d’achat offer public of buying ‘takeover bid’); c. the role of nouns in forming multi-word sequences pertaining to other parts of speech such as a) complex determiners as for instance un nuage de (lait) a cloud of (milk) ‘a small quantity of milk’, of which there are more than three thousand, b) adjectivals (à l’abandon to the abandonment ‘abandoned’), adverbials (la bouche ouverte the mouth open ‘with one’s mouth open’), etc.; d. the different primary functions assumed by nominal sequences: They can be predicates: x a lancé une offre publique d’achat sur les actions de la société y, ‘x launched a takeover bid for the stocks of company y’, arguments: Le bon vieux temps est derrière nous, ‘The good old days are behind us’,

41. Multi-word units in French actualisers: x a pris une kyrielle de décisions. ‘X made a string of decisions’.

3.2. Verbal sequences The most frequent types of multi-word expressions quoted in works on fixedness are verbal sequences. This can be explained by the large number of manipulations that they are subject to. As they only lack a subject in order to constitute fixed sentences, they are the foundation of transformations entailing sentence structure. If the two following sentences are compared: (7)

a. Tous les chemins mènent à Rome ‘All roads lead to Rome’ b. Pierre n’y va pas par quatre chemins, ‘Pierre gets straight to the point’,

it is clear that fixedness in (7a) covers the totality of the sentence whereas it excludes the subject in (7b). In the position of the subject, paradigmatic substitution is free: Pierre (the neighbor, this person in charge …) n’y va pas par quatre chemins (lit. ‘does not go there on four paths’). The most common specific transformations used as tests to measure the fixedness of such sequences are: (8)

a. passivisation: Paul a perdu les pédales Paul lost the pedals ‘Paul got all mixed up’ vs. *Les pédales ont été perdues The pedals were lost, but Il a vite pris le pli / Le plis a été vite pris He quickly took the fold / The fold was quickly taken ‘He quickly got the habit’, b. extraction of the nominal element: Paul cherche à noyer le poisson Paul seeks to drown the fish ‘Paul seeks to cloud the truth’ vs. *C’est le poisson que Paul cherche à noyer, *It is the fish that Paul seeks to drown, c. topicalisation of the nominal element: Paul a pris la mouche Paul took the fly ‘Paul flew off the handle’ vs. *La mouche, Paul l’a prise The fly, Paul took it, d. relativisation: Paul a eu froid Paul was cold vs. *le froid qu’il a eu the cold that he had, e. the deletion of the verb: Paul a eu froid Paul was cold vs. *le froid de Paul the cold of Paul, Paul a faim Paul is hungry / la faim de Paul the hunger of Paul (in the second case, it is a light verb actualising a predicative noun), f.

interrogation: Paul a pris la tangente Paul took the tangent ‘Paul made himself scarce’ vs. *qu’est-ce que Paul a pris? what did Paul take?.

These sequences do not all have the same degree of fixedness. This is why some admit certain transformations and reject others. But compared to other sequences, verbal se-

735

736

III. Units and processes in word-formation II: Special cases quences are never completely fixed since they comprise the verbal element that must necessarily be actualised. However this actualisation is not completely free. The use of tense is constrained by the meaning of the sequence. In the sentence Paul mange les pissenlits par la racine Paul eats the dandelions by the root ‘Paul is pushing up daisies’, it is not possible to accept the periphrastic past tense of the verb manger ‘to eat’. As a consequence of its meaning ‘to be dead and buried’, the sequence has already the perfective aspect; hence, adding the periphrastic form would be redundant. All these characteristics do not relate uniquely to verbal sequences in Standard French. They also relate to other varieties of French (cf. for example Lamiroy 2010).

3.3. Adjectival sequences These represent the least studied sequences. Tradition has limited them to very restricted paradigms such as that of color adjectives of the type bleu clair blue light ‘light blue’. But the studies of fixedness show that they are as important as the other sequences. Due to the lack of the affixal morphology typical of adjectives, multi-word units like à la mode ‘in fashion’ are not normally considered as adjectives. However, this sequence behaves like a simple adjective, for example élégant ‘elegant’. It can function: (9)

a. as a nominal predicate Paul est à la mode/élégant ‘Paul is in fashion/elegant’, b. as an apposition of the name to which it serves as a support Paul, à la mode/élégant, force l’admiration de ses camarades ‘Paul, in fashion/elegant, forces the admiration of his friends’, c. be an epithet Un garçon à la mode/élégant. ‘An elegant boy/a boy in fashion’, and d. can be referred to by the pronoun le Paul est à la mode/élégant, Léa l’est également. ‘Paul is in fashion/elegant, Lea too’.

There are many internal structures of these adjectives. M. Gross (1988) and G. Gross (1996) provide an exhaustive list of them. Suffice it to remember that, even if prepositional phrases are the most frequent (à l’état liquide ‘in liquid-state’; près du but ‘close to the goal’; en noir et blanc ‘in black and white’; entre la vie et la mort ‘between life and death’; sous le choc ‘under the shock’; dans son élément ‘in his element’), other structures are common: Adj N: fait maison made home ‘homemade’. Certain sequences have the particularity of being ambivalent; they function either as adjectives or as adverbs according to their syntactic environment: (10) a. On avait eu un service à la carte (adj.) ‘We had had a service a la carte’

41. Multi-word units in French b. On était servi à la carte (adv.) ‘We were served a la carte’

3.4. Adverbial sequences We refer to previous studies of Guimier (1996) for the characteristics of adverbs − especially those ending in -ment −, linked diachronically to multi-words, and those of M. Gross (1986) and G. Gross (1996) for multi-word adverbs. We have already mentioned that the most productive structures are those of the pattern Prep (Det) N and that adverbs can play a role both in the structuring of sentences and in intersentential sequences. As we have shown previously, their construction can bring them closer to adjectival usages. But they can also be grouped together with prepositional or conjunctive phrases as connectors, what we call connecting adverbs: (11) a. sentential usage Il parle à tort et à travers ‘He talks a lot of nonsense’ b. intersentential usage […] ses doigts étaient tourmentés d’un tremblement fébrile; en outre, il transpirait abondamment des paumes […] ‘his fingers were tormented of a nervous trembling; moreover, he sweated abundantly from his palms’. (Duhamel, cited by the Grand Robert 2001) Certain multi-word adverbs can appear in the form of discontinuous units, thus making it possible to structure entire passages: (12) a. d’un côté… de l’autre (côté) ‘on the one hand’ … ‘on the other (hand)’ b. d’abord…ensuite…enfin ‘initially’… ‘then’… ‘finally’ c. certes…mais ‘certainly’… ‘but’, etc. (cf. inter alia Siepmann 2008)

3.5. Prepositional and conjunctive sequences Relegated to the rank of simple markers of syntactic functions, in particularly circumstantial complements, these sequences have, thanks to the works of Gross and Prandi (2004), seen their status change completely. These authors were able to demonstrate that a prepositional sequence as à cause de ‘because of’ is on the contrary the principal element in a sentence like:

737

738

III. Units and processes in word-formation II: Special cases (13) Paul n’a pas pu monter son projet à cause de ses difficultés avec les concurrents. ‘Paul could not put together his project because of his difficulties with the competitors.’ The predicative analysis shows that there are three predicates in this sentence: (14) Paul n’a pu monter son projet (B) ‘Paul could not put together his project’, Paul a des difficultés avec les concurrents (A) ‘Paul has difficulties with the competitors’, Ceci (A) a causé cela (B) ‘This caused that’, and that the first two predicates are actually the arguments of the prepositional phrase à cause de. This is why it is considered as a predicate of second order. Two points are to be kept in mind on this subject: some linguists do not consider such sequences as fixed because they admit several variations (dans le but de in the goal/target of ‘with the aim to’, dans ce but ‘to this end’, dans le seul but ‘in the only goal’, etc.), but they omit that fixedness here is of a functional nature, i.e. the grammatical function assigned to it by language. The rest is a question of degree of fixedness: if we compare à cause de ‘because of’ and grâce à ‘thanks to’, we notice that the second is more fixed than the first. The second point, related to the first, relates to grammaticalisation: all the works agree on the fact that fixedness is the only multi-word process that equips languages with grammatical tools.

3.6. Determinative sequences Buvet (2012) describes predeterminers in French, specifying that their number reaches 4,000 of which 3,000 have a nominal core (for example, un tas de livres ‘a lot (lit. heap) of books’), and 2,000 determinative fixed sequences that imply a predeterminer and a modifier (for example, une mémoire d’éléphant ‘a memory of an elephant’). As actualisers of the noun, he studies them according to whether they determine predicates (Il a pris une kyrielle de décisions ‘He made a string of decisions’) or arguments (Il a lu un tas de livres ‘He read a lot of books’).

3.7. Multi-word interjections While the subject of interjections has not yet been addressed by many researchers, analysing them as predicates could indicate some possible avenues for future research: as “phrasoïdes” (synthetic phrases), they could felicitously accept a predicative analysis. To be convinced it is enough to compare the long form to the truncated one: Va au diable ! ‘Go to hell!’, Au diable ! ‘To hell!’. From the very strong links that they have

41. Multi-word units in French to the situation of utterance, they generally function as pragmatemes whose significance is firmly anchored in verbal exchange.

3.8. Other types of unit Until now no special reference has been made to multi-word pronouns, which appear in language in small numbers since they belong to what is traditionally called “closed paradigms”. It should be kept in mind however that besides multi-word pronouns as demonstratives (ceux-ci/là, celles-ci/à … ‘this/that one’) or relatives (à laquelle ‘to which’), grammars classify a whole series of multi-word pronouns as indefinites: l’un ‘the one’, quelqu’un ‘somebody’, etc. (cf. Riegel, Pellat and Rioul 2009 [1994]). Descriptive multi-word sequences could also be included in the list of multi-word sequences. Oddly enough, they can have all kinds of functions (titles of paintings, names of bars, restaurants, etc.) as well as a wide variety of structures (cf. Bosredon 1998). Finally hybrid sequences which do not belong to a part of speech nor are sentences, as for instance La balle est dans le camp de X ‘the ball is in X’s court’, could also be mentioned here.

4. Multi-word sequences at the crossroads of linguistic preoccupations The centrality of fixedness and its massive presence both in language and in discourse raises a number of questions relating to: (15) a. the theoretical dimension regarding the status of multi-word units: Are they words? If not, what status should be granted to them? b. the automatic recognition imposed by the development of corpus research: How are multi-word units to be dissociated from the corresponding syntactic units with the same signifier? c. the construction of a lemmatiser that takes into account all the degrees of fixedness and the multiple variations of the sequences imposed consequently in discourse; d. the lexicographic description of multi-word units imposed by the great vacuum left by traditional lexicographic practices: Is it necessary to compile dictionaries specific to multi-word units? If so, what needs to be taken into account? Syntax? Semantics? Variations in discourse? Stereotypes? Inferences?, etc. e. the combinatory profile (Blumenthal 2008) of the single-word units as the basis of the most recurring collocations: Should we see this phenomenon, if it is systematic, as the beginnings of fixedness which leads, in the long run, to the constitution of new multi-word units?

739

740

III. Units and processes in word-formation II: Special cases All the issues raised so far will probably suggest new approaches for future study and lead to a serious questioning of the morphological, syntactic, semantic, lexicological or pragmatic approaches which have been formulated up to now!

Acknowledgements I would like to thank Dimitra Alexandridou, Zoe Gavriilidou and John Humbley for the English version of this paper.

