Eurotyp: 7 Noun Phrase Structure in the Languages of Europe 9783110197075, 9783110157482

Citation preview

Noun Phrase Structure in the Languages of Europe

≥

Empirical Approaches to Language Typology Eurotyp

Editors Georg Bossong Bernard Comrie

Mouton de Gruyter Berlin · New York

20-7

Noun Phrase Structure in the Languages of Europe

edited by Frans Plank

Mouton de Gruyter Berlin · New York

2003

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines of the 앪 ANSI to ensure permanence and durability.

ISBN 3-11-015748-9 Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at ⬍http://dnb.ddb.de⬎.  Copyright 2003 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Typesetting: Christoph Eyrich, Berlin. Printing: Hubert & Co, Göttingen. Binding: Lüderitz & Bauer, Berlin. Printed in Germany.

✐ v

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #1

✐

✐

General Preface

The present volume is one of a series of nine volumes in which the results of the European research project “Typology of Languages in Europe” (EUROTYP) are published. The initiative for a European project on language typology came from a proposal jointly submitted to the European Science Foundation (ESF) by Johannes Bechert (University of Bremen), Claude Buridant (University of Strasbourg), Martin Harris (University of Salford, now University of Manchester) and Paolo Ramat (University of Pavia). On the basis of this proposal and following consultations with six experts the Standing Committee for the Humanities of the ESF decided to organize a workshop (Rome, January 1988), in which this idea was further explored and developed. The results of this workshop (published by Mouton, 1990) were sufficiently encouraging for the Standing Committee to appoint a preparatory committee and entrust it with the tasks of drawing up a preliminary proposal, of securing interest and participation from a sufficiently large number of scholars and of finding a suitable programme director. The project proposal formulated and sent out by Simon Dik (University of Amsterdam) as chair of this committee met with very supportive and enthusiastic reactions, so that the Standing Committee for the Humanities recommended the funding of a planning stage and the General Assembly of the ESF approved a year zero (1989) for an ESF Programme in Language Typology. During this planning phase all major decisions concerning the management structure and the organisation of the work were taken, i.e., the selection of a programme director, the selection of nine focal areas around which the research was to be organized, the selection of a theme coordinator for each theme and the selection of the advisory committee. The first task of the programme director was to draw up a definitive project proposal, which was supplemented with individual proposals for each theme formulated by the theme coordinators, and this new proposal became the basis of a decision by the ESF to fund the Programme for a period of five years (1990–1994). Language typology is the study of regularities, patterns and limits in crosslinguistic variation. The major goal of EUROTYP was to study the patterns and limits of variation in nine focal areas: pragmatic organization of discourse, constituent order, subordination and complementation, adverbial constructions, tense and aspect, noun phrase structure, clitics and word prosodic systems in the languages of Europe. The decision to restrict the investigation to the languages of Europe was imposed for purely practical and pragmatic reasons. In the course of the project an attempt was made, however, to make as much sense of this restriction as possible, by characterizing the specific features of European languages against the background of

✐

✐ ✐

✐

✐ vi

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #2

✐

✐

vi

General Preface

non-European languages and by identifying areal phenomena (Sprachbünde) within Europe. More specifically, the goals of the EUROTYP project included the following: – to contribute to the analysis of the nine domains singled out as focal areas, to assess patterns and limits of cross-linguistic variation and to offer explanations of the patterns observed. – to bring linguists from various European countries and from different schools or traditions of linguistics together within a major international project on language typology and in doing so create a new basis for future cooperative ventures within the field of linguistics. More than 100 linguists from more than 20 European countries and the United States participated in the project. – to promote the field of language typology inside and outside of Europe. More specifically, an attempt was made to subject to typological analysis a large number of new aspects and domains of language which were uncharted territory before. – to provide new insights into the specific properties of European languages and thus contribute to the characterization of Europe as a linguistic area (Sprachbund). – to make a contribution to the methodology and the theoretical foundations of typology by developing new forms of cooperation and by assessing the role of inductive generalization and the role of theory construction in language typology. We had a further, more ambitious goal, namely to make a contribution to lingustic theory by uncovering major patterns of variation across an important subset of languages, by providing a large testing ground for theoretical controversies and by further developing certain theories in connection with a variety of languages. The results of our work are documented in the nine final volumes: Pragmatic Organization of Discourse in the Languages of Europe (edited by G. Bernini) Constituent Order in the Languages of Europe (edited by A. Siewierska) Subordination and Complementation in the Languages of Europe (edited by N. Vincent) Actance et Valence dans les langues d l’Europe (edited by J. Feuillet) Adverbial Constructions in the Languages of Europe (edited by J. van der Auwera) Tense and Aspect in the Languages of Europe (edited by Ö. Dahl) Noun Phrase Structure in the Languages of Europe (edited by F. Plank) Clitics in the Languages of Europe (edited by H. van Riemsdijk) Word Prosodic Systems in the Languages of Europe (edited by H. van der Hulst) In addition, the EUROTYP Project led to a large number of related activities and publications, too numerous to be listed here. At the end of this preface, I would like to express my profound appreciation to all organizations and individuals who made this project possible. First and foremost, I must mention the European Science Foundation, who funded and supported the

✐

✐ ✐

✐

✐ vii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #3

✐

✐

General Preface

vii

Programme. More specifically, I would like to express my appreciation to Christoph Mühlberg, Max Sparreboom and Geneviève Schauinger for their constant and efficient support, without which we would not have been able to concentrate on our work. I would, furthermore, like to thank my colleague and assistant, Martin Haspelmath, and indeed all the participants in the Programme for their dedication and hard work. I finally acknowledge with gratitude the crucial role played by Johannes Bechert and Simon Dik in getting this project off the ground. Their illness and untimely deaths deprived us all of two of the project’s major instigators. Ekkehard König, Programme Director

✐

✐ ✐

✐

✐ viii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #4

✐

✐

✐

✐ ✐

✐

✐ ix

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #5

✐

✐

Contents Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiii Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxv I.

Introduction

Frans Plank Noun phrase structure: An und für sich, in time, and in space . . . . . . . . . II.

3

On inflection

Aleksandr E. Kibrik Nominal inflection galore: Daghestanian, with side glances at Europe and the world . . . . . . . . . . .

37

Edith Moravcsik Inflectional morphology in the Hungarian noun phrase: A typological assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 Frans Plank The selective elaboration of nominal or pronominal inflection . . . . . . . . . 253 Greville G. Corbett Types of typology, illustrated from gender systems . . . . . . . . . . . . . . 289 III.

On (over-)determination

Frans Plank Double articulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 Edith Moravcsik Non-compositional definiteness marking in Hungarian noun phrases . . . . . 397 David Gil English goes Asian: Number and (in)definiteness in the Singlish noun phrase . . . . . . . . . . . 467

✐

✐ ✐

✐

✐ x

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #6

✐

✐

x

Contents

Maria Koptjevskaja-Tamm A woman of sin, a man of duty, and a hell of a mess: Non-determiner genitives in Swedish . . . . . . . . . . . . . . . . . . . . . . 515 IV.

On amplification

James R. Hurford The interaction between numerals and nouns . . . . . . . . . . . . . . . . . . 561 Maria Koptjevskaja-Tamm Possessive noun phrases in the languages of Europe . . . . . . . . . . . . . . 621 Maria Koptjevskaja-Tamm Action nominal constructions in the languages of Europe . . . . . . . . . . . 723 Leon Stassen Noun phrase conjunction: The coordinative and the comitative strategy . . . . 761 Indexes Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 821 Language index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 827 Author index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 837

✐

✐ ✐

✐

✐ xi

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #7

✐

✐

Contents detailed Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiii Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxv I.

Introduction

Frans Plank Noun phrase structure: An und für sich, in time, and in space 1. Us . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Typology by itself . . . . . . . . . . . . . . . . . . . . . 3. Typology and diachrony . . . . . . . . . . . . . . . . . . 4. Typology and geography . . . . . . . . . . . . . . . . . . Appendix: Further publications of the EUROTYP Noun Phrase Group Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . II.

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

3 3 13 18

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

22 27 30

On inflection

Aleksandr E. Kibrik Nominal inflection galore: Daghestanian, with side glances at Europe and the world 1. Introducing the Daghestanian family . . . . . . . . . . . . . . . . . 2. Categories of nominal inflection . . . . . . . . . . . . . . . . . . . 2.1. Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Grammatical cases . . . . . . . . . . . . . . . . . . . . . . . 2.2.1. Core cases . . . . . . . . . . . . . . . . . . . . . . 2.2.2. Peripheral cases . . . . . . . . . . . . . . . . . . . . 2.2.3. European analogies . . . . . . . . . . . . . . . . . . 2.2.4. Assorted analogies outside Europe . . . . . . . . . . 2.3. Spatial forms . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1. Localization . . . . . . . . . . . . . . . . . . . . . . 2.3.2. Direction . . . . . . . . . . . . . . . . . . . . . . . 2.3.3. Uralic analogies . . . . . . . . . . . . . . . . . . . 2.3.4. Turkic analogies . . . . . . . . . . . . . . . . . . . 2.3.5. Assorted analogies outside Europe . . . . . . . . . .

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

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

37 38 38 38 38 39 41 42 43 43 44 46 48 48

✐

✐ ✐

✐

✐ xii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #8

✐

✐

xii

Contents detailed

2.4. Double case marking . . . . . . . . . . . . . . . . . . . . 2.4.1. Attributes with head present . . . . . . . . . . . 2.4.2. Attributes with head missing . . . . . . . . . . . 2.5. Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5.1. Classification . . . . . . . . . . . . . . . . . . . 2.5.2. Class agreement . . . . . . . . . . . . . . . . . 3. Systems of nominal inflection . . . . . . . . . . . . . . . . . . 3.1. Word classes and paradigm types . . . . . . . . . . . . . 3.1.1. Inflectional peculiarities of pronouns of 1st and 2nd person . . . . . . . . . . . . . . . . . . . . 3.1.2. Ordinary vs. attributive declension . . . . . . . . 3.2. Patterns of stem formation . . . . . . . . . . . . . . . . 3.2.1. The two-stem pattern . . . . . . . . . . . . . . . 3.2.2. The one-stem pattern . . . . . . . . . . . . . . . 3.2.3. Intermediate patterns . . . . . . . . . . . . . . . 3.2.4. The pure oblique stem strategy . . . . . . . . . . 3.2.5. The unmarked/marked root opposition . . . . . . 3.2.6. The modules of stem formation . . . . . . . . . 3.3. Declensions . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1. Co-existing stem formation patterns . . . . . . . 3.3.2. Diversity of stem formatives . . . . . . . . . . . 3.3.3. Paradigmatic versatility . . . . . . . . . . . . . 4. How nominal and how inflectional is nominal inflection? . . . . 4.1. Similarities of nouns to other word classes . . . . . . . . 4.1.1. The inflection of postpositions . . . . . . . . . . 4.1.2. Spatial adverbs . . . . . . . . . . . . . . . . . . 4.1.3. Spatial converbs . . . . . . . . . . . . . . . . . 4.1.4. Temporal converbs . . . . . . . . . . . . . . . . 4.1.5. Possessive locatives . . . . . . . . . . . . . . . 4.1.6. Cases or postpositions? . . . . . . . . . . . . . . 4.2. Inflection or derivation? . . . . . . . . . . . . . . . . . . 4.2.1. Spatial forms . . . . . . . . . . . . . . . . . . . 4.2.2. Genitive or attributive? . . . . . . . . . . . . . . 4.2.3. A motive for inflectional or derivational leanings 5. Using nominal inflection . . . . . . . . . . . . . . . . . . . . . 5.1. Frequency . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Variation . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3. Paradigm gaps . . . . . . . . . . . . . . . . . . . . . . . 5.3.1. Structural gaps . . . . . . . . . . . . . . . . . . 5.3.2. Semantic gaps . . . . . . . . . . . . . . . . . . 5.3.3. Closing gaps . . . . . . . . . . . . . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

50 50 51 53 53 53 55 55

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

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

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

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

56 58 60 60 61 62 63 64 66 68 68 69 70 72 72 72 72 73 73 74 75 75 75 83 85 86 86 86 87 87 88 91

✐

✐ ✐

✐

✐ xiii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #9

✐

✐

xiii

Contents detailed

5.4. Appendix: Semantic classes of nouns with defective number . 6. Tracing the history of nominal inflection . . . . . . . . . . . . . . . 6.1. Present traces . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.1. Evidence from pronouns . . . . . . . . . . . . . . . 6.1.2. Evidence from obliques . . . . . . . . . . . . . . . 6.1.3. Evidence from the plural . . . . . . . . . . . . . . . 6.1.4. Evidence from cases . . . . . . . . . . . . . . . . . 6.1.5. Evidence from localizations . . . . . . . . . . . . . 6.1.6. Evidence from spatial forms . . . . . . . . . . . . . 6.2. Origins and progress . . . . . . . . . . . . . . . . . . . . . . 6.2.1. Proto-Daghestanian . . . . . . . . . . . . . . . . . . 6.2.2. After separation . . . . . . . . . . . . . . . . . . . . 7. Afterword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Edith Moravcsik Inflectional morphology in the Hungarian noun phrase: A typological assessment 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2. Synopsis of the inflections . . . . . . . . . . . . . . . . . 2. Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Meaning and form . . . . . . . . . . . . . . . . . . . . . 2.1.1. Meanings . . . . . . . . . . . . . . . . . . . . . 2.1.2. Forms . . . . . . . . . . . . . . . . . . . . . . . 2.1.3. Zero meanings, zero forms . . . . . . . . . . . . 2.1.4. Synonymy . . . . . . . . . . . . . . . . . . . . 2.1.5. Homonymy . . . . . . . . . . . . . . . . . . . . 2.1.6. Cumulative exponence . . . . . . . . . . . . . . 2.1.7. Extended exponence . . . . . . . . . . . . . . . 2.1.8. Non-compositionality . . . . . . . . . . . . . . 2.2. Distribution . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1. Cooccurrence . . . . . . . . . . . . . . . . . . . 2.2.1.1. Choice of morphemes . . . . . . . . . . . . . . 2.2.1.2. Choice of allomorphs . . . . . . . . . . . . . . 2.2.2. Linear order . . . . . . . . . . . . . . . . . . . . 2.2.3. Bonding . . . . . . . . . . . . . . . . . . . . . . 3. Typological assessment . . . . . . . . . . . . . . . . . . . . . . 3.1. Meaning and form . . . . . . . . . . . . . . . . . . . . . 3.1.1. Meanings . . . . . . . . . . . . . . . . . . . . .

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

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

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

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

92 97 98 98 99 99 100 100 101 101 101 103 103 104 105

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

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

113 113 116 119 119 119 120 130 132 135 139 144 150 153 153 154 163 174 177 180 180 180

✐

✐ ✐

✐

✐ xiv

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #10

✐

✐

xiv

Contents detailed

3.1.2. Forms . . . . . . . . . . . . . . . . . . . 3.1.3. Zero meanings, zero forms . . . . . . . . 3.1.4. Synonymy . . . . . . . . . . . . . . . . 3.1.5. Homonymy . . . . . . . . . . . . . . . . 3.1.6. Cumulative exponence . . . . . . . . . . 3.1.7. Extended exponence . . . . . . . . . . . 3.1.8. Non-compositionality . . . . . . . . . . 3.2. Distribution . . . . . . . . . . . . . . . . . . . . 3.2.1. Cooccurrence . . . . . . . . . . . . . . . 3.2.2. Linear order . . . . . . . . . . . . . . . . 3.2.3. Bonding . . . . . . . . . . . . . . . . . 4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . Appendix: Summary list of generalizations about Hungarian nominal inflections . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . .

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

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

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

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

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

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

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

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

187 188 190 191 201 205 209 211 211 223 227 228

. . . . . . . . 232 . . . . . . . . 242 . . . . . . . . 246

Frans Plank The selective elaboration of nominal or pronominal inflection 1. Variable extents and domains of inflection . . . . . . . . . 2. Interdependency in categorial elaboration across domains . 3. Number and person . . . . . . . . . . . . . . . . . . . . . 4. Number and distance . . . . . . . . . . . . . . . . . . . . 5. Person and distance . . . . . . . . . . . . . . . . . . . . . 6. Pronominal and nominal languages . . . . . . . . . . . . 7. Ach, Europa . . .: Verkehrte Welt . . . . . . . . . . . . . . Appendix: The sample of languages with a dual (n=205) . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

253 254 256 259 263 266 274 278 282 285

Greville G. Corbett Types of typology, illustrated from gender systems 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 2. Resolution in its typological space . . . . . . . . . . . 3. Types of gender resolution . . . . . . . . . . . . . . . 3.1. Semantic gender resolution . . . . . . . . . . . 3.2. Mixed semantic and syntactic gender resolution 3.3. Syntactic gender resolution . . . . . . . . . . . 3.4. More on the typology of resolution systems . . . 4. Types of gender assignment . . . . . . . . . . . . . . 4.1. Semantic assignment . . . . . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

289 290 295 295 303 305 308 310 310

. . . . . . . . .

. . . . . . . . .

✐

✐ ✐

✐

✐ xv

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #11

✐

✐

Contents detailed

4.2. Predominantly semantic assignment . . . 4.3. Morphological assignment . . . . . . . . 4.4. Phonological assignment . . . . . . . . 5. The relation between resolution and assignment Notes . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . III.

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

xv

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

310 311 316 317 320 329

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

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

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

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

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

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

337 338 338 340 342 347 352 353 360 363 365 369 371 374 376 377

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

378 379 382 383 383 385 386 392

On (over-)determination

Frans Plank Double articulation 1. What to expect . . . . . . . . . . . . . . . . . . . . . . . . 2. Phenomenology of excess . . . . . . . . . . . . . . . . . . 2.1. The more noun phrases, the more articles . . . . . . . 2.2. Apposition, sort of: One noun phrase or two? . . . . . 2.3. Attributives re-ordered to follow their nouns . . . . . 2.4. Highlighting adjectives in their accustomed position . 2.5. Attributives in a row . . . . . . . . . . . . . . . . . . 2.6. Adjectives pure and simple . . . . . . . . . . . . . . 2.7. Superlative adjectives . . . . . . . . . . . . . . . . . 2.8. Ordinal numerals . . . . . . . . . . . . . . . . . . . . 2.9. Blends of alternative orders due to disorderly adverbs 2.10. Definiteness and indefiniteness blended . . . . . . . . 2.11. Noun phrase downgraded to modifier . . . . . . . . . 2.12. Strong nouns on their own . . . . . . . . . . . . . . . 2.13. Emphatically indefinite . . . . . . . . . . . . . . . . 2.14. Specific reference or discourse prominence . . . . . . 2.15. Doubly (in)definite for no (good) reason other than the availability of several markers . . . . . . . . . . . . . 2.16. Sole nouns with layered (in)definiteness marking . . 2.17. Fossilized markers . . . . . . . . . . . . . . . . . . . 3. Etiology of excess . . . . . . . . . . . . . . . . . . . . . . 3.1. Seven reasons for overmarking . . . . . . . . . . . . 3.2. Favourable conditions . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Edith Moravcsik Non-compositional definiteness marking in Hungarian noun phrases 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397

✐

✐ ✐

✐

✐ xvi

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #12

✐

✐

xvi

Contents detailed

2. Undermarking . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. One nominal, no marker . . . . . . . . . . . . . . . . . . 2.2. One marker, more than one nominal . . . . . . . . . . . . 2.2.1. Coordinate structures . . . . . . . . . . . . . . . 2.2.2. Comparing coordinate and subordinate structures 2.2.3. Subordinate structures . . . . . . . . . . . . . . 2.2.3.1. Adjectival constructions . . . . . . . . . . . . . 2.2.3.2. Possessive constructions . . . . . . . . . . . . . 2.2.4. Summary . . . . . . . . . . . . . . . . . . . . . 2.3. Undermarking: Crosslinguistic parallels and rationale . . 3. Overmarking . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Co-presence of more than one article . . . . . . . . . . . 3.2. Co-presence of articles and determiners . . . . . . . . . . 3.3. Co-presence of articles and lexically marked nouns . . . . 3.4. Overmarking: Crosslinguistic parallels and rationale . . . 4. Contradictory marking . . . . . . . . . . . . . . . . . . . . . . 5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

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

405 406 407 407 410 414 414 425 434 434 438 440 441 450 456 458 461 462 464

David Gil English goes Asian: Number and (in)definiteness in the Singlish noun phrase 1. English goes to Singapore . . . . . . . . . . . . . . . . . . . . . . . . 2. Some methodological preliminaries . . . . . . . . . . . . . . . . . . . 3. Determiners and possessives . . . . . . . . . . . . . . . . . . . . . . . 4. Reification and modification . . . . . . . . . . . . . . . . . . . . . . . 5. What is a noun phrase in Singlish? . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

467 470 472 480 495 500 509

Maria Koptjevskaja-Tamm A woman of sin, a man of duty, and a hell of a mess: Non-determiner genitives in Swedish 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2. Formal features of noun phrases with genitives in Swedish 2.1. Determiner genitives vs. adjectives . . . . . . . . . 2.2. Swear genitives . . . . . . . . . . . . . . . . . . . 2.2.1. Swear genitives as noun phrases . . . . . . 2.2.2. Co-occurrence with articles . . . . . . . . 2.2.3. Co-occurrence with adjectives . . . . . . . 2.2.4. Co-occurrence with genitives/possessives .

515 517 517 519 519 519 519 520

. . . . . . . .

. . . . . . . .

. . . . . . . .

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

. . . . . . . .

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

. . . . . . . .

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

. . . . . . . .

. . . . . . . .

✐

✐ ✐

✐

✐ xvii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #13

✐

✐

xvii

Contents detailed

2.2.5. Swear genitives as adverbials . . . . . . . . . . . . . . . 2.2.6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Measure genitives . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1. Measure genitives as noun phrases . . . . . . . . . . . . 2.3.2. Co-occurrence with articles . . . . . . . . . . . . . . . 2.3.3. Co-occurrence with adjectives . . . . . . . . . . . . . . 2.3.4. Co-occurrence with genitives/possessives . . . . . . . . 2.3.5. Measure genitives outside noun phrases . . . . . . . . . 2.3.6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Inserted genitives . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1. Inserted genitives as noun phrases . . . . . . . . . . . . 2.4.2. Co-occurrence with articles . . . . . . . . . . . . . . . 2.4.3. Co-occurrence with adjectives . . . . . . . . . . . . . . 2.4.4. Co-occurrence with genitives/possessives . . . . . . . . 2.4.5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 2.5. Determiner genitives vs. non-determiner genitives vs. adjectives: Synchrony and diachrony . . . . . . . . . . . . . . . . . . . . . 3. Semantics of noun phrases with inserted genitives . . . . . . . . . . . . 3.1. Types of noun phrases with inserted genitives . . . . . . . . . . . 3.2. Definiteness of inserted genitives . . . . . . . . . . . . . . . . . 3.3. Inserted genitives vs. determiner genitives and discourse structure 4. Why both determiner and non-determiner genitives? . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV.

520 521 521 521 523 524 525 525 525 526 526 527 528 528 529 529 537 537 544 545 549 554 556

On amplification

James R. Hurford The interaction between numerals and nouns 1. Introduction . . . . . . . . . . . . . . . . . . . . . 1.1. Aims . . . . . . . . . . . . . . . . . . . . . 1.2. Organization . . . . . . . . . . . . . . . . . 1.3. Sample . . . . . . . . . . . . . . . . . . . . 1.4. Terminology . . . . . . . . . . . . . . . . . 2. Simple lexical numerals . . . . . . . . . . . . . . 2.1. Arithmetic range of simple lexical numerals 2.2. Distinct counting forms . . . . . . . . . . . 2.3. Ordinals . . . . . . . . . . . . . . . . . . . 3. Simple lexical numerals modifying nouns . . . . . 3.1. Multi-word numeral-noun constructions . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

561 561 561 562 563 563 563 564 566 567 567

✐

✐ ✐

✐

✐ xviii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #14

✐

✐

xviii

Contents detailed

3.1.1. Prepositional constructions . . . . . . . . . . . . . . 3.1.2. Numeral classifier constructions . . . . . . . . . . . 3.2. Word order . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Morphological interactions . . . . . . . . . . . . . . . . . . 3.3.1. Definiteness . . . . . . . . . . . . . . . . . . . . . . 3.3.1.1. Definiteness marked on (end of) whole noun phrase . 3.3.1.2. Definiteness marked on numeral . . . . . . . . . . . 3.3.2. Case . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2.1. Case assigned to numerals by clause-level structure . 3.3.2.2. Case assigned to noun by numeral-noun structure . . 3.3.3. Number . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3.1. Number assigned to noun by sister numeral . . . . . 3.3.3.2. Number assigned to numeral by sister noun . . . . . 3.3.3.3. Meaning of whole noun phrase determines number numeral and noun . . . . . . . . . . . . . . . . . . . 3.3.4. Gender (alias noun class) . . . . . . . . . . . . . . . 3.4. A note on “uninflected” . . . . . . . . . . . . . . . . . . . . 4. Internal structure of complex numerals . . . . . . . . . . . . . . . 4.1. Overall phrase structure and semantics . . . . . . . . . . . . 4.2. Multiplication . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1. Decimal and vigesimal bases . . . . . . . . . . . . . 4.2.2. Word order in multiplicative structures . . . . . . . . 4.2.3. 1-Deletion . . . . . . . . . . . . . . . . . . . . . . 4.2.4. Other noun-like behaviour of Ms . . . . . . . . . . . 4.3. Addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1. Single word versus several words . . . . . . . . . . 4.3.2. Absence or presence of an overt connective . . . . . 4.3.3. Low-high versus high-low word order . . . . . . . . 4.3.4. Combinations of addition-related features: Summary 4.4. Minor arithmetical operations . . . . . . . . . . . . . . . . . 5. Complex numerals modifying nouns . . . . . . . . . . . . . . . . . 5.1. Word order . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Preservation versus regularization of irregular processes . . . 5.3. Complex numerals as local or global targets, or non-targets . 5.3.1. Local targets . . . . . . . . . . . . . . . . . . . . . 5.3.1.1. Ordinalization . . . . . . . . . . . . . . . . . . . . 5.3.1.2. Definiteness . . . . . . . . . . . . . . . . . . . . . . 5.3.1.3. Case . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1.4. Gender . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2. Global targets . . . . . . . . . . . . . . . . . . . . . 5.3.2.1. Ordinalization . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

568 570 572 576 577 577 577 579 579 581 582 582 584 584 590 590 592 592 594 594 595 596 597 597 598 599 600 600 601 602 602 603 605 605 605 606 606 607 607 607

✐

✐ ✐

✐

✐ xix

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #15

✐

✐

xix

Contents detailed

5.3.2.2. 5.3.2.3. 5.3.2.4. 5.3.3. 5.3.3.1. 5.3.3.2. 5.3.3.3. 5.3.4. 5.3.5.

Case . . . . . . . . . . . . . . . . . . . . . . . . . . . Number . . . . . . . . . . . . . . . . . . . . . . . . . Gender . . . . . . . . . . . . . . . . . . . . . . . . . Natural constraints on globality . . . . . . . . . . . . Limited resources . . . . . . . . . . . . . . . . . . . . Case assigned within a numeral . . . . . . . . . . . . Gender assigned within a numeral . . . . . . . . . . . Coincidence of global processes in complex numerals Complex numerals opting out of morphological processes . . . . . . . . . . . . . . . . . . . . . . . . 5.4. Complex numerals influencing a sister noun . . . . . . . . . . 5.4.1. Whole complex numeral influences sister noun . . . . 5.4.2. One word in complex numeral influences sister noun . 5.5. Residual conflicts and puzzles . . . . . . . . . . . . . . . . . . 6. In conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maria Koptjevskaja-Tamm Possessive noun phrases in the languages of Europe 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Typically European . . . . . . . . . . . . . . . . . . . . . . . 3. Synthetic constructions . . . . . . . . . . . . . . . . . . . . . 3.1. Dependent-marking . . . . . . . . . . . . . . . . . . . 3.1.1. Prototypical genitives . . . . . . . . . . . . . . 3.1.2. Deformed genitives . . . . . . . . . . . . . . . 3.1.3. Added indexers . . . . . . . . . . . . . . . . . 3.2. Double-marking . . . . . . . . . . . . . . . . . . . . . 3.2.1. Prototypically European . . . . . . . . . . . . 3.2.2. Less frequent variants . . . . . . . . . . . . . 3.3. Head-marking . . . . . . . . . . . . . . . . . . . . . . 3.4. No marking: Juxtaposition . . . . . . . . . . . . . . . . 3.5. Compounding . . . . . . . . . . . . . . . . . . . . . . 3.6. Overview of morphological marking . . . . . . . . . . 4. Analytic constructions . . . . . . . . . . . . . . . . . . . . . 4.1. Prepositions and inflecting possessive “articles” . . . . 4.2. “Linking” pronouns . . . . . . . . . . . . . . . . . . . 4.3. Synthetic and analytic analogues . . . . . . . . . . . . 5. Cross-categorial uses of construction markers . . . . . . . . . 5.1. Other types of heads: Clauses and adpositional phrases 5.1.1. Dependent-marking and dependent-association

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

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

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

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

. . . . . . . .

609 609 610 610 610 612 613 613

. . . . . . . .

614 615 615 615 616 617 618 618

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

621 623 626 626 626 628 633 639 639 644 645 649 652 653 659 660 665 670 676 678 678

✐

✐ ✐

✐

✐ xx

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #16

✐

✐

xx

Contents detailed

5.1.1.1. Restricted use of construction markers with non-nominal heads . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1.2. Shared encoding and syncretisms . . . . . . . . . . . . 5.1.2. Head-marking . . . . . . . . . . . . . . . . . . . . . . 5.2. Other types of dependents . . . . . . . . . . . . . . . . . . . . . 5.3. Synchrony vs. diachrony . . . . . . . . . . . . . . . . . . . . . . 6. Alienable vs. inalienable possession . . . . . . . . . . . . . . . . . . . 6.1. Maltese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2. Adyghe and the Besleney dialect of Kabardian . . . . . . . . . . 6.3. Khinalug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4. Others . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maria Koptjevskaja-Tamm Action nominal constructions in the languages of Europe 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. The subject . . . . . . . . . . . . . . . . . . . . . . . . . 1.2. The sample . . . . . . . . . . . . . . . . . . . . . . . . . 2. The typology of action nominal constructions . . . . . . . . . . 3. Major types . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Type 1: Sentential . . . . . . . . . . . . . . . . . . . . . 3.2. Type 2: Possessive-Accusative . . . . . . . . . . . . . . 3.3. Type 3: Ergative-Possessive . . . . . . . . . . . . . . . . 3.4. Type 4: Nominal . . . . . . . . . . . . . . . . . . . . . . 3.4.1. Double-Possessive . . . . . . . . . . . . . . . . 3.4.2. Possessive-Adnominal . . . . . . . . . . . . . . 4. Minor and restricted types . . . . . . . . . . . . . . . . . . . . 4.1. Mixed . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Possessive-Relative . . . . . . . . . . . . . . . . . . . . 4.3. Possessive-Incorporating . . . . . . . . . . . . . . . . . 4.4. Restricted Double-Possessive . . . . . . . . . . . . . . . 4.5. One argument only . . . . . . . . . . . . . . . . . . . . . 5. Generalizations about action nominal constructions . . . . . . . 5.1. Complement-deranking vs. complement-balancing as type predictors . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. The form of constructions . . . . . . . . . . . . . . . . . 5.2.1. Lack of autonomy . . . . . . . . . . . . . . . . 5.2.2. Word order . . . . . . . . . . . . . . . . . . . . 5.2.3. Head-marking . . . . . . . . . . . . . . . . . .

678 679 688 692 697 701 702 703 705 706 707 710 711

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

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

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

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

723 723 725 725 726 726 728 729 731 731 735 736 737 738 738 739 741 744

. . . . .

. . . . .

. . . . .

. . . . .

744 747 747 748 749

✐

✐ ✐

✐

✐ xxi

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #17

✐

✐

xxi

Contents detailed

5.2.4. 5.2.5.

Dependent-marking . . . . . . . . . . . . . Relations between dependent-marking and head-marking . . . . . . . . . . . . . . . . 6. Europe vs. the world . . . . . . . . . . . . . . . . . . . . Appendix: The European sample . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . 749 . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

750 750 754 755 757

Leon Stassen Noun phrase conjunction: The coordinative and the comitative strategy 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Definition of the domain of investigation . . . . . . . . . . . . 3. The Coordinative Strategy . . . . . . . . . . . . . . . . . . . . 4. The Comitative Strategy . . . . . . . . . . . . . . . . . . . . . 5. AND-languages and WITH-languages . . . . . . . . . . . . . . 6. Areal distribution . . . . . . . . . . . . . . . . . . . . . . . . . 7. Correlational parameters . . . . . . . . . . . . . . . . . . . . . 8. The languages of Europe . . . . . . . . . . . . . . . . . . . . . Appendix: The sample . . . . . . . . . . . . . . . . . . . . . . . . Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

761 763 766 778 781 799 802 806 808 809 811

Indexes Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 821 Language index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 827 Author index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 837

✐

✐ ✐

✐

✐ xxii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #18

✐

✐

✐

✐ ✐

✐

✐ xxiii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #19

✐

✐

Contributors Greville G. Corbett Department of Linguistic and International Studies University of Surrey Guildford Surrey GU2 5XH Great Britain E-mail: [email protected] David Gil Linguistics Max Planck Institute for Evolutionary Anthropology Inselstraße 22 D-04103 Leipzig Germany E-mail: [email protected] James R. Hurford Department of Linguistics University of Edinburgh Adam Ferguson Building 40 George Square Edinburgh EH8 9LL Great Britain E-mail: [email protected] Aleksandr E. Kibrik Filologiˇceskij fakultet 953 Moskovskij gosudarstvennyj universitet (MGU) 119 899 Moskva Russia E-mail: [email protected] Maria Koptjevskaja-Tamm Institutionen för lingvistik Stockholms Universitet Fiskartorpsvägen 160 C S-10691 Stockholm Sweden E-mail: [email protected]

✐

✐ ✐

✐

✐ xxiv

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #20

✐

✐

xxiv

Contributors

Edith A. Moravcsik Department of Linguistics University of Wisconsin-Milwaukee Milwaukee WI 53201-0413 USA E-mail: [email protected] Frans Plank Fachbereich Sprachwissenschaft Universität Konstanz D-78457 Konstanz Germany E-mail: [email protected] Leon Stassen Instituut Algemene Taalwetenschap Katholieke Universiteit Erasmusplein 1, K. 5.15 NL-6525 GG Nijmegen The Netherlands E-mail: [email protected]

✐

✐ ✐

✐

✐ xxv

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #21

✐

✐

Abbreviations 1, 2, 3 (. . .) (a) 1st, 2nd, 3rd person (b) genders/noun classes I, II, III, . . . (a) genders/noun classes (b) distance distinctions in demonstratives A ABL ABS ABSL ACC ACT AD ADESS A(DJ) ADJCT ADV AFF AGR AL ALL AN ANC ANIM AOR AP APL APUD ART ASP ASRT ASSOC ATTR AUX CAUS CLn

transitive subject (agent) ablative case (a) absolutive case (b) absolute noun form absolute state accusative case actor localization ‘adjacent to OP’ adessive case adjective adjectivalizing marker (a) adverb (b) adverbializing marker affix agreement marker alienable possession allative case (a) action nominal (b) action nominal marker action nominal construction animate gender aorist tense/aspect adpositional phrase associative plural localization ‘near to, but not touching OP’ article aspect assertive mood associative (or linking) form attributive marker auxiliary causal case noun class n

CLF CLT CM CMPR COM COMM COMP COMPAR CONJ CONST CONT CONV COP CUM

D DAT DCT DEF DEL DEM DERIV DET DIST DISTR DP DS DU DUR DYN ELAT EMPH ENUM EQU ERG

classifier clitic construction marker comparative degree comitative case common (uter) gender complementizer comparative case conjunction construct state localization ‘contiguous, in complete contact with OP’ converb copula localization ‘near to, but in an unspecified position relative to inanimate OP, together with animate OP’ dependent dative case direct case(s) (a) definite (b) definite article delative case demonstrative derivational affix determiner (a) distal demonstrative (b) distant (past) distributive(-temporal) case “determiner phrase” different-subject marker dual number durative aspect dynamic aspect elative case emphatic marker enumerative marker equative case ergative case

✐

✐ ✐

✐

✐ xxvi

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #22

✐

✐

xxvi ESS EXCL EXT EZ FAC F(EM) FIN

Abbreviations

essive case exclusive 1st person extended localization ezafe marker factive case feminine/female gender (a) final case (b) finite verb form FOC focus marker FORM formal case FUNC functive case FUT future tense GEN genitive case GEN.DCT direct genitive case GEN.OBL oblique genitive case GER gerundial marker GNR generic H head ILL illative case IMP imperative mood IMPRF imperfective aspect IN localization ‘inside, within a closed space of OP’ INAL inalienable INCL inclusive 1st person IND indicative mood INDEF (a) indefinite (b) indefinite article INDV individualizing marker INESS inessive case INF infinitive INGR ingressive aspect INST instrumental case INT interrogative marker INTER localization ‘within a filled continuous space of OP’ IO indirect object LAT lative case LOC (a) locative case (b) localization ‘in a typical position relative to OP’ (LO ‘below’, HI ‘above’, LEV ‘same level’) LOCAL localization category

M M(ASC) MHUM

multiplier (or base) in numerals masculine/male gender masculine human gender (NON_MHUM not masculine human) MOD modal case N noun NARR narrative verb form NEG negative, negation N((EU)T) neuter gender NFIN non-finite verb form NMNL nominalizer NONFUT non-future tense NONM non-marked NOM nominative case NP noun phrase NUM (a) numeral (b) “number” constituent in numerals NUMB number OBJ object OBL (a) oblique stem (b) oblique case(s) OP orientation point ORD (a) ordinal numeral (b) ordinal-forming marker ORIENT orientation marker ORLOC orientational locative case P transitive object (patient) PART participial marker PARTIT partitive case PASS passive PAT patient marker PERF perfective aspect PERMUT permutative case PERS person PL plural number PNP possessive noun phrase POSS (a) possessive case (b) possessed noun form (i.e., marked for possessor) (c) localization ‘belonging to animate OP’ POSSR possessor

✐

✐ ✐

✐

✐ xxvii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #23

✐

✐

Abbreviations POST PPL PR.ART PRED PREF PREP PRET PRO PROG PROX PRS PST PTCL QUANT QUOT RAT RECIP RFL RM

localization ‘behind OP’ possessive plural (pre)proprial article predicative form prefix preposition preterite tense pronoun or pronominal form progressive aspect (a) proximal demonstrative (b) proximal (past) present tense past tense particle quantifier quotative rational gender/noun class reciprocal reflexive relational marker

xxvii

S

intransitive subject (agent/patient) SBJ subject SG singular number SOC sociative case SOV, etc. subject–object–verb, etc. SS same-subject marker STAT stative form SUB localization ‘below OP’ SUBL sublative case SUBORD subordinate SUPER localization ‘on OP’ SUPESS superessive case TEMP temporal case TERM terminative case TOP topic marker TRANSL translative case UNS unspecified tense/aspect VOC vocative case VP verb phrase

Separators in examples and glosses = . :

morpheme boundary class marker boundary (in Daghestanian languages) (a) semantic components expressed cumulatively (b) morphological components not segmented in example morphological components not segmented in example

✐

✐ ✐

✐

✐ xxviii

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #24

✐

✐

✐

✐ ✐

✐

✐ 1

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #25

✐

✐

I.

Introduction

✐

✐ ✐

✐

✐ 2

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #26

✐

✐

✐

✐ ✐

✐

✐ 3

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #27

✐

✐

Frans Plank

Noun phrase structure: An und für sich, in time, and in space

1. Us From roughly 1990 to 1995, the contributors to this volume were the principal partners in a broad-mindedly collaborative enterprise devoted to noun phrases, itself part of a larger joint venture, the EUROTYP programme of the European Science Foundation.1 In addition to pursuing their individual interests in noun phrase matters, they would confer twice a year, and correspond about twice a day, in the furtherance of concerns they were sharing. Of the ideas that were broached some came to a state of fruition where they could be put down in writing, for good or to be rewritten. Thus accrued a batch of working papers, two thematic collections, and at long last this volume. The contents of EUROTYP Working Papers VII/1–25 (1990–95),2 of Double Case (1995), and of The Maltese Noun Phrase Meets Typology (1996), complementing the present volume as the literary legacy of five years of cooperation, are detailed in the Appendix to this introduction.3 My emphasis in this introduction will be less on chronicling our progress and digressions than on giving a general characterization of the sort of approach that we took and the sort of outlook that it would afford or withhold. Although inevitably also retrospective, the basic mood will be mixed introspective and prospective. The views expressed on typology by itself (Section 2) are likely to be shared by everybody in this volume, while my reflections on typology’s relation to diachrony and areal studies (Sections 3 and 4), occasioned by the EUROTYP experience, are of a more personal nature.

2. Typology by itself Naturally, the approach of the EUROTYP Noun Phrase Group was typological. By common assent, approaching a structural domain typologically means, first, to identify what is individually variable about this domain across the linguistic universe; second, to determine whether individual, logically independent variables co-vary with any others, in the same domain or in others, thereby setting limits on crosslinguistic diversity; and third, to explain such findings. There can be no co-variation

✐

✐ ✐

✐

✐ 4

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #28

✐

✐

4

Frans Plank

unless there is variation; still, universals which proscribe variation unconditionally define limits on diversity just like universals of co-variation do. These latter are typically stated as material implications, often relativized statistically, but there are also other significant conditional relationships between variables, such as those of licensing or favouring. Co-variation is typically inferred from samples: language samples are typologically the more meaningful the more representative they are of the full range of structural diversity, at least within that part of the linguistic universe which happens not to have vanished without a trace.4 Explanations of co-variation are typically cast in terms of higher-level structural generalizations, but they are ultimately also sought in perception or cognition, in storage, production, or processing, in acquisition or evolution, in communicative functions or genetic predispositions. Our assigned structural domain was a particular kind of phrase, best known as NOUN PHRASE (NP). Even for a domain so circumscribed and familiar (under whichever name), a typological profile is not drawn at one sitting. Getting in control of all conceivable variables of NP structure, and of a sufficiently diverse range of languages to be sufficiently confident when claiming co-variation, is too tall an order to fill at the present state of knowledge, even when prolonged concerted action is taken. To get anywhere, you have to make a choice and give some variables priority over others – guided by the emphases of your theoretical framework, by your expectations about which variables hold the greatest promise of co-variation, and also by practical considerations about the feasibility of obtaining enough reliable information. In our case, for various reasons including that of labour being better divided (among the several EUROTYP groups) than repeated, we did not give priority to the EXTERNAL grammar of NPs. Referring to persons, things, and further kinds of referents liable to receive similar grammatical treatment (e.g., abstract notions, events, actions, or also properties), independent NPs form part of larger syntactic expressions: in particular, NPs occur as core arguments of verbal and other predicating expressions (as subjects and direct and indirect objects or their equivalents in other types of relational systems), yielding verb phrases or clauses; as complements of adpositions, yielding adpositional phrases; as coordinate conjuncts of other NPs; or also as integral parts of complex NPs (as attributes, to use a neutral designation). Although bona fide NPs may also be used in yet further kinds of constructions (e.g., as adverbials, predicate nominals, or absolutely), and although expressions other than NPs may also be used as parts of constructions just mentioned (e.g., clauses, verb phrases, or adpositional phrases as subjects), these co-occurrences with particular kinds of other expressions in particular kinds of grammatical relations constitute the prototypical external grammar of expressions of the class traditionally subsumed under the label NP. Although we by no means expect the external grammar of NPs to be invariable across languages nor only to vary randomly, systematic variation in these respects

✐

✐ ✐

✐

✐ 5

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #29

✐

✐

Noun phrase structure

5

is given rather short shrift in the published work of the EUROTYP Noun Phrase Group. Nonetheless, there are many ways in which the external grammar of NPs is reflected internally and vice versa, and this is reflected in this volume in contributions on the outside agreement with NPs at clause level (in Corbett’s chapter,5 and also in Stassen’s), on NP coordination as opposed to other constructions for the joint involvement of NP referents (in Stassen’s chapter), on the morphological coding of the external relations of NPs by case and similar inflectional categories (in Kibrik’s and Moravcsik’s chapters in Part II, and in all chapters in Part IV), on the marking for such a prime discourse category as definiteness (Part III), and on the shading of clauses into NPs when contracting the prototypical external relations of the latter (in Koptjevskaja-Tamm’s chapter on nominalization, Part IV). Elsewhere the external NP relations of apposition (in Double Case) and predication (in Stassen’s chapter in The Maltese Noun Phrase Meets Typology6) have been attended to. As to the external linear ordering of NPs, their external grammatical relations,7 their behaviour in topicalizing and focusing constructions, their sensitivity to tense, aspect, and actionality, their not-so-independent realization as clitics, and the distinction between NPs and clauses in relationships of subordination and complementation, the interested reader is referred to the relevant companion volumes in the EUROTYP series. Our own priority was the INTERNAL grammar of NPs. To cover this subdomain comprehensively, and postponing questions of outside implicational connections, one would have to look at cross-linguistic diversity in the following general respects: (i) how phrasehood is manifested (if it is); (ii) what syntagmatic relations obtain between NP constituents and how they are encoded; (iii) which kinds of words, phrases, and clauses are found as NP constituents; (iv) which subclasses of such forms are distinguished; (v) how word formation enriches the fund of nominal words; (vi) how form classes match up with syntagmatic relations; and (vii) which nominal words inflect for which categories. All items of this agenda were covered as EUROTYP proceeded, though some less comprehensively than others.8 Our focal points were systems of inflection (Part II), the morphology, syntax, and semantics of determination (Part III), and the relational structure of NPs variously amplified (Part IV). As to (i), the phrasehood of NPs as such was a central issue especially in Double Case, with echoes in Plank’s and Moravcsik’s chapters in Part III of the present volume. The point was made, and illustrated in great (often extra-European) detail, that NP-internal cohesion can be tighter or looser, with some languages, sometimes characterized as having “flat” or appositive word-based syntax, arguably not evincing such a phrase type at all. Habitual discontinuity may be indicative of a low degree of phrasal integration; but it does not suffice to disprove phrasehood.9 Phrasal cohesion can show in linear ordering, with co-constituents subject to adjacency and cooccurrence constraints (requiring or excluding one another), but also, simultaneously or alternatively, in the marking of phrasal parts for agreement or cross-reference with one another or for governing or commanding one another, in the possibility of being

✐

✐ ✐

✐

✐ 6

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #30

✐

✐

6

Frans Plank

combined in morphological rather than only in syntactic construction (compounding, incorporation), or also, if perhaps less directly, in phonological phrasing. While overall questions of NP-internal order were delegated to another EUROTYP group10 and phonological phrasing was largely disregarded, agreement/cross-reference and government/command within NPs and their overt manifestations, and more peripherally also the combining of NP parts at word level, are recurrent themes in this volume and in our other publications.11 As to (ii), provided parts cohere tightly enough one way or another to form phrases, they can hold different syntagmatic relations within their phrases. Basic elements of relational architecture are arguably selected from a fund that is universal, and is to some extent shared by constructions of all kinds: phrases (NPs and other) have heads, whose most characteristic prerogative is self-sufficiency, grammatically and semantically; if heads don’t come alone, they are accompanied by dependents, or also by notso-dependents (sometimes called “adjuncts”, like adjoined relative clauses) or other independents (conjuncts). For NPs, dependency subsumes a range of more particular relations, including primarily those of determination (or specification), modification (or attribution, perhaps at several levels of closeness to heads), complementation, and apposition, and yet more particularly others such as quantification or classification, or also specific semantic relations of modifiers and determiners such as possessor, agent, part-whole, material, or origin. Having NPs does not perforce mean that all of these relations will play a role in the grammar of all such languages (for example, adjuncts do not seem to be very frequent at NP level); nor will the distinctions between those relations selected always be equally clear-cut (that between determination and modification, for example, can be rather elusive12). Least invariable are the syntactic manifestations and the overt encoding of NP-internal relations. The bulk of the present volume is given over to the exploration of such variation for the relations of determination (Part III) and “amplification”, to coin a maximally neutral term meant to subsume all expansions of NPs beyond their lexical core without prejudging their precise nature (Part IV). A general chapter on the dependency structure in NPs was also planned for this volume, to examine the claims which elements other than nouns have on the status of head (especially determiners, but probably also quantifiers and adjectives, as suggested in the generative DP literature and before) and to find distributional, inflectional, and semantic motivation for hierarchical layering inside NPs; unfortunately it did not materialize. However, aspects of this question have variously been addressed elsewhere by members of the EUROTYP Noun Phrase Group, including in a separate collective volume, Corbett et al. (1993), where Payne (1993) makes a strong empirical case against an analysis of NPs as DPs. Also, clarifying the dependency structure within NPs was a main theme in Double Case. Quantification, in this volume essentially limited to numerals, was the subject of several working papers (by Gil, Corbett, Kibrik et al.) and of chapters in The Maltese Noun Phrase.

✐

✐ ✐

✐

✐ 7

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #31

✐

✐

Noun phrase structure

7

As to (iii), the constituents to be typically reckoned with in NPs include these: words of such major classes as nouns, adjectives, numerals, and perhaps certain adverbs; function words such as independent pronouns (sometimes with null realization), quantifiers (or also special number words), articles, classifiers, and all kinds of particles (including intensifiers and other qualifiers, positional and directional indicators, focus particles, ligatives linking the parts of NPs, mere markers of NPhood); phrases of various kinds (headed by typically nominal words, namely adjective phrases, possessive phrases, NPs themselves, but also adpositional phrases); and clauses of various kinds (relative and other attributive clauses, complement clauses). Few of these classes of NP-internal words, phrases, and clauses are universal, although many are widespread, even if not always differentiated from one another equally strictly. Nouns and personal pronouns (at least those of 1st and 2nd person, with demonstratives often doing duty for 3rd person deictic and phoric pro forms) have the best claims to universality. But even these parts are not undisputed. Conceivably, there might be only one single, syntactically undifferentiated class of lexemes, with a Standard Average European transitive sentence like ‘The boy sees the goat’ rendered by a sequence of three uniform predications where such lexemes are combined with (perhaps bound) pronouns, ‘he is young, he sees it, it stinks’.13 However, upon closer inspection of relevant languages, the grammatical and semantic potentials of such lexemes and their overt morphosyntactic properties in particular constructions have always turned out to be different, massively or at least subtly (as in another rendering of the above sentence, ‘he sees it, the youth, the stinker’, which clearly contains two nouns, neither however unqualifiedly nouny), thus reconfirming that not to distinguish nouns from non-nouns is not an option. In other languages where nouns have been questioned (e.g., Salish, Wakashan, or Philippine), the questions are about the predispositions of lexemes, being allegedly so versatile as not to require any special marking to adapt them for referring, predicating, or modifying uses, rather than about their actual grammatical uses, indubitably manifesting distinctions of form and distribution.14 Doubts whether 1st and 2nd person pronouns are universal seem better grounded: deictic reference to speaker and hearer is sometimes made by means of barely grammaticalized nouns for social roles such as ‘servant’ and ‘master’ or also by local deictics. Turning to point (iv), although subclasses of NP constituents are rarely entirely language-particular either, there is cross-linguistic variation as to whether and how nouns are distinguished as proper and common, count and mass, individual and collective, animate and inanimate, concrete and abstract, or relational and absolute; adjectives as denoting essential or accidental properties, as denoting value, size/dimension, age, colour, material, human propensity, or other kinds of property concepts, as gradable or ungradable, descriptive or limiting; numerals as cardinal, ordinal, distributive, multiplicative, and fractional, as counting and adnominal forms; pronouns as personal, possessive, reflexive, reciprocal, logophoric, demonstrative,

✐

✐ ✐

✐

✐ 8

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #32

✐

✐

8

Frans Plank

indefinite, interrogative, and relative; articles as definite, indefinite, specific, generic, proprial, and partitive; complement and attributive clauses as finite and non-finite, etc. Variation in the distinction of form classes of NP constituents, especially of nominal words and their subclasses, has been a pervading concern in our work. In this volume, this is reflected in chapters devoted to personal and demonstrative pronouns (Plank, Part II), articles (all of Part III), numerals (Hurford), possessive phrases (Koptjevskaja-Tamm), and adpositional phrases (especially comitatives, Stassen) (Part IV); and noun subclassification for inflectional purposes figures prominently in Part II. Adding to this what we have published in this rubric in our working papers and other collective volumes (e.g., on noun subclasses in relation to number marking, or on the delimitation of cases from adpositions and concomitantly of NPs from adpositional phrases),15 there still remain serious lacunae, such as pronominal systems in overview or proper names looked at from a typological angle. As to (v), although basic nominal or pronominal words can in all languages be supplemented by complex ones, the ways and means of word formation differ a great deal from language to language, and few efforts have been made in typology to reduce such diversity to order. The delimitation of word formation from inflection is not equally strict everywhere to begin with, and within word formation at its most typical, compounding is not always equally clearly distinguished from derivation. The uses of compounding and the categories, basic units (stems, words), systemic properties, and formal devices of derivation show great and largely unmapped variation, which is evidently not altogether random, although the evidence for constraints is rarely ample – to illustrate: if words other than nouns form compounds, then nouns will do so too; if there are diminutives and augmentatives of adjectives, then nouns will also have such derivatives; verb-derived nouns will primarily include categories of clausal roles such as agent, patient, beneficiary, instrument, place; if there is polysemy, nomina agentis are likeliest to share their affixal encoding with nomina instrumenti and loci. One area where chance variation has been thought to reign supreme is increasingly emerging as being rather well-structured: languages do not differ without limit in which meanings are taken care of by core vocabularies of basic words and which need complex words for their expression, the best-studied lexical fields to date being those of colours and numbers. The formation of numerals (Hurford), actional nominals (one of KoptjevskajaTamm’s chapters in Part IV), and stem formation accompanying inflection proper (Kibrik) are three issues of word formation directly addressed in this volume. Complex attributive forms of nouns (genitives, possessive adjectives, and other) and their derivational or inflectional nature are another relevant topic touched on in chapters by Kibrik and Koptjevskaja-Tamm (Part III, on non-determiner genitives16), but Double Case has much more on this.17

✐

✐ ✐

✐

✐ 9

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #33

✐

✐

Noun phrase structure

9

As to point (vi) of our agenda of variation, there is much that is universal about which classes of constituents are eligible for which syntagmatic relations within NPs; after all, to some extent the rationale of form classes lies in such relations. For example, adjectives are words specializing in modifiying words which themselves specialize in establishing reference to persons and things (by naming, describing, or pointing to such referents). Still, there is variation in how relationally versatile certain form classes are, and in what formal adaptations are required if words or complex expressions are used in relations others than those they are predestined for – for example, in whether nouns can be used as modifiers, unchanged or only upon overt adjectivalization. In this volume, such issues of the matching of form classes and NP-internal relations are raised repeatedly, most specifically with respect to attributive relations and the special marking of forms when used attributively in Daghestanian languages (Kibrik), to the word-class provenance of quantifying words (the higher the number the nounier the numeral; Hurford), to the use of genitives as modifiers rather than determiners (Koptjevskaja-Tamm’s chapter in Part III), and to the poly- or perhaps rather macro-functionality of certain kinds of formal markers of NPs and their constituents (Gil). Finally, as to (vii), one of the most productive sources of variation is the expression of certain nominal grammatical (or “functional”) categories – such as definiteness, grammatical relations, gender/class, possession, number – by either function words (articles, adpositions, classifiers, possessive pronouns, quantifiers or other number words) or inflections. To begin with, bound morphology may or may not be easy to distinguish from free forms which are clitic. If there is nominal inflection, there may be more or less of it, as measured in inflectional categories, in terms realizing categories (for example, with number only realized by singular and plural, or also by dual, trial, quattral, paucal, multal, singulative, collective, associative), and in the extensions of domains (for example, with only some pronouns and nouns or all pronouns and count nouns inflecting for number; or with only determiners or also with modifiers agreeing in number). Further, overt exponents, coming in a variety of forms (affixes, reduplication, segmental or suprasegmental modification, suppletion, subtraction, metathesis, zero – as also used in derivation), may express separate categories (such as number and case) separately or they may cumulate them, and one category may be expressed in one place or it may extend over several co-occurring segments of a word. Inflectional exponents may be distinct or syncretic, invariant for all relevant words or variant, giving rise to inflection classes. Inflectional categories may be paradigmatically independent of each other or interdependent, exerting all kinds of influences on each other’s expression. All in all, the potential for diversity of whole inflectional systems is vast. Accordingly, Part II of this volume and parts of Part III are probing actual differences and similarities between inflectional systems, looking at a wide range of

✐

✐ ✐

✐

✐ 10

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #34

✐

✐

10

Frans Plank

nominal inflectional categories18 and paying special attention to such parameters as have traditionally figured in morphological typology, where languages were characterized along two dimensions: as (predominantly) analytic (or isolating), synthetic, or polysynthetic (incorporating), and as (predominantly) agglutinative, flective, or introflective (non-concatenative).19 One objective of Double Case had been to examine the typological significance of what has been called group inflection or phrase marking (as opposed to word marking) but turned out to show an unmistakable family resemblance with agglutination. So much for the thematic agenda of five years of NP research within EUROTYP. Now, doing typology, with whatever structural focus, you ineluctably find yourself doing chores which are sometimes believed to be the responsibility of others – description, classification, and theory. (And to history and geography I will turn in a moment.) Typologists are out to discover and explain co-variation, but getting under way in that direction can be laborious, requiring much preparatory spadework. Typological research is dependent on reliable descriptions of languages, ideally representing the full gamut of structural diversity. Sometimes good descriptions are available for languages which one would like to have in one’s sample, some even informed by typological expectations about variation. These are conveniently transferred to typological questionnaires and checklists. Often, however, and not only for out-ofthe-way languages and arcane structures, typologists themselves have to first obtain data and to describe for particular languages what they then want to compare. Much of our own work in the EUROTYP years, eventually finding its way into chapters of this volume or into our other publications, was of that nature. The Daghestanian family (Kibrik et al.),20 Maltese (The Noun Phrase Group in association with local linguists), English in a Southeast Asian environment (especially Singlish; Gil), and Swedish (Koptjevskaja-Tamm), all investigated in situ with the assistance of native consultants, were the main beneficiaries of our descriptive preparations. Co-variation presupposes variation. However, tracing variation for individual parameters and classifying languages accordingly, in preparation for the real typological business of establishing co-variation, can be an arduous task in itself. While some parameters are straightforward (e.g., having or lacking a trial number, although even that one can be tricky), others are internally complex (e.g., Suffixaufnahme, whose prototype and varieties involve a bundle of syntactic and morphological features), and surveying them across languages cannot be done by glancing at one designated paragraph in the respective grammars. Although we take credit for a goodly number of implications, put forward, examined, or rejected in this volume and elsewhere,21 much effort has gone into mapping out dimensions of variation, yet awaiting to be correlated with others. Good typology feeds on theory and in turn feeds into good theory. Linguistic theory is about the laws of language, and this is what implicational universals, typol-

✐

✐ ✐

✐

✐ 11

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #35

✐

✐

Noun phrase structure

11

ogists’ objects of desire, are or can be reduced to. Typology would be ill-advised to expect theory to be taken care of elsewhere: insight does not perforce come with a “framework”, subjecting a structural domain to its own distinctive terminology and formalism. However, typologists willing to shoulder theoretical responsibility themselves should perhaps not content themselves with such properties of languages as are easy to see at a glance, but also take on the question of the mental representation of grammars – of the rules, constraints, or whatever other organizing principles of linguistic behaviour. The work of the EUROTYP Noun Phrase Group has been done without an allegiance to any particular current school or framework. Our descriptive analyses might be characterized as “concrete”, insofar as the categories and structures posited are intended to relatively directly reflect elementary formal and distributional patterns of forms in constructions of particular languages; no theoretical framework, however averse to taxonomy and given to facile shorthand (of the sort of [±N, ±V]), should be able to do without them in whatever notational guise. Explanations were typically sought in higher-level descriptive generalizations, first within languages and then across them. However, when we were able to state cross-linguistic generalizations as implications, licensing some combinations of properties but proscribing one, then they would typically involve concrete rather than abstract variables, linked directly rather than indirectly. The most general explanatory principles with the greatest appeal for us would be longstanding ones which have often been termed “functional”, such as simplicity (formal economy, markedness), clarity (ambiguity avoidance), expressiveness, or iconicity (form being motivated by function). Nothing in our approach, however, militates against explanations being genetic, invoking inherited traits which are evolutionarily advantageous or also arbitrary. When concrete analyses are favoured over abstract ones in our own typical manner, the risk is incurred of overlooking certain grammatical generalizations which are very real – and which, ironically, a typological approach is best suited to uncover. Let me illustrate this danger with an example of word order in NPs.22 When facing variation in the preferred, unmarked, non-contrastive relative ordering of adjectives like that between English (1a), Bahasa Indonesia (1b), and Maltese (1c) – that is, in languages which have attributive adjectives to begin with and which can stack them, if sometimes reluctantly – the priority in our approach would be to make sure that these are indeed the only cross-linguistic options (and ones about equally well represented), with a rather obvious fourth alternative (1d), related to (1c) as (1b) is to (1a), being unattested, and then to look for explanations of one’s language-particular and cross-linguistic findings. (1) a. b.

a beautiful big red ball bola merah besar tjantik ‘ball red big beautiful’

✐

✐ ✐

✐

✐ 12

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #36

✐

✐

12

Frans Plank

c.

ballun sabiè kbir aèmar ‘ball beautiful big red’ d. *‘red big beautiful ball’ Disregarding potential interferences from factors such as phonological weight (light tending to come before heavy) or the inclination of inherently emphatic adjectives (such as ‘big’) to jump the queue, adjectives themselves are evidently ordered in terms of their meanings; adding further semantic classes, rarely met with all at once, would only confirm this conclusion (e.g., a beautiful big heavy new red woollen Swiss medicine ball). This insight represents a generalization, since particular adjectives like ugly etc., small etc., green etc. are now seen as falling under the same general rule as beautiful, big, and red respectively. In English, VALUE adjectives precede SIZE adjectives, which in turn precede COLOUR adjectives. In Bahasa Indonesia it is the other way round. Maltese has its adjectives in the same order as English. Considering only languages like English and Bahasa Indonesia, where the mirror-image difference in the ordering among adjectives comes with a difference in the ordering of adjectives relative to their head noun (itself sometimes erroneously believed to follow from an all-encompassing setting of a head-beforedependent/dependent-before-head parameter, also regulating the basic order of object and verb etc.23), a higher-level generalization suggests itself, to the effect that the relative distance of semantic classes of adjectives to their head is the same, regardless of heads being final or initial. Generalizing further, the adjective classes involved could be arranged on a scale of nouniness, also determining other grammatical properties of words with such meanings: COLOUR would be nouniest, VALUE least nouny, and SIZE intermediate – a difference also manifesting itself in word class differences, the use of special derivational morphology (thus, in English there are colour terms which can be used as modifiers but which are nouns rather than adjectives and others which need to be formed from nouns by derivation or compounding: a silver/golden/dove-coloured ball), or also in adjectives of nounier meaning being unavailable to begin with in particular languages reserving this word class only for words of the most pertinent meanings. Aiming yet higher in one’s explanatory aspirations, this mirror-image pattern of (1a/b) might then be explained iconically, as an instance of conceptual affinity (nouniness) motivating linear closeness.24 The non-attestation of (1d) is only to be expected from this explanation. Alas, Maltese casts serious doubt on it: the way its adjectives are ordered among each other, they are on the wrong side of the noun. And Maltese is by no means a loner: in Europe, the Celtic fringe is equally wrongheaded (with a minor modification, apparently preferring Noun – SIZE – VALUE – COLOUR, but this is also an option in the Maltese family, Semitic), and Romance varies this theme by inserting the noun somewhere in the middle, otherwise conforming to the English rather than the Indonesian inter-adjectival order (e.g., Italian una bella grande palla rossa ‘a beautiful big ball red’).

✐

✐ ✐

✐

✐ 13

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #37

✐

✐

Noun phrase structure

13

On a concrete analysis, a representation of NPs in Maltese et al. with the head noun in final position, VALUE – SIZE – COLOUR – Noun, would not normally be countenanced: such an ordering is not even used as a marked alternative in this language! And yet, it is only on such an abstract analysis that the iconic explanation of the relative distance of semantic classes of adjectives from nouns can be upheld: Maltese et al. are – abstractly – exactly like English and exact mirror-images of Bahasa Indonesia. The distinctive feature of the grammar of such languages is an extra syntactic rule of Noun Fronting (or Halfway Fronting in Romance), but this too is not without language-particular structural justification.25 In fact, when asking oneself why it is that some languages have an abstract rule of (Halfway) Noun Fronting while others don’t, one is pointed to further differences in their respective grammars, some abstract and others fairly concrete (such as, allegedly, inflectional differences in the marking for gender and number). The ultimate objects of comparison, then, must be grammars: however concrete or abstract, it is what is represented in the mind which determines the “properties” we are wont to ascribe to “language(s)”. Especially when links between variables in co-variation are not self-evidently direct, as for example in the case of correlations between the architectures of NPs and other kinds of constructions,26 the ways in which they are mediated cannot be reconstructed other than through the workings of mental grammar.27 The next question is whether the involvements that naturally come with doing typology are merely threefold, encompassing description, classification, and theory, or indeed fivefold, also extending to history and geography.

3. Typology and diachrony Like uniformity or perhaps even more so, diversity can be the result of change.28 In the present volume and elsewhere, all kinds of things that are variable about NP structure are shown or conjectured to have changed in time: from appositive origins NP may become more tightly integrated; agreement and genitive-like marking on dependents in NPs may get grammaticalized from anaphoric pronouns in the process of looser constructions being tightened up (ulterior motives throughout Double Case); in the right neighbourhood, NPs may revert and show signs of structural disintegration (see Gil’s chapter on Singaporean English); phrases in determiner function may shift towards modifier status (see Koptjevskaja-Tamm in Part II); demonstratives and numerals may get recruited as definite and indefinite personal pronouns and articles (Working Papers VII/20 and 23); single-function words may become poly- or macro-functional or perhaps the other way round (see Gil and Stassen); inflections may spread among nominal word classes, especially from pronouns to nouns, and may undergo all sorts of semantic and formal reanalyses, after having

✐

✐ ✐

✐

✐ 14

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #38

✐

✐

14

Frans Plank

been grammaticalized from non-bound sources (see Kibrik’s and Plank’s chapters in Part II). However, no comprehensive portrayal of typological developments of NP structure has been attempted. What a specifically typological perspective on diachrony would have to focus on is concurrent rather than individual changes: if variables are contingent on one another, they should not be able to undergo change independently. Now, there are no two or three or four types of NPs that could change into one another: the parameters along which NPs can differ, and hence change, are legion, and they are not easily ranked in significance either. A full typological scenario of co-variation as co-evolution in NPs is as yet unrealistic. And it is not really obvious how rewarding it would be for typologists to turn historians. On one view, their only benefit from such further involvement could be heuristic: if two variables undergo change simultaneously (as for example when case inflection is weakening while articles are emerging), this might suggest that they are contingent on one another. Thus alerted, it would still fall on typologists to establish, across all (known) languages at all (known) times (representatively sampled), whether the simultaneity of these changes was a coincidence or whether these variables are indeed implicationally related, and if so, to figure out why. This view, currently espoused in many circles, casts typology as the master and diachrony as the servant. On the one hand, there are universals, unconditional and conditional, which define the limits of cross-linguistic diversity and which are valid everywhere and at all times and any time; these timeless laws of language are the subject matter of typology. On the other hand, there is much that can be observed about changes (their actuation, transition, embedding, and evaluation), but what is regular about it and amenable to lawlike explanation is essentially only that it is constrained by timeless laws: no language can change so as to infringe upon a law, or at any rate not without subsequent changes swiftly redressing the balance one way or another. Alternatively, however, diachrony can itself be seen to be in charge – and this is how it used to be seen of old, when types were conceived of as stages of evolution.29 Assuming that particular targets (forms, categories, constructions, rules, constraints) can only result from particular mechanisms of change operating on particular sources, this would impose limits on how languages can differ: they can only be what they could become. If change itself is lawful, timeless laws of typology could be redundant. The search for systematic co-variation would turn out to be a comparative diachronic endeavour. An extreme way of diachrony being in charge would be in another scenario, but that is not supported by much evidence. Suppose former diversity got drastically reduced without a trace through mass extinction of languages or rather their speakers, then the grammatical traits shared and retained by the fortuitous survivors would have ended up universal.

✐

✐ ✐

✐

✐ 15

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #39

✐

✐

Noun phrase structure

15

To see from an NP example how responsibility for co-variation can conceivably be apportioned among typology and (non-terminal) diachrony, compare singular and plural forms of indefinite articles in English (3), vis-à-vis those of the definite article (2): (2) I read

SG

PL

the poem

the poems

(3) a. b.

I read I read

a poem Ø poems sm poem which . . . sm poems which . . . " " There is an overt form of the definite article in both singular and plural – indeed the same, the, not agreeing with nouns in number in English. The same holds for indefinite some [sm], while corresponding to singular a(n) there is nothing in the plural.30 " This is no freak peculiarity of English but arguably reflects a law of language: (4)

If there is an overt indefinite article in plural NPs, then there is one in singular NPs (provided there is an indefinite article at all).

Definite (personal, (5)) and indefinite (6) pronouns may show an analogous asymmetry, here illustrated from varieties of German: (5)

SG

(6) a. b. c. d.

PL

Aus Gozo ‘from Gozo

ist er/sie has he/she

gekommen come

sind sie have they

Aus Gozo

ist

gekommen

‘from Gozo Aus Gozo ‘from Gozo

has someone come

sind sind sind have ist has

eine(r)

gekommen come’

welche gekommen eine [oa] gekommen Ø gekommen some come’ man gekommen one come’

Definite pronouns have overt forms in both singular and plural; and so have indefinite pronouns in Standard German (with suppletive interrogative welche or formally singular man serving as plurals, (6a/d)) and in Bavarian (formally regular plural eine, (6b)). In the Palatinate dialect, however, the indefinite plural pronoun is zero even as subject, governing plural verb agreement (6c).31 Again this is no idiosyncrasy of one dialect of one language: whenever there is a zero form of indefinite pronouns in only one number, then it seems to be where it is in Palatinate German – which suggests this implicational universal: (7)

If there is an overt plural indefinite pronoun, then there is an overt singular indefinite pronoun.

✐

✐ ✐

✐

✐ 16

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #40

✐

✐

16

Frans Plank

Markedness is such a pliable explanatory notion that it accommodates (4) and (7), despite first impressions: while plural is marked vis-à-vis singular in definite NPs, requiring some extra formal expenditure, it is probably the other way round in indefinite NPs, with indefiniteness and non-individuation (i.e., non-singular) being mutually conducive, rendering singular the marked number in these circumstances. What we are faced with, thus, is an instance of the general phenomenon of markedness being reversed, from singular being unmarked and plural marked, in marked contexts.32 But why invoke timeless laws, no matter how convincing their explanation, when the pattern to be accounted for follows from laws of change anyhow? Indefinite pronouns and especially articles are typically grammaticalized from one particular source: the cardinal numeral ‘one’. Owing to its meaning, this numeral will typically be confined to singular uses. Thus, if a word meaning ‘one’, and perhaps increasingly used in NPs of specific reference (‘a certain poem’), acquires the meaning ‘indefinite’ (that is, if it comes to be more or less obligatory in NPs whose referent is not assumed by the speaker to be identifiable by the hearer on the descriptive information (s)he is given), and perhaps undergoes some reductive formal change as well (as in English one > a(n)), and provided nothing else happens, there just won’t be any form around to express indefiniteness in numbers other than the singular. The asymmetric distribution of zero and non-zero marking of indefiniteness over non-singular and singular thus ensues automatically from how indefinite pronouns and articles are grammaticalized. Typically grammaticalized from (distal) demonstratives, which are equally comfortable with all numbers, definite articles will not suffer from a comparable lack of non-singular forms.33 In actual fact, however, there are all kinds of other things that may happen. First, the numeral ‘one’ is not the only possible source for the grammaticalization of indefinite pronouns and articles: others include generic nouns, existential quantifiers, or interrogative pronouns, all seen in the English and German examples above (man, some, welch-, respectively), and none as dedicated to the singular number as the numeral ‘one’. Second, even such a dedicated singular word as the numeral ‘one’ may have or may acquire non-singular uses – to express higher numerical or quantificational meanings (e.g., ‘one-DUAL’ meaning ‘two’, ‘one-PLURAL’ meaning ‘several’), to form a distributive numeral (‘one each’), to group accompanying nouns (‘one-DUAL shoe’, i.e. ‘one pair of shoes’, vs. ‘one-SG shoe’), to agree with dualia/ pluralia tantum nouns in number (‘one-PLURAL scissors’), to number-agree with any nouns when part of a complex higher numeral (‘twenty-one-PL pages’), or also upon conversion to (pro-) nounhood (‘Which poems did he read? – The Japanese one-PL’), or just so (as in the case of the Bavarian plural pronoun [oa]).34 Third, dedicated non-singular words utilized for indefiniteness marking, such as arguably the quantifier some,35 may equally be extended to the singular. With no other regulations reining in such a host of possible changes, there could thus come about all conceivable distributions of zero and non-zero indefinite markers

✐

✐ ✐

✐

✐ 17

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #41

✐

✐

Noun phrase structure

17

over singular and non-singular numbers. Yet only such distributions seem ever to be encountered as are licensed by (4) and (7). This finding can only mean that change is indeed superintended by such timeless laws or the explanatory principle behind them (markedness, as suggested above). However, this superintendence does not consist in INDIVIDUAL changes being held in check by timeless laws, but in SEPARATE changes – grammaticalizations of indefiniteness markers from various sources, extensions across numbers in opposite directions – not being permitted to be effectuated INDEPENDENTLY of one another. In this sense co-variation is here explained by co-evolution, with concomitance of changes guaranteed by a constraint on co-existing structures rather than on change itself. This is not to deny that there are genuine regularities of change, such as the semantic change from ‘one’ to ‘indefinite’ (with the numerical meaning perhaps persisting) being indeed as unidirectional and irreversible as grammaticalization theory would like to have it for all changes within its domain. Still, even when one only looks at indefinites developing so regularly from this particular numeral source, the explanation of the number asymmetry cannot be exclusively diachronic either. Again, further possible changes have to be taken into consideration, especially ones that would lead to the discontinuation of a ‘one’-based marker of indefiniteness: through its complete phonetic erosion, its simply falling into disuse, or its reanalysis as something else (like a pure NP marker, independent of indefiniteness). Conceivably, such discontinuation could be the fate of a single form, but this apparently never happens, or at least not to singulars when plurals hang on. Again, it would seem to be a structural constraint on what can co-exist at any time – no non-singular indefinites without corresponding singulars – which is superintending what can be discontinued by successive generations of speakers. On the strength of such examples, there is little hope for typologists that they can have their work wholly done by historians. Being about co-variation in co-evolution, it is, however, best done in cooperation. And typologist-historian teams should be under no illusions what it really is that may show co-variation in co-evolution: grammars, not languages. It is of course only a manner of speaking to say that languages, or their properties, change in time: languages are not individuals possessing temporal (or also spatial) continuity. What really only exists in time (and space, sort of), other than individual speech acts, are representations in the minds of speakers, i.e. grammars. It is ultimately about these that claims are made when laws of change or variation are being posited. In customary parlance, laws of change are universal generalizations about which particular states of a language can or cannot result from which particular prior states and about the mechanisms potentially effectuating the transitions. Put less metaphorically, diachronic (or simply chronic) laws are ones which force particular grammatical or lexical representations upon learners or more advanced speakers, or put

✐

✐ ✐

✐

✐ 18

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #42

✐

✐

18

Frans Plank

such representations out of their reach, whenever they encounter data of a particular kind, and these data have been produced by previous speakers on the basis of different internal representations. By contrast, when laws are said to be achronic (or panchronic), the proper understanding is that there can never be different representations of the same data. Learners or more advanced speakers abiding by such achronic laws may also need temporal experiences to trigger a representation (if a law is implicational, the implicans will need to be encountered as a trigger of the implicatum); but what is invariable, and in this sense timeless, are the representations which successive generations may form of such data. The fieldwork of both the historian and the typologist, thus, has to be done in the field of the acquisition and restructuring of grammar(s). The universe from which they have to draw a representative sample to meaningfully extrapolate universals is not that of languages but the vast one of individual restructurings (including the terminal one of discontinuing a whole grammar): in toto, these are what ultimately accounts for diversity and uniformity across languages.36

4.

Typology and geography

The prepositional phrase subjoined to the main noun phrase of the title of this volume, “in the languages of Europe”, and the name of the series in which this book is included raise contradictory expectations. Discerning readers might want to be certain what they are letting themselves in for, typology or areal studies, for on the face of it the two have got next to nothing inherently to do with each other. Residing in the heads of speakers and hearers, grammars are as mobile as these. Nothing linguistic can prevent a European grammar from thus migrating to, say, Tasmania. Nothing that is universal about it will change. No change of location will license the combination of null singular with non-null plural indefinites or COLOUR before SIZE before VALUE adjectives before nouns. If valid, universals of grammar, which is what typology is about, are not only true at all times (achronically or chronically), but at all places where human languages are spoken. There are no genuinely areal laws of grammar. All that is in essence local about grammar, or for that matter language, is that speech acts have got to take place somewhere. Admittedly, for chronic universals which determine how successive grammars can or must be constructed differently in light of the same linguistic experiences, or also for those achronic ones which need to be triggered, it does matter that those occasioning a learning experience by engaging in speech acts and those acquiring or restructuring their own grammars on this basis are in sufficiently close spatiotemporal contact to establish communication. But this was not the motivation for the areal slant of the EUROTYP programme. The idea simply was to shed more light on the grammatical uniformity and diversity

✐

✐ ✐

✐

✐ 19

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #43

✐

✐

Noun phrase structure

19

of a group of some 130–150 languages spoken of old in an area which is geographically and politically reasonably well delimitable, though not always unarbitrarily: Eurasia from the Atlantic to the Urals, including the Caucasus and Transcaucasia. The genetic affiliations of these EUROTYP languages, hailing in very unequal numbers from seven well established families (Afroasiatic, Indo-European, Uralic, Altaic, Northwest Caucasian, Northeast Caucasian, South Caucasian) joined by a lone survivor (Basque) and a few long defunct apparent isolates without known descendants (Etruscan and suchlike Trümmersprachen as Iberian and Ligurian), are not nearly as diverse and controversial as those in many other neighbourhoods of the world, and the histories of many of them are reasonably well documented; so one could also expect to get an advantageous angle on the traditional historical concern of the Sprachbund, namely on how languages in areal proximity, however historically different, converge.37 In compliance with the EUROTYP areal remit, the Noun Phrase Group was paying special in-depth attention to selected languages of Greater Europe, ensured especially good coverage of the full set in cross-linguistic comparisons, and was seeking to put this set in perspective by holding it against other sets (worldwide samples, including or excluding Europe, or also other areal sets). We nonetheless remained in the business of typology. We would sometimes draw maps out of general curiosity, but such geographical involvement did not seem to us (or at any rate, me) conducive to insight of a specifically typological nature. As typologists, whose business is the discovery of co-variation, we might with equal or indeed better justification have drawn contour maps showing distributions of grammatical features by numbers of speakers, by altitude levels of the habitats of speech communities, or by average annual local rainfall – for who knows whether such extra-grammatical variables might not co-vary with grammatical ones, directly or indirectly. Though perhaps sometimes more Eurocentric in our cross-linguistic coverage than usual, we remained aware of the need to base typological claims on STRUCTURALLY diversified samples. This is not to say that typological samples would a priori be useless unless well-balanced in areal terms. Areal diversity as such is as little guarantee of structural diversity as is genetic diversity – and structural diversity is what counts for typology. But it so happens that for many grammatical parameters the EUROTYP set, although spread over a relatively large area, is known to be seriously lacking in this respect.38 Many valid implications would remain unsuspected on so unvariegated a basis; or, even worse, implications can come out as impeccably valid whose exact opposite is the worldwide rule (like that of inclusive-exclusive and a dual mutually excluding each other in European pronominal inflection, as observed in Plank’s chapter in Part II). Though instrumental in coining and spreading the term “euroversal” and other such areo-versal EUROTYP neologisms, it eluded us what theoretical or practical

✐

✐ ✐

✐

✐ 20

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #44

✐

✐

20

Frans Plank

value such notions could have for typology. Assuming it is valid for the intended areal domain and provided it is not valid everywhere else too (like (ii) presumably is), information like the following might conceivably be found worthy to be treasured in the great storehouse of human knowledge: for all European languages, (i) if basic word order is VSO, then there will be verbal nouns; (ii) there is a distinction of plural and singular number; (iii) actions, unlike necessarily states, are expressed verbally, hence distinct from nouns (three of the euroversals of Haarmann 1976: 105–112); (iv) in Europe, grammaticalized words for distinguishing definite and indefinite nonpronominal NPs (i.e., articles) are more frequent than anywhere else (apocryphal); (v) in Europe, plural words (as opposed to inflections and quantifiers) are less frequent than in some other areas (inferred from Dryer 1989); (vi) Northern Europe is one of the very few places, and perhaps the only one partnering Australia, where there is agreement within NPs but not at clause level (EUROTYP Working Papers VII/23: 11); (vii) for all Greater European languages, if personal pronouns distinguish inclusive and exclusive, then they do not inflect for dual, and vice versa (Plank in Part II of this volume). However, such truths about Europe, which could easily be multiplied, are hardly useful knowledge for typologists on duty: their truths have to be universal. More appreciative should be historians, taking stock of the current grammatical state of affairs in Europe and comparing it to earlier states, with languages appearing in this macro-area and others disappearing and with those staying on changing around their grammars in situ. No real effort, therefore, went into elaborating upon the notion of a “typically European NP”, collecting the traits shared by the NPs of all languages in the area and by none or few outside, at least in this combination. In a companion EUROTYP volume such a characterization has been attempted by Rijkhoff (1998), listing such traits as the well-integrated, non-appositive structure of European NPs; their relatively clearcut major word class distinctions, with adjectives more on the nominal side; their aversion to expressing cardinality by verbs; the rarity of numeral classifiers; the existence of such subclasses as individual, collective, and mass nouns, but not that of “conceptual” nouns; the rich diversity of kinds of modifiers, and the virtually unlimited possibility of stacking them. Again, however, knowing that such traits of NPs characteristically combine in the languages of Europe has no typological surplus value. It is factual information, generalizing over two areally circumscribed subsets of languages, those of Greater Europe vs. the rest or vs. all languages including those of Europe. A priori, no further conclusions can be drawn from this. No comparisons with other areal subsets will lead typologists anywhere either. Only historical comparisons will, focusing on both grammars and places. What behoves typologists, however, is to take note of the outcome of such comparisons and perhaps to give counsel. There can be only three historical reasons for grammatical traits of different languages coming to be in geographical proximity: population movements, linguistic

✐

✐ ✐

✐

✐ 21

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #45

✐

✐

Noun phrase structure

21

diffusion (in the extreme case amounting to language shift), and independent developments fortuitously coinciding. If traits are universal, these will of course always be the joint possessions of all neighbours. Roaming populations, pushed or on the move of their own free will, may enter or vacate areas, and as they bring along and retain their own languages, new contiguities or rifts will appear on maps of grammatical features. This should teach typologists caution in generalizing. What they perceive as universal or widespread may lack any inherently linguistic rationale; conceivably, what they are dealing with may be grammatical traits which happen to be shared by expansive populations, having driven others off the face of the earth, or into such marginal neighbourhoods where there was little point for them in continuing their old language or where they easily escape typological notice. Controlling typological samples in terms of known areas of expansion39 is an important and long-neglected safeguard against positing laws of grammar where the law that has asserted itself was that of the jungle. In this respect maps matter for typology. With or without peoples moving, grammatical traits (or of course lexical items) can be “borrowed” among languages whose speakers have come in contact, which usually implies areal proximity. Borrowing speakers will to some extent be bi- or multilingual; but in the extreme case of an entire language being thus borrowed, as for instance one brought in by a newly dominant élite, later generations may again be monolingual, though with certain traits of the language which their ancestors used to speak preserved as a substratum. The typological interest of such local reductions of cross-linguistic diversity in the wake of diffusion depends on the kind of traits borrowed, or also those retained as a substratum. Traits which are individually universal will be omnipresent anyhow, hence need no special arealist attention. Traits which are individually variable can be independent of others, but what typology is about are those which universally co-vary. Now, among languages in contact plenty of traits get borrowed which are not interpedendent. Surveying the lists of grammatical characteristics defining acknowledged Sprachbünde, especially those of small and medium size, very few of these areally shared, non-inherited innovations are known to be implicated in universal co-variation.40 No known set of Sprachbund traits is typologically consistent. On the other hand, when one trait of two or more established as being in co-variation is getting borrowed, the other(s) would be expected to be borrowed as well – on the assumption that languages in contact are not exempt from the general (achronic or chronic) laws of language. Such concomitant borrowings are also on record, if perhaps distributed in macro- rather than micro-areal dimensions.41 As to substrata on record, some of their traits are typologically rather salient and some sets of such traits are typologically rather consistent (for instance, those attributed to Afroasiatic in Celtic42), while others are typologically as random as Sprachbund traits tend to be.

