The Syntax-Morphology Interface: A Study of Syncretism (Cambridge Studies in Linguistics)

This page intentionally left blank THE SYNTAX–MORPHOLOGY INTERFACE Syncretism – where a single form serves two or mo...

Author: Matthew Baerman | Dunstan Brown | Greville G. Corbett

116 downloads 1268 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

This page intentionally left blank

THE SYNTAX–MORPHOLOGY INTERFACE

Syncretism – where a single form serves two or more morphosyntactic functions – is a persistent problem at the syntax–morphology interface. It results from a ‘mismatch’ whereby the syntax of a language makes a particular distinction, but the morphology does not. This pioneering book provides the first full-length study of inflectional syncretism, presenting a typology of its occurrence across a wide range of languages. The implications of syncretism for the syntax–morphology interface have long been recognized: it argues either for an enriched model of feature structure (thereby preserving a direct link between function and form), or for the independence of morphological structure from syntactic structure. This book presents a compelling argument for the autonomy of morphology, and the resulting analysis is illustrated in a series of formal case studies within Network Morphology. It will be welcomed by all linguists interested in the relation between words and the larger units of which they are a part. M A T T H E W B A E R M A N is Research Fellow at the University of Surrey. He is author of The evolution of fixed stress in Slavic (1999) and has written on the subject of syncretism, morphology and prosody for a wide range of journals including Language, Lingua, Russian Linguistics, and Natural Language and Linguistic Theory. D U N S T A N B R O W N is Senior Lecturer in Linguistics at the University of Surrey and currently Treasurer of the Linguistics Association of Great Britain, with research interests in computational linguistics, linguistic typology and morphology. He has written widely on these subjects for journals such as Language, Lingua, Journal of Slavic Linguistics and Yearbook of Morphology. G R E V I L L E G . C O R B E T T is Distinguished Professor of Linguistics at the University of Surrey, and formerly President of the Linguistics Association of Great Britain. Among other books he is author of Gender (Cambridge University Press, 1991), and Number (2000), and co-editor of Heads in Grammatical Theory (1993).

In this series 66 67

ANTHONY R. WARNER: P. H.

MATTHEWS:

English auxiliaries: structure and history Grammatical theory in the United States from

Bloomfield to Chomsky 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99

LJILJANA PROGOVAC:

Negative and positive polarity: a binding approach Ergativity Y A N H U A N G : The syntax and pragmatics of anaphora K N U D L A M B R E C H T : Information structure and sentence form: topic, focus, and the mental representation of discourse referents L U I G I B U R Z I O : Principles of English stress J O H N A . H A W K I N S : A performance theory of order and constituency A L I C E C . H A R R I S A N D L Y L E C A M P B E L L : Historical syntax in cross-linguistic perspective L I L I A N E H A E G E M A N : The syntax of negation P A U L G O R R E L : Syntax and parsing G U G L I E L M O C I N Q U E : Italian syntax and universal grammar H E N R Y S M I T H : Restrictiveness in case theory D . R O B E R T L A A D : Intonational morphology A N D R E A M O R O : The raising of predicates: predicative noun phrases and the theory of clause structure R O G E R L A S S : Historical linguistics and language change J O H N M . A N D E R S O N : A notional theory of syntactic categories B E R N D H E I N E : Possession: cognitive sources, forces and grammaticalization N O M T E R T E S C H I K - S H I R : The dynamics of focus structure J O H N C O L E M A N : Phonological representations: their names, forms and powers C H R I S T I N A Y . B E T H I N : Slavic prosody: language change and phonological theory B A R B A R A D A N C Y G I E R : Conditionals and predication: time, knowledge and causation in conditional constructions C L A I R E L E F E B V R E : Creole genesis and the acquisition of grammar: the case of Haitian Creole H E I N Z G I E G E R I C H : Lexical strata in English: morphological causes, phonological effects K E R E N R I C E : Morpheme order and semantic scope: word formation and the Athapaskan verb A P R I L M C M A H O N : Lexical phonology and the history of English M A T T H E W Y . C H E N : Tone Sandhi: patterns across Chinese dialects G R E G O R Y T . S T U M P : Inflectional morphology: a theory of paradigm structure J O A N B Y B E E : Phonology and language use L A U R I E B A U E R : Morphological productivity T H O M A S E R N S T : The syntax of adjuncts ELIZABETH CLOSS TRAUGOTT and RICHARD B. DASHER: Regularity in semantic change M A Y A H I C K M A N N : Children’s discourse: Person, space and time across languages D I A N E B L A K E M O R E : Relevance and linguistic meaning: The semantics and pragmatics of discourse markers R. M. W. DIXON:

100 101 102 103 104 105 106 107 108 109

I A N R O B E R T S A N D A N N A R O U S S O U : Syntactic change: a minimalist approach to grammaticalization D O N K A M I N K O V A : Alliteration and sound change in early English M A R K C . B A K E R : Lexical categories: verbs, nouns and adjectives C A R L O T A S . S M I T H : Modes of discourse: the local structure of texts R O C H E L L E L I E B E R : Morphology and lexical semantics H O L G E R D I E S S E L : The acquisition of complex sentences S H A R O N I N K E L A S a n d C H E R Y L Z O L L : Reduplication: doubling in morphology S U S A N E D W A R D S : Fluent aphasia B A R B A R A D A N C Y G I E R a n d E V E S W E E T S E R : Mental spaces in grammar: conditional constructions MATTHEW BAERMAN, DUNSTAN BROWN AND GREVILLE G . C O R B E T T : The syntax–morphology interface: a study of syncretism.

Earlier issues not listed are also available

CAMBRIDGE STUDIES IN LINGUISTICS General editors: P. AUSTIN, J. BRESNAN, B. COMRIE, S. CRAIN, W. DRESSLER, C. J. EWEN, R. LASS, D. LIGHTFOOT, K. RICE, I. ROBERTS, S. ROMAINE, N. V. SMITH

The Syntax–Morphology Interface A Study of Syncretism

T H E SY N T A X – MORPHOLOGY INTERFACE A STUDY OF SYNCRETISM

MATTHEW BAERMAN University of Surrey

DUNSTAN BROWN University of Surrey

GREVILLE G. CORBETT University of Surrey

   Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo Cambridge University Press The Edinburgh Building, Cambridge  , UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521821810 © Matthew Baerman, Dunstan Brown, and Greville G. Corbett 2005 This publication is in copyright. Subject to statutory exception and to the provision of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published in print format 2005 - -

---- eBook (EBL) --- eBook (EBL)

- -

---- hardback --- hardback

Cambridge University Press has no responsibility for the persistence or accuracy of s for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

To our families

Contents

Preface List of abbreviations and symbols

page xv xvii

1

Introduction

1.1 1.2 1.3

History of the notion Delimiting the notion of syncretism Scope of the investigation 1.3.1 Typological methodology 1.3.2 Selection of forms Accidental versus systematic homophony Using this book 1.5.1 Supporting materials 1.5.2 Glossing conventions 1.5.3 Structure of the book

3 4 7 7 8 9 10 10 11 12

2

Characteristics of syncretism

13

2.1

2.5

Syncretic paradigms 2.1.1 Types 2.1.2 Implications Domains of comparison 2.2.1 Morphological classes 2.2.2 Feature values Morphological characteristics A typology of interpretations 2.4.1 Syncretism as neutralization 2.4.2 Syncretism as uninflectedness 2.4.3 Canonical syncretism Conclusion

13 13 17 17 17 19 23 27 28 30 33 35

3

Cross-linguistic typology of features

37

3.1

Case 3.1.1 3.1.2

38 38 40

1.4 1.5

2.2

2.3 2.4

Introduction Types of case syncretism

1

xi

xii

3.2

3.3

3.4

3.5

3.6

3.7

3.8

Contents 3.1.3 Conclusion Person 3.2.1 Introduction 3.2.2 Patterns of syncretism 3.2.3 Directional effects 3.2.4 Diachrony 3.2.5 Summary Person syncretism in two-place verbs 3.3.1 Introduction 3.3.2 Syncretism of subject person 3.3.3 Syncretism of object person 3.3.4 Other patterns 3.3.5 Summary Gender 3.4.1 Introduction 3.4.2 Gender and number 3.4.2.1 Smaller systems 3.4.2.2 Larger systems 3.4.3 Syncretism restricted by target 3.4.4 Summary Number 3.5.1 Introduction 3.5.2 Values 3.5.3 Directionality 3.5.4 Summary Tense-aspect-mood 3.6.1 Introduction 3.6.2 Affix suppression 3.6.3 Syncretic affixes 3.6.4 Compound systems 3.6.5 Summary Polarity effects 3.7.1 Introduction 3.7.2 Morphological systematicity 3.7.3 Semantic systematicity 3.7.4 Summary The interaction of features 3.8.1 Introduction 3.8.2 Syncretisms and their contexts 3.8.3 Nominal feature interactions 3.8.3.1 Typologies of interaction 3.8.3.2 Exploring interaction 3.8.3.3 Constraints on nominal features 3.8.4 Verbal feature interactions

56 57 57 59 63 70 75 75 75 76 79 80 81 81 81 82 83 86 90 91 92 92 93 94 95 95 95 96 100 101 103 103 103 105 108 111 111 111 112 113 113 114 118 119

Contents

xiii

3.9

3.8.4.1 Verbal paradigms 3.8.4.2 Exploring the verb data 3.8.5 Interpreting the generalisations 3.8.5.1 Number, case and gender on nominals 3.8.5.2 TAM and agreement on verbs 3.8.6 Summary Conclusion

120 121 123 123 123 124 124

4

Formal representation

126

4.1 4.2

Introduction Defining sets of values 4.2.1 Natural classes 4.2.2 Unnatural classes 4.2.3 A note on polarity effects Symmetrical versus directional rules 4.3.1 Convergent bidirectional syncretism 4.3.2 Divergent bidirectional syncretism 4.3.3 Symmetrical versus directional rules: a summary 4.3.4 Ranked constraints as an alternative to directional syncretism Possible constraints on syncretism 4.4.1 Variants on hierarchical structures 4.4.2 Carstairs (1987), Carstairs-McCarthy (1998a, b) 4.4.3 Impoverishment 4.4.4 Stump (2001) and Zwicky (2000) Summary 4.5.1 Predictions and counter-examples 4.5.2 Towards a model of syncretism

126 126 126 131 132 133 136 139 144

5

Formal framework and case studies

171

5.1

Network Morphology and syncretism 5.1.1 Inferential-realizational theories and morphology 5.1.2 Default inheritance 5.1.3 Underspecification and semantic naturalness 5.1.4 Systematicity in Network Morphology Case study 1: Dhaasanac Case study 2: The Dalabon verbal system 5.3.1 The structure of the Dalabon intransitive paradigm 5.3.2 The Dalabon transitive paradigm and the inadequacy of underspecification 5.3.3 The Dalabon transitive paradigm: a generalized referral analysis 5.3.4 The verbal hierarchy 5.3.5 The shape of the verbal paradigm

172 175 177 180 182 183 186 187

4.3

4.4

4.5

5.2 5.3

145 150 150 151 160 163 166 166 169

188 194 199 200

xiv

Contents

5.5

5.3.6 Referral of 1 > 2sG to 3 > 2sG 5.3.7 Referral of 2 > 1 to 3 >1 5.3.8 Dalabon: summing up Case study 3: The Russian nominal system 5.4.1 Domains of syncretism 5.4.1.1 Phonologically determined ‘syncretism’ 5.4.1.2 Lexically determined syncretism 5.4.1.3 Morphologically determined syncretism 5.4.1.4 Candidates for syntactically determined syncretism 5.4.2 Orthogonal specification of syncretism Conclusion

203 203 203 204 206 206 206 207 213 216 217

6

Conclusion

219

6.1 6.2 6.3

Taking stock Results Consequences

219 220 221

5.4

Appendix 1: Case syncretism in the World Atlas of Language Structures sample Appendix 2: Person syncretism in the World Atlas of Language Structures sample Appendix 3: Syncretism in two-place verbs in the World Atlas of Language Structures corpus Appendix 4: DATR fragment for Dhasaanac case study Appendix 5: DATR fragment for Dalabon case study Appendix 6: DATR fragment for Russian case study References Author index Language index Subject index

223 228 233 236 242 248 254 271 275 279

Preface

This book has an interesting history of collaboration. It began life in research done by Greville Corbett and Norman Fraser on the morphology of Russian, starting in 1990, research which was inspired by the work of Roger Evans and Gerald Gazdar on DATR. The ESRC and Leverhulme Trust provided funding, which brought Dunstan Brown and Andrew Hippisley to Surrey, and the work developed into a more general theoretical framework, Network Morphology. We found syncretism of increasing importance in the development of the framework and gave presentations at the following places: Krems (Austria), University of Sussex, Linguistics Association of Great Britain (at the University of Surrey), University of California (Berkeley), Gregynog (Wales), HeinrichHeine-Universita¨t (Du¨sseldorf), University of Edinburgh, University of Cologne, University of Helsinki, La Trobe University, Norsk Forening for Spra˚kvitenskap (Oslo), Institutt for Østeuropeiske og Orientalske Studier (University of Oslo), Moscow University, University of Oxford, Cornell University, Twelfth International Conference on Historical Linguistics (University of Manchester), Conference on Lexical Structures (Wuppertal), British Association for Slavonic and East European Studies (Cambridge), University of Sheffield, University of Essex, University of Pennsylvania, Leipzig University, Association for Linguistic Typology (University of Amsterdam), Second Mediterranean Meeting on Morphology (University of Malta), Max Planck Institute for Psycholinguistics (Nijmegen), Second Winter Typological School (Istra, Moscow district), Ninth International Morphology Meeting (Vienna), University of California (Santa Barbara), University College London, Second Northwest Conference on Slavic Linguistics (Berkeley), Stockholm University, 37th Meeting of the Chicago Linguistic Society, Fakultetets Forsknings Fredage (University of Copenhagen), University of Melbourne, Scandinavian Slavists’ Summer School (Kunga¨lv, Sweden), University of Leeds, School of Oriental and African Studies (London), University of Manchester Institute of Science xv

xvi

Preface

and Technology, Lancaster University, University of Catania, Second International Seminar ‘Computer Treatment of Slavonic Languages’ (Bratislava), and University of York. We are very grateful for all the comments we have received on these occasions. We applied for funding to work specifically on syncretism within this framework, which allowed Matthew Baerman to join the Surrey Morphology Group. He undertook the careful typological work which led to the Surrey Syncretisms Database, and which is a basis for the book. Collaboration with Nicholas Evans on Dalabon added an important impetus to the work. As it became clear that the research on syncretism had more substance than could fit into a journal article, Baerman’s role became increasingly important. He is the book’s first author. Brown and Corbett are together the second author, with Brown’s role being particularly significant in the development of the formal side of Network Morphology, while Corbett’s role was most important at the start of the project. Portions of this book have been adapted from previously published material, specifically Evans, Brown and Corbett (2001) (Chapter 5: x5.3); Baerman, Brown and Corbett (2002b) (Chapter 3: x3.1); Corbett, Baerman and Brown (2002) (Chapter 5: x5.4.1); Baerman (2005) (Chapter 3: x3.2, Chapter 4: x4.4.2 and x4.5); and Baerman (2004) (Chapter 1: x1.5 and Chapter 4: x4.3). The material here supercedes the earlier works. We are very grateful to our friends and colleagues who read the book in draft and gave us helpful comments from their different perspectives: Jim Blevins, Jonathan Bobaljik, Andrew Carstairs-McCarthy, Martin Haspelmath, Andrew Spencer, Greg Stump. The following gave helpful feedback on specific sections: Helma van den Berg, Michael Cysouw, Nicholas Evans, Roger Evans, while Lisa Mack substantially improved the presentation of the draft. For help with the maps we are indebted to Hans-Jo¨rg Bibiko, and we thank Tom Khabaza and the Clementine software for help with the data analysis in x3.8, and Marina Chumakina for assistance with Russian data. We are very grateful to the ESRC for funding, under grant R000237939 and partially under grants R000271235 and RES000230082. The University of Surrey Research Committee also provided timely support.

Abbreviations and symbols

1 2 3 A ABESS ABL ABS ACC ADIT ADJ ALL AN CAR CAUS CMP COM CONT COORD DAT DEF DES DIS DU EL ERG ESS EXCL F GEN

first person second person third person transitive subject (where forms may differ from those of the intransitive subject) abessive ablative absolutive accusative aditive adjective allative animate caritative causative comparative comitative contactive coordinative dative definite designative disharmonic dual elative ergative essive exclusive feminine genitive xvii

xviii

Abbreviations and symbols

HARM HON HUM ILL IMPRF INAN INCL INDF INESS INS INTR LOC M N NNARR NOM OBJ PER PFV PL PRF PROL PRS PST REL S SBJ SBJV SG SUBORD SUPERESS TAM TR TRI TRANS VOC

harmonic honorific human illative imperfect inanimate inclusive indefinite inessive instrumental intransitive locative masculine neuter non- (e.g. NSG for non-singular) narrative nominative object perlative perfective plural perfect prolative present past relative intransitive subject (where the forms may differ from those of the transitive subject) subject subjunctive singular subordinate superessive tense-aspect-mood transitive trial translative vocative

Abbreviations and symbols -

xy x/y x>y [x] (x)

xix

used to form compound names for morphosyntactic values, e.g. NOM-ABS = ‘nominative-absolutive case’ (a single morphosyntactic case in the language in question) form ‘x’ alternates with form ‘y’ feature value ‘x’ is syncretic with feature value ‘y’ in a transitive verb, ‘x’ is the subject and ‘y’ is the object the feature value ‘x’ has no overt morphological expression (e.g. English dog would be glossed as ‘dog[SG]’) the feature value ‘x’ is inherent to the lexeme, and has no overt expression (e.g. French plage would be glossed as ‘beach(F)’)

1 Introduction

We might expect that a language’s sentence-structure and word-structure would mesh rather straightforwardly. For instance, if the syntax of a particular language distinguishes different arguments of the verb, and the morphology distinguishes different cases, it seems natural to assume that the two systems will line up. In real languages the situation is often more complex. One of the most persistent and interesting problems at this syntax–morphology interface is syncretism. As a first informal characterization, syncretism is the situation where the morphology ‘lets down’ the syntax. To make that more concrete, let us take some Russian examples. (Normally we shall give detailed glosses but here, so as not to build our assumptions into the examples, we shall give only the basics.) (1)

Masˇ a cˇitaet knigu Masha reads book ‘Masha reads a book.’

(2)

Na stole lezˇit kniga on table lies book ‘A book lies on the table’, ‘There’s a book on the table.’

These sentences are representative, in that they show what they seem to show. Russian distinguishes subject from object in its syntax, by a variety of means (for example, the verb agrees with its subject but not with its object). This appears to be reflected in the inflectional morphology. In (1) we have knigu ‘book’, in the accusative case, as opposed to kniga in (2) when it is in the nominative. (Similarly Masˇa in (1) is in the nominative.) As any reasonable non-linguist would expect, the two systems work hand in hand to distinguish subject and object and to facilitate the task of the hearer. But now compare: (3)

Masˇ a cˇitaet pis 0 mo Masha reads letter ‘Masha reads a letter.’ 1

2 (4)

The Syntax–Morphology Interface Na stole lezˇit pis 0 mo on table lies letter ‘A letter lies on the table’, ‘There’s a letter on the table.’

Here the syntactic structures are as in the earlier examples, but the noun fails to show the expected morphological distinction. We say that pis 0 mo ‘letter’ in (3) and (4) shows syncretism of nominative and accusative case. We understand syncretism therefore as a mismatch between syntax and morphology. We know that Russian syntax requires reference to subject and object. This is reflected in the morphosyntactic category of case, which distinguishes nominative and accusative (as in (1) and (2) above). However, the morphology of pis 0 mo in (3) and (4) fails to make this distinction. The key components of the definition are: a. a morphological distinction which is syntactically relevant (i.e. it is an inflectional distinction) b. a failure to make this distinction under particular (morphological) conditions c. a resulting mismatch between syntax and morphology. Thus syncretism is the failure to make a morphosyntactically relevant distinction. A good way to look at it is to say that examples (1) and (2) set up the expectation that there will be two forms of pis 0 mo in (3) and (4). Syncretism is the breaking of that expectation; the nominative singular and the accusative singular of pis 0 mo are identical. As Spencer (1991: 45) puts it ‘a single inflected form may correspond to more than one morphosyntactic description.’ A similar definition is: ‘Identity in form between two grammatically different inflections’ (Trask 1997: 215). There are various questions to be asked about our example pis 0 mo ‘letter’. For example, is it an odd exception, going against the general trend? No, there are thousands of nouns in Russian which behave similarly; there are also many thousands of the kniga ‘book’ type. And pis 0 mo does not fail to draw certain other inflectional distinctions. It has a distinct locative, as shown by: v pis 0 me ‘in the letter’. We might think that we could simply divide Russian nouns into those which distinguish nominative and accusative and those like pis 0 mo which do not. But here we find that kniga ‘book’, which marks the distinction in the singular, fails to do so in the plural (both forms are knigi). We shall investigate which distinctions can fail to be drawn. We have seen an instance where case is involved, but there are several other

Introduction

3

inflectional distinctions which may be treated similarly. We shall ask when this occurs. It may be that our expectation is based on some lexemes (like kniga) and is not met by others (like pis 0 mo). Or it may be that other features provide the circumstances (kniga ‘should’ distinguish nominative and accusative, as the singular shows, but it fails to do so in the plural). It is not self-evident that all the phenomena which fall under the broad umbrella of syncretism should be modelled in the same way, and so we shall look carefully at the arguments for particular ways of treating different instances of syncretism. While we shall be inclusive in our coverage we shall ensure that we are indeed dealing with inflection. That is, the expectation of a difference in form must arise from the syntax and morphology of the given language. Languages may have a distinct locative case, as in Russian, but there is no language-internal evidence for such a morphological case in languages like English. We shall not, therefore, treat English book as syncretic between nominative and locative. Nor shall we be concerned with derivational morphology. There are, of course, interesting coincidences of form in derivational morphology, but we do not have the same expections of consistency and completeness for derivational morphology as for inflection. Syncretism is also distinct from lexical homonymy, where there is a single form with distinct meanings (as in bank (of river) and bank (financial institution); this could be characterised as the lexicon letting down the semantics.

1.1

History of the notion

The term ‘syncretism’ ultimately descends from the Greek tio´& ‘union, federation of Cretan communities’ (Liddell and Scott 1996), referring to the practice that the continually feuding Cretan communities had of laying aside their differences and banding together in the face of a common enemy. In post-classical times Erasmus of Rotterdam reintroduced the term, using it to designate ‘the coherence of dissenters in spite of their difference of opinions, especially with reference to theological divisions’ (Herbermann, Pace, Pallen, Shahan and Wynne 1907–18). At some point, through learned folk etymology, the term was confounded with "ńni ‘to mix’ and its derivatives (e.g. ´ to& ‘mixed’), so acquiring a more general meaning of a mixture of originally different elements or viewpoints, typically with respect to religion (Grimm and Grimm 2004). In the nineteenth century, use of the term became

The Syntax–Morphology Interface

4

fashionable in other realms as well. Pott (1836) is generally credited with introducing the term into linguistics (Curtius 1863: 160; Wackernagel 1924: 32; Hjelmslev 1935–7: 59),1 where it is used to mean the diachronic collapse of originally distinct inflectional forms, either through merger of the forms, or through the merger of their underlying functions. Pott himself uses the term only in passing (p. 638),2 but it is more generously applied in the index (p. 792) – written not by Pott but by Heinrich Bindseil – where it refers the reader to sections discussing the merger (typically through sound change) both of case marking in nouns and person marking in verbs. In spite of this early use of the term, it did not enter into general use until the 1890s (Meiser 1992: 212, fn. 2). In the twentieth century, this diachronic approach to syncretism was recast in synchronic terms within the framework of structuralist linguistics (Hjelmslev 1935–7: 60; Jakobson 1936 [1971]: 67). On this approach, syncretism involves the contrast not between an earlier and a later stage of a language, but between an underlying system and its concrete realization. This is how we understand syncretism in the present work, though of course we shall not ignore the diachronic processes that may have led to such a state of affairs. 1.2

Delimiting the notion of syncretism

Some authors have suggested that the term ‘syncretism’ should be reserved for the products of certain diachronic developments. Roughly speaking, the diachronic merger of forms within an inflectional paradigm can have two sources, either as the result of blind phonological change, or the result of a more complex morphosyntactic readjustment. Blind phonological change can be illustrated by the merger of nominative and accusative singular in first declension nouns in Vulgar Latin (5), a result of the regular loss of wordfinal -m. Note that in other declension classes the two forms remained distinct, as their difference was not solely due to the presence or absence of -m. 1

2

Curtius (1863: 160) attributes to Pott the notion of a ‘syncretic case’, i.e. a case historically descended from two or more cases, as with the Greek dative or genitive (see (6)). However, the reference he gives (‘Pott Et. Forsch. I1, 22’, namely Pott 1859: 22), although it deals with this topic, does not include the term ‘syncretic case’ as such. ‘Im Lat. sog. Abl. und Dat. Plur. scheinen die Functionen des eig. Abl. [ . . . ], Instrumentalis [ . . . ], Locativs [ . . . ], und endlich Dativs [ . . . ] vereinigt; dabei wird ebenfalls theilweise Formen-, theilwiese vielleicht bloßer Begriffs-Synkretismus obgewaltet haben.’ (‘In Latin the so-called ablative and dative plural appear to have united the functions of the original ablative, instrumental, locative and dative; here too [as with the Greek dative] we see, in part, syncretism of the forms, and perhaps also, in part, sheer syncretism of the concepts.’)

Introduction

5

(5) Nominative/accusative singular in Vulgar Latin (Coleman 1976: 50–4) Classical Latin first declension

NOM SG ACC SG

second declension

NOM SG ACC SG

third declension

NOM SG ACC SG

Vulgar Latin

>

luna lunam annus annum pater patrem

luna luna annus anno pater patre

Merger resulting from morphosyntactic change is illustrated by the development of the Proto-Indo-European dative and locative singular, which were combined in Ancient Greek into the case traditionally known as the dative (6). By the laws of sound change that applied between Proto-IndoEuropean and Ancient Greek, these two forms should be distinct in o-stem (6) Dative/locative singular in Ancient Greek (Buck 1933: 180–5) Proto-Indo-European o-stem consonant stem

>

Greek

DAT SG LOC SG

-o i -oi, -ei

-o i

DAT SG LOC SG

-ei -i

-i

and consonant-stem nouns, so their merger cannot be attributed to phonological change. Nor can it be attributed to the outright loss of one of the cases, since both the dative and locative have contributed to the syncretic forms: *

*

3

In the o-stems, the Greek form descends from the ProtoIndo-European dative singular. The Proto-Indo-European locative singular is marginally preserved in some adverbialized forms, such as oikoi ‘at home’, originally from oikos ‘house’. In consonant stems, the Greek form descends from the ProtoIndo-European locative singular. The original dative form has been lost completely.3

In the a-stems, these two forms were already syncretic in late Proto-Indo-European, presumably the result of sound change: early Proto-Indo-European dative singular *- a-ei and locative singular *- a-i both developed into *- ai (Szemereńyi 1989: 200).


6

Thus, the Vulgar Latin example represents a superficial inflectional homophony, the result of blind sound change. The Ancient Greek dative singular results from a blending of both the forms and the functions of the Proto-IndoEuropean dative and locative, the result of some fundamental reanalysis of the system of morphosyntactic oppositions. Meiser (1992: 190) and Luraghi (2000) both suggest that the term ‘syncretism’ be restricted to the second kind of change. Luraghi suggests that the product of phonological change should be called ‘homophony’ while Meiser (p. 190) proposes reviving the term ‘synemptosis’, used by ancient Greek grammarians for situations where a morphological expression belonged to several grammatical categories. However, though this distinction is one of undoubted theoretical significance, in practice it is often difficult to draw the line. Consider the merger of ablative and dative singular in second declension nouns in Latin (7). This seems to have been the result of two independent sound changes. On the one hand, final consonants were lost after a long vowel, so that ablative (7)

Ablative and dative singular in Latin (Buck 1933: 176, 181) Proto-Indo-European/ Italic5

second declension (o-stems)

ABL SG DAT SG

first declension (a-stems)

ABL SG DAT SG

-o d -o i -a d -a i

>

Latin -o -o -a -ai (-ae)

singular -o d became -o (Buck 1933: 157). On the other hand, long diphthongs were monophthongized, losing their second element, so that dative singular -o i became -o (Buck 1933: 90). The first change was quite regular and is also found in the first declension (thus, ablative singular - ad became - a). However, the development of the original long diphthongs in Latin turns out to have been erratic. Sometimes they were monophthongized, but in other contexts they were shortened; the conditions which determined which change took place remain obscure (Leumann 1977: 271–2).4 While the dative singular of the second declension underwent the first change, the dative singular of the first declension underwent the second change, with - ai becoming -ai (orthographically -ae), though there 4 5

According to Leumann, V is the expected prepausal reflex, -Vi elsewhere. The Proto-Indo-European a-stem dative singular ending was - ai. The ending - ad is an innovation within Italic, formed on analogy with the ablative of the o-stems.

Introduction

7

is evidence from early inscriptions that monophthongization took place sporadically as well (Buck 1933: 176). Consequently, there is no syncretism between ablative and dative singular in the first declension. Although the merger of ablative and dative singular can be portrayed as the result of sound change, a key element remains unaccounted for, namely why did the long diphthong develop one way in one declension class and another way in the other? For some reason syncretism was favoured in the second declension but not in the first, in a way that does not obviously follow from phonological developments. Such examples show that there is not always a clear distinction between phonological and morphological change, much less a way to classify phenomena whose history remains unknown. It seems useful, then, to retain ‘syncretism’ as a cover term that will apply to all instances of inflectional homophony, regardless of their origin or interpretation; indeed, this is how the term was first used by Pott (and Bindseil) in 1836. 1.3

Scope of the investigation

1.3.1 Typological methodology The core of the present book is a cross-linguistic typological investigation of syncretism, with two complementary goals. On the one hand, we explore the logical space of syncretism: what features may be involved, and what sort of patterns do these describe? On the other hand, we have aimed for a diverse sampling of the world’s languages. In particular, we have brought the evidence of non-Indo-European languages to bear, since these remain relatively under-represented, a legacy of the fact that the notion of syncretism was born in comparative Indo-European studies. To ensure genetic breadth we followed, in part, the selection of languages used for the World Atlas of Language Structures (Haspelmath, Dryer, Gil and Comrie 2005). While genetic breadth can be achieved by the application of apriori criteria, typological breadth – in order to see how many of the logical possibilities are in fact attested – can be attained only by sifting through masses of information, unconstrained by prior notions of what one may find. This is a task which will never and can never be completed, but the present study represents at least an introduction. 1.3.2 Selection of forms In this book we focus our attention on syncretism between inflected whole word forms. In principle, one could speak of syncretism between the

8


individual components of inflected words, e.g. by comparing prefixes to prefixes, stems to stems and suffixes to suffixes, regardless of whether there is homophony at the level of the word form as a whole. For example, in the Nilo-Saharan language Mursi (8), the verb stem is identical in the first and second singular of the indefinite aspect (baaio), and in the first and third singular of the indefinite aspect (baaka). However, in neither aspect is there syncretism at the level of the whole word, because 1SG is distinguished by the prefix ka-. (8) Singular indicative forms of the verb ‘eat’ in Mursi (Turton 1981: 341–2) indefinite aspect 1SG 2SG 3SG

ka-

baagio baagio baato

definite aspect wawawa-

ka-

baaka baaku baaku

Stump (2001: 217) terms this block syncretism, because the pattern obtains within a given block of rules. So long as a word form is transparently segmentable, such an approach has the advantage of widening the field of investigation. But the question of segmentation into components (rule blocks, morphemes or formants) is not always easy to resolve. For example, take Pike’s (1965) analysis of the six present tense forms and the infinitive of the German verb sein ‘to be’ (9). At the level of the whole word, these show 1PL/3PL syncretism alone. (9) INF

1SG 2SG 3SG 1PL 2PL 3PL

German sein ‘to be’ sein bin bist ist sind seid sind

Within these forms, however, eight distinct patterns of identity can be isolated, as shown in (10): the elements b-, s-, -ei-, -ist-, -in-, -i-, -n- and -t

Introduction

9

(10) Patterns of identity in German sein (adapted from Pike 1965: 198) 3SG 2SG 1SG 3PL 1PL 2PL

b b s s s s

INF

i i i i i i i

e e

s s n n n

t t d [t] d [t] d [t]

n

(note that orthographic d in (10) is indistinguishable from t in word-final position) each combine different person-number values. Whatever the merits of such an analysis, it is not one which is compatible with most morphological models. The drawback to such an approach is that the more a word is broken up into components, the more the resulting parts are peculiar to a specific analysis. This is not desirable in the context of a large-scale typological investigation such as ours. On the other hand, the status of the whole word, while hardly self-evident, is nevertheless more uncontroversial than that of such elements as morphemes or formants. By limiting the investigation to whole word forms, we aim to keep the typological part of the investigation theory-neutral. 1.4

Accidental versus systematic homophony

The focus of this book is inflectional morphology: what interests us are instances of inflectional homophony that might be seen as systematic, that is somehow represented in morphological structure. However, it is undoubtedly the case that not all instances of homophony within inflectional paradigms are morphologically encoded as such. Consider the Russian forms in (11).

NOM/ACC SG GEN SG

a. stem-stress ‘place’

b. end-stress ‘wine’

orthographic

phonetic

orthographic

phonetic

mesto mesta

:mje.st :mje.st

vino vina

vji.:no vji.:na

e e

(11)

For the noun mesto ‘place’, the genitive singular form is identical to the nominative/accusative singular, while for vino ‘wine’, the genitive singular is

10


distinct. However, the difference between the two nouns is easily explained in phonological terms. Russian has a general rule whereby /a/ and /o/ are distinguished only under stress. In this declension class, the nominative/accusative singular ending is -o and the genitive -a, as reflected in the orthography. For vino, these endings are stressed, and remain distinct, while for mesto they are unstressed, and hence homophonous. The collapse of the genitive with the nominative/accusative is a superficial by-product of phonology and need not be reflected in a morphological analysis. In general we have excluded such obvious examples of accidental homophony from consideration. However, it should be borne in mind that the question is seldom so clear-cut. On the one hand, a pattern of syncretism may be restricted to a particular phonological environment without there being any generally applicable phonological rule that would account for it. On the other hand, there is evidence that originally accidental homophony may be ascribed by speakers to a morphological rule, a reanalysis which remains covert until revealed by diachronic processes (see Chapter 4: x4.5.1). Therefore, if we have erred in our selection of material for presention, it has been on the side of inclusiveness.

1.5

Using this book

1.5.1 Supporting materials The text is supported by a range of additional materials which have been made available to the reader. There is an annotated bibliography of syncretism (Baerman 2002a), which contains details of 100 items. This is freely available online at: http://www.surrey.ac.uk/LIS/MB/Bibliography.htm. Since this bibliography is available, we can restrict the references section in the book to those items which are discussed in the text. It is important that our account of syncretism is based on a wide range of languages rather than on the few usual suspects. Our work is grounded on an investigation of syncretism in a sample of genetically diverse languages. These data are available in the Surrey Syncretisms Database (Baerman, Brown and Corbett 2002a), which records all instances of syncretism in thirty genetically diverse languages, comprising 1,256 separate entries. This can be searched online at: http://www.smg.surrey.ac.uk. Its rationale is explained in Brown (2001). The provision of this database means that the reader can frame hypotheses about syncretism and investigate them on-line. The essential information for using the database is provided in readme files at the web address given. There is a second database covering

Introduction

11

a more limited phenomenon over a much broader range of languages, cataloguing syncretism of subject person in intransitive verbs in 111 languages. Person marking was chosen for this large-scale survey because, of all inflectional features, it is the one which is most comparable crosslinguistically, the same core values being nearly universally present (Baerman 2002b); this too can be searched over the web at http:// www.smg.surrey.ac.uk/. We participated in the research of the World Atlas of Language Structure (Haspelmath, Dryer, Gil and Comrie 2005) and contributed to maps, found in chapter 3 of the book, keyed to data found in Appendices 1–3. We wish to stress too that our analyses are not ‘cute’ accounts of convenient selections of data. We have worked out full accounts of large and complex morphological systems, within the Network Morphology framework (see Corbett and Fraser 1993, Evans, Brown and Corbett 2001 and references there). We have implemented these analyses, so that others can check that the analyses do indeed account for all the relevant inflectional forms of the languages analysed. Sample fragments are given in Appendices 4–6; the samples are simplified in parts to help the reader to see the essentials. Full versions, to enable the reader to check the detail, are available at http://www.surrey.ac.uk/LIS/SMG/Syncretism. 1.5.2 Glossing conventions We adopt the Leipzig Glossing Rules (available at http://www.eva.mpg.de/ lingua/files/morpheme.html). In some ways these conventions codify best practice, but where different possibilities are found they opt for one alternative. This is helpful: to take a trivial example, there may be more than one reasonable abbreviation for a feature value and it then makes sense to standardize on one. Most of the abbreviations we use are from the Leipzig list, but we have had to add a few (see our full list, p. xvii–xix). Here, for illustration, is an example similar to those discussed earlier: (12)

Na polk-e lezˇ-it knig-a on shelf-LOC.SG lie.PRS-3SG book-NOM.SG ‘A book lies on the shelf’, ‘There’s a book on the shelf.’

If we were writing about word order, for example, this might be considered an adequate gloss. For a book on syncretism, however, it is important to be clear how potential ambiguities are handled in glosses. In example sentences, linguists generally give a morphosyntactic gloss. This means that the syntactic context is taken into account to resolve potential

12


ambiguities. Given the example John bid eighty pounds at the auction yesterday, most linguists would gloss bid as past tense, even though out of context it could be the imperative, among other things. In (12), polke is glossed as locative singular. In principle it could be dative singular or locative singular. We might gloss it as locative singular because the preposition na ‘on’ takes the locative case (as we know from nouns where dative and locative are distinct). Where the theoretical ambiguity matters, as generally it will, we either discuss it explicitly in the text, or we include all the possibilities: LOC SG/DAT SG. Where forms are discussed in the text, then the default is to give all the morphological descriptions, since there is no syntactic context to favour one of them. Note that we are using the slash for alternative glosses; it is available for this use because the other function it may fulfil, namely to gloss fused subject and object markers, is carried out in the Leipzig conventions by ‘>’. 1.5.3 Structure of the book The book progresses from an empirical to a formal investigation of syncretism. In Chapter 2 we lay out the typological parameters and define basic concepts. Chapter 3 is a cross-linguistic survey of syncretism. In xx3.1–3.7 we look at how syncretism affects individual features, and in x3.8 we look at how different features interact. In Chapter 4 we look at the implications these observations will have on how a formal model of morphology accounts for syncretism, investigating some previous proposals. In Chapter 5 we propose our own formal model, within the framework of Network Morphology. The appendices contain the detailed data for the statistical claims made in Chapter 3 and are based on our work for the World Atlas of Language Structures.

2 Characteristics of syncretism

We shall map out the range of data to be considered. We include problems familiar from the widely cited languages, but we shall also considerably extend the scope of the discussion and of the languages investigated. We first look at the patterns of syncretism and their implications (Chapter 2: x2.1), and then examine the domains which allow us to compare paradigms (Chapter 2: x2.2). This permits us to begin considering the types of analysis available (to be discussed fully in Chapter 4). In Chapter 2: x2.3 the important issue of directionality is raised. Then we consider the ‘extreme’ interpretations in Chapter 2: x2.4, namely neutralization and uninflectedness. These prove to describe the easy instances: the more challenging ones lie in between and form the subject of the remaining chapters. 2.1

Syncretic paradigms

2.1.1 Types As we have characterized it, syncretism involves the identity of cells within an assumed morphosyntactic paradigm. Graphically, we shall represent this by generating a complete inflectional paradigm, and enclosing the identical forms within a box.1 Before considering what morphosyntactic values may form the parameters, let us outline a brief typology of the ways paradigmatic cells may be united. These are illustrated below with examples of case syncretism. The different types will turn out to have important consequences for the representation of feature structure. In the simplest pattern, which we call simple syncretism, two or more cells with different values for a feature are merged. For example, in Central Alaskan Yup’ik (1), the absolutive and relative cases are identical in the plural and dual. 1

Note that we do not ascribe any significance directly to the geometry. However, the graphic representation gives a clear indication of the morphosyntactic description of the cells.

13

14


(1) Simple syncretism in Central Alaskan Yup’ik (Jacobson 1995: 469–71)

ABS REL LOC ABL ALL PER CMP

singular

dual

plural

nuna nunam nunami nunamek nunamun nunakun nunatun

nunak nunak nunagni nunagnek nunagnun nunagnegun nunagtun

nunat nunat nunani nunanek nunanun nunatgun nunacetun

‘land’

If simple syncretism is compounded across different environments, the result is nested syncretism. In the West Slavonic language Upper Sorbian (2), a-stem nouns have syncretism of the dative and locative in the singular, while all nominals have syncretism of the dative, locative and instrumental in the dual. Thus the syncretic pattern of the singular can be said to be nested within the larger syncretic pattern of the dual. (2)

NOM ACC GEN DAT LOC INS

Nested syncretism in Upper Sorbian (Sˇołcína and Wornar 2002: xx 4, 10, 12) plural

singular

dual

zˇony zˇony zˇonow zˇonam zˇonach zˇonami

zˇona zˇonu zˇony zˇonje zˇonje zˇonu

zˇonje zˇonje zˇonow zˇonomaj zˇonomaj zˇonomaj

‘wife’

Multiple patterns of syncretism are also found in the paradigms of Nuer, a Nilo-Saharan language (3). However, they are not nested within each other as in Upper Sorbian. Each case may be syncretic with each other case, without there being an implicational hierarchy. Thus the genitive is syncretic with the locative in ‘egret’, but with the nominative in ‘girl’, while the locative is syncretic with the nominative in ‘bug’.2 Following Williams 2

The default singular pattern is represented by ‘egret’, while the default plural pattern involves no case distinction, e.g. bo— o— Nn— i ‘egrets.NOM/GEN/LOC.PL’. The other patterns are restricted to small sets of nouns. It should be noted, though, that the default pattern accounts for a relatively small portion of the Nuer noun lexicon. Frank (1999) notes that


15

(1981), we term this contrary syncretism, because the pairings in each paradigm are mutually exclusive. (3) Nuer (Frank 1999: 84–6)

NOM SG GEN SG LOC SG

‘dog’

‘egret’

‘girl’

‘bug’

jio¨k jio— k jio— o— k

boö¨N boö¨Nka— boö¨Nka—

nyal nyal nyaal

baan baanka¨ baan

The juxtaposition of contrary syncretic patterns may imply the existence of a morphosyntactic distinction which itself has no unique expression. For example, in the Dardic language Phalura (4), the distribution of inflected forms warrants the assumption of five cases: absolutive (or nominative), accusative, dative-locative, ergative and genitive-ablative. However, in the material described by Morgenstierne (1941), no paradigm displays more than four distinct case forms. Neither the accusative nor the dative-locative has a distinct form in any paradigm. The accusative is syncretic with the absolutive in nouns, with the dative locative in plural pronouns. The dativelocative is syncretic with the ergative in nouns and with the accusative in pronouns. (Singular pronouns display only two forms, e.g. first person absolutive/accusative/dative-locative ma, ergative/genitive-ablative mi .) Following Zaliznjak (1973 [2002]: 629) we can speak of the accusative and dative-locative as non-autonomous values. (Similarly, as pointed out by Jim Blevins (personal communication), we might speak of non-autonomous inflection classes, defined by their patterns of syncretism – compare ‘egret’ and ‘bug’ in the Nuer example in (3) above.) The identification of non-autonomous values as syncretism is not always certain, since in some instances this surface effect may be attributed to a syntactic asymmetry. For example, most modern descriptions of Latvian assume seven cases (nominative, accusative, genitive, dative, locative, vocative). However, there are a few prepositions which govern the accusative or genitive case in the singular, but the dative in the plural (Mathiassen 1997: 184–5). By distributional criteria one might wish to postulate two additional cases to account for this. However, the fact is only 22 out of his corpus of 264 nouns are fully regular; in order to fully account for the nouns in the corpus, one would need 207 distinct inflectional classes.

16

The Syntax–Morphology Interface (4) Non-autonomous cases in Phalura (Morgenstierne 1941: 16–18)

ABS ACC DAT-LOC ERG GEN-ABL

noun ‘man’

pronoun ‘we’

mi:sˇ mi:sˇ mi:sˇ a mi:sˇ a mi:sˇ i

be asa:m asa:m asim asi:

that all prepositions govern the dative case in the plural, so it may be just as well to define the case government of individual prepositions only in the singular, alongside a global stipulation about their case government in the plural (see Fennell 1975 for a discussion).3 Finally, it is also possible to identify syncretism on the basis of an unusual distribution of forms within the paradigm of an individual lexeme. Consider the Old Irish paradigm in (5), representative of masculine o-stem nouns. What concerns us here is the behaviour of the form fir, which appears in both the singular and plural. Within the case paradigm for each of these numbers, the form is not syncretic: in the singular it serves as the genitive and in the plural it serves as the nominative. But all the same, by our criteria fir is syncretic, since it fills two cells in the morphosyntactic paradigm of the lexeme. Nor can these two cells easily be collapsed, representing as they do distinct values for both case and number. Adapting a term from Cushitic studies, we call such patterns polarity effects (see Chapter 3, x3.7). (5)

NOM ACC GEN DAT

3

Polarity effect in Old Irish (Pokorny 1923: 77) singular

plural

dual

fer fer fir fiur

fir firu fer feraib

fer fer fer feraib

‘man’

Traditionally, grammatical descriptions of Latvian have concerned themselves only with the behaviour of the preposition ar ‘with’. Historically, this governed the instrumental case, since fallen together with the accusative in the singular. Its persistence as a separate case in the Latvian grammatical tradition is thus due to a mixture of diachronic and distributional criteria.


17

2.1.2 Implications Each of these patterns has different implications for the possible relationship between the feature values involved. Simple syncretism suggests a link between two values, as in the Central Alaskan Yup’ik example, represented in (6a). Nested syncretism suggests a branching hierarchy of links, as in the Upper Sorbian example, represented in (6b). Contrary syncretism suggests multiple independent links, as in Nuer, represented in (6c).

(6)

a.

b.

ABS

REL

DAT

c.

LOC

INS

NOM

GEN

LOC

Polarity effects cannot readily be portrayed as a relationship between values of a feature. The ramifications that these patterns have for morphological models are discussed in depth in Chapter 4.

2.2

Domains of comparison

In order to identify syncretism, we typically compare different paradigms involving the same feature values and note a discrepancy between them in the distribution of forms. In the previous section we looked at the different sorts of discrepancy that are possible. In this section we look at the domains of comparison, that is what kinds of paradigms can be compared with each other? We illustrate the possibilities with syncretism of the feature person. 2.2.1 Morphological classes Syncretism may be concomitant with variation in the inflectional morphology. That is, some affixes may be syncretic, and others not, as in Abipon (7), an extinct Guaicuruan language formerly spoken in Argentina. There are seven conjugation classes, each associated with different person-marking prefixes; the second person suffix -i remains constant throughout. In class IV, the first person and third person prefixes are syncretic. (In all conjugation classes the form of the stem remains constant.)

18


(7) Verbal affixes in Abipon (Najlis 1966: 32–3)4

1 2 3

I

II

III

IV

V

VI

VII

ei n n- -i n-

haØ- -i r-

rigr- -i n-

mm- -i m-

ehe- -i ei-

ahØ- -i ai-

lil- -i l-

Alternatively, the inflectional affixes may be the same for all verbs but interact differently with different stem classes. For example, in the Carib language Macushi (8), the distinction between 2SG a- and 3SG i- is lost when these are prefixed to vowel-initial stems.5

(8)

Verbal prefixes in Macushi (Abbot 1991: 101) C initial stem

1SG 2SG 3SG

V initial stem

initial heavy syllable

(VC, V: or VV)

initial light syllable (V)

uai-

Øawaw-

Ø(stem-initial V:) (stem-initial V:)

A related possibility is where a single set of affixes is used, but occupy different positions in different inflection classes, and so interact differently with the stem. For example, in Somali (9), person-number markers are prefixed to a few verb stems, and otherwise are suffixed. When prefixed, 1SG Ø- is distinct from 3SG M y-, but when suffixed, this distinction is lost. (2SG and 3SG F are syncretic in either case.)

4

5

Some morphological variation is not shown in the table: (i) before vowel-initial stems, the prefix-final vowels of I and II are deleted, and second person prefix of II is h-, (ii) the first person prefix of III is gr- when the stem takes a plural suffix, and (iii) the third person prefix in II and III may be i-, under conditions which are not explained by Najlis (1966). Although the stem classes can be characterized in phonological terms, and the syncretic patterns appear to originate in the phonological interaction of prefix and stem, the syncretism itself cannot entirely be reduced to a synchronic phonological rule. Thus, while the structure of the prefix aw- seems to be due to constraints on syllable structure (V + heavy V yields two syllables, and non-initial syllables must be C-initial), there is no automatic rule whereby i fi a before w (e.g. i-wa0 ka-ri 3-axe-POSSESSION ‘his axe’ (Abbot 1991: 85)).

Characteristics of syncretism (9)

19

Simple past in Somali (Saeed 1999: 86, 98)

1SG 2SG 3SG F 3SG M

prefix conjugation ‘say’

suffix conjugation ‘wait’

Ø-idhi t-idhi t-idhi y-idhi

sug-Ø-ay sug-t-ay sug-t-ay sug-Ø-ay

Finally, the full range of inflectional distinctions may depend on the interaction of a stem alternation with affixation, where both exhibit block syncretism (see above, x1.4.2). If there is a class of stems lacking the stem alternation, this leads to syncretism. For example, in German (10), the ending -t marks both 3SG and 2PL. Some stems have a vowel alternation in the 2SG and 3SG, which disambiguates the forms in -t (10a). Other stems lack this alternation, yielding 3SG/2PL syncretism (10b).

(10)

1SG 2SG 3SG 1PL 2PL PL

Present tense in German a. ‘give’

b. ‘live’

geb-en gib-st gib-t geb-en geb-t geb-en

leb-en leb-st leb-t leb-en leb-t leb-en

2.2.2 Feature values Person features occur on nominals, typically to mark possession, and on verbs to refer to arguments. Given its distribution across different word classes, person occurs in the presence of a wide variety of other morphosyntactic features (including case, number, gender, tense, voice, negation), thereby allowing us to see how the paradigm of one feature may be influenced in a variety of contexts. The presence of particular values of these accompanying features can affect the differentiation made in the person paradigm.

20


Case In languages which mark case and possession on nouns, where possession marking realizes the person (and number) of the possessor, we can observe the interaction of case and person. In the Samoyedic language Selkup (11) a variety of cases use the same form for the second person singular possessor and third person singular possessor. (11) Syncretism in the context of case in Selkup (Helimski 1998: 558–89)6 possessor

NOM GEN ACC INS CAR TRANS COORD DAT-ALL ILL, LOC, EL PROL

1SG

2SG

3SG

nommı¨ nomnı¨ nommı¨ nomnı¨ sa¨ nomnı¨ ka˚a˚lı¨ k nomno(qo) nomnı¨ sˇ j ak nomnı¨ nı¨ k nopqa¨k nomma¨k

nomlı¨ nomtı¨ nomtı¨ nomtı¨ sa¨ nomtı¨ ka˚a˚lı¨ k nomtoo(qo) nomtisˇ j ak nomtı¨ nı¨ k nopqa¨ntı¨ nomma¨ntı¨

nomtı¨ nomtı¨ nomtı¨ nomtı¨ sa¨ nomtı¨ ka˚a˚lı¨ k nomtoo(qo) nomtisˇ j ak nomtı¨ nı¨ k nopqı¨ ntı¨ nommı¨ ntı¨

‘God, heaven’

Person Syncretism of person in the context of person may occur in verbs that mark more than one argument. In the Papuan language Yimas (12), of the SepikRamu family, second and third person singular subject values in transitive verbs are syncretic when the object person is third singular. (12) Syncretism in the context of person in Yimas (Foley 1991: 217)

subject

object

6

1SG 2SG 3SG

1SG

2SG

3SG

— maNatayłcut naNatayłcut

kampantayłcut — nanantayłcut

nakatayłcut nantayłcut nantayłcut

Some of the forms with 1SG possessor are optionally based on the stem nuu-; these have been omitted.


21

Gender Similar to case, gender is a feature which is prone to syncretism (see Chapter 3, x3.1, x3.4 and x3.8.3), but gender may still provide feature values which determine the form of other features, such as person. In the Indo-Aryan language Sindhi (13) the gender value ‘feminine’ in the plural determines syncretism of second and third person. (13) Syncretism in the context of gender in Sindhi (Khubchandani 2003: 647) masculine singular plural 1 lIkh-«nd-UsI 2 lIkh-«nd-e˜ 3 lIkh-«nd-o

lIkh-«nd-ası˜ lIkh-«nd- lIkh-«nd-a

singular

feminine plural

lIkh-«nd-«sı˜ lIkh-«nd-ı˜ «˜ lIkh-«nd-i

1Ikh-«nd-yu˜sı˜ lIkh-«nd-yu˜ lIkh-«nd-yu˜

‘write’

Number In the Papuan language Kobon (of the Trans-New Guinea phylum), the value ‘dual’ of the number paradigm accompanies a loss of distinction between the second and third person for all verbs (here illustrated with a past tense form).

(14)

1 2 3

Syncretism in the context of number in Kobon (Davies 1981: 166) singular

plural

dual

arno¨ arna ara

arno arbe arla

arlo arlo¨ arlo¨

‘went’

Tense-aspect-mood Person syncretism can also occur in the presence of tense-aspect-mood. In Aymara (spoken primarily in Bolivia), verbs may have syncretism between first and second person in the future perfect but have separate forms for this in the present tense (15). Voice The verbal feature of voice can also be a context for person syncretism. In Gothic (16), the distinction between first and third person singular is lost in

22

The Syntax–Morphology Interface (15) Syncretism in the context of tense-aspect in Aymara (Deza Galindo 1992: 103–5)

1SG 2SG 3SG

(16)

present

future perfect

muntua muntahua munihua

munchiya¨ta munchiya¨ta munchı¨ na

Syncretism in the context of voice in Gothic (Wright 1930: 135–6) indicative


subjunctive

active

passive

active

passive

nima nimis nimiþ nimam nimiþ nimand

nimada nimaza nimada nimanda nimanda nimanda

nimau nimais nimai nimaima nimaiþ nimaina

nimaidau nimaizau nimaidau nimaindau nimaindau nimaindau

‘take’

the passive voice. That this is systematic is suggested by the fact that this syncretism also occurs in the subjunctive mood, but with different forms. The passive combined with the plural provides a context for complete loss of person distinctions in both the indicative and the subjunctive. Negation Where negation is realized in the morphology it is possible to find person interacting with negation. In the positive paradigms of the Nilo-Saharan language Dongola Nubian (17), here represented by the present indicative, four person-number forms are distinguished, while in the negative, only two forms are distinguished, one for 3PL and one for all other values. Some observers have seen the relationship between feature values and syncretism in terms of markedness. That is, the more complex ‘marked’ feature value is associated with diminished variability (Moravcsik and Wirth 1986: 1–3). One explanation put forward for loss of distinctions in the presence of a ‘marked’ value is that the system avoids an excessive number of ‘marked’ values (for instance, Cairns 1986: 18, McCarthy 2002: 81).


23

(17) Syncretism in the context of negation in Dongola Nubian (Armbruster 1960: 195, 205) present indicative ‘drink’


positive

negative

niri nin nin niru niru niran

nimunun nimunun nimunun nimunun nimunun nimunan

As Cairns (1986: 18) puts it, ‘in a marked member of any given opposition, fewer other oppositions are likely to occur, because, for each further opposition, there would have to be a marked member.’ Some of the examples do seem to be compatible with this approach, e.g. syncretism under dual number in Kobon, the passive in Gothic, or negation in Dongola Nubian. On the other hand, the predictive value of markedness should not be exaggerated, as the quirky distribution of forms in Selkup shows. In addition, one should note that in all the examples adduced above – indeed, in most examples – differences in feature values are concomitant with differences in the inflectional morphology. The interpretation of such examples thus confronts an inherent ambiguity: is the syncretism a property of the conditioning feature value, or of the morphology that realizes that paradigm? 2.3

Morphological characteristics

The morphological realization of a syncretic pattern may in itself influence its interpretation. Below we consider three characteristics, regularity, directionality and morphological unmarkedness. By regularity we mean the repretition of a syncretic pattern across multiple exponents. Consider the syncretism of dative and ablative in Latin (18). In the plural, these cases are syncretic for all declension classes, realized in some by the ending -is and in others by the ending -ibus. Not only do these endings look different, they have distinct diachronic origins (Buck 1933: 182, 186). In Kashmiri (19), ergative and ablative plural are also syncretic for all declension classes. However, this ending, -av, is the same in all declension classes, even though other endings might differ.

24 (18)

The Syntax–Morphology Interface Latin second delension ‘star’

NOM ACC GEN DAT ABL

third declension ‘chief’

singular

plural

singular

plural

stella stellam stellae stellae stella

stellae stellas stellarum stellis stellis

princeps principem principis principi principe

principes principes principium principibus principibus

(19) Kashmiri (Kachru 1969: 112–16) first declension: ‘child’

ABS ERG ABL DAT

second declension ‘tree’

singular

plural

singular

plural

gobur gobran gobri gobur

gobar gobrav gobrav gobran

kul kul’ kuli kulis

kul’ kul’av kul’av kul’an

The regularity of the syncretic pattern in Latin over two distinct endings suggests that this pattern is systematic. Otherwise, we would have to assume two instances of accidental homophony which happen to coincide. The Kashmiri example is ambiguous; since only one ending is involved, the notion that this is accidental is more plausibly entertained than in the case of Latin. The second characteristic, directionality, concerns the possible morphological affiliation of the syncretic form to one of its component values. Consider the imperfective paradigms from the Dagestanian language Lak, shown in (20). The syncretic 1SG/2SG past tense ending is not morphologically related to any of the endings in the non-syncretic present tense paradigm. Now consider the paradigms from the Tungusic language Udihe, likewise illustrated in (20). The syncretic 1SG/2SG ending of the future tense is identical to the 2SG ending as found in the past tense. The situations in Lak and Udihe present different descriptive tasks. In the case of Lak, there is a static relationship between the endings and the morphosyntactic values they express; we need only state that the


25

(20) Lak (Xajdakov 2001: 354) and Udihe (Nikolaeva and Tolskaya 2001: 212–13) Lak imperfective ‘stand up’

1SG 2SG 3SG

Udihe ‘sing’

present

past

past

future

izan-na izan-ssara izan-ssar

izajssij-av izajssij-av izajssij-a

jexe:-mi jexe:-i jexe:-ni

jexezeNe-i jexezeNe-i jexezeNe-ni

ending -av expresses both 1SG and 2SG. In the case of Udihe, the relationship between the ending and the morphosyntactic values varies. We must somehow account for the fact that the ending -i sometimes functions solely as 2SG, and sometimes as both 1SG and 2SG. We term this a ‘directional effect’, because it looks as if the form for one value is the source of the form for the other value; e.g. in Udihe, it looks as if the 1SG future form is based on the 2SG. The third characteristic, unmarkedness, concerns the possible relationship within a paradigm between a morphologically unmarked form (the bare stem, lacking affixes or evidence of other morphological operations) and syncretism. For example, in the Francisco Leon dialect of Zoque (a Mixe-Zoquean language), the bare stem is used for 1SG and 3SG. This is illustrated in (21) using the preterite form; the person-number markers do not vary with tense. (Nyindicates nasalization of the initial consonant of the stem, followed by -y-.)

(21) Francisco Leon Zoque (Engel, Allhiser de Engel and Mateo Alvarez 1987: 378) poyu ‘ran’

1 2 3

underlying forms of affixes

singular

plural

singular

poyu mbyoyu poyu

potyamu mbyotyamu poyaju

Ø NyØ

plural Ø NyØ

-tam -tam -yaj

Here, syncretism coincides with the morphologically unmarked form, which may suggest a causal connection between the lack of inflectional marking and syncretism: Francisco Leon Zoque lacks 1SG and 3SG

26


morphology and uses the bare stem to fill in this gap.7 In this case the use of the bare stem form also coincides with what are commonly seen as the default morphosyntactic values for person and number, namely third singular. But one also finds instances where morphological and morphosyntactic unmarkedness do not line up. The present tense forms of English are a familiar example (walks versus walk). A similar example comes from the far past conjugation of the Papuan language Orokaiva (Trans-New Guinea phylum) (22), where only the 3SG and 2PL have overt endings, while the bare stem is used elsewhere (the final -a is the indicative mood marker). On the other hand, syncretism may fail to coincide with zero morphology – in the other tense-aspect-mood paradigms (roughly two dozen), Orokaiva uses a different set of person-number endings, where 1PL and 3PL are syncretic, but it is the 2SG form which lacks an overt person marker. This is shown by the mid-past B form in (22).

(22)

Orokaiva ‘walk’ (Healey, Isoroembo and Chittleborough 1969: 40, 59, 62) far past

1 2 3

mid-past B

singular

plural

singular

plural

hembu -a hembu -a hembu -n -a

hembu -a hembu -w -a hembu -a

hembu -ha -n -a hembu -ha -a hembu -ha -j -a

hembu -ha -r -a hembu -ha -w -a hembu -ha -r -a

In this case of the mid-past B forms in (22), the syncretic pattern cannot so readily be attributed to the extension of a morphological default. Such patterns may also be found where the morphological default seems to coincide with the morphosyntactic default, as in the Niger-Congo language Dogon, where the bare stem is used for the 3SG, but syncretism affects the 1PL/2PL. (The 3PL ending varies in form with aspect-mood, but otherwise there is a single set of person-number endings for all aspectmood paradigms.)

7

Engel, de Engel and Alvarez (1987: 379) report that the enclitic first person pronoun -0 ijtzi is used to distinguish between 1SG and 3SG; judging by the examples in the text, its use is not obligatory

Characteristics of syncretism (23)

1 2 3

2.4

27

Dogon, habitual form of ‘stay’ (Plungian 1995: 30–1) singular

plural

wada-de-m wada-de-w wada-de

wada-de-y wada-de-y wada-d-iN

A typology of interpretations

It has been shown in Chapter 2: x2.1.1 that the comparison of multiple paradigms allows us to determine the presence in a language of particular feature values and determine where the distinctions represented by those values may be collapsed in the morphology. In other words, it can be determined that particular morphosyntactic distinctions are present underlyingly, even if absent morphologically. However, if we mechanically combine all the logically possible morphosyntactic distinctions which have an expression in form, without considering the relationship between syntax and morphology, this will produce an abundance of information, not all of which is of equal interest. Some of the maximal paradigms predicted by comparing multiple paradigms will turn out not to be relevant for syntax. This is neutralization, which is discussed in Chapter 2: x2.4.1. At the other end of the spectrum, morphology may fail to distinguish any values of a syntactically relevant feature. This is uninflectedness, which is discussed in Chapter 2: x2.4.2. Neutralization and uninflectedness are violations of two properties of what we may call ‘canonical inflection’: (a) that feature values found in a language combine mechanically into fully distinguished combinations, and (b) that morphology always distinguishes syntactically relevant feature values. Neutralization violates property (a), and uninflectedness is a straightforward violation of property (b). There is a more challenging violation of property (b), which is canonical syncretism. This is discussed in Chapter 2: x2.4.3. While uninflectedness involves a total failure to distinguish syntactically relevant values, canonical syncretism involves partial failure to do so. We therefore need to distinguish between these three related ways of understanding syncretism. In order to do this we will, in examples (24)–(31), have two separate glosses for the feature values differentiated by morphology and by syntax.

28


2.4.1 Syncretism as neutralization Across the spectrum of possibilities for syncretism, certain instances can be accounted for easily. For example, Russian pronouns, adjectives and verbs all distinguish three genders (masculine, feminine and neuter) and two numbers (singular and plural). As gender is an agreement category for these three word classes, it is therefore syntactically relevant. However, in Russian, gender is only distinguished in the singular; there is no context in which gender is distinguished in the plural.8 (24) morphology: syntax:

knig-a book(F)-SG.NOM book(F)-SG.NOM ‘the book is interesting’

interesn-a interesting-F.SG interesting-F.SG

(25) morphology: syntax:

rasskaz tale(M)[SG.NOM/ACC] tale(M)[SG.NOM] ‘the tale is interesting’

interesen interesting.M.SG interesting.M.SG


predlozˇeni-e proposal(N)-SG.NOM/ACC proposal(N)-SG.NOM ‘the proposal is interesting’

interesn-o interesting-N.SG interesting-N.SG

In examples (24)–(26) gender is syntactically relevant in that this information must be available to the predicate adjective in order for it to agree. Seen from another perspective, the presence of a particular agreement marker on the adjective constrains the choice of noun which the adjective may occur with, and this also disambiguates the case syncretism represented by the morphological glossing for the nouns in (25) and (26), because the nouns in (25) and (26) are required by the syntax to be in the nominative case so as to control predicate agreement. We can contrast the role of gender in the singular with the situation in the plural. (27) morphology: syntax:

8

knig-i book(F)-PL.NOM/ACC book(F)-PL.NOM ‘the books are interesting’

interesn-y interesting-PL interesting-PL

The one possible exception to this is the quantifier ‘both’ oba obe which, in the prescribed written standard at least, has oblique forms which distinguish gender.