5. References Blanco, Xavier 2010 Traduction des pragmatèmes dans les guides de conversation en russe: Contenus conceptuels et enjeux culturels. Synergies [Tunisie] 2: 75−84. Blumenthal, Peter 2008 Combinatoire des mots: Analyses contrastives (français/allemand). In: Peter Blumenthal and Salah Mejri (eds.), Les séquences figées. Entre langues et discours, 173−194. Stuttgart: Steiner. Bosredon, Bernard 1998 Les signalétiques de nomination ou quand le discours se fige. In: Le figement lexical. 1ères Rencontres linguistiques méditerranéennes, 209−218. Tunis: RLM, CERES. Buvet, Pierre-André 2012 La dimension lexicale de la détermination du français. Paris: Champion. Darmesteter, Arsène 1964 [1887] La vie des mots étudiée dans leur signification. Paris: Delagrave. Dictionnaire Larousse Illustré 1993 Paris: Larousse. Dubois, Jean 2001 Le vocabulaire politique et social en France de 1869 à 1872. Paris: Larousse. Grand Robert 1962 Paris: Le Robert. Gréciano, Gertrud 1983 Signification et dénotation en allemand. La sémantique des expressions idiomatiques. Metz: Centre d’analyse syntaxique, IX, Université de Metz. Gross, Gaston 1996 Les expressions figées en français. Paris: Ophrys. Gross, Gaston and Taoufik Massoussi 2011 Figement et transparence. In: Jean-Claude Anscombre and Salah Mejri (eds.), Le figement linguistique. La parole entravée, 95−108. Paris: Champion. Gross, Gaston and Michele Prandi 2004 La finalité. Fondements conceptuels et genèse linguistique. Bruxelles/Louvain La Neuve: De Boeck/Duculot. Gross, Maurice 1982 Une classification des phrases figées du français. Montréal: Université du Québec à Montréal. Gross, Maurice 1986 Grammaire transformationnelle du français. Vol. 3: Syntaxe de l’adverbe. Paris: ASSTRIL.

41. Multi-word units in French

741

Gross, Maurice 1988 Les limites de la phrase figée. Langages 90: 7−22. Guilbert, Louis 1965 La formation du vocabulaire de l’aviation. 2 vol. Paris: Larousse. Guilbert, Louis 1967 Le vocabulaire de l’astronautique. Paris: Larousse. Guilbert, Louis 1975 La créativité lexicale. Paris: Larousse. Guimier, Claude 1996 Les adverbes du français. Le cas des adverbes en -ment. Paris: Ophrys. Kleiber, Georges 2000 Sur le sens des proverbes. Langages 139: 39−58. Lamiroy, Béatrice (éd.) 2010 Les expressions verbales figées de la francophonie. Belgique, France, Québec et Suisse. Paris: Ophrys. Mathieu-Colas, Michel 1996 Essai de typologie des noms composés français. Cahiers de Lexicologie 69: 71−125. Mejri, Salah 2003 Introduction: Polysémie et polylexicalité. In: Salah Mejri (ed.), Polysémie et polylexicalité, 13−30. Caen: Presses Universitaires de Caen. Mejri, Salah 2005 Figement, néologie et renouvellement du lexique. Linx 52: 163−174. Mejri, Salah 2010 Le figement chez Alain Rey: Une pratique lexicographique théorisante. In: Giovanni Dotoli (ed.), Alain Rey de l’artisanat du dictionnaire à une science du mot, 97−105. Fasano-Paris: Schena/Hermann. Mejri, Salah forthc. Les pragmatèmes, des universaux phraséologiques très idiomatiques: Le cas du doua en arabe. In: Antonio Pamies (ed.), La parémiologie contrastive. Colloque d'AUROPHRAS 2010, Granada. Mel’čuk, Igor 1998 Collocations and Lexical Functions. In: Antony P. Cowie (ed.), Phraseology. Theory, Analysis, and Applications, 23−53. Oxford: Clarendon Press. Navarro Dominguez, Fernando 2000 Analyses du discours et des proverbes chez Balzac. Paris: L’Harmattan. Neveu, Franck 2004 Dictionnaire des sciences du langage. Paris: Colin. Petit, Gérard 2008 Dénomination et figement. In: Peter Blumenthal and Salah Mejri (eds.), Les séquences figées. Entre langue et discours, 131−144. Stuttgart: Steiner. Polguère, Alain 2008 Lexicologie et sémantique lexicale. Notions fondamentales. Montréal: Les Presses de l’Université de Montréal. Rey, Alain and Sophie Chantreau 1989 Dictionnaire des expressions et locutions. Paris: Robert. Riegel, Martin, Jean-Christophe Pellat and René Rioul 2009 [1994] Grammaire méthodique du français. Paris: Presses Universitaires de France. Siepmann, Dirk 2008 Idiomaticité et traduction: Essai d’une systematisation. In: Peter Blumenthal and Salah Mejri (eds.), Les séquences figées. Entre langues et discours, 173−194. Stuttgart: Steiner.

Salah Mejri, Paris (France)

742

III. Units and processes in word-formation II: Special cases

42. Multi-word expressions and univerbation in Slavic 1. 2. 3. 4. 5.

Introduction General overview Univerbation in contemporary Czech, Slovak and Polish Multiverbation References

Abstract Univerbation can be generally defined as the formation of a single-word unit based on a semantically equivalent multi-word denomination. In Slavic linguistics, however, the term has undergone certain specifications: in contemporary studies, it most often comprises the transition of multi-word denominations into one-word units (formally accompanied by ellipsis of the head and conversion or suffixation of the modifier). The activity of similar processes can be partially explained by the lower productivity of composition in Slavic languages. Differences in the interpretation of univerbation can be explained by the individual conceptions of word-formation as word-based or based on transformations of complex expressions. The introduction of this article deals with terminological issues. The second section briefly outlines the research background in individual Slavic languages, the late 1950s and early 1960s being selected as starting point in Czech linguistics. Section 3 focuses on univerbation in contemporary Czech, Slovak and Polish and its linguistic description. In section 4, a correlative process, i.e. multiverbation, is mentioned.

1. Introduction One of the most common definitions of “univerbation” in the Slavic languages, describing the concept of the phenomenon from a genetic and procedural point of view, is given by the multilingual terminological Slovník slovanské lingvistické terminologie. Vol. 1 [Dictionary of Slavonic Linguistic Terminology. Vol. 1] (Jedlička 1977: 184): “Univerbation − formation of a single-word denomination based on a synonymic multi-word denomination, e.g., Cz. plenární schůze [‘plenary meeting’] → plenárka [‘id.’]” (translated from Czech). What we can observe here is the replacement of a multi-word expression by a shorter synthetic form. This is the way in which the term is used in most articles on Slavic languages in this handbook, in contrast to other articles where the term refers to a procedure and the result of word-formation consisting in the contraction of the words in a fixed word combination into a single word (cf. E. anytime, whenever, everyday). However, in the literature on Slavic word-formation, univerbation is not only based on derivation as the above-mentioned Czech example might imply. There are also other types of denominations regarded as univerbation where the head of the semantically identical multi-word expression is formally deleted, but its meaning is preserved in the univerbate (cf. Russ. vychodnoj ‘day off’ ← vychodnoj den’ ‘id.; lit. go out-ADJ day’,

42. Multi-word expressions and univerbation in Slavic Cz. trvalá ‘perm’ ← trvalá ondulace ‘permanent wave; lit. permanent undulation’), or synecdoche (e.g., Slk. žrď ‘pole vault; lit. pole’ ← skok o žrdi ‘pole vault; lit. jump with pole’), blends, etc. As will be shown in this article, a narrower or broader definition of univerbation can also be connected with individual traditions of the description of wordformation as word-based or based on syntagms. The above-mentioned examples are formed from nominal word combinations. The univerbate is identical with the modifier and represents the same meaning as the complex denomination (cf. Russ. vychodnoj). In other cases, such as Cz. plenárka, the modifier becomes the basis of a derivative which is semantically identical with the complex base. Interestingly, the formal deletion of the head which is specific to numerous results of univerbation in Slavic, has only in some cases served as criterion for the definition of univerbates. Since Isačenko’s works of the 1950s which also include special types of compounds (see section 2.1), terms such as “semantic condensation”, “compression” (of multi-word units) have been applied as hyperonyms of the different univerbation processes, only some of which can be regarded as structurally based on the ellipsis of one of the components. Keeping to this tradition, the term “ellipsis” will be used in this article in accordance with the terminology of the respective authors (e.g., Isačenko 1958b; Kučerová 1973a, 1974; Buttler 1981). With this term, the authors refer to the ellipsis of the head and the nominalization of the modifier of the corresponding multiword expression, sometimes also to the ellipsis of the modifier of multi-word units. In Slavic studies on word-formation, the term “ellipsis” is usually not applied to the univerbation of multi-word expressions by means of suffixation with the modifier becoming the derivational base. For this reason, we use the term “ellipsis” in a broader sense exclusively with the objective of an additional explanation (as laid out in the abstract of this article) and a cross-reference (section 3.3). Terminological issues also concern the naming of the process and the results of univerbation. In Czech, Slovak, Polish, Bulgarian, and Ukrainian linguistics, the term univerbization (e.g., Cz. univerbizace, Pol. uniwerbizacja) is used instead of univerbation. The term (in either variant) is ambiguous. In Czech linguistics, for example, it is mainly used to designate a) a process of denomination consisting in the transition of a multiword denomination into a single-word unit, b) a way to form words (i.e. in relation to procedures such as derivation, composition, etc., leading to univerbation). The descriptive nature of the term covers the phenomenon from a genetic and procedural point of view, but much less from a structural viewpoint that focuses on the result of univerbation, etc. (see section 2.1 and 3.1). For univerbized denominations the term univerbate (Cz. univerbát, Pol. uniwerbat) is used (recently preferred in West Slavic languages), the term univerb, but also univerbat, can be found in Bulgarian and Russian word-formation. In Russian it refers either to suffixal transformations of attributive constructions of the A+N type (see Neščimenko 2003: 293), or it is also used in a more general meaning.

2. General overview 2.1. Czech studies The origins of research on univerbation in Czech linguistics go back to the Russian linguist A. V. Isačenko who lived in former Czechoslovakia. Isačenko’s papers (1956,

743

744

III. Units and processes in word-formation II: Special cases 1958a, 1958b) had a significant influence on the works of Dokulil (1962), Helcl (1963), and Jedlička (1965, 1969, 1973) with respect to theoretical problems of univerbation, its manifestations, formation processes and tendencies. Because of their importance for Czech linguistics and their reception and influence on research of univerbation in other Slavic languages we will discuss these works in more detail. Influenced by the word-formation theory of Jan Rozwadowski (1904), Isačenko starts from the thesis of the contradiction between the formally binominal denominations and the unity of their meaning. According to Isačenko, a solution to this contradiction can be found in the loss of formal and semantic segmentation, which is one of the rules of lexical development. Isačenko considers different processes related to the loss of formal and semantic segmentation as “semantic condensation” and supposes that condensation processes can best be researched on so-called complex (i.e. multi-word) denominations. “Complex denominations, consisting of several lexical units have a clear tendency to univerbation, i.e. to the compression of the semantic content into one word.” (Isačenko 1958b: 340; translated from Russian). Isačenko distinguishes the following processes of “semantic condensation” (which manifest themselves in various types of denomination): 1. Composition: Russ. mesto žitel’stva ‘place of domicile; lit. place domicile-GEN’ → mestožitel’stvo ‘id.; lit. place-domicile-NOM’, Slk. svetový názor ‘world view; lit. world-REL.ADJ view’ → svetonázor ‘id.’; 2. Juxtaposition (merger): Russ. uma lišennyj lit. ‘mind-GEN deprived of’ → umališennyj ‘mad, insane’, Cz. pravdě podobný lit. ‘truth-DAT similar’ → pravděpodobný ‘probable’. (Juxtaposition (merger) is the only procedure regarded as univerbation in English, German and French linguistics.); 3. Ellipsis: a) ellipsis of the head, i.e. nominalization of an adjective: e.g., Russ. mostovaja doroga ‘paved way’ → mostovaja ‘pavement’, Cz. taneční hodiny lit. ‘dance-REL.ADJ hours’ → taneční ‘dancing classes’; b) ellipsis of the modifier: Russ. grammofonnaja plastinka lit. ‘grammophoneREL.ADJ disc’ → plastinka ‘record’, Cz. železná dráha ‘railway; lit. iron-REL.ADJ way’ → dráha ‘id.’; 4. Affixal derivation: Russ. setčataja oboločka ‘retina; lit. net-ADJ membrane’ → setčatka ‘id.’, Cz. železná dráha ‘railway; lit. iron-REL.ADJ way’ → železn-ice ‘id.’; 5. Formation of compounds (with an abbreviated component) and abbreviations: Russ. medicinskaja sestra ‘nurse; lit. medicine-REL.ADJ sister’ → medsestra ‘id.’; Moskovskij gosudarstvennyj universitet ‘Moscow State university; lit. Moscow-REL.ADJ stateREL.ADJ university’ → MGU ‘id.’. Dokulil (1962; the following quotations are translated from Czech) explicitly addresses the d e l i m i t a t i o n of the term univerbation. From his point of view, univerbation is a “transformation of already existing multi-word denominations into single-word denominations” (p. 21); from the perspective of designation it is “a naming procedure altering an existing denomination” instead of forming a new one (p. 27). Dokulil also adapts Isačenko’s original conception of univerbation from the functional and structural points of view. According to Dokulil, the distinctive feature of a univerbized denomination is the coexistence of a synonymous multi-word denomination, which should − a further criterion emphasized by Dokulil − be a “real”, i.e. a fixed, established two-word denomi-