✐

✐ ✐

✐

✐ 22

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #46

✐

✐

22

Frans Plank

The benefit or areal studies for typology would thus be that sometimes, when relevant traits are getting borrowed or are retained as a substratum, confirmation is provided of the expected, namely that co-variation shows historically in concomitant changes or retentions. It is again change (or also non-change) that is at issue, then, that is, the restructuring of grammars in the course of acquiring more than one language – not geography. Drawing maps of grammatical traits in geographical space is not the primary research that could shed light on this matter. In seeking to clarify the relationship between typology and diachrony, chronic laws have above been distinguished from achronic ones. The former are about linguistic experience lawfully calling for different analyses in successive grammars in acquisition; the latter impose the same grammatical representations regardless of particular experiences. A real question which areally minded typologists might help to answer, in association with historical linguists and others professionally interested in the acquisition of first and further grammars, would be whether concomitant restructurings of grammars differ in substance depending on whether or not the grammar for another set of linguistic experiences (that is, for another language) is interfering. Universal constraints specific to “borrowing”, as opposed to same-language learning and all chronic in the above sense, have been suggested by Moravcsik (1978), excluding for example the borrowing of only non-nouns, or of dependents (such as adpositions) without the ordering relative to their heads. But this needs further empirical examination. However pretty geographers’ maps of linguistic structures will be to look at, and however much they reveal about population movements, grammatical diffusion, and language shifts, ultimately mental maps or brain imagings might afford typologists even more telling insights into their own particular subject matter – structural covariation.

Appendix: Further publications of the EUROTYP Noun Phrase Group A. EUROTYP Working Papers VII/1–25. 1. Frans Plank, Suffix copying as a mirror-image phenomenon. (February 1990.) [Published in: Linguistics 28: 1039–1045 (1990).] 2. Frans Plank, On the selective elaboration of nominal or pronominal inflection. (May 1990.) [Published in this volume.] 3. Greville G. Corbett, Gender and gender systems. (June 1990.) [Published in: R. E. Asher et al. (eds.), The encyclopedia of language and linguistics. Oxford: Pergamon Press, vol. 3, 1347–1353 (1994).] 4. Edith A. Moravcsik, Descriptors of noun-phrase-internal structure. (August 1990.)

✐

✐ ✐

✐

✐ 23

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #47

✐

✐

Noun phrase structure

23

5.

Greville G. Corbett, Agreement: An overview. (September 1990.) [Published in: R. E. Asher et al. (eds.), The encyclopedia of language and linguistics. Oxford: Pergamon Press, vol. 1, 54–60 (1994).]

6.

Frans Plank, Review of Agreement in Natural Language, ed. by Michael Barlow & Charles A. Ferguson, Stanford: CSLI, 1988. (October 1990.) [Published in: Journal of Linguistics 27: 532–542 (1991).]

7.

Maria Koptjevskaja-Tamm, Action nominal constructions in the European languages. (November 1990.) [Published in this volume.]

8.

Jim Hurford, An inventory of noun phrase universals: Format and specimen list. (November 1990.) [Later developed by Simon Kirby into the searchable EUROTYP Noun Phrase Universals Archive, available on the internet at: http://www.ling.ed.ac.uk/˜eurotyp/]

9.

Greville G. Corbett, The head of the noun phrase: Evidence from Russian numeral expressions. (January 1991.) [Published in: Greville G. Corbett, Norman M. Fraser, & Scott McGlashan (eds.), Heads in grammatical theory. Cambridge: Cambridge University Press, 11–35 (1993).]

10. Frans Plank, Inflection and derivation. (March 1991.) [Published in abridged form in: R. E. Asher et al. (eds.), The encyclopedia of language and linguistics. Oxford: Pergamon Press, vol. 3, 1671–1678 (1994).] 11. Frans Plank, On determiners. 1. Ellipsis and inflection; 2. Co-occurrence of possessives. (April 1991.) [Published in part as: “Possessives and the distinction between determiners and modifiers”, Journal of Linguistics 28: 453–468 (1992).] 12. David Gil, Universal quantifiers: A typological study. (April 1991.) [Published as part of: “Universal quantifiers and distributivity”, in: Emmon Bach, Eloise Jelinek, Angelika Kratzer, & Barbara H. Partee (eds.), Quantification in natural languages. Dordrecht: Kluwer, 321–362 (1995).] 13. Frans Plank, From cases to adpositions. (July 1991.) [Published in: Nicola Pantaleo (ed.), Aspects of English diachronic linguistics. Fasano: Schena, 17–61 (1992).] 14. Kristiina Jokinen, On the two genitives in Finnish. (July 1991.) 15. Greville G. Corbett, A typology of number systems. (March 1992.) 16. Aleksandr E. Kibrik, Defective paradigms: Number in Daghestanian. (June 1992.) [Incorporated into his chapter in this volume.] 17. Frans Plank, Advantage Albanian: Grouping in multiple attribution. (June 1992.) [Published in: Heinz Vater et al. (eds.), Festschrift für Winfried Boeder. Frankfurt am Main: Lang (2002).]

✐

✐ ✐

✐

✐ 24

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #48

✐

✐

24

Frans Plank

18. Aleksandr E. Kibrik (ed.), The noun phrase in the Andalal dialect of Avar as spoken at Sogratl. (May 1993.) With chapters by Aleksandr Kibrik, Yakov Testelec, Sergey Tatevosov, Konstantin Kazenin, Ekaterina Bogdanova, Elena Kalinina. 19. David Gil (ed.), Studies in number and quantification. (December 1993.) Containing: Greville G. Corbett, Systems of grammatical number in Slavonic. [Published in: Slavonic and East European Review 72: 201–217 (1994).] David Gil, Conjunctive operators: Areal phenomena or semantic universals? [Published as: “Conjunctive operators: A unified semantic analysis”, in: Peter Bosch & R. van der Sandt (eds.), Focus and natural language processing. Volume 2: Semantics (Working Papers of the Institute for Logic and Linguistics, Working Paper 7, IBM TR-80.94007), 311–322 (1994).] David Gil, Some principles governing the number marking of anaphors. Geraint Wong, The use of numbers in idiomatic expressions in Mandarin Chinese. 20. Conversations on noun phrases. (April 1994.) Containing: David Gil and the LINGUIST List, Numeral classifiers: An e-mail conversation. David Gil and the LINGUIST List, Adjectives without nouns: An e-mail conversation. Frans Plank and the Extended Noun Phrase Group, A panel on the lawfulness of the double life of the definite article. Frans Plank, Edith Moravcsik, Greville G. Corbett, Johan van der Auwera, & Wolfgang Schellinger, On Greenberg 45, mostly. Seventh Expert, Definite double life: An afterthought. 21. Aleksandr E. Kibrik (ed.), Godoberi’s noun phrase. (April 1994.) [Published as parts of: Aleksandr E. Kibrik (ed.), Godoberi. München: LINCOM Europa (1996).] With chapters by Aleksandr E. Kibrik, Sandro V. Kodzasov, Olga Fedorova, J. Gisatullina & S. Toldova, Konstantin I. Kazenin, Sergej G. Tatevosov, Olga Fedorova & Julia Sidorenko, T. Sosenskaja. 22. The noun phrase sketch book. (May 1994.) Containing: Gianguido Manzelli, Descriptive sketches: Italian, Albanian, KomiZyrian, Chuvash, Maltese, Lingala, Summary. David Gil, Hebrew noun phrase checklist. Maria Koptjevskaja-Tamm, Noun phrases in Swedish and Russian.

✐

✐ ✐

✐

✐ 25

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #49

✐

✐

Noun phrase structure

25

Igor’ V. Nedjalkov, Syntactic types of Evenki noun phrases. Frans Plank, Teaching myself noun phrases: (Scottish) Gaelic, Samoan. Robert Mullally, The internal structure of the noun phrase in Scottish Gaelic. John R. Payne, Persian and Lithuanian noun phrases. Edith Moravcsik, Inflectional morphology in the Hungarian noun phrase – a typological assessment. [Published in this volume.] 23. Agreement gender number genitive &. (November 1994.) Containing: Greville G. Corbett, Types of typology, illustrated from gender systems. [Published in this volume.] Frans Plank, What agrees with what in what, generally speaking? Frans Plank, The unlikely plurals of ONE in Bavarian and Miskito. Frans Plank, Homonymy vs. suppletion: A riddle. Frans Plank & Wolfgang Schellinger, More and less trouble for Greenberg 45. [Published as: “The uneven distribution of genders over numbers: Greenberg Nos. 37 and 45”, Linguistic Typology 1: 53–101 (1997).] David Gil, Genitives, number and (in)definiteness: Some data from English, Singlish, Mandarin and Malay. David Gil, “Everything also must grab”: A unified semantic analysis for Singlish also. [Published as: “Singlish also: A unified semantic representation”, in: Proceedings, INTELEC ’94, International English Language Education Conference, National and International Challenges and Responses. Kuala Lumpur: Language Centre, Universiti Kebangsaan Malaysia, 197–210 (1995).] 24. Frans Plank (ed.), Overdetermination. (May 1995.) Containing: Frans Plank, Double articulation. [Published in this volume.] Kersti Börjars, Double determination in Swedish (mainly). [Published as part of: “Swedish double determination in a European typological perspective”, Nordic Journal of Linguistics 17: 219–252 (1994).] Maria Koptjevskaja-Tamm, A woman of sin, a man of duty, and a hell of a mess: Non-determiner genitives in Swedish. [Published in this volume.] Johan van der Auwera, On double determination in Dutch, Lebanese Arabic, Swedish, and Danish. Edith Moravcsik, The distribution of definiteness marking in Hungarian noun phrases. [Published in this volume.] David Gil, Multiple (in)definiteness marking in Hebrew, Mandarin, Tagalog, Indonesian and Singlish.

✐

✐ ✐

✐

✐ 26

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #50

✐

✐

26

Frans Plank

David Gil, Noun-phrase constructions in Singlish: A questionnaire survey. [Published as: “Patterns of macrofunctionality in Singlish noun phrases: A questionnaire survey”, in: S. L. Chelliah & W. J. de Reuse (eds.), Papers from the Fifth Annual Meeting of the Southeast Asian Linguistics Society. Tempe: Arizona State University, 147–182 (1998).] 25. Albert Borg & Frans Plank (eds.), The Maltese noun phrase meets typology. (June 1995.) [Published as: Rivista di Linguistica 8–1 (1996).] Containing: Albert Borg, The structure of the noun phrase in Maltese. Manwel Mifsud, The collective in Maltese. David Gil, “Collective nouns”: A typological perspective. Greville G. Corbett, Minor number and the plurality split. Edward Fenech, Functions of the dual suffix in Maltese. Albert Borg, Distributive and universal quantification in Maltese. David Gil, The Maltese universal quantifier: An areal-diachronic perspective. John R. Payne, The syntax of Maltese cardinal numerals. Ray Fabri, The construct state and the pseudo-construct state in Maltese. Maria Koptjevskaja-Tamm, Possessive noun phrases in Maltese: Alienability, iconicity, and grammaticalization. Martin Haspelmath & Josephine Caruana, Indefinite pronouns in Maltese. Frans Plank, Edith Moravcsik, & David Gil, The Maltese article: Language-particulars and universals. Leon Stassen, The switcher’s paradise: Nonverbal predication in Maltese. B. Frans Plank (ed.) (1995), Double case: Agreement by Suffixaufnahme. New York: Oxford University Press. Containing: Frans Plank, (Re-)Introducing Suffixaufnahme. Gernot Wilhelm, Suffixaufnahme in Hurrian and Urartian. Ilse Wegner, Suffixaufnahme in Hurrian: Normal cases and special cases. Winfried Boeder, Suffixaufnahme in Kartvelian. Aleksandr E. Kibrik, Direct-oblique agreement of attributes in Daghestanian. Ol’ga Ju. Boguslavskaja, Genitives and adjectives as attributes in Daghestanian. Francisco Villar, Indo-European o-stems and feminine stems in -¯ı. Greville G. Corbett, Slavonic’s closest approach to Suffixaufnahme: The possessive adjective. John R. Payne, Inflecting postpositions in Indic and Kashmiri. Maria Koptjevskaja-Tamm, Possessive and relational forms in Chukchi.

✐

✐ ✐

✐

✐ 27

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #51

✐

✐

Noun phrase structure

27

Robert Hetzron, Genitival agreement in Awngi: Variation on an Afroasiatic theme. Fritz Schweiger, Suffixaufnahme and related case marking patterns in Australian languages. Peter Austin, Double case marking in Kanyara and Mantharta languages, Western Australia. Alan Dench, Suffixaufnahme and apparent ellipsis in Martuthunira. Nick Evans, Multiple case in Kayardild: Anti-iconic suffix ordering and the diachronic filter. Anthony Rodrigues Aristar, Binder-anaphors and the diachrony of case displacement. Edith A. Moravcsik, Summing up Suffixaufnahme. C. Albert J. Borg & Frans Plank (eds.) (1996), The Maltese noun phrase meets typology. Torino: Pacini (=Rivista di Linguistica 8-1.) Containing: Albert Borg, The structure of the noun phrase in Maltese. Manwel Mifsud, The collective in Maltese. David Gil, Maltese “collective nouns”: A typological perspective. Edward Fenech, Functions of the dual suffix in Maltese. Greville G. Corbett, Minor number and the plurality split. Frans Plank, Domains of the dual, in Maltese and in general. Albert Borg, Distributive and universal quantification in Maltese. David Gil, Maltese kull: An areal-diachronic perspective. John R. Payne, The syntax of Maltese cardinal numerals. Frans Plank & Edith Moravcsik, The Maltese article: Language-particulars and universals. Martin Haspelmath & Josephine Caruana, Indefinite pronouns in Maltese. Ray Fabri, The construct state and the pseudo-construct state in Maltese. Maria Koptjevskaja-Tamm, Possessive noun phrases in Maltese: Alienability, iconicity, and grammaticalization. Leon Stassen, The switcher’s paradise: Nonverbal predication in Maltese.

Notes 1. Speaking for myself, for five very special years of EUROTYP I am especially grateful to the late Simon Dik, Christoph Mühlberg (then with the ESF), my noun phrase friends, and those who helped to bridge the caesura afterwards. Collectively, we thank the ESF and those who initiated and guided EUROTYP for setting us to work on noun phrases. 2. Many items have meanwhile been revised and properly published, but connoisseurs will always treasure the green-covered grey originals. 3. Follow-up work of ours will be mentioned en passant.

✐

✐ ✐

✐

✐ 28

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #52

✐

✐

28

Frans Plank

4. However elaborately samples are controlled in genetic, areal, and other non-structural terms (see for example Rijkhoff & Bakker 1998), independence in these respects obviously is no guarantee of structural diversity. But see below, Sections 3 and 4, on how diachrony and geography might relate to typology. 5. See further Corbett (1999). 6. See further Stassen (1997). 7. A more basic parameter of external grammar, seemingly trivial but typologically potent, is the NP density per clause; see Munro & Gordon (1982: Section 4.3). 8. In EUROTYP Working Paper VII/4 (1990) Moravcsik provided an initial checklist of descriptors of NP-internal structure couched in maximally elementary terms; but putting this format into comparative practice was not always found convenient. In EUROTYP Working Papers VII/18 (1993), 21 (1994), 22 (1994), and 25 (1995), NPs of particular languages have been described, or sketched, in this format and in others. 9. See recently Lehmann (1991) for a distinction between “nominal group” and “noun phrase” on the criterion of contiguity, suggesting that Latin has a nominal group but no NP. A nominal group supposedly behaves as a unit in terms of its external grammar – which is what is questionable with really flat, non-phrasal syntax, as found in Australia. See Double Case. 10. See especially Rijkhoff (1998). 11. See further Moravcsik (1993) on government. 12. See further Plank (1992). 13. This example is inspired by Tuscarora, as portrayed in Mithun Williams (1976). In fact, strong arguments for a word class of nouns in Iroquoian languages have later been given by Mithun (2000), contra Sasse (1988, 1993) and other lumpers. 14. See further Moravcsik (2000) on nouns, Anward, Moravcsik, & Stassen (1997) on elementary questions of word class distinction, and Gil (1994) for a dissenting voice concerning the universality of noun vs. non-noun or indeed whatever. 15. See now also Stassen (1997) on subclasses of adjectives, some rather nouny or verby; and Dahl & Koptjevskaja-Tamm (2001) on kin terms. 16. See further Koptjevskaja-Tamm (2001). 17. Concerning adjectival lexical fields, see further Davies, Sosenskaja, & Corbett (1999) on colour, adding a new basic term (‘turquoise’) to the universally permissible inventory, Plank & Plank (1995) on smell terminology, often wrongly suspected to be only derived, and Sutrop (1998) and Koptjevskaja-Tamm & Raxilina (1999) on basic and derived temperature terms. 18. See further Corbett & Mithun (1996) and Plank & Schellinger (2000) on less common numbers, the associative and the dual respectively. 19. See further Plank (1999). 20. See further Kibrik (ed.) (1999). 21. Implicational universals relating to NPs, including our own, were collected as part of our EUROTYP collaboration; see Kirby (1995). A more extensive and up to date archive of universals, likewise searchable on the internet (http://ling.uni-konstanz.de/pages/proj/ sprachbau/htm), has since been built up at Konstanz; see Plank & Filimonova (2000) for an introduction.

✐

✐ ✐

✐

✐ 29

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #53

✐

✐

Noun phrase structure

29

22. The voluminous relevant cross-linguistic literature includes, most importantly, Hetzron (1978) and Sproat & Shih (1990). I have broached this topic in my contribution to The Noun Phrase Sketch Book (EUROTYP Working Papers VII/22 (1994)). 23. The independence of adjective–noun and verb–object ordering has been demonstrated by Dryer (1988). 24. And it is not obvious that this explanation would be of a functional rather than a formal nature. 25. Such analyses, drawing some inspiration from Semitic construct state constructions (Ritter 1988) and assuming heads of NPs other than nouns (it is allegedly to head positions that nouns are moving), have become standard in Romance and Celtic generative syntax since Cinque (1993), Longobardi (1994), and Rouveret (1994). An extra explanandum here is why there is Noun Fronting but no analogous Noun Backing producing (1d) from (1b); presumably, there are general principles of grammar determining the direction of such movements and possible landing sites. 26. Or also when implications go across domains, such as phonology on the one hand and morphology and syntax on the other (as surveyed in Plank 1998). 27. This conclusion seems to me ineluctable regardless of whether mental grammar is conceived of as rules, constraints, or connectionist networks. 28. This is not to deny, obviously, that it is conceivable for instances of variation to be primordial and immune to change. 29. See Plank (2001) for the early history of typology-as-byproduct-of-diachrony. Most prominently today, though often overlooked by too selective readers of the relevant half of his oeuvre (as sampled in Greenberg 1990), Greenberg’s “dynamicization” of typology is in this spirit. From a contemporary grammaticalization angle this view is most radically articulated in Heine (1997), though with less emphasis that what matters is systematic CO-evolution, not systematic evolution per se. See further discussion in Plank (1999), with particular reference to morphological typology. In this volume, Plank’s chapter in Part II seeks a diachronic explanation for interdependencies of nominal and pronominal inflection. 30. As this example is only intended for illustration of a general point, many details are ignored, including the indefinite articulation of mass nouns, which may align with plural or with singular count nouns (as in English and Bavarian German, respectively). 31. As objects, zero indefinite plural pronouns are more common. See Glaser (1993) on the dialectal variation in plural indefinites. 32. Though more familiar from phonology, markedness reversal is at home in syntax and morphology too; see Plank (1977). 33. Recently this has been the essence of the story of Heine (1997: Chapter 4). 34. On such ways of extending ‘one’ beyond singular, in Europe and elsehwere, see further my “The unlikely plurals of ONE in Bavarian and Miskito” in EUROTYP Working Papers VII/23 (1994). 35. “Dedicated non-singular” because it is plural (or mass) when used on its own (e.g., Some left) and needs the support of one to be singularized (Someone left). 36. This insight is Joseph Greenberg’s (Bell 1978: 146). 37. EUROTYP has since spawned two other such ventures surveying areas included in or

✐

✐ ✐

✐

✐ 30

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #54

✐

✐

30

38.

39. 40.

41. 42.

Frans Plank overlapping with Greater Europe, the Circum-Baltic and the Circum-Mediterranean ones; see Dahl & Koptjevskaja-Tamm (eds.) (2001) and Cristofaro & Putzu (eds.) (2000). For certain domains, such as verbal voice, even single languages can be like microcosms, comprising virtually everything that is cross-linguistically possible, if sometimes only incipiently grammaticalized. Even if this were true for NP structures too, glimpsing the possibilities of universal variation in an individual language would not be tantamount to recognizing the laws of co-variation. As in Dryer (1989b) and much subsequent work, including in particular Nichols (1992), substantiating the idea that linguistic diversity is patterned in terms of macro-areas. The Sprachbund literature is voluminous, but a good concise survey is Campbell (1996), who can certainly not be accused of giving short shrift to typologically relevant traits. Authors referring to areal studies as “areal typology” (including recently, with EUROTYP background, Ramat 1998) tend to gloss over the scarcity of such traits. Work like that by Nichols (1992) or Stassen (1997, also in this volume) points in this direction. See Gensler (1993) and Comrie (1999).

References Anward, Jan, Edith Moravcsik, & Leon Stassen 1997 “Parts of speech: A challenge for typology”, Linguistic Typology 1: 167–183. Bell, Alan 1978 “Language samples”, in: Greenberg, Ferguson, & Moravcsik (eds.), 123–156. Campbell, Lyle 1996 “Typological and areal issues in grammar”, in: Keith Brown & Jim Miller (eds.), Concise encyclopedia of syntactic theories. Kidlington: Elsevier, 339–343. Cinque, Guglielmo 1993 “On the evidence for partial N movement in the Romance DP”, unpublished manuscript, Università degli Studi di Venezia. Comrie, Bernard 1999 “Typology and the history of language”, unpublished paper (given at the DGfS-Tagung Konstanz). Corbett, Greville G. 1999 “Resolution rules for gender agreement in Tsakhur”, in: Rakhilina & Testelets (eds.), 400–411. Corbett, Greville G., Norman M. Fraser, & Scott McGlashan (eds.) 1993 Heads in grammatical theory. Cambridge: Cambridge University Press. Corbett, Greville G. & Marianne Mithun 1996 “Associative forms in a typology of number systems: Evidence from Yupik”, Journal of Linguistics 32: 1–17. Cristofaro, Sonia & Ignazio Putzu (eds.) 2000 Languages in the Mediterranean area: Typology and convergence. Milano: FrancoAngeli. Dahl, Östen & Maria Koptjevskaja-Tamm 2001 “The grammar of kin terms”, in: Irène Baron & Michael Herslund (eds.), Dimensions of possession. Amsterdam: Benjamins, 201–225.

✐

✐ ✐

✐

✐ 31

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #55

✐

✐

Noun phrase structure

31

Dahl, Östen & Maria Koptjevskaja-Tamm (eds.) 2001 The Circum-Baltic languages: Their typology and contacts. Amsterdam: Benjamins. Davies, Ian R. L., Tat’jana Sosenskaja, & Greville G. Corbett 1999 “Colours in Tsakhur: First account of the basic colour terms of a Nakh-Daghestanian language”, Linguistic Typology 3: 179–207. Dryer, Matthew S. 1988 “Object–verb order and adjective–noun order: Dispelling a myth”, Lingua 74: 77–109. 1989a “Number words”, Linguistics 27: 865–895. 1989b “Large linguistic areas and language sampling”, Studies in Language 13: 257–292. Gensler, Orin 1993 A typological evaluation of Celtic/Hamito-Semitic syntactic parallels. Ph.D. dissertation, University of California, Berkeley. Gil, David 1994 “The structure of Riau Indonesian”, Nordic Journal of Linguistics 17: 179–200. Glaser, Elvira 1993 “Syntaktische Strategien zum Ausdruck von Indefinitheit und Partitivität im Deutschen (Standardsprache und Dialekt)”, in: Werner Abraham & Josef Bayer (eds.), Dialektsyntax. Opladen: Westdeutscher Verlag, 99–116. Greenberg, Joseph H. 1990 On language. (Edited by Keith Denning & Suzanne Kemmer.) Stanford: Stanford University Press. Greenberg, Joseph H., Charles A. Ferguson, & Edith A. Moravcsik (eds.) 1978 Universals of human language, vol. 1: Method and theory. Stanford: Stanford University Press. Haarmann, Harald 1976 Grundzüge der Sprachtypologie: Methodik, Empirie und Systematik der Sprachen Europas. Stuttgart: Kohlhammer. Heine, Bernd 1997 Cognitive foundations of grammar. New York: Oxford University Press. Hetzron, Robert 1978 “On the relative order of adjectives”, in: Hansjakob Seiler (ed.), Language universals. Tübingen: Narr, 165–184. Kibrik, Aleksandr E. (ed.) 1999 Elementy caxurskogo jazyka v tipologiˇceskom osvešˇcenii. Moskva: Nasledie. Kirby, Simon 1995 The EUROTYP Noun Phrase Universals Archive. http://www.ling.ed.ac.uk/˜eurotyp/ Koptjevskaja-Tamm, Maria 2000 “Romani genitives in a cross-linguistic perspective”, in: Viktor Elšík & Yaron Matras (eds.), Grammatical relations in Romani: The noun phrase. Amsterdam: Benjamins, 123–149. Koptjevskaja-Tamm, Maria & Ekaterina V. Raxilina 1999 “S samymi teplymi cˇ uvstvami”, in: Rakhilina & Testelets (eds.), 462–487. Lehmann, Christian 1991 “The Latin nominal group in typological perspective”, in: Robert Coleman (ed.), New studies in Latin linguistics. Amsterdam: Benjamins, 203–232. Lehmann, Christian & Edith Moravcsik 2000 “Nouns”, in: Geert Booij, Christian Lehmann, & Joachim Mugdan (eds.), Morphology: An international handbook on inflection and word-formation. Berlin: De Gruyter, vol. 1, 732–757.

✐

✐ ✐

✐

✐ 32

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #56

✐

✐

32

Frans Plank

Longobardi, Giuseppe 1994 “Reference and proper names: A theory of N-movement in syntax and logical form”, Linguistic Inquiry 25: 609–665. Mithun, Marianne 2000 “Noun and verb in Iroquoian”, in: Petra M. Vogel & Bernard Comrie (eds.), Approaches to the typology of word classes. Berlin: Mouton de Gruyter, 397–420. Mithun Williams, Marianne 1976 A grammar of Tuscarora. New York: Garland. Moravcsik, Edith 1978 “Language contact”, in: Greenberg, Ferguson, & Moravcsik (eds.), 93–122. 1993 “Government”, in: Joachim Jacobs et al. (eds.), Syntax: An international handbook of contemporary research. Berlin: De Gruyter, vol. 1, 705–721. Munro, Pamela & Lynn Gordon 1982 “Syntactic relations in Western Muskogean: A typological perspective”, Language 58: 81–115. Nichols, Johanna 1992 Linguistic diversity in space and time. Chicago: University of Chicago Press. Payne, John 1993 “The headedness of noun-phrases: Slaying the nominal hydra”, in: Corbett, Fraser, & McGlashan (eds.), 114–139. Plank, Frans 1977 “Markiertheitsumkehrung in der Syntax”, Papiere zur Linguistik 17/18: 6–66. 1992 “Possessives and the distinction between determiners and modifiers”, Journal of Linguistics 28: 453–468. 1998 “The co-variation of phonology with morphology and syntax: A hopeful history”, Linguistic Typology 2: 195–230. 1999 “Split morphology: How agglutination and flexion mix”, Linguistic Typology 3: 279– 340. 2001 “Typology by the end of the 18th century”, in: Sylvain Auroux et al. (eds.), Handbook on the history of the language sciences. Berlin: De Gruyter. Plank, Frans & Elena Filimonova 2000 “The Universals Archive: A brief introduction for prospective users”, Sprachtypologie und Universalienforschung 53: 109–123. (http://ling.uni-konstanz.de/pages/proj/ sprachbau.htm) Plank, Frans & Wolfgang Schellinger 2000 “Dual laws in (no) time”, Sprachtypologie und Universalienforschung 53: 46–52. Plank, Sigrid & Frans Plank 1995 “Unsägliche Gerüche: Versuche, trotzdem vom Riechen zu sprechen”, in: B. Busch & U. Brandes (eds.), Das Riechen: Von Nasen, Düften und Gestank. Göttingen: Steidl, 59–72. Rakhilina, Ekaterina V. & Yakov G. Testelets (eds.) 1999 Typology and linguistic theory: From description to explanation. For the 60th birthday of Aleksandr E. Kibrik. Moskva: Jazyki russkoj kul’tury. Ramat, Paolo 1998 “Typological comparison and linguistic areas: Some introductory remarks”, Language Sciences 20: 227–240. Rijkhoff, Jan 1998 “Order in the noun phrase of the languages of Europe”, in: Anna Siewierska (ed.), Constituent order in the languages of Europe. (EALT EUROTYP 20–1.) Berlin: Mouton de Gruyter, 321–382.

✐

✐ ✐

✐

✐ 33

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #57

✐

✐

Noun phrase structure

33

Rijkhoff, Jan & Dik Bakker 1998 “Language sampling”, Linguistic Typology 2: 263–314. Ritter, Elizabeth 1988 “A head movement approach to construct state nominals”, Linguistics 26: 909–929. Rouveret, Alain 1994 Le syntaxe du gallois. Paris: Editions CNRS. Sasse, Hans-Jürgen 1988 “Der irokesische Sprachtyp”, Zeitschrift für Sprachwissenschaft 7: 173–213. 1993 “Das Nomen – eine universale Kategorie?”, Sprachtypologie und Universalienforschung 46: 187–221. Sproat, Richard & Chilin Shih 1990 “The cross-linguistic distribution of adjective ordering restrictions”, in: Carol Georgopoulos & Roberta Ishihara (eds.), Interdisciplinary approaches to language: Essays in honor of Sige-Yuki Kuroda. Dordrecht: Kluwer, 565–593. Stassen, Leon 1997 Intransitive predication. Oxford: Clarendon Press. Sutrop, Urmas 1998 “Basic temperature terms and subjective temperature scale”, Lexicology 4: 60–104.

✐

✐ ✐

✐

✐ 34

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #58

✐

✐

✐

✐ ✐

✐

✐ 35

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #59

✐

✐

II.

On inflection

✐

✐ ✐

✐

✐ 36

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #60

✐

✐

✐

✐ ✐

✐

✐ 37

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #61

✐

✐

Aleksandr E. Kibrik

Nominal inflection galore: Daghestanian, with side glances at Europe and the world

1. Introducing the Daghestanian family The languages of Europe are extremely diverse with regard to the inflectional resources of nominals. There are languages, including English, which are very poor in inflection, and there are families with an abundance of nominal forms, compared not only to English but also to other languages of the world. At the top of the table, the Daghestanian family enjoys the reputation of having the richest case system, and not entirely undeservedly, if the notion of case is taken in a wide sense. Historically derived from a single pattern of nominal inflection, different Daghestanian languages have modified this original pattern in many respects, sometimes quite radically. For present purposes, this richness and diversity is well suited to shed light on the nature of systems of nominal inflection in general. While the focus of this chapter will therefore be on Daghestanian languages, other languages, from Europe and outside, will sometimes be taken into consideration in order to see how ordinary or extraordinary the inflectional phenomena under investigation are. The Daghestanian family, which together with the three Nakh languages (Chechen, Ingush, Batsbi) forms one of the two branches of North-East Caucasian (or NakhDaghestanian), has traditionally been assumed to include 26 languages, subgrouped as follows:1 Avaric:

Tsezic: Lakic-Dargic: Lezgic:

Avar Andic: Akhvakh, Karata, Andi, Botlikh, Godoberi, Bagvalal, Tindi, Chamalal Khvarshi, Tsez, Hinukh, Bezhta, Hunzib; Lak, Dargwa; Lezgian, Tabasaran, Agul, Rutul, Tsakhur, Budukh, Kryz, Khinalug, Udi, Archi

This classification is unsatisfactory in two respects. First, in the case of Lezgic and Lakic-Dargic, it unites in single groups languages which are both genetically and typologically widely separated. Second, in the case of Avar and Dargwa, it unites in single languages dialects whose genetic and typological distance is comparable to the distance between dialects which have the status of separate languages in other

✐

✐ ✐

✐

✐ 38

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #62

✐

✐

38

Aleksandr E. Kibrik

branches (Andic and Tsezic) of this classificaton. I therefore prefer to subgroup the Daghestanian family as follows, on the basis of the relative distance between its members: Avaric: (literary) Avar, Andalal, Antsukh, etc. Andic: Akhvakh, Karata, Andi, Botlikh, Godoberi, Bagvalal, Tindi, Chamalal Tsezic: Khvarshi, Tsez, Hinukh, Bezhta, Hunzib Dargic: (literary) Dargwa, Chirag, Itsari, Megeb, Kubachi, etc. Lezgic: Lezgian, Tabasaran, Agul, Rutul, Tsakhur Kryzic: Kryz, Budukh “Isolates”: Lak, Archi, Khinalug, Udi Most of the data discussed in this chapter come from files collected in field expeditions to Daghestan between 1967 and 1994, undertaken by the Department of Linguistics of Moscow State University under the supervision of the present author.

2. Categories of nominal inflection In most Daghestanian languages nouns inflect for four grammatical categories: number, case, localization, and class.

2.1. Number All Daghestanian languages have a morphological opposition for number, and most of them distinguish two values of this category. For countable nouns the natural interpretation of this opposition is ‘one N’ (SG) vs. ‘more than one N’ (PL); e.g., Akhvakh beca SG ‘snake’ vs. bec-e PL ‘snakes’.

2.2. Grammatical cases Daghestanian languages have often been claimed to have extremely elaborate systems of more than forty cases (see, e.g., Hjelmslev 1935/1937: 139). However, their so-called cases do not represent a homogeneous set of forms and there is every reason to subdivide them into two groups. One group – grammatical cases, in turn distinguished as core and peripheral – is similar in many respects to ordinary cases as well known elsewhere too; but the other group has special formal and functional features, and it will here be dealt with under the rubric of spatial forms. 2.2.1. Core cases The main function of grammatical cases is to specify the grammatical relations between nouns (noun phrases) and their heads (verbs and nouns). Daghestanian lan-

✐

✐ ✐

✐

✐ 39

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #63

✐

✐

Nominal inflection galore

39

guages are well-known for their ergative pattern of clause structure; thus, their core cases, encoding the NPs of a transitive clause with a two-place verb, are nominative (also called absolutive) and ergative, rather than nominative and accusative as in the accusative pattern. A further core case found in all Daghestanian languages is traditionally called dative; its main function is to encode the beneficiary and experiencer roles (e.g., the objects of verbs such as ‘give’ and ‘like’). And most Daghestanian languages also have a genitive for nouns in an attributive construction headed by a noun. Table 1 illustrates this standard set of core grammatical cases from Lezgian. Lak deviates from this standard, limiting itself to three core cases, with the ergative of nouns (but not of pronouns) formally non-distinct from the genitive (Table 2; see Table 20 in Section 3.1.1 for personal pronouns). This is reminiscent of some FinnoUgric languages, in particular modern Finnish, where only personal pronouns have a distinct accusative while nouns use the genitive/nominative in the singular and the nominative in the plural to mark the direct object (see Eliseev 1993: 99; Timberlake 1975). The opposite situation with the standard set supplemented by a fifth case, the affective (used with some verba sentiendi, such as ‘see’), is encountered in Bagvalal (Table 3). Some languages, such as Rutul, lack the genitive but add a comitative to the system of grammatical cases (Table 4).2 Rutul is usually described as having a genitive (see Ibragimov 1978: 51), but this is just a bow in the direction of tradition: the so-called genitive in Rutul is an attributivizer -d/-d1, deriving attributes from nouns and verbs (see further Section 4.3.2). 2.2.2. Peripheral cases Individual Daghestanian languages have additional grammatical cases. The most frequent is an instrumental, as in Bezhta: (1)

cˇ it’a-d Xor muXoc. ö-di boy-OBL . ERG knife-INST sheep.NOM slaughter.PRS ‘The boy slaughters the sheep with the knife.’

Languages lacking a distinct instrumental express the instrumental meaning through a syncretic ergative. It is then possible to have two ergative NPs in the same clause, as in Godoberi: (2)

im-u-di ìudi buq’ida kurt’-i-di. father-OBL - ERG firewood.NOM cuts axe-OBL - ERG ‘Father cuts firewood with an axe.’

Languages with an instrumental reserve the ergative for agentive relations. In that event, as a rule, only a restricted class of nouns, viz. that of animates, has the ergative (see Section 4.3.2 for Tsakhur).

✐

✐ ✐

✐

✐ 40

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #64

✐

✐

40

Aleksandr E. Kibrik

Table 1. Grammatical cases of sew ‘bear’ in Lezgian NOMINATIVE (ABSOLUTIVE) ERGATIVE DATIVE GENITIVE

sew sew-re sew-re-z sew-re-n

Table 2. Grammatical cases of ars ‘son’ in Lak NOMINATIVE ERGATIVE / GENITIVE DATIVE

ars ars-na-l ars-na-n

Table 3. Grammatical cases of zin ‘cow’ in Bagvalal NOMINATIVE ERGATIVE DATIVE GENITIVE AFFECTIVE

zin zin-a-r zin-a-la zin-a-ì¯ zin-a-ba

Table 4. Grammatical cases of edemi ‘man’ in Rutul NOMINATIVE ERGATIVE DATIVE

[ COMITATIVE

edemi edemi-je-r edemi-je-s edemi-je-k◦ an]

The other option is to add a comitative to the inventory of grammatical cases, as does Archi (3), or sometimes also a causal, as does Bezhta (4). (3)

¯ di:ìabu bac a=b=ˇcu zari. jamu bošor-mu-ìu this husband-OBL . SG-COM seven month.NOM III=spend I.ERG ‘I spent seven months with this husband.’

(4)

mi di-l egähäl aqePe. abo-bza father-CAUS you I.OBL-DAT see.INF can.NEG ‘Because of father I could not see you.’

Some peripheral grammatical cases are attested only in single languages, with the maximum found in Archi: a comparative (5), a partitive (6), an equative (7), and a permutative (8).

✐

✐ ✐

✐

✐ 41

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #65

✐

✐

Nominal inflection galore

(5)

niIš-i-Xur t"i. dogi donkey.NOM horse-OBL-COMPAR small ‘A donkey is smaller than a horse.’

(6)

edi. tor laha-s jemim marˇci-me-qIiš os lo L’an-ši that girl-DAT they all-OBL - PARTIT one guy love-CONV. PRS Aux.PST ‘That girl loved one guy out of them all.’

(7)

¯ qer-ši w=i. ìann-a-qIdi tow he.NOM woman-OBL - EQU dance-CONV. PRS I=Aux.PRS ‘He dances like a woman.’

(8)

k’◦ i-s Hadur-ši w=i. zon wa-L’ana I.NOM you.OBL - PERMUT die.I-INF ready-CONV. PRS I=Aux.PRS ‘I am ready to die for you (i.e., instead of you).’

41

The locative was not included in the list of peripheral cases, and this was not an accidental omission (Section 2.3). Khinalug is the only language with a genuine locative (and an ablative), but this language has not retained the prototypical scheme of Daghestanian case systems, as will be seen in Section 4.4.1. 2.2.3. European analogies European languages which have the inflectional category of case have a comparable set of core cases; to mention some other ergative languages: Georgian, Svan (both from the Kartvelian family), and Basque (isolate, Saltarelli 1988: 300). Accusative languages have the following core cases: nominative, accusative, dative, and genitive. Such an inventory is abundantly represented in Indo-European (e.g., in Ancient Greek, Latin, German, Russian) and Turkic languages, including Azerbaijani (Gadžieva 1966: 73), Gagauz (Pokrovskaja 1966: 118), Karachai-Balkar (Xabiˇcev 1966: 219), Nogai (Baskakov 1966: 286). Kalmyk (Mongolian family, Todaeva 1968: 39) realizes the same pattern with syncretism of dative and locative. Uralic languages usually have a reduced set of core cases. The function of the accusative is often divided between other cases, such as genitive and nominative in Finnish (Tauli 1966; Timberlake 1975), or genitive, nominative, and partitive, depending on the number and definiteness of the noun, in Estonian (Estonskij jazyk 1993: 121). However, some Uralic languages, including the Kildian dialect of Saami (Kert 1971: 162) and Hungarian, do have a separate accusative. Some Uralic languages also lack the genitive (Mansi, see Rombandeeva 1993: 290) or syncretize it with the dative (Hungarian, see Moravcsik 1994: 11). Similar reduced systems are also found in some Turkic languages. For example, Bashkir (Juldašev 1966: 179) and Tatar (Zakiev 1966: 144) lack the dative as a

✐

✐ ✐

✐

✐ 42

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #66

✐

✐

42

Aleksandr E. Kibrik

separate case, and there is syncretism of accusative and dative in Chuvash (Andreev 1966: 48), of dative and lative in Karaim (Musaev 1966: 264), and of accusative and genitive in Kumyk (Magomedov 1966: 199). A maximally reduced system is found in Adyghe.3 Only two cases are distinguished here: direct and oblique, with the unmarked direct as equivalent to the nominative/absolutive of ergative languages and with the oblique being used in all other actant functions, including the ergative and other core grammatical cases. Most of the peripheral cases attested in Daghestanian languages have analogues in other European languages. An instrumental as a separate case is present in most Slavic languages, including Russian, Serbo-Croat, Slovene, Polish (Comrie & Corbett 1993). A comitative is found in Uralic languages such as Finnish and Estonian, while Hungarian has a combined instrumental-comitative and also a causalpurposive (Moravcsik 1994: 11). A partitive is a common case in most Uralic languages. Kalmyk has instrumental, comitative, and conjunctive (soedinitel’nyj, Todaeva 1968: 39) as peripheral cases.

2.2.4. Assorted analogies outside Europe Most Australian languages have the following core and peripheral cases: nominative, ergative, accusative (for pronouns), dative, and causal (Blake 1987: 32). Most commonly an instrumental coincides in form with the ergative; in a minority of languages it instead coincides with the locative (Dixon 1976: 314). The existence of an inflectional genitive is problematic (cf. Blake 1987: 31; Dench & Evans 1988: 10– 13). Its function is expressed by what is sometimes called a “pre-case” suffix, which is followed by a copy of the case of the head noun (exemplifying Suffixaufnahme, see Plank 1995). A similar case inventory is attested in Kiranti languages (Sino-Tibetan; Ebert 1994): nominative, ergative/instrumental, genitive, comitative, and dative (limited to some contexts). Palaeo-Siberian languages, both isolates and families within this areal group, also have grammatical cases with by now familiar functions and syncretisms. Thus, Naukan Eskimo has nominative, ergative/genitive/causal (also known as “relative”, dative/allative, instrumental/ablative, and comparative (Menovšˇcikov 1975: 79–100). Koryak has nominative, ergative/instrumental, dative, comitative (expressed by a circumfix rather than a suffix, untypically for genuine cases), causal, and designative (‘become somebody’) (Žukova 1972: 98–123). Aleut has a minimal set of cases, distinguishing only between nominative (or absolutive, zero-marked) and relative (marker -m, attached only to singular nouns), with relative combining the functions of genitive and ergative (Menovšˇcikov 1968: 391, 403).

✐

✐ ✐

✐

✐ 43

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #67

✐

✐

Nominal inflection galore

43

Table 5. Localization concepts SUPER SUB POST IN INTER AD APUD CONT CUM LOC POSS

‘on OP’ ‘below OP’ ‘behind OP’ ‘inside, within a closed space of OP’ ‘within a filled continuous space of OP’ ‘adjacent to OP’ ‘near to but not touching OP’ ‘contiguous, in complete contact with OP’ ‘near to but in an unspecified position relative to inanimate OP, together with animate OP’ ‘in a typical position relative to OP’ ‘belonging to animate OP’

2.3. Spatial forms Considering only grammatical cases, Daghestanian languages thus have only modestly sized systems, commensurate with those of other European languages. The celebrated richness of Daghestanian case systems is due to the numerous spatial forms of nominals, sometimes amounting to some three dozen, comprising local cases and localization.4 However, whereas the grammatical cases lack systematic formal and functional interrelations among each other, spatial forms have a clear internal structure and are formally and semantically analysable as combinations of items. On these grounds it is advisable to divide up case into two categories, cases proper, as dealt with in the preceding section, and spatial forms, comprising local cases and localization. What I call localization is a set of forms which has to do with the nature of an area relative to a point of orientation, as defined by the object denoted by the noun. Local cases by contrast express directional meanings.5 The standard spatial nominal form thus has the structure STEM-{LOCALIZATION-DIRECTION}. 2.3.1. Localization Although Daghestanian languages differ greatly with respect to the meanings falling under the category of localization, there are primary localization concepts which recur in different languages, entering various combinations. It is currently impossible to be confident about the comprehensiveness of a list of accurately described localization concepts expressed in the Daghestanian languages, but Table 5 should include most of them.6

✐

✐ ✐

✐

✐ 44

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #68

✐

✐

44

Aleksandr E. Kibrik

By localization the position of an object is specified relative to an orientation point, but sometimes there is additional marking for non-specific or approximate localization. Thus, Bezhta has a suffix -da (directly following the primary localization marker) for ‘in (the direction of) the area of whatever is specified by localization proper’, called “extended localization” (EXT) in Kibrik & Testelec (in press): (9) a.

b.

2.3.2.

do märä-L’ä-Ø eˇcenäj. I mountain-SUPER-ESS stay.PRS ‘I stand on the top of the mountain.’ eˇcenäj. do märä-L’ä-dä:-Ø I mountain-SUPER-EXT-ESS stay.PRS ‘I stand (somewhere) in the area of the top of the mountain.’ Direction

Markers of direction specify different kinds of movement, or also non-movement, of a spatially oriented object relative to the localization value (symbolized by LO CAL). In Daghestanian the category of direction standardly comprises three terms, as illustrated by Lezgian, with APUD as the value of localization: (10) a.

b.

c.

ESSIVE : ‘to be in LOCAL (OP)’ sew-re-w-Ø ‘near the bear’ LATIVE: ‘to move toward LOCAL (OP)’ sew-re-w-di ‘toward the bear’ ELATIVE: ‘to move away from LOCAL (OP)’ sew-re-w-aj ‘away from the bear’

But there are also further distinctions of direction expressed in Daghestanian, summarily listed in Table 6. Semantically, lative, allative, and terminative share the notion of approaching the orientation point, while elative and ablative share that of moving away from it, and specific terms in both groups distinguish degrees of contact in the process of movement with whatever is specified by localization relative to the orientation point. Thus, lative, unlike allative, implies some contact during the motion, while terminative implies final contact. Elative, unlike ablative, implies primary contact; translative implies spreading all over the area of localization. Archi overtly distinguishes most of these directional meanings. The five lines of Table 7 show the localization distinctions and the six columns of the local cases, which defines a potential maximum of 30 spatial forms, but there are two gaps in the

✐

✐ ✐

✐

✐ 45

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #69

✐

✐

Nominal inflection galore

45

Table 6. Directional concepts ESSIVE LATIVE ALLATIVE ELATIVE ABLATIVE TRANSLATIVE TERMINATIVE

‘to be in LOCAL (OP)’ ‘to move toward LOCAL (OP)’ ‘to move in the direction of LOCAL (OP) without necessarily reaching it’ ‘to move away from LOCAL (OP)’ ‘to move in the direction away from LOCAL (OP)’ ‘to move across, along LOCAL (OP)’ ‘to move to LOCAL (OP) and reach it, but not further’

Table 7. Spatial forms in Archi

Direction

Localization

IN INTER SUPER CONT SUB

ESSIVE

ELATIVE

LATIVE

ALLATIVE

TERMINATIVE

TRANSLATIVE

-a -qI -t — -L’

-aš -qIaš -¯tiš -raš -L’aš

-ak -qIak -¯tik -rak -L’ak

-aši -qIaši -¯tiši -raši -L’aši

-akana -qIakana -¯tikana -rakana -L’akana

-aXut -qIaXut -¯tiXut — -L’aXut

line for contiguous localization. Usually the meaning of spatial forms is fully compositional, as shown in (11), but for the translative its context of use is a contributing factor (12). (11) a.

b.

c.

d.

cˇ ’eleli-t-Ø stone-SUPER - ESSIVE ‘on the stone’ cˇ ’eleli-¯t-ik stone-SUPER - LATIVE ‘to the stone’ cˇ ’eleli-L’ stone-SUB - ESSIVE ‘under the stone’ cˇ ’eleli-L’-iš stone-SUB - ELATIVE ‘from under the stone’

✐

✐ ✐

✐

✐ 46

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #70

✐

✐

46

Aleksandr E. Kibrik

¯ Qajriplan dunilli-¯ti-Xut parXar. plane sky-SUPER - TRANSL fly.PRS ‘The plane is flying across the sky.’

(12)

While it is the rule for local cases to refer to the localization area relative to the orientation point, other orientations are not excluded. Thus, in Tabasaran, movement away from the starting point (i.e., LOCAL (OP)) can be specified also with reference to an end point, oriented relative to the speaker. The elative is then followed by an exponent consisting of localization (LOCAL (Speaker)) plus allative markers (Magometov 1965: 119): (13) a. b. c. d. (14) a.

b. c.

-an-mi-na -ELATIVE-here-ALL -an-ti-na -ELATIVE-there-ALL -an-Ri-na -ELATIVE-uphill-ALL ¯ -an-ki-na -ELATIVE-downhill-ALL daR-ˇZi-l-an mountain-OBL - SUPER - ELAT ‘from the mountain’ daR-ˇZi-l-an-mina ‘to here from the mountain’ daR-ˇZi-l-an-tina ‘to there from the mountain’, etc.

The categorial status of these exponents is not clear. Magometov calls them “adverbpostpositions”, although they are not used independently. It is quite probable that in contemporary Tabasaran such forms are somewhere intermediate between analytic and synthetic constructions, and it will be seen subsequently that spatial forms in general are of a rather ambiguous categorial nature. 2.3.3. Uralic analogies Among European languages an elaborate system of spatial nominal forms is also found in Uralic. Uralic languages usually distinguish between the two inflectional localizations of internal (IN) and external (AD) and the three directional meanings of essive, lative, and elative, as shown in Table 8 for Hungarian (after Moravcsik 1994: 11–12; Majtinskaja 1994: 265). The morphological structure of these forms, which is not immediately apparent, becomes clearer when compared with the inflection of spatial

✐

✐ ✐

✐

✐ 47

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #71

✐

✐

Nominal inflection galore

47

Table 8. Spatial nominal inflections in Hungarian

localization

IN AD

direction ESSIVE

LATIVE

ELATIVE

-ban/-ben -nál/-nél

-ba/-be -hoz/-hez/-höz

-ból/-b˝ol -tól/-t˝ol

Table 9. Inflection of spatial postpositions in Hungarian POSTPOSITION

ESSIVE

LATIVE

ELATIVE

‘under’ ‘in front of’ ‘among’ ‘behind’

al-att el-t˝ot köz-ött mög-ött

al-á el-é köz-é mög-é

al-ól el-ól köz-ül mög-ül

Table 10. Spatial nominal inflections in Finnish (with vowel harmony variants omitted)

localization

IN AD

direction ESSIVE

LATIVE

ELATIVE

-s-sa -l-la

-hVn -l-le

-s-ta -l-ta

postpositions (see Table 9), inflecting regularly for directional meanings – i.e., -Ø/ -V for lative, -tt/-Vtt for essive, and -l/-Vl for elative (ignoring vowel harmony and the problem of whether the vowel belongs with the stem or the suffix). Thus, some nominal suffixes in Table 8 can be recognized as being morphologically complex: -b-a/-e and -b-ól/-˝ol, with inessive -b-an/-en having an idiosyncratic essive exponent -an/-en. INAD cases, however, resist such an analysis. In Finnish (Eliseev 1994: 99), the external spatial cases, with the localization meaning AD, preserve the complex morphological structure in all forms, while the IN meaning has an amalgated marker -hVn in the illative, with the quality of the vowel being phonologically determined; see Table 10. In general, nominal spatial forms in Uralic are not as morphologically transparent as they normally are in Daghestanian languages, and the category of localization has therefore not found favour in traditional descriptions, despite the regular use of case labels such as inessive, illative, and elative. The tendency to express spatial meanings in nominal inflection is something the Uralic and Daghestanian languages

✐

✐ ✐

✐

✐ 48

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #72

✐

✐

48

Aleksandr E. Kibrik

have in common, but the kind of their expression is more case-like in the former than in the latter. 2.3.4. Turkic analogies Turkic languages have a reduced system of spatial forms similar to that of Khinalug. Localization lacks inflectional recognition, while locative and ablative are usually distinguished as directional meanings, the latter being formally derived from the former by -n (Gadžieva 1966: 71; Pokrovskaja 1966: 117; Magomedov 1966: 199; Andreev 1966: 48). As is also shown in Table 11, Turkic languages with a tripartite directional distinction express the lative meaning by an independent marker (Zakiev 1966: 144; Juldašev 1966: 179; Musaev 1966: 265). Karachai-Balkar has a translative case for movement along the orientation point (suffix -(t)Vn; Xabiˇcev 1966: 219) in addition to locative and ablative. Inflected spatial postpositions, as characteristic of Daghestanian and Uralic, are not found in Turkic. 2.3.5. Assorted analogies outside Europe In Naukan Eskimo there are separate locative and prolative (‘along’) cases, while basic directional meanings are combined with grammatical functions: allative with dative and ablative with instrumental (Menovšˇcikov 1975: 84–98). Koryak (Žukova 1972: 99–101) and Ket (Krejnoviˇc 1968: 457) have a standard set of directional cases: locative, allative, ablative, and also prolative. The way of expressing orientational meanings in Koryak and the related languages Alutor and Chukchi (Kibrik et al. 2000: 254, Skorik 1961: 315–318) is unusual, as it involves derivational nominal suffixes of local orientation. For example, in Koryak, suffixes such as -lq SUPER, -giN SUB, or cˇ @ku IN derive new nouns with a full nominal paradigm including a nominative (Žukova 1972: 88–90, 110):

Table 11. Spatial nominal inflections in Turkic languages

Azerbaijani Gagauz Kumyk Chuvash Tatar Bashkir Karaim

LOCATIVE

ABLATIVE

LATIVE / DATIVE

-dV -dV/tV -dV -rV/tV -dV/tV -dV -dV/tV

-dV-n -dV-n/tV-n -dV-n -rV-n/tV-n -dV-n/tV-n/nV-n -dV-n -dVn/tV-n

-(j)V -(j)V -ge/Ra/V -V -gV/kV -RV/qV -RV/kV

✐

✐ ✐

✐

✐ 49

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #73

✐

✐

Nominal inflection galore

49

Table 12. Spatial inflections in Hua

Localization

IN AD

(15) a.

b.

c.

d.

Direction ESSIVE

LATIVE

ELATIVE

-vi’ -ro’

-vin-ga -ro-ga

-vin-ga-ri/vi-ti’ -ro-ga-ri/ro-ti’

utt@-ut tree-NOM ‘tree’ utt@-k tree-LOC ‘near the tree’ utt@-giN-k@ tree-SUB-LOC ‘under the tree’ utt@-giN-@n tree-SUB-NOM ‘surface under the tree’

There are languages with a similar formal distinction of localization and direction in various other parts of the world. For example, in Papua New Guinea, Hua has markers of localization for IN and AD which are conjoined with the directional markers of essive (-Ø), lative (-ga), and elative (-ti’/ga-ri); see Table 12 (after Haiman 1980: 232, where all spatial forms are called cases). There are also Papuan languages with a single locative marker and derived ablative and allative cases, e.g., Kâte (LOC -o, ABL -o-pek, ALL -o-nek) and Selepet (see Foley 1986: 101). Most of the Kiranti languages (Sino-Tibetan) may introduce an additional orientation point in localization, which is also possible in Tabasaran. Bantawa, Camling, Thulung, and Khaling have the set of localizations in Table 13 (“vertical cases” in the terms of Ebert 1994). These localizations are combined with three directional meanings of locative, allative, and ablative, as shown in Tables 14a/b and as exemplified for Camling in (16). (16) a.

b.

khim-dhi-ka house-HI-ABL ‘from the house higher up’ khim-ya-ni house-LEV-ALL ‘over to the house’

✐

✐ ✐

✐

✐ 50

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #74

✐

✐

50

Aleksandr E. Kibrik

Table 13. Localizations in Kiranti

unmarked, ‘be in some space adjacent to OP’ with the additional meaning ‘below the position of the speaker’ ‘above the position of the speaker’ ‘on the same level as the speaker’

LOC

loLOC hiLOC levLOC

Table 14. Spatial forms in Kiranti

Additional OP

a.

Localizations unmarked LO HI LEV

b.

Languages Bantawa

Camling

Thulung

Khaling

-da -yu -du -ya

-da -i -dhi/-di -ya

-da -yu -la -no

-bi -yu -tü -yo

-Ø -nin -ka

-Ø -ni -ka

-Ø

-Ø (-tha) -ka/-laka

Spatial cases LOCATIVE ALLATIVE ABLATIVE

-m

2.4. Double case marking Adding to morphological complexity, some of the Daghestanian languages have double case marking. 2.4.1. Attributes with head present In Bezhta the genitive (distinguished as direct and oblique, depending on the direct or oblique case of the head noun) can be added to nouns that are already case-marked, especially with the comitative: (17) a.

b.

Qädäm-li-Rör-sö jeL’ä-Pä. kid girl.NOM man-OBLl-COM-GEN.DCT leave-NEG ‘The girl (who is) with the man does not leave.’ wahalo Qädäm-li-Rör-lä suk’o-l jeˇgä:näj. wahago kid man-OBL-COM-GEN . OBL person-DAT see.PRS this girl.NOM that ‘The person (who is) with that man sees this girl.’

✐

✐ ✐

✐

✐ 51

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #75

✐

✐

Nominal inflection galore

51

Following upon a comitative suffix (-Ror), the genitive is direct (-so) in (17a), being an attribute of a nominative head, and oblique (-la) in (17b), where the head noun is dative, the oblique case governed by the verb. (The vowels of genitive suffixes harmonize with the root vowels.) 2.4.2. Attributes with head missing If its head is missing, the genitive behaves like an independent nominal and may, along with other cases, acquire a further genitive marker. To illustrate again from Bezhta: (18) a.

b.

abo-s k’obala biXalo gäxi, usup-i-la Jusup-OBL-GEN . OBL father-GEN . DCT stick.NOM long be, bäsijö gäxi. ibrahim-i-la-s Ibrahim-OBL-GEN . OBL-GEN . DCT short be ‘The stick of Jusup’s father is long, that of Ibrahim’s is short.’ abo-s k’obala ibrahim-i-la Ibrahim-OBL-GEN . OBL father-GEN . DCT stick.NOM ‘the stick of Ibrahim’s father’

As the non-elliptical version of the subject NP of the second part of this sentence shows (18b), the first genitive suffix on Ibrahim is oblique because its (elided) immediate head (abo-s), being itself an attribute, would be in the genitive (i.e., in an oblique case), and the second genitive is direct because the (elided) topmost head (k’obala) would be nominative (i.e., in the direct case). In such recursive attributions in Bezhta it is always only the immediate head which is reflected in the direct or oblique forms of genitives. In Archi an attributive noun with an adjectivizing suffix may occur independently, and is then inflected like the nominal head of an NP: (19)

rang haIt@r-ˇce-qIa-š-du-m-mi-n river-OBL-INTER-ABL - ADJCT- III - OBL - GEN colour.NOM ‘the colour of the one from the river’

In (19) the attribute is used as an NP in its own right, with its head elided, and it functions as a nominal attribute, requiring the oblique stem suffix -mi (which causes assimilation in the preceding class marker) and the genitive -n. Kubachi is among the languages where two genitives may come to co-occur within a single word in this manner (cf. Magometov 1963: 109): (20) a.

milq’a a:s-la silver-GEN spoon.NOM ‘silver spoon’

✐

✐ ✐

✐

✐ 52

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #76

✐

✐

52

Aleksandr E. Kibrik

b.

q¯ ulluq a:s-la-l-la silver-GEN - OBL - GEN business.NOM ‘the business of the one of silver’

Only the oblique stem formative, required by nouns taking an oblique case in Kubachi, intervenes between the two genitives on the attribute in (20). Locatives and an orientation marker are crucial components of a special type of such multiple case marking in Archi, viz., possessive locatives. Thus, in (21), olo, the suppletive genitive of the 1st person plural pronoun nen, acts like an independent nominal. (21)

cˇ ’eba olo-ma-ši! go.IMP we.GEN - ORIENT- ALL ‘Come to our house!’

It is the orientation marker -ma, preceding a spatial case suffix, including allative, which adds the meaning component ‘typical place for X’, where X is the referent of the relevant pronoun or human noun. Whereas locative forms are normally based on the oblique stem, the basic genitive here would seem to suggest an implicit head, ‘the one of X’. Considering that the genitive prototypically expresses possessive relations and that among one’s most typical possessions is the place one habitually occupies, the meaning ‘place where X lives’ is the most obvious one to infer. As to the historical origin of such forms with -ma, it should be noted (cf. Kibrik 1977: 105) that Archi has similar forms of locative gerunds with -ma as a localization marker (‘there where’) followed by standard locative case markers, as in (22). (22)

L’am-ma-ši uqIa! was DAT ORIENT . ALL likego.IMP you. ‘Go where you like!’

Nouns usually have no such marker. The word biq’◦ ‘place’ is one of the few exceptions, having the regular oblique marker -li (biq’◦ -li ERG), but showing a more archaic form in the inessive, biq’◦ -m-a ‘in the place’, where -m is OBL, -a is a marker of orientation for ‘in’, and the essive remains unmarked. The oblique -mi is regular for nominalized attributes such as demonstrative pronouns (e.g., to-r ‘that’, i.e., woman, belonging to class II, marked by -r), which, when occurring independently, can be used in the oblique cases (e.g., to-r-mi-n that-II-GEN . OBL ‘her’). Thus, -ma is not originally simple, but consists of an oblique and an orientation marker. Accordingly, converbs as in (22) can be historically analysed as nominalizations of verbs (L’an ‘to love, like’), inflected with the aid of the oblique marker -m(i), the orientation marker -a, and case endings. The history of the oblique, orientation, and case markers of personal pronouns, as in (21), would seem to be the same. What remains to be explained here is the genitive of the pronoun. Lacking a verb ‘to have’, Archi puts the possessor into the genitive in clauses with an existential verb, from which a locative

✐

✐ ✐

✐

✐ 53

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #77

✐

✐

Nominal inflection galore

53

gerund can be derived by means of -ma. Inflected genitives such as olomaši would then be explainable as historical reductions of full clauses such as (23), with a noun redundantly identifying the possession and an existential verb omitted. (23)

noL’ i-ma-ši olo we.GEN house be-ORIENT- ALL ‘towards where we have a house; where our house is’

2.5. Class 2.5.1. Classification With the exception of Lezgian, Agul, Udi, and partly Tabasaran, Daghestanian languages share the classificatory category of class or gender.7 Normally, class membership is a lexical property of nouns which has no overt expression on the nouns themselves (exceptions to be discussed below) but makes itself felt in their syntactic environment, controlling the form of other words – the targets of rules of agreement. Moreover, sometimes the class membership of a noun predetermines its inflectional class (see Section 3.1.2 below, and compare Corbett 1982). Most widespread in Daghestanian languages is a system with four classes, as found in Tsakhur, Rutul, Khinalug, Kryz, and elsewhere. The first two are semantically highly transparent, with class I comprising all nouns denoting male humans and class II all nouns denoting females. Membership in classes III and IV is often arbitrary, but there are many minor homogeneous clusters, such as big or small, domestic or wild animals, domestic or wild plants, insects, trees, instruments, abstract nouns. The three-class system, as exemplified in Avar and in the Andic group (Godoberi, Akhvakh, Bagvalal, etc.), is more transparent, insofar as the male and female classes are complemented by a class whose membership is also semantically defined, comprising nouns denoting everything not male or female (i.e. neuter). The two-class system, as in the northern dialects of Tabasaran, divides nouns into human and non-human. In some languages classes are more numerous. For example, Andi and Chamalal distinguish five, Hunzib six (Bokarev 1959), and Archi eight (Kibrik 1994). 2.5.2. Class agreement In all Daghestanian languages class serves as an agreement category, with verbs, adjectives, certain other words, and sometimes also nouns themselves agreeing in class (expressed cumulatively with number) with a noun inherently specified for class. Limiting ourselves to such nominal agreement inflection, we thus find nouns in the essive, a spatial case, also showing class (and number) agreement in Megeb (Dargic

✐

✐ ✐

✐

✐ 54

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #78

✐

✐

54

Aleksandr E. Kibrik

Table 15. Partial paradigm of bidra ‘bucket’ in Megeb

AD

SUPER

Table 16.

AD

AD

ESSIVE

LATIVE

ELATIVE

bidra-li-šu=b/r OBL - AD - ESS . III. SG / PL ‘near the bucket’ bidra-li-ˇcu=b/r OBL - SUPER - ESS . III. SG / PL ‘on the bucket’

bidra-li-šu-Ø OBL - AD - LAT ‘to the bucket’ bidra-li-ˇcu-Ø OBL - SUPER - LAT ‘up to the bucket’

bidra-li-šu-la OBL - AD - EL ‘from the bucket’ bidra-li-ˇcu-la OBL - SUPER - EL ‘off the bucket’

spatial forms in Dargwa LATIVE

ESSIVE

ELATIVE

ALLATIVE

-ˇc

-ˇci=b

-ˇci=b-ad

-ˇci=b-eH

Table 17. Partial paradigm of aa ‘house’ in Lak

IN

Table 18.

IN

IN

ESSIVE

LATIVE

ALLATIVE

q¯ a¯t-lu-wu OBL - IN . ESS

q¯ a¯t-lu-wu-n OBL - IN - LAT

q¯ a¯t-lu-wu-n=m/n/Ø-aj OBL - IN - LAT = III / II / I - ALL

spatial forms in Andalal ESSIVE

LATIVE

ELATIVE

TRANSLATIVE

=w/j/b=Ø

=w/j/b=e

-sa

-sumo

group, Table 15), whereas latives and elatives do not undergo such agreement. Somewhat more complicatedly, in literary Dargwa (Table 16, Abdullaev 1967) the lative and the essive have zero marking and class marking respectively, like in Megeb, but the elative is derived from the essive (i.e. contains class markers), as is the allative. Lak (Table 17, Murkelinskij 1967: 493) differs from literary Dargwa in that class exponents appear only in the allative case, in addition to the normal allative marker, with the allative being based on the more elementary lative. Andalal (Sogratl dialect) also uses class exponents in the spatial forms, but in the domain of localization rather than the spatial cases, for marking IN localization (in the context of essive and lative; Kibrik 1993b: 27); see examples in Tables 18 and 19. In the Chadakolob dialect of Antsukh (Table 20) class exponents are used in forming the dative.

✐

✐ ✐

✐

✐ 55

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #79

✐

✐

Nominal inflection galore

55

Table 19. Partial paradigm of roso ‘village’ in Andalal ESSIVE IN

LATIVE

ELATIVE

TRANSLATIVE

ros-no=w/j/b=Ø ros-no=w/j/b=e ros-no-sa ros-no-sumo OBL = I / II / III = ESS OBL = I / II / III = LAT OBL - ELAT OBL - TRANSL ‘in the village’ ‘to the village’ ‘from the village’ ‘through the village’

Table 20. Partial paradigm of dede ‘father’ in Chadakolob

dede-Ø father-NOM

dede-d father-ERG

dede-l father-GEN

dede=w/j/b/r=e father=I / II / III / PL = DAT

Table 21. Partial paradigm of ima ‘father’ in Botlikh

ima-Ø father-NOM

im-u-di father-OBL - ERG

im-u=b father-OBL - GEN 2

im-u-j father-OBL - DAT

The Andic languages use class exponents with the second genitive, as exemplified from Botlikh in Table 21 (Gudava 1967: 297). In Archi, finally, the 1st person personal pronouns ‘I’ and ‘we’ have irregular dative and genitive forms accompanied by agreement class markers; see Table 24 in Section 3.1.1.

3. Systems of nominal inflection In a Daghestanian language nominals do not all inflect alike; there are different nominal word classes and paradigm types, different patterns of stem formation, and different declensions.

3.1. Word classes and paradigm types The category of nominals subsumes the following morphological word classes: primary (underived) nominals: secondary (derived) nominals:

nouns pronouns (of 1st and 2nd person) pronouns (of 3rd person) adjectives participles numerals masdars

✐

✐ ✐

✐

✐ 56

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #80

✐

✐

56

Aleksandr E. Kibrik

Nouns are the most natural, prototypical nominals, possessing most nominal properties. Other nominal word classes are usually defined in relation to primary nouns. Personal pronouns, as elsewhere in the languages of the world, deviate formally in some respects from primary nouns. Derived nominals are of verbal or attributive origin. Masdars result from the nominalization of the verb or predicate adjective in a clause with a nominal role (see Koptjevskaja-Tamm 1992), and usually are overtly marked as derived; cf. masdar forms such as haQ-ir from the verb haQa ‘to see’ or hadu-ìir from the adjective hadu ‘deep’ in Godoberi. Nominalized attributives usually are not overtly marked as derived. 3rd person pronouns are derived from demonstrative pronouns. Headless adjectives, participles, and numerals behave like nouns, taking the inflectional categories surveyed above. Nevertheless, pronouns and secondary nouns usually have some inflectional peculiarities which deserve to be mentioned. 3.1.1. Inflectional peculiarities of pronouns of 1st and 2nd person First, in contrast to other nominals, 1st and 2nd person pronouns tend to neutralize the distinction between nominative and ergative. Such neutralization, providing crucial evidence that nominative and ergative were non-distinct for 1st and 2nd person pronouns in Proto-Daghestanian (see below, 6.1.1), is found in Andi and Godoberi in the Andic group, in Tsez and Bezhta in the Tsezic group, in Tabasaran and Agul in the Lezgic group, in Kryz and Budukh in the Kryzic group, and among the “isolates” in Lak and Udi. Chamalal, Hinukh, Chirag, Khinalug neutralize nominative and ergative only in the plural of 1st and 2nd person pronouns, while Tsez does so only in their singular, and Archi and Botlikh only in the singular of the 1st person. However, several contemporary Daghestanian languages – Avar, Karata, Akhvakh, Bagvalal, Tindi, Khvarshi, Lezgian, Rutul, Tsakhur – do distinguish nominative and ergative in the pronouns as well as in common nouns. Table 22 presents relevant fragments of pronominal paradigms, giving nominative, ergative, and dative forms (in that order, separated by a slash). Second, personal pronouns sometimes have suppletive inflection with oblique cases using a separate root, unrelated to the direct (nominative) form of the same pronoun. This is the case in Lak (Table 23); suppletive roots are also seen in Table 22 (e.g., ‘thou’ in Bezhta mi/du-, Tsez mi/deb-, Archi un/wa-, Khvarshi mo/dub-) and in Table 24 (‘we’ in Archi nen/olo). Finally, pronouns are notable for various other irregularities in their inflection: compare the inflection of the regular noun ‘cow’ with that of personal pronouns in Archi (Table 24), which is typical in this respect for Daghestanian languages. The oblique cases of the noun use the oblique stem, derived from the root with the exponent -i, with regular case suffixes added (-Ø for ergative, -n for genitive, -s for dative, -Xur for comparative). The oblique stems of the pronouns are clearest and

✐

✐ ✐

✐

✐ 57

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #81

✐

✐

57

Nominal inflection galore

Table 22. Types of syncretism in the paradigms of personal pronouns

Bezhta Chirag Tsez Botlikh Khvarshi

‘I’

‘thou’

‘we’

‘you’

do/do/di-l du/di-¯ce/dam-i: di/di/dä-r den/iškur/di-j do/d-e/di-l

mi/mi/du-l Qu/Qi-¯ce/Qat mi/mi/deb-er min/min/du-j mo/m-e/dub-ul

ile/ile/ilo-l nu¯sa/nuža/nu¯s-i: eli/el-a:/el-ur iˇ¯si/iˇ¯si/iˇ¯si-j ilo/il-e/il-ul

miže/miže/mižo-l nuˇ¯sa/nuˇ¯sa/nuˇ¯s-i: meži/mež-a:/mež-ur bišti/bišti/bišti-j mižo/miž-e/miž-ul

Note: For Chirag and Botlikh, which distinguish inclusive and exclusive 1st plural pronouns, only the exclusive forms are given.

Table 23. Personal pronouns in Lak

‘I’ ‘thou’

NOM

ERG

GEN

DAT

na ina

na ina

¯tu-l wi-l

¯tu-n wi-n

Table 24. Partial paradigms of nouns and personal pronouns in Archi

NOM ERG GEN DAT COMPAR

‘cow’

‘I’

‘thou’

‘we (EXCL)’

‘ye’

XIon XIin-i XIin-i-n XIin-i-s XIin-i-Xur

zon za-ri b=is b=ez za-Xur

un un wit wa-s wa-Xur

nen nen b=olo b=el la-Xur

ž◦ en ž◦ en wiš wež ž◦ a-Xur

most regular in the comparative (the same oblique stem is used in other oblique cases not included in this partial paradigm): za-, wa-, la-, and ž◦ a-, respectively. However, these oblique stems do not correspond regularly to those in the core cases of ergative, genitive, and dative, with the latter two especially irregular. The only exceptions are dative wa-s ‘thou’ and the analogically re-formed ergative za-ri ‘I’. Formal similarities between stems that can possibly be detected (e.g., between zon and za-, un and wa-, ž◦ en and ž◦ a-) are merely historical. Unlike nouns, pronouns do not have any separable suffixal markers for genitive and dative (with the exception of dative wa-s ‘thou’). Moreover, the 1st person pronouns ‘I’ and ‘we’ have class exponents in the genitive and dative (with b= for class III in Table 24).

✐

✐ ✐

✐

✐ 58

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #82

✐

✐

58

Aleksandr E. Kibrik

3.1.2. Ordinary vs. attributive declension The distinction between primary and secondary (basic and derived) nouns is a matter of derivational morphology, but sometimes it also matters for inflection. For example, Godoberi distinguishes between an ordinary declension for primary nouns and masdars and so-called attributive declensions for primary attributes, i.e. nominalized adjectives, participles, and numerals, and also a few primary masculine nouns (such as ‘director’), as shown in Table 25. The ordinary declension is illustrated by a partial paradigm of the primary noun ‘house’, whose ergative and genitive are formed with the suffixes -di and -Li (the so-called genitive 2). When they lack a head, adjectives like ‘thin’, numerals like ‘one’, and demonstrative pronouns like ‘that’ inflect according to the attributive declension. They inherit the class/gender of the implied head noun and have anaphoric class suffixes in the nominative (e.g., ce=w/j ‘one’) and class prefixes in all cases (for overtly marked classes; e.g., w=ollara-š-¯ti ‘thin (masculine ergative)’). Unlike the ordinary declension, the attributive declension regularly distinguishes masculine from feminine and neuter in ¯ as oblique exponents and genitive class/gender markers the singular, using -šu vs. -łi (genitive 1) vs. the more general -Li marker (genitive 2). In the plural of the attributive declension the human and neuter genders are differentiated by the oblique plural markers -r-du vs. -r-di and by genitive 1 vs. genitive 2, as shown in Table 26.

Table 25. Ordinary vs. attributive declension in Godoberi (singular, partial)

‘house (NEUT )’ ‘director (MASC )’ NOM ERG GEN

NOM ERG GEN

‘thin (MASC )’

‘thin (FEM )’

hanq’u hanq’u-di hanq’u-Li

derektor w=ollara=w j=ellara=j derektor-aš-¯ti w=ollara-š-¯ti j=ellara-ì-¯ti ¯ derektor-ašu=w/j/b w=ollara-šu=w/j/b j=ellara-ìi-Li

‘one (MASC )’

‘one (FEM )’

‘that, he (MASC )’

‘that, she (FEM )’

ce=w ce-š-¯ti ce-šu=w/j/b

ce=j ce-ì-¯ti ¯ ce-ìi-Li

ho=w ho-š-¯ti ho-šu=w/j/b

ho=j ho-ì-¯ti ¯ ho-ìi-Li

Table 26. Attributive declension in Godoberi (plural, partial) HUMAN ( MASC NOM ERG GEN

& FEM )

b=ellara(l) b=ellara-r-du-di b=ellara-r-du=b/j

NEUTER

r=ellara(l) r=ellara-r-di-di r=ellara-r-di-Li

✐

✐ ✐

✐

✐ 59

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #83

✐

✐

Nominal inflection galore

59

This declensional distinction, with some variation, is highly developed in the Andic languages. In the Avaric languages there is no opposition of two genitives, but one of oblique markers distributed according to the class/gender of a nominal. There are traces of the attributive declension also in other groups, for example in Tabasaran (South dialects, see Magometov 1965: 151–153), Archi, and Khinalug. In Archi the oblique markers -mu/-mi distinguish class I (male) from all other singular classes of nominalized demonstrative pronouns, participles, and numerals (see Table 27). In Khinalug, secondary nouns derived from attributes differ from primary nouns in that they distinguish class I (male), II (female), and III–IV (non-human) oblique markers in the singular, and an oblique marker for class I/II (human) from that of non-human classes (III–IV) in the plural. See Table 28 for the inflection of the substantivized demonstrative pronoun.