Characteristics of syncretism (28) morphology: syntax:

rasskaz-y tale(M)-PL.NOM/ACC tale(M)-PL.NOM ‘the tales are interesting’



predlozˇeni-ja9 proposal(N)-PL.NOM/ACC proposal(N)-PL.NOM ‘the proposals are interesting’


29

In (27)–(29) the morphological gloss of the form interesny does not contain any gender values. As an alternative, we could have added the choice of values F/M/N to the morphological gloss for interesny. Because this choice would involve all possible values for gender, its presence is syntactically no more constraining than its absence. This demonstrates that, although the noun lexemes have inherent lexical gender, syntax does not need to make reference to gender in plural contexts.10 It can be concluded that, for Russian, gender in plural contexts is syntactically irrelevant. This can therefore be easily accounted for as an instance of syntactic neutralization, which, because the form reflects the irrelevance of the feature for syntax, it is not the role of inflectional morphology to account for.11

9

10

11

We have transliterated, rather than transcribed, the Russian examples. This affects the segmentation of morphological glossing, because certain of the so-called ‘soft’ vowel letters may actually represent the last element of a root or stem as part of the vowel of the ending. This is the case with the form predlozˇenija, for example, where /j/ is actually the last phoneme of the stem but is represented as combined with the final vowel /a/ by the letter ~, which represents combinations of /j/ plus /a/ or indicates that the consonant before /a/ is palatal. As we have transliterated the examples, we have accordingly made the morpheme divisions, as if they corresponded to the written words. Similar comments can be made about other Russian examples in this chapter, such as (33), but this issue is of no material importance for the points being made here. This generalization includes the agreement of the morphologically plural, but semantically singular, honorific vy ‘you’ with long-form predicate adjective in the singular. Vy molcˇaliv-aja You silent-F.SG.NOM ‘You are silent.’ (addressed to a woman) (Corbett 1983: 53) The challenge for syntactic theory here is to account for the singular agreement. Given the singular form of the adjective, we expect gender agreement. It is of interest for typology, as shown by Greenberg and many authors following him. Absolute neutralization is covered by typological claims of the type ‘A language never has more gender categories in nonsingular numbers than in the singular’ (Greenberg 1963: 95).

30


The total absence of distinct values of a particular feature, in a given context, is necessary, but not sufficient to define neutralization. Because neutralization involves lack of syntactic relevance, total absence of distinctions for one set of syntactic objects in a given syntactic context is not sufficient, as other syntactic objects may still realize that set of distinctions. Hence, syntactic relevance means absence of distinctions within the morphology of all word classes. Neutralization Neutralization is defined as follows: i. In the presence of a particular combination of values of one or more other features (the context), there is a general loss of all values of a particular feature F found elsewhere in the language. ii. No syntactic objects distinguish any values of feature F in the given context, and feature F is therefore syntactically irrelevant in that context. Given the examples in (27)–(29), neutralization represents the most straightforward explanation: the lack of formal distinction merely reflects the irrelevance of the feature in question for syntax. 2.4.2 Syncretism as uninflectedness Whereas it is possible on the one hand to account for lack of formal differentiation as reflecting syntactic irrelevance, it is also possible to interpret it as reflecting a lack of response by morphology to syntactically relevant distinctions. This is uninflectedness. For example, the Russian noun pal 0 to ‘coat’ is indeclinable and so has no distinct number or case forms. However, a typical Russian noun does inflect for case and number, as does any adjective which agrees with pal 0 to. Nouns like pal 0 to therefore constitute a particular morphological class which is insensitive to distinctions which are still syntactically relevant. With the Russian noun pal 0 to, there is no syntactic feature value which is the context for its uninflectedness. This contrasts with neutralization, where at least one feature value is present as a context, because otherwise, if all syntactic objects failed to make distinctions for a given feature when there is no context, there would be no evidence for the existence of that feature. Uninflectedness may still occur in a particular morphosyntactic context, however. In Polish, nouns such as muzeum ‘museum’ do not inflect in the singular, but do in the plural (Kotyczka 1980: 95, 105–6;


31

Tokarski 1993: 257). Importantly, the uninflected singular forms can still be used in different syntactic contexts. (30) morphology: syntax:

now-ego new-M.SG.GEN new-M.SG.GEN ‘of (the/a) new museum’

muzeum museum(M)[SG] museum(M)[SG.GEN]


now-ych new-PL.GEN new-PL.GEN ‘of (the) new museums’

muze-o´w museum(M)-PL.GEN museum(M)-PL.GEN

In (30) the same form can be used in the genitive singular context as the nominative singular, even though there is no genitive singular inflection as such, whereas there is a specific genitive plural inflection in (31).12 This cannot be treated as neutralization, because a typical Polish noun will inflect for case in the singular, and hence this feature is syntactically relevant. Therefore while the value ‘genitive’ is absent in the morphological gloss of muzeum in (30), the syntax treats the form as though it were genitive (as with similar forms of the Russian pal 0 to ‘coat’). Occasionally it may be more difficult to distinguish between instances of uninflectedness and neutralization. The verbal system of Russian is an example of this. Russian verbs have two synthetic paradigms, past and non-past.13 Past tense forms mark gender and number (32) but do not mark person. (32)

M F N

Russian verb ‘played’ singular

plural

igral igrala igralo

igrali igrali igrali

In contrast, non-past forms mark person and number (33).

12

13

That this contrast between uninflectedness in the singular and inflectedness in the plural is reflected in use is easily checked by searching Polish language websites for the different number and case combinations. For instance, we can find example sentences containing the noun phrases nowego muzeum and nowych muzeo´w. Simple non-past forms have either a future interpretation, if a verb is perfective, or a present tense interpretation, if a verb is imperfective.

32

The Syntax–Morphology Interface (33)

1 2 3

Russian verb ‘play’ singular

plural

igraju igraesˇ 0 igraet

igraem igraete igrajut

However, in contrast with the examples (27)–(29), where we saw that no agreement targets in Russian mark gender in the plural, we cannot say that gender is neutralized in the non-past, because it may be relevant for other agreement targets. (34)

Esli vopros okazyva-et-sja if question(M) turns.out-3.SG-REFL ‘If the question turns out to be stupid . . . ’

glup-ym . . . stupid-M.SG.INS

In the past tense equivalent of (34) the verb would also require the correct gender form to agree with the subject, but the non-past form in (34) never marks gender. In (34) the non-past form of the verb okazyvat 0 sja ‘to turn out’, which has a reflexive marker added to the non-past person and number marking, occurs with a predicate adjective which marks gender. Hence, in this example, although the verb does not distinguish any gender features in the non-past, this differs from neutralization, as gender features may still be syntactically relevant. Therefore, we say that gender is present syntactically, but that the non-past verb forms are uninflected for gender. Of course, syntactic relevance or presence will depend on the view one has of syntactic structure, and so there are examples of uninflectedness which are harder to distinguish from neutralization, as is the case for the Russian verb. Uninflectedness, in common with neutralization, involves total absence of distinctions for a given feature. In contrast with neutralization, this total absence of distinctions may be limited to a particular morphological or syntactic class. As there are other classes which maintain the appropriate distinctions in the same context, this feature is still relevant for syntax. Hence, whereas neutralization is about syntactic irrelevance as reflected in morphology, uninflectedness is about morphology being unresponsive to a feature that is syntactically relevant.


33

Uninflectedness is defined as follows: i. There is, in certain lexemes only, a loss of all values of a particular feature F found elsewhere in the language. This loss may depend on the presence of a particular combination of values of one or more other features (the context). ii. Other syntactic objects distinguish values of feature F, either generally or in the given context, and feature F is therefore syntactically relevant. Uninflectedness and neutralization deal with the interaction of morphosyntactic features. They are both important theoretically, because they suggest possible ways of treating syncretism. They are also the most straightforward parts of the spectrum of morphology-syntax interaction which involve loss of feature distinctions, as they can be accounted for by constraints on feature combination. The more problematic part of the spectrum, canonical syncretism, cannot be treated in terms of uninflectedness or neutralization, unless the feature set is modified in some way. 2.4.3 Canonical syncretism Uninflectedness and neutralization cannot readily account for the challenge of syncretism involving the partial collapse of syntactically relevant feature distinctions. For uninflectedness, where the feature involved is syntactically relevant, the problem is accounting for the fact that the morphology is only partially inert, as opposed to totally inert. The problem for neutralization, in contrast, is that the feature is still syntactically relevant. The partial loss of distinctions can be illustrated by Slovene, where the l-participle, as other agreement targets, marks masculine, feminine and neuter genders in the plural and singular (35).

(35)

M N F

Slovene l-participle of the verb ‘write’ singular

dual

plural

pisal pisalo pisala

pisala pisali pisali

pisali pisale pisala

34


In the dual, the distinction between neuter and feminine is collapsed, while there is still a separate masculine form. This is therefore an instance of gender syncretism, which cannot be treated in terms of gender neutralization or uninflectedness, as we have seen for Russian in the plural and past tense respectively. Canonical syncretism is defined as follows: i. There is, in certain contexts, a loss of distinctions between some but not all values of a particular feature F. This loss may depend on the presence of a particular combination of values of one or more other features (the context). ii. Other syntactic objects distinguish those values of feature F, and they are therefore syntactically relevant. The Slovene example (35) involves syncretism of gender in the dual, where the gender distinctions are only partly lost, which naturally entails that gender is still syntactically relevant in the dual. The fact that the loss of feature distinctions is not total, and therefore still involves values of the feature in question, means that this part of the spectrum of morphology– syntax interaction is the most challenging. Unless the inventory of feature values is further augmented, it cannot be claimed that the loss of distinction reflects lack of relevance for syntax (as would be the case with neutralization), nor is it just a fact about morphology that it is featurally inert in dual contexts (as would be the case with uninflectedness). It may be tempting to extend the notions of uninflectedness and neutralization to instances of canonical syncretism. For instance, the Russian long-form adjective collapses distinctions between masculine and neuter genders in the oblique cases (36), but retains a partial contrast between masculine and neuter on the one hand and feminine on the other. (36) Russian ‘new’ in the singular


M SG

N SG

F SG

novyj novyj novogo novogo novomu novom novym

novoe novoe novogo novomu novom novym

novaja novuju novoj novoj novoj novoj(u)


35

Gender is neutralized in the plural in Russian, which suggests that there is a constraint on feature interaction whereby number is a context for loss of gender. If case is added as a potential context, then the masculine/neuter syncretism can also be treated as an example of feature interaction (underspecification of gender in the presence of particular case and number values). The problem that arises here is that case itself is not fully distinguished in the presence of the feminine gender. So the genitive, dative, locative and, if the archaic form novoju is not used, the instrumental, all share the syncretic form novoj. This is syncretism, because these cases have distinct forms for the masculine and neuter genders. If novoj is treated as underspecified for case (F.SG), and novomu as underspecified for gender (SG.DAT), there is a problem determining which form to use for the feminine dative singular, as both specifications are compatible with this. This indicates that, on its own, the most natural formal construct associated with uninflectedness and neutralization, namely underspecification, may well be insufficient to deal with the more challenging examples of canonical syncretism.

2.5

Conclusion

Syncretism exhibits different patterns: the simple and nested syncretism patterns might suggest a hierarchical structure for the features involved, contrary syncretism and polarity effects demonstrate that such structures are insufficient for describing all patterns of syncretism. We have also seen that these patterns may depend on the domains of comparison. A syncretic pattern may be associated with a particular morphological class; it may arise because of the effects of different stems on the same affix (as in the Macushi example in (8)), it could be the result of different orderings of the same affix, or it could arise in some classes and not others, because of the presence or absence of alternation to the stem. Finally, we have identified a spectrum of morphology–syntax interaction involving loss of feature distinctions, which could broadly be described using the term ‘syncretism’. We will naturally be concentrating on that part of the spectrum which is left over when we have removed the obvious explanations for the absence of a morphological distinction. These obvious explanations are irrelevance for syntax (neutralization) and morphological inertness (uninflectedness). They can both be explained by the constraints on the ordering of features, and both neutralization and uninflectedness, because of what they tell us about the

36


ordering of features, give us a hint about how to deal with some examples of syncretism. At other times it is not self-evident how to analyse examples of syncretism. Under particular models of grammar it is possible to take some or all instances of syncretism and analyse them in terms of neutralization or uninflectedness, by creating further structure within a feature. But if one chooses to analyse what we have termed canonical syncretism in this way, additional assumptions are required which pose problems and need justification.

3 Cross-linguistic typology of features

Although the theoretical interest implicit in inflectional syncretism has been recognized for some time, the range of languages that have contributed to its study has remained limited. The notion was first applied to Indo-European languages, and it is from these that the bulk of examples has been drawn. Languages from other families have been brought to bear in theoretically or typologically oriented works, but to a lesser extent, e.g. Afro-Asiatic (Carstairs 1987, Fradkin 1991, Johnston 1997, Noyer 1997), Uralic (Ba´tori 1990, McCreight and Chvany 1991, Kiparsky 2001, Blevins 2003), Tibeto-Burman (Gvozdanovic´ 1991), Altaic (Carstairs 1987), and more broadly defined, languages of the Caucasus (Hjelmslev 1935–7, Boeder 1976, Carstairs 1987, Carmack 1997, Kibrik 1997, Helmbrecht 1999), of Australia (Zaliznjak 1973, Goddard 1982, Bavin and Shopen 1987, Heath 1991, Noyer 1997, Evans, Brown and Corbett 2001), of the Americas (Boeder 1976, Hewson 1989, Heath 1998, Laka¨mper and Wunderlich 1998, Harbour 2003), of Siberia (Kibrik 1997, Spencer 2000) and New Guinea (Kibrik 1997, Noyer 1998, Wunderlich 2001b). However, these works address at most several languages. Cross-linguistic studies relevant to syncretism are rare: Cysouw (2003) is a comprehensive treatment of the morphological and lexical expression of pronominal features (person and number), while Arkadiev (2003) discusses case syncretism. Thus, there remains a real gap in our empirical understanding of syncretism, which the following chapter should begin to redress. We have gathered examples from a wide range of languages. The sample is based in part on that used in the World Atlas of Language Structures (see Baerman and Brown, 2005 a and b) and partly on our own databases (Baerman, Brown and Corbett 2002a, Baerman 2002b), and has been supplemented by additional languages chosen to illustrate special points. In xx3.1–3.7 we look at syncretism with respect to major morphosyntactic features (case, person, gender, number, tense-aspect-mood). The selection of features was determined partly by their cross-linguistic frequency in inflectional systems – thus, one 37

38


can imagine syncretism with respect to degrees of honorificity, but it is neither sufficiently frequent as an inflectional feature, nor are the values sufficiently comparable across languages, to make any confident crosslinguistic generalizations. Also relevant was the availability of more than two distinct values for the feature, which follows from the first part of our definition of ‘canonical’ syncretism (Chapter 2: x2.4.3.); this eliminates from consideration features with inherently binary distinctions such as definiteness. In x3.7 we consider syncretism which cannot be construed as involving the simple collapse of feature values. In x3.8 we look at the role of feature values as conditioning factors for the syncretism of other features. 3.1

Case

3.1.1 Introduction Inflectional syncretism is often implicitly understood as case syncretism; e.g. the entry under ‘Syncretism’ in the recent Morphology: an international handbook on inflection and word-formation (Luraghi 2000) deals exclusively with case. For its own part, case syncretism is seen mostly in terms of IndoEuropean languages. Since Indo-European patterns of case syncretism have, rightly or wrongly, contributed so much to our understanding of syncretism as a whole, it makes sense to take them as a starting point. Case syncretism in Indo-European languages is nearly ubiquitous: it was already present in Proto-Indo-European and carried through to the daughter languages that retained case. Among contemporary languages, the Slavonic family achieves the peak of variety and complexity. In Russian, one can enumerate at least six regular patterns, as shown in (1), correlated with word class, inflection class, number and animacy. In (1a), nominative and accusative are syncretic, in (1b) accusative and genitive, in (1c) genitive and locative, in (1d) locative and dative, in (1e) genitive, locative and dative, and in (1f) genitive, locative, dative and instrumental. (Further combinations of these patterns are possible as well.) The challenge of accounting for such an array of patterns led to Jakobson’s (1936 [1971]) study of Russian case semantics, which remains, directly or indirectly, one of the most influential works on case syncretism. The six primary cases are broken down into three semantic primitives, expressing directionality, scope and peripherality (see (2)). The values for individual cases are construed as composites of these primitives. For example, the genitive expresses scope, the locative expresses scope and peripherality, the dative expresses directionality and peripherality, and so on. Case

Cross-linguistic typology of features (1)

Patterns of syncretism in Russian nominals a.‘table’

NOM ACC GEN LOC DAT INS

39

stol stol stola stole stolu stolom

b.‘student (M)’ student studenta studenta studente studentu studentom

c.‘new.PL’ novye novye novyx novyx novym novymi

d.‘book’ kniga knigu knigi knige knige knigoj

e.‘mother’ 0

mat mat0 materi materi materi materju

f.‘forty’ sorok sorok soroka soroka soroka soroka

syncretism results from suppressing the expression of the semantic primitives. For example, nominative/accusative results from suppressing the expression of directionality, and genitive/locative results from suppressing the expression of peripherality. On this view, case syncretism is a reflection of the underlying network of semantic values which make up the case system.

(2)

directionality

scope

peripherality

NOM

+ +

+ +

+ + +

ACC GEN LOC DAT INS

The particular schema in (2) was revised by Jakobson himself (1958) and further by Neidle (1988). Although the particular features proposed by Jakobson have found only limited application beyond Russian – e.g. to Polish by Schenker (1964) and Slovene by Miller (1990) – the basic notion remains widely accepted, namely that individual morphosyntactic cases are the reflection of some underlying semantic network, and that case syncretism reveals otherwise covert aspects of this network (e.g. Bierwisch 1967, Wiese 1996, 2003, Halle 1997, Calabrese 1998, Ivanov 2001 and Kiparsky 2001). Now let us consider non-Indo-European languages. Although not as ubiquitous as in Indo-European, syncretism is still common. In a controlled sample, described in Appendix 1, out of sixty-four non-IndoEuropean languages which mark case inflectionally, thirty evince case

40


syncretism. However, the sorts of patterns seen in (1) are not equally well represented. The distribution of attested patterns is clearest when we describe them in terms of core and peripheral cases. On this basis we distinguish three types of case syncretism: * *

*

Type 1: syncretism of the core grammatical cases, as in (1a). Type 2: syncretism of a core case (typically the accusative or ergative) with a peripheral case, as in (1b). Type 3: syncretism of peripheral cases, as in (1c–f).

In Appendix 1, languages where case syncretism is restricted to type 1 are distinguished from those where types 2 and 3 are found. Their geographic distribution is illustrated in Map 1. All three types are well represented in Indo-European. Non-Indo-European languages, on the other hand, favour type 1. Out of the seventeen languages in the sample where case syncretism is restricted to type 1, sixteen are non-Indo-European. 3.1.2

Types of case syncretism

Type 1: syncretism of the core grammatical cases Syncretism of the cases representing the core grammatical functions of subject and object is the most common type in all languages. Conventionally, two primary systems of core case distinction are recognized, nominative accusative and ergative absolutive. In a nominative accusative system, there is a subject case (nominative) and object case (accusative). In an ergative absolutive system, there is a case for transitive subjects (ergative) and a case for intransitive subjects as well as objects (absolutive).1 Syncretism occurs under both systems. Nominative/accusative syncretism is represented in nearly every IndoEuropean language. In Proto-Indo-European this pattern characterized neuter nominals, as reflected in Ancient Greek adjectives (3a), and was also innovated independently in the daughter languages (see Ringe 1995), as in Latvian, where it is found in feminine plurals (3b). Nominative/accusative syncretism is also widespread in non-IndoEuropean languages, e.g. in plural pronouns in the Algic language Yurok (4a), and in inanimate nouns in the Dravidian language Telugu (4b).2 1

2

Additionally, one may note active stative systems, where the treatment of intransitive subjects is variable. Accusative may be optionally marked by the ending -ni in inanimates, but the preferred option is not to do so.

Map 1. Case syncretism

Inflectional marking of case is absent Syncretism of core cases only Syncretism involving both core and non-core cases Inflectional marking of case is not syncretic

42


(3) a. Greek adjective

NOM SG ACC SG GEN SG DAT SG

b. Latvian nouns (Mathiassen 1997: 43, 46)

(neuter)

‘wise’ (masculine)

soph-on soph-on soph-ou soph-oi

soph-os soph-on soph-ou soph-oi

NOM PL ACC PL GEN PL DAT PL LOC PL

(4) a. Yurok pronouns (Robins 1958: 20–1)

NOM ACC COM LOC

‘sister’ (feminine)

‘father’ (masculine)

masas masas masu masam masas

tevi tevus tevu teviem tevos

b. Telugu (Krishnamurti and Gwynn 1985: 88–9)

plural ‘we’

singular ‘I’

nekah nekah neka:noØ (?)neya:?ik

nek nekac neka?aØ (?)neya:?ik

NOM ACC GEN DAT

inanimate ‘houses’

animate ‘dogs’

il::lu il::lu il::la il::laki

kukkalu kukkalani kukkala kukkalaki

Absolutive/ergative syncretism is rather better represented in non-IndoEuropean languages than in Indo-European, though this is presumably due to the limited scope of ergative case marking in Indo-European languages (it is an innovation within Indo-Iranian languages, though here it is widespread). For example, both the non-Indo-European Dagestanian language Tsakhur (see (5a)) and Indo-European Domaaki (see (5b)) display absolutive/ergative syncretism in pronouns (in the case of Domaaki, only 2SG and 3SG pronouns). Some languages juxtapose nominative accusative and ergative absolutive systems and represent a special instance of core case syncretism familiar under the name ‘split ergativity’. For example, in the PamaNyungan language Guugu Yimidhirr, pronouns appear to display a nominative accusative system, and nouns an absolutiveergative system (see (6)). However, as argued by Zaliznjak (1973), Goddard (1982) and Comrie (1991), we can treat these as different morphological realizations of a single underlying system which distinguishes three core cases: accusative for


(5)

ABS ERG DAT

a. Tsakhur (Kibrik 1999: 130, 201) pronoun ‘I’

noun ‘idiot’

zi zi zas

baIcˇar baIcˇare baIcˇaris

43

b. Domaaki (Lorimer 1939: 44, 76)

ABS ERG GEN ACC DAT

pronoun ‘you.SG’

noun ‘man’

tu tu te tUs tUsˇ u

mnisˇ « mnisˇ «an mnisˇ «ei mnisˇ «(ek) mnisˇ «sˇ o

objects, ergative for transitive subjects and another for intransitive subjects, which we shall give the non-committal label ‘nominative-absolutive’. On this interpretation, split ergativity can be treated as the result of two different patterns of case syncretism: the nominative-absolutive is a nonautonomous case, syncretic with the ergative in pronouns and with the accusative in nouns (see (7)). (6)

Guugu Yimidhirr (Haviland 1979: 47–51, 66–7) ‘I’ ngayu nganhi

NOM ACC

(7)

ABS ERG

gabiir gabiirrngun

Guugu Yimidhirr (alternative representation)

ERG NOM-ABS ACC

‘girl’

‘I’

‘girl’

ngayu ngayu nganhi

gabiirrngun gabiir gabiir

This interpretation is strengthened by the fact that some languages have a system similar to Guugu Yimidhirr’s, but with a set of nominals where all three core case distinctions are morphologically overt. For example, in the Pama-Nyungan language Wagaya, the nominative-absolutive is syncretic with the ergative for first and second person pronouns and with the

44


accusative for demonstratives and nouns, while for the third person pronouns it is distinct (see (8)). Morphologically, the Wagaya paradigms illustrate an important feature of many languages that display split ergativity: the accusative and ergative are morphologically marked (accusative -iny and ergative -l:), while the nominative-absolutive is morphologically unmarked. The three paradigms in (8) can then be described in terms of differential marking of the accusative and ergative. First and second person pronouns mark a distinct accusative, while third person pronouns, demonstratives and nouns mark a distinct ergative. (8) Wagaya (Breen 1976: 591)

ERG NOM-ABS ACC

I ‘you.PL’

II ‘he’

III ‘that.M’

ir ir iriny

yuw«l: yuwu yuwiny

bul«l: bulu bulu

I first and second person pronouns II third person pronouns III demonstratives, nouns

The split between accusative and ergative marking is typically correlated with a hierarchy of properties first proposed for Australian languages (Silverstein 1976), and now commonly known as the animacy hierarchy. This name is somewhat misleading, in as much as animacy is only one component, and it is perhaps better seen as a hierarchy of ‘inherent referential content’ (Goddard 1982) or ‘individuation’ (Timberlake 1975). At one extreme, specific, animate entities have a distinct accusative form, and at the other extreme, non-specific, inanimate entities have a distinct ergative form. A representative example comes from the Pilbara languages, a group of Pama-Nyungan languages spoken in Western Australia, where split ergativity shows different cut-off points in the different languages of the group, permitting the extrapolation of the hierarchy in (9): (9)

1SG > 2SG > 1INCL DU > 1INCL PL > 2DU > 2PL > 3 >this > that > indefinite > animate > meat, vegetable > other inanimate (Dench 2001: 122)


45

One can see this as the conjunction of several component hierarchies, for example literal animacy (animate > inanimate), person (1 > 2 > 3), number (singular > non-singular), word class (pronouns > demonstratives > nouns) and definiteness (definite > indefinite). Because the interpretation of split ergativity seems to be relatively clear, it is tempting to extend it to the analysis of languages which have syncretism of the core cases but do not display split ergativity as such, as in examples (6)–(8). If valid, this would manifest itself in two respects. *

*

Morphologically, the accusative and ergative should be marked, while the nominative and absolutive should be the unmarked default. Core case syncretism should then manifest itself as the extension of this default from the nominative-absolutive to both core cases. The division between syncretic and non-syncretic paradigms should be sensitive to the animacy hierarchy. That is, given a hierarchy such as in (9), all instances of nominative/accusative syncretism should fall towards the right-hand side, and all instances of absolutive/ergative syncretism should fall towards the left-hand side.

It turns out that these expectations are only partly met. The morphological correlation appears often to be true, with syncretic nominative/accusative being realized by a default form otherwise used solely for the nominative, and syncretic absolutive/ergative realized by a default form otherwise used solely for the absolutive. For example, this is the case in Russian when one compares student ‘student’ and stol ‘table’, shown in (1a) above: nominative/accusative stol is morphologically the bare stem, and this corresponds to the distinct nominative singular form student, likewise the bare stem. The identification of bare stems as the morphological default here might seem unobjectionable.3 However, neither is it automatic. Consider the plural forms of Russian sapog ‘boot’ and soldat ‘soldier’ in (10), which represent a class of Russian nouns whose genitive plural ending is zero. The syncretic nominative/accusative sapogi corresponds to the distinct nominative soldaty (the alternation i y is conditioned by the stem-final consonant), but it is rather the genitive plural (or accusative/genitive in the case of soldat) which represents the 3

An assumption often made within recent works within the framework of Optimality Theory, e.g. Kiparsky (2001) and Wunderlich (2004).

46


bare stem. While we may still want to equate the nominative plural with the morphological default, we must also admit that there is no independent evidence for doing so.

(10) Russian masculines with zero genitive plural

NOM PL ACC PL GEN PL LOC PL DAT PL INS PL

inanimate ‘boots’

animate ‘soldiers’

sapog-i sapog-i sapog sapog-ax sapog-am sapog-ami

soldat-y soldat soldat soldat-ax soldat-am soldat-ami

Further, there are examples which show the opposite pattern, where nominative/accusative syncretism appears to suggest that it is the nominative which is marked with respect to the accusative. This is familiar from Indo-European, for example in Latvian, where a plural ending of the shape -Vs is used for the accusative in the masculine declension classes and for the nominative/accusative in the feminine declension classes (see (3b)). Sometimes, both patterns occur in a single language. Thus, in the second declension of Latin, the syncretic neuter ending -um corresponds to the accusative of masculines, while in some third declension nouns, the syncretic neuter form is the bare stem, corresponding to the nominative of masculines (see (11)). (11) Latin second declension

NOM SG ACC SG GEN SG DAT SG ABL SG

neuter ‘war’

masculine ‘slave’

bell-um bell-um bell-i bell- o bell- o

serv-us serv-um serv-i serv- o serv- o

third declension

NOM PL ACC PL GEN PL DAT PL LOC PL

neuter ‘sea’

masculine ‘conqueror’

aequor aequor aequor-is aequor-i aequor-e

victor vict or-em vict or-is vict or-i vict or-e


47

The predictions suggested by the animacy hierarchy about the distribution of syncretic paradigms are met only in certain respects. We shall look at the three components of the hierarchy which are best represented in our corpus: (i) literal animacy, (ii) word class and (iii) number. There is evidence of a positive correlation between literal animacy and the distinction of nominative and accusative, for example in Telugu (4b) and Eastern Armenian (Minassian 1980). Russian has a similar correlation, but it is embedded within the system of infectional classes. Thus for certain classes, nominative and accusative are distinct only for animates (as in student ‘student(M)’ in (1b)), while for other classes animacy is irrelevant (as in the singular paradigm of mat0 ‘mother’ in (1e)). The syncretism displayed by neuter nouns in Indo-European (see the Greek example in (3a)) may represent a largely morphologized version of the same correlation, in that these nouns typically have inanimate referents.4 However, there is no direct correlation between semantic animacy and gender: while the neuter gender is composed of inanimates, many inanimates are found in the masculine and feminine gender as well, and so may have distinct nominative and accusative forms. Curiously, in languages with exclusively ergative absolutive marking, the analogous correspondence does not seem to obtain. That is, we do not find languages (at least in our sample) where absolutive/ergative syncretism is directly correlated with inanimacy. This suggests that sensitivity to inanimacy is contingent on sensitivity to animacy: a language can have case marking determined by animacy, or both by animacy and inanimacy, but not by inanimacy alone. A correlation between inanimacy and absolutive/ergative syncretism can occur only in the context of split ergativity. The word class hierarchy predicts, broadly speaking, an opposition between pronouns (possibly restricted to first and second person) and other nominals, with core case syncretism being found in one group but not the other. This seems to be largely true. For example, Indo-European languages regularly have nominative/accusative syncretism in nouns but may lack it in pronouns. Non-Indo-European languages may also show a similar distribution. For example, in Finnish, where nouns and pronouns take essentially the same case endings, the accusative is always syncretic with some other case for nouns, while pronouns have the distinct accusative ending -t (Fromm 1982). In Nenets, nominative and accusative are 4

Exceptions include Sanskrit mitra ‘friend’ and vr:tra ‘foe’ (Gregory Stump (personal communication), citing Whitney 1889: x1185c).

48


frequently syncretic for nouns (namely in all dual paradigms, all plurals with a possessive suffix, and singulars with a first person possessive suffix), but distinct for all personal pronouns in all numbers (Salminen 1997). In Yurok, nominative and accusative are syncretic for the 3SG pronoun but distinct for 1SG and 2SG pronouns (Robins 1958). Correspondingly, absolutive/ergative syncretism affects pronouns more than nouns. Thus in Tsakhur, pronouns have it and nouns do not (see (5a)). In Yup’ik, 1SG and 2SG pronouns have relative/absolutive syncretism and the 3SG does not (Jacobson 1995); the relative case in Eskimoan languages combines the functions of ergative and genitive. Of course, the correspondence between word class and core case syncretism is not always perfect. For example, in the Papuan language Suena (of the Trans-New Guinea phylum), there is a distinct accusative form for interrogative pronouns and demonstratives, but not for personal pronouns (Wilson 1974), which would seem to be the reverse of what the hierarchy predicts. Nevertheless, one can extract a rough implicational hierarchy: if syncretism of nominative and accusative is found somewhere among pronouns, it is found somewhere among casemarked nouns too; and if syncretism of absolutive and ergative is found somewhere among nouns, it is found somewhere among case-marked pronouns as well. The one element of the animacy hierarchy which clearly does not apply to languages without split ergativity is the number correlation. We should expect non-singular numbers to favour nominative/accusative syncretism, and the singular to favour absolutive/ergative. Instead, we find simply that both patterns of syncretism are more likely in the non-singular. Thus nominative/accusative syncretism in the dual and plural is quite general in Indo-European. The same is common in non-Indo-European languages, as seen in the languages described in the sample in Appendix 1 (see references there): in Finnish (nouns), in the Australian language Mangarayi (demonstratives), in the Nilo-Saharan language Murle (in one class of nouns), in the Samoyedic language Nenets (plural nouns with a possessive suffix, and all dual nouns), in the Uto-Aztecan language Yaqui and in Yurok (first and second person pronouns). While this corresponds to the expectation, by the same token, we find absolutive/ ergative in the non-singular alone in Basque (proximate nouns, demonstratives and the second person pronoun), Domaaki (nouns; see (5b)) and Yup’ik (nouns). Thus the correlation seems to be a more general one between non-singular number and syncretism as such (familiar since Hjelmslev 1935–7; see also Chapter 2: x2.2.2.), and nothing to do with


49

the relationship between core case marking and the animacy hierarchy. Even this correlation does not hold without exception, for example in Elliot’s (1999) description of Rincoń Luiseno (Uto-Aztecan; cited in Kathol 2002), nominative and accusative are syncretic for nouns in the singular but not in the plural (12); note that number is not distinguished by the non-core cases. (12)

NOM ACC ILL ABL LOC INS

Rincoń Luiseno ‘sling’: Elliot (1999: 21) singular

plural

pı´ ı´ vanla-sh pı´ ı´ vanla-sh

pı´ ı´ vanla-chum pı´ ı´ vanla-shmi pı´ ı´ vanla-yk pı´ ı´ vanla-Nay pı´ ı´ vanla-Na pı´ ı´ vanla-tal

Type 2: syncretism of a core case with a non-core case In type 2, the core cases are kept distinct from each other, but one of the core cases is itself syncretic with one of the peripheral cases. Typically it is the marked core case which is affected, that is the accusative in a nominative accusative system and the ergative in an ergative absolutive system. For example, in Finnish, the ending -n, which is used solely for the genitive in pronouns, is used for both genitive and accusative in singular nouns (see lukko in 13a). In the Chukotko-Kamchatkan language Koryak the locative ending found in the first declension (see ‘(a) grandmother’ in 13b) is used for both locative and ergative-instrumental in the second declension (see ‘grandmother’ in 13b). (The second declension pattern is used for specific human individuals, the first declension pattern is used for all other nouns; some nouns with human referents may follow either pattern.) A possible explanation of the motivation for the type 2 pattern comes from the Slavonic languages, such as Russian (see Klenin 1983 for a detailed discussion). In general, it seems to have been used to restore the distinction between nominative and accusative that had been lost in a number of nominal paradigms. The earliest records show that nominative and accusative were syncretic in most nominal paradigms, while the interrogative/relative pronoun had type 2 syncretism of the genitive/accusative.

50


(13) a. Finnish core cases (Fromm 1982: 89; foldout 1)

NOM ACC GEN

noun ‘lock’

pronoun ‘I’

lukko luko-n luko-n

mina¨ minu-t minu-n

b. Koryak core cases (Zˇukova 1972: 99)

ABS ERG-INS LOC

second declension ‘grandmother’

first declension ‘(a) grandmother’

anja anja-na-k anja-na-k

anja anja-ta anja-k

This type 2 pattern was extended to paradigms with nominative/accusative syncretism, following a word-class hierarchy (pronouns > nouns) and, within nouns, a hierarchy of animacy or individuation (singular > plural, male > female, human > animal), where it has remained restricted to animate arguments.5 Eastern Armenian underwent a similar development, extending the accusative/dative syncretism of pronouns to animate nouns, which otherwise have nominative/accusative syncretism. The diachronic layering of type 1 and type 2 syncretism produced the synchronic alternation of the two patterns we still see in Russian (see (1a, b)) and similarly in Eastern Armenian, where nominative/accusative alternates with accusative/dative. The same sort of alternation, involving the ergative case, is found in Lak, where the ergative is syncretic with the genitive in nouns and the absolutive in pronouns (see (14); note this is just a portion of the unusually extensive case paradigm). (14)

ABS ERG GEN DAT SUPERESS COM

5

Lak (Zˇirkov 1955: 36, 64–6) noun ‘house’

pronoun ‘I’

k’atta k’atlu-l k’atlu-l k’atlu-n k’atlu-j k’atlu-sˇ sˇ a

na na ttu-l ttu-n ttu-j ttu-sˇ sˇ a

In pronouns, the extension of genitive/accusative syncretism was extended even to the point of replacing a distinct accusative ending, as happened with the 3SG feminine pronoun. In nouns, however, the presence of a distinct accusative ending seems to have blocked the spread of this pattern.


51

More unusually, we may find an alternation between two type 2 patterns, as in the Mongolic language Bonan, where the accusative is syncretic with the genitive in nouns and with the dative-locative (a single case in Bonan) in pronouns (see (15)). Note that for both languages the case endings are independent of number. (15) Bonan (Todaeva 1997: 35)

NOM GEN ACC DAT-LOC ABL INS-COM

noun ‘foliage’

pronoun ‘he’

labcˇoN labcˇoN-ne labcˇoN-ne labcˇoN-de labcˇoN-se labcˇoN-Gale

ndzˇaN ndzˇaN-ne ndzˇaN-de ndzˇaN-de ndzˇaN-se ndzˇaN-Gale

However, there are also examples of type 2 syncretism where this functional explanation – the restoration of core case distinctions – clearly does not apply. In the Pama-Nyungan language Ngiyambaa, the distinct ergative-instrumental ending has been replaced by the dative in the speech of younger speakers. Since the core cases were already distinct in the older system, this has no obvious functional motivation. Rather, it seems to be an instance of morphophonological simplification. In the older system, the dative ending was -gu for all stem types, while the ergative-instrumental had a series of allomorphs, depending on stem type: -dhu for stems ending in -y or -yN, -du for stems ending in -n, -u for stems ending in -r or -l, and -gu (identical to the dative) elsewhere. What the younger speakers appear to have done is to generalize the default allomorph -gu, eliminating the morphophonological alternation (Donaldson 1980: 84–5; Austin 1986). In defining type 2 syncretism we have spoken simply of syncretism with a peripheral case, but the choice of peripheral case does appear to be limited. With the accusative, we typically find either the genitive (as in the Slavonic languages, Finnish6 and the Pama-Nyungan language

6

Presumably owing to a phonological accident: the original accusative *-m and genitive *-n fell together as a result of the merger of these sounds (Anttila 1972: 103), a process common to all of Baltic Finnic.

52


Martuthunira (see Appendix 1)) or the dative (as in Eastern Armenian, Ngiyambaa, and Phalura) or both (Bonan).7 It is generally held that these two cases, along with the core cases, together constitute the structural or syntactic cases, which express basic syntactic functions. These are opposed to the semantic cases, such as directional cases or the instrumental, which typically serve as adjuncts. Thus, the choice of case form to stand in for the marked core case under type 2 syncretism is constrained by this division. Within that, though, it is difficult to find a motivation for the choice of genitive versus dative. This is especially striking when one compares Slavonic with Eastern Armenian. Both have essentially the same inventory of cases and employ type 2 syncretism to the same end (marking of animate accusatives), but one uses the genitive and the other the dative. With the ergative, type 2 syncretism in our sample most typically joins it with the genitive, as in the Tacanan language Araona, the isolate Burushaski, Lak and the Tibeto-Burman language Limbu (see Appendix 1). It is likely that this is not a random choice, in that there are languages which have cases which inherently combine the functions of ergative and genitive (e.g. the relative case of the Eskimoan languages). Such constructions may have their origin in nominalizations, with the agent expressed by the genitive. However, although diachronic explanations may be found, it is unlikely that a direct, synchronic motivation can be demonstrated for most type 2 patterns. Type 3: Peripheral case syncretism Whereas both Indo-European and non-Indo-European languages display similar behaviour with respect to syncretism involving the core cases (types 1 and 2), they show significant differences when it comes to syncretism of peripheral cases. Many Indo-European languages display multiple conflicting patterns of peripheral case syncretism, such as we have seen in Russian (1). It is just these sorts of patterns which have led to the idea that case values can be decomposed into an intersecting network of component values, as in Jakobson’s (1936 [1971]) analysis shown in (2). However, evidence for such networks is exceedingly sparse outside of IndoEuropean. The sorts of patterns which are found in non-Indo-European are themselves fairly marginal within Indo-European languages.

7

One also finds languages where both these case functions are combined in a single case, e.g. the dative-accusative of Pengo.


53

One pattern involves the collapse of all peripheral cases and is found in Ingush (16) and in Georgian (see (20) below). In both instances the rich declensional paradigm is reduced to two forms, one for the absolutive form and one for everything else. This presupposes type 2 syncretism, in addition to the collapse of peripheral cases, since the ergative patterns with the peripheral cases. (16)

Ingush (Nichols 1994: 99) ‘big village’ joqqa joqqacˇa joqqacˇa joqqacˇa

ABS ERG GEN DAT

jurt jurtuo jurta jurtaa

Similar patterns have arisen occasionally in individual branches of IndoEuropean (no comparable pattern was present in Proto-Indo-European), where it likewise co-occurs with core case syncretism (either type 2, as in Ingush or Georgian, or type 1). For example, in Russian, the numerals ‘40’, ‘90’ and ‘100’ have only two forms, one for the core cases nominative and accusative, and one for the peripheral cases (see above, 1f). In the middle Iranian language Sogdian, the plural paradigm of light stem nouns shows essentially the same pattern, collapsing the nominative and accusative on the one hand, and the genitive, locative and ablative on the other (Sims-Williams 1982). In German, weak adjectives, exemplified by bo¨se ‘bad’, likewise display only two forms: nominative singular versus a form which serves for the genitive/dative singular as well as all plural cases (17). The behaviour of the accusative varies with the gender of the modified noun. If neuter or feminine, (17) German ‘the bad child’ (neuter) NOM ACC GEN DAT

das das des dem

bo¨se bo¨se bo¨sen bo¨sen

Kind Kind Kindes Kind(e)

‘the bad man’ (masculine) der den des dem

bo¨se bo¨sen bo¨sen bo¨sen

Mann Mann Mannes Mann(e)

54


it follows the type 1 pattern and so is identical to the nominative. If masculine, it follows the type 2 pattern and is identical to the peripheral form. The second type of peripheral case syncretism which is found relatively often in non-Indo-European languages involves syncretism of a directional case (e.g. illative or allative) with the dative or locative, which may well have a semantic motivation. Thus, in the Finnic language Erzja Mordvin, the dative ending is used for the illative in singular definite nouns (and optionally in plural definite nouns).

(18)

NOM GEN DAT ILL ABL INESS EL PROL CMP ABESS

Erzja Mordvin (Feoktistov 1966: 180) ‘the house’

‘(a) house’

kudos’ kudont’ kudonten’ kudonten’ kudodont’ kudosont’ kudostont’ kudovant’ kudosˇ kant’ kudovtomont’

kudo kudon’ kudonen’ kudos kudodo kudoso kudosto kudova kudosˇ ka kudovtomo

In the Pama-Nyungan language Diyari, the allative case is non-autonomous: with singular nouns and male personal names it is syncretic with the dative, and otherwise it is syncretic with the locative; cf. ‘boy’ versus ‘people’ in (19).

(19)

ERG ABS ACC LOC ALL DAT ABL

Diyari (Austin 1981: 51) ‘boy’

‘people’

kankuli kanku kanku kankun5i kankuya kankuya kankundu

kan:awar:ali kan:awar:a kan:awar:ana kan:awar:aNu kan:awar:aNu kan:awar:an:i kan:awar:aNundu ;


55

There is some cross-linguistic evidence that the use of the dative or locative case form in a directional function may have some semantic motivation. For example, in Turkish, the dative is used for goal of motion as well as for indirect objects (cf. Blake 1994: 145), and the alternation of location and direction is familiar from Indo-European prepositions such as Latin in ‘in, to’ or in Russian v ‘in, to’, where the different functions are correlated with different patterns of case government. Nevertheless, the examples of peripheral case syncretism from non-Indo-European languages have no direct parallel in Indo-European languages, as the latter typically lack a distinct directional case. Probably the closest thing to an Indo-European style system of multiple syncretic patterns is found in Georgian, which has two patterns. The first is found in the so-called Old Georgian plural paradigm, inherited from Old Georgian but little used in the contemporary language. (The normal system of case marking in the plural is identical to the singular.) It involves the collapse of all peripheral cases, as seen in buzni ‘flies’ in (20), though note that the vocative does not pattern with the peripheral cases. This paradigm is also defective, in that it is not used for the instrumental or adverbial case. The second syncretic pattern is found with attributive adjectives and involves syncretism of absolutive/genitive/instrumental and dative/adverbial. This combination of cases does not correspond to any pattern found in an Indo-European language (or any other non-Indo-European language). (20)

Georgian (Aronson 1991: 219, 232–3) ‘flies’

ABS INS GEN ERG DAT ADV VOC

buzni — buz-t buz-t buz-t — buz-o

‘old book’ dzveli dzveli dzveli dzvelma dzvel dzvel dzvelo

c’igni c’igni-t c’igni-s c’ignma c’ign-s c’ign-ad c’igno

Morphologically, the syncretism in these paradigms reflects a sort of agglutinative structure quite unlike what is found in Indo-European. Both paradigms involve partial morphological truncation. The syncretic form of the Old Georgian plural lacks case endings altogether, substituting the

56


(typically verbal) plural marker -t. The syncretic forms of the attributive adjectives result from deletion of the final elements -s ‘GEN’, -t ‘INS’, -s ‘DAT’ and -ad ‘ADV’ (Aronson 1991: 236; Carstairs-McCarthy 1994: 778–9), leaving a consonant-final base (dzvel) or a vowel-final base (dzveli). 3.1.3 Conclusion In Indo-European languages, case syncretism displays a wide variety of patterns, as illustrated in (1), which has led to the proposal that there is a network of semantic primitives that underlie case: individual cases are composites of these primitives, and case syncretism within a language represents their different combinatorial possibilities. Cross-linguistically, however, most instances of case syncretism can be described in terms of the behaviour of the marked core cases (accusative and ergative), which are either identical to the unmarked core case (type 1), or to one of the peripheral cases (type 2). Less frequently, peripheral case distinctions are effaced. These patterns show a sensitivity to basic syntactic relations but are not striking evidence for a network of semantic relations. That said, the semantic value of cases probably does play a direct role in some instances of case syncretism (as in (18) or (19)), but sporadically enough that no extensive, cross-linguistically viable system of semantic primitives can be derived from them. We have speculated that the motivation for types 1 and 2 may lie in the differential marking of objects or agents, such as is found in languages with split ergativity, where accusative marking may be omitted low in the animacy hierarchy, and ergative marking omitted high in the hierarchy. Overall, there is a correlation, but it is quite inexact. For example, in the variety of Luisenõ described by Malećot (1963) (21), if one were to look solely at the contrast in accusative case marking between the animate noun ja?a´sˇ and the inanimate noun sˇamUt, one might say simply say that animates mark the accusative and inanimates do not. But the agreeing adjective alaxwIsˇ marks the accusative in both instances. Thus, there is differential object marking here, though it is correlated with semantic properties and still requires a morphological account. (21)

Luisenõ (Malećot 1963: 288) animate: man + bad = ‘bad man’ inanimate: grass + bad = ‘bad grass’

NOM ACC

ja?a´sˇ ja?a´tsˇ I

alaxwIsˇ alaxwItsˇ I

sˇ amUt sˇ amUt

alaxwIsˇ alaxwItsˇ I


57

As often as not, patterns of case syncretism are thoroughly morphologized and yield only dim echoes of this putative original motivation. At any one synchronic stage multiple versions of the same pattern may be found. For example, in Russian, nominative/accusative has at least three different sources, depending on nominal class,8 and further, may or may not be sensitive to animacy, also dependent on nominal class. We interpret the cross-linguistically dominant patterns primarily as the unmotivated, morphologized residue of diachronically recurring, motivated patterns. 3.2

Person

3.2.1 Introduction More so than any other feature, the values of person are relatively constant cross-linguistically. Thus, the values first person (possibly distinguishing inclusive and exclusive), second person and third person are largely sufficient to characterize any language, with further values involving distinctions subordinate to these (e.g. honorific versus familiar, proximate versus obviative, and same subject versus different subject). This constrained range of values allows us to make a fairly direct comparison between syncretic patterns in different languages. This is of particular interest because it can be used to assess claims that patterns of syncretism are correlated with underlying morphosyntactic or semantic relationships, since these relationships are presumably common across languages. The most common domain for person marking is as the subject marker on verbs, and it is on this that we will focus in this section. (The interaction with object marking involves a distinct set of conditions and is treated in x3.3.) Syncretism of subject person is common: in the language sample from the World Atlas of Language Structures, illustrated in Appendix 2, out of 141 languages with inflectional marking of subject person, 61 display syncretism in some portion of the verbal paradigm (see Map 2). In this section we shall ask two questions: (i) do patterns of person syncretism reflect morphosyntactic or semantic natural classes, and (ii) do directional effects reflect markedness relationships between person values?

8

Thus, looking at nouns alone, nominative/accusative was inherited from Proto-IndoEuropean in neuters (fourth declension), innovated in Common Slavonic as a result of sound change in singular o-stem masculines (first declension), singular i-stems (third declension), and the plural of all but o-stem masculines, and then further innovated in the Old Russian period through morphological change in the plural of o-stem masculines.

Map 2. Syncretism in the marking of subject person on verbs

Inflectional marking of person is syncretic Inflectional marking of person is not syncretic

Inflectional marking of person is absent


59

3.2.2 Patterns of syncretism It will be useful to distinguish two contexts. First, a given set of values may be syncretic for all paradigms, which we will term ‘complete’ syncretism. Second, a set of values may be syncretic only in some paradigms, restricted, for example, to a particular tense or conjugation class. This we term ‘partial’ syncretism. Note that nothing further is implied by this distinction; we do not assume that two different interpretations are warranted. Nor is the distinction always meaningful; for example, if a language has only one set of person markers, it achieves complete syncretism vacuously. Nevertheless, it is a useful criterion, allowing some distinct tendencies to emerge. Where syncretism is complete, there is a sharp contrast between the behaviour of person in the singular versus non-singular. Complete syncretism of person solely in the singular is uncommon. In the sample described in Baerman (2002b; sources listed there), out of twenty-nine examples of complete person syncretism in the sample (from twenty-seven languages), only six are restricted to the singular (see (22)). Of these, two come from languages where person is not distinguished in the plural, so that it is not so much the case that the syncretic pattern is restricted to the singular, but that person marking itself is restricted to the singular. In the non-singular, 1/2 and 2/3 both occur in roughly equal measures, while 1/3 is less common. A similar distribution is found where number is irrelevant, with examples of 1/2 and of 2/3 predominating. (22)

Examples of complete syncretism, by language singular

non-singular

1/3

Koiari*, Zoque

Aleut, German, Hindi

2/3

Atakapa, Hindi, Nivkh*, Nubian

Amele, Kapau, Kewa, Kobon, Slovene

Chitimacha, Guambiano, Kiwai, Wambon

Burarra, Dogon, Manchad, Nubian, Prinmi, Tetun

Hunzib, Ingush, Nez Perce, Sango, Waskia

1/2

number-neutral

*

Person distinguished in singular only.

Thus, cross-linguistically, there seems to be a preference for syncretism of first with second person, and of second person with third, in both cases restricted to non-singular or number-neutral contexts. Interestingly, this

60


corresponds well with the sorts of patterns found in free pronouns, as shown in (23). (23)

Compound person values in free pronouns singular

non-singular

number-neutral

1/3

Dakar Wolof

2/3

Amele* , Kalam* , Kamoro, Kobon, Korafe* , Meyah, Mansim, Nganasan, Nez Perce* , Sango* , Warekena* , Wolof

Kawesqar

1/2

Awa* , Fongbe, Slave* , Yimas

Winnebago

*Examples from Cysouw (2003). Other sources: Dakar Wolof from Nussbaum, Gage and Varre (1970), Fongbe from Lefebvre and Brousseau (2002), Kamoro from Voorhoeve (1975), Kawesqar in Clairis (1985), Mansim and Meyah from Reesink (2002), Winnebago from Lipkind (1945), Yimas from Foley (1991).

As with inflectional marking, nearly all the examples of combined person values in free pronouns involve 1/2 and 2/3 in the non-singular; none involve syncretism of singular person values alone. (Note that the evidence from number-neutral pronouns is exceedingly thin: we are aware of only two examples, the 1/2 emphatic personal pronoun of the Souian language Winnebago (Lipkind 1945: 29) and the 2/3 pronoun of the Patagonian language Kawesqar (Clairis 1985: 465); the latter is only partial, in that there are distinct possessive forms for these two persons.) The preference for certain patterns across languages of different families and with different morphological structure, and the parallel between these patterns and the evidence of free pronouns, suggests a connection. Since it is reasonable to suppose that free pronouns, as discrete lexical items, are semantically coherent, this connection may lie in the morphosyntactic or semantic structure of person values. However, we know of no model of person features that will account precisely for both the rarity of 1/3 syncretism and the rarity of syncretism which is restricted to the singular. For example, Harley and Ritter (2002), in a model designed to account for the person values of pronominal forms, predict that 1/2 pronouns should


61

be found, but not 2/3.9 However, as we have seen, not only are 2/3 pronouns found, they appear to be more common than 1/2 pronouns. They further predict that if a language does display a 2/3 or 1/2 pronoun, this is a result of accidental homophony and will be disambiguated by verb agreement (p. 513, note 42). This prediction is contradicted by Amele, Kobon (dual only) and Nez Perce. Even if this prediction were true, it is curious that accidental homophony would be more frequent than systematic syncretism. Further, their model predicts that number values should have no influence on patterns of person syncretism (because number is construed as a separate node in their model of feature structure and does not interact with person). If we choose to ascribe a semantic rationale to these patterns, it is probably significant that non-singular numbers favour syncretism, since this is precisely the context where there may be referential overlap, and hence ambiguity. Thus, in a language without an inclusive exclusive distinction, first person plural may or may not include the addressee, so blurring the distinction between first and second person. Likewise, second person plural may or may not be construed as including some nonaddressees, so blurring the distinction between second and third person. Nevertheless, there are reasons to be cautious about overemphasizing the semantic naturalness of such syncretic combinations, which are especially apparent when one considers the behaviour of the first person inclusive. Since semantically it overlaps with first and second person, we should expect syncretism with those two persons. While this does occur, instances of 1INCL/1EXCL syncretism are far more frequent than 1INCL/2, which is not readily accounted for if semantic overlap is deemed to license the syncretism; signficantly, 1INCl/2 syncretism is no more frequent than 1INCL/3, which cannot be accounted for by the notion of semantic overlap (Cysouw forthcoming). An example of such an ‘unnatural’ pattern comes from the Austronesian language Kwamera (24), where the first inclusive is syncretic with the third person in the dual, precisely the pattern one would expect not to find.10 9

10

Harley and Ritter (2002) model person features as a binary branching tree structure. The base node (‘Referring expression’) is interpreted by default as third person, the node dependent on that (‘Participant’) is interpreted by default as first person, and the node in turn dependent on that (‘Addressee’) is interpreted as second person. Thus, the only syncretisms that can be expressed by this model are 1/2 (through underspecification for the Addressee node) and 1/2/3 (through complete underspecification). Note though that Kwamera has an impersonal marker k-, and a morphological relationship between impersonals and the first plural is known from other languages. Thus, in some Athapaskan languages, the unspecified person marker is used for the first plural

62


1 2 3 1INCL

Kwamera verbal prefixes (Lindstrom and Lynch 1994: 10) singular

dual

plural

iakikr—

iak-rouik-rouk-rouk-rou-

iak-haik-hahasa-ha-

Conversely, in the Tibeto-Burman language Kinnauri (25), the first person inclusive has a distinct form (dual and plural), while the first person exclusive and second person non-singular are syncretic (1INCL PL is optionally syncretic with this form). (25)

1INCL 1 2 3

Kinnauri verbal suffixes (Sharma 1988: 132–3)11 singular

dual

plural

– -k -n -n˜ (HON) -Ø -t -d -sˇ (HON)

-icˇ -e -cˇ -cˇ -sˇ

-cˇ -e -cˇ -cˇ -sˇ

Since 1EXCL and second person are mutually exclusive, there can be no question of a semantic or functional overlap between the values of the syncretic form (see Noyer 1997 for discussion of a similar pattern in Mam).12 Partial syncretism does not lend itself to the same generalizations as complete syncretism. Most strikingly, instances of syncretism affecting solely singular person values predominate. Clearly, no strong case can be made for the sort of semantic motivation discussed above. In part, the cause must lie in the morphology of the forms. One banal observation is

11

12

(Rice 2000: 201), in Ngiti, the third person indefinite form is used for first person inclusive when preceded by a free pronoun (Kutsch Lojenga 1994), in colloquial Finnish, impersonal forms may substitute for first plural (Karlson 1999: x97), and, of course, the use of the French impersonal on for first plural is well known. Second and third person singular have honorific forms (in the case of third person, the plural is used). Sharma (1988) does not explain what conditions the other allomorphs. Other examples of 1EXCL/2 syncretism alongside a distinct 1INCL come from Udihe (see (37) below), Burarra (see (36) below), Aymara, Nunggubuyu and Tiwi (Baerman 2002b).


63

that many of the instances of partial syncretism involve affixes which are simply phonologically less robust than the non-syncretic affixes, for example they are shorter, or vocalic as opposed to consonantal, so making them more susceptible to phonological decay or merger. Of course, the incompleteness of a syncretic pattern does not preclude its having a semantic motivation, any more than completeness is a proof of semantic motivation. The lack of clear patterning under partial syncretism is at best an indication of multiple causation. 3.2.3 Directional effects Directional effects indicate a morphologically asymmetric relationship between two values. It has often been suggested that this asymmetry obtains at the level of content too. For example, the default status of the third person has long been recognized; e.g. Watkins’ Law (Watkins 1962) holds that 3SG serves as the base for analogical remodelling of conjugational paradigms. More recently, under Impoverishment Theory (Noyer 1998), it is held that third person is unmarked with respect to other persons, and that directional effects result from the replacement of first or second person forms by third person forms. Thus, under this framework, Bobaljik (2002) argues that the syncretic 2SG/3SG aorist and imperfect forms in Macedonian (26) result from the replacement of second person by third person. (The example is taken from Stump (1993) and is identified as being directional because (i) the person marker is -Ø, characteristic of the third person in the present, and (ii) it lacks the element -v-, whose absence is characteristic of the third person (singular and plural) in both the aorist and imperfect.) (26)


Macedonian present

aorist

imperfect

-am -sˇ -Ø -me -te -at

-v -Ø -Ø -vme -vte -a

-ev -esˇ e -esˇ e -evme -evte -ea

Under such an approach, it is assumed that feature values are arranged in a hierarchy of markedness, and that synchronic directional effects are

64


prima facie evidence of this hierarchy. Thus, where first or second person are syncretic with third, the resulting syncretic form should somehow be identifiable as third person rather than first or second (provided directional effects are observable at all). In the case of 1/2 syncretism, opinions vary about the markedness hierarchy that obtains; for example, Noyer (1997: 114) and Helmbrecht (2003: 6) argue that first person is marked, Harley and Ritter (2002: 486) argue the opposite. However, when we look at a larger set of examples, we find little evidence that directional effects reflect any consistent hierarchy. Below we review what we consider to be the reasonably convincing examples of directionality that we have come across; the corpus could be expanded or shrunk depending on what one considers to be a convincing example of directionality. We have limited ourselves to examples where the syncretism is realized by an overt morphological marker, rather than by a bare stem. In so doing we eliminate instances which would be transparently interpretable as underspecification under practically any formal model. First person / third person There are not many convincing examples of directionality involving these values. The Papuan language Koiari of the Trans-New Guinea phylum (27) seems to confirm the expectation that the third person form should prevail, in as much as the 1SG obligatory mood may optionally be identical to the 3SG (elsewhere, first and third person are always identical). Livonian (see (42) below) presents a diachronic example where the third person form prevailed. (27) Koiari (Dutton 2003: 345, 351) obligatory mood

1SG 2SG 3SG PL

option

option 2

imperfect

perfect

-ahina -ihama -ahima -ihava

-ahima -ihama -ahima -ihava

-ma -a -ma -a

-nu -nua -nu -nua

!

But there are also examples where the first person form seems to prevail. In the Nilo-Saharan language Murle (28), first person (inclusive) and third


65

person are syncretic (in both numbers) in the subjunctive but not the perfect. In the perfect, the first person is characterized by prefixed k-, while the third person has no prefix. The syncretic form in the subjunctive has a prefixed k-, just as the distinct first person of the perfect. Thus, it appears as if an overtly first person element is serving for third person as well.13 (28) Murle (Arensen 1971: 83) perfect 1SG 2SG 3SG 1INCL PL 1PL 2PL 3PL

k- -a -u (-un) k- -it k- -da -tu -it

subjunctive

! !

kØ kk--it k--da -it k--it

Second person / third person Here there are rather more clear examples of directionality than with 1/3 syncretism, though the results are mixed. In some examples the third person form prevails, e.g. in Macedonian, as discussed in (26). Likewise, in the Nilo-Saharan language Nobiin (29), the 2SG appears to take the form of the 3SG in both tenses (present and past). In Dutch (30), the syncretic pattern is correlated with the position of the subject pronoun. When subject pronouns are preposed, 2SG takes the ending -t, identical to that of the 3SG. When the subject pronoun is postposed (as occurs in questions and in subordinate clauses), only 3SG takes -t; thus jij kom-t kom je ‘you’re coming are you coming?’ versus zij kom-t kom-t zij ‘she’s coming is she coming?’ (see Ackema and Neeleman (2003) for a recent discussion of the conditions).15

13

14 15

Note though that the related language Mursi (also a member of the Surmic branch of NiloSaharan) has a similar pattern, in which the 3SG may have a stem alternant distinct from the others (Turton 1981: 344). The suffix -un is found optionally in the perfect. This pattern has some exceptions, as there is one verb which displays a stem alternation which disambiguates the 2SG and 3SG (jij heb-t ‘you have’ versus zij heef-t ‘she has’), and another verb where 2SG has -t even though 3SG does not ( jij ben-t ‘you are’ versus zij is ‘she is’).


66

(29) Nobiin present past (Werner 1987: 147–9, 153–4)


(30)

indicative

interrogative

-ir -is -nam -onam -i -o -ir -is -rokom -sokom -inna -sa

-re -i -o -i -o -ro -so -ro -so -inna -sa

!


Dutch ‘come’ verb + pronoun

pronoun + verb

-Ø -Ø -t -en -en -en

-Ø -t -t -en -en -en

!

Although there are examples where the third person form prevails, there are even more which favour second person. For example, in Callahuaya, a Quechua-based language, the original second person marker -nki is sometimes found with third person; contrariwise, third person -n is not used for second person (Muysken 1997: 437–8). In the Bantu language Kongo (31), indicative and subjunctive prefixes are distinguished in the 3SG only, allowing for differences in syllabicity.16 The syncretic 3SG prefix o- of the indicative can be identified with the distinct 2SG w- of the subjunctive. In Old Icelandic (32), the syncretic 2SG/3SG ending -er of the present (31) Kongo (Carter and Makoondekwa 1979: 6–11, 19–21)

subjunctive 1SG 2SG 3SG 1PL 2PL 3PL

16

ywktwnwb-

(32) Old Icelandic weak verb (Noreen 1923: 353–4) subjunctive, preterite

indicative

!

iootunube-


-a -er -e -em -eþ -e

present indicative

!

-a -ar -ar -om -eþ -a

The alternation between syllabic indicative prefixes and asyllabic subjunctive prefixes is phonologically regular: indicative prefixes attach to the (consonant-initial) verb stem, while the subjunctive prefixes precede the subjunctive marker a-, which is prefixed to the verb stem.


67

indicative is the same as the distinct 2SG ending found in the other tensemood paradigms. (The indicative subjunctive contrast is marked by a vowel alternation in the ending for all but 1SG and 2PL.) The Papuan language Dani (Trans-New Guinea phylum) (33) shows a syncretic 2PL/3 ending -ep in the hypothetical mood which matches the distinct 2PL ending -ip of the past. (The hypothetical mood is also characterized by a lowering of the vowel of the ending.) In Carib (Kalihna) (34), the interrogative form of the copula distinguishes 2PL and 3PL, while elsewhere they are syncretic, displaying the form of the 2PL. The Papuan language Suena (Trans-New Guinea phylum) presents a particularly striking example (35): the syncretic 2DU/3DU and 2PL/3PL of the remote tense have the same element -w- that is found in the 2DU and 2PL endings found in (33)


Dani (Bromley 1981: 192) past

hypothetical

-i -in -e -u -ip -a

-e -en -ep -o -ep -ep

!