42. Multi-word expressions and univerbation in Slavic nation, but not any free combination of words (p. 115). Simultaneously, he points out that a motivational relation between a fixed two-word denomination and a derivative is only one possible motivational relation among others (Dokulil 1962: 114−117). Helcl (1963) introduces a stylistic approach to the delimitation and definition of univerbation. Parallel denomination types, i.e. univerbized and multi-word denominations, usually differ stylistically: the formation of univerbized expressions is characteristic of the colloquial language, whereas multi-word denominations are typical of the literary language, cf. Cz. kulturák (coll.) and kulturní dům ‘a building where cultural events, etc. take place; lit. house culture-GEN’ (Helcl 1963: 31). Jedlička (1965, 1969, 1973) approaches univerbation from the point of view of the synchronic developmental dynamics of the lexicon. He considers univerbation to be specific to the state and development of the contemporary vocabulary of Czech (as well as of other Slavic languages). Together with the elaboration of general issues of univerbation and its specific features in Czech (including stylistic aspects) he also deals with the problems of multiverbation as the counterpart of univerbation (Jedlička 1969; see also section 4). In the following period, attention is paid to particular occurrences and manifestations of univerbation, i.e. in slang and professional speech. Univerbation is studied in connection with abbreviation and composition, within the area of proper names, etc. The interest in univerbation is also reflected by the inclusion of a special entry into the “Encyclopaedic dictionary of Czech” (Hladká 2002: 505 f.). A recent description of univerbation has been provided by Kolářová (2011).

2.2. Slovak studies In Slovakia, the research on univerbation began to develop in the 1970s, especially with the works of Kučerová (1973a, 1973b, 1974). Her comparison of Russian and Slovak is accompanied by the study of theoretical questions concerning the delimitations of univerbation, univerbation by means of nominalization, and the relations between univerbation and ellipsis. The studies of Bosák (1984, 1987, 1989) who approaches univerbation from a sociolinguistic point of view as a dynamic phenomenon of contemporary communication, establish a new orientation of research. Thus, the following topics come to the foreground: the relatedness of univerbation and its types to individual varieties of the national language, the stylistic and communicative characteristics of univerbized words, and the cases of their stylistic revaluation (for instance, stylistic neutralization). Moreover, Bosák suggests distinguishing univerbation in the broad sense of the term, i.e. as a general tendency based on the principle of linguistic economy in communication, on the one hand, and, in the more specific sense, univerbation as a phenomenon applied to the lexical level of language. Bosák also provides a new typology of univerbized denominations (see section 3.2.3). The focus on contemporary spoken communication is reflected in numerous works dealing with univerbation in slang, professional speech and expert communication. Among other topics, Furdík (2004) gives his view on the boundaries between univerbation and word-formation. According to Furdík, it is not possible to separate univerba-

745

746

III. Units and processes in word-formation II: Special cases tion from word-formation because the motivational relations underlying a univerbized denomination are the same as in a common derivative. A specific semantic condensation is not only a typical feature of univerbized denominations, but also a potential feature of motivated words in general.

2.3. Polish studies A deep insight into the issues of univerbation is provided by the works of Polish linguists in the 1960s and 1970s, e.g., Siatkowska (1964), Satkiewicz (1969), Miodek (1976), and Buttler (1977, 1981). Numerous Polish authors mention that the approaches to and the delimitations of univerbation (including the term itself) were adopted from Czech works, e.g., Isačenko, Dokulil and Jedlička (see, e.g., Satkiewicz 1969; Buttler 1977; Szczepańska 1994). Buttler highlights, above all, the dubiousness of Dokulil’s (1962) main criterion of univerbation, i.e. a derivative is acknowledged to be a “univerb” only in the case that it is based on a f i x e d two-word expression. According to Buttler, on the contrary, the boundaries of fixed and free word combinations are fluent; not only two-word expressions, but also multi-word expressions can be motivationally related to one-word denominations. She suggests specifying the criteria of univerbation so that units based on syntactic combinations of a certain structural-semantic type like, for instance, ‘author of’ (cf. Pol. felietoniarz ‘feature writer; lit. feature-AGENT’, tekściarz ‘songwriter; lit. textAGENT’, etc.) can also be treated as univerbized denominations (← Pol. autor felietonu, autor tekstu (songu) lit. ‘author feature-GEN, author text-GEN’). With reference to Miodek (1976), Buttler (1977) also doubts whether it is adequate to limit univerbation solely to nouns. Therefore her conception of univerbation considers not only nouns, but also verbs (miedziować ‘to copper’ ← pokrywać miedzią ‘to cover with copper’), adjectives (przyrynkowy lit. ‘marketplace-ADJ’ ← przy rynku ‘at the marketplace’) as well as adverbs, derived from relational adjectives, as in the following example: wagowo (adv.) ← wagowy (adj.) ‘referring to scales’, e.g., (ustalić) wagowo ‘(to determine [the weight]) by scales’ ← za pomocą wagi ‘with the help of scales’. Buttler particularly focuses on the so-called disintegrational derivation (that formally corresponds to backformation, i.e. word-formation with the omission of the final morpheme or part of the base), e.g., długograj ‘LP’ (← płyta długogrająca ‘long-playing record; lit. record longplay-PART/ ADJ’), a question that continually arises in studies on Polish as well as on other Slavic languages (cf. among the recent publications on this topic: Sojda 2011). Valuable findings on the development and changes in the use of individual types of univerbation in contemporary Polish are provided by Jadacka (2001). Polish linguistics constantly deals with abbreviations which, by their nature, can also be regarded as univerbates. The literature on this topic is too extensive to be treated in this article. Attention has also been paid to the contrastive research of univerbation, among others, by Szczepańska (1994; Czech-Polish).

2.4. Russian studies In Russian linguistics of the 1960s and the early 1970s numerous works were dedicated to individual ways of forming single-word denominations from multi-word expressions

42. Multi-word expressions and univerbation in Slavic (suffixal univerbation, nominalization of adjectives, etc.). However, the term “univerbation” (Russ. univerbacija) is not yet used as a common term; instead, we can find terms such as “synthetic compression” (Russ. sintetičeskoe sžatie), “semantic condensation” (Russ. semantičeskaja kondenzacija), “inclusion” (Russ. vključenie), cf. Janko-Trinickaja (1964). The focus on the necessary terminological solution to the issue as well as on the analysis of other ways of replacing word combinations by one-word denominations is most thoroughly reflected by Lopatin (1973, 1978). He prefers the term univerbacija and delimits the means of univerbation (suffixation, nominalization, composition and abbreviation). In connection with suffixation, typical word-formation types are mentioned, e.g., nouns in -k(a) such as Russ. Literaturka ‘Literary Newspaper’ (← Literaturnaja gazeta ‘id.’), or motorka ‘motor boat’ (← motornaja lodka ‘id; lit. motor-REL.ADJ boat’), cf. Lopatin (1973: 44). Other aspects were introduced into the research on univerbation by the works of Zemskaja (1973, 1981). In connection with her investigations into colloquial Russian, the author views univerbation as a characteristic means of word-formation in this register. She describes it as “compressive” word-formation (cf. also Kopot’ 2005). (In contrast to Lopatin, Zemskaja does not regard abbreviation as a type of univerbation.) Research on univerbation as an active word-formation process holds a significant position in today’s widely branched examination of the phenomenon (Zemskaja 1997, 2000; Osipova 1999). At the same time, research into the means of univerbation from the point of view of dynamic tendencies in contemporary languages and communication emerged (Neščimenko 2000, 2003).

2.5. Bulgarian studies For a relatively long time in Bulgarian linguistics, univerbation was related entirely to dialectal phenomena. As late as the end of the 1970s univerbation began to be dealt with as a phenomenon specific to colloquial language. In this context, general issues of univerbation were examined from the point of view of denomination and word-formation (Videnov 1976; Murdarov 1983; Karastojčeva 1984; Radeva 1991). Another specific trait of the research into univerbation in Bulgaria can be seen in the fact that numerous works focus on the comparison of the phenomenon in two or more Slavic languages. The increased tendency towards univerbation has been reflected not only in numerous surveys on the dynamics of univerbation, again with contrastive aspects (Blagoeva 2003, 2004; Avramova 2003; Likomanova 2003), but also in works on theoretical issues (Avramova 2007, 2008).

2.6. Contrastive studies Monitoring research into univerbation in the Slavic languages, we see that it has often had a constrastive aspect. Attention has been paid to the comparison of univerbation in Czech and Polish (Siatkowska 1964; Szczepańska 1994), Slovak and Polish (Pančíková 1993), Slovak and Ukrainian (Sidorenko 1993), Russian and Slovak (Kučerová 1973a, 1974), Russian and Czech (Neščimenko 2000; Rudincová 2002), Bulgarian and Czech

747

748

III. Units and processes in word-formation II: Special cases (Videnov 1976; Avramova 2003), Bulgarian and Russian (Malkova and Nekezova 1996; Blagoeva 2004), Bulgarian, Czech and Russian (Blagoeva 2003); cf. also Neščimenko (2003) and Likomanova (2003) on current tendencies in univerbation in the West, East and South Slavic languages in a broader sociolinguistic context.

2.7. Conclusions The following short list of topics characterizing the research on univerbation in the individual countries summarizes the main subjects that have been examined so far. Research has predominantly been carried out with respect to the issues of: a) b) c) d) e) f)

conceptions and delimitations of univerbation and related terminological questions; means and procedures on the base of which univerbized denominations are formed; structural types of univerbized denominations; problems concerning the determination of the motivational basis of univerbation; delimitation between univerbation and other means of word-formation; assignment of univerbation and univerbized denominations to certain varieties of the national languages; g) stylistic and other characteristics of univerbized denominations and stylistic neutralization; h) dynamics in the development of individual types of univerbation; i) variability of univerbates; polysemy and expressivity of univerbized units.

3. Univerbation in contemporary Czech, Slovak and Polish In Czech, Slovak and Polish works dealing with univerbation, it is consistently stated that univerbation is used more and more often (Jedlička 1965: 20; Satkiewicz 1969: 182; Buttler 1981: 215; Bosák 1987: 231; Pančíková 1993: 150; Szczepańska 1994: 7; Jadacka 2001: 147). The characteristics of univerbation and univerbized denominations from the point of view of their formation and their belonging to different varieties of the contemporary languages as well as from the point of view of communication and style show the following similarities: univerbized denominations arise mainly within the colloquial language and the sphere of professional and special-interest communication (professional speech, slang). They penetrate into literary communication in which they can maintain their stylistic markedness as well as undergo neutralization. As has been stated in numerous studies, not only univerbized words, but also univerbation as a wordformation procedure shows a tendency towards stylistical neutralization in the contemporary Slavic literary languages (cf. Jedlička 1973: 219 on Czech; Buttler 1981: 219 on Polish; Bosák 1989: 305 on Slovak).