Table 27. Attributive declension in Archi (partial)

‘working’ PART

I II III IV

‘that, he’, ‘that, she’ ‘that, it’, ‘that, it’

I II III IV

NOM . SG

ERG . SG

¯ ◦ innu w=irX ¯ ◦ innu=r d=irX ¯ ◦ innu=b b=irX ¯ irX◦ innu=t

¯ ◦ innu=m=mu w=irX ¯ ◦ innu=r=mi d=irX ¯ ◦ innu=m=mi b=irX ¯ irX◦ innu=t=mi

to=w to=r to=b to=t

to=w-mu to=r-mi to=m-mi to=t-mi

Note: The final class marker =w is lost in polysyllabic words, and =w and =b are assimilated to a following /m/.

Table 28. Attributive inflection of demonstrative pronoun in Khinalug (partial paradigm)

NOM ERG COMPAR COM

I SG

II SG

III SG

IV SG

I-II PL

III-IV PL

du R-i Ro-q’ Ro-škili

dä Ru-i Rä-q’ Rä-škili

dä

Zˇ i

dur Roz-i Roz-1q’ Roz-1škili

Zˇ it sedr-i sedr-iq’ sedr-iškili

s-i sä-q’ sä-škili

Note: For some speakers ergative I and II singular forms (R-i, Ru-i) are synonymous. Labialization characterizes the non-male class in the ergative (Rui) and the male class in the comparative and comitative (cf. Archi -mu/-mi). Idiolectal hesitations and the inversion of labialization are evidence for the continuing loss of this opposition.

✐

✐ ✐

✐

✐ 60

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #84

✐

✐

60

Aleksandr E. Kibrik

3.2. Patterns of stem formation 3.2.1. The two-stem pattern Four things can be seen from the partial paradigm of the Archi noun gel ‘cup, noggin’ in Table 29: first, plural is the marked term of the category of number and is expressed everywhere by -um; second, case markers (zero for nominative, ergative, and essive) are in final position and are independent of number; third, the localization marker precedes the marker of spatial case and usually has two allomorphs, one for final and one for non-final position (e.g., -t/-¯ti SUPER, -L’/-L’a SUB); fourth, there are “empty” morphemes -li/-l in the singular and -ˇcaj/-ˇce/-ˇc in the plural oblique cases. The latter observation is of special interest. The paradigm in question contains not only morphemes which separately express number, case, and localization, but also morphemes which combine number and the general meaning of oblique case, with a single category thus expressed more than once in one word (“one-to-many syntagmatic deviation” in the terms of Carstairs-McCarthy 1992: 195). Thus, -li/-l

Table 29. Partial paradigm of gel ‘cup, noggin’ in Archi

Case (+ localization)

SINGULAR

PLURAL

NOMINATIVE

gel gel-li gel-li-n gel-li-s gel-li-Xur ¯ gel-li-ìu gel-li-L’ana gel-li-qIiš gel-l-a gel-l-a-š gel-l-a-k gel-li-t gel-li-¯ti-š gel-li-¯ti-k gel-li-L’ gel-li-L’a-š gel-li-L’a-k gel-li-ra-š gel-li-ra-k

gel-um gel-um-ˇcaj gel-um-ˇce-n gel-um-ˇce-s gel-um-ˇce-Xur ¯ gel-um-ˇce-ìu gel-um-ˇce-L’ana gel-um-ˇce-qIiš gel-um-ˇc-aj gel-um-ˇc-ej-š gel-um-ˇc-ej-k gel-um-e-t gel-um-ˇce-¯ti-š gel-um-ˇce-¯ti-k gel-um-ˇce-L’ gel-um-ˇce-L’a-š gel-um-ˇce-L’a-k gel-um-ˇce-ra-š gel-um-ˇce-ra-k

ERGATIVE GENITIVE DATIVE COMPARATIVE COMITATIVE PERMUTATIVE PARTITIVE IN - ESSIVE IN - ELATIVE IL - LATIVE SUPER - ESSIVE SUPER - ELATIVE SUPER - LATIVE SUB - ESSIVE SUB - ELATIVE SUB - LATIVE CONT- ELATIVE CONT- LATIVE

✐

✐ ✐

✐

✐ 61

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #85

✐

✐

Nominal inflection galore NOM . SG

oblique cases SG

= ←

ROOT ↓ OBL . SG

→

PL

↓ OBL . PL

=

NOM . PL

→

oblique cases PL

61

Schema 1. Stem formation

can be glossed as OBL . SG and -ˇcaj/-ˇce/-ˇc as OBL . PL. This is tantamount to saying that in different parts this paradigm is based on two stems: a direct stem coincides with the root of the noun in the singular and with the pluralized root in the plural; each oblique stem is derived from the direct stem by adding -li to the singular direct stem and -um to the plural direct stem. Schema 1 represents the construction of paradigms of this kind, with arrows symbolizing derivational processes where the element following the arrow is added to the string preceding it and the equal sign means the form is unchanged.

3.2.2. The one-stem pattern While the two-stem pattern is the most widespread, occurring in all groups of Daghestanian languages, it is not the only one. Characteristic of Khinalug is a one-stem pattern, illustrated in Table 30 and schematized in Schema 2. Here all noun forms are directly derived from the root in the singular and from the pluralized root in the plural. (Vowel-final stems lose this vowel before a vocalic suffix, while consonantfinal stems add a vowel before a consonantal suffix; cf. the allomorphs m1da/m1d and -r/-ir/-1r.) While in Khinalug all nouns follow the one-stem principle, in other languages there are many instances where some nouns do, while others don’t.

Table 30. Partial paradigms mida ‘mountain’ and gis ‘roof’ in Khinalug

NOMINATIVE ERGATIVE

1 GENITIVE 2 GENITIVE DATIVE COMITATIVE LOCATIVE ELATIVE

SG

PL

SG

PL

m1da m1d-i m1d-i m1d-e m1d-u m1da-škili m1da-r m1da-lli

m1da-d m1da-d-i m1da-d-i m1da-d-i m1da-d-u m1da-d-1škili m1da-d-1r m1da-d-1lli

gis gis-i gis-i gis-i gis-u gis-iškili is-ir gis-illi

gis-irdir gis-irdir-i gis-irdir-i gis-irdir-i gis-irdir-u gis-irdir-iškili gis-irdir-ir gis-irdir-illi

✐

✐ ✐

✐

✐ 62

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #86

✐

✐

62

Aleksandr E. Kibrik

NOM . SG

=

ROOT

→

PL

=

NOM . PL

oblique cases SG

oblique cases PL

Schema 2. Stem formation

Table 31. Partial paradigm of nuX ‘cave’ in Lak

NOMINATIVE GENITIVE

SG

PL

nuX nuX-li-l

nuX-ru ¯ nuX-ardi-l

3.2.3. Intermediate patterns There are yet further types of inflectional stem formation in Daghestanian languages (reviewed in greater detail in Kibrik 1991). First, the two-stem pattern may be restricted to only the singular or only the plural, yielding two types intermediate between the standard and reduced patterns of Schemas 1 and 2. Second, the singular and plural oblique stems, as well as the nominative plural, may be directly derived from the root, as with the Lak noun nuX ‘cave’ (Table 31). There is here no single consistent marker of plural, whose expression is instead distributed between the cumulative morphemes -ru NOM . PL and -ardi OBL . PL (the final ¯ regains its original quality in prevocalic position). As represented stem consonant Xin Schema 3, in this type oblique plural stem as well as oblique singular stem are derived directly from the root (by adding -ardi and -li, respectively). This method of forming the oblique plural stem can be combined with the onestem pattern in the singular, as exemplified by the noun maIL’u ‘beak’ in Khvarshi (Table 32). Finally, the opposition of direct and oblique stems can be separated from the category of number, as exemplified by p’1z ‘lip’ in Rutul (Table 33). The primary direct stem is equally expanded in all oblique cases, irrespective of number. It is natural NOM . SG

oblique cases SG

←

=

ROOT

→

OBL . SG

← →

OBL . PL

NOM . PL

→

oblique cases PL

Schema 3. Stem formation

✐

✐ ✐

✐

✐ 63

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #87

✐

✐

Nominal inflection galore

63

Table 32. Partial paradigm of maIL’u ‘beak’ in Khvarshi

NOMINATIVE GENITIVE

SG

PL

maIL’u maIL’u-s

maIL’u-bo maIL’u-za-s

Table 33. Partial paradigm of p’1z ‘lip’ in Rutul

NOMINATIVE ERGATIVE

SG

PL

p’1z p’1z-1-r

p’1z-b1r p’1z-1-m1-r

Table 34. Partial paradigm of barc’ ‘wolf’ in Lak

NOMINATIVE GENITIVE

SG

PL

barc’ burc’-i-l

barc’-ru burc’-urdi-l

to regard this exponent (-1) as a pure oblique stem marker which is augmented in the plural by the special OBL . PL marker -m1, while in the singular there is no special oblique marker.

3.2.4. The pure oblique stem strategy A more complex pattern, represented in Schema 4, is found in Lak. For nouns as in Table 34 the direct/oblique opposition is expressed by the (irregular) vowel alternation a/u, and additionally by the oblique singular marker -i and the oblique plural marker -urdi. Rutul exemplifies a reduced variant of Schema 4, lacking the OBL . SG stage (Table 33 above).

=

NOM . SG

oblique cases

←

OBL . SG

←

ROOT ↓ OBL

→ →

OBL . PL

NOM . PL

→

oblique cases

Schema 4. Stem formation

✐

✐ ✐

✐

✐ 64

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #88

✐

✐

64

Aleksandr E. Kibrik

3.2.5. The unmarked/marked root opposition Inflectional patterns as in Table 35 seem highly unusual vis-à-vis the previous schemas. A striking peculiarity of these paradigms is the presence of an extra marker (-je/-j, -am, -a), which sets apart all other forms from the nominative singular. None of the familiar glosses (such as OBL or PL) are suitable to characterize its function. One possible interpretation (proposed in Kibrik 1991) assumes that the initial stem is present everywhere except in NOM . SG, with NOM . SG thus as a derived form. Formally similar paradigms would then be the 3rd declension of Latin and Greek (Tables 36 and 37). The difference between the nominal stems in the nominative singular and the other forms can be explained in terms of the one-stem inflection,8 where the primary stem is present in all forms except NOM . SG, the latter being derived by morphophonological rules (e.g., dropping of final /t, b/ in Latin and /t, d, th/ in Greek before NOM . SG -s). Russian also has several nouns of a similar inflectional pattern: e.g., imja NOM . SG – imen-i GEN . SG – imen-a NOM . PL. Another interpretation of the pattern in Table 35 (see Kibrik & Kodzasov 1990: 255–256) is that nominative singular itself has initial status and that the other basic stem is derived. Formally, the data in Table 35 can be naturally accounted for as ordinary morphological derivations from the bare root used for NOM . SG, while an opposite derivation seems less natural. General considerations about inflectional systems, with bare roots as frequent representatives of NOM . SG in Daghestanian, also support this interpretation. Further, the markers of the “second main stem” are identical to the OBL . SG or PL marker in other words of the same languages (see the Khvarshi paradigms in Table 38). The marker -a (present in all forms of ezol ‘eye’ in Table 35) is the OBL . SG marker of monu ‘nail’ and the PL marker of L’eL’el ‘saddle’ (Table 38). This homonymy of the marker -a (and some others in Khvarshi) can be explained as the result of the loss of a previous stage where this marker had a unified function. The formal naturalness of this second interpretation of paradigms as in Table 35 is complemented by the functional sense that it makes. Such a system is based on the

Table 35. Partial noun paradigms from Rutul, Khinalug, and Khvarshi

Rutul Khinalug Khvarshi

NOM . SG

ERG . SG

NOM . PL

ERG . PL

t’1la k’1l ezol

t’1la-je-r k’1l-am-i ezol-a-s**

t’1la-j-ma-r k’1l-am-z1r ezol-a-ba

t’1a-j-ma-ši* k1l-am-z1r-i ezol-a-za-s**

‘dog’ ‘hand’ ‘eye’

Notes: * Ergative in the plural has no overt expression. ** Genitive (instead of ergative) used for exemplification.

✐

✐ ✐

✐

✐ 65

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #89

✐

✐

Nominal inflection galore

65

Table 36. Partial paradigms of homo ‘man’, iter ‘road’, aetas ‘age’, nubes ‘cloud’ in Latin

‘man’ ‘road’ ‘age’ ‘cloud’

NOM . SG

GEN . SG

NOM . PL

GEN . PL

homo iter aeta:-s nu:be-s

homin-is itiner-is aeta:t-is nu:b-is

homin-e:s itiner-a aeta:t-e:s nu:b-e:s

homin-um itiner-um aeta:t-um nu:b-ium

Table 37. Partial paradigms of só:ma ‘body’, lampás ‘lamp’, órni:s ‘bird’ in Greek

‘body’ ‘lamp’ ‘bird’

NOM . SG

GEN . SG

NOM . PL

GEN . PL

só:ma lampá-s órni:-s

só:mat-os lampád-os órnith-a

só:mat-a lampád-es órnith-es

so:mát-o:n lampád-o:n orníth-o:n

Table 38. Partial noun paradigms in Khvarshi

‘nail’ ‘saddle’

NOM . SG

GEN . SG

NOM . PL

GEN . PL

monu L’eL’el

mon-a-s L’eL’el-o-s

monu-bo L’eL’el-a-ba

monu-za-s L’eL’el-a-za-s

opposition between the uninflected form of a noun (in the present-day systems this is the nominative singular) and the inflected forms. It is motivated at a stage where there was no systematic distinction between the singular and plural as separate terms of a single category of number. Arguably, such genetically diverse groups as Andic (Andi, Tindi), Tsezic (Khvarshi, Bezhta, Hunzib), Lezgic (Rutul), as well as the “isolates” Lak, Khinalug, and Udi reflect such a stage, to a greater or lesser extent. The pattern for k’1l ‘hand’ in Khinalug can accordingly be represented as in Schema 5, distinguishing unmarked and marked root. It is similar to that in Schema 2 (with the exception of the unmarked root), and is regular for Khinalug in general. It is easy to see that ezol ‘eye’ in Khvarshi and t’1la ‘dog’ in Rutul (Table 35) are inflected similarly to Schemas 3 and 1, respectively (with the exception of the unmarked root and the absence of a separate OBL . SG), augmented by a separate marker of NOM . PL in the case of ‘dog’. It is only Schema 4 that has no corresponding variant with unmarked/marked root. However, there are also examples which are only superficially identical to those described so far. Thus, in Tindi (illustrated in Table 39) it is possible to separate out an exponent of a marked root (-a) in accordance with the above interpretation, but it is also possible to choose cˇ ’ink’◦ a as the primary root, assuming that the nominative

✐

✐ ✐

✐

✐ 66

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #90

✐

✐

66

Aleksandr E. Kibrik

Table 39. Partial paradigm of cˇ ’ink’u ‘elbow’ in Tindi

‘elbow’

NOM . SG

=

NOM . SG

ERG . SG

NOM . PL

ERG . PL

cˇ ’ink’u

cˇ ’ink’◦ -a-l-di

cˇ ’ink’◦ -a-bi

cˇ ’ink’◦ -a-b-a-ji

unmarked ROOT

→

marked ROOT

oblique cases SG

→

PL

→

NOM . PL

oblique cases PL

Schema 5. Stem formation

singular is derived from the root by the narrowing and labialization of the final vowel without any addition of morphemic material. Possibly, two interpretations do not mutually exclude each other in such instances, but are available simultaneously. 3.2.6. The modules of stem formation The full inventory of patterns of stem formation in Daghestanian languages is large, and perhaps not all patterns have been discovered yet.9 What should have become apparent by now, however, is that the diversity of patterns largely results from the different ways of in which the same few elementary modules can be combined. Summarizing the previous sections, these modules of stem formation can be distinguished: STEM - FORMATION STRATEGIES :

(a) (b) (c)

use or non-use of marked vs. unmarked (primary) root; one-stem or two-stem declension (direct vs. oblique) in singular and/or plural; use or non-use of plain oblique stem (unspecified for number); PLURAL EXPRESSION STRATEGIES : (d) use or non-use of unified separate plural marker; NOMINATIVE FORMATION STRATEGIES : (e) NOM . SG coincides with or is derived from the (unmarked) root (usually by phonological rules); NOM . PL is derived from the (marked) root/plural stem or coincides with the (f) (marked) root/plural stem; OBLIQUE DERIVATION STRATEGIES : (g) OBL . PL, if any, is derived from the (marked) root or plural stem or plainoblique stem; (h) OBL . SG, if any, is derived from the (marked) root or plain-oblique stem.

✐

✐ ✐

✐

✐ 67

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #91

✐

✐

Nominal inflection galore Pattern A NOM . SG = (unmarked ROOT → marked) oblique cases SG

ROOT → PL = NOM . PL ↓ ↓ (OBL . PL ) → oblique (OBL . SG ) cases PL

Pattern B NOM . SG = (unmarked ROOT → marked) ROOT oblique

Pattern C NOM . SG

(OBL . SG )

=

oblique ← ( OBL . SG ) ← OBL → cases cases

→ OBL . PL

→

ROOT ↓

67

OBL . PL

NOM . PL

→ oblique cases PL

NOM . PL

→ oblique

Schema 6. Stem formation summarized

The alternatives with different modules are sometimes in complementary distribution: e.g., one-stem declension in singular and plural in module (b) excludes a positive choice for module (c) or modules (g)–(h); the derivation of NOM . PL directly from the root in module (f) is incompatible with the one-stem formation of the plural in module (b). Sometimes, one choice presupposes another: e.g., any choice in (g)–(h) presupposes the two-stem inflection choice in (b). Still, the majority of alternatives of different modules are freely combinable, hence the great number of attested patterns in Daghestanian languages. The restrictions on the mutual compatibility of the module alternatives (but only the dominant choices of nominative formation strategies are taken into account) make it possible to reduce the variety of schemas to three generalized patterns: A, B, and C, as in Schema 6 (with optional steps in parentheses). The main differences among them reduce to three different strategies of packaging and ordering number and case. In pattern A the order of categories (most explicit in the plural) is this: ROOT–NUMBER (PL)–CASE . Pattern C goes for the opposite order where the unspecified separate oblique marker (which opposes direct case to the oblique cases) precedes the (usually amalgamated) number markers: ROOT–CASE (OBL)– NUMBER . Pattern B differs from A and C in that it combines number and case, with no need to order these categories relative to each other: ROOT–NUMBER.CASE .

✐

✐ ✐

✐

✐ 68

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #92

✐

✐

68

Aleksandr E. Kibrik

3.3. Declensions Having surveyed the individual patterns of forming inflectional stems, we shall now consider Daghestanian noun inflection from the viewpoint of its paradigmatic uniformity or diversity. 3.3.1. Co-existing stem formation patterns Individual languages tend to utilize more than one pattern of stem formation. Only Lezgian and the Dargic languages (in our sample Chirag, Megeb, and Itsari) limit themselves to a single pattern, namely that of Schema 1. Some instances of diversity can be explained as the natural historical development from one type to another. For example, in the Avaric languages two patterns of stem formation co-exist, illustrated from the Chadakolob dialect of Antsukh in Table 40: that of Schema 1, with a separate OBL . SG exponent (-i), and a modification of it skipping this derivational step. The latter type, used by about 12 % of nouns (those with a vowel-final stem), is the result of a drift away from the basic type 1, helped along by phonological rules affecting OBL . SG markers most of which are vocalic (-i, -o, -a, -e). By contrast, Akhvakh has a clear preference for type 2 (80 % of nouns), with types 1 and 3 (omitting the OBL . PL derivational step) as marginal alternatives (Table 41). They are all variations of pattern A; diachronically, type 2 appears to be an intermediate stage of a shift from type 1 to type 3 which coincides with the one-stem pattern dominant in Khinalug (Schema 2). Matters are more complicated when languages simultaneously use the mutually exclusive patterns A and C (Rutul, Lezgic) or A, B, and C (Lak, isolate). Patterns A

Table 40. Partial noun paradigms in Antsukh (Chadakolob dialect)

Type 1 Type 2

NOM . SG

GEN . SG

NOM . PL

GEN . PL

mat šuL’u

mat-i-l šuL’u-l

mat-a-l šuL’-ja-l

mat-a-ž-ul šuL’-ja-ž-ul

’nail’ ’squirrel’

Table 41. Partial noun paradigms in Akhvakh

Type 1: Type 2: Type 3:

NOM . SG

ERG . SG

NOM . PL

ERG . PL

k’eha mina leRi

k’eh-o-de mina-de leRi-de

k’eh-i mina-di leR-a

k’eh-i-le-de mina-di-le-de leR-a-de

‘eye’ ‘head’ ‘thigh’

✐

✐ ✐

✐

✐ 69

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #93

✐

✐

Nominal inflection galore

69

and C are hard to reduce to one source, and their co-existence in one and the same language requires a different historical explanation (see below, Section 5.1.2). 3.3.2. Diversity of stem formatives Daghestanian nominal inflection is notable not only for the diversity of patterns of stem formation, but also that of stem formatives themselves – in striking contrast to the uniformity of case and localization markers.10 Diachronically, this would seem to point to a previous inflectional system of a different kind. All Daghestanian languages that build paradigms on an OBL . SG have several formatives for this category (disregarding allomorphy). Thus, Archi uses -li, -mu, -mi, -á, -é, -í, -¯té, -¯taj, -iri, -ró, -rá, and -maj for OBL . SG (Table 42). Lak even has over 30 such formatives. There is a comparable richness in plural formatives. Archi, for example, has -¯tu, -mul, -óm, -ór, -ót, -ló, -til, -q’ul, -rul, -šul, -bur, and some others (see Table 39). Lak has around 50 plural markers.11 Plural marking is uniform only in Tabasaran (-ar) and Agul (Richa dialect: -(j)ar). OBL . PL formation is less diverse; thus, in Khvarshi there is only the single formative -za (Table 35). It is important to note that OBL . PL markers are cumulative and are never identical with OBL . SG markers. This way of forming an oblique stem in the plural looks redundant, because plural is usually expressed independently by a separate PL marker (see discussion below). Co-existing OBL . SG and PL markers can be distinguished as productive or unproductive and as regular or more or less irregular (Bybee 1985a: 115, 132; Mel’ˇcuk 1993: 317, 263). Productive exponents cover the the majority of nouns or a significant number of them, and most importantly, they can be used with newly derived or borrowed nouns. Unproductive exponents can only be added to a closed list of nouns. Regular exponents are added to stems conforming to rules of gram-

Table 42.

OBL . SG

‘house’ ‘husband’ ‘knife’ ‘sun’ ‘horse’ ‘beard’ ‘moon’

and PL formatives in Archi

NOM . SG

DAT. SG

noL’ bošór k’os barq noIš moc’ór bac

nóL’-li-s bošór-mu-s k’a¯s-á-s berq-é-s niIš-í-s mež-dé-s boc-ró-s

‘dog’ ‘ball’ ‘name’ ‘cup’ ‘tooth’ ‘mouth’ ‘man’

NOM . SG

NOM . PL

baHrí dump c’or gel sot s¯ob adám

baHrí-¯tu dúmp-mul c’or-óm gél-um so¯t-ór s¯ ob-ót adám-til

✐

✐ ✐

✐

✐ 70

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #94

✐

✐

70

Aleksandr E. Kibrik

mar. The choice of stem formatives is not usually rule-governed, and their regularity varies from (Daghestanian) language to language. There are regular OBL . SG markers (Bezhta -li), regular PL markers (Antsukh -as/-aì), both (Archi -li OBL . SG, -¯tu/-mul PL), or neither (Andi). Regularity can be a matter of different kinds of rules. First, a highly productive marker can become dominant and is then chosen by a default rule (Wurzel 1990: 207; Zwicky 1985), adding this marker unless an item is lexically specified otherwise. This is the case with OBL . SG -li in Archi, covering about 80 % of all nouns, and with OBL . SG -li in Bezhta, covering some 40 % of nouns (in competition with three irregular yet productive strategies: formatives -a and -i, covering about 20 % of nouns each, and a one-stem pattern, covering 10 %). In Antsukh (Chadakolob) 70 % of nouns use the marker -a in the plural and so do 60 % in Andalal (Sogratl). 80 % of Akhvakh nouns are inflected in the singular according to the one-stem pattern, and the remaining 20 % according to a two-stem pattern (using ten irregular markers of OBL . SG). Second, rules may be subject to phonotactic restrictions. Thus, the plural markers -¯tu/-mul in Archi and -jar/-ar in Agul (Richa dialect) are in complementary distribution between vowel-final and consonant-final stems. (In Agul there are no other PL markers; in Archi there are several irregular ones, requiring lexical specification.) Third, there may be morphological conditions on rules. Thus, the attributive declension is usually regular in Godoberi, with all nominals derived from attributes using specific OBL . SG and PL markers. Fourth, rules for stem formatives can be semantic. Thus, in Antsukh (Chadakolob dialect) all male human nouns use -as for OBL . SG and female human nouns use -aì. In Rutul there are two regular ways of plural formation: animate nouns usually use -ma for PL (plus the marker -r in the nominative or -i in the oblique cases), while inanimate nouns add NOM . PL -b1r to the primary stem and (in the oblique cases) OBL . PL -m1 to the oblique stem (Table 43). Fifth, there may be implicational rules referring to stem formatives added to the primary stem.12 They are common in OBL . PL stem formation, where the choice of a marker is determined by the PL or NOM . PL marker. For example, in Bezhta the plural can be formed with a PL marker (with zero for NOM . PL) or independently in the nominative plural and other oblique cases, corresponding to patterns A and B, respectively (Table 44): if nominative plural is expressed by -bo or -rol, then OBL . PL is -la or -lara, respectively; if nominative plural is zero, then OBL . PL is -ra. 3.3.3. Paradigmatic versatility Not only are declensional patterns diverse, nouns are also mobile between them. While in most Daghestanian languages a limited number of nouns is inflectionally versatile, some (especially Avaric) permit such variation quite regularly. Thus, nouns

✐

✐ ✐

✐

✐ 71

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #95

✐

✐

Nominal inflection galore

71

Table 43. Partial noun paradigms in Rutul NOM . SG

NOM . PL

DAT. PL

‘cook’

dadal

‘nose’

xyexy

dadal-ma-r ROOT-PL - NOM . PL xyexy-b1r ROOT-NOM . PL

dadal-ma-ši-s ROOT-PL - OBL . PL - DAT xyexy-i-m1-s ROOT-OBL-OBL . PL - DAT

Table 44. Partial noun paradigms in Bezhta NOM . SG

NOM . PL

DAT. PL

‘kid’

t’öq’ä

‘face’

mot’o

‘net’

q’i

t’öq’-bö ROOT-NOM . PL mot’o-rol ROOT-NOM . PL q’i-ja ROOT-PL ( NOM )

t’öq’-lä-s ROOT-OBL . PL - DAT mot’-lara-s ROOT-OBL . PL - DAT q’i-ja-ra-s ROOT-PL - OBL . PL - DAT

Table 45. Declensional variation of nouns in Andalal (Sogratl dialect) NOM . SG

SUPER - ESS

‘flower’

t’eh

t’éh-da t’oh-ó-da

‘grass’

Xer

‘saddle’

L’ilí

‘axe’

Qiždí

CONT- ESS

ERG

Xir-í-X Xér-du-X L’ilí-ìa L’ol-ó-ìa L’ilí-du-ìa Qiždí-ìa Qožd-ó-ìa Qiždí-du-ìa

can have alternative case forms in Andalal (Table 45). Different forms differ not only with respect to stem formatives (e.g., zero, -o, or -du with ‘saddle’ and ‘axe’), but also in accent alternations (stem or suffix accent) and attendant segmental changes of stems (t’eh → t’oh, L’ili → L’ol, Xir → Xer, etc.). A more complicated situation arises when a noun has different (and/or complementarily distributed) sets of secondary stem markers in the context of different cases, as is illustrated by one noun in Andalal in Table 46. The nominative plu-

✐

✐ ✐

✐

✐ 72

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #96

✐

✐

72

Aleksandr E. Kibrik

Table 46. Partial paradigm of roso ‘village’ in Andalal (Sogratl dialect) NOM . SG NOM . PL GEN . SG SUPER - LAT CONT- LAT IN - LAT III

roso ros-u-bi || ros-a-bi || ros-i-bi ros-da-l ros-d-e ros-du-X-e || ros-u-X-e ros-no-b-e

ral has three variants, with -u, -a, or -i for PL. In the singular there are different OBL . SG markers (-da, -du, -no, -u) in different cases, some in free variation (-du/-u in cont-lative) and others in complementary distribution (e.g., -da and -no in genitive and in-lative). In the super-lative the markers -da/-du are neutralized before the lative case marker -e.

4. How nominal and how inflectional is nominal inflection? Nominal inflection in Daghestanian languages, as surveyed in the preceding sections, gives the impression of not really being one well-integrated homogeneous system. It is especially spatial forms and the genitive case that raise such doubts.

4.1. Similarities of nouns to other word classes First of all, the way nouns inflect for spatial cases and localization is not unique to this word class. Other word classes show systematic similarities.

4.1.1. The inflection of postpositions The Daghestanian languages usually have a class of spatial postpositions which share the category of spatial case with nouns. Comparing, for example, spatial postpositions in Archi (Table 47) to the spatial subparadigm of nouns (Table 7), they are seen to be almost identical. The postpositional roots correspond to the nominal category of localization, and the inflectional suffixes are the same as the nominal markers of spatial cases: -k LATIVE, -š ELATIVE, -ši ALLATIVE, -Xut TRANSLATIVE. The only difference is that postpositions do not have the terminative form.

✐

✐ ✐

✐

✐ 73

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #97

✐

✐

Nominal inflection galore

73

4.1.2. Spatial adverbs Postpositions, in turn, have a lot in common with spatial adverbs, both in their roots and in inflection. Compare Tables 47 and 48 for Archi. It is evident that adverbs and postpositions have the same origin and inflectional meanings, but different paradigm gaps for cognate words and different syntactic environments (which is an argument for cognate adverbs and postpositions being different words).

4.1.3. Spatial converbs Deverbal spatial converbs are widespread among Daghestanian languages, and they also use the same set of directional oppositions and corresponding markers. For Archi, compare the spatial converb ‘where (he) worked’ (Table 49), formed from the verb wirX◦ ni ‘work (AOR )’. The derivational suffix (orientation marker) is -ma, the essive marker is zero, the lative suffix is -k, the elative is -š, etc.

Table 47. Inflection of spatial postpositions in Archi

‘between’ ‘on’ ‘near’ ‘under’ ‘behind’ ‘in front of’

ESSIVE

LATIVE

ELATIVE

ALLATIVE

TRANSLATIVE

qI’on jat – – Xir –

qI’ana-k ja-¯tik ì¯◦ a-k L’ara-k Xara-k hara-k

qI’ana-š – ì¯◦ a-š L’ara-š Xara-š hara-š

qI’ana-ši – ì¯◦ a-ši – – –

qI’ana-Xut ja¯ti-Xut ì¯◦ a-Xut L’ara-Xut Xara-Xut hara-Xut

Table 48. Inflection of spatial adverbs in Archi

‘above’ ‘together’ ‘beneath’ ‘behind’ ‘in front of’

ESSIVE

LATIVE

ELATIVE

ALLATIVE

TRANSLATIVE

jat ì¯◦ a – Xir –

– ì¯◦ a-k L’ara-k Xara-k hara-k

ja¯ti-š ì¯◦ a-š L’ara-š Xara-š hara-š

ja¯ti-ši ì¯◦ a-ši L’ara-ši Xara-ši hara-ši

ja¯ti-Xut ì¯◦ a-Xut L’ara-Xut Xara-Xut hara-Xut

Table 49. Partial paradigm of the spatial converb wirX◦ ni ‘work’ in Archi ESSIVE

LATIVE

ELATIVE

wirX◦ ni-ma

wirX◦ ni-ma-k

wirX◦ ni-ma-š

✐

✐ ✐

✐

✐ 74

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #98

✐

✐

74

Aleksandr E. Kibrik

4.1.4. Temporal converbs Along with regular spatial converbs, a converb is attested in Godoberi which is derived from the infinitive by attaching the nominal APUD localization marker -Xa (historically originating from the meaning POST ‘behind’). This marker is used adnominally and adverbally in (24) and (25), respectively: (24) a.

b.

(25) a.

b.

ce=w adami w=aPa. iLi-Xa we.OBL . SG - APUD one=MALE man.NOM MALE=come.AOR ‘One man came to us.’ uns-e:-Xa kalXoz-u-Li. den w=uk’a I.NOM MALE=be.AOR bull-OBL . PL-APUD kolkhoz-OBL . SG-GEN ‘I was tending the kolkhoz bulls.’ b=uL-i-Xa jašu-di heìe k’eda saQati today two hour.NOM NEUT=finish-INF-APUD daughter-ERG hanq’u harˇc’-al-iLi-bu-da. house.NOM clean-CAUS - FUT- PART- COP ‘Today the daughter will start cleaning the house in two hours.’ b=al-i-Xa, quˇca di-ˇc’u maQana den b=eRu I NEUT=much NEUT=read-INF-APUD book.NOM I-CONT more b=iˇc-anta-da. N=understand-PRS-COP ‘The more I read the book, the better I understand it.’ (lit. ’Along with/behind my much reading, I understand the book more.’)

4.1.5. Possessive locatives In addition to the standard set of word classes sharing the category of spatial case with nouns, some languages idiosyncratically extend the applicability of spatial case markers. An example is provided by Archi possessive locatives (mentioned above, Section 2.4.2), which are derived from nominal genitives by means of the orientation marker -ma and case markers. Thus, from the noun dos ‘friend’ the genitive dos-li-n can be derived, which serves as a base for possessive locatives inflected for spatial cases as set out in Table 50.

Table 50. Inflection of possessive locative of dos ‘friend’ in Archi ESSIVE

LATIVE

ELATIVE

dos-li-m-ma ‘at . . .

dos-li-m-ma-k ‘to . . .

dos-li-m-ma-š ‘from the friend, the friend’s place’

✐

✐ ✐

✐

✐ 75

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #99

✐

✐

Nominal inflection galore

75

4.1.6. Cases or postpositions? Historically, spatial nominal forms originate from postpositions. This is synchronically evident, for example, when such localization markers as -L’/-¯t SUB / SUPER in Archi are compared with the postpositions L’ara- ‘under’ and jat ‘on’. It is therefore often difficult, in Daghestanian no less than elsewhere,13 to distinguish nominal case forms from postpositional phrases. In Plank (1991b) a long list of criteria was suggested for distinguishing cases from postpositions, of which these three are of primary importance for present purposes: (i)

Is the relational marker (RM) prosodically bound to elements (words or stems) of particular (word or stem) classes, especially nouns, or is it free to associate with elements of different classes?

(ii)

Is the phrase carrying RM an NP or a syntactically somewhat different type of phrase (in particular, an adpositional phrase)?

(iii)

Is it impossible or possible for the RM to co-occur with another RM of the same formal kind in a simple NP or AP?

These criteria favour an adpositional interpretation of directional markers in Daghestanian: they can associate with elements of different classes; they are prototypically used in syntactic positions equivalent to adverbial phrases; and sporadically (cf. the Archi possessive locatives) they co-occur with the genitive case marker. Overall, directional markers come out as occupying an intermediate position on the gradient scale from prototypical case to prototypical postposition.

4.2. Inflection or derivation? It does not go without saying that Daghestanian nominal morphology is inflectional in its entirety, rather than derivational. The feasibility of a distinction between inflection and derivation as such is a complex issue; but on the strength of some of the criteria traditionally used to draw it (as collected in Plank 1991a), the resulting picture is not a homogeneous one. 4.2.1. Spatial forms If morphological categories are considered inflectional or derivational depending on whether or not they are limited to bases of a single word class, then localization is inflectional and direction (spatial cases) derivational, provided they are both accepted as forming part of the nominal paradigm. However, if localization is considered derivational on the grounds of its changing word class, turning nouns into adverbs,

✐

✐ ✐

✐

✐ 76

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #100

✐

✐

76

Aleksandr E. Kibrik

then direction in turn has to be considered inflectional, with only adverbs having this category. Localization allows an inflectional interpretation insofar as the word class of words localized is uniquely determined by this category, while spatial case can be considered derivational on this criterion (contradicting the principle “inflection follows derivation” (Anderson 1988: 41)). On the criterion that only derivational categories can be assigned more than once to the same base, possessive locatives in Archi are derivational. On the criterion that only inflectional categories can be assigned by syntactic government, localization and/or direction qualify as inflectional, since they are assigned to nouns by some verbs. If the existence of synonymous expressions, especially ones with the same base, is characteristic of derivation, then examples like (26), with a spatial nominal form replaced by an equivalent spatial postpositional phrase, suggest that spatial case (in Archi) is derivational. (26)

jat / mulli-t boXI obsdili i. mulli-s mountain-DAT on / mountain-SUPER - ESS goat.NOM stand.CONV be ‘The/a Caucasian goat is standing on the mountain.’

Derivational categories are typically limited to subsets of relevant words rather than being fully productive. There are such semantic limitations, for example, of Archi localizations IN (‘within an empty or closed space’) and INTER (‘in a filled, dense space’), with the choice between them determined by the meaning of nouns: cˇ ejdan ‘kettle’, cˇ itir ‘tent’, buXar ‘fireplace’, lagi ‘stomach’ are typical IN nouns, and ¯ìan ‘water’, bi ‘blood’, huq’ ‘smoke’, goIroIrˇci ‘hot cereal’ typical INTER nouns. Localization exponents are morphophonemically different from the regular inflectional exponents: they inherit the shape of the postpositional stems they are derived from. Genuine inflectional markers usually use vowels, glides, liquids, and sibilants, while roots prefer stops and affricates. Thus, the sound shapes of forms suggest that they are different in kind. Inflectional categories tend to recur cross-linguistically, while derivational categories may be more idiosyncratic. Being peculiar to North Caucasian (see Kilby 1981), localization shows another derivational feature. Overall, then, there is some evidence for a derivational interpretation of localization, with spatial form as denominal adverbs. However, there is also much to be said for its inflectional status. In particular, taking it for derivational creates descriptive problems. First, most Daghestanian languages express some non-spatial meanings by means of spatial forms. Thus, in Archi, the IN - ESSIVE has the meaning ‘external cause’ in (27), the CONT- LATIVE and CONT- ELATIVE denote a person who ‘becomes/ceases

✐

✐ ✐

✐

✐ 77

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #101

✐

✐

Nominal inflection galore

77

to be a temporary possessor’ in (28), the SUPER - ESSIVE and SUPER - ELATIVE have the temporal meanings ‘when’ and ‘after’ in (29)–(30): (27)

baIraj+owXu. jasqi zon Lerkul-l-a today I.NOM heat-OBL-INESS bathe.AOR ‘Today I took a bath because of the heat.’

(28) a.

¯ ◦ alli za-ra-k boq’i X I-CONT-LAT bread.NOM give.IMP ‘Give me some bread (perhaps for somebody else).’ ¯ ¯ ◦ al-li-n me¯c’e obXa. X towmu za-ra-š he-ERG I-CONT-ELAT bread-OBL-GEN piece.NOM take-AOR ‘He took a piece of bread from me.’

b.

(29) a.

b.

(30)

tot s¯aQal-li-t that time-OBL-SUPER . ESS ‘that time’ kino abu-qi. sreda-li-t Wednesday-OBL-SUPER . ESS movie.NOM make-FUT ‘On Wednesday there will be a movie.’ ¯ bocr-o-¯ti-š dakra-s ìonnol month-OBL-SUPER-ELAT wife.NOM take-INF ‘to marry a month later’

Second, some verbs govern particular spatial forms, in the manner that case markers are syntactically governed. Table 51 lists some spatial forms and their governing verbs in Archi. Thus, there is conflicting evidence in the same language with respect to the inflectional or derivational interpretation of spatial forms, jeopardizing any neat compartmentalization. More adequate is the assumption of a continuum between derivation and inflection, along which spatial forms are drifting in time, originating as syntactic constituents (viz., adverbial postpositional phrases) and developing into derivational forms (denominal adverbs) and eventually coming close to inflectional forms (like nouns and case markers are). Different Daghestanian languages have reached different stages of this development. Comparing Khinalug and Godoberi to Archi shows well how far apart they can be. The case system of Khinalug, shown in Table 52, is much more homogeneous than that of Archi (Kibrik et al. 1972: 50). The characteristically Daghestanian pattern of spatial forms has been lost. First, there are no different localization meanings and no opposition of essive and lative; only the elative in opposition to the locative is preserved. As illustrated in (31) and (32), locative means ‘to be found at, or moving towards, a position relative to X’ and elative ‘to move from a typical position relative to X’.

✐

✐ ✐

✐

✐ 78

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #102

✐

✐

78

Aleksandr E. Kibrik

Table 51. Governed spatial forms in Archi

Governed spatial form

SUPER - ESSIVE

Governing verbs On subject

On direct or oblique object

hejb i ‘be guilty’

¯ L’ili eìas ˇ¯sak qIes wiX maIrš ¯ sakas oj-aˇcas Linˇc’ar Xustar-as barka-bos k◦ aˇc’et-bos

SUPER - LATIVE SUB - ELATIVE CONT- ALLATIVE

‘saddle’ ‘doubt’ ‘ride’ ‘believe’ ‘envy’ ‘look at’ ‘hear’ ‘be afraid of’ ‘save from’ ‘congratulate’ ‘tell lies to’

Table 52. Khinalug cases NOMINATIVE ERGATIVE

1 2

GENITIVE GENITIVE DATIVE

COMITATIVE LOCATIVE ORIENTATIONAL LOCATIVE POSSESSIVE LOCATIVE COMPARATIVE

(31) a.

b.

c.

1

-Ø -i -i -e -u -škili -r -X -š -q’

ELATIVE ORIENTATIONAL ELATIVE POSSESSIVE ELATIVE COMPARATIVE

2

-illi -X-illi -š-illi -q’-illi

bädrä-r xu qomä. bucket-LOC water.NOM be ‘There is water in the bucket.’ xu loži. bädrä-r bucket-LOC water.NOM pour.IMP ‘Pour water into the bucket.’ ¯ m1kar qomä. an-¯t1r ground-LOC stick.NOM be ‘There is a stick on the ground.’

✐

✐ ✐

✐

✐ 79

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #103

✐

✐

Nominal inflection galore

d.

(32) a.

b.

79

¯ Xä¯ˇculkwit omä. d1var-1r Häjnä be wall-LOC mirror.NOM hang ‘A mirror is hanging on the wall.’ gis-illi k’a¯c1n alkšämä. roof-ELAT stone.NOM fall ‘A stone fell off the roof.’ q’urq’or qal¯t1rkušämä. x1n-1lli water-ELAT frog.NOM jump.out ‘A frog jumped out of the water.’

Second, orientational locative/elative forms combine spatial (approximate location near X) with grammatical (goal, cause) meanings: (33) a.

b.

c.

tapä-X k1za omä. m1d-e mountain-GEN 2 top-ORLOC snow.NOM be ‘Snow is lying on/near the top of the mountain.’ azadl1-X daQvakwižmä. hini He.ERG freedom-ORLOC fight ‘He fought for freedom.’ pXra-X läk’širetmä. z1 I.NOM dog-ORLOC look.at ‘I am looking at the dog.’

Third, the possessive and comparative locative/elative markers have predominantly grammatical meanings. Comparative 1 and 2 are usually interchangeable. Thus, the uniform -Ø/-illi distinction is clearly present only in the context of the locative/elative meaning and partially in those of the orientational meanings. In combination with possessiveness this distinction is metaphorically interpreted as the difference between possession and loss of possession, while in the context of comparatives there is no semantic difference between these markers at all. This suggests that no category of localization should be postulated in Khinalug and that all nominal forms should be seen as separate cases in spite of their partial formal identities. Turning to postpositions, Khinalug has innovated a system of spatial postpositions, mostly borrowed from Azerbaijani, which distinguish between locative (-r) and elative (-illi) forms (Table 53). Etymologically these postpositions are derived from nominal spatial forms and, as former heads of genitive constructions, they govern the genitive (1) of the dependent noun: (34)

noXut’-urdur šiqä. lik’ebir-i iˇci-lli grain-GEN inside-ELAT pea-NOM . PL pick.out.IMP ‘Pick out the peas from the grain.’

✐

✐ ✐

✐

✐ 80

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #104

✐

✐

80

Aleksandr E. Kibrik

Table 53. Khinalug postpositions

Meaning

Postposition (LOC / ELAT)

Azerbaijani source

‘on’ ‘in, inside’ ‘under, near’ ‘behind’ ‘in front of’

üstü-r/-lli iˇci-r/-lli dibi-r/-lli dal1-r/-lli qabaR1-r/-lli

ust iˇc dib dal/dal1 qabaq

‘top’ ‘interior’ ‘bottom’ ‘back side’ ‘fore-part’

While Khinalug thus demonstrates a well-integrated system of nominal morphology, Godoberi (Table 54) could seem reminiscent of Archi (Table 7 above) in the way it sets apart and elaborates spatial forms. However, the formal regularity of spatial forms in Godoberi is at odds with their usage. Most of the localizations do not have independent spatial meanings and are lexically or syntactically conditioned. Thus, SUPER is used in the sense of LOC (‘in a typical position relative to OP’) with a restricted number of nouns for various kinds of flat, limited surfaces (e.g., miq’i ‘road’, bazar ‘marketplace’, Qagar ‘town’, bajdan ‘square’). Similarly, IN localization is used exclusively in a LOC function and only with a few nouns (e.g., sanuqi ‘trunk’, ami ‘big basin’, tarsi ‘basin’, cˇ aqi ‘pot’, zanXi ‘jug’, qati ‘handful’, lamur ‘roof’, bešen ‘bed’). The localization PLACE (marker -alda borrowed from Avar) is used only in LOC function and only with nouns denoting towns, villages etc. (e.g., Moskwa ‘Moscow’, gorod/ šahru ‘town’), as in (35).

Table 54. Spatial forms in Godoberi

Localization

Direction ESSIVE / ELATIVE ALLATIVE TRANSLATIVE LATIVE

SUPER CONT IN INTER APUD AD PLACE

(-Ø) -la -ˇc’u -o -Li -Xa -qi -alda

(-ru) -laru -ˇc’uru -oru -Liru -Xaru -qiru -aldaru

(-Xu) -laXu – -oXu -LiXu -XaXu -qiXu -aldaXu

(-ruXu/-ridi/-rudi/-riL’i) -laruXu/-laridi/-larudi/-lariL’i – -oruXu/-oridi/-orudi/-oriL’i -LiruXu/-Liridi/-Lirudi/-LiriL’i -XaruXu/-Xaridi/-Xarudi/-XariL’i -qiruXu/-qiridi/-qirudi/-qiriL’i -aldaruXu/-aldaridi/-aldarudi/-aldariL’i

✐

✐ ✐

✐

✐ 81

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #105

✐

✐

Nominal inflection galore

81

den gorod-alda/šahar-alda w=aPa. MALE=come. AOR I.NOM town-PLACE.ESS ‘I came to the town.’

(35)

The localization INTER in its primary semantic function obligatorily requires a semantically redundant postposition hinu ‘in’. It is used, as in (36), with nouns of appropriate meaning, such as those denoting unbounded substances (sira ‘sand’, leni ‘water’, anzi ‘snow’, raL’i ‘ground’) and wholes (q’oq’a ‘group’, rexedi ‘herd’, ludi ‘firewood’, hani ‘village’, uškur ‘school’). hinu sira ida. raL’i-Li ground-INTER in sand.NOM COP ‘The sand is in the ground.’

(36)

The localization AD is used semantically, viz., for location ‘near X’ (37a), and syntactically, substituting for all localizations in the plural (37b); and it can be governed by verbs (such as gulati ‘talk with’, hališi ‘look at’, cˇ ’ani ‘tie to’). (37) a.

b.

den ihur bahi-qi w=aPa. bel-qi mountain-AD I lake.NOM border-AD MALE=come.AOR ‘In the mountains I came near the border of the lake.’ den bajda-me-qi j=aPa. I square-PL-AD FEM=come.AOR ‘I came to the squares.’

Although the localization CONT has an orientational meaning ‘to be in contact with X’ (38a), such semantically motivated uses are rare; it occurs frequently, however, since CONT is governed by numerous verbs (q’ami ‘bite’, ubadi ‘kiss’, buzi ‘believe’, wuva ‘understand’, etc.), as illustrated in (38b), and is often used with postpositions of spatial meaning (Table 55). (38) a.

b.

ìib-u šišuXa-riL’i ho-rd-i-ˇc’u-ru he-OBL . PL-PL-CONT-ELAT fear-CONV.PST Shishuha-through den. w=aPa MALE=go. AOR I ‘I ran [went] from them in fright down through Shishuha (village).’ ubadi. Qali-di jaš-u-ˇc’u Ali-ERG girl-OBL-CONT.ESS kiss.AOR ‘Ali kissed the girl.’

These postpositional constructions, and not synthetic spatial nominal forms, are the main means for expressing spatial relations:

✐

✐ ✐

✐

✐ 82

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #106

✐

✐

82

Aleksandr E. Kibrik

Table 55. Spatial postposition constructions in Godoberi

Meaning

Postposition (ESSIVE)

Example noun-CONT / INTER postposition-CASE

‘on’ ‘over’ ‘under’ ‘near’* ‘near’* ‘among’** ‘in front of’** ‘behind’** ‘in’***

hila hilaridi hiL’i isibala isibaqi w/j/b=aL’u se=w/j/b=a wo/je/bo=Xut’u hinu

ustura-ˇc’u hila ustura-ˇc’u hilaridi ustura-ˇc’u hiL’i ustura-ˇc’u isibala minsa-ˇc’u isibaqi ruši-be-ˇc’u b=aL’u ustura-ˇc’u se=b=a ustura-ˇc’u bo=Xut’u burali-Li hinu

‘on the table’ ‘over the table’ ‘under the table’ ‘near the table’ ‘on the bank of the river’ ‘between the trees’ ‘in front of the table’ ‘behind the table’ ‘in the pitcher’

Notes: * While isibala is neutral, isibaqi preferably takes “long” objects and refers to either their right or left side. ** These postpositions require a class-number marker controlled by the external noun in the nominative. ‘Among’ demands a plural noun or conjoined nouns. *** The meaning ‘in’ (cf. localization IN) is expressed by means of the INTER localization of the corresponding noun (including nouns usually combined with the IN localization in synthetic constructions).

(39)

ho-š-¯ti, hingura-ˇc’u hiL’i-la he-OBL.MALE-ERG window-CONT. ESS under-AND ¯ w=uL-u, hant’uk’ya rik-u-da. MALE=stay-CONV. PST ear listen-PST-COP ‘He stood under the window and listened (to something).’

(40)

Qalik’ya-ˇc’u be=Xut’u ba=L-u-da. gurži-di Georgian-PL Alik’ya-CONT PL=behind PL=go-PST- COP ‘The Georgians went after Alik’ya.’

(41)

haPa jaša waš-u-ˇc’u-la jaš-u-ˇc’u-la di-ra I.OBL - AFF see.AOR girl boy-OBL - CONT- AND girl-OBL - CONT- AND j=aL’u. FEM=between ‘I saw the girl (standing) between the boy and the girl.’

All of this shows that the original system of Godoberi spatial forms is in disarray, in spite of the preservation of morphological markers. In fact, these preserved forms are not irrelevant for the overall morphological technique, insofar as in modern Godoberi there is a tendency to amalgamate postpositional phrases into single synthetic forms resulting in new localization terms (see Table 56).

✐

✐ ✐

✐

✐ 83

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #107

✐

✐

Nominal inflection galore

83

Table 56. Innovative localizations in Godoberi

ESSIVE ELATIVE ALLATIVE

New SUPER

New SUB

-ˇc’u + hila → -ˇc’ila -ˇc’u + hilaru → -ˇc’ilaru -ˇc’u + hilaXu → -ˇc’ilaXu

-ˇc’u + hiL’i → -ˇc’iL’i -ˇc’u + hiL’iru → -ˇc’iL’iru -ˇc’u + hiL’iXu → -ˇc’iL’iXu

inflection (cases)

derivation (adverbs)

Bezhta – Rutul – Archi – Megeb – Avar – Chirag – Kryz – Budukh – Khinalug Tabasaran Godoberi Schema 7. Daghestanian spatial forms on the inflection-derivation continuum

To sum up, depending on the inflectional or derivational inclinations of their spatial forms, Daghestanian languages can be arranged as in Schema 7. Bezhta and Khinalug are diametrically opposed, Tabasaran is between Bezhta and Rutul, and Archi is in the middle. Among other things, Bezhta differs from Archi in its lack of non-spatial meanings of spatial forms and in having a few verbs which govern spatial forms. Standard Godoberi, now close to Khinalug, is set to move left provided it continues to create new localizations from postpositional phrases. 4.2.2. Genitive or attributive? Another case where Daghestanian nominal morphology vacillates between inflection and derivation is in marking attributive constituents. The question here is whether there is a genuine genitive case or a more general derivational marker of attribution. Archi and Rutul will illustrate the difference (see further Kibrik 1995). Archi nouns have a regular genitive case, marked by -n, whose morphological behaviour is similar to that of other cases. Adjectives are a minor word class, comprising only a few primary adjectives (bare roots) plus demonstrative pronouns (listed in Table 57). Attributives, on the other hand, are an open word class. They can be formed by means of the suffix -¯tu from verbs, adverbs, postpositions, and also nouns (Table 58). Attributives derived from nouns add the attributivizer to the genitive case of the noun. The genitive and the attributive are two separate forms. Although traditional descriptions (Ibragimov 1978: 51) include the genitive in the noun paradigm of Rutul, a different interpretation seems more appropriate. All modifiers of nouns have the same marker -d/-d1 which derives attributives from words of various word classes (Table 59). There is no special genitive case marker. The only difference between attributives derived from verbs, adverbs, and postpositions,

✐

✐ ✐

✐

✐ 84

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #108

✐

✐

84

Aleksandr E. Kibrik

Table 57. Primary adjectives in Archi

ja¯tan ‘Avar’ aquš ‘Dargwa’ haman ‘Lak’ gurži ‘Georgian’ marˇci ‘all’ os para ‘some’ na:q’ukan ‘many’

meLle ‘masculine’ ¯ ìenne ‘feminine’ kulu ‘orphan’ cˇ ’ere ‘barren (of a sheep)’ biši ‘someone else’s’ but’u ‘other’

araIˇc’ ‘big, enough’

jo=w/ja=r ‘this (close to speaker)’ jamu/jamu=r ‘this (close to hearer)’ to=w/to=r ‘that’ and others

Table 58. Types of attributives in Archi

Stative verbs:

Dynamic verbs: Adverbs: Postpositions: Nouns:

Primary form

Attributive

hiba mu t’i ¯ ◦ in wirX abc’u hinc jat qI’on kutak san

hiba-¯tu-t mu-¯tu-t t’i-¯tu-t ¯ ◦ in-nu wirX abc’u-¯tu-b hinc-du-t ja¯t-u-t qI’on-nu-t kutak-li-n-nu-t san-ni-n-nu-t

‘be good’ ‘be beautiful’ ‘be small’ ‘work (PRS , I)’ ‘fill (AOR , III)’ ‘now’ ‘above’ ‘among’ ‘force’ ‘year’

‘good (IV)’ ‘beautiful (IV)’ ‘small (IV)’ ‘working (I)’ ‘filled (III)’ ‘present (IV)’ ‘upper (IV)’ ‘middle (IV)’ ‘strong (IV)’ ‘annual (IV)’

on the one hand, and nominal attributives, on the other, is that the latter are derived not from a bare stem (as adverbs and postpositions are) or an aspect stem (as verbs are), but from an oblique stem, which is also used for case inflection. Thus, in (42) this attributivizer is used twice: with the noun bilax ‘spring’ and the stative verb t’am1r ‘be clean’ (analytic present, formed with a participle of the main verb and an auxiliary); in (43) it is attached to both the dynamic verb ‘do’ and the noun ‘son’; and in (44) (see also Table 60), in very un-case-like manner, it is seen to be added to a dative. (42)

xad t’am1r-d1 i. bilax-a-d spring-OBL - ATTR water clean-ATTR AUX.PRS ‘Spring water is clean.’

(43)

duX-ar-d1 d1d dawat waPa-d marry do.PERF - ATTR son-OBL . SG-ATTR father ‘the father of a married son’

✐

✐ ✐

✐

✐ 85

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #109

✐

✐

Nominal inflection galore

85

Table 59. Types of attributives in Rutul

Primary form stative verbs: dynamic verbs: adverbs: postpositions: nouns:

Attributive

h1IX1 diri Rirxe-rPa ma: uliky saPbaj waz

‘be good’ ‘be quick’ ‘hit (PRS )’ ‘again’ ‘before’ ‘under’ ‘moon’

g◦ ašag

‘bag’

h1IX-d1 diri-d Rirxe-d ma:-d1 uliky-d1 saPbaj-d1 waz-1l-d1 (OBL - ATTR) g◦ ašag-1-d (OBL - ATTR)

‘good’ ‘quick’ ‘hitting’ ‘repeated’ ‘former’ ‘lower’ ‘of the moon’ ‘of the/a bag’

Table 60. Dative attributives in Rutul

Root

Dative

Attributive

uXn-a-s dawat-a-s

uXn-a-s-d1 dawat-a-s-d1

uXun dawat

‘dress’ ‘wedding’

(44)

uXn-a-s-d1 parˇcä dress-OBL . SG - DAT- ATTR fabric ‘fabric for the/a dress’

‘intended for the/a dress’ ‘intended for the/a wedding’

4.2.3. A motive for inflectional or derivational leanings The main non-verbal word classes of noun, adverb, and adjective are motivated by their primary syntactic functions: a noun prototypically fills the slot of a verbal argument, an adverb that of a verbal modifier, and an adjective that of a noun modifier.14 Languages with a preference for derivational morphology are keeping these correspondences between syntactic functions and word classes intact: when words which are basically nouns are used in functions other than that of verbal argument, they are derived to form words which are less nouny. In languages where genitives and spatial forms are inflectional, on the other hand, the morphological principle of the unity of word classes is reasserting itself in all syntactic functions of nouns: as argument, inflecting for grammatical cases; as verbal modifier, inflecting for spatial cases (or accompanied by adpositions); and as nominal modifier, inflecting for genitive. In view of such conflicting principles it should not be too surprising to find inflection and derivation as hard to separate as they are in Daghestanian nominal morphology.

✐

✐ ✐

✐

✐ 86

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #110

✐

✐

86

Aleksandr E. Kibrik

5. Using nominal inflection From categories and formal systems of Daghestanian nominal inflection we now turn to questions of the use, or also non-use, of inflections.

5.1. Frequency Without going into details of how frequency bears on inflectional categories and systems,15 it is worth mentioning that there are not only differences between inflectional forms as to their frequency but also between languages as to their frequency distributions for particular words and word forms. Godoberi, for example, is remarkable for its large number of word forms with extremely low frequency. The main reasons for low frequency of inflections are as follows. First, an inflection can be lexically restricted, in particular to lexemes which are themselves infrequent. Thus, the localization BELOW in Godoberi, marked by -L’i, is only applicable to a very restricted set of words including ‘sky’ (rešina-L’i ‘in the sky’, lit. ‘under the sky’) and ‘ceiling’ (muq’uda-L’i ‘on the ceiling’). Second, there may be syntactic restrictions, as with the affective case in Godoberi, which is only governed by three or four verbs (‘see’, ‘hear’, ‘know’, and perhaps ‘forget’). Third, it may be for pragmatic reasons, owing to the marginality of their meanings, that some inflections are less in demand than others (like the translative vis-à-vis the other directional cases of lative and elative). Fourth, a category may be about to be lost, surviving only in the speech of the older generation.

5.2. Variation As mentioned previously (Section 3.3.3), there is considerable variability of inflectional systems and exponents in Avaric, Lezgic, and other Daghestanian groups, especially with respect to stem formation. This variability may take two forms: there may not be a unique norm for a speech community, and there may even be variability for one and the same speaker. The latter is particularly widespread in Antsukh (Chadakolob), where speakers often judge forms in isolation ungrammatical which they have earlier produced themselves. Oblique singular in Tabasaran is a category notable for such vacillation (cf. Magometov 1965: 100). Although such variation is an indicator of the instability of a system, and of the precarious condition of a language in general, it can also be remarkably long-lived. Thus, although great diversity and variation of oblique stem formation in Lezgian were already recorded by Uslar (1876: 28), some fifteen exponents of this category still co-exist more than a hundred years later.

✐

✐ ✐

✐

✐ 87

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #111

✐

✐

Nominal inflection galore

87

In our field work we often noticed that our consultants would produce and/or understand different inflectional forms, presented out of context, with different degrees of ease, which would seem to point to different ways of storing or accessing them. Comparatively easy to deal with out of context are unmarked inflectional categories, markers which are both productive and regular, markers which are neither productive nor regular (i.e., which need to be memorized individually as forms of the relevant lexical items), markers which are formally and semantically compositional. Inflections come less naturally if a category is productive but the means of its expression are multiple and irregular or also if composite markers are not fully compositional. It would be instructive to see serious psycholinguistic work done on languages with really elaborate inflection of Daghestanian dimensions.16 Presumably, what would be found is both, generation of inflectional forms by rule and learning of entire (and to some extent unanalysed) word forms by rote. Of course, the richer an inflectional system, the more it would be expected to be relying on rules, and agglutinative morphology should favour the rule over the rote strategy, too. Still, on structural grounds and on the evidence of consultants’ reactions, there would also seem to be (not so transparent but textually very frequent) parts of large inflectional systems like the Daghestanian ones which are more amenable to holistic storage and retrieval.

5.3. Paradigm gaps Elaborate inflectional systems are prone to be defective, not providing forms which speakers could conceivably have a use for. Gaps in paradigms may be of different kinds. 5.3.1. Structural gaps Ceteris paribus, there ought to be as many forms in a paradigm as there are terms of each category multiplied by each other: the number of cases times the number of localizations times the number of numbers. However, some combinations may (unaccountably) be illicit. Thus, in Archi a form is missing for the combination of the CONT localization with the essive and translative cases, for no apparent semantic reasons (Table 61, reproducing parts of Table 7). The case paradigm of the cognate spatial adverb ‘in front of’ only partially coincides with the nominal CONT cases (Table 62). By definition, pluralia tantum are defective, lacking singular forms. There are not many nouns of this kind in Daghestanian languages, and some lack a plural for semantic reasons (see below, Sections 5.3.2 and 5.4). But arguably there may also be a structural reason, since some pluralia tantum, exemplified in (45), are count nouns and thus ought to be able to take part in the number opposition.

✐

✐ ✐

✐

✐ 88

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #112

✐

✐

88

Aleksandr E. Kibrik

Table 61. Spatial forms of Archi

IN CONT

ESS

LAT

ELAT

ALL

TERM

TRANSL

-a –

-ak -rak

-aš -raš

-aši -raši

-ak@na -rak@na

-aXut –

Table 62. Spatial inflection of the adverb ‘in front of’ in Archi

‘in front of’

(45) a.

b.

c. d. e. f.

ESS

LAT

ELAT

ALL

TRANSL

–

hara-k

hara-š

hara-ši

hara-Xut

Tabasaran: marXI-ar ‘sledge’ (marXI ‘sledge runner’), mar¯c-ar ‘small horned livestock’ (being the plural of ‘sheep’) Archi: so-¯tor ‘beads’ (cf. sot ‘bead’), šar-um ‘cemetery’ (cf. šar ‘grave’), iXI-mul ‘play’ (cf. iXI ‘joke’), marzi-k’ol-or ‘weaving loom’ (marzi and k’ol being different parts of the tool) Agul (Burshag): c’arm-ar ‘bullock cart’, XIu¯p-ar ‘small horned livestock’ (see above) Tabasaran: pirpi-jar ‘riding saddle’, qIur-ar ‘carting saddle’ Lezgian: q’änar-ar ‘bridle’, Xy¯p-er ‘small horned livestock’ Agul (Burshag): ¯ ◦ -ar ‘weaving loom’, Qaj-ar ‘scabbard’, ruq’-ar ‘lock’, durq’-ar ‘trap’, aX Zˇ an-ar ‘bellows’

What makes these nouns special, and motivates their plural marking, is that they denote entities which are conceptualized as being composed of functionally important parts. Some languages use singularia tantum for precisely such meanings (Bezhta of Khoshar-Khota q’or ‘trap’, Akhvakh tori ‘net’, izo ‘weaving loom’, Chadakolob pil ‘bellows’), suggesting that there is some kind of a semantic rationale to the number defectiveness in such instances. 5.3.2. Semantic gaps In a sense paradigms are idealizations: frequently words will lack, or not use, inflections which a paradigm theoretically provides because they do not suit the meaning of the word to be inflected. Thus, vocatives will not be used a great deal with inani-

✐

✐ ✐

✐

✐ 89

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #113

✐

✐

Nominal inflection galore

89

Table 63. Partial case paradigms in Tsakhur

‘mother’ ‘ox’ ‘apple’

NOM . SG

ERG . SG

INST. SG

ERG . PL

INST. PL

jedy jac eˇc

jed-e: jac-a-n –

– – e-ˇce-n

jedy-a:-š-e: jac-a:-š-e: –

– – eˇc-e:-š1-n

mates in languages which have such a case; no single noun will use both an ergative and spatial cases in languages (such as Burushaski, Lorimer 1935: 56) where the former is limited to animates and the latter to inanimates; etc. In Daghestanian languages, those which have a pure ergative, only expressing the semantic role of agent, limit this case to animate nouns, leaving inanimates with a corresponding gap in their case paradigm. This is true for Tsakhur (Table 63), though with an analytic hitch. There are two exponents for ergative in the singular, one for human (-e, irrespective of number) and the other for non-human animate nouns (-n in singular, -e: in plural). Inanimate nouns have an instrumental form instead of the ergative, expressed by -n (irrespective of number) – which is homonymous with the non-human animate ergative suffix in the singular. More commonly, however, the primary function of the ergative is combined with some other meanings, which precludes such gaps: instrumental in Godoberi, Archi, Andi, etc., and genitive in Lak (excluding personal pronouns, which distinguish ergative from genitive, see Table 23). Languages which have a pure instrumental case restricted to the semantic role of instrument, as in Tsakhur, use this case only with inanimate nouns. Owing to its semantics the affective is always a very restricted case. It is allowed with a small closed set of verbs and is compatible only with animate nouns. Godoberi has four verbs governing it (‘see’, ‘hear’, ‘know’, and ‘forget’, the latter vacillating between affective and dative), and Tsakhur has five (‘see’, ‘hear’, ‘know’, ‘say’, and ‘show’). Although the nominative would seem to be the most neutral, primary case common to all nouns, some Daghestanian languages (including Avaric, illustrated from the Sogratl dialect of Andalal in (46), after Kibrik 1993a, 1993b) have a peculiar inflection for place names, using the form of the spatial case essive as their primary stem; such nouns are after all primarily used in locative functions, and they retain the localization marker in the nominative. (46)

suRraì ‘(in) Sogratl’ (village name, -ì INTER, the essive does not have ¯ ‘(in) an overt marker), idib ‘(in) Idib’ (village name, -b IN), nak’azuX Nakazukh (village name, -X CONT)

✐

✐ ✐

✐

✐ 90

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #114

✐

✐

90

Aleksandr E. Kibrik

A striking example of a semantic restriction imposed on the use of localizations can be seen in the complementary distribution of IN and INTER in Archi. Having no non-spatial meanings, IN and INTER are pure localization terms. Although both have the meaning ‘in a typical position relative to OP’, they distinguish three nonintersecting classes of orientation points: that compatible with IN (nouns as those in (47a), with INTER (nouns as those in (47b), often non-count singularia tantum), and with neither (a negatively defined and semantically correspondingly heterogeneous class of nouns) (Kibrik 1977: 158–163). (47) a.

b.

flat or three-dimensional objects which are potential containers (stove, purse, glove, spoon, tray); flat areas of a limited size (courtyard, roof, back); limited periods of time (week, life); messages of a limited size (letter, book, talk); substances (blood, hay, salt, water); undifferentiated aggregations of objects (pile, flock of birds, village); unlimited water spaces (river, sea); unlimited periods of time (mourning, rebellion); collective names of plants (apples, nuts, nettles).

Count nouns have a straightforward number opposition of singular and plural (‘one’ vs. ‘more than one’). Non-count nouns are typically (though not categorically) defective in number, and Daghestanian languages differ to some extent in treating them as singularia tantum or, far less commonly, pluralia tantum. In (48) are some examples of non-count singularia tantum, with the unmarked number (singular) used where there is no contrast: (48) a.

b.

c.

‘copper’: Chadakolob pax, Andi hir, Akhvakh haˇc’a, Tsez jez, Bezhta bak, Chirag du:ši, Archi duk’, Kryz, Budukh, Khinalug mis; ‘salt’: Chadakolob c’am, Bezhta ca, Lak c’u, Archi orXIi, Tsakhur q’ew, Budukh q’el, Khinalug q’ä, Udi el; ‘pus’: ¯ ◦ a:r, Lak lal, Archi šin, Udi mar. Chadakolob rec’i, Chamalal a:, Tindi X

Nouns with the meaning ‘the totality of countable entities’, where the idea of totality is emphasized, also belong here. For example, Archi lur means ‘eye’ and ‘(a pair of) eyes of one person’, with the plural form lurum reserved for the latter meaning. Other relevant nouns are: (49) a.

‘small horned livestock’: Chadakolob, Bezhta boc’i, Tindi bocar, Khvarshi biL’, Chirag ma¯ca, Archi ri:c’i;

✐

✐ ✐

✐

✐ 91

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #115

✐

✐

Nominal inflection galore

b. c. d. e.

91

‘cattle’: Khvarshi Rolo, Lak Ra¯tara; ‘horses (collective)’: Tsez širšaRIu; ‘goats (collective)’: Tindi cena; ‘sheep droppings’: Chadakolob xil, Akhvakh šili, Chamalal (Gakwari dialect) buši, Chamalal ¯ Tindi hili, Hunzib ìar, Lak šulu. (Gigatli dialect) L’eXa,

Non-count pluralia tantum are commonest in Chirag (Dargic group), with only a very few examples attested elsewhere (50), except again in the case of the meaning ‘totality of countable entities’ (51). (50) a.

b. c.

d. (51) a.

b. c.

Chirag: taIH-ne ‘urine’, nerR-be ‘tears’, šerš-ne ‘saliva’, baIt’baIt’-e ‘liquid excrement’; Archi: nejnej-mul ‘small potato’, zuL’er-¯tu ‘dried apricots’; Tabasaran: ruRan zimz-ar ‘raspberries’ (zimz ‘ant’) (Kondik dialect), ziwz-ar ‘strawberries’ (ziwz ‘ant’) (Burkikhan dialect); Agul: c’iw-ar ‘hips’. ‘small horned livestock’: Andi Xoro-l, Tsez, Kidero dialect beL’ (with no overt PL marker in the nominative, but its plural status shows in the oblique cases, cf. beL’-za-s OBL . PL - GEN ), Lak ja¯tu (ditto, ja¯t-i-l OBL . PL - GEN ); ‘cattle’: Andi Hijda-l; ‘clothes’: Archi k’ob (cf. k’ab-aj OBL . PL - ERG).

5.3.3. Closing gaps With some inflectional categories it is somewhat dubious whether they can be meaningfully applied to a lexical item. For example, its directional meaning makes the translative a natural inflection for nouns such as ‘road’, ‘field’, ‘house’, ‘wall’. But with other nouns it is hard to tell in the abstract whether translatives are possible or impossible; it is often up to the speaker’s intention on a particular occasion whether (s)he finds it appropriate. Extending cases normally limited to animate nouns (ergative, affective) to inanimates needs similar pragmatic or contextual incentives (e.g.,

✐

✐ ✐

✐

✐ 92

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #116

✐

✐

92

Aleksandr E. Kibrik

the context of fairy tales). A counterexample to the complementary distribution of IN and INTER localizations in Archi, mentioned above, is the noun X¯ ◦ ak ‘forest’, which in appropriate contexts permits both localizations, conditioning a reanalysis of its ¯ (INTER) means ‘in the forest, i.e., between the trees’, while meaning: X¯ ◦ ak-e-qI ¯ ¯ X◦ ak-a (IN) means ‘in the forest (as a limited space)’. Given regular inflectional means, pragmatics may thus overrule semantics.

5.4. Appendix: Semantic classes of nouns with defective number Taking any given noun meaning, of those prone to being conceived of as non-count and thus not to participate in the regular number opposition (as touched upon above), it is difficult to find agreement across Daghestanian languages as to the actual number inflection of the noun expressing it. Based on the comparative Daghestanian dictionary of Kibrik & Kodzasov (1990), this section therefore surveys the number behaviour of such nouns by semantic classes. (52) a.

b. c. d. e.

Homogeneous substances without any natural division into discrete units: liquids: ‘water’, ‘broth’, ‘milk’, ‘whey’, ‘beesting’, ‘malt’, ‘beer’, ‘blood’, ‘bile, gall’, ‘urine’, ‘sweat’, ‘liquid excrement’, ‘pus’, ‘pitch’, ‘honey’, ‘sour cream’; gaseous substances: ‘smoke’, ‘fog’; mass nouns: ‘foam’, ‘mould’, ‘dough’, ‘butter’, ‘dirt’, ‘dew’, ‘hoar-frost’, ‘snow’; dry substances that can be poured because they consist of very small units: ‘flour’, ‘oat flour’, ‘salt’, ‘sand’, ‘dust’, ‘soil’, ‘cinders’, ‘ashes’, ‘soot’; solid substances: ‘meat’, ‘cheese’, ‘fat’, ‘gold’, ‘silver’, ‘copper’, ‘lead’, ‘tin’, ‘iron’, ‘pressed dung used as fuel’, ‘charcoal’.

Nouns of this class tend to be treated as singularia tantum (with the exception of Chirag ¯taIH-ne PL ‘urine’, baIt’baIt’-e PL ‘liquid excrement’), like in fact in Russian (with the exception of slivki ‘cream’) and English (with the exception of ashes). It is instructive to see what the plurals of such “mass” nouns actually mean, if a language provides one: (53) a.

DIFFERENT KINDS : ¯ Akhvakh ìe-dila, Khinalug x1n-1b1r ‘kinds of water’; Tabasaran (Kondik dialect) XaX-ar ‘kinds of sour cream’; Chirag XIal-e ‘kinds of fat’, Qanˇcbe ‘kinds of soil’.

✐

✐ ✐

✐

✐ 93

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #117

✐

✐

Nominal inflection galore

b.

c. d. e.

(54)

93

QUANTA OF THE SUBSTANCE : Akhvakh t’oš-i, Chamalal (Gigatli dialect) t’uš-ib, Chirag cˇ ‘it’-re, Kryz q’urq’ušim-bi ‘pieces of lead’; Rutul jak-b1r ‘pieces of meat’, XIadaXIulb1r ‘portions of whey’; Tabasaran (Djubek dialect) XaX-ar ‘portions of sour cream’, k’am-jar ‘portions of butter’, XIamr-ar ‘portions of dough’; Tabasaran (Kondik dialect) c¯ aw-ar ‘portions of beesting’; Andi šiw-odu-l ¯ ◦ aš-ne ‘small groups of cinders’, rug-re ‘portions of milk’; Chirag XI ‘small clumps of dust’. MUCH: ¯ ‘much milk’. Tabasaran (Kondik dialect) -š¯tar ‘much water’, nik-ar DIFFERENT TIMES : Chirag mag-ne ‘sweat (during several days)’. DIFFERENT PLACES : Tabasaran (Djubek dialect) ši¯t-ar ‘water (in different places)’, cˇ ’irw-ar ¯ ‘pus (in different places)’; Akhvakh ìe-dila ‘springs’; Chirag k’um-re ‘sour cream (in different vessels)’, magaš-e ‘cheese (in plates)’; Rutul r1q’-1m1r ‘broth (in different vessels)’.

Homogeneous substances that can be conceptualized as divisible into smaller quanta: ‘brains, cerebrum’ – ‘brain of one person’, ‘tears’ – ‘tear’, ‘saliva’ – ‘spittle, saliva’, ‘snot’ – ‘a snot’, ‘dry snot’ – ‘a piece of dry snot’.

While nouns of this semantic class have a full number paradigm in Russian (with the singular meaning a quantum of the substance), they are singularia tantum in Akhvakh, Chamalal, Tindi, Bezhta, and Lak, pluralia tantum in Chirag, and fully number-inflecting elsewhere (e.g., Andi moGo ‘tear’ – moGo-lidi ‘tears’, toto ‘spittle’ – toto-l ‘saliva’). (55)

Substances or collections composed of homogeneous (usually small) entities not functioning separately: ‘grain’, ‘seeds’, ‘roasted grain’, ‘grass’, ‘hay’, ‘straw’, ‘chaff’, ‘awn’, ‘sheep’s wool’, ‘charcoal’, ‘firewood’, ‘moth’ (see below, s. v. insects).

Nouns in this class are usually singularia tantum. The exceptions found in Archi – nejnej-mul PL ‘small potatoes’ and zuL’er-¯tu PL ‘dried apricots’ – denote entities which are eaten one by one. Sometimes such nouns have a singulative, denoting an individual from the collective. The singulative of ‘grain’ is usually a separate word, for example Chadakolob buQa ‘grain’ – muhPa ‘grain (kernel)’. Lak has a lexical singulative for ‘straw’: c’ajnda ‘straw’ – sun ‘a piece of straw’. If there are plurals for nouns in this class, which there rarely are, these are their possible meanings:

✐

✐ ✐

✐

✐ 94

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #118

✐

✐

94

Aleksandr E. Kibrik

(56) a. b.

c.

KIND : Khvarshi ižu-bo ‘kinds of seed’; Chamalal Zˇ aˇZ-ib ‘kinds of grain’. QUANTA : ¯ Andi sit’-ol ‘heaps of straw’, nik-odul ‘heaps of chaff’; Tabasaran (Djubek) ritn-ar ‘pieces of charcoal’. PLURAL FROM SINGULATIVE: Chirag ar¯sam ‘straw’ – ars-me ‘pieces of straw’; Chamalal (Gakwari dialect) h˜ucˇ ’a ‘grass’ – h˜ucˇ ’a-be ‘blades of grass’, ìunni ‘firewood’ – ìunn-e ‘heap of firewood’, hab ‘charcoal’ – hab-e ‘some amount of charcoal’.

The variation between languages is greatest with nouns for insects. ‘Moth’ is uncountable at least in Chadakolob, Andi, Akhvakh, Chamalal, Tindi, Lak, Archi, Agul (Burshag dialect), Lezgian, Kryz, and Udi. Other insects are mainly countable, but there are exceptions. For example, the nouns for these insects are singularia tantum in the Andic group and in Archi: ‘flea’: Chamalal (Gigatli) cˇ ’anni; ‘louse’: Archi nac’; ‘nit’: Archi nat’, Chamalal (Gigatli) t’ana; ‘tick’: Akhvakh šarašo; ‘ant’: Tindi žunžu; ¯ ‘mosquito’: Andi k’ara; ‘bee’: Tindi, Chamalal (Gigatli) pera.

(57)

Without exception countable are ‘fly’, ‘horse-fly’, ‘spider’, ‘butterfly’, ‘grasshopper’. That flies and their kind tend to be countable, moths and their kind uncountable, and fleas and their kind undecided, would seem to reflect general principles of human categorization (see Wierzbicka 1988). (58) a. b. c. d.

e.

Specific plants and plant products: cereals: ‘wheat’, ‘rye’, ‘millet’, ‘oats’, ‘barley’; grasses and herbs: ‘nettles’, ‘wild onion’, ‘caraway’, ‘mint’, ‘cane’, ‘weed’; vegetables: ‘onion’, ‘garlic’, ‘pea’; fruits and berries: ‘apple’, ‘pear’, ‘apricot’, ‘plum’, ‘cherry-plum’, ‘strawberry’, ‘raspberry’, ‘rose-hip’, ‘grape’, ‘cornel’, ‘barberry’; trees: ‘birch’, ‘oak’, ‘lime’, ‘willow’, ‘nut’.

Nouns in this class are typically uncountable. Cereals are mostly singularia tantum, even in Agul, which has otherwise a very productive plural. Only ‘oats’ is a

✐

✐ ✐

✐

✐ 95

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #119

✐

✐

Nominal inflection galore

95

plurale tantum in the Richa dialect of Agul: alˇcix-ar. Kvarshi has plurals for cereal nouns, usually with the meaning DIFFERENT PLACES (at’-ba ‘plots of wheat’, Ruˇc-bo ‘plots of rye’, niXa-ba ‘plots of oats’); another possible meaning for such plurals is QUANTA, as in Tabasaran (Djubek) (daXIn-ar ‘heaps of wheat’). Nouns for grasses and herbs do not usually denote individualized objects and have only singular forms in Hunzib (miˇc ‘nettles’, so ‘wild onion’, mu¯cer ‘caraway’, sam ‘mint’), Akhvakh, Tindi, Bezhta, Lak and no doubt other languages. Where such plant nouns do have plurals, their meanings are as in (59): (59) a.

b. c.

QUANTA : Khvarshi miˇc-bo ‘bushes of nettles’; Chamalal (Gakwari dialect) tizk’uldi ‘mint plants’. MUCH: Udi meˇ¯c -uruX ‘patches of nettles’. DIFFERENT PLACES : Chamalal (Gakwari dialect) s’im-di ‘cane in different places’.