(34)


Carib copula (Hoff 1968: 212) interrogative

present

waN maN naN kı¨ ta:toN mandoN nandoN

wa ma:na maN, na kı¨ ta:toN mandoN mandoN

!

the other tenses. (The forms in the first column are used with the future, present, today’s past, yesterday’s past and past tenses. The symbol -Vdenotes the variable mood marker.) (35) Suena (Wilson 1974: 59)

1SG 2SG 3SG 1INCL DU 1EXCL DU 2DU 3DU 1INCL PL 1EXCL PL 2PL 3PL

default

remote

-n-V -s-V -i-V -n-V-ge -n-V-to -w-V-to -r-V-to -n-V-kai -n-V-kare -w-V -r-V

-n-V -s-V -nu-V -n-V-ge -n-V-to -w-V-to -w-V-to -n-V-kai -n-V-kare -w-V -w-V

! !

68


First person / second person In the light of the failure of 1/3 and 2/3 syncretism to reflect any consistent morphological hierarchy between these values, we should not be surprised to find the evidence of 1/2 syncretism to be equally inconclusive. In some examples the second person form prevails. Thus, in the non-PamaNyungan Australian language Burarra (36), 1/2 augmented (plural) and unit-augmented (dual) are marked by the same prefix nyi- which serves for second person in the singular. (36) Burarra (Glasgow 1984, cited in Cysouw 2003)

1INCL 1(EXCL) 2 3

minimal

augmented

unit augmented

arrngunyi(a-)

ngu-burrnyi-burrnyi-burra-burr-

a-rrinyi-rrinyi-rri(a)birri-

!

In the Tungusic language Udihe (37), first and second person are syncretic in both the singular and the plural in various paradigms, and the form corresponds to the distinct second person form as found in other paradigms.17 (37) Udihe (Nikolaeva and Tolskaya 2001: 212–13)

1SG 2SG 3SG 1INCL PL 1PL 2PL 3PL

17

past, past participle

present

permissive, subjunctive

-mi -i -ni -fi -mu -u -ti

-mi -i -ini, -ili -fi -u -u -iti, -du-

-mi -i Ø -fi -u -u -du-

!

!

future, converbs, present and future perfect, conditional participles -i -i Ø -ti -u -u -du-

-i -i -ni -fi -u -u -ti

Etymologically, the second person had the form -sV, with the -s- lenited to -h- in some varieties of Udihe (Sunik 1997: 238) and to Ø in others.


69

In the Omotic language Shinassha (38), the 1PL subjunctive appears to be based on the 2PL form, in both the prefix conjugation and the suffix conjugation.

(38)

Shinassha (Lamberti 1993: 149–53, 163–5) (prefix conjugation suffix conjugation)

1SG 2SG 3SG M 3SG F 1PL 2PL 3PL

default

subjunctive

tı` - -e` nı´ - -ı´ bı´ - -e´ bı` - -a` no`- -o` ı´ t- -«t bo´- -noó´

ni- -ee ni- -ii ni- -ee bi- -aane it- -«te it- -«te bo- -noo

!

On the other hand, there are languages where first person form seems to prevail over second. Thus, Nobiin shows this pattern in the plural interrogative; in (39) we reproduce the paradigm shown above in (29); note that there, the second person is involved in a different directional effect in the singular. In literary Kannada (40), 1SG is distinguished from 2SG by the addition of the element -nu in the future. In the past, the -nu element is extended to 2SG.

(40) (39) Nobiin


(present past) indicative

interrogative

-ir -is -nam -onam -i -o -ir -is -rokom -sokom -inna -sa

-re -i -o -i -o -ro -so -ro -so -inna -sa

!

Literary Kannada (Sridhar 1989: 221–2)

future 1SG 2SG 3SG M 3SG F 3SG N

-enu -e -anu -alu _ -uu

past

!

-enu -enu -anu -alu _ -itu

70


First / second / third person In the future tense in Gujarati (41), 2SG has no distinct form: it is either identical to the 1SG or to the third person, the two forms being in free variation (Cardona 1964: 142). Note that Dutch displays a similar alternation, at least superficially (see above, 30); we do not treat this example as parallel to Gujarati because the 1SG/2SG pattern is transparently a default form. (41)

Gujarati future (Cardona and Suthar 2003: 682, 684) option 1

1SG 2SG 3 1PL 2PL

-isˇ -isˇ -sˇ e -sˇ (i)u˜ -sˇ o

option 2

!

-isˇ -sˇ e -sˇ e -sˇ (i)u˜ -sˇ o

3.2.4 Diachrony Some compelling instances of directionality can be found in diachrony, where the expected form for one value is replaced by the form for another value. A familiar example, adduced in KuryØowicz’s famous article on analogy (1949 [1960]), involves Old Icelandic, cited above in (32). These paradigms are interpreted as the result of the extension of the 2SG ending to the 3SG in place of the expected *-þ (Haugen 1982: 129). The syncretic pattern itself is attributed to analogy with verbs in -l and -n, where it was the result of a general phonological development (syncope of the theme vowel and assimilation of the ending into the stem-final consonant; KuryØowicz 1949 [1960: 81]). We can ask the same questions about this diachronic scenario as we have asked above about synchronic patterns: (i) where does the syncretic pattern come from?, and (ii) why does one form prevail over the other?18 Let us look at a few examples with these two questions in mind.

18

Of course, we can also ask ‘Why was the syncretic pattern extended?’ This is not a question of syncretism per se, so much as one of a general theory of analogy. In all of the examples known to us, an obvious functional explanation (such as that offered for the genitive/ accusative in Slavonic; see x3.1.2) is lacking. Since avoidance of homophony is invoked at least as often as extension of homophony in order to account for unusual morphological


71

Some examples involve substituting a third person form for the original form, which would follow from the assumption that third person serves as a default of some sort. For example, in Livonian (42), the 3SG present tense ending -b is found in place of the expected 1SG ending *-n ! -Ø (Viitso 1998: 112). Compare the paradigm from the closely related Estonian, where the original 1SG ending is found.19 Note that in monosyllabic stems in West Livonian dialects, vacillation was recorded between the original 1SG form and the innovative one, thus the verb ‘to be’ has the singular forms uo or uob ‘1SG’, uod ‘2SG’, uob ‘3SG’ (Kettunen 1938: lx). Kettunen (1938: lx–lxii) attributes this to analogy with the preterite paradigm, where 1SG and 3SG fell together as the result of regular sound change. Thus, as with Old Icelandic, the syncretic pattern was already established in the language by regular sound change. (Note that the 2PL/3PL pattern of the preterite was not extended.) (42) present


preterite

Livonian

Estonian

Livonian

Estonian

lugub lugud lugub lu’gg«m lu’gg«t lu’gg«b«d

loen loed loeb loeme loete loevad

lugiz lugist lugiz lugizm« lugist(«) lugist(«)

lugesin lugesid luges lugesime lugesite lugesid

German likewise shows what appears to be the extension of third person forms. In Middle High German (MHG), all persons were distinct in the present tense plural; by the Early New High German (ENHG) period, four different syncretic patterns were found, varying according to dialect (43). In type I, the third person form has been extended to second person. In type II, it has been extended to all three persons. In type III (ultimately established as the literary norm), 1PL and 3PL are syncretic. Though the form looks like the original 1PL, the source is more likely to have been the -n found in the 3PL preterite. In IV, this -n is extended to all plural persons.

19

innovations (cf. Bentley and Eytho´rsson 2001), we prefer not to address this question. Note that KuryØowicz (1949) does not address it either. Although this example appears in an article on analogy, it is used to illustrate quite a different point; the analogical extension of the syncretic pattern is itself taken for granted. Though note that the expected reflex of final -n should be -Ø in Estonian as well. Its retention in 1SG forms is anomalous (Anttila 1972: 79).

72


Present tense endings in Middle and Early New High German (Wegera 2000: 1546; Grosse 2000: 1333) ENHG

1PL 2PL 3PL

MHG

I

II

III

IV

-n -t -nt

-n -nt -nt

-nt -nt -nt

-n -t -n

-n -n -n

The third person also prevails in the Kumta dialect of Kannada (44), where, in place of the historically expected 1PL ending -vV (found in other dialects, as well as the literary language), the 3PL ending -ru is found. As in the German examples above (43), this same ending may replace the 2PL ending -ri as well.

(44) Kannada ‘kept’ (Upadhyaya 1976: 130–2)

1PL 2PL 3PL

Bellary

Kumta

ittive it_t_ri it_t_ru (M/F) it_t_uvu (N) __

ittru it_t_ru ittri __ it_t_ru __

A similar development has been reconstructed for the plural passive of Gothic: 1PL and 3PL fell together by regular sound change, and the 2PL was analogically altered to match them, leading to the attested 1PL/2PL/3PL ending -anda (Szemereńyi 1989: 255). A particularly striking example of the extension of third person forms comes from the Oceanic language Anejom (45). In the nineteenth century, the auxiliary had distinct forms for first person (inclusive and exclusive), second and third in the dual, trial and plural. Between the nineteenth and twentieth centuries the language underwent catastrophic change (due in part to population loss). The result is a system with considerable variation, if not to say confusion. Lynch (2000) hypothesizes that it is moving in the direction of generalizing the third person plural for all non-singular persons.


73

Anejom auxiliary (Lynch 2000: 91–5) (aorist, typical for other TAM paradigms) 19th century

1SG 2SG 3SG 1INCL DU 1DU 2DU 3DU 1INCL TRI 1TRI 2TRI 3TRI 1INCL PL 1PL 2PL 3PL

ek na et intau ecrau ekau erau intaj ektaj, ektij ahtaj ehtaj inta ecra eka era

20th century

>

ek na et

era

!

However, not all such examples favour the third person. Thus, Romani shows evidence of multiple directionality (46). In the present tense, syncretism of 2PL/3PL is common to all the dialects. (The origin of this pattern is unclear, so we take it as a given.)20 What is of interest to us is the extension of this pattern into the perfect, which occurs in a few dialects. Significantly, it is sometimes the 2PL ending which prevails (Sinti), and sometimes the 3PL ending (Northeastern Romani, and partly in Welsh Romani). (46)

Romani (Matras 2002: 145) perfect

1PL 2PL 3PL

20

present

reconstructed

Central, Finnish, Balkan, Vlax

Sinti

Welsh

Northeast

-as -en -en

*-am *-an *-e

-am -an, -en -e

-am -an -an

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

-am -e, -an -e ................

-am -e -e

The syncretic 2PL/3PL corresponds to what would be expected for the 3PL. This is unexplained, but possibly the resemblance between the 2PL perfect ending -an and the 3PL present ending -en had something to do with it.

74


Thus, while there is some diachronic evidence that third person forms can replace others diachronically, it is hardly an exceptionless generalisation. Especially striking are the examples that appear to show second person replacing third, for which no explanation readily comes. Nevertheless, there is one area where the default status of third person is more firmly established, namely the wholesale loss of person marking, as seen above in German, the Kumta dialect of Kannada and in Gothic; in all these examples it is the etymological third person form which prevails. A seeming counterexample to this observation comes from the present tense in Modern East Scandinavian languages (Swedish, Danish and Norwegian), where person distinctions have been eliminated, and there is but a single form in -r, which, as discussed above, was originally second person. However, it was only after it was extended to third person singular that it ultimately predominated (47); that is, the form which was extended to all person values served as a third person form (among other things) immediately prior to its extension.

(47) Eastern Scandinavian (present indicative)

1SG 2SG 3SG

proto-ES

>

Old ES

*-(V) *-(V)r *-(V)þ

!

-(V) -(V)r -(V)r

>

!

Modern ES -(V)r -(V)r -(V)r

This suggests that two types of operation were at work: (i) an idiosyncratic, morphologically specified extension of the second person to the third, and (ii) a cross-linguistically unexceptional extension of the 3SG form. In the examples above, a form originally associated with one value was extended to other values. Such developments, where incomplete, may account for some of the instances of synchronic directionality effects that we have seen. However, in some cases we may be seeing the result of the reverse development, as suggested by the historical development from Old Nubian (first attested in the seventh century; Browne 2002: 1) to its direct descendant, Nobiin, examples from which are cited above ((29) and (39)). Old Nubian had regular 2SG/3SG and 1PL/2PL syncretism in all its verbal paradigms (48); the origins of this pattern are unknown. In Nobiin, 2SG and 2PL suffixes were innovated and added to the indicative and negative paradigms, but not to the interrogative or conditional. Thus, the syncretic


1 2 3

75

Old Nubian present indicative (Bechhaus-Gerst 1996: 237) SG

PL

-ire -ina, -ena -ina, -ena

-iro -iro -iran, -eran

paradigms represent the older state of affairs, and the directional effect that results is a reflection of diachronic layering. 3.2.5 Summary Seen in terms of the syncretic values, some common patterns emerge, which are most easily expressed negatively: syncretism of first person and third is rare, and syncretism which is confined to the singular is rare. Directional effects, from which we might deduce hierarchical relationships between feature values, do not lend themselves to any generalization, except for the observation that, diachronically, when person distinctions are effaced completely, the form which prevails is likely to have originated as the third person form.

3.3

Person syncretism in two-place verbs

3.3.1 Introduction Verbs which mark both subject and object (two-place verbs) are a case apart from verbs which mark only a single argument (one-place verbs). The interaction of subject and object person appears to encourage person syncretism, especially that of the subject. Significantly, syncretism in two-place verbs is independent of syncretism in one-place verbs. Thus, in the languages of the World Atlas of Language Structures sample, illustrated in Appendix 3, twenty languages evince person syncretism in two-place verbs (out of sixty-nine languages which mark both subject and object person), and of these, thirteen have person syncretism only in two place verbs. Note that in the remaining seven, the syncretic pattern in two-place verbs is distinct from the one found in one-place verbs; instances where a syncretism found in one-place verbs is simply carried over to two-place verbs have not been recorded. Below we distinguish three patterns, arranged according to their frequency: syncretism of subject person, syncretism of object person, and other patterns.


76

3.3.2 Syncretism of subject person Syncretism of subject person is the most common pattern affecting twoplace verbs, found in thirteen out of the twenty languages in the World Atlas of Language Structures sample. In many instances first or second person objects are involved. It is known that first and second person objects trigger exceptional morphological marking (see Heath 1991, 1998, Kibrik 1997, Noyer 1997). Kibrik (1997) accounts for this in terms of the animacy hierarchy and the notion of prototypical agents and patients: arguments of high animacy or prominence (such as first and second person) are prototypical agents, and arguments of low animacy or prominence (such as third person) are prototypical patients. When the former are in the patient role, this may lead to some special morphological treatment, the most familiar example being the inverse marking of Algonquian languages. Syncretism in this context implies the complete neutralization of subject–person distinctions, since only two subject-person values are possible (second or third person with a first person object, and first or third person with a second person object). Various means are available for neutralizing subject person. Subject person not marked In some cases it is evident that subject marking is simply absent. For example, in Guarani (49), all values of subject person are syncretic when the object is first person. Note that the marker which is used is identical to the so-called ‘personal reference’ form for first person. (The personal reference form is used for the subject of stative verbs and for possessors.) Thus, in the context of a first person object, subject person is not marked at all.

subject

(49) Person prefixes in Guarani (Gregores and Sua´rez 1967: 131–2)

1SG 1PL 2SG 2PL 3

object

‘personal reference’

1SG

1PL

2SG

2PL

Ø or 3

sˇ eorenepenei-

— — sˇ esˇ esˇ e-

— — oreoreore-

roro— — ne-

popo— — pene-

arorepeo-


77

Nevertheless, these forms still seem to be sensitive to the presence of subject person, if not its person value, in as much as we find the ‘personal reference’ prefixes sˇe- ‘1SG’ and ore- ‘1PL’, and not simply the intransitive subject prefixes a- ‘1SG’ and ro- ‘1PL’. In other languages we may find a system which allows for distinctions in the value of subject person even where the subject is not overtly marked, as in the Chukotko-Kamachatkan language Koryak. In (50) and (51) the person-number markers of Koryak are shown. (Dual is shown in preference to plural because it is morphologically basic in Koryak; their behaviour is the same with respect to subject and object marking.) First consider the suffixes shown in (50). Where the object is 1DU or second person, subject person is not distinguished; where the object is third person, first person, 2SG and 3DU subject person are not distinguished from each other. We can interpret these forms as marking the object person only, unspecified for subject. (50) Koryak verbal suffixes (Zˇukova 1972: 232, 252–4)

subject

object

1SG 1DU 2SG 2DU 3SG 3DU

Ø

1SG

1DU

2SG

2DU

3SG

3DU

-k

— — -j -t«k -j -g«m

— — -m«k -m«k -m«k -m«k

-gi -gi — — -gi -gi

-t«k -t«k — — -t«k -t«k

-n -n -n -tk« -nin -n

-net -net -net -tk« -nin -net

-j -t«k -j -g«hi

Now consider the distribution of the prefix ne- in (51). (51) Koryak verbal prefixes (Zˇukova 1972: 232, 252–4)

subject

object

1SG 1DU 2SG 2DU 3SG 3DU

Ø

1SG

1DU

2SG

2DU

3SG

3DU

t«m«t-

— — ineineinene-

— — nenene ne-

t«m«t— — nene-

t«m«t— — nene-

t«m«t-

t«m«t-

ne-

ne-

78


Except where the object is 1DU, it serves to distinguish subject person, though it is clear from its distribution that it does not overtly mark subject person as such. Following Comrie (1980), we can interpret them as a kind of inverse marking, indicating that the relationship of subject to object runs counter to the person-number hierarchy 1 > 2 > 3SG > 3DU/PL. The prefix ne- is found when the subject is lower on the hierarchy (i.e. to the right) than the object, or when both are at the bottom of the hierarchy. The forms with the prefix ine- likewise lack overt marking of the subject person but have a different structure from those in ne-. Morphologically, the ine- forms are intransitives (Zˇukova 1972: 254–5), apparently based on the antipassive paradigm (Spencer 2000: 210). Nevertheless, because these forms are used only with a 1SG subject, subject person is unambiguously implied. Extension of subject markers In some cases the range of the overt subject marker is quite extensive. The Gunwinyguan (Australian, non-Pama-Nyungan) language Dalabon employs this approach when singular subjects are combined with a first person object. In that case the third person transitive subject form kahis used for both second and third person subjects (52). Note further that kahis the form used when the object is lower in animacy than the subject. We discuss this in our case study of Dalabon in Chapter 5: x5.3. (Where the object is second person, only the object person is marked, as discussed above.) (52)

Dalabon singular transitive verbal prefixes (Evans, Brown and Corbett 2001: 199)

subject

object

1SG 2SG 3SG

1SG

2SG

3SG

— kahkah-

djah— djah-

ngahdahkah- (bvkah-)

The Tibeto-Burman language Trung likewise employs substitution of subject person when the object is first person (53). In this case, though, the directionality is reversed: the second person marker -n« is used for third person subjects as well.


79

Trung verbal markers (Watters 2003: 396)

subject

object

1 2 3

1

2

3 or ø

— n«- -N n«- -N

-N — n«-

-N n«Ø

3.3.3 Syncretism of object person Syncretism of the object person is less common than syncretism of subject person. In the World Atlas of Language Structures sample it is found in seven out of the twenty languages displaying syncretism in two-place verbs. In contrast with subject-person syncretism, the interaction of subject and object person does not necessarily play a role in syncretism of object person. This is due to the existence of systems where both arguments are overtly marked, with distinct subject and object forms. Thus transitive verbs provide an environment for a distinct series of person-marking forms, which in turn may be syncretic. For example, in Barbarenõ Chumash (54), an extinct language of California, subject markers are prefixed and object markers are suffixed, and each series is distinct. Subject prefixes distinguish three persons (and three numbers), while object suffixes have syncretism of first and second person (as well as dual and plural number). Thus, syncretism of the object marker is inherent in its form and not conditioned by the value of subject person. (54)

1SG 2SG 3SG 1DU 2DU 3DU 1PL 2PL 3PL

Subject and object markers in Barbarenõ Chumash (Beeler 1976: 255) subject

object

kpskisˇ pisˇ sisˇ kiypiysˇ iy-

-it -in -us (or Ø) -iyuw -iyuw -wun -iyuw -iyuw -wun

80


Otherwise, the examples are not numerous enough to make any meaningful generalizations.

3.3.4 Other patterns Other patterns may be found besides the simple syncretism of subject or object person, either by transposing subject and object roles, or by combining incompatible subject and object values. Mojave illustrates both possibilities (55). First, the combination of second person subject and third person object is not distinguished from a third person subject and second person object, both being marked with m-. Given that m- otherwise functions as a second person prefix, it is clear that where both second and third person are involved, only the second person argument is marked. Subject or object status plays no role. With a first person argument, the situation is more complex. The prefix ny- marks any first person argument involved in a transitive construction, except where the object is third person. Where the subject is second person, this too is marked, yielding the sequence ny-m-. As a result, the prefix ny- used by itself marks the values ‘first person > second person’ and ‘third person > first person’. Thus, although both the functions of ny- involve first person, in one it acts as the subject, in the other as the object.

(55)

Verbal prefixes in Mojave (Munro 1976: 12–13)

subject

object

1 2 3

1

2

Ø or 3

— ny-mny-

ny— m-

?mØ

The Nilo-Saharan language Turkana goes even one step further in combining incompatible values (56). The prefix kI- marks (i) a first plural argument with any other argument, (ii) a second person subject with a first person object, and (ii) a third person subject with a second person object. In the case of Mojave, ny- was identifiable as some kind of a first person marker. Here, the prefix kI- is not uniquely associated with any one person


81

Transitive verbal prefixes in Turkana (Dimmendaal 1983: 282)

subject

object

1SG 1PL 2 3

1SG

1PL

2

3

— — kIka-

— — kIkI-

kakI— kI-

akIIe-

value. At best it can be defined disjunctively: it involves first and/or second person.21 3.3.5 Summary Most examples of syncretism in two-place verbs involve subject-person distinctions when the object is first or second person, a phenomenon which some have seen as involving the animacy hierarchy: high animacy objects disfavour the explicit marking of the subject. To a much more limited extent we find syncretism of object person, without any obvious connection to the animacy hierarchy. Other syncretic patterns may involve underspecification for the distinction between subject and object, or remain difficult to characterize. 3.4

Gender

3.4.1 Introduction Gender categories differ greatly across languages. At the core of all systems is some distinction of sex or animacy, but beyond that, the principles 21

In fact, Dimmendaal (1983) treats it as a case of accidental homophony, attributing to it three separate underlying morphological representations, predicated on the following assumptions: (a) where neither argument is third person, the subject is marked in preference to the object; (b) first and second person are marked in preference to third; (c) k- is a sort of inverse marker (not termed so by Dimmendaal, though), used whenever first or second person are objects: kI! kI1 1PL (1PL > 3) k + kI! kI2 1PL ‘inverse’ (1PL object or 1PL > 2) k + I! kI3 second person ‘inverse’ (2 > 1SG object or 3 > 2 object) However, this analysis crucially depends on the existence of the distinct inverse prefix k-, for which there is no independent evidence, so the analysis remains speculative. Heine (1981: 116) likewise treats a similar pattern in the related language Camus as a result of accidental homophony.

82


by which nouns are assigned to gender may vary wildly, determined by language-specific semantic and morphological principles. Nevertheless, allowing for a certain degree of abstraction, we can make some valid cross-linguistic observations. We will look at gender syncretism in two domains: across numbers, and across different targets.22 3.4.2 Gender and number Gender shows a special affinity for syncretism across number. It is probably no exaggeration to say that, where gender marking is morphologically distinct between singular and non-singular, gender syncretism is more the rule rather than the exception.23 Corbett (1991: 155–6) describes two patterns associated with gender syncretism across number, convergent and crossed, based on the Africanist practice of drawing lines to connect singular gender forms with plural gender forms. In a convergent system, there is a many-to-one relationship between singular and plural forms. The Dagestanian language Karata (57a) displays this configuration, where masculine and feminine have distinct forms in the singular but are associated with a single form in the plural. In this case, there is syncretism in the plural. A crossed system is one where there is a many-to-many relationship between singular and plural forms (Heine 1982: 197). Thus, in Albanian (57) a. Karata ‘that.ABS’ (Magometbekova 1967: 329)

M F N

SG

PL

hadiw hadij hadib

hadibaj

b. Albanian ‘this.NOM’ (Newmark 1982: 263) SG M F

ky kjo

PL

N

ke¨ta ke¨to

hadiraj

(57b) there are two distinct forms in each number; the masculine singular form may be associated with either of the plural forms, while the feminine plural form may be associated with either of the singular forms. This is the result of the existence of a large and productive class of nouns which seem to switch gender across number, taking masculine forms in the singular and the feminine forms in the plural (Newmark 1982: 133). Although the 22

23

In as much as the present book concerns syncretism in inflection, we look only at gender agreement marking, and not the overt marking of gender on controller nouns. And even where gender marking is not morphologically distinct, as in the Andi dialects discussed below.


83

Albanian grammatical tradition does not sanction this label, by distributional criteria this class justifies being seen as a distinct gender, which we can call neuter, which is syncretic with the masculine in the singular and with the feminine in the plural (as is done with Romanian, which has essentially the same system; see (62) below). These schemata are useful to keep in mind, especially as there is a large body of literature on Niger-Congo languages (particularly rich in gender, and gender syncretism) in which gender marking is presented precisely in this way. However, for the remainder of this section we will present such paradigms in the terms outlined in Chapter 2: x2.1.1, where masculine and feminine plural in Karata are an example of simple syncretism (58a), and the neuter in Albanian is an example of a non-autonomous value (58b).

(58) a. Karata (= 57a)

M F N

Albanian (= 57b)

SG

PL

hadiw hadij hadib

hadibaj hadibaj hadiraj

M F N

SG

PL

ky kjo ky

ke¨ta ke¨to ke¨to

3.4.2.1 Smaller systems In order to isolate cross-linguistically valid generalizations, it will be useful if we first look at simple systems with only three genders, as in the examples above. The first observation we can make is a paraphrase of Greenberg’s familiar universal 37 (‘A language never has more gender categories in non-singular numbers than in singular’; 1963: 112): convergence typically goes in the direction of singular to plural, and not the reverse, with the result that we may find more distinct gender forms in the singular than the plural, but not more forms in the plural than the singular. As we shall see below (in Kisi and Ju/ 0 hoan), there are convincing counterexamples to this claim, but it does describe a strong tendency. For example, while complete neutralization of gender distinctions in the plural is not uncommon (e.g. Russian or German, with three gender distinctions in the singular and none in the plural), the reverse is all but unknown (the Nilo-Saharan language Fur, with a distinction of human versus non-human gender in the plural but not the singular, is one of the rare examples; Jakobi 1990: 102–3).

84


Gender convergence in the plural itself seems to be under certain constraints. Consider the examples in (59). Both are languages with distinct masculine, feminine and neuter forms in the singular, and with syncretism of two genders in the plural. In Laal (59a), an unclassified language of Chad, masculine and feminine plural are regularly syncretic, illustrated here with the possessive pronoun. In the Iroquoian language Seneca (59b), feminine and neuter dual and plural are regularly syncretic.24 Both these are common patterns; less frequently found are patterns such as that found in the Nilo-Saharan language Turkana (60), where gender convergence results in syncretism of the masculine and neuter in the plural. (59) a. Laal possessive pronoun (Boyeldieu 1982: 14)

M F N

singular

plural

da:r do`:g da`:na´

de`:rı´ de`:rı´ dua`:na´

(60)

b. Seneca (Chafe 1997: 563) singular M F N

dual plural

hahni- hati-, he˛nyekni- wati-, we˛nka-, w-, y- kni- wati-, we˛n-

Turkana (Dimmendaal 1983: 217) (restrictive non-restrictive)25 singular plural

M F N

e- loa- nai- ni-

Ni- luNa- naNi- lu-

As further examples below will illustrate, the plural appears to accord a privileged status to such semantic categories such as male personal, masculine, human or animate. In this respect, masculine and feminine form a natural grouping (‘human’) as do feminine and neuter (as the remainder after masculine has been isolated), but masculine and neuter do not. Of course, different languages have different means for assigning gender, and 24

25

Where multiple forms are listed for Seneca, these are allomorphs conditioned by the initial segment of the following stem. Restrictiveness is a notion allied with, but not exactly equivalent to, definiteness (Dimmendaal 1983: 218).


85

these are not always semantic but may be based on phonology, morphology, or outright lexical stipulation for some items. Nevertheless, there is typically a semantic core which justifies gender designations such as ‘masculine’ or ‘human’, and the generalization above still applies to these core genders, regardless of the contribution of non-semantic principles to gender assignment. The two patterns represented by Laal and Seneca may themselves be combined with any pattern in the singular. For example, the Dravidian language Telugu (61a) behaves like Laal in the plural, with syncretism of masculine and feminine, but it also has regular syncretism of feminine and neuter in the singular. The system found in the Kru (Niger-Congo) language Grebo (61b) can be seen as the mirror-image. It has three genders: one for humans, one for large, valuable things (‘non-human 1’), and one for all other things (‘non-human 2’). There is clearly a scale of prominence, with humans as the most prominent and the second non-human gender as the least. In the plural, the two less prominent genders are syncretic, while in the singular the two most prominent genders are syncretic. (61)

a. Telugu copular suffix (Krishnamurti and Gwynn 1985: 138)

M F N

singular

plural

-d:u -di -di

-ru -ru -y(i)

b. Grebo third person pronoun (Innes 1966: 52–3)

HUM NHUM‘1’ NHUM

‘2’

singular

plural

O O e

o e e

Most significantly, masculine/neuter syncretism may occur in the singular. Romanian, illustrated in (62), is a familiar example, where it is combined with feminine/neuter syncretism in the plural (this pattern is consistent on all targets). Romanian can serve as a good example of the contrast between (62)

M F N

Romanian third person pronoun, nominative case (Rothe 1957: 78) singular

plural

el ea el

ei ele ele

86


gender syncretism in the singular and the plural, if we look briefly at its history (the following account is based on Windisch 1973). Roughly speaking, the gender affiliation of nouns in contemporary Romanian continues that of their Latin forebears, masculine, feminine and neuter. There is no good evidence that the syncretic patterns as such had any but a phonological or morphological basis. In Latin, masculine and neuter were only weakly distinguished from each other in the first place, while in the plural, phonological merger has been proposed, though there is no consensus (Windisch 1973: 157–69). Thus, we can take as a point of departure a system with two semantically arbitrary patterns of gender syncretism, one in the singular and one in the plural. However, over time there has been a continual migration of inanimate nouns from the masculine to the neuter, so that the masculine gender has become increasingly restricted to male animates (Windisch 1973: 117, 196). Because masculine and neuter are syncretic in the singular, morphologically, this has consequences only in the plural: the switch from masculine to neuter involves using the ‘feminine’ agreement form rather than the masculine. As a result, the distribution of forms in the plural has a much greater degree of semantic predictability than in the singular. A similar system has also developed in Albanian (see (58b) above), and was also characteristic of Tokharian (Windisch 1973: 18–20). 3.4.2.2 Larger systems When we move to systems with a larger inventory of genders, we largely see the same themes repeated, namely the consolidation and realignment in the plural of gender distinctions into groups with a more-or-less transparent semantic justification. A particularly clear illustration of the ‘semanticization’ of gender forms in the plural comes from dialects of the Dagestanian language Andi, discussed by Corbett (1991: 198–9). The dialects termed types A, B and C all display four distinct gender forms in the singular and plural. The conservative type A dialect has the simplest system (63a), distinguishing human males (gender I, with the marker w), human females (gender II, with the marker j), most non-human animates (gender III, with the marker b) and the remainder (gender IV, with the marker r). These gender markers are the same for singular and plural. In the more innovative type B dialect (63b), animate members of gender III pattern with female nouns in the plural, taking the marker j. In the type C dialect (63c), animate members of gender IV do the same. As a result, in the type C dialect, even though the repertory of gender markers is the same in


87

the singular and plural, there is a greater degree of semantic predictability in the plural, since all non-male animates take the same marker j, whereas in the singular they are distributed across the three markers j, b and r.

(63) Andi (Xajdakov 1980, cited in Corbett 1991: 198–9) a. Type A

I II III IV

b. Type B

SG

PL

w j b r

w j b r

I II III AN III INAN IV

c. Type C SG

PL

w j b b r

w j j b r

I II III AN III INAN IV AN IV INAN

SG

PL

w j b b r r

w j j b j r

The Maasinankore dialect of the Atlantic (Niger-Congo) language Fula, described by Breedveld (1995), represents a particularly extreme example of the consolidation of gender forms in the plural. The basic system is given in (64); minor genders consisting of one member are omitted. There are twelve distinct gender markers in the singular, but only four in the plural:

(64)

Fula

singular 'O 'O

(Breedveld 1995: 457) plural BE

NGE

DI

NDU

NGAL

'DI 'DI 'DE 'DE 'DE 'DE 'DE 'DE 'DE

NGEL

KOY

NGOL NDI NDE NGO BA KI KA

human

'DI largely animate

inanimate

augmentatives

88


most genders take the markers 0 DI or 0 DE in the plural. The paradigm in (64) in fact represents an oversimplification. All of the singular gender markers which are associated with 0 DE may take 0 DI as well (with the exception of the augmentative gender NGEL). 0 DE is used strictly for inanimates (though including ‘nasty animals that bite or sting’; p. 446). 0 DI is used for animates, but also appears to function as the default, in that even inanimate nouns which used to take 0 DI are now showing a tendency to use it. To the extent that the reorganization of gender groupings in the plural involves such consolidation, the predictions of Greenberg’s universal 37 are maintained. However, reorganization in the plural can also involve massive splitting, to the extent that there are more distinct gender forms in the plural than in the singular. This is the case with the Atlantic (NigerCongo) language Kisi, which we will look at in some detail, based on Childs (1995: 162–70). It has seven genders, realized by three forms in the singular and five forms in the plural. The approximate semantic composition of the genders is given in (65), listed in order of the number of nouns they contain (out of a total sample of 910). (65) Kisi (Childs 1995: 162–8) SG

PL

semantics

o

la

o i

a N

le le

i la

le

ma

o

i

no semantic core; default class for borrowed inanimates not denoting liquids virtually all animates inanimate, but little semantic cohesiveness (possible semantic core: small and round objects) long and thin, string-like objects inanimates; productive for deverbal or denominal abstract nouns 1. liquids (productive for borrowings) 2. pointed objects trees and tree-like plants

43.4% 27.3%

15.4% 4.3% 3.2% 3.1% 3%

The i/N gender is a semantically diffuse class of inanimates, which is distinct from other genders in both the singular and plural. Of the remaining genders, the plural forms lend themselves much more readily to semantic characterization than the singular ones, with three of the forms having a relatively clear semantic basis, and the remaining one functioning as a default:

Cross-linguistic typology of features * * * *

89

a is for animates i is for long objects ma is for (i) liquids and juicy plants, and (ii) sharp, pointed objects la is the default for inanimates (except those belonging to the i/N gender)

Neither of the singular forms can be so characterized. The marker le is used exclusively for inanimate genders but otherwise can be paired with any of the plural markers, while o can only be defined negatively, being used for anything but long, thin, string-like objects. The most salient distinction between the two appears rather to be morphological, that between a productive and unproductive marker. The o class is large (74 per cent of nouns) and is productive for borrowings except for words denoting liquids, while the le is small (11 per cent) and is unproductive except for nouns denoting liquids. In Kisi, the syncretic gender forms in the singular mark agreement with a heterogeneous group of nouns, while in the syncretic forms of the plural there is a strong semantic component to the grouping. It thus conforms to the generalization that plural forms show a more semantically coherent grouping of gender than the singular ones. However, even this is not always the case, as demonstrated by the Khoisan language Ju/ 0 hoan, which violates both this generalization and Greenberg’s universal 37. Ju/ 0 hoan distinguishes five genders. The first three are relatively easy to characterize: I is for humans, II for animals, as well as the names of (non-Ju/ 0 hoan) nations and ethnicities, and III is for (most) plants and foodstuffs. The other two genders are rather harder to characterize on a semantic basis. Both are exclusively inanimate; IV contains many long objects, but otherwise shows ‘little semantic coherence’ (Dickens 1992: 16). Gender V is likewise heterogeneous, being used, among other things, for deverbal nouns, for reference to clauses, and for body parts. Gender is syncretic in both the singular and the plural, whereby free pronouns have more distinct gender forms in the plural than in the singular (four versus three). What is more, the syncretic gender form in the singular seems semantically more coherent than that in the plural. In the singular, genders I, II and III are syncretic, thus embracing all animates, as well as living, or at any rate growing, inanimates. In the plural, genders II and IV are syncretic; this grouping defies any obvious semantic characterization. Gu¨ldemann and Vossen (2000: 112) account for this unusual distribution of forms by positing that the forms ha and hı` perform distinct functions depending on the animacy of the noun. For inanimate nouns they mark

90


I II III IV V

Ju/0 hoan (Dickens 1992: 12–16) pronoun singular

plural

ha ha ha hı` ka´

sı` la hı` ha hı` ka´

gender without reference to number:26 ha for plants and hı` for long objects. For animate nouns they mark number: ha for singular and hı` for plural. (The other two forms, ka´ and sı`la, are monofunctional.) On this view, what we see in Ju/ 0 hoan is not so much gender syncretism as syncretism between the features of gender and number; in either case, it is an unusual system. 3.4.3 Syncretism restricted by target The syncretic patterns described above are restricted to a particular morphosyntactic context, namely to a particular number value. Gender syncretism may also be restricted on the basis of the target; that is, some lexemes display the syncretism and others do not. For example, in Latin, there are some third declension adjectives where masculine and feminine are syncretic; compare the nominative singular forms fortis ‘strong.M/F’ versus acer ‘sharp.M’, acris ‘sharp.F’. Given that distinctions between target types involve language-specific lexical and morphological features, the expectation of finding any valuable cross-linguistic generalizations is slim. We will therefore content ourselves with indicating the variety of patterns that can result. A good illustration (and a particularly complex one) of the sort of variation that can be found comes from the non-Pama-Nyungan Australian language Gaagudju. The language has four genders: I for human males, II for human females, III for plants and IV for remaining inanimates. The basic set of gender-marking prefixes distinguishes all four;

26

Note that, although the pronouns for genders III–V do not distinguish between singular and plural, it is not because number is irrelevant to nouns of these genders; e.g. uto` ~ uto`sı` ‘car ~ cars’ (gender III), n!a`ma` ~ n!a`ma`sı` ‘road ~ roads’ (gender IV) and !aìhn ~ !aìhnsı` ‘trees’ (gender V).


91

it is used for declension 1 adjectives, and a similar set is found in demonstratives, and in most verb paradigms. Otherwise, various patterns of syncretism are involved. Second declension adjectives display syncretism of genders III and IV. Free pronouns, as well as indirect object enclitics, display syncretism of genders II, III and IV. Third person subjects of intransitive verbs in the realis present tense display syncretism of genders II and IV, while in transitive realis verbs with a first person subject, third person objects show syncretism of genders I and IV. While the nominal patterns may be explicable semantically (second declension adjectives show syncretism of the non-human genders, and pronouns oppose male humans to everything else), the verbal patterns are quirkier. Why the patterns should be distributed across the targets in this way remains unclear.

(67) Gaagudju (Harvey 2002: 144, 157, 224–5) realis verbs adjectives

I (male) II (female) III (plant) IV (residue)

first declension

second declension

pronoun

present (intransitive subject)

Ø njiNmagu-

nanjiNnaNnaN-

naawu ngaayu ngaayu ngaayu

Ø nj-djama-yanj-dja-

present and unmarked tense object with first person subject arranji-rrama-raarra-

3.4.4 Summary Cross-linguistically, gender syncretism shows some strong tendencies when considered in the context of number. The typical pattern is the consolidation of genders in the non-singular into syncretic classes with a more-or-less transparent semantic basis (or at any rate, more transparent than in the singular). However, this is only a tendency, and languages such as Ju/ 0 hoan show that this need not be so. Gender syncretism that is restricted to particular targets is, by its nature, language-specific, and even within a language (such as Gaagudju), shows a mixture of semantic and morphological motivation.


92

3.5

Number

3.5.1 Introduction Syncretism within the feature number is a surprisingly straightforward affair, typically dictated by the semantically natural grouping of nonsingular number values (see Corbett 2000: 39–42). Number syncretism is often correlated with the animacy hierarchy (Corbett 2000: 55–66), with items lower on the hierarchy showing a smaller range of non-singular values. For example, in Gothic, dual and plural are distinct only for first and second person, reflected both in free pronouns and in person-number agreement on verbs (68). All other nominals (pronouns, nouns and adjectives) distinguish only singular and non-singular. (See Cysouw 2003 for further cross-linguistic data.) (68) Number values in Gothic (Wright 1930: 91, 120, 135)

SG DU PL

pronoun (nominative) ‘I’ ‘he’

noun (nominative) ‘guest’

verb (‘take’, present) 1 2

ik wit weis

gast gasteis gasteis

nima nimo s nimam

is eis eis

nimiþ nimand nimand

In the Gunwinyguan (Australian, non-Pama-Nyungan) language Bininj Gun-Wok, the collapse of number values is correlated not with the value of person, but with the morphological context. Bound pronouns (illustrated here by intransitive subject markers) distinguish three number values for each person (dual, trial and plural in the case of the first inclusive). Free pronouns distinguish only singular and plural; the distinction between first person exclusive and inclusive is also neutralized.27 The Slovene example cited below in (63) in Chapter 4: x4.4.4 and (9) in Chapter 5: x5.1.1 represents a collapse of dual/plural which is even more deeply embedded in language-specific morphology. Given the regularity with which the semantically natural pattern is manifested cross-linguistically, it is all the more interesting to consider examples which deviate from the expected pattern, either in terms of the 27

This also obtains for part of the bound pronominal system, namely in transitive verbs which mark both subject and object.


93

Person and number in Bininj Gun-Wok (Evans 2003: 263)

1SG 1DU 1PL 1INCL DU 1INCL TRI 1INCL PL 2SG 2DU 2PL 3SG 3DU 3PL

subject (past tense)

free pronoun

nganganingarringarrganigarriyinguningurribabanibarri-

ngaye

ngad

ngudda wudda nungga (M), ngaleng (F) bedda

values which are combined, or in terms of the apparent markedness relationship between number values. 3.5.2 Values If we limit ourselves to languages with singular, dual and plural, we should expect that dual and plural should fall together. The unexpected patterns then involve the dual falling together with the singular, or the singular and plural falling together, leaving the dual distinct. The first pattern is found in the distal habitual paradigm of the East Papuan language Yele (70). In Yele, person, number and tense-aspect-mood are marked on the auxiliary (which Henderson (1995) calls a ‘prenuclear element’). The distal habitual auxiliary is the same as the near past with the addition of the suffix -mo (though note that 1DU nyi- corresponds to the immediate past form). The form dpıˆ is used for 2DU/3DU in the near past paradigm, while in the distal habitual it is extended to the 3SG as well. Regular syncretism of singular and plural (but not dual) does not occur in our material, except alongside other patterns of number syncretism.28 A particularly striking example comes from the Kiowa-Tanoan language Kiowa (discussed in greater detail in x3.7), which illustrates all three logical possibilities. Number syncretism is dependent on noun class. Class I 28

A large part of the possessed noun paradigm in Yup’ik displays singular/plural syncretism, in the absence of any other patterns of number syncretism. However, it appears to have a fairly transparent phonological or morphophonological basis.

94


1SG 1DU 1PL 2SG 2DU 2PL 3SG 3SG 3PL

Auxiliary forms for continuous events in Yele (Henderson 1995: 36) immediate past

near past

distal habitual

nıˆ nyi nmıˆ nyi dpıˆ nmyi Ø Ø Ø

nıˆ ny:oo nmıˆ nyi dpıˆ nmyi dıˆ dpıˆ dnyi

nıˆ -mo nyi-mo nmıˆ -mo nyi-mo dpıˆ -mo nmyi-mo dpıˆ -mo dpıˆ -mo dnyi-mo

(largely animates) has singular/dual syncretism, class II (inanimates) has ‘normal’ dual/plural syncretism, and class III has singular/plural syncretism. The coexistence of these patterns, and particularly the unusual class III pattern (this class consists of only four nouns), is bound up with the system of inverse number marking (see x3.7). (71)

SG DU PL

Kiowa nouns (Watkins 1984: 82–9) I ‘horse’

II ‘tree’

III ‘apple’

ce˛ˆ : ce˛ˆ : ` ce˛ˆ :-gO

` a´:-dO a´: a´:

a´lO` :-bO` a´lO` : a´lO` :-bO`

3.5.3 Directionality Common assumptions about the markedness relations between number values would lead us to expect that dual forms should be based on plural forms. However, there are examples where the syncretic dual/plural form looks like the dual instead. One such example comes from Koryak, where the plural is morphologically derived from the dual (see (50–1) in x3.3.2). For first and second person arguments, the plural is formed from the dual plus the suffix -la. A careful reading of the sources suggests that -la is unstable in some contexts (most often where the plural subject acts on a


95

third person object) and liable to be omitted.29 In such cases dual and plural are syncretic, but it is the form otherwise associated with the dual which prevails. 3.5.4 Summary Most instances of the collapse of number values can be characterized as the consolidation of non-singular number values, and this is largely the same across different languages. Since this is in keeping with the number hierarchy, most examples can be treated as the result of a language employing different cut-off points on the hierarchy for different items. However, as the system of plural marking in Koryak shows, this semantic hierarchy is not necessarily mirrored by the morphological hierarchy. Nor is the number hierarchy an absolute constraint on number syncretism, but it comes close. In this respect the behaviour of number is far more predictable than that of other features.

3.6

Tense-aspect-mood

3.6.1 Introduction In this section we consider syncretism of values of tense, aspect and mood features, which we shall consolidate under the single rubric of ‘tenseaspect-mood’ (TAM). The principles behind TAM systems are particularly heterogeneous from language to language, and we are unlikely to be able to make many cross-linguistic generalizations by looking at the values of individual features. However, it turns out that there are some distinct tendencies in morphological realization of TAM syncretism, which we shall focus on. Consider English weak verbs with the ending -ed, as in walked. This realizes the syncretic combination of preterite and past participle, otherwise distinct in strong verbs, as in drove versus driven. In a small number of verbs (monosyllabic roots ending in a dental consonant, with a lax high root vowel), the ending -ed, though possible, is typically absent, resulting in syncretism of all three possible distinctions.

29

The variation is partly determined by the source and, within a particular source, the tensemood paradigm (where the system of person-number marking should not vary, in principle); within Zˇukova (1972) the variation also depends on the chapter where the material is presented. See the entry under Koryak in Baerman, Brown and Corbett (2002a) for details.

96

(72)


PRESENT PRETERITE PAST PARTICIPLE

drive drove driven

walk walked walked

quit quitted quitted

quit quit quit

Thus English illustrates two contrasting morphological patterns. In the first case there is an affix which directly encodes the syncretic values: -ed typically marks the syncretic combination of preterite and past participle. Where these two are distinct, their morphological realization is by other means (ablaut, and possibly the ending -en). In the second case, the syncretism is realized by not applying a rule (see Zwicky 1991: 117). The syncretic form is the morphological base to which nothing has been done. It is our impression that most examples of TAM syncretism that we are familiar with are of this second type 3.6.2 Affix suppression Some of the clearest examples of the suppression of what appears to be a regular TAM affix come from Oceanic languages. For example, both Gapapaiwa and the related language Tawala have a marker -na-, identified as a future tense marker in Gapapaiwa and as a potential mood marker in Tawala. It is regularly inserted after the person-number markers, except that in Gapapaiwa it is not found with the 1INCL PL (73a), and in Tawala it is not found in conjunction with any of the first person markers (73b). This leads to syncretism with the morphologically simplest TAM form, the subjunctive in Gapapaiwa and the simple aspect in Tawala.

(73) a. Gapapaiwa (McGuckin 2002: 308)


subjunctive

future

akuitakakoi-

a-naku-nai-nataka-nako-nai-na-

b. Tawala (Ezard 1997: 117)


simple aspect

potential mood

auitatoohi-

au-nai-natatoo-nahi-na-


97

Note that, at least in the case of Tawala, the absence of -na- cannot be attributed to any inherent incompatibility between first person and the potential mood, as there are verbs such as ‘meet together’ (74) whose stems alternate for all persons in the potential mood, a result of what appears originally to have been incorporation of the -na- marker into the stem. (74)

Tawala ‘meet together’ (Ezard 1997: 117)


simple aspect

potential mood

a-‘m-boina u-‘m-boina i-‘m-boina ta-‘m-boina to-‘m-boina o-‘m-boina hi-‘m-boina

a-n‘om-boina u-n‘om-boina i-n‘om-boina ta-n‘om-boina to-n‘om-boina o-n‘om-boina hi-n‘om-boina

In this respect it is interesting to note that Loniu, another Oceanic language, shows a similar effect with its potential mood marker, in this case affecting the second person. The potential mood differs from the present/ past by the addition of the prefixed element kV- (in the plural, the vowel depending on conjugation class) or k-. But the 2SG lacks this k-, resulting in regular syncretism of the potential and present/past. (75) Loniu ‘come’ (Hamel 1994: 73)

1SG* 2SG 3SG PL

present/past

potential

i-me, u-me e-me i-me me

ki-me, ku-me e-me ki-me ke-me

*free variation in the vowel of the 1SG prefix

Another Oceanic language, Ura (76), displays tense syncretism correlated with number. Ura has three basic tense-mood paradigms which are distinguished solely by their person-number endings: recent past, distant past and optative. The recent past and distant past are identical for all plural persons. These syncretic forms resemble, morphologically, the

98


recent past: while 1SG and 2SG distant past differ from the recent past by the addition of -mi-, this element is lacking in the plural forms (the -mi- of the 1PL exclusive belongs to the pronominal element). The third person form in the plural has the initial consonant c- characteristic of the recent past of the third person form in the singular. There is no obvious semantic reason why these two tenses should be undistinguished in the plural; note that in the closely related language Sye, they are distinct (Crowley 1998: 90, 95). (76)

Ura (Crowley 1999: 157–61)


recent past

distant past

optative

yauki(c)i(q)urqimirqir(c)ir-

yaumikami(y)i(q)urqimirqir(c)ir-

yaupikapipiqispir——* qipirpir-

*not attested

An analogue to affix suppression can also be found involving stem alternations (see (10) in Chapter 2: x2.2.1). For example, in the NiloSaharan language Fur (77), this occurs with an accentual alternation. Verbs distinguish three TAM paradigms (subjunctive, perfect and present), marking person-number in all three. Within each TAM paradigm, in addition to the person-number prefixes, there are two stem suffixes, one for the 3PL and one for all other person-number values. These suffixes are always distinct for the present tense, but, for most verb classes, the non-3PL suffix is the same for the subjunctive and perfect. In class I verbs, subjunctive and perfect for the non-3PL forms are distinguished by a tonal alternation realized on the suffix vowel (low tone is indicated by a grave accent, high tone is unmarked). In class II verbs, there is no tonal alternation, so that subjunctive and perfect are syncretic for all but the 3PL.30 Similarly, stem vowel ablaut may be involved. In Latvian (78) there is an interesting interplay between stem allomorphy and affix allomorphy 30

This does not seem to have an obvious phonlogical explanation, as -i can receive low tone, e.g. the 1/2 person stem of class II,2 verbs has the shape HH-ı`, where H=high tone syllable (Jakobi 1990: 111).


99

Fur (Jakobi 1990: 104, 109) ‘descend’

1SG 2SG 3SG 1PL 2PL 3PL NHUM 3PL HUM

‘show’

subjunctive

perfective

subjunctive

perfective

u`fuNo ju`fuNo buuNo ku`fuNo bu`fuNo buuNo`l ku`fuNo`l

u`fuNo` ju`fuNo` buuNo` ku`fuNo` bu`fuNo` buuNu`l ku`fuNu`l

aùli jaùli faali kaùli baùli faalia`l kaùlia`l

aùli jaùli faali kaùli baùli faalie` kauliee`

across tense. Latvian has three synthetic tenses: past, present and future. The future has a distinct set of endings and so is never syncretic with the others. The present and past are partly distinguished by their endings, but the first singular ending -u is the same for both tenses. Thus distinction of the two tenses in the first singular is dependent on a stem alternation. First conjugation verbs may undergo root vowel ablaut, or an alternation of the stem-final consonant, while third conjugation verbs have a vowel extension in the past tense, as in ‘wish’ in (78). Second conjugation verbs, such as ‘wash’ in (78), and many first conjugation verbs undergo no stem alternation, and so have syncretism of present and past in the first singular. (78)

Latvian (Mathiassen 1997: 104–5) third conjugation ‘wish’

1SG 2SG 3SG/PL 1PL 2PL

second conjugation ‘wash’

present

past

present

past

gribu gribi grib gribam gribat

gribeju gribeji gribeja gribej am gribej at

mazgaju mazga mazga mazgajam mazgajat

mazgaju mazgaji mazgaja mazgajam mazgajat

There are two phenomena to note here. First, it has been claimed (Karin¸sˇ 1994: 113, citing Graudin¸a 1969: 16–20), that some speakers reanalyse second conjugation verbs as third conjugation verbs, precisely in order to circumvent this homophony. Second, observe the second singular ending.

100


In the past it is -i, while in the present there are two allomorphs: -i, identical to the past tense ending, and -Ø, as in the third person. The choice of allomorph is partly determined by the presence or absence of a present past stem alternation. Stems lacking an alternation have -Ø, as in ‘wash’, thus maintaining the contrast between the two tenses: present mazg a versus past mazg aji (-j- is inserted intervocalically by a general phonological rule). Since -i is the historically expected ending, and its loss does not follow from regular sound change, it has been argued that -i ! Ø was a morphologically driven change, motivated by avoidance of syncretism (Karin¸sˇ 1994: 113, citing Endzelins 1922). Note, however, that the range of the new Ø ending extends in the first conjugation even to verbs which do have a present past stem alternation, so the correlation is not complete. The syncretism with the third person form which resulted in many cases (in some verbs they have distinct stem alternants) was, evidently, more tolerable. 3.6.3 Syncretic affixes Now let us turn to examples such as English -ed, where an overtly syncretic affix is used. English -ed is in fact fairly exceptional, in that it marks verbal features without any indication of person.31 Most examples involve markers with cumulative marking of TAM and person. For example, in the Otomanguean language Chichimeco, TAM and person are marked cumulatively by prefixes (which may differ according to conjugation class), and, for some verbs, by a stem alternation as well (number is separately marked). Example (79) illustrates a verb of the default conjugation class I, showing only singular forms. There are eight TAM values; in the first and in the third persons all eight TAM values have distinct prefixes. In the second person the same prefix ki- is used for the anterior past, recent past, future and present. The present may be distinguished by a stem alternation32 (as in this example, where the stem has three alternants: ’e, te´, nde´), but the first three TAM values are always syncretic in this conjugation class.33

31

32 33

The modern English forms were historically the base forms on which further inflection was added (so that in Old English, past tense forms included person-number marking following the -d). The prosodic alternation manifest on the prefix is associated with the verb stem as well. One class of verbs has su- for the second person present in place of ki- (p. 171), another has su- for both the present and anterior past (p. 172).


101

(79) Chichimeco ‘give’ (de Angulo 1932: 165)

1SG 2SG 3SG

I

II

III

IV

V

VI

VII

VIII

tu´-’e kı` -te´ u`-nde´

ga´-’e` kı` -te´ ga`-nde´

ku´-’e kı` -te´ ku`-nde´

e´-’e` kı´ -’e` e´-’e`

u´-’e` ı` -te´ zu´-’e`

nu`-’e´ mı` -te´ mu`-’e´

ra´-’e` gı` -te´ ru´-’e`

su´-’e` sı` -te´ su´-’e`

I

anterior past

II

future

III

recent past

IV

present

V

immediate past

VI

potential

VII contemporaneous VIII negative

3.6.4 Compound systems Both of the morphological patterns described above – affix suppression and the use of syncretic affixes – may occur within the same language, as is the case with English. A particularly striking example comes from the Papuan language Kiwai of the Trans-New Guinea phylum (Table 1), with ten finite TAM values distinguished both by TAM affixes and by the person prefixes (number is marked separately). In the non-singular numbers, the TAM affixes by themselves suffice to distinguish all the values, but some of these affixes are not found in the singular, namely -ru- (definite past and present habitual), -du- (present habitual) and -duru- (present and near past habitual). The person prefixes in turn distinguish TAM, but only partly. First person is invariant n-, while 2/3 person marking is distributed as follows: r- present, remote future, present habitual g- definite past, past habitual, near past habitual w- near past, immediate future, indefinite future, future habitual

Omission of -du-, -ru- and -duru- in the singular interacts with TAM syncretism in the person markers. With the first person prefix n-, which does not distinguish TAM, the near past, definite past and present are syncretic, likewise the present habitual, past habitual and near past habitual. Tense marking on the 2/3 person prefix largely disambiguates these

n-S r-S n-S-duru-do

present n-S-ri w-S-ri ni-do-S-ri

ni-do-S-ri wi-do-S-ri ni-du-do-S-ri

immediate indefinite future future remote future

‘S’ here stands for any verb stem

trial number (not shown) is distinguished from the plural by the additional affix -bi-

*

n-a-S-go r-a-S-go n-a-du-doS-go r-a-du-doS-go n-a-du-moS-go r-a-du-moS-go

present habitual

*

Note:

2/3PL

1PL

2/3DU

n-S g-S n-S-ru-do

definite past

ni-mi-S-ri ri-mi-S-ri ni-mi-dudo-S-ri w-S-do g-S-ru-do r-S-duru-do wi-do-S-ri wi-du-do-S-ri ri-mi-dudo-S-ri n-S-mo n-S-ru-mo n-S-duru-mo ni-mo-S-ri ni-du-mo-S-ri ni-mi-dumo-S-ri w-S-mo g-S-ru-mo r-S-duru-mo wi-mo-S-ri wi-du-mo-S-ri ri-mi-dumo-S-ri

n-S 1SG 2/3SG w-S 1DU n-S-do

near past

Table 1: Island Kiwai (Wurm 1975: 338–9)

n-a-S-go g-a-S-go n-a-ru-doS-go g-a-ru-doS-go n-a-ru-moS-go g-a-ru-moS-go

past habitual

n-a-S-go g-a-S-go n-a-durudo-S-go g-a-durudo-S-go n-a-durumo-S-go g-a-durumo-S-go

near past habitual

wi-du-m-a-S

ni-du-m-a-S

wi-du-d-a-S

ni-d-a-S wi-d-a-S ni-du-d-a-S

future habitual


103

values, except for the past habitual and near past habitual, which share the prefix g- and so are syncretic for the 2/3 person as well. 3.6.5 Summary Although we have concentrated on the morphological manifestation of TAM syncretism, it is worth considering what relation these patterns might have to the values which are syncretic. Affix suppression reflects a morphological hierarchy: the syncretic form is basic, the non-syncretic forms are morphologically derivative. Whether this can also be seen as reflecting a hierarchy of feature values depends on whether one believes that morphological structure reflects feature structure. Seen in terms of the TAM values, the examples shown above are compatible with that assumption, for example in Gapapaiwa (80a), Tawala (80b) and Kiwai (Table 1) it is the future which is morphologically derived, which we can reasonably construe as a non-basic tense value as well. In Ura (76) the distant past is derived from the recent past, in Latvian (78) the past from the present. Whether or not these reflect genuine tense hierarchies is uncertain, but it is not improbable. However, the environment in which these patterns occur is harder to account for on this view, e.g. why only the singular should be affected in one case, the plural in another, first person in another, and so on. In none of these cases does there seem to be a natural relationship between, on the one hand, the feature value which conditions the suppression of the affix and, on the other hand, the value expressed by the affix. Likewise, the syncretism encoded by syncretic affixes, such as English -ed, does not necessarily express a natural class of values. Indeed, the identity of the past tense and the participle in English is a well-known example of a morphosyntactically unnatural class, used by Aronoff (1994: 23–5) to argue for the existence of morphosyntactically arbitrary morphological forms (‘morphomes’). Nor do classes of related values seem to be involved in the Kiwai example (see the distribution of r-, g- and w- outlined above) or the Chichimeco example (recent past, anterior past and future – and only for second person, at that).

3.7

Polarity effects

3.7.1 Introduction So far we have looked at examples where syncretism can be described as the collapsing of distinct values of a single feature, which has allowed us to consider whether this is a reflection of some closer relationship between the

104


given feature values. But examples which show polarity effects (see Chapter 2: x2.1.1) defy description in these terms, because the syncretism seems to operate over a paradigm rather than features. Somali provides a well-known example. Determiners, which are suffixed to nouns, have two distinct forms: broadly speaking, those in k- attach to nouns which are masculine singular or feminine plural, while those in t- attach to nouns which are feminine singular or masculine plural. (Note that there are exceptions: class 3 and 4 nouns take k- for both numbers.) (80)

F M

Somali definite article (Saeed 1999: 112) singular

plural

-ta -ka

-ka -ta

In its canonical formulation (e.g. Hetzron 1967: 184) polarity involves just this sort of perfect mirror-image identity of non-contiguous paradigmatic cells. As such, it is a rare phenomenon (the well-known examples are from Semitic and Cushitic, but similar effects may be found in Oceanic; see Corbett 2000: 162–4). However we might profitably think in looser terms of generalized polarity effects, whose shared property is that their morphosyntactic description requires reference to more than one feature. This is what Pike (1965: 219) terms ‘second degree neutralization’, defined by the fact that the syncretic forms can be made adjacent to each other only through a linear arrangement, but not through a grid with two axes. We can distinguish three types of polarity: (i) full polarity, (ii) partial polarity, and (iii) mediated polarity. Full polarity is represented by Somali above. Partial polarity is simply one half of full polarity, that is a syncretic form which occupies non-adjacent cells,34 without a corresponding mirrorimage form. This is found in subject person marking on verbs in Siuslaw (81), an extinct language of Oregon, where 2SG and 3PL are regularly syncretic. The third type, mediated polarity, is a relaxed version of partial polarity: the form can be represented as adjacent cells, but a full description still 34

We stress that it is not the geometry of the paradigmatic representation per se which is important, but the fact that reference must be made to different values of two (or more) features.


1INCL 1 2 3

105

Siuslaw (Frachtenberg 1922: 468) singular

plural

dual

— -n -nx Ø

-nØ -nxan -tcıˆ -nx

-ns -auxuˆn, -axuˆn, -ts -aux

requires reference to multiple conflicting feature values. For example, in the Papuan language (82) Wojokeso, of the Trans-New Guinea phylum, same-subject non-future medial verbs have a single form which serves for first and second person in the singular, and second and third person in the dual. (82)

1 2 3

Wojokeso (West 1973: 10) singular

dual

plural

-onji -onji -i

-ontae -onji -onji

-ontone -ontØfi -ontØfi

The existence of polarity effects is undeniable. However, as we shall see in Chapter 4: x4.2.3, many morphological models exclude them outright and so assume that, where they occur, they must be accidental. Therefore, it is worth asking whether there is any morphological or semantic evidence to suggest that they are ever systematic. 3.7.2 Morphological systematicity As with other syncretic patterns, there are two ways of demonstrating morphological systematicity: (i) the pattern is regular, that is, it is repeated across distinct inflectional exponents (see Chapter 2: x2.3), and (ii) there is diachronic evidence that the pattern was extended analogically. Examples of the first type are hard to come by. For example, the system in Somali, pervasive as it may be, involves a single set of inflectional markers (t- and k-). This is typical of most examples – indeed, as we have observed, this is typical of most examples of syncretism in general. Nevertheless, there is at least one striking example from another Cushitic

106


language, Dhaasanac (data from Tosco 2001: 123–206). Subject-person marking on verbs (see (83)) is found in the perfect and imperfect positive, the dependent positive and the short past, and is realized as a stem alternation: one form (labelled ‘A’) is found in the first person singular, third person masculine, first person inclusive plural and the third plural; the other form (labelled ‘B’) is found in the second person, third singular feminine and the first exclusive plural. (83)

Dhaasanac stem alternations

1INCL 1 2 3F 3M

singular

plural

— A B B A

A B B A A

The stem alternation takes a number of different shapes, illustrated in (84), involving changes in the stem-final consonant, (e.g. kufi kuyyi) steminitial consonant (yes ces), as well as vowel alternations (seð sieti). (84) Dhaasanac stem forms (Tosco 2001: 123–206) form A

form B

leeði kufi guurma ?uufumi seð yes

leeti kuyyi guuranna ?uufeeni sieti ces

‘fall down.PRF’ ‘die. PRF’ ‘migrate. IMPRF’ ‘cough. PRF’ ‘walk. PRF’ ‘kill. PRF’

The systematicity of the ‘A’ ‘B’ stem alternation is undeniable, as is the heterogeneity of the syncretic values. Even if ‘A’ is treated as an elsewhere form (Tosco 2001 considers it the morphological base form), ‘B’ remains an example of mediated polarity. We analyse Dhaasanac in Chapter 5: x 5.2. Another, perhaps less striking, example comes from Pashto, where the distinct feminine and masculine forms show regular syncretism of the oblique singular and direct plural cases, here illustrated with adjectival inflection:


107

Pashto ‘ripe’ (Grjunberg and Edel’man 1987: 15) masculine feminine

direct oblique

singular

plural

pox paxa pa:x« paxe

pa:x« paxe paxo

Although evidence of synchronic morphological systematicity is hard to find, there are some plausible diachronic examples. For example, the pattern in Dhaasanac, as suggested by Sasse (1976), originally arose by regular sound change. Verbs fall into two classes, one where personnumber markers were suffixed, and a much smaller class where they are prefixed. The related language Arbore represents something like the older stage of Dhaasanac (86). Regular sound change led to the assimilation of a stem-final resonant (n, m, or l) with a following -t or -n suffix, thus yielding syncretism of the first plural with the second person and third singular feminine in resonant-final stems. This syncretic pattern was then extended by analogy to all the other stem classes, as well as the small class of verbs with person-marking prefixes, as in yes ces in example (84) above (Sasse 1976: 219–20). (86)

Arbore (Hayward 1984: 254 ) prefix conjugation suffix conjugation singular plural

1 2 3F 3M

?- -Ø t- -t t- -t y- -Ø

n- -n t- -t y- -Ø y- -Ø

A similar example comes from Sami declension. In the North Sami dialects, the comitative singular and the locative plural are syncretic in nouns. In the Western Finnmark dialects, which preserve the more archaic pattern, these two forms remain distinct in pronouns. However, in the Eastern Finnmark dialects, this syncretic pattern has been extended to pronouns as well (Hansson 1996: 15–16).