3.1. Conception and delimitation of univerbation Different authors deal with the conception and delimitation of univerbation in different ways. Univerbation is explained as the removal of contradictions between the formal

42. Multi-word expressions and univerbation in Slavic binominal structure of the motivating word combination and the unity of meaning, and as a manifestation of the tendency towards language economy, but also − in the terminology of the Prague School − towards democratization of language, etc. The differences in the delimitations of univerbation depend on the methodological approaches to the phenomenon. They can be based on the development and the “synchronic dynamics” of contemporary language as well as on sociolinguistic and communicative approaches. (On the diachronical aspects of univerbation see Němec 1968: 38−39.) There are two views that influence the differentiation and delimitation of univerbation either in a broader or a more specific sense: a) In the broader sense of the term, univerbation is described as a process typical for the state and the development of the vocabulary of the contemporary language (Jedlička 1965, 1969, 1973; see also section 2.1) or, even broader, as a “general tendency that stems from the principle of the economy of communication […] which is applied at all levels of language and has a systemic character” (Bosák 1987: 231; translated from Slovak). From the structural point of view, Buttler (1981: 215) has characterized univerbation as the most typical manifestation of the tendency towards condensation. b) Univerbation in the more specific sense of the term is defined − as already mentioned above − as the formation of a single-word denomination (a word) from a multi-word denomination, i.e. as a process which does not form a new denomination but adjusts an existing one. From the point of view of the typology of denomination procedures, univerbation is a specific kind of word-formation as it makes use of the same means which are characteristic of word-formation in general, with the distinction that univerbates are always based on complex motivating units. It is also described as a substitution of a multi-word denomination by a single-word unit (Furdík 2004: 66; Nagórko 2002: 211) or as the transition of a multi-word denomination into a single-word expression (Kučerová 1973a). Finally, it is approached as means of secondary naming (“transnomination”) which means assigning a new name to an already named object (Kaliszan 1986). As to the nature of the formation of such designations (words), it is specified as an adjustment of an existing designation based on condensation (Dokulil 1962: 27).

3.2. Different views on the units underlying univerbation A prerequisite for the formation of univerbized words (see section 2.1), is the existence of synonymous multi-word units to which such words are motivationally related. The standpoints on the nature and essence of the multi-word unit are again very different in the treatment of individual authors (and partly within individual national linguistics as well). The concept used first of all is based on the criterion that the underlying (motivating) unit must be a fixed two-word denomination (e.g., Cz. jehličnatý strom ‘conifer; lit. needle-ADJ tree’ → jehličnan ‘conifer’). However, motivation by an underlying complex denomination, consisting of more than two members, is admitted (Cz. závodník v sdruže-

749

750

III. Units and processes in word-formation II: Special cases ném závodě ‘racer in a combined race’ → sdruženář ‘id.’). In either case, the criterion of stability and the semantic function of the motivating word combination as a naming unit is fulfilled. Other conceptions assume that one-word denominations are also based on syntactic combinations (motivational entities) of a certain structural-semantic type, e.g., Pol. specjalista w okresie historii nowożytnej ‘expert in the field of contemporary history; lit. expert in the field of history contemporary’ → nowożytnik ‘id.’ (cf. also Pol. materiał na ubranie, materiał na koszule, … ‘fabric for suits, fabric for shirts, …’ → ubraniówka, koszulówka, … ‘id.’). Being typical for Polish linguistics, this approach leads to the fact that, for instance, names of persons with the suffix -ak (warszawiak based on the synonymous word combination mieszkaniec Warszawy ‘resident of Warsaw’), and nouns with the suffix -ina (cielęcina ← mięso cielęce ‘veal; lit. meat calf-REL-POSS.ADJ’) are considered to be univerbates. Both approaches lead to differences when determining particular units (and even whole semantic word-formation types) as results of univerbation. In Czech word-formation, for instance, personal nouns such as Cz. fotbalista ‘footballer’, hokejista ‘hockey player’; Varšavan/Varšavák ‘resident of Warsaw’; teletina ‘veal’, etc. are not regarded as univerbates but as denominal derivatives (cf. Dokulil 1962: 115−117), whereas within the second approach they are ranked among univerbates (see Szczepańska 1994: 31, 32).

3.3. Different views on the range of procedures leading to univerbation The typical features in the approach to univerbation in Czech, Slovak and Polish also include a standpoint concerning the range of univerbation. Generally, Isačenko’s typology (see section 2.1) is accepted, which includes both procedures resting on the semantic and formal condensation of a multi-word expression into one word (univerbation by derivation, nominalization, etc.), and the transformation of a multi-word expression into a single word without change of meaning (juxtaposition, etc.), cf. the typology of Buttler (1981), Hladká (2002), and Kaliszan (1986). Buttler (1981: 215) ranks the following procedures among univerbation: 1. Derivation (Pol. odblaskówka ‘reflector’ ← lampa odblaskowa ‘reflector; lit. lamp reflection-REL.ADJ’); 2. Composition (Pol. mrozoodporność ‘frost resistance’ ← odporność na mróz ‘id.; lit. resistance against frost’, klubokawiarnia (coordinative compound with connecting vowel) ‘café-club’ ← kawiarnia-klub ‘id.’); 3. Ellipsis (Pol. karny ‘penalty throw; lit. penal’ ← rzut karny ‘id.; lit. throw penal’) [here, the ellipsis of the head is accompanied by the nominalization of the modifier]; 4. Disintegration (Pol. długograj ‘LP’ ← płyta długogrająca ‘long-playing record; lit. record longplay-PART/ADJ’) [a special case of backformation of the modifier + ellipsis of the head of the multi-word expression, see section 2.3]; 5. Abbreviation (Pol. Cepelia ← Centrala Przemysłu Ludowego i Artystycznego ‘Folk and Art Industry Cooperative’).

42. Multi-word expressions and univerbation in Slavic The typology of univerbation procedures proposed by Hladká (2002) seems to be similar. Bosák (1987: 232) provides the following typology for Slovak. It includes: 1. Formation of single-word denominations on the basis of motivational multi-word denominations by: a) derivation (Slk. presilovka ‘power play’ ← presilová hra ‘id.; lit. power-REL.ADJ. play’); b) nominalization (Slk. triedna ‘class teacher f.; lit. class-REL.ADJ-FEM’ ← triedna učiteľka ‘id.; lit. class-REL.ADJ-FEM teacher-FEM’) [the modifier is nominalized after the ellipsis of the head]; c) synecdoche (Slk. žrď ‘pole vault; lit. pole’ ← skok o žrdi ‘id.’); d) composition (Slk. prvoligista ‘first-league player; lit. first-league-PERS’ ← hráč prvej ligy ‘id.; lit. player first-GEN league-GEN’); 2. Simplification of multi-word denominations, e.g.: a) contextual (Slk. vláda ‘the government [of the ČSSR]’ ← vláda ČSSR ‘id.’; b) by abbreviation, acronymization (Slk. SAV ‘Slovak Academy of Sciences’ ← Slovenská akadémia ved ‘id.’). Cf. also Bosák (1987: 235 f.; translated from Slovak): “Sometimes all procedures of abbreviation […] are also ranked as univerbation. However, in such cases it is not univerbation in the genuine sense of the term but a mechanical process which is above all dependent on the context (the situation respectively). For this reason, we consider this process as simplification.”

3.4. Univerbation by derivation and composition The most widespread and frequent procedure of univerbation in Czech, Slovak and Polish is univerbation by derivation (Jedlička 1965; Satkiewicz 1969; Bosák 1987; Pančíková 1993; Szczepańska 1994; Jadacka 2001). The univerbates are formed from nominal two-word combinations whose (adjectival or nominal) attribute becomes the basis of derivation. In their formation a narrow range of semantically non-specific and frequent suffixes are used: in Czech -ka (minerál-k-a ‘mineral water’ ← minerální voda ‘id.; lit. mineral-REL.ADJ water’), -ák (trestň-ák ‘penalty kick’ ← trestný kop ‘id.; lit. penaltyREL.ADJ kick’), sometimes also -ař/-ář (oč-ař ‘oculist’ ← oční lékař ‘id.; lit. eye-REL.ADJ doctor’). (In slangs the means of univerbation are more diversified.) The situation is similar in Slovak: most often the suffix -ka (presilov-k-a ‘power play’, see section 3.3) is used, -ár/-iar with masculine personal nouns (oč-iar ‘oculist’), in slang expressions also -ák/-iak (zim-ák ‘ice-rink’ ← zimný štadión ‘id.; lit. winter-REL.ADJ stadium’) (Bosák 1987). In Polish, the suffixes -(ów)ka (podstawów-k-a ‘elementary school’ ← szkoła podstawowa ‘id.; lit. school basis/foundation-REL.ADJ’), -ak (poprawcz-ak ‘penitentiary, community home’ ← dom poprawczy ‘id.; lit. house (to) better-/improve-REL.ADJ’), -(ow)iec (odrzutowiec ‘jet’ ← samolot odrzutowy ‘id.; lit. aircraft jet-propelled’) are typically used. Due to the broader understanding of univerbation in Polish linguistics, the suffixes include also -ista (polemista ‘author of polemic works; lit. polemic-PERS’ ← autor utworów polemicznych ‘id.’), -(ow)iec, also for personal nouns (radiowiec ‘employee of a broadcasting company; lit. radio-PERS’ ← pracownik radia ‘id.; lit. employee

751

752

III. Units and processes in word-formation II: Special cases radio-GEN’), and -acz in names of devices or machines (opiekacz ‘toaster’ ← maszynka do opekania ‘id.; lit. machine-DIM for toasting-GEN’, etc.), cf. Szczepańska (1994), Pančíková (1993), Jadacka (2001). Univerbation is also based on the nominalization of the attribute in a two-word denomination (e.g., Cz. vysoká ‘college’ ← vysoká škola ‘id.; lit. high school’, Slk. nočná ‘night shift’ ← nočná smena ‘id.; lit. night-REL.ADJ shift’, Pol. oddziałowa ‘charge nurse’ ← siostra oddziałowa ‘id.; lit. nurse department-REL.ADJ’). Compared to derivational univerbation, nominalization is less frequent in all of the three languages, although not to a similar extent. According to Szczepańska (1994: 33, 54), it is very common in Polish (for instance, in names of soups, alcoholic beverages, and cigarettes, but also languages, trains, dances, etc.; see also article 156 on Polish, although the respective formations are not regarded as results of univerbation, but as suffixal derivatives). In Czech and Slovak it is less used. A widely used type-forming procedure can be observed in Czech and Slovak, e.g., bronz ‘bronze medal; lit. bronze (noun)’ (← bronzová medaile/bronzová medaila ‘bronze medal; lit. bronze-REL.ADJ medal’), cf. also Pol. kompakt ‘compact disc; lit. compact (noun)’ (← płyta kompaktowa ‘compact disc; lit. disc compact (adj.)’). The formation of these words is described either as de-suffixation of the modifying adjective (Cz. bronzová (medaile) → bronz), in Polish word-formation as “disintegration”, or − in Slovak studies − as metonymically motivated formation (Bosák 1987). This type of naming raises doubts about whether it can be regarded as univerbation. Nagórko (2002: 192) points out that it is necessary to distinguish cases where the one-word expression has an established synonymous multiword denomination, e.g., Pol. noga (lit. ‘foot’) in the meaning of ‘football’ (coll.) and piłka nożna ‘id.; lit. ball foot-REL.ADJ’, and where the multi-word combinations are periphrases (e.g., arab ‘Arab’ (equest.) and koń arabski ‘Arabian horse’ / koń rasy arabskiej lit. ‘horse of the Arabian breed’, with the relational adjective postponed, etc.). Jadacka (2001: 147) emphasizes that in Polish, the last twenty years were a period in which univerbation by suffixation increased considerably (e.g., Pol. czasówka ‘time trial’ ← jazda czasowa ‘id.; lit. drive-N time-REL.ADJ’, see also poprawczak, odrzutowiec above). However, at the end of the 1970s, nouns going back to clipped adjectives, sometimes accompanied by suffixation, started to spread (Polish rzeczowniki postadiektywne ‘postadjectival nouns’), e.g., Pol. komórka ‘mobile telephone; lit. cell’ (← telefon komórkowy ‘id.; lit. telephone cell-REL.ADJ’), etc. Univerbation by composition includes different structural types (e.g., Cz. stará čeština ‘Old Czech’ → staročeština ‘id.’, fotografický papír ‘photographic paper’ → fotopapír ‘id.’, profesionální armáda ‘professional army’ → profiarmáda ‘id.’). From the point of view of motivation, it is possible to treat them either as cases of univerbation or as proper compounds. There are different views among individual authors. Pančíková (1993), for instance, regards formations like Slk. volnoštýliar ‘free style wrestler; lit. free-o-style-PERS’ as univerbized compounds (its base is the two-word denomination volný štýl ‘free style’) (see also article 33 on synthetic compounds in German). According to Pančíková, the number of formations of this type is increasing in Polish and Slovak. In some works, different formations of acronyms, partly in combination with suffixation, are also ranked under univerbation, e.g., Cz. jednotka intenzivní péče ‘intensive care unit; lit. unit intensive-GEN care-GEN’ → jip and jip-ka ‘id.’, Slk. Slovenská akadémia vied ‘Slovak Academy of Sciences’ → SAV or Savka (coll., slang) ‘id.’, Pol. Polska akademia nauk ‘Polish Academy of Sciences’ → PAN ‘id.’), although this point of view is not generally accepted (see Avramova 2007; Bosák 1987: 235 f., and section 3.2).