Among vegetables, ‘pea’ is uncountable in Chirag, Akhvakh, Chamalal, Khvarshi, Tsez, Bezhta, Hunzib, Lak, Archi, Rutul, Tsakhur, and Udi. ‘Onion’ and ‘garlic’ are uncountable in Tsez, Hunzib, Lak, Archi, Tsakhur, and Udi. If there are plurals for plants, they mean PLURAL OF THE SINGULATIVE (Tabasaran (Djubek) gul-ar ‘heads of onion’, serz-er ‘heads of garlic’). Fruits and berries are sometimes classified as ¯ kul ¯ ‘apri˜ ‘pear’, ku uncountable and singularia tantum (e.g., Tindi eˇci ‘apple’, hiha cot’, aXi ‘grapes’, k◦ ank◦ i ‘plum’, q’arat’i ‘strawberry’, g◦ abi ‘raspberry’, k’ye¯ˇsu ‘rose-hip’; similarly in Chamalal (Gigatli), Tindi, Khvarshi, Tsez), or more rarely also pluralia tantum (Tabasaran, Agul). Sometimes a full number opposition is found (e.g., Andi inˇci ‘apple’ – inˇci-l, hi˜hi˜ ‘pear’ – hi˜hi˜-l, Gurˇci ‘apricot’ – Gurˇc-ol, etc.) In general, the bigger the fruit, the likelier it is to be countable. Finally, nouns for trees sometimes denote the kind of tree and are then singularia tantum (e.g., Andi beXu ‘birch’, X◦ o ‘lime’; Tsez me ‘birch’, naHu ‘oak’, waqIar ‘walnut’). When necessary, compound expressions are used for single exemplars: Chamalal (Gakwari) beXuL woha ‘birch tree’. (60)

Unique objects: ‘sky’, ‘sun’, ‘moon’.

Nouns of this class are typically singularia tantum. A plurale tantum for ‘sky’ (zaw-ar) in Agul (Burshag dialect) is exceptional. Full paradigms for ‘sky’ are found in Chirag (¯cab – c¯ ab-re), Tabasaran and some Agul dialects (zaw – zaw-ar), Lezgian (¯cew – c¯ ew-er), Tsakhur (XIaw – XIaj-b1), and Kryz (zow – zow-ni). Where plurals for ‘sun’ and ‘moon’ are possible, they effectuate metaphorical shifts of the core meanings (‘sun’ → ‘day’, ‘moon’ → ‘month’): Chirag banr-e ‘days’; Tabasaran

✐

✐ ✐

✐

✐ 96

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #120

✐

✐

96

Aleksandr E. Kibrik

riR-ar, Lezgian ra¯q-ar ‘sunny days’; Agul raR-ar ‘a lot of sun’; Khvarshi buc-olobo, Hunzib b1c-@wa, Tabasaran waz-ar, etc. ‘months’. (61)

Atmospheric conditions, natural phenomena: ‘wind’, ‘light’, ‘shade’, ‘lightning’, ‘fire’, ‘snow’, ‘rain’.

Such nouns are typically categorized as uncountable and singularia tantum in some Daghestanian languages. For ‘wind’ see Chadakolob huri, Andi moˇci, Chamalal muš, Tindi muˇcu, Tsez ìaci, Bezhta (Tljadal) zaLo, Archi R◦ IalqIi ‘north wind’, luqIe ‘south wind’, Udi muš. Other languages, however, can “count” winds: Khvarshi haca – haca-ba, Bezhta (Khoshar-Khota) zaLo – zaLo-ra (the Tljadal dialect has singulare tantum!), Lak maˇc◦ – maˇc◦ -ri. Similarly ‘light’ is usually singulare tantum: Chadakolob kanti, Akhvakh k◦ ani, Chamalal kunłir, Tsez, Hunzib, Bezhta ¯ (Khoshar-Khota) kanìi (but Tljadal dialect kanìi – kanìi-ja), Lak cˇ ani, Archi akon, ˇ Tsakhur iš1R. If there is a plural, it means DIFFERENT PLACES (e.g., Chirag ¯sala – ¯ˇsal-me ‘lights in different places’, Tindi k◦ ana – k◦ an-ibi ‘lights far away’). ‘Shade’ is uncountable: Chadakolob raštak’u, Tindi an¯cˇu, Tsez Redu, Hunzib h˜udula. The meaning of this noun in these languages can be paraphrased as ‘an area without direct sunlight’ – which is only one of the meanings of the corresponding noun ten’ in Russian: the other is ‘shadow’, and that meaning seems more amenable to pluralization. ‘Lightning’ is uncountable: Chadakolob piri, Tsez maq’Lu, Bezhta (Tljadal) pir (cf. Khoshar-Khota piri – piri-la PL), Hunzib pir, Archi par@la, Agul (Richa dialect) c’arf (cf. Burkikhan dialect c’arp – c’arp-ar PL), Tsakhur jild1r1m, Kryz šimšek, Udi ca¯qI. ‘Fire’ may also be conceptualized as uncountable: Tindi c’a:, Tsez c’i, Bezhta c’o, Hunzib c’@, Archi oc’, Udi aruX. Its plural is usually interpreted as DIFFERENT PLACES : Chadakolob cˇ ’a-jol, Chamalal c’a:-be, Lak c’u-rdal, Chirag c’a-me ‘fire in different places’. ‘Snow’ is a singulare tantum in Avaric, Andic, and Tsezic (e.g., Chamalal anz). Its plural means ‘snowfalls’ (e.g., Udi ižI – ižI-uruX). ‘Rain’ is a singulare tantum: Chamalal (Gigatli) c’aj (cf. Gakwari s’a: – s’a:-be), Tsez qema, Archi t’aIp’an ‘long rain’, teret ‘rain with snow’, Kryz cˇ ’ebiˇZ. However, in Chirag dusre takes plural agreement. (62) a. b. (63) a. b.

Periods: seasons of the year and parts of the day: ‘winter’, ‘spring’, ‘summer’, ‘autumn’; ‘morning’, ‘night’; year and day themselves. The former tend to be uncountable and the latter countable nouns. Wholes composed of countable parts: generic designations of classes of animals: ‘cattle’, ‘small horned livestock’, ‘goats’, ‘horses’; generic designations of the class of bread products: ‘bread’;

✐

✐ ✐

✐

✐ 97

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #121

✐

✐

Nominal inflection galore

c.

97

generic designations of classes of countable heterogeneous entities: ‘clothes’, ‘footwear’, ‘bedding’, ‘dowry’, ‘plates and dishes’.

With nouns of this type (cf. Sapir 1930) it is possible to emphasize the unity of the whole, yielding singularia tantum, or the the plurality of the members of the whole, yielding pluralia tantum. Animals have already been covered. Generic nouns in general are uncommon in Daghestanian languages, which prefer specific names of concrete entities. Bread is one of the rare exceptions because it is an important type of food; each language has many countable nouns for specific bread products. ‘Bread’ is a singulare tantum: Chamalal (Gigatli) rezi, Tsez magalu, Archi X¯ ◦ alli, Tsakhur ginej, Udi šIum, Chirag azajne ‘round bread products with filling’. For the last subcategory both singularia tantum and pluralia tantum are attested: singulare tantum is Chadakolob raQ-miˇcil ‘plates and dishes’ (lit. ‘jug-churn-staff’), Lak t’aIhni-k’iˇc’u ‘jug-bowl’, Budukh Gab-GaˇZaR; plurale tantum is Agul (Burshag) Qab-ar lit. ‘plates’, (Richa) q¯ ab-ar lit. ‘cups’, and Rutul Gab-ar lit. ‘vessels’. To summarize, there is considerable variation across Daghestanian languages and indeed dialects as to the number behaviour of nouns of such semantic classes tending towards uncountability. “Mass” and “total” nouns tend to have a defective paradigm of the singularia tantum type, while nouns which denote “compound entities” favour the pluralia tantum pattern. It is the specific lexical semantics of nouns which is decisive, and translations into Russian or English can here be misleading. The morphology of number is also an influence, insofar as languages with highly regular plural inflection (e.g., Tabasaran, Agul, Lezgian, Rutul) prefer to have full number paradigms regardless of the lexical semantics of nouns, while languages with irregular plural inflection (e.g., Andi, Chamalal, Tindi, Bezhta) are more sensitive to count vs. non-count noun meanings, with paradigms frequently remaining defective. Non-count nouns are, however, productively reinterpreted as count, with meanings such as KINDS OF SUBSTANCE, QUANTA OF SUBSTANCE, MUCH OF SUBSTANCE, PRESENCE IN DIFFERENT PLACES OR TIMES , PLURAL FROM SINGULATIVE, not unfamiliar from other languages, including Russian and English. Sometimes, variation along the lines traced in this section can be made some kind of sense of as reflections of historical changes.

6.

Tracing the history of nominal inflection

The complexities of Daghestanian nominal inflection in general would seem to find some motivation in their history: “Much more of what one finds in morphological systems is the result of the interplay of other areas of grammar, and especially of diachronic change; and it is in these other domains that the basic parameters of a language’s structure are presumably to be sought” (Anderson 1990: 281).17 The synchronic comparison of subgroups ought to provide a window to the past.

✐

✐ ✐

✐

✐ 98

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #122

✐

✐

98

Aleksandr E. Kibrik

6.1. Present traces We first summarize those inflectional phenomena that are relevant for the historical reconstruction of Daghestanian inflection.

6.1.1. Evidence from pronouns It is well known that personal pronouns belong to the core and most conservative part of nominal vocabulary. Therefore the nominative/ergative syncretism in the system of personal pronouns characteristic of many Daghestanian languages (Section 3.1.1) is a crucial piece of evidence for the historical priority of the neutral, non-distinct encoding of pronominal agent and patient arguments. The development of the nominative/ergative distinction in these pronouns varies across the different groups but is in every single instance an innovation under the pressure of a general ergative alignment pattern. Where this pressure has been resisted, it is in line with the animacy hierarchy that (1st and 2nd person) personal pronouns, on top of this hierarchy as being typically used in agentive relation, should be the last nominals to be specifically marked as agents, with inanimate nouns, at the bottom of the hierarchy, spearheading the introduction of agent marking.18 According to a survey of 34 languages with some ergative alignment by Nichols (1993), 26 of them use case marking on both nouns and pronouns, with 16 having ergative alignment of both nouns and pronouns, five having accusative alignment only for pronouns (Sumerian and four Australian languages), one each having accusative alignment for both pronouns and nouns (Sahaptin, Western North America), three-way constrastive alignment only for pronouns (Siuslaw, Western North America), split ergative/accusative alignment for both pronouns and nouns (Burushaski), and with two having split contrastive/accusative alignment for pronouns and ergative alignment for nouns (Djungili and Yukulta, Australia). In another survey of 49 languages with dominant accusative alignment (Johanna Nichols, personal communication), pronouns had accusative alignment in 48 and neutral alignment in only one. In a survey of 103 accusative languages with neutral noun inflection there were twelve with accusative pronoun inflection, indicating that pronouns are more inclined to the accusative pattern than nouns are (cf. the pronominal paradigm in Finnish, which has a pure accusative, while nouns merge the accusative with the genitive or the nominative; see Eliseev 1993: 99). In view of such cross-lingustic evidence, the neutral, non-distinct agent/patient inflection of personal pronouns, though in some sense natural, appears highly unusual. Personal pronouns in Daghestanian languages usually have a suppletive oblique stem. This can be interpreted as a trace of a stage when the oblique syntactic position of personal pronouns did not have the status of a core argument.

✐

✐ ✐

✐

✐ 99

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #123

✐

✐

Nominal inflection galore

99

6.1.2. Evidence from obliques Two-stem inflection distinguishing the oblique cases from the nominative presupposes a stage when this opposition was alive, on the assumption that inflectional systems are structured according to the principle of “one function – one form” (Wheeler 1993: 96; see also Wurzel 1987: 92–93). Along these lines, there is a binary opposition of direct and oblique cases also for attributes in Tsakhur (Kibrik 1995) and for demonstrative pronouns in Khinalug. The sporadically preserved opposition of unmarked vs. marked root excluding the category of number (Section 3.2.5) also testifies to the ancient tendency to give nominative and singular (being the unmarked terms of case and number) special inflectional significance. Categories which are semantically unmarked ought to be encoded as less feature-bearing than semantically marked categories (Wheeler 1993: 109). The multiplicity of oblique singular markers preserves a stage when the direct/ oblique case opposition was not yet generalized and only existed in the form of various noun classes differently interacting with syntactic positions. The historical inference here is similar to that of Skorik (1961: 145–151) for Chukchi, where numerous co-existing nominative singular exponents (zero; partial stem reduplication; suffixes -n, -lg@n, -l@N@n, -@tl@N@n, -N@) are interpreted as traces of an archaic nominal class system. In general, two-stem inflection appears to be rare. In Burushaski, where this phenomenon is attested, only female nouns can have oblique singular stem markers added to all the oblique cases, excluding the ergative; the genitive coincides with the oblique stem in -mo (Lorimer 1935: 61; Table 64). 6.1.3. Evidence from the plural In addition to the multiplicity of plural markers in Daghestanian these also occur in different orders relative to oblique markers, or they may also be cumulated with them (with OBL . PL markers being relatively regular). Further, oblique cases can be derived from a single plural stem, or plural can be expressed independently for the nominative and oblique cases. The order oblique + number contrasts with generally rather regular cross-linguistic patterns of number being closer to stems than case Table 64. The two-stem paradigm of gus ‘woman’ in Burushaski NOM ERG GEN DAT ABL

gus gus-e gus-mo gus-mu-r (o → u before suffix) gus-mu-cum

✐

✐ ✐

✐

✐ 100

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #124

✐

✐

100

Aleksandr E. Kibrik

and other relational marking (cf. Bybee 1985b; Matthews 1991: 171). In one of the rare instances of case being closer to stems than number, seen in Kajtitj (an Arandic Australian language, Koch 1980, Plank 1985: 78), dual and plural markers originate from independent words ‘two’ and ‘many’ and, not yet fully grammaticalized, are still outside nominal case markers. On such evidence plural as a unified category can be inferred not to have been original in Daghestanian. 6.1.4. Evidence from cases The core grammatical cases in Daghestanian, besides the nominative, are the ergative, dative, and genitive (Section 2.2).19 However, their markers do not have common origins. The ergative usually coincides with oblique, but some languages have developed a special ergative marker: -d (Antsukh, Chamalal), -di (Andi), -de (Akhvakh), -ji/-di (Tindi), -l(o) (Hunzib), -a: (Tsez), -e (Chirag), -i (Khinalug), -en (Udi), -r (Kryz), -re (Budukh). The dative is attested everywhere, but there is some formal variation: -l (Bezhta), -j/-li (Chirag), -l/-li (Tindi), -la (Chamalal), -(L)a (Akhvakh), -j (Andi), the class-number marker (Avaric), -s (Archi, Tabasaran, Agul), -z (Hinukh), -r (Tsez), lengthening of the final vowel (Hunzib), -n (Lak), and -V(X) (Udi). The genitive markers, with some languages distinguishing two and with two (Rutul, Tsakhur) not having any, cannot be reduced to a single source either: -(u)l (Antsukh), the class-number marker (genitive 1) and -Li (genitive 2) (Andi), -s and -z (Tsez), -s and -la (Bezhta), -l (Lak), -la (Chirag), -n (Archi, Tabasaran, Agul), zero (Kryz), -u (Budukh, Khinalug), -i and -e (Khinalug), and -Vj/Vn (Udi). The other cases are even less comparable, like the affective markers -ba (Tindi) and -k’le (Tsakhur). The picture is similar for the spatial cases with their own exponents. However, worth noting is the identity of dative and allative markers and genitive and elative markers in Bezhta (Kibrik & Testelec, in press). 6.1.5. Evidence from localizations Corresponding localization markers also have different sources in different languages, often spatial adverbs. Sometimes some localizations are morphologically irregular. Thus, in Archi there are many nouns which preserve an ancient oblique stem in the IN and INTER localizations while using the regular -li marker elsewhere (Table 65). It is also significant that the stressed IN marker -a coincides with one of the OBL . SG markers, so that some nouns using this OBL . SG marker have the same form for the ergative (zero marked) and the in-essive: e.g., ‘sieve’ c’om NOM – c’am-a ERG / IN - ESS, ‘pod’ ìob NOM – ìab-a ERG / IN - ESS .

✐

✐ ✐

✐

✐ 101

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #125

✐

✐

Nominal inflection galore

101

Table 65. Partial paradigms with IN / INTER localizations in Archi

‘hand’ ‘field’ ‘roof’ ‘fire’ ‘village’

NOM . SG

ERG . SG

IN - ESS . SG

INTER - ESS . SG

kul uX harq oc’ XIor

kul-li uX-li harq-li oc’-li XIol-li

kur-a uX-a harq-a – –

– – – c’er-e-qI XIer-e-qI

6.1.6. Evidence from spatial forms The spatial forms show different stages of their development. Some languages, such as Bezhta, preserve what must be the initial transparent system, with localizations and spatial cases freely combinable, while other languages, such as Khinalug, have lost localization altogether. However, the basics of such a spatial system seems so deeply ingrained that it is cyclically reproduced by new morphological means.

6.2. Origins and progress It is self-evident that Daghestanian noun inflection has undergone many stages of development. It is, however, possible to draw a line between a period of protoDaghestanian unity and that of its dissolution. 6.2.1. Proto-Daghestanian In what follows a long period of the historical development of Proto-Daghestanian as such is assumed. The properties attributed to the proto-stage, enumerated in this section, should therefore not be assumed to have evolved simultaneously; but their relative chronology is a subject for future research. Daghestanian inflection combines the agglutination of case and localization marking (with minor allomorphy) with a highly redundant synonymy of basic-stem marking patterns (with semantically impoverished oppositions of singular and plural and direct and oblique). Clearly these opposite morphological techniques will have originated in different periods. On the assumptions that grammatical forms are most transparent and productive at the initial stages of their grammaticalization and that forms grammaticalized earlier are expressed closer to stems than forms grammaticalized later, the modern case inflection must be a later innovation than number and oblique stem marking.

✐

✐ ✐

✐

✐ 102

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #126

✐

✐

102

Aleksandr E. Kibrik

The co-existence of plural-before-oblique and oblique-before-plural order and plural-and-oblique cumulation in contemporary Daghestanian suggests that in ProtoDaghestanian a number opposition was not yet grammaticalized. The contemporary elaborate case systems suggest that they are the results of long evolution at whose origin was a state with no spatial and few grammatical cases. Lacking case and number in the present sense, Proto-Daghestanian arguably had different inflections for different classes of nouns. Gender or class as a determinant of inflectional variants is cross-linguistically quite common.20 Noun class distinctions have survived into the modern period: there is regularly agreement in class (and number) and sometimes class-number agreement markers also serve as ersatz cases. The argument thus is that the singular/plural opposition emerged as a reinterpretation of noun class oppositions. The plural originally appeared, at least in some noun classes, as a derivational category sensitive to meanings of noun roots. This hypothesis is supported by several arguments, both typological and from within the Daghestanian languages. For one thing, there is a correlation between the absence of number and the presence of numeral classifiers.21 Then, the co-existing irregular and unproductive plural markers in Daghestanian, especially in the order pluralbefore-oblique, make diachronic sense as reinterpreted nominal classifiers. Last, in languages with elaborate class systems such as Fula (West-Atlantic group of the Niger-Congo macrofamily; Koval 1979), case is lacking and number is co-expressed with class, class functioning both derivationally and inflectionally. The origin of the case system is primarily connected with the direct/oblique opposition, where the (unmarked) direct covered the core grammatical relations and oblique the more peripheral relations. Though differing from contemporary Daghestanian, such two-way case systems with such a distribution of labour are encountered elsewhere, for instance in Bare (Arawak family, Brazil; Aikhenvald, 1995) and in Riau Indonesian (Sumatra; Gil 1994: 196). The oblique case was split into an ergative (inheriting the oblique form if unmarked) and completive cases (with adverbial or locative meaning), along with the restriction of the direct case to factitive/absolutive meaning (in the sense of Kibrik 1979, 1991b, 1997). This yields the ergative pattern which the Daghestanian languages have preserved to the present time. This portrait of Proto-Daghestanian noun inflection is usefully compared with the reconstructions of Proto-Indo-European by Uhlenbeck (1937) and Trubetzkoy (1939) (as summarized by Stepanov 1975: 176–178). Proto-Indo-European, or the Sprachbund at its origin, supposedly integrated and inherited the conflicting properties of two different types of languages which were in contact with it, a type A language which had one-stem noun inflection, an elaborate system of spatial cases, and nominative-accusative alignment, and a type B language which had two-stem noun inflection, a two-term case system, and nominative-ergative alignment. Tru-

✐

✐ ✐

✐

✐ 103

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #127

✐

✐

Nominal inflection galore

103

betzkoy suggested that type A properties are shared by Uralic and Altaic, while type B properties belong to the so-called Mediterranean group of languages, including, among others, North Caucasian. As outlined here, Proto-Daghestanian bears great similarity to type B. The absence of a system of spatial forms in Proto-Daghestanian does not exclude the possibility of the essence of this pattern having been borrowed from outside, e.g., from contact languages of type A, and from having been developed independently in descendant languages. 6.2.2. After separation All individual Daghestanian languages acquired the grammatical categories of number, case, and localization, drifting in the same direction from their common past. The previous multiplicity of oblique and plural markers is reduced and sometimes fossilized, and regular, productive markers are developed, distributed morphonologically (Archi -mul/-¯ıu PL) or by default rules (Archi -li OBL . SG ); or the previous pattern may also be obliterated entirely (Tabasaran unique plural marker -ar; Khinalug one-stem inflection covering both singular and plural). Somewhat different grammatical cases have been independently developed in different Daghestanian languages; in particular, some have one genitive, others two, yet others none for encoding attributive relations. The greatest changes affected spatial forms, arguably lacking entirely in ProtoDaghestanian. Their development, however, follows universal patterns of grammaticalization, with nouns fusing with postpositions, which in turn derived from spatial adverbs, themselves inflected for directionality in Daghestanian. Once reduced to obligatory affixal markers of localization, such forms are prone to change further (by reduction of the number of oppositions and/or by fusion of the markers of localization and spatial case), and the category of localization itself is endangered or indeed lost in most contemporary Daghestanian languages. But there is also the potential of its re-creation in the course of the cycle of grammaticalization, realized in Godoberi.22 The unbalancing of symmetrical spatial forms is accompanied by their reinterpretation. For example, the Archi partitive -qIiš probably originates from the interelative (contemporary -qIa-š). In general, innovations in the case system follow the invariant Daghestanian principle of using surface cases in strict correspondence to semantic relations, as is typically the case in languages with agglutinative morphology.

7. Afterword Nominal inflection of Daghestanian dimensions – with its proliferation of categories and formal exponents and all kinds of interrelations between them and sensitivities to

✐

✐ ✐

✐

✐ 104

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #128

✐

✐

104

Aleksandr E. Kibrik

the nominals to be inflected – is reasonably well described for most of the members of this family. Such exuberance is a challenge to morphological theory, typology, historical linguistics, and psycholinguistics, wont to feed on more frugal staple diet and consequently sometimes somewhat undernourished.

Acknowledgments I would like to thank Bernard Comrie and Georg Bossong for their very valuable comments. I am especially grateful to Robert Channon and Frans Plank who scrupulously read the paper and provided many very helpful criticisms and suggestions.

Notes 1. In the scheme of Nikolayev & Starostin (1994), Lak, Dargwa, and Khinalug each represent an independent branch, and Inkhokvari (usually seen as a dialect of Khvarshi) is recognized as an additional language within Tsezic. 2. All non-local cases are here subsumed under the rubric of “grammatical” cases. 3. According to the analysis of Jakovlev & Axamaf (1940: 385). But see Kumaxov (1971: 59–64) for a different analysis of the Adyghe case inventory. 4. This approach is also adopted in Comrie & Polinsky (1998). Strangely, this kind of formal system is often ignored in general works devoted to spatial semantics, such as Svorou (1994). 5. As defined here, the notions “localization” and “direction” correspond to Kilby’s (1981) “dimensionality” and “directionality”. In the Russian tradition spatial forms have variously been analysed as “localization” + “case” (Kibrik 1977), “series” + “case” (Madieva 1967), “case” + “series” (Saidov 1958, which is terminologically particularly unfortunate, since the forms called “cases” are precisely those whose functions are not the prototypically syntactically-relational ones normally associated with this category), or simply as cases in their own right (Xanmagomedov 1958, following the precedent of Hjelmslev’s classic), multiplying their number enormously. 6. OP is short for orientation point, elsewhere called “ground/reference object” (Talmy 1983: 232) or “landmark” (Langacker 1986). It should be mentioned that from the point of view of localization concepts as exemplified by Daghestanian languages possession belongs to that domain. 7. For a recent synopsis of this category, traditionally referred to as class rather than as gender in Daghestanian studies, see Corbett (1992). 8. This, however, is counter to two-stem analyses of such nouns or of the archaic lexicon of Indo-European in general, as suggested by Carstairs (1987: 209) or Stepanov (1975: 177). 9. While Kibrik & Kodzasov (1990) only recorded eleven patterns, Kibrik (1991) already had seven general patterns and some twenty subtypes.

✐

✐ ✐

✐

✐ 105

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #129

✐

✐

Nominal inflection galore

105

10. They exemplify what has been called “extended exponence” (Matthews 1972), “one-tomany syntagmatic deviation” (Carstairs-McCarthy 1992: 195), or “sensitive exponence” (Plank 1986: 32). 11. This is much more than is allowed by Plank’s (1986: 46) formal economy constraint, limiting nominal inflectional exponents to about 30. 12. Cf. the peripherality constraint by Carstairs (1987: 193). 13. For Uralic languages, for example, cf. Hajdú (1993: 13), or for Limbu, Van Driem (1987: 34). Such uncertainties are the main reason for variation in case numbers from one grammar to the next. 14. The genitive has on these grounds sometimes been denied casehood, since it does not, or not primarily, express a noun’s relation to verbs (e.g., Volodin 1974: 285). 15. Not infrequently rare cases, like the locative in Latin or the partitive genitive in Russian, are excluded from case inventories in descriptive grammars. For some discussion see Comrie (1991). 16. Taking cues from work such as Stemberger & MacWhinney (1988) and Derwing (1990), but also from structural considerations like those of Bybee (1988), Perlmutter (1988), and Mel’uk (1993). 17. See also Givón (1971), Lehmann (1985), Hopper (1991), McMahon (1994), and Anderson (1992); also Koch (1996) for a contemporary survey of the method of morphological reconstruction. 18. See Silverstein (1976), Croft (1990), Garrett (1990) for the general validity of this reasoning. If personal pronouns innovate any distinctive marking, it ought to be for patient uses, as in Australian (Dixon 1972, Comrie 1991). 19. Cf. the comprehensive study of pre-Lezgic noun inflection by Alekseev (1985). 20. See, e.g., Corbett (1982) and Corbett & Fraser (1993) on Russian, Aronoff (1994: 79– 87) on Latin, or also Lorimer (1935: 25–39) on Burushaski, with its enormous number of about 40 plural suffixes, 25 of them uniquely correlated with only one of the three genders. 21. See Sanches & Slobin (1973), Gil (1987), Lucy (1992: 74). 22. In general see Meillet (1912), Lehmann (1985), Heine, Claudi, & Huennemeyer (1991), Traugott & Heine (1991), Hopper & Traugott (1993); also Plank (1992) for a brief history of the idea of cycles of grammaticalization.

References Abdullaev, Zapir G. 1967 “Darginskij jazyk [Dargwa]”, in: Jazyki narodov SSSR. Tom IV. Iberijsko-kavkazskie jazyki. Moskva: Nauka, 508–523. Aikhenvald, Alexandra Y. 1995 Bare. (Languages of the World/Materials, 100.) München: Lincom Europa. Alekseev, Mixail E. 1985 Voprosy sravnitel’no-istorižeskoj grammatiki lezginskix jazykov. Morfologija, sintaksis [Issues in comparative grammar of the Lezgic languages: Morphology, syntax]. Moskva: Nauka.

✐

✐ ✐

✐

✐ 106

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #130

✐

✐

106

Aleksandr E. Kibrik

Anderson, Stephen R. 1988 “Inflection”, in: Hammond & Noonan (eds.), 23–43. 1990 “Sapir’s approach to typology and current issues in morphology”, in: Dressler et al. (eds.), 277–295. 1992 A-morphous morphology. Cambridge: Cambridge University Press. Andreev, I. A. 1966 “Chuvašskij jazyk [Chuvash]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 43–65. Aronoff, Mark 1994 Morphology by itself: Stems and inflectional classes. Cambridge, Mass.: MIT Press. Baskakov, Nikolaj A. 1966 “Nogajskij jazyk [Nogai]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 280–300. Blake, Barry J. 1977 Case marking in Australian languages. Canberra: Australian Institute of Aboriginal Studies. 1987 Australian aboriginal grammar. Sydney: Croom Helm. Bokarev, Evgenij A. 1959 Cezskie (didojskie) jazyki Dagestana [Tzezic (Dido) languages of Daghestan]. Moskva: Izdatel’stvo AN SSSR. Bybee, Joan L. 1985a Morphology: A study of a relation between meaning and form. Amsterdam: Benjamins. 1985b “Diagrammatic iconicity in stem-inflection relations”, in: John Haiman (ed.), Iconicity in syntax. Amsterdam: Benjamins, 11–48. 1988 “Morphology as lexical organization”, in: Hammond & Noonan (eds.), 119–141. Carstairs, Andrew 1987 Allomorphy in inflection. London: Croom Helm. Carstairs-McCarthy, Andrew 1992 Current morphology. London: Routledge. Comrie, Bernard 1991 “Form and function in identifying cases”, in: Plank (ed.), 41–55. Comrie, Bernard & Greville G. Corbett (eds.) 1993 The Slavonic languages. London: Routledge. Comrie, Bernard & Maria Polinsky 1998 “The great Daghestanian case hoax”, in: Anna Siewierska & Jae Jung Song (eds.), Case, typology and grammar: In honor of Barry J. Blake. Amsterdam: Benjamins, 95–114. Corbett, Greville G. 1982 “Gender in Russian: An account of gender specification and its relationship to declension”, Russian Linguistics 6: 197–232. 1992 Gender. Cambridge: Cambridge University Press. Corbett, Greville G. & Norman M. Fraser 1993 “Network morphology: A DATR account of Russian nominal inflection”, Journal of Linguistics 29:113–142. Croft, William 1990 Typology and universals. Cambridge: Cambridge University Press. Dench, Alan & Nicholas Evans 1988 “Multiple case marking in Australian languages”, Australian Journal of Linguistics 8: 1–48. Derwing, Bruce L. 1990 “Morphology and the mental lexicon: Psycholinguistic evidence”, in: Dressler et al. (eds.), 249–265.

✐

✐ ✐

✐

✐ 107

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #131

✐

✐

Nominal inflection galore

107

Dixon, R. M. W. 1972 The Dyirbal language of North Queensland. Cambridge: Cambridge University Press. Dixon, R. M. W. (ed.) 1976 Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies; New Jersey: Humanities Press. Dressler, Wolfgang U., et al. (eds.) 1990 Contemporary morphology. Berlin: Mouton de Gruyter. Driem, George van 1987 A grammar of Limbu. Berlin: Mouton de Gruyter. Ebert, Karen 1994 The structure of Kiranti languages: Comparative grammar and texts. Zürich: ASAS. Eliseev, Jurij S. 1993 “Finskij jazyk [Finnish]”, in: Jazyki mira: Ural’skie jazyki. Moskva: Nauka, 90–115. Estonskij jazyk (without author) 1993 “Estonskij jazyk [Estonian]”, in: Jazyki mira: Ural’skie jazyki. Moskva: Nauka, 115– 134. Foley, William A. 1986 The Papuan languages of New Guinea. Cambridge: Cambridge University Press. Gadžieva, Ninel Z. 1966 “Azerbajdžanskij jazyk [Azerbaijani]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 66–90. Garrett, Andrew 1990 “The origin of NP split ergativity”, Language 66: 261–296. Gil, David 1987 “Definiteness, noun phrase configurationality, and the count-mass distinction”, in: E. J. Reuland & A. B. ter Meulen (eds.), The representation of (in)definiteness. Cambridge, Mass.: MIT Press, 254–269. 1988 “The structure of Riau Indonesian”, Nordic Journal of Linguistics 17: 179–200. Givón, Talmy 1971 “Historical syntax and synchronic morphology: An archeologist’s field trip”, in: CLS 7: 394–415. Gudava, Togo E. 1976 “Botlixskij jazyk [Botlikh]”, in: Jazyki narodov SSSR. Tom IV. Iberijsko-kavkazskie jazyki. Moskva: Nauka, 293–306. Haiman, John 1980 Hua: A Papuan language of the Eastern Highlands of New Guinea. Amsterdam: Benjamins. Hajdú, Peter 1993 “Ural’skie jazyki [Uralic languages]”, in: Jazyki mira: Ural’skie jazyki. Moskva: Nauka, 7–19. Hammond, Michael & Michael Noonan (eds.) 1988 Theoretical morphology: Approaches in modern linguistics. San Diego: Academic Press. Heath, Jeffrey 1991 “Pragmatic disguise in pronominal-affix paradigms”, in: Plank (ed.), 75–89. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer 1991 Grammaticalization: A conceptual framework. Chicago: University of Chicago Press. Hjelmslev, Louis 1935/1937 La catégorie des cas. (2 volumes.) Aarhus: Universitetsforlaget. Hopper, Paul J. 1991 “On some principles of grammaticalization”, in: Traugott & Heine (eds.), vol. 1, 17–35.

✐

✐ ✐

✐

✐ 108

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #132

✐

✐

108

Aleksandr E. Kibrik

Ibragimov, Garun X. 1978 Rutul’skij jazyk [Rutul]. Moskva: Nauka. Jakovlev, Nikolaj & D. Ašxamaf 1940 Grammatika adygejskogo literaturnogo jazyka [A grammar of literary Adyghe]. Moskva: Izdatel’stvo AN SSSR. Juldašev, Axmed A. 1966 “Baškirskij jazyk [Bashkir]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 173–193. Kert, Georgij M. 1971 Saamskij jazyk (kil’dinskij dialekt) [Saami (Kildian dialect)]. Leningrad: Nauka. 1993 “Saamskij jazyk [Saami]”, in: Jazyki mira: Ural’skie jazyki. Moskva: Nauka, 115–134. Kibrik, Aleksandr E. 1977 Opyt srukturnogo opisanija arˇcinskogo jazyka [Structural description of Archi]. Vol. 2. Moskva: Izdatel’stvo MGU. 1979 “Canonical ergativity and Daghestanian languages”, in: Frans Plank (ed.), Ergativity. New York: Academic Press, 61–77. 1991a “Organizing principles for nominal paradigms in Daghestanian languages: Comparative and typological observations”, in: Plank (ed.), 255–274. 1991b “Semantically ergative languages in typological perspective”, Working papers of North Dakota SIL session 35. Grand Forks, North Dakota: 67–90. 1993a “Sogratl text with interlinear translation and notes”, in: Aleksandr E. Kibrik (ed.), The noun phrase structure in the Andalal dialect as spoken at Sogratl (=EUROTYP Working Papers VII/18), 1–19. 1993b “Andalal noun inflection”, in: Aleksandr E. Kibrik (ed.), The noun phrase structure in the Andalal dialect as spoken at Sogratl (=EUROTYP Working Papers VII/18), 20–30. 1994 “Archi”, in: Rieks Smeets (ed.), Indigenous languages of the Caucasus. Vol. 3. NorthEast Caucasus. II. Delmar, N.Y.: Caravan, Part II, 297–365. 1995 “Direct-oblique agreement of attributes in Tsez and elsewhere in Daghestanian”, in: Plank (ed.), 216–229. 1997 “Beyond subject and object: Toward a comprehensive relational typology”, Linguistic Typology 1: 279–346. Kibrik, Aleksandr E. & Sandro V. Kodzasov 1990 Sopostavitelnoe izuˇcenie dagestanskix jazykov. Imja, fonetika [Comparative studies in Daghestanian languages. Noun. Phonetics]. Moskva: Izdatel’stvo MGU. Kibrik, Aleksandr E., Sandro V. Kodzasov, & Irina A. Muravjova 2000 Jazyk i folklor alutorcev [Language and folklore of the Alutor people]. Moskva: Nasledie. Kibrik, Aleksandr E., Sandro V. Kodzasov, & Irina P. Olovjannikova 1972 Fragmenty grammatiki xinalugskogo jazyka [Studies in Khinalug grammar]. Moskva: Izdatel’stvo MGU. Kibrik, Aleksandr E. & Jakov G. Testelec in press “Bezhta”, in: Michael Job & Rieks Smeets (ed.), Indigenous languages of the Caucasus. Vol. 4. North-East Caucasus. II. Delmar, N.Y.: Caravan. Kilby, David 1981 “On case markers”, Lingua 54: 101–133. Koch, Harold J. 1980 “Kaititj nominal inflection: Some comparative notes”, in: Bruce Rigsby & Peter Sutton (eds.), Papers in Australian linguistics No. 13: Contributions to Australian linguistics. Canberra: Pacific Linguistics A59: 259–276.

✐

✐ ✐

✐

✐ 109

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #133

✐

✐

Nominal inflection galore

109

1990

“Do Australian languages really have morphemes?”, in: Peter Austin, R. M. W. Dixon et al. (eds.), Language and history: Essays in honour of Luise A. Hercus. Canberra: ANU, 193–208. 1996 “Reconstruction in morphology”, in: Mark Durie & Malcolm Ross (eds.), The comparative method reviewed: Regularity and irregularity in language change. New York: Oxford University Press, 218–263. Koptjevskaja-Tamm, Maria 1992 Nominalization. London: Routledge. Koval, Antonina I. 1979 “O znaˇcenii morfologiˇceskogo pokazatelja klassa v fula [On the meaning of the morphological class marker in Fula]”, in: Morfonologija i morfologija klassov slov v jazykax Afriki. Moskva: Nauka. Krejnoviˇc, Evgenij A. 1968 “Ketskij jazyk [Ket]”, in: Jazyki narodov SSSR. Tom V. Mongolskie, tunguso-manˇcžurskie i paleoaziatskie jazyki. Moskva: Nauka, 453–473. Kumaxov, Muxadin A. 1971 Slovoizmenenie adygskix jazykov [Inflection of the Adyghean languages]. Moskva: Nauka. Langacker, Ronald 1986 “An introduction to cognitive grammar”, Cognitive Science 10: 1–40. Lehmann, Christian 1985 “Grammaticalization: Synchronic variation and diachronic change”, Lingua e stile 20: 303–318. Lorimer, D. L. R. 1935 The Burushaski language. Vol.1. Oslo: H. Aschehoug & Co. Lucy, John A. 1992 Grammatical categories and cognition: A case study of the linguistic relativity hypothesis. Cambridge: Cambridge University Press. Madieva, Gjuldžaxan I. 1967 “Avarskij jazyk [Avar]”, in: Jazyki narodov SSSR. Tom IV. Iberijsko-kavkazskie jazyki. Moskva: Nauka, 255–271. Magomedov, A. G. 1966 “Kumykskij jazyk [Kumyk]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 194–212. Magometov, Aleksandr A. 1963 Kubaˇcinskij jazyk [Kubachi]. Tbilisi: Georgian Academy of Sciences. 1965 Tabasaranskij jazyk [Tabasaran]. Tbilisi: Mecniereba. Majtinskaja, Klara E. 1994 “Vengerskij jazyk [Hungarian]”, in: Jazyki mira: Ural’skie jazyki. Moskva: Nauka, 256–279. Matthews, Peter H. 1972 Inflectional morphology. Cambridge: Cambridge University Press. 1991 Morphology: An introduction to the theory of word-structure. Cambridge: Cambridge University Press. McMahon, April M. S. 1994 Understanding language change. Cambridge: Cambridge University Press. Meillet, Antoine 1912 “L’évolution des formes grammaticales”, in: Antoine Meillet, Linguistique historique et linguistique générale. Paris: Champion, 130–148. Mel’ˇcuk, Igor 1993 Cours de morphologie générale. Montréal: CNRS Editions.

✐

✐ ✐

✐

✐ 110

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #134

✐

✐

110

Aleksandr E. Kibrik

Menovšˇcikov, Georgij A. 1968 “Aleutskij jazyk [Aleut]”, in: Jazyki narodov SSSR. Tom V. Mongol’skie, tungusomanˇcžurskie i paleo-aziatskie jazyki. Moskva: Nauka, 386–406. 1975 Jazyk naukanskix eskimosov [Naukan Eskimo]. Leningrad: Nauka. Moravcsik, Edith 1994 “Inflectional morphology in the Hungarian noun phrase”, in: The noun phrase sketch book (=EUROTYP Working Papers VII/22). Revised in this volume. Murkelinskij, Gadi B. 1967 “Lakskij jazyk [Lak]”, in: Jazyki narodov SSSR. Tom IV. Iberijsko-kavkazskie jazyki. Moskva: Nauka, 488–507. Musaev, Kenisbaj M. 1966 “Karaimskij jazyk [Karaim]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 260–279. Nichols, Johanna 1993 “Ergativity and linguistic geography”, Australian Journal of Linguistics 13: 39–89. Nikolayev, Sergej L. & Sergej A. Starostin 1994 Etymological dictionary of North-Caucasian languages. Moskva: Asterisk Publishers. Perlmutter, David M. 1988 “The split morphology hypothesis: Evidence from Yiddish”, in: Hammond & Noonan (eds.), 79–100. Plank, Frans 1985 “On the reapplication of morphological rules after phonological rules and other resolutions of functional conflicts between morphology and phonology”, Linguistics 23: 45–82. 1986 “Paradigm size, morphological typology, and universal economy”, Folia Linguistica 20: 29–48. 1991a “Inflection and derivation”, EUROTYP Working Papers VII/10. 1991b “From cases to adpositions”, EUROTYP Working Papers VII/13. 1991c “Of abundance and scantiness in inflection: A typological prelude”, in: Plank (ed.), 1–39. 1992 “Language and earth as recycling machines”, in: Bernd Naumann, Gottfried Hofbauer, & Frans Plank (eds.), Language and earth. Amsterdam: Benjamins, 221–269. Plank, Frans (ed.) 1991 Paradigms: The economy of inflection. Berlin: Mouton de Gruyter. 1995 Double case: Agreement by Suffixaufnahme. New York: Oxford University Press. Pokrovskaja, Lidija A. 1966 “Gagauzskij jazyk [Gagauz]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 112–138. Rombandeeva, Evdokija I. 1993 “Mansijskij jazyk [Mansi]”, in: Jazyki mira: Ural’skie jazyki. Moskva: Nauka, 283– 301. Saidov, Magomedsaid D. 1958 Avarsko-russkij slovar’ [Avar-Russian dictionary]. Moskva: Enciklopedija. Saltarelli, Mario 1988 Basque. London: Croom Helm. Sanders, M. & Dan Slobin 1973 “Numeral classifiers and plural marking: An implicational universal”, Working Papers on Language Universals 11: 1–22. Sapir, Edward 1930 Totality. (Language Monographs, No. 6.)

✐

✐ ✐

✐

✐ 111

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #135

✐

✐

Nominal inflection galore

111

Silverstein, Michael 1976 “Hierarchy of features and ergativity”, in: Dixon (ed.), 112–171. Skorik, Petr Ja. ˇ 1961 Grammatika cˇ ukotskogo jazyka. Cast’ pervaja. Fonetika i morfologija imennyx cˇ astej reˇci [Chukchi grammar. Part one. Phonetics and nominal morphology]. Leningrad: Nauka. Stemberger, Joseph P. & Brian MacWhinney 1988 “Are inflected forms stored in the lexicon?”, in: Hammond & Noonan (eds.), 101–116. Stepanov, Jurij S. 1975 Osnovy obšˇcego jazykoznanija [Foundations of general linguistics]. Moskva: Prosvešcˇ enie. Svorou, Soteria 1994 The grammar of space. Amsterdam: Benjamins. Talmy, Leonard 1983 “How language structures space”, in: L. Pick & Linda P. Acredolo (eds.), Spatial orientation: Theory, research and application. New York: Plenum Press. Tauli, Valter 1966 Structural tendencies in the Uralic languages. The Hague: Mouton. Timberlake, Alan 1975 “The nominative object in Finnish”, Lingua 35: 201–230. Todaeva, B. X. 1968 “Kalmyckij jazyk [Kalmyk]”, in: Jazyki narodov SSSR. Tom V. Mongol’skie, tungusomanˇcžurskie i paleo-aziatskie jazyki. Moskva: Nauka, 34–52. Trubetzkoy, Nikolaj S. 1939 “Gedanken über das Indogermanenproblem”, Acta linguistica 1: 81–89. Uhlenbeck, C. C. 1937 “The Indogermanic mother language and mother tribe complex”, American Anthropologist 39: 385–393. Uslar, Petr K. 1876 Kjurinskij jazyk [Lezgian]. (Etnografia Kavkaza. Jazykoznanie. VI.) Tiflis. Volodin, Aleksandr P. 1974 “Padež: forma i znaˇcenie ili znaˇcenie i forma? [Case: Form and meaning or meaning and form?]”, in: Sklonenie v paleoaziatskix i samodijskix jazykax. Leningrad: Nauka, 261–291. Wheeler, Max W. 1993 “On the hierarchy of naturalness principles in inflectional morphology”, Journal of Linguistics 9: 95–111. Wierzbicka, Anna 1988 “Oats and wheat: Mass nouns, iconicity, and human categorization”, in: Anna Wierzbicka, The semantics of grammar. Amsterdam: Benjamins, 499–560. Wurzel, Wolfgang U. 1987 “System-dependent morphological naturalness in inflection”, in: Wolfgang U. Dressler (ed.), Leitmotifs in natural morphology. Amsterdam: Benjamins: 59–96. 1989 Inflectional morphology and naturalness. Dordrecht: Kluwer. 1990 “The mechanism of inflection: Lexicon representations, rules, and irregularities”, in: Dressler et al. (eds.), 203–216. Xabiˇcev, M. A. 1966 “Karaˇcaevo-balkarskij jazyk [Karachai-Balkar]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 213–233.

✐

✐ ✐

✐

✐ 112

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #136

✐

✐

112

Aleksandr E. Kibrik

Xanmagomedov, Bejdullax G.-K. 1958 Sistema mestnyx padeˇcej v tabasaranskom jazyke [The system of spatial cases in Tabasaran]. Maxachkala. Zakiev, M. Z. 1966 “Tatarskij jazyk [Tatar]”, in: Jazyki narodov SSSR. Tom II. Tjurkskie jazyki. Moskva: Nauka, 139–154. Žukova, Alevtina N. 1972 Grammatika korjakskogo jazyka [Koryak grammar]. Leningrad: Nauka. Zwicky, Arnold M. 1990 “Inflectional morphology as a (sub)component of grammar”, in: Dressler et al. (eds.), 217–236.

✐

✐ ✐

✐

✐ 113

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #137

✐

✐

Edith Moravcsik

Inflectional morphology in the Hungarian noun phrase: A typological assessment

1. Introduction 1.1. Issues Must the structural characteristics of a language be the way they are or could they be otherwise? The answer that typology has to offer to this seminal question is this: A structural feature is necessarily present in a language either if it holds for all languages or if it holds for a structurally or otherwise delimitable subset of them which that language is a member of. If the occurrence of a structural feature is not fully predictable in this way, it may still be rendered more or less probable by a crosslinguistic tendency. The purpose of this chapter is to approach Hungarian nominal inflection with this question in mind and to seek answers in typology. The nominal subsystem of Hungarian inflection consists of number, case, and the possessor’s person and number marking on the possessum. (1) provides examples.1 (1)

a fiú-k-nak a ház-a the boy-PL-GEN the house-SG3 ‘the boys’ house’

Our task will thus be to explore which properties of number, case, and possessor marking turn out to be either necessary because universal and typological generalizations require them or at least likely because of crosslinguistic preferences. Some of the generalizations that we will appeal to are unrestricted universals of the sort “In all languages, there is (or there is likely to be) X”. Other generalizations will be restricted to proper subsets of languages. The various factors that delimit proper subsets of languages within which a structural feature holds may in principle be aspects of synchronic structure, or features of prior historical stages of the language or of other languages that the one in question has been in contact with; or they may be the geographic area where the language is spoken, or traits of the culture surrounding the language. The present study will consider generalizations of two kinds. On the one hand, we will invoke typological generalizations of the synchronic structural sort: “In all languages, if there is X, there is also (or there is also likely to be) Y”.

✐

✐ ✐

✐

✐ 114

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #138

✐

✐

114

Edith Moravcsik

On the other hand, we will utilize generalizations where the implicans is areal: “In all languages, if they are European, there is (or there is likely to be) X”. The latter kind will be termed Euroversal. Typological assessments of Hungarian morphology have been offered before (Skaliˇcka 1979a; Dressler 1985) featuring Hungarian as a more-or-less agglutinating language. The hoped-for contribution of the present study is to provide a typological account that is more comprehensive – albeit only with respect to a subpart of Hungarian morphology – and more systematic. Informed by the discussions to follow here, we will return to the issue of agglutination in the closing section. Before we can begin to place the Hungarian facts in a typological perspective, a presentation of the facts themselves is called for. How might one distill an inflectional system so that the resulting profile is systematic and comprehensive yet concise?2 In surveying the facts of Hungarian nominal inflection, we will posit twelve parameters designed to capture salient features of meaning-form relations and distribution. The first two parameters concern the meanings and forms that constitute the inflections: (i)

AFFIX MEANINGS

What meanings do affixes express? (ii)

AFFIX FORMS

What are the forms of affixes like? The next seven parameters pertain to the relationship of meanings and forms. Meaning-form relations may be simple or complex. The relationship is simple if the correspondence between the two is either one-to-one or many-to-many and compositional. The relationship may be one-to-one either in a paradigmatic sense or in a syntagmatic sense. Syntagmatically, a meaning-form relation is one-to-one if, in a given context, each form carries a single component of the total meaning and every meaning component is paired with one and only one bit of form (i.e., no portmanteaus and no redundancy). Paradigmatically, the relation is one-to-one if each meaning component has only one possible expression across all contexts in the language and each form can convey only a single meaning across all contexts (i.e., no homonymy and no synonymy). A syntagmatic many-to-many relation is compositional if the meanings of the part forms and of their relations add up to the total meaning. Actual instances of meaning-form correspondence typically depart from these simple patterns. It is these departures that will be the focus of our interest: their kinds and degrees may be taken to form part of the fingerprint of a language. Departures from simple meaning-form correspondence patterns may be the following.3 If the correspondence involves either one meaning or one form, the following six asymmetries may arise:

✐

✐ ✐

✐

✐ 115

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #139

✐

✐

Inflectional morphology in the Hungarian noun phrase PATTERN OF EXPRESSION :

MEANING

FORM

empty morpheme zero marking cumulation extended exponence homonymy synonymy

0 1 or many many (co-present) 1 many (alternative) 1

1 or many 0 1 many (co-present) 1 many (alternative)

115

If there is more than one form and more than one associated meaning, a seventh asymmetry may arise: non-compositionality. To probe into the occurrence and nature of these seven types of mismatch between meaning and form in an inflectional system, we need to ask the following questions: (iii)

ZERO MEANING

Are there any affix forms that have no meaning? (iv)

ZERO FORM

Are there any affix meanings for which there is no overt form? (v)

SYNONYMY

Are there any affix meanings for which there is more than one alternative form? (vi)

HOMONYMY

Are there any affix forms that can alternatively carry more than one meaning? (vii) CUMULATIVE EXPRESSION Are there any affix forms that simultaneously carry more than one meaning component? (viii) EXTENDED EXPONENCE Are there any affix meanings that are repetitiously expressed by more than one form component? (ix) NON - COMPOSITIONALITY Are there any multi-morphemic inflectional sequences where the meanings of the morphemes and of their relations do not add up to the meaning of the entire complex? In addition to meanings, forms, and meaning-form relations, the one remaining aspect of the grammar of inflections is distribution. The distribution of any constituent is completely characterized along three parameters: its cooccurrence possibilities, the linear order in which it appears relative to its co-constituents, and the morphophonological bonding relation that holds between the constituent and its context. Accordingly, we will complete the listing of the twelve parameters by adding the following: (x)

COOCCURRENCE

Given an affix, what stems and what other affixes can it occur with? If an affix or a stem shows allomorphic variation, how are the allomorphs distributed?

✐

✐ ✐

✐

✐ 116

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #140

✐

✐

116 (xi)

Edith Moravcsik LINEAR ORDER

Given an affix, what is its linear position relative to the stem and to the cooccurring affixes? (xii) BONDING Given an affix, how closely is it bonded to the stem and to the cooccurring affixes? Following the listing of the inflections themselves (Section 1.2), it is along these twelve parameters that salient aspects of Hungarian nominal inflection will be surveyed (Section 2).

1.2. Synopsis of the inflections As noted above and illustrated in (1), Hungarian NP-internal inflection consists of number markers, case markers, and the possessor’s person-number marking on the possessed noun.4 The four number markers are the following: (2)

MEANING

singular basic plural possessive plural associative plural

AFFIX FORM ( S ) Ø -k -ak -ok -ek -ök -i -ai -ei -jai -jei -ék

EXAMPLE

kapa kapá-k kapá-i-m János-ék

‘hoe’ ‘hoes’ ‘my hoes’ ‘János and associate(s)’

The three plurals will be abbreviated as PL, PPL, and APL throughout the chapter. Since the possessive plural occurs if and only if the noun to be pluralized is possessed, it may be viewed as a grammatically conditioned variant of the basic plural; although, interestingly, it may choose a different stem allomorph than what the basic plural takes (e.g., ajtó-k door-PL ‘doors’, ajta-jai-m door-PPL-SG1 ‘my doors’). A second inflectional category of Hungarian nouns is case. Case affixes are not easily delimitable in Hungarian. Most of the uncertainty comes from having to decide whether some of the less productive markers are genuine case inflections or adverb-forming derivational affixes. Accordingly, the number of affixes recognized as cases differs from linguist to linguist ranging over 17 (as in Antal 1961: 49 and Abondolo 1988: 180), 18 (Kiefer 2000: 580), 25 (Kiefer 1987) and 27 (Tompa 1968: 206–209). I will assume the following 22 cases. (3)

LABEL

Nominative: Accusative:

AFFIX FORM ( S ) Ø -t -ot -at -et -öt, or Ø

EXAMPLE

fa fá-t fá-m-Ø

‘tree’ ‘tree (ACC)’ ‘my tree (ACC)’

✐

✐ ✐

✐

✐ 117

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #141

✐

✐

Inflectional morphology in the Hungarian noun phrase

Inessive:

-ban -ben, or -ba -be Elative: -ból -b˝ol Illative: -ba -be Superessive: -n -on -en -ön Delative: -ról -r˝ol Sublative: -ra -re Adessive: -nál -nél Ablative: -tól -t˝ol Allative: -hoz -hez -höz Terminative: -ig Dative/Genitive: -nak -nek, or Ø (Genitive only) Temporal: -kor Formal: -kép -képpen Essive-Modal: -ul -ül Essive-Formal: -ként TranslativeFactive: -vá -vé . . .* InstrumentalComitative: -val -vel . . .* Causal-Final: -ért Distributive-nta -nte -onta -anta Temporal: -ente -önte Sociative: -stul -stül -ostul -astul -estül -östül

fá-ban fá-ba fá-ból fá-ba fá-n fá-ról fá-ra fá-nál fá-tól fá-hoz fá-ig fá-nak fa hat-kor fa-kép fá-ul fa-ként

‘in tree’ ‘in tree’ ‘out of tree’ ‘into tree’ ‘on tree’ ‘off tree’ ‘onto tree’ ‘at tree’ ‘from tree’ ‘to tree’ ‘up to tree’ ‘to/of tree’ ‘of tree’ ‘at six’ ‘as tree’ ‘as tree’ ‘as tree’

fá-vá

‘into tree’

fá-val fá-ért

‘with tree’ ‘for tree’

nap-onta

‘every day’

fá-stul

‘with tree’

117

Regarding case labels: they are more or less suggestive of the meanings of the cases they stand for except the term “sublative” which, denying its etymology, is traditionally used to refer to the ‘onto’ relation. There are two glosses that are given for more than one case and thus some explanation is called for. The gloss ‘as’ is assigned to three distinct cases. More precise glosses are ‘in the manner of’ for the Formal case, ‘for the purpose of’ for the Essive-Modal, and ‘in the function of’ for the EssiveFormal. The gloss ‘into’ is assigned to both the Illative and the Translative-Factive. The Illative meaning is spatial directionality; the Translative-Factive designates what something changes into. The chart in (3) lists all allomorphs except for those of the two asterisked affixes for which only two allomorphs are shown: -vá -vé and -val -vel. These two inflections actually have 48 allomorphs each. This is because following consonant-final stems, affix-initial /v/ undergoes total assimilation to the stem-final consonant; e.g., ág ‘branch’, ággal ‘with branch’ (*ágval). Since there are 23 consonants each poten-

✐

✐ ✐

✐

✐ 118

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #142

✐

✐

118

Edith Moravcsik

tially noun-stem-final, and since each assimilated affix variant has both a back-vowel and a front-vowel version, the total number of allomorphs is 2 (for vowel-final stems) + 2 × 23 = 48. The assimilation rule is not a general phonological pattern in the language: it is restricted to the nominal paradigm. There are two /v/-initial verbal affixes (-ve -ve, and -ván -vén) and they do not undergo it. The choice among allomorphs that differ in their vowels is governed by vowel harmony, about which more will be said in Section 2.2.1.2. In addition to affixal case, Hungarian also marks similar content with postpositions, of which it has about 50 (Marácz 1989: 362–363; cf. Keresztes 1975, Korponay 1986b); e.g., a fa alatt the tree under ‘under the tree’ or a fá-n túl the treeSUPESS beyond ‘beyond the tree’. These clitics differ from case affixes in their degree of bondedness to the stem. Nonetheless, since they both resemble and may collocate with case affixes, they will crop up in the discussion. In addition to number and case affixes, the third semantic category of Hungarian nominal inflection consists of markers which indicate the possessor’s person and number. They are the following: (4)

MEANING

AFFIX FORM ( S )

EXAMPLE

SINGULAR POSSESSUM POSSR

SG1 SG2 SG3 PL1 PL2 PL3

-m -om -am -em -öm -d -od -ad -ed -öd -a -e -ja -je -nk -unk -ünk -tok -tek -tök -otok -atok -etek -ötök -uk -ük -juk -jük

könyv-em könyv-ed könyv-e könyv-ünk könyv-etek

‘my book’ ‘your (SG) book’ ‘his book’ ‘our book’ ‘your (PL) book’

könyv-ük

‘their book’

könyv-ei-m könyv-ei-d könyv-ei könyv-ei-nk könyv-ei-tek könyv-ei-k

‘my books’ ‘your (SG) books’ ‘his books’ ‘our books’ ‘your (PL) books’ ‘their books’

PLURAL POSSESSUM POSSR

SG1 SG2 SG3 PL1 PL2 PL3

-m -d Ø -nk -tok -tek -tök -k

These twelve affixes may be analysed into eight morphemes. This is because the SG1, SG2, PL1, and PL2 forms in the plural possessum paradigm are identical to allomorphs of the corresponding markers in the singular possessum paradigm. The only special markers used with plural possessum are for singular and plural 3rd person possessors: -Ø and -k. The eight morphemes are therefore the following: (5)

SG1: SG2:

-m -om -am -em -öm -d -od -ad -ed -öd

✐

✐ ✐

✐

✐ 119

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #143

✐

✐

Inflectional morphology in the Hungarian noun phrase

SG3:

singular possessum: plural possessum:

PL1: PL2: PL3:

singular possessum: plural possessum:

119

-a -e -ja -je Ø -nk -unk -ünk -tok -tek -tök -otok -atok -etek -ötök -uk -ük -juk -jük -k

In sum: we will assume a total of 34 nominal inflections: – 4 numbers; – 22 cases; – 8 person-number markers on the possessum. We are now ready to discuss these affixes from the point of view of the twelve salient features of meaning, form, and distribution laid out in Section 1.1. The survey will result in a profile of Hungarian nominal inflection consisting of 60 summary statements numbered H-1, H-2 etc. and presented piecemeal at the ends of the relevant sub-sections of Section 2 and as a single list in the Appendix. Section 3 will then proceed to examine the facts from a typological point of view to see which are derivable from generalizations of crosslinguistic scope. Section 4 lays out some conclusions.

2. Facts 2.1. Meaning and form 2.1.1. Meanings The first parameter of our descriptive framework is meaning. As shown by the lists in Section 1.2, the meanings of Hungarian nominal inflections are the following: AFFIX MEANINGS : CATEGORIES Affixes mark number, case, and the possessor’s person and number on the possessum.

H-1.

AFFIX MEANINGS : SUBCATEGORIES

H-2. a. b. c.

The 4 number affixes stand for singular, basic plural, possessive plural, and associative plural. There are 22 affixal cases assumed here (or 17–27, depending on the analysis). The 8 markers for the possessor’s number and person differentiate three persons and two numbers.

✐

✐ ✐

✐

✐ 120

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #144

✐

✐

120

Edith Moravcsik

2.1.2. Forms The second parameter of our framework is form: that is, the phonological make-up of the affixes and their allophonic variation. A glance at the charts in Section 1.2 shows that affix forms are generally short: one-syllable-long or less than that. There is no affix which is bisyllabic in all of its variants and there are only five (the Formal, the Distributive-Temporal, the Sociative, the Possessive Plural, and the PL2 possessor marker) that have any bisyllabic allomorphs at all. From the point of view of length, affixes differ from stems. First, an affix form may be a single consonant while stems are always at least monosyllabic. Second, stems may consist of many syllables (e.g., el-visel-het-etlen-ség-etek-ért VERBAL.PREFbear-able-PRIVATIVE-NOMINALIZER-PL2-for ‘for your intolerability’) while affix length, as just noted, is limited to two syllables. Affixes and stems may, however, be similar in phonological make-up. (6) illustrates this with examples of formal identity or near-identity between roots and affixes: (6) a. b. c. d.

affix: root: affix: root: affix: root: affix: stem:

-kor kor -tok tok -nak nyak -nál nyál

‘at the time of (TEMP)’ ‘age’ ‘PL2 on singular possessum’ ‘case’ ‘to (DAT)’, ‘of (GEN)’ ‘neck’ ‘at (ADESS)’ ‘saliva’

Of the 34 affix morphemes, nine are invariant in form: the three zeros and six overt markers: (7)

Singular: Nominative: SG3 on plural possessum: PL3 on plural possessum: Associative plural: Terminative: Temporal: Essive-Formal: Causal-Final:

Ø Ø Ø -k -ék -ig -kor -ként -ért

The remaining 25 affixes have allomorphs. The chart in (8) presents the 24 inflections whose allomorphs are conditioned (rather than freely varying)5 and it indicates the types of allomorphic differences and their clusterings. The affixes are given in the order of increasing numbers of allomorph-differentiating features. The column headings indicating the types of allomorphic differences are as follows:

✐

✐ ✐

✐

✐ 121

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #145

✐

✐

Inflectional morphology in the Hungarian noun phrase

121

FRONT vs. BACK: allomorphs differ in whether they include front or back vowels; RD vs. UNRD: allomorphs differ in whether, from among the front vowels, they include rounded or unrounded vowels; /a/ vs. /o/: allomorphs differ in whether, from among back vowels, they contain /a/ or /o/; #VC- vs. #C: allomorphs differ in whether their initial consonant is or is not preceded by a linking vowel; #VV vs. #V: allomorphs differ in whether the /i/ vowel that they all share is or is not preceded by another vowel; #jV vs. #V: allomorphs differ in whether they do or do not include an initial /j/; /v/-ASSIM: allomorphs differ in the variants of the initial /v/ resulting from assimilation to the stem-final consonant. As can be seen from the chart, apart from the allomorphic overabundance of the two /v/-initial case markers (48 allomorphs each; cf. Section 1.2), the number of allomorphs ranges from two to seven. Except for the presence of the affix-initial prevocalic vowel, which is unique to the possessive plural (-ai and -ei versus -i), all difference types recur over at least two affixes. The prominence of vowels is an overriding theme in allomorphic variation. First, there are more types of vocalic differences than consonantal ones: five of the former against two of the latter. Second, vocalic differences are more general than consonant-based ones: all consonantal differences are segment-specific involving a single consonant type only, while at least some of the vocalic differences pertain to classes of sounds. Third, vowel-based differences are more widespread across affixes than consonantal ones. The restrictedness of consonant-based allomorphy in comparison to vowel-based allomorphy is further shown by the fact that some of vowel-based allomorphy (such as vowel harmony) extends beyond nominal inflection to all of Hungarian morphology, both inflectional and derivational, while the consonant-based differences are specific to nominal inflection. Is the existence and type of allomorphy predictable for an affix? Let us first consider the predictability of the very existence of allomorphy. If we compare the phonological make-up of the nine invariant inflections with the 25 variant ones, some canonical shapes turn out to be unique to one or the other class. For example, if an affix consists of /i/ followed by a consonant, it must be invariant (cf. invariant -ig ‘up to (TERM)’). In other ways, however, members of the two sets are deceptively similar. For example, there are both variant and invariant affixes with /o/ (Allative -hoz and Temporal -kor, respectively). Similarly, the vowel é occurs in both sets. Given that we already know that an affix has allomorphs, the next question is whether the particular shape of the allomorph(s) is predictable. There are indeed some regularities. First of all, the front-back vowel variation is universal across all 24 nominal affixes that have conditioned allomorphs. Some of the other allomorph-

✐

✐ ✐

✐

✐ 122

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #146

✐

✐

122

Edith Moravcsik

(8)

Affix allomorphy DIFFERENCES IN VOWELS

AFFIXES

FRONT

vs. BACK

CASE: -ra -re -ban -ben -ból -b˝ol -ba -be -ról -r˝ol -nak -nek -nál -nél -tól -t˝ol -ul -ül -val -vel... -vá -vé. . . POSS: -unk -ünk -nk -uk -ük -juk -jük -a -e -ja -je PPL: -i -ai -ei -jai -jei CASE: -hoz -hez -höz -on -en -ön -n POSS: -om -em -öm -am -m -od -ed -öd -ad -d -otok -etek -ötök -atok -tok -tek -tök Basic PL: -ok -ek -ök -ak -k CASE: -ot -et -öt -at -t -ostul -astul -estül -östül -stul -stül -onta -anta -ente -önte -nta -nte

‘onto’ ‘in’ ‘out of’ ‘into’ ‘from’ ‘to/of’ ‘at’ ‘from’ ‘as’ ‘with’ ‘into

+ + + + +

‘1PL’ ‘3PL’ ‘3SG’

+ + +

CONSONANTS

RD

/v/-

vs.

/a/ #VC- #VV- #jVvs. vs. vs. vs. UNRD /o/ #C #V #V

ASSIM

+ + + + +

+ + + + +

+

+

‘to’ ‘on’

+ +

+ +

‘1SG’ ‘2SG’

+ +

+ +

+ +

+ +

‘2PL’

+

+

+

+

+

+

+

+

‘ACC’

+

+

+

+

‘with’

+

+

+

+

‘at’

+

+

+

+

+

+

✐

✐ ✐

✐

✐ 123

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #147

✐

✐

Inflectional morphology in the Hungarian noun phrase

123

differentiating features show dependencies. Two examples: if an affix includes an /o/ vowel and it has any allomorphs at all, it must have one containing /e/; and if the /a/ versus /o/ difference is operative across the allomorphs of an affix, the presence versus absence of an affix-initial pre-consonantal vowel is operative among them as well. How does affix allomorphy compare with stem allomorphy? While, as just seen, the overwhelming majority of affixes (25 out of 34) have allomorphs, stems tend to be invariant. Within Papp’s extensive sample including over 30,000 nouns, about 80 % were found to have no variant forms (Papp 1975: 22). Allomorphy is not only less widespread across stems than across affixes but also more meager in terms of the number of variants for any one stem. With few exceptions (e.g., apa ‘father’ with variants apá- and ap-), nouns have only two allomorphs if they have any at all. Nouns that do have allomorphs fall into classes depending on how the oblique (non-Nominative and/or possessed) form differs phonologically from the unpossessed Nominative form, which we will call the base or basic form. Papp (1975: 22–23) assumes a total of nine stem classes depending on the nature of this difference, most of which are unproductive (cf. also Vago 1980: 110–123 and Abondolo 1988: 179–210). Here are examples of the nine allomorphic stem types with the oblique forms exemplifed with the singular 3rd person possessed form (‘his X’): (9)

BASIC FORM

OBLIQUE FORM

MEANING

alma id˝o apa madár kapocs ló tó tetü kehely

almá-ja ide-je ap-ja madar-a kapcs-a lov-a tav-a tetv-e kelyh-e

‘apple’ ‘time’ ‘father’ ‘bird’ ‘snap’ ‘horse’ ‘lake’ ‘louse’ ‘chalice’

The phonological properties that differentiate the oblique form from the basic form in these nine types range over the presence, length, and quality of vowels, the presence of the phoneme /v/, the order of consonants, and combinations of these. Just as in affixes, vowel-based allomorphy dominates: there are more types of it, and they are part of each of the nine patterns, while consonant-based differences show up only in a few types always coupled with a vocalic difference. The nine types of allomorphic variation are listed in (10) under the following headings: differences involving only vowels: V1 # vs. V:1 #: the stem-final vowel varies in length; V1 # vs. V2 #: the stem-final vowel varies in quality;

✐

✐ ✐

✐

✐ 124

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #148

✐

✐

124

Edith Moravcsik

V# vs. Ø#: the stem-final vowel is absent in the oblique form; CV:1 C# vs. CV1 C#: the vowel preceding the final consonant varies in length; CVC# vs. CC#: the vowel preceding the final consonant is absent in the oblique form; differences involving both vowels and consonants: V:1 # vs. V1 v#: the stem-final vowel varies in length; the oblique form has a /v/; V1 # vs. V2 v#: the stem-final vowel varies in quality; the oblique form has a /v/; V# vs. v#: the stem-final vowel is absent in the oblique form, the oblique form has a /v/; C1 VC2 # vs. C2 C1 #: the order of the last two consonants of the stem varies, the vowel between them is absent in the oblique form. (10)

Stem allomorphy

EXAMPLE

DIFFERENCES IN VOWELS ONLY

BOTH VOWELS

V1 # V1 # V# CV:1 C# CVC# vs. vs. vs. vs. vs. V1 :# V2 # Ø# CV1 C# CC# alma almá+ id˝o ideapa apmadár madarkapocs kapcsló lovtó tavtetü tetvkehely kelyh-

V:1 # V1 # vs. vs. V1 v# V2 v#

& CONSONANTS V# vs. v#

C1 VC2 # vs. C2 C1 #

+ + + + + + + +

As seen above, both the existence and shape of affix allomorphs were to some extent predictable. Is this also true for stems? Predictions are possible but only in some instances. First, note that generally it is the oblique form that is predictive of the existence and shape of a separate basic form rather than in reverse. For example, base forms of nouns never end in á and thus from word forms like almá-ban ‘in apple’, the existence and the shape of a distinct base form alma is predictable. Similarly, kelyh ‘chalice’ is not a possible base form and thus from the inflected (Accusative) form kelyh-et, at least the existence – if not the exact form – of a distinct base form follows. Also, if a stem has an oblique form that ends in a vowel followed by /v/, as in tav- ‘lake’, and it does have a base form distinct from the oblique, the base form will drop the /v/ (tó ‘lake (NOM)’).

✐

✐ ✐

✐

✐ 125

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #149

✐

✐

Inflectional morphology in the Hungarian noun phrase

125

There are also dependencies among the particular allomorph-differentiating features so that once one feature differentiating two allomorphs of a stem morpheme has been identified, the presence of another differentiating feature automatically follows. To continue with the example of tó/tav- ‘lake’, it is not only that the oblique form predicts the elision of the /v/ in the base form but this difference in turn entails a difference in the quality or length of the preceding vowel (tó as opposed to ta-). A further example: allomorphs that differ in consonant metathesis (e.g., kehely ‘chalice (NOM)’, kelyh-et ‘chalice (ACC)’) always differ also in the presence versus absence of a vowel between the inverted consonants. The most comprehensive of these dependencies is the one noted above: stem allomorphs are never differentiated by a consonantal feature only; any consonantal difference is always paired with some vocalic difference. However, the predictability of the existence and particular nature of stem allomorphy is not perfect. For example, in the madár/madar- ‘bird’ type, where the short-vowel form is the base form and the long-vowel form is the oblique form, no prediction is possible either way regarding the existence of one allomorph based on the other. Because of the existence of base forms such as zavar ‘embarrassment (NOM)’, the short-vowel form madar- turns out to be a possible base form and is thus not predictive of a different base form. Similarly, because of words such as határ ‘border (NOM)’, which has the same type of base form as madár but which does not shorten the vowel in the Accusative (határ-t ‘border (ACC)’), the long-vowel form madár is a possible oblique form and therefore does not call for a variant. The lack of complete predictability of allomorphic variation on the basis of phonological stem shape is further shown by the fact that homonymous stems may differ in whether they do or do not show allomorphic variation. First, an already inflected stem never has allomorphs even if it happens to be homonymous with an uninflected stem that does show allomorphic variation: (11) a.

b.

NOMINATIVE

ACCUSATIVE

korom soot kor-om age-SG1 torok throat tor-ok funeral.meal-PL

korm-ot soot-ACC kor-om-at (*korm-at) age-SG1-ACC tork-ot throat-ACC tor-ok-at (*tork-at) funeral.meal-PL-ACC

‘soot’ ‘my age’ ‘throat’ ‘funeral meal’

Second, a polysemous stem may show allomorphic variation for one meaning but not the other:

✐

✐ ✐

✐

✐ 126

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #150

✐

✐

126

Edith Moravcsik

(12)

NOMINATIVE

ACCUSATIVE

daru daru

darv-at daru-t

‘crane (bird)’ ‘crane (machine)’

Third, for all but one stem class (the alma/almá- ‘apple’ type), the very same stem morpheme resists allomorphic variation when used as a proper name or as quoted material: (13) a.

b.

NOMINATIVE

ACCUSATIVE

kehely Kehely “kehely” alma Alma “alma”

kelyh-et Kehely-t “kehely”-t almá-t Almá-t “almá”-t

‘chalice’ ‘Mr. Chalice’ ‘(the word) “chalice” ’ ‘apple’ ‘Mr. Apple’ ‘(the word) “apple” ’

Fourth, for some stems a special oblique form is optional even in their normal use: (14)

NOMINATIVE

ACCUSATIVE

falu sátor

falv-at or falu-t sátr-at or sátor-t

‘village’ ‘tent’

While, as (13) and (14) show, differential meaning or use may induce different allomorphy, stems show the same allomorphs when they occur as the final constituent of a compound: (15)

NOMINATIVE

a.

b.

c.

d.

ló hinta-ló swing-horse egér b˝or-egér leather-mouse kapocs irat-kapocs document-clasp madár jó-madár good-bird

ACCUSATIVE

lov-at hinta-lov-at swing-horse-ACC eger-et b˝or-eger-et leather-mouse-ACC kapcs-ot irat-kapcs-ot document-clasp-ACC madar-at jó-madar-at good-bird-ACC or jó-madár-t good-bird-ACC

‘horse’ ‘hobby horse’ ‘mouse’ ‘bat’ ‘clasp’ ‘staple’ ‘bird’ ‘scoundrel’

As will be discussed in Section 2.2.1.1, nominal inflection occurs not only on nouns but also on pronouns, numerals, and adjectives. The various word classes

✐

✐ ✐

✐

✐ 127

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #151

✐

✐

Inflectional morphology in the Hungarian noun phrase

127

differ, however, in their tendency to have stem allomorphs. It will be seen that, by and large, numerals are like nouns in having alternative stem forms while adjectives are less prone to allomorphy and pronouns tend to show more of it. The following statements serve to summarize features of affixal and stem form in Hungarian nominal inflection: AFFIX FORM : SKELETON

H-3. a. b. c.

All 31 overt affixes have at least one monosyllabic allomorph. 7 affixes have subsyllabic allomorphs. 5 affixes have bisyllabic allomorphs. STEM FORM : SKELETON

H-4. a. b.

Stems are minimally monosyllabic and can be many syllables long. Most – but not all – syllabic affix forms are possible root forms.

H-5.

AFFIX FORM : ALLOMORPHY: NUMBER OF AFFIXES WITH ALLOMORPHS Out of the 31 overt affixes, 25 have (non-grammatically-conditioned) allomorphs while 6 are invariant.

H-6.

AFFIX FORM : ALLOMORPHY: NUMBER OF ALLOMORPHS PER AFFIX Except for the two /v/-initial case markers, which have 48 allomorphs each, the number of affix allomorphs per affix ranges from 2 to 7.

H-7.

AFFIX FORM : ALLOMORPHY: DIFFERENTIATING FEATURES

Allomorphs of an affix may differ in the following seven ways: vowels: – front versus back; – if front: rounded versus unrounded; – if back: /a/ versus /o/; – presence versus absence of an affix-initial pre-consonantal vowel; – presence versus absence of an affix-initial pre-vocalic vowel; consonants: – presence versus absence of /j/; – choice of affix-initial consonant. H-8.

AFFIX FORM : ALLOMORPHY: GENERALITY OF DIFFERENTIATING FEA TURES ACROSS AFFIXES

6 of the 7 difference types recur across affixes. H-9.

AFFIX FORM : ALLOMORPHY: CONSONANTAL VERSUS VOCALIC DIF FERENCES AMONG AFFIX ALLOMORPHS

The primacy of vocalic over consonantal differences is borne out in four ways:

✐

✐ ✐

✐

✐ 128

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #152

✐

✐

128

Edith Moravcsik

a. b.

c. d.

There are more types of vocalic differences (5) than of consonantal differences (2). Vowel-based differences are more general than consonantal ones: some of the former hold over classes of vowels while each of the latter involves a single consonant segment type. Vocalic differences are more widespread across affixes than consonantal differences: they affect more affixes. The presence of a consonantal difference implies that of a vocalic one. AFFIX FORM : ALLOMORPHY: PREDICTABILITY OF THE EXISTENCE

H-10.

OF ALLOMORPHS

The existence versus non-existence of affix allomorphs is predictable from some affix forms but not from others. AFFIX FORM : ALLOMORPHY: PREDICTABILITY OF THE SHAPE OF AL -

H-11.

LOMORPHS

a. b.

STEM FORM : ALLOMORPHY: NUMBER OF STEMS WITH ALLOMORPHS

H-12. a. b. H-13.

With one exception, all allomorph sets show vocalic front/back distinction. There are dependencies among allomorphs with respect to allomorphdifferentiating features. Over 80 % of nouns have no allomorphs. Some pronoun stems show more allomorphy than nominal ones; nouns and numerals show more allomorphy than adjectives. STEM FORM : STEM ALLOMORPHY: NUMBER OF ALLOMORPHS PER STEM

The overwhelming majority of nouns with allomorphs have only a single variant in addition to the base. H-14.

STEM FORM : ALLOMORPHY: DIFFERENTIATING FEATURES All stem allomorphs differ in their vowels; some additionally differ in consonants as well. Vowel-related differences: – length of stem-final vowel; – quality of stem-final vowel; – presence or absence of stem-final vowel; – length of the vowel in the stem’s final closed syllable; – presence of the vowel in the stem’s final closed syllable. Consonant-related differences: – presence or absence of stem-final /v/ – order of the stem’s last two consonants.

✐

✐ ✐

✐

✐ 129

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #153

✐

✐

Inflectional morphology in the Hungarian noun phrase

129

STEM FORM : ALLOMORPHY: CONSONANTAL VERSUS VOCALIC DIF -

H-15.

FERENCES

a. b. c.

The primacy of vocalic over consonantal differences is borne out in three ways: There are more types of only-vocalic differences (5) than both-consonantal-and-vocalic ones (4). Vocalic differences are more widespread across stems than consonantal ones: they affect more stems. No stem allomorphs differ only consonantally, i.e., without a concomitant vocalic difference also present. STEM FORM : ALLOMORPHY: PREDICTABILITY OF THE EXISTENCE OF

H-16.

ALLOMORPHS

The existence versus non-existence of additional stem allomorphs is predictable from some stem allomorph shapes but not from others, with the oblique form a better predictor of the existence of a distinct base form than vice versa. STEM FORM : ALLOMORPHY: PREDICTABILITY OF THE SHAPE OF AL -

H-17.

LOMORPHS

a.

b.

The actual shape of a stem allomorph is in some cases (but not always) predictive of that of the others, with the oblique form once again a better predictor of the base form than vice versa. There are also dependencies between one feature of an allomorphic difference and another such feature. STEM FORM : ALLOMORPHY: MONOMORPHEMIC AND INFLECTED

H-18.

STEMS

Inflected stems have no special oblique forms even if they are homonymous with uninflected stems that do. H-19. a.

b.

STEM FORM : ALLOMORPHY: SUSPENSION Given a yet uninflected stem S with distinct base and oblique forms, – an inflected stem that is homonymous with the uninflected stem S may not show the same allomorphic variants; – stem S when used as a quoted word or as a proper name generally (but not always) forgoes the oblique form that it otherwise has; – the use of the distinct oblique form may be optional for stem S even in its normal use. Nouns that have a special oblique form either may or must show this variant even when they occur as the final part of an inflected compound.