108


(87) North Sami nominal declension, ‘eye’

NOM ACC/GEN ILL LOC COM ESS

(88)

ACC/GEN LOC COM ESS

plural

cˇalbmi cˇalmmi cˇalbma´-i cˇalmmi-s cˇalmmi-in cˇalbmi-n

cˇalmmi-t cˇalmmi-id cˇalmmi-de cˇalmmi-in cˇalmmi-iguin cˇalbmi-n

North Sami pronominal declension, ‘who’

NOM

ILL

singular

Western Finnmark (conservative)

Eastern Finnmark (innovative)

singular

plural

singular

plural

gii gea-n gea-sa gea-s gea-inna gea-nin

gea-t gea-id gea-idda gea-in gea-iguin gea-nin

gii gea-n gea-sa gea-s gea-inna gea-nin

gea-t gea-id gea-idda gea-inna gea-iguin gea-nin

Such patterns may be surprisingly tenacious over time, in spite of their apparent ‘unnaturalness’ (Maiden 1992). 3.7.3 Semantic systematicity We have singled out polarity effects as a distinct phenomenon because they cannot be attributed to natural classes in any obvious sense. Nevertheless, there are some examples where it is not implausible to seek a coherent semantic motivation. Singulative number in the Nilo-Saharan language Camus is an illustration of this principle, though not strictly speaking syncretic under our definition (examples from Serzisko 1982: 197, citing Heine 1981). The singulative endings, -i, -ni and -o mark the singular on nouns whose default value is plural, e.g. l-to´to ‘grinding stones’ l-to´to-i ‘grinding stone’ (the choice among singulative endings is lexically


109

determined by the noun). These same endings are also used to mark the plural of nouns whose default value is singular, e.g. l-me´rekec ‘ram’ l-me´rekec-i ‘rams’. Broadly speaking, there is a coherent semantic distinction between the two types of nouns: those that take singulative marking describe things that occur in groups or are otherwise unindividuated (‘tears’, ‘flies’, ‘stars’), while those that take plural marking are individuated, e.g. animates (Heine 1981: 107–8). Thus, the endings -i -ni and -o could be said to mark ‘unexpected’ number, whose specific value is contingent on the semantics of the noun. However, for purposes of number agreement (as occurs on verbs), the two types are not distinguished (Heine 1981: 106), so the semantic principle underlying the lexical distribution of these endings is not active in the syntax of Camus. In Kiowa, the category of ‘unexpected’ number is syntactically active, involving a three-way number opposition with a somewhat unusual semantic basis. Kiowa nouns are divided into a number of different classes on the basis of how they mark number (both on themselves and in terms of agreement). Watkins (1984: 92) distinguishes four major classes. By default (as judged by productivity), class I is for animates, class II for inanimate count nouns and class IV for inanimate non-count nouns. Class III is a closed class, consisting of the words for ‘plum/apple’, ‘orange’, ‘tomato’ and ‘hair’. What concerns us is the category known as ‘inverse’ number, whose value depends on the class of the noun it is associated with. See Corbett (2000: 159–62) for references on the origin of the term and details of related languages with inverse number, as well as discussion of Kiowa as marking the unexpected number value. Its value is plural for class I, singular for class II and both singular and plural for class III. Class IV does not take inverse number marking. The distribution of inverse marking, as well as more conventional marking of singular, dual and plural, is shown in third person intransitive subject verb prefixes illustrated in (89), where e- is the inverse prefix, e˛- the dual prefix and gya- the plural. As Watkins describes them, the noun classes can be characterised as having an inherent number value. Thus, class I nouns are inherently singular, class II inherently non-singular and class III nouns inherently dual. Inverse marking occurs where the referential value of number does not correspond to the inherent value. Broadly speaking, the inherent number values of the individual classes is not too surprising: individuated entities (animates) are preferentially singular, unindividuated entities (inanimates) are plural, and non-count nouns do not mark number. The class III pattern is unusual, in that it marks singular and plural together as

110


(89) Kiowa, intransitive third person subject marking (Watkins 1984: 82–9)

I IIa IIb III IVa IVb IVc

singular

plural

dual

Ø eeeØ Ø gya-

egyaØ egyaØ gya-

e˛e˛e˛e˛e˛e˛gya-

inverse; perhaps significantly, this is an unproductive class. Nevertheless, one should perhaps avoid overemphasizing the semantic naturalness of this system, in that the assignment of nouns to one class or another may be idiosyncratic in some cases, e.g. ‘leg’ belongs to class I but ‘foot’ to class II. Thus, there is some evidence (though not entirely conclusive) that inflection can mark the unexpected or marked value of a particular feature, the specific value being determined in conjunction with the lexeme itself. Similar accounts in terms of markedness have been proposed for classic polarity, i.e. that what is being marked is relative markedness (Serzisko 1982: 198, citing Smith 1979). In Somali, for example, k- would indicate the expected relationship of gender and number (masculines are singular and feminines are plural), while t- would indicate the unexpected relationship (namely the reverse). However, the naturalness of these relationships is less obvious than in Kiowa, and still less than in Camus. The more this analogy is extended the more tenuous it becomes, and it is doubtful whether this is a useful perspective to take on patterns such as we find in Sami ((87)–(88)). There is one more area where a semantic explanation has been suggested for polarity effects, namely person-number marking. Thus, Foley notes the frequency of 2SG/1NONSG syncretism in Papuan languages (found even in free pronouns) and speculates that there may be a semantic connection between the two values (Foley 1986: 72–3). He claims that this pattern does not occur in languages with a distinct first inclusive and proposes that in such instances a singular addressee potentially enters into the semantic range of first non-singular, hence motivating the syncretism. However, there are two possible objections to this interpretation. First, based on our sample (Baerman 2002b ), there does not seem to be any noticeable crosslinguistic preference for this particular polarity pattern. However, the examples are too few to make any serious statistical claims. Second, as


111

we have noted (Chapter 3, x3.2.2), the presence of a distinct first inclusive is not a barrier to syncretism of first and second person, at least outside of New Guinea. 3.7.4 Summary In spite of some tantalizing examples involving number marking (Camus, Kiowa), there is little reason to believe that polarity effects on the whole can be analysed as representing any sort of semantic natural class. On the other hand, there is some good evidence that such patterns may be morphologically encoded. What is more, the best examples of morphological systematicity are probably among the worst examples of semantic systematicity. This makes polarity effects – where their systematicity can be demonstrated – compelling evidence that syncretism may be morphologically encoded.

3.8

The interaction of features

3.8.1 Introduction In xx3.1–3.7 we looked at individual features. However, we also saw, specifically for person in x3.2 and gender in x3.4 in relation to number, that there may be interactions with other features. Rather than look at the distinctions made within a feature (the ‘within-sets’ approach) we specifically concentrate in this section on the interactions between features (the ‘between-sets’ view). In contrast to the within-sets view, it is also possible to look at the environments in which syncretism occurs, as pointed out in Chapter 2: x2.2.2. For example, Carstairs (1984) talks of ‘conditioning’ and ‘neutralised’ properties (features). As an example he cites the syncretism between the dative and ablative plural in Latin, treating the distinction between dative and ablative as ‘neutralised’ and number as the conditioning property. The work by Aikhenvald and Dixon (1998), which among other things uses examples of syncretism to study dependencies between grammatical features, also belongs to this tradition. Rather than considering specific values belonging to a feature, we distinguish here between features serving as a context for a syncretism, similar to Carstairs’ ‘conditioning’ property, and features which serve as the content of the syncretism itself. In this way we abstract away from the issue of the particular values involved and look at behaviour of the features to which they belong. In order to do this we will look at a small sample, namely the thirty languages from the Surrey Syncretisms Database (Baerman, Brown and Corbett 2002a).

112


3.8.2 Syncretisms and their contexts Distinctions in form for one feature may be distributed unevenly in the presence of values of another feature. Hence, in Russian (90), singular number can be seen as a context for the syncretism of dative and locative cases (and dative, locative and genitive for class III). (90) Case syncretism in Russian

GEN SG DAT SG LOC SG GEN PL DAT PL LOC PL

class I ‘table’

class II ‘room’

class III ‘bone’

class IV ‘window’

stola stolu stole stolov stolam stolax

komnaty komnate komnate komnat komnatam komnatax

kosti kosti kosti kostej kostjam kostjax

okna oknu okne okon oknam oknax

This contrasts with the plural in Russian where the dative, locative and genitive have distinct forms for all nouns. This is why number can be viewed as a context for case syncretism. From the between-sets perspective, it does not matter which particular value provides the context for a syncretism. It matters only that there is a difference in the pattern of syncretism across different values of the feature which is the context. As we saw for gender in x3.4, there is an interaction with number. There can also be interaction with case. Gender is an inherent feature for nouns but is an inflectional feature which is manifested on agreement targets, such as verbs and adjectives. In the West Slavonic language Lower Sorbian, as in other Slavonic languages, there is syncretism of masculine and neuter gender in the oblique cases in the singular, as illustrated in (91). This is the same pattern as for the Russian example we saw above in Chapter 2 ((36) in Chapter 2: x2.4.3). As the masculine and neuter are formally distinguished in the nominative case, they are viable grammatical distinctions for Lower Sorbian. Consequently, failure to distinguish them in the genitive, dative, instrumental and locative can be taken as instances of syncretisms, as was true for the comparable Russian example we saw in Chapter 2: x2.4.3. Furthermore, case here appears to be a context for gender syncretism. (For the time being we leave the question of the feminine singular locative and dative and return to it in x3.8.3.3.)


113

(91) The adjective dobry ‘good’ in Lower Sorbian (Stone 1993: 630)

NOM SG ACC SG GEN SG DAT SG INST SG LOC SG

3.8.3

masculine

neuter

feminine

dobry dobry dobrego dobrego dobremu dobrym dobrem

dobre dobre dobrego dobremu dobrym dobrem

dobra dobru dobreje dobrej dobreju dobrej

Nominal feature interactions

3.8.3.1 Typologies of interaction Within Network Morphology, Brown (1998b) discusses the relationship between number and case in terms of a feature dependency of case on number. Specifically, the idea is that, ‘where case and number occur together, a number feature may determine the number of case distinctions, but not the other way round’ (Brown 1998b: 190). Of course, the statement of a dependency relation between the two may be interpreted as implying that number is causal in influencing the distinctions available. The type of interactions seen in the previous section might lead one to propose a hierarchical ordering of features. (92)

A possible hierarchy gender syncretism > case syncretism > number syncretism

In other words, if a language has number syncretism, it must have case syncretism and gender syncretism; or if it has case syncretism, then it must have gender syncretism. But this hierarchy needs further refinement. First, if a language has gender morphology, but not case morphology, for example, then this hierarchy does not apply. In other words, it does not mean that if a language exhibits gender syncretism, it must have case morphology, and there must be syncretism somewhere within that. A second refinement would be to restrict the hierarchy in (92) to purely inflectional features, so that, for example, the generalisation about gender syncretism would not apply to nouns. So it also has to be clear which morphological domain the hierarchy in (92) applies to. Another problem with (92) is that there are counter-examples to the implications from which it is constructed.

114


One set of counter-examples to the relation between case and number in (92) can be found in the Konstanz universals archive. Universal 115 is given as follows: ‘It is common for certain case distinctions to be neutralized for a given number, while the opposite phenomenon perhaps never occurs.’ This is attributed to Uspensky’s (1965: 210) reading of Greenberg (1963: 103). Counter-examples were suggested by Hjelmslev (1935–7). Koryak represents a counter-example to (92), because the singular and plural are conflated in all cases except the absolutive in one set of nouns, but not in all nouns (93). Furthermore, as there is still a number distinction in the other cases in the second declension example in (93), this cannot readily be treated as an example of number neutralization, as there is a set of syntactic objects which distinguish number in these contexts. (Though note that these do conflate dual and plural.) There are a number of possible reactions to such counter-examples. One of them is to argue that feature ranking plays a role but may differ from language to language. Stump (2001: 239) develops a Feature Ranking Principle whereby the specification of the ranking of features is languagespecific; this is discussed in Chapter 4: x4.4.4. As we will attempt to show, there is something about the relation between the features of case, gender and number in terms of syncretism which may not be just language-specific. 3.8.3.2 Exploring interaction In this section we specifically look at cross-linguistic data on interaction between features, analysing the thirty languages from the Surrey Syncretisms Database. The sample in the database was not chosen randomly; instead the languages were selected to explore the logical space of possibilities, including extremes, and so the conclusions are indicative rather than absolute. The method used involves abstracting away from the particular feature values. Consider the partial Slovene paradigm in (94). Dative and locative cases are distinct in the plural in Slovene, and so it can therefore be considered to have these case values. The syncretism in (94) is an instance of a more abstract relationship between case and number, namely the relationship where number is the context and case is syncretic. We have seen this pattern in other examples, of course. This can be represented as in (95). For Slovene any instance of number as context with case as syncretic can be represented as in (95), thereby yielding one instance of this type of syncretism for that language. We abstracted information of this type from the Surrey Syncretisms Database (Baerman, Brown and Corbett 2002a).

CONT

NARR-CAUS

DES

ADIT

DAT

TRANS

ABL

LOC

dual en’picˇi-t en’picˇi-k en’picˇi-te an’pecˇe-Nqo an’pecˇe-jp«N an’pecˇe-N an’pecˇe-jt«N en’picˇi-nu en’picˇi-kjit en’picˇi-jite

en’picˇ en’picˇi-k en’picˇi-te an’pecˇe-Nqo an’pecˇe-jp«N an’pecˇe-N an’pecˇe-jt«N en’picˇi-nu en’picˇi-kjit en’picˇi-jite

first declension ‘father’

singular

Koryak noun declensions

ERG-INS

ABS

(93)

en’picˇi-w

plural appa appa-na-k appa-na-k appa-na-Nqo appa-na-jp«N appa-na-N appa-na-jt«N appa-na-no appa-na-kjet appa-jeta

singular

appa-nte appa-j«k appa-j«k appa-j«ka-Nqo appa-j«ka-jp«N appa-j«k-«N appa-j«ka-jt«N appa-j«cˇge-no appa-j«-kjet appa-j«ka-jeta

dual

second declension ‘papa’

appa-w

plural

116


Locative/dative syncretism in a-stem nouns Slovene ‘grove’

NOM SG ACC SG GEN SG LOC SG DAT SG INS SG

(95)

dobrava dobravo dobrave dobravi dobravi dobravo

Feature interaction (number and case)

NUMBER

CASE

Context

Syncretic

As there are thirty languages in the database, the potential maximum number of occurrences of this type of syncretism is thirty. But this does not mean that the potential maximum number of occurrences in the database of number as context or case as syncretic, is only thirty, as these features may occur as contexts or syncretic in the presence of other features, such as gender, for example. If the syncretisms represented in the database are abstracted in this way for all languages and word classes, the number of records in the database is reduced to about one-quarter of the original size, indicating that this is a powerful means for making generalizations. If we further select from these records those which apply only to nominals, that is adjectives, nouns, pronouns and numerals, then this figure is further reduced. We can visualize the occurrences of the nominal features in syncretisms as a web of binary pairs of connections between the context values and the syncretic values for a given pair of features. We then rate these links between the pairs of values as strong if they co-occur more than thirty-five times, as medium if they co-occur more than fifteen times, and as weak if they co-occur less than fifteen times. In Figure 1 the dotted lines are the weak connections, the medium connections are thin solid lines and the one strong connection is a thick solid line. An instance of a weak connection is that between context gender


117

Figure 1. Web of nominal features and syncretism

and syncretic definiteness, for example. This means that there are very few examples where gender values serve as a context for syncretism of definiteness values. It should also be noted that there may be multiple connections where context is involved more than once, or where syncretic values are connected more than once. For instance, where the dative plural of gender III has the same form as the dative singular of gender III in Tsakhur, this involves both gender and case as contexts for number syncretism. That is why there are connections between context gender and context case, for example. Equally, syncretism between masculine singular and feminine plural in Somali involves syncretic number and syncretic gender simultaneously. The commonest tendency observed for nominal features in the database is represented by the one strong connection in Figure 1: number is the context and case values are syncretic. This suggests that case being syncretic in the context of number may be commonly found within languages which have syncretism. It should be noted that person and definiteness have only weak links associated with them as either context or syncretic. This is because these features do not occur with sufficiently high frequency in nominals for any conclusions to be drawn. Gender can only be an inflectional feature for nominals which are not nouns (i.e. adjectives, pronouns and numerals), unless involved in possession constructions, and it is not a feature for all the languages in the database.

118


Consequently it does not have any strong links. It does have medium strength links, and these all involve gender as syncretic, typically with case and number as contexts, but also alongside number as a syncretic feature. This fits with the observation in x3.4 that gender syncretism is more the rule than the exception. The other medium strength links fill out the rest of the logical space of connections between case and number as syncretic or context, so that case can serve with number as a context for syncretism, or itself have syncretism in the context of number, indicating that the common tendency for case to be syncretic in the presence of number is indeed just a tendency. 3.8.3.3 Constraints on nominal features The hierarchy introduced in (92) can be interpreted as indicating the likelihood of a feature to show syncretism, with the left-most feature being most likely to do so. Equally, the rightmost feature in (92), namely number, may have a greater tendency to act as the context for syncretism. This tendency is reflected in the analysis of the languages in the Surrey Syncretisms Database, as represented in Figure 1. The Russian example (90) fits this hierarchy, with case distinctions (here locative and dative) lost in the presence of singular number. In the Sorbian example in (91) the combinations of singular number with genitive, dative, instrumental and locative cases are contexts for the loss of distinction between masculine and neuter genders. Such an approach allows for a hierarchical ordering of the features involved. However, we find that in the Sorbian data in (91) it is clear that there is also dative-locative syncretism in the feminine. In this instance gender is the context and case is syncretic, the opposite way round from what is predicted by (92). So, if we are to argue that ranking of features alone is tenable, this leaves us with an apparent paradox. If we accept that this is not possible, then are there any possible generalizations to be made which do not have counter-examples ? We shall propose that, yes, there are indeed two generalisations about syncretism of case and gender which do appear to hold for the languages in the database. Furthermore, it should be relatively straightforward to find counter-examples, if they exist. The notions of syncretic and context can be defined as follows: the syncretic values of a syncretism S are the two feature values V1 and V2 of the same feature, whereby V1 6¼ V2. In contrast, the context values of a syncretism S are the two feature values V1 and V2 of the same feature whereby V1 ¼ V2. Of course, a syncretism does not necessarily require a


119

context, whereas by definition it requires a content. Given these two definitions we are now placed to make our generalizations. (96)

a. For a word-class W in language L If values from the feature gender serve as context for a syncretism in W, then values from the feature gender must be syncretic elsewhere in W. b. For a word-class W in language L If values from the feature case serve as context for a syncretism in W, then values from the feature case must be syncretic elsewhere in W.

A word class can range from an individual lexeme and its paradigm through inflectional classes to nouns, nominals (a combination of nouns, pronouns and adjectives) and verbs. In the database of thirty syncretism languages there are no instances which violate the generalizations in (96). It should be noted that these generalizations do not rank case and gender relative to each other. However, it does follow from (96b) that, if case is a context for gender syncretism, as in (91), then it must also be syncretic somewhere within the same word class. Although this fact about case must follow from (96b), in principle it has nothing to do with gender being syncretic. Of course, if gender and number are the only other two features in W, then in order to satisfy (96b) we will find in a language which cumulates gender, case and number that case must necessarily be syncretic in the presence of these, and in the Sorbian example in (91) we find that this occurs in the feminine singular. Hence, we find that the apparent paradox of (91) is not at all a paradox. It is the result of superimposing a requirement for the co-occurrence of certain features onto the generalizations in (96) which are logically independent of this requirement. We should also note from (96a) that the case syncretism in the feminine singular, where gender is serving as (part of) the context, then logically requires gender to be syncretic somewhere. And this is satisfied by the masculine/neuter syncretisms of that paradigm, where case serves as context. Even Koryak obeys the constraint (96b), as the word for ‘papa’ in (93) has case syncretism between locative and the ergative-instrumental case, which means that case is itself also syncretic somewhere within the noun domain, as well as being the context for number syncretism. 3.8.4 Verbal feature interactions The number of potential features involved in verbal inflection is greater than for nominal inflection and this should mean that the space of logical

120


possibilities is greater. As we shall see in x3.8.4.2, there do appear to be a number of constraints on which verbal features can be context and which syncretic. We shall look at tense, aspect and mood (TAM) and agreement. 3.8.4.1 Verbal paradigms In the classification of Booij (1996: 2), tense, aspect and mood are features which are inherent to the verb. In contrast, person, nominal number and gender are agreement features which encode information about the arguments of the verb. Voice and negation (or affirmation) must also be considered relevant to the verb as they involve the number of arguments of the verb and the truth conditions of the predication. (97)

A verbal hierarchy Agreement Syncretism > TAM Syncretism

Where verbal systems have both TAM marking and agreement marking on the verb this appears to be a reasonably strong generalisation, as long as it is understood that syncretism within tense, aspect and mood need not be accompanied by agreement syncretism within the same word. For instance, in the Tibeto-Burman language Limbu past and non-past are syncretic in the presence of a second person singular subject and third person object of either number.

(98)

Tense syncretism in Limbu (van Driem 1987 : 370) ‘teach’

2SG > 3SG 2SG > 3PL

non-past

past

kehu?ru kehu?rusi

kehu?ru kehu?rusi

In (98) we see that for the particular combinations involved there is no syncretism of the agreement features, but there is syncretism of tense. However, within the detailed study of thirty languages in the database there is no example of a language which has some TAM syncretism in part of its verbal paradigm and does not have syncretism of some agreement feature somewhere within the verbal morphology. In Limbu, for example, there is syncretism of the 2 > 1PL with 2PL > 1SG, which cross-cuts the


121

Figure 2. Web of verbal features and syncretism

TAM paradigm. Consequently, (97) may be true of the verbal domain within languages but is not true at the word level. Furthermore, we cannot easily make equivalent claims for the verbal domain as those in (96). More concretely, although (97) appears to hold, we cannot say that TAM will be a context, if it is syncretic. 3.8.4.2 Exploring the verb data Figure 2 is a web of connections for the verbal domain and its features. Again, we count connections which occur more than thirty-five times as strong (bold) links between pairs of syncretic or context values. Medium links occur less than thirty-five times, but more than fifteen, and weak links (dotted lines) occur less than fifteen times. Features which occur frequently may have a number of strong links for both syncretic and context values. What is important are the differences between the strengths of their links. If we consider person, it has three strong links as syncretic: person often occurs as syncretic with TAM as a context, with number as a context, and along with number when the person feature’s values are syncretic. Person also often occurs as a context when number is syncretic. Number behaves in almost exactly the same way as person, having the same strength of connections as person. The pattern of involvement of gender on verbs is both expected in some respects and surprising in others.

122


As gender does not occur as much as person and number, it does not have any strong links, and, as expected, its strongest links are medium ones where its values are syncretic. In particular, one of its strongest links is with TAM as a context. While we might expect gender to be a syncretic feature par excellence, it contrasts with the other agreement features person and number in never co-conspiring with TAM, voice or negation in a syncretism pact (i.e. where two or more features are syncretic, as we see with polarity effects). This can be summed up as in (99). (99)

TAM is not syncretic simultaneously with gender

We could assume that (99) is the result of lack of co-occurrence of gender and TAM marking together. But in the Surrey Syncretisms Database (Baerman, Brown and Corbett 2002a), ten of the languages (one-third) have gender marking on verbs. Of these, there are four languages – Burushaski, Classical Arabic, Kashmiri and Somali – where gender and TAM are realized together in forms which involve syncretism of some feature. In Burushaski we can find examples where there is gender syncretism in the presence of a particular tense, and this may also be accompanied by person or number syncretism (i.e. more than one feature being syncretic). In Classical Arabic the distinction between imperfect jussive, indicative and subjunctive is neutralized in the second and third person plural, while there is still a gender distinction. This gender distinction could be argued to be syncretic, because of the difference between gender marking in the plural and singular, but even so, the mood distinctions are neutralized entirely rather than syncretic. Also, in the second person singular feminine the jussive and subjunctive are syncretic. Here, then, we have an example of TAM syncretism in the presence of gender as one of the contexts, although the interactions are complex. The overall picture is that gender syncretism can occur in the presence of TAM (as context), and that, more rarely, TAM syncretism can occur in the presence of gender (as context), but TAM syncretism and gender syncretism do not appear to co-occur. Finally, it could be argued that the apparent tendency for TAM and gender on the verb not to syncretize together could be attributed to the fact that gender marking on verbs is typically associated with participles, diachronically or synchronically. However, it should be noted that there are examples of gender marking in both synthetic and periphrastic verbal constructions, suggesting that we cannot attribute this tendency solely to this fact. It is also a fact which does not contradict (97), which is a statement about the verbal system as a whole.


123

3.8.5 Interpreting the generalizations In this section we consider why the generalizations in (96) and (97) appear to hold. 3.8.5.1 Number, case and gender on nominals One possible explanation for the difference between case and gender on the one hand, and number on the other is the cardinality of the respective sets. One interpretation which can be put forward is that those sets which show a tendency for syncretism between values are ones which are high cardinality (i.e. which typically are many-valued). While this may be an explanation for case, it is less plausible for gender, which can easily occur as a twovalued system. This is indicated by Corbett’s (2005) survey where 45 per cent (50/112) of those languages with gender systems have two-valued systems. Of course, we cannot tell whether there is syncretism once distinctions in a two-valued system are lost. Gender is also associated with a wide variety of possible values which can be heterogeneous. Heterogeneity of features can be manifested cross-linguistically in languages with twovalued systems. For instance, two-valued gender systems could differ in terms of the two different values involved. In features which have a high cardinality there is also a possibility for greater variation, that is heterogeneity, of feature values. One explanation for the generalizations in (96) might be that the wide variety of possible values for gender and case mean that there just is a greater probability of syncretism occurring among these features. But there does seem to be something in the nature of these features which lends them to syncretism, irrespective of the overall heterogeneity of the features cross-linguistically. Another possible explanation is grounded in the overall syntactic function of the features. Gender, which in our treatment is an agreement feature, is non-inherent and is imposed on agreement targets by a noun head or noun phrase. Equally, case is imposed on nominal words by syntactic dependency or constituency. In other words, the rules of syntax must clearly make reference to the features of gender and case if they are present in a language. Hence, a property of both these features is that they are contextual inflection, rather than inherent inflection in Booij’s (1996: 2) terms, which may indicate that this is an important factor for syncretism. 3.8.5.2 TAM and agreement on verbs As indicated, the generalization in (97) holds for the verbal domain but not within the same word form. Verbal morphology may be a combination of

124


inherent TAM and marking of the agreement features for person, number and gender, among other things. As we have seen, there is an expectation that syncretism within the TAM system somewhere in the verbal domain will entail syncretism within the agreement system. Again, it is the contextual inflection which is most likely to be affected, but the fact that gender appears not to be syncretic simultaneously with TAM – on the rarer occasions when TAM is syncretic – suggests another important dimension to the interaction of feature sets on the verb, namely that gender is the best representative of agreement features and is even less likely to behave in a similar way to the non-agreement features than person or number. 3.8.6 Summary Looking at syncretism in terms of the interaction of the feature sets involved proves particularly fruitful. Our examination of the database shows that the use of hierarchies of the feature sets is insufficient at the word level. However, the nominal features of case and gender are much more prone to syncretism than number, and the agreement features on the verb are also more susceptible to syncretism than the TAM features, if one considers this question at the level of the word domain. This aspect of syncretism suggests a particular paradigmatic design, which can be obscured in particular instances, but the general tendencies are clear. 3.9

Conclusion

The preceding sections have illustrated something of the morphological variety and cross-linguistic scope of inflectional syncretism. In particular, we have considered the role played by semantics or feature structure: patterns repeated across unconnected languages are likely to have a basis in semantic properties of inflectional features that are shared across languages. In this respect different features yield different results. On the one hand, patterns of number syncretism are quite similar across languages (collapse of non-singular number distinctions), suggesting a basis in the semantics of number itself (x3.5). Gender values, while themselves language- specific, show a general tendency to fall together in non-singular numbers (x3.4). On the other hand, patterns of tense-aspect-mood syncretism are typically bound to the particulars of verbal morphology and seem not to have a direct connection with meaning (x3.6). Other features show a mixture of the two; for example, patterns of core case syncretism show


125

many similarities cross-linguistically, but non-core case syncretism shows language-specific peculiarities (x3.1). It will be important to keep these considerations in mind in the following chapter, as we consider different formal models of syncretism, which often take the semantic structure of features as a point of departure.

4 Formal representation

4.1

Introduction

It is a generally shared assumption that any adequate formal model of morphology ought to take some account of inflectional syncretism. Most investigators who have addressed syncretism overtly have taken this assumption one step further: it is not enough to describe syncretism, one should also constrain it. This goal is motivated by two factors. First, syncretism is something of an aberration: by default we assume a one-to-one relationship between morphosyntactic function and form, and syncretism is a violation of this assumption (Carstairs 1987). Second, syncretism displays preferred patterns, as we have seen throughout Chapter 3. Ultimately, the morphological description of a particular syncretism must contain two elements: (i) a list of the set of values which are syncretic, and (ii) a way of associating this set with a form. To a large extent, constraints on syncretism are a product of how these elements are treated. For example, the syncretic set may be a natural class of values or simply a stipulated disjunction, while the form itself may be defined over the whole set, or defined in terms of one of the constituent members. In x4.2 below we examine the inherent properties of different rule types, and in x4.3 we see how these have been employed in particular accounts of syncretism.

4.2

Defining sets of values

4.2.1 Natural classes Probably the most common approach to syncretic sets of values is to treat them as a reflection of underlying feature structure. That is, feature values are themselves composed of more basic sub-values, which can be grouped into natural classes (a view first formalized by Bierwisch 1967). Given a particular model of feature structure, one can predict what will be the possible and impossible patterns of syncretism. Three basic types of feature structure 126


127

are possible: (i) flat, (ii) hierarchical and (iii) cross-classifying; the last term is taken from Johnston (1997). While cross-classifying feature structure is, in principle, unlimited in what it can describe, flat and hierarchical feature structures have some inherent restrictions. Therefore it will be revealing to contrast what can and cannot be described by these two models. As an illustration we use two examples of gender syncretism, from the Atlantic (Niger-Congo) language Noon and the non-Pama-Nyungan Australian language Gaagudju, which reveal both the power and the limitations of the two models of feature structure. Noon distinguishes seven genders (see (1)). They are all distinct in the singular, and show a convergent pattern in the plural (see Chapter 3: x3.4.2): genders 1, 2 and 3 are syncretic, and genders 4, 5 and the diminutive gender are syncretic.

(1)

Noon gender (Soukka 2000: 66)

1 2 3 4 5 diminutive animate

singular modifier

plural modifier

wfmkpjy-

cccttt”-

The Noon paradigm can be easily described using hierarchical feature structure such as in (2), made up of binary branching nodes, which have been given the arbitrary labels A–H. Syncretism of genders 4, 5 and the diminutive can be accounted for by a rule which associates the prefix t- with node D, assuming that the animate form is itself overtly defined. Syncretism of genders 1, 2 and 3 follows if the prefix c- is associated with node A; that is, it is the general default. (Other orderings of the genders in this hierarchical structure are also possible.) Flat feature structure is illustrated in (3). This cannot be used to model gender syncretism in Noon. The only locus for rules other than the daughter nodes is the single mother node (‘gender’), which will allow us to describe one instance of syncretism (as an ‘elsewhere’ form), but not two.


128 (2)

Hierarchical model of feature structure in Noon A B C D E F G

2

1

(3)

3

4

5

6

diminutive

H animate

Flat model of feature structure in Noon gender

1

2

3

4

5

dimin.

animate

The reverse situation is presented by the Gaagudju paradigms in (4), repeated here from Chapter 3 (x3.4.3), where a flat feature structure will work, but not a hierachical one. (4) Gaagudju gender (Harvey 2002: 144, 157, 224–5) realis verbs adjectives

I (male) II (female) III (plant) IV (residue)

first declension

second declension

Ø njiNmagu-

nanjiNnaNnaN-

pronoun

naawu ngaayu ngaayu ngaayu

present and unmarked tense object with first present (intransitive person subject subject) Ø nj-djama-yanj-dja

arranji-rrama-raarra-


129

By assuming the flat feature structure in (5), all the patterns can be described by simple underspecification. That is, the non-syncretic forms are fully specified for gender, while the syncretic forms naN-, ngaayu, nj-dja- and arra- are ‘elsewhere’ forms, unspecified for gender, and so are used wherever there is no more specific form. (5)

Flat model of feature structure in Gaagudju gender

I

II

III

IV

Now let us see how this might be represented with a hierarchical feature structure, consisting of binary branching nodes as suggested above for Noon. Example (6) represents what seems to be a plausible arrangement of values; for convenience, we can label the node which joins III and IV ‘inanimate’, and the node which joins II, III and IV as ‘non-male’. (6)

Hierarchical model of feature structure in Gaagudju gender non-male inanimate

plant I

II

IV

III

By simply underspecifying for lower nodes, we can describe the patterns found in the nominal word classes: III/IV syncretism results from a rule specifying an ‘inanimate’ form, while II/III/IV syncretism results from a rule specifying a ‘non-male’ form. This accounts for the nested pattern that these syncretisms display, since each higher node subsumes the ones underneath. This structure can also be used to describe one of the patterns found with verbs, namely the syncretism of genders II and IV found in the present tense intransitive subject forms. This will result if the inanimate node is left unspecified, as illustrated in (7).

130

The Syntax–Morphology Interface (7) node

form

gender non-male plant

Ø nj-dja ma-ya-

However, this accounts for only one of the two syncretic patterns found with verbs; I/IV syncretism cannot be accommodated with this hierarchy. Only values which are adjacent on the hierarchy can be described as syncretic, since any unspecified value will default down the hierarchy to the next most specific value. This translates into a graphic rule of thumb: there is a single linear order of values, and only values which are adjacent can be described as syncretic (Chvany 1986, Johnston 1997). In this case, no single linear order will bring all the syncretic values adjacent to each other; the order in (8), which accommodates most of the patterns, still leaves genders I and IV separated from each other: (8)

I

II

III

IV

Thus, we have seen the constraints inherent in the two models. If feature structure is construed as flat, there can be only one syncretic form per paradigm, namely the elsewhere form. Hierarchical feature structure allows multiple syncretisms within a single paradigm but restricts the number of ways that feature values may be combined in different paradigms. Of course, hybrid structures are possible as well. For example, one could recast the model of Noon gender along the lines of (9), where nodes A, B and C/D constitute a hierarchy, while the structure beneath nodes C and D is flat. Such a hybrid model simply combines the properties of hierarchical and flat feature structure and does not add any new descriptive potential. Cross-classifying feature structure, by contrast, is potentially unconstrained. For example for Gaagudju we might posit a system of four feature values: ‘non-male’, ‘inanimate’, and two additional values, which we can provisionally term ‘value x’ and ‘value y’. The individual gender values can thus be portrayed as composites of these four sub-values, as shown in (10).

Formal representation (9)

131

Hybrid model of feature structure in Noon A

B C

animate

1

D

2

3

4

5

diminutive

The individual syncretic patterns result from rules which make reference to these component values. The number of syncretic patterns that can be described is constrained only by the number of values employed, which is not subject to any inherent limit. (10) non-male I II III IV

inanimate

value x

value y þ

þ þ þ

þ þ þ

þ

þ

4.2.2 Unnatural classes The other approach to defining syncretic classes of values is simply to stipulate them. The resulting set of values thus represents an unnatural class, in that its members share no common feature other than subjection to that stipulation; a recent work employing this approach within LexicalFunctional Grammar is Dalrymple and Kaplan (2000). Obviously, such a technique can describe any possible set of values. In practice, the full power of stipulation is seldom employed; most investigators impose some constraints, whether overt or covert. For example, Brown (1998b: 87) requires that the values be of the same feature or features, e.g. {nominative, accusative} or {second person singular, first person plural}. Such a restriction seems to be tacitly in place for some other approaches which employ unnatural classes, e.g. Stump (2001) or Zwicky (2000).

132


A somewhat less direct way of stipulating sets has emerged in Optimality Theory. For example, in Kiparsky (2001), the syncretism of genitive and accusative in singular nouns in Finnish is accounted for by morphological markedness constraints which block the use of the expected case endings for the realization of an underlying accusative. Instead, an underlying accusative is forced to use the genitive ending. Thus, the constraints and their ranking conspire to define a set consisting of the accusative and genitive. 4.2.3 A note on polarity effects In the preceding sections we have concentrated on syncretic patterns confined to values of a single feature. Polarity effects represent a special problem: since they involve combinations of values which are not directly related in terms of feature structure, it would appear at first glance that they could only be described by overt stipulation. However, some investigators have proposed that inflectional rules can make reference to abstract properties of feature structure, which allow some polarity effects to be systematically represented in featural terms. The crucial notion here is markedness: as Serzisko notes, some polarity effects can be described in terms of markedness (1982: 194–6). For example, the default system of gender-number marking in Somali (11a) can be viewed in terms of markedness coordination (11b): k- is used where the markedness values of both features is the same, and t- is used where the values differ, on the assumption that feminine and plural are the marked values.

(11) a.

SG

PL

b.

unmarked

M

kt-

tk-

unmarked marked

U/ U

U/M

M/U

M/M

F

marked

(based on Serzisko 1982: 185, 195)

If the features are construed as being hierarchically structured, the values break down as in (12a); if cross-classifying, they break down as in (12b). Both models now lend themselves to a more abstract representation using variables. With hierarchical features, the feature itself is made a variable (13a), with cross-classifying features it is the value which is made a variable (13b).


133

(12) a. hierarchical

b. cross-classifying

feature structure M SG F SG M PL F PL

feature structure F PL þ F PL F þPL þ F þPL

M SG

þF þPL þF þPL

F SG M PL F PL

(13) a. hierarchical

b. cross-classifying

abstract feature structure M SG F SG M PL F PL

abstract feature structure M SG

þa þa þa þa

F SG M PL F PL

aF aPL aF aPL aF aPL aF aPL

The structure in (13a) now allows us to describe t- as the form which represents a single degree of markedness (þa), while k- is an elsewhere form (such an argument is advanced by Be´jar and Hall 1999). The structure in (13b) allows both forms to be explicitly described: k- is the form used where the value for both features matches (aF aPL), and t- is used where they do not match (aF aPL). Thus, if the use of variables is allowed, some instances of polarity can be attributed to feature structure. However, as pointed out in Chapter 3 (x3.7), it is doubtful whether many examples can be profitably seen in terms of neat markedness relationships. 4.3

Symmetrical versus directional rules

As we have defined it, syncretism involves a set of values associated with a single form. Given this definition, the question of how this association is effected remains open. In principle, there are two options: (i) the form is associated with the set as a whole, as in the schematic rule in (14a), or

134


(ii) the form is associated with one of the component values and is ‘borrowed’ by the other members of the set, as in the schematic rule in (14b). (14)

a. there is a set of values {a, b} which have the form x b. there is a set of values {a, b} which have the same form; the form of a is x

It will be useful (borrowing the terminology of Stump 2001) to distinguish these as ‘symmetric’ rules (14a) versus ‘directional’ (14b) rules. On the face of it, symmetric rules are conceptually simpler, since they have the same structure as an ‘ordinary’ morphological rule: there is an element in morphosyntax which is directly associated with a form. Directional rules, by contrast, entail a two-step process. The most familiar breed of directional rule is the ‘rule of referral’ (Zwicky 1985). In describing (14b) above, a rule of referral would state that the form associated with the value b is the same as the form associated with the value a (that is, b refers to a for its form). One of the main reasons for assuming directional rules is to account for directional effects, that is those contexts where the syncretic form looks as if it is ‘belongs’ to one of the component values, numerous examples of which have been seen in the previous sections (especially involving case and person). However, it has been argued that directional effects can be described without directional rules, so that, for reasons of formal economy, they should be rejected. We can illustrate a directional analysis, as well as a nondirectional counterproposal, with a familiar example from Latin, shown in (15). Nominative and accusative are syncretic in neuter nouns and distinct in masculine nouns (at least in the singular). In the singular of the second declension, the form taken by the syncretic nominative/accusative of the neuters, -um, is the same as that of the distinct accusative case of the masculines.

(15)

Latin second declension


neuter ‘war’

masculine ‘slave’

bell-um bell-um bell-i bell-o bell-o

serv-us serv-um serv-i serv-o serv-o


135

Thus, it looks as if the neuter nominative singular has taken the form of the accusative. Using a directional rule, this apparent ‘borrowing’ can be incorporated directly in the analysis. First, one can assume two general rules that apply to all second declension nouns: the nominative singular ending is -us (16a), and the accusative singular ending is -um (16b). For neuters, there is a rule stating that the nominative singular takes the form of the accusative (16c). (16)

a. NOM SG ¼ stem þ -us b. ACC SG ¼ stem þ -um c. NOM SG in neuter ¼ ACC SG

Because the domain of the rule in (16c) is more specific than the domain of the rule in (16a), in that it additionally contains reference to gender, the rule in (16c) takes precedence by Panini’s principle (otherwise known as the ‘elsewhere condition’, ‘blocking’ and the ‘Subset principle’ – we will use the name ‘Panini’s principle’). The morphological description above contains two rule types, one which associates a morphosyntactic value with a form (16a, b), and one which associates one morphosyntactic value with another (16c). In the terminology introduced by Zwicky (1985), the former type is a ‘rule of exponence’, the latter a ‘rule of referral’. Rules of exponence state the relationship between a morphosyntactic value and a form, while rules of referral state the relationship between morphosyntactic values. Some linguists have suggested that rules of referral are unnecessary and should be dispensed with for reasons of formal economy. Thus, Zwicky (2000) suggests that all syncretism can be described as symmetrical syncretism, with directional effects derived as a by-product of underspecification. We can illustrate his proposal with the same Latin data from (15). First, declare that nominative and accusative singular form a class, which is assigned the index ‘X’ (17a). This class is associated with the ending -um (17b). In (17c) a more specific rule is introduced, stating that the nominative singular of the masculines has the ending -us, which takes precedence over the rule in (17b) by Panini’s principle. (17)

a. {NOM SG ¨ ACC SG} ¼ X b. X ¼ stem þ -um c. NOM SG in masculine ¼ stem þ -us

Thus, under a symmetrical analysis, the specifically nominative singular masculine ending -us is opposed to the default nominative/accusative

136


ending -um.1 What looks like the spread of the accusative form to the nominative is simply the emergence of the unmarked form. This alternative analysis will work for examples of what Stump (2001) calls unidirectional syncretism, that is directional syncretism where the directional effect seems to move in only one direction (accusative ! nominative in the above example). However, Stump (2001) identifies another type, bidirectional syncretism, where the directional effect seems to move in two directions. This does not readily lend itself to the same analysis. It will be useful here to distinguish between two types, which we will call convergent bidirectional syncretism and divergent bidirectional syncretism, to be defined below. Each one presents a distinct problem for Zwicky’s model. 4.3.1 Convergent bidirectional syncretism Under convergent bidirectional syncretism, there is a feature value x which takes the form associated with feature value y in some contexts, and in other contexts takes the form associated with feature value z. Some examples have already appeared in Chapter 3, involving the accusative in Russian and in Bonan, and the ergative in Lak (all in x3.1.2) and the second person singular in Gujarati (x3.2.3). Bonan presents an especially clear and straightforward example, and this is what we analyse below. Recall that in Bonan (the paradigms are given again in (18)), the accusative case does not have a distinct form: in nouns it is syncretic with the genitive, while in pronouns it is syncretic with the dative-locative (a single case in Bonan). (18) Bonan (Todaeva 1997: 35) noun ‘foliage’ NOM GEN ACC DAT-LOC ABL INS-COM

1

labcˇoN-Ø labcˇoN-ne labcˇoN-ne labcˇoN-de labcˇoN-se labcˇoN-Gale

pronoun ‘he’

!

ndzˇaN-Ø ndzˇaN-ne ndzˇaN-de ndzˇaN-de ndzˇaN-se ndzˇaN-Gale

One could construe the arbitrary class ‘X’ as some kind of natural class, e.g. ‘direct case’, but this does not affect the argument.


137

Seen in directional terms, it looks as if the accusative takes the form of the genitive in nouns, while in pronouns it takes the form of the dativelocative. The following represents a possible description using directional rules. There are two rules of referral, shown in (19). In nouns, the accusative takes the form of the genitive (19a). With pronouns, the accusative takes the form of the dative-locative (19b). (19)

a. b.

ACC ACC

in nouns ¼ GEN in pronouns ¼ DAT-LOC

The relevant rules of exponence are shown in (20); note that no accusative forms are defined, because these are derived by the rules of referral in (19) (20)

a. NOM ¼ stem þ -Ø b. GEN ¼ stem þ -ne c. DAT-LOC ¼ stem þ -de d. ABL ¼ stem þ -se e. INS-COM ¼ stem þ -Gale

The interaction of the rules of referral with the rules of exponence is graphically represented in (21). In nouns, the rule of referral in (19a) causes the form of the genitive to be extended to the accusative, while in pronouns, the rule of referral in (19b) causes the form of the dative-locative to be extended to the accusative.

(21) noun ‘foliage’ NOM ¼ labcˇoN-Ø GEN ¼ labcˇoN-ne ACC ¼ DAT-LOC ¼ labcˇoN-de ABL ¼ labcˇoN-se INS-COM ¼ labcˇoN-Gale

pronoun ‘he’ NOM ¼ ndzˇaN-Ø GEN ¼ ndzˇaN-ne ACC ¼ DAT-LOC ¼ ndzˇaN-de ABL ¼ ndzˇaN-se INS-COM ¼ ndzˇaN-Gale

Now let us see how we might describe the same phenomenon without directional rules. The syncretisms are represented in (22) as the classes ‘X’ and ‘Y’. In (23), these symmetrical rules define the syncretic forms. (22)

a. {ACC ¨ GEN} ¼ X b. {ACC ¨ DAT-LOC} ¼ Y

138

(23)

The Syntax–Morphology Interface a. X ¼ stem þ -ne b. Y ¼ stem þ -de

Obviously, this is an incomplete description, because nothing yet has been said about the difference between nouns and pronouns. As it stands, the rules for X and Y conflict for the definition of the accusative form. This conflict is not resolved by Panini’s principle, because neither rule is more specific. Two options are available for clarifying the distribution of X and Y. One would be to incorporate rule ordering, with nouns and pronouns exhibiting different ordering for the rules involving X and Y, as sketched in (24): the rule for X precedes the rule for Y in nouns, while the rule for Y precedes the rule for X in pronouns.

(24) rule ordering for nouns i. X ¼ stem þ -ne ii. Y ¼ stem þ -de

rule ordering for pronouns i. Y ¼ stem þ -de ii. X ¼ stem þ -ne

In each case, the rule conflict is resolved in favour of the prior rule. An obvious objection to this approach is that it substitutes one formal device (rule ordering) for another (rules of referral). Whether this is a serious problem depends on the status one gives to rule ordering; for example, in approaches such as Stump’s (2001) Paradigm Function Morphology and Corbett and Fraser’s Network Morphology (1993), rule ordering is absent. The other possibility would be to incorporate information about the lexical class within the set of syncretic values (thanks to Jonathan Bobaljik for pointing out this option). In (25), the accusative is specified as nominal or pronominal.2

2

Of course, one could leave one of these rules underspecified for lexical class; the argument remains the same, though, since at least one rule will have to incorporate reference to lexical class.

Formal representation (25)

139

revision of (22)

a. {ACC noun ¨ GEN} ¼ X b. {ACC pronoun ¨ DAT-LOC} ¼ Y Technically, this would yield the correct output, but such rules seem to hold to the letter but not the spirit of a symmetrical analysis, in that they encode an obvious asymmetry. For example, (25a) states that form X serves by default for the genitive, and for one class of lexemes it is also used for the accusative. In its own fashion, this states that form X is primarily a genitive form, and secondarily an accusative form. In summary, the existence of convergent bidirectional syncretism makes certain demands on a model of morphology: they require rule ordering or directional rules (or something very much like them).

4.3.2 Divergent bidirectional syncretism As we have seen above, symmetrical rules can be used to describe convergent directional syncretism, provided other elements in the morphological model comply. As for divergent bidirectional syncretism, symmetrical rules simply cannot be used to describe it in any systematic fashion. Under divergent bidirectional syncretism, there is a feature value x which takes the form associated with feature value y in some contexts, while in other contexts y takes the form associated with x. For an illustration, we can return to the Latin example from (15), adding further data as shown in (26). In addition to the default masculine and neuter types of the second declension, there are a few nouns (the most prominent being vulgus ‘crowd’, virus, ‘poison’ and pelagus ‘sea’) which display an accusative in -us (Neue and Wagener 1902: 972).3 If we include these nouns in the picture,

3

The origin of this type is mixed. Pelagus is borrowed from the Greek s-stem pelagos, where -os is the final part of the stem, and not an inflectional ending. Clearly, however, when borrowed into Latin, it was interpreted as an ending, since pelagus was assigned to the second declension (with the ending -us), and not to the third declension s-stem type, where -us is instead the stem-final element (e.g. genus ‘nation, race’ and corpus ‘body’, where stemfinal -s is realized as -r when followed by endings, as in the genitive singular forms gener-is, corpor-is). This justifies our treatment of the -us in pelagus and the -us in servus as representing the same ending. Other borrowings from Greek s-stems followed this pattern on occasion, such as c etus ‘large sea animal, whale’ and chaus ‘chaos’ (Neue and Wagener 1902: 502–4). Vulgus and virus are native Latin items, and the origin of their exceptional declension is not known.

140


(26) Latin second declension


default neuter ‘war’

default masculine ‘slave’

bell-um bell-um bell-i bell-o bell-o

serv-us serv-um serv-i serv-o serv-o

accusative in -us ‘crowd’

!

vulg-us vulg-us vulg-i vulg-o vulg-o

then the Latin second declension is not simply an example of unidirectional syncretism, but rather has two mirror-image patterns: in the default neuter type, the nominative looks like the accusative, and in the small class represented by vulgus, the accusative looks like the nominative. Using directional rules, this sort of pattern does not differ in principle from unidirectional syncretism. To the rules from (16) we simply add a fourth rule (27d), stipulating that the accusative takes the form of the nominative for this small class. (27)

a. NOM SG ¼ stem þ -us b. ACC SG ¼ stem þ -um c. NOM SG in neuter ¼ ACC SG d. ACC SG in ‘vulgus’ type ¼ NOM SG

For symmetrical rules, divergent bidirectional syncretism imposes more constraints than does convergent bidirectional syncretism. Rule ordering is no longer an option, because the same values are syncretic both in the neuters and in the ‘vulgus’ type. For example, in (28) we represent the

This type is also defective, lacking plural forms, except for the occasional pelag e, whose ending is transparently Greek, unincorporated into the Latin declensional system. All of these nouns show a tendency to be reinterpreted as masculine nouns of the servus type, with nominative -us and accusative -um. One interesting variation occurs in Late Latin (sixth century), where pelagus is reinterpreted as masculine without altering its declension pattern: the example furentem pelagus ‘raging sea’ (from the Variae of Cassiodorus, cited in Neue and Wagener (1902: 503)) has the agreeing participle with the non-syncretic masculine accusative ending -em. This shows that nominative/accusative syncretism in the singular was not necessarily tied to neuter gender.


141

syncretic nominative/accusative forms, using the index ‘X’ to represent {NOM SG ¨ ACC SG}. (28)

a. {NOM SG ¨ ACC SG} ¼ X b. X in neuter ¼ stem þ -um c. X in ‘vulgus’ type ¼ stem þ -us

But what about masculines, which require both -us and -um endings? The description of a masculine noun would need some version of both rule (28b) and (28c). In (29) we show what these rules would look like with reference to the masculine declension added. (29)

Revision of (28) a. {NOM SG ¨ ACC SG} ¼ X b. X in {neuter ¨ masculine} ¼ stem þ -um c. X in {‘vulgus’ type ¨ masculine} ¼ stem þ -us

But this is unworkable: rules (29b) and (29c) conflict within the masculine declension. Rule ordering cannot resolve this, because either rule would bleed the other completely. The only way a symmetrical analysis can represent this pattern is to treat some aspect of it as accidental. One possibility would be to treat the syncretism of nominative and accusative as systematic, but to treat the identity of -us in the masculines and in the ‘vulgus’ type as accidental, by assuming two distinct but homophonous endings -us1 and -us2, as in (30). (30)

a. {NOM SG ¨ ACC SG} ¼ X b. X in neuter ¼ stem þ -um c. NOM SG in masculine ¼ stem þ -us1 d. X in ‘vulgus’ type ¼ stem þ -us2

This analysis treats the paradigms as an example of unidirectional syncretism (between instances of -um forms), while the apparent identity of the endings in servus and vulgus is treated as an accident. In effect, this amounts to a denial that there is such a thing as divergent bidirectional syncretism. Alternatively, one might treat the identity of endings across different declension classes as systematic but treat the different instances of the same case syncretism as accidental. That would be the consequence of recapitulating the analysis sketched above in (25), where the syncretic set of values includes reference to the lexical class. This is shown in (31), where the class

142


X describes the distribution of all instances of ‘stem þ -um’, and class Y describes all instances of ‘stem þ -us’. (31)

a. {NOM SG neuter ¨ ACC SG} ¼ X b. {NOM SG masculine ¨ ACC SG ‘vulgus’ type} ¼ Y c. X ¼ stem þ -um d. Y ¼ stem þ -us

On this analysis, the fact that both X and Y involve syncretism of nominative and accusative is treated as accidental.4 Because of what they imply for the modelling of syncretism, examples of divergent bidirectional syncretism are particularly important, so it is worth considering some additional data. So far, the only examples which seem to have been adduced in the literature are those found in Stump (1993, 2001), namely the accusative/dative singular in Old Icelandic, the genitive singular/nominative plural in Russian and the first singular/third plural in Romanian. Some of these examples have been subject to criticism by Feldstein (2003) and Wunderlich (2004), who argue that these patterns are accidental and hence not represented in the morphology. However, the Latin example above cannot be too readily dismissed, and below we adduce some further examples which appear to be good examples of divergent bidirectional syncretism. Classical Arabic has three cases: nominative, genitive and accusative (see (32)). In the default pattern (singulars and the broken plurals, which are formed by an alternation in the vocalic/syllabic pattern) all three cases have distinct endings: -u, -i and -a. In the so-called sound plurals (formed by suffixation), genitive and accusative are syncretic, marked by the ending -i:, which corresponds to the distinct genitive of the default type.5 The so-called diptotic declension pattern (displayed by certain adjectival stems, some broken plurals and some personal names; the name reflects the fact that there are two rather than three case forms) likewise has a syncretic genitive/accusative, but the ending is -a, corresponding to the distinct accusative of the default type. The diptotic pattern is found only with indefinites; the definite form and the construct form (head of an adnominal construction) revert to the default pattern. 4

5

This is unfortunate, in as much as there is an important generalization that applies to all nouns with nominative/accusative syncretism in the singular, namely that they belong to the neuter gender, taking neuter agreement (see the preceding footnote). The dual likewise displays genitive/accusative syncretism; e.g. nominative dual, construct state mu’min-a:, genitive/accusative mu’ min-ay ‘believers.’


143

(32) Classical Arabic noun declension (Fischer 1997: 196) triptotic (default) pattern

plural ‘believers.PL’ (construct) NOM GEN ACC

mu’min-u: mu’min-i: mu’min-i:

‘believer’ (construct)

‘black one’ (construct)

diptotic ‘black one’ (indefinite)

mu’min-u mu’min-i mu’min-a

’aswad-u ’aswad-i ’aswad-a

’aswad-u ’aswad-a ’aswad-a

!

The plural endings -u: and -i: are assumed to derive from lengthening of the corresponding singular endings (Kienast 2001: 143). The origin of the diptotic pattern is unclear; Kienast (2001: 142) cites Brockelmann’s (1908–13) theory that it started among personal names: some names ended in -u (e.g. Iazi:du) and some in -a (e.g. Sˇammara), and these were reinterpreted as fragments of a case paradigm, which was then fleshed out. This proposal assumes that the diptotic endings -u and -a are etymologically distinct from the corresponding case endings -u and -a, but that at some later point the two sets were equated with each other. As with the Latin example in (26), the analytical problem here is that, if we say that both -i and -a are genitive/accusative, how can the two endings be combined in a single paradigm? On the other hand, using directional rules, one could say that the accusative takes the form of the genitive in sound plurals, and that the genitive takes the form of the accusative in diptotic nouns. A further example is found in the Pama-Nyungan language Diyari. Dual and plural nouns and pronouns, and female personal names, have an accusative ending -n5a which is distinct from the absolutive (see (33)). In male personal names the ending -n5a is found in both the absolutive and the accusative. A zero ending (the bare stem) is found in the ergative/absolutive of non-singular pronouns, the ergative of non-singular nouns, and the absolutive/accusative of singular nouns. As with the previous examples, the problem faced by symmetrical rules is posed by the overlapping range of the inflectional forms. The zero ending ranges across all three core cases (ergative, absolutive, accusative), while -n5a is found in both the absolutive and the accusative. Using symmetrical rules, we would associate {ERG, ABS, ACC} with -Ø and {ABS, ACC} with -n5a. Since the two rules overlap in the absolutive and accusative, they cannot coexist in the same paradigm. Again, directional rules are the

144


(33) Diyari noun declension (Austin 1981: 47–50, 61) I ERG ABS ACC LOC ALL DAT ABL

-(ya)li -Ø -Ø -n5i -ya -ya -ndu

II

-li -Ø -n5a -Nu -Nu -n?i -Nundu

!

III

IV

V

-Ø -Ø -n5a -Nu -Nu -n?i -Nundu

-ndu -ni -n5a -n5aNu -n5aNu -n5aNka -Nundu

-li -n5a -n5a -Nu -Nu -n?i -Nundu

!

I singular nouns II non-singular nouns, non-singular third person pronouns, singular pronouns III non-singular first and second person pronouns IV female personal names, singular pronouns6 V male personal names

only device that can give -Ø and -n5a a unified representation across all of the paradigms. One possible analysis is to identify -Ø as the absolutive ending and -n5a as the accusative ending. In type I, the accusative takes the form of the absolutive, in type III, the ergative takes the form of the absolutive, and in type V, the absolutive takes the form of the accusative.7 (In type IV a distinct absolutive ending is found.) 4.3.3 Symmetrical versus directional rules: a summary. Zwicky (2000) has argued that symmetrical rules are sufficient to generate the surface effect of directionality, thus making directional rules unnecessary. However, we have shown that the rejection of directional rules brings real consequences in the range of facts that can be described (further arguments may be found in Evans, Brown and Corbett 2001, using somewhat 6

7

Singular pronouns, besides the third person non-feminine, display this pattern in the ergative, absolutive and accusative, while in the remaining cases they behave like nonsingular pronouns, with the element -ka- preceding the endings. The third person nonfeminine behaves essentially like a non-singular noun, with ergative n5ulu, absolutive n5awu, accusative n5in5a, and remaining cases based on the stem n5u ka- (Austin 1981: 61). A plausible alternative in this case would be to combine symmetrical and directional rules by viewing -Ø as the default core case ending. Types I–IV, then, follow a familiar split ergative pattern, with a distinct ergative in some paradigms and a distinct accusative in others. In this case, only type V requires a directional rule. In either case, though, directional rules are required.


145

different evidence). Although symmetrical rules can describe unidirectional syncretism, convergent directional syncretism can be described only by substituting another formal device (rule ordering), or by combining lexical and morphosyntactic information within a single syncretic value, which is undesirable on at least some approaches. Divergent bidirectional syncretism cannot be systematically described with symmetrical rules, requiring that some element of the pattern be treated as accidental. 4.3.4 Ranked constraints as an alternative to directional syncretism Alongside Zwicky’s (2000) discussion, the argument that directional effects can be derived without using directional rules has come from the perspective of Optimality Theory (OT). The issue is addressed most directly by Wunderlich (2004), who offers an OT alternative to directional rules. Most relevant is his treatment of what we have termed convergent syncretism, for which he analyses the nominative/accusative genitive/ accusative alternation of Russian, described in x3.1.1 and x3.1.2. To recapitulate: for most declension classes, the accusative case does not have a distinct form. In most classes, in inanimate nouns it appears to take the form of the nominative, and in animate nouns it appears to take the form of the genitive. For the present purposes, it will suffice to consider the singular of o-stems (which show the animacy-based alternation) and i-stems (which do not), as shown in (35). (34)

NOM ACC GEN

Nominative/accusative genitive/accusative alternation in Russian a.‘table’

b.‘student(M)’

c.‘mother’

stol stol stola

student studenta studenta

mat0 mat0 materi

The feature values that Wunderlich proposes for the morphosyntactic cases are given in (35). The nominative is simply unmarked. The other cases are rendered in terms of semantic roles: the accusative is [(þhr)v] (‘there is a higher role (verbal)’), which expresses its subordinate role within a transitive verbal clause, while the genitive is [(þhr)N] (‘there is a higher role (nominal)’), which expresses its subordinate role within a nominal clause.

146


(35) case name

feature value

nominative accusative genitive

[] [(þhr)v] [(þhr)N]

i.e. ‘there is a higher role (verbal)’ i.e. ‘there is a higher role (nominal)’

The relevant case endings are given in (36). The i-stem genitive -i is fully specified in terms of the underlying morphosyntactic features. The o-stem ending -a is underspecified: the value [(þhr)], which omits reference to nominal or verbal contexts, represents genitive/accusative together as a natural class. The default ending, -Ø, has no feature specifications.

(36) o-stem ending -Ø -a

feature value [] [(þhr)]

i-stem ending -Ø -i

feature value [] [(þhr)N]

The link between the underlying morphosyntactic features and the endings is effected by three ranked constraints: *

* *

Compatibility: the categorical specification of input and output must match. That is, genitive case (with a nominal specification) cannot be used where the input requires accusative case (with a verbal specification), and vice versa. *(þhr)/V inanimate: do not mark accusative case for inanimates. Max (þhr): if the feature (þhr), common to the accusative and genitive, is present in the input, it must be realized in the output.

The interaction of the above elements is shown in the tableaux in (37)–(39). In animate o-stems, the constraint Max (þhr) blocks the ending -Ø, because it is not specified for (þhr), thus causing the genitive/accusative to be selected. In the inanimate o-stems, however, the genitive/accusative is blocked by the constraint *(þhr)/V inanimate, so the unmarked nominative is selected, in spite of its violation of Max (þhr). In the animate i-stems, Compatibility


147

blocks the use of the overtly genitive ending -i for the input accusative, so the nominative is selected; the issue of animacy does not even arise. (37) accusative of animate o-stem input: [(þhr)v], animate

Compatibility

*(þhr)/V inanimate

student-Ø [ ]

Max (þhr) *!

F student-a [(þhr)]

(38) accusative of inanimate o-stem input: [(þhr)v], inanimate

Compatibility

*(þhr)/V inanimate

F stol [ ]

Max (þhr) *

stol-a [(þhr)]

*!

(39) accusative of animate i-stem input: [(þhr)v], animate

Compatibility

F mat0 [ ] mater-i [(þhr)N]

*(þhr)/V inanimate

Max (þhr) *

*!