42. Multi-word expressions and univerbation in Slavic

4. Multiverbation In studies on univerbation it is often pointed out that univerbation requires the formation of new multi-word units as a permanent source of future univerbation (see, e.g., Jedlička 1969: 94). Within this context, attention is also paid to multiverbation. The problems of multiverbation are a border phenomenon which is involved either in the area of vocabulary or in the area of syntax. Multiverbation becomes relevant in relation to research in analytical constructions, including phrasemes, and their syntactic and stylistic specifics. Hereafter, we will shortly introduce the conception elaborated in Czech linguistics by Jedlička (1969), which has been continued by Slovak and Polish researchers (Ondrejovič 1989; Buttler 1978, 1981; Mietła 1998). Multiverbation is defined as the formation of analytical multi-word denominations in parallel to synonymous one-word denominations (e.g., Cz. čtenářská obec ‘the reading public’ − čtenáři ‘readers‘, provádět výzkum ‘to do research’ − zkoumat ‘to research’). Although multiverbation and univerbation are processes which are typical for the contemporary language, the former is not a counterpart of the latter: multiverbized units coexist with synonymous one-word denominations, but they are not considered to be formed out of single-word denominations. They also differ from univerbized units with respect to use. Very often they are limited to the scope of literary language, the sphere of public communication and expert discourse, etc. Multiverbation units can also be described under formal, semantic and stylistic aspects. Based on their formal aspects, they encompass several types: a) Verb + dependent noun (Cz. provést (pf.) / provádět (ipf.) výzkum ‘to do research’ − prozkoumat (pf.) / zkoumat (ipf.) ‘to research’); b) Noun + dependent adjective (Cz. podnikatelská obec ‘business community; lit. entrepreneur-REL.ADJ community’ − podnikatelé ‘business people, entrepreneur-PLUR’); c) Noun + dependent adjective, motivationally coexisting with an adverb (Cz. zajímavým způsobem ‘in an interesting way; lit. interesting-INSTR way-INSTR’ − zajímavě ‘interestingly’); d) Noun + dependent noun (Cz. otázka výchovy ‘a matter of education; lit. question education-GEN’ − výchova ‘education’). There are differences in the manifestation of multiverbation in the individual Slavic languages based on the structure of the respective denomination models and according to their frequency and distribution.

5. References Avramova, Cvetanka 2003 Slovoobrazuvatelni tendencii pri sӑštestvitelnite imena v bӑlgarskija i českija ezik v kraja na XX vek. Sofija: Heron Press. Avramova, Cvetanka 2007 Univerbáty v lexikálním systému jazyka. Bohemistyka 7(3): 190−206. Avramova, Cvetanka 2008 Univerbatite v leksikalnata sistema na ezika. Sӑpostavitelno ezikoznanie 33(1): 99−115.

753

754

III. Units and processes in word-formation II: Special cases Blagoeva, Diana 2003 Univerbizacijata v naj-novoto slavjansko slovoobrazuvane (vӑrchu material ot bӑlgarskija, ruskija i češkija ezik). Sӑpostavitelno ezikoznanie 28(3): 5−21. Blagoeva, Diana 2004 Edin tip neologizmi univerbati v bӑlgarskata razgovorna reč (v sӑpostavka s ruski). In: Christina Stanova (ed.), Problemi na bӑlgarskata razgovorna reč. Kniga šesta, 45−52. Veliko Tӑrnovo: Universitetsko izdatelstvo “Sv. sv. Kiril i Metodij”. Bosák, Ján 1984 Hovorovosť ako dynamický faktor. Slovenská reč 49: 65−73. Bosák, Ján 1987 O vymedzení univerbizácie. Slovenská reč 52: 231−237. Bosák, Ján 1989 Univerbizácia. In: Ján Horecký, Klára Buzássyová and Ján Bosák (eds.), Dynamika slovnej zásoby súčasnej slovenčiny, 298−306. Bratislava: Veda. Buttler, Danuta 1977 Niektóre problemy opisu zjawisk uniwerbizacji. Slavistična revija 25: 435−448. Buttler, Danuta 1978 Procesy multiwerbizacji we współczesnej polszczyźnie. Poradnik językowy (2): 54−62. Buttler, Danuta 1981 Tendencje rozwojowe w zasobie słownym powojennej polszczyzny. In: Halina Kurkowska (ed.), Współczesna polszczyzna, 187−219. Warszawa: Państwowe Wydawnictwo Naukowe. Dokulil, Miloš 1962 Tvoření slov v češtině. Vol. 1: Teorie odvozování slov. Praha: Nakladatelství ČSAV. Furdík, Juraj 2004 Slovenská slovotvorba. Prešov: Náuka. Helcl, Miloš 1963 Univerbizace a její podíl při růstu dnešní slovní zásoby. Slovo a slovesnost 24: 29−37. Hladká, Zdeňka 2002 Univerbizace. In: Petr Karlík, Marek Nekula and Jana Pleskalová (eds.), Encyklopedický slovník češtiny, 505−506. Praha: Nakladatelství Lidové noviny. Isačenko, Aleksandr V. 1956 O některých zákonitostech v oblasti pojmenování. In: Václav Křístek (ed.), Sborník Vysoké školy pedagogické v Olomouci. Jazyk a literatura 3, 17−23. Praha: SPN. Isačenko, Aleksandr V. 1958a Obecné zákonitosti a národní specifičnost ve vývoji slovní zásoby slovanských jazyků. In: Jiří Danˇhelka (ed.), K historickosrovnávacímu studiu slovanských jazyků, 143−151. Praha: Sta´tni´ pedagogicke´ nakladatelství. Isačenko, Aleksandr V. 1958b K voprosu o strukturnoj tipologii slovarnogo sostava slavjanskich literaturnych jazykov. Slavia 27: 334−352. Jadacka, Hanna 2001 System słowotwórczy polszczyzny (1945−2000). Warszawa: Wydawnictwo Naukowe PWN. Janko-Trinickaja, Nadija A. 1964 Processy vključenija v leksike i slovoobrazovanii. In: Iosif P. Mučnik and Michail V. Panov (eds.), Razvitie grammatiki i leksiki sovremennogo russkogo jazyka, 18−35. Moskva: Nauka. Jedlička, Alois 1965 K charakteristice slovní zásoby současné spisovné češtiny. Slavica Pragensia 7: 13−27.

42. Multi-word expressions and univerbation in Slavic Jedlička, Alois 1969 Univerbizace a multiverbizace v pojmenovacích strukturách. Slavica Pragensia 11: 93− 101. Jedlička, Alois 1973 Dynamika současného spisovného jazyka a jeho stylová diferenciace. In: Bohuslav Havránek (ed.), Československé přednášky pro VII. mezinárodní sjezd slavistů ve Varšavě. Lingvistika, 211−221. Praha: Academia. Jedlička, Alois (ed.) 1977 Slovník slovanské lingvistické terminologie. Vol. 1. Praha: Academia. Kaliszan, Jerzy 1986 Semantiko-kondensacionnaja univerbacija sostavnych naimenovanij v sovremennom russkom jazyke. Poznań: Uniwersytet im. Adama Mickiewicza. Karastojčeva, Cvetana 1984 Univerbizacija s kalamburen efekt. Ezik i literatura (39)5: 37−49. Kolářová, Ivana 2011 Univerbizace. In: František Štích (ed.), Kapitoly z české gramatiky 1, 273−410. Praha: Academia. Kopоť, Lilija V. 2005 Univerbacija kak vid kompressivnogo slovoobrazovanija. Ph.D. dissertation, Adyghe State University Majkop. Kučerová, Eleonóra 1973a Univerbizácia substantivizáciou dvojslovných pomenovaní mužských osôb. (Na ruskom materiáli v porovnaní so slovenským). Slavica Slovaca 8(1): 3−14. Kučerová, Eleonóra 1973b Univerbizácia odvodzovaním dvojslovných pomenovaní v ruštine. Československá rusistika 18(3): 109−113. Kučerová, Eleonóra 1974 Z problematiky univerbizácie substantivizáciou. Podstatné mená so ženským gramatickým rodom. Slavica Slovaca 9(1): 23−34. Likomanova, Iskra 2003 Projavi na tendencijata za ezikova ikonomija v južnoslavjanskite ezici. In: Ingeborg Ohnheiser (ed.), Komparacja systemów i funkcjonowania współczesnych języków słowiańskich. 1. Słowotwórstwo/Nominacja, 307−315. Opole: Uniwersytet Opolski. Lopatin, Vladimir V. 1973 Roždenie slova. Neologizmy i okkazional’nye obrazovanija. Moskva: Nauka. Lopatin, Vladimir V. 1978 Suffiksal’naja univerbacija i smežnye javlenija v sfere obrazovanija novych slov. In: Nadežda Z. Kotelova (ed.), Novye slova i slovari novych slov, 72−80. Leningrad: Nauka. Malkova, Vera and Jordanka Nekezova 1996 O nekotorych obščich tendencijach v sovremennom russkom i bolgarskom slovoobrazovanii. Sӑpostavitelno ezikoznanie 21(1): 14−29. Mietła, Joanna 1998 Multiwerbizacja w języku czeskim i polskim. Toruń: Uniwersytet Mikołaja Kopernika. Miodek, Jan 1976 Syntetyczne konstrukcje leksykalne w języku polskim. Wrocław: Zakład Narodowy im. Ossolińskich. Murdarov, Vladko 1983 Sӑvremenni slovoobrazovatelni procesi. Sofija: Nauka i izkustvo. Nagórko, Alicja 2002 Zarys gramatyki polskiej. Warszawa: PWN. Němec, Igor 1968 Vývojové postupy české slovní zásoby. Praha: Academia.