✐

✐ ✐

✐

✐ 130

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #154

✐

✐

130

Edith Moravcsik

2.1.3. Zero meanings, zero forms The third and fourth issues that our framework requires us to address are the absence of meaning paired with the presence of form, and the absence of form paired with the presence of meaning. There are no meaningless stems or meaningless affixes in the nominal inflection paradigms in Hungarian. Let us therefore turn to the opposite pattern: stem and affix meanings without overt expression. Affixes and roots, as seen above, share most aspects of their canonical form. There is, however, one kind of phonological realization that shows up exclusively in affixes: zero form. In other words, Hungarian has zero affixes but – apart from one possible exception (see Section 2.2.2) – there are no words that contain overt affixes but lack a stem. Among affixes, there are three inflectional categories whose sole exponent is zero: the singular, the Nominative, and the 3rd person possessor marker on the plural possessum. Examples: (16)

SINGULAR NOMINATIVE

kalap-Ø cf. kalap-ok kalap-Ø cf. kalap-ot

‘hat (SG)’ ‘hats’ ‘hat (NOM)’ ‘hat (ACC)’

SG3 POSSESSOR ON THE PLURAL POSSESSUM kalap-jai-Ø ‘his hats’ cf. kalap-jai-m ‘my hats’ In addition, there are two affixes whose range of exponents includes zero as an option alongside with overt variants: the Accusative and the Genitive. The Accusative is obligatorily marked for unpossessed nouns with the -t affix but the affix is optional for nouns that are possessed by 1st or 2nd person singular possessor: (17)

Látom a ház-am-at. I:see the house-SG1-ACC ‘I see my house.’ Látom a ház-am. I:see the house-1SG cf. Látom a ház-at I:see the house-ACC ‘I see the house.’ *Látom a ház. I:see the house

Regarding the Genitive, Hungarian has no unique affix for this case: the nominal possessor either carries the Dative case affix or it remains unmarked (cf. Szabolcsi

✐

✐ ✐

✐

✐ 131

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #155

✐

✐

Inflectional morphology in the Hungarian noun phrase

131

1992: 43–44; Elekfi 1993). Overt Genitive marking is optional if the possessor immediately precedes the possessum but it is obligatory if the possessor is displaced in that either it precedes the possessum with some other material intervening, or it follows it. Whether it is best to say that Hungarian has no Genitive case and the Dative is pressed into service to mark the Possessor, or that there is a Genitive case which is marked with the Dative affix is a subtle question left undecided here. For expository purposes, however, we will label the affix of the Possessor as Genitive. (18) a.

Possessor directly precedes Possessum: A fiú(-nak a) könyvét olvasom. the boy-(GEN the) book:SG3:ACC read:SG1 b. Possessor indirectly precedes Possessum: A fiú-nak olvasom a könyvét. the boy-GEN read:SG1 the book:SG3:ACC *A fiú olvasom a könyvét. the boy read:SG3 the book:SG3:ACC c. Possessor follows Possessum: Olvasom a könyvét a fiú-nak. read:SG1 the book:SG3:ACC the boy-GEN *Olvasom a könyvét a fiú. read:SG1 the book:SG3:ACC the boy ‘I am reading the boy’s book.’

For personal-pronominal possessors, zero-marking is virtually the only option: (19)

az én könyv-em the I book-SG1 *az én-nek a könyv-em the I-GEN the book-SG1 *én-nek a könyv-em I-GEN the book-SG1 ? nekem a könyv-em to:me the book-SG1 ‘my book’

In sum: H-20.

ZERO MEANING

There are no semantically empty nominal affixes or stems. H-21.

ZERO FORM : AFFIXES AND STEMS

There are no words that contain overt affixes and a zero stem.

✐

✐ ✐

✐

✐ 132

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #156

✐

✐

132

Edith Moravcsik

H-22.

ZERO FORM : ZERO AS EXCLUSIVE OR ALTERNATIVE EXPONENT

a. b.

The singular, the Nominative, and the 3rd person singular possessor on the plural possessum can only be zero-marked. The Accusative and the Genitive occur both overtly-marked and zeromarked.

2.1.4. Synonymy Having considered the meanings and forms of the nominal inflections of Hungarian, let us proceed to the first of the two paradigmatic many-to-one relations that can hold between meaning and form: synonymy. Given affixes, uninflected stems, and inflected stems, synonymy may in principle hold within various pairs of terms. First, if affixes or stems have allomorphs, they will, by definition, be synonymous – an option already explored in Section 2.1.2 above. In addition, there may be morpheme-to-morpheme synonymies: (a) (b) (c) (d) (e) (f)

among affix morphemes; among stem morphemes ; among affix morphemes and stems; as well as synonymies involving inflected words: among inflected words; among affix morphemes and inflected words; among stem morphemes and inflected words.

There are no examples of (e) and (f): an affix morpheme is never semantically equivalent to an inflected word and an uninflected stem never means the same thing as an inflected word (the latter does not occur for analytic reasons: in instances that might look like it we would posit a zero affix). Stem-to-stem synonymy (b) does occur of course (e.g., divány and sezlon ‘sofa’) and so does (c) occasionally: the Dative case affix -nak -nek means in some contexts the same as the Benefactive postposition számára. As far as affix-to-affix synonymy is concerned ((a) above), there are three instances of it: (20)

Accusative: Genitive: Inessive:

-t -at -ot -et -öt, and Ø -nak -nek, and Ø -ban -ben, and -ba -be

Lastly, let us turn to (d): synonymy among inflected words. The most obvious way in which inflected words can come to convey the same meaning is if they include morphemes that are themselves synonymous ((a) and (b) above). The very fact that a language has synonymous affixes or synonymous stems suggests that it should have synonymous inflected words as well. Examples are iskolá-ban and iskolá-ba both meaning ‘in school’ (where the affixes overlap in their meaning ranges) and

✐

✐ ✐

✐

✐ 133

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #157

✐

✐

Inflectional morphology in the Hungarian noun phrase

133

divány-on and sezlon-on both meaning ‘on the sofa’ (where the stem morphemes are synonymous). More interestingly, there are conceivable instances of synonymy between inflected words which do not involve synonymous component morphemes. These are the following: (a) (b) (c) (d)

optional presence versus absence of a morpheme; alternative affix allomorphs occurring with the same stem; alternative stem allomorphs occurring with the same affix; morphemes occurring in alternative orders.

As (21) shows, all four types occur in Hungarian: (21) a.

b.

c.

optional presence versus absence of a morpheme in pronominal case marking: (i) én-nek-em I-DAT-SG1 nek-em DAT-SG1 ‘to me’ (ii) t˝ol-em-t˝ol ABL-SG1-ABL t˝ol-em ABL-SG1 ‘from me’ alternative choice of affix allomorphs in SG3 possessor marking: (i) szék-e chair-SG3 szék-je chair-SG3 ‘his chair’ (ii) anyag-a material-SG3 anyag-ja material-SG3 ‘its material’ alternative choice of stem allomorph in the Accusative: (i) falu-t village-ACC falv-at village-ACC ‘village (ACC)’

✐

✐ ✐

✐

✐ 134

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #158

✐

✐

134

Edith Moravcsik

(ii)

sátor-t tent-ACC sátr-at tent-ACC ‘tent (ACC)’ alternative ordering of morphemes in pronominal plural: enyé-i-m mine-PPL-SG1 enyé-m-ek mine-SG1-PL ‘mine (PL)’

Note that (c) and (d), the choice of alternative stem allomorphs and the differential ordering of morphemes, imply the choice of alternative affix allomorphs as well. There is some tendency for synonymy to be eliminated: affixes that are synonymous on some stems may acquire different meanings on others and types of stem variants synonymous in some instances may semantically diverge in others (cf. Abondolo 1988: 234–239). For example, while of the two pairs of examples in (21b), the first pair – szék-je and szék-e – are entirely synonymous, the second – anyag-ja and anyag-a – differ slightly in that the former is ‘its material’ in an inalienable sense, referring to what something is made of, while anyag-ja is alienable (cf. Kiefer 1985). Similarly, falu-k and falv-ak, and sátor-t and sátr-at in (21c) both mean ‘villages’ and ‘tent (ACC)’, respectively, but a similar word behaves differently: the plural of daru ‘crane’ is either darv-ak or daru-k, but the former must refer to the bird and the latter to the construction machinery. One might expect synonymy to spare pronominal paradigms. Nonetheless, there are interesting instances of alternative pronoun forms even beyond those given in (21d) above. For example, the Instrumental case forms of the demonstratives ez ‘this’ and az ‘that’ have two variants: ez-zel, az-zal, and ev-vel, av-val. While the former two follow the rule of progressive assimilation of the Instrumental -val -vel which is the general rule in nouns (e.g., kéz-zel ‘with hand’, *kév-vel), the latter forms are the result of the same total assimilation applying regressively – a directionality specific to demonstratives but universal for all demonstrative forms. Another instance of pronominal synonymy is the Accusative of the PL1 and PL2 personal pronouns: in addition to the more regular forms of mink-et ‘us’ and titek-et ‘you (PL.ACC)’, there is also bennünk-et and bennetek-et, respectively. These forms are really odd since they involve the Inessive as stem (bennünk ‘in us’, bennetek ‘in you (PL)’) with the Accusative affix (-et) added on. The following statements provide a summary:

✐

✐ ✐

✐

✐ 135

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #159

✐

✐

Inflectional morphology in the Hungarian noun phrase

135

SYNONYMY: AFFIXES , STEMS , AND INFLECTED WORDS

H-23. a.

b.

There is synonymy – between affix morphemes; – between stems; – between affixes and stems; – between inflected words. There is no synonymy – between an affix and an inflected word; or – between a root and an inflected word.

H-24.

SYNONYMY: ZERO INFLECTION The zero marker is synonymous with overt exponents of the Accusative and of the Genitive.

H-25.

SYNONYMY: INFLECTED WORDS There are instances of synonymous inflected nominals differing in each of the following ways: – the presence versus absence of a morpheme; – the choice of affix allomorph, with the stem morpheme remaining the same; – the choice of the stem allomorph and the concomitant choice of the affix morpheme; – the order of morphemes within the word.

H-26.

SYNONYMY: SEMANTIC SPLIT

a. b.

There are pairs of affix allomorphs that are synonymous when added to some stems but show a semantic contrast when added to other stems. There are types of stem variation that preserve meaning in some cases but are semantically distinctive in others.

2.1.5. Homonymy The sixth parameter in the description of inflectional systems we adopted has to do with the second of the two paradigmatic many-to-one relations between meanings and forms identified in Section 1.1: homonymy.6 In Hungarian nominal inflection, there are examples of a form being homonymous between any of the following: (22) a. b.

two uninflected stems: szél ‘wind’ and ‘edge’ an uninflected and an inflected stem: korom ‘soot’ and ‘my age’ (kor-om) állam ‘state’ and ‘my chin’ (áll-am) torok ‘throat’ and ‘funeral feasts’ (tor-ok)

✐

✐ ✐

✐

✐ 136

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #160

✐

✐

136

Edith Moravcsik

c.

d.

a stem and an affix: kor ‘age’ and ‘at the time of (TEMP)’ ön ‘you (FORMAL)’ and ‘on (SUPES)’ two inflected stems hat-ok ‘sixes’ and ‘I influence’

In addition, there are forms that are homonymous between two affix meanings. First of all, as seen above, the zero marker is five-ways homonymous as a noun affix: it is the exclusive expression of the singular, the Nominative, and the 3rd person singular possessor on a plural possessum and it is one of the two ways to mark the Accusative and the Genitive. Also, it indicates the 3rd person singular subject on most verbs. (23) a.

b.

c.

d.

e.

zero as singular marker: hajó-Ø boat-SG ‘boat’ zero as Nominative marker: hajó-Ø boat-NOM ‘boat (NOM)’ zero as alternative Accusative marker: hajó-m-at boat-SG1-ACC hajó-m-Ø boat-SG1-ACC ‘boat (ACC)’ zero as alternative Genitive: az ember-nek a hajója the man-GEN the SG3:boat az ember-Ø hajója the man-GEN SG3:boat ‘the man’s boat’ zero as the 3rd person singular possessor marker: on the plural possessum: kabát-ai-Ø coat-PL-SG3 ‘his coats’ cf. kabát-ai-m coat-PL3-SG1 ‘my coats’

✐

✐ ✐

✐

✐ 137

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #161

✐

✐

Inflectional morphology in the Hungarian noun phrase

f.

137

kabát-ai-d coat-PL3-SG2 ‘your coats’ zero as verbal person and number marker: jár-Ø walk-SG3 ‘he walks’ cf. jár-ok walk-SG1 ‘I walk’ jár-sz walk-SG2 ‘you (SG) walk’

In addition to the zero marker, the following six affixes are homonymous: (24) a.

b.

c.

d.

-ja -je -a -e as both singular and plural 3rd person possessor on the singular possessum if the possessor is nominal: az ember-nek a könyv-e the man-GEN the book-SG3 ‘the man’s book’ az ember-ek-nek a könyv-e the man-PL-GEN the book-SG3 ‘the men’s book’ -nak -nek as both Dative and Genitive marker: Tomi-nak adom a tortát. Tommy-DAT I:give the cake:ACC ‘I give the cake to Tommy.’ Tomi-nak a tortá-ja Tommy-GEN the cake-SG3 ‘Tommy’s cake’ -ba -be as both Inessive and Illative: a könyv-be the book-INESS ‘in the book’ a könyv-be the book-ILL ‘into the book’ -k as both basic plural marker and 3rd person plural possessor marker on the plural possessum: kefé-k brush-PL ‘brushes’

✐

✐ ✐

✐

✐ 138

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #162

✐

✐

138

Edith Moravcsik

e.

f.

kefé-i-k brush-PPL-PL3 ‘their brushes’ -i as both possessive plural and denominal adjectivalizer: Budapest-é-i Budapest-POSS-PPL ‘those of Budapest’ Budapest-i Budapest-ADJCT ‘being from Budapest’ -ék as both associative plural and deverbal nominalizer: Bori-ék Bori-APL ‘Bori and associate(s)’ borít-ék cover-ADJCT ‘envelope’

Some of these cases exemplify non-accidental homonymy, or polysemy, while others seem coincidental. Note also that some of the examples show homonymy of a single allomorph of a morpheme only: the denominal adjectivalizer and the possessive plural overlap in form only in -i, which is the sole form of the former but only one of the several allomorphs of the latter. A striking fact is the multiple homonymy of the zero marker. In sum: HOMONYMY: AFFIXES , STEMS , AND INFLECTED WORDS

H-27. a. b. c. d. e.

There are forms that are homonymous between uninflected stem meanings. There are forms that are homonymous between affix meanings. There are forms that are homonymous between a stem meaning and an affix meaning. There are forms that are homonymous between inflected and uninflected stem meanings. There are forms that are homonymous between inflected stem meanings.

H-28.

HOMONYMY: ALL ALLOMORPHS OR ONLY ONE Homonymy may hold over all allomorphs of a morpheme or involve one allomorph only.

H-29.

HOMONYMY: ZERO INFLECTION

Zero marking shows more extensive homonymy than any other affix.

✐

✐ ✐

✐

✐ 139

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #163

✐

✐

Inflectional morphology in the Hungarian noun phrase

139

2.1.6. Cumulative exponence Of the various logically possible complex symbolization patterns listed in Section 1.1, we have so far considered zero meaning, zero form, and the two paradigmatic many-to-one relations: synonymy and homonymy. We will turn next to one of the two many-to-one relations that may hold syntagmatically: cumulation – i.e., the joint expression of more than one meaning element by a single form. Stems of course vastly outdo affixes in this regard: the pattern of a single morpheme conveying a cluster of meanings is prevalent across the lexicon of any language. In addition, however, affixes also show cumulation in Hungarian nominal inflection and, furthermore, stem and affix meanings may also be expressed cumulatively. The following components cumulate: (a) (b) (c) (d)

position and orientation in spatial case affixes; person and number in possessor marking; association, alikeness, and inclusiveness in the associative plural; stem and affix in pronominal case marking.

Let us consider them in turn. C UMULATION OF POSITION AND ORIENTATION IN SPATIAL CASE AFFIXES Case affixes, as well as postpositions, of spatial content show a fairly consistent three-way contrast between location, source, and destination (Abondolo 1988: 248– 250): (25) a.

case affix triplets: AFFIXES

Inessive:

EXAMPLES

Elative: Illative:

-ban -ben, or -ba -be -ból -b˝ol -ba -be

fá-ban fá-ba fá-ból fá-ba

‘in tree’ ‘in tree’ ‘out of tree’ ‘into tree’

Superessive: Delative: Sublative:

-n -on -en -ön -ról -r˝ol -ra -re

fá-n fá-ról fá-ra

‘on tree’ ‘off tree’ ‘onto tree’

Adessive: Ablative: Allative:

-nál -nél -tól -t˝ol -hoz -hez -höz

fá-nál fá-tól fá-hoz

‘at tree’ ‘from tree’ ‘to tree’

✐

✐ ✐

✐

✐ 140

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #164

✐

✐

140

Edith Moravcsik

b.

a postposition triplet: POSTPOSITIONS

EXAMPLES

‘under’

fa alatt fa alól fa alatt-ról fa alá

‘from under’ ‘to under’

alatt alól or alatt-ról alá

‘under tree (location)’ ‘from under tree’ ‘under tree (direction)’

The meaning components cumulated are position relative to an object (such as ‘inside’ or ‘on the surface of’ or ‘near’) and orientation (such as ‘location’, ‘source’, and ‘destination’).7 The degree of cumulation, however, differs. There are two patterns showing non-cumulation. While among the three positions, ‘inside’ and ‘on the surface of’ do show some subregularities in expressing all three orientational categories: location (-n), source (-ól, -˝ol), and direction (-á -é), affixes for the third position – ‘near’ – show -ól for source but have no marker for location and direction. In other words, the position ‘near’ shows more complete cumulation than ‘inside’ and ‘on the surface of’, and the orientation categories ‘location’ and ‘destination’ cumulate more than ‘source’. P ERSON AND NUMBER IN POSSESSOR MARKING Recall the paradigm of possessor markers on the possessum given in (5) (Section 1.2) and repeated below: (5)

SG1: SG2: SG3:

singular possessum: plural possessum:

PL1: PL2: PL3:

singular possessum: plural possessum:

-m -om -am -em -öm -d -od -ad -ed -öd -a -e -ja -je Ø -nk -unk -ünk -tok -tek -tök -otok -atok -etek -ötök -uk -ük -juk -jük -k

In marking possessed nouns for the number and person of the possessor, there are three categories involved: the person of the possessor, the number of the possessor, and the number of the possessum (the possessum’s person does not figure since only 3rd person entities may be possessed). Of these three categories, the possessum’s number does not cumulate with the possessor’s number or person: there is a distinct marker for the possessum’s plural (the so-called possessive plural). The possessor’s person and plurality, however, do receive joint expression: while the person markers for the singular possessor each mark a single meaning element: SPEAKER, AD DRESSEE, and OTHER , respectively, the three plural affixes each convey a complex meaning:8

✐

✐ ✐

✐

✐ 141

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #165

✐

✐

Inflectional morphology in the Hungarian noun phrase

141

PL1: SPEAKER AND AT LEAST ONE OTHER PL2: ADDRESSEE AND AT LEAST ONE OTHER ( NOT SPEAKER ) PL3: MORE THAN ONE ( NOT SPEAKER OR ADDRESSEE ) While it would in principle be possible for these meaning clusters to receive separatist expression – with SPEAKER, ADDRESSEE, and OTHER each having its own affix – the affixes listed above opt for cumulation. C UMULATION OF ASSOCIATION , LIKENESS , AND INCLUSIVENESS IN THE ASSO CIATIVE PLURAL

The associative plural, as in Bélá-ék ‘Béla and his associate(s)’, conveys more than one meaning element: X (where X is one or more unique humans) AND AT LEAST ONE Y SUCH THAT Y IS LIKE X AND IS ASSOCIATED WITH X. The associative plural thus appears to be a true instance of cumulation. For a detailed discussion of the semantics involved, see Gil (1994b). There is nonetheless an analysis under which this affix does not cumulate. According to this view, the marker is bimorphemic. Lotz (1968) equates the -é of -ék with the possessum-deriving affix in Bélá-é ‘that of Béla’, and the -k with the regular nominal plural marker. There are three advantages of the bimorphemic analysis. First, it explains the invariance of -ék: why it is that the -é of -ék does not disappear after vowel-final stems as is the case in other vowel-initial affixes, and why is it that it does not participate in vowel harmony as vowels of most affixes do. The explanation that the bimorphemic analysis affords is by reference to the fact that the affix -é, which under this analysis is a component of -ék, is itself invariant: it does not disappear after vowel-final stems and it is exempt of vowel harmony. (26) shows the identical phonological behavior of the affixes -é and -ék in contrast to the basic plural marker, whose initial vowel comes and goes depending on whether the stem is consonant- or vowel-final and changes quality depending on the backness and frontness of the stem vowels. (26) a.

b.

associative plural -ék and possessive -é: front-vowel stem: Péter-ék Péter-é back-vowel stem: Sándor-ék Sándor-é vowel-final stem: Sárá-ék Sárá-é basic plural: front-vowel stem: egér eger-ek back-vowel stem: madár madar-ak

‘Péter and associate(s)’ ‘Péter’s’ ‘Sándor and associate(s)’ ‘Sándor’s’ ‘Sára and associate(s)’ ‘Sára’s’ ‘mouse’ ‘mice’ ‘bird’ ‘birds’

✐

✐ ✐

✐

✐ 142

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #166

✐

✐

142

Edith Moravcsik

vowel-final stem:

vese vesé-k

‘kidney’ ‘kidneys’

The second attraction of the bimorphemic analysis is that it explains why -ék occurs with the particular stem allomorphs that it does. There is once again an exact correspondence between the two affixes -é and -ék: they both occur with the base form of the stem while other affixes, as seen in Section 2.1.2, often require a special oblique form. (27) illustrates this; the associative plural forms are given in quotes since they are formed from unpossessed common nouns and are therefore semantically anomalous – although they are possible if the nouns serve as family names: (27) a.

b.

ló lov-ak ló-é “ló-ék” madár madar-ak madár-é “madár-ék”

‘horse (SG.NOM)’ ‘horse (PL)’ ‘(the) horse’s’ ‘(the) horse and associate(s)’ ‘bird (SG.NOM)’ ‘bird (PL)’ ‘(the) bird’s’ ‘bird and associates’

If the -é affix is seen as part of the -ék affix, the correspondence between the stem allomorphs that the two select is explained. Thus, all in all, the bimorphemic analysis accounts for the alikeness between the -é and the -ék affixes both in regard to their own allomorphic invariance and in regard to the stem variants they take. Thirdly, there is evidence from some Dunántúl dialects of Hungarian that shows the independence of -é and -k in -ék. In these varieties, the regular plural affix -k is used in the associative meaning; e.g. standard Hungarian Pistá-ék would be expressed as Pistá-k (Korchmáros 1995: 296). Alongside its virtues, the bimorphemic analysis nonetheless has a drawback: it does not make for compositional structure. Compositionally, the meaning of Bélá-ék should be ‘those of Béla’ since Bélá-é means ‘that of Béla’ and -k means ‘(basic) plural’. The actual meaning, however, departs from this in two ways: first, ‘those’ have to be people that are close associates; second, the set of Bélá-ék must include Béla himself as well. (The actual translation of ‘those of Béla’ would be Bélá-é-i where -i is the possessive plural marker.) Thus, the implied PLURAL component can be attributed to the -k and the ASSOCIATED component to the -é but the remainder of the meaning – ‘LIKE X’ and ‘INCLUDES X’ – is not linkable to any portion of the form. The associative plural marker is not an extreme case of non-compositionality: it is not that the component meanings have nothing to do with the total meaning; however, they constitute only a subpart of the semantics of the entire expression. The form-meaning relation manifested in the associative plural thus turns out to be special under both the monomorphemic and the bimorphemic analysis. Under the

✐

✐ ✐

✐

✐ 143

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #167

✐

✐

Inflectional morphology in the Hungarian noun phrase

143

monomorphemic view (adopted by Tompa 1968: 128; Kálmán 1985: 257; Korchmáros 1995; Bartos 2000: 694–699), -ék is an example of cumulative exponence: a single morpheme expressing more than one bit of meaning. Under Lotz’s bimorphemic analysis (adopted also by Abondolo 1988: 182), it is an instance of noncompositional expression: more than one morpheme shouldering the total meaning, with the meanings of the parts, however, falling short of the entire meaning. C UMULATION OF STEM AND AFFIX IN PRONOMINAL CASE MARKING While stem and case are normally expressed through two distinct morphemes, in 1st and 2nd person personal pronouns – especially in the singular – the two are not clearly distinguishable. Consider the following accusative forms: (28) a. b. c. d.

én engem te téged mi minket ti titeket

‘I’ ‘me’ ‘you (SG.NOM)’ ‘you (SG.ACC)’ ‘we’ ‘us’ ‘you (PL.NOM)’ ‘you (PL.ACC)’

The accusative forms are best considered suppletive and thus examples of cumulation. Here is a set of summary statements: CUMULATION : AFFIX MEANINGS

H-30.

The following affix meanings cumulate: – position and orientation in spatial case affixes; – person and number in possessor marking; – association, alikeness, and inclusiveness in the associative plural. H-31. a. b.

H-32.

CUMULATION : AFFIX MEANINGS : SPATIAL CASES Of the three positions (‘inside’, ‘on the surface of’, and ‘near’), ‘near’ cumulates most completely with orientation. Of the three orientations (‘location’, ‘source’, and ‘destination’), ‘source’ cumulates least with position. CUMULATION : STEM AND AFFIX The only instances of stem-affix portmanteaus are the 1st and 2nd person pronouns’ Accusative and plural.

✐

✐ ✐

✐

✐ 144

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #168

✐

✐

144

Edith Moravcsik

2.1.7. Extended exponence The second of the two syntagmatic many-to-one relations – the sixth parameter on the list given in Section 1.1 – involves multiple expressions of the same meaning either all within a single word or distributed over a larger domain. This pattern – extended exponence – crops up in all three inflectional categories: number, case, and the possessor’s person and number marking. Let us look at each in some detail. Redundant case marking is frequent in the Accusative of personal pronouns. Note first that, while the 3rd person singular pronoun, just like nouns and demonstrative pronouns, shows the regular nominal accusative marker -t, the 1st and 2nd person pronouns form their basic Accusative by suppletion (see also (28) above): (29) a.

b.

c.

d.

e.

noun: n˝o ‘woman (NOM)’ n˝o-t ‘woman (ACC)’ demonstrative pronoun: ez ‘this (NOM)’ ez-t ‘this (ACC)’ 3rd person singular pronoun: o˝ ‘he’ o˝ -t ‘him (ACC)’ 1st person singular pronoun: én ‘I’ engem *én-t ‘me (ACC)’ 2nd person singular pronoun: te ‘you (SG)’ téged *te-t *té-t ‘you (SG.ACC)’

However, colloquial speech allows for the optional addition of the regular Accusative

✐

✐ ✐

✐

✐ 145

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #169

✐

✐

Inflectional morphology in the Hungarian noun phrase

145

suffix to the 1st and 2nd person suppletive forms bringing them somewhat more in line with the regular Accusative marking pattern (Vago 1980: 125) but at the same time introducing redundancy: (30) a.

engem-et I:ACC-ACC ‘me (ACC)’ téged-et you(S):ACC-ACC ‘you (SG.ACC)’

Interestingly, the colloquial option of doubly marking the Accusative of singular 1st and 2nd person pronouns also extends to the 3rd person singular pronoun and the demonstrative pronoun even though they form their Accusative regularly. These pronouns thus end up with a sequence of two separate Accusative markers: (30) b.

o˝ he ‘he’ o˝ -t he-ACC o˝ -t-et he-ACC-ACC ‘him’ ez this ‘this (NOM)’ ez-t this-ACC ez-t-et this-ACC-ACC ‘this (ACC)’

Double marking of pronouns occasionally involves other cases as well (cf. Tompa 1968: 195, 197, 198): (31) a.

double Adessive: nál-am than-SG1 nál-am-nál than-SG1-than ‘than me’

✐

✐ ✐

✐

✐ 146

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #170

✐

✐

146

Edith Moravcsik

cf.

b.

ház-nál house-than ‘than (the) house’ double Allative: t˝ol-em ALL-SG1 t˝ol-em-t˝ol ALL-SG1-ALL ‘from me’ cf. ház-tól house-ALL ‘from (the) house’

A second occurrence of extended exponence is in person and number marking. First, in some oblique case forms of personal pronouns, person and number may optionally be expressed twice: once as what appears to be the stem and once as an affix (for further discussion, see Section 2.2.2): (32) a.

b.

t˝ol-em ALL-SG1 én-t˝ol-em I-ALL-SG1 ‘from me’ t˝ol-e ALL-SG3 o˝ -t˝ol-e he-ALL-SG3 ‘from him’

Second, similar redundancy shows up in the possessor’s person and number marking on the possessum. Within the word, the possessor’s person and number is marked only once. However, the possessum, already marked by the possessor’s person and number affix, may optionally be accompanied by a free possessor pronoun as well. Thus, ‘my book’ is alternatively a könyv-em or az én könyv-em. De Groot (1989: 67– 94) convincingly analyses the free pronoun as an appositive to the inflection, rather than as a constituent that controls inflectional agreement. A pattern asserts itself across both instances of redundant marking of person and number just seen. In both cases – one involving case-marked pronouns and the other, possessed nouns – one of the two synonymous person-number markers is optional and in both cases it is the free pronoun, rather than the affix, that is dispensable. Apparently, where multiple synonymous exponents cooccur with one exponent optional and with one exponent more free than the other, it is the freer element that is dispensable and the bound one is obligatory.

✐

✐ ✐

✐

✐ 147

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #171

✐

✐

Inflectional morphology in the Hungarian noun phrase

147

In two of the examples of extended exponence seen so far, the multiple markers occurred all in the same word. The redundant personal pronoun possessor with possessed nouns, however – such as in az én könyvem the I book-my ‘my book’ – illustrates repetitive exponence that extends beyond a single word. As we now turn to the third inflectional category – number – that instantiates extended exponency, we will find yet another example of extended exponency spread over the phrase. In Hungarian NPs, number, as well as case, are obligatorily expressed on both the noun and its demonstrative determiner: (33)

ez-ek-et a kék madar-ak-at this-PL-ACC the blue bird-PL-ACC ‘these blue birds’

The demonstrative is the only case-and-number-marked noun satellite in Hungarian: no other NP-internal constituent agrees with the noun in anything. This feature of the demonstrative connects with some other facts that all point to its somewhat headlike nature and to the concomitant appositive structure of the demonstrative-noun construction (cf. Moravcsik 1997; for a recent analysis of Hungarian demonstrative constructions, see Payne & Chisarik 2000). First, the demonstrative can cooccur with other determiners, such as the definite article or a possessor. Given that one would expect a noun to be “determinable” only once, such constructions are best analysed with one of the two determiners forming a separate NP all by itself. That it is the demonstrative that is the independent element, rather than the article or the possessor, is shown not only by the fact that the demonstrative is case- and number-marked while the article and the possessor are not, but also by the linear order of the determiners: the demonstrative precedes both the definite article and the (pronominal or zero-marked nominal) possessor and thus can be analysed as falling outside the principal domain of the phrase. In addition to these bits of evidence from cooccurrence, case- and number-marking, and linear order, a fourth indication of the independence of the demonstrative is that, unlike the noun, it does not show the possessor’s person and number marking and thus appears to stand outside the scope of the possessor: (34) a.

ez az én fia-m this the I son-SG1 ‘this son of mine’ *ez-em az én fia-m this-SG1 the I son-SG1 b. ez-ek az én fia-i-m this-PL the I son-PPL-SG1 ‘these sons of mine’ *ez-ei-m az én fia-i-m this-PPL-SG1 the I son-PPL-SG1

✐

✐ ✐

✐

✐ 148

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #172

✐

✐

148

Edith Moravcsik

These four facts all suggest that the demonstrative acts at least in some ways as a separate NP followed by the rest of the phrase in appositive-like manner.9 The redundant expression of number in demonstrative-noun constructions is all the more striking since number marking is otherwise parsimonious in Hungarian. First, numerated nouns take no plural marking (35a). Second, verbs whose subjects are numerated nouns are also in the singular (35b): (35) a.

b.

lány ‘girl’ lány-ok ‘girls’ két lány ‘two girls’ öt lány ‘five girls’ hatszázhúsz lány ‘six hundred and twenty girls’ A hat lány jön. the six girl come:SG1 ‘The six girls are coming.’ cf. A lány-ok jön-nek. the girl-PL come-PL3 ‘The girls are coming.’ A lány jön. the girl come:SG1 ‘The girl is coming.’

Third, paired body part nouns are preferentially used in the singular. Paired body parts seem construable in Hungarian both as two organs and as a single organ: an expression like ‘half of his ear’ is actually ambiguous between half of one ear and one of two ears, with the latter – especially if ‘half ear’ is pronounced with a single word stress – the more likely reading: (36) a.

b.

Levágták a fülét. they:cut:off the ear:SG3:ACC ‘They cut off one of his ears.’ or ‘They cut off both of his ears.’ Levágták a fél fülét. they:cut:off the half ear:SG3:ACC ‘They cut off one of his ears.’ or ‘They cut off half of one of his ears.’

Fourth, generic nouns are frequently in the singular (cf. Moravcsik, this volume): (37)

Szeretem az almát. I:like the apple:ACC ‘I like apples.’

✐

✐ ✐

✐

✐ 149

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #173

✐

✐

Inflectional morphology in the Hungarian noun phrase

149

Fifth, note an interesting case, which can also be seen as a symptom of economy in plural marking (James Hurford, p.c.). The possessor’s plurality is obligatorily marked both on the possessum and on the possessor if it is 1st or 2nd person: (38)

a mi könyv-ünk the PL1 book-PL1 ‘our book’ a ti könyv-etek the PL2 book-PL2 ‘your (PL) book’

If, however, the possessor is 3rd person plural, its plurality is marked only once: in case of a pronominal possessor, it is marked on the possessum only, and in case of a nominal possessor, it is marked on the possessor only, as shown in (39). (39) a.

pronominal possessor: az o˝ kert-jük the SG3 garden-PL3 *az o˝ -k kert-jük the SG3-PL garden-PL3 ‘their garden’ b. nominal possessor: a szomszéd-ok kert-je the neighbor-PL garden-SG3 *a szomszéd-ok kert-jük the neighbor-PL garden-PL3 ‘the neighbors’ garden’

In other words, ‘their garden’ is his their:garden and ‘the neighbors’ garden’ is the neighbors’ his:garden. There is number conflict in both constructions resulting from syncretism but the site of syncretism is different: if the possessor is a 3rd person pronoun, the singular-plural distinction is neutralized on the free pronoun while maintained inflectionally; if, however, the possessor is a noun, number distinction is neutralized in the inflection but maintained on the possessor. In sum: H-33.

EXTENDED EXPONENCE : TYPES Extended exponence occurs in number, case, and person marking.

H-34.

EXTENDED EXPONENCE : NUMBER OF FORMS INVOLVED

There are no more than two exponents per meaning.

✐

✐ ✐

✐

✐ 150

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #174

✐

✐

150

Edith Moravcsik

H-35. a. b.

EXTENDED EXPONENCE : OVER THE PHRASE Extended exponence may occur spread over the NP. The possessor’s person and number for which the possessum is inflected may be additionally marked by a free possessor pronoun. In NPs that include a noun as well as a demonstrative, case and number of the noun must be marked both on the noun and on the demonstrative.

H-36.

EXTENDED EXPONENCE : OPTIONALITY If the components of extended exponence are a free form and a bound form and extended exponence is optional, it is always the free form that is dispensable.

H-37.

EXTENDED EXPONENCE : NON - REDUNDANCY OF PLURAL MARKING Some crosslinguistically recurrent redundancies in plural marking do not in fact occur: Numerated nouns are in the singular. Verbs whose subject is a numerated noun are in the singular. Paired body parts may be used in the singular in reference to both members of the pair. Generic nouns may be in the singular. While for plural 1st and 2nd person possessors, the plurality of the possessor is marked both on the possessor and on the possessum, for the plural 3rd person possessor, plurality is marked only once: either on the possessor itself (if it is a noun), or on the possessum (if the possessor is a pronoun).

a. b. c. d. e.

2.1.8. Non-compositionality If the relationship between meaning and form is many-to-many and thus quantitatively symmetrical, the two may still be mismatched if the meanings of the part forms and those of their relations do not add up to the meaning of the whole form. This pattern evidences non-compositionality – the last of the seven complex meaning-form relations identified in Section 1.1. Meaning-form relations in inflected Hungarian nominals are generally compositional. One possible example of non-compositional inflection is the associative plural if viewed as bimorphemic: the -ék of Bélá-ék ‘Béla and his associate(s)’ analysed as consisting of -é, which derives a possessum out of a possessor, and -k, the basic plural. As already noted in Section 2.1.6, under this analysis the associative plural shows an interesting conflict: it is semantically non-compositional since the meaning of the whole goes beyond the sum of the meanings of the parts, but structurally compositional in that the form properties of the whole complex are predictable from the form properties of the parts.

✐

✐ ✐

✐

✐ 151

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #175

✐

✐

Inflectional morphology in the Hungarian noun phrase

151

Another instance of non-compositionality shows up in case marking. On Hungarian nouns, case may be indicated by affix, adposition, zero marking, or their combinations, of which six occur: (a) (b) (c) (d) (e) (f)

Affix Adposition Affix + Adposition Adposition + Affix Affix + Adposition + Affix Zero marking

(40) illustrates each: (40) a.

b.

c.

d.

e.

f.

Affix: a ház-ban the house-INESS ‘in the house’ Adposition: a ház alatt the house under ‘under the house’ Affix + Adposition: a ház-on túl the house-SUPESS beyond ‘beyond the house’ Adposition + Affix: a ház mögött-r˝ol the house behind-DEL ‘from behind the house’ Affix + Adposition + Affix: a ház-on túl-ról the house-SUPESS beyond-from ‘from beyond the house’ Zero marking: a ház-Ø teteje-Ø the house-GEN its:roof-NOM ‘the house’s roof’

Of these, (a), (b), and (f) involve a single form only and are thus not even candidates for non-compositional expression. (d) involves two case markers – a postposition and an affix – but it is compositional: each case marker contributes part of the total meaning and the two jointly account for the whole meaning. In (c), however, the entire meaning ‘beyond’ is expressed by the postposition: the meaning of the

✐

✐ ✐

✐

✐ 152

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #176

✐

✐

152

Edith Moravcsik

Superessive affix is not part of the total meaning. Thus, (c) is an instance of noncompositionality, and so is (e) – a combination of (c) and (d). There is a total of 18 postpositions (20 according to Marácz 1989: 374) that, like túl ‘beyond’ (see (40c), (40e)), can mark case only in cooccurrence with a case affix on the noun. Here is the list (adapted from Marácz 1989: 374), with examples: (41) a.

b. c.

d.

e.

f.

postpositions governing the Instrumental: együtt ‘together’ Kati-val együtt szembe ‘across to’ Kati-val szembe szembe(n)‘across from’ Kati-val szembe(n) szemközt ‘across from’ Kati-val szemközt postposition governing the Sublative: nézve ‘regarding’ Kati-ra nézve postpositions governing the Allative: képest ‘compared to’ Katihoz képest közel ‘near’ Katihoz közel postpositions governing the Superessive: alul ‘below (metaphoric)’ Kati-n alul át ‘through’, ‘across’ Kati-n át belül ‘inside’ Kati-n belül felül ‘over and above’ Kati-n felül keresztül ‘across, by means Kati-n keresztül of’ kívül ‘outside, besides’ Kati-n kívül túl ‘beyond’ Kati-n túl postpositions governing the Adessive: fogva ‘as a result of’ ennél fogva nélkül ‘without’ Kati nélkül postpositions governing the Ablative: fogva ‘from’ ett˝ol fogva kezdve ‘starting from’ ett˝ol kezdve

‘together with Kati’ ‘to across from Kati’ ‘across from Kati’ ‘across from Kati’ ‘regarding Kati’ ‘compared to Kati’ ‘near Kati’ ‘under Kati’ ‘through Kati’ ‘inside Kati’ ‘over and above Kati’ ‘by means of Kati’ ‘besides Kati’ ‘beyond Kati’ ‘as a result of this’ ‘without Kati’ ‘from this point on’ ‘from this point on’

The six case affixes governed by these postpositions are all spatial except for the Instrumental. The Superessive occurs with the largest number of postpositions (7); the Sublative occurs with one postposition only. While these postpositions cannot mark case by themselves, the case affixes that cooccur with postpositions also function as case affixes on their own. In other words, given a postposition, both the presence and the choice of the cooccurring case affix are predictable; but given a case affix, neither the presence nor the choice of the cooccurring postposition can be predicted. In sum:

✐

✐ ✐

✐

✐ 153

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #177

✐

✐

Inflectional morphology in the Hungarian noun phrase

153

H-38.

NON - COMPOSITIONALITY: EXTENT Inflected words are generally compositional.

H-39.

NON - COMPOSITIONALITY: CASE When a case affix and a postposition are joint exponents of a case meaning, the case affix involved also occurs as an independent case marker but the postposition does not; the postposition selects a unique cooccurring case affix on the noun, but, with one exception, the case affix may cooccur with several different postpositions.

a. b.

H-40.

NON - COMPOSITIONALITY: FORM VERSUS MEANING

Semantic and formal compositionality do not always go hand-in-hand. The associative plural marker, if analysed as bimorphemic, is formally but not semantically compositional: form properties of the whole are the sum of the form properties of the parts, but the meaning of the whole is not the sum of the meanings of the parts.

2.2. Distribution 2.2.1. Cooccurrence In Section 2.1, facts about Hungarian nominal inflection were surveyed from the point of view of meaning, form, and the seven complex ways in which the two can be related to each other. The descriptive framework outlined in the beginning of this paper raises three more issues all related to the distribution of inflections. The first is cooccurrence. This parameter pertains to the choice of the constituents with which inflections occur. What are the types of constraints that govern cooccurrence? Since there are five ways in which components of an inflected word may be incongruous, there can be five types of cooccurrence constraints. Examples of the five kinds of cooccurrence violations are the following: (42) a.

wrong choice of an affix relative to another affix, e.g., the basic plural is used together with the associative plural: *János-ék-ok János-APL-PL b. wrong choice of a stem relative to an affix, e.g., a common unpossessed noun stem used with the associative plural: *kabát-ék coat-APL ‘coat and associates’

✐

✐ ✐

✐

✐ 154

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #178

✐

✐

154

Edith Moravcsik

c.

wrong choice of an affix allomorph relative to the stem, e.g., -ok rather than -ak used for the plural of madár ‘bird’: *madar-ok bird-PL ‘birds’ d. wrong choice of a stem allomorph relative to the affix, e.g., madár- rather than madar- used with the plural affix: *madár-ak bird-PL ‘birds’ e. wrong choice of an affix allomorph relative to another affix, e.g., -ot rather than -at used after a possessor marker: *madar-am-ot bird-SG1-ACC ‘my bird (ACC)’

In what follows, we will explore the specific constraints of each of these five types in Hungarian nominal inflection. Section 2.2.1.1 will discuss morpheme-tomorpheme constraints (cf. (a) and (b) above) and Section 2.2.1.2 will turn to allomorphic choice (cf. (c), (d), and (e) above). 2.2.1.1. Choice of morphemes Two questions arise: First, what affix morphemes can occur with what other affix morphemes? Second, what affix morphemes can occur with what stem morphemes? Let us take them up in turn. C OOCCURRENCE OF AFFIX MORPHEMES The three main affix types: basic and possessive plural, case, and the possessor’s person-number marking on the possessum, can occur in any combination: (43)

PL

or PPL + + + +

PERS - NUMB

+ + + +

CASE

+ + + +

EXAMPLES (ház ‘house’) ház-ak ‘houses’ ház-am ‘my house’ ház-ban ‘in house’ ház-ai-m ‘my houses’ ház-ak-nak ‘to my houses’ ház-am-nak ‘to my house’ ház-ai-m-nak ‘to my houses’

The cooccurrence possibilities are thus the following:    (PL) (CASE)   (( PPL ) PERS - NUMB )

✐

✐ ✐

✐

✐ 155

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #179

✐

✐

Inflectional morphology in the Hungarian noun phrase

155

This formula does not yet include the associative plural. The associative plural is different from the other two plurals in its cooccurrence options: while the basic and possessive plural are in complementary distribution, the associative plural can occur with the possessive plural but not with the basic plural: *a testvér-ek-ék ‘the siblings and associate(s)’ is ungrammatical but a testvér-ei-m-ék ‘my siblings and associate(s)’ is well-formed. This must be for semantic reasons: the associative plural can be formed only from nouns that refer to a unique set of individuals and possessed common nouns are “more unique” than unpossessed ones. As Bartos astutely observes (2000: 698), a further restriction on the use of the associative plural is that, if the stem is possessed, it must be possessed by a singular possessor. Thus, tanár-om-ék teacher-my-APL ‘my teacher and his associates’ is grammatical but *tanár-unk-ék teacher-our-APL ‘our teacher and his associates’ is not! The full range of the associative plural’s cooccurrence possibilities is as follows: (44)

APL

PPL

PERS-

CASE

EXAMPLES

NUMB

+

János-ék

+

+

+

+

+

+

+

+

+

+

+

+

János-ék-nak testvér-em-ék

+

testvér-em-ék-nek testvér-ei-m-ék

+

testvér-ei-m-ék-nek

‘János and associate(s)’ ‘to János and associate(s)’ ‘my sibling and associate(s)’ ‘to my sibling and associate(s)’ ‘my siblings and associate(s)’ ‘to my siblings and associate(s)’

An extended version of the above chart showing the cooccurrence options of all nominal inflections is as follows:    (PL) (CASE)   (( PPL ) PERS - NUMB ) ( APL ) In other words, any affix can cooccur with any other affix except for the following: – the possessive plural occurs if and only if the noun is possessed – i.e., in the company of possessor marking; – there can only be a single exponent of each affix category (i.e., one number, one case, one possessor marking), except that the associative plural may cooccur with the possessive plural.

✐

✐ ✐

✐

✐ 156

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #180

✐

✐

156

Edith Moravcsik

The only pattern that conflicts with the above schema constitutes a deviation from the single-exponent constraint: as discussed in Section 2.1.7, there are multiple occurrences of case markers in the same word, e.g. o˝ -t-et he-ACC-ACC ‘him’ or t˝olem-t˝ol ABL-SG1-ABL ‘from me’. C OOCCURRENCE OF STEM MORPHEMES AND AFFIX MORPHEMES Nominal inflections might reasonably be expected to occur on all nouns and pronouns and only on nouns and pronouns. Hungarian nominal inflections, however, do not fulfill either expectation: on the one hand, there are restrictions on the occurrence of inflections with nominals and, on the other hand, the affixes can occur on members of word classes other than nominals. Let us first consider the limitations on the distribution of nominal inflections over nominal stems. Some of the constraints are semantic in nature. Thus, as noted before, the associative plural can occur only on stems that designate unique humans: mostly proper names, titles, or names of relatives: (45) a.

János-ék János-APL ‘János and associate(s)’ b. Nagy úr-ék Nagy Mr.-APL ‘Mr. Nagy and associate(s)’ c. anyá-m-ék mother-SG1-APL ‘my mother and associate(s)’ d. *az ember-ék the man-APL ‘the man and associate(s)’

Another semantically-based constraint is that the Distributive-Temporal case affix -nta -nte ‘at every (time unit)’ can only occur with temporal nouns. Additional constraints, such as that proper nouns and personal pronouns cannot take the EssiveModal -ul -ül ‘as’ or the Essive-Formal -ként ‘in the role of’, and that animate nouns and pronouns resist the Terminative -ig ‘up to’, are less obviously meaning-based. While these facts show certain limitations on the use of nominal inflections relative to nouns and pronouns, in other ways these inflections have a very broad distribution: they can occur not only on nouns and pronouns but also on stems belonging to any word class whose members can fill nominal slots. (46) shows the occurrence of the basic plural on nouns, personal pronouns, demonstratives, and adjectives. (On the nominal nature of Hungarian adjectives, see Moravcsik 2001.)

✐

✐ ✐

✐

✐ 157

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #181

✐

✐

Inflectional morphology in the Hungarian noun phrase

(46) a.

b.

c.

d.

157

the basic plural on a noun: A fiú-k megérkeztek. the boy-PL arrived ‘The boys arrived.’ the basic plural on a personal pronoun: o˝ -k megérkeztek. he-PL arrived ‘They arrived.’ the basic plural on a demonstrative pronoun: Ez-ek megérkeztek. this-PL arrived ‘These arrived.’ the basic plural on an adjective: A kék-ek megérkeztek. the blue-PL arrived ‘The blue ones arrived.’

(47) illustrates the similarly cross-categorial use of case. (47) a.

b.

affixal case: (i) on a noun: a ház-ban the house-INESS ‘in the house’ (ii) on a numeral: a három-ban the three-INESS ‘in the three’ (iii) on a demonstrative: eb-ben this-INESS ‘in this’ (iv) on an adjective: a feketé-ben the black-INESS ‘in the black one’ affixal-postpositional case: (i) on a noun: a ház-on túl the house-SUPESS beyond ‘beyond the house’

✐

✐ ✐

✐

✐ 158

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #182

✐

✐

158

Edith Moravcsik

(ii)

on a numeral: a hárm-on túl the three-SUPESS beyond ‘beyond the three’ (iii) on a demonstrative: ez-en túl this SUPESS beyond ‘beyond this’ (iv) on an adjective: a feketé-n túl the black-SUPESS beyond ‘beyond the black one’ Finally, (48) shows the cross-categorial applicability of the third inflection type: possessor marking. (48) a.

b.

c.

possessor marker on a noun: ház-unk house-PPL1 ‘our house’ on a numeral: öt-ünk five-PPL1 ‘five of us (out of more than five)’ on an adjective: kék-ünk blue-PPL1 ‘our blue one’

The distribution of possessor markers is even broader than that of the other nominal inflections. In addition to occurring on nouns, affixes (near-)identical to the possessor markers are also used in three other syntactic contexts: to show person and number in case-marked pronouns and to show the subject’s person and number on finite verbs and on infinitives (de Groot 1989: 67–94). (49) first illustrates the use of these markers on possessed nouns and then their use in the three other contexts just mentioned. (49) a.

possessed nouns: kert-em ‘my garden’ kert-ed ‘your garden’ kert-je ‘his garden’

kert-ünk kert-etek kert-jük

‘our garden’ ‘your garden’ ‘their garden’

✐

✐ ✐

✐

✐ 159

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #183

✐

✐

Inflectional morphology in the Hungarian noun phrase

b.

c.

d.

case-marked pronouns: benn-em ‘in me’ benn-ed ‘in you (SG)’ benn-e ‘in him’ infinitives (-ni/-n INF): Men-n-em kell. ‘I have to go.’ Men-n-ed kell. ‘You have to go.’ Men-ni-e kell. ‘He has to go.’ verbs with definite object: Emel-em. ‘I lift it.’ Emel-ed. ‘You lift it.’ Emel-i. ‘He lifts it.’ verbs with indefinite or no object: Emel-ek. ‘I lift.’ Emel-sz. ‘You lift.’ Emel. ‘He lifts.’

benn-ünk benn-etek benn-ük

159

‘in us’ ‘in you (PL)’ ‘in them’

Men-n-ünk kell. ‘We have to go.’ Men-n-etek kell. ‘You have to go.’ Men-ni-ük kell. ‘They have to go.’ Emel-jük. Emel-itek. Emel-ik.

‘We lift it.’ ‘You lift it.’ ‘They lift it.’

Emel-ünk. Emel-tek. Emel-nek.

‘We lift.’ ‘You lift.’ ‘They lift.’

In addition to the fact that they all take person-number affixes, case-inflected pronouns (also referred to as “conjugated postpositions”), infinitives, and verbs further resemble possessed nouns in that they may all be accompanied by a free pronoun reinforcing reference to person and number. As (50) shows for én ‘I’, optional free pronouns stand in the Nominative except with the infinitive, where they must be in the Dative. (50)

Noun: Pronoun: Verb: Infinitive:

az (én) kertem (én)bennem (én) emelem. (Nekem) mennem kell.

‘my book’ ‘in me’ ‘I lift it.’ ‘I have to go.’

While, as just seen, nominal inflections are cross-categorially applicable, their actual exponents may differ with the word class of the stem. Three such differences stand out, one having to do with overt versus zero case marking, the second with the affixal versus stem use of the person-number morpheme, and the third with the occurrence of governed case. Let us consider each in turn. First, as discussed above (Section 2.1.3), nouns may or may not be zero-marked for the Genitive case. Some pronouns, however, differ from nouns in this regard: personal pronouns must be zero-marked for the Genitive and demonstrative pronouns must be overtly-marked for it: (51) a.

Genitive marking on nouns: overt or zero a fiú könyv-e the boy book-SG3

✐

✐ ✐

✐

✐ 160

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #184

✐

✐

160

Edith Moravcsik

a fiú-nak a könyv-e the boy-GEN the book-SG3 ‘the boy’s book’ b. Genitive marking on a personal pronoun: zero only az én könyv-em the I book-SG1 *az én-nek a könyv-em the I-GEN the book-SG1 ? nekem a könyv-em I:GEN the book-SG1 ‘my book’ c. Genitive marking on a demonstrative pronoun: overt only (i) demonstrative as head: ez this ‘this (one)’ en-nek a könyv-e this-GEN the book-SG3 *ez a könyve this the book-SG3 ‘the book of this (one)’ (ii) demonstrative as determiner: ez a fiú this the boy ‘this boy’ en-nek a fiú-nak a könyv-e this-GEN the boy-GEN the book-SG3 *ez a fiú-nak a könyv-e this the boy-GEN the book-SG3 *ez a fiú könyv-e this the boy book-SG3 ‘this boy’s book’ Second, the case marking of personal pronouns differs from that of nouns also in whether case functions as a stem or an affix. The relevant difference between nominal and personal pronominal case marking can be read off from the following schemata: Noun (Pronoun)

+ +

CaseMarker CaseMarker

+

PossessivePronounAffix

For example:

✐

✐ ✐

✐

✐ 161

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #185

✐

✐

Inflectional morphology in the Hungarian noun phrase

(52) a.

b.

161

case marking of nouns (i) by affix: n˝o-vel woman-INST ‘with (the) woman’ (ii) by postposition: n˝o mellett woman next:to ‘next to (the) woman’ case marking of pronouns (i) by affix: (én-)vel-em (I)-INST-SG1 ‘with me’ (ii) by postposition: (˝o-)mellett-e (he)-next:to-SG3 ‘next to him’

Thus, while case-marked nouns show a noun stem followed by a case affix or postposition – woman-with for ‘with woman’ – case-marked pronouns consist of an optional pronoun followed by the case affix or postposition functioning as stem which is marked for agreement with the (optional) pronominal stem: I-with-my or with-my for ‘with me’. In nominal case marking, the noun provides the stem and case is appended as an affix or clitic. In pronouns, the case marker forms the stem with the pronoun appended to it as an affix as well as optionally showing up preposed. The third difference between nominal and pronominal case marking has to do with affixal case government by postpositions. As was seen in Section 2.1.8, some postpositions must occur with a case affix on the noun while others must not do so (Marácz 1989: 363–381). This pattern is, however, relaxed for some postpositions when they mark pronouns. For example, the postposition nélkül ‘without’, which must not take a case affix on nouns, may or may not occur with a case affix on pronouns (53). (53) a.

Noun: kert nélkül garden without *kert-nél nélkül garden-ADESS without ‘without (the) garden’

✐

✐ ✐

✐

✐ 162

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #186

✐

✐

162

Edith Moravcsik

b.

Pronoun: (én)-nélkül-em (I)-without-SG1 or: nál-am nélkül ADESS-SG1 without ‘without me’

Similarly, the postposition kívül ‘outside’, which must occur with a case affix on the noun, as in (54a), may or may not do so with pronouns (54b). (54) a.

b.

kert-en kivül garden-SUPESS outside ‘outside (the) garden’ (én-)rajt-am kivül (I-)SUPESS-SG1 outside or: (én-)kivül-em (I-)outside-SG1 ‘outside me’

In sum: COOCCURRENCE : AFFIX TO AFFIX

H-41. a.

b. c.

COOCCURRENCE : AFFIX TO STEM : CROSS - CATEGORIAL INFLECTION

H-42. a.

b. H-43.

Of the four inflection types of number (basic and possessive), associative plural, possessor’s marking, and case, any type can cooccur with any other type. A possessed noun in the plural must be marked with the possessive plural. Apart from some instances of pronominal case marking, more than one exponent of an affix type may not occur in a word.

With the exception of the associative plural, all three affix types are crosscategorial in distribution: they can coocur not only with nouns but with a member of any word class when it fills a nominal slot, such as pronouns, numerals, and adjectives. Possessor markers also mark the subject of infinitives and verbs. COOCCURRENCE : AFFIX TO STEM : CROSS - CATEGORIAL INFLECTION : VARYING EXPONENCE

Cross-categorially applicable inflections may have cross-categorially differing exponents.

✐

✐ ✐

✐

✐ 163

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #187

✐

✐

Inflectional morphology in the Hungarian noun phrase

163

2.2.1.2. Choice of allomorphs As illustrated in (42) in Section 2.2.1, the proper selection of the formatives contained in an inflected word depends not only on the proper choice of the constituent morphemes but also on the proper choice of the allomorphs of these morphemes. A word may end up allomorphically ill-formed either if the choice of an affix allomorph is wrong relative to the stem, or if it is inappropriate relative to another affix, or if the choice of the stem allomorph is wrong relative to an affix. Correspondingly, we will explore constraints on allomorphic choice by taking up the following three questions: (a) Given an uninflected stem and an affix that has allomorphs, which affix allomorph will the stem select? (b) Given two affixes A and B cooccurring in any order where B has allomorphs, which allomorph of B will A select? (c) Given a nominal affix and a stem that has allomorphs, which stem allomorph will the affix select? T HE CHOICE OF AFFIX ALLOMORPH BY THE STEM Allomorph choice for an affix is least constrained if variant forms of an affix are all in free variation for any stem. This hypothetical option would hold if, say, with the stem toll ‘pen’, the basic plural could be formed with any of the allomorphs of the basic plural yielding not only toll-ak but also *toll-ok, *toll-ek, *toll-ök, and *toll-k as equally possible alternatives. On the other hand, allomorph choice is maximally constrained if every stem can occur with only one affix allomorph. The actual situation falls between these two extremes but closer to the constrained end of the scale. The general rule is that stems select a single allomorph of each affix they cooccur with. Nonetheless – and this is why constrainedness is not maximal – there are some stems which can occur with two alternative allomorphs of an affix. Examples: (55) a.

b.

c.

both front- and back-vowel forms of an affix are possible: férfi-nak / férfi-nek man-DAT / man-DAT ‘to man’ both /o/- and /a/-forms of an affix are possible: arany-ok / arany-ak gold-PL / gold-PL ‘pieces of gold’ both /j/- and no-/j/ forms of an affix are possible: barát-jai-k / barát-ai-k friend-PPL-PL3 / friend-PPL-PL3 ‘their friends’

✐

✐ ✐

✐

✐ 164

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #188

✐

✐

164

Edith Moravcsik

No stem relaxes the single-allomorph-only constraint for all of the affixes that it can occur with, and never more than two allomorphs of an affix can alternate freely on a given stem. As we saw before in Section 2.1.4 (H-26), in some cases, the two alternative forms may split in meaning. Given that most stems require a single allomorph of each affix, the question is how that single allomorph is selected. In most instances, the choice of affix allomorphs is simply a function of the phonological make-up of the stem. For example, the 1st person plural possessor marker has the following allomorphs: -m, -om, -am, -em, -öm. The choice between the vowel-initial variants and -m is fully dependent on whether the stem’s last segment is a consonant or a vowel: (56)

kefé-m brush-SG1 ‘my brush’ kerek-em brush-SG1 ‘my wheel’

Consonant-final stems are then left with a choice of four vowel-initial allomorphs differing in vowel quality. The choice is determined by the cross-categorial rules of vowel harmony. Hungarian vowel harmony is conditioned by the backness versus frontness of the vowels, and, among front vowels, by rounding. Thus, if the last vowel of the stem is back, the affix allomorph also has to contain a back vowel; and if the last vowel of the stem is front, the affix has to contain a front vowel, with the rounding feature matching that of the last stem vowel for some – but not all – stems and just in case both round-vowel and unrounded-vowel allomorphs are available for the affix (but see Note 10). The /i/ vowel’s front-back classification varies by lexeme. (For detailed analyses of Hungarian vowel harmony, see Vago 1980; Ringen 1988; for a recent overview, see Rebrus 2000: 786–803.) (57)

kanal-am spoon-SG1 ‘my spoon’ kerek-em wheel-SG1 ‘my wheel’ ökr-öm ox-SG1 ‘my ox’

However, the phonological shape of the stem all by itself does not always determine the choice among affix allomorphs. Consider the following:

✐

✐ ✐

✐

✐ 165

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #189

✐

✐

Inflectional morphology in the Hungarian noun phrase

(58) a. b.

NOMINATIVE

ACCUSATIVE

határ madár varázs parázs

határ-t madar-at varázs-t parazs-at

165

‘border’ ‘bird’ ‘magic’ ‘ember’

These examples show two allomorphs of the Accusative affix: -t and -at. The choice between them is not dependent on the last segment of the stem: as (58a) shows, stem-final /r/ may be followed either by the -t version (határt ‘border’) or by the -at version (madarat ‘bird’). (58b) illustrates the same for /ž/-final stems. Nor would the last two segments serve to predict the choice: for example, the noun kar ‘arm’, which, like the oblique version of madár, ends in -ar, takes the -t while madaroccurs with -at. Nonetheless, the choice between -t and -at is predictable: the -at version occurs on stems whose base form differs from the oblique form in the vowel of the last CVC segment of the stem being long (Abondolo 1988: 19). Even though, as these examples show, affix allomorph choice is not fully predictable from stem form, stems can be sorted into classes so that rules regarding the choice of affix allomorphs can achieve some generality: the rules apply to more than one stem at the same time. Another way in which these rules could gain in generality would be if they predicted the choice of allomorphs for more than one affix at the same time. The instances where cross-affix applicability of allomorph-choosing rules would appear most likely are those where the allomorphs of the different affixes are phonologically similar. Thus, for example, since the variants of the Accusative affix (-t, -at, -ot, -et, -öt) and those of the basic plural (-k -ak, -ok, -ek, -ök) show the same phonological differences having to do with the presence and quality of an affixinitial vowel, one would not expect two different rules to govern the choice among the variants. This expectation is generally borne out in the case of three of the six allomorphdifferentiating features (cf. Section 2.1.2) that recur across affixes; the remaining three fall substantially short of it. The three allomorph types whose distribution by and large obeys the same rules across affixes are front- versus back-vowel forms, rounded- versus unrounded-vowel forms, and the variant forms of the affix-initial /v/ governed by total assimilation.10 (59) illustrates all three. (59a) and (59b) show that the Accusative and the basic plural affixes co-vary in exhibiting their back-vowel versus front-vowel forms (59a) and their front-rounded versus front-unrounded forms (59b). (59c) illustrates a similar co-variation regarding the first consonant of the Instrumental and Translative-Factive case allomorphs (szelep ‘valve’, kalap ‘hat’, török ‘Turk’): (59) a.

ACCUSATIVE

BASIC

PLURAL

front versus back vowel:

szelep-et kalap-ot

szelep-ek kalap-ok

✐

✐ ✐

✐

✐ 166

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #190

✐

✐

166

Edith Moravcsik

b.

unrounded versus rounded: szelep-et török-öt

c.

/v/-assimilation:

szelep-ek török-ök

INSTRUMENTAL

TRANSLATIVE - FACTIVE

szelep-pel

szelep-pé

The other three affix-allomorph-differentiating features – the choice between /a/ and /o/ (as in basic plural -ok versus -ak), the presence versus absence of an affixinitial vowel (as in basic plural -k versus -ok/-ek/-ök), and the presence versus absence of an affix-initial /j/ (as in the possessive plural: -ai versus -jai) – often obey different rules in different affixes. While in these cases, too, there are some overarching generalities that hold across affixes, exceptions abound. Take first the occurrence of the affix-initial vowel. It occurs in seven inflections: (60) basic plural: Accusative: Superessive: possessor SG1: possessor SG2: possessor PL1: possessor PL2:

INITIAL VOWEL

NO INITIAL VOWEL

-ok -ek -ök -ak -ot -et -öt -at -on -en -ön -om -em -öm -am -od -ed -öd -ad -unk -ünk -otok -etek -ötök

-k -t -n -m -d -nk -tok -tek -tök

The basic regularity that holds for all of these affixes is that after a vowel-final stem, there is no initial vowel in any of these affixes, as shown in (61) on the example of kefe ‘brush’: (61)

basic plural: Accusative: Superessive: possessor SG1: possessor SG2: possessor PL1: possessor PL2:

kefé-k kefé-t kefé-n kefé-m kefé-d kefé-nk kefé-tek

However, the reverse of the rule does not hold: the vowel-initial allomorph is not always the one used after consonants. The regularity holds for the plural and the possessive affix but not for the accusative (ablak ‘window’, lap ‘page’, föld ‘land’, madár ‘bird’, asztal ‘table’, pénz ‘money’):

✐

✐ ✐

✐

✐ 167

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #191

✐

✐

Inflectional morphology in the Hungarian noun phrase

(62)

PATTERNS

167

EXAMPLES

PLURAL POSSESSIVE ACCUSATIVE

PLURAL

POSSESSIVE ACCUSATIVE

has #V + + + +

has #V + + + +

has #V + + + +/−

has #V ablak-ok lap-ok föld-ek madar-ak

has #V ablak-om lap-om föld-em madar-am

+ +

+ +

− −

asztalok pénz-ek

asztal-om pénz-em

has #V ablak-ot lap-ot föld-et madár-t or madar-at asztal-t pénz-t

While the choice between the presence and absence of the affix-initial vowel in the Accusative depends to a large part on the final segment of the stem and is thus not fully random (see Papp 1975: 63–66; Vago 1980: 92; Abondolo 1988: 221–223), the point is that the presence versus absence of the affix-initial vowel is not dictated by the same conditions across affixes. Similar discrepancies hold in the choice of allomorphs differing in /a/ versus /o/ (see Papp 1975: 36, 61–107; especially 92–96, Abondolo 1988: 214–227) and of allomorphs differentiated by the presence and absence of the initial /j/ – a topic to which Papp devotes 54 pages of his monograph (1988: 109–163; see also Abondolo 1988: 214–225) and which is the focus of a separate paper by Kiefer (1985). That the choice of /a/ versus /o/ is not always the same across affixes is exemplified by láb ‘foot’: (63)

Plural: Accusative: SG1: SG2:

láb-ak láb-at láb-am or láb-ad or

(*láb-ok) (*láb-ot) láb-om láb-od

(64) shows the same kind of inter-paradigm inconsistency with respect to the presence and absence of the /j/ in possessor marking for the noun barát ‘friend’: while in general both variants are possible, in the SG3 form only the /j/-full form is grammatical (DNA = does not apply). (64)

SINGULAR POSSESSUM

/j/ SG1 SG2 SG3 PL1 PL2 PL3

no /j/

DNA DNA barát-ja (*barát-a) DNA DNA barát-juk or bará-tuk

PLURAL POSSESSUM

/j/ barát-jaim barát-jaid barát-jai barát-jai-nk barát-jai-tok barát-jai-k

or or or or or or

no /j/ barát-aim barát-aid barát-ai barát-ai-nk barát-ai-tok barát-ai-k

✐

✐ ✐

✐

✐ 168

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #192

✐

✐

168

Edith Moravcsik

As these examples show, some phonologically analogous allomorph sets are subject to the same rules of distribution while others are governed by different rules. However, even if the rules differ across affixes, unidirectional implications are still operative: once the correct allomorph choice has been identified for one affix in the context of a given stem type, there are predictions to make regarding allomorph choice for another affix in the same context. One such dependency holds with respect to the presence versus absence of the affix-initial vowel: if the stem takes a vowel-initial allomorph of the Accusative marker, it will also take vowel-initial allomorphs of other stem-altering affixes; but not vice versa (cf. Papp 1975: 61; Dressler 1985: 10–11). (65) illustrates this: (65)

NOM

ACC

POSS

PL

SUPESS

méz pénz

méz-et pénz-t

méz-em pénz-em

méz-ek pénz-ek

méz-en pénz-en

‘honey’ ‘money’

Furthermore, there is also a subregularity involving the possessive and the basic plural: the initial vowel of the possessive affix predicts the presence of the initial vowel of the plural, but, as shown in (66), not vice versa11 (budapesti ‘being from Budapest’): (66)

NOM

ACC

POSS

PL

budapesti

budapesti-t

budapesti-m

budapesti-ek

A third inter-affix dependency is pointed out in Vago (1980: 11). If a stem cooccurs with an affix that has allomorphs differentiated by the choice of /a/ versus /o/ and a stem selects the /a/-type allomorph, then that same stem will select an affixinitial vowel in the Accusative affix and will not select /j/-initial allomorphs of the affixes that have such: If ACC -at, then both PL *-k and PPL *-j-. In other words, the forms of the word madár ‘bird’ are in an implicational relationship:12 madar-at

>

madar-ak *madár-k

>

madar-ai *madár-jai

In sum, the choice of affix allomorphs by the stem is neither fully free nor fully restricted: there are instances of a given stem putting up with either of two allomorphs of an affix but never with more than two and never for all of its affixes. The meanings of the resulting slightly variant word forms may in some cases acquire differential semantics. In those cases where the choice of allomorphs is restricted, unidirectional implications nonetheless hold from the choice of allomorphs of one affix for a given stem type to allomorphic choice for another affix in the same context. The predictions are not idiosyncratic to individual stems: stems that take the same affix allomorph fall into classes.

✐

✐ ✐

✐

✐ 169

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #193

✐

✐

Inflectional morphology in the Hungarian noun phrase

169

C HOICE OF AFFIX ALLOMORPH BY ANOTHER AFFIX So far, we have considered the choice of affix allomorphs relative to a non-inflected stem. What if a stem already has an inflection on it? Does an inflected stem condition the same allomorph of the next affix to be put on it as a phonologically analogous non-inflected stem would? Throughout, we will consider an already inflected stem to form a stem for possible further inflection. In many cases, the answer is affirmative: a non-inflected stem and an inflected stem that are alike in phonological make-up generally condition the same allomorph choice. For example, állam is homonymous between an uninflected and inflected stem; as the former, it means ‘state’, as the latter, it means ‘my chin’, and both words take some of the same allomorphs: (67)

állam-mal állam-tól

‘with (the) state’ ‘from (the) state’

áll-am-mal áll-am-tól

‘with my chin’ ‘from my chin’

However, this is not always the case: in other instances, the first affix on the stem dictates its own choice of the following affix. Here are two examples. First, if a back-vowel stem is inflected for the possessive and then takes the Accusative, the Accusative will always show the allomorph containing /a/ and not the one containing /o/. (68) a.

b.

bot-om-at / *bot-om-ot cane-SG1-ACC ‘my cane (ACC)’ bot-od-at / *bot-od-ot cane-SG2-ACC ‘your cane (ACC)’

Second, if a back-vowel stem is inflected for the basic plural and then takes the Accusative, the Accusative allomorph will be -at rather than -ot: bot-ok-at caneBPL-ACC ‘canes (ACC)’. That it is not the root portion of the stem which calls for the -at form across the intervening affix is shown by the fact that the root by itself would take the -ot form (see (69a)). Furthermore, that it is not the mere phonological form of the inflected stem that dictates the -at can also be proven: uninflected stems of similar phonological make-up would take -ot (69b): (69) a.

bot-om-at cane-SG1-ACC ‘my cane (ACC)’ but: bot-ot / *bot-at cane-ACC / cane-ACC ‘cane (ACC)’

✐

✐ ✐

✐

✐ 170

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #194

✐

✐

170

Edith Moravcsik

b.

bot-om-at cane-SG1-ACC ‘my cane (ACC)’ but: forum-ot / *forum-at forum-ACC / forum-ACC ‘forum (ACC)’

The general pattern that emerges is that affixes with /a/- and /o/-containing alternants that follow another affix (as it turns out, only the Accusative qualifies) always show the /a/-containing allomorph even if the uninflected stem all by itself would require the /o/-version and even if an uninflected stem (near-)homonymous with the inflected one would require the /o/-version. In these examples, an “earlier” affix conditions the “later one”. This is the direction in which allomorphic conditioning generally applies in Hungarian nominal inflection. There are nonetheless two cases of affix-to-affix conditioning which go in the opposite direction. In both cases, the possessor’s person-number marker is the conditioning affix: (70) a.

the possessor’s marking selects the possessive plural over the basic plural to precede it: asztal-ok table-PL ‘tables’ asztal-om table-SG1 ‘my table’ asztal-ai-m table-PPL-SG1 *asztal-ok-om table-PL-SG1 ‘my tables’ b. the possessor’s marking selects the right allomorph of the infinitival affix (-ni or -n) to precede it (jár-ni ‘to go’):13 SG1: jár-n-om *jár-ni-om SG2: jár-n-od *jár-ni-od SG3: jár-ni-a *jár-n-a PL1: jár-n-unk *jár-ni-unk PL2: jár-n-otok *jár-ni-otok PL3: jár-ni-uk *jár-n-uk

Thus, the generalization regarding the directionality of Hungarian nominal suffix-tosuffix conditioning is that it is always to the right except for the possessor’s person

✐

✐ ✐

✐

✐ 171

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #195

✐

✐

Inflectional morphology in the Hungarian noun phrase

171

and number markings, which can condition to the left. Having discussed the conditioning of affix allomorphy by inflected and uninflected stems, let us now turn to the third and final issue of cooccurrence: the choice of stem allomorphs. C HOICE OF THE STEM ALLOMORPH BY THE AFFIX Given a nominal affix and a stem that has multiple allomorphs, which stem allomorph will be selected by the affix? Just as in the case of allomorphic choice between affixes, there are two conceivable extreme scenarios: the choice of stem allomorphs is entirely unconstrained: any stem allomorph can occur with any one affix; or it is fully constrained so that it is never the case that more than one stem allomorph can alternatively occur with the same affix. As in the case of affixes, the truth is between the two extremes but closer to the constrained end. Examples of alternative stem allomorphs occurring with the same affix are the following:14 (71) a.

b.

c.

sátor ‘tent’ Accusative:

sátor-t sátr-at

or

falu ‘village’ Accusative: falu-t or falv-at daru ‘crane’ Accusative: daru-t or darv-at

Once again, just as with affix allomorphs, semantic specialization occurs in some instances. As noted in Section 2.1.2, most stems have no allomorphs and the overwhelming majority of those that do have only a single variant form in addition to the base form. Whether the base or the oblique form occurs depends on both the affix and the stem (Abondolo 1988: 182–210). There are eleven affixes – the basic plural, the possessive plural, the Accusative, the Sociative, the Distributive-Temporal, and the six overt possessor markers on the singular possessum – that require the non-base allomorph for almost all stems that have one.15 In other words, if a stem has an oblique form, these affixes almost always select it over the base. For example: (72)

NOM

PL

PPL

POSS(SG1) ACC

SOC

alma nyár kapocs ló kehely

almá-k nyar-ak kapcs-ok lov-ak kelyh-ek

almá-i-m nyar-ai-m kapcs-ai-m lov-ai-m kelyh-ei-m

almá-m nyar-am kapcs-om lov-am kelyh-em

almá-stul nyar-astul kapcs-ostul lov-astul kelyh-estül

almá-t nyar-at kapcs-ot lov-at kelyh-et

‘apple’ ‘summer’ ‘clasp’ ‘horse’ ‘chalice’

✐

✐ ✐

✐

✐ 172

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #196

✐

✐

172

Edith Moravcsik

The rest of the affixes require a special stem allomorph for some stem types but not the others. For example, the Superessive requires a non-base shape for nouns such as alma ‘apple’ or kapocs ‘clasp’ but not nouns of the nyár-type (nyár ‘summer’). The Inessive, on the other hand, requires the oblique allomorph in the alma-class (alma ‘apple’) but not in any of the others. (73)

NOM

SUPESS

INESS

alma kapocs ló nyár

almá-n kapcs-on lov-on nyár-on

almá-ban kapocs-ban ló-ban nyár-ban

‘apple’ ‘clasp’ ‘horse’ ‘summer’

Do the affixes that require the oblique form of all stems form a natural class? There is one trait that they all share: they have a large number of allomorphs themselves. The following chart lists all nominal inflections grouped by the number of their overt allomorphs. The bold-faced affixes are the ones that demand the oblique form of all stems that have one. (74)

NUMBER OF VARIANT FORMS

48 7 6 5 4 3 2 1

AFFIXES

TRANS-FAC, INST-COM PL2 DISTR-TEMP, SOC PL, PPL, ACC, SG1, SG2 SUPESS, SG3, PPL3 ALL, PPL1 INESS, ELAT, ILL, DEL, SUBL, ADESS, ABL, DAT/GEN, FORM, ESS-MOD APL, TERM, TEMP, ESS-FORM, CAUSFIN

If we discount the Translative-Factive and Instrumental-Comitative (the two /v/initial suffixes), which head the list but whose allomorphic conditioning is special, we find that the base-altering affixes all tend to bunch up in the top part of the list. There is thus a clear correlation between the allomorphic richness of an inflection and the extent of allomorphic demands it places on its stem: those inflections that are themselves sensitive to the stem also require sensitivity on the part of the stem. The facts about allomorphic choice may be summarized as follows: H-44.

COOCCURRENCE : AFFIX ALLOMORPHS : GRAMMATICAL VERSUS PHONOLOGICAL CONDITIONING

The only grammatically conditioned affix variants are the plural, which differs in shape depending on whether the noun is possessed or not, and the 3rd person singular and plural possessor markers on the possessum, which differ with the possessum being singular or plural.

✐

✐ ✐

✐

✐ 173

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #197

✐

✐

Inflectional morphology in the Hungarian noun phrase

H-45.

COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM No affix allomorph is specific only to a single stem type.

H-46.

COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM :

173

ALTERNATIVES

Generally, every stem selects a single allomorph of an affix; but there are some cases of a stem alternatively occurring with multiple allomorphs of an affix. COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM :

H-47.

MUTUAL DEPENDENCIES AMONG AFFIXES

The choice from among allomorphs that differ in harmonizing vowel quality or the presence/absence of the initial preconsonantal vowel is generally – but not always – determined the same way across all affixes. COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM :

H-48.

UNIDIRECTIONAL DEPENDENCIES AMONG AFFIXES

a.

b.

c.

H-49.

There are dependencies among affixes regarding the choice of their allomorphs. For example: If a stem selects the vowel-initial allomorph of the Accusative affix, it will also select the vowel-initial allomorph of the SG1, SG2, and PL2 possessive affixes in the singular-possessum paradigm; but not vice versa. If a stem selects the vowel-initial allomorph of the SG1, SG2, and PL2 possessive affixes in the singular-possessum paradigm, it will select the vowel-initial allomorph of the basic plural; but not vice versa. If a stem cooccurs with an affix that has allomorphs differentiated by /a/ versus /o/ and the stem selects the /a/-containing allomorph, then that same stem will select a vowel-initial allomorph in the Accusative affix; but not vice versa. COOCCURRENCE : AFFIX ALLOMORPHS : PREDICTING AFFIX ALLOMORPH FROM STEM TYPE

If a stem’s base form ends in CV:C and its oblique form ends in the corresponding CVC, then the stem takes the /a/-containing allomorph of the Accusative affix. H-50.

COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY ANOTHER AFFIX VERSUS CHOICE BY THE UNINFLECTED STEM

Inflected words do not always behave the same way as their homonymous non-inflected counterparts in choosing among allomorphs of an additional affix. After back-vowel stems that already include an inflectional affix, any additional inflection must occur in its /a/-containing form if it has /a/- versus /o/-containing allomorphs.

✐

✐ ✐

✐

✐ 174

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #198

✐

✐

174

Edith Moravcsik

H-51.

COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY ANOTHER AFFIX : DIRECTION OF CONDITIONING

As a rule, affixes condition cooccurring affix allomorphs to the right. The possessor markers are the only affixes that can select affix allomorphs to the left. H-52.

COOCCURRENCE : STEM ALLOMORPHS : GENERALITY No stem allomorph is usable with a single affix only.

H-53.

COOCCURRENCE : STEM ALLOMORPHS : CHOICE BY THE AFFIX : ALTERNATIVES

Generally, every affix selects a single allomorph of a stem; but there are some cases of an affix occurring with alternative stem allomorphs. COOCCURRENCE : STEM ALLOMORPHS : BASE - ALTERING AND NON -

H-54.

BASE - ALTERING AFFIXES

a.

b. H-55.