To the extent that the factual coverage is the same, the choice between this analysis and one which employs directional rules (e.g. Corbett and Fraser 1993) has significance only within the context of the theoretical and descriptive programme of the individual investigator. Wunderlich’s (2004) analysis of Russian is part of a larger model of the differential marking of objects and agents, whereby arguments high on the animacy/prominence scale tend to mark the accusative, and arguments low on this scale tend to mark the ergative. Whatever the cross-linguistic insights of this approach, it fails to capture some fairly striking generalizations that obtain within the languages actually under analysis. In the Russian example, the genitive/ accusative syncretism displayed by the animate accusative is attributed to underspecification of the ending, so it is, in effect, a lexical idiosyncrasy. But it is not just one ending which needs to be underspecified, but rather at least seven (the four which are treated by Wunderlich, plus three distinct adjectival

148


and pronominal endings). This is why the animate accusative in Russian and other Slavic languages has long been used as an example of systematic syncretism, and particularly of directional rules (starting with Perlmutter and Oresˇ nik 1973). The alternative as offered by Wunderlich is to treat the sevenfold repetition of this pattern as purely accidental.8 However, Wunderlich’s analysis differs from a directional approach not only in its theoretical goal, but in its factual coverage as well, at least in the form in which it is presented. In effect, the analysis involves symmetrical rules plus rule ordering, as sketched above in (24); in this case, constraint ranking substitutes for rule ordering. The combination of lexical specifications and constraints state, in effect, that a rule for genitive/accusative precedes a rule for nominative/accusative. Under certain conditions the genitive/accusative rule is blocked, allowing the nominative/accusative rule to be first. Naturally, such a model can easily describe unidirectional syncretism as well. But, as we suggested above, it cannot describe divergent bidirectional syncretism. Wunderlich raises the issue but does not propose an analysis, instead rejecting Stump’s (2001) purported example of divergent bidirectional syncretism. However, other examples could be offered in its place (as was done above in Chapter 2: x2.4 and x2.5). Thus, the empirical problem created by divergent bidirectional syncretism remains. However, the difficulty caused by rejecting directional rules is not really a problem for Wunderlich’s model, since it in fact contains them, even though they are not directly exploited. To demonstrate this, let us review some of the key points in the above analysis. The core of the morphological model consists of three elements: the input, the output and the Max constraint 8

A similar atomization results when we consider another of his proposals, namely that the constraint *(þhr)/V inanimate can generally be used to account for the nominative/accusative syncretism typical of Indo-European languages. Recall that this constraint blocks the use of the accusative ending for inanimates ( neuters), which instead use the default form (informally, the nominative). But consider Latin once again. It is true that this analysis is possible for the singular of some nouns of the third declension, examples of which were given in Chapter 3 (x3.1.2): victor ‘conqueror’ versus aequor ‘sea’). The nominative singular of masculines is the bare stem, and the accusative is the stem plus -em. In neuters, the nominative/accusative is the bare stem, which could easily be described as the result of the omission of the accusative ending. But such an analysis cannot be applied to the second declension nouns described above in (26). The neuter is distinguished from the masculine by the lack of the nominative singular ending -us. Undoubtedly, this could be accounted for by some constraint, but, equally undoubtedly, this is not the same constraint *(þhr)/V inanimate found in the third declension, which blocks the accusative. Thus, the observation that neuter nouns in Latin have nominative/accusative syncretism is reduced to the concatenation of two apparently unrelated facts.


149

which mediates between them. The input is fully specified (accusative, i.e. [(þhr)v]), while the Max constraint is underspecified (genitive/accusative, i.e. (þhr)). The behaviour of the output form varies: in o-stems it is treated as underspecified, in i-stems it is treated as fully specified, as shown in (40). (40)

Analysis of the Russian accusative per Wunderlich (2004)

input output Max

o-stems

i-stems

[(þhr)v] fully specified [(þhr)] underspecified (þhr) underspecified

[(þhr)v] fully specified [(þhr)v] fully specified (þhr) underspecified

However, this variation in the lexical specification of the output forms makes no difference in the selection of the winning candidate. Both the underspecified output and the fully specified output satisfy the underspecified Max constraint. Thus, the tableau in (41) corresponds to those in (37)–(38), and represents symmetrical syncretism: there is a syncretic genitive/accusative form, which can be used for the accusative (as well as the genitive). The tableau in (42) corresponds to (39) and represents directional syncretism: the genitive form can be used for the accusative (as well as the genitive). (41)

Tableau with underspecified output

input: ACC [(þhr)v] NOM

[]

Max (þhr) *!

F GEN/ACC [(þhr)N] (42)

Tableau with fully specified output

input: ACC [(þhr)v] NOM

[]

Max (þhr) *!

F GEN [(þhr)N]

Thus, the Max constraint defines a syncretic class of values, while the absence of a distinct accusative in the output produces a paradigmatic gap, which is filled by the best available candidate. The effects of this directional

150


rule do not surface in this particular instance in Wunderlich’s analysis, because they are blocked by a higher-ranking constraint (Compatibility), which bars the use of the genitive for the accusative. But directional rules are nonetheless inherent in the model and, indeed, are implied wherever a fully specified output interacts with an underspecified Max constraint. Thus, Wunderlich’s model cannot be seen as offering an alternative to directional rules, because it already contains them. 4.4

Possible constraints on syncretism

In the preceding section we have looked at the properties inherent in different descriptive mechanisms. In this section we look at how these have been employed to describe syncretism, reviewing some representative proposals. More particularly, we look at their restrictiveness, as this is a central concern of most approaches: it is generally assumed that an adequate model will distinguish between possible and impossible patterns of syncretism. Below we evaluate these claims in the light of patterns surveyed in Chapter 3: x3. 4.4.1 Variants on hierarchical structures A number of proposals have been, in effect, claims that feature values are hierarchically structured, and that syncretism is necessarily a reflection of this structure. The most restrictive version is offered by Williams (1981, 1994), which does not make use of the elsewhere condition. As we have shown in Chapter 2: x2.1, this implies that if there are multiple syncretic patterns in a language, they will be nested within each other, which Williams (1994) expresses with the claim that ‘when there are multiple related paradigms, there will be one instantiated paradigm, and all others will have its syncretic structure, and perhaps some more. But no other related paradigm will have a contrary syncretic structure, making distinctions where that one does not’ (p. 27). Chvany (1986) and Johnston (1997) argue for loosening this model by, in effect, allowing underspecification, as described above for Noon (2) and Gaagudju (6)–(7) The effect is to allow ‘contrary syncretic structures’, but only in as much as they can be described given a single linear order of features.9 McCreight and Chvany (1991) take the logical next step and extend this proposal along multiple dimensions.

9

Plank (1991) observes that the quest for a single invariant linear order that would capture all the major syncretic patterns of a language is implicit throughout the history of grammatical description.


151

For example, given a paradigm delimited by two features and drawn as a grid, only those cells which can be brought next to each other can be described as syncretic, as in (43a), while (43b) is an impossible pattern. (43) a. X Y Z

x

y

a a b

a c d

b. X Y Z

x

y

a b c

b d e

These models remain agnostic about the actual content of feature structure; what they constrain is the co-occurrence of multiple patterns of syncretism involving the same feature values. All of the proposals leave a good amount of material unaccounted for. Thus, what Williams (1994) describes as ‘contrary syncretic structure’ is in fact common, especially with case, person and gender. The more relaxed model of Chvany (1986) and Johnston (1997) largely holds for case (which they were designed to describe) and person, but not for gender, as seen above in x4.2.1. McCreight and Chvany’s (1991) model, which is specifically designed to handle the interaction of multiple features, cannot describe polarity effects (see Chapter 3: x3.7). Finally, these proposals are all predicated on symmetrical rules, so the observations in x4.3 above apply. A less restrictive model of linear ordering is offerred by Plank (1991). In a survey of case syncretism in several Indo-European languages, he notes that although strict linear ordering will not capture all the syncretic patterns for any of the languages, the deviations from linear ordering are fairly minimal, so that cases which are syncretic with each other can at least be represented as adjacent on a two-dimensional model. 4.4.2 Carstairs (1987), Carstairs-McCarthy (1998a, b) Carstairs (1987) divides inflectional homophony (‘syncretism’ in the sense we have been using it) into two types: *

Syncretism. This occurs where the feature whose value is neutralized and the feature whose value is the determining context are realized simultaneously (the inflectional marker exhibits cumulative exponence). For example, dative/ablative plural in Latin constitutes syncretism for Carstairs because the neutralized feature (case) and the conditioning feature (number) are realised simultaneously by

152


*

the endings -is, -ibus, which cannot be broken up into distinct case and number morphemes. This is the most common and diachronically stable type of inflectional homophony. Take-overs. These occur where an inflectional marker that realizes the morphosyntactic value a in one context realises both values a and b in some other context.

He proposes the Systematic Homonymy Claim (Carstairs 1987: 123), which states that all inflectional homophony is either (i) a syncretism or (ii) a takeover, where the conditioning feature is lower on the relevance hierarchy than the feature whose values are neutralized. The relevance hierarchy, proposed by Bybee (1985), ranks inflectional features on their semantic relevance to the meaning of the root, as manifested in cross-linguistic frequency and position. Carstairs cites the following as contexts for take-overs, where the features to the left of the arrows are higher on the relevance hierarchy: (44) * * *

tense, mood > person-number definiteness > person-number number > case

The following schematic noun paradigms illustrate what is possible under this system. Paradigms (45) and (46) both conform to the Systematic Homonymy Claim. Syncretism (in Carstairs’ sense) is represented in (45): genitive and dative are syncretic in the plural, expressed by the form g, which simultaneously realizes case and number. The paradigm in (46) illustrates a take-over, with homophony of the genitive singular and genitive plural. First, there is separate realization of number (p) and case (a–d). Second, the neutralized feature (number) is higher on the relevance hierarchy than the conditioning feature (case).

(45)

(46)

NOM ACC GEN DAT

SG

PL

a b c d

e f g g

(Carstairs 1987: table 4.15, 113)

NOM ACC GEN DAT

SG

PL

a b pc d

pa pb pc pd


153

The paradigms in (47) and (48) represent impossible, or at any rate unsystematic, patterns. The paradigm in (47), with syncretism of the genitive and dative in the plural, is not possible, because, although the distribution of the genitive marker c makes it look like a take-over, the neutralized feature (case) is not higher on the relevance hierarchy than the conditioning feature (number), in violation of the Systematic Homonymy Claim. Finally, the paradigm in (48) is not possible, because it is not a take-over (e does not serve a non-syncretic function elsewhere), nor is it a syncretism, because number (p) and case (a–d) are realized separately. (47)

NOM ACC GEN DAT

(48) SG

PL

a b c d

pa pb pc pc

(Carstairs 1987: table 4.10, 111)

NOM ACC GEN DAT

SG

PL

a b c d

pa pb pe pe

(Carstairs 1987, table 4.11, 110)

Before considering the important revisions to this model made by Carstairs-McCarthy (1998a, b), let us consider the predictions made so far. The statistical predominance of what Carstairs-McCarthy terms syncretism is undeniable. What we have termed directional effects – Carstairs’ take-overs – are observable only in a minority of cases. Carstairs (1987: 111–12) states that this follows from the reasonable assumption that the normal state of affairs is for an inflectional marker to have a consistent function. Thus, in (45), g is always genitive/dative plural, while in (46), p is usually plural but then sometimes it is singular/plural. However, the banning of (47) and (48) seems less well motivated. The sort of pattern in (47) is not uncommon; we shall return to it below. On the other hand the pattern in (48) is indeed rare, though we can cite an example involving case in the Dravidian language Pengo (49). (The initial d of the singular endings is the post-consonantal variant of t and n; compare the vocalic stem singular doNri, doNri-tin, doNri-to, doNri-ni, doNri-taN ‘mountain’.) Though this pattern is rare, that fact may follow from more general principles of morphological organization than from any specific constraint on syncretism. Where features are realized separately, feature conditioned allomorphy is the exception rather than the rule. The rarity of (48) is then a facet of the rarity of the sort of pattern in (50), where case and

154


(49) Pengo noun declension, ‘hand’ (Burrow and Bhattacharya 1970: 32–3)

NOM ACC-DAT LOC GEN INS-ABL

singular

plural

key key-diN key-do key-di key-daN

key-ku key-ku-kaN key-ku-kaN key-ku-ni key-ku-taN

(50)

NOM ACC GEN DAT

singular

plural

a b c d

pa pb pe pd

number are marked separately, but the genitive, and only the genitive, has a distinct plural allomorph. Carstairs-McCarthy (1998a, b) offers a somewhat different analysis of these phenomena, which result in a somewhat different range of predictions. The underlying principle is that inflectional meaning should be governed by the same constraints that obtain for lexical semantics, which leads to four axioms, summarized below from Carstairs-McCarthy (1998b): A. Lexical items do not contain meanings consisting of incompatible disjuncts, e.g. *‘apple OR banana’. By the same token, the meaning of an inflectional marker should not contain incompatible disjuncts consisting of competing values for the same feature, e.g. *‘ablative OR locative’. (Carstairs-McCarthy assumes that feature structure is flat.) B. Lexical items do not have meanings containing negations, e.g. *‘not apple’. By the same token, inflectional markers should not contain negations in their meanings, e.g. *‘not past’. C. Lexical items may contain compatible disjuncts. For example, the different senses of climb in (a) the boy climbed up the tree


155

(b) the boy climbed down the tree (c) the snake climbed up the tree can be reconciled by characterizing its semantics as ‘go, upward OR clambering’. Sentence (b) contains only the element ‘clambering’ and sentence (c) only the element ‘upward’, but the two meanings are compatible with each other, as witnessed by sentence (a). (Carstairs-McCarthy takes this example from Jackendoff 1985.) D. Special vocabularies may exist where the meanings of otherwise distinct lexical items may be collapsed; Carstairs-McCarthy cites the example of the Dyirbal ‘mother-in-law language’, where, for example, the normal terms nudin ‘cut deeply’ and gunban ‘cut less deeply’ are combined in the single term dyalnggan ‘cut’. By the same token, where distinct allomorphs of inflectional markers are found in special contexts, for example where case marking is cumulated with plural marking in plural contexts, related meanings which elsewhere have distinct realizations may share the same realization. E. No rule can make overt reference to the unmarked value of a feature. Axiom D now accounts for ‘syncretism’ as characterized in Carstairs (1987) and is defined as inflectional homophony that occurs under conditions of allomorphy. Cumulative exponence, by definition, provides a context for allomorphy, but is not itself the crucial factor – for example, distinct inflection classes are also relevant contexts.10 Another property of syncretism under axiom D is that the neutralized meanings be related (p. 292); however, since feature structure is construed as flat (p. 294), relatedness of meaning cannot be formalized, so it is not clear what constraints emerge from this. The other axioms account for directional effects, largely replacing takeovers with underspecification. Because of axiom E, which states that the unmarked value is, quite literally, unmarked, underspecification for any value always entails homophony with the unmarked value. This yields two types of directional effect, depending on whether axiom C is invoked. Where axiom C is not involved, it is predicted that the form associated with the unmarked value will prevail in unmarked contexts, as in 10

If, as Carstairs-McCarthy suggests (1998b: 290), inflection class is to be construed as part of the meaning of an inflectional marker, then this too constitutes cumulative exponence.

156


example (51), taken from the Chibchan language Ika. First and third person singular are syncretic in all tense paradigms except the distal past. The syncretic 1SG/3SG form has no overt person-marking affix, which makes it look like the 3SG form of the distal past. If we assume that third person and ‘elsewhere’ are the unmarked values for person and tense, respectively, this is what the axioms should yield, since the form associated with the unmarked third person value is extended in the unmarked context. (51)

Ika (Frank 1990) distal past

1SG 2SG 3SG

(stem) -rua nL-(stem) (stem)

elsewhere -na -na -na

(stem) nL- (stem) (stem)

This can be represented by the rules in (52), which are in accordance with the axioms above. The ending -rua is specified as the first person distal past ending, nL- as the second person prefix, unspecified for tense, and Ø is the general default, unspecified for tense and person.

(52) -rua nLØ

1 distal past 2

Carstairs-McCarthy (1998b) illustrates the effects of axiom C with an example from Hungarian conjugation (53), which had been described in Carstairs (1987) as a take-over of the 1SG indefinite by the definite in the past. (53) present 1SG INDF 1SG DEF

va´r-o-k va´r-o-m

past

!

va´r-t-a-m va´r-t-a-m


157

At issue is the distribution of the 1SG affixes -k and -m, which are isolated in (54). In the present, -k marks the indefinite and -m the definite, but in the past -m marks both. (54)

1SG INDF 1SG DEF

present

past

-k -m

-m -m

Carstairs-McCarthy attributes the distribution of the affixes to the rules: (55) -m -k

1 definite OR past 1

Since the values ‘past’ and ‘definite’ can co-occur, this is a compatible disjunction; -k is simply an elsewhere form. Although these rules result in neutralisation both of definiteness and of tense, at the level of the word, it is only the past tense forms which are identical. This is a secondary effect, resulting from the fact that there is a separate layer of affixes which distinguish the two tenses (the distinct theme vowels, plus the past tense marker -t-), thus breaking up the potential homophony between present and past. To summarize, Carstairs-McCarthy’s (1998a, b) model accommodates two types of non-accidental syncretism: (i) syncretism in the strict sense, involving a combination of values represented by a distinct allomorph, and (ii) underspecification. (The continued need for take-overs is acknowledged (1998b: 299) but not directly incorporated into the model.) Does this model allow for testable predictions? In the case of ‘syncretism’ (in the sense defined above), the prediction would appear to be that where we find an affix (or inflectional operation of some sort) which always combines some set of values wherever it is found, these values should be related to each other, that is they should have related meanings. However, without an explicit theory as to what constitutes

158


‘related’ meanings, we cannot readily assess this prediction. However, in the case of underspecification, the model implies two clear predictions about the behaviour of directional effects: (56)

I. Directional effects will always involve the unmarked value of the feature in question (because of axiom E). II. Directionality will be of two sorts; either:

a. The form associated with the unmarked value of some feature will prevail in the context of the unmarked value of some other feature.

–or– b. The form associated with the marked value of some feature will prevail in the context of the marked value of some other feature.

Assessing the validity of these predictions is complicated by the fact that they all revolve around the notion of markedness, which is itself hard to pin down with precision. Therefore, our observations can only be tentative. Nevertheless, working with commonly held assumptions about what should be construed as the unmarked value, we find no general support for these constraints. We summarize our observations below. Note that we are merely looking at whether the data conform, superficially, to the predictions. Whether the individual examples even admit of a comprehensive analysis in terms of Carstairs-McCarthy’s (1998a, b) model is a question we have not addressed. Case. Prediction I is met in all instances of what we have called ‘core case syncretism’ (Chapter 3: x3.1.2), which is indeed the most common type. The prediction is violated wherever directional effects can be observed which do not affect nominative or absolutive, and a number of these have been seen so far, or will be adduced below, e.g. genitive/accusative in Russian and Classical Arabic (see (32) above); ergative/locative in Koryak ((13b) in Chapter 3: x3.1.2) and the ergative/genitive in Lak ((14) in x3.1.2); the gentive/accusative dative-locative/accusative alternation in Bonan ((15) in Chapter 3: x3.1.2); and dative/illative in Erzja Mordvin ((18) in Chapter 3: x3.1.2). Within those examples which conform to prediction I, prediction IIa seems to be met in cases of differential object or agent marking, where accusative or ergative marking is absent for some set of nominals. Examples which relate to prediction IIb are hard to intepret. For example, the accusative form seems to prevail in Greek and Latvian ((3) in Chapter 3: x3.1.2); whether these conform to prediction IIb depends on whether one interprets neuter (in the case of Latin) and feminine (in the case of Latvian) as marked values.


159

Person. Directional effects involving person are outlined in Chapter 3: x3.2.3; the results are summarized in (57), where ‘U’ indicates unmarked and ‘M’ indicates marked. The symbol to the left of the slash (‘/’) stands for the syncretic value, the symbol to the right of the slash represents the context. Thus U/U means ‘the form associated with the unmarked person value is extended to a marked value in the unmarked context’, M/M means ‘the form associated with a marked person value is extended to the unmarked value in a marked context’, and so on.

(57)

Directional effects involving person (from Chapter 3: x3.2.3)

a. Burarra b. Callahuaya c. Carib d. Dani e. Dutch f. Gujarati g. Koiari h. Kongo i. Literary Kannada j. Murle k. Nobiin l. Old Icelandic m. Shinassha n. Suena o. Udihe

1/2

2/3

?/?

M/M M/? M/U M/M U/M U/?

1/3

U/? M/U ?/? M/M ?/M

U/M M/U

?/M M/M ?/?

If third person is construed as the default value, then prediction I is violated where directional effects do not involve third person (in Gujarati, Literary Kannada, Nobiin, Udihe and Shinassha). In evaluating prediction II, we have assumed that values such as singular, present tense, indicative and declarative constitute the unmarked values of the contextual features, though in a number of cases it is not obvious what markedness values to assume. Possible examples conforming to prediction IIa are found in Gujarati and Koiari: in both cases the third person form prevails, but it is not clear whether the context should be construed as unmarked. Examples conforming to prediction IIb come from Burrara, Dani, Shinassha and

160


Suena, while Callahuaya and Udihe are possible examples as well. Neither component of prediction II is met by Dutch or by Nobiin (since the unmarked third person form prevails in the marked context), nor by the Carib, Old Icelandic and Kongo examples (because the marked second person form prevails in the unmarked context). Number. Assuming a flat feature structure, we have found no good examples conforming to either prediction. Were one to admit a hierarchical features structure, with dual as the marked counterpart to plural, then the Koryak example cited in Chapter 3: x3.5 seems on the surface to conform to both predictions: the dual form prevails in certain transitive combinations, which may be construed as the marked context. However, one should recall that in Koryak, morphologically, it is the plural which is marked with respect to the dual (by the addition of the suffix -la-). Tense-aspect-mood. The convincing examples of directionality conform to prediction I but not to II. Thus Gapapaiwa, Tawala and Loniu (cited in Chapter 3: x3.6.2) all involve syncretism with what may reasonably be construed as an unmarked tense, in the context of marked person (nonthird person), but the form which prevails is the zero marking associated with the unmarked tense. The intuition behind Carstairs-McCarthy’s predictions is that directional effects can be attributed to underspecification, and that the possibilities for underspecification are themselves limited by markedness relationships. However, as we have seen, directionality does not seem to be generally predictable on the basis of the values of the features involved. CarstairsMcCarthy does allow the possibility that apparent counter-examples may be analysable as rules of referral (1998a: 18–19). In this case, the model makes no generally applicable predictions about directional effects, since there is no theory-independent way of identifying which phenomena should be construed as rules of referral. 4.4.3 Impoverishment Impoverishment (Noyer 1997, 1998, 2001) is a device allowing for the constrained alteration of feature values and plays a key role in some descriptions of syncretism. Bobaljik (2002) writes that ‘[i]mpoverishment . . . admit[s] of predictions about impossible syncretisms crosslinguistically, predictions that appear to be largely borne out.’ Specifically, impoverishment is designed to account for directional effects. As an illustration, Bobaljik adduces an example from Stump (1993), syncretism of the 2SG and 3SG in the two past tenses of Macedonian:


161

(58) Macedonian (Stump 1993) present 1SG 2SG 3SG 1PL 2PL 3PL

padn padne padne padne padne padn

-am -sˇ -me -te -at

aorist

imperfect

padna -v padna padna padna -v -me padna -v -te padna- -a

padne -v padne -sˇ e padne -sˇ e padne -v -me padne -v -te padne- -a

In Stump (1993) this example was used to illustrate a stipulated directional rule, namely a rule of referral. The syncretic form is interpreted as looking like the third singular, both for its zero person-number ending and for the absence of the past tense marker -v-.11 The rule of referral stipulates that 2SG takes the form of 3SG (i.e. 2SG and 3SG form a set, whose form is determined by the 3SG). Bobaljik offers instead an interpretation in terms of impoverishment, which illustrates its two key aspects: (i) deletion of features, and (ii) the assumption of unmarked or default feature values. In this case, impoverishment deletes the feature second person in the singular of the past tenses. This deleted value is replaced by the default value, construed as third person. What is not possible is the replacement of the third person form by the second, which would be perfectly feasible under Stump’s model. Thus, impoverishment makes two predictions: I. Directional effects will involve the unmarked value of a feature. II. The form associated with the unmarked value will prevail. Note that prediction I coincides with that of Carstairs-McCarthy (1998a, b), while II is the opposite. However, an assessment of the validity of these predictions is contingent on the particular model of feature structure employed, and this is an issue distinct from the device of impoverishment as such. As discussed above, different models of feature structure make different predictions, and so, the effects of impoverishment will differ according to

11

The absence of -v- in the 3PL forms has a different diachronic explanation from that of the 2SG/3SG form. Historically, the -v- descends from -x-. There is no evidence that -x- was ever present in the singular forms at any stage in the history of the Slavonic languages. Its loss in the 3PL is relatively recent (post-sixteenth century; Koneski 1996 [1986]: 92), due to the regular loss of intervocalic -x- in Macedonian.

162


the feature structure assumed. If we assume flat feature structure, then the observations above about Carstairs-McCarthy’s prediction I apply, and the observations about prediction II apply in the reverse: Carstairs-McCarthy’s positive examples are counter-examples to impoverishment’s predictions, and vice versa. More elaborate models of feature structure create more markedness relationships, and so more domains for impoverishment to operate in. But, as Carstairs-McCarthy points out (1998b: 288), this also opens up the range of surface effects that can be produced. He illustrates this with Noyer’s account of number syncretism in the verbal prefixes of the Papuan language Nimboran. The prefix k-, which is characteristic of the dual alone for first and third persons, is found for the plural as well in second person and first inclusive (59). (59) person

[þSG -PL] [-SG -PL] PL [-SG þPL]

SG

DU

1, 3

2, 1INCL

Ø ki-

Ø kk-

Noyer’s analysis is illustrated in (60). Number values are treated as compounds – significant for the present example is that dual and plural are each treated as further modifications of the value non-singular. Impoverishment works by deleting the plural value, yielding the underlying non-singular. Separately, the prefix i- is construed as overtly plural, while k- is the more general non-singular, yielding what looks like the ‘take-over’ of the plural by the dual in the second person and first inclusive.

(60) impoverishment [þPL] ! Ø/ [þ2 -SG]

inflectional rules i k

[þPL] [-SG]


163

But, as Carstairs-McCarthy points out, the reverse situation can just as easily be described by transforming the dual into non-singular, and treating i- as the generic non-singular marker, as in (61)–(62): (61) person

[þSG PL] [SG PL] PL [SG þPL]

SG

DU

1, 3

2, 1INCL

Ø ki-

Ø ii-

(62) impoverishment [PL] ! Ø/ [þ2 SG]

inflectional rules i k

[SG] [SG PL]

If, as Noyer assumes, dual is marked with respect to plural, then it would seem that impoverishment makes no consistent predictions in this instance about directional effects. 4.4.4 Stump (2001) and Zwicky (2000) Stump (2001) presents a model in which overt stipulation of syncretic sets of values plays a major role; Zwicky (2000) is a reaction to and adaptation of this model. Stump assumes flat feature structure, with no markedness asymmetries. He divides syncretism into three types, each with its own formal representation: i. Unstipulated syncretism. This is described by simple underspecification for the feature in question. Because of the assumptions about feature structure, this is effectively limited to instances where all the values of a feature are syncretic. ii. Unidirectional and bidirectional syncretism. These are described by rules of referral. The Bidirectional Referral Principle (p. 219) stipulates that any rule of referral also entails its inverse. That is, if

164

The Syntax–Morphology Interface a rule of referral states that value x of feature a takes the form of value y, this implies that value y takes the form of value x, all else being equal. iii. Symmetrical syncretism. As with rules of referral, this involves a stipulated set of values, but here the form is defined across the whole set rather than in terms of the individual members. In a sense, symmetrical syncretism can be defined negatively: it is syncretism which does not lend itself to being described by underspecification (because some but not all values of a feature are involved) nor by rules of referral (because no directional effects are in evidence).

These devices can describe any syncretic pattern. However, Zwicky 2000 argues that it is too complex, imputing underlying distinctions to phenomena which are, in his view, epiphenomenal. First, he rejects the distinction between unstipulated and stipulated syncretism: what Stump terms unstipulated syncretism is a convenient representational short cut, but no less stipulative than simply declaring the identity of all the values. Second, he rejects the distinction between directional and symmetrical syncretism: directional effects can be seen as a by-product of underspecification (see x4.4.3 above). Thus, Zwicky proposes retaining Stump’s symmetrical syncretism as the sole type. However, as pointed out above in x4.3 and its subsections, a symmetrical approach can only readily describe unidirectional syncretism: what we have described as convergent bidirectional syncretism cannot be described without assuming further devices, and divergent bidirectional syncretism cannot be described at all. Thus, there are inherent limits to what Zwicky’s model can describe. On the other hand, though Stump’s basic system can describe any pattern, he imposes a distinct constraint onto this, namely the Feature Ranking Principle, which reads: For any language ‘, there is a ranking > of morphosyntactic features in ‘ which satisfies the following condition: for every stipulated syncretism S in ‘, if the dominant properties of S include a specification of the feature Fd and the subordinate properties of S include a specification of the feature Fs, then Fd > Fs. (Stump 2001: 239)

In effect, this states that, within a given language, if you have stipulated syncretism of values of feature x in the context of feature y, then you cannot have stipulated syncretism of values of y in the context of feature x. Empirically, this represents a limit on the number of different syncretic


165

patterns that can be accommodated in a language, and it appears, on the whole, to be borne out. However, we know of one possible counterexample, namely case and number syncretism in Slovene, which seem to require stipulated syncretism of number in the context of case, and of case in the context of number. Slovene distinguishes three numbers and six cases. Number syncretism in the context of case is systematically found in nouns and adjectives, where dual and plural are syncretic for the genitive and locative. As we argue below in Chapter 5: x5.1.1, this pattern is best modelled as a rule of referral, as evidenced by the suppletive pattern of the word cˇlovek. Case syncretism in the context of number is widespread; in particular, the dual always has two syncretic forms, though the actual patterns may vary both across and within word classes. All nominals have dative and instrumental syncretism in the dual. Nouns and adjectives additionally show nominative/accusative syncretism.

(63)


Declension of Slovene ‘person’ singular

dual

plural

cˇlovek cˇloveka cˇloveka cˇloveku cˇloveku cˇlovekom

cˇloveka cˇloveka ljudi ljudeh cˇlovekoma cˇlovekoma

ljudje ljudi ljudi ljudeh ljudem ljudmi

Pronouns have accusative/genitive syncretism, while the behaviour of the locative dual varies. It never has a distinct form of its own in Slovene; the literary standard allows syncretism either with the dative/locative, or with the genitive (Toporisˇ icˇ 1976: 241–2).12

12

Genitive/locative syncretism in the dual is an archaism artificially introduced into the literary language in the nineteenth century (Tesnie`re 1925: 305). Nonetheless, it is given as the preferred option in such authoritative grammars as Toporisˇ icˇ (1976). It is interesting to note that the third person pronoun, whose declension has both pronominal and adjectival characteristics, has three options for its locative dual: it can be identical to the dative/ instrumental dual (njima), to the genitive dual (naju) or to the locative plural (njih) (Toporisˇ icˇ 1976: 241–2).

166


Declension of Slovene ‘I’


singular

dual

plural

jaz mene mene meni meni menoj

midva naju naju naju nama nama nama

mi nas nas nas nam nami

If we assume, along with Stump, that feature structure is flat, then one of the patterns of case syncretism in the dual must be stipulated, since underspecification for case in the context of number can only be invoked once. 4.5

Summary

4.5.1 Predictions and counter-examples The contrast is often made between a restrictive theory, which is able to make predictions about possible and impossible structures, and a descriptive framework, which is open-ended. The formal models we have reviewed above aim to provide a restrictive theory of syncretism. In no case, however, do the predictions correspond precisely to the empirical evidence – there is always a residue of counter-examples. On a literal-minded approach this would seem to invalidate all such attempts straightaway. However, we can assume that most investigators who have proposed constraints are aware of the possibility of counter-examples, so a more nuanced assessment of such proposals must focus on how these are treated. One way of dealing with exceptions would be to treat them as the result of accidental homophony, and thus beyond the reach of morphological analysis. A clear example of accidental homophony was cited above in Chapter 1 (x1.5), where vowel reduction in Russian leads to the homophony of the case-number endings -o and -a when unstressed. However, more often than not the examples we see are ambiguous, so that it is no more natural to interpret them as phonologically accidental than as morphologically systematic.13 More importantly, there is diachronic evidence that the boundary 13

As an illustration of the danger of such an approach, we can take Noyer’s (2001) treatment of syncretism (primarily of gender and number) in the non-Pama-Nyungan Australian language Nunggubuyu, where the formal model forces systematic syncretism to be treated


167

between accidental and systematic is fluid: syncretic patterns which arose as by-products of regular sound change may be reinterpreted as morphologically systematic, and then extended to environments that were not subject to the original sound change. Sologub (1983) discusses such an example from Russian. In Standard Russian, the genitive singular of a-stem nouns is distinct from the dative/locative: genitive singular zˇeny ‘wife’ versus dative/ locative singular zˇene. In some dialects, the two forms have collapsed, resulting in genitive/dative/locative singular zˇeny or zˇene (depending on the dialect). This does not follow from any sound change but is rather the result of analogy with the i-stems (as in NOM/ACC kost 0 , GEN/DAT/LOC kosti ‘bone’; see (90) in Chapter 3: x3.8.2), where this collapse was phonologically regular (p. 86). Other examples that have been regarded as the morphological extension of originally accidental patterns include some forms cited in Chapter 3, namely the Dhaasanac B forms (examples (83) and (84) in x3.7.2), the second singular / third singular present in Old Icelandic (x3.2.4), the first singular / third singular present in Livonian (x3.2.4), as well as the genitive singular/ as the concatenation of separate instances of accidental homophony. Nunggubuyu shows systematic syncretism of first exclusive and second person non-singular in person marking on verbs, alongside a non-syncretic first inclusive form. This occurs in one of the two sets of subject prefixes (the choice is determined by tense, aspect and polarity (negation)), and in the object prefixes. According to Noyer, these two person values do not constitute a natural class (indeed, it is hard to imagine how they would; see the discussion at the end of x3.2.2). Instead, he treats subject-and object-person syncretism as two separate instances of accidental homophony. Syncretism of the subject markers is construed as the accidental by-product of a morphological rule. The rule itself is not given, but the derivations (Noyer 2001: 811) are given below, where nV:- is construed as the first person non-singular subject marker, nV- is the second person non-singular marker, and wu- as the non-singular intransitive subject marker. (Note that in Noyer’s analysis, nV- is in fact construed as unmarked for number; its function as a non-singular marker is due to the existence of a series of overtly singular second person markers.) underlying form 1DU F 2DU F 1DU M 2DU M 1PL 2PL

nV:- NinV- wu- NinV:- ninV- wu- ninV:- runV- wu- ru-

surface form ! ! ! ! ! !

ni:-Nini:-Nini:-Nini:-Ninu-runu-ru-

The syncretic 1PL/2PL object markers have the form na-. This is assumed to result from the existence of two homophonous prefixes: second person non-singular nV- (as seen above) and first person non-singular object nV- (Noyer 2001: 760).

168


accusative plural in Common Slavonic (Meillet and Vaillant 1934: 398) and the collapse of singular and plural in neuter nouns in Old High German (Wurzel 1987: 69). If such diachronic interpretations are held to be valid, then we must admit the possibility that accidental homophony may be reinterpreted as morphological systematicity. This is not to claim that all apparent instances of syncretism must necessarily be treated as morphologically systematic, but rather that such an interpretation should be available for any apparent instance, to be examined on its individual merits rather than excluded a priori. A formal model which instead dismisses certain phenomena from the outset can only be a poor representation of linguistic reality. A more inclusive approach would characterize the counter-examples not as accidental, but rather as ‘marked’. That is, the morphological model can produce them, but only in a complex or indirect fashion. This approach is subject to the same observations made above: it is a covert distinction, and so has no necessary empirical ramifications. While neither approach can successfully determine the possibility of a given pattern in a given language, they do imply statistical and diachronic claims. Systematic patterns are common, accidental or marked patterns are rare and diachronically unstable. Both these points invite caution. The causal connection between formal markedness and rarity is often assumed but remains undemonstrated (especially since formal markedness is not subject to direct observation). Further, as the examples in Chapter 3 show, the standard assumptions about what are common and what are rare patterns are in need of revision. The related issue of diachrony is also problematic. For example, Carstairs (1987: 128–31) gives the example of the syncretism of 1SG/2SG/ 3SG in the imperfect indicative in medieval dialects of Italian, which was resolved in Modern Italian by the creation of new, unsyncretic forms. This is taken as evidence that the original syncretic pattern was unsystematic. He observes, though, that unsystematic patterns may be diachronically persistent. Further, we may note that systematic syncretism may be resolved over time as well – consider the Old Nubian paradigm cited in (48) in Chapter 3: x3.2.4, where 2SG/3SG and 1PL/2PL are syncretic. This constitutes systematic syncretism in Carstairs’ (1987) terms, as it involves cumulative exponence, and its diachronic stability is attested by its retention over ten centuries (Browne 2002: 1) in contemporary Nubian languages such as Dongola Nubian (Armbruster 1960). But in the Nubian language Nobiin this syncretism was resolved through the creation of new,


169

distinct second person forms. What this seems to indicate is simply that homophony of any sort may be resolved over time; underlying systematicity plays no necessary role. 4.5.2 Towards a model of syncretism In this chapter we have shown that no attempt to impose formal constraints on the description of syncretism is wholly successful. On the other hand, there are some decided tendencies, remarked on throughout Chapter 3, most especially the prevalence of certain patterns of case, person and number syncretism, which have a plausible basis in the structure of the features in question. The challenge for morphological theory is to allow the formal model to be open-ended, while still giving an account of the fact that some patterns are common, and others rare. Below we suggest that the statistical distribution of patterns is a reflection of diachronic processes. Consider first the change undergone by Anejom, discussed in (45) in Chapter 3: x3.2.4, where a four-way person distinction (first inclusive and exclusive, second and third) and three-way number distinction (dual, trial and plural) was collapsed into a single form, etymologically the third plural. Though striking in its rapidity, the direction of change does not seem unexpected: third person substitutes for the other forms, and plural substitutes for dual and trial. The basis for the innovative system of twentieth-century Anejom need not be sought in language-specific morphological quirks. Rather, it follows from quite widespread assumptions about the unmarkedness of the third person with respect to other person values, and of plural with respect to other non-singular number values. Contrast this with the developments in the Cushitic language Dhaasanac, shown in (83) and (84) in Chapter 3: x3.7.2, where syncretism of the first plural exclusive form with the 2/3SG F form is extended from one class of stems (where it was the result of regular sound change) to all stems. This combination of values cannot plausibly be argued to have any basis in the structure of person or number. Rather, this accidental disjunction of values was taken over as a pattern by the morphology. The change in Anejom yielded a ‘natural’ result, while the change in Dhaasanac yielded an unnatural result. But in essence the two kinds of change are the same. The original paradigmatic space was reorganized according to a new morphological pattern. The difference lies in the source of the morphological pattern. In the case of Anejom, we can suppose its source lay in (universal?) properties of feature structure, whereby 3PL is unmarked with respect to other non-singular person-number values. Thus,

170


this morphological pattern is shared across most (all?) languages. In the case of Dhaasanac, the morphological pattern was created by a phonological change that affected one stem class. By definition, this morphological pattern is language-specific.14 The difference between natural (i.e. common) and unnatural (i.e. uncommon) syncretic patterns thus need not be reflected in the formal model. We suggest that contrast between these two diachronic routes is sufficient to account for the statistical predominance of morphosyntactically natural patterns. Paradigmatic morphological patterns based on common or universal elements of feature structure (i) are available to all languages, (ii) can arise spontaneously (as in Anejom) and (iii) are self-regenerating in case of disruptions. Morphological patterns based on phonological change are language-specific, and always in competition with morphological patterns based on feature structure. It only follows that the former type should be more widespread. On this view, even if we manage to construct a comprehensive model of feature structure that will account for some syncretic patterns, the possibility still remains that patterns may be codified which are independent of feature structure. The crucial question here is whether there are any constraints on the production of unnatural syncretic patterns. At present we do not have enough information to decide this question; as the discussion above will have shown, it can only be resolved empirically.

14

This distinction roughly corresponds to that between ‘system-independent’ and ‘systemdependent’ morphological naturalness, as discussed within Natural Morphology (Mayerthaler 1987, Wurzel 1987).

5 Formal framework and case studies

As we saw in the previous chapter, the challenge for a formal analysis of syncretism is twofold: to represent the cross-linguistically more general patterns of syncretism in terms of feature structure, and to account for patterns which occur independently of feature structure. In this chapter we lay out a formal framework and demonstrate it with three cases studies. When introducing the formal framework, we show in x5.1 that inferential-realizational frameworks, such as Network Morphology and Paradigm Function Morphology, are capable of capturing syncretic patterns which may arise as the result of underspecification, or as the result of referrals (i.e. independently of feature structure). One advantage of such approaches is that referrals and underspecification can be used simultaneously. We shall see when we come to the second case study that this is just what is required for the avoidance morphology of Dalabon. Indeed, we show that generalized referrals – where sets of paradigm cells can refer to sets of paradigm cells – which frameworks such as Network Morphology allow for, are an important requirement. We consider the relationship between underspecification (a feature structure-based constraint) and semantic naturalness. As we saw in Chapter 4: x4.2.1, underspecification can be used with atomic feature values (i.e. in a ‘flat’ structure) where the syncretism is the default to the ‘elsewhere’. In such cases, in the absence of other representational means, only one syncretic pattern can be described for any domain. Underspecification in such instances need not be equated with semantic naturalness, as the elsewhere is the residue of what is not specified in the morphology. Where a hierarchical feature structure is assumed, underspecification may represent the resort to an intermediate ‘natural class’, but as we saw in our discussion of Gaagudju in Chapter 4: x4.2.1, sometimes it is not possible to order feature values in such a way that we can always treat syncretism as a default to an intermediate semantic class. Our case studies in x5.2–x5.4 concentrate on significant fragments of individual languages. In x5.2 we show that the Dhasaanac data in 171

172


Chapter 3: x3.7.2 can be analysed straightforwardly in the Network Morphology framework. In x5.3 we turn to the verbal system of Dalabon to show that syncretism which may originally have had a semantic or pragmatic motivation has been morphologized, and demonstrate that an adequate account requires a combination of referral and underspecification in order to see how that language’s morphological systematicity relates to more general patterns. In x5.4 we look at the nominal system of Russian to show how the different degrees of morphological regularity can be represented in terms of the domains to which they apply, and how morphology, syntax and semantics interact in the animacy-based syncretisms. In each case, we have implemented our analysis of the languages in question, an important step, because it allows the reader to check that the analyses are valid, and so to have a firm basis for evaluating them.

5.1

Network Morphology and syncretism

We shall express our analyses within the Network Morphology framework. It is based on the following principles: (1) Morphology is lexeme-based. This means that lexical items in a Network Morphology analysis are treated as generalizations over their inflectional paradigms.1 Informally, the rules of morphology are therefore functions from lexical items into sets of inflected forms. A second important principle associated with Network Morphology is (2). (2) Morphology is inferential-realizational. Network Morphology does not treat affixes (or other inflectional operations) as lexical entries. Instead these are represented as generalizations in an inheritance hierarchy. This is a natural way to treat inflection, as it is therefore straightforward to treat exceptionality and irregularity as associated with the lexeme in question, rather than having to encode in each affix the information about the most irregular item that it can combine with. We illustrate this further in x5.1.1.

1

We are excluding from this definition idioms and cliche´s (such as ‘throw down the gauntlet’) which are listed lexically. However, even for items such as these it is still important to provide their inflectional variants (such as ‘John threw down the gauntlet’, and so on).


173

Another property of Network Morphology is that it uses default inheritance networks which can distribute information about morphology in different parts of the network. (3) Network Morphology uses default inheritance. We illustrate how default inheritance works in x5.1.2 The main advantage of default inheritance theories is that one is able to specify the general inflectional properties of a language and at the same time include exceptions. It also allows one to see the domains over which a particular phenomenon applies. We shall illustrate this in our case study on Russian, where we consider the domains of syncretism in that language. As with all formal frameworks we make a distinction between the fully specified morphosyntactic paradigm and the particular formal analysis which accounts for it. The fully specified morphosyntactic paradigm has the following characteristics: (4) Feature values are atomic. (5) Features are ordered relative to each other.2 When we talk of underspecification we mean the relationship of the formal analysis to the fully specified morphosyntactic paradigm. (6) The formal analysis may be underspecified relative to the fully specified paradigms. Any underspecification of the formal analysis in (6) will have to obey the constraints of the ordering of features in (5). The ordering of the features reflects cross-linguistic tendencies, such as those illustrated in our analysis in Chapter 3: x3.8. Network Morphology uses underspecification and referrals and combines both as generalized referrals. (7) Generalized referral: a. One feature specification may refer to another feature specification for its realization. 2

There are at least two, logically independent, roles which might be attributed to feature ordering: (i) underspecification-based syncretism; (ii) the ordering of inflectional elements. The ordering of features, as we have seen in our discussion of uninflectedness and neutralization in Chapter 2 and the analysis in x3.8, is ideal for underspecification-based syncretism. We are not claiming that all rules of inflectional morphology can be accommodated with ordered features. Finkel, Shen, Stump and Thesayi (2002) develop KATR a multisetbased extension of DATR in order to treat instances of morphology for which feature ordering is not relevant.

174

The Syntax–Morphology Interface b. As with other realization rules, referrals may be underspecified. c. Extensions of the referring specification will be realized by extensions of the referred-to specification.

The consequence of (7a) is that there can be instances of directional syncretism. It follows from (7b) and (7c) that this directionality need not be limited to individual paradigm cells, but, because referrals are underspecified, may involve whole sub-paradigms. While underspecification and generalized referrals, as we have formulated them, still involve reference to features, we allow for morphomic indexes to capture non-directional patternings which are independent of features. The relationship between these indexes and morphosyntax is still constrained by the ordering of features. In other words, indexed forms must still map to well-formed feature specifications. (8) Morphomic indexes must be associated with well-formed feature specifications. We shall see how indexes are used in x5.2 in the case study for Dhaasanac. An important part of the Network Morphology philosophy is that the analyses should be implementable and testable. For this purpose the lexical knowledge representation language DATR is used (Evans and Gazdar 1996). An attractive property of DATR is that it can be used to represent default inheritance networks, a key organizing principle of the Network Morphology framework.3 Default inheritance provides for elegant treatments of many problems within morphology. Corbett and Fraser (1993) and Fraser and Corbett (1995) showed how default inheritance could be used to analyse animacy, gender and declensional class assignment in Russian, and this work was applied to a substantial fragment of Russian nominal morphology. Brown (1998a) adopted this Network Morphology approach for analysing gender assignment in Polish, and Fraser and Corbett (1997) implemented a Network Morphology analysis of gender and morphological class assignment in Arapesh, a language of the Torricelli family, spoken on the north

3

DATR has been used for lexical knowledge representation for a variety of languages. The DATR-based work on German by Bleiching, Drexel and Gibbon (1996) and Cahill and Gazdar (1997, 1999) is of particular relevance for syncretism. Cahill and Gazdar (1997: 220–3) discuss rules of exponence and rules of referral in their work on adjectives, determiners and pronouns. This work is extended to nouns in Cahill and Gazdar (1999).


175

coast of Papua New Guinea. Evans, Brown and Corbett (2002) give a detailed account of the gender and morphological class assignment system of Bininj Gun-Wok (a non-Pama-Nyungan language of northern Australia, discussed above in Chapter 3: x3.5.1). This work included the mother-in-law, or avoidance register, and was applied to a sample of nouns in the language. The notion of layered defaults was important for that analysis, as certain nouns require access to what generally holds for their class and to what is the overall default for nouns as a whole. Other work has shown how the stress system of Russian can be analysed using default inheritance (Brown, Corbett, Fraser, Hippisley and Timberlake 1996), and Hippisley (1997, 2001) has applied the Network Morphology framework to word-formation. 5.1.1 Inferential-realizational theories and morphology Network Morphology is an inferential-realizational theoretical framework. This term is due to Stump (2001: 1–30), who develops a typology of morphological theories which divides them according to two dimensions: lexical versus inferential, and incremental versus realizational. Lexical theories treat morphological formants, such as affixes, in the same way as stems and their associated meanings, by listing both in the lexicon. Inferential theories, on the other hand, treat the relation between lexical stems and their paradigm of word forms as a matter of inference, that is of rules or formulas. Incremental theories require of inflectional morphology that it be information-increasing so that words build their morphosyntactic specification as the result of acquiring affixes. Realizational theories, on the other hand, associate a set of morphosyntactic specifications with lexical stems and these specifications license morphological exponents, such as affixes. Stump (2001: 17–27) gives a number of fundamental reasons for preferring inferential-realizational theories over other theories of inflectional morphology. What concerns us here are the specific properties of inferential-realizational theories which make them particularly well suited for analysing syncretism. Consider the Slovene noun paradigm in (9), reproduced from Chapter 4: x4.4.4, which has been discussed by Corbett and Fraser (1997) in relation to rules of referral. The question here is how to guarantee that the plural stem ljud- occurs in the correct contexts in the dual, namely genitive and locative, and not in the other cases of that paradigm. This is a problem for an incremental framework, because a narrow feature specification for the stem is the most natural way of specifying the contexts in which they occur, but this would stop the addition of the affixes -i and -eh, because they would not

176



Paradigm of the Slovene noun cˇlo´vek ‘person’ (based on Priestly 1993: 401) singular

dual

plural

cˇlovek cˇloveka cˇloveka cˇloveku cˇloveku cˇlovekom

cˇloveka cˇloveka ljudi ljudeh cˇlovekoma cˇlovekoma

ljudje ljudi ljudi ljudeh ljudem ljudmi

contribute any new information and would therefore not be informationincreasing. Lexical theories rely on the subcategorization properties of the affixes which they treat as having lexical entries. The combined suppletion and syncretism in (9) poses a number of problems for such theories. The stem ljud- requires a different set of plural affixes from those which would typically go with a regular noun of cˇlovek’s type. As we can see from the other cells in the dual paradigm, the stem cˇlovek- can be used in the dual. The question is what blocks the expected genitive plural/dual and locative plural/dual affixes for nouns of cˇlovek’s type, -ov and -ih respectively, from combining with the stem cˇlovek-? The answer for the plural paradigm is clear enough; there is a lexically specified suppletive plural stem, which is associated with a different set of endings. So the subcategorization information of affixes must be sensitive to the number information on the stem, even though most nouns do not have different singular and plural stems. At this point, a lexical framework already has to start listing within its regular affixes elements of potential exceptionality. For inferentialrealizational theories this problem does not arise, because they do not treat affixal exponents as lexical entries that require exhaustive information on what they may combine with. In fact, inferential-realizational theories can treat exceptionality where it arises, with the stems of the lexical items in question, for which the appropriate affixes or other exponents can be inferred. When we continue to the more challenging question of the dual paradigm, the problems for lexical theories increase. If we assume that the stem ljud- is specified by some disjunction as also being the genitive dual and locative dual stem, then the affixes have to be sensitive to this information. If we do this, of course, then we have encoded in the affixes information about the most irregular items they can combine with. If our framework is a lexical-incremental one then things cannot be


177

worse, because the output information is no more informative than the subcategorization information. Yet, there is a simple answer to why we obtain the pattern in (9): for all nouns the genitive and locative dual and plural are always identical in form. Because they can make reference to sets of paradigm cells, inferential-realizational theories have no problem making simple generalizations like this. While Network Morphology is an inferential-realizational framework, it also combines this property with the ability to incorporate varying degrees of regularity, by analysing morphology in terms of default inheritance, to which we now turn. 5.1.2 Default inheritance Networks consist of nodes and connections between them. In default inheritance networks, information is inherited from higher nodes unless it is specifically overridden. In Figure 3 we present a simple default inheritance network. The diagram is based on a DATR example from Evans and Gazdar (1996: 176).4 In Figure 3, VERB, EN_VERB, Love, Do, Mow, Sew and Be are all nodes in the network. The nodes Love, Do, and EN_VERB inherit from VERB. The nodes Mow, Sew and Be inherit from EN_VERB, and therefore also from VERB. The nodes in Figure 3 are locations for information about the morphological paradigms of the items in question. As such, they generalize the information which classes of lexemes have in common. The relationships between nodes also make it possible to characterize the degree of exceptionality or lexical idiosyncrasy involved. VERB

EN_VERB

Love

Do

Mow

Sew

Be

Figure 3. A default inheritance hierarchy for some English verbs (based on Evans and Gazdar 1996) 4

This section discusses default inheritance in DATR. For the purposes of illustration using English, the examples are based on the fragment published in Evans and Gazdar (1996), including their use of attributes. This fragment was written to demonstrate DATR, and not Network Morphology.

178


A property that all English verbs share is that the passive participle has the same form as the past participle. Examples (10) and (12) involve a past participle, while examples (11) and (13) involve passive participles. (10) (11) (12) (13)

I have loved. I am loved. I have done this. This is often done.

We should be able to state this fact as a general property of verbs. The form of the past tense is generally -ed, but this can be overridden by particular items, such as do, whose past tense is did. Often, but not always, the past participle will have the same form as the past tense. There are also subregular classes, such as the one where the past participle is formed using -en. Default inheritance allows for a concise treatment of these facts. Evans and Gazdar (1996: 176) state the following at the node VERB in (14), where we have omitted some information, as indicated by the ellipses. What is given in (14) is a representation of the information associated with the top node in Figure 3. (14)

VERB: <syn cat> == verb <syn type> == main <mor past> == "<mor root>" ed <mor passive> == "<mor past>" <mor present> == "<mor root>" <mor present participle> == "<mor root>" ing <mor present tense sing three> == "<mor root>" s ...

The node name VERB is placed before the colon. Each line containing ‘==’ is a DATR equation. Each left-hand side of a DATR equation contains paths. Paths contain a combination of ordered attributes. The right-hand side of the equation may contain values, such as ‘verb’. Alternatively it may contain paths, or node names, or it may contain a combination of paths, values and node names. The first equation at VERB states that the syntactic category of items belonging to this class is ‘verb’. The equation after this states that the syntactic type of verb is ‘main’ (i.e. a typical verb is a main verb rather than an auxiliary). The next equation says that the past is a concatentation of -ed onto what Evans and Gazdar call the morphological root. The equation after that says that the passive has the same form as the past. This is the way referrals are represented in DATR. The equation after the


179

statement about the passive says that the present uses the morphological root. This brings us to another important property of DATR: in the absence of any information to the contrary, we can infer that the value of a path will be the same as the value for the most specific path of which that path is an extension. In (15) we give examples of path extension. (15)

<mor present> <mor present participle> <mor present tense sing one> <mor present tense sing two> <mor present tense sing three> <mor present tense plur>

All of the paths listed after <mor present> in (15) are extensions of it. At the node VERB in (14) only two of these extensions are found on the left-hand side of equations, in addition to the path <mor present> itself. These are: (16)

<mor present participle> <mor present tense sing three>

This means that we can infer that the values for the first person singular present tense, the second person singular present tense and the plural present tense are the same as for the present tense as a whole, namely the morphological root. This is an example of default inference. The values for the present participle and third person present tense are not inferred in this way, as they are already specified at VERB. From the equation at VERB with left-hand side <mor past> it can be inferred that the past tense and past participle (which are extensions of <mor past>) are both formed by suffixing -ed to the morphological root. We have already seen a referral-based way of stating the identity of the past participle and passive participle. Here the default syncretism of the past tense and past participle is the result of underspecification. We also need to override the default syncretism of the past tense and past participle. Verbs which have a past participle in -en must inherit the suffixation of -ed for the past tense but override it for the past participle. This can be stated at the node EN_VERB which, as we can see from Figure 3, inherits from the node VERB. Evans and Gazdar (1996: 176) represent this in DATR as follows: (17)

EN_VERB: == VERB <mor past participle> == "<mor root>" en.

180


In (17) the path indicates that EN_VERB will inherit all information from VERB unless otherwise specified. EN_VERB therefore inherits the specification that <mor past> suffixes -ed onto the morphological root. This means that the past tense forms for EN_VERB will still be formed using -ed. As the extension <mor past participle> has a specific value at EN_VERB, this means that the past participles of verbs of this type will suffix -en. This is one way in which the default syncretism of the participle and the past tense forms at VERB is overridden by a particular class of verbs, and it is represented in terms of path extension, which can be interpreted as being a strict form of underspecification. It is strict, as the order of attributes in the path is important for the definition of extension. (17) also illustrates another point about theories which use default inheritance, namely that they can involve layered or cascaded defaults. The overall default for verbs is to have the past participle and past tense form the same, but another layer of verbs maintains the default -ed for the past tense, while introducing its own default for the past participle. Syncretism can be represented in terms of underspecification and referral in Network Morphology. An important advantage of this is that it is possible to combine referrals and underspecification to pick out sets of cells which are syncretic. In the following sections we discuss how underspecification relates to semantic naturalness and how to treat systematic syncretisms. The beauty of an inheritance-based approach, such as Network Morphology, is that it can represent what occurs over a large domain of morphology, representing degrees of regularity, from the domain of a particular word class, or higher, right down to individual lexical items. 5.1.3 Underspecification and semantic naturalness Underspecification is often associated with a ‘semantic’ approach to syncretism. However, the feature values involved in syncretism may or may not be semantically assigned. This is true of gender, for example, in systems which combine formal and semantic assignment. Under such systems the relationship between genders x and y might be formal (in that they share some morphology), and between genders y and z semantic (in that they are related by meaning). Or genders x, y and z may have a semantic core complemented by formal assignment for each. Given the cross-linguistic propensity of genders to collapse (Chapter 3: x3.8), particularly when number is involved (Chapter 3: x3.4.2), we might also wish to treat gender


181

syncretism as underspecification for this reason. This means that there are therefore two different interpretations of what underlies underspecification: (18) (19)

Syncretisms based on underspecification reflect semantic naturalness. Syncretisms based on underspecification reflect the cross-linguistic tendency of a feature to syncretize (i.e. for different values of that feature to share identical morphology).

The more appropriate interpretation of underspecification is that in (19). The reason for this is that underspecification is the natural way of treating both uninflectedness and neutralization, as we saw in Chapter 2: x2.4. If we consider the examples of neutralization and uninflectedness given in Chapter 2: x2.4, none of them can be explained entirely in terms of the structure inherent to the feature involved. As we have seen in Chapter 3: x3.4.2, there appears to be some involvement of number – typically plural – and the restructuring of gender, which may sometimes have a semantic basis. Uninflectedness is a morphological property which is sensitive to morphosyntactic structure. Under such circumstances underspecification should be put in the context of other morphosyntactic features and is itself not directly interpretable as reflecting a semantic motivation for syncretism. It is logical to ask whether syncretism can be determined solely by semantics. Animacy in Slavonic appeared to be a possible instance. But as we saw in Chapter 3: x3.1, it is difficult to isolate an on-going semantic pattern underlying the instances of type 2 syncretism (core case with noncore case), to which this Slavonic example belongs. Furthermore, the animacy rule is itself subject to being overridden when items which should be subject to it belong to a particular inflectional class, showing that morphology plays a role.5 As Claire Bowern points out (personal communication) there are instances in various other languages of place names having their nominative form identical to the locative. However, these are typically instances of lexically determined syncretism: individual items have the syncretism, but not all the items in a semantic class. In Tsez, proper nouns denoting places, provided they are native words, will have the absolutive identical to one of six local cases (Bernard Comrie, personal communication). For example, the name of the village Asakh, asaq, includes the marker -q ‘on (a vertical surface)’ and may function both as

5

The noun djadja ‘uncle’ in Russian, for example, is a masculine animate noun but is not subject to the animacy rule in the singular, because of its inflectional class membership.

182


an absolutive and as a local case form. Hence the semantic type allows the prediction that there will be a syncretism but does not determine which case will be involved, since this varies from place name to place name. Moreover, though the form may be used as an absolutive, speakers show some reluctance here, and prefer to use the name in apposition to a common noun, which has a clear absolutive. Here again, then, semantics does not uniquely determine a domain of syncretism. To date we have not found a language in which a noun denoting a place will necessarily have a specific syncretism. And in general, we have not found instances of semantics uniquely providing a domain for syncretism. 5.1.4 Systematicity in Network Morphology Below we offer a model in which systematic syncretism is represented by the devices of underspecification, indexing and referral, supplemented by the DATR notion of attribute ordering. In the analysis which we are proposing underspecification is related to the ordering of attributes in a path, which in turn is connected with the crosslinguistic patterns of ordering of features which were highlighted in Chapter 3: x3.8. Hierarchies which rank inflectional features have played an important role in other accounts of particular areas of the syncretism spectrum: for example, Carstairs (1987) treatment of ‘take-overs’ and the Systematic Homonymy Claim as discussed in Chapter 4: x4.4.2, or Stump’s (2001: 239) Feature Ranking Principle, as discussed in Chapter 4: x4.4.4. Our formal analysis is based on the following interpretations: (20)

Network Morphology analysis a. The order of attributes in a path is associated with cross-linguistic tendencies of particular features to syncretize. The further to the right an attribute occurs, the more likely it is to syncretize. b. The order of attributes in a path will line up with neutralization and uninflectedness. c. Underspecification involves resort to an elsewhere form only. d. Referrals represent directional systematic syncretism within the language. e. Indexes represent non-directional syncretism within the language. f. Syncretisms which violate the order of attributes as required by (20a) are language-specific referrals (but still systematic). g. For underspecification-based syncretism there is generally one form which has more than one function as a default property.


183

h. For referral-based syncretism there is a default form associated with a default primary function. The other function of that form will typically be manifested in a subset of lexical items or morphosyntactic contexts. i. The adequate treatment of some syncretic patterns requires simultaneous use of underspecification and referrals, thereby justifying the theoretical requirement for both. The Slovene nominal paradigm in (9) and the Dalabon verbal system (x5.3) involve examples of this type. The order of attributes in (20a) is established according to cross-linguistic tendencies, such as those observed in Chapter 3: x3.8. Given atomic values, we can easily find examples of less straightforward syncretism which line up with uninflectedness and neutralization, as expected by (20b). Breaking features up further may obscure this tendency, as we will allow for explanations based more on uninflectedness and neutralization. Because of the constraint in (20a) and the requirement for atomic values in (5), the number of possible elsewhere forms is constrained by the number of features realized. For example, if we considered a language which has two number values and three or more case values, then it potentially has two elsewhere forms, a singular one and a plural one. This arises from ordering number before case (as suggested by the data in Chapter 3: x3.8). In contrast with this, the potential number of elsewhere forms for Russian adjective morphology is twelve (a grid of two number values and six case values), if number is ordered before case, and case is ordered before gender. If we ignore animacy, then Russian comes close to making use of the full range of elsewhere forms in the adjectival paradigm: syncretism of masculine and neuter in the four oblique cases in the singular, and neutralization of all gender distinctions in the plural, leaving just six plural cells. But this is not the end of syncretism in Russian, because there are additional patterns related to animacy. Referral-based syncretism is not constrained by an underlying paradigmatic design which leads to a limit on the number of elsewhere forms. However, we would also expect to find some indication of directional effects.

5.2

Case study 1: Dhaasanac

As we saw in Chapter 3: x3.7.2 the Dhaasanac verbal paradigm is an example of morphological regularity where the syncretized values are

184


heterogeneous. This is a very good example of ‘morphology by itself’ (Aronoff 1994), as the use of indexes below shows. In the case study we show how we can capture this autonomous morphological regularity and still associate it with the relevant morphosyntax. The DATR representation associated with this analysis is given in Appendix 4. We repeat the scheme of syncretism in Dhaasanac from Chapter 3: x3.7.2 in (21). (21)

Dhaasanac syncretisms

1INCL 1 2 3F 3M

singular

plural

— A B B A

A B B A A

This pattern generalizes across the positive imperfect, positive perfect, positive dependent and positive simple past paradigms in Dhaasanac. This is strong evidence for morphological systematicity, which is further reinforced by the fact that different verbs have different A and B variants but still obey the pattern in (21). The other important aspect of the Dhaasanac data is that they are an instance of mediated polarity, a term which we introduced in Chapter 3: x3.7.1. Importantly, the B form is always used for the second person, both singular and plural. This is therefore an example of underspecification of number in the presence of second person, but which still needs to be combined with indexing of the forms. The fully specified paradigms which we assume for Dhaasanac will be as in (22). Here ‘pos’ stands for positive, ‘prf’ for perfect, and ‘1st sg’ for first person singular, and so on. (22)

<mor pos prf 1st sg> <mor pos prf 2nd sg> <mor pos prf 3rd sg f> <mor pos prf 3rd sg m> <mor pos prf 1st_excl pl> <mor pos prf 1st_incl pl> <mor pos prf 2nd pl> <mor pos prf 3rd pl>

The ordering of attributes in (22) reflects the fact that it is negation (or its absence) and TAM which determine whether there are A and B forms,


185

because the negated paradigms and the imperative do not distinguish A and B forms. It is the role of our formal analysis to infer the correct forms for the positive paradigms. In (23) we give the lexical entry for the verb fu´r ‘to open’, which belongs to the class of verbs whose stem ends in a coronal consonant. (23)

F˙r: == VERB == CORONALS == ‘open’ <stem type> == one_vowel <stem cons_1> == f <stem vowel_1> == _u <stem final> == _r.

The node Fu´r inherits from the node VERB, which includes the information in (24). (We have omitted some of the information given at VERB.) (24)

VERB: == <syn> == verb == _A == _B == _B == _B <mor pos imprf> == "" <mor pos prf> == ""

...