755

756

III. Units and processes in word-formation II: Special cases Neščimenko, Galina P. 2000 Slovoobrazovanie razgovornogo jazyka v svete tendencii jazykovoj ėkonomii. In: Ingeborg Ohnheiser (ed.), Wortbildung. Interaktiv im Sprachsystem − interdisziplinär als Forschungsgegenstand. Slovoobrazovanie v ego otnošenijach k drugim sferam jazyka, 253−270. Innsbruck: Institut für Sprachwissenschaft. Neščimenko, Galina P. 2003 Jazykovaja ėkonomija kak impul’s dinamiki nominacionnoj sistemy. In: Ingeborg Ohnheiser (ed.), Komparacja systemów i funkcjonowania współczesnych języków słowiańskich. 1. Słowotwórstwo/Nominacja, 289−306. Opole: Uniwersytet Opolski. Ondrejovič, Slavomír 1989 Multiverbizačné procesy v spisovnej slovenčine. In: Ján Horecký, Klára Buzássyová and Ján Bosák (eds.), Dynamika slovnej zásoby súčasnej slovenčiny, 251−259. Bratislava: Veda. Osipova, Ljudmila I. 1999 Aktivnye processy v sovremennom russkom slovoobrazovanii. Suffiksal’naja univerbacija, usečenie. Moskva: Russkij jazyk. Pančíková, Marta 1993 Univerbizované pomenovania v slovenčine a polštine. In: Stanisław Gajda (ed.), Języki słowiańskie wobec współczesnych przemian w krajach Europy Środkowej i Wschodniej. Materiały konferencji z 23−25 IX 1992 r., 149−153. Opole: Wyższa Szkoła Pedagogiczna. Radeva, Vasilka 1991 Slovoobrazuvaneto v bӑlgarskija knižoven ezik. Sofija: UI “Sv. Kliment Ochridski”. Rozwadowski, Jan 1904 Wortbildung und Wortbedeutung. Eine Untersuchung ihrer Grundgesetze. Heidelberg: Winter. Rudincová, Blažena 2002 Processy univerbizacii v sovremennom russkom jazyke. In: Jiří Gazda (ed.), Jazykovědná rusistika na počátku nového tisíciletí, 192−200. Brno: UM. Satkiewicz, Halina 1969 Produktywne typy słowotwórcze współczesnego języka ogólnopolskiego. Warszawa: Wydawnictwa Universytetu Warszawskiego. Siatkowska, Ewa 1964 Syntetyczne i analityczne nazwy v języku czeskim i polskim. Prace Filologiczne 18(2): 219−237. Sidorenko, Evgenij N. 1993 K voprosu o motivacionnoj baze univerbov (na materiale slovackogo i ukrainskogo jazykov). In: Lev N. Smirnov (ed.), Problemy razvitija i funkcionirovanija sovremennych literaturnych jazykov, 110−124. Moskva: Institut slavjanovedenija i balkanistiki RAN. Sojda, Sylwia 2011 Procesy uniwerbacyjne w języku słowackim i polskim. Katowice: Uniwersytet Śląski. Wydawnictwo Gnome. https://wydawnictwo.us.edu.pl/sites/wydawnictwo.us.edu.pl/ files/1663_pdf.pdf [last access 12 Jan 2012]. Szczepańska, Elżbieta 1994 Uniwerbizacja w języku czeskim i polskim. Kraków: Universitas. Videnov, Michail 1976 Univerbizacijata − prisӑšta čerta na bӑlgarskija razgovoren stil. Ezik i literatura 31(4): 28−35. Vlková, Věra 1978 K problematice tzv. multiverbizačních spojení, zvláště v odborném stylu. Slovo a slovesnost 39: 106−115.

43. Compounds and multi-word expressions in Slavic Zemskaja, Elena A. 1973 Russkaja razgovornaja reč’. Moskva: Nauka. Zemskaja, Elena A. 1997 Aktivnye tendencii slovoproizvodstva. In: Evgenij N. Širjaev (ed.), Russkij jazyk. Najnowsze dzieje języków słowiańskich, 169−201. Opole: Uniwersytet Opolski − Instytut Filologii Polskiej. Zemskaja, Elena A. 2000 Aktivnye processy sovremennogo slovoproizvodstva. In: Elena A. Zemskaja (ed.), Russkij jazyk konca XX stoletija (1985−1995), 90−141. 2nd ed. Moskva: Jazyki russkoj kul’tury. Zemskaja, Elena A., Margarita V. Kitajgorodskaja and Evgenij N. Širjaev 1981 Russkaja razgovornaja reč’. Obščie voprosy. Slovoobrazovanie. Sintaksis. Moskva: Nauka.

Olga Martincová, Prague (Czech Republic)

43. Compounds and multi-word expressions in Slavic 1. 2. 3. 4. 5. 6. 7.

Introduction Relations between modifier and head in the traditional Slavic compounds Relations between modifier and head in combinations of relational adjectives and nouns Types of compounds without a linking vowel and their adjectival competitors Derivatives from multi-word expressions (RA+N) and denominal derivatives Conclusion References

Abstract Slavic languages, as compared to English or German, show significant limitations with respect to the formation of nominal compounds. These limitations are often compensated for by complex designations consisting of a relational adjective and noun or by other types of multi-word expressions. To a significant extent the increase of new vocabulary in modern Slavic languages feeds on borrowings and loan translations or hybrid words, a large part of which consists of different types of compounds and compound elements of foreign origin. Most of the new compounds continue to exhibit parallel designations in terms of multi-word expressions consisting of a relational adjective+noun or noun+noungen/prep.case . Their occurrence is often more frequent than that of compounds and thus represents the continuation of typical preferences in designation.

1. Introduction In a study on the structural typology of the vocabulary of various Slavic languages, Isačenko (1958) addressed the interaction of different methods of designation in modern Slavic

757

758

III. Units and processes in word-formation II: Special cases languages. In his study he referred primarily to different structures of multi-word expressions (also in relation to different loan influences experienced by Slavic languages) as well as to possible transformations of complex designations into one-word designations (compounds or derivatives). So, while there are multi-word expressions including formations of the genitive type in Russian, Czech frequently exhibits word combinations consisting of a relational adjective (RA) and noun (N) or suffixations, e.g., Cz. srdeční vada lit. ‘heart-RA defect’ vs. R. porok serdca ‘heart defect; lit. defect heartgen’, Cz. železnice ‘rail way; lit. iron(N/RA-stem)-SUFF’ vs. R. železnaja doroga ‘id.; lit. iron-RA way’. Polish designations (genitival and prepositional types, according to Isačenko) like trzęsienie ziemi ‘earthquake; lit. quake earthgen’ and dziurka do klucza ‘keyhole; lit. hole-DIM for key’, might be paralleled with compounds in Russian (zemletrjasenie ‘earth-quake’) or RA+N (zamočnaja skvažina lit. ‘lock-RA hole’). Other examples presented by Isačenko have meanwhile experienced changes: so alongside the original RA+N-combination tisková konference ‘press conference’, Czech also uses borrowings without relational adjectives (press konference), as was formerly the case in Russian (press-konferencija) next to the new direct borrowing brifing ‘briefing’, to denote a press meeting. Even in the 1950s Isačenko (1958: 352) considered complex designations as an exceptionally productive method of lexical expansion, in particular in the field of terminology, found in all standard Slavic languages. The adoption of such formations for everyday use often goes hand in hand with structural simplifications as well as consequent stylistic revaluations, i.e. that colloquial formations and professionalisms may enter the standard language. Similar issues were later reviewed by Dokulil (1962) in terms of the intralinguistic stylistic comparison of different structures (multi-word expression, compound, derivative, cf. also Štekauer 2005, 2009), e.g., Cz. růže čajová (N+RA) ‘tea rose’ in botanical nomenclature compared to the colloquial derivative čajovka (N ← RA) ‘id.’ (Dokulil 1962: 116). In more recent Slavic grammars, the description of word-formation is usually undertaken according to word classes which are in turn classified according to methods of word-formation (compounding, derivation, etc.). The description of the derivation of the noun, adjective or verb mostly involves a subclassification according to the base word (N ← V, N ← N, etc.) and, furthermore, on the basis of the word-formation category (e.g., in the case of deverbal nouns: abstract nouns, agent nouns, etc.). Relations between the individual word-formation methods or − in a wider sense − designation methods are, for the most part, only considered in reference to word combinations as forming a basis for synthetic compounds and derivatives (from relational adjectives), partly also in reference to compounds as a basis for derivatives, but not in view of the relationship between compounds and multi-word expressions consisting of a relational adjective and noun, cf. the Russian examples in Table 43.1: Tab. 43.1: Semantically identical compounds and multi-word expressions in Russian Compound

RA+N

sud-o-mechanik (20,000 occurences in Yandex)

sudovoj mechanik (21,000)

‘ship mechanic’

sud-o-remont (784,000)

sudovoj remont (1,000)

‘ship repair’

sud-o-dvigatel’ (41)

sudovoj dvigatel’ (1 m.)

‘marine engine’

sud-o-stroenie (2 m.)

sudovoe stroenie (238)

‘ship building’

43. Compounds and multi-word expressions in Slavic

759

For Slavic languages there is no investigation in which compounds and designations consisting of a relational adjective and noun have been jointly considered as “complex nominals” on the basis of a corresponding inventory of relations between modifier and head, as was undertaken for English by Levi (1978). This may be due to the fact that the formation types of nominal determinative compounds are limited in terms of possible relations between head and modifier. The focus of the relationship between composition and multi-word designation − (i) N+N including a nominal modifier in the genitive case or (ii) a prepositional case or else (iii) RA+N − is therefore rather reserved for individual contrastive studies (cf. Gladrow 1998 on Russian and German, Szymanek 2009 on the comparison of Polish and English, Rainer 2013 also including evidence from Slavic languages). When considering the examples provided by Szymanek (2009: 466) alongside Russian and Czech equivalents we can see that indeed they do not have compounds − just as their Polish equivalents. There is, however, also no consistent structural correspondence between the equivalents in the closely related Polish and Czech languages, and the frequency of use of the examples can vary strongly when it comes to formal correspondences (for Polish and Czech determined via Google, for Russian via Yandex; the same holds for the following Tables), cf. Table 43.2: Tab. 43.2: English compounds and their equivalents in Polish, Czech and Russian (adapted after Szymanek 2009: 466) a)

E. telephone number

N+Ngen

P. i. numer telefon-u

Čz. [číslo telefon-u]

R. nomer telefon-a (33 m.)

N+Nprep.case

P. ii. *numer do telefon-u

Čz. −

−

N+RA/RA+N

P. iii. *numer telefon-iczn-y

Čz. telefon-n-í číslo

R. telefon-n-yj nomer (3 m.)

b)

E. computer paper

N+Ngen

P. i. *papier komputer-a

−

−

N+Nprep.case

P. ii. papier do komputer-a (5 m.)

Čz. [papír pro počítač]

R. bumaga dlja kompjutera (less frequent)

N+RA/RA+N

P. iii. papier komputerow-y (124,000)

Čz. počítač-ov-ý papír (11,000)

R. komp’juter-n-aja bumaga

c)

E. tooth paste

N+Ngen

P. i. *pasta zęb-ów

−

−

N+Nprep.case

P. ii. pasta do zębów

[Čz. pasta na zuby]

R. pasta dlja zubov (20,000)

N+RA/RA+N

P. iii. *pasta zęb-ow-a

Čz. zub-n-í pasta

R. zub-n-aja pasta (3 m.)

760

III. Units and processes in word-formation II: Special cases Szymanek (2009: 467) simply states: “[…] certain functions that are served by compounding in other languages tend to be realized by syntactic, inflectional and/or derivational means in Polish”. At the same time he points out that the designations which he classified as “fixed nominal phrases” like, for instance, drukarka laserowa (N+RA) ‘laser printer’, are, according to other Polish authors (e.g., Jadacka 2006), considered as juxtapositions (P. zestawienia), which is “a special type of a generally conceived category of compounding”. Are there certain regularities regarding the choice of the designation method depending on the relationships that may exist between head and modifier? On the basis of the examples presented in Table 43.2, we can first understand Isačenko’s differentiation between Polish as “genitival type” and “prepositional type” and Czech as “adjectival type” (Isačenko 1958: 340). Russian, however, does not clearly appear as “genitival type” in the above examples. As will be demonstrated later on, also in Russian or Czech, general (nonspecific) relations or purposive relations, as in b. and c., are likewise not expressed by compounds. But what then about the Polish and Czech equivalents corresponding to the Russian examples, provided in Table 43.1, which additionally exhibit “object relations” between the constituents? Equally, the Polish equivalent here is never a compound: designations of the type N+Ngen probably represent the only equivalents (e.g., budowa statków ‘ship building’), and in those cases where a purposive relation is expressed and the formation of N+RA is possible, this is less frequently attested, e.g., mechanik statków (11,000) vs. mechanik statkowy (3,150) ‘ship mechanic’. Similarly there are no compounds in Czech, but RA+N-equivalents (lodní mechanik ‘ship mechanic’, lodní motor ‘ship engine’) and besides N+Ngen.pl (stavba lodí ‘ship building’) a synonymous abstract noun: loďař-ství (← loďař ‘shipbuilder’ ← loď ‘ship’). This does not, however, imply that there are no compounds at all: the documented types nevertheless need to be analysed with respect to the relationships realized between the constituents. And in light of Szymanek’s statement: “However, there is a wealth of evidence to demonstrate that nominal compounding is, relatively speaking, a living process in Polish, too”, it remains to be examined which tendencies emerge in the formation of neologisms compared to the established compounds. This question should also be considered in regard to Czech. As a basis for the following discussion we will focus our attention on Russian which, in adapting more recent borrowings, appears to be especially open to compounds.