Eleven affixes require the non-base form of all nouns that have such: the basic and possessive plurals, three cases: the Accusative, the TemporalDistributive, the Sociative, and the six possessor markers on the singular possessum. All other affixes take the base form of some stems and the oblique form of others. COOCCURRENCE : CORRELATION OF AFFIX ALLOMORPHY AND BASE ALTERING CAPABILITY

The affixes that require the oblique form of all allomorphic nouns regardless of their inflectional class are among the ones that have the largest numbers of allomorphs themselves. 2.2.2. Linear order Cooccurrence is only one aspect of distribution. A word may be in compliance with the rules of both morphemic and allomorphic choice and still be ungrammatical for one of two reasons: either the linear order of the components is wrong or the phonological bonding relation that holds among them is incorrect. Correspondingly, we will complete the characterization of the distributional properties of Hungarian nominal inflections by discussing their linear order (in the present section) and their bonding patterns (in the next section). The discussion of these two parameters will bring to a head the exploration of the twelve-parameter agenda set at the beginning of this chapter. The order of any one affix relative to the uninflected stem is governed by a simple generalization in Hungarian: all affixes follow the stem. This is actually part of a larger picture: all of Hungarian inflectional morphology is suffixal and, in fact,

✐

✐ ✐

✐

✐ 175

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #199

✐

✐

Inflectional morphology in the Hungarian noun phrase

175

almost all Hungarian affixes whether inflectional or derivational are suffixed – with the possible exception of the so-called verbal prefixes, which, however, may actually be clitics rather than affixes or they may form compound-like structures with the verb, rather than derived words (cf. Kiefer 1992: 72–74; Kiefer & Ladányi 2000). The order of affixes relative to each other is somewhat more complex. The chart given in Section 2.2.1.1 and repeated below captures not only the cooccurrence options of the affixes but their linear orderings as well:    ( PL ) ( CASE )   (( PPL ) PERS - NUMB ) ( APL ) There are, however, some exceptions. A seemingly deviant ordering of possessor marking and case is manifested in case-inflected pronouns such as (én-)vel-em ‘with me’ (cf. Section 2.1.7). The problem arises under an analysis according to which the case affix – vel ‘with (INST)’ – and the person-number agreement are both suffixed to the optional pronoun stem. Under this analysis, the order of case affix and person-number agreement is deviant: case precedes person-number agreement rather than following it. An attractive feature of this account is that it explains why most – but not all (e.g., hozzá-m ‘to me’ (*hezzé-m)) – case-affix forms that occur in casemarked pronouns have front vowels: since all person pronouns include front vowels, this fact would follow from the normal rules of vowel harmony. However, in other ways this is not an optimal analysis. First, it does not explain those forms that are based on the back-vowel allomorph of the case suffix (such as hozzá-m ‘to me’). Also, it is in conflict with the fact that the prefixed pronoun is optional (cf. (én)hozzám ‘to me’): how can one posit a word with an “absentee stem”? This analysis would run counter to the general pattern stated in H-21, according to which Hungarian inflected words always contain an overt stem. The alternative account takes the case affix to be the stem claiming that, while case is affixal for nouns, in personal pronouns it is realized as a stem. This view eliminates the exceptionality of the order pattern: now the case morpheme that irregularly precedes person-number agreement is not an affix and thus not subject to the affix order rule. It is interesting that, as noted above, some pronominal forms have an alternative, partially regularized case form: nál-am-nál in addition to nál-am and t˝ol-em-t˝ol in addition to t˝ol-em. These forms involve a suffixal restatement of the case marker that, under this analysis, forms the stem. However, there are other pronominal forms where the exceptional order of case before possessor marking cannot be explained away. In some case-marked pronouns, the person-number affix and case may occur in either order (Tompa 1968: 195): (75) a.

elé-m-be in:front-SG1-ILL or

✐

✐ ✐

✐

✐ 176

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #200

✐

✐

176

Edith Moravcsik

b.

elé-be-m in:front-ILL-SG1 ‘in front of me (direction)’ elé-d-be in:front-SG2-ILL or elé-be-d in:front-ILL-SG2 ‘in front of you (SG) (direction)’

Thus, inflectional forms of personal pronouns differ from those of nouns in linear order in that the order “case before person-number marking”, which is never occurrent in nouns, is possible in some pronominal forms (75), and – according to one analysis – necessary in other pronominal forms (as in nek-em DAT-SG1 ‘to me’). Having considered the precedence relations between stem and affix, let us now turn to their proximity relation. Inflectional affixes are always contiguous forming a single complex, except for one particular affix which may intrude. This is the derivational affix -é (mentioned in Sections 2.1.6 and 2.1.8 as a possible component of the associative plural marker; cf. Antal 1961: 51–76; Lotz 1968; Korchmáros 1995). This affix is derivational in that it derives a possessum from a possessor; e.g., ember is ‘man’, and ember-é is ‘that of the man’. The base to which it is added may be a singular or plural noun and it may be possessed or unpossessed. (76) shows the various stem options for the -é affix: (76)

BASE SG

EXAMPLES SG

PL

PL

APL

UNPOSS POSS UNPOSS POSS

+ + + +

a lány-é a lány-om-é a lány-ok-é a lány-ai-m-é + Kati-ék-é

‘that of the girl’ ‘that of my daughter’ ‘that of the girls’ ‘that of my daughters’ ‘that of Kathy and associate(s)’

The -é affix may, to an extent, be used recursively: Kati-é ‘that of Kathy’, Katié-é ‘that of that of Kathy’. More than two instances of the affix in a sequence are unusual but possible. The possessum derived by the affix -é may also be pluralized: a lány-é-i ‘those of the girl’. For pluralization, the possessive plural must be used which then forms a semantic unit with -é for purposes of recursive use: a lány-é-i-é ‘that of those of the girl’, ? a lány-é-é-i ‘those of that of the girl’, ? a lány-é-i-é-i ‘those of those of the girl’. Case endings follow the -é affix. The formula below includes all options of the occurrence of the -é affix (* indicates recursion):

✐

✐ ✐

✐

✐ 177

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #201

✐

✐

Inflectional morphology in the Hungarian noun phrase

 ( PL )

 





(( PPL ) PERS - NUMB ) ( APL )

177

(- É ( PPL ))* ( CASE )

The following generalizations summarize all these facts: H-56.

LINEAR ORDER : PRECEDENCE : INVARIANCE The order of morphemes in inflected words is invariant except for certain pronouns where case and possessor may occur in either order.

H-57.

LINEAR ORDER : PRECEDENCE : STEM AND AFFIX All nominal inflectional affixes follow the stem.

H-58. a. b. H-59.

LINEAR ORDER : PRECEDENCE : AFFIX AND AFFIX Number and possessor marking precede case on noun stems. Number precedes possessor marking but the associative plural follows it. LINEAR ORDER : ADJACENCE Inflectional affixes form a continuous sequence except for the possessumderiving affix, which, along with a subsequent optional plural marker, may occur immediately preceding the case affix.

2.2.3. Bonding On to the last of the twelve issues raised initially: the phonological bonding relation that holds between components of a word. Hungarian NP-internal inflectional affixes differ in how closely bonded they are to the stem: whether their relationship is more “agglutinating” or more “fusional”. Obvious criteria of bondedness relate to the option of the affix also occurring as a free form and allomorphic variation both as shown by and as conditioned by an inflection. As was seen above (Section 2.2.1.2), the characteristics of having and requiring allomorphs tend to go hand in hand. The possessor-marking affixes, the basic and possessive plural, and some of the case endings are, by this joint criterion, more bonded to the stem, while the rest of the cases and the associative plural are less bonded. Postpositions neither have conditioned allomorphs nor require any on the part of the noun they stand with. To this extent, their non-affixal status is fully supported. Nonetheless, the distinction between case affixes and adpositions in Hungarian is somewhat tenuous. This is for two reasons. First, the criterion of occurrence as a free form that pits case inflections against adpositions to begin with is not completely clear-cut. On the one hand, some case affixes may occur as stem forms. (77) shows that a Noun + CaseAffix complex, just as a Noun + Postposition complex, may be paraphrased with either type of case marker preceding the noun as a separate word when combined with a possessive affix.

✐

✐ ✐

✐

✐ 178

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #202

✐

✐

178

Edith Moravcsik

(77) a.

b.

c.

d.

a ház mellett the house next:to ‘next to the house’ mellett-e a ház-nak next:to-SG3 the house-GEN ‘next to the house’ A fiú-ért megyek. the boy-CAUS I:go ‘I am going to pick up the boy.’ Ért-e megyek a fiú-nak. CAUS-SG3 I:go the boy-GEN ‘I am going to pick up the boy.’

While (77d) illustrates the stem-like use of a case affix, there are also instances of the opposite scenario: postpositions may appear as bound elements. While postpositions are not bound to nouns (see (78)), when it comes to marking pronouns for case, both case suffixes and postpositions are bound (see (78b)). (78) a.

b.

a ház mellett the house next:to ‘next to the house’ a ház-zal the house-INST ‘with the house’ (˝o-)mellett-e (he-)next:to-SG3 ‘next to him’ (˝o-)vel-e (he-)INST-SG3 ‘with him’

Secondly, and more strikingly, postpositions may actually exceed case affixes in the degree to which they are bonded to their host: the derivational affix -i ‘belonging to’, ‘coming from’, normally added to bare noun stems only, can also be added to most Noun + Postposition complexes but not to case-inflected nouns (cf. Marácz 1989: 356–357; Laczkó 2000: 629–636). (79a) and (79b) show this option for bare and postpositional nouns and (79c) shows its non-availability for a case-inflected noun. (79) a.

a kert-i út the garden-ADJCT road ‘the garden road’

✐

✐ ✐

✐

✐ 179

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #203

✐

✐

Inflectional morphology in the Hungarian noun phrase

179

b.

a kert mellett the garden next:to ‘next to the garden’ a kert mellett-i út the garden next:to-ADJCT road ‘the road next to the garden’ c. a kert-ben the garden-INESS ‘in the garden’ *a kert-ben-i út the garden-INESS-ADJCT road ‘the road in the garden’ With the free versus bound distinction between case suffixes and postpositions seen so far as less than categorical, we are left with one criterion of differentiating affixal and cliticized case markers: allomorphic interaction between case marker and host. Case affixes tend to have allomorphs and may call for stem allomorphy while adpositions are invariant and cooccur with the basic stem form. However, since not all case affixes have allomorphs, even this criterion cannot do more than provide a sufficient but not necessary criterion for affixhood. Not only is the bonding difference between case affixes and postpositions tenuous: postpositions themselves also differ among themselves in degree of bondedness. Those postpositions that are accompanied by a case suffix on the noun are less rigidly adjacent and less rigidly postposed to the noun than those that govern no affixal case: at least one form considered by Marácz a postposition – végig ‘to the end’ (Marácz 1989: 374) – can also occur preposed. The three sentences in (80) are synonymous. (80) a.

b.

c.

postposition immediately follows the NP: Hajók úsztak a folyó-n végig. boats swam the river-SUPESS to:the:end ‘Boats floated along the river.’ postposition non-immediately follows the NP: A folyó-n hajók úsztak végig. the river-SUPESS boats swam to:the:end postposition precedes the NP: Hajók úsztak végig a folyó-n. boats swam to:the:end the river-SUPESS

If we accept Marácz’ analysis of végig ‘to the end’ as a postposition, this element shows that at least some case-governing postpositions are somewhat adverbial in nature. However, it is exactly because of the more independent, adverbial nature of végig that I did not include it in the list of postpositions in Section 2.1.8. (On the

✐

✐ ✐

✐

✐ 180

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #204

✐

✐

180

Edith Moravcsik

categorial blurring of postpositions vis-à-vis case markers and adverbs, see Marácz 1989: 353–362; Bartos 2000: 702–704; Payne & Chisarik 2000: 195–197.) In sum: H-60.

BONDING

a. b. c.

All affixes occur only as bound form except for some case affixes which may, in possessed form, also occur as free forms. All postpositions occur as bound stems when marking personal pronouns. Postpositions, adverbs, and case affixes are not always clearly distinguishable.

3. Typological assessment As stated in the beginning of this chapter, our goal is to explain – i.e., render necessary or, short of that, render probable – as many as possible of the characteristics of Hungarian nominal inflection. So far the facts have been surveyed according to a twelve-parameter framework and summarized in statements H-1 through H-60. For all we know at this point, any of these facts may be either accidental or necessary or probable or improbable. How can we tell the difference? As noted in Section 1.1, from the perspective of linguistic typology, a structural property is necessary if it occurs either in all languages or in all languages of a particular sort and it is probable if it occurs in the majority of all languages or of languages of a particular sort. If, however, it is specific to a particular language only, it is accidental. The purpose of Section 3 is to consider the status of the Hungarian facts relative to these options. We will next embark on a search for relevant crosslinguistic generalizations.

3.1. Meaning and form 3.1.1. Meanings Statements H-1 and H-2 gave the semantic types of Hungarian nominal inflections. (For easy reference, statements H-1 through H-60 have been culled from Section 2 and are jointly presented in the Appendix.) Viewed from a typological perspective, these facts raise two questions: (a) (b)

Why does Hungarian have NP-internal inflection at all? Why does it have the semantic types that it has?

Nominal inflection is not universal. Thus, in order to explain its occurrence in Hungarian, we would need a typological generalization that features it as its implicatum: “All languages that have X also have nominal inflection.” Since I know of no

✐

✐ ✐

✐

✐ 181

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #205

✐

✐

Inflectional morphology in the Hungarian noun phrase

181

such generalization, there is no other choice but viewing the occurrence of nominal inflection in Hungarian as accidental from a synchronic typological point of view. While NP-internal inflection does not appear to figure as an implicatum in any typological statement, it may be an implicans. There is some indication that while there are languages with both nominal and verbal inflection (e.g., Russian), languages with inflection in neither category (e.g., Vietnamese), as well as languages with verbal but no nominal inflection (e.g., Japanese), the existence of nominal inflection without verbal inflection may not occur. Should the tentative generalization, according to which if a language has nominal inflection, it also has verbal inflection, hold up under systematic crosslinguistic scrutiny, the existence of nominal inflection in Hungarian would render the presence of verbal inflection necessary. Before we turn to the predictability of the presence of individual semantic types of inflection, let us try to explain why Hungarian does NOT have certain types. This question arises because the inflection types of Hungarian do not exhaust the totality of those attested across languages. By content, inflection in general may be one of four types (Nichols 1986: 58–59, 1992: 48–56; cf. also Plank 1995). First, an inflectional affix may spell out information about the inflected constituent itself, such as gender affixes of nouns do. Self-referencing, non-relational markers contrast with inflection that marks a relationship that the inflected constituent bears to something else in the sentence. Such relational markers may in turn be of three types: they may simply register the existence of a relation, or they may go further and code the nature of the particular relation that holds, or they may, by cross-referencing, provide information about the other constituent that participates in the relationship. The three kinds of NP-internal inflection in Hungarian – number, case, and the possessor’s person-number marking on the possessum – exemplify three of these four general types. Nominal number exemplifies the first, self-referencing type; case is an instance of the third, relation-coding type; and person-number marking may in some contexts function as cross-reference. Hungarian has no affixes of the second kind that simply register a relationship. The relative distribution of the four types of inflection across languages has, to my knowledge, not yet been studied and thus the lack of one type in Hungarian has no typological explanation.16 Within the self-referencing type of inflection, there are again several types, of which Hungarian only manifests one: number. Another crosslinguistically common self-referencing inflection, which typically also figures in cross-reference, is gender. The fact that Hungarian does not have gender has been proposed to be linked to its agglutinating nature by Skaliˇcka (1979b: 223). It is nonetheless impossible to claim a strict – or even a statistical – implication connecting the two properties. On the one hand, agglutination cannot be said to imply the lack of gender in the light of Bantu, which shows some symptoms of being agglutinating (note, however, that Skaliˇcka 1979b views Bantu as only partially agglutinating and the Bantu nounclassificational system as distinct from gender). On the other hand, lack of gender

✐

✐ ✐

✐

✐ 182

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #206

✐

✐

182

Edith Moravcsik

does not imply agglutination either since lots of genderless languages are isolating. No other structural characteristics have been proposed as linked to the occurrence of gender whether in absolute or in probabilistic terms: Corbett’s recent comprehensive survey of gender systems points to genetic and areal factors that correlate with gender but proposes no concomitant structural features (Corbett 1991: 1–2). Thus, the lack of gender, just as the lack of relation-registering inflection, remains a typologically random characteristic of Hungarian. Next, let us consider the existing semantic types of nominal inflection in Hungarian to see if they figure in any typological generalizations. First, is there any way to predict the occurrence of nominal number in a language? David Gil proposed six structural features correlating with obligatory nominal plurality marking (Gil 1987, 1991: 38–39). They are: the presence of obligatory definiteness marking, of rigid NP-internal word order, and of hierarchically interpreted stacked adjectives; and the absence of obligatory numeral classifiers, of distributive numerals, and of stacked adnominal numeral constructions. Although there are indeed languages that bear out the clustering of these properties (such as English versus Japanese, as illustrated in Gil 1991), the exact distributional dependency relations among them are yet to be established. Hungarian, for example, which does have obligatory nominal plural marking (except after numerals), does bear out the predicted presence of definiteness markers and of hierarchically interpreted stacked adjectives as well as the predicted absence of classifiers and stacked numerals, but it departs from the proposed typology in that NP-internal order is not quite rigid and there are distributive numerals. While the existence of nominal number thus remains unpredicted by any universal, two relevant areal generalizations about European languages (EU for “Euroversal” henceforth) have been proposed by Décsy (1973: 210; see also Haarmann 1976: 110) and Corbett (1992: 34), respectively. EU-1.

All European languages make a singular-plural distinction.

EU-2.

No European language has number words.

The presence of inflectional number in Hungarian is derivable from the joint content of these two statements. Let us now consider the predictability of the various semantic subtypes of number inflection. The fact that Hungarian has no trial follows from a well-known hypothesis of Greenberg’s (Greenberg 1963: No. 34): U-1.

No language has a trial number unless it has a dual. No language has a dual unless it has a plural.

U-1 would, however, allow for the occurrence of a dual number.

✐

✐ ✐

✐

✐ 183

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #207

✐

✐

Inflectional morphology in the Hungarian noun phrase

183

Within the plural, we found two semantic subtypes: the basic plural and the associative plural. That the basic plural is frequent across languages is widely recognized. The associative plural has received less attention in crosslinguistic discussions. Nonetheless, this construction is also widespread: recent studies (Corbett & Mithun 1994; Gil 1994b; Moravcsik 1994; Corbett 2000: 101–111; Danièl’ 2000) offer documentation and analysis of it in over 80 genetically and areally diverse languages. The following implicational statement stating the primacy of the basic plural over the associative plural is consistent with known evidence: U-2.

If a language has a grammaticalized expression of the associative plural, it also has a grammaticalized expression of the basic plural utilizing either the same form as the associative plural or a distinct one.

This generalization renders the presence of the basic plural in Hungarian a necessary rather than accidental, fact, while the occurrence of the associative plural remains unexplained. The second of the three semantic types of inflection in Hungarian NPs is case. An oft-cited Greenbergian universal predicts the occurrence of case inflection statistically (Greenberg 1963: No. 41): U-3.

If in a language the verb follows both the nominal subject and nominal object as the dominant order, the language almost always has a case system.

It is not clear, however, whether this generalization applies to Hungarian. Although verb-final order is common in this language, it is not a dominant one in terms of pragmatic neutrality and probably not in terms of frequency, either. (On the principles of Hungarian sentence-level order, see Horvath 1986; É. Kiss 1987, 1994.) Hungarian nevertheless does exhibit a number of characteristics that are often present in verb-final languages, such as having suffixes, postpositions, and mostly preposed modifiers, quantifiers, and determiners. Thus, whether this statement does or does not serve to explain the presence of case in Hungarian depends on how literally we interpret its implicans. As stated in H-2, Hungarian has 17–27 affixal cases. Is this a large case inventory as languages go? And, if so, why does Hungarian have so many cases? The size of the case paradigm in Hungarian is indeed extensive in comparison with other languages even if we just count the 22 affixal cases assumed in this chapter. Plank’s crosslinguistic survey of inflectional case paradigms (Plank 1986) lists Hungarian as having an “extra large” inventory. As Plank shows, the distribution of smaller versus larger case paradigms across languages is not random: it correlates with whether a language observes a one-meaning-per-form “separatist” inflection pattern or whether it is “cumulating”; and whether inflections are sensitive in that they show stemconditioned allomorphy or whether their forms are insensitive to stem choice. In particular, inflectional systems that are either cumulative or separatist-but-sensitive

✐

✐ ✐

✐

✐ 184

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #208

✐

✐

184

Edith Moravcsik

are restricted to smaller case paradigms while case paradigm size in insensitive separatist languages may range from small to large. Plank explains this tendency by economy: if a language with cumulative inflection were to adopt a large number of case distinctions, it would need a very large number of distinct affixes, while in separatist languages, where, because of affix combinability, a small set of affixes goes a long way, large paradigms can be afforded. The following implication emerges: U-4.

Languages with large inflectional case paradigms have inflections that are separatist (i.e., each conveys one meaning component) and insensitive (i.e., have no stem-conditioned allomorphs).

What this means is that the large number of case distinctions in Hungarian is not itself predictable but it in turn serves to make a prediction with respect to the meanings of the case affixes as well a prediction as to their forms. Each of these two predictions is borne out in Hungarian in part but not fully. The first makes for a closer fit with the facts: most Hungarian nominal affixes can indeed be viewed as involving a single meaning component, although, for example, members of the spatial triplets involve two features: position and orientation (cf. Section 2.1.6). The second prediction fares worse: even apart from vowel harmony, which may perhaps be disregarded since it is not specific to case affixes, some of the case inflections – the Accusative (-t -ot -at -et -öt) as well as the Superessive, the Instrumental-Comitative, the Distributive-Temporal, and the Sociative (cf. Section 2.1.2) – do have stem-sensitive allomorphs. Although we have not found a way to predict the existence and number of inflectional cases in general, the existence of some case inflections is predictable from others. According to Lehmann 1985, the zero versus affixal versus adpositional expression of a case relation is predictable by the general principle that says that more concrete cases fall towards the adpositional end of the scale and less concrete, more grammaticalized cases towards the zero end. A large number of specific implicational statements follow from the theory, the relevant subset of which follows (in the chart below I replaced Lehmann’s “Locative” with “Inessive” on the assumption that they are equivalent terms). U-5. a. b.

c.

Given the following sets of cases:   Adessive Allative Benefactive   Perlative Accusative Dative   Benefactive  Genitive Ablative   Dative

✐

✐ ✐

✐

✐ 185

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #209

✐

✐

Inflectional morphology in the Hungarian noun phrase

d. e. f.

 Genitive

185



Dative Allative Ablative Inessive Comitative Instrumental, the affixal marking of a case on the left implies the affixal marking of the case(s) on the right in the same language.

The eleven cases that figure in these predictions include nine affixal cases of Hungarian as well as two of its adpositional cases: the Perlative (át, keresztül ‘through’) and the Benefactive (számára ‘for’). Affixal or zero form is predicted for six of the nine affixal cases of Hungarian. Below, the left column lists the nine cases; the middle column indicates whether the affixal form of a given case is or is not an implicans by predicting or not predicting the affixal form of some other case in terms of Lehmann’s generalizations; the right column shows whether the affixal form of a given case is an implicatum – i.e., whether it is or is not predictable from the affixal form of some other case. For example, the Accusative’s affixal marking is listed as not predictive of the affixal marking of any other case; but its affixal nature is indicated as predictable (from that of the Perlative or the Dative).

Accusative Inessive Instrumental Allative Dative Genitive Ablative Adessive Comitative

PREDICTIVE OF AN

ITS AFFIXAL FORM

AFFIXAL CASE FORM ?

PREDICTABLE ?

no no no yes yes yes yes yes yes

yes yes yes yes yes yes no no no

Thus, the bound form of the Accusative, Inessive, Instrumental, Allative, Dative, and Genitive in Hungarian follows from a crosslinguistic pattern. Since the Perlative and the Benefactive, figuring as implicantia in (a), (b), and (c), are adpositions in Hungarian, they do not participate in predicting bound marking for the cases to their right. They do serve, however, as predictors of adpositional marking for cases more “concrete” than they are themselves. In the spirit of Lehmann’s theory, we might hypothesize the following: U-6.

If a language has adpositional marking for the Benefactive case, it also has adpositional marking for any case more concrete than the Benefactive.

U-7.

If a language has adpositional marking for the Perlative case, it also has adpositional marking for any case more concrete than the Perlative.

✐

✐ ✐

✐

✐ 186

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #210

✐

✐

186

Edith Moravcsik

U-6 correctly predicts that the case relations el˝ott ‘in front of’ or mögött ‘behind’, both undoubtedly more concrete than the Benefactive, are adpositionally expressed. On the other hand, other adpositions – such as Causal miatt ‘because of’ – are not within the scope of U-6 or U-7 since the Causal does not appear more concrete in meaning than either the Benefactive or the Perlative. Another fact that remains unexplained is why some equally concrete case markers vary in having free versus bound form. Why is it, for example, that the ‘on’ relation is expressed with an affix but ‘under’, ‘in front of’ and ‘behind’ are postpositions? Such questions must be left open here. Bernard Comrie suggests (p.c.) that the key notion may be markedness: it is the less marked cases that are affixally marked and the more marked ones are adpositional. Let us finally turn to the third semantic type of Hungarian nominal inflection: the possessor’s number and person on the possessum. De Groot (1989: 69–94) proposes a typology of person-number cross-reference systems along several dimensions. One of the claims put forth and tested by de Groot & Limburg (1986; cf. also Radics c. 1980) has to do with the stem categories that exhibit such agreement. The generalization is that if a language shows person-number marking on adpositions, it also shows the same marking on nouns; if it shows person-number marking on nouns, it also shows such marking on verbs. Lehmann (1985: 97) proposes a somewhat different version, according to which person-number marking on adpositions implies either person-number marking on nouns, or object or ergative marking on verbs, both of which imply the existence of verb-subject agreement. Celtic, as pointed out by Carstairs-McCarthy (p.c.), is a counterexample to either claim since, in spite of the occurrence of inflected adpositions, there is neither possessor marking on the noun nor object or ergative marking on the verb. What remains uncountered is the following generalization: U-8.

If in a language adpositions are person-number-marked, so are at least some verbs of the language.

U-8 does not serve to predict person-number marking in Hungarian for any of the three categories that have it: adpositions, nominals, and verbs, but it correctly predicts verb agreement on the basis of person-number marking on adpositions.17 Given the complex, “displaced” nature of cross-referential inflection – such inflection says something about another constituent rather than itself – it would seem plausible that the presence of cross-referencing inflection in a language should imply the presence of some self-referencing expression of the same categories. According to this hypothesis, if verb inflection cross-references person and number, some nouns and pronouns of the language should wear their own person and number on their sleeves. If this hypothesis were to hold, the occurrence of number inflection – and, indeed, the very existence of personal pronouns – would be rendered predictable in Hungarian.

✐

✐ ✐

✐

✐ 187

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #211

✐

✐

Inflectional morphology in the Hungarian noun phrase

187

In sum: while many facts about the meanings of Hungarian nominal inflections remain unexplained, the following four facts about the existence and subtypes of nominal inflection in Hungarian are derivable from crosslinguistic generalizations: – the existence of number (cf. EU-1, EU-2); – the existence of basic plural (cf. U-2); – the existence of affixal case (?) (cf. U-3); – the non-adpositional expression of six cases (cf. U-5). 3.1.2. Forms As discussed in Section 2.1.2 and summarized in H-3, Hungarian nominal inflections are short, mostly consisting of one syllable. Dressler (1985: 17) proposes a connection between affix length and the general morphological type of the language. U-9.

Affixes in agglutinating languages tend to be shorter than in flectional languages.

If agglutinating languages are generally “separatist” in that affixes convey single meaning components, while flectional languages tend to cumulate, then the pattern stated in U-9 turns out to be iconic: the greater length of inflections in flectional languages corresponds to their richer semantics. Some correlation between semantic complexity and phonological length is also evidenced within Hungarian itself. For example, the associative plural marker’s form is somewhat heavier than that of the basic plural marker and the possessor’s person-number markings tend to consist of more sound segments in the plural than in the singular. A possible corollary of U-9 would be that a semantically complex inflection never has a simpler phonological form than a semantically simple one in the same language. The importance of this tentative generalization, if true, would lie in the fact that such an iconic correlation certainly does not hold for stem morphemes: it is not generally true that the word for say ‘animal’ is shorter than the words for its hyponyms ‘bear’ or ‘fly’. A number of observations were made about allomorphic form (H-5 through H-19). Affix allomorphy was seen to be much more widespread and involving more variants than stem allomorphs and vocalic differentiation among allomorphs of both affixes and stems was shown to be dominant over consonantal differences. Allomorphic form was seen to be not entirely random: given one allomorph, the existence and type of another is often predictable. The typological status of many of these observations remains unclear. For example, is the predominantly vocalic differentiation of both stem and affix allomorphs (H-7, H-8, H-9, H-14, H-15) a peculiarity of Hungarian? Alternatively, it may plausibly hold for vowel harmony languages in general as a result of “system congruity”

✐

✐ ✐

✐

✐ 188

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #212

✐

✐

188

Edith Moravcsik

(Wurzel 1989: 74 ff.).18 Or, it may be a general tendency across all languages resulting from a greater functional load placed upon consonants than vowels (Gregory Iverson, p.c.). One aspect of allomorphy that is derivable from a crosslinguistic generalization is the number of allomorphs. More affixes were seen to have allomorphs than stems (H-5, H-12) and the number of allomorphs per morpheme was also seen to be larger for affixes: most stems do have only two variants (H-13) while affix allomorphs range in number from 2 to 7, with two affixes having as many as 48 (H-6). This pattern is in accordance with the following claim of markedness theory (Greenberg 1966; Moravcsik & Wirth 1986): U-10.

More frequent and syntagmatically simple forms show more paradigmatic complexity; less frequent and syntagmatically complex forms show less paradigmatic complexity.

Given that affixes are generally simpler in meaning than stems and assuming that any one affix is more frequent than any one stem, this generalization explains why affixes vary in shape more than stems do. It also explains why, among affixes, it is the associative plural (-ék), and some of the more esoteric case affixes such as the Essive -ként that lack allomorphs while ubiquitous affixes such as the Accusative or the Superessive show rich allomorphy. 3.1.3. Zero meanings, zero forms Let us turn to the facts noted about zero meanings and zero forms in Hungarian noun inflection (Section 2.1.3). Our first observation (H-20) was that there are no affixes or stems in the Hungarian nominal inflection system that would not make some semantic contribution. I am not aware of any crosslinguistic generalization that would be predictive of this fact. The second observation – H-21 – was that, while there are zero-form inflections, there are no zero-form stems (cf. Section 2.2.2 for the one possible exception). An instance of zero stem would be a word that contains only affixes. Since a crosslinguistic asymmetry favoring the use of zero marking for affixes over stems is both plausible and consistent with evidence from well-known languages, the following universal hypothesis may be tentatively proposed from which the Hungarian case would follow: U-11.

Zero marking within an inflected word is an option for inflections but not for stems.19

Of the five inflectional meanings that were seen to receive zero expression – singular, Nominative, Accusative, Genitive, and the singular 3rd person possessor on plural possessum (cf. H-22) – the possiblity (but not the necessity) of non-overt

✐

✐ ✐

✐

✐ 189

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #213

✐

✐

Inflectional morphology in the Hungarian noun phrase

189

marking of the Nominative is predicted by a Greenbergian implication (Greenberg 1963: No. 38): U-12.

Where there is a case system, the only case which ever has only zero allomorphs is the one which includes among its meanings that of the subject of the intransitive verb.

The necessity of zero marking in the Nominative, along with the possible zero marking of the Accusative, follows from Lehmann’s hypothesis (1985; cf. also Blake 1994: 91) discussed in Section 3.1.1. The relevant portion of the theory may be stated as follows: U-13.

Given the following scale: Nominative < Accusative < Genitive < other cases, if a language has zero marking for a case, it also has zero marking for all the cases to the left.

Thus, from among the three zero-marked cases in Hungarian, it is only the zero Genitive that remains unexplained. The Hungarian evidence would actually be consistent with a stricter version of U-13 as well, according to which not only the occurrence but also the obligatoriness versus optionality of zero marking obeys the hierarchy, with obligatoriness either stagnating or increasing – but never decreasing – to the left. What about the zero marking of the singular noun? The following Greenberg universal is relevant (1963: No. 35): U-14.

There is no language in which the plural does not have some nonzero allomorphs, whereas there are languages in which the singular is expresssed only by zero.

This generalization does not serve to predict zero marking for the singular in a language: it only licenses it. However, it does predict the non-zero, overt marking of the plural. The zero marking of the 3rd person singular possessor on the plural possessum is more problematic. Recall the facts: (81) a.

b.

a kutyá-nak a nyakörv-Ø-e the dog-GEN the collar-SG-SG3 ‘the dog’s collar’ a kutyá-nak a nyakörv-ei-Ø the dog-GEN the collar-PPL-SG3 ‘the dog’s collars’

What is of interest in (81) is the final suffix of the possessum. It is shown that, while the singular possessum is marked for the 3rd person singular possessor with the affix

✐

✐ ✐

✐

✐ 190

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #214

✐

✐

190

Edith Moravcsik

-a/-e/-ja/-je (81a), the plural possessum is zero-marked for the person and number of the possessor (81b). First of all, there is no reason why the singular 3rd person possessor should be zero-marked only on the plural possessum but not on the singular possessum. Second, even the fact that zero marks a 3rd person possessor, rather than 1st or 2nd person possessor, is underivable from any crosslinguistic generalization: a plausible guess according to which if any person is zero-marked it must be the 3rd person fails in its absolute form in the light of English present-tense verb inflection – a crosslinguistically very rare pattern. Thus, the most that can be said is that the Hungarian facts here conform to a pattern that is dominant within the language (cf. Wurzel 1989: 74–111): in Hungarian verbal inflection, as in most languages, it is only the singular 3rd person that is ever zero-marked. However, since it cannot be claimed that “system congruity” always prevails, no complete explanation emerges regarding the zero marking of the 3rd singular possessor the plural possessum. In sum: zero marking of the Nominative and of the Accusative and the non-zero marking of the plural follow from crosslinguistic generalizations. The zero option in Genitive marking, the zero marking of the singular of the noun and that of the singular 3rd person possessor on the plural possessum remain unexplained. 3.1.4. Synonymy Statements H-23 through H-26 in Section 2.1.4 highlighted a number of facts about synonymy in Hungarian nominal inflection. It was stated that synonymy relations can be found both on the morphemic and word level. A variety of ways were seen for inflected words to end up having the same meaning and a tendency was noted for the semantic differentiation of synonymous affix forms and synonymous stem forms. From among these patterns, one that has clear crosslinguistic parallels is the tendency for allomorphs to turn into separate morphemes of distinct meanings. Examples of this tendency were found in Hungarian both for affixes and for stems. Thus, as was shown in Section 2.1.4 (cf. H-26), the /j/-ful and /j/-less variants of the 3rd person singular possessor marker are often synonymous but sometimes they express alienable versus inalienable possession (e.g. ablak-ja ‘its (alienable) window’, ablak-a ‘its (inalienable) window’). Similarly, stem variants of the hó/hav‘snow’ type are generally synonymous but in some cases there is a semantic split between them (e.g. daru-t Accusative of ‘crane (machine)’, darv-at Accusative of ‘crane (bird)’). These examples are similar to the semantic split in German Wört-er and Wort-e (closely paralleled by Hungarian szó-k ‘(unconnected) words’ and szavak ‘(connected or unconnected) words’) or English mice and (computer) mouses. The following typological statement of the existential, rather than implicational, sort does not predict, but at least allows for, the Hungarian facts indicating that they are not entirely unique to Hungarian:

✐

✐ ✐

✐

✐ 191

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #215

✐

✐

Inflectional morphology in the Hungarian noun phrase

U-15.

191

Forms that are allomorphic in some contexts but semantically contrastive in others do occur in languages.

3.1.5. Homonymy Homonymy in Hungarian nominal inflection was discussed in Section 2.1.5 and summarized in H-27 through H-29. We noted that forms may be homonymous between affix meanings or between uninflected stem meanings, or between an affix meaning and an uninflected stem meaning, or between an inflected and an uninflected stem meaning, or between two inflected stem meanings. Homonymy may furthermore involve only a single allomorph of a morpheme or the entire range of allomorphs. We found that the zero marker showed most homonymy. In what follows, we will consider the seven instances of homonymy among Hungarian nominal affixes and consider how well they conform to the various crosslinguistic constraints that have been proposed in the literature regarding homonymy. Let us first survey these constraints. What are the variables involved in homonymy that may possibly be subject to constraints? In order to be able to identify them, I will follow Carstairs-McCarthy’s helpful presentational technique by starting off with a well-known case in Latin and taking stock of the various ways in which it could be different from what it is. The Latin case involves the homonymy of the plural Dative and Ablative illustrated here on the third-declension noun uxor ‘wife’: (82)

Latin: Dative Ablative

SINGULAR

PLURAL

uxor-i uxor-e

uxor-ibus uxor-ibus

First, note the context in which the Dative-Ablative distinction is neutralized: neutralization happens in the plural. The meaning element ‘plural’, however, has no linearly separable expression: it forms a single cumulative expression with case. This is not a necessary fact: it is conceivable for the context to be expressed separately, as it would be in the following Pseudo-Latin forms, where -t is an imaginary plural affix: (83)

Pseudo-Latin (A): Dative Ablative

SINGULAR

PLURAL

uxor-i uxor-e

uxor-t-ibus uxor-t-ibus

Thus, one parameter along which homonymy patterns may vary is whether the context is co-lexicalized with the neutralized meanings or separately expressed. Second, notice that the homonymous affix -ibus is distinct from both of the two unambiguous case exponents in the singular: -i and -e. An alternative would be for

✐

✐ ✐

✐

✐ 192

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #216

✐

✐

192

Edith Moravcsik

the homonymous form also to function as the singular expression of one of the two cases whose distinction is suspended in the plural: (84)

Pseudo-Latin (B) (-l imaginary ‘singular’): Dative Ablative

SINGULAR

PLURAL

uxor-l-ibus uxor-l-e

uxor-ibus uxor-ibus

These considerations so far yield two parameters of variation for inflectional homonymy: (a) (b)

Is the context in which the distinction between the two meanings is neutralized separately expressed or does it cumulate with the neutralized meanings? Is the homonymous affix that expresses both A and B unique or is it “borrowed” from another context where it expresses only A but not B?

While these two variables have to do with formal properties of the target and of the context of homonymy, there are also semantic aspects of the Latin example which could be otherwise. The cases – Dative and Ablative – whose distinction is suspended are relatively marked and so is the context where homonymy arises: the plural. Alternatives would be for the Ablative to cumulate with say the Nominative as in (85a) or for homonymy to arise in the singular, as in (85b): (85) a.

b.

Pseudo-Latin (C): (SINGULAR) Nominative (uxor) Ablative (uxor-e) Pseudo-Latin (D): SINGULAR

Dative Ablative

uxor-i uxor-i

PLURAL

uxor-es uxor-es (PLURAL) (uxor-ibus) (uxor-ip)

Furthermore, in the real-Latin example the two meanings – Dative and Ablative – are obligatorily merged: there is no way to express them unambiguously in the plural. But it is conceivable for homonymy to be optional: one or the other meaning (or both of them) might also have an additional unambiguous expression. For example: (86)

Pseudo-Latin (E): (SINGULAR) Dative: (uxor-i) Ablative (uxor-e)

PLURAL

uxor-ibus uxor-ibus or uxor-ip

These considerations yield three additional parameters of homonymy: (c) (d)

What are the types of meanings that may serve as context for homonymy? What are the kinds of meanings whose distinction may be suspended?

✐

✐ ✐

✐

✐ 193

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #217

✐

✐

Inflectional morphology in the Hungarian noun phrase

(e)

193

Is homonymy obligatory in that the neutralized meanings do not have unambiguous expression, or is it optional in that one or both of the meanings also have an unambiguous exponent?

All the constraints on homonymy that have been proposed in the literature seem to pertain to one or more of these five issues. Here is a sample of them. C ONTEXT SEPARATELY OR CUMULATIVELY EXPRESSED Based on crosslinguistic evidence, Carstairs (1987: 114–116) has noted that in most instances, the meaning that forms the context for homonymy is included in the meaning of the homonymous form rather than having a separate exponent. U-16.

In most cases of inflectional homonymy, the context in which a meaning contrast is suspended is simultaneously expressed together with the neutralized meaning elements of the homonymous affix rather than as a separate affix.

Carstairs-McCarthy labels this type of homonymy syncretism. Of the examples above, the actual Latin case in (82) is an instance of it, while Pseudo-Latin (A) in (83) is not. H OMONYMOUS FORM UNIQUE OR “ BORROWED ” (= “ TAKEN OVER ”) This parameter is logically independent from the choice between the context being simultaneously or sequentially expressed: all four combinations of the two alternatives along each dimension are logically possible: (87)

CHOICES

EXAMPLE

I

context cumulates +

form borrowed −

II

−

+

III

+

+

IV

−

−

SINGULAR

PLURAL

DAT: uxor-i ABL: uxor-e DAT: uxor-ibus ABL: uxor-e DAT: uxor-l-ibus ABL: uxor-l-e DAT: uxor-i ABL: uxor-e

uxor-ibus uxor-ibus uxor-t-ibus uxor-t-ibus uxor-ibus uxor-ibus uxor-t-ibus uxor-t-ibus

Of the four patterns, (I), the actual situation in Latin, exemplifies syncretism, which U-16 claimes to be the most frequent type. Do the other three types occur? Carstairs-McCarthy has two relevant points to make. First, he proposes that (IV) is either very rare or non-existent at all. The scarcity of (IV) may be stated as follows (cf. Carstairs 1987: 123; Carstairs-McCarthy & Stemberger 1988):

✐

✐ ✐

✐

✐ 194

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #218

✐

✐

194 U-17.

Edith Moravcsik

Inflectional homonymies which are neither syncretisms nor take-overs are rare or non-existent.

This leaves (II) and (III). Carstairs-McCarthy finds that, of these two, (II), evidencing “take-over”, is second to syncretism in frequency: U-18.

In those cases of inflectional homonymy where the context in which a meaning contrast is suspended is separately, rather than simultaneously, expressed, and which are therefore not instances of syncretism, the homonymous form is generally a “take-over” in that it also serves as an unambiguous expression of one of the two meanings in another context.

Whether the remaining pattern (III) – one that unites the properties of syncretism and take-over – occurs at all is an open question (Carstairs 1987: 118–119). T YPES OF MEANINGS FORMING THE CONTEXT FOR HOMONYMY U-18 posits two properties of take-overs: the context of homonymy is separately, rather than cumulatively, expressed, and the homonymous form also functions as an unambiguous expression of one of the two neutralized meanings in another context. Carstairs-McCarthy’s characterization of take-overs includes one more component. He proposes that the (separately expressed) meaning that supplies the context for a take-over is semantically less relevant to the stem than the meanings whose contrast is suspended (Carstairs 1987: 123): U-19.

In take-overs, the context in which the meaning distinction is suspended is semantically more marginal relative to the stem’s meaning than the neutralized meaning is.

A real instance of take-over will illustrate the point. In Finnish, the distinction between singular and plural nouns is suspended in the Comitative and Instructive cases with a single form expressing each of these cases in both numbers (Carstairs 1987: 119–120). Since, according to Bybee’s Relevance Hierarchy (Bybee 1985a: 13–16, 1985b), case – the context in this instance of homonymy – is semantically less relevant to the stem than the target of homonymy: number, the Finnish example bears out the proposal. If in turn the case distinction were suspended in the context of separately spelled-out number, as in Pseudo-Latin (A)’s DAT/ABL uxor-t-ibus, this would go against the proposed generalization. While U-19 places a constraint on the relationship between the neutralized meaning and the context meaning, another proposal stated by Plank (1991a: 22–23) proposes a tendency regarding the contexts where homonymy may arise: U-20.

Marked semantic contexts are more conducive to homonymy than unmarked ones.

✐

✐ ✐

✐

✐ 195

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #219

✐

✐

Inflectional morphology in the Hungarian noun phrase

195

For example, the contrast of Dative and Ablative is more likely to be suspended in the plural (as in the Latin example in (82)) than in the singular (as in Pseudo-Latin (D) in (85b)). This once again follows from the general theory or markedness (cf. U-10 above). The set of contexts in which homonymy may arise is also constrained from the point of view of disambiguability. Plank (1980) proposed that homonymies must be disambiguable by context as follows: U-21.

Homonymy is tolerable if it is disambiguated by the lexical cooccurrence constraints of a co-constituent. For example, subject-object ambiguities are tolerable since the selectional demands of the verb will in many cases give away what is subject and what is object; but genitive-head ambiguities are not tolerable since there often is no comparable disambiguator available.

T YPES OF MEANINGS NEUTRALIZED IN HOMONYMY Markedness theory further generates a prediction regarding the meanings themselves whose contrast is suspended in homonymy: U-22.

Distinctions between marked semantic categories are suspended preferentially over distinctions between unmarked categories.

Thus, the distinction between plural and dual is more likely to be suspended than that between singular and plural. This again follows from the general pattern of marked categories showing no more subdistinctions than marked ones. A further proposal limits the relationship between meanings whose difference may be neutralized (cf. Plank 1991a: 19): U-23.

The meanings whose contrast is suspended by a homonymous form tend to be similar.

For example, two semantically related cases – e.g., spatial and temporal location – are more likely to be united under a single form than, say, spatial location and the Accusative. O BLIGATORINESS VERSUS OPTIONALITY OF HOMONYMOUS EXPRESSION There are instances of optional homonymy of the kind hypothetically illustrated in Pseudo-Latin (E). For example, the 3rd person singular present form of the English verb need is either needs or need, of which the latter is homonymous with the other person-and-number forms. The 3rd person singular category is thus involved in homonymy optionally only because it also has an unambiguous expression.

✐

✐ ✐

✐

✐ 196

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #220

✐

✐

196

Edith Moravcsik

In instances of optional homonymy, an additional variable is introduced: of the two meaning categories involved, which is the one that will also have an unambiguous expression alongside with the ambiguous one? Markedness theory yields the following prediction (Greenberg 1966): U-24.

If the contrast between a marked and unmarked meaning is optionally suspended with one of the meanings having also an unambiguous exponent, it is the marked meaning that has the unambiguous expression. In other words, the unmarked member of an opposition may stand for the marked member but not vice versa.

Having marshalled nine constraints on homonymy that have been proposed in the literature (U-16–U-24), let us now see how they allow us to make sense of the seven homonymous nominal affixes of Hungarian, which are repeated here: (88)

-i: -ék: -nak/-nek: -ba/-be: -ja/-je/-a/-e: -k: Ø:

possessive plural marker and denominal adjectivalizer associative plural marker and deverbal nominalizer Dative and Genitive marker Inessive and Illative marker singular and plural 3rd person nominal possessor marker on the possessum plural noun marker and 3rd person plural possessor marker marker of singular, Nominative, Accusative, Genitive, and 3rd person singular possessor on the possessum

First of all, -i and -ék can be dismissed as instances of accidental homonymy (cf. Carstairs 1987: 93–102). Their coincidental nature is borne out by two facts. First, the meanings that converge on these forms are very different from each other and thus blatantly violate the tendency expressed in U-23. Second, the homonymy of the possessive plural -i does not extend to the rest of the allomorphs of this morpheme. This leaves us with the remaining five homonymous forms, of which we will first consider the four non-zero ones: the two possessor markers and the two case markers. T HE HOMONYMY OF -k Recall what is involved: -k is the basic plural marker as well as the marker of the 3rd person plural possessor on the plural possessum: (89)

basic plural: possessive plural with PL3 possessor:

almá-k

‘apples’

almá-i-k

‘their apples’

The peculiarity of this morpheme is that its homonymy never results in homonymous words: the -k is the marker of the basic plural if and only if it occurs without the

✐

✐ ✐

✐

✐ 197

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #221

✐

✐

Inflectional morphology in the Hungarian noun phrase

197

possessive plural -i -jai -jei -ai -ei preceding it and it is the marker of the plural possessor if and only if it is preceded by the possessive plural. The Pseudo-Latin parallel would be a pattern where the Ablative meaning of -ibus is possible only in the context of -im ‘my’: (90)

Pesudo-Latin (F): Dative Ablative possessed: unpossessed:

uxor-ibus uxor-im-ibus uxor-utat

‘to wives’ ‘from my wives’ ‘from wives’

The meaning distinction suspended is that of possessed and unpossessed entity and the context is the plural. Since the context is cumulatively expressed, it is a case of syncretism. The two meanings are similar and disambiguating context is on hand. T HE HOMONYMY OF -ja -je -a -e As seen in Section 2.1.7, while the plurality of 1st and 2nd person possessors is marked both on the dependent (the possessor) and the head (the possessum) for nominal possessors, constructions involving 3rd person pronominal and nominal possessors are singly marked only: the plurality of the nominal possessor is marked on the dependent constituent of the construction only, while the plurality of the corresponding pronominal possessor is indicated solely by a displaced marker on the head. As a result, the 3rd person singular marker on the possessum is ambiguous in the number of the possessor: (91)

az ember-nek a ház-a the man-GEN the house-SG3/PL3 ‘the man’s house’ az ember-ek-nek a ház-a the man-PL-GEN the house-SG3/PL3 ‘the men’s house’

The meanings involved – singular and plural – are similar. The context is a nominal possessor: for pronominal posessors, the number distinction is expressed on the possessum, while lost on the possessor pronoun: (92)

az o˝ ház-a the he house-SG3 ‘his house’ az o˝ ház-uk the he house-PL3 ‘their house’

Disambiguation is provided by the nominal possessor, which is marked for number. Since the context is sequentially expressed and the homonymous form in another

✐

✐ ✐

✐

✐ 198

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #222

✐

✐

198

Edith Moravcsik

context does express singular possessor unambiguously – namely, if the possessor is a 3rd person pronoun – it is a case of take-over. As predicted, it is the unmarked – singular – form which is stretched to stand for the plural as well. Such phrase-internally contradictory plural marking is apparently not unique to Hungarian. Koptjevskaja-Tamm (this volume) comments on this pattern in Hungarian and provides some crosslinguistic analogues. In Kabardian possessive phrases, too, the possessor’s plurality is indicated only once, by a marker on the possessum. Thus, ‘the girls’ book’ is expressed as the girl’s their:book rather than what one would expect: the girls’ their book. This displaced marking pattern is what Hungarian has with 3rd person pronominal possessors, but not with nominal possessors. Similarly, in Turkish, the 3rd person possessor’s plurality (whether it is a noun or a pronoun) is shown only once: either on the possessum only (if there is no overt possessor): çocuk-lar-1 child-PL-SG3 ‘their child’,20 or on the possessor only: on-lar-1n çocug-u he-PL-GEN child-SG3 ‘their child’ (Lewis 1967: 40, 69; Underhill 1976: 92–93). In Tatar, the suffixes used on the possessum in reference to the possessor are the same for the 3rd person singular and plural possessor while they are distinct for the other two persons (Poppe 1963: 31). The issue of phrase-internally contradictory plural marking will be further explored in Section 3.1.7. T HE HOMONYMY OF -nak -nek Examples: (93)

Az ember-nek adom a könyvet. the man-DAT I:give the book:ACC ‘I give the book to the man.’ az ember-nek a könyve the man-GEN the book:SG3 or az ember könyve the man book:SG3 ‘the man’s book’

This is a case of optional homonymy because the Genitive is not necessarily expressed with the Dative marker: it may be expressed with a zero marker as well. Pseudo-Latin (E) (in (86)) exemplifies this pattern with Dative uxor-ibus and Ablative uxor-ibus or uxor-ip. The context in which homonymy occurs is nouns: for personal pronouns, the two cases Dative and Genitive have only distinct expressions. Corresponding to the prediction, it is a take-over: in pronouns, -nak -nek is a Dative marker only. The claim about the relevance distinction between neutralized and contextual meanings for take-overs probably does not apply since it is defined on inflectional, rather than stem, contexts. Disambiguation is effected by the adnominal versus adverbal position of the two cases, in line with Plank’s observation (1991a:

✐

✐ ✐

✐

✐ 199

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #223

✐

✐

Inflectional morphology in the Hungarian noun phrase

199

21) that case neutralization is especially likely when it involves ad-verbal cases since the syntactic context provides for disambiguation. The requirement of homonymy to occur in marked contexts is fulfilled provided nouns can be demonstrated to be marked over pronouns – a hypothesis that is supported by multiple evidence, such as the preference for pronouns over nouns to show number and case distinctions. T HE HOMONYMY OF -ba -be As was discussed in Section 2.1.6, Hungarian has three spatial affix triplets where the triplets differ from each other with respect to spatial position (i.e., whether they have to do with ‘in’, ‘on’, or ‘near’) and the affixes within each triplet represent the three options of spatial orientation: ‘location’, ‘source’, and ‘destination’. (94) below shows the positional categories as columns and the orientational ones as rows: (94) LOCATION

DESTINATION SOURCE

IN

ON

NEAR

-ban -ben or -ba -be ‘in’ -ba -ba ‘into’ -ból -b˝ol ‘from inside’

-n -on -en -ön

-nál -nél

‘on’ -ra -re ‘onto’ -ról -r˝ol ‘off’

‘at’ -hoz -hez -höz ‘to’ -tól -t˝ol ‘from’

The portion of this display that involves homonymy is the upper left: in contemporary colloquial Hungarian, locational -ban/-ben ‘in’ is being superseded by directional -ba/-be ‘into’ so that the latter is acquiring the Inessive meaning while still having the Illative meaning as well. Three aspects of the homonymy of -ba/-be run counter to markedness considerations. First, recall U-20, which constrains the context of homonymy. It says marked semantic contrasts are more conducive to homonymy than unmarked ones. But here, of the three positional categories – ‘in’, ‘on’, and ‘near’ – the ‘in’ series is the context in which the neutralization of location and destination occurs. This is counter to the tendency stated in U-20 because the ‘in’ series otherwise seems unmarked relative to ‘on’ and ‘near’ and is thus predicted to be less prone to homonymy. Second, recall U-22, which states a preference regarding the meanings suspended in homonymy in favor of marked meanings over unmarked ones. U-22 would predict that, of the three orientational categories – ‘location’, ‘destination’, and ‘source’ – ‘location’, which is the least marked of the three, is not likely to be involved in homonymy unless both of the two directional markers are. The facts are, however, otherwise: homonymy affects ‘location’ and ‘destination’, rather than ‘destination’ and ‘source’.

✐

✐ ✐

✐

✐ 200

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #224

✐

✐

200

Edith Moravcsik

Third, recall U-24, which stipulates which member of an opposition may become homonymous. According to U-24, if the contrast between an unmarked and a marked meaning is suspended, it is the unmarked form that should assume the meaning of the marked form rather than the other way around. The facts are once again problematic: if ‘location’ is unmarked relative to ‘destination’, it is the Inessive -ban/-ben that should have additionally acquired the meaning of the Illative -ba/-be rather than the the Illative doing so. Let us now see if we can find answers to these puzzles. The first question regarding the neutralization of -ban/ben ‘in’ and -ba/-be ‘into’ is why it takes place in the ‘in’ position as opposed to the ‘on’ or ‘near’ positions. Parallels to this odd pattern offer themselves from other languages as well. While English is in line with markedness considerations since neutralization happens in all three positional categories ‘in’, ‘on’, and ‘near’ (that is, in, on, and near can all be used both for ‘location’ and ‘direction’), in German, neutralization happens in ‘inside’ and ‘on’ (in meaning both ‘in’ and ‘into’, and auf meaning both ‘on’ and ‘onto’, with a difference in governed case) but not in ‘near’. The German configuration holds in Russian as well, where both the ‘in’ preposition v and the ‘on’ preposition na are ambiguous between ‘location’ and ‘destination’, with governed case once again different, but for ‘near’, there are different forms (u ‘at’ and k ‘to’). Also in Russian (Bernard Comrie, p.c.), there are special ablative forms for ‘behind’ and ‘under’ but not for ‘in front of’ and ‘above’. No explanation for these examples of semantic neutralization favoring unmarked contexts can be proposed. The second question raised above was this: if two of the three meaning categories forming a triplet merge, why is the unmarked locational member involved? One would expect the marked categories of destination and source to merge with each other before any of them merges with location. Once again, Hungarian has company: it is not the only language not to bear out this expectation. English, German, Russian, and Latin are examples of languages that are also counter to it. In all these languages, ‘in’ and ‘into’ are collapsed (with the cooccurring affixal case disambiguating in German, Russian, and Latin), rather than ‘into’ and ‘from’ merging. In Yidiny (Blake 1994: 153–155), too, it is ‘location’ and ‘destination’ that merge. This discrepancy between markedness theory and the facts can be made sense of in terms of U-21. This statement which requires that ambiguity be resolved in terms of the lexical cooccurrence constraints of the surrounding constituents, offers an explanation. Verb semantics typically predicts whether the complement will be locational (such as in the case of ‘stay’, ‘live’, ‘work’) or directional (e.g., ‘go’, ‘come’, ‘arrive’), while the semantics of directional-complement verbs does not predict whether the direction is ‘from’ or ‘to’ Thus, merger between ‘to’ and ‘from’ would play havoc with the intended message while merger between ‘in’ and either of the two directional markers does not result in irrecoverable ambiguity (Skaliˇcka 1979a: 74).

✐

✐ ✐

✐

✐ 201

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #225

✐

✐

Inflectional morphology in the Hungarian noun phrase

201

The last question was why the form for ‘destination’ does double duty for both ‘destination’ and ‘location’, rather than in reverse, with ‘location’ taking over for ‘destination’. In English, this latter pattern holds: it is locational in that may be used in contexts that would call for destinational into (e.g. Put the apples in the fridge and not *The apples are into the fridge). The English alternative is what markedness would predict: if in is unmarked relative to into, then in should step in for into. Notice, however, that in terms of phonological form, the Hungarian expression of ‘destination’ is unmarked over ‘location’ since the former (-ba/-be) is shorter than the latter (-ban/-ben). (The same holds in Turkish: -da -de ‘in’, -a -e ‘into’). Thus, when the Allative takes over for the Inessive in Hungarian, formal markedness apparently wins out over semantic markedness. An analogous case is provided by English, where originally directional home has become an optional variant of locational at home. In sum: Of the three questions asked about the merger of -ban/-ben ‘in’ and -ba/ -be ‘into’, (c) can be answered if the priority of formal markedness over semantic markedness is independently motivatable; (b) can be answered by reliance of antiambiguity considerations; and (a) remains unexplained. 3.1.6. Cumulative exponence Different morphological paradigms of a language need not all represent the same symbolization type. For example, paradigms of Basque, Estonian, and Old English differ both along the sensitive-insensitive and the cumulative-separatist dimension (Plank 1986: 36–37). The same point can be made for Hungarian. As seen in Section 2.1.2, affixes differ in how sensitive they are to stems – i.e., in the number of their allomorphs. They furthermore differ along the separatist/cumulativist scale. The very fact that Hungarian has inflectional affix sequences at all shows a separatist bent: certain categories are expressed by separate affixes. Cumulation nonetheless occurs: possessor markers express person and number cumulatively, several spatial case markers cumulate position and orientation, and meaning components of the associative plural are also cumulatively conveyed (H-30, H-31). Furthermore, even stem meanings and affix meanings may have portmanteau expression in the 1st and 2nd person pronoun’s plural and Accusative (H-32). In what follows, we will take up the four instances of cumulative expression found in Hungarian nominal paradigms to see if there are any crosslinguistic generalizations to explain their existence. C UMULATION OF NUMBER AND PERSON Recall the Hungarian pronominal paradigm:

✐

✐ ✐

✐

✐ 202

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #226

✐

✐

202 (95)

Edith Moravcsik

SG1 SG2 SG3 PL1 PL2 PL3

én te o˝ mi ti o˝ k

The cumulative expression of number in the 1st and 2nd person pronoun in contrast to the separatist expression of number in the 3rd person pronoun reflects a clear crosslinguistic pattern. The regularity, consistent with all of the crosslinguistic evidence presented in Forchheimer (1953) and in fact implied by Forchheimer (1953: 64), may be stated as follows: U-25.

If a language forms the plural of one of the three persons by inflection rather than by co-lexicalization, that person will be the 3rd.

Thus, there are languages of the German type where no personal pronoun has inflectional plural marking; there are languages of the Latin, French, Spanish, Russian, or Serbo-Croatian type where the inflectional plural marking is restricted to the 3rd person pronoun; and there are languages such as Turkish or Mandarin Chinese where all persons form their plural by inflection (variant forms indicate gender or politeness distinctions). Standard English is of the German type; but note dialectal forms for the 2nd person plural such as you-all, youse, and thouse. (96) illustrates the main types. (96) SG1 SG2 SG3 PL1 PL2 PL3

German ich du er, sie, es wir ihr sie

Russian ja ty on, ona, ono my ty on-i, on-a

Turkish ben sen o biz siz on-lar

Mandarin wo ni t’a wo-men ni-men t’a-men

Hungarian follows the Russian pluralization type. U-25 thus correctly predicts the inflectional plural in the 3rd person of the Hungarian pronoun. Furthermore, if two of the three persons have identical inflectional plurals, such formal identity between the pluralizing affixes is to be expected in the 3rd and 2nd person over the 3rd and 1st person. This is illustrated by Cree possessive affixes: (97)

SG1 SG2 SG3 PL1 PL2 PL3

nikioni-. . .-inan ki-. . .-iwawa o-. . .-iwawa

✐

✐ ✐

✐

✐ 203

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #227

✐

✐

Inflectional morphology in the Hungarian noun phrase

203

Furthermore, given languages such as Turkish where only the 3rd person pronoun forms its plural using the nominal plural marker, with the 1st and 2nd person using a special plural marker, while Mandarin Chinese uses the same nominal plural marker for all persons, an additional generalization suggests itself: U-26.

If a language indicates number in the 1st and/or 2nd person pronoun by using a plural marker also used for nouns, it does the same in the 3rd person.

The fact that person and number tend to be expressed cumulatively is not surprising if one considers that, as has often been pointed out in the literature, 1st and 2nd person plurals are semantically different from 3rd person and nominal plural. While nominal and 3rd person pronominal plurals involve homogeneous sets, 1st and 2nd person plurals are like associative plurals: they designate heterogeneous sets. Thus, ‘we’ does not normally designate multiple ‘I’-s but, rather, ‘I and at least one other’; and plural ‘you’ is frequently used in reference to the addressee and at least one 3rd person rather than referring to multiple listeners. (For a general discussion of the relationship between person and number, see Smith-Stark 1974; Corbett 2000: 61–66 and passim; Danièl’ 2000; and Cysouw 2001.) In Hungarian, as just seen, the independent 3rd person pronoun does not cumulate person and number. However, the affixal 3rd person possessor marker does just as the 1st and 2nd person markers do. Here is the paradigm for these markers occurring on singular possessum, to be contrasted with the independent pronouns in (95) above: (98)

SG1 SG2 SG3 PL1 PL2 PL3

-m -om -am -em -öm -d -od -ad -ed -öd -a -e -ja -je -nk -unk -ünk -tok -tek -tök -otok -atok -etek -ötök -uk -ük -juk -jük

The cumulative nature of these inflections even in the 3rd person as opposed to the non-cumulative plural of the 3rd person independent pronoun plausibly derives from a crosslinguistic tendency for person and number to have separatist expression in free pronouns preferentially over pronominal affixes. The validity of this tendency, however, once again remains to be tested. C UMULATION OF POSITION AND ORIENTATION Recall once again the spatial case affixes, discussed in Sections 2.1.6 and 3.1.5:

✐

✐ ✐

✐

✐ 204

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #228

✐

✐

204

Edith Moravcsik

(99) LOCATION

DESTINATION SOURCE

IN

ON

NEAR

-ban -ben or -ba -be ‘in’ -ba -ba ‘into’ -ból -b˝ol ‘from inside’

-n -on -en -ön

-nál -nél

‘on’ -ra -re ‘onto’ -ról -r˝ol ‘off’

‘at’ -hoz -hez -höz ‘to’ -tól -t˝ol ‘from’

As noted in Section 2.1.6, minor subregularities do exist in the expression of the various positional and locational categories (as noted by Kibrik 1994; cf. also Skaliˇcka 1979a: 68–71), but none of the six meaning categories involved have consistently recurring distinct exponents. Languages differ in whether position and orientation are cumulatively or separately expressed. As shown in Kibrik 1994, Finnish, similar to Hungarian, shows only some subregularities (Kibrik: this volume), while Avar (Blake 1994), Archi and Hua (Kibrik: this volume) show a consistently transparent pattern. The various positional and orientational categories in Hungarian are not all alike from the point of view of cumulation. The following scales hold (cf. H-31): (100)

CUMULATIVE POSITION ORIENTATION

‘near’ ‘destination’

TRANSPARENT

‘on’ ‘location’

‘in’ ‘source’

Markedness theory predicts that unmarked forms should be more cumulative than marked ones. Although the markedness relations among the categories of position and orientation are far from clear, the following would appear to be a reasonable hypothesis: (101)

UNMARKED POSITION: ORIENTATION:

‘in’ ‘location’

MARKED

‘on’ ‘destination’

‘near’ ‘source’

The comparison of the two charts shows that only two predictions of markedness theory match the facts: ‘source’ is indeed the most transparent orientation and, among positions, ‘on’ occupies an intermediate point in that it is neither maximally cumulative nor fully transparent. The ranking of ‘near’ and ‘in’, and that of ‘destination’ and ‘location’ is, however, counterpredicted. C UMULATIVE EXPRESSION OF THE ASSOCIATIVE PLURAL The expression of the associative plural marker in Hungarian is either monomorphemic or bimorphemic depending on how we analyse it (cf. Section 2.1.6). Under either analysis, however, it involves a complex pattern: either cumulation or noncompositionality. Here is an example:

✐

✐ ✐

✐

✐ 205

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #229

✐

✐

Inflectional morphology in the Hungarian noun phrase

(102)

205

apám-ék ‘my father and his associate(s)’

Since, as discussed in Section 2.1.6, the set defined by the associative plural includes the referent of the noun that forms the stem as well as at least one other similar – human – entity that is associated with the noun referent, the meaning elements of the -ék affix that need to be posited are ASSOCIATED , ALIKE, and INCLUSIVE. Even the bimorphemic analysis cannot assign each of these four meaning components to a separate portion of the affix. A crosslinguistic survey of the associative plural indicates that its expression is never fully separatist: in a sample of over 80 languages, none were found where the associative plural marker consisted of three morphemes, one for each of the three meaning components (cf. Corbett & Mithun 1994; Gil 1994b; Moravcsik 1994; Corbett 2000: 101–111; Danièl’ 2000). In most languages, the affix is monomorphemic, represented by a pronominal or nominal plural marker or a case marker or a conjunction or a plural definite article or a plural pronoun. If the associative plural marker is analysable, it always consists of a plural suffix preceded either by a possessive case marker (as in Lezgian) or a possessive adjectivalizer (e.g., Bulgarian), or a special associative marker (as in Central Alaskan Yup’ik). This suggests that the Hungarian marker, which offers some justification both for a monomorphemic and a bimorphemic analysis, straddles the line between the crosslinguistically predominant monomorphemic and less frequent bimorphemic pattern and that, under either analysis, it bears out the following tentative universal hypothesis: U-27.

In no language is the marker of the associative plural fully transparent, with each meaning element expressed with a separate affix.

C UMULATIVE EXPRESSION OF CASE - MARKED PERSONAL PRONOUNS In Hungarian, cumulation in the 1st and 2nd person pronoun holds not only with respect to the plural (e.g., én ‘I’ versus mi ‘we’) but, as seen in Section 2.1.6, also with respect to the Accusative: while the 3rd person pronouns have regular nominal separatist Accusative marking, e.g., o˝ -t ‘him’, the 1st and 2nd person pronouns cumulate with the Accusative, e.g., engem ‘me’. This, too, is a crosslinguistically common pattern; cf., for example, Latin, Classical Greek, or Russian. Thus, it is possible that the propensity of 1st and 2nd person pronouns toward cumulation stated in U-25 and U-26 holds not only for number but also for case. 3.1.7. Extended exponence As discussed in Section 2.1.7, Hungarian has instances of redundant marking for all three basic inflectional categories: number, case, and the possessor’s person-number marking on the possessum. Multiple markers may occur within the same word or

✐

✐ ✐

✐

✐ 206

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #230

✐

✐

206

Edith Moravcsik

within the same phrase. Examples: ez-t-et this-ACC-ACC ‘this (ACC)’, én-t˝ol-em S1-from-S1 ‘from me’, ez-ek-et a madar-ak-at this-PL-ACC the bird-PL-ACC ‘these birds’. Of the five statements summarizing the facts of extended exponence in Hungarian NPs (H-33 through H-37), there are three that have clear parallels and thus stand a good chance of having a measure of crosslinguistic validity. One of them is H36, which states that, if a category has two exponents in a word or a phrase such that one is free and the other bound and one of the two exponents is optional, the optional exponent is always the free form. This regularity holds widely in “pro-drop languages” that have object and/or subject reference marking on the verb: in such cases, the free object or subject pronoun is omissible but the reference marker is not. Nonetheless, this pattern may not be exceptionless. Hebrew is a partial counterexample: the use of nominal plural with round-number numerals is optional (David Gil, p.c.). Notice, however, that the nominal plural marker and the numerals are not exactly identical in meaning: the plural marker conveys only part of the meaning of the numerals. Tatar comes closer to a genuine departure from the regularity: in possessive phrases involving an overt pronominal possessor, the suffix on the possessum referring to the possessor is optional; e.g., both min-em defter-em I-SG1 notebookSG1 and min-em defter I-SG1 notebook mean ‘my notebook’ (Poppe 1963: 115). Note, however, that the possessive suffix is not the only optional constituent here: the possessive pronoun is also optional if in turn the suffix is present. Thus, defterem notebook-SG1 is also a proper expression of ‘my notebook’. Thus, the Tatar data are still consistent with the generalization according to which if a category has two exponents in a word or a phrase such that one is free and the other bound and only one of the two exponents is optional, that one optional exponent is always the free form. The second observation that has crosslinguistic parallels is H-35 (b), which states that the demonstrative shows the number and case of the noun. As noted, there are several arguments for accounting for this fact by analysing the demonstrative as a separate NP, rather than as an agreeing noun satellite. Apart from the fact that it shows the same case and number inflection as the noun itself, we found three facts in support of this analysis: the demonstrative cooccurs with determiners such as the article or the possessor; it does not show cross-reference to the possessor; and it is peripherally ordered within the NP. At least some of these four facts are paralleled in other languages (cf. Gil: this volume). Thus, in many languages – such as Breton or Abkhaz – the demonstrative is in construction, rather than in complementary distribution, with other determiners (cf. Lyons 1985; Plank 1992). Also, the demonstrative is typically peripheral in the linear sequence of the NP (cf. Greenberg 1963: No. 20; Rijkhoff 1990.). Furthermore, in languages where noun satellites show any agreement at all with the noun, the demonstrative is one of the agreeing constituents – such as it is in English. Thus, the NP-like nature of the demonstrative in Hungarian is not unparalleled across languages (for a general survey, see Diessel 1999: 59–78).

✐

✐ ✐

✐

✐ 207

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #231

✐

✐

Inflectional morphology in the Hungarian noun phrase

207

However, no typological implication is available to actually predict, rather than just allow, this pattern in Hungarian. The third phenomenon with some crosslinguistic support is one noted in H-37. This statement inventories the various ways in which Hungarian number marking does not show the kinds of redundancies that occur in some other languages: the plural is not used on numerated nouns, it is not used on verbs whose subject is a numerated noun, it is optional with paired body parts, it is generally not occurrent on generic nouns, and it is marked only once for the plural possessor. Of these instances of non-redundant number marking, at least one does follow from a crosslinguistic generalization. The pattern of numerated nouns being in the singular has been proposed to be one of the hallmarks of a nominal type called “set noun” by Rijkhoff (1992: 85–87). Given that another characteristic of set nouns is proposed to be singular verb agreement after numerated nouns (Rijkhoff 1992: 94), the following unidirectional generalization may be formulated: U-28.

If in a language numerated noun subjects take singular verb agreement, numerated nouns themselves don’t show plural nominative marking.

The proviso “plural nominative marking” is to make the statement consistent with Russian, where, as pointed out to me by Bernard Comrie, a subject consisting of a numeral ‘5’ or above and a noun in the plural genitive as required after such numerals, may cooccur with a verb in neuter singular. Since in Hungarian, the verb shows singular agreement after a numerated subject, U-28 renders the use of the singular noun with numerals a necessary, rather than random, fact. What, if any, implicational relationship might hold among the rest of the five features of number-marking economy remains an open question. In contrast with the three aspects of extended exponence in Hungarian nominal inflection which are in line with crosslinguistic tendencies, there is one feature of this expression type that runs counter to a proposed crosslinguistic generalization. Kilby notes that adpositions occur only once per NP: i.e., in contrast with affixal case markers, they do not participate in phrase-internal agreement (1981: 115, 118; cf. also Plank 1991c: 17). Although Kilby bases the proposal only on a small sample of languages and does not claim it to be universal, we may extend the point to a tentative crosslinguistic hypothesis: U-29.

NP constituents do not show agreement in adpositionally expressed case.

U-29 is not true for Hungarian because of the way NPs with demonstratives are case-marked. The demonstrative must repeat some or all of the case marking of the noun as follows. First, if the noun’s case is affixal, the demonstrative takes the affix: (103)

ez-en a könyv-ön this-SUPESS the book-SUPESS ‘on this book’

✐

✐ ✐

✐

✐ 208

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #232

✐

✐

208

Edith Moravcsik

Second, if the noun’s case marking consists of both an affix and a postposition, the demonstrative takes at least the affix and optionally the postposition as well: (104)

ez-en a könyv-ön kívül this-SUPESS the book-SUPESS besides or ez-en kívül a könyv-ön kívül this-SUPESS besides the book-SUPESS besides ‘besides this book’

Third, if the noun’s case is postpositional only, the demonstrative takes the postposition: (105)

e-fölött a könyv fölött this-above the book above ‘above this book’

Of these three patterns, the latter two go against U-29: the postposition is repeated on the demonstrative rather than just showing up on the noun as U-29 would require it. (This is a problem for U-29 only if the demonstrative is indeed a constituent of the NP rather than a separate phrase in appositional relation to the NP.) Let us therefore see how U-29 could be relaxed accordingly. First consider the third pattern. On closer analysis, it is not certain that it is entirely counter to U-29. This is because the free status of the adposition following the demonstrative is dubious. Standard Hungarian orthography views the postposition following the case-affixed demonstrative as a free form (103) but the postposition following a bare demonstrative (104) as part of the same word. The affixal nature of the postposition following the bare demonstrative is also supported by the fact that it requires a special reduced allomorph of the demonstrative – one which does not occur if the demonstrative precedes a separate word which similarly starts with a consonant: (106) a.

ez this ‘this’ e-felett this-above *ez-felett this-above ‘above this’ b. Ez finom. this fine

✐

✐ ✐

✐

✐ 209

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #233

✐

✐

Inflectional morphology in the Hungarian noun phrase

209

*E finom. this fine ‘This is fine.’ If we consider the postposition following a bare demonstrative an affix, then the facts are in line with a somewhat modified version of U-29, according to which if the noun has a postposition, NP constituents either do not take it or they take it as an affix. Next, let us turn to (104) above, which is the other pattern that seems counter to U29. Here, the postposition that follows the demonstrative is clearly not affixal. Thus, to accommodate case agreement in affix + postpositional case within U-29, what is needed is appending to it the possibility of optional agreement in postpositional case if the agreeing NP constituent also agrees in affixal case. All in all, the following revision of U-29 would be consistent with the Hungarian facts: U-29

NP constituents do not show agreement in adpositionally expressed case unless either – the adposition becomes affixed, rather than cliticized, to the agreeing constituent, or – the adposition cooccurs with an affixal case marker on the noun, in which case the agreeing constituent may take not only the affix but also the adposition.

3.1.8. Non-compositionality Compositionality in inflection, which as we noted widely occurs in Hungarian nominal inflection (H-38), is a general feature of inflectional systems (cf. for example Plank 1991b: point (19)): U-30.

Inflected words are generally compositional in that the meaning of the word is the sum of the meaning of the stem and the meaning(s) of the affix(es).

There are two constructions that we encountered that lack compositionality: the case marking pattern that involves both a postposition and an affix, and the expression of the associative plural taken under the bimorphemic analysis. The expression a ház-on túl ‘beyond the house’ is non-compositional since the meaning ‘beyond’ is expressed by túl; the Superessive ‘on’ meaning has nothing to do with it. Thus, the meanings of the affixes go beyond the total meaning. The associative plural construction, e.g., János-é-k ‘John and associate(s)’, is non-compositional since the meaning ‘John and associate(s)’ is not all spelled out in the affixes: the sum of the affix meanings falls short of the total meaning.

✐

✐ ✐

✐

✐ 210

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #234

✐

✐

210

Edith Moravcsik

In H-40 we noted a conflict between compositionality with respect to form versus meaning: the associative plural is non-compositional in meaning but compositional in its structure. What we mean by structural compositionality is a construction where components of a larger construction preserve the formal characteristics that they have when occurring all by themselves. The lack of structural compositionality is in turn manifested in a construction where a part does not behave as it would when occurring independently. The non-coincidence of structural and semantic compositionality is not unique either to Hungarian nor to the expression of the associative plural. (107) gives some English examples. (107)

COMPOSITIONALITY IN MEANING

IN FORM

+ − + −

+ − − +

a. b. c. d.

ENGLISH EXAMPLE

twenty-first (*twenty-oneth) sabertooths (*saberteeth) ? underwent (*undergoed)

In (a), both the structural and semantic characteristics of the components are preserved: e.g., twenty-first includes the meaning of ‘one’ and the ordinal form of twenty-one is predictable from knowing the ordinal form of one. In reference to numerals, Greenberg calls this feature of complex expressions – where the components of the complex expression show the same formal properties as what they do when occurring in isolation – the principle of incorporation (1978: 270–280). In (b) in turn, neither semantic nor structural features of the incorporated element are preserved; the meaning of sabertooths in reference to sabertooth tigers is not the sum of its parts nor does tooth keep its regular plural. Other examples of (b) (from Marcus et al. 1993: 9; Kim et al. 1994) are still lifes, walkmans (‘walkman cassette players’), and lowlifes. In (c), the meaning of a component is palpable but its formal properties do not assert themselves. This does not seem to occur in English but ordinal numeral formation in some other languages exemplifies it. In English, Latin, German, or Russian, the ordinals ‘first’ and ‘second’ are suppletive or otherwise irregular forms of ‘one’ and ‘two’. When ordinal forms of higher numerals are formed that include the expression of ‘one’ as the last constituent of the numeral, the same irregular form – first and its equivalents, rather than *oneth – will be used. This is an example of pattern (a). However, in other languages – including Hungarian – the ordinal will be regularized: (108) a.

Hungarian: tíz ‘ten’

tiz-edik ten-ORD ‘tenth’

✐

✐ ✐

✐

✐ 211

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #235

✐

✐

Inflectional morphology in the Hungarian noun phrase

egy ‘one’

b.

husz-on-egy-edik twenty-SUPESS-one-ORD ‘twenty-first’ Basque (Tovar 1957: 74): bi ‘two’ bat ‘one’

ogeita-bat-garren twenty-one-ORD ‘twenty-first’

211

els˝o *egy-edik one-ORD ‘first’ *husz-on-els˝o twenty-SUPESS-first

bi-garren two-ORD ‘second’ len *bat-garren one-ORD ‘first’ *ogeita-len twenty first

Finally, in (d), part meanings are dead but the grammatical behavior of the parts is alive: the past of undergo is underwent even though there is hardly any ‘going’ involved in ‘undergoing’. Other English examples are understood, became, undertook. The associative plural, if analysed as bimorphemic, illustrates this pattern. Hungarian -ék, though, is not quite like the English examples. In understand etc., the meanings of the parts have nothing to do with the total meanings. In the case of the Hungarian associative plural marker, the part meanings are components of the total meaning but they fail to do full justice to it. In sum: while no generalization could be found from which the mismatch between semantic and structural compositionality in the Hungarian associative plural could be derived, the non-uniqueness of this phenomenon has been demonstrated from other constructions and other languages.

3.2. Distribution 3.2.1. Cooccurrence Of the fifteen generalizations about the cooccurrence options of Hungarian nominal inflections (H-41 through H-55), some are more likely to have crosslinguistic validity than others. For example, the non-recursiveness of inflections (H-41c) turns out to be common across languages (although not universal; for multiple case marking, cf. Plank (ed.) 1995). Other patterns – such as the existence of inflected infinitives (H-42b) – are not unique to Hungarian (cf. Portuguese) but they are not frequent,

✐

✐ ✐

✐

✐ 212

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #236

✐

✐

212

Edith Moravcsik

either. The exact crosslinguistic distribution of these patterns is to my knowledge yet undetermined. In what follows, we will focus on areas where explicit crosslinguistic generalizations have been proposed, to see the degree of match between these statements and the Hungarian facts. The following three issues will be discussed: – cross-categorial distribution of nominal affixes – the choice of affix allomorphs by the stem – the choice of affix allomorphs by another affix. T HE CROSS - CATEGORIAL DISTRIBUTION OF NOMINAL AFFIXES As noted in H-42a, all three affix types – number (with the exception of the associative plural), case, and possessor marking on the possessum – are cross-categorially applicable: they occur not only on nouns but on members of any other word class that may fill a nominal slot: pronouns, numerals, and adjectives. This is not a random fact about the language. First of all, the cross-categoriality of NP-internal inflection follows from an areal universal which links the sparsity or absence of nominal subclasses to the very absence of nouns and pronouns themselves as a morphological class distinct from adjectives (Haarmann 1976: 130–131): EU-3.