Recall from (22) that extensions of the path <mor pos prf>, for example, will involve attributes for person and number (in that order), and for the third person singular, also gender. Hence, in the absence of any more specific information, the form of the perfect paradigm will be determined by looking at the equation associated at VERB with <mor pos prf>, namely <mor pos prf> == "". The right-hand side of the equation "" involves an evaluable path (the part). This equation basically means that the morphology of the positive perfect is determined by looking for the form of the perfect and inserting the appropriate index for that form. Let us consider the morphology for the path <mor pos prf 2nd sg>. The rules of inference mean that, among other things, we can infer (25) from the equation <mor pos prf> == "". (25)

We can infer that: <mor pos prf 2nd sg> == ""

186


From the information in (24) we can infer (25), and because of the equation == _B in (24), and the inference made in (25), we can infer (26). (26)

<mor pos prf 2nd sg > == ""

The other TAM, person and number combinations work in the same way. This is a realizational approach to morphology because it separates out the morphosyntactic specification from the actual form with which it may be associated. We should note from (23) that the lexical item Fu´r inherits information about its associated forms from the node CORONALS. Even though Fu´r’s A and B forms may differ from those of other verbs, the equations at VERB can generalize across all verbs, irrespective of the specific realization of the A and B forms. In the case of Fu´r we can obtain the forms in (27) for the perfect paradigm. (27)

Fu¤r:<mor pos prf 1st sg> = f _u _r -i. Fu¤r:<mor pos prf 2nd sg> = f _u _ºº -i. Fu¤r:<mor pos prf 3rd sg f> = f _u _ºº -i. Fu¤r:<mor pos prf 3rd sg m> = f _u _r -i. Fu¤r:<mor pos prf 1st_excl pl> = f _u _ºº -i. Fu¤r:<mor pos prf 1st_incl pl> = f _u _r -i. Fu¤r:<mor pos prf 2nd pl> = f _u _ºº -i. Fu¤r:<mor pos prf 3rd pl> = f _u _r -i.

Thus, by treating the inventory of forms available as separate from the associated morphosyntax, it is possible to capture the morphological systematicity of the Dhasaanac verb. At the same time the morphology is still constrained, because these indexed items have to match up with wellformed feature specifications.

5.3

Case study 2: The Dalabon verbal system

This case study sets out the core details of syncretism within Dalabon. It is based on Evans, Brown and Corbett (2001) and uses the same orthography. The DATR representation associated with this Network Morphology analysis is given in Appendix 5. Dalabon is an Australian language of the Gunwinyguan family, spoken in central Arnhem Land by a declining population which is now reduced to about twenty fluent speakers. Dalabon’s verbal system marks intransitive subjects using bound prefixal morphology, and transitive subjects and objects using a combination of bound prefixes and clitics. As we saw in


187

Chapter 3: x3.3 there is strong cross-linguistic evidence to show that transitive verbs pattern very differently from intransitives in terms of the person syncretism. This would suggest that transitivity is an important factor in syncretism on verbs. Dalabon’s verbal system involves a morphologization of what was originally a semantically or pragmatically motivated avoidance strategy, a phenomenon investigated by Heath for Australian languages (Heath 1991) and languages of the Americas (Heath 1998). The avoidance strategy in Dalabon has produced a situation in which certain combinations of first person subject on second person object or second person subject on first person object are avoided by the use of person syncretism of the subject, whereby the appropriate form of the third person subject is used. There is a directional effect, because the forms used in the syncretism have third person subject as their primary function, as can be demonstrated by looking at the intransitive paradigm. Furthermore, the referrals are simultaneously combined with underspecification and predict whole sets of cells in the extended Dalabon paradigm. This demonstrates that the best treatment of some syncretisms requires a combination of referral and underspecification, and so an adequate framework requires both. 5.3.1 The structure of the Dalabon intransitive paradigm Dalabon pronominal prefixes mark tense, aspect, mood and clause status. There are six relevant distinctions: REALIS, SUBORD1, SUBORD2, IRREALIS, APPREHENSIVE and HORTATIVE. With the exception of certain forms in the APPREHENSIVE, the six series for tense, aspect, mood and clause status can be derived by regular rules. The other series can be derived from the SUBORD1 forms. The prefix forms shown in this case study are the REALIS prefixes, which differ from the SUBORD1 forms in that they terminate with a marker -h.6 The forms linked to certain cells within the transitive paradigm are based on the corresponding intransitive ones, and so we look at the features of the intransitive paradigm first. We represent the features involved as a 4 4 matrix, with four person values (FIRST EXCLUSIVE, FIRST INCLUSIVE, SECOND and THIRD), combined with three values for the number feature: SINGULAR, 6

The values used here are based on those in Evans, Brown and Corbett (2001). Evans has recently revised the inventory of values, merging SUBORD1 and IRREALIS, and adding a further value, PURPOSIVE (Nicholas Evans, personal communication). As the relevant issue for our case study is person syncretism which generalizes across the tense, aspect, mood and clause status series, this change is not significant for the formal analysis presented here.

188


DUAL, PLURAL, with an additional feature of harmonicity which has two values, HARMONIC and DISHARMONIC. This feature marks kinship relations and is normally restricted in its interpretation to the subjects of intransitive and transitive verbs, but may occasionally be interpreted in relation to the object (see Evans, Brown and Corbett 2001: 197–8). Typically, harmonicity can be seen as a distinction made in the presence of dual number. The Dalabon intransitive paradigm (REALIS) is given in (28).

(28)

The Dalabon intransitive paradigm (REALIS series)

SG DU DIS DU HARM PL

1 EXCL

1 INCL

2

3

ngahngehyarrahyalah-

— djehyahngarrah-

djahdehnarrahnalah-

kahkehbarrahbalah-

Logically there cannot be a first person inclusive singular, hence the gap in the paradigm in (28). From this it follows that person in Dalabon determines number marking to an extent. We can therefore represent the relationship between person, number and harmonicity in Dalabon as a hierarchy in Figure 4. The ordering of attributes in the paths directly reflects the hierarchy in Figure 4. The hierarchy in Figure 4 is therefore a representation of the structure of the verbal paradigm for which our formal analysis should predict the forms. The FIRST INCLUSIVE cannot have singular forms, and so the ordering allows us to state that there are no SINGULAR extensions of FIRST INCLUSIVE. It is important to make a distinction here between the fully specified paradigm in Figure 4, on the one hand, and the statements required in the formal analysis to infer the correct forms in that fully specified paradigm. When we turn to the structure of the transitive paradigm in the next section, we will see that we need to refer one person value to another, together with its possible extensions for number. This illustrates the point that referrals and underspecification are required simultaneously. 5.3.2

The Dalabon transitive paradigm and the inadequacy of underspecification There are 102 distinct subject/object combinations, which in turn generate 102 6 = 612 forms when all TAM combinations are included. This


1ST

SG

PL

DIS

2ND

1ST INC

DU

PL

HARM

DU

DIS

SG PL

HARM

DIS

189

3RD

DU

SG PL

HARM

DU

DIS

HARM

Figure 4. Paths for person, number and harmonicity in Dalabon

number takes account of the fact that coreferential combinations are ruled out, as these are encoded by the use of a reflexive/reciprocal suffix with the intransitive paradigm. There is also a contrast between ‘higher’ and ‘lower’ third person objects when both subject and object are third person singular. We shall analyse the paradigm section by section. In (29) we give the forms of the paradigm for the third person transitive subject, when the object is singular. For comparison, the intransitive subject forms are given in square brackets above the columns with the transitive forms. The forms for the second person singular object are (29)

The third person REALIS subject paradigm in Dalabon subject

object

[none 1SG 2SG 3SG

3SG

3 DIS

3DU

3PL

kahkahdjahkahbvkah7-

kehkehdjirrahkeh-

barrahburrahdjirrahburrah-

balah-] bulahdjilahbulah-

special portmanteau forms which combine information about the subject and the object. In the case of the paradigm for the second person singular object it is possible to interpret the syncretism of 3 DIS > 2SG with 7

Used for a higher animacy object.

190


3DU > 2SG as resulting from the underspecification of the values for harmonicity. The disyllabic dual and plural transitive subject forms which occur in (29) with the first person objects and third person objects are derived from the intransitive subject forms by the following rule: (30)

CaLah- ! CiLah- if C = y !CuLah- elsewhere (where L is a liquid, i.e. l or rr)

The forms derived by rule (30) could be interpreted as marking transitivity, subject person and number, while being underspecified for the actual person and number of the object. This is because they can be used with a third or first person singular object or with non-singular objects of any person. In the latter case a variety of proclitics are used to mark the object. The analysis of the forms burrah- and bulah- has consequences for our analysis of the forms involving the second person singular object. If djirrah- and djilah- are treated as realizing second singular object combined with the number of the subject, then the competition between the forms burrah- and djirrah-, and between the forms bulah- and djilah-, cannot be resolved using Panini’s principle, as explained in Chapter 4: x4.3, because the feature specifications would overlap but neither would be more specific than the other (31). (31)

burrah- (A 3DU), bulah- (A 3PL) djirrah- (OBJ 2SG, A DU), djilah- (OBJ 2SG, A PL)

The reason that the competition cannot be resolved using Panini’s principle is that the forms burrah- and bulah- in (31) contain a subject feature value for third person. (32)

CONFLICT A: which forms should be used for 3DU > 2SG and 3PL > 2SG?

However, we can avoid CONFLICT A, if the value for the subject person in burrah- and bulah- is deleted, treating these prefixes as marking only subject number. We would then obtain (33). (33)

burrah- (A DU), bulah- (A PL) djirrah- (OBJ 2SG, A DU), djilah- (OBJ 2SG, A PL)

Now the specification for burrah- and bulah- is contained in that for djirrah- and djilah- and so the competition between the two can be resolved according to Panini’s principle. So far it looks as though the specifications in (33) will obtain the right result. This appears to be borne out when we


191

look at other parts of the transitive paradigm, such as that for the first exclusive subject in (34). (34)

The first person exclusive REALIS subject paradigm in Dalabon subject

object

[none 2SG 3SG

1SG

1DIS

1DU

1PL

ngahdjahngah-

ngehdjirrahngeh-

yarrahdjirrahyirrah-

yalah-] djilahyilah-

The forms for 1DU > 3SG and 1PL > 3SG are obtained from the intransitive subject forms by the rule in (30). This means that we can assume that they have the specifications in (35). (35)

yirrah- (A 1DU), yilah- (A 1PL)

Recall from (33) the specifications for djirrah- (OBJ 2SG, A DU) and djilah(OBJ 2SG, A PL). As things stand, the specifications for djirrah- and djilahoverlap with those for yirrah- and yilah- in (35), but neither is more specific than the other. (36)

CONFLICT B: which forms should be used for 1DU > 2SG and 1PL > 2SG?

The natural resolution of CONFLICT A by underspecification in (32) was to treat the forms burrah- (A DU) and bulah- (A PL), which are used for 3DU > 2SG and 3PL > 2SG, as not marking subject person. This means that the option of underspecifying the subject person of yirrah- (A 1DU) and yilah- (A 1PL) is not open to us in order to resolve CONFLICT B in (36), as this would then be the same specification as for burrah- (A DU) and bulah(A PL). Hence the resolution of CONFLICT A is itself in conflict with the resolution of CONFLICT B. We could claim that this problem can be avoided by resort to a principle which prioritizes the realization of the 2SG object. While the second person singular object is clearly being treated as special, the problem is that first person objects are not realized by bound prefixal morphology, and so the putative principle cannot be associated with a position to the left on a person hierarchy, and would therefore involve ad hoc stipulation. There is a further problem for an underspecification analysis when we turn to the second person subject paradigm in (37).

192


The second person REALIS subject paradigm in Dalabon subject

object

[none 1SG 3SG

2SG

2DIS

2DU

2PL

djahkahdah-

dehkehdeh-

narrahburrahnurrah-

nalah-] bulahnulah-

Like yirrah- and yilah- the forms nurrah- and nulah- are derived by rule (30) from the intransitive forms and therefore have the feature specifications in (38). (38)

nurrah- (A 2DU), nulah- (A 2PL)

However, the question arises why the forms burrah- and bulah- are used for the 2DU > 1SG and 2PL > 1SG respectively. To see this, let us compare the feature specifications with those of burrah- and bulah- which were given in (33). (39)

nurrah- and nulah- should fill the 2DU > 1SG and 2PL > 1SG cells according to Panini’s principle burrah- (A DU), bulah- (A PL) nurrah- (2A DU) nulah- (A 2PL)

As the feature specifications in (38) for nurrah- (A 2DU) and nulah- (A 2PL) are more specific than the feature specifications we assumed for burrah(A DU) and bulah- (A PL) in (33), we would expect nurrah- and nulah- to fill the 2DU > 1SG and 2PL > 1SG cells. However, even if we eliminated the person information of nurrah- and nulah-, this would not make them less specific than burrah- and bulah-. Hence, underspecification cannot avoid the incorrect prediction that nurrah- and nulah- will fill the 2DU > 1SG and 2PL > 1SG cells. (40)

CONFLICT C: which forms should be used for 2DU > 1SG and 2PL > 1SG?

One way to resolve CONFLICT C in (40) might be to alter the information content in (38) so that nurrah- and nulah- specify the person of the object, namely third person. However, a further problem, which is perhaps not insurmountable, is that the rule in (30) would then be associated with different morphosyntactic content, depending on the prefix. For


193

nurrah- and nulah- rule (30) would not only involve the addition of information that the prefix is transitive, but also that the person of the object is third person. On the other hand, for yirrah- (A 1DU) and yilah- (A 1PL) rule (30) would not involve the addition of information about the person of the object. So far we have concentrated on the non-singular subject forms which are derived by rule (30) from their intransitive counterparts. But it turns out that the problems we have highlighted for underspecification and the nonsingular forms generalize to all forms, that is to singular and disharmonic as well. In (30) we present the paradigm of singular subjects and singular objects.

(41)

The paradigm of singular subjects and singular objects (REALIS) subject

object

[none 1SG 2SG 3SG

1SG

2SG

3SG

ngah— djahngah-

djahkah— dah-

kah-] kahdjahkahbvkah- (higher animacy object)

For the singular subjects similar problems arise for underspecification. If djah- may mark the second singular intransitive subject and the object in the 1SG > 2SG cell (an ergative absolutive pattern), then why is kah- used in the 2SG > 1SG cell, rather than djah-? We can, of course, specify kah- as marking a singular subject (with person underspecified), but even then djah-, which marks second person singular (subject or object), would be expected to fill the 2SG > 1SG cell, being more specific. This problem is in some senses worse than CONFLICT A, as in this case it is not that the feature specifications overlap, but that one is more specific than the other, and the wrong one (the less specific kah-) wins. The conflict between djah(2SG) and ngah- (1SG) to realize the 1SG > 2SG cell can be seen as an extension of CONFLICT B. Again, we cannot underspecify ngah- for subject person, as this would then mean it had exactly the same specification as kah- (which is the same problem we encountered for yirrah- (A 1DU) and yilah- (A 1PL)), and here it is not clear that our putative principle, which

194


treats the marking of second person objects as special, can work, as djah-, because of its alignment, is not just an object marker. So for a variety of reasons underspecification fails when used on its own.8 5.3.3 The Dalabon transitive paradigm: a generalized referral analysis As we showed in Chapter 3: x3.3, the person syncretisms which arise in transitive verbs are a case apart from intransitives. This suggests that, where we seek explanations for syncretisms in transitive verbs, these should be justified on the basis of combinations which are brought about by transitivity. At the same time we should not dispense with the insights of underspecification, if it can be applied naturally. So far we have looked only at those parts of the Dalabon verbal paradigm involving singular objects. If we lay out the whole paradigm (Table 2), then a clear pattern emerges. (42)

(43)

GENERALIZATION 1: The 1SG > 3SG, 3SG > 3SG, 1 DIS > 3SG and 3 DIS > 3SG forms are the same as the forms for the corresponding intransitive subject (with the exception of bvkah- which marks 3SG > 3SG with higher animacy objects). GENERALIZATION 2: The 1DU > 3SG, 3DU > 3SG, 1PL > 3SG, 3PL > 3SG forms are transparently derived from the intransitive subject forms by rule (30).

Generalizations 1 and 2 both have a phonological basis since the forms referred to in (42) are all monosyllabic, and those referred to in (43) are polysyllabic. It should also be noted that third person singular object marking cannot be treated as neutralization, as we have defined it in Chapter 2: x2.4.1, for two reasons. First, third person agreement occurs with a noun phrase, as in (44) where the comitative verb ‘appear with’ agrees with the noun Naworneng. (44)

8

‘‘djarra Naworneng nga-h-ye-burlhmung’’ kah-yini-nj here Naworneng 1.SG>3.SG-R-COM-appear.PST.PFV 3.SG-say-PST.PFV ‘‘‘Here I’ve brought Naworneng’’ he said.’ (Evans, Brown and Corbett 2001: 200)

This conclusion finds independent justification in that Wunderlich (2001a) introduces ‘taboos’ as a constraint in his correspondence-theoretic account of Dalabon. The mechanisms involved in this analysis basically replace certain feature values with others by deleting particular values in the input. This is actually very similar to referrals, thereby demonstrating that underspecification alone cannot account for the Dalabon data.

ngah-

3[—]

bulu ngah-

3PL [bulu]

bulu dah-

njel kah-

bulu kah-

nol kah-

njel kah-

bulu ngeh-

nol ngeh-

bunu ngeh-

norr ngeh-

bulu deh-

njel keh-

bunu deh-

njerr keh-

bulu keh-

nol keh-

njel keh-

bunu keh-

norr keh-

njerr keh-

keh-

deh-

ngeh-

keh-

3 DIS [keh- ]

djirrah-

keh-

2 DIS [deh- ]

djirrah-

1 DIS [ngeh- ]

bulu yirrah-

nol yirrah-

bunu yirrah-

norr yirrah-

yirrah-

djirrah-

1DU [yarrah- ] [A yirrah- ]

bulu nurrah-

njel burrah-

bunu nurrah-

njerr burrah-

nurrah-

burrah-

2DU [narrah- ] [A nurrah- ]

bulu burrah-

nol burrah-

njel burrah-

bunu burrah-

norr burrah-

njerr burrah-

burrah-

djirrah-

burrah-

3DU [barrah- ] [A burrah- ]

bulu yilah-

nol yilah-

bunu yilah-

norr yilah-

yilah-

djilah-

1PL [yalah- ] [A yilah- ]

bulu nulah-

njel bulah-

bunu nulah-

njerr bulah-

nulah-

bulah-

2PL [nalah- ] [nulah- ]

bulu bulah-

nol bulah-

njel bulah-

bunu bulah-

norr bulah-

njerr bulah-

bulah-

djilah-

bulah-

3PL [balah- ] [A bulah- ]

The combinations with first inclusive have been omitted from this table (for which see Evans, Brown and Corbett 2001: 199). The prefixal marking in bold is determined by the referrals (56) and (57).

nol ngah-

2PL [nol]

1PL[njel]

bunu kah-

bunu ngah-

3DU [bu(l)nu]

bunu dah-

norr kah-

njerr kah-

BVKAH-

kah-

djah-

kah-

3 [kah- ]

norr ngah-

njerr kah-

dah-

kah-

2 [djah- ] [A dah- ]

2DU [norr]

1DU [njerr]

djah-

1 [ngah- ]

2 [njing]

1 [ngey]

subject [INTR [/TR] form] object [free pronoun]

Table 2: Paradigm of Dalabon subject þ object combinations, including object clitics; realis TAM series (based on Evans, Brown and Corbett. 2001: 199).

196


A second reason is the form bvkah- which marks a higher animacy third person object. As this form is used uniquely for third person objects (in the presence of a third person subject), it suggests that third person is a valid syntactic distinction for objects in Dalabon. Both of these facts indicate that object person marking in the singular is not irrelevant for syntax, which therefore means that this cannot be seen as neutralization. The next set of generalizations concerns the first person singular object. (45) (46) (47)

(48)

GENERALIZATION 3: The 3SG > 1SG and 3 DIS > 1SG forms are the same as the forms for the corresponding intransitive subject. GENERALIZATION 4: The 3DU > 1SG and 3PL > 1SG forms are transparently derived from the intransitive subject by rule (30). GENERALIZATION 5: Generalizations 3 and 4 mean that the forms of 3 > 1SG are the same as the forms of the 3 > 3SG paradigm (with the exception of bvkah-). GENERALIZATION 6: Generalizations 3 and 4 mean that there is no marking of first person singular objects in the bound pronominal morphology.

Generalizations 3–6 might lead one to assume that first person singular object is not a morphosyntactically relevant entity. However, if it played no role, or indeed were non-existent, then CONFLICT C in (40) could not arise. This point itself is independent of how we actually choose to treat CONFLICT C, by underspecification or otherwise. If, for example, first person object were a syntactically non-existent entity then we would have no need to refer to it in the paradigms and also in the formal analysis associated with them. But this would mean that the contrast between the forms kah- (2SG > 1SG) and dah- (2SG > 3SG) in (41) could not exist. Of course, the prefixes kah- and dah- are used for other purposes and would not be treated in most theories as fully specified for the feature values 2SG > 1SG and 2SG > 3SG respectively. Examination of Table 2 shows that there is always a contrast between 2 > 1 and the 2 > 3 paradigm, and this means that reference must be made by morphology to first person objects. As for the second person, we know that objecthood is important there, because of the special portmanteau forms. Having demonstrated the inadequacy of underspecification alone, we now turn to the analysis in Evans, Brown and Corbett (2001) and our representation of it. That analysis resolves CONFLICT A by specifying the feature values for burrah-, bulah-, djirrah- and djilah- as in (49).

Formal framework and case studies (49)

197

burrah- (A 3DU), bulah- (A 3PL) djirrah- (OBJ 2SG, A 3DU), djilah- (OBJ 2SG, A 3PL)

What we have done in (49) is to add the second person object values to djirrah- and djilah-. This means that djirrah- and djilah- are more specific and will therefore fill the 3DU > 2SG and 3PL > 2SG cells. We now turn to the resolution of CONFLICTS B and C. These are both resolved by the use of rules of referral which are related to the combination of subject and object persons. Recall that we have seen cross-linguistic evidence to suggest that person syncretism in two-place verbs is different from that in intransitives, which strongly suggests that transitivity plays an important role. Because the rules of referral are formulated in terms of the combination of subject and object person they account for the syncretisms observed as resulting in part from transitivity. The rules of referral are as in (50) (50)

a. 2[n=] > 1[n=] uses the form for 3[n=] > 1[n=] b. 1[n=] > 2SG uses the form for 3[n=] > 2SG where n is the number value (Evans, Brown and Corbett 2001: 207)

The effect of (50a) is to say that the 2 > 1 paradigm uses the forms of the 3 > 1 paradigm. The second line (50b) states that the 1 > 2SG paradigm uses the forms of the 3 > 2SG paradigm. Hence, in this case the original motivation for the referral is the avoidance of combinations of first and second person objects, the top end of a person hierarchy. The use of variables , here is for expository purposes. The point is that the referral does not just involve one cell of the paradigm referring to another. Rather it is a combination of a referral and underspecification. A generalized referral of this kind can then be used to predict whole sub-paradigms. The referrals in (50) are based on the notion of ‘avoidance’ as elaborated by Heath (1991 and 1998). The assorted mechanisms . . . have in common the fact that they obscure the ‘objective’ relationship between speaker and addressee . . . The 1st « 2nd combinations are doubly dangerous because they not only contain the most pragmatically sensitive pronominals, they also combine them into a syntagmatic structure and thereby necessarily focus attention on the speaker–addressee relationship . . . The Australian languages . . . play down the speaker–addressee relationship by omission, substitution, or skewing of the normal, most transparent, hence also bluntest first and second person morphemes. (Heath 1991: 86)

198


Dalabon has actually taken this a step further by morphologizing the avoidance strategy, so that there are now no longer unique forms for the 2 > 1 and 1 > 2SG paradigms. This means that one can no longer treat these effects as facts of the ethnography of communication in the Dalabon community, and therefore in the same way as the use of polite Sie in German, for example. The politeness rules in German and similar languages can refer directly to the free pronoun (regardless of whether a verb is present in the utterance): the primary effect is that one free pronoun is replaced by another (du by Sie, and so on), and the effects on the morphology are secondary, simply involving the verb agreeing with the new pronoun in its (etymological) person and number – third plural in German (for interesting complications see Comrie 1975). However, in Dalabon, free pronouns themselves are unaffected by the referrals in (50) which only affect the agreement patterns in verbal prefixation. Consequently, what was once pragmatically systematic in Dalabon has become morphologically systematic syncretism. Under this analysis, CONFLICT B, in both its specific and its generalized variants, does not arise, because the 1SG > 2SG, 1DU > 2SG and 1PL > 2SG cells are subject to the referral in (50b). The correct forms are selected by rule (50b). As djah- is the form used for 3SG > 2SG in (29), it will fill the 1SG > 2SG cell in (41). As djirrah- is the form used for the 3DU > 2SG in (29) it will be used for the 1DU > 2SG cell in (34), and as bulah- is used in the 3PL > 2SG cell in (29), it will fill the 1PL > 2SG cell in (34). These correct predictions all follow from the single referral in (50b). Conflict C, in both its specific and generalized variants, is resolved by the rule of referral (50a). The forms used for the 2DU > 1SG and 2PL > 1SG cells are burrah- and bulah-, because these are the forms used for the 2DU > 1SG and 2PL > 1SG cells in (29). The reason why kahis used for the 2SG > 1SG cell in (40) is that this is the form used for the 3SG > 1SG cell. In the case of referral (50a) it generalizes to all first person object cells where the subject is second person. This can be verified by checking Table 2 where the dual first person object proclitic njerr and the plural first person object proclitic njel occur in the non-singular first person object portion of the 2 > 1 paradigm, with in each instance the bound pronominal being the same as for the corresponding 3 > 1 paradigm (for example, njerr kahis both 2SG > 1DU and 3SG > 1DU, and njel burrah- is both 2DU > 1PL and 3DU > 1PL). With the adoption of the referrals in (50), the conflicts resulting from the use of underspecification on its own are resolved. However, the referrals


199

themselves involve a degree of underspecification, because the number of the argument is underspecified (as indicated by the use of the variables in our informal representation in (50)). Underspecification of number in the presence of person makes cross-linguistic sense, of course, because we know that number marking correlates with person values. The referrals themselves also imply a directionality, in that burrah- (A 3DU), bulah- (A 3PL), djirrah- (OBJ 2SG, A 3DU) and djilah- (OBJ 2SG, A 3PL), for example, are now treated as the primary exponents of third person. There is evidence for this. The forms barrah- (S 3DU), and balah- (S 3PL), from which burrah- (A 3DU) and bulah(A 3PL) are derived by rule (30) are uniquely associated with the third person in the intransitive paradigm. In the singular, kahis also uniquely associated with the third person in the intransitive paradigm, while djah-, which marks second person singular object in the transitive paradigm, is uniquely associated with the second person singular subject in the intransitive paradigm. There is no direct evidence that the forms djirrah- and djilah- are primary exponents of 2SG > 3DU and 2SG > 3PL respectively, but as they fill these cells, the referral (50b) will make reference to them in any event. While underspecification alone does not work, the generalized (underspecified) referral approach derives the correct results, is favoured by the paradigminternal evidence and also has a potential original motivation in terms of pragmatically based avoidance based on the top two values of a person hierarchy. We now turn to a brief account of the implementation of this analysis using DATR. 5.3.4 The verbal hierarchy This section takes one sample Dalabon verb nan ‘to see’. Figure 5 illustrates a small part of the lexemic hierarchy in which the verbal lexical items inherit from a node VERB. Like every other verb, verb nan ‘see’ inherits from the node VERB. Its lexical entry is given in (51).

VERB

Nan

Figure 5. Dalabon verbal hierarchy

OTHER VERBS

200 (51)

The Syntax–Morphology Interface Nan: == VERB == see == nan.

In (51) the ‘empty path’ is paired with the node VERB, on its right-hand side. This simply means that information which is not specified at the node 9 Nan will be inherited from VERB. In (52) we give the node VERB. (52)

VERB: == undefined <mor> == MOR_VERB <syn> == SYNTAX <syn cat> == verb.

This node serves to bring together the morphological and syntactic specification of verbs in Dalabon. In (52) the path <mor> refers to the path <mor>, and all of its extensions, at MOR_VERB, the node which provides information about the morphology of verbs. It is the information represented at this node which forms the core of our analysis of the Dalabon verbal system and it is to this that we now turn in the next section. 5.3.5 The shape of the verbal paradigm Recall from x5.1.2 that the order of the attributes in our DATR representation is important, because, as a highly constrained kind of underspecification, it determines what can be inferred by default. We saw in (28) in x5.3.1 that there is no person syncretism in the intransitive paradigm. However, there are still constraints on the possible combinations of person and number. Specifically, person imposes certain constraints. First person inclusive is incompatible with singular, because the lowest cardinality associated with it is two, although morphologically the first inclusive dual and the first inclusive plural look like the singular and dual forms respectively of the other persons. As person determines which numbers are available, we order the person attributes before the number attributes, for a given argument (intransitive subject, transitive subject and object). 9

A path cannot be paired with a node without also making reference to another path and its extensions at that node. By convention, where a path is paired with another node and no overt reference is made to a particular path at the node referred to, then the referring path refers to the identical path at the node referenced. In sum, the empty path at the node Nan refers to the empty path, and its extensions, at VERB. Furthermore, the value for any extension of a path which is not already specified at Nan will be found by looking for a matching path at VERB.

Formal framework and case studies (53)

201

For a given argument of the verb, person attributes are ordered before number attributes.

Given the assumption that underspecification is related to what is crosslinguistically general, (53) entails that there should be some reflex of this cross-linguistically general relationship between person and number in Dalabon. Typically, this would be syncretism of number in the presence of person, but this is not the case for Dalabon. The ordering in (53) is related to our informal description of the referrals in (50). There we used variables to show that the referral does not just involve one cell of the paradigm referring to another, but in fact picks out whole sub-paradigms. The fact that it picks out whole sub-paradigms, of course, shows that the sub-paradigms are defined in terms of person. While avoidance of 1 > 2SG and 2 > 1 is more specific to Dalabon, the involvement of whole subparadigms defined by person will follow from the cross-linguistically general principle in (53). The next step is to determine whether the object attributes should occur before the transitive subject attributes in the paradigm. This question obviously does not arise for intransitive subjects. If we were to base our decision on the ordering of affixes, then the fact that non-singular object clitics appear before the bound prefixes might lead us to claim that object attributes (followed by their person and number attributes) should occur first in the path. We argue, in fact, that this is the case for third and second person object, but not for first person exclusive object. Only when the object is third or second person may there be clear marking of the person of that object on the bound pronominal prefixes. So the second person singular object is clearly marked on the portmanteau bound pronominal prefix. Furthermore, in the presence of second person singular object marking, the disharmonic transitive subject is syncretic with the dual transitive subject. There is also an example of a bound pronominal prefix clearly marking third person: the third person singular higher animacy object (Table 2). The form bvkah- is used only when the object is third person singular, not when it is first person exclusive (41), as indicated by the contrast between the 3 > 1 and 3 > 3 paradigms. Hence this form definitely identifies a third person object. In contrast with this there is no bound pronominal prefix form where an object is clearly marked as first person. From this we conclude that objecthood triggers marking of second and third person singular. For the second person singular it also has an effect on the number marking of transitive subjects.

202


We therefore order attribute combinations of object with second and third person before any attribute combinations of transitive subject (and any associated person and number marking). There is no unambiguous bound prefixal marking of the first person exclusive object. The bound pronominal system has the same forms for the 3 > 3SG paradigm and the 3 > 1SG paradigm (Generalizations 3–6). In other words, for the bound pronominal paradigm third person transitive subjects condition loss of distinction between first person singular and third person singular objects. This leads us to claim that, in contrast with the other persons, the object attributes for first person exclusive occur last in the path. We shall relate these facts to the referrals in (50). The resulting combination of values in the paradigm is therefore as in (54) (54)

Second and third person object attributes are ordered before transitive subject attributes10 Examples: <mor infl o 3rd sg a 2nd sg> <mor infl o 3rd du a 2nd sg> <mor infl o 2nd pl a 1st_excl sg>

In contrast, first person objects are never distinguished by the bound pronominal morphology and so these are treated as occurring last in the path. (55)

First person object attributes are ordered after transitive subject attributes Examples: <mor infl a 2nd sg o 1st_excl sg> <mor infl a 3rd sg o 1st_excl du> <mor infl a 2nd du o 1st_excl pl>

In (54) and (55) we have examples of the shape of the fully specified paradigm. The attribute ordering reflects what is most likely to be underspecified. We can contrast the avoidance-motivated referral involving the top two positions on the person hierarchy (first and second) with the orderings in (54) and (55), which oppose the lower two positions on the person hierarchy (second and third) with the top position, first person exclusive. The analysis which derives the fully specified combinations, such as those in (54) and (55), can be underspecified in relation to these combinations.

10

Following Dixon (1994) in our representation we shall use the attributes a, transitive subject, intransitive subject and object respectively.

s

and

o

for


203

5.3.6 Referral of 1 > 2SG to 3 > 2SG The referral of the 1 > 2SG to the 3 > 2SG combinations illustrates that syncretisms can arise from what is essentially a referral with a degree of underspecification. The underspecified element is the number of the transitive subject. Note that this referral also fits in with the attribute ordering we have determined in (54). As the second singular object attributes are toward the left of the path, they are less likely to be involved in underspecification-based syncretism. This referral is given in (56) below. (56)

<mor infl o 2nd sg a 1st_excl> == <mor infl o 2nd sg a 3rd>

All extensions of paths for 1 > 2SG can be obtained by referring to the extensions of 1 > 3SG. These extensions will, of course, be those for number of the transitive subject, as this is what is not specified in the referral. 5.3.7 Referral of 2 > 1 to 3 > 1 We claimed in (55) that the first exclusive object attributes should be specified as occurring last in any path. We argued that this had to do with the fact that there is no person-specific bound pronominal marking for the first exclusive object, in contrast with the second and third person at least, where the second singular object is clearly marked and there is marking of the higher animacy third singular object. As this means that first exclusive object marking is an extension of the third person paradigm, the referral of 2 > 1 to 3 > 1 takes the form in (57). (57)

<mor infl a 2nd> == <mor infl a 3rd>

This states that the second person transitive subject finds its extensions from the third person transitive subject. If third person object attributes were also ordered after transitive subject attributes, this would mean takeover of the 3 > 3 paradigm by the 2 > 3 paradigm. But this is not the case, because from (55) it follows that only first exclusive object attributes can extend a path which starts with transitive subject attributes, such as in (57). Note the high degree of underspecification involved. The referring paradigm of second person transitive subject obtains its transitive subject number, together with the first exclusive object person and number information, from the third person paradigm. 5.3.8 Dalabon: summing up The referral-based approach allows us to capture the intuition that the syncretism is motivated by the potential inappropriate combination of first

204


and second person. It is transitivity which conditions the loss of person distinction. It also allows us to treat number marking in the transitive and intransitive paradigms in a similar way. Number distinction is affected by person (first inclusive) in the intransitive paradigm, which makes full person distinctions. The best account requires the use of referrals and underspecification so that whole sub-paradigms may be systematically syncretic. Thus both underspecification and referral are required. It is important to check that our Network Morphology analysis does indeed provide the correct forms for the substantial paradigms of Dalabon. Hence the need for an implementation, a part of which was presented above. The full implementation in Appendix 5 does give exactly the right forms for Dalabon, and only these, and to that extent the Network Morphology account is valid. Having seen how underspecification can be combined with referral to provide the best account of syncretism in the Dalabon verbal system, we now go on to consider the relationship between position on a morphological class hierarchy and the domains of syncretism. For this we shall use Russian as the main case study, referring to other languages when data from Russian is insufficient. 5.4

Case study 3: The Russian nominal system

The complex inflectional morphology of Russian offers an illuminating case study (58). Its system of inflectional classes provides helpful data for (58)

Main paradigms of Russian nouns

NOM SG ACC SG GEN SG LOC SG DAT SG INS SG NOM PL ACC PL GEN PL LOC PL DAT PL INS PL

class I ‘law’

class II ‘room’

class III ‘bone’

class IV ‘wine’

zakon zakon zakona zakone zakonu zakonom zakony zakony zakonov zakonax zakonam zakonami

komnata komnatu komnaty komnate komnate komnatoj komnaty komnaty komnat komnatax komnatam komnatami

kost0 kost0 kosti kosti kosti kost0 ju kosti kosti kostej kostjax kostjam kostjami

vino vino vina vine vinu vinom vina vina vin vinax vinam vinami


205

discussing the notions of domain and regularity. Russian has four basic noun paradigms, as given in (58). In a traditional account, these paradigms might be treated as monolithic units. In a Network Morphology account, we would analyse the oblique plural forms as being shared across the paradigms; thus the dative plural of a noun consists of its stem plus -am, irrespective of inflectional class. We would also treat the nominative plural as being stem plus -i (or its orthographic variant -y), with this being overridden just for inflectional class IV. At a lower level, we would capture the shared forms of inflectional classes I and IV. That is, we would have a hierarchy of defaults, the highest applying very generally, in fact applying to more than just nouns, the lowest having smaller domains, and at the bottom of the hierarchy would be lexical items which must contain some idiosyncratic information. A possible structure is given in Figure 6. In addition to the four lexical items given, there are thousands of other nouns which inherit from the four inflectional class nodes. For the detail see Corbett and Fraser (1993). The implementation on which this case study is based is from Brown (1998b), and a simplified version of this is given in Appendix 6.11 The relevance of this approach is that it suggests a range of possible domains for syncretism.

MOR_NOMINAL

MOR_NOUN

MOR_ADJECTIVE

N_0

N_I zakon …

N_IV

N_II

vino … karta …

N_III kost´…

Figure 6. An inheritance structure for Russian nominals 11

The fragment rusnoms.dtr at the Sussex DATR archive http://www.cogs.sussex.ac.uk/lab/ nlp/datr/datr.html precedes Brown (1998b) but still involves separate hierarchies. Corbett and Fraser (1993) gave a single hierarchy. Later papers use a network of hierarchies; in particular a syntactic hierarchy is added, which is concerned with the syntactic category of items; by default, items which are, for example, syntactic nouns will inherit information from the noun section of the morphological hierarchy, but this is not always so.

206


The higher up the hierarchy that a syncretism is stated, the larger the domain of syncretism. And the larger the domain, the more systematic the syncretism. 5.4.1 Domains of syncretism Given that we find numerous instances of syncretism, what are the possible domains? We first separate off phonology and then move to the central potential domains. 5.4.1.1 Phonologically determined ‘syncretism’ Certain syncretisms can be excluded on the grounds that they result solely from a phonological rule; for example, the collapse of genitive singular with nominative/accusative singular in some stem-stressed nouns in Russian, discussed in Chapter 1: x1.5, which is due to the reduction of unstressed vowels. Such instances can be straightforwardly accounted for by the phonology and therefore excluded from treatment within morphology proper. 5.4.1.2 Lexically determined syncretism The least systematic type of syncretism (that involving the smallest domain) would be a syncretism for a single lexical item. We do not find such an instance in Russian. To illustrate what this would look like, consider the paradigm of Italian essere ‘be’ in (59). (59)

Italian essere ‘be’ in comparison with a regular verb ‘be’

1 2 3

‘speak’

singular

plural

sono sei e`

siamo siete sono

1 2 3

singular

plural

parlo parli parla

parliamo parlate parlano

According to Davide Ricca (personal communication) the verb essere ‘be’, which has first person singular and third person plural both as sono, is the only verb in the language to have this identity. This then is a case of lexically determined syncretism, which is equivalent to Coleman’s (1991) first degree syncretism. Other languages show further unique syncretisms, though these are rare.


207

In principle the situation is the same, whether it affects a single item or several items in an unmotivated way: that is, the syncretism would need to be specified in the lexical entry for each of them. Here is a further example of lexically determined syncretism. Numerals in Russian vary in the number of case distinctions made, but the normal minimal paradigm distinguishes three forms. Consider these numerals against that general background: While syncretism of nominative and accusative is widespread in Russian numerals, syncretism of all the remaining case forms is unusual, and is the only morphological feature these items have in common. This combination of syncretisms including the instrumental must be lexically specified for these three items only.12 (60) Lexically determined syncretism in Russian numerals


‘forty’

‘hundred’

‘ninety’

sorok sorok soroka soroka soroka soroka

sto sto sta sta sta sta

devjanosto devjanosto devjanosta devjanosta devjanosta devjanosta

5.4.1.3 Morphologically determined syncretism In this part of our case study of Russian we illustrate how morphological regularity is itself a matter of degree. Referring back to the Russian paradigms in (58) we now consider morphological domains, starting from the smallest. The smallest morphological domain above that which is lexically determined is syncretism within a single inflectional class. We find this in the class we have labelled class III in (58), where there is, among other things, identity of genitive singular and locative singular, repeated here in (61).

12

Sorok ‘forty’ and sto ‘hundred’ are the clear cases; because of vowel reduction all the forms of devjanosto ‘ninety’ are pronounced identically, thus there is an additional phonologically induced syncretism of the nominative and accusative with all the other cases, giving only one phonological form for this item. Note also that GEN/DAT/LOC/INS syncretism is also characteristic of the singular feminine adjectives; see (67) below.

208


We could reflect this in the lexical entry. However, we would be suggesting, counterfactually, that the identity is an idiosyncratic fact about the particular noun kost0 ‘bone’, which is not shared with other items, in other words that the syncretism is lexically determined. In fact, the syncretism of (61) Syncretism within a single inflectional class in Russian class III ‘bone’ NOM SG ACC SG GEN SG DAT SG LOC SG INS SG

kost0 kost0 kosti kosti kosti kost0 ju

locative and genitive singular is one of the characteristics shared by all members of the inflectional class (Coleman’s 1991 second degree syncretism). We should therefore state this syncretism at a higher point in the inheritance hierarchy in Figure 6, at the node N_III, from which kost0 inherits (as do over 4,000 other nouns). This syncretism is evidently more systematic than any lexically stipulated syncretism. It is, however, less systematic than instances we shall come to shortly. There is also a good reason to treat the syncretism in (61) as directional. This is represented with the DATR equation in (62). Again, we have omitted some information, as indicated by the ellipses. (62)

N_III: == MOR_NOUN <mor sg loc> == "<mor sg gen>"

...

The equation in (62) is a referral which states that for class III the form of the locative singular is determined by the form of the genitive singular. The reason for this is that the form of the genitive singular -y in class II, although orthographically different in (58), can be treated as phonologically identical with the exponent -i of class III, suggesting that the stem +i combination is primarily the exponent of genitive singular, and secondarily in class III the exponent of locative singular. This analysis is further justified when we consider syncretism which ranges over more than one inflectional class.


209

In Russian, as can be seen in (58), dative and locative singular are identical in two inflectional classes (II and III). We repeat the relevant paradigms in (63). (63)

Syncretism within a subset of the inflectional classes of a word class

NOM SG ACC SG GEN SG DAT SG LOC SG INS SG

II ‘map’

III ‘bone’

karta kartu karty karte karte kartoj

kost0 kost0 kosti kosti kosti kost0 ju

While the case forms are identical within the inflectional classes, the inflections involved differ between inflectional classes. This example is more systematic than that involving just class III. It is more systematic in two ways. First, many more nouns are involved, all of those in classes II and III. We state the syncretism at the MOR_NOUN node (and then override it for the remaining nouns which inherit from N_O in Figure 6, see Brown (1998b: 257)). Second, the syncretism holds true for two quite different morphological realizations: those nouns like karte ‘map.DAT SG/LOC SG’ and those nouns like kosti ‘bone.DAT SG/LOC SG’, as shown in (63). If we refer back to the example paradigms in (58), we see that the locative singular realization combining a stem with the ending -e is the default for nouns as a whole, because it is found in classes I, II and IV. Furthermore, in classes I and IV, the combination of a stem with the ending -e is reserved solely for the locative singular. This is repeated in (64). (64)

The ending -e as primary exponent of locative singular

NOM SG ACC SG GEN SG DAT SG LOC SG INS SG

I ‘law’

IV ‘wine’

zakon zakon zakona zakonu zakone zakonom

vino vino vina vinu vine vinom

210


Examination of the paradigms in (63) and (64) suggest that the combination of stem and ending -e should primarily be associated with locative singular as a default for nouns. The DAT/LOC syncretism is therefore a referral of dative singular to locative singular. This can be expressed as in (65). (65)

MOR_NOUN: == MOR_NOMINAL <mor sg dat> == "<mor sg loc>" <mor sg loc> == "<stem sg>" e "<stress sg>"

...

Class II, which includes nouns such as karta ‘map’, inherits both equations and therefore combines the syncretic pattern with the default ending -e. The intermediate class N_O, from which class I and class IV inherit, overrides the referral by stipulating that the realization of dative singular is the stem plus the ending -u. Class III, on the other hand, overrides the default realization of the locative singular, while still inheriting the default referral of dative singular to locative singular. When the referral in (65) is combined with the referral specific to class III in (62) we obtain a chain in which the dative singular is based on the locative singular, and the locative singular is based on the genitive singular, thereby yielding the collapse of three case distinctions in class III. Analytically this is important, because the identity of genitive and dative in class III is the product of combining the noun default referral for DAT/LOC with the class III referral for LOC/GEN. With the original Jakobsonian features the DAT/LOC syncretism can be captured only by leaving the realization specified as +peripheral. However, this would also include the instrumental case. In contrast, it is easier to pick out the LOC/GEN syncretism. The GEN/DAT syncretism in class III is problematic for the Jakobsonian feature system, as these cases do not share any values under the unmodified version. It is possible to express some of these generalizations using a modified variant of the Jakobsonian approach (Chapter 3: x3.1.1). The problem is that a decision must be made about which generalizations are to be captured, and this involves precluding other phenomena. For instance, in Mu¨ller’s (forthcoming) approach to syncretism and inflectional allomorphy in Russian, there are two separate vocabulary items, /e/4 and /e/5, the former used for the dative/locative of class II and the latter used for the locative of classes I and IV. The reason for this is that, even though dative and locative form a natural class [-subj, +obl], Mu¨ller also attempts to capture sharing between inflectional


211

classes by using variables, and the variables for classes I and IV exclude those for class II. In fact, scaling up underspecification-based approaches is often a problem when taken in the context of other features. Breaking the features up even further to make underspecification work brings with it the potential for massively increasing the number of elsewheres, which the system underutilizes in any event. When taken in isolation this problem for theories which manipulate the feature values is not as great as when number and case, or lexical information about inflection class, are actually combined, thereby yielding a massive number of potential combinations which never appear. In contrast, the analysis of the noun syncretisms based on atomic values and referrals actually captures the licensing effect that one syncretism may have on others. Having seen an example of a syncretism which covers more than one morphological class within a word class, we now turn to a more general domain, namely to potential examples of syncretism throughout a word class. For a clear case we turn to Russian’s South Slavonic relative, SerboCroat. Here the accusative plural and genitive plural are syncretic for the personal pronouns, but not for nouns and adjectives. In (66) we illustrate this syncretism using the first and third person pronouns (non-clitic).

(66)


Syncretism throughout one word class in Serbo-Croat 1PL

3PL

mi nas nas nama nama nama

oni (M), ona (N), one (F) njih njih njima njima njima

The other pronouns behave in the same way with regard to the PL ACC/PL GEN syncretism. Thus we have an instance which involves all members of the word class. While the word class involved in Serbo-Croat is small with respect to the number of lexical items involved, we have already seen an instance where the lexical class was much larger, namely Finnish (example (13) in Chapter 3: x3.1.2), where syncretism affects all nouns, but not pronouns.

212

The Syntax–Morphology Interface MOR_NOMINAL

LOC PL / GEN PL

MOR_ADJECTIVE

MOR_NOUN

LOC SG / DAT SG

N_0 LOC SG / GEN SG

N_I

N_IV

zakon … vino …

N_II

N_III

karta …

kost ′…

Figure 7. The location of syncretism patterns in the Russian inheritance hierarchy

We also find syncretisms which generalize across more than one word class. In Russian all adjectives and pronouns have the genitive plural syncretic with the locative plural. This does not extend to nouns, where the inflected forms differ, as can be seen in (58). Finally, it is possible to find examples of syncretism across all potentially relevant word classes. In Slovene, the dative dual and instrumental dual are syncretic for nouns, adjective and pronouns (Priestly 1993: 399). That is to say, these forms are syncretic for anything which can mark them. No lexical item of Slovene has a dative dual which is distinct from the instrumental dual. The type of inheritance hierarchy illustrated in Figure 6 suggested possible morphological domains for syncretism, and we have found instances of all of them. In Figure 7 we show where the generalizations are located on the inheritance hierarchy for Russian. The syncretism of locative plural and genitive plural is a property of nominals, because it occurs for pronouns as well as adjectives, but it is overridden by nouns. The locative singular and dative singular syncretism is a default for nouns, but overridden at the node N_O, and the syncretism of the locative singular and genitive singular is a default for class N_III only and is specified there. The locations in the hierarchy represent different degrees of regularity. The higher up the hierarchy a syncretism is stated, the greater the number of items that are likely to be affected. The second aspect to regularity, namely that with inflectional classes a syncretism may generalize over different realizations, was illustrated in (63) above, and we shall discuss a more dramatic instance of it in x5.4.2. Before that we must


213

consider what would be a highly systematic type of syncretism, but one for which we have found no convincing evidence. 5.4.1.4 Candidates for syntactically determined syncretism There are various types of situation which at first appear to be syntactic domains for syncretism. On closer examination they seem to us to be better analysed in other ways. In the earlier discussion we talked of ‘noun’, ‘adjective’ and so on, which suggests that syntactic categories can provide the domain for syncretism. However, the hierarchy given in Figures 6 and 7 simplifies the position. We need a separate hierarchy for syntactic categories, though by default items in a given category inherit their morphological information from a node dominated by a morphological category corresponding to this syntactic category (nouns from MOR_NOUN, adjectives from MOR_ADJECTIVE, and so on). The question then arises as to what happens when syntactic and morphological categories fail to match. Which provides the domain for syncretism? Slavonic provides helpful data here. Consider Russian items like stolovaja ‘dining room’. Syntactically these behave like nouns. However, they behave morphologically as adjectives (67).

(67)

Russian nouns and adjectives

syntactically: morphologically: NOM SG ACC SG GEN SG DAT SG LOC SG INS SG

‘map’ noun noun

‘dining room’ noun adjective

‘new’ (F.SG) adjective adjective

karta kartu karti karte karte kartoj

stolovaja stolovuju stolovoj stolovoj stolovoj stolovoj

novaja novuju novoj novoj novoj novoj

The syncretisms found with morphological nouns and morphological adjectives are different in Russian. Stolovaja ‘dining room’ and similar items have the syncretism of morphological adjectives. This shows that it is the morphological category which provides a domain for syncretism. To demonstrate that a syntactic category is a possible domain, we would need to find the converse situation to that found in Russian, that is a situation in

214


which morphological category and syntactic category differed for certain items, and where these items took the syncretisms appropriate for the syntactic category but not for the morphological category.13 The other area where we might consider syntactic influence on syncretism is with features which may have a role both in morphology and in syntax. (They are to be distinguished from purely morphological features, such as features indicating inflectional class.) Typical examples are number and gender, which are relevant to syntax, notably in agreement, and also to morphology. To see the role of such features we shall consider the interesting issue of animacy, discussed above in Chapter 3: x3.1.2 and Chapter 4: x4.3.4 (also see Perlmutter and Oresˇ nik 1973, Huntley 1980, Corbett and Fraser 1993). Table 3 includes two examples for each of the four main inflectional classes (see (58) above); each of the eight examples is representative of a group of nouns. If we look at any of the examples individually, we find an instance of syncretism. Take just the first example, student. There is syncretism of accusative and genitive singular (conditioned by animacy). More generally, the singular accusative depends on the gender and animacy of the item in question in Table 3, and the plural accusative on the animacy of the item in question. Evaluation of this information allows us to state that the singular accusative of masculine animates is the same as the singular genitive, and the plural accusative of animates is the same as the genitive. The different inflectional paradigms in Table 3 share the same pattern of identity, even though the particular inflections differ. It would clearly be inadequate to state the identity of forms separately for each inflectional class; that would imply that the cases involved could equally well differ from inflectional class to inflectional class. This regularity is captured in the Network Morphology account by a statement high up the inheritance

13

Andrew Carstairs-McCarthy (personal communication) points out a possible example of syntactically conditioned syncretism in Latin. Some third declension adjectives have syncretism of the dative and ablative singular (with the ending -i) when used adjectivally, but when used substantivally, these forms are distinct (dative -i versus ablative -e), e.g. a sapienti viro ‘by a wise man’, a sapiente ‘by a philosopher’ (Kennedy 1955: 40). However, this alternation also obtains within nouns of the third declension. The third declension is divided into two subtypes, the i-stems and the consonant stems: i-stems have the syncretic dative/ablative singular -i, while consonant stems have a distinct dative singular -i versus ablative singular -e. Thus, the choice of forms is not necessarily correlated with syntax, and the alternation between syncretic and non-syncretic patterns in the third declension adjectives might better be described as an alternation between i-stem and consonant-stem declension patterns.

GEN PL

ACC PL

NOM PL

GEN SG

ACC SG

NOM SG

‘law’ inanimate

zakon zakon zakona zakony zakony zakonov

‘student’ animate

student studenta studenta studenty studentov studentov

‘map’ inanimate karta kartu karty karty karty kart

‘teacher(F)’ animate ucˇitel0 nica ucˇitel0 nicu ucˇitel0 nicy ucˇitel0 nicy ucˇitel0 nic ucˇitel0 nic

Table 3 The morphological effect of animacy in Russian

misˇ misˇ misˇ i misˇ i misˇ ej misˇ ej

‘mouse’ animate kost0 kost0 kosti kosti kosti kostej

‘bone’ inanimate

cˇudovisˇ cˇe cˇudovisˇ cˇe cˇudovisˇ cˇa cˇudovisˇ cˇa cˇudovisˇ cˇ cˇudovisˇ cˇ

‘monster’ animate

vino vino vina vina vina vin

‘wine’ inanimate

216


tree. Since animacy also affects the agreement of adjectives, and the form of pronouns, the logical place for it would be at the top of the hierarchy in Figure 7, namely at the MOR_NOMINAL node. There are two questions here, the first is the type of feature we are dealing with, and the second, our main concern, is the domain of syncretism. Animacy appears to be a semantic feature, in that the nouns involved denote entities which live and move (thus insects are animate but plants are not). The match with this semantic definition is close in Russian (less so in some other Slavonic languages). There are some interesting borderline cases, for instance pokojnik ‘the deceased’ is grammatically animate. Such instances are animate for agreement purposes and for the morphological matter of syncretism. But we should also capture the fact that the personal pronouns regularly have accusative/genitive syncretism, whether or not they are referring to an animate entity. This may be captured by stipulating that they are grammatically animate. Our conclusion is that the animacy feature is a morphosyntactic one (albeit one with strong semantic motivation in Russian, and with lesser semantic motivation in some other Slavonic languages). In any case, while animacy is a major determining feature for accusative/genitive syncretism, the syncretism depends on the interaction of animacy with number and gender (see Fraser and Corbett 1995 for more details). Moreover, it is overridden by morphological considerations (inflectional class II nouns have accusative singular in -u irrespective of animacy). In our previous examples the domain given uniquely specified the syncretism in question. The animacy feature does not do this. What then is the domain of the syncretism we have just examined? We shall see in the next section. 5.4.2 Orthogonal specification of syncretism The clearest domains we have encountered were well described in terms of an inheritance hierarchy. However, the interesting data on animacy prove significantly different. The point is that generalizations about syncretism determined by animacy can be stated high up at the level of MOR_NOMINAL in the hierarchy in Figure 7. They apply to different paradigms, giving identities of pattern rather than any phonological identity. It is not the case that accusative/genitive syncretism is found with all items in the nominal domain, rather only those that are animate (whether as a result of their semantics, or being specified as animate, or being marked as animate as a consequence of agreement). Thus the specification is orthogonal to the specification for inflectional class.


217

This captures the common-sense view that we do not wish to claim there are eight inflectional classes for nouns in Russian, rather that there are four main classes, with animacy affecting each of them. The natural way of thinking of domains is in hierarchical terms, and this was our approach. However, syncretism may require specification which is orthogonal to the morphological hierarchy. How then does such orthogonal specification of syncretism differ from lexical specification? Lexical specification means that the individual items must each be specified, in other words that they are exceptional in this regard. Orthogonal specification can be regular (animacy syncretism in Russian shows a very high degree of regularity). The distinguishing point is that it depends on a feature which is otherwise irrelevant to inflection, because it affects only the accusative case, and which cross-cuts the features which determine inflection. We can specify a domain within which it operates, but not a domain where it uniquely determines a syncretism. Animacy in Russian shows that regularity of syncretism is quite a subtle notion. On the one hand, syncretism based on animacy is systematic, because its evaluation potentially determines the accusative plural of every lexical item which is nominal, and therefore it is stated high up in the hierarchy. Furthermore, it is realized in several different ways which have no phonology in common. On the other hand, there are many nouns, for which in the singular it has no effect (as already mentioned, inflectional class II nouns have accusative singular in -u irrespective of animacy).

5.5

Conclusion

We have examined three case studies of unrelated languages, illustrating different aspects of syncretism. The Dhaasanac study required the generalization of the A/B pattern across different tenses. Furthermore, this was a pattern which, while systematic, could not be accounted for using morphosyntactic features alone. By separating the realization of the forms from the morphosyntactic feature specification, Network Morphology makes it possible to account for such systematicity. As it places constraints on feature specification, Network Morphology is still able to relate this language-specific morphological systematicity to the general patterns: the choice of A/B patterns is determined by tense, aspect, mood and negation. In our case study of Dalabon we showed that what was originally a pragmatically determined avoidance of particular forms – namely those involving combinations of first person and second person in the transitive

218


paradigm – had become morphologized. Underspecification cannot deal with this syncretism on its own. The Network Morphology construct of generalized referral involves a degree of underspecification and is therefore able to refer one set of cells in the paradigm to another set of cells. This means that it is not just stating the facts directly, but predicting whole sets of forms. Finally, our case study of Russian showed that degrees of morphological regularity can be defined by locating syncretisms at different places in an inheritance hierarchy. Each of the case studies in this chapter has an implementation associated with it. The fragment for Dhaasanac in Appendix 4 generates the positive perfect and imperfect paradigms, as well as the positive imperative singular, for three example verbs, two of the coronal type and one of the non-coronal type. The Dalabon fragment in Appendix 5 covers 117 forms (102 transitive and 15 intransitive) derived for the verb ‘to see’, and the Russian fragment in Appendix 6 covers a significant proportion of the nominal lexicon. Our approach therefore has justification on two grounds. We have put forward linguistic arguments which favour these approaches, and we have implemented our analyses to demonstrate that linguistic arguments are indeed tenable.

6 Conclusion

6.1

Taking stock

It is worth reflecting on how we have reached this point. We began from a logical typology of syncretism. We asked what might theoretically be found, working up from the simplest possibilities (Chapter 2). This approach is feasible because the phenomenon is finite, and particularly worthwhile because syncretism is complex, and the reality does not correspond to many linguists’ hunches about it. Given this basis we could move on to the cross-linguistic typology of Chapter 3, where we saw that extending the range of languages beyond those most usually treated for syncretism changes the picture dramatically. Among other things it shows that some of the patterns of syncretism which are familiar from Indo-European are actually somewhat exotic. We relied largely on two samples. First of all, our own sample for the Surrey Syncretisms Database, with thirty languages chosen for genetic and geographic diversity but with the entry condition that they must show instances of syncretism. These were investigated in great detail in preparation for this book. The second sample is that of the World Atlas of Language Structures. This sample of 200 languages, also chosen to avoid genetic and areal bias, provides a useful balance in that it was externally selected; of course, some of the languages provided no helpful data since they lack inflectional morphology. While the breadth of these samples is invaluable in ensuring coverage, so that unusual phenomena do not fall through the net, we were not tied to them. This is because the slight differences found between closely related languages (which would not figure in the same sample) can be of great significance too. There are instances in the literature of a good deal of theoretical weight being put on the set-up of a particular language’s syncretisms, which is shown to be misplaced by the inability of the analysis to handle the slightly different system of a closely related language. Having established a solid empirical base we moved on to considering the formal 219

220


tools available for describing the patterns observed (Chapter 4). We established the essential types of pattern, and the basic theoretical devices employed to model them. We noted that some claims for simplicity of analysis were undermined by the need to reintroduce complexity by other means. Some false trails have been introduced into the literature by analyses of interesting but rather limited selections of inflectional systems. We suggest that such analyses need to be justified within an account of the whole system. And given the complexity of some of these systems, the convincing way to validate an analysis is through implementation. In Chapter 5, therefore, we discussed full analyses and implementations of inflectional systems which are particularly significant for the syncretisms they exhibit.

6.2

Results

When viewed in terms of morphosyntactic features, certain patterns are repeated across unrelated languages with sufficient frequency to suggest that there is a basis in function or meaning. Syncretism of case most typically affects the core cases used to express subject (intransitive and transitive) and object. One pattern involves collapse of core case distinctions, often correlated with lower animacy or specificity of the argument. The other major pattern involves the use of some non-core case form in the object or transitive subject function; which non-core case is used is subject to cross-linguistic variation. Syncretism of person, which we examined through subject-person marking on verbs, usually affects non-singular values, most often combining first and second or second and third person. In the case of transitive verbs which agree with both core arguments, syncretism usually involves the collapse of subject-person distinctions in the presence of certain object values (most typically first or second person). Gender syncretism does not lend itself to very clear cross-linguistic generalizations; however, there is a decided tendency for gender values which are distinct in the singular to be syncretic in the non-singular, often with a plausible semantic basis. Syncretism of number usually involves collapse of non-singular values. In Chapter 3: x3.8, analysis of interactions showed that cross-linguistically some features are more likely to be syncretic than others. However, if the prevalence of certain patterns is evidence for a morphosyntactic or semantic basis to syncretism, there is also no lack of isolated, language-specific patterns. Most strikingly, there is evidence of the

Conclusion

221

propagation of patterns that were originally accidental, the result of regular sound change (see especially Chapter 3: x3.7). This suggests that syncretism need not have a basis in the meaning or function of the morphosyntactic values that have been collapsed. As we have seen in Chapter 4, the existence of what appears to be morphologically stipulated syncretism has posed a problem for formal models which attempt to link syncretism directly to feature structure. Nor do more abstract constraints on morphological structure, for example the rejection of directional rules, seem to provide an accurate picture of the sorts of syncretic patterns which do or do not occur.

6.3

Consequences

In a sense, syncretism can be seen as a minor morphological phenomenon, a brief detour on the path between function and form. As such, an account of syncretism need be nothing more than a makeshift patch on an idealized morphological model. But when we take into account the full range of syncretic patterns as seen in the languages of the world, we find that their implications stretch down to the foundations of our conception of morphology. Without syncretism, the structure of inflectional morphology need be nothing more than a direct link between morphosyntactic values and forms. Syncretism, in all its variety, argues for the existence of morphological structure which is, at least in some degree, independent of meaning. This independence is reflected in the assumption of three morphological devices: underspecification, indexing and referrals. Failure to incorporate these devices in a model of morphology will undermine any attempt to develop a sound and coherent theory of features. Without them, we would be required to stretch morphosyntactic features beyond what is reasonable and consistent in order to account for examples of syncretism which involve some degree of morphological systematicity. A final benefit of taking the full range of syncretisms into account is that it gives us a clearer picture of the interaction between morphology and syntax. Syncretism is morphology failing syntax, and it does this in interestingly different ways. The three formal devices (underspecification, indexing and referral) represent successively more extreme deviations from a one-to-one correspondence between morphosyntax and morphology. Underspecification is uninformative, in that it fails to honour all of the relevant functional distinctions. Indexing compounds uninformativeness with an additional, autonomous structure which cuts across

222


morphosyntactic natural classes. Referrals display both these violations, and are misleading as well, in as much as a form which has the function x in one context has the syncretic function x/y in some other context. These interactions between the core components of morphology and syntax provide a unique window into the internal workings of language.

Appendix 1: Case syncretism in the World Atlas of Language Structures sample Indo-European languages are marked with ‘*’ in Appendix 1. 1.1

Languages with syncretism between core cases only

1. Basque 2. Boumaa Fijian 3. English* 4. Harar Oromo

5. Mangarayi 6. Murle 7. Paumari 8. Pitjantjatjara 9. Suena

10. Wambaya 11. Warao

Absolutive/ergative in a number of plural paradigms (Hualde and Ortiz de Urbina 2003). Nominative/accusative in first exclusive (dual, paucal, plural) pronouns (Dixon 1988). Subject/object in singular neuter pronouns. Nominative/absolutive in plural pronouns; absolutive here is an object case, also used for predicate nominals (Owens 1985). Nominative/accusative in the collective demonstrative (Merlan 1982). Nominative/accusative for some nouns (Arensen 1971). Absolutive/ergative in pronouns (Chapman and Derbyshire 1990). Absolutive/ergative in pronouns, nominative/ accusative in nouns (Bowe 1990). Nominative/accusative in personal pronouns. These cases are distinct in interrogative and demonstrative pronouns (Wilson 1974). Absolutive/ergative in pronouns, nominative/ accusative in other nominals (Nordlinger 1998). Nominative/accusative in nouns, 1SG, 1PL and 2SG pronouns do distinguish them (RomeroFigueroa 1997). (Nouns distinguish other cases though: dative, locative, allative, ablative.) 223

224

Appendix 1

12. West Greenlandic 13. Yaqui

14. Yidiny

15. Yukaghir

16. Yup’ik

17. Yurok

1.2

Essentially the same system as in Yup’ik described below (Fortescue 1984). Nominative/accusative-possessive in the plural of all nominals except for the third person pronoun (Dedrick and Casad 1999). Absolutive/ergative in pronouns, nominative/ accusative in other nominals and inanimate deictics (Dixon 1977). Nominative/predicative in proper names, possessed nouns and third person pronouns. The predicative marks intransitive subject or transitive object under focus; otherwise, distinct nominative and accusative are marked (Maslova 2003). Absolutive/relative in non-singular nouns, and in first and second person pronouns. The relative case combines the functions of ergative and genitive (Jacobson 1995). Nominative/accusative in plural and third person pronouns. Only pronouns mark case (Robins 1958).