2. Relations between modifier and head in the traditional Slavic compounds In the following we will confine ourselves to a brief outline of the formation of determinative noun-noun compounds (with and without the linking vowel o/e, particularly in recent calques), because only here can we establish parallels to word combinations consisting of RA+N (possibly also N+Ngen or N+Nprep.case). Referring to English nominal compounds, Plag (2009: 148) presents different possibilities of classifying semantic relations between constituents, as, e.g., ‘location’, ‘cause’, ‘possessor’, ‘material’, ‘instrument’, whilst considering such attempts as “somewhat fu-

43. Compounds and multi-word expressions in Slavic tile”. One should rather ask the question as to which interpretations are possible, since N+N-compounds are basically ambiguous and can take on different meanings depending on the context (cf. contributions to the topic in Lieber and Štekauer 2009). This evaluation of English might also be true of German, yet cannot be applied to compounding in Slavic languages. Plag’s observation concerning the description of isolated compounds in English (i.e. without context) − which are usually interpreted in terms of a typical relationship between the constituents − is, however, more likely to apply to Slavic compounds. Slavic grammars exclusively proceed from isolated compounds, provided this question is at all subject to discussion. This relates to the restrictions on the formation of compounds as well as to the limited, if not impossible, textually dependent semantic interpretations. Consequently, none of the German compounds with a nominal modifier presented in article 59 on schemata and semantic roles in word-formation in Table 59.1 “Types of noun-noun and verb-noun compounds in German” can have a compound equivalent in Russian, Polish or Czech. In Slavic languages the formation of compounds including a verbal modifier is impossible.

2.1. N+N-compounds In its description of determinative N+N-compounds (e.g., R. zvuk-o-režisser ‘sound operator’, most formations including a deverbal head noun, e.g., R. dač-e-vladelec ‘dacha owner’, tepl-o-otdača ‘heat emission’), the “Russian Grammar” (Russkaja grammatika 1980: 242 ff.) completely ignores the semantic relations between the constituents. In “pure compounds” of the N+N-type, the “Czech Grammar” (Mluvnice češtiny 1986: 456 ff.) distinguishes between: a) compounds whose heads do not denote actions, e.g., Cz. vod-o-znak ‘watermark’, dřev-o-průmysl ‘wood(working) industry’, including numerous technical terms as well as older loan translations (-o- is the linking vowel). What is very revealing in this context is that some compounds marked as archaic have been replaced by RA+N (par-o-stroj ‘steam engine’ > parní stroj ‘id.; lit. steam-RA machine’), by a prepositional phrase (kav-o-mlýnek ‘coffee mill’ > mlýnek na kávu ‘id.; lit. mill-DIM for coffee’) or by a derivative (par-o-loď ‘steamboat’ [> parní loď lit. ‘steam-RA boat’] > parník ‘steamer’), see also section 5; b) compounds whose heads denote an action (mostly action nouns), including calques, e.g., Cz. jazyk-o-věda ‘linguistics; lit. language-o-science’, těl-o-výchova ‘physical education; lit. body-o-education’, vod-o-lečba ‘water cure’, beside vodní lečba lit. ‘water-RA cure’. Complex words comparable to Czech formations belonging to the group described under b) are, according to the “Grammar of Modern Polish” (Gramatyka 1998: 457 ff.), among the most frequently attested patterns of determinative compounds in Polish. The syntactic relations between the constituents are taken as a basis for the description of compounds representing the structure “N(N1+N2)” (with the linking vowel -o-), e.g.: “N1 denotes the object of the action performed by N2” (P. mit-o-twórca ‘myth-o-creator’, projekt-odawca ‘project developer; lit. project-o-giver’)”; “N1 = medium, instrument of the action

761

762

III. Units and processes in word-formation II: Special cases Tab. 43.3: Russian compounds and synonymous RA+N combinations and their equivalents in Polish and Czech Russian Compound

Polish, Czech RA+N

Compound, derivative, RA+N/N+RA, N+Ngen

N+Nnon-derived

R. zvuk-o-režisser (2 m.) ‘sound editor’

R. zvukovoj režisser (600)

P. N+RA operator dźwiękowy (← dźwięk ‘sound’) (321) P. N+Ngen operator dźwięku (23,000) Cz. derivative zvukař (← zvuk ‘sound’)

N+N scrip-tor); possessor, plausor and tonsor result from possed-tor, plaudtor and tond-tor by a series of events, which are assumed to be the following: voicing assimilation (/d+t/ > /t+t/); /s/ epenthesis to break the /t+t/ sequence; assimilation of the /t/s on both sides to the epenthetic /s/; degemination of /ss/ in certain phonotactic contexts (after sonorants, long vowels and diphthongs). Synchronically a number of facts remain unclear under this analysis (even in the cases for which it would be possible to give a diachronic account): for example, the contrast between cubitor and scriptor (why

784

III. Units and processes in word-formation II: Special cases not *scribitor or *cuptor?). Furthermore, the form of the verb chosen by one and the same suffix appears to vary between a stem with thematic vowel (ama-), the bare root (scrib-, etc.), and a string lacking a predesinential -e- which appears in the present tense forms (tond- vs. tonde-; in the case of possed- vs. posside-, the stem that is assumed to be the base of the -tor derivative differs from the one that surfaces in present tense forms also in the quality of a stem-internal vowel). All these problems disappear if one assumes synchronic derivation of the agent nouns from third stems, by way of a vowel initial suffix -or (and likewise for all the other processes).

2.3. Other processes possibly based on the third stem Further derivational processes that might be argued to be based, at least in some cases, on the third stem are those listed in (5): (5)

4th declension masc. nominalizations without overt affix feminine 3rd declension agent nouns in -rīx Adjectives in -īvus, -a, -um Adjectives in -ōrius, -a, -um Adverbs in -im Adjectives in -ibilis

sensus ‘feeling’ victrix ‘conqueress’ activus ‘active’ decretorius ‘decisive’ cursim ‘quickly, hastily’, corruptibilis ‘liable to decay’

In addition, through a process whose exact origin has been fiercely debated (see Georges 1968 for a critical review), deverbal action nouns homophonous with the feminine form of the past participle, such as expēnsa, remissa, developed in early Romance, and are continued in Italian (cf. section 3.11). Each of these derivational categories deserves comment.

2.4. 4th declension masculine nominalizations The productivity of 4th declension masculine nominalizations based on the third stem and lacking an overt derivational affix in Classical Latin is debated (Georges 1968: 376− 380; Panagl 1987); in Italian some of these nouns are continued (senso, frutto, atto …), but they are not homophonous with the past participles of the corresponding verbs (sentito vs. senso from sentire ‘to feel’, agito vs. atto from agire ‘to act’).

2.5. Feminine agent nouns in -rīx Feminine agent nouns in -rīx (about 350 in Lewis and Short 1879) are sometimes considered to be derived from the corresponding masculine noun by suffix substitution. In most cases, such an analysis is empirically undistinguishable from one assuming that -rix is

44. Paradigmatically determined allomorphy suffixed to the third stem: substitution of -(t)rix for -(t)or or suffixation of -rix to the third stem give the same result in cases such as victor/victrix ‘conqueror/conqueress’. Three derivatives appear to contain a stem which is phonologically slightly distinct from the base verb’s third stem: genetrix ‘mother’ vs. genit- (gigno ‘to beget, bear, bring forth, produce’), meretrix ‘she who earns money; hence, a prostitute’ vs. merit- (mereo ‘to deserve’), obstetrix ‘midwife’ vs. obstat- (obsto ‘to stand before or against sth.’) or obstit- (obsisto ‘to place one’s self before sth.’). Although the by-forms genitrix and obstitrix are also attested, albeit less frequently than genetrix and obstetrix, meretrix has no *meritrix counterpart. These three forms would also be problematic for an analysis assuming that the masculine agent noun is the base, as there is no corresponding masculine for meretrix and obstetrix (i.e. “bei nur weiblichen Berufen” [with jobs that are only feminine] according to Leumann 1977: § 329.3), and the masculine noun corresponding to genetrix is genitor. It must be observed that a genet- stem appears also in genetivus ‘genitive’ (which has a less frequent by-form genitivus). Whatever the reason for this irregularity, it appears in three nouns that designate almost exhaustively the roles traditionally assigned to women in an ancient society, thereby forming a close network. A particularly interesting subset of nouns in this class is represented by a handful of forms ending in -strix, such as tonstrix ‘female hair-cutter or barber’, plaustrix ‘female applauder’. Unfortunately most of these nouns are attested only by grammarians or in glossaries, and are therefore of a dubious nature; however, at least tonstrix, suppostrix ‘she that fraudulently exchanges or substitutes’, ambestrix ‘female consumer’ and persuastrix ‘female persuader’ are accepted lectiones in the works of Plautus (although, since Plautus is well known for the humorous character of his language, Latinists usually dismiss these forms as insignificant). Several analyses can be offered for these formations. Those who assume the suffix to be -trix observe that when /t/ is followed by /r/ it doesn’t assimilate to the epenthetic /s/ that arises between a stem-final /d/ or /t/ and the /tr/ cluster: therefore, tond-trix regularly develops to tonstrix (presumably by way of the following stages: tond-trix > tont-trix > tont-s-trix, tons-s-trix > tonstrix). Assuming instead that the suffix is -rix and the base is the third stem requires an explanation for the appearance of /t/ between base and suffix. Such an explanation can in fact be found in Priscian’s (1999) Partitiones. Priscian assumes that the suffix is -rix, and that the feminine agent nouns are derived by substituting -rix for the -or suffix found in the masculine nouns, in his account derived from the past participle; if the masculine ends in -sor, a feminine derivative is better avoided “propter asperitatem pronuntiationis” [because of the harshness of pronunciation]. If it is absolutely necessary to form a -rix derivative from a stem ending in /s/, the /sr/ cluster is broken by inserting a /t/: omnia participia praeteriti temporis us in or conuertentia faciunt nomen uerbale in omni coniugatione masculinum ex quo iterum or in rix mutantes facimus femininum, nisi euphonia, id est sonus, prohibeat, quod euenit in illis quae in sor desinunt ut pransor cursor tonsor. nemo enim dicit pransrix cursrix tonsrix propter asperitatem pronuntiationis. unde et Terentius tonstrina dixit euphoniae causa addens contra regulam t. sicut enim a doctore doctrina consonantes eas habuit quas suum primitiuum, sic debuit etiam tonstrina absque t esse nisi sonoritas coegisset. «defenstrix» quoque Cicero in Timaeo protulit addita t. (Keil III, 463, 11−20; Passalacqua 1999, 51, 13−22) [All perfect participles, by converting -us to -or, make a masculine verbal noun in every conjugation, from which in turn, by changing -or to -rix, we make a feminine [noun], unless euphony, i.e. sound prevents it, which happens in those nouns which end in -sor, such as

785

786

III. Units and processes in word-formation II: Special cases pransor, cursor, tonsor. For no one says pransrix, cursrix, tonsrix on account of the harshness of [their] pronunciation. This is why Terence says tonstrina, adding -t- against the rule for the sake of euphony. For just as doctrina has gotten from doctor those consonants which its base has, so ought tonstrina also have been without a -t- were it not for the fact that sound had made it necessary. Also by adding a -t- Cicero produced the word defenstrix in the Timaeus.]