European languages with fewer than three gender distinctions do not differentiate between nominal and adjectival inflection.

Note, however, that – as pointed out to me by Bernard Comrie – Scandinavian languages are exceptions since they have two genders but distinct nominal and adjectival inflections. More generally, the cross-categorial use of inflection has been claimed to be a characteristic of agglutinating languages (Skaliˇcka 1979c: 336). Assuming that the implication is bidirectional, the following two statements capture the idea: U-31.

If a language is (to an extent) agglutinating, at least some of its inflections can be used on members of more than one word class.

U-32.

If a language uses at least some of its inflections cross-categorially, then the inflectional system of the language is agglutinating at least to an extent.

Plank (1991a: 4–5) suggests a logical tie: the non-discriminating character of the affix may be a manifestation of the generally loose relationship between stem and affix – a symptom of agglutination.

✐

✐ ✐

✐

✐ 213

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #237

✐

✐

Inflectional morphology in the Hungarian noun phrase

213

As was also noted, there is one word class which carries nominal inflection even when it is used not in the place of, but jointly with a noun. This class is demonstratives, which agree with nouns in number and case although they do not take possessive agreement. Is there a relationship between a constituent being nominally inflected when in construction with a noun and its ability to stand for a noun? One of the two possible implications definable over the two options clearly does not hold: it is not the case that any constituent that can fill a nominal position is also nominally inflected when functioning as a noun satellite. For example, Hungarian numerals and adjectives take on the inflectional garb of a noun when used as nouns but they are not inflected when cooccurring with nouns. Would the implication hold the other way around, with agreement with the noun implying the possibility of independent nominal use? Gil’s informal crosslinguistic survey of adjectival form and adjectival use (1994a) yields the following statistical generalization (cf. also Gil, this volume): U-33.

In most languages, if the adjective is inflected when it is in construction with a noun, it can also function as a noun all by itself.

This statement does not apply to Hungarian since the implicans is not met: adjectives are not inflected when they occur with nouns. The only NP satellite that does agree with the noun is the demonstrative. Should a similar generalization hold for demonstratives – or perhaps for all NP constituents – the fact that demonstratives can be used nominally in Hungarian would be explained or at least rendered probable. Walter (1981) proposed a number of generalizations regarding the cross-categorial use of person-number inflection in possessed nouns and verbs. Hungarian meets the implicans of four of his implications. Each of the four statements, to be given below, points at some section of a verbal inflectional system where person-number markers are prone to formal identity with possessive person-number markers. The first is about the identity of possessive affixes with intransitive versus transitive verb inflections. U-34.

If person marking in possessed nouns and intransitive verbs is phonologically identical, this will also be the case for possessed nouns and transitive verbs. (Walter 1981: 113)

As the data below show, the implicans is fulfilled in Hungarian in the plural 1st and 2nd person: these are the only slots in the intransitive verbal paradigm where – allowing for allomorphic variation – the person-number marker is the same as for the possessed noun. This works, however, only if the transitive verb paradigm invoked is the one used with indefinite, rather than definite, objects: this paradigm is exactly identical with that of an intransitive verb (note lap ‘page’, áll ‘stand’, ír ‘write’):

✐

✐ ✐

✐

✐ 214

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #238

✐

✐

214

Edith Moravcsik

(109) SG1 SG2 SG3 PL1 PL2 PL3

POSSESSED

INTRANSITIVE

NOUN

VERB

INDEF OBJ .

TRANSITIVE VERB DEFINITE OBJ .

lap-om lap-od lap-ja lap-unk lap-otok lap-juk

áll-ok áll-sz áll áll-unk áll-tok áll-nak

ír-ok ír-sz ír ír-unk ír-tok ír-nak

ír-om ír-od ír-ja ír-juk ír-jatok ír-ják

Walter’s second hypothesis is about possessive inflection and perfective versus non-perfective verb forms: U-35.

If person marking is phonologically identical for possessed nouns and intransitive verbs of non-perfective tenses, this will also be the case in perfective tenses. (Walter 1981: 114)

Hungarian has no perfective and non-prefective tenses and thus U-35 does not apply to this language. However, a similar tense-related implication pertaining to past and non-past, rather than perfective and imperfective, would be supported by the Hungarian facts. As (110) shows, in the plural 1st and 2nd persons, the inflections of the possessed noun are the same as those of the present and past tenses (not counting allomorphic variation). These facts by themselves do not show either that possessive inflections are preferentially identical with past-tense inflections or that, in reverse, the preference is for identity with present-tense inflections. However, if U-35 is indeed valid, then, given the semantic relatedness of past and perfective, we may hypothesize that the general pattern is preference for formal identity between possessive and past inflections over possessive and present inflections. (110)

POSSESSED

SG1 SG2 SG3 PL1 PL2 PL3

INTRANSITIVE VERB

NOUN

PRESENT

PAST

lap-om lap-od lap-ja lap-unk lap-otok lap-juk

áll-ok áll-sz áll áll-unk áll-tok áll-nak

áll-t-am áll-t-ál áll-t áll-t-unk áll-t-atok áll-t-ak

The third hypothesis has to do with mood: U-36.

If person marking is phonologically identical for possessed nouns and intransitive verbs of non-indicative moods, this will also be the case in the indicative mood. (Walter 1981: 114)

✐

✐ ✐

✐

✐ 215

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #239

✐

✐

Inflectional morphology in the Hungarian noun phrase

215

In (111), non-indicative moods are represented by the subjunctive and the conditional. The prediction does hold for both; relevant forms are, again, those in the plural 1st and 2nd person. (111)

POSSESSED

SG1 SG2 SG3 PL1 PL2 PL3

INTRANSITIVE VERB

NOUN

SUBJUNCTIVE

CONDITIONAL

INDICATIVE

lap-om lap-od lap-ja lap-unk lap-otok lap-juk

áll-j-ak áll-j(-ál) áll-j-on áll-j-unk áll-j-atok áll-j-anak

áll-n-ék áll-n-ál áll-na áll-ná-nk áll-ná-tok áll-ná-nak

áll-ok áll-sz áll áll-unk áll-tok áll-nak

Finally, U-37 compares subordinate and non-subordinate forms: U-37.

If person marking is phonologically identical for possessed nouns and intransitive verbs of the non-subordinate sort, this will also be the case for intransitive verbs of the subordinate sort. (Walter 1981: 116)

Once again, the prediction does come true; see the plural 1st and 2nd person forms (kell ‘must’): (112)

POSSESSED

SG1 SG2 SG3 PL1 PL2 PL3

INTRANSITIVE VERB

NOUN

NON-SUBORD.

SUBORD.

lap-om lap-od lap-ja lap-unk lap-otok lap-juk

áll-ok áll-sz áll áll-unk áll-tok áll-nak

áll-n-om kell áll-n-od kell áll-ni-a kell áll-n-unk kell áll-n-otok kell áll-ni-uk kell

In sum: the Hungarian evidence bears out, or at least is consistent with, a preference for possessive nominal inflection to show identity with transitive over intransitive verb inflection, with past over present verb forms, with indicative over non-indicative verb affixes, and with the subordinate over non-subordinate verbal paradigm. T HE CHOICE OF AFFIX ALLOMORPHS BY THE STEM In recent years, three proposals have been made with the intent of constraining the possible distributional patterns of affix variants relative to stems: one by Wurzel and two by Carstairs-McCarthy. Wurzel proposed (1989: 114–121) that there is always some predictability from the choice of one exponent of an inflectional category – say, the Accusative – to exponent choice in another inflectional category, such as the Genitive. Thus, the following inflectional system should never occur (C1, C2 stand

✐

✐ ✐

✐

✐ 216

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #240

✐

✐

216

Edith Moravcsik

for inflectional categories such as cases; A, B, C etc. stand for stem types such as the various declensional classes in Latin; a, b, c, etc. are inflectional affixes): STEM CLASSES INFLECTIONAL CATEGORIES:

C1 C2

A a b

B b a

C b c

D c b

E a c

F c a

This system is counterpredicted since here the applicability of an exponent of an inflectional category to a stem does not imply anything about what exponent will be selected for another inflectional category. None of the logically possible implications predicting the exponent of C2 (e.g., the Accusative) from C1 (e.g., the Nominative) hold: *if C1 is a, then C2 is b *if C1 is a, then C2 is c *if C1 is b, then C2 is a *if C1 is b, then C2 is c *if C1 is c, then C2 is a *if C1 is c, then C2 is b For example, while exponent “a” goes with exponent “b” in class A, the choice of “a” does not predict the choice of “b” since “a” occurs without “b” in class E. Similarly, none of the six reverse implications predicting C1 from C2 hold, either. The minimal way for exponent-to-exponent predictions – dubbed “paradigm structure conditions” by Wurzel – to be in evidence is for there to be at least one exponent of at least one inflectional category which predicts the exponent of at least one other inflectional category for at least one stem class. But the requirement that in every inflectional system there be at least one such paradigm structure condition does not begin to capture the actual limitedness of inflectional systems as encountered in various languages: actual paradigms far exceed this minimal stipulation. Hungarian is an example: as shown in H-48, several dependencies hold among exponents. Since assuming a minimal paradigm structure condition does not do justice to the actual nature of inflectional paradigms, we might entertain a maximal constraint. The maximal way in which paradigm structure conditions could constrain paradigms would be if the exponent of every inflectional category predicted exponents of all other inflectional categories for a stem class. This would be true if all exponents were entirely unique, with no recurrence of the same inflectional form across stem classes. That this requirement errs in the direction of being too restrictive is also shown by Hungarian. Hungarian falls short of having a maximally constrained nominal inflection system in that not all exponents are predictive of the rest of the paradigm.

✐

✐ ✐

✐

✐ 217

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #241

✐

✐

Inflectional morphology in the Hungarian noun phrase

217

For example, if a noun has -ok in the basic plural, it may still have either -ja or -a for the 3rd person possessor marker: (113) a.

b.

patak-ok creek-PL ‘creeks’ horgony-ok anchor-PL ‘anchors’

patak-ja creek-SG3 ‘his creek’ horgony-a anchor-SG3 ‘his anchor’

The question is this: where exactly do Hungarian nominal paradigms fall between the two extremes of being minimally constrained and being maximally constrained? The minimal paradigm structure constraint (“If for stem class A, C1 is expressed by exponent ‘a’, then C2 is expressed by exponent ‘b’ ”) may gain generality in two ways. One is if the same constraint applies to more than one stem class; for example, the form of the Genitive is predictable from the form of the Accusative not only in Declension One but also in Declensions Two and Three. The second source of generality is exponent choice in an inflectional category predicting exponent choice in more than one other inflectional category in the same stem class; for example, the Accusative marker in Declension One predicts not only the choice of the Genitive marker but also the choice of the Ablative marker. As we saw in Section 2.2.1, in Hungarian nominal inflection there are examples of paradigm structure constraints that are general in each of these two ways. First, some dependencies hold across stem classes. For example, the generalization that, if the stem selects the vowel-initial allomorph of the Accusative affix, it will also select the vowel-initial allomorph of the 1st person singular possessor affix, holds both for the stem class that requires the vowel to be /a/ and for the stem class that requires /o/ (cf. H-48 a/): (114) Accusative: SG1:

‘foot’ ‘creek’ láb-at patak-ot láb-am patak-om *láb-m *patak-m

Second, some inflectional categories predict exponent choice for more than one other inflectional category. For example, the presence of the affix-initial vowel in the Accusative predicts the presence of the vowel not only for the 1st person singular possessor marker (as shown in (114)) but also, as (115) shows, for the 2nd person singular and plural possessor markers (H-48a):

✐

✐ ✐

✐

✐ 218

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #242

✐

✐

218

Edith Moravcsik

(115) Accusative: SG2: PL2:

‘foot’ ‘creek’ láb-at patak-ot láb-ad patak-od *láb-d *patak-d láb-atok patak-otok *láb-tok *patak-tok

The question is whether there are any specific crosslinguistic generalizations from which the particular nature of these dependencies follows. The two proposals by Carstairs-McCarthy alluded to above strengthen the minimal paradigm structure constraint in both directions just discussed. Here is a statement of one of them: the Paradigm Economy Principle (Carstairs 1984; Carstairs-McCarthy 1987): U-38.

Paradigm Economy Principle: In any inflectional system of any word type, for any stem class there must be at least one inflectional category whose exponents predict the rest of the paradigm; and that inflectional category must be the same for all stem classes.

What this means is that, given the entire inflectional system charted with stem classes as columns and inflectional categories as rows, there must be one row where the exponents of the different inflectional categories are all distinct. For example, if a language has five nominal stem classes and five cases, one case – for example, the Accusative – must be predictive of all four other case exponents in each stem class. This principle strengthens the minimal paradigm structure constraint schema: In stem class A, if C1 has a, then C2 has b, in the following way: For all stem classes A, B, C . . ., there is at least one inflectional category C1 such that if C1 has the affix a, then all the other inflectional categories C2, C3 . . . have b, c . . ., respectively (where C1 is the same category across stem classes and where a, b, c . . . are distinct). While the Paradigm Economy Principle imposes a stipulation on one row of an inflectional chart, Carstairs-McCarthy’s second hypothesis, the No Blur Principle (Carstairs-McCarthy 1994), complements the first by setting up a constraint across all rows in the chart. While the Paradigm Economy Principle imposes a maximal constraint on one row by requiring that all exponents in one row be fully predictive of the rest of the exponents, the No Blur Principle imposes a less than maximal constraint on all rows by requiring that, in any row, no more than one exponent be allowed to fail to be fully predictive of all other exponents within the stem class.

✐

✐ ✐

✐

✐ 219

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #243

✐

✐

Inflectional morphology in the Hungarian noun phrase

U-39.

219

No Blur Principle: In any inflectional system of any word class and for any inflectional category, there can be no more than one exponent that fails to uniquely predict the rest of the paradigm.

Putting it again schematically, the No Blur Principle strengthens the minimal schema: In stem class A, if C1 has a, then C2 has b, as follows: For all inflectional categories C1, C2, C3, . . . if inflectional category C1 is expressed as a, b, c, . . . in the different stem classes, then each of a, b, c, except one must be predictive of the exponents of all other inflectional categories C2, C3, C4, . . . within the same stem class. Let us now examine how Hungarian nominal paradigms stack up against the principles of Paradigm Economy and No Blur. Consider the following: (116)

Stem class A (pokróc ‘blanket’) PL: -ok SG3: -a

Stem class B (patak ‘creek’) -ok -ja

Stem class C (fal ‘wall’) -ak -a

Stem class D (híd ‘bridge’) -ak -ja

It is easy to see that the No Blur Principle does not hold: in each row, there are exponents that fail to identify uniquely the rest of the paradigm, while the principle would admit only one such exponent. However, the validity of the Paradigm Economy Principle cannot be assessed based only on (116). If the two rows displayed in (116) exhausted all of Hungarian nominal inflection, the principle would be clearly violated. This is because the exponents in neither row predict exponent choice in the other row: both -ok and -ak can paradigmatically cooccur with either -a or -ja, with all four possible combinations of the four exponents possible. Thus, (116), being a partial paradigm, does not in itself disprove the Paradigm Economy Principle: the possibility is left open that the rest of the paradigm would reveal a row that meets the requirement. In actuality, however, there is in fact no inflectional category each of whose exponents is unique to its stem class and thus the Hungarian facts violate Paradigm Economy as well.21 T HE CHOICE OF AFFIX ALLOMORPHS BY ANOTHER AFFIX As noted earlier (H-51), Hungarian affixes generally condition allomorphs to the right. This is in conformity with a crosslinguistic tendency: as first noted by Carstairs (1984), allomorph choice tends to be dictated “outwards” rather than “inwards”. In other words, suffix allomorphs tend to be choices based on preceding context and prefix allomorphs on following context. That some languages, Hungarian among

✐

✐ ✐

✐

✐ 220

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #244

✐

✐

220

Edith Moravcsik

them, violate this principle was recognized by Carstairs-McCarthy (1984: 51–52; cf. also 1987 and 1992: 212–213): in Hungarian, as discussed in Section 2.2.1.2 (cf. H-51) the choice between the basic and possessive plural depends on whether the plural marker is followed by a possessor marker or not: (117)

asztal-ok table-PL *asztal-i table-PPL ‘tables’ asztal-ai-m table-PPL-SG1 *asztal-ok-om table-PL-SG1 ‘my tables’

To accommodate such facts, Carstairs-McCarthy proposed the following constraint on the directionality of conditioning: U-40.

Peripherality (or No Look Ahead) Constraint: Allomorphic choice is never conditioned by more peripherally positioned environment unless the conditioning environment includes an entire paradigm.

Since in Hungarian the possessed plural must be chosen before all possessor markers and not just, say, before the 1st person singular, leftward conditioning is condoned by the constraint. However, Hungarian offers an additional instance of inwards conditioning which violates even this relaxed stipulation. As noted in Section 2.2.1.2 (cf. H-51), the leftwards-conditioning power of the possessor markers is manifested not only in that they require a special allomorph of the plural but also in that they condition allomorphic variation for the infinitival affix -ni -n: (118)

ír-ni write-to ‘to write’ ír-n-om write-to-SG1 ‘for me to write’ ír-n-od write-to-SG2 ‘for you (SG) to write’

✐

✐ ✐

✐

✐ 221

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #245

✐

✐

Inflectional morphology in the Hungarian noun phrase

221

ír-ni-a write-to-SG3 ‘for him to write’ ír-n-unk write-to-PL1 ‘for us to write’ ír-n-otok write-to-PL2 ‘for you (PL) to write’ ír-ni-uk write-to-PL3 ‘for them to write’

Unlike in the case of the possessive plural, where all members of the inwardsconditioning paradigm were united in opting for the possessive plural over the basic plural, here some members of the inwards-conditioning paradigm select -ni and others select -n. These facts, therefore, run counter to Peripherality. There are two general ways in which the Peripherality Constraint might in principle be further loosened so as to become consistent with these facts. On the one hand, one might try to exempt some of the affected affixes: the ones that are subject to leftward conditioning. On the other hand, one might attempt to exempt the affecting affix type: the affixes that may perform inwards conditioning. Exploration of the first option – stipulating what may be affected inwards – is successful in part only. Note first that the possessive plural is used not only preceding the possessor markers but also following another affix, namely the possessum-deriving -é:

(119) a.

lány-ok girl-PL ‘girls’ lány-ai-m girl-PPL-SG1 *lány-ok-om girl-PL-SG1 ‘my girls’ b. lány-é girl-AFF ‘that of the girl’ lány-é-i girl-AFF-PPL

✐

✐ ✐

✐

✐ 222

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #246

✐

✐

222

Edith Moravcsik

*lány-é-k girl-AFF-PL ‘those of the girl’ In other words, choice of the possessive plural over the basic plural may be called for from either direction: -i must be chosen over -k either if the plural morpheme is followed by a possessive affix or if it is preceded by the possessum-deriving -é. Thus, one might hypothesize the following: An affix may be inwards-conditioned by an entire paradigm only if it is also outwards-conditioned in another context in the same language. However, this formulation would take care only of the possessive plural: it would do nothing to allow for the inbound conditioning of the infinitival affix since that suffix is never conditioned by preceding environment. The other alternative, that of constraining the conditioning rather than the conditioned affix, is more successful. There are two facts about the possessor markers that are relevant. First, the possessor markers do not condition allomorphs only on their left: they can also condition allomorphs to the right. Thus, the Accusative affix (on consonant-final back-vowel stems) always occurs as -at even if the stem without the possessor marker would select -ot (cf. Section 2.2.1.2 and H-50): (120)

kalap-ot hat-ACC ‘hat (ACC)’ kalap-om hat-SG1 ‘my hat’ kalap-om-at hat-SG1-ACC *kalap-om-ot hat-SG1-ACC ‘my hat (ACC)’

Second, outwards conditioning figures in the life of the possessor markers in another way as well: they themselves are conditioned from the left. The choice among the various allomorphs of each possessor marker is outward-conditioned by the stem: (121)

kalap-om hat-SG1 ‘my hat’ madar-am bird-SG1 ‘my bird’ kalap-ja hat-SG3 ‘his hat’

✐

✐ ✐

✐

✐ 223

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #247

✐

✐

Inflectional morphology in the Hungarian noun phrase

223

madar-a bird-SG3 ‘his bird’ Thus, a minimal reformulation of the Peripherality Constraint that is consistent with the Hungarian facts would be the following: U-40 .

Allomorphic choice is never conditioned by more peripherally positioned evironment unless either – the conditioning environment includes an entire paradigm, or – members of the conditioning paradigm both condition allomorphs of some other morphemes outwards, and are themselves selected through outwards conditioning.

This revision preserves the spirit of the original formulation in that it continues to uphold the priority of outwards conditioning. Rather than altogether exempting certain affix types from outwards conditioning, it allows for conditioning in the marked direction only if the same affix type also engages in the unmarked pattern by being both affected by and affecting allomorph choice outwards. 3.2.2. Linear order As seen in Section 2.2.2 (cf. H-56 through H-59), all nominal affixes in Hungarian are suffixed; they form a continuous sequence except for the intrusion of the possessum-deriving -é, and they are, apart from one exception, invariantly ordered with respect to each other, with number preceding possessor marking and possessor marking preceding case. The suffixing nature of Hungarian morphology is in line with a general crosslinguistic preference for suffixing (cf. Cutler, Hawkins, & Gilligan 1985; Hall 1988; Hawkins & Gilligan 1988). Furthermore, the existence of at least some suffixes in Hungarian directly follows from a Haarmannian Euroversal (Haarmann 1976: 139– 140): EU-4.

All European languages have productive suffixing.

There is, however, no crosslinguistic generalization known to me from which the suffixhood of the three particular affix types – number, case, and possessor marking – would necessarily follow. Of the three types, the one that comes closest to being necessarily postposed is case: case affixes are suffixed in the overwhelming majority of languages (Hawkins & Gilligan 1988: 222); exceptions are some Semitic and Bantu languages (Hetzron 1980: 278; Blake 1994). U-41.

If a language has case affixes on nouns, they are almost always suffixed.

✐

✐ ✐

✐

✐ 224

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #248

✐

✐

224

Edith Moravcsik

As far as number and person-number cross-referencing affixes are concerned, they are not universally suffixed. Nevertheless, there are some crosslinguistic regularities that predict the Hungarian order. First, nominal number inflection shows a strong tendency toward being postposed in postpositional languages – which Hungarian clearly is – as well as in languages with dominant SOV order, which is one of the frequent orders in Hungarian (Hawkins & Gilligan 1988: 223). U-42.

In postpositional languages, nominal number affixes, if any, are suffixed with considerably greater than chance frequency (in 94 % of the sample used).

U-43.

If a language has SOV order, nominal number affixes, if any, are suffixed with overwhelmingly greater than chance frequency (in 97 % of the language sample used).

Second, regarding the possessor-marking affixes, a Euroversal stating a European tendency is available (Manzelli 1990: 66): EU-5:

With the exception of some North-Western Caucasian languages, no European language has person-number cross-reference prefixed to possessed nouns.

If, pending a non-affixal analysis of verbal prefixes (cf. Ackerman 1987; Kiefer 1992; Kiefer & Ladányi 2000), suffixing turns out to be the only affixation pattern of Hungarian, then this feature of the language serves as a predictor of postpositions (Greenberg 1963: No. 27): U-44.

If a language is exclusively suffixing, it is postpositional.

Having considered the position of affixes relative to the stem, let us now turn to linear order within the affix complex. Greenberg offers a statistical prediction of the relative order of number and case inflection which fits the Hungarian facts (Greenberg 1963: No. 39; cf. also Booij 1993: 35): U-45.

Where morphemes of both number and case are present and both follow or both precede the noun base, the expression of number almost always comes between the noun base and the expression of case.

This order pattern is part of a more general one. According to Bybee (1985a: 33–35), the overarching regularity is that affix order reflects degrees of semantic relevance to the stem, with relevance decreasing with distance. What this says in fact is that there is no affix-to-affix ordering principle; all affix ordering falls out of affix-to-stem principles. U-46.

Affixes are ordered so that their increasing distance from the stem reflects decreasing semantic relevance to it.

✐

✐ ✐

✐

✐ 225

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #249

✐

✐

Inflectional morphology in the Hungarian noun phrase

225

What U-46 predicts is that, given any pair of affixes, there is only one possible ordering for them both within any one language and also across languages, unless alternative ordering is accompanied by alternative meanings so that the less peripheral affix is the one that ends up interpreted as more relevant to the stem’s meaning. If we consider crosslinguistic evidence, what emerges is that the order of affixes is not always uniform across languages or even within a single language. First, affixes whose order varies from one language to another include case and possessor marker, and number and possessor marker. Thus, while in Hungarian, possessor marker precedes case, the opposite order holds in Finnish (Haarmann 1976: 135; Blake 1994: 106; Harms 1964: 2): cf. Finnish kirko-lla-mme church-ADESS-PL1 versus Hungarian templom-unk-nál church-PL1-ADESS ‘at our church’. (On the synchronic rationale and historical evolution of the varying orders of case and possessor affixes in Uralic languages, cf. Nichols 1973; Comrie 1980.) Similarly, number comes before possessor marking in Hungarian as well as in Turkish, but in Chuvash, possessor marking precedes number (Johanson 1973: 91). Since there is no concomitant difference in meaning correlated with the different order patterns, these examples compromise the extent to which the Relevance Hypothesis holds. Second, the order of affixes may vary even within a single language. In some cases, alternative order is accompanied by alternative meaning as predicted by the Relevance Hypothesis. Here are two examples. In Turkish, the plural marker -lar/ -ler taken by itself is ambiguous between the basic plural and the associative plural meaning. If, however, the noun is a possessed noun, the order of the personnumber agreement affix and the plural disambiguates between the two meanings (Lewis 1967: 40): if the plural affix precedes the possessor marking, the word expresses a regular plural, while the plural following the possessor marking has the associative reading: (122)

karde¸s-ler brother-PL ‘brothers’ karde¸s-im brother-SG1 ‘my brother’ karde¸s-ler-im brother-PL-SG1 ‘my brothers’ karde¸s-im-ler brother-SG1-PL ‘my brother and his family’

A somewhat similar example involves the ordering of the possessum-deriving -é

✐

✐ ✐

✐

✐ 226

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #250

✐

✐

226

Edith Moravcsik

in Hungarian. As seen in Section 2.2.2, the plural affix may precede or follow the -é suffix but with a difference in semantics: (123)

ember-é-i man-AFF-PPL ‘those of (the) man’ ember-ek-é man-PL-AFF ‘that of (the) men’

While in these two examples of language-internally variant affix order there is a concomitant difference in meaning as predicted by U-46, there are other instances where no semantic difference is present. In Nenets, possessive suffixes generally follow case suffixes but in some instances of dual possessum, the order is the reverse (Hajdú 1963: 65–66). In Hungarian, as noted in Section 2.2.2, there is optional variation in the ordering of case and possessor marking for some pronouns with no accompanying meaning difference (Tompa 1968: 195): (124)

elé-m-be in:front-SG1-into ‘in front of me (direction to)’ elé-be-m in:front-into-SG1 ‘in front of me (direction to)’

This clearly goes against relevance-based ordering. In the light of such evidence, the Relevance Principle of affix ordering can be taken as a tendency but not as a strict universal. As such, however, it is very useful since it is suggestive of a number of facts. For example, it suggests a reason for the fact that, as seen before, two plural markers, the basic plural and the associative plural, take up different positions in the Hungarian affix sequence: (125)

lány-ai-m girl-PPL-SG1 ‘my girls’ lány-om-ék girl-SG1-APL ‘my daughter and her associates’

Since in a phrase like lány-om-ék ‘my daughter and her associates’ the associative plural -ék identifies the group through reference to the possessum lány-om ‘my daughter’, the possessive affix is maximally relevant to the stem and this is why it precedes the associative plural affix. The Relevance Principle also explains a violation of a very general order pattern in Hungarian. Greenberg proposed the following (1963: No. 28):

✐

✐ ✐

✐

✐ 227

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #251

✐

✐

Inflectional morphology in the Hungarian noun phrase

U-47.

227

If both the inflection and the derivation follow the root, or they both precede the root, the derivation is always between the root and the inflection.

Whereas this generally holds for Hungarian – derivational suffixes do precede inflectional ones – we encountered one derivational affix which interrupts the inflectional affix sequence: the -é affix deriving a possessum out of a possessor (Section 2.2.2). For example: (126)

ember-ek-é-i-b˝ol man-PL-DERIV-PPL-ELAT ‘out of those of the people’

The affix -é (if indeed derivational, as shown by its semantics) does violate U-47 since it follows the plural inflection. However, looking at things from the point of view of the Relevance Principle, the order makes perfect sense since -é has the plural within its scope rather than in reverse. What seems exceptional here is the fact that derivational meaning applies to the meaning of an inflected stem. Once this fact is accepted, the order of the affix is in accordance with the Relevance Principle. While U-47 is generally a consequence of the Relevance Principle, in those cases where the two are in conflict, the Relevance Principle prevails. Booij (1993: 42) proposes an alternative constraint on the linear order of derivational and inflectional affixes that is less restrictive than U-47: U-48.

Non-contextual inflection may feed word formation.

What this says is that inflections that express properties inherent to the stem, such as nominal number and gender, may be closer to the stem than derivational affixes, but inflections of the non-inherent, relational sort, such as case and agreement, may not. The fact that in Hungarian, the (possessive) plural suffix may precede possessumderiving -é is clearly in line with U-48. Whether the precedence of the possessor markers relative to the -é affix is also permitted by U-48 depends on the exact delimitation of the notion “non-contextual inflection”. The possessor markers are not agreement morphemes since they are present even when there is no overt external possessor to agree with: thus they are “non-contextual”. On the other hand, they are relational, rather than inherent, since they point to the possessor which is a semantic entity distinct from the possessum. 3.2.3. Bonding The distinction between affixes and adpositions in Hungarian was seen to be scalar rather than all-or-none (Section 2.2.3). The gradual nature of the distinction between case markers of the free and bound type is common across languages (Kahr 1976, Kilby 1981) and it has been demonstrated by Plank to hold along a large number of parameters (Plank 1991c).

✐

✐ ✐

✐

✐ 228

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #252

✐

✐

228

Edith Moravcsik

In addition, there are also different degrees of bondedness among affixes themselves: we saw that affixes differ in the extent to which their form responds to the choice of the stem and in the extent to which the stem’s form responds to their presence. Bybee (1985a: 36–37, 1985b) proposes that relevance to the stem correlates not only with affix order but also with the degree to which affixes are bonded to the stem. U-49.

The degree of bonding between stem and affix is in proportion to the semantic relevance of the affix to the stem.

In order for U-49 to explain different degrees of bondedness among nominal inflections in Hungarian, it would have to be shown on independent grounds that affixes with the highest numbers of allomorphs and the widest stem-allomorph-provoking capability – such as the Accusative as opposed to the Superessive – are indeed semantically more relevant to the stem. It is, however, not clear to me how relevance distinctions can be independently motivated among case markers. There is one more fact about Hungarian nominal inflection that seems to fall out from a crosslinguistic generalization. As discussed in Section 2.2.3, the degree of bonding between stem and affix is also affected by whether the stem is a noun or a pronoun. Inflected pronouns show a much tighter bonding. This clearly corresponds to a widely observable pattern: U-50.

If noun stems and pronoun stems show a difference in the degree to which inflections are bonded to them, the difference is in the direction of more bonding in pronouns.

4. Conclusions This chapter first attempted to highlight some aspects of Hungarian nominal inflection according to twelve general parameters taken to be seminal to morphological description. Second, it attempted to place the Hungarian facts in a typological context with the goal of explaining as many of them as possible through crosslinguistic generalizations from which the facts would necessarily or statistically follow. From the point of view of their language-internal distribution, the facts of Hungarian nominal inflection differed from each other. Some of them – such as /v/assimilation (cf. Section 1.2) – were restricted to the nominal subdomain of Hungarian morphology while others – e.g., vowel harmony as governing the distribution of affix allomorphs, or economy in number marking – extended to the entire morphology and in some cases even to the syntax of the language. Similarly, from the point of view of their crosslinguistic distribution, the Hungarian facts have been found to be diverse. They fall into four broad classes. Some facts

✐

✐ ✐

✐

✐ 229

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #253

✐

✐

Inflectional morphology in the Hungarian noun phrase

229

are language-specific, such as the particular phonological forms -ban -ben expressing the Inessive. Others turned out to be crosslinguistically recurrent: parallels from other languages were available but the class of languages in which the particular facts occur could not be characterized. The homonymy of the expression of the Inessive and the Illative, occurring in Hungarian, English, Russian, etc., is an example (Section 3.1.5); the tendency for synonymous forms to acquire different meanings is another (Section 3.1.4). Facts of the third kind are also crosslinguistically recurrent and, to boot, they are universal or universally dominant albeit only within a limited subclass of human languages, where that subclass is either structurally or areally definable. The existence of such facts could thus be derived – and thus in a sense explained – or rendered likely from generalizations of universal scope. For example, that the Accusative has an affixal rather than postpositional exponent in Hungarian necessarily follows from the structurally-restricted universal according to which this is so in all languages if the Dative is also affixal (Section 3.1.1); and the fact that Hungarian has a singularplural distinction is derivable from the areally-restricted generalization according to which all languages have such a distinction if they are European (Section 3.1.1). Finally, the fourth class is made up by facts that are derivable from statements that are universal or universally preferred within the entire class of human languages. The various constraints on homonymy (Section 3.1.5) or the non-compositionality of the associative plural markers (Section 3.1.8) are examples. One morphological characteristic that has been proposed by several linguists to be useful in predicting a number of features of Hungarian is agglutinativity (e.g., Skaliˇcka 1979c: 335; Korhonen 1979: 16–19; cf. also Bechert 1990: 128 and Haspelmath 1999). In view of the facts presented, let us return to this issue raised in the beginning of this chapter (Section 1.1) and attempt to determine the extent to which this concept indeed explains facts about Hungarian. According to Skaliˇcka, agglutinating languages exhibit a very large number of characteristics (Skaliˇcka 1979a: 117, 1979b: 223, 1979c: 335–337). We will select for consideration just thirteen of these (cf. Carstairs-McCarthy 1992: 167). Some of these features have to do with familiar patterns of meanings, forms, and distribution of the inflections: Form: (a) Inflections are generally at least one syllable long. (b) Inflections are invariant in form. Meaning: (c) Semantic types of inflection do not include gender. Synonymy: (d) Inflections show little synonymy. Homonymy: (e) Inflections show little homonymy.

✐

✐ ✐

✐

✐ 230

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #254

✐

✐

230

Edith Moravcsik

Cumulation: (f) Inflections show separatist (non-cumulative) expression. Extended exponence: (g) Inflections do not show one kind of extended exponency: agreement. Distribution: (h) Inflections are cross-categorially applicable; if cross-categorial applicability is not total, then nouns and adjectives are classed against verbs. Other proposed members of the agglutinating feature cluster have to do with noninflectional aspects of morphology: (i) (j)

The language has derivational affixes. There are many similarities between inflectional and derivational affixes,

and with syntax: (k) (l) (m)

Closed-class items are absent. Subordinate clauses are not conspicuous. Word order is fixed.

If we take the assertion literally that these features mutually imply each other, the claim turns out to amount to a large number of dependencies of the sort “If a language has inflections with little synonymy, it has fixed word order.” Since, of the thirteen traits, each can be an implicans for the other twelve, the total is 13 × 12 = 156 implications. Eliminating those that are redundant because of the transitivity of the implication relation still leaves us with at least 24 logically distinct unidirectional implications, each making a separate empirical claim about the relative crosslinguistic distribution of two structural traits. Our exploration of Hungarian nominal inflection, taken together with some information outside this particular domain of Hungarian morphology, allows us to form an idea of the extent and ways in which Hungarian bears out the claimed clustering of features. Since several of the characteristics are stated in terms of agglutinating languages having “more” or “less” of something, the claims can only be tested loosely.

✐

✐ ✐

✐

✐ 231

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #255

✐

✐

Inflectional morphology in the Hungarian noun phrase TYPICAL CHARACTERISTICS

231

ARE THEY PRESENT IN HUNGARIAN?

Inflections show (a) at least monosyllabic form (b) formal invariance (c) no gender (d) little synonymy (e) (f) (g) (h)

little homonymy separatist (non-cumulative) expression no agreement cross-categorial applicability

Other aspects of the grammar exhibit (i) derivational affixes (j) many similarities between inflectional and derivational affixes (k) no closed-class items (l) few subordinate clause types (m) fixed word order

mostly YES mostly NO YES

for morphemes, mostly YES; but there is allomorphy YES YES

and NO

NO YES

YES YES

NO NO YES

and NO

As this chart shows, the label “agglutinative”, in the sense of referring to a cluster of thirteen features that all mutually imply each other, cannot be applied to Hungarian: the language is agglutinative in some respects and to a degree but not in other respects and not fully. This is in spite of the fact that Skaliˇcka considers Hungarian – along with Turkish, Finnish, Armenian, Basque, Georgian, and Eskimo – a clear token of this type (Skaliˇcka 1979c: 335). Clearly, much more empirical work is needed to establish possible crosslinguistically consistent cooccurrence patterns of grammatical properties that may end up characterizing agglutinating languages (cf. Plank 1999 and Haspelmath 1999; on patterns of co-variation between phonology and morphosyntax in general and in agglutinating languages in particular, see Plank 1998). If we grant that the label “partially agglutinative” is usefully applied to Hungarian in that it suggests at least some aspects of its structure, the next question is whether the partially agglutinating nature of Hungarian itself is further predictable from something else about the language. Dressler observes (Dressler 1985) that languages with vowel harmony are always (somewhat) agglutinating and, furthermore, that this is for a reason: vowel harmony is there to make up for the otherwise loose bond between stem and affix. If there is indeed a valid typological implication between the presence of vowel harmony and a degree of agglutination, then the ultimate

✐

✐ ✐

✐

✐ 232

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #256

✐

✐

232

Edith Moravcsik

predictor of some of the morphological profile of Hungarian turns out to be a feature of its phonology.

Appendix: Summary list of generalizations about Hungarian nominal inflections I.

AFFIX MEANINGS

H-1.

(cf. Section 2.1.1)

AFFIX MEANINGS : CATEGORIES

Affixes mark number, case, and the possessor’s person and number on the possessum. AFFIX MEANINGS : SUBCATEGORIES

H-2. a. b. c. II.

AFFIX FORMS

H-3.

(cf. Section 2.1.2)

AFFIX FORM : SKELETON

a. b. c.

All 31 overt affixes have at least one monosyllabic allomorph. 7 affixes have subsyllabic allomorphs. 5 affixes have bisyllabic allomorphs.

a. b.

STEM FORM : SKELETON Stems are minimally monosyllabic and can be many syllables long. Most – but not all – syllabic affix forms are possible root forms.

H-4.

H-5.

The 4 number affixes stand for singular, basic plural, possessive plural, and associative plural. There are 22 affixal cases assumed here (or 17–27 depending on the analysis). The 8 markers for the possessor’s number and person differentiate three persons and two numbers.

AFFIX FORM : ALLOMORPHY: NUMBER OF AFFIXES WITH ALLOMORPHS

Out of the 31 overt affixes, 25 have (non-grammatically-conditioned) allomorphs while 6 are invariant. H-6.

AFFIX FORM : ALLOMORPHY: NUMBER OF ALLOMORPHS PER AFFIX

Except for the two /v/-initial case markers, which have 48 allomorphs each, the number of affix allomorphs per affix ranges from 2 to 7. H-7.

AFFIX FORM : ALLOMORPHY: DIFFERENTIATING FEATURES

Allomorphs of an affix may differ in the following 7 ways:

✐

✐ ✐

✐

✐ 233

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #257

✐

✐

Inflectional morphology in the Hungarian noun phrase

vowels:

consonants:

233

front versus back; if front: rounded versus unrounded; if back: /a/ versus /o/; presence versus absence of an affix-initial pre-consonantal vowel; presence versus absence of an affix-initial pre-vocalic vowel; presence versus absence of /j/; choice of affix-initial consonant.

AFFIX FORM : ALLOMORPHY: GENERALITY OF DIFFERENTIATING

H-8.

FEATURES ACROSS AFFIXES

6 of the 7 difference types recur across affixes. H-9.

AFFIX FORM : ALLOMORPHY: CONSONANTAL VERSUS VOCALIC DIFFERENCES AMONG AFFIX ALLOMORPHS

a. b.

c. d.

The primacy of vocalic over consonantal differences is borne out in four ways: There are more types of vocalic differences (5) than types of consonantal differences (2). Vowel-based differences are more general than consonantal ones: some of the former hold over classes of vowels while each of the latter involves a single consonant segment type. Vocalic differences are more widespread across affixes than consonantal differences: they affect more affixes. The presence of a consonantal difference implies that of a vocalic one. AFFIX FORM : ALLOMORPHY: PREDICTABILITY OF THE EXISTENCE

H-10.

OF ALLOMORPHS

The existence versus non-existence of affix allomorphs is predictable from some affix forms but not from others. AFFIX FORM : ALLOMORPHY: PREDICTABILITY OF THE SHAPE OF

H-11.

ALLOMORPHS

a. b. H-12. a. b.

With one exception, all allomorph sets show a vocalic front/back distinction. There are dependencies among allomorphs with respect to features that differentiate them. STEM FORM : ALLOMORPHY: NUMBER OF STEMS WITH ALLOMORPHS Over 80 % of nouns have no allomorphs. Some pronoun stems show more allomorphy than nominal ones; nouns and numerals show more allomorphy than adjectives.

✐

✐ ✐

✐

✐ 234

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #258

✐

✐

234

Edith Moravcsik

H-13.

STEM FORM : STEM ALLOMORPHY: NUMBER OF ALLOMORPHS PER STEM

The overwhelming majority of nouns with allomorphs have only a single variant in addition to the base. H-14.

STEM FORM : ALLOMORPHY: DIFFERENTIATING FEATURES All stem allomorphs differ in their vowels; some additionally differ in consonants as well. Vowel-related differences: length of stem-final vowel; quality of stem-final vowel; presence or absence of stem-final vowel; length of the vowel in the stem’s final closed syllable; presence of the vowel in the stem’s final closed syllable. Consonant-related differences: presence or absence of stem-final /v/; order of the stem’s last two consonants.

H-15.

STEM FORM : ALLOMORPHY: CONSONANTAL VERSUS VOCALIC DIFFERENCES

a. b. c.

The primacy of vocalic over consonantal differences is borne out in three ways: There are more types of vocalic differences (5) than consonantal differences (4). Vocalic differences are more widespread across stems than of consonantal ones: they affect more stems. No stem allomorphs differ only consonantally, i.e., without a concomitant vocalic difference also present. STEM FORM : ALLOMORPHY: PREDICTABILITY OF THE EXISTENCE OF

H-16.

ALLOMORPHS

The existence versus non-existence of additional stem allomorphs is predictable from some stem allomorph shapes but not from others, with the oblique form being a better predictor of the existence of a distinct base form than vice versa. STEM FORM : ALLOMORPHY: PREDICTABILITY OF THE SHAPE OF

H-17.

ALLOMORPHS

a.

The actual shape of a stem allomorph is in some cases (but not always) predictive of that of the others, with the oblique form once again a better predictor of the base form than vice versa.

✐

✐ ✐

✐

✐ 235

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #259

✐

✐

Inflectional morphology in the Hungarian noun phrase

b.

235

There are also dependencies between one feature of an allomorphic difference and another such feature. STEM FORM : ALLOMORPHY: MONOMORPHEMIC AND INFLECTED

H-18.

STEMS

Inflected stems have no special oblique forms even if they are homonymous with uninflected stems that do. H-19. a.

b. III.

STEM FORM : ALLOMORPHY: SUSPENSION Given a yet uninflected stem S with distinct base and oblique forms, – an inflected stem homonymous with the uninflected stem S may not show the same allomorphic variants; – stem S when used as a quoted word or as a proper name generally (but not always) forgoes the oblique form that it otherwise has; – the use of the distinct oblique form may be optional for stem S even in its normal use. Nouns that have a special oblique form either may or must show this variant even when they occur as the final part of an inflected compound.

ZERO MEANING

H-20.

(cf. Section 2.1.3)

ZERO MEANING

There are no semantically empty nominal affixes or stems. IV. ZERO FORM (cf. Section 2.1.3) H-21.

ZERO FORM : AFFIXES AND STEMS There are no words that contain overt affixes and a zero stem.

H-22.

ZERO FORM : ZERO AS EXCLUSIVE OR ALTERNATIVE EXPONENT

a. b. V.

The singular, the Nominative, and the 3rd person singular possessor on a plural possessum can only be zero-marked. The Accusative and the Genitive occur both overtly-marked and zeromarked.

SYNONYMY

H-23. a.

(cf. Section 2.1.4)

SYNONYMY: AFFIXES , STEMS , AND INFLECTED WORDS There is synonymy – between affix morphemes – between stems – between affixes and stems – between inflected words.

✐

✐ ✐

✐

✐ 236

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #260

✐

✐

236

Edith Moravcsik

b.

There is no synonymy – between an affix and an inflected word, or – between a root and an inflected word.

H-24.

SYNONYMY: ZERO INFLECTION The zero marker is synonymous with overt exponents of the Accusative and of the Genitive.

H-25.

SYNONYMY: INFLECTED WORDS There are instances of synonymous inflected nominals differing in each of the following ways: – the presence versus absence of a morpheme – the choice of affix allomorph, with the stem morpheme remaining the same – the choice of the stem allomorph and the concomitant choice of the affix morpheme – the order of morphemes within the word.

H-26.

SYNONYMY: SEMANTIC SPLIT

a. b. VI.

There are pairs of affix allomorphs that are synonymous when added to some stems but show a semantic contrast when added to other stems. There are types of stem variation that preserve meaning in some cases but are semantically distinctive in others.

HOMONYMY

H-27.

(cf. Section 2.1.5)

HOMONYMY: AFFIXES , STEMS , AND INFLECTED WORDS

a. b. c. d. e.

There are forms that are homonymous between uninflected stem meanings. There are forms that are homonymous between affix meanings. There are forms that are homonymous between a stem meaning and an affix meaning. There are forms that are homonymous between inflected and uninflected stem meanings. There are forms that are homonymous between inflected stem meanings.

H-28.

HOMONYMY: ALL ALLOMORPHS OR ONLY ONE Homonymy may hold over all allomorphs of a morpheme or involve one allomorph only.

H-29.

HOMONYMY: ZERO INFLECTION

Zero marking shows more extensive homonymy than any other affix.

✐

✐ ✐

✐

✐ 237

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #261

✐

✐

Inflectional morphology in the Hungarian noun phrase

VII.

CUMULATIVE EXPONENCE

H-30.

237

(cf. Section 2.1.6)

CUMULATION : AFFIX MEANINGS The following affix meanings cumulate: – position and orientation in spatial case affixes – person and number in possessor marking – association, likeness, and inclusiveness in the associative plural.

H-31. a. b.

CUMULATION : AFFIX MEANINGS : SPATIAL CASES Of the three positions (‘inside’, ‘on the surface of’, and ‘near’), ‘near’ cumulates most completely with orientation. Of the three orientations (‘location’, ‘source’, and ‘destination’), ‘source’ cumulates least with position. CUMULATION : STEM AND AFFIX The only instances of stem-affix portmanteaus are the 1st and 2nd person pronoun’s Accusative and plural.

H-32.

VIII. EXTENDED EXPONENCE (cf. Section 2.1.7) H-33.

EXTENDED EXPONENCE : TYPES Extended exponence occurs in number, case, and person marking.

H-34.

EXTENDED EXPONENCE : NUMBER OF FORMS INVOLVED There are no more than two exponents per meaning.

H-35.

EXTENDED EXPONENCE : OVER THE PHRASE Extended exponence may occur spread over the NP. The possessor’s person and number for which the possessum is inflected may be additionally marked by a free possessor pronoun. In NPs that include a noun as well as a demonstrative, case and number of the noun must be marked both on the noun and on the demonstrative.

a. b. H-36.

EXTENDED EXPONENCE : OPTIONALITY If the components of extended exponence are a free form and a bound form and extended exponence is optional, it is always the free form that is dispensable.

H-37.

EXTENDED EXPONENCE : NON - REDUNDANCY OF PLURAL MARKING

a. b. c.

Some crosslinguistically common redundancies in plural marking do not in fact occur: Numerated nouns are in the singular. Verbs whose subject is a numerated noun are in the singular. Paired body parts may be used in the singular to indicate both members of the pair.

✐

✐ ✐

✐

✐ 238

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #262

✐

✐

238

Edith Moravcsik

d. e.

Generic nouns may be in the singular. While for plural 1st and 2nd person possessors, the plurality of the possessor is marked both on the possessor and on the possessum, for the plural 3rd person possessor, plurality is marked only once: either on the possessor itself (if it is a noun), or on the possessum (if the possessor is a pronoun).

NON - COMPOSITIONALITY

IX. H-38.

(cf. Section 2.1.8)

NON - COMPOSITIONALITY: EXTENT

Inflected words are generally compositional. H-39.

a. b.

NON - COMPOSITIONALITY: CASE When a case affix and a postposition are joint exponents of a case meaning, the case affix involved also occurs as an independent case marker but, with one exception, the postposition does not; the postposition selects a unique cooccurring case affix on the noun, but the case affix may cooccur with several different postpositions. NON - COMPOSITIONALITY: FORM VERSUS MEANING

H-40.

Semantic and formal compositionality do not always go hand-in-hand. The associative plural marker, if analysed as bimorphemic, is formally but not semantically compositional: form properties of the whole are the sum of the form properties of the parts, but the meaning of the whole is not the sum of the meanings of the parts. X.

COOCCURRENCE

H-41.

(cf. Section 2.2.1)

COOCCURRENCE : AFFIX TO AFFIX

a.

b. c.

H-42. a.

Of the four inflection types of number (basic and possessive), associative plural, possessor’s marking, and case, any type can cooccur with any other type. A possessed noun in the plural must be marked with the possessive plural. Apart from some instances of pronominal case marking, more than one exponent of an affix type may not occur in a word. COOCCURRENCE : AFFIX TO STEM : CROSS - CATEGORIAL INFLECTION With the exception of the associative plural, all three affix types are crosscategorial in distribution: they can coocur not only with nouns but with a member of any word class when it fills a nominal slot, such as pronouns, numerals, and adjectives.

✐

✐ ✐

✐

✐ 239

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #263

✐

✐

Inflectional morphology in the Hungarian noun phrase

b.

239

Possessor markers also mark the subject of infinitives and verbs. COOCCURRENCE : AFFIX TO STEM : CROSS - CATEGORIAL INFLECTION :

H-43.

VARYING EXPONENCE

Cross-categorially applicable inflectional morphemes may have crosscategorially differing exponents. COOCCURRENCE : AFFIX ALLOMORPHS : GRAMMATICAL VERSUS

H-44.

PHONOLOGICAL CONDITIONING

The only grammatically conditioned affix variants are the plural, which differs in shape depending on whether the noun is possessed or not, and the 3rd person singular and plural possessor markers on the possessum, which differ with the possessum being singular or plural. H-45.

COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM No affix allomorph is specific only to a single stem type.

H-46.

COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM : ALTERNATIVES

Generally, every stem selects a single allomorph of an affix; but there are some cases of a stem alternatively occurring with multiple allomorphs of an affix. COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM :

H-47.

MUTUAL DEPENDENCIES AMONG AFFIXES

The choice from among allomorphs that differ in harmonizing vowel quality or the presence/absence of the initial preconsonantal vowel is generally – but not always – determined the same way across all affixes. COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY THE STEM :

H-48.

UNIDIRECTIONAL DEPENDENCIES AMONG AFFIXES

a.

b.

c.

There are dependencies among affixes regarding the choice of their allomorphs. For example: If a stem selects the vowel-initial allomorph of the Accusative affix, it will also select the vowel-initial allomorph of the SG1, SG2, and PL2 possessive affixes in the singular-possessum paradigm; but not vice versa. If a stem selects the vowel-initial allomorph of the SG1, SG2, and PL2 possessive affixes in the singular-possessum paradigm, it will select the vowel-initial allomorph of the basic plural; but not vice versa. If a stem cooccurs with an affix that has allomorphs differentiated by /a/ versus /o/ and the stem selects the /a/-containing allomorph, then that same stem will select a vowel-initial allomorph in the Accusative affix; but not vice versa.

✐

✐ ✐

✐

✐ 240

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #264

✐

✐

240

Edith Moravcsik

H-49.

COOCCURRENCE : AFFIX ALLOMORPHS : PREDICTING AFFIX ALLOMORPH FROM STEM TYPE

If a stem’s base form ends in a CV:C and its oblique form ends in the corresponding CVC, then the stem takes the /a/-containing allomorph of the Accusative affix. COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY ANOTHER AFFIX

H-50.

VERSUS CHOICE BY THE UNINFLECTED STEM

Inflected words do not always behave the same way as their homonymous non-inflected counterparts in choosing among allomorphs of an additional affix. After back-vowel stems that already include an inflectional affix, any additional inflection must occur in its /a/-containing form if it has /a/- versus /o/-containing allomorphs. COOCCURRENCE : AFFIX ALLOMORPHS : CHOICE BY ANOTHER AFFIX :

H-51.

DIRECTION OF CONDITIONING

As a rule, affixes condition allomorphs of cooccurring affixes to the right. The possessor markers are the only affixes that can select affix allomorphs to the left. H-52.

COOCCURRENCE : STEM ALLOMORPHS : GENERALITY No stem allomorph is usable with a single affix only.

H-53.

COOCCURRENCE : STEM ALLOMORPHS : CHOICE BY THE AFFIX : ALTERNATIVES

Generally, every affix selects a single allomorph of a stem; but there are some cases of an affix occurring with alternative stem allomorphs. COOCCURRENCE : STEM ALLOMORPHS : BASE - ALTERING AND NON -

H-54.

BASE - ALTERING AFFIXES

a.

b.

H-55.

Eleven affixes require the non-base form of all nouns that have such: the basic and possessive plurals, three cases: the Accusative, the TemporalDistributive, the Sociative, and the six possessor markers on the singular possessum. All other affixes take the base form of some stems and the oblique form of others. COOCCURRENCE : CORRELATION OF AFFIX ALLOMORPHY AND BASE ALTERING CAPABILITY

The affixes that require the oblique form of all allomorphic nouns regardless of their inflectional class are among the ones that have the largest numbers of allomorphs themselves.

✐

✐ ✐

✐

✐ 241

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #265

✐

✐

Inflectional morphology in the Hungarian noun phrase

XI.

LINEAR ORDER

241

(cf. Section 2.2.2)

H-56.

LINEAR ORDER : PRECEDENCE : INVARIANCE The order of morphemes in inflected words is invariant except for certain pronouns where case and possessor may occur in either order.

H-57.

LINEAR ORDER : PRECEDENCE : STEM AND AFFIX All nominal inflectional affixes follow the stem.

H-58.

LINEAR ORDER : PRECEDENCE : AFFIX AND AFFIX

a. b.

LINEAR ORDER : ADJACENCE Inflectional affixes form a continuous sequence except for the possessumderiving affix, which, along with a subsequent optional plural marker, may occur immediately preceding the case affix.

H-59.

XII.

Number and possessor marking precede case on noun stems. Number precedes possessor marking but the associative plural follows it.

BONDING

H-60.

(cf. Section 2.2.3)

BONDING

a. b. c.

All affixes occur only as bound form except for some case affixes which, in possessed form, may also occur as free forms. All postpositions occur as bound stems when marking personal pronouns. Postpositions, adverbs, and case affixes are not always clearly distinguishable.

Acknowledgments A preliminary version of this paper appeared in The Noun Phrase Sketch Book, EUROTYP Working Papers VII/22, 1994. I am grateful to members of EUROTYP’s Group 7 and especially to Greville Corbett and David Gil for corrections and feedback, to Andrew Carstairs-McCarthy for his very helpful comments, and to Bernard Comrie for an extremely thorough and insightful running commentary.

✐

✐ ✐

✐

✐ 242

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #266

✐

✐

242

Edith Moravcsik

Notes 1. Unless noted otherwise, Hungarian data reflect my own native, standard, middle-class, relatively-current Budapest dialect and are given in standard Hungarian (near-phonemic) orthography. Gender-wise neutral forms will be glossed as ‘he’, ‘him’, or ‘his’. Letters generally stand for their IPA value with the following exceptions: ü front high rounded ö front mid rounded ny palatal nasal ty voiceless palatal stop gy voiced palatal stop sz voiceless alveolar fricative z voiced alveolar fricative s voiceless palatal fricative zs voiced palatal fricative cs voiceless palatal affricate j and ly both stand for the same voiced palatal glide. Accents consisting of dashes rather than dots (as in í, o˝ , ú, u˝ ) indicate the long version of the corresponding short vowel; but é is not only longer but also higher than e and á is not only longer than a but also unrounded (Vago 1980: 1–3). 2. For comprehensive analyses of Hungarian NP-internal morphology, see Ferenc Papp’s monograph on Hungarian nominal inflection written in Hungarian (Papp 1975) and two English-language discussions: Chapter 4 of Robert Vago’s Hungarian (morpho-)phonology (Vago 1980) and Chapter 3 of Daniel Abondolo’s Hungarian morphology (Abondolo 1988). Brief general overviews of Hungarian inflectional morphology are provided by Kálmán 1985 and Kiefer 2000. Additional relevant references on Hungarian are as follows: on nominal inflection: Tompa 1968: 174–210; on case: Antal 1961; Keresztes 1975; Korponay 1986a, 1986b; Marácz 1989; on adjectivally-governed case in Hungarian and four other languages: Laczik et al. 1989; on verb agreement: Moravcsik 1984; de Groot 1989: esp. Chapter 3; on compounds: Kiefer 1990, 1992, 1993; on Hungarian morphology in general: Elekfi 1994; Kiefer (ed.) 2000; on Hungarian NP structure: Szabolcsi 1994; É. Kiss 2000; on Hungarian morphosyntax in general: Horvath 1986; É. Kiss 1987; Szabolcsi 1992; Kiefer & É. Kiss (ed.) 1994. Kiefer (ed.) 2000 is a comprehensive work (over 1000 pages) that includes 15 studies on derivation, inflection, and compounding in Hungarian, three of which – Bartos 2000, Kiefer 2000, and Rebrus 2000 – focus more specifically on nominal inflection. Although occasional references will be made to these studies, the recency of this publication has prevented me from fully integrating the data and analyses offered in the volume with the content of this chapter. 3. Cf. Matthews 1974: 147–150; Carstairs 1984: 19; Carstairs-McCarthy 1987, 1992: 193– 214; Lehmann 1984: 248; Mayerthaler 1988. For a comprehensive analysis of symbolization patterns within morphology, see Mel’ˇcuk 1973. 4. David Gil called my attention to an additional morphological phenomenon in Hungarian which might be analysed as involving inflection. This is the total reduplication of nu-

✐

✐ ✐

✐

✐ 243

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #267

✐

✐

Inflectional morphology in the Hungarian noun phrase

5.

6. 7.

8. 9.

243

merals to express distributivity, such as egy-egy alma one-one apple ‘one each’ in Adtam nekik egy-egy almát ‘I gave them an apple each’. However, while partial reduplication (not found in Hungarian) does resemble regular inflection, total reduplication seems a sufficiently different pattern to justify its exclusion from inflection. Considering total reduplication a type of inflection would require the identification of a stem part and an affix part in the reduplicated structure and this would be problematic for at least two reasons. First, I cannot think of ways to determine which of the two egy-s in egy-egy one-one is the stem and which is the affix. Second, regardless of how this problem is solved, when total reduplication is applied to complex numerals such as husz-on-három twenty-SUPESS-three (as in huszonhárom-huszonhárom ‘23 each’), the inflectional analysis would call for assigning affixal status to a polymorphemic word. The case affix Formal -kép -képpen ‘by way of’ differs from other allomorphic affixes. First, the two variant forms are in free, rather than conditioned, variation. Second, the difference may be morphological, rather than phonological: -képpen may be analysed as bimorphemic consisting of -kép ‘picture, image’ followed by -en, which may be the Superessive or the de-adjectival manner-adverbial affix. Note also that the variant end of the affix is the “outer” one rather than the one adjacent to the stem as in other cases. For psycholinguistic experiments involving inflectional homonymy in Hungarian, see Gergely & Pléh 1994. For ‘inside’, ‘on’, ‘near’, etc. Kilby (1981) uses the term “dimensionality” and Kibrik (1994) calls them “localization”. For location, source, and destination, Kilby uses the term “directionality” and Kibrik uses “direction”. At the very real risk of further compounding terminological proliferation and confusion, I will use two different terms: “position” and “orientation”, respectively. The reason is that they seem to be somewhat more descriptive of the intended content than either of the other two pairs of terms. It would be possible to view the plural forms as bimorphemic but the resulting analysis would be complex. The demonstrative agrees with the noun not only in affixal but also in (some) postpositional cases (cf. Section 3.1.7). This fact leads to an interesting effability problem (cf. Bartos 2000: 702–704). The quandary arises when the postposition applies to a set of conjoined nouns such as ‘the house and the yard’. If the postposition applies separately to each noun, there is no problem: (i)

a ház mellett és a kert mellett the house next:to and the yard next:to ‘next to the house and next to the yard’

(ii)

emellett a ház mellett és emellett a kert mellett next:to:this the house next:to and next:to:this the yard next:to ‘next to this house and next to this yard’

But suppose the postposition applies to a single entity made up of the referents of two nouns, such as in ‘next to the house-and-yard’ or ‘between the house and yard’. As long as the nouns are not accompanied by demonstratives, there is no problem: the postposition simply follows the conjoined set:

✐

✐ ✐

✐

✐ 244

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #268

✐

✐

244

Edith Moravcsik (iii)

a ház és a kert között the house and the yard between ‘between the house and the yard’

However, if one or both nouns are accompanied by a demonstrative – such as in ‘between this house and the garden’ – a conflict arises. Semantics requires that the postposition should occur only after everything within its scope – i.e., both conjuncts – have been pronounced. Syntax, however, prescribes the occurrence of the postposition on the demonstrative as well. Both desiderata cannot be fulfilled at the same time. While (iv) correctly places the postposition for scope, it violates the case agreement rule between noun and demonstrative and thus ends up ungrammatical: (iv)

*ez a ház és a kert között this the house and the yard between

(v), on the other hand, complies with case agreement but places the postposition on the demonstrative which does not exhaust its scope: (v)

*eközött a ház és a kert között between:this the house and the yard between

As a result, speakers tend to stumble over the construction or avoid it altogether. 10. This generalization is exceptionless for the rule governing the /v/-initial affixes but it has exceptions with respect to front-back and roundedness harmony. Regarding front-back harmony, Vago (1980: 11) points out that there are nouns in Hungarian where the choice of vowel-harmonic affix allomorphs differs from one affix to another. The noun derék ‘waist’ takes back-vowel allomorphs of some affixes (e.g., derek-am ‘my waist’, derek-at ‘waist (ACC)’), but accepts either front or back-vowel allomorphs in other cases: e.g., derék-ban or derék-ben ‘in waist’, derék-tól or derék-t˝ol ‘from waist’. Compare this with the noun fenék ‘bottom’ which, even though it is very similar in phonological makeup to derék, is regular and takes only front-vowel allomorphs of all affixes. Rounding harmony is not consistent across affixes, either; for example, the noun öv ‘belt’ chooses the rounded form of the Superessive: öv-ön, but the unrounded form of the Accusative: öv-et (*öv-öt; but török-öt ‘Turk (ACC)’). 11. Budapesti-m ‘my person from Budapest’ is given erroneously as Budapesti-em in Dressler 1985: 11, which is why the possessive-plural hierarchy remained hidden. 12. An exception is híd ‘bridge’: hid-at, hid-ak, but both hid-ai and híd-jai. 13. There are two questions that arise in connection with the paradigm of the inflected infinitive. First, why is the infinitival marker -ni not preserved intact throughout the paradigm yielding *jár-ni-m, *jár-ni-d, etc., instead of jár-n-om, jár-n-od? Second, granted the i of -ni gets lost, why does it not get lost in the 3rd persons as well, yielding *jár-n-a and *jár-n-uk, instead of jár-ni-a and jár-ni-uk? Tentative answers may be the following. First, since all verb stems in Hungarian are consonant-final, chopping off the i of -ni may be an attempt to bring the resulting stem in line with this prevailing pattern. Second, if the i were to disappear in the 3rd person singular, homonymy would arise: the resulting form jár-n-a is also a conditional form meaning ‘he would go’. Given, however, that homonymy does occur in other areas of the Hungarian verbal paradigm, this explanation is merely suggestive.

✐

✐ ✐

✐

✐ 245

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #269

✐

✐

Inflectional morphology in the Hungarian noun phrase

245

14. For a diachronic study of such unstable stems, see É. Kiss 1993. 15. Exceptions are id˝o ‘time’, which shows the base form in the basic plural and the Accusative: id˝o-k and id˝o-t, but an oblique stem in the possessed form SG3: ide-je; and híd ‘bridge’, which uses the short-vowel oblique form in the basic plural and the Accusative: hid-ak, hid-at, but either the oblique form or the base form in the possessed form S3: hid-a or híd-ja. Another exception is szó ‘word’, which has two alternative plurals: one using its oblique form szavak, the other using the base form of the stem szók. As in the case of German Worte and Wörter, the first refers to connected words in a text and the second refers to isolated words as in a dictionary. 16. The distinctness of the relation-registering versus the relation-coding types is actually somewhat tenuous. For example, one could consider a genitive marker an inflection of the coding sort since it specifies the genitive relation within the set of case relations in general; but one could also consider it as a mere relation-indicating kind since it glosses over the various semantic types of the genitive relation. 17. For a non-implicational typology of inflected adpositions, which simply registers the fact that in some languages, Hungarian among them, person-number-inflected adpositions may optionally be associated with a free pronoun form as well, while in other languages, such as in Irish, this is not possible, see Marácz 1989: 399–401. 18. Crosslinguistic discussions of vowel harmony (e.g., Lightner 1965; Aoki 1968; Vago 1973; Clements 1980; Goldsmith 1985; Ringen 1988) shed little light on the issue since, if they are typologically oriented at all, they focus on the typology of vowel harmony systems stopping short of considering the typology of languages with vowel harmony systems; i.e., what other language properties the existence and particular nature of a vowel harmony system may be implicationally related to. 19. Bernard Comrie has pointed out (p.c.) that counterexamples to U-11 may be found in Amele and other Papuan languages, where there is a single zero stem for the verbs ‘give’ and ‘strike’; and to U-12 in Yuman and some Nilotic languages. 20. Çocuklar1 is actually three-ways ambiguous, with the plural affix marking either the plurality of the possessum (‘his children’) or that of the possessor (‘their child’) or both (‘their children’) (Lewis 1967: 40). 21. I wish to thank Andrew Carstairs-McCarthy for directing my attention to these data as potentially problematic for his two paradigm structure constraints.

✐

✐ ✐

✐

✐ 246

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #270

✐

✐

246

Edith Moravcsik

References Abondolo, Daniel Mario 1988 Hungarian inflectional morphology. Budapest: Akadémiai Kiadó. Ackerman, Farrell 1987 “Pronominal incorporation: the case of prefixal preverbs”, in: Kenesei (ed.), 213–260. Alberti, Gábor & István Kenesei (eds.) 2000 Approaches to Hungarian. Papers from the Pécs conference. Szeged: JATEPress. Antal, László 1961 A magyar esetrendszer. [The Hungarian case system.] Budapest: Akadémiai Kiadó. Aoki, Haruo 1968 “Toward a typology of vowel harmony”, International Journal of American Linguistics 34: 142–145. Bartos, Huba 2000 “Az inflexiós jelenségek szintaktikai háttere” [The syntactic background of inflectional phenomena], in: Kiefer (ed.), 653-762. Bechert, Johannes 1990 “The structure of the noun in European languages”, in Bechert, Bernini, & Buridant (eds.), 115–140. Bechert, Johannes, Giuliano Bernini & Claude Buridant (eds.) 1990 Toward a typology of European languages. Berlin: Mouton de Gruyter. Blake, Barry 1994 Case. Cambridge: Cambridge University Press. Booij, Geert 1993 “Against split morphology”, in: Booij & van Marle (eds.), 27–49. Booij, Geert & Jaap van Marle (eds.) 1993 Yearbook of morphology. Dordrecht: Kluwer. Bybee, Joan L. 1985a Morphology. A study of the relation between meaning and form. Amsterdam: Benjamins. 1985b “Diagrammatic iconicity in stem-inflection relations”, in: Haiman (ed.), 11–47. Carstairs(-McCarthy), Andrew 1984 Constraints on allomorphy in inflexion. Bloomington, IN: Indiana University Linguistics Club. 1987 Allomorphy in inflexion. London: Croom Helm. 1992 Current morphology. London: Routledge. 1994 “Inflection classes, gender, and the Principle of Contrast”, Language 70: 737–788. Carstairs-McCarthy, Andrew & Joseph P. Stemberger 1988 “A processing constraint on inflectional homonymy”, Linguistics 26: 601–617. Clements, George N. 1980 Vowel harmony in nonlinear generative phonology: An autosegmental model. Bloomington, IN: Indiana University Linguistics Club. Comrie, Bernard 1980 “The order of case and possessive suffixes in Uralic languages: An approach to the comparative-historical problem”, Lingua Posnaniensis 23: 81–86. Corbett, Greville G. 1991 Gender. Cambridge: Cambridge University Press. 1992 “A typology of number systems”, EUROTYP Working Papers VII/15. 2000 Number. Cambridge: Cambridge University Press.

✐

✐ ✐

✐

✐ 247

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #271

✐

✐

Inflectional morphology in the Hungarian noun phrase

247

Corbett, Greville G. & Marianne Mithun 1996 “Associative forms in a typology of number systems: Evidence from Yup’ik”, Journal of Linguistics 32: 1–7. Cutler, Anne, John A. Hawkins, & Gary Gilligan 1985 “The suffixing preference: A processing explanation”, Linguistics 23: 723–758. Cysouw, Michael 2001 The paradigmatic structure of person marking. Catholic University of Nijmegen doctoral dissertation. Danièl’, Mikhail Aleksandroviˇc 2000 Tipologija associativnoj množestvennosti. [The typology of associative plurals.] Moscow University doctoral dissertation. Décsy, Gyula 1973 Die linguistische Struktur Europas. Wiesbaden: Harrassowitz. Diessel, Holger 1999 Demonstratives. Form, function, and grammaticalization. Amsterdam: Benjamins. Dressler, Wolfgang U. 1985 “Typological aspects of natural morphology”, Wiener Linguistische Gazette 33-35: 326. Eckman, Fred R., Edith A. Moravcsik, & Jessica R. Wirth (eds.) 1986 Markedness. New York: Plenum. É. Kiss, Katalin 1987 Configurationality in Hungarian. Budapest: Akadémiai Kiadó. 1994 “Sentence structure and word order”, in: Kiefer & É. Kiss (eds.), 1–90. 2000 “The Hungarian noun phrase is like the English noun phrase”, in: Alberti & Kenesei (eds.), 119–149. Elekfi, László 1993 “Eltün˝oben van-e a magyar birtokos határozó?” [Is the Hungarian possessive dative marker in the process of disappearing?], in: Horváth & Ladányi (eds.), 35–43. 1994 Dictionary of Hungarian inflections. Budapest: MTA Nyelvtudományi Intézete. Fisiak, Jacek (ed.) 1980 Historical morphology. The Hague: Mouton. Forchheimer, Paul 1953 The category of person in language. Berlin: de Gruyter. Gergely, György & Csaba Pléh 1994 “Lexical processing in an agglutinative language and the organization of the lexicon”, Folia Linguistica 28: 176–204. Gil, David 1987 “Definiteness, noun-phrase configurationality, and the count-mass distinction”, in: Reuland & ter Meulen (eds.), 254–269. 1991 “Nouns, verbs, and quantification”, EUROTYP Working Papers X/l. 1994a “Adjectives without nouns: an e-mail conversation”. Conversations on noun phrases: 31–48. EUROTYP Working Papers VII/20. 1994b “Association”. Paper given at the meeting of the Association for Linguistic Typology, Konstanz. Goldsmith, John 1985 “Vowel harmony in Khalkha Mongolian, Yaka, Finnish, and Hungarian”, Phonology Yearbook 2: 253–275. Greenberg, Joseph H. 1963 “Some universals of grammar with particular reference to the order of meaningful elements”, in: Greenberg (ed.), 73–113. 1966 Language universals. The Hague: Mouton.

✐

✐ ✐

✐

✐ 248

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #272

✐

✐

248

Edith Moravcsik

1978 “Generalizations about numeral systems”, in: Greenberg (ed.), 249–295. Greenberg, Joseph H. (ed.), 1963 Universals of language. Cambridge: MIT Press. 1978 Universals of human language, Volume 3, Word structure. Stanford, CA: Stanford University Press. Groot, Casper de 1989 Predicate structure in a functional grammar of Hungarian. Dordrecht: Foris. Groot, Casper de & Machiel J. Limburg 1986 “Pronominal elements: Diachrony, typology and formalization in Functional Grammar”, Working Papers in Functional Grammar 12. Amsterdam: University of Amsterdam. Haarmann, Harald 1976 Grundzüge der Sprachtypologie. Methodik, Empirie und Systematik der Sprachen Europas. Stuttgart: W. Kohlhammer. Haiman, John (ed.) 1985 Iconicity in syntax. Amsterdam: Benjamins. Hajdú, Péter 1963 The Samoyed peoples and languages. Bloomington, IN & The Hague: Indiana University & Mouton. Hall, Christopher 1988 “Integrating diachronic and processing principles in explaining the suffixing preference”, in: Hawkins (ed.), 321–349. Harms, Robert T. 1964 Finnish structural sketch. The Hague: Mouton. Haspelmath, Martin 1999 “The agglutination hypothesis: A belated empirical investigation”, paper given at the Third Meeting of the Association for Linguistic Typology, Amsterdam. Hawkins, John A. & Gary Gilligan 1988 “Prefixing and suffixing universals in relation to basic word order”, Lingua 74: 219– 259. Hawkins, John A. (ed.) 1988 Explaining language universals. Oxford: Basil Blackwell. Hetzron, Robert 1980 “Universals of human language: A Monteverdian quartet in forty-six movements”, Lingua 50: 249–294. Horvath, Julia 1986 Focus in the theory of grammar and the syntax of Hungarian. Dordrecht: Foris. Horváth, Katalín & Mária Ladányi (eds.) 1993 Állapot és történet – szinkrónia és diakrónia – viszonya a nyelvben. [The relationship of state and history – synchrony and diachrony – in language.] Budapest: ELTE. Johanson, Lars 1973 “Sprachbau und Inhaltssyntax am Beispiel des Türkischen”, Orientalia Suecana 22: 82–106. Kahr, Joan Casper 1976 “The renewal of case morphology: Sources and constraints”, Working Papers on Language Universals 20: 107–151. Kálmán, László 1985 “Inflectional morphology”, in: Kenesei (ed.), 247–262. Kefer, Michel & Johan van der Auwera (eds) 1992 Meaning and grammar. Cross-linguistic perspectives. Berlin: Mouton de Gruyter.

✐

✐ ✐

✐

✐ 249

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #273

✐

✐

Inflectional morphology in the Hungarian noun phrase

249

Kenesei, István (ed.) 1985 Approaches to Hungarian, Vol. 1. Szeged: JATE. 1987 Approaches to Hungarian, Vol. 2. Szeged: JATE. Keresztes, Kálmán 1975 Hungarian postpositions vs. English prepositions: A contrastive study. The HungarianEnglish Contrastive Linguistics Project, Working Paper 7. Budapest: Linguistic Institute of the Hungarian Academy of Sciences and Center for Applied Linguistics. Kiefer, Ferenc 1985 “The possessive in Hungarian: A problem for natural morphology”, Acta Linguistica Academiae Scientiarum Hungaricae 35: 85–116. 1987 “The cases of Hungarian nouns”, Acta Linguistica Academiae Scientiarum Hungaricae 37: 93–101. 1990 “Noun incorporation in Hungarian”, Acta Linguistica Academiae Scientiarum Hungaricae 40, 1–2. 1992 “Compounding in Hungarian”, Rivista di Linguistica 4: 61–78. 1993 “Thematic roles and compounds”. Folia Linguistica 27: 45–55. Kiefer, Ferenc (ed.) 2000 Structurális magyar nyelvtan. 3. kötet: Morfológia. [Structural grammar of Hungarian. 3rd volume: Morphology.] Budapest: Akadémiai Kiadó. Kiefer, Ferenc & Katalin É. Kiss (eds.) 1994 The syntactic structure of Hungarian. (Syntax and Semantics 27.) New York: Academic Press. Kiefer, Ferenc & Mária Ladányi 2000 “Morfoszintaktikailag semleges képzések” [Derivations that are morphosyntactically neutral], in: Kiefer (ed.), 165–214. Kilby, David 1981 “On case markers”, Lingua 54: 101–133. Kim, John J., Gary F. Marcus, Steven Pinker, Michelle Hollander, & Marie Coppola 1994 “Sensitivity of children’s inflection to grammatical structure”, Journal of Child Language 21: 173–209. Kiss, Gábor 1993 “ ‘Sátort’ vagy ‘sátrat’? a nyelvi változás dinamizmusáról” [‘Sátort’ or ‘sátrat’? About the dynamism of linguistic change], in: Horváth & Ladányi (eds.), 105–114. Korchmáros, Valéria 1995 “Az -ék: többesjel!” [-ék: a plural marker!], Néprajz és Nyelvtudomány [Folklore and Linguistics] 36: 295-307. Korhonen, Mikko 1979 “Entwicklungstendenzen des finnisch-ugrischen Kasussystems”, Finnisch-Ugrische Forschungen 43: 1–21. Korponay, Béla 1986a Outlines of a Hungarian-English case grammar. Budapest: Tankönyvkiadó. 1986b Postpositions. Hungarian Studies in English XIX. Debrecen: Kossuth Lajos University. Laczik, Mária, Árpád Mihalovics, Tibor Szabó, & Gábor Székely (eds.) 1989 Melléknévi vonzatok és mellékneves szószerkezetek öt nyelven (magyar, angol, francia, német, orosz). [Adjectival complements and adjectival constructions in five languages (Hungarian, English, French, German, Russian).] Budapest: Tankönyvkiadó. Laczkó, Tibor 2000 “Az ige argumentumszerkezetét meg˝orz˝o f˝onévképzés” [Noun derivation that preserves the argument structure of the verb], in: Kiefer (ed.), 293 – 407.

✐

✐ ✐

✐

✐ 250

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #274

✐

✐

250

Edith Moravcsik

Lehmann, Christian 1984 Der Relativsatz. Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: Gunter Narr. 1985 “Latin case relations in typological perspective”, in: Touratier (ed.), 81–104. Lewis, G.L. 1967 Turkish grammar. Oxford: Clarendon Press. Lightner, Theodore 1965 “On the description of consonant and vowel harmony”, Word 19: 376–387. Lotz, John 1968 “Grammatical derivability. The pronominal suffix -é ‘that of . . .’ in Hungarian”, Lingua 21: 627–637. Lyons, Christopher 1986 “The syntax of English genitive constructions”, Journal of Linguistics 22: 123–143. Manzelli, Gianguido 1990 “Possessive adnominal modifiers”, in Bechert, Bernini, & Buridant (eds.), 63–111. Marácz, László 1989 Asymmetries in Hungarian. Groningen: Rijksuniversiteit doctoral dissertation. Marcus, Gary, Ursula Brinkmann, Harald Clahsen, Richard Wiese, Andreas Woest, & Steven Pinker 1993 “German inflection: The exception that proves the rule”, Occasional Paper No. 47, MIT: Center for Cognitive Science. Matthews, Peter H. 1974 Morphology. An introduction to the theory of word-structure. Cambridge: Cambridge University Press. Mayerthaler, Willi 1988 Morphological naturalness. Ann Arbor, MI: Karoma. Mel’ˇcuk, Igor A. 1973 “The structure of linguistic signs and possible formal-semantic relations between them”, in: Rey-Debore (ed.), 103–135. Moravcsik, Edith A. 1984 “On grammatical classes: The case of ‘definite’ objects in Hungarian”, Working Papers in Linguistics (Honolulu: University of Hawaii) 15: 75–107. 1994 “Associative plurals”, Paper given at the meeting of the Association for Linguistic Typology, Konstanz. (Revised version given at the Conference on Functionalism, Summer of 1995, Albuquerque, NM.) 1997 “Parts and wholes in the Hungarian noun phrase – a typological study”, in: Palek (ed.), 307–324. 2001 “On the nouniness of Hungarian adjectives”, in: Schaner-Wolles, Rennison, & Neubarth (eds.), 337–346. Moravcsik, Edith A. & Jessica R. Wirth 1988 “Markedness – an overview”, in: Eckman, Moravcsik, & Wirth (eds.), 1–11. Moreno, Juan Carlos 1992 “Polarization as a universal of linguistic organization”, in: Kefer & van der Auwera (eds.), 123–136. Nichols, Johanna 1987 “Suffix ordering in Proto-Uralic”, Lingua 32: 227-238. 1986 “Head-marking and dependent-marking grammar”, Language 62: 56–119. 1992 Linguistic diversity in space and time. Chicago: The University of Chicago Press. Palek, B. (ed.) 1997 Proceedings of LP ’96. Prague: Charles University Press.