Languages with syncretism involving core cases and peripheral cases

18. Araona 19. Armenian*

20. Burushaski 21. Chukchi

22. Comanche

Ergative/genitive-possessive for all nominals except 1SG and 2SG pronouns (Pitman 1980). There is no distinct accusative case: inanimate nouns have nominative/accusative, animate nouns and pronouns have dative/accusative. Some declension classes have dative/genitive in the singular, while the pronoun ‘who’ has accusative/dative/genitive (Minassian 1980). Ergative/genitive in all nominals except for singular feminine nouns and modifiers (Berger 1998). Locative/ergative-instrumental in second declension nouns, and in the interrogative pronoun ‘who’ (Skorik 1961–77). Nominative/accusative in some nouns (nominative is -Ø, and one of the allomorphs of the accusative is -Ø). Genitive is marked by a morphophonological

Case syncretism in the WALS sample

23. Finnish 24. French*

25. Georgian

26. German*

27. Greek (Modern)*

28. Hindi*

225

alternation using the accusative as a base (singular, and one of the dual paradigms) or the nominative as a base (plural, and one of the dual paradigms). This alternation affects only final -v and -r. Thus in stems ending in other segments, nominative/ genitive or accusative/genitive (Charney 1993). Accusative/genitive in singular nouns, nominative/ accusative in plural nouns (Fromm 1982). Accusative/dative (direct/indirect object) for nonthird person pronouns. (No case distinctions for first and second plural pronouns, unless possessive form is considered.) Absolutive/instrumental/genitive and dative/ adverbial in attributive adjectives. Genitive/ ergative/dative in the so-called Old Georgian plural form (Hewitt 1995). In the definite article and strong adjectives (which display the maximal range of distinctions), nominative/accusative in the feminine and neuter singular, and in the plural. Feminine singular also has dative/genitive. Nouns mark only the dative in the plural, and in the masculine and neuter singular, the genitive and (marginally) the dative. N-stems (nouns and weak adjectives) distinguish nominative (masculine) or nominative/accusative (feminine and neuter) from a peripheral casecum-plural form (masculine and neuter n-stem nouns may distinguish genitive singular). Nominative/accusative in the singular of feminine and neuter nouns, and in the plural of i-stem nouns. Accusative/genitive in the singular of masculine i-stem nouns and in the plural of first and second person pronouns (Holton, Mackridge and Philippaki-Warburton 1997). Direct/oblique in the singular of feminine nouns and the plural of first and second person pronouns. Oblique singular/direct plural in masculine nouns. Compounds of the previous patterns are found in other paradigms (McGregor 1977).

226

Appendix 1

29. Ingush

30. Irish*

31. Kayardild 32. Krongo 33. Lak1 34. Latvian*

35. Lezgian

36. Martuthunira 37. Nenets

38. Ngiyambaa

1

Nominative/genitive in the plural of some nouns, ergative/genitive in the singular of others. All peripheral cases are syncretic in attributive adjectives (Nichols 1994). Both cases (nominative-accusative and genitive) syncretic in singular of one noun class. The irregular noun bean ‘woman’ has two forms: nominative-accusative and one serving the other functions (Bammesberger 1982). Nominative/locative in first declension nouns (Evans 1995). Nominative-accusative/dative in pronouns (Reh 1985). Absolutive/ergative in pronouns, ergative/ genitive in noun. (Zˇirkov 1955). Nominative/accusative in the plural of feminine nouns and adjectives. Nominative/genitive in the singular of some declension classes. Various types of ‘polarity effects’ between singular and plural cases (Mathiassen 1997). Ergative/inessive in the singular of nouns with stems in -a or -e, and the 1SG pronoun (Haspelmath 1993). Accusative/genitive in all but vocalic stem nominals, and in 1SG and 2SG pronouns (Dench 1995). Accusative/genitive in the singular of second person pronoun and nouns with a first person possessor. Nominative/accusative in the dual of nouns with second or third person possessor, and in the plural of any possessed noun. Combination of both patterns in the dual of unpossessed nouns, and nouns with a first person possessor (Salminen 1997). Absolutive/ergative in first and second person pronouns, absolutive/accusative in other nominals. Ergative/dative in some noun classes (Donaldson 1980).

Note that this value has changed from that found in Baerman and Brown (2005a).

Case syncretism in the WALS sample 39. Russian* 40. Spanish*

1.3

227

Wide variety of patterns; see Chapter 3: x3.1 for a sample. Pronouns maximally distinguish three forms (subject, object and prepositional). The last two are distinct only for 1SG and 2SG (Butt and Benjamin 2000).

Languages where inflectional case marking is absent or minimal

(A language is classed as having minimal case marking if the nominal paradigm contains no more than two forms. In such a language the absence of any case distinction is equivalent to the complete absence of case inflection.) Abipon, Abkhaz, Acoma, Ainu, Alamblak, Amele, Apurina, Arabic (Egyptian), Asmat, Awa Pit, Aymara, Bagirmi, Bambara, Barasano, Beja, Bribri, Bukiyip (Mountain Arapesh), Burmese, Cambodian, Canela-Kraho, Carib, Cayuvava, Chalcatongo Mixtec, Chamorro, Coast Tsimshian, Cree, Daga, Dani, Dehu, Diola-Fogny, Ekari, Ewe, Fur, Gooniyandi, Grebo, Guarani, Haida, Hanis Coos, Hebrew, Hixkaryana, Hmong Njua, Igbo, Indonesian, Iraqw, Jakaltek (Popti’), Japanese, Kanuri, Kapau, Karo Batak, Karok, Kawesqar, Kayah Li, Kera, Kewa, Khasi, Khmu, Kilivila, Kiowa, Kiribatese, Kobon, Kongo, Korean, Koromfe, Koyraboro Senni Songay, Kunama, Kutenai, Lakhota, Lango, Lavukaleve, Lealao Chinantec, Luvale, Maba, Malagasy, Mandarin Chinese, Maori, Mapuche, Maranungku, Marind, Mataco, Maung, Maybrat, Mundari, Nama, Navajo, Ndyuka, Ngiti, Nkore-Kiga, Otomi, Paamese, Pasamaquoddy, Persian, Piraha, Rama, Rapanui, Sango, Sanuma, Selknam, Sentani, Shipibo-Konibo, Slave, Squamish, Supyire, Swahili, Taba, Tagalog, Tamazight, Tetelcingo Nahuatl, Thai, Tiwi, Tlingit, Tukang Besi, Tunica, Ungarinjin, Urubu-Kaapor, Usan, Vietnamese, Wari, Wichita, !Xu, Yagua, Yimas, Yuchi, Zulu (123 languages). 1.4

Languages where inflectional case marking is not syncretic

Bawm, Brahui, Cahuilla, Copainala Zoque, Dongolese Nubian, Epena Pedee, Evenki, Garo, Hausa, Hungarian, Hunzib, Ika, Imonda, Kannada, Ket, Khalkha, Koasati, Ladakhi, Lepcha, Maricopa, Meithei, Nez Perce, Nivkh, Nunggubuyu, Paiwan, Quechua (Imbabura), Sierra Miwok, Southeastern Pomo, Trumai, Turkish, Una, Wardaman, Witoto, Yoruba (34 languages).

Appendix 2: Person syncretism in the World Atlas of Language Structures sample

2.1

Languages with syncretism of subject person in verbs which mark a single argument

1. Abipon 2. Amele

1/3 in fourth conjugation verbs (Najlis 1966). 2/3 in dual and plural. 1SG/2SG in remote past and negative past. In some other paradigms these patterns are combined with syncretism of number (Roberts 1987). 3. Arabic (Egyptian) 2SG M/3SG F, and 2DU/3DU F in imperfect paradigms (Mitchell 1962). 4. Awa Pit 2/3 in statements and 1/3 in questions (Curnow 1997). 5. Aymara 1/2 in singular and plural in the subjunctive, and in the so-called ‘third form’ paradigms (Deza Galindo 1992). 6. Bagirmi In consonant-intitial stems, 1SG/2SG/3 in the indefinite aspect, and 2SG/3PL in the definite aspect. 1PL either is included in the previous syncretic values, or is syncretic with 2PL. 1PL/3PL optional in vowel-initial stems (Stevenson 1969). 7. Beja 1SG/3SG M in past tense (Hudson 1974). 8. Burushaski Many different patterns depending on tense/ mood: 1/2 and 2/3 in both singular and plural, and some paradigms have 1/2/3 in plural (Berger 1998). 9. Canela-Kraho 1/3 in one class of verbs (Popjes and Popjes 1986). 10. Carib 2PL/3PL in present tense copula (Hoff 1968). 11. Cayuvava 1INCL PL/1EXCL PL in copula (Key 1967). 12. Chukchi 2SG/3SG (Skorik 1961–77). 13. Daga 1SG/3PL in past (Murane 1974). 228

Person syncretism in the WALS sample 14. Dani 15. Diola-Fogny 16. Dongolese Nubian

17. Ekari 18. English

19. Ewe 20. French 21. German

22. Harar Oromo 23. Hebrew 24. Hindi 25. Hixkaryana 26. Ika

27. Iraqw 28. Irish

29. Kannada 30. Kapau

31. Karok 32. Ket 33. Kewa

229

2PL/3 in hypothetical mood (Bromley 1981). 2SG/PL in the so-called ‘full form’, 3SG/1PL the ‘stripped form’ (Sapir 1965). 2SG/3SG and 1PL/2PL. In the negative indicative, all values except 3PL are syncretic (Armbruster 1960). 1SG/3SG F and 2SG/1PL (Drabbe 1952). 1SG distinct from 3SG and ‘elsewhere’ form in the verb ‘be’, otherwise 1SG not distinct from ‘elsewhere’ form. 2SG/3SG optional for verbs in sentence-medial position (Westermann 1930). 2SG/3SG and 1SG/2PL in future, 2SG/3SG/3PL in conditional, 1SG/2SG/3(SG) in present. 1PL/3PL throughout, 1SG/3SG in preterite, and 3SG/2PL of verbs lacking vowel alternation in 2SG/3SG. 1SG/3SG M and 2SG/3SG F (Owens 1985). 2SG M/3SG F in future (Berman 1997). 2SG/3SG and 1PL/3PL (McGregor 1977). 1PL EXCL/3 (Derbyshire 1979). 1PL/2PL except in future and with experiencer verbs, 1SG/3SG has the same restriction, plus it does not occur in the distal past (Frank 1990). 2SG/3SG F in main verbs, and 1/2 in the selector and the locative existential verb (Mous 1993). 3SG/2PL in preterite subjunctive, conditional, and in habitual past, 2/3 in present indicative, and 1SG/ 2/3 elsewhere (Bammesberger 1982). 1SG/2SG in literary past tense, and optionally in negative (Sridhar 1989). 2/3 in dual and plural of a number of the numerous tense/aspect/mood paradigms, 1SG/ 2PL/3PL in others (Oates and Oates 1968). 2SG/3SG and 2PL/3PL in negative verbs (Bright 1957). 1SG/3SG M where person-marking prefix precedes a vowel-initial base (Verner 1999). 1SG/2SG in the perfect of active verb bases marking egocentric benefaction (Franklin 1971).

230

Appendix 2

34. Kiowa

35. Kobon 36. Kongo 37. Kunama 38. Lak

39. Lango

40. Latvian 41. Lavukaleve 42. Lealao Chinantec 43. Luvale

44. Marind 45. Murle 46. Nama 47. Nenets 48. Ngiti 49. Nivkh 50. Nkore-Kiga 51. Nunggubuyu

1DU-PL/3, where the number value of ‘3’ depends on the noun class of the referent: plural if class I, singular if class II, singular and plural if class III (Watkins 1984). 2DU/3DU; 2SG/3SG for medial same-subject verbs (Davies 1981). 2SG/3SG for indicative verbs (Carter and Makoondekwa 1979). 2PL/3PL in suffix conjugation, 1SG/2SG in class IIa irregulars (Bender 1996). 1/2 in either singular or both singular and plural in some tense/aspect paradigms, 2/3 in transitives, except for present imperfective (Xajdakov 2001). 1DU/3PL in subjunctive, plus 1PL in habitual and progressive, plus the 3SG form used in the perfect with the 3SG pronoun and relativized subjects (Noonan 1992). 2SG/3 in the future, and in the present of some stem classes. (Mathiassen 1997). 1INCL DU-PL/2PL (Terrill 2003). 1SG/3 in transitives with an animate object (Rupp 1989). 2SG/3SG, where 3SG is of class II, and, under some conditions, of class I (animates). 1PL/3PL, where 3PL is of class VII (Horton 1949). 3SG/2PL in various paradigms, depending on tense and choice of prefixes (Drabbe 1955). 1SG/3SG/1INCL PL in the subjunctive; 2SG/3 in the imperfect in one conjugation class (Arensen 1971). 2SG F/3SG F (Hagman 1977). 2DU/3DU in transitives where the object is dual or plural (Salminen 1997). 1INCL PL/3 indefinite, when an overt pronoun is used (Kutsch Lojenga 1994). 2SG/3SG versus another form for all other values (Gruzdeva 1997). 1SG/3PL, where 3PL is of class 12/13 (Taylor 1985). 1/2 in dual and plural of a number of tense/aspect/ mood paradigms (Heath 1984).

Person syncretism in the WALS sample 52. Otomi1 53. Sentani 54. Spanish 55. Suena 56. Swahili 57. Tiwi 58. Usan 59. Yimas

60. Yukaghir

61. Zulu

2.2

231

2/3 in the antecopreterite paradigm (Hekking 1995). 2SG/3DU in the adhortative (Cowan 1965). 1SG/3SG in the present subjunctive and imperfect (Butt and Benjamin 2000). 2/3 in the dual and plural of the remote tense (Wilson 1974). 2SG/3SG, where 3SG is of class 3/4 or 11/10 (Ashton 1947, Myachina 1981). 1PL/2PL; 1SG/3SG F in past (Osborne 1974). 2/3 in both singular and plural of medial switch reference verbs (Reesink 1987). 1/2 in paucal, though note that the corresponding free pronoun also combines these values (Foley 1991). 1SG/2SG in transitive verbs with object focus. All values except 3PL syncretic in intransitives with subject focus (Maslova 2003, Nikolaeva and Xelimskij 1997). 2SG/3SG, where 3SG is animate or of class III/IV (Oxotina 1961).

Languages where subject person is marked inflectionally, but is not syncretic

(Note that this does not include instances where subject-person marking interacts with object-person marking, for which see Appendix 3.) Abkhaz, Acoma, Ainu, Alamblak, Apurina, Armenian, Asmat, Barasano, Basque, Bawm, Brahui, Bukiyip (Mountain Arapesh), Cahuilla, Chalcatongo Mixtec, Chamorro, Coast Tsimshian, Copainala Zoque, Cree, Evenki, Finnish, Fur, Georgian, Gooniyandi, Grebo, Greek (Modern), Guarani, Hanis Coos, Hungarian, Hunzib, Ingush, Jakaltek (Popti’), Kanuri, Karo Batak, Kilivila, Koasati, Krongo, Kutenai, Lakhota, Maba, Malagasy, Mangarayi, Mapuche, Maranungku, Maricopa, Mataco, Maung, Maybrat, Mundari, Navajo, Nez Perce, Oneida, Paamese, Paiwan, Pasamaquoddy, Paumari, Persian, Quechua (Imbabura), Russian, Sango, Sierra Miwok, Slave, Squamish, Taba, Tamazight, Tetelcingo Nahuatl, 1

Data for Otomi added subsequent to Baerman and Brown (in press b).

232

Appendix 2

Tlingit, Tukang Besi, Tunica, Turkish, Una, Ungarinjin, Urubu-Kaapor, Wambaya, Wardaman, West Greenlandic, Wichita, Witoto, Yuchi, Yup’ik, Yurok (80 languages). 2.3

Languages with no inflectional marking of subject person

Araona, Bambara, Boumaa Fijian, Bribri, Burmese, Cambodian, Comanche, Dehu, Epena Pedee, Garo, Haida, Hausa, Hmong Njua, Igbo, Imonda, Indonesian, Japanese, Kawesqar, Kayah Li, Kayardild, Kera, Khalkha, Khasi, Khmu, Kiribatese, Korean, Koromfe, Koyraboro Senni Songay, Ladakhi, Lepcha, Lezgian, Mandarin Chinese, Maori, Martuthunira, Meithei, Ndyuka, Ngiyambaa, Piraha, Pitjantjatjara, Rama, Rapanui, Sanuma, Selknam, Shipibo-Konibo, Southeastern Pomo, Supyire, Tagalog, Thai, Trumai, Vietnamese, Warao, Wari, !Xu, Yagua, Yaqui, Yidiny, Yoruba (57 languages).

Appendix 3: Syncretism in two-place verbs in the World Atlas of Language Structures corpus Languages marked with ‘þ’ also have person syncretism in one-place verbs (which may be recapitulated in two-place verbs, but is not recorded separately here). 3.1

Languages where a distinct pattern of subject-person syncretism is found in two-place verbs

1. Bininj Gun-Wok

2. Georgian

3. Gooniyandi 4. Guarani 5. Maba

6. Mangarayi 7. Ungarinjin 8. West Greenlandic 9. Yimasþ

Subject person not distinguished where object is first or second person, unless the subject is 1SG, or the object is 1SG and the subject singular (Evans 2003). In 1 > 2 constructions, where one argument is plural, the form is the same as 3SG > 2PL. This applies only to the present tense of verbs with 3SG ending -s (Hewitt 1995). Subject person not distinguished when object is first person (McGregor 1990). Subject person not distinguished where object is first person (Gregores and Sua´rez 1967). Subject person not distinguished where object is first person. Syncretism of 1/3PL where object is second person (Tucker and Bryan 1966). Person of subject not distinguished where object is first or second person (Merlan 1982). Subject person not distinguished where object is first or second person (Rumsey 1982). In conditional, syncretism of 1SG/2SG subject where object is 3rd person (Fortescue 1984). Syncretism of 2SG/3SG subject where object is 3SG (Foley 1991). 233

234

Appendix 3

10. Yurok

3.2

Languages with syncretism of object person in two-place verbs

11. Iraqwþ

12. Nkore-Kigaþ 13. Nunggubuyuþ 14. Kanuri

3.3

Syncretism of 2/3SG subject where object is 3PL. (Here a 1PL passive form is used; where object is 3SG, 1PL active intransitive form is used (Robins 1958).)

The selector distinguishes object person, with syncretism of 1SG/2SG M/3PL, and of 3SG M/2SG F – except that where both arguments are third person, both subject and object are distinguished (Mous 1993). Syncretism of 2PL/3PL object, regardless of subject (Taylor 1985). Syncretism of 1PL/2PL/1INCL DU object where subject is 3rd person (Heath 1984). In conjugation class I, syncretism of 1SG/3rd object where subject is 3SG, and syncretism of 1/3rd object where subject is 3PL (Hutchison 1981).

Languages with both subject and object person syncretism

15. Karokþ

16. Marind

Subject is not distinguished if object is 1PL or 2PL. Syncretism of 2SG/3 object where subject is 1PL. In the positive indicative and optative, syncretism of 3PL/1PL object where subject is 3PL. In the negative indicative, syncretism of 2PL/3PL subject where object is 3SG, and of 2PL/3 subject where object is 3PL (Bright 1957). For all paradigms, syncretism of 2PL/3PL object where subject is third person. Other patterns limited by tense/aspect/mood (i) person of plural subject not distinguished where object is first person, (ii) syncretism of 2SG/3SG subject where object is 3SG, (iii) object person not distinguished where subject is first person, (iv) singular subject not distinguished where object is 2SG, and (v) syncretism of 1SG/3SG subject, regardless of object (Drabbe 1955).

Syncretism in two-place verbs in the WALS corpus 17. Sentaniþ

3.4

Person of dual and plural subject not distinguished where object is second person. Syncretism of 3SG/2DU subject regardless of object. Syncretism of 1SG/3SG object regardless of subject (Cowan 1965).

Languages with other patterns of person syncretism in two-place verbs

18. Asmat 19. Caribþ 20. Maricopa Nunggubuyu (see no. 13 above)

3.5

235

There is a single form for 3SG > 1SG/2SG and 3PL > 2PL (Voorhoeve 1965). Where both first and second person are involved, the 1INCL form is used (Hoff 1968). The form for 1 > 2 is the same as that for 3 > 1 (Gordon 1986). With the A series prefixes, the form for 1 > 2SG is the same as that for 3SG F > 3SG M, and the form for 2PL > 1 is the same as that for 1INCL PL > 3, where 3 belongs to the ‘ANAwu’ gender (Heath 1984).

Languages with two-place verbs, where this is not associated with a distinct pattern of person syncretism

Abiponþ, Acoma, Ainu, Alamblak, Apurina, Basque, Bawm, Bukiyip, Cahuilla, Cayuvavaþ, Chukchiþ, Coast Tsimshian, Copainala Zoque, Cree, Daniþ, Ekariþ, Eweþ, Grebo, Hanis Coos, Hixkaryanaþ, Ket, Kilivila, Kiowaþ, Kongoþ, Kunamaþ, Kutenai, Lakhota, Lavukaleveþ, Limbu, Maung, Mundari, Nez Perceþ, Oneida, Paamese, Paiwan, Sierra Miwok, Squamish, Tetelcingo Nahuatl, Tlingit, Tukang Besi, Tunica, Wambaya, Wardaman, Wichita, Yuchi (45 languages).

Appendix 4: DATR fragment for Dhasaanac case study File: Purpose: Author: Documentation:

dhaasanac.dtr A and B form syncretism in Dhaasanac Dunstan Brown (March 2004) University of Surrey, Guildford GU2 7XH Chapter 5: x5.2; data from Tosco (2001)

Here and in Appendices 5 and 6, examples which appear in Chapter 5 are noted as such by the use of bold in the comments which precede them. DATR is distinguished from comments by use of courier font. This fragment generates the positive perfect and imperfect paradigms for three example verbs, two coronals ( fu´r ‘open’ and leét ‘fall down’), and one non-coronal (rok ‘grind’), as well as the positive imperative singular, which is the stem plus high tone (Tosco 2001: 114). The positive imperative singular is the citation form and is therefore used for labelling the example lexical entries. The purpose of this fragment is to demonstrate that we can generate different A and B forms for verbs of different types. The nodes FORMS, CORONALS, NON_CORONALS: these form a hierarchy, with the nodes CORONALS and NON_CORONALS inheriting from FORMS. FORMS therefore makes generalisations about the forms for both types. The node FORMS: the first equation states that the perfect ends in -i (see Tosco 2001: 113). This ending will be preceded by the appropriate consonant grade, which will depend on whether the form is the A or B form, and whether the verb inherits from CORONALS or NON_CORONALS. FORMS: = = "<stem initial> " " " -i

The next equation at FORMS creates the forms for the imperfect. This is done by evaluating the path <stem type> for each lexical item. In this 236

DATR fragment for Dhasaanac case study

237

fragment either a lexical item will have one vowel associated with its stem, or two. = = ""

If the lexical item has one vowel associated with it, then it will combine that one vowel with the values for <stem initial> The values for <stem initial> are defined at the VERB node, but their effect is to create an initial stem which has one vowel, if the lexical item has only one vowel, and two vowels, if the lexical item has two vowels. Hence, for items which have only one vowel, this vowel is repeated for (associated with the imperfect). This gives us a partial implementation of the Bimoraic Filter (Tosco 2001: 125–6), namely by Stem Adjustment. Nasal extension and reduplication are not dealt with here. The evaluable path at the end of the equation deals with vowel harmony. = = "<stem initial> " "<stem vowel_1> " " " " "

The equation below concatenates <stem initial>, the appropriate consonantal grade () and evaluates the final vowel of the stem to determine the vowel of the imperfect. In this case, stems with two vowels do not need to be augmented for the imperfect. = = "<stem initial> " " " " ".

The node CORONALS: this inherits from the node FORMS. CORONALS: = = FORMS

In the two equations below, the stem final element in the lexical entry is evaluated to determine the A and B forms. = = = =

The equations below represent the consonant gradation for the B forms of verbs. Consonant gradation assigns the final-stem consonant for B forms. This means that by default the grade associated with

238

Appendix 4

will be the same as what is specified for a lexical item as its final element of the stem. = = "<stem final>"

The spirantization-II rule (Tosco 2001: 21) has been approximated here as a rule which entails that the A form of stems ending in -t will be _ð. = = _ð

The equations below correspond to Table 6 in Tosco (2001: 128). As an example, the equation = = _ºº can be understood as follows: the B form of a stem ending in r has ºº in place of r. The other equations are interpreted in the same way. = = _ºº = = _t = = _ºº = = _nn = = _t = = _ll.

The nodes NON_CORONALS: this inherits from the node FORMS. NON_CORONALS: = = FORMS = = non_coronal = = "<stem initial>" "" " " = = "<stem initial> " " " " "

Non-coronals drop their last stem consonant when it is followed by a suffix. This means that the non-coronal is dropped in all verbal forms with the exception of the imperative singular, and the a-form of the short past (Tosco 2001: 132). The B form of non-coronals affix yy before the paradigmatic vowel (Tosco 2001: 134). We have treated these two phenomena together, by associating the consonantal grade for A forms as truncated, and the consonantal grade for B forms as yy. If the short past were included in this fragment, it would require the first of these statements to be overridden.


239

= = = = _yy.

The node VERB (example (24) in Chapter 5: x5.2) VERB: = = <syn> = = verb

Example (24) in Chapter 5: x5.2 The path and its extensions associates the morphosyntax with the appropriate index. Second person, third person singular feminine, and first person plural are associated with index B. All other person and number combinations are associated with index A by default. = = _A = = _B = = _B = = _B

Example (24) in Chapter 5: x5.2 The forms of the imperfect positive and perfect positive are determined by evaluating the index associated with the particular person and number information. For example, the form of the second person (of either number) associated with <mor pos imprf> – the morphosyntax of the imperfect positive – will require the form for , once the path is evaluated. On the other hand, the form of the first person singular, for example, will require the form for . These indexed forms are then determined by the lexical item’s membership of a particular form class (coronal or non-coronal). <mor pos imprf> = = " " <mor pos prf> = = " "

By default there is only one stem vowel and so the the value for is ‘undefined’.

<stem

vowel_2>

<stem vowel_2> = = undefined

The initial form of the stem <stem a second vowel, <stem vowel_2>.

initial>

depends on whether there is

<stem initial> = = <stem "<stem vowel_2> " >

As the default is for <stem vowel_2> to be undefined, the stem will consist of a consonant and a vowel.

240

Appendix 4 <stem undefined> = = "<stem cons_1> " "<stem vowel_1> "

Otherwise, if <stem vowel_2> is specified lexically, the stem will consist of a consonant and two vowels, together with a value for tone <stem tone>. <stem> = = "<stem cons_1> " "<stem vowel_1> " "<stem tone> " "<stem vowel_2> "

However, as positive verbal forms are unaccented (Tosco 2001: 41), the default for stem tone is set to lack of tone. This is done purely because we are dealing with the positive paradigms. But the value could be different, if this analysis were extended. <stem tone> = =

The form of the positive imperative singular is the the initial stem plus high tone, and the final element of the stem. <mor pos impv sg> = = "<stem initial high_tone> " "<stem final> "

The position of the high tone is determined by evaluating whether the stem has two vowels or not. <stem initial high_tone> = = <stem high_tone "<stem vowel_2> " >

If the vowel for <stem vowel_2> is ‘undefined’, then the high tone will be associated with the vowel of the mononsyllabic stem. <stem high_tone undefined> = = "<stem cons_1> " "<stem vowel_1> " "<stem tone high_tone> "

High tone is represented by . <stem tone high_tone> = =

The equations below implement the vowel harmony rule for the imperfect. If the final vowel of the stem is o, then the affix of the imperfect is o. If the final vowel of the stem is e, then the affix is e. Otherwise, the imperfect affix is a. = = -e = = -o = = -a.


241

Example verb lexical entries The verb ro´k belongs to the class of non-coronals. The fragment generates forms for the positive perfect and imperfect paradigms and positive imperative singular. These have been checked against Tosco (2001: 451). However, the B form of the imperfect generated is as given in Tosco (2001: 134). Reduplicated variants are not generated by this fragment. R˘k: = = VERB = = NON_CORONALS = = ‘grind’ <stem type> = = one_vowel <stem cons_1> = = r <stem vowel_1> = = _o <stem final> = = _k.

The verb Leét belongs to the class of coronals. The fragment generates forms for the positive perfect and imperfect paradigms and positive imperative singular. These have been checked against Tosco (2001: 432). L¯et: = = VERB = = CORONALS = = ‘fall down’ <stem type> = = two_vowels <stem cons_1> = = l <stem vowel_1> = = _e <stem vowel_2> = = _e <stem final> = = _t.

Example (23) in Chapter 5: x5.2 The verb fu´r belongs to the class of coronals. The fragment generates forms for the positive perfect and imperfect paradigms and positive imperative singular. These have been checked against Tosco (2001: 383). F˙r: = = VERB = = CORONALS = = ‘open’ <stem type> = = one_vowel <stem cons_1> = = f <stem vowel_1> = = _u <stem final> = = _r.

Appendix 5: DATR fragment for Dalabon case study

File: Purpose: Author: Documentation:

dalvbs.dtr referrals and avoidance in Dalabon Dunstan Brown (September. 1999) Chapter 5 (x5.3); data from Evans, Brown and Corbett (2001)

Dalabon is an Australian language of the Gunwinyguan family, spoken in central Arnhem Land by a dwindling population now reduced to about twenty fluent speakers. This fragment implements the analysis given in case study 2 in x5.3 and the associated work in Evans, Brown and Corbett (2001a). We show that syncretism of 2 > 1 (second person subject and first person object) with the 3 > 1 paradigm can be accounted for by the use of referrals of the 2 > 1 paradigm to the 3 > 1 paradigm. Similarly, the 1 > 2 sg paradigm refers to the 3 > 2 sg paradigm. This is a formalization of the idea that ‘person disguise’ is involved, in that the combinations of second and first person are too direct and need to be disguised. In Dalabon this strategy has been morphologized. Our analysis also assumes a particular structure to the Dalabon verbal paradigm. This means that the object attributes are ordered first in the path for all persons except first person exclusive, for which they are ordered last. This analysis generates 117 forms (102 transitive and 15 intransitive forms). No TAM marking is included, and so the forms generated belong to the the subord1 series. The difference between these forms and those found in Table 2 is that the latter have the marker for realis attached, namely -h. These realis forms have also been checked and are correct. The associated theorem dumps for this fragment can be viewed at http:// www.surrey.ac.uk/LIS/SMG/dalabon/. DATR is distinguished from comments by use of courier font. 242

DATR fragment for Dalabon case study

243

The node VERB Example (52) in Chapter 5 In a complete analysis of the morphology of Dalabon the lexeme node VERB would inherit from a higher node WORD information about words in general. As this fragment covers only part of the verbal system, we have stated that any other extensions of the empty path for VERB, if not given at the VERB node, are undefined. The second equation states that the morphology of verbs is found at the MOR_VERB node. The third refers to a node SYNTAX, which defines a verb phrase, and the fourth states that the syntactic category of this class of items is ‘verb’. VERB: ¼ ¼ undefined <mor> ¼ ¼ MOR_VERB <syn> ¼ ¼ SYNTAX <syn cat> ¼ ¼ verb.

Verbal morphology defaults to nothing and verbal inflection consists of some cluster (cl) before a verbal root. MOR_VERB: <mor> ¼ ¼ <mor infl> ¼ ¼ <mor cl> ""

Referrals Example (56) in Chapter 5 The 1 > 2 sg object paradigm refers to the 3 > 2 sg object paradigm. <mor infl o 2nd sg a 1st_excl> ¼ ¼ <mor infl o 2nd sg a 3rd>

Example (57) in Chapter 5 For extensions of the second person transitive subject paradigm refer to the third person transitive subject paradigm. Because of our ordering of attributes, the first person exclusive object attributes constitute the only possible extension. This accounts for the 2 > 1 paradigm. <mor infl a 2nd> ¼ ¼ <mor infl a 3rd>

The 3 dis > 2 sg cell refers to the 3 du > 2 sg cell. <mor infl o 2nd sg a 3rd dis> ¼ ¼ <mor infl o 2nd sg a 3rd du>

244

Appendix 5

The Morphological Cluster The morphological cluster will default to bound prefixal (<mor prefix>) marking. Clitic forms are followed by a plus sign. With the exception of the first person exclusive clitics, the clitic forms are specifically given as being followed by bound prefixal morphology (<mor prefix>). The realizations of first exclusive dual and plural just specify the clitic pronoun forms. Another equation specifies that third person bound prefixes may be preceded by clitics (see formants and equations for the ‘front’ element of third person prefixes). Of course, first person exclusive object clitics are also followed by the bound prefixal morphology, but we state this later in terms of them preceding it. The attribute o means ‘object’. <mor cl> ¼ ¼ <mor prefix> <mor cl o 1st_excl du> ¼ ¼ njerrþ <mor cl o 1st_excl pl> ¼ ¼ njelþ <mor cl o 1st_incl du> ¼ ¼ njehþ <mor prefix> <mor cl o 1st_incl pl> ¼ ¼ ngorrþ <mor prefix> <mor cl o 2nd du> ¼ ¼ norrþ <mor prefix> <mor cl o 2nd pl> ¼ ¼ nolþ <mor prefix> <mor cl o 3rd du> ¼ ¼ bunuþ <mor prefix> <mor cl o 3rd pl> ¼ ¼ buluþ <mor prefix>

Intransitive prefixes Subject prefixes are created by putting together front and final elements (see formants). The front element may combine person and number marking, and the final element may combine intransitive subject (s) or transitive subject (a) marking with number. Note that the 1st_inclusive dual combines a front formant which marks first inclusive dual, and a final element which marks singular subject. This captures the similarity with an augmented / unit augmented system. <mor prefix s 1st_excl> ¼ ¼ <mor prefix front 1st_excl> <mor prefix final s> <mor prefix s 1st_incl du> ¼ ¼ <mor prefix front 1st_incl du> <mor prefix final s sg>


245

<mor prefix s 1st_incl pl> ¼ ¼ <mor prefix front 1st_excl sg> <mor prefix final s du> <mor prefix s 1st_incl dis> ¼ ¼ <mor prefix front 2nd sg> <mor prefix final s dis> <mor prefix s 2nd> ¼ ¼ <mor prefix front 2nd> <mor prefix final s> <mor prefix s 3rd> ¼ ¼ <mor prefix front 3rd> <mor prefix final s>

Transitive prefixes The prefixes of the transitive paradigm are similar to the intransitive, having a front and final element which mark person and number. <mor prefix a> ¼ ¼ <mor prefix s> <mor prefix a 1st_excl> ¼ ¼ <mor prefix front 1st_excl> <mor prefix final a> <mor prefix a 1st_incl pl> ¼ ¼ <mor prefix front 1st_excl> <mor prefix final a du> <mor prefix a 2nd> ¼ ¼ <mor prefix front 2nd> <mor prefix final a> <mor prefix a 2nd sg> ¼ ¼ <mor prefix front 2nd dis> <mor prefix final a sg> <mor prefix a 3rd> ¼ ¼ <mor prefix front 3rd> <mor prefix final a>

Object marking The special 2 sg object marking involves the 2 sg front element combined with the expected marking of the transitive subject (a) in the final element.

246

Appendix 5

The third person object adds bv_ if it is higher animacy than the subject, and otherwise has no extra marking, if it is lower animacy. <mor prefix o 2nd sg a 3rd> ¼ ¼ <mor prefix front 2nd sg> <mor prefix final a> <mor prefix o 3rd sg> ¼ ¼ <mor prefix> <mor prefix lower> ¼ ¼ <mor prefix> <mor prefix higher> ¼ ¼ bv_ <mor prefix>

Formants Front of Prefix It should be noted that we have tried to break down the prefixes beyond a standard analysis of morphemes. We have analysed prefixes as containing a front element, which marks person and also number. Note the left-hand side paths: <mor prefix front 3rd sg>, <mor prefix front 3rd pl>, <mor prefix front 3rd du>. These have a corresponding reference on the right-hand side to <mor cl> as well as the front element which marks third person for the appropriate number. This allows for the combination of third person subjects with the first person exclusive object clitics. <mor prefix front 1st_excl> ¼ ¼ ng_ <mor prefix front 1st_incl du> ¼ ¼ y_ <mor prefix front 2nd sg> ¼ ¼ dj_ <mor prefix front 2nd dis> ¼ ¼ d_ <mor prefix front 2nd> ¼ ¼ n_ <mor prefix front 3rd sg> ¼ ¼ <mor cl> k_ <mor prefix front 3rd pl> ¼ ¼ <mor cl> b_ <mor prefix front 3rd du> ¼ ¼ <mor prefix front 3rd pl> <mor prefix front 3rd dis> ¼ ¼ <mor prefix front 3rd sg> <mor prefix front 1st_excl pl> ¼ ¼ <mor prefix front 1st_incl du> <mor prefix front 1st_excl du> ¼ ¼ <mor prefix front 1st_excl pl>

Final Part of Prefix The final part of the prefix is the same for singular transitive subjects (a) and intransitive subjects (s). The disyllabic plural and dual formant has a_ for the intransitive subject and _U/I_for the transitive. The form _U/I_ is realized as /i/ after palatals and /u/ otherwise.


247

<mor prefix final a sg> ¼ ¼ <mor prefix final s sg> <mor prefix final s> ¼ ¼ a_ <mor prefix final> <mor prefix final a> ¼ ¼ _U/I_ <mor prefix final> <mor prefix final s dis> ¼ ¼ e_ <mor prefix final a dis> ¼ ¼ <mor prefix final s dis> <mor prefix final du> ¼ ¼ rra_ <mor prefix final pl> ¼ ¼ la_.

Example (51) in Chapter 5 Nan ‘see’ is our example lexical entry. Nan: ¼ ¼ VERB ¼ ¼ see ¼ ¼ nan.

Recall that this analysis generates the appropriate 117 forms of the paradigm. The forms can be seen at http://www.surrey.ac.uk/LIS/SMG/ dalabon/.

Appendix 6: DATR fragment for Russian case study

File: Purpose: Author: Documentation:

rusnoms7.dtr (simplified version) the nominal system of Russian Dunstan Brown, Greville Corbett, Norman Fraser and Andrew Hippisley Chapter 5 (x5.4); data from Brown (1998b)

This appendix contains a simplified version of the DATR fragment which implements the analysis given in case study 3 in Chapter 5: x5.4. It is based on rusnoms7.dtr which is discussed in detail in Brown (1998b), where the full DATR fragment is given. The original fragment was checked on the first 1,500 noun lexemes from Zasorina’s (1977) frequency dictionary, a sample of adjectives, and first, second and third person pronouns. Among other things, we have made the following simplifications: *omitted a number of nodes which deal with morphonological interdependencies stress assignment *omitted paths which determine stress patterns *omitted concatenation marking DATR is distinguished from comments by use of courier font. Declensional class hierachy The declensional class hierarchy represented in Figure 6 in Chapter 5: x5.4 is implemented in DATR below. The node MOR_NOMINAL This contains equations which give much of the plural morphology: pl acc, pl nom, pl gen, pl dat, pl inst, pl loc. It is also at this point in the hierarchy that the forms of the accusative, singular and plural, are determined. The accusative singular evaluates both gender and animacy, whereas the accusative plural needs only to evaluate animacy. The node ACCUSATIVE has been omitted from this simplified 248

DATR fragment for Russian case study

249

appendix. It is used to state that animates use the genitive form for accusatives, and that inanimates use the nominative form for accusatives. The oblique plural forms require a theme vowel, <mor theme_vowel>, which is a for nouns and i for adjectives. As these statements are at the node MOR_NOMINAL they apply to both nouns and adjectives, unless overridden. MOR_NOMINAL: ¼ ¼ MOR_WORD <mor sg acc> ¼ ¼ ACCUSATIVE:< sg "<syn gender>" "<syn animacy>" > <mor pl acc> ¼ ¼ ACCUSATIVE:< pl "<syn animacy>" > <mor pl nom> ¼ ¼ "<stem pl nom>" i "<stress pl nom>" <mor pl gen> ¼ ¼ "<mor pl loc>" <mor pl dat> ¼ ¼ "<stem pl>" "<mor theme_vowel>" "<stress pl>" m <mor pl inst> ¼ ¼ "<stem pl>" "<mor theme_vowel>" "<stress pl>" m’i <mor pl loc> ¼ ¼ "<stem pl>" "<mor theme_vowel>" "<stress pl>" x.

The node MOR_NOUN (example (65) in Chapter 5: x5.4) There is a noun default for singular dative, which is the referral to singular locative. There is a noun default for the singular locative. The other equations do not appear in example (65). One requires the evaluation of the noun’s stem type to determine its genitive plural, and the other states that the theme vowel for nouns is -a-. MOR_NOUN: ¼ ¼ MOR_NOMINAL <mor sg dat> ¼ ¼ "<mor sg loc>" <mor sg loc> ¼ ¼ "<stem sg>" e "<stress sg>" <mor pl gen> ¼ ¼ MGP: <mor theme_vowel> ¼ ¼ a.

The node N_O This node generalizes over classes I and IV, which are exemplified by zakon and vino in Chapter 5: x5.4. N_O states the realization of singular

250

Appendix 6

genitive, singular dative and the singular instrumental for both types of noun. N_O: ¼ ¼ MOR_NOUN <mor sg gen> ¼ ¼ "<stem sg gen>" a "<stress sg>" <mor sg dat> ¼ ¼ "<stem sg>" u "<stress sg>" <mor sg inst> ¼ ¼ "<stem sg>" om "<stress sg>".

The node N_I N_I inherits from N_O. The formal gender assigned, if there is no assignment by semantics, is masculine. In the full DATR representation the nominative plural form is partly determined by the stress pattern. The node NOM_PL is not given in this simplified appendix. The genitive plural of nouns with hard stems is the stem plus ov. N_I: ¼ ¼ N_O <mor formal gender> ¼ ¼ m <mor pl nom> ¼ ¼ NOM_PL: <mor hard pl gen> ¼ ¼ "<stem pl>" ov "<stress pl>".

The node N_IV N_IV inherits from the node N_O. The formal gender of this class is neuter. The singular nominative is the stem plus o. The plural nominative is the stem plus a. The plural genitive involves evaluation of the final element of the stem. N_IV: ¼ ¼ N_O <mor formal gender> ¼ ¼ n <mor sg nom> ¼ ¼ "<stem sg nom>" o "<stress sg>" <mor pl nom> ¼ ¼ "<stem pl>" a "<stress pl nom>" <mor pl gen> ¼ ¼ GEN_PL:< "<stem pl final>" >.

The node N_II N_II inherits from MOR_NOUN. The formal gender of this class is feminine. The singular nominative is the stem plus a. N_II overrides the assignment of accusative singular which would otherwise be inherited from MOR_NOMINAL. The accusative singular is stem plus u. The genitive singular is stem plus i. The instrumental singular is stem plus Vj (u), where V is a vowel o and (u) is the archaic alternative.


251

The plural genitive involves evaluation of the final element of the stem and stress information. The node STEMSTRESS is not given in this simplified appendix. N_II: ¼ ¼ MOR_NOUN <mor formal gender> ¼ ¼ f <mor sg nom> ¼ ¼ "<stem sg nom>" a "<stress sg>" <mor sg acc> ¼ ¼ "<stem sg>" u "<stress sg acc>" <mor sg gen> ¼ ¼ "<stem sg>" i "<stress sg>" <mor sg inst> ¼ ¼ "<stem sg inst>" "<mor vowel sg>" "<stress sg>" j ‘(‘ u ’)’ <mor pl gen> ¼ ¼ STEMSTRESS: .

The node N_III (example (62) in Chapter 5: x5.4) This node inherits from MOR_NOUN. All members of this class are morphologically soft (not given in (62)). The singular genitive is inherited orthogonally from N_II (not given in (62)). The singular locative refers to the singular genitive. N_III: ¼ ¼ MOR_NOUN <mor stem hardness> ¼ ¼ soft <mor sg gen> ¼ ¼ N_II <mor sg loc> ¼ ¼ "<mor sg gen>" <mor sg inst> ¼ ¼ "<stem sg inst>" ju <mor formal gender> ¼ ¼ f.

Adjective inflection The node MOR_ADJ: this inherits from MOR_NOMINAL, including the the generalizations about animacy and the accusative, as well as most plural forms. MOR_ADJ specifies most of the singular forms of adjectives. It also specifies the theme vowel which is used in the oblique plural forms given at MOR_NOMINAL. MOR_ADJ: ¼ ¼ MOR_NOMINAL <mor sg gen> ¼ ¼

252

Appendix 6 "<stem>" "<mor vowel sg>" "<stress sg>" vo

<mor sg gen f> ¼ ¼ "<mor sg loc f>" <mor sg dat> ¼ ¼ "<stem>" "<mor vowel sg>" "<stress sg>" mu <mor sg dat f> ¼ ¼ MOR_NOUN <mor sg inst> ¼ ¼ "<stem>" i "<stress sg>" m <mor sg inst f> ¼ ¼ N_II <mor sg loc> ¼ ¼ "<stem>" "<mor vowel sg loc>" "<stress sg>" m <mor sg loc f> ¼ ¼ "<stem>" "<mor vowel sg>" "<stress sg>" j <mor theme_vowel> ¼ ¼ i.

The node A_I: this inherits from the node MOR_ADJ. It accounts for the majority of attributive adjectives. The singular nominative masculine combines the singular stem with ij or oj, depending on the stress pattern of the adjective. (ADJ_VOWEL is not given in this appendix, but it determines the vowel on the basis of stress.) The singular nominative feminine refers to N_II and combines this with the augment -ja. The singular neuter inherits in part from the node N_IV and combines the value with the augment -je. The plural nominative combines the plural theme vowel with je. The singular accusative feminine combines the value for singular accusative at N_II with the augment ju. A_I: ¼ ¼ MOR_ADJ <mor sg nom m> ¼ ¼ "<stem sg>" ADJ_VOWEL: "<stress sg>" j <mor sg nom f> ¼ ¼ N_II ja <mor sg nom n> ¼ ¼ N_IV je <mor sg acc f> ¼ ¼ N_II ju <mor pl nom> ¼ ¼ "<stem pl>" "<mor theme_vowel>" "<stress pl>" je.

The nodes A_II and A_III: these inherit from MOR_ADJ. They define possessive adjectival classes which combine some noun morphology with the default adjectival morphology. This is why they consist almost entirely of inheritance relations with the noun classes. The class A_III is rare.


253

A_II: ¼ ¼ MOR_ADJ <mor sg nom f> ¼ ¼ N_II <mor sg nom n> ¼ ¼ N_IV <mor sg acc f> ¼ ¼ N_II. A_III: ¼ ¼MOR_ADJ <mor sg nom f> ¼ ¼ N_II <mor sg nom n> ¼ ¼ N_IV <mor sg acc f> ¼ ¼ N_II <mor sg gen n> ¼ ¼ "<mor sg gen m>" <mor sg dat n> ¼ ¼ "<mor sg dat m>" <mor sg gen m> ¼ ¼ N_I <mor sg dat m> ¼ ¼ N_I.

6.3

A note on Russian phonology

The Russian examples in Chapter 5: x5.4 and elsewhere in the book have been given in transliteration. In contrast, the DATR fragment which this simplified appendix is based on generates forms in phonological transcription. It is therefore important to outline the phonological correspondences assumed for the DATR fragment. This note is based on Corbett and Fraser (1993: 114). The phoneme transcribed as /i/ is retracted to its back allphone [ç] after nonback hard (unpalatalized) consonants. Thus the nominative plural /zakoni/ is realized with [ç], but /kost 0 i/ is realised with [i], since [t 0 ] is soft. This difference is reflected in the writing system, with the inflection of /zakoni/ being written z, and the inflection for /kost 0 i/ written h. Underlyingly it is assumed that both items actually share an identical inflection. All consonants which can be palatalized are automatically palatalized before /e/. Thus the locative singular of /zakon/, namely /zakone/, will be realized with a palatalized [n 0 ]. If the consonant is already palatalized as in genitive plural /kost 0 -ej/, it simply remains palatalized. Some consonants are always hard (/sˇ , zˇ, c/), and remain so before /e/. On the other hand, /cˇ 0 / and /sˇ cˇ 0 /are always soft (palatalized), and naturally remain so before /e/. The velars /k, g, x/are palatalized before /i/, so that the genitive form /knigi/ ‘book’, will be realized with palatalized [g 0 ] (which then demands the front allphone of /i/, namely [i]).

References

This section lists works cited in the preceding text. An annotated bibliography on syncretism (including some of the works below, but not restricted to them) may be found in Baerman (2002a). Abbot, Miriam. 1991. Macushi. In Desmond Derbyshire and Geoffey K. Pullum (eds.), Handbook of Amazonian languages (vol. III). Berlin: Mouton de Gruyter. 23–160. Ackema, Peter and Ad Neeleman. 2003. Context-sensitive spell-out. Natural Language and Linguistic Theory 21. 681–735. Aikhenvald, Alexandra Y. and Robert M. W. Dixon. 1998. Dependencies between grammatical systems. Language 74. 56–80. Anttila, Raimo. 1972. An introduction to historical and comparative linguistics. New York: Macmillan. Arensen, Jon. 1971. A Murle grammar. Linguistic Monograph Papers No. 7. Khartoum: Sudan Research Unit, Faculty of Arts, University of Khartoum. Arkadiev, Pe¨tr M. 2003. Tipologija padezˇnogo sinkretizma. Ms., Russian State University for the Humanities. Armbruster, Charles Herbert. 1960. Dongolese Nubian: a grammar. Cambridge: University Press. Aronoff, Mark. 1994. Morphology by itself. Cambridge: MIT Press. Aronson, Howard I. 1991. Modern Georgian. In Alice C. Harris (ed.), The indigenous languages of the Caucasus (vol. I: Kartvelian languages). Delmar: Caravan Books. 219–312. Ashton, E. O. 1947. Swahili grammar (second edition). London: Longmans, Green and Co. Austin, Peter. 1981. A grammar of Diyari, South Australia. Cambridge: Cambridge University Press. 1986. Structural change in language obsolescence: some Eastern Australian examples. Australian Journal of Linguistics 6. 201–30. Baerman, Matthew. 2002a. Syncretism: an annotated bibliography. Available on-line at http://www.surrey.ac.uk/LIS/MB/Bibliography.htm. 2002b. The Surrey person syncretism database. Available on-line at: http:// www.smg.surrey.ac.uk/personsyncretism/index.aspx. 2004. Directionality and (un)natural classes in syncretism. Language 80. 807–827. 254

References

255

2005. Typology and the formal modelling of syncretism. In Geert Booij and Jaap van Marle (eds.), Yearbook of morphology 2004. Dordrecht: Kluwer. 41–72 . Baerman, Matthew and Dunstan Brown. 2005a. Case syncretism. In Martin Haspelmath, Matthew Dryer, David Gil and Bernard Comrie (eds.), World atlas of language structures. Oxford: Oxford University Press. 2005b. Verbal person/number syncretism. In Martin Haspelmath, Matthew Dryer, David Gil and Bernard Comrie (eds.), World atlas of language structures. Oxford: Oxford University Press. Baerman, Matthew, Dunstan Brown and Greville G. Corbett. 2002a. The Surrey syncretisms database. Available online at: http://www.smg.surrey.ac.uk/ syncretism/index.aspx. 2002b. Case syncretism in and out of Indo-European. In Mary Andronis, Christopher Ball, Heidi Elston and Sylvain Neuvel (eds.), CLS 37: The main session. papers from the 37th meeting of the Chicago Linguistic Society (vol. I). Chicago: Chicago Linguistic Society. 15–28. Bammesberger, Alfred. 1982. A handbook of Irish. Heidelberg: Carl Winter. Ba´tori, Istvań. 1990. Die Markierung des Objekts am Verb im Mordwinischen: morphologische Unterbestimmtheit und Homomorphie. Nyelvtudomańyi Ko¨zlemeńyek 91. 15–23. Bavin, Edith L. and Tim Shopen. 1987. Innovations and neutralizations in the Warlpiri pronominal system. Journal of Linguistics 23. 149–75. Bechhaus-Gerst, Marianne. 1996. Sprachwandel durch Sprachkontakt am Beispiel des Nubischen im Niltal. Ko¨ln: Ru¨diger Ko¨ppe. Beeler, Madison S. 1976. Barbarenõ Chumash grammar: a farrago. In Margaret Langdon and Shirley Silver (eds.), Hokan studies. The Hague: Mouton. 251–70. Be´jar, Susana and Daniel Currie Hall. 1999. Marking markedness: the underlying order of diagonal syncretisms. Paper presented at the Eastern States Conference on Linguistics, University of Connecticut. Bender, Lionel M. 1996. Kunama. Munich: Lincom. Bentley, Delia and Tho´rhallur Eytho´rsson. 2001. Alternation according to person in Italo-Romance. In Laurel J. Brinton (ed.), Historical linguistics 1999. Amsterdam/Philadelphia: John Benjamins. 63–74. Berger, Hermann. 1998. Die Burushaski-Sprache von Hunza und Nager (vol. I: Grammatik). Wiesbaden: Harrassowitz. Berman, Ruth. 1997. Hebrew. In Robert Hetzron (ed.), The Semitic languages. London: Routledge. 312–34. Bierwisch, Manfred. 1967. Syntactic features in morphology: general problems of so-called pronominal inflection in German. In To honor Roman Jakobson: essays on the occasion of his seventieth birthday. The Hague: Mouton. 239–70. Blake, Barry J. 1994. Case. Cambridge: Cambridge University Press. Bleiching, Doris, Guido Drexel and Dafydd Gibbon. 1996. Ein Synkretismusmodell fu¨r die deutsche Morphologie. In Dafydd Gibbon (ed.), Natural language processing and speech technology. Results of the third KONVENS Conference, Bielefeld. Berlin: Mouton de Gruyter. 237–48. Blevins, James. 2003 Stems and paradigms. Language 79. 737–67.

256

References

Bobaljik, Jonathan. 2002. Syncretism without paradigms: remarks on Williams 1981, 1994. In Geert Booij and Jaap van Marle (eds.), Yearbook of morphology 2001. Dordrecht: Kluwer. 53–86. Boeder, Winfried. 1976. Morphologische Kategorien. In K. Braunmu¨ller and W. Ku¨rschner (eds.), Grammatik. Akten des 10. Linguistischen Kolloquiums, Tu¨bingen 1975 (vol. II). Tu¨bingen: Niemeyer. 117–26. Booij, Geert. 1996. Inherent versus contextual inflection and the split morphology hypothesis. In Geert Booij and Jaap van Marle (eds.), Yearbook of morphology 1995. Dordrecht: Foris. 1–16. Bowe, Heather. 1990. Categories, constituents, and constituent order in Pitjantjatjara. London: Routledge. Boyeldieu, Pascal. 1982. Deux e´tudes laal. Berlin: D. Reimer. Breedveld, J. O. 1995. Form and meaning in Fulfulde: a morphophonological study of Maasinankoore. Leiden: CNWS. Breen, J. Gavan. 1976. Wagaya. In Robert M. W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies. 590–4. Bright, William. 1957. The Karok language. Berkeley: University of California Press. Brockelmann, Carl. 1908–13. Grundriss der vergleichenden Grammatik der semitischen Sprachen. Berlin: Reuther and Reichard. Bromley, H. Myron. 1981. A grammar of Lower Grand Valley Dani. (Pacific Linguistics C63). Canberra: Australian National University. Brown, Dunstan. 1998a. Defining ‘subgender’: virile and devirilised nouns in Polish. Lingua 104. 187–233. 1998b. From the general to the exceptional: a Network Morphology account of Russian nominal inflection. PhD thesis, University of Surrey. 2001. Constructing a typological database for inflectional morphology: the SMG database for syncretism. In Steven Bird, Peter Buneman and Mark Liberman (eds.), Proceedings of the IRCS Workshop on Linguistic Databases. Philadelphia: Institute for Research in Cognitive Science, University of Pennsylvania. 56–64 (paper available at: http:// www.ldc.upenn.edu/annotation/database/proceedings.html). Brown, Dunstan, Greville Corbett, Norman Fraser, Andrew Hippisley and Alan Timberlake. 1996. Russian noun stress and network morphology. Linguistics 34. 53–107. Browne, Gerald M. 2002. A grammar of Old Nubian. Munich: Lincom. Buck, Carl D. 1933. Comparative grammar of Greek and Latin. Chicago: University of Chicago Press. Burrow, Thomas and Sudhibhushan Bhattacharya. 1970. The Pengo language. Oxford: Clarendon Press. Butt, John and Carmen Benjamin. 2000. A new reference grammar of modern Spanish. London: Arnold. Bybee, Joan. 1985. Morphology. Amsterdam/Philadelphia: John Benjamins. Cahill, Lynne and Gerald Gazdar. 1997. The inflectional phonology of German adjectives, determiners, and pronouns. Linguistics 35. 211–45.

References

257

1999. The German noun inflection. Journal of Linguistics 35. 1–42. Cairns, Charles E. 1986. Word structure, markedness, and applied linguistics. In Fred R. Eckman, Edith A. Moravcsik and Jessica R. Wirth (eds.), Markedness. Proceedings of the twelfth annual linguistics symposium of the University of Wisconsin-Milwaukee, March 11–12, 1983. New York: Plenum Press. 13–38. Calabrese, Andrea. 1998. Some remarks on the Latin case system and its development in Romance. In Jose´ Lema and Esthela Trevinõ (eds.), Theoretical analyses on Romance languages. Amsterdam/Philadelphia: John Benjamins. 71–126. Cardona, George. 1964. A Gujarati reference grammar. Philadelphia: University of Pennsylvania Press. Cardona, George and Babu K. Suthar. 2003. Gujarati. In George Cardona and Dhanesh Jain (eds.), The Indo-Aryan languages. London: Routledge. 658–97. Carmack, Stanford. 1997. Blocking in Georgian verb morphology. Language 72. 314–38. Carstairs, Andrew. 1984. Outlines of a constraint on syncretism. Folia Linguistica 18. 73–85. 1987. Allomorphy in inflection. London: Croom Helm. Carstairs-McCarthy, Andrew. 1994. Inflection classes, gender, and the principle of contrast. Language 70. 737–88. 1998a. How lexical semantics constrains inflectional allomorphy. In Geert Booij and Jaap van Marle (eds.), Yearbook of morphology 1997. Dordrecht: Kluwer. 1–24. 1998b. Comments on the paper by Noyer. In Steven G. Lapointe, Diane K. Brentari, and Patrick M. Farrell (eds.), Morphology and its relation to phonology and syntax. Stanford: CSLI Publications. 286–301. Carter, Hazel and Joaõ Makoondekwa. 1979. Kongo course: maloo`ngi makikoóngo (dialect of Zoombo, Angola). London: School of Oriental and African Studies. Chafe, Wallace. 1997. Sketch of Seneca, an Iroquoian language. In Ives Goddard (ed.) Handbook of North American Indians (vol. XVII: Languages). Washington: Smithsonian Institute. 551–79. Chapman, Shirley and Desmond C. Derbyshire. 1990. Paumarı` . In Desmond C. Derbyshire and Geoffrey K. Pullum (eds.), Handbook of Amazonian languages (vol. III). Berlin: Mouton de Gruyter. 161–352. Charney, Jean Ormsbee. 1993. A grammar of Comanche. Lincoln: University of Nebraska Press. Childs, George Tucker. 1995. A grammar of Kisi: a southern Atlantic language. Berlin: Mouton de Gruyter. Chvany, Catherine V. 1986. Jakobson’s fourth and fifth dimensions: on reconciling the cube model of case meanings with the two-dimensional matrices for case forms. In Richard D. Brecht and James S. Levine (eds.), Case in Slavic. Columbus: Slavica. 107–29. Clairis, Christos. 1985. El Qawasqar: lingu¨ı´stica fueguina, teorıá y descripcioń. (Estudios filolo´gicos, Anejo 12.) Valdivia, Chile: Facultad de Filosofı´ a y Humanidades, Universidad Austral de Chile.

258

References

Coleman, Robert. 1976. Patterns of Syncretism in Latin. In Anna M. Davies and Wolfgang Meid (eds.), Studies in Greek, Italic and Indo-European linguistics offered to Leonard R. Palmer. Innsbruck: Institut fu¨r Sprachwissenschaft der Universita¨t Innsbruck. 47–56. 1991. The assessment of paradigm stability: some Indo-European case studies. In Frans Plank (ed.), Paradigms: the economy of inflection. Berlin: Mouton de Gruyter. 197–211. Comrie, Bernard. 1975. Polite plurals and predicate agreement. Language 51. 406–18. 1980. Inverse verb forms in Siberia: evidence from Chukchee, Koryak, and Kamchadal. Folia Linguistica Historica 1/1. 61–74. 1991. Form and function in identifying cases. In Frans Plank (ed.), Paradigms: the economy of inflection. Berlin: Mouton de Gruyter. 41–55. Corbett, Greville G. 1983. Hierarchies, targets and controllers: agreement patterns in Slavic. London: Croom Helm. 1991. Gender. Cambridge: Cambridge University Press. 2000. Number. Cambridge: Cambridge University Press. 2005. The number of genders. In Martin Haspelmath, Matthew Dryer, David Gil and Bernard Comrie (eds.), World atlas of language structures. Oxford: Oxford University Press. Corbett, Greville G., Matthew Baerman and Dunstan Brown. 2002. Domains of syncretism: a demonstration of the autonomy of morphology. In Mary Andronis, Christopher Ball, Heidi Elston and Sylvain Neuvel (eds.), CLS 37: The Panels. Papers from the 37th meeting of the Chicago Linguistic Society (vol. I). Chicago: Chicago Linguistic Society. 385–98. Corbett, Greville G. and Norman M. Fraser. 1993. Network morphology: a DATR account of Russian nominal inflection. Journal of Linguistics 29. 113–42. 1997. Komp0 juternaja lingvistika i tipologija. Vestnik Moskovskogo Universiteta (ser. 9. Filologija, no. 2). 122–40. Cowan, H. K. J. 1965. Grammar of the Sentani language. ’s-Gravenhage: Martinus Nijhoff. Crowley, Terry. 1998. An Erromangan (Sye) grammar. Honolulu: University of Hawaii Press. 1999. Ura: a disappearing language of southern Vanuatu. (Pacific Linguistics C156). Canberra: Australian National University. Curnow, Timothy J. 1997. A grammar of Awa Pit (Cuaiquer): An indigenous language of South-western Colombia. PhD thesis, Australian National University. Curtius, Georg. 1863. Erlaüterungen zu meiner Griechischen Schulgrammatik. Prague: Tempsky. Cysouw, Michael. 2003. The paradigmatic structure of person marking. Oxford: Oxford University Press. Cysouw, Michael. Forthcoming. Syncretisms involving clusivity. In Elena Filimonova (ed.), Clusivity. Amsterdam/Philadelphia: John Benjamins. Dalrymple, Mary and Ronald M. Kaplan. 2000. Feature indeterminancy and feature resolution. Language 76. 759–98.