Indeed, /sr/ is a universally disfavoured sequence (Passino 2007), and almost unattested in Latin (the only words containing it listed by Lewis and Short 1879 are disraro, disrumpo, Cisrhenanus, Transrhenanus and Israel). Insertion of an intrusive stop to enhance syllable contact is a well-attested repair strategy (e.g., in Old French essre [< Latin *essere] developed to estre ‘to be’ [Modern French être]). This repair strategy is only very marginally available in Italian, as will be shown in section 3.10.

2.6. Adjectives in -īvus Adjectives in -īvus (about 280 in Lewis and Short 1879) can take as bases nouns, verb roots, or verb third stems: examples are aestīvus ‘pertaining to summer’, internecīvus ‘deadly, murderous’, and actīvus ‘active’. The majority of derivatives are from third stems, particularly from 1st conjugation verbs (125 items in -atīvus); many derivatives are metalinguistic terms; often the adjectives have been nominalized after ellipsis of a head noun, as in the names of cases (genetīvus, datīvus, accūsātīvus, ablātīvus, all attested from Quintilian onwards).

2.7. Adjectives in -ōrius Adjectives ending in -ōrius (about 350 in Lewis and Short 1879) are for the most part derived from agent nouns in -ōr by way of the suffix -ius, as in the cases of praetorius, quaestorius, censorius, etc. However, in some cases a corresponding agent noun in -ōr is not attested, and derivation from a verb’s third stem is the only possible analysis, as in the cases of decretorius ‘decisive’ (← decerno, decret- ‘to decide’), meritorius ‘that brings in money’ (← mereo, merit- ‘to earn’). In other cases, even if a corresponding agent noun in -or is attested, an analysis that has the -orius adjective based on the verb is semantically more plausible. This is often the case with nominalized adjectives resulting in neuter nouns in -orium that designate places where an activity is carried out, such as deversorium ‘inn, lodging place’, more likely from deverto, devers- ‘to lodge’ than from the hapax deversor ‘lodger’.

2.8. Adverbs in -im Adverbs in -im have been extensively studied (cf. Ricca 2010 and references therein, particularly Schaffner-Rimann 1958). It is commonly accepted that the suffix originated through the reanalysis of an originally inflectional ending (considered to be an accusative

44. Paradigmatically determined allomorphy ending by most scholars, but an instrumental according to Bonfante 1937). According to Schaffner-Rimann (1958) the reanalysis was favoured by the occurrence of forms such as partim in appositional contexts, a construction which appears throughout the history of the language. She cites examples such as cives partim expellere, partim interficere (Nepos, Thras. 1,5). Partim quickly became an invariable noun, used also in contexts where a nominative would have been expected, and finally it was reanalyzed as an adverb (for example, in contexts such as nostri … partim fugientes ab equitatu interficiuntur, partim integri procumbunt (Caesar, Bell. civ. 2,42,2)). Once -im had been reanalyzed as an adverbial suffix, some deverbal adverbs (or rather, converbs) were formed based on the third stem, such as cursim, passim (e.g., Plautus, Poen. 567 hoc cito et cursim est agendum). Through a further reanalysis of derivatives from 1st conjugation verbs, such as nominatim ‘by name’, that could be analyzed as a denominal derivative from nomen, -inis ‘name’ as well as a derivative from the verb nominare ‘to call by name’ based on its third stem, a suffix -atim was isolated and applied to nominal bases to form distributive adverbs such as centuriatim ‘by companies’. None of these processes survives in Italian; unadapted Latin borrowings such as passim, verbatim are used in academic or bureaucratic writing.

2.9. Adjectives in -bilis Adjectives in -bilis (over 700 in Lewis and Short 1879) were mostly formed from the present stem (always in the case of 1st conjugation verbs, which represent over two thirds of the bases: admirabilis, amabilis, execrabilis …). Some derivatives from verbs of other conjugations seem to be based on the present root rather than its stem, showing no theme vowel: solubilis, volubilis … Possibilis, also from a present root, has an i between root and suffix that can be analyzed as either epenthetic or analogical (legunt : legibilis = possunt : possibilis, according to Leumann 1977: § 312), as do a number of derivatives based on third stems, such as corruptibilis, destructibilis, sensibilis, visibilis … The suffix seems to be able to select both present stems (or roots) and third stems as bases, and quasi-minimal pairs such as inerumpibilis ‘not to be torn’ vs. corruptibilis ‘liable to decay’ exist. Steriade (forthc.) has investigated these derivatives thoroughly alongside the related derivatives in -ilis, and has observed that a syntactico-semantic condition on the selection of the third stem (in her terms, the “t-stem”) holds: derivatives in -bilis and -ilis that select a participial (also known as t- / third) stem “refer only to arguments that corresponding t-participles [i.e. perfect passive participles] refer to” (Steriade forthc. § 5.2). These arguments are “passive subjects (e.g., duct-ibilis ‘leadable’; ductus ‘led’), and active subjects of intransitive or deponent verbs (e.g., pass-ibilis ‘who can suffer’; passus ‘having suffered’)” (Steriade forthc. § 5). In Steriade’s view, then, the selection of participial / t- / third stems to form derivatives with -bilis is not due to an arbitrary morphological (morphomic) condition, but reflects “the passive syntax of t-participles” (Steriade forthc. § 5). No such condition holds for -bilis derivatives formed on the root or the present (infectum) stem: “Root and infectum-based forms can refer to any argument of the verb that occupies its subject or object position, while forms containing the t-stem

787

788

III. Units and processes in word-formation II: Special cases refer only to arguments that corresponding t-participles refer to: e.g., ducibilis means both ‘who can lead’ and ‘leadable’ while ductibilis can only mean ‘leadable’” (Steriade forthc. § 5.2). Steriade’s hypothesis is that the relation “between pairs like ductus-ductibilis, passus-passibilis is the default relation between any base and its derivative: derivatives inherit both the phonology and the syntactic and semantic structure of the bases they contain” (Steriade forthc. § 5). In other words, she rejects the hypothesis that semantically void, purely morphomic, bases operate in derivation.

2.10. An alternative analysis Steriade’s account, by which the semantics of -bilis derivatives from participial stems is a function of the passive syntactic content of their bases, of course begs the question of why we don’t find similar semantic properties in other derivatives from participial stems, such as agent nouns in -(t)or. Her answer is the following: the derivatives in (2) are not formed by concatenating a verb’s third stem with a vowel-initial suffix, as in Aronoff’s (1994) analysis; rather, they are formed by selecting a verb’s root or infectum / present stem and a t-initial suffix, in which this initial consonant is a “stem extension”. The structure of a Latin deverbal derivative, in Steriade’s analysis, is the following: [[[[root] + buffer or theme vowel] + first C of a derivational suffix] + the rest of the derivational suffix]; an example of this structure is audītor, to be analyzed as [[[[aud] + ī] t] + or]. The string up to and including the initial consonant of the suffix (e.g., [[[aud] + ī] t]) is called a “minimal stem”. A derivative such as pulsor would have the structure [[[[pell] + i] t] + or]. But why, then, doesn’t it surface as *pellitor? According to Steriade, this is due to the workings of a series of correspondence constraints, by which, if two minimal stems headed by the same verb root end in homorganic [α sonorant] segments, they stand in correspondence, and must merge in order to surface as identical. Minimal stems such as puls- (the participial stem) and pellit- (the stem of the derivatives with t-initial suffixes), both from the verb pello ‘to push, drive away’ (i.e., “headed by the same verb root”), meet these criteria as they both end in a coronal obstruent; pellit- is then changed to puls- to obey the correspondence requirement. The change doesn’t happen the other way around (i.e. we don’t get *pellitus as a past participle) because “the direction of the merger is determined by correspondence to a listed word form” (§ 4.1), in this case puls-. Listed word forms are assumed to be those “deviating unpredictably from regular trends” (§ 4.1). Such an account explains the lack of inheritance of the passive past participle’s syntactic features in the derivatives in (2): the appearance of the participial / third / t- stem in these derivatives is not due to its selection by lexeme-formation rules, but “is the consequence of a phonological chain of events taking place, presumably, in a postsyntactic component” (Steriade forthc. § 4.5). Steriade’s analysis shifts the arbitrariness from the morphological component to the phonological one: the specification of which entities stand in correspondence to one another, and under which conditions this occurs, is as arbitrary and language-specific as the specification of which lexeme-formation processes select a given morphomic stem. Further level-internal (phonological or morphological) regular “laws” operate in each case. Steriade’s account is preferable only in models that assume no separate morpholog-

44. Paradigmatically determined allomorphy ical realization component, but aim at reducing all operations to phonology and itemand-arrangement syntax.

2.11. Denominal derivatives An important component of Steriade’s analysis is the hypothesis that the Latin suffixes involved in the derivational processes in (2) and (5) are /t/-initial. She claims that further evidence for this is given by occasional denominal derivatives with these suffixes, which invariably exhibit a /t/. Her examples of agent nouns are iānitor ‘gate keeper’ (← iānua ‘gate’), funditor ‘sling fighter’ (← funda ‘sling’), fīcitor ‘fig planter’ (← fīcus ‘fig’), vindēmitor ‘grape-gatherer, vintager’ (← vindemia ‘grape harvest’), adversitor ‘one who goes to meet another’ (← adversus ‘towards’). To her list I can add āleator ‘dice player’ (← ālea ‘dice’), senātor ‘senator’ (← senex ‘old, aged’), holitor ‘kitchen gardener’ (← holus ‘kitchen or garden herbs’), vīnitor ‘vine-dresser’ (← vinum ‘wine’). Two more dubious examples are viātor ‘wayfarer, traveller’, which could derive from via ‘way, road’ or from its denominal verb vio ‘to go, travel’, and gladiātor ‘fighter in the public games, gladiator’, which could derive from gladius ‘sword’ or from a possible denominal verb *gladior, unattested (but cf. digladior ‘to fight for life and death’). Altogether, these denominal derivatives are very few, in comparison with over a thousand deverbal agent nouns in -tor or -sor. The denominal derivatives cluster into a few semantic categories (officers, such as iānitor, adversitor, senātor; agricultural workers, such as fīcitor, vīnitor, holitor, vindēmitor (note that vindēmiātor is also attested, from vindēmio ‘to gather the vintage’); fighters and gamesters, such as funditor, āleator, and gladiātor [if this is indeed denominal]). This semantic clustering is typical of analogical formations. In my view, the model of hundreds of agent nouns in -ātor and -itor derived from regular verbs is sufficient to explain the appearance of -ātor and -itor in the denominal derivatives, particularly when specific forms, such as viātor, could be analyzed both as deverbal and as denominal derivatives.

3. Italian reflexes of the Latin processes based on the third stem 3.1. Popular vs. learned Italian derivatives In Italian, most of the deverbal derivational processes which could be analyzed as based on the third stem in Latin have reflexes. Both lexemes inherited from Latin, either by direct (“popular”) transmission or by learned borrowing, and lexemes newly created in Italian are in usage. Learned borrowings display certain minor phonological adaptations, but can be distinguished from forms inherited by way of continuous transmission because they display certain phonological peculiarities. For example, truly popular continuants of derivatives in -TIONE(M) end in -/(t)tsone/: canzone < CANTIONE(M) ‘song’, stazzone < STATIONE(M) ‘stopping place’, where the glide /j/ (< Ĭ) is not preserved, but disappears after causing intervocalic gemination and affrication of /t/. Learned borrowings, instead, display affrication, gemination and the glide, as in adulazione /adulat'tsjone/ ‘flattery’