✐

✐ ✐

✐

✐ 251

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #275

✐

✐

Inflectional morphology in the Hungarian noun phrase

251

Papp, Ferenc 1975

A magyar f˝onév paradigmatikus rendszere. [The paradigmatic system of the Hungarian noun.] Budapest: Akadémiai Kiadó. Payne, John & Erika Chisarik 2000 “Demonstrative constructions in Hungarian”, in: Alberti & Kenesei (eds.), 170 – 198. Plank, Frans 1980 “Encoding grammatical relations: Acceptable and unacceptable non-distinctness”, in: Fisiak (ed.), 289–325. 1986 “Paradigm size, morphological typology, and universal economy”, Folia Linguistica 20: 29–48. 1991a “On abundance and scantiness in inflection: A typological prelude”, in: Plank (ed.), 1–39. 1991b “Inflection and derivation”, EUROTYP Working Papers VII/10. 1991c “From cases to adpositions”, EUROTYP Working Papers VII/13. 1992 “Possessives and the distinction between determiners and modifiers (with special reference to German)”, Journal of Linguistics 28: 453–468. 1995 “(Re-)introducing Suffixaufnahme”, in: Plank (ed.), 3–110. 1998 “The co-variation of phonology with morphology and syntax: A hopeful history”, Linguistic Typology 2: 195–230. 1999 “Split morphology: How agglutination and flexion mix”, Linguistic Typology 3: 279– 340. Plank, Frans (ed.) 1991 Paradigms. The economy of inflection. Berlin: Mouton de Gruyter. 1995 Double case. Agreement by Suffixaufnahme. Oxford: Oxford University Press. Poppe, Nicholas 1963 Tatar manual. Bloomington, IN & The Hague: Indiana University Press & Mouton. Radics, Katalín c. 1980 A személyegyeztetés tipológiája és története. [The typology and history of person agreement.] Unpublished manuscript. Rebrus, Péter 2000 “Morfofonológiai jelenségek” [Morphophonological phenomena], in: Kiefer (ed.), 763–947. Reuland, E.J. & A.G.B. ter Meulen (eds.) 1987 The representation of (in)definiteness. Cambridge: MIT Press. Rey-Debore, J. (ed.) 1973 Recherches sur les systemes significants. Symposium de Varsovie 1968. The Hague: Mouton. Rijkhoff, Jan 1990 “Explaining word order in the noun phrase”, Linguistics 28: 5–42. 1992 The noun phrase. A typological study of its form and structure. University of Amsterdam doctoral dissertation. To appear at Oxford University Press. Ringen, Catherine O. 1988 Vowel harmony: Theoretical implications. New York: Garland. Schaner-Wolles, Chris, John Rennison, & Friedrich Neubarth (eds.) 2001 Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60th birthday. Torino: Rosenberg & Sellier. Skaliˇcka, Vladimir 1979a “Zur ungarischen Grammatik”, in: Skaliˇcka 1979d, 59–125. 1979b “Über die Typologie der Bantusprachen”, in: Skaliˇcka 1979d, 198–237. 1979c “Ein ‘typologisches Konstrukt’ ”, in: Skaliˇcka 1979d, 335–341. 1979d Typologische Studien. Braunschweig: Vieweg.

✐

✐ ✐

✐

✐ 252

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #276

✐

✐

252

Edith Moravcsik

Smith-Stark, T. Cedric 1974 “The plurality split”, in: Papers from the Tenth Regional Meeting, Chicago Linguistic Society, Chicago, IL: Chicago Linguistic Society, 657–671. Szabolcsi, Anna 1992 A birtokos szerkezet és az egzisztenciális mondat. [The possessive construction and the existential sentence.] Budapest: Akadémiai Kiadó. 1994 “The noun phrase”, in: Kiefer (ed.), 179–274. Tompa, József 1968 Ungarische Grammatik. The Hague: Mouton. Touratier, Christian (ed.) 1985 Syntaxe et latin. Aix en Provence: Université de Provence. Tovar, Antonio 1957 The Basque language. Philadelphia: University of Pennsylvania Press. Underhill, Robert 1976 Turkish grammar. Cambridge, MA: MIT Press. Vago, Robert M. 1973 “Abstract vowel harmony in Uralic and Altaic languages”, Language 49: 579–605. 1980 The sound pattern of Hungarian. Washington, DC: Georgetown University Press. Walter, Heribert 1981 Studien zur Nomen-Verb-Distinktion aus typologischer Sicht. (Structura, Band 13.) München: Wilhelm Fink. Wurzel, Wolfgang U. 1989 Inflectional morphology and naturalness. Dordrecht: Kluwer.

✐

✐ ✐

✐

✐ 253

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #277

✐

✐

Frans Plank

The selective elaboration of nominal or pronominal inflection

1. Variable extents and domains of inflection Languages with inflection may differ in HOW MUCH of it they have and WHERE they have whatever they have of it. Some inflecting languages have few inflectional categories, possibly only person and number, while others have many, possibly including case, state, possession, gender or class, animacy, topicality, switch reference, definiteness, specificity, politeness or formality, distance, location, direction or orientation, comparison, tense, aspect, mood, evidentiality, negation, diathesis, inversion, valency, finiteness, or a few others. While some of these inflectional categories are invariably realized by only a few terms in all languages which share them, others vary considerably. Thus, while switch reference, definiteness, or inversion rarely get beyond the minimum of two terms (same subject, different subject, with different subject sometimes differentiated further; definite, indefinite, with definite sometimes differentiated further; direct, inverse), case ranges from two terms to about two dozen (with case systems arguably expanding along a rather limited number of dimensions).1 After categories and terms, the exponents expressing them are a third parameter for quantitative variation: a given term or term bundle (say, nominative plural) may be expressed by only a single exponent in one language and by several synonymous ones in another; and the exponents of two or more terms or term bundles (say, nominative plural and genitive singular) may be distinct in one language and homonymous in another.2 Some languages have inflection almost everywhere: on verbs, auxiliaries, nouns, pronouns, articles, demonstratives, quantifiers, adjectives, adverbs, and even on adpositions and conjunctions, or also on units other than words. Others practise moderation and limit inflection, for example, to verbal as opposed to nominal words, to heads as opposed to dependents or the other way round, to words as opposed to phrases, to words at the margin of phrases, especially their end, as opposed to phrase-internal words, or to special constituents whose sole purpose is to collect all inflections of a clause. In many ways cross-linguistic variation in the extents and domains of inflection is systematic rather than random: among the thousands of choices which languages take in these respects a good deal are contingent on others.3 Continuing the search for

✐

✐ ✐

✐

✐ 254

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #278

✐

✐

254

Frans Plank

generalizations in this area, what this chapter4 explores are INTERDEPENDENCIES between inflectional categories as to the extents of their ELABORATION relative to their DOMAINS. In particular, we will consider number, person, and distance (spatial and related deixis) and compare nouns and pronouns as to how richly or poorly they inflect for these categories. A so far unappreciated pattern which emerges is that in addition to domain-neutral languages there are languages which favour the elaboration either of only pronominal or, less commonly, of only nominal inflection.

2. Interdependency in categorial elaboration across domains Logically the elaboration of one inflectional category is independent of that of any other, within the same domain or across domains. Taking number and person as examples, it is conceivable (i) that neither category distinguishes more than the minimal number of terms, viz. two numbers (most likely singular and plural5) and three persons (speaker, addressee, non-speech-act participant, with non-singulars including referents lower on the hierarchy of speech-act roles);6 (ii) that both categories distinguish more than the minimal number of terms, with number perhaps realized by singular, dual, trial, paucal, and multal (and perhaps associative and collective, if these are counted as numbers), and with person adding a contrast of inclusive and exclusive to 1st and perhaps even 2nd person non-singular; (iii) that only number or (iv) only person distinguishes more than the bare minimum of terms. It is an empirical question whether all four logical possibilities are attested in the languages which inflect for the relevant categories; the absence or relative infrequency of one or more of them would suggest interdependency between the categories. Such interdependency could take two forms: the elaboration (or non-elaboration) of one category could go hand in hand with that of another (possibilities (i) and (ii)), or it could be inversely proportional to that of another ((iii) and (iv)). It has variously been claimed – among others by Royen (1929: 549), Capell (1965), and Klimov (1977) – that categories whose domains are different phrase types, in particular nominal and verbal phrases, will not be equally elaborate but will either be equally inelaborate or flourish alternatively. Accordingly, numerous terms for case, number, gender, and perhaps further nominal categories should not be found in languages rich in terms for tense, aspect, mood, diathesis, and perhaps further verbal categories, nor vice versa. For categories whose joint domain is the noun phrase, on the other hand, the elaboration of some categories has been claimed to encourage, rather than to discourage, that of some others. Thus, Wilhelm Wundt (1904: 427–448) maintained that if there are more than two numbers, with singular and plural accompanied by a dual and perhaps trial, there will tend to be a contrast of inclusive and exclusive in the paradigm of personal

✐

✐ ✐

✐

✐ 255

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #279

✐

✐

Nominal or pronominal inflection

255

pronouns as well as more distance distinctions than that between proximal and distal with demonstratives and local adverbs, and vice versa. Wundt was primarily thinking of pronominal inflection, but, aiming at a holistic typology of “concrete” and “abstract” languages, he extended the correlation also to nominal number (as well as to adverbial cases of nouns, among many other parameters). In the same vein, Wilhelm Schmidt (1926) saw the presence of a dual and perhaps a trial, irrespective of their pronominal or nominal domain, as largely coinciding with that of the inclusiveexclusive contrast of 1st person pronouns; but he admitted that a dual was unaccompanied by inclusive-exclusive in his northern Primärsprachenkreis. The correlation Schmidt envisaged, thus, was not biconditional: inclusive-exclusive appeared to imply dual, but not vice versa.7 To Henri Frei (1944: 127), surveying demonstrative systems, the correlation between the elaboration of distance with demonstratives and nominal number in general seemed “assez net, malgré des exceptions”, and he also, rather implausibly, hinted at a further correlation with numeral classifiers. The evidence for such claims has at best been impressionistic. Their fate accordingly has been either not to be taken note of or to be dismissed as far-flung or fantastic, sprung from long obsolete preconceptions about concreteness and abstractness in language and thinking. Although they are claims about how languages can differ, they could seem reminiscent of the ravings of a Jacob Grimm who would not see any real space for variation in categorial elaboration, and for whom all grammatical categories – genders, numbers, persons, declensions, grades of comparison, tenses, diatheses, basic vowels, and what not – were always triadic, or aimed to be.8 First impressions are often wide of the mark, but not always. In the case at hand, more systematic comparisons across sizeable language samples do show some interdependency, if only of a probabilistic kind, between the extents to which the nounphrase categories of number, person, and distance are elaborated. Contrary to earlier assumptions, however, the mutual influences here can be discouraging as well as encouraging. And what matters crucially are the domains of inflection. As probes into reality I will draw on the samples of Forchheimer (1953), as adapted by Ingram (1978), and Perkins (1980), consisting of 69 and 50 languages respectively,9 and above all on the sample of 205 languages listed in the Appendix to this chapter. Forchheimer (1953) and Ingram (1978) were seeking cross-linguistic generalizations about the inflection of personal pronouns, especially for number. Perkins’s (1980) interest was in deixis, and some further inflectional categories such as number and gender, and in the possibility of correlating inflectional complexity with cultural complexity. The focus of our own database (called DuDa) is on the dual and its forms and uses, but some further inflectional categories as well as some syntactic patterns are also taken into account in order to determine whether dual variables co-vary with other variables. Having a dual, in the sense of any grammaticalized form or combination of forms for reference to two,10 was the criterion for inclusion in this sample. Actually, no strong genetic biases ensue from this cri-

✐

✐ ✐

✐

✐ 256

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #280

✐

✐

256

Frans Plank

terion, since there are only few higher-level families which were ruled out owing to their current or historical lack of a dual: Northeast and Northwest Caucasian, extinct Hurro-Urartean, Altaic other than Tungusic, Sinitic, Cushitic and Omotic within Afroasiatic, some North-American Indian families including Salish, and isolates such as Basque, Burushaski, and Korean and Japanese. Families with attested duals not represented in the current DuDa sample include Mande, Kartvelian, Tungusic, Dravidian, Algonquian, Oto-Manguean, as well as pidgins and creoles.11 Despite all kinds of imperfections these three samples should be reasonably adequate as a basis from which to draw conclusions about interdependencies between the inflectional categories at issue.

3. Number and person Neither dual nor inclusive-exclusive are typically present in minimal number and person inflection.12 Considering them independently, such enrichments are not exceedingly rare, though. In the Forchheimer-Ingram sample the languages which have a dual with personal pronouns and those which lack one are about equally numerous (35 and 34, respectively). Perkins (1980: 169–170) gives figures for the dual only for bound pronouns: of 35 languages with bound pronouns, eight inflect them for dual and 27 do not. Presumably, if information were available for the inflection of nouns, and for the Perkins sample also of independent pronouns, the proportion of languages with dual to those without would increase noticeably. According to a recent census (Schellinger 1995a), at least one third, but possibly about half, of the languages of the world are confirmed possessors of duals, in one domain or another. As to inclusive and exclusive, 37 languages in the Forchheimer-Ingram sample have this contrast and 32 lack it. The proportion is somewhat lower in the Perkins sample: here fifteen languages distinguish inclusive and exclusive (for 1st person) in independent personal pronouns and ten do so in bound pronouns, whereas 32 and 25 languages respectively do not. Turning now to interrelations between these elaborations of number and person inflection, Perkins’s (1980: 169–170) data for bound pronouns suggest that if there is no dual there will very likely be no inclusive-exclusive contrast either, and vice versa; see Table 1 for the figures. Further, the proportion of dual languages to languages without dual is higher for languages which distinguish inclusive-exclusive (4 : 6, 40 % : 60 %) than for languages which do not (4 : 21, 16 % : 84 %). The proportion of languages with inclusive-exclusive to those without is likewise much higher for languages with dual (4 : 4, 50 % : 50 %) than for those without (6 : 21, 22.2 % : 77.8 %). From what Perkins (1980: 103, 119) summarily concludes about the correlation of cultural complexity with the presence of inclusive-exclusive on the one hand and

✐

✐ ✐

✐

✐ 257

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #281

✐

✐

Nominal or pronominal inflection

257

Table 1. Dual and inclusive-exclusive in bound pronouns; sample of Perkins (1980). Note: In all tables percentages are given in parentheses; those calculated vertically are in roman type and those calculated horizontally are in italic. INCLUSIVE -

DUAL

EXCLUSIVE

+

+

4

− (50.0)

(40.0) −

6

(22.2)

10 (28.6)

21

(77.8)

25 (71.4)

(60.0) 4

(50.0)

(16.0) total

total

(84.0) 8

27

(22.9)

35

(77.1)

Table 2. Dual and inclusive-exclusive in independent personal pronouns; sample of Forchheimer (1953) and Ingram (1978). INCLUSIVE -

DUAL

EXCLUSIVE

+

+

26

− (74.3)

(70.3) −

9

(25.7)

(32.4)

37 (53.6)

23

(67.6)

32 (46.4)

(71.9) 35

(50.7)

11 (29.7)

(28.1) total

total

34

69

(49.3)

of dual in independent pronouns on the other, it can be inferred that the results for independent pronouns would be similar to those in Table 1. The data of Forchheimer (1953) and Ingram (1978) for independent personal pronouns, summarized in Table 2, point in the same direction. Here one need not even resort to comparing proportions in languages with and without the respective number and person terms in order to see that their elaborations are mutually conducive. What Table 2 suggests is that if personal pronouns distinguish a dual they will very likely, in 74.3 % of the languages in the sample, also distinguish inclusive and exclusive (of 1st person); and if they distinguish inclusive and exclusive they will very likely, in 70.3 % of the languages sampled, also distinguish a dual. In the DuDa sample of 205 languages, all of which have a dual in personal pronouns (independent and/or bound), nouns, or elsewhere, 122 (59.5 %) distinguish inclusive and exclusive in personal pronouns (independent and/or bound), and 83 (40.5 %) do not (Table 3) – which is a result intermediate between those of Perkins and Forchheimer-Ingram.

✐

✐ ✐

✐

✐ 258

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #282

✐

✐

258

Frans Plank

EXCLUSIVE

INCLUSIVE -

Table 3. Dual in different domains and inclusive-exclusive; DuDa sample. DUAL

nominal and pronominal

+

only pronominal

39 (51.3) (32.0)

−

37 (48.7)

total

31 (28.4)

10 (90.9)

4 (44.4)

122 (59.5)

5 (55.6)

83 (40.5)

(6.0) 11

(5.4)

total

(3.3)

(12.0) 109

(53.2)

only verbal

1 (9.1) (0.8)

(37.3) 76

(37.1)

only nominal

78 (71.6) (63.9)

(44.6)

by domains

9

205

(4.4)

What the breakdown of these last figures in Table 3 reveals is that the dependence of person elaboration on that of number is even clearer when nominal and pronominal domains of the dual are kept apart. Of the 109 languages which have the dual only in pronouns,13 78 (71.6 %) distinguish inclusive and exclusive, a contrast only relevant to personal pronouns, and 31 (28.4 %) do not; and these percentages resemble those for the Forchheimer-Ingram sample rather closely, where nominal inflection has been disregarded. Having a dual only with nouns, on the other hand, DIScourages the distinction of inclusive and exclusive in personal pronouns: as many as ten of the pertinent eleven languages (90.9 %) do not make this distinction. While there are, thus, good statistical implications if the dual is limited either to pronouns or to nouns – if only pronouns have a dual, they will very likely also have inclusive-exclusive; if only nouns have a dual, personal pronouns are very likely to lack inclusive-exclusive – little follows from the indiscriminate use of a dual with both pronouns and nouns. Among such languages (76 in the 205-sample) the inclusive-exclusive contrast is about as frequently present (39 times) as it is absent (37 times) – which may still be slightly more than can be said for languages without dual, to judge from the Forchheimer-Ingram and Perkins samples.14 From the nature of our own data it is impossible to determine whether inclusiveexclusive implies anything about duals. Nonetheless, among dual languages which do have inclusive-exclusive, the likeliest domains of the dual are, in this order, the purely pronominal one (63.9 %), the nominal plus pronominal one (32.0 %), and, trailing far behind, the purely verbal (3.3 %) and the purely nominal ones (0.8 %). Among dual languages which lack inclusive-exclusive, the dual is most likely to be found with both nouns and pronouns (44.6 %), followed by the purely pronominal domain (37.3 %) and the purely nominal one (12.0 %), which does much better here.

✐

✐ ✐

✐

✐ 259

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #283

✐

✐

Nominal or pronominal inflection

259

24 languages in the 205-sample elaborate number even further by also inflecting for trial. Trials are only found with personal pronouns and not with nouns,15 and it is in line with the previous results that the presence of a trial encourages the inclusive-exclusive distinction in pronouns even more strongly: nineteen trial languages (79.2 %) have inclusive-exclusive and only five (20.8 %) lack it, whereas the proportion is 103 : 78 (56.9 % : 43.1 %) for trial-less languages which have a dual in some domain, and 66 : 30 (68.7 % : 31.3 %) for trial-less languages with a purely pronominal dual. Eleven languages in the 205-sample have inflectional singulative, an elaboration of number peculiar to nouns. Among these, six distinguish inclusive-exclusive in pronouns and five do not, reversing the minimal preponderance obtained for dual languages in general and thus lending further (though in itself minimal) support to the hypothesis that number elaboration limited to nouns discourages person elaboration in pronouns. No such support derives from those languages in the dual sample which further differentiate a paucal from a multal, which is mostly done only with nouns. In this subset of twelve confirmed (plus several doubtful) cases, the pronominal distinction of inclusive and exclusive predominates massively (11 : 1, with a better balance among the doubtfuls), far more so than in the set of dual languages as a whole.

4. Number and distance Demonstrative pronouns are akin to personal pronouns insofar as they may do what 3rd person personal pronouns do, viz. phorically or deictically refer to (by then) definite referents other than speech-act participants; and often no separate personal pronouns are available for these purposes to begin with. The cardinal paradigmatic category of demonstratives, apart from number and perhaps gender and case, is spatial deixis and notions metaphorically derived from it, and this too is often transparently related to deictic distinctions of speech-act roles (or “persons”) fundamental to paradigms of personal pronouns. The range of variation in the deictic elaboration of demonstrative paradigms, however, is much wider than that in personal pronoun paradigms. And there do not seem to be great differences in this respect depending on whether demonstratives are free or bound, independent or adnominal, nominal, or adverbial. (Spatial deixis is of course also at home in domains other than demonstrative pronouns, although only rarely in genuine noun inflection; but these will be ignored here.) Some languages have been claimed, not uncontroversially, to forgo spatial deictic distinctions altogether, allowing themselves only a single, deictically neutral demonstrative. Binary and ternary demonstrative systems differentiate the relative proximity of referents to the deictic centre (usually the speaker) and/or locate them

✐

✐ ✐

✐

✐ 260

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #284

✐

✐

260

Frans Plank

in relation to the speaker, the addressee, or others not participating in the speech event (i.e., 3rd persons). More elaborate demonstrative systems involve parameters such as present/absent, visible/invisible, above/below, in front/behind, left/right, north/south/east/west, inside/outside, sitting/standing/moving (away from/towards); they rarely distinguish distance as such beyond the three terms of proximal, medial, and distal (such as far-away and not-so-far-away distal). Attributions of binary, ternary, or richer demonstrative systems to particular languages have sometimes also been controversial, owing to often massive diachronic and dialectal variation. Another reason for uncertainty is the frequent availability of a deictically neutral demonstrative in addition especially to a proximal and a distal one, which makes it difficult to decide whether a system is binary or ternary. To judge from the wide-ranging surveys of demonstratives by Frei (1944), Anderson & Keenan (1985), and Diessel (1999), ternary systems seem somewhat more popular than binary ones, which in turn possibly outrank quaternary and all richer systems lumped together, although perhaps only slightly. Charitably interpreted in terms of notions of cultural complexity like those used by Perkins (1980) and Schellinger (1995b), Frei’s hypothesis of an inverse correlation between degrees of civilization and of spatial deictic elaboration is not, en gros, implausible.16 For almost 20 % percent of the languages in the DuDa sample reliable information about spatial and related deictic distinctions in demonstratives, here summarily referred to as distance, is missing. It is conceivable, though unlikely, that distributions of distance degrees over dual domains in this substantial residue are such as to upset the overall pattern; therefore, in order not to preclude any eventualities, percentages have been calculated separately for the entire sample (Table 4) and for only those languages where information on distance is available (Table 4 ). It seems safe to conclude that among languages which have a dual binary and ternary distance are about equally frequent, with more elaborate demonstrative systems being about half as frequent as either. This is a distribution which, presumably, is not radically different from that likely to be found with languages in general, regardless of whether or not they have a dual. On the assumption that a dual and elaborate deictic systems individually correlate negatively with cultural complexity, as shown by Schellinger (1995b) and Perkins (1980), one might perhaps have expected a more marked predominance of ternary, quaternary, and richer demonstratives in a sample of dual languages; possibly, quaternary-plus demonstratives indeed are more frequent than is their overall average. It is again if the domains of dual inflection are differentiated that a more conspicuous pattern emerges, as seen in Tables 4 and 4 . If the dual is limited to personal pronouns, ternary distance is almost twice as frequent as binary distance (38.5 % : 22.0 % in Table 4, 49.4 % : 28.2 % in Table 4 ), whereas it is the other way round if the dual is limited to nouns (63.6 % binary : 27.3 % ternary). Or, looked at from the angle of distance in languages with a dual (horizontal percentages in

✐

✐ ✐

✐

✐ 261

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #285

✐

✐

261

Nominal or pronominal inflection

DISTANCE

degrees

Table 4. Dual in different domains and distance in demonstratives; DuDa sample. DUAL

nominal and pronominal

I

only pronominal

0

by domains only nominal

3 (2.8)

only verbal

0

0

total 3 (1.5)

(100.0) II

28 (36.8) (45.2)

III

24 (22.0) (38.7)

20 (26.3) (29.8)

IV+

42 (38.5) (62.7)

? total

1 (9.1)

24 (22.0)

76

2 (22.2)

67 (32.7)

3 (33.3)

38 (18.5)

1 (11.1)

35 (17.1)

(7.9) 0

(68.6)

(37.1)

62 (30.2)

(3.0)

(2.6)

10 (13.2) (28.6)

3 (27.3)

16 (14.7) (42.1)

3 (33.3) (4.8)

(4.5)

18 (23.7) (47.4)

7 (63.6) (11.3)

(2.9) 109

11

(53.2)

(5.4)

9

205

(4.4)

DISTANCE

degrees

Table 4 . Dual in different domains and distance in demonstratives; DuDa sample, with those languages omitted where reliable information on distance degrees is missing. DUAL

nominal and pronominal

I

only pronominal

0

by domains only nominal

3 (3.5)

only verbal

0

0

total 3 (1.8)

(100.0) II

28 (42.4) (45.2)

III

24 (28.2) (38.7)

20 (30.3) (29.8)

IV+

42 (49.4) (62.7)

18 (27.3) (47.4)

total

3 (27.3)

16 (18.8)

1 (9.1)

2 (25.0)

67 (39.4)

3 (37.5)

38 (22.3)

(7.9) 11

(6.5)

62 (36.5)

(3.0)

(2.6) 85

(50.0)

3 (37.5) (4.8)

(4.5)

(42.1) 66

(38.8)

7 (63.6) (11.3)

8

170

(4.7)

Tables 4/4 ): while it is overall far more likely that the domain of the dual is confined to pronouns than to nouns (53.2/50.0 % : 5.4/6.5 %), this likelihood is even more strongly in favour of the exclusively pronominal domain if distance is ternary

✐

✐ ✐

✐

✐ 262

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #286

✐

✐

262

Frans Plank

than if it is binary (62.7 % : 4.5 %, as opposed to 38.7 % : 11.3 %). Quaternary and richer distinctions of distance too are more common in the company of exclusively pronominal (14.7/18.8 %) than of exclusively nominal duals (9.1 %). If nouns as well as pronouns have a dual, the distribution of binary, ternary, and richer distance is more balanced (36.8/42.4 % : 26.3/30.3 % : 23.7/27.3 %). The elaboration of number only in personal pronouns is, thus, conducive to the elaboration of distance in demonstrative pronouns (and vice versa), while the elaboration of number only in nouns tends to inhibit that of distance in demonstratives. On the slender evidence available, the richer demonstrative systems appear to be relatively most popular where rich number distinction is the sole responsibility of verbal inflection. What does not quite fit into the pattern of Tables 4/4 is that the least elaborate demonstrative systems, viz. the single-member ones (as purportedly found in Bongu, Nengone, and North-East Aoban), should be concentrated in languages distinguishing a dual in personal pronouns but not in nouns. The presence of a trial, always confined to personal pronouns, does not increase the likelihood of distance distinctions in demonstratives being more than two-way. Of the 24 trial languages in the 205-sample, one has no distance contrast, seven (29.2 %) have binary distance, seven (29.2 %) have ternary distance, four (16.7 %) have more than ternary distance, and for five information is lacking. These are about the same percentages as for the 181 languages with a dual but no trial: 1.1 % unary, 29.8 % binary, 33.7 % ternary, 18.2 % quaternary or richer, 17.1 % no information. The distinction of a paucal from a multal, although it is more common with nouns than with pronouns, does not curb distance elaboration: among the at least twelve relevant languages in the DuDa sample, ternary and quaternary-plus distance are about equally frequent as binary distance. Of the eleven languages with a singulative, a number exclusive to nouns, as many as six have ternary distance, but only one has binary distance (and another one quaternary-plus), which runs counter to the tendency of the inflectional elaboration of nouns impeding that of pronouns. If distance elaboration in demonstrative pronouns is interdependent with number elaboration in nouns or personal pronouns, what about number in demonstratives itself? There is considerable evidence that independent demonstratives are likely to inflect for number as such (see Diessel 1999: 25–28, 171–173), and that adnominal demonstratives are equally keen to agree in number with their head nouns (Plank 1994). What is less certain (and this is currently also a gap in DuDa) is how likely the number inflection of demonstratives is to include that for dual, and whether this is influenced by limitations of the dual to personal pronouns or nouns. The indications are that there are no such interdependencies. Among the eleven languages with an exclusively nominal dual, demonstratives inflecting for dual are a minority: they do in Awa, Gadsup, Eastern Libyan Arabic, and Anatolian Arabic, but they don’t in Önge, Hopi, Modern Irish, Maltese, Biblical Hebrew, and probably Biblical Aramaic

✐

✐ ✐

✐

✐ 263

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #287

✐

✐

Nominal or pronominal inflection

263

and Akkadian. It should be borne in mind here that the majority of these languages in fact limit their dual to semantic subclasses of nouns (which often include that of natural pairs, or also the complementary class); and unless this subclass includes animates (which it does only rarely) it is actually implausible for the dual to extend also to demonstratives: demonstratives proper seem to be most frequently used with animate reference.17 Among the 109 languages with a dual confined to personal pronouns, duals on demonstratives seem more frequent, though only marginally so. An impediment here is that about half of these languages limit their pronominal dual to the 1st and/or 2nd person; but it would be the 3rd person which forms a natural class with demonstratives. Thus, it seems essentially only by virtue of the category of distance that demonstrative pronouns are interdependent with personal pronouns in inflectional elaboration. On the side of personal pronouns, another category which could conceivably co-vary with distance in demonstratives and number and person in personal pronouns themselves is formality or politeness; but relevant evidence has yet to be collected. Within the limits of the present study, evidence is already in concerning distance and person, as seen in the next section.

5. Person and distance For languages with a dual the elaborations of the two pronominal categories of person and distance tend to reinforce each other, as is to be expected after the previous results and as is seen in Tables 5/5 and 6/6 . If personal pronouns distinguish inclusive and exclusive forms, in all DuDa languages altogether, demonstrative pronouns are somewhat likelier to have ternary than binary distance (38.5 % : 27.0 % in Table 5, 46.1 % : 32.3 % in Table 5 ), whereas binary prevails over ternary distance (34.9/42.6 % : 24.1/29.4 %) if the inclusiveexclusive contrast is lacking. Considering only those 119 languages of the DuDa sample which limit the dual to personal pronouns, given in Tables 6 and 6 , those with an inclusive-exclusive contrast now show an even higher ratio of ternary to binary distance (42.2 % : 19.3 %, or, ignoring those languages with no reliable information on distance, 52.2 % : 23.9 %) than all dual languages altogether, while those among them without inclusive-exclusive show about the same slight reverse predominance of binary over ternary distance (30.6/40.7 % : 27.8/37.0 %). Of the eleven languages which limit the dual to nouns, only one (Awa) distinguishes inclusive and exclusive, and it also misbehaves by indulging in demonstrative distinction (quaternaryplus). What slightly mars the overall correlation of person and distance elaborations are the relatively higher frequencies of four or more distance degrees among languages without inclusive-exclusive (22.9/27.9 %, as compared to 15.6/18.6 % with inclusive-exclusive in all duals, 16.7/22.2 % : 15.7/19.4 % in pronominal-only du-

✐

✐ ✐

✐

✐ 264

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #288

✐

✐

264

Frans Plank

Table 5. Inclusive-exclusive and distance in demonstratives, in dual languages regardless of the domain of the dual; DuDa sample. INCLUSIVE - EXCLUSIVE

DISTANCE

degrees

+

I

3

(2.5)

33

(27.0)

−

total

0

3 (1.5)

(100.0) II (53.2) III

47

(38.5)

(70.1) IV+

19

(15.6)

62 (30.2)

20

(24.1)

67 (32.7)

19

(22.9)

38 (18.5)

15

(18.1)

35 (17.1)

(50.0) 20

(16.4)

(57.1) total

(34.9)

(29.9)

(50.0) ?

29 (46.8)

(42.9) 122

83

(59.5)

205

(40.5)

Table 5 . Inclusive-exclusive and distance in demonstratives, in dual languages regardless of the domain of the dual; DuDa sample, with languages omitted where reliable information on distance degrees is missing. INCLUSIVE - EXCLUSIVE

DISTANCE

degrees

+

I

3

(2.9)

33

(32.3)

−

total

0

3 (1.8)

(100.0) II (53.2) III

47

(46.1)

(70.1) IV+

19

(18.6)

62 (36.5)

20

(29.4)

67 (39.4)

19

(27.9)

38 (22.3)

(50.0) 102

(60.0)

(42.6)

(29.9)

(50.0) total

29 (46.8)

68

170

(40.0)

als). Also, undifferentiated distance is in the DuDa sample confined to languages with inclusive-exclusive. Looked at from the angle of distance, languages with inclusive-exclusive outnumber those without more clearly if they have ternary distance (70.1 % : 29.9 %)

✐

✐ ✐

✐

✐ 265

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #289

✐

✐

Nominal or pronominal inflection

265

Table 6. Inclusive-exclusive and distance in demonstratives, in languages limiting the dual to personal pronouns; DuDa sample. INCLUSIVE - EXCLUSIVE

DISTANCE

degrees

+

I

3

(3.6)

16

(19.3)

−

total

0

3 (2.5)

(100.0) II (59.3) III

35

(42.2)

(77.8) IV+

13

(15.7)

16

(19.3)

10

(27.8)

45 (37.8)

6

(16.7)

19 (16.0)

9

(25.0)

25 (21.0)

(36.0) 83

36

(69.7) Table 6 .

27 (22.7)

(31.6)

(64.0) total

(30.6)

(22.2)

(68.4) ?

11 (40.7)

119

(30.3)

Inclusive-exclusive and distance in demonstratives, in languages limiting the dual to personal pronouns; DuDa sample, with languages omitted where reliable information on distance degrees is missing. INCLUSIVE - EXCLUSIVE

DISTANCE

degrees

+

I

3

(4.5)

16

(23.9)

−

total

0

3 (3.2)

(100.0) II (59.3) III

35

(52.2)

(77.7) IV+

13

(19.4)

27 (28.7)

10

(37.0)

45 (47.9)

6

(22.2)

19 (20.2)

(31.6) 67

(71.3)

(40.7)

(22.2)

(68.4) total

11 (40.7)

27

94

(28.7)

than if they have binary distance (53.2 % : 46.8 %); and in the subset of languages which limit the dual to personal pronouns these differences are even bigger (ternary: 77.8 % : 22.2 %, binary: 59.3 % : 40.7 %). This trend is again reversed by languages with undifferentiated distance, all three of which have an inclusive-exclusive con-

✐

✐ ✐

✐

✐ 266

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #290

✐

✐

266

Frans Plank

trast. In languages with more than three-way distance, inclusive-exclusive contrasts are rather less frequent (50.0 %) across the entire dual sample; however, among those quaternary-plus distance languages which limit the dual to personal pronouns, they are about as frequent (68.4 %) as might be expected on the assumption that the likelihood of person elaboration increases steadily the more degrees of distance are being distinguished.

6.

Pronominal and nominal languages

To summarize, there is evidence, rather than only flights of fancy, to suggest that the extents are interdependent to which the inflectional categories of number, person, and distance are elaborated. In two respects these interdependencies are less global than has sometimes been conjectured. First, they are not equally valid for all terms of these categories. Among number terms it is apparently the dual which is most conducive to person and distance elaboration; and for distance it is the difference between binary and ternary systems which correlates best with number and person elaboration. Second and most importantly, such interdependencies are relative to domains of inflection. The elaboration of number in both personal pronouns and nouns presumably does not massively increase the likelihood of person distinguishing inclusive and exclusive and of distance being ternary. Elaborating number only in personal pronouns, on the other hand, does favour inclusive-exclusive in personal pronouns and ternary distance in demonstrative pronouns, while the elaboration of number only in nouns significantly discourages the distinction of such terms in pronominal inflection. Being both pronominal categories, person and distance (especially if ternary) tend to be elaborated in unison. These interdependencies are all mere tendencies; none are categorical. They are therefore unlikely to be explicable by hard and fast principles of grammar. Still, there is an overarching theme of universal grammar which one might have hoped to be able to turn to in order to make sense especially of differences between domains. As often observed, there is a general preference for pronominal rather than nominal inflection: if nouns inflect, pronouns and especially personal ones will inflect too.18 This asymmetry may have functional reasons or also diachronic ones, with pronouns often serving as the analogical or grammaticalized sources of noun inflection (see below). Whatever its reasons, while this generalization seems largely valid for all relevant inflectional CATEGORIES, some TERMS of some categories may, however, show opposite possibilities or indeed predilections. Thus, local cases, unlike grammatical and some other adverbial cases, would generally seem to favour nouns (especially ones denoting localities) over personal pronouns.19 And, as seen above, the

✐

✐ ✐

✐

✐ 267

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #291

✐

✐

Nominal or pronominal inflection

267

dual number sometimes favours nouns too,20 even though it generally favours personal pronouns, being limited to them to the exclusion of nouns in 119 languages in our 205-sample, with only eleven languages limiting it to nouns. In the light of this general inflectional privilege of pronouns, it is languages which, more unusually, elaborate the inflection of nouns at the expense of that of pronouns that command special attention. The unorthodox preference for nominal inflection could of course be an idiosyncrasy of a few random languages which do not share anything else, structurally or otherwise, thus defying any general explanation. However, there are indications that this peculiarity tends to run in families. Of the eleven relevant languages in the DuDa sample no less than six are Semitic (Akkadian, Biblical Hebrew, Biblical Aramaic, Eastern Libyan Arabic, Anatolian Arabic, Maltese) and two are from the Papuan Kainantu family (Gadsup, Awa). However, the remaining three languages (Modern Irish, Önge, Hopi) are unrelated to these as well as among each other, which precludes an exclusively genetic account of the preference for nominal inflection as a somewhat unusual shared heritage.21 But then, even if the incidence of an unusual phenomenon could be circumscribed genetically, there would still be a crucial question waiting to be answered: How come generations of learners have been continuing to acquire something so unusual, rather than to restructure their grammars along more conventional lines, as has happened elsewhere in the families concerned? To the extent that exclusively nominal duals and what comes with them show areal concentrations, being primarily found in North Africa and the Near East and in the Eastern Highlands of New Guinea, these follow from the genetic affiliations of the languages concerned (Semitic, Kainantu). As with the dual in general, when history matters, it is in terms of inheritance (and increasing cultural complexity) rather than borrowing. There do not seem to be linguistic neighbourhoods where duals would perforce wax or wane once languages with whatever categorial endowments have moved there. Being in contact with a language with or without a dual (of one kind or another) is an insignificant factor by comparison to being genetically derived from a language with or without dual (and to being the language of a community of a certain level of cultural complexity).22 Apart from certain genetic (and cultural) predispositions, it is only structural correlates which are left as possible pointers to an explanation of the preference for elaborating nominal inflection more than that of pronouns. Four such correlates emerge from DuDa, the first three to do with the domains, uses, and forms of the dual itself.23 First, as briefly alluded to above, limiting the dual to nouns to the exclusion of personal pronouns greatly favours its further limitation to subsets of nouns, and in particular to that of natural pairs or also that of standard units for counting and measuring (Tables 7 and 8). Limiting the dual to personal pronouns likewise favours its

✐

✐ ✐

✐

✐ 268

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #292

✐

✐

268

Frans Plank

Table 7. Use of the dual with all or only some nouns, in those languages where it is used with nouns; DuDa sample. DUAL

with nouns

DUAL

by domains

nominal and pronominal

all

52

only nominal

(68.4)

(91.2) some

5

(45.5)

57 (65.5)

6

(54.5)

26 (29.9)

(8.8) 20

(26.3)

(76.9) ?

total

(23.1) 4

(5.3)

0

4

11

87

(4.6)

(100.0) total

76 (87.4)

(12.6)

Table 8. Use of the dual with subsets of nouns, in those languages where it is not used with all nouns; DuDa sample. (It is possible for single languages to limit the dual to more than one subset of nouns.) DUAL

with nouns

DUAL

nominal and pronominal

animates

16

only nominal

(66.7)

(88.9) pairs

total

2

(22.2)

18 (54.5)

4

(44.4)

10 (30.3)

(11.1) 6

(25.0)

(60.0) non-pairs

by domains

(40.0) 2

(8.3)

0

2

(6.1)

3

(9.1)

(100.0) measures

0

3

(33.3)

(100.0) total

24 (72.7)

9

33

(27.3)

further limitation to 1st and/or 2nd person, although this remains a minority pattern (Table 9). Second, when reference is being made to two the use of a dual with words possessing one may be obligatory or only optional, with plural, paucal, a numberneutral basic form, or even singular as alternatives. If the dual is limited to personal pronouns, the likelihood is great that it will be used obligatorily, whereas exclusively nominal duals are almost as frequently optional as obligatory (Table 10). Third,

✐

✐ ✐

✐

✐ 269

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #293

✐

✐

269

Nominal or pronominal inflection

Table 9. Use of the dual with all or only some personal pronouns, in those languages where it is used with pronouns; DuDa sample. DUAL

DUAL

with nouns

by domains

nominal and pronominal

all

68

only nominal

(89.5)

(50.0) some

68

(62.4)

136 (73.5)

41

(37.6)

49 (26.5)

(50.0) 8

(10.5)

(16.3) total

total

(83.7) 76

109

(41.1)

185

(58.9)

Table 10. Obligatory or optional use of dual depending on domain; DuDa sample. DUAL

use

DUAL

nominal and pronominal obligatory

only pronominal

22 (28.6) (22.4)

optional

32 (41.5)

?

11 (10.2)

23 (29.9)

total

77 (37.6)

4 (36.4)

31 (28.7)

2 (18.2)

98 (47.8)

2 (22.2)

49 (23.9)

2 (22.2)

58 (28.3)

(3.4) 11

(5.4)

5 (55.6)

(4.1)

(3.4) 108

(52.7)

total

(5.1)

(8.2)

(53.4)

only verbal

5 (45.4) (5.1)

(22.4)

(39.7)

only nominal

66 (61.1) (67.3)

(65.3)

by domains

9

205

(4.4)

while the general tendency is for actual dual and plural exponents to be formally independent of each other, there are instances of exponents of dual being based on those for plural as well as the other way round. However, plural forms based on dual forms are only found with personal pronouns (most frequently if languages lack a nominal dual) but not with nouns; and dual forms based on plural forms are not found with nouns either in languages which lack a pronominal dual (Table 11). On all three counts, languages with a pronoun-only dual do not differ greatly from languages whose dual is in both domains, which confirms that it is languages with noun-only duals that are special.24 There is a fourth structural correlate which on the face of it is rather enigmatic: basic word order in clauses. Yet it may hold the most important clue to what is

✐

✐ ✐

✐

✐ 270

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #294

✐

✐

270

Frans Plank

Table 11. Formal relationship between dual and plural marking of nouns and independent personal pronouns, relative to domain limitations; DuDa sample. (Languages are counted multiply if different pronouns or forms of pronouns behave differently with respect to the criterion at issue. Bound personal pronouns show a similar pattern as independent ones.) and forms

of nouns

independent DU based on PL PL based on DU total

77 (91.7) 7 (8.3) 0 (0.0) 84

DUAL

PLURAL

of nouns if there is no pronominal dual 11 0 0 11

(100) (0) (0)

of pronouns

of pronouns if there is no nominal dual

127 33 33 193

65 20 27 112

(65.8) (17.1) (17.1)

(58.0) (17.9) (24.1)

behind the elaboration of the inflection of nouns rather than pronouns – or perhaps rather the under-elaboration of pronominal inflection relative to that of nouns. As all large-scale word-order surveys suggest,25 the most common basic order of S(ubject), O(bject), and V(erb) is SOV, favoured by well over 40 % of the languages of the world, followed by SVO (around 30 %),26 and with VSO some way behind in third position (around 10 %). On the evidence of DuDa, which also includes information about basic word order (allowing for single languages to have more than one basic order or also free order, although this distinction is sometimes difficult to draw in practice), dual languages as such are not a particularly nonconformist lot. As shown in the right-hand column of Table 12, the ranking of basic word orders among dual languages is about the same as that among languages in general, with SVO perhaps somewhat disfavoured. The subset of languages which have both pronominal and nominal duals is not strikingly different either. Once more, what does make a real difference is whether the dual is limited to either personal pronouns or nouns. If its domain is exclusively pronominal, SVO is almost level with SOV (32.5 % : 37.6 %), with VSO (9.4 %) far behind and with VOS (5.1 %) doing relatively well for an otherwise extremely marginal order. If the dual is exclusively nominal, on the other hand, SOV is as predominant as ever (46.1 %), but VSO now comes second best at not such a great distance (30.8 %), and SVO is conspicuously infrequent (23.1 %). Obviously, absolute numbers are small for the nouns-only group, and percentages can change correspondingly fast. For instance, replacing a single VSO language in the DuDa sample by an SVO one (say, Biblical Hebrew by Israeli Hebrew) would reverse the VSO-SVO ranking. Still, the unusually good showing of VSO relative to SVO, by comparison with languages of any other dual domains, is unlikely to be an artefact of sampling and small numbers. Among nouns-only languages, there are

✐

✐ ✐

✐

✐ 271

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #295

✐

✐

271

Nominal or pronominal inflection

Table 12. Dual in different domains and basic word order; DuDa sample, with single languages showing up more than once when they have more than one basic word order. BASIC WORD ORDER

DUAL

nominal and pronominal

SOV

only pronominal

39 (48.1) (41.1)

SVO

16 (19.7)

VSO

38 (32.5)

5 (6.2)

VOS

1 (1.2) (14.3)

OVS

1 (1.2) 1 (1.2) 9 (11.1) 9 (11.1)

4 (30.8) (19.0)

1 (8.3)

21 (9.4)

(4.8)

6 (5.1)

0

3 (2.6)

0

0

7 (3.1)

1 (8.3)

5 (2.2)

(20.0) 2 (1.7)

0

0

3 (1.3)

3 (2.6)

0

0

12 (5.4)

10 (8.5)

0

0

19 (8.5)

(52.6) 81

(36.3)

61 (27.3)

(25.0)

(47.4) total

4 (33.3)

(66.6)

(75.0) ?

95 (42.6)

(6.6)

(60.0)

(33.3) free

6 (50.0)

(85.7)

(20.0) OSV

3 (23.1)

11 (9.4)

total

(6.3)

(4.9)

(52.4)

only verbal

6 (46.1) (6.3)

(62.3)

(23.8)

only nominal

44 (37.6) (46.3)

(26.2)

by domains

117 (52.5)

13 (5.8)

12

223

(5.4)

not so many SVO candidates that could have made it into the DuDa sample instead of their VSO relatives. And for this minority group, basic word orders, of which all three major ones are actually instantiated, are not neatly distributed along genetic lines either: SOV is found in Semitic (Akkadian, Biblical Aramaic), Kainantu (Awa, Gadsup), as well as in Önge and Hopi; VSO is mostly found in Semitic (Biblical Aramaic, Biblical Hebrew, Eastern Libyan Arabic), but also in Modern Irish; and only SVO is only found in one family, Semitic (Eastern Libyan Arabic, Anatolian Arabic, Maltese). By comparison with languages with nominal and pronominal duals, free word order is underrepresented with both limitations of duals (11.1 % vs. 2.6 % pronounsonly and even nil nouns-only).27 Permitting themselves alternative basic orders, if not complete freedom, is somewhat more common among pronouns-only languages in DuDa: there are five such languages with co-existing SVO and VSO (SVO being universally a most popular partner of basic VSO28), and one each with SVO/VOS, SVO/SOV, SOV/VOS, VSO/VOS; nouns-and-pronouns languages com-

✐

✐ ✐

✐

✐ 272

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #296

✐

✐

272

Frans Plank

bine SOV/SVO five times and SOV/OSV once; there are two nouns-only languages with co-existing VSO/SOV (Biblical Aramaic) and VSO/SVO (Eastern Libyan Arabic); and two verbs-only languages with co-existing SOV/SVO/OVS and SOV/OSV. The connection, however tenuous, with the preference or dispreference for VSO relative to SVO as basic word order adds typological substance to the distinction between languages which preferably elaborate only nominal or only pronominal inflection. Basic word order is after all assumed to be tied up with much else. Perhaps, among the direct correlates of basic word orders, there are ones that throw light on the mystery of the link with domain limitations of the dual and its own inflectional correlates. Several studies (including Steele 1978, Bybee 1985, Hawkins & Gilligan 1988, Nichols 1992, Siewierska & Bakker 1996) have suggested correlations between basic word order and (affixally or clitically) bound verbal morphology. Among much that is controversial, what seems beyond doubt is, first, that verbal agreement or cross-reference licenses free or at least flexible word order of core arguments: for obvious functional reasons (Avoid relational ambiguity but do so economically!), free/flexible word order strongly implies verbal agreement/cross-reference marking of at least the subject but very likely also of object(s). Second, of word order types not classified as free, verb-mediality (especially SVO, but also OVS) most strongly discourages verbal agreement/cross-reference, and verb-initiality (especially VSO) encourages it most strongly, with verb-finality (most commonly SOV) somewhat undecided.29 A functional motivation that has been suggested (by Nichols 1992: 108–109) for the link between verb-initiality and verbal marking is that the relational frame of a clause is best established right at its beginning – and this is what agreement/cross-reference marking on (initial) verbs does; when the verb comes last or also second, relations are better clarified earlier, by rigid order or case marking on NPs themselves.30 One of the correlates of elaborate verbal agreement/cross-reference, which typically identifies arguments in terms of speech-act role (person), number, distance or other deictic qualities, gender/class, and/or grammatical relation, in turn is that these arguments themselves tend to be omissible if they do not add content to the bound verbal marking – i.e., if they are personal (or also demonstrative) pronouns. Thus, simplifying greatly, free basic word order and verb-initiality are licensed by verbal agreement/cross-reference, which also licenses or indeed encourages pro-drop. Determining whether a language is pro-drop is notoriously difficult because the omissibility of independent pronouns is not a question of all or nothing: they may or may not be omissible in a given language depending on all kinds of morphological, syntactic, and pragmatic circumstances; when they are in principle omissible in given circumstances, they may be omitted or retained with varying degrees of frequency rather than categorically; and personal pronouns of different kinds (e.g., 1st and 2nd person vs. 3rd; expletive vs. full; definite vs. indefinite) may differ in their

✐

✐ ✐

✐

✐ 273

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #297

✐

✐

Nominal or pronominal inflection

273

dropping propensities. On the understanding, however vague and gross, that a language counts as resolutely pro-drop if the normal expression of basic, not strongly context-bound declarative main clauses with finite verbs is lacking independent definite subject pronouns for all three persons, whose addition would add special emphasis, the cross-linguistic evidence would indeed seem to justify making this link: resolute pro-drop needs to be licensed by verbal agreement/cross-reference marking, but although it is encouraged it is not enforced by it.31 Now, if the chances are that independent pronouns will not be used a great deal to begin with, why go to the trouble of providing such low-frequency words with elaborate inflection? On the above reasoning, then, languages with verbal agreement/ cross-reference, and concomitantly with free or verb-initial basic order and resolute pro-drop, are the ones that can most easily afford not to have elaborate pronoun inflection, although they might still indulge, in acknowledgment of the general inflectional privilege of pronouns. To the extent that demonstratives perform functions similar to those of (3rd person) personal pronouns, they ought to show similar dropping behaviour; but it remains to be seen whether distance (in-)elaboration indeed also correlates with (in-)frequency of use. The supporting evidence that we can cite at least for personal pronouns is that of the 205 languages in the DuDa sample, twenty have VSO as the basic order or as one basic order, and as many as four of them (20 %) limit the dual to nouns – which is a lot by comparison with all 205 DuDa languages, of which only 5.4 % have a purely nominal dual. Indulging languages are still a majority even among the VSO subset, though: twelve (60 %) have purely pronominal and four (20 %) pronominal and nominal duals – as compared to 53.2 % and 37.1 %, respectively, for all dual languages. And, not wasting their number inflections on words rarely used, all of the four VSO languages in the 205-sample with a purely nominal dual (Biblical Aramaic, Biblical Hebrew, Eastern Libyan Arabic, Modern Irish32), are resolutely pro-drop. So are all the nominal-dual-only languages with other basic orders, including SVO (Maltese, Anatolian Arabic, Eastern Libyan Arabic: all relatively recent converts to SVO from earlier VSO). It should be noted that the argument invoking pro-drop is simply that ANY inflections of independent pronouns would be underemployed; it is not that verbal agreement/cross-reference marking would render the SAME inflectional distinctions on independent pronouns superfluous. In actual fact, at least duals are far more frequently distinguished on independent pronouns than on bound (i.e., cross-referencing) pronouns or in (non-pronominal) verb agreement.33 Of the 204 languages in DuDa which have independent personal pronouns, 185 (90.1 %) inflect these for dual, while of the 144 languages with bound personal pronouns, only 68 (47.2 %) have a dual for them; only Western Desert has a dual for bound but not for independent pronouns. Of the 135 languages where verbs agree in number with subject

✐

✐ ✐

✐

✐ 274

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #298

✐

✐

274

Frans Plank

and/or object, with the agreement markers being of a non-pronominal nature about half of the time, only 83 (61.5 %) have dual agreement. It is, thus, one option for languages with inflection, instead of letting independent pronouns exercise their prerogatives in all matters inflectional, to afford a nominal dual but not to elaborate pronominal number and person (and distance) beyond the minimum, on condition that such pronominal inflection would be underemployed owing to pro-drop. Although functional motives have been mentioned for some links in this chain of interdependencies, the rationale of this pro-nominal, or antipronominal, option is probably at heart diachronic. In barest outline, an overall long-term diachronic scenario would have inflectional elaboration begin with independent personal (and probably demonstrative) pronouns. Either by analogy or by pronouns being actually added to nouns, the same inflections would then be extended to nouns. Within limits, especially those of “minor” numbers, nouns themselves, or at least subsets of them, would also be able to independently procure themselves inflections by grammaticalizing suitable words (such as numerals, quantifiers, or collective nouns). In the right syntactic circumstances independent pronouns would eventually get grammaticalized as verbal cross-reference and agreement markers, licensing the dropping of independent pronouns. Utilized as agreement/cross-reference markers, inflections would be prone to neutralization and eventual obliteration. With such bound verbal marking falling into decay, sufficiently expressive inflection would be re-innovated on independent pronouns (of whatever provenance: newly grammaticalized from nouns, reactivated relics, or borrowed); and so on. A dual in nominal inflection but inelaborate pronominal inflection (no dual, no inclusive-exclusive, binary distance) would be what could be expected of languages caught at that stage of this scenario where older-generation pronouns-turned-boundmorphology are looking forward to be complemented by a new generation of elaborately inflected independent pronouns, and nouns hold on to a dual and perhaps other minor numbers.

7. Ach, Europa . . .: Verkehrte Welt In comparison to the world of (dual) languages at large Europe is anomalous, especially if circumscribed as generously as in EUROTYP. Singling out the parameters which were claimed to be interdependent in the previous sections, Table 13 provides profiles for all European languages currently boasting a dual, and some extinct ones representing (sub-)families where a dual has been lost. Of these languages, nine were in the 205-sample on which the claims had been based that nominal and pronominal inflection can be elaborated selectively, getting the western quarter of Eurasia a reasonably fair hearing: Maltese (Afroasiatic), Ancient Greek, Slovene, Upper Sorbian,

✐

✐ ✐

✐

✐ 275

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #299

✐

✐

Nominal or pronominal inflection

275

Lithuanian, Gothic, Old English, Modern Irish (all Indo-European), and Northern Saami (Uralic34). Perhaps the most striking generalization to emerge is that not only are both a dual and an inclusive-exclusive distinction rather unpopular in Greater Europe, but they in fact mutually exclude each other, regardless of dual domains. The only European (or EUROTYP) languages where an inclusive-exclusive distinction is attested are located in the Caucasus: Abkhaz (Northwest Caucasian), Chechen, Ingush, Bats (North-Central Caucasian), Avar, the Andic group, about half of the Lezgic group (Northeast Caucasian), and Svan (South Caucasian). Occasional rumours to the contrary and with the single exception of the Khevsurian dialect of Georgian, Caucasian families have no dual in pronouns or anywhere else, and never had one, so far as one can tell. Reassuringly, they tend to have ternary or richer distance distinctions with demonstratives; and not disconcertingly either, basic word order is SOV or sometimes also SVO. With some exceptions (in Northeast and NorthCentral Caucasian), verbs inflect for person, or class, or person and class, or person and number, or number and class, or person and number and class. The languages with a dual are listed in Table 13, arranged by dual domains: none has inclusive-exclusive; and ironically, those with an exclusively nominal dual fairly consistently come with ternary distance while those with an exclusively pronominal dual favour binary distance. Reassuringly, SVO does only moderately well and verb-initiality finds no favour at all among pronominal-dual languages, while verb-initiality outranks SVO and indeed SOV among nominal-dual languages. Verbs agree without exception, as a rule for person and number (sometimes including for dual), while pro-drop is somewhat erratic. Radical though the differences are between our world findings (Sections 3–6) and our European findings (Section 7), they arguably give only little cause for typological concern. Typology is about structural affinities between logically unrelated traits. Therefore, on the assumption that genetic and/or areal homogeneity does not guarantee structural homogeneity – in other words, that genetically or areally related languages can nonetheless differ on any single variable permitted by universal grammar to vary – a collection of languages with a genetic or areal bias does not a priori disqualify as a reasonable typological sample: it may happen to encompass all the structural variation that is to be found under the sun. Still, concerning the structural traits at issue here, there are good reasons for typologists interested in determining whether they show or do not show co-variation to stake their money on the world rather than Europe. First, European duals hail from only four families, and are indeed confined to subgroups of them: Uralic (Saami and Samoyedic), Indo-European (Celtic, Germanic, Baltic, Slavonic, Hellenic), Afroasiatic (Semitic), and South Caucasian. In genetic terms, whatever they are worth, this is hardly a representative sample; potentially, frequency results on this basis could easily be overturned by casting one’s net wider.

✐

✐ ✐

✐

✐ − anim − − 12 12 12 12 12/− 1 −− − − non-pairs 23 −− − −− − −− − − −/non-pairs − −− − II III II III II II

−i/e

indep, du the during the [a] century and a half which followed the Conquest the [a] dozen Europeans who have crossed the Sahara the cocktail takes the place of the [a] cake and a glass of wine

a.

the + the > the any sleepers on the [the] Prince of Wales Road side of the house

(120)

(121) (122) a. (123) a. b. c.

371

a + the > a some of his former associates established a new [the] Lord Chamberlain’s company an + a > a a[n a] month old baby would have eaten more36 I wish it was a[n a] little heavier one this looked like a[n a] good deal bigger affair than he had anticipated

Once the article that belongs with the whole NP is at a distance from an inherent article, the latter may resurface, however uneasily: (120 ) a.

the ten or [the] a dozen Europeans who . . .37

(121 ) a.

any sleepers on the noisy the Prince of Wales Road side of the house

(122 ) a.

some of his former associates established a new the Lord Chamberlain’s company

(123 ) a.

a healthy a month old baby would have eaten more

However, there are arguably two NPs corresponding to the two articles in virtually all such constructions, except (123b).

2.11. Noun phrase downgraded to modifier NPs may get degraded, though. Earlier we have met with situations where a second NP, full or nounless, is amplifying a first NP and the relationship between them is shading from that of loose apposition into that of closer-knit attribution, with the NP-hood of the amplifying constituent being increasingly less obvious (§§2.2–5). An alternative scenario of NP-integration is for an NP that is the head of another NP to be reanalysed as a modifier, with the erstwhile modifier correspondingly rising to the rank of the head. In the process of downgrading, the constituent affected may partly or fully lose its NP-hood while retaining its article. This is what has happened, and is still happening, to ein bißchen ‘a bit’ in German. This expression is used as a modifier of predicative adjectives (124), of mass

✐

✐ ✐

✐

✐ 372

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #396

✐

✐

372

Frans Plank

nouns, themselves without article, or of nominalizations headed by indefinite pronouns (125), but also of certain count nouns (126) and especially in conjunction with attributive adjectives with excessive zu ‘too’ (127). (124)

Der Schnaps ist ein bißchen warm the schnapps is a bit warm

(125) a.

Gib mir ein bißchen Sauerkraut! give me a bit sauerkraut! Gib mir ein bißchen was zu lesen! give me a bit something to read!

b. (126)

Gib mir ein bißchen einen Vorsprung! give me a bit an advantage!

(127) a.

Er gab mir einen ein bißchen zu großen Vorsprung he gave me an a bit too big advantage Er gab mir ein bißchen einen zu großen Vorsprung he gave me a bit a too big advantage

b.

The current status of ein bißchen is not unequivocal. Like a noun, including the near synonym Stückchen (129), bißchen admits modifying adjectives of suitable meaning (128a) and is accompanied by the indefinite article, governing their gender and number (neuter singular); and together with its own amplifiers it precedes those constituents to which it is to be related semantically (like zu groß in (129a)). However, it differs from nouns in being consistently spelled with a lower-case initial (like words other than nouns are), in permitting its own modifying adjective to drop its agreement inflection (128b/129b), and in being able to move ahead of entire articulated NPs without requiring them to be in the genitive case (126/129c, 127b/129e);38 furthermore, while bißchen is transparently a diminutive of the noun Biß ‘bite’, its meaning has been extended to that of a general quantifier for any small quantity or extent. (128) a. b.

ein klein-es bißchen Sauerkraut a little-NEUT.SG bit sauerkraut ein klein-Ø bißchen Sauerkraut a little-Ø bit sauerkraut

(129) a.

ein klein-es Stückchen Schnitzel a little-NEUT.SG morsel schnitzel b. *ein klein-Ø Stückchen Schnitzel *a little-Ø morsel schnitzel c. *Gib mir ein Stückchen einen Vorsprung! *give me a morsel an advantage!

✐

✐ ✐

✐

✐ 373

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #397

✐

✐

Double articulation

373

d.

Er gab mir einen ein Stückchen zu großen Vorsprung he gave me an a morsel too big advantage e. *Er gab mir ein Stückchen einen zu großen Vorsprung *he gave me a morsel a too big advantage Thus, despite its indefinite article, ein bißchen is not – diachronically speaking, no longer – a full-blown NP in its own right, and doubly articulated NPs containing this newly developed modifier, such as (126) and (127), must accordingly be considered to lack, to some extent, an NP constituent. Being obligatorily indefinite, the article of bißchen is more of a lexical than of a syntactic nature, though. Although the equivalents of ein bißchen in languages such as Dutch (een beetje),39 English (a bit), French (un peu), or Italian (un poco/po’) would all seem to be drifting towards modifierhood, none is as resolute in retaining its indefinite article while shedding much of its NP-hood. For instance, there is an inclination to continue to structurally subjugate the NP quantified by means of prepositions (a bit of, un peu de, un po’ di NP), but this is perhaps weakening with English a bit, and it has subsided completely with een beetje in Dutch: (130)

Percival is a bit ?? (of) a bungler

(131)

Perceval is een beetje (*van) een broddelaar

When in construction with an attributive adjective, een beetje is also found in front of the NP’s own article (132), as is possible in German (127b), whereas a bit stays by its adjective (133a), but is not entirely happy there either because English would much rather suppress one of two articles in a row (see (123) above); shifting the NP’s article after the adjective when this is permitted by an intensifier like too provides some relief (133b). (132) (133) a. b.

Perceval is een beetje een (te) oud paard Perceval is a bit a (too) old horse Percival is (?? an) a bit (too) old horse is a bit too old a horse

? Percival

One English modifier of this kind, and perhaps the only one, that is clearly nonnominal is little, which presumably derives by ellipsis from little bit, whence its unusual article; but this article too tends to get dropped in the company of another (123b), such is the horror English has of any double articulation of simple and even some complex NPs. In the Egyptian Arabic variation on this theme, NPs are sometimes more permissive (Woidich 1995: 267). When nouns such as hitta ‘piece’, habba ‘grain’, and ˙ ˙ of gasoline’ > šuwayya ‘small thing’ are grammaticalized as quantifiers (‘a grain ‘a little gasoline’), they are turned from heads into modifiers on a par with numerals. However, while NPs with a numeral take only a single definite article, prefixed

✐

✐ ✐

✐

✐ 374

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #398

✐

✐

374

Frans Plank

or proclitic to the numeral (134a), ex-noun quantifiers in substandard Cairo Arabic retain their definite article upon the combination with a noun itself marked as definite in what are clearly simple modifier-head constructions on all relevant evidence (134b). (134) a.

b.

it-talat kutub DEF-three book.PL ‘the three books’ il-habbit il-banz¯in ˙ DEF-little DEF-gasoline ‘the/this little gasoline’

2.12. Strong nouns on their own In our survey so far the nouns in doubly articulated NPs were always accompanied by lexical material of one kind or another, and it was such additional constituents which could, in one way or another, be held responsible for re-articulation. Amplification of an NP is not, however, a necessary condition for double articulation.40 In the Bavarian dialect of German, especially as spoken by the more masculine parts of the indigenous population, NPs are not infrequent where all that follows the definitely articulated head noun is another definite article (or also a 2nd or 1st person personal pronoun, if this is what precedes the head noun in lieu of an article41), and it makes no difference whether or not the noun itself is accompanied by an adjective: (135) a. b. (136) a. b.

die (bsuffane) Sau die! the (drunken) swine the! Hast die (bsuffane) Sau die gsehng? have:you the (drunken) swine the seen? du/ich (alter) Depp du/ich! you/I (old) fool you/I! Woher weißt du (alter) Depp du das? how know you (old) fool you that?

Especially in their second occurrence, the definite articles here appear in their strongest form, which they share with the independent or determiner demonstrative pronoun – as is seen when the orthography is truer to dialectal sound than it was in (135) and (136):42 (137) a.

Brennd a Haus / (a)s Haus / (a)s/dees Haus wo am Sepp is.on.fire a house / the house / the house which to:the Sepp gheard / dees oder dees Haus? belongs / this or that house (Neuter)?

✐

✐ ✐

✐

✐ 375

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #399

✐

✐

Double articulation

b.

375

Brennd ?? (a)s/dees Haus *as/dees etzat endli? house the now after.all? is.on.fire the

The force of such article or pronoun repetition, with NPs used independently as exclamations (135a/136a) or also properly embedded in complete clauses (135b/ 136b/137b), is to impart a characteristically emotional flavour to one’s words. To be able to receive such special emphasis, an NP should have a certain inherent potential for affective charge. An indefinite article can only be repeated, for similar effects of spirited emphasis, when it is in the company of an intensifying adjective: (138)

ein solcherner (alter) Depp ein solcherner a such (old) fool a such

The lack of any extra lexical material would seem to argue against the NP-hood of the plain definite article or the indefinite article plus an intensifying element that are following the properly articulated NPs in such constructions. On the other hand, it will be remembered that the purely referential value of non-overt heads in hypostasis formations (§2.1.2) did not preclude their recognition as NPs in their own right either. Moreover, it is possible, with no great difference in meaning or force, to repeat an adjective, if the NP contains one, along with the article, or indeed to have an adjective only after the second article, yielding a pattern in Bavarian that is reminiscent of that illustrated earlier from Greek (Merkle 1975: 168f.): (135 ) a. b.

die (bsuffane) Sau die bsuffane! the (drunken) swine the drunken! Hast die (bsuffane) Sau die bsuffane gsehng? have:you the (drunken) swine the drunken seen?

While the preference in such more substantial constructions is again for the strongest forms of the definite article (those shared with the demonstrative) in both its occurrences, their weaker alternatives seem more acceptable than when there is no adjective to support them: (137 ) b.

Brennd ? (a)s/dees Haus ? as/dees varreggde etzat endli? is.on.fire the house the damned now after.all?

At any rate, should articles (or whatever kinds of pronouns they turn out to be on closer examination of such formal subtleties) which are repeated on their own in the Bavarian style be analysable as remnants of NPs, the ties between such rudimentary NPs and those they accompany are far tighter than in afterthoughts or run-of-the-mill apposition.

✐

✐ ✐

✐

✐ 376

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #400

✐

✐

376

Frans Plank

2.13. Emphatically indefinite While the repetition of an article in Bavarian lends special emotional force to the entire NP, the effect of repeated marking for definiteness, or rather indefiniteness, may also be confined to this category itself. To place emphasis on indefiniteness as such has been claimed to be an effect that is produced by multiple marking. According to Krámský (1972: 142–145), who gives no source and does not identify the relevant dialects, Kurdish (Iranian) has two ways of marking plural indefiniteness, the suffix (or enclitic) -n@, derived from the indefinite pronoun hın@ ‘some, several’, and this indefinite pronoun itself, serving as an indefinite article. The indefinite suffix and article are used interchangeably under normal circumstances, but they combine when the NP’s indefiniteness needs emphasizing: (139) a. b. c.

xort-n@ youth-PL.INDEF hın@ xort some ([sm]) youth(s) " hın@ xort-n@ some youth-PL.INDEF

In Ubykh (North-West Caucasian), indefiniteness is marked by the omission of the definite prefix/proclitic or by the cardinal numeral ‘one’ used as an indefinite article, which can be an independent word or a proclitic. In order to emphasize the indefiniteness of an NP a further indefinite element can be added (enclitically or by way of compounding) to the noun in an NP already equipped with an indefinite article (140b); and if the NP has further constituents, the indefinite article can, in proclitic form, be repeated with all of them for particularly strong emphasis (140c), and it can also be combined with the added indefinite itself (140d) (Krámský 1972: 152, after Dumézil 1931: 13–16).43 (140) a. b.

c.

d.

za-t@t a-man za-t@t-g◦ ara a-man-INDEF ‘a (certain) man’ za-ˇc’@@ za-bGa-s za-t@t-g◦ ara a-horse a-on-sitting a-man-INDEF ‘a (certain) man sitting on horseback’ za-g◦ @ˇcáq’a za-g◦ ara a-word a-INDEF ‘a (certain) word’

✐

✐ ✐

✐

✐ 377

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #401

✐

✐

Double articulation

377

There are no indications of added structural complexity in NPs in Kurdish and Ubykh with more than one marker of indefiniteness (unless za-g◦ ara in (140d) is to be taken for the independent indefinite pronoun, which is indeed formed by combining these two components, and the whole expression accordingly consists of two NPs in apposition: ‘a word, some(thing)’). What should perhaps not be taken for granted is that the purpose of such double marking really is to lay emphasis on indefiniteness; judging by the translations provided, it might also be a way of endowing indefinite NPs with specific reference.

2.14. Specific reference or discourse prominence In fact, it is not uncommon for specific reference, a category rarely if ever endowed with marking entirely of its own, to be expressed by double marking for (in)definiteness. In Sinhalese (Indo-Aryan), this is done by combining markers of indefiniteness, viz. the suffixal (or enclitic) indefinite article grammaticalized from the numeral ‘one’ and that numeral itself (a definite article is lacking; Geiger 1938: 115–116): (141)

ek-goviy-ek one-cultivator-INDEF ‘a certain cultivator’

Elsewhere markers of indefiniteness and definiteness are combined for this purpose. Thus, while non-specific indefiniteness is simply marked by prefix (or proclitic) ši- in Moroccan Arabic, specific indefiniteness requires two forms, the indefinite article waèed (identical to the numeral ‘one’) plus the regular definite prefix (Harrell 1962):44 (142) a.

b. c.

ši-ktab INDEF-book ‘some book (or other)’ l-bent DEF-girl waèed l-bent a DEF-girl ‘a (specific) girl’

Direct objects in Kirghiz (Northern Turkic) show a four-way contrast: when inherently uncommitted as to definiteness they are unarticulated and in the zero-marked absolutive (143a); when indefinite they press the numeral ‘one’ into service as an article (143b); when definite they take the accusative ending (143c); and for specific reference they avail themselves of both the indefinite article and the definite accusative ending at the same time (143d) (Krámský 1972: 170).45

✐

✐ ✐

✐

✐ 378

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #402

✐

✐

378

Frans Plank

(143) a.

b.

c.

d.

alma ˇjedim apple I.ate ‘I ate a/the apple; I was apple-eating’ bir alma ˇjedim a apple I.ate ‘I ate an (i.e. any) apple’ ˇjedim alma-nı apple-ACC.DEF I.ate ‘I ate the apple’ ˇjedim bir alma-nı a apple-ACC.DEF I.ate ‘I ate some specific apple’

However, as is pointed out in Comrie (1981: 169) apropos of Tajik (South-West Iranian), what direct objects like that in (143c) are marked for may not be definiteness per se but the relevance of an NP for the subsequent discourse. If the sole business of articles and equivalent bound forms is taken to be the distinction of definite and indefinite (‘referent assumed or not assumed by the speaker to be identifiable by the addressee’), then NPs like the Tajik specimen in (144) would not strictly speaking qualify as doubly articulated, because the precise contribution of the “definite” accusative ending is to signal referential prominence; but that would arguably be too narrow a view on definiteness and articles, drawing a strict line through an overall semantic-pragmatic domain where several focal areas shade into each other almost imperceptibly. (144)

parda-e-ro curtain-SG.INDEF-ACC.DEF ‘a certain curtain’

On the present evidence, it seems that at least one of the markers which may join forces in simple, unamplified NPs in order to render the NP’s reference specific or its discourse status prominent must be a bound form.

2.15. Doubly (in)definite for no (good) reason other than the availability of several markers Markers of (in)definiteness may also combine redundantly rather than for a semantic or pragmatic purpose. As in amplified NPs, a mechanism that may occasion such vacuous over-articulation is that of blending: when there are different forms and/or positions for markers of (in)definiteness, a single NP may conflate such alternatives. Like Kurdish (§2.13), Modern Persian is another Iranian language to have two overt forms for indefiniteness: the enclitic -i (grammaticalized from the Old Iranian

✐

✐ ✐

✐

✐ 379

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #403

✐

✐

Double articulation

379

numeral ‘one’) and the article yek (identical to the current numeral ‘one’), favoured in formal and colloquial style respectively (145b/c). The bare noun without any suffix or determiner can be interpreted as indefinite or definite depending on the context (145a). Especially in colloquial language both indefinite markers may also show up simultaneously (145d), but, unlike in Kurdish or also Sinhalese (§2.14), with no attendant difference in meaning or force (Lazard 1992: 66, 71f.):46 (145) a. b. c. d.

ketâb ‘book, a (or some) book(s), the book (in question)’ ketâb=i book=INDEF yek ketâb a book yek ketâb=i a book=INDEF ‘a book, whatever book, a certain book’

When the noun is modified by an adjective, the indefinite enclitic may attach to the adjective or (without ezâfe) to the noun (146a/b), and in both cases there may also be an indefinite article (146d/e), but the indefinite article apparently does not occur without its clitic partner (146c): (146) a. b. c. d. e.

ketâb-e kamyâb=i book-EZÂFE rare=INDEF ketâb=i kamyâb book=INDEF rare *yek ketâb-e kamyâb *a book-EZÂFE rare yek ketâb-e kamyâb=i a book-EZÂFE rare=INDEF yek ketâb=i kamyâb a book=INDEF rare

2.16. Sole nouns with layered (in)definiteness marking In principle it may also be the normal and only legitimate state of affairs in simple NPs for unaccompanied nouns to be doubly marked for plain definiteness or indefiniteness. Languages with such a penchant for overdoing articulation appear to be rare, though. And it is not always clear, to speakers as well as grammarians, how the relevant patterns are best analysed. What may happen now and again is that two generations of articles come to cluster around nouns; but since such co-occurring forms should be at far-apart stages in what

✐

✐ ✐

✐

✐ 380

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #404

✐

✐

380

Frans Plank

has been claimed to be the typical life cycle of articles – demonstrative > definite article > marker of definite/indefinite specific reference > gender/class marker or mere noun marker (Greenberg 1978, 1991), and mutatis mutandis for indefinites – they will hardly qualify as instances of essentially the same thing being done twice to a simple NP.47 2.16.1. Where permissible, genuine overmarking, being recognizable as such synchronically, may be something that not all nouns (or other relevant words) of a language are prepared to engage in. Thus, while the general mode of definiteness marking with Albanian nouns is inflectional (147), a subset of nouns that comprises kin terms as well as zot ‘master’ and zojë ‘lady’ in addition takes the prepositive “article” (148), whose main function otherwise is to link attributives to their heads, regardless of their definiteness (Buchholz & Fiedler 1987: 200, 290f.). (147) a. b. (148) a. b.

djal-i boy-SG.NOM.DEF një djalë-Ø a/one boy-SG.NOM.INDEF i nip-i the nephew-SG.NOM.DEF një nip-Ø a/one nephew-SG.NOM.INDEF

It has been shown above (§2.6) that the attributive article does have a share in the coding of definiteness, and it follows from the contrast with indefinites (148a/b) that marking definiteness indeed is what this article is doing with kin terms, in concert with nominal inflection.48 It is also prone to do something else, however. Presumably owing to the article’s redundancy as a marker of definiteness, it has been reinterpreted as a 3rd person possessive pronoun, and as such has become equivalent to genuine possessive pronouns, except that unlike these it does not distinguish gender and number of the possessor: (149) a. b.

i nip-i the/his/her/their nephew-SG.NOM.DEF nip-i i tij / saj / tyre nephew-SG.NOM.DEF ART his / her / their

Another way of dealing with the redundant definite article turned possessive pronoun in recent Albanian is to omit it in contexts that suggest a natural identification of the possessor, which partly realigns kin terms with ordinary nouns that are not marked twice for definiteness:

✐

✐ ✐

✐

✐ 381

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #405

✐

✐

Double articulation

(150)

381

Vajti te (i) nip-i he.has.gone to (his/the) nephew-SG.NOM.DEF ‘He has gone to his nephew’

Cardinal numerals, when occurring on their own in an NP, are another kind of word to be doubly marked for definiteness in Albanian (Buchholz & Fiedler 1987: 233), without reinterpretation or loss threatening to befall the prepositive article: (151)

të gjashtëmbëdhjetë-t ART sixteen-PL.NOM.DEF ‘the sixteen’

2.16.2. It is also conceivable for regular overmarking to be completely general, affecting all articulable nouns (or other relevant words) under all circumstances. This is indeed what we appear to find again in Albanian, where all nouns are obligatorily distinguished as definite and indefinite by their inflection, and when they are indefinite they are accompanied, frequently if not obligatorily, by the indefinite article, as was seen in (147b) and (148b). In Arabic it is definiteness which arguably has two exponents even in unamplified NPs, one negative and the other positive. Many singular and most plural inflections of nouns signal indefiniteness by nunation, whose absence is therefore a fairly reliable indicator of definiteness; and when definite, an unnunated noun requires the definite prefix (or proclitic): (152) a. b.

kalbu-n dog-INDEF al-kalbu-Ø DEF-dog-DEF

And to rehash yet another language of those given to re-articulation when NPs are amplified (§2.6), Swedish might be argued on the same grounds to mark indefiniteness twice on bare nouns. By virtue of the absence of a definite suffix a noun is already recognizable as indefinite, and this status is confirmed by the indefinite article: (153) a. b.

hus-et house-DEF ett hus-Ø a house-INDEF

What renders the analysis of such nouns as doubly (in)definite somewhat suspect is that their overt marking is always simple. It is only if the absence of an exponent of definiteness (Albanian, Swedish) or of indefiniteness (Arabic) is taken to contribute to the marking of the opposite category – indefiniteness in Albanian and Swedish

✐

✐ ✐

✐

✐ 382

✐ $Id: M-plank.tex,v 1.29 2002/10/31 14:01:27 eyrich Exp $

| 7/11 19:11 | #406

✐

✐

382

Frans Plank

and definiteness in Arabic – that this latter category comes out as doubly marked. However, when zeroes are as systematic as in our three cases, perhaps they wish to be taken literally and do not want their nothingness to be imbued with meaning. There is a further question of analysis here, even if the present evidence makes it seem academic. If a language were like Albanian, Arabic, or Swedish except that the zeroes in examples such as (147b/148b), (152b), (153b) were replaced by something more tangible, would this perforce amount to genuinely double, repeated marking of nouns for (in)definiteness? An obvious alternative would be to analyse such patterns as involving simple marking, superficially split into two interdependent parts. Presumably, owing to its very generality, such bipartite marking would be prone to be unified, with the two parts that occur in tandem on all nouns being linked to each other as the extended and possibly discontinuous exponents of a single marker. Another, syntactic way for co-occurring forms of the same categories to be regularly partnered, when at least one of them is bound and its host is in syntactic construction with its partner, is to be distributed by rules of government or agreement. Accordingly, when nouns inflect for definiteness and indefiniteness, this would naturally be analysed as being due to agreement or government if there are words (i.e., articles) in their permanent company which are also distinguished as definite and indefinite.

2.17. Fossilized markers Finally, an article may be added to a noun whose basic lexical form includes what used to be an article but was re-analysed as an integral part of the noun. Synchronically, therefore, such nouns do not count as articulated and are no liability to their NPs. Cases in point are Maltese ilma and nu˙cc˙ ali, which historically consist of the definite prefix and a noun stem (154) (Sutcliffe 1936: 19), and English newt and nickname, which historically incorporate a fragment of an article, with the final consonant of the prevocalic form of the indefinite article having been missegmented as the initial consonant of these nouns (155).49 (154) a. b. (155) a. b.

l-ilma DEF-water in-nu˙cc˙ ali DEF-glasses a newt a nickname

< < <