References

259

Davies, John. 1981. Kobon (Lingua Descriptive Studies 3). Amsterdam: NorthHolland. de Angulo, Jaime. 1932. The Chichimeco language (Central Mexico). International Journal of American Linguistics 7. 152–94. Dedrick, John M. and Eugene H. Casad. 1999. Sonora Yaqui language structures. Tucson: University of Arizona Press. Dench, Alan C. 1995. Martuthunira: a language of the Pilbara region of western Australia. Canberra: Australian National University (Pacific Linguistics C125). 2001. Descent and diffusion: the complexity of the Pilbara situation. In Alexandra Y. Aikhenvald and Robert M. W. Dixon (eds.), Areal diffusion and genetic inheritence: problems in comparative linguistics. Oxford: Oxford University Press. 105–33. Derbyshire, Desmond. 1979. Hixkaryana (Lingua Descriptive Studies 1). Amsterdam: North-Holland. Deza Galindo, Juan Francisco. 1992. Grama´tica de la lengua aymara. Lima: Artex Editores. Dickens, Patrick. 1992. A Ju/0 hoan grammar. Ms., Nyae Development Foundation of Namibia. Dimmendaal, Gerrit J. 1983. The Turkana language. Dordrecht: Foris. Dixon, Robert M. W. 1977. A grammar of Yidi . Cambridge: Cambridge University Press. 1988. A grammar of Boumaa Fijian. Chicago: University of Chicago Press. 1994. Ergativity. Cambridge: Cambridge University Press. Donaldson, Tamsin. 1980. Ngiyambaa: the language of the Wangaaybuwan. Cambridge: Cambridge University Press. Drabbe, Peter. 1952. Spraakkunst van het Ekagi. ’s-Gravenhage: Martinus Nijhoff. 1955. Spraakkunst van het Marind, zuidkust Nederlands Nieuw-Guinea. Mo¨dling: Missiehuis St. Gabrie¨l (Reprinted 1966 by Johnson Reprint Corp., New York). van Driem, George. 1987. A grammar of Limbu. Berlin: Mouton de Gruyter. Dutton, Tom. 2003. A dictionary of Koiari, Papua New Guinea, with grammar notes. Canberra: Australian National University. Elliot, Eric. 1999. Dictionary of Rincoń Luisenõ. PhD thesis, University of California at San Diego. Endzelıns [Endzelin], J anis. 1922. Lettische Grammatik. Riga: Gulbis. Engel, Ralph, Mary Allhiser de Engel and Jose´ Mateo Alvarez. 1987. Diccionario zoque de Francisco Leon. Hidalgo: Instituto Lingu¨ı´ stico de Verano. Evans, Nicholas. 1995. A grammar of Kayardild: with historical-comparative notes on Tangkic. Berlin: Mouton de Gruyter. 2003. Bininj Gun-Wok: a pan-dialectal grammar of Mayali, Kunwinjku and Kune (Pacific Linguistics 541). Canberra: Australian National University. Evans, Nicholas, Dunstan Brown and Greville G. Corbett. 2001. Dalabon pronominal prefixes and the typology of syncretism: a Network Morphology analysis. In Geert Booij and Jaap van Marle (eds.), Yearbook of morphology 2000. Dordrecht: Kluwer. 187–231. 2002. The semantics of gender in Mayali: partially parallel systems and formal implementation. Language 78. 111–55.

¸

260

References

Evans, Roger and Gerald Gazdar. 1996. DATR: a language for lexical knowledge representation. Computational Linguistics 22. 167–216. Ezard, Bryan. 1997. A grammar of Tawala: an Austronesian language of the Milne Bay Area, Papua New Guinea (Pacific Linguistics C137). Canberra: Australian National University. Feldstein, Ronald. 2003. On the structure of syncretism in Romanian conjugation. Ms., Indiana University. Fennell, Trevor G. 1975. Is there an instrumental case in Latvian? Journal of Baltic Studies 6. 41–8. Feoktistov, Aleksandr p. 1966. Erzjanskij jazyk. Jazyki narodov SSSR (vol. III: Finno-ugorskie jazyki). Moscow: Nauka. 177–98. Finkel, Raphael, Lei Shen, Gregory Stump and Suresh Thesayi, 2002. KATR: a setbased extension of DATR (Technical report 346–02). University of Kentucky Department of Computer Science, Lexington, KY. Fischer, Wolfdietrich. 1997. Classical Arabic. In Robert Hetzron (ed.), The Semitic languages. London: Routledge. 187–219. Foley, William A. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge University Press. 1991. The Yimas language of New Guinea. Stanford: Stanford University Press. Fortescue, Michael. 1984. West Greenlandic (Croom Helm Descriptive Grammars). Dover, New Hampshire: Croom Helm. Frachtenberg, Leo J. 1922. Siuslaw. In Franz Boas (ed.), Handbook of American Indian languages (vol. II) (Smithsonian Institution Bureau of American Ethnology Bulletin). Washington: Government Printing Office. 431–629. Fradkin, Robert A. 1991. Marking, markedness, and person-gender-number patterning in the Arabic tenses and moods. Folia Linguistica 25. 609–64. Frank, Paul. 1990. Ika syntax (Studies in the Languages of Colombia 1). Dallas: Summer Institute of Linguistics. Frank, Wright Jay. 1999. Nuer noun morphology. MA thesis, State University of New York, Buffalo. Franklin, Karl James. 1971. A grammar of Kewa, New Guinea (Pacific Linguistics C16). Canberra: Australian National University. Fraser, Norman M. and Greville G. Corbett. 1995. Gender, animacy and declensional class assignment: a unified account for Russian. In Geert Booij and Jaap van Marle (eds.), Yearbook of morphology 1994. Dordrecht: Kluwer. 123–50. 1997. Defaults in Arapesh. Lingua 103. 25–57. Fromm, Hans. 1982. Finnische Grammatik. Heidelberg: Carl Winter. Glasgow, Kathleen. 1984. Burarra word classes. In Kathleen Glasgow et al. (eds.), Papers in Australian linguistics 16 (Pacific Linguistics A68). Canberra: Australian National University. 1–54. Goddard, Cliff. 1982. Case systems and case marking in Australian languages: a new interpretation. Australian Journal of linguistics 2. 167–96. Gordon, Lynn. 1986. Maricopa morphology and syntax (University of California Publications in Linguistics 108). Berkeley: University of California Press. Graudin¸ a, A. 1969. Liter ar as valodas un izloksˇ n¸ u mijiedarbıba. Latviesˇu valodas kult uras jauta¯jumi. 16–20.

References

261

Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Joseph H. Greenberg (ed.), Universals of language. Cambridge, Massachusetts: MIT Press. 73–113. Gregores, Emma, and Jorge A. Sua´rez. 1967. A description of colloquial Guarani. The Hague: Mouton. Grimm, Jacob and Willhelm Grimm. 2004. Deutsches Wo¨rterbuch (Digital edition). Frankfurt am Main: Zweitausendeins. Grjunberg, A. L. and D. I. Edel0 man. 1987. Afganskij jazyk. In Osnovy iranskogo jazykoznanija (vol. IV: Novoiranskie jazyki: vostocˇnaja gruppa). Moscow: Nauka. 6–154. Grosse, Siegfried. 2000. Morphologie des Mittelhochdeutschen. In Werner Besch, Anne Betten, Oskar Reichmann and Stefan Sonderegger (eds.), Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung (vol. 1.2). Berlin: Mouton de Gruyter. 1332–40. Gruzdeva, Elena Ju. 1997. Nivxskij jazyk. In Jazyki mira: paleoaziatskie jazyki. Moscow: Indrik. 139–54. Gu¨ldemann, Tom and Rainer Vossen. 2000. Khoisan. In Bernd Heine and Derek Nurse (eds.), African languages. Cambridge: Cambridge University Press. 99–122. Gvozdanovic´, Jadranka. 1991. Syncretism and the paradigmatic patterning of grammatical meaning. In Frans Plank (ed.), Paradigms: the economy of inflection. Berlin: Mouton de Gruyter. 133–60. Hagman, Roy S. 1977. Nama Hottentot grammar. Bloomington: Indiana University. Halle, Morris. 1997. Distributed morphology: impoverishment and fission. MIT Working Papers in Linguistics 30. 425–49. Hamel, Patricia J. 1994. A grammar and lexicon of Loniu, Papua New Guinea (Pacific Linguistics C103). Canberra: Australian National University. Hansson, Gunnar O. 1996. Productive syncretism and morphological theory: a case study of North Sami inflectional morphology. Ms., University of California, Berkeley. Harbour, Daniel. 2003. Elements of number theory. PhD thesis, Massachusetts Institute of Technology. Harley, Heidi and Elizabeth Ritter. 2002. Person and number in pronouns: a feature-geometric analysis. Language 78. 482–526. Harvey, Mark. 2002. A grammar of Gaagudju. Berlin: Mouton de Gruyter. Haspelmath, Martin. 1993. A grammar of Lezgian. Berlin: Mouton de Gruyter. Haspelmath, Martin, Matthew Dryer, David Gil and Bernard Comrie (eds.), 2005. World atlas of language structures. Oxford: Oxford University Press. Haugen, Einar. 1982. Scandinavian language structures: a comparative historical survey. Minneapolis: University of Minnesota Press. Haviland, John. 1979. Guugu Yimidhirr. In Robert M. W. Dixon and Barry J. Blake (eds.), Handbook of Australian languages. Canberra: Australian National University Press. 27–180. Hayward, Richard. 1984. The Arbore language. Hamburg: H. Buske. Healy, Alan, Ambrose Isoroembo and Martin Chittleborough. 1969. Preliminary notes on Orokaiva grammar. In Papers in New Guinea Linguistics 9 (Pacific Linguistics A18). 33–64.

262

References

Heath, Jeffrey. 1984. Functional grammar of Nunggubuyu. Atlantic Highlands NJ: Humanities Press (and Canberra: Australian Institute of Aboriginal Studies). 1991. Pragmatic disguise in pronominal-affix paradigms. In Frans Plank (ed.), Paradigms: the economy of inflection. Berlin: Mouton de Gruyter. 75–90. 1998. Pragmatic skewing in 1 « 2 pronominal combinations in Native American languages. International Journal of American Linguistics 64. 83–104. Heine, Bernd. 1981. The non-Bantu languages of Kenya. Berlin: Reimer. 1982. African noun class systems. In Hansjakob Seiler and Christian Lehman (eds.), Apprehension: das sprachliche Erfassen von Gegensta¨nden (vol. I). Tu¨bingen: Narr. 189–216. Hekking, Ewald. 1995. El Otomı´ de Santiago Mexquititlań: desplazamiento lingu¨ı´stico, pre´stamos y cambios gramaticales. Amsterdam: Institute for Functional Research into Language and Language Use. Helimski, Eugene. 1998. Selkup. In Daniel Abondolo (ed.), The Uralic languages. London: Routledge. 548–79. Helmbrecht, Johannes. 1999. The typology of 1st person marking and its cognitive background. In Masako K. Hiraga, Chris Sinha and Sherman Wilcox (eds.), Cultural, psychological and typological issues in cognitive linguistics. Amsterdam/Philadelphia: John Benjamins. 285–97. 2003. Ikonizita¨t in Personalpronomina (Arbeitspapiere des Seminars fu¨r Sprachwissenschaft der Universita¨t Erfurt nr. 5). Erfurt: Seminar fu¨r Sprachwissenschaft der Universita¨t Erfurt. Henderson, James. 1995. Phonology and grammar of Yele, Papua New Guinea (Pacific Linguistics B112). Canberra: Australian National University. Herbermann, Charles G., Edward A. Pace, Conde´ B. Pallen, Thomas J. Shahan and John J. Wynne. 1907–18. The Catholic encyclopedia. New York: Robert Appleton. Hetzron, Robert. 1967. Agaw numerals and incongruence in Semitic. Journal of Semitic Studies 12. 169–97. Hewitt, Brian G. 1995. Georgian: a structural reference grammar. Amsterdam/ Philadelphia: John Benjamins. Hewson, John. 1989. Motivated syncretism. Journal of the Atlantic Provinces Linguistic Association/Revue de l‘Association de Linguistique des Provinces Atlantique 11. 39–56. Hippisley, Andrew. 1997. Declarative derivation: a Network Morphology account of Russian word formation with reference to nouns denoting ‘person’. PhD thesis, University of Surrey. 2001. Word formation rules in a default inheritance framework. In Jaap van Marle and Geert Booij (eds.), Yearbook of morphology 1999. Dordrecht: Kluwer. 221–61. Hjelmslev, Louis. 1935–7. La cate´gorie des cas: e´tude de grammaire geńe´rale (Acta Jutlandica VII/1). Aarhus: Universitetsforlaget. Hoff, B. J. 1968. The Carib language. The Hague: M. Nyhoff. Holton, David, Peter Mackridge and Irene Philippaki-Warburton. 1997. Greek: a comprehensive grammar of the modern language. London: Routledge.

References

263

Horton, A. E. 1949. A grammar of Luvale. Johannesburg: Witwatersrand University Press. Hualde, Jose´ Ignacio and Jon Ortiz de Urbina (eds.). 2003. A grammar of Basque. Berlin: Mouton de Gruyter. Hudson, Richard A. 1974. A structural sketch of Beja. In D. W. Arnott (ed.), African language studies. London: School of Oriental and African Studies. 111–42. Huntley, David. 1980. The evolution of genitive-accusative animate and personal nouns in Slavic dialects. In Jacek Fisiak (ed.), Historical morphology. The Hague: Mouton. 189–212. Hutchison, John P. 1981. The Kanuri language: a reference grammar. Madison: African Studies Program, University of Wisconsin. Innes, Gordon. 1966. An introduction to Grebo. London: School of Oriental and African Studies. Ivanov, Vjacˇeslav V. 2001. Xettskij jazyk (second edition). Moscow: URSS. Jackendoff, Ray S. 1985. Multiple subcategorization and the u-criterion: the case of climb. Natural Language and Linguistic Theory 3. 271–95. Jacobson, Steven A. 1995. A practical grammar of the Central Alaskan Yup’ik Eskimo language. Fairbanks: Alaska: Alaska Native Language Center, University of Alaska. Jakobi, Angelika. 1990. A Fur grammar: phonology, morphophonology, and morphology. Hamburg: Helmut Buske. Jakobson, Roman O. 1936 [1971]. Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutung der russischen Kasus. Reprinted in Selected writings (vol. II: Word and language). The Hague: Mouton. 23–71 (originally in Travaux du Cercle Linguistique de Prague VI. 240–99). 1958. Morfologicˇeskie nabljudenija nad slavjanskim skloneniem. In American contributions to the Fourth International Congress of Slavists, Moscow. ’s-Gravenhage: Mouton. 127–56. Johnston, Jason. 1997. Systematic homonymy and the structure of morphological categories: some lessons from paradigm geometry. PhD thesis, University of Sydney. Kachru. Braj B. 1969. A reference grammar of Kashmiri. Urbana: Department of Linguistics, Univerity of Illinois. Karin¸ sˇ , A. Krisˇ j anis. 1994. Functionalism and linguistic change in Latvian verb morphology. Linguistica Baltica 3. 109–20. Karlson, Fred. 1999. Finnish: an essential grammar. London/New York: Routledge. Kathol, Andreas. 2002. Nominal head-marking constructions: two case studies from Luisenõ. In Frank Van Eynde, Lars Hellan and Dorothee Beermann (eds.), Proceedings of the eighth international conference on Head-Driven Phrase Structure Grammar. Stanford: CSLI Publications. 189–201. Kennedy, Benjamin Hall. 1955. The revised Latin primer (edited and further revised by Sir James Mountford). London: Longmans, Green and Co. Kettunen, Lauri. 1938. Livisches Wo¨rterbuch mit grammatischer Einleitung. Helsinki: Suomalais Ugrilainen Seura. Key, Harold H. 1967. Morphology of Cayuvava. The Hague: Mouton.

264

References

Khubchandani, Lachman M. 2003. Sindhi. In George Cardona and Dhanesh Jain (eds.), The Indo-Aryan languages. London: Routledge. 622–58. Kibrik, Aleksandr E. 1997. Ierarxii, roli, nuli, markirovannost0 i ‘anomal0 naja’ upakovka grammaticˇeskoj semantiki. Voprosy Jazykoznanija 4. 27–57. (ed.) 1999. E`lementy caxurskogo jazyka v tipologicˇeskom osvesˇcˇenii. Moscow: Nasledie. Kienast, Burkhart. 2001. Historische semitische Sprachwissenschaft. Wiesbaden: Harrassowitz. Kiparsky, Paul. 2001. Structural case in Finnish. Lingua 111. 315–76. Klenin, Emily. 1983. Animacy in Russian: a new interpretation. Columbus: Slavica. Koneski, Blazˇe. 1996. Istorija na makedonskiot jazik (reprint of 1986 edition). Skopje: Detska Radost. Kotyczka, Jozef. 1980. Kurze polnische Sprachlehre. Berlin: Volk und Wissen. Krishnamurti, Bhadriraju and John P. L. Gwynn. 1985. A grammar of modern Telugu. Delhi: Oxford University Press. Kuryłowicz, Jerzy. 1949 [1960]. De la nature des proce`s dits ‘analogiques’. Reprinted in: Esquisses linguistiques. Wroctław: Zaktład narodowy im. Ossolin´skich. 66–86. Kutsch Lojenga, Constance. 1994. Ngiti: A Central-Sudanic language of Zaire. Cologne: Ru¨diger Ko¨ppe. Laka¨mper, Renate and Dieter Wunderlich. 1998. Person marking in Quechua – a constraint-based minimalist analysis. Lingua 105. 113–48. Lamberti, Marcello. 1993. Die Shinassha-Sprache. Heidelberg: Carl Winter. Lefebvre, Claire and Anne-Marie Brousseau. 2002. A grammar of Fongbe. Berlin: Mouton de Gruyter. Leumann, Manu. 1977. Lateinische Laut- und Formenlehre (fifth edition). Munich: Beck. Liddell Henry G. and Robert Scott. 1996. A Greek–English lexicon. Oxford: Clarendon Press. Lindstrom, Lamont and John Lynch. 1994. Kwamera. Munich: Lincom. Lipkind, William. 1945. Winnebago grammar. New York: King’s Crown Press. Lorimer, David L. R. 1939. The Dum aki Language: outlines of the speech of the Doma, or B ericho, of Hunza. Nijmegen: Dekker and van de Vegt. Luraghi, Silvia. 2000. Synkretismus. In Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphologie: ein Handbuch zur Flexion und Wortbildung. Berlin: Mouton de Gruyter. 638–47. Lynch, John. 2000. A grammar of Anejom. Canberra: Australian National University. Magometbekova, Z. M. 1967. Karatinskij jazyk. In E. A. Bokarev and K. V. Lomatidze (eds.), Jazyki narodov SSSR (vol. IV: Iberijsko-kavkazskie jazyki). Moscow: Nauka. 323–35. Maiden, Martin. 1992. Irregularity as a determinant of morphological change. Journal of Linguistics 28/2. 285–312. Malećot, Andre´. 1963. Luisenõ, a structural analysis II: morphosyntax. International Journal of American Linguistics 29. 196–210. Maslova, Elena. 2003. A grammar of Kolyma Yukaghir. Berlin: Mouton de Gruyter.

References

265

Mathiassen, Terje. 1997. A short grammar of Latvian. Columbus: Slavica. Matras, Yaron. 2002. Romani: a linguistic introduction. Cambridge: Cambridge University Press. Mayerthaler, Willi. 1987. System-independent morphological naturalness. In Wolfgang U. Dressler (ed.), Leitmotifs in natural morphology. Amsterdam/ Philadelphia: John Benjamins. 25–58. McCarthy, John J. 2002. A thematic guide to Optimality Theory. Cambridge: Cambridge University Press. McCreight, Katherine and Catherine V. Chvany. 1991. Geometric representation of paradigms in a modular theory of grammar. In Frans Plank (ed.), Paradigms: the economy of inflection. Berlin: Mouton de Gruyter. 91–112. McGregor, Ronald S. 1977. Outline of Hindi grammar. Delhi: Oxford University Press. McGregor, William. 1990. A functional grammar of Gooniyandi. Amsterdam: John Benjamins. McGuckin, Catherine. 2002. Gapapaiwa. In John Lynch, Malcom Ross and Terry Crowley (eds.), The Oceanic languages. Richmond: Curzon. 297–321. Meillet, Paul and Andre´ Vaillant. 1934. Le slave commun. Paris: Institut d’e´tudes slaves. Meiser, G. 1992. Syncretism in Indo-European languages. Transactions of the Philological Society 90/2. 187–218. Merlan, Francesca. 1982. Mangarayi (Lingua Descriptive Studies 4). Amsterdam: North-Holland. Miller, Raymond. 1990. Form and function of the peripheral cases in some Slovene dialects. Slovene Studies 12/1. 5–22. Minassian, Martiros. 1980. Grammaire d’Armenien oriental. Delmar: Caravan Books. Mitchell, Terence F. 1962. Colloquial Arabic: the living language of Egypt. London: English Universities Press. Moravcsik, Edith and Jessica Wirth. 1986. Markedness – an overview. In Fred R. Eckman, Edith A. Moravcsik and Jessica R. Wirth (eds.), Markedness: proceedings of the twelfth annual linguistics symposium of the University of Wisconsin-Milwaukee, March 11–12, 1983. New York: Plenum Press. 1–11. Morgenstierne, Georg. 1941. Notes on Phal ura, an unknown language of Chitral. Oslo: Hos Jacob Dybwad (Skirfter utgit av Det Norske Videnskaps-Akademi i Oslo II. Hist.-Filos. Klasse. No. 5). Mous, Maarten. 1993. A grammar of Iraqw (Kuschitische Sprachstudien 9). Hamburg: Buske. Mu¨ller, Gereon. Forthcoming. A Distributed Morphology approach to syncretism in Russian noun inflection. In Olga Arnaudova, Wayles Browne, Maria Luisa Rivero and Dejan Stojanovic´ (eds.), Proceedings of Formal Approaches to Slavic Linguistics 12. Ann Arbor: Michigan Slavic Publications. Munro, Pamela. 1976. Mojave syntax. New York: Garland. Murane, Elizabeth. 1974. Daga grammar: from morpheme to discourse. Norman: Summer Institute of Linguistics.

266

References

Muysken, Pieter. 1997. Callahuaya. In: Sally G. Thomasan (ed.), Contact languages. Amsterdam/Philadelphia: John Benjamins. 427–48. Myachina, E. N. 1981. The Swahili language: a descriptive grammar (translated by G. L. Campdell). London: Routledge. Najlis, Elena. 1966. Lengua abipona (vol. I). Buenos Aires: Centro de Estudios, Universidad de Buenos Aires. Neidle, Carol J. 1988. The role of case in Russian syntax. Dordrecht: Kluwer. Neue, Friedrich and C. Wagener. 1902. Formenlehre der lateinischen Sprache (vol. I: Das Substantivum). Leipzig: O. R. Reisland. Newmark, Leonard. 1982. Standard Albanian: a reference grammar for students. Stanford: Stanford University Press. Nichols, Johanna. 1994. Ingush. In Rieks Smeets (ed.), The indigenous languages of the Caucasus (vol. IV). Delmar, New York: Caravan Books. 79–145. Nikolaeva, Irina and Evgenij A. Xelimskij. 1997. Jukagirskij jazyk. In Jazyki mira: paleoaziatskie jazyki. Moscow: Indrik. 155–68. Nikolaeva, Irina and Maria Tolskaya. 2001. A grammar of Udihe. Berlin: Mouton de Gruyter. Noonan, Michael. 1992. A grammar of Lango. Berlin: Mouton de Gruyter. Nordlinger, Rachel. 1998. A grammar of Wambaya, Northern Territory (Australia) (Pacific Linguistics C140). Canberra: Australian National University. Noreen, Adolf. 1923. Altislandische und altnorwegische Grammatik. Halle: Max Niemeyer. Noyer, Rolf. 1997. Features, positions and affixes in autonomous morphological structure. New York: Garland. 1998. Impoverishment theory and morphosyntactic markedness. In Steven G. Lapointe, Diane K. Brentari, and Patrick M. Farrell (eds.), Morphology and its relation to phonology and syntax. Stanford: CSLI Publications. 264–85. 2001. Clitic sequences in Nunggubuyu and PF convergence. Natural Language and Linguistic Theory 19/4. 751–826. Nussbaum, Loren V., William W. Gage and Daniel Varre. 1970. Dakar Wolof: a basic course. Washington, DC: Center for Applied Linguistics. Oates, William J. and Lynette F. Oates. 1968. Kapau pedagogical grammar (Pacific Linguistics C10). Canberra: Australian National University. Osborne, Charles R. 1974. The Tiwi language. Canberra: Australian Institute of Aboriginal Studies. Owens, Jonathan. 1985. A grammar of Harar Oromo (northeastern Ethiopia). Hamburg: Buske. Oxotina, Natalija V. 1961. Jazyk zulu. Moscow: Nauka. Perlmutter, David M. and Janez Oresˇ nik. 1973. Language-particular rules and explanation in syntax. In Steven R. Anderson and Paul Kiparsky (eds.), A festschrift for Morris Halle. New York: Holt Rinehart. 419–59. Pike, Kenneth. 1965. Non-linear order and anti-redundancy in German morphological matrices. Zeitschrift fu¨r Mundartforschung 32. 193–221. Pitman, Donald. 1980. Bosquejo de la gramatica araona (Notas Lingu¨ı´ sticas, no. 9). Riberalta, Bolivia: Instituto Linguistico de Verano.

References

267

Plank, Frans. 1991. Rasmus Rask’s dilemma. In Frans Plank (ed.), Paradigms: the economy of inflection. Berlin: Mouton de Gruyter. 161–96. Plungian, Vladimir. 1995. Dogon. Munich: Lincom. Pokorny, Julius. 1923. A historical reader of Old Irish. Halle: Max Niemeyer. Popjes, Jack, and Jo Popjes. 1986. Canela-Kraho. In Desmond C. Derbyshire and Geoffrey K. Pullum (eds.), Handbook of Amazonian languages (vol. I). Berlin: Mouton de Gruyter. 128–99. Pott, August. 1836. Etymologische Forschungen auf dem Gebiete der IndoGermanischen Sprachen (vol. II). Lemgo: Meyer’sche Hof-Buchhandlung. 1859. Etymologische Forschungen auf dem Gebiete der Indo-Germanischen Sprachen (vol. I: Prepositionen). Lemgo and Detmold: Meyer’sche HofBuchhandlung. Priestly, T. M. S. 1993. Slovene. In Bernard Comrie and Greville G. Corbett (eds.), The Slavonic languages. London: Routledge. 388–451. Reesink, Ger P. 1987. Structures and their functions in Usan: a Papuan language of Papua New Guinea. Amsterdam/Philadelphia: John Benjamins. (ed.), 2002. Languages of the Eastern Bird’s Head. Canberra: Australian National University. Reh, Mechthild. 1985. Die Krongo-Sprache (Niino Mo-Di). Berlin: Dietrich Reimer. Rice, Keren. 2000. Morpheme order and semantic scope. Cambridge: Cambridge University Press. Ringe, Donald. 1995. Nominative-accusative syncretism and syntactic case. Penn Working Papers in Linguistics 2. 45–81. Roberts, John R. 1987. Amele. London: Croom-Helm. Robins, R. H. 1958. The Yurok language. Berkeley: University of California Press. Romero-Figueroa, Andres. 1997. A reference grammar of Warao. Munich: Lincom. Rothe, Wolfgang. 1957. Einfu¨hrung in die historische Laut- und Formenlehre des Ruma¨nischen. Halle: Max Niemeyer. Rounds, Carol. 2001. Hungarian: an essential grammar. London/New York: Routledge. Rumsey, Alan. 1982. An intra-sentence grammar of Ungarinjin, North-Western Australia (Pacific linguistics B86). Canberra: Australian National University. Rupp, James E. 1989. Lealao Chinantec syntax (Studies in Chinantec Languages 2). Dallas: Summer Institute of Linguistics. Saeed, John. 1999. Somali. Amsterdam/Philadelphia: John Benjamins. Salminen, Tapani. 1997. Tundra Nenets inflection (Me´moires de la Socie´te´ FinnoOugrienne 227). Helsinki: Suomolais-Ugrilainen Seura. Sapir, J. David. 1965. A grammar of Diola-Fogny (West African Language Monographs 3). Ibadan: Cambridge University Press in association with the West African Languages Survey and the Institute of African Studies. Sasse, Hans-Ju¨rgen. 1976. Dasenech. In M. Lionel Bender (ed.), The non-Semitic languages of Ethiopia. East Lansing: African Studies Center, Michigan State University. 196–221. Schenker, Alexander M. 1964. Polish declension. The Hague: Mouton.

268

References

Serzisko, Fritz. 1982. Numerus/Genus-Kongruenz und das Pha¨nomen der Polarita¨t am Beispiel einiger ostkuschitischer Sprachen. In Hansjakob Seiler and Franz Josef Stachowiak (eds.), Apprehension: das sprachliche Erfassen von Gegensta¨nden (vol. II). Tu¨bingen: Narr. 179–200. Sharma, D. D. 1988. A descriptive grammar of Kinnauri. Delhi: Mittal Publications. Silverstein, Michael. 1976. Hierarchy of features and ergativity. In Robert M. W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies. 112–71. Sims-Williams, Nicholas. 1982. The double system of nominal inflection in Sogdian. Transactions of the Philogical Society 80. 67–76. Skorik, Pe¨tr Ia. 1961–77. Grammatika cˇukotskogo iazyka (vols. I and II). Leningrad: Izdatel0 stvo Akademii nauk SSSR. Smith, Lawrence R. 1979. Labrador Inittut inverted number marking, exchange rules and morphological markedness. Linguistics 17/1–2. 153–67. Sˇołcína, Jana and Edward Wornar. 2002. Obersorbisch im Selbststudium: ein Sprachkurs fu¨r Unerschrockene. Budysˇ in: LND (on-line edition available at http://www.serbski-institut.de/). Sologub, A. I. 1983. O sinkretizme form v sklonenii susˇ cˇestvitel0 nyx zˇenskogo roda edinstvennogo cˇisla po dialektnym dannym. In Ruben I. Avanesov (ed.), Russkie narodnye govory. Moscow: Nauka. 82–8. Soukka, Maria. 2000. A descriptive grammar of Noon: a Cangin language of Senegal. Munich: Lincom. Spencer, Andrew. 1991. Morphological theory. Oxford: Blackwell. 2000. Agreement morphology in Chukotkan. In Wolfgang U. Dressler, Oskar E. Pfeiffer, Markus A. Po¨chtrager and John R. Rennison (eds.), Morphological analysis in comparison. Amsterdam/Philadelphia: John Benjamins. 191–222. Sridhar, Shikaripur N. 1989. Kannada. London: Routledge. Stevenson, R. C. 1969. Bagirmi grammar (Linguistic Monograph Series No. 3). Khartoum: Sudan Research Unit, University of Khartoum. Stone, Gerald. 1993. Sorbian. In Bernard Comrie and Greville G. Corbett (eds.), The Slavonic languages. London: Routledge. 593–686. Stump, Gregory T. 1993. On rules of referral. Language 69. 449–79. 2001. Inflectional morphology. Cambridge: Cambridge University Press. Sunik, O. P. 1997. Ude`gejskij jazyk. In Jazyki mira: mongol0 skie jazyki, tunguso-man0 cˇzˇurskie jazyki, japonskij jazyk, korejskij jazyk. Moscow: Indrik. 236–48. Szemereńyi, Oswald. 1989. Einfu¨hrung in die vergleichende Sprachwissenchaft. Darmstadt: Wissenshaftliche Buchgesellschaft. Taylor, Charles V. 1985. Nkore-Kiga. London: Croom Helm. Terrill, Angela. 2003. A grammar of Lavukaleve. Berlin: Mouton de Gruyter. Tesnie`re, Lucien. 1925. Les formes du duel en slove`ne. Paris: Champion. Timberlake, Alan. 1975. Hierarchies in the genitive of negation, Slavic and East European Journal 19. 123–38. Todaeva, Buljasˇ X. 1997. Baoan0 skij jazyk. In Jazyki mira: mongol0 skie jazyki, tunguso-man0 cˇzˇurskie jazyki, japonskij jazyk, korejskij jazyk. Moscow: Indrik. 29–36.

References

269

Tokarski, J. 1993. Schematyczny indeks a tergo polskich form wyrazowych (revised and edited by Zygmunt Saloni). Warsaw: Wydawnictwo Naukowe PWN. Toporisˇ icˇ, Jozˇe. 1976. Slovenska slovnica. Maribor: Obzorja. Tosco, Mauro. 2001. The Dhaasanac language. Ko¨ln: Ru¨diger Ko¨ppe. Trask, R. L. 1997. A student’s dictionary of language and linguistics. London: Arnold. Tucker, Archibald N. and Margaret A. Bryan. 1966. Linguistic analysis: the nonBantu languages of North-Eastern Africa. London: Oxford University Press. Turton, David. 1981. Le Mun (Mursi). In G. Manessy (ed.), Les langues de l’Afrique subsaharienne (Les langues dans le monde ancien et moderne, vol. 1.1). Paris. Editions du CNRS. 335–49. Upadhyaya, U. Padmanabha. 1976. A comparative study of Kannada dialects: Bellary, Gulbarga, Kumta, and Nanjangud dialects. Mysore: Prasaranga, University of Mysore. Uspensky, Boris A. 1965. Strukturnaja tipologija jazykov. Moskva: Nauka. Verner G. K. [Werner, Heinrich]. 1999. Ketskij jazyk. In Jazyki mira: paleoaziatskie jazyki. Moscow: Indrik. 177–87. Viitso, Tiit-Rein. 1998. Fennic. In Daniel Abondolo (ed.), The Uralic languages. London: Routledge. 96–114. Voorhoeve, C. L. 1965. The Flamingo Bay dialect of the Asmat language. ’s-Gravenhage: Martinus Nijhoff. 1975. Central and western Trans-New Guinea phylum languages. In Stephan A. Wurm (ed.), Papuan languages and the New Guinea linguistic scene (Pacific Linguistics C38). Canberra: Australian National University. 345–459. Wackernagel, Jacob. 1924. Vorlesungen u¨ber Syntax (vol. I). Basel: Birkhaüser. Watkins, Calvert. 1962. Indo-European origins of the Celtic verb. Dublin: Institute of Advanced Studies. Watkins, Laurel J. 1984. A grammar of Kiowa. Lincoln: University of Nebraska Press. Watters, David E. 2003. A grammar of Kham. Cambridge: Cambridge University Press. Wegera, Klaus Peter. 2000. Morphologie des Fru¨hneuhochdeutschen. In Werner Besch, Anne Betten, Oskar Reichmann and Stefan Sonderegger (eds.), Sprachgeschichte: ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung (vol. 1.2). Berlin: Mouton de Gruyter. 1313–22. Werner, Roland. 1987. Grammatik des Nobiin (Nilnubisch). Hamburg: Helmut Buske. West, Dorothy. 1973. Wojokeso sentence, paragraph and discourse analysis (Pacific Linguistics B28). Canberra: Australian National University. Westermann, Diedrich. 1930. A study of the Ewe language. London: Oxford University Press. Whitney, William Dwight. 1889. Sanskrit grammar. Cambridge: Harvard University Press. Wiese, Bernd. 1996. Iconicity and syncretism. In Robin Sackmann (ed.), Theoretical linguistics and grammatical description: papers in honour of HansHeinrich Lieb on the occasion of his 60th birthday. Amsterdam/Philadelphia: John Benjamins. 323–44.

270

References

2003. Zur lateinischen Nominalflexion: die Form-Funktion-Beziehung. Ms., Institut fu¨r deutsche Sprache, Mannheim. Williams, Edwin. 1981. On the notions ‘lexically related’ and ‘head of a word’. Linguistic Inquiry 12/2. 245–74. Williams, Edwin. 1994. Remarks on lexical knowledge. Lingua 92. 7–34. Wilson, Darryl. 1974. Suena grammar (Workpapers in Papua New Guinea languages 8). Ukarumpa, Papua New Guinea: Summer Institute of Linguistics. Windisch, Rudolf. 1973. Genusprobleme im Romanischen: das Neutrum im Ruma¨nischen. Tu¨bingen: Pra¨zis. Wright, Joseph. 1930. Grammar of the Gothic language. Oxford: Oxford University Press (reprint of 1910 edition). Wunderlich, Dieter. 2001a. A correspondence-theoretic analysis of Dalabon transitive paradigms. In Geert Booij and Jaap van Marle (eds.), Yearbook of morphology 2000 Dordrecht: Kluwer. 233–52. 2001b. How gaps and substitutions can become optimal: the pronominal affix paradigms of Yimas. Transactions of the Philological Society 99. 315–66. 2004. Is there any need for the concept of directional syncretism? In L. Gunkel, G. Mu¨ller and G. Zifonun (eds.), Explorations in nominal inflection. Berlin: Mouton de Gruyter. 373–395. Wurm, Stephan A. 1975. The Trans-Fly (sub phylum level) stock. In Stephan A. Wurm (ed.), Papuan languages and the New Guinea linguistic scene (Pacific Linguistics C38). Canberra: Australian National University. 323–44. Wurzel, Wolfgang U. 1987. System-dependent morphological naturalness in inflection. In Wolfgang U. Dressler (ed.), Leitmotifs in natural morphology. Amsterdam/Philadelphia: John Benjamins. 59–96. Xajdakov, S. M. 1980. Principy immennoj klassifikacii v dagestanskix jazykax. Moscow: Nauka. 2001. Lakskij jazyk. In M. E. Alekseev (ed.), Jazyki mira: kavkazskie jazyki. Moscow: Academia. 347–57. Zaliznjak, Andrej A. 1973 [2002]. O ponimanii termina ‘padezˇ’ v lingvisticˇeskix opisanijax. In Russkoe imennoe slovoizmenenie. Moscow: Jazyki slavjanskoj kul0 tury. 613–47 (originally in Andrej A. Zaliznjak (ed.), Problemy grammaticˇeskogo modelirovanija. Moscow: Nauka. 53–87). Zasorina, L. N. 1977. Cˇastotnyi slovar 0 russkogo jazyka. Moscow: Russkij jazyk. Zˇirkov, L. I. 1955. Lakskij jazyk. Moscow: Izdatel0 stvo Akademii Nauk SSSR. Zˇukova, Alevtina N. 1972. Grammatika korjakskogo iazyka: fonetika, morfologija. Leningrad: Nauka. Zwicky, Arnold. 1985. How to describe inflection. In Proceedings of the eleventh annual meeting of the Berkeley Linguistics Society. 372–86. 1991. Systematic versus accidental phonological identity. In Frans Plank (ed.), Paradigms: the economy of inflection. Berlin: Mouton de Gruyter. 113–32. 2000. Describing syncretism: rules of referral after fifteen years. Presentation at the annual meeting of the Berkeley Linguistics Society (26), University of California, Berkeley.

Author Index

Abbot, Miriam 18 Ackema, Peter 65 Aikhenvald, Alexandra Y. 111 Allhiser de Engel, Mary 25, 26 Anttila, Raimo 51, 71 Arensen, Jon 65, 223, 230 Arkadiev, Pe¨tr M. 37 Armbruster, Charles Herbert 23, 168, 229 Aronoff, Mark 103, 184 Aronson, Howard I. 55, 56 Ashton, E. O. 231 Austin, Peter 51, 54, 144 Baerman, Matthew 10, 11, 37, 59, 62, 95, 110, 111, 114, 122, 226, 231 Bammesberger, Alfred 226, 229 Ba´tori, Istvań 37 Bavin, Edith L. 37 Bechhaus-Gerst, Marianne 75 Beeler, Madison S. 79 Be´jar, Susana 133 Bender, Lionel M. 230 Benjamin, Carmen 227, 231 Bentley, Delia 71 Berger, Hermann 224, 228 Berman, Ruth 229 Bhattacharya, Sudhibhushan 154 Bierwisch, Manfred 39, 126 Bindseil, Heinrich 4, 7 Blake, Barry J. 55 Bleiching, Doris 174 Blevins, James P. 15, 37 Bobaljik, Jonathan 63, 138, 160, 161 Boeder, Winfried 37 Booij, Geert 120, 123 Bowe, Heather 223 Boyeldieu, Pascal 84 Breedveld, J. O. 87 Breen, J.Gavan 44 Bright, William 229, 234

Brockelmann, Carl 143 Bromley, H. Myron 67, 229 Brousseau, Anne-Marie 60 Brown, Dunstan 10, 11, 37, 78, 95, 111, 113, 114, 122, 131, 144, 174, 175, 186, 187, 194, 196, 197, 205, 209, 226, 231, 236, 242, 248 Browne, Gerald M. 74, 168 Bryan, Margaret A. 233 Buck, Carl D. 6, 7, 23 Burrow, Thomas 154 Butt, John 231 Bybee, Joan 152 Cahill, Lynn 174 Cairns, Charles E. 22–3 Calabrese, Andrea 39 Cardona, George 70 Carmack, Stanford 37 Carstairs, Andrew see Carstairs-McCarthy, Andrew Carstairs-McCarthy, Andrew 37, 56, 111, 126, 151, 161–3, 168, 182, 214 Carter, Hazel 66, 230 Casad, Eugene H. 224 Chafe, Wallace 84 Chapman, Shirley 223 Charney, Jean Ormsbee 225 Childs, George Tucker 88 Chittleborough Martin 26 Chvany, Catherine V. 37, 130, 150, 151 Clairis, Christos 60 Coleman, Robert 5, 72, 208 Comrie, Bernard 7, 11, 42, 78, 181, 198 Corbett, Greville G. 10, 11, 29, 37, 78, 82, 86, 87, 92, 95, 104, 109, 111, 114, 122, 123, 138, 144, 147, 174, 175, 186, 187, 194, 196, 197, 205, 214, 216, 236, 242, 248 Cowan, H. K. J. 231, 235 Crowley, Terry 98

271

272

Author Index

Curnow, Timothy J. 228 Curtius, Georg 4 Cysouw, Michael 37, 60, 61, 68, 92 Dalrymple, Mary 131 Davies, John 21, 230 de Angulo, Jaime 101 Dedrick, John M. 224 Dench, Alan C. 44, 226 Derbyshire, Desmond 223, 229 Deza Galindo, Juan Francisco 22, 228 Dickens, Patrick 89, 90 Dimmendaal, Gerrit J. 81, 84 Dixon, Robert M. W. 111, 202, 223, 224 Donaldson, Tamsin 226 Drabbe, Peter 229, 230, 234 Drexel, Guido 174 Dryer, Matthew 7, 11 Dutton, Tom 64 Edel 0 man, D. I. 107 Elliot, Eric 49 Endzelins, Janis 100 Engel, Ralph 25, 26 Evans, Nicholas 11, 37, 78, 93, 144, 175, 186, 187, 188, 194, 196, 197, 226, 233, 236, 242 Evans, Roger 174, 177, 178, 179 Eytho´rsson, Tho´rhallur 71 Ezard, Bryan 96, 97 Feldstein, Ronald 142 Fennell, Trevor G. 16 Feoktistov, Aleksandr P. 54 Finkel, Raphael 47, 173 Fischer, Wolfdietrich 143 Foley, William A. 20, 60, 110, 231, 233 Fortescue, Michael 224, 233 Frachtenberg, Leo J. 105 Fradkin, Robert A. 37 Frank, Paul 156, 229 Frank, Wright Jay14 Franklin, Karl James 229 Fraser, Norman 11, 138, 147, 174, 175, 205, 214, 216, 248 Fromm, Hans 47, 50, 225 Gage, William W. 60 Gazdar, Gerald 174, 177, 178, 179 Gibbon, Dafydd 174 Gil, David 7, 11, 136 Glasgow, Kathleen 68 Goddard, Cliff 37, 42, 44 Gordon, Lynn 235 Graudin¸a, A. 99

Greenberg, Joseph H. 27, 29, 83, 88, 89, 114 Gregores, Emma 76, 233 Grimm, Jacob 3 Grjunberg, A. L. 107 Grosse, Siegfried 72 Gruzdeva, Elena Ju. 230 Gu¨ldemann, Tom 89 Gvozdanovic´, Jadranka 37 Gwynn, John P. L. 42 Hagman, Roy S. 230 Hall, Daniel Currie 133 Halle, Morris 39 Hamel, Patricia J. 97 Hansson, Gunnar O. 107 Harbour, Daniel 37 Harley, Heidi 60, 61, 64 Harvey, Mark 91, 128 Haspelmath, Martin 7, 11 Haugen, Einar 70 Haviland, John 43 Hayward, Richard 107 Healey, Alan 26 Heath, Jeffrey 37, 76, 187, 197, 230, 234, 235 Heine, Bernd 81, 82, 108, 109 Hekking, Ewald 231 Helimski, Eugene see Xelimskij, Evgenij A. Helmbrecht, Johannes 37, 64 Henderson, James 93–4 Herbermann, Charles G. 3 Hetzron, Robert 104 Hewitt, Brian G. 225, 233 Hewson, John 37 Hippisley, Andrew 175 Hjelmslev, Louis 4, 37, 48, 114 Hoff, B. J. 67, 228, 235 Holton, David 225 Horton, A. E. 230 Hualde, Jose´ Ignacio 223 Hudson, Richard A. 228 Huntley, David 214 Hutchison, John P. 234 Innes, Gordon 85 Isoroembo, Ambrose 26 Ivanov, Vjacˇeslav V. 39 Jackendoff, Ray S. 155 Jacobson, Steven A. 14, 48 Jakobi, Angelika 83, 98, 99 Jakobson, Roman O. 4, 39, 52, 210 Johnston, Jason 37, 127, 130, 150, 151 Kachru, Braj B. 24 Kaplan, Ronald M. 131 Karin¸sˇ , A. Krisˇ janis 99, 100

Author Index Karlson, Fred 62 Kathol, Andreas 49 Kennedy, Benjamin Hall 214 Kettunen, Lauri 71 Key, Harold H. 228 Khubchandani, Lachman M. 21 Kibrik, Aleksandr E. 37, 43, 76 Kienast, Burkhart 143 Kiparsky, Paul 37, 39, 45, 132 Klenin, Emily 49, 226 Koneski, Blazˇe 161, 182, 214 Kotyczka, Jozef 30 Krishnamurti, Bhadriraju 42 Kuryowicz, Jerzy 70, 71 Kutsch Lojenga, Constance 62, 230 Laka¨mper, Renate 37 Lamberti, Marcello 69 Lefebvre, Claire 60 Leumann, Manu 6 Liddell, Henry G. 3 Lindstrom, Lamont 62 Lipkind, William 60 Lorimer, David L. R. 43 Luraghi, Silvia 5, 38 Lynch, John 62, 72, 73 Mackridge, Peter 225 Magometbekova, Z. M. 82 Maiden, Martin 108 Makoondekwa, Joaõ 66, 230 Malećot, Andre´ 56 Maslova, Elena 224, 231 Mateo Alvarez, Jose´ 25, 26 Mathiassen, Terje 15, 42, 99, 226, 230 Matras, Yaron 73 Mayerthaler, Willi 170 McCarthy, John J. 22 McCreight, Katherine 37, 150, 151 McGregor, Ronald S. 225, 229 McGregor, William 233 McGuckin, Catherine 96 Meillet, Paul 168 Meiser, Gerhard 4, 5 Merlan, Francesca 223, 233 Miller, Raymond 39 Minassian, Martiros 47, 224 Mitchell, Terence F. 228 Moravcsik, Edith 22 Morgenstierne, Georg 15–16 Mous, Maarten 229, 234 Mu¨ller, Gereon 210 Munro, Pamela 80 Murane, Elizabeth 228 Muysken, Pieter 66 Myachina, Ekaterina N. 231

273

Najlis, Elena 18, 228 Neeleman, Ad 65 Neidle, Carol J. 39 Neue, Friedrich 139, 140 Newmark, Leonard 82 Nichols, Johanna 53, 226 Nikolaeva, Irina 25, 68, 231 Noonan, Michael 230 Nordlinger, Rachel 223 Noreen, Adolf 66 Noyer, Rolf 37, 62, 63, 64, 76, 160, 162–3, 166–7 Nussbaum, Loren V. 60 Oates, Lynette F. 229 Oates, William J. 229 Oresˇ nik, Janez 148, 214 Ortiz de Urbina, Jon 223 Osborne, Charles R. 231 Owens, Jonathan 223, 229 Oxotina, Natalija V. 231 Pace, Edward A. 3 Pallen, Conde´ B. 3 Perlmutter, David M. 148, 214 Philippaki-Warburton, Irene 225 Pike, Kenneth 8–9, 104 Pitman, Donald 224 Plank, Frans 150, 151 Plungian, Vladimir 27, 29, 83, 88, 89, 114 Pokorny, Julius 16 Popjes, Jack 228 Popjes, Jo 228 Pott, August 4, 7 Priestley, T. M. S. 176, 212 Reesink, Ger P. 60, 224, 231 Reh, Mechthild 226 Ricca, Davide 206 Rice, Keren 62 Ringe, Donald 40 Ritter, Elizabeth 60, 61, 64 Roberts, John R. 228 Robins, R. H. 42, 48, 224, 234 Romero-Figueroa, Andres 223 Rothe, Wolfgang 85 Rumsey, Alan 233 Rupp, James E. 230 Saeed, John 19, 104 Salminen, Tapani 48, 226, 230 Sapir, J. David. 229 Sasse, Hans-Ju¨rgen 107 Schenker, Alexander 39 Scott, Robert 3

274

Author Index

Serzisko, Fritz 108, 110, 132 Shahan, Thomas J. 3 Sharma, D. D. 62 Shen, Lei 47, 173 Shopen, Tim 37 Silverstein, Michael 44 Sims-Williams, Nicholas 53 Skorik, Pe¨tr Ia. 224, 228 Smith, Lawrence R 110 Sˇocína, Jana 14 Sologub, A. I. 167 Soukka, Maria 127 Spencer, Andrew 2, 37, 78 Sridhar, Shikaripur N. 69, 229 Stevenson, R. C. 228 Stone, Gerald 113 Stump, Gregory T. 8, 47, 63, 114, 131, 134, 136, 138, 142, 148, 160, 161, 163–6, 173, 182 Sua´rez, Jorge A. 76, 233 Sunik, O. P. 68 Suthar, Babu K. 70 Szemereńyi, Oswald 5, 72, 208 Taylor, Charles V. 230, 234 Terrill, Angela 230 Tesnie`re, Lucien 165 Thesayi, Suresh 47, 173 Timberlake, Alan 175 Todaeva, Buljasˇ X. 51 Tokarski, J. 30 Tolskaya, Maria 25, 68 Toporisˇ icˇ, Jozˇe 165 Tosco, Mauro 106, 236, 237, 238, 240, 241 Trask, R. Lawrence 2 Tucker, Archibald. N. 233 Turton, David 8, 65 Upadhyaya, U. Padmanabha 72 Uspensky, Boris A. 114

Vaillant, Andre´ 168 van Driem, George 120 Varre, Daniel 60, 100 Verner G. K. see Werner, Heinrich Viitso, Tiit-Rein 71 Voorhoeve, C. L. 60, 235 Vossen, Rainer 89 Wackernagel, Jacob 4 Wagener, C. 139 Watkins, Calvert. 63 Watkins, Laurel J. 94, 109, 110, 230 Watters, David E. 79 Wegera, Klaus Peter 72 Werner, Heinrich 229 Werner, Roland 66 West, Dorothy 105 Westermann, Diedrich 229 Whitney, William Dwight 47 Wiese, Bernd 39 Williams, Edwin 14, 150, 151 Wilson, Darryl 48, 67, 223, 231 Windisch, Rudolf 86 Wirth, Jessica 22 Wornar, Edward 14 Wright, Joseph 22, 92 Wunderlich, Dieter 37, 45, 142, 145, 147, 148, 149, 150, 194 Wurm, Stephan A. 102 Wurzel, Wolfgang U. 170 Wynne, John J. 3 Xajdakov, S. M. 25, 87, 230 Xelimskij, Evgenij A. 20, 231 Zaliznjak, Andrej A. 15, 37, 42 Zasorina, L. N. 248 Zˇirkov, L. I. 50, 226 Zˇukova, Alevtina N. 50, 77, 78, 95 Zwicky, Arnold 96, 131, 134, 135, 136, 144, 145, 163–4

Language index

The index includes languages discussed in the text or appendices. It does not feature languages which appear listed in the appendices but are not overtly discussed. Sources for these latter languages may be found in Baerman and Brown (2005a, b). Abipon (Guaicuruan) 17–18, 228 Albanian (Indo-European, Albanian) 82–3, 86 Aleut (Eskimo-Aleut) 59 Amele (Trans-New Guinea, Madang) 59, 60, 61, 228 Andi (Nakh-Dagestanian, Dagestanian, Avaro-Andi-Dido) 82, 86–7 Anejom (Austronesian, Oceanic) 72, 73, 169 Arabic (Afroasiatic, Semitic) 122, 142–3, 158, 228 Arabic, Classical see Arabic Arabic, Egyptian see Arabic Araona (Tacanan) 52, 224 Arapesh (Torricelli) 174 Arbore (Afroasiatic, Cushitic) 107 Armenian (Indo-European, Armenian) 47, 50, 52, 224 Asmat (Trans-New Guinea, Central and South New Guinea) 235 Atakapa (Isolate) 59 Awa (Trans-New Guinea, East New Guinea Highlands) 60 Awa Pit (Paezan, Barbacoan) 228 Aymara (Aymara) 21–2, 62, 228 Bagirmi (Nilo-Saharan, Bongo-Bagirmi) 228 Barbarenõ (Chumash) 79 Basque (Isolate) 48, 223 Beja (Afroasiatic, Beja) 228 Bininj Gunwok (Gunwinyguan) 92, 93, 175, 233 Bonan (Altaic, Mongolian) 51, 52, 136–8, 158 Boumaa Fijian (Austronesian, Oceanic) 223 Burarra (Burarran) 59, 62, 68, 159 Burushaski (Isolate) 52, 122, 224, 228

Callahuaya (Quechua based) 159, 160 Camus (Nilo-Saharan, Nilotic) 81, 108, 109, 110, 111 Canela-Kraho (Ge-Kaingang) 228 Carib (Carib) 67, 159, 160, 228, 235 Cayuvava (Cayuvava) 228 Chichimeco (Oto-Manguean, Otopamean) 100, 103 Chitimacha (Isolate) 59 Chukchi (Chukotko-Kamchatkan) 224, 228 Comanche (Uto-Aztecan, Numic) 224 Daga (Trans-New Guinea, Central and Southeast New Guinea) 228 Dalabon (Gunwinyguan) 78, 171, 172, 183, 186–204, 217, 218, 242–7 Dani (Trans-New Guinea, Dani-Kwerba) 67, 159, 229 Danish (Indo-European, Germanic) 74 Dhasanaac (Afroasiatic, Cushitic) 106, 107, 167, 169–70, 171, 174, 183–6, 217, 218, 236–41 Diola-Fogny (Niger-Congo, Northern Atlantic) 229 Diyari (Pama-Nyungan) 54, 143–4 Dogon (Niger-Congo, Dogon) 26, 59 Domaaki (Indo-European, Dardic) 42, 43, 48 Dulong see Trung Dutch (Indo-European, Germanic) 65, 70, 159, 160 Dyirbal (Pama-Nyungan) 155 Eastern Armenian see Armenian Egyptian Arabic see Arabic Ekari (Trans-New Guinea, Wissel Lakes-Kemandoga) 229

275

276

Language Index

English (Indo-European, Germanic) 3, 26, 95–6, 100, 101, 103, 180, 223, 229 Erzja Mordvin (Uralic, Finnic) 54, 158 Estonian (Uralic, Finnic) 71 Ewe (Niger-Congo, Kwa) 229 Finnish (Uralic, Finnic) 47, 48, 49, 50, 51, 62, 132, 211, 225 Fongbe (Niger-Congo, Kwa) 60 French (Indo-European, Italic) 62, 225, 229 Fula (Niger-Congo, Northern Atlantic) 87–8 Fur (Nilo-Saharan, Fur) 83 Gaagudju (Gaagudju) 90–1, 127, 128–9, 130, 150, 171 Gapapaiwa (Austronesian, Oceanic) 96, 103, 160 Georgian (Kartvelian) 53, 225, 233 German (Indo-European, Germanic) 8–9, 19, 53, 59, 71, 72, 74, 83, 168, 198, 225, 229 Gooniyandi (Bunaban) 233 Gothic (Indo-European, Germanic) 21–2, 23, 74, 92 Grebo (Niger-Congo, Kru) 85 Greek (Indo-European, Greek) 3, 5, 38–9, 40, 42, 47, 139, 140, 158, 225 Greek, Ancient see Greek Greek, Modern see Greek Guambiano (Paezan, Barbacoan) 59 Guarani (Tupi-Guarani) 76, 233 Gujarati (Indo-European, Indic) 70, 136–60 Guugu Yimidhirr (Pama-Nyungan) 42–3 Harar Oromo (Afroasiatic, Cushitic) 223, 229 Hebrew (Afroasiatic, Semitic) 229 Hindi (Indo-European, Indic) 59, 225 Hixkaryana (Carib) 229 Hungarian (Uralic, Ugric) 156–7 Hunzib (Nakh-Dagestanian, Dagestanian, Avaro-Andi-Dido) 59 Icelandic, Old (Indo-European, Germanic) 66, 70, 71, 142, 159, 160, 167 Ika (Chibchan, Aruak) 156, 229 Ingush (Nakh-Dagestanian, Nakh) 53, 59, 226 Iraqw (Afroasiatic, Cushitic) 229, 234 Irish (Indo-European, Celtic) 16, 226, 229 Irish, Old see Irish Italian (Indo-European, Italic) 206 Ju/ 0 hoan (Khoisan, Northern Khoisan) 83, 88–90, 89, 91

Kalam (Trans-New Guinea, East New Guinea Highlands) 60 Kamoro (Trans-New Guinea, Central and South New Guinea 60 Kannada (Dravidian) 69, 72, 74, 159, 229 Kanuri (Nilo-Saharan, Saharan) 234 Kapau (Trans-New Guinea, Angan) 59, 229 Karata (Nakh-Dagestanian, Dagestanian, Avaro-Andi-Dido) 82–3 Karok (Hokan, Karok) 229, 234 Kashmiri (Indo-European, Dardic) 23–4, 122 Kawasqar (Qawesqar) 60 Kayardild (Tangic) 226 Ket (Ket) 229 Kewa (Trans-New Guinea, East New Guinea Highlands) 59, 229 Kinnauri (Sino-Tibetan, Western Himalayish) 62 Kiowa (Aztec-Tanoan, Tanoan) 93–4, 109, 110, 111, 230 Kisi (Niger-Congo, Southern Atlantic) 83, 88–9 Kiwai (Trans-New Guinea, Trans-Fly) 59, 101, 102, 103 Kobon (Trans-New Guinea, East New Guinea Highlands) 21, 23, 59, 60, 61, 230 Koiari (Trans-New Guinea, Central and Southeast New Guinea) 59, 64, 159 Kongo (Niger-Congo, Bantu) 66, 159, 160, 230 Korafe (Trans-New Guinea, Binanderean) 60 Koryak (Chukotko-Kamchatkan) 49, 50, 94, 95, 114, 115, 119, 158, 160 Krongo (Kordofanian) 226 Kunama (Nilo-Saharan, Kunama) 230 Kwamera (Austronesian, Oceanic) 61, 62 Laal (Unclassified) 84, 85 Lak (Nakh-Dagestanian, Dagestanian, Lak-Dargwa) 24, 50, 52, 136, 158, 226, 230 Lango (Nilo-Saharan, Nilotic) 230 Latin (Indo-European, Italic) 4–5, 6, 23, 24, 46, 55, 86, 90, 111, 134, 135, 139–42, 148, 151, 158, 214 Latvian (Indo-European, Baltic) 15–16, 40, 42, 46, 98–100, 103, 158, 226, 230 Lavukaleve (East Papuan, Solomons East Papuan) 230 Lealao Chinantec (Oto-Manguean, Chinantecan) 230 Lezgian (Nakh-Dagestanian, Dagestanian, Lezgic) 226 Limbu (Tibeto-Burman, Kiranti) 52, 120

Language Index Livonian (Uralic, Finnic) 64, 71, 167 Loniu (Austronesian, Oceanic) 97, 160 Luisenõ (Aztec-Tanoan, Uto-Aztecan, Takic) 49, 56 Luvale (Niger-Congo, Bantu) 230 Maba (Nilo-Saharan, Maban) 233 Macedonian (Indo-European, Slavonic) 63, 65, 161 Macushi (Carib) 18, 35 Mam (Mayan) 62 Manchad (Tibeto-Burman, Kiranti) 59 Mangarayi (Mangarayi) 48, 223, 233 Mansim (West Papuan) 60 Maricopa (Hokan, Yuman) 235 Marind (Trans New Guinea, Marind) 230, 234 Martuthunira (Pama-Nyungan) 51, 226 Mayali see Bininj Gun-Wok Meyah (East Bird’s Head) 60 Mojave (Hokan, Yuman) 80 Murle (Nilo-Saharan, Surma) 64, 65, 159, 223, 230 Mursi (Nilo-Saharan, Surma) 8, 65 Nama (Khoisan, Central Khoisan) 230 Nenets (Uralic, Samoyedic) 47, 48, 226, 230 Nez Perce (Penutian, Sahaptian) 59, 60, 61 Nganasan (Uralic, Samoyedic) 60 Ngiti (Nilo-Saharan, Balendru) 62, 230 Ngiyambaa (Pama-Nyungan) 51, 52, 226 Nimboran (Trans-New Guinea, Nimboran) 162–3 Nivkh (Isolate) 59, 230 Nkore-Kiga (Niger-Congo, Bantu) 229, 230, 234 Nobiin see Nubian Noon (Niger-Congo, Northern Atlantic) 127–8, 129, 130, 150 Norwegian (Indo-European, Germanic) 74 Nubian (Nilo-Saharan, Nubian) 22, 23, 59, 65, 69, 74, 75, 159, 160, 168, 229 Nubian, Dongolese see Nubian Nubian, Old see Nubian Nuer (Nilo-Saharan, Nilotic) 14–15, 17 Nunggubuyu (Nunggubuyu) 62, 72, 166–7, 230, 234, 235 Old English see English Old High German see German Orokaiva (Trans New Guinea, Binanderean) 26 Otomı´ (Oto-Manguean, Otomian) 231 Pashto (Indo-European, Iranian) 106, 107 Paumari (Arauan) 223

277

Pengo (Dravidian) 52, 153 Phalura (Indo-European, Dardic) 15, 52 Pitjantjatjara (Pama-Nyungan) 223 Polish (Indo-European, Slavonic) 30–1, 39, 174 Prinmi (Sino-Tibetan, Qiangic) 59 Pumi see Prinmi Quechua (Quechua) 66 Rawang see Trung Romani (Indo-European, Indic) 73 Romanian (Indo-European, Italic) 83, 85–6, 142 Russian (Indo-European, Slavonic) 1–3, 9–10, 28–9, 30, 31–2, 34, 38–9, 40, 42, 47, 49, 50, 52, 53, 55, 57, 83, 112, 136, 139, 140, 142, 145–9, 158, 166, 167, 172, 173, 174, 175, 181, 183, 204, 214, 215, 218, 227, 248–53 Sami (Uralic, Finnic) 107–8, 110 Sango (Niger-Congo, Adamawa-Ubangian) 59, 60 Selkup (Uralic, Samoyedic) 20, 23 Seneca (Macro-Siouan, Iroquoian) 84, 85 Sentani (Trans-New Guinea, Sentani) 231, 235 Serbo-Croat (Indo-European, Slavonic) 211 Shinassha (Afroasiatic, Omotic) 69, 159 Sindhi (Indo-European, Indic) 21 Siuslaw (Penutian, Siuslaw) 105 Slave (Nadene, Athapaskan-Eyak) 60 Slovene (Indo-European, Slavonic) 33–4, 39, 59, 92, 114, 116, 165–6, 175–7, 183, 212 Sogdian (Indo-European, Iranian) 53 Somali (Afroasiatic, Cushitic) 18–19, 104, 105, 110, 117, 122, 132 Sorbian (Indo-European, Slavonic) 14, 112, 118, 119 Spanish (Indo-European, Italic) 227, 231 Suena (Trans-New Guinea, Binanderean) 48, 67, 159, 223, 231 Swahili (Niger-Congo, Bantu) 231 Swedish (Indo-European, Germanic) 74 Sye (Austronesian, Oceanic) 98 Tawala (Austronesian, Oceanic) 96–7, 103, 160 Telugu (Dravidian) 40, 42, 47, 85 Tetun (Austronesian, Timor) 59 Tiwi (Tiwi) 62, 231 Tokharian (Indo-European, Tokharian) 86 Trung (Tibeto-Burman, Nungish) 78, 79 Tsakhur (Nakh-Dagestanian, Dagestanian, Lezgic) 42, 43, 48, 117

278

Language Index

Tsez (Nakh-Dagestanian, Dagestanian, Avaro-Andi-Dido) 181–2 Turkana (Nilo-Saharan, Nilotic) 80, 81, 84 Turkish (Altaic, Turkic) 55

West Greenlandic (Eskimo-Aleut) 224, 233 Winnebago (Macro-Siouan, Siouan) 60 Wojokeso (Trans-New Guinea, Angan) 105 Wolof (Niger-Congo, Northern Atlantic) 60

Udihe (Altaic, Tungusic) 24, 62, 68, 159, 160 Ungarinjin (Wororan) 233 Ura (Austronesian, Oceanic) 97–8, 103 Usan (Trans-New Guinea, Adelbert Range) 231

Yaqui (Aztec-Tanoan, Uto-Aztecan, Taracahitic) 48, 224 Yele (East Papuan, Solomons East Papuan) 93–4 Yidiny (Pama-Nyungan) 224 Yimas (Sepik-Ramu, Nor-Pondo) 20, 60, 231, 233 Yukaghir (Yukaghir) 224, 231 Yup’ik, Central Alaskan (Eskimo-Aleut) 13–14, 17, 48, 93, 224 Yurok (Algic, Yurok) 40, 42, 48, 224, 234

Wagaya (Pama-Nyungan), 43–4 Wambaya (West Barkly) 223 Wambon (Trans-New Guinea, Central and South New Guinea) 59 Warao (Isolate) 223 Warekena (Maipurean) 60 Waskia (Trans-New Guinea, Madang) 59

Zoque (Mixe-Zoquean) 25, 59 Zulu (Niger-Congo, Bantu) 231

Subject Index

ablaut 98–9 affix suppression 96–100, 101 agglutinative morpology 55, 56 agreement 28, 32, 33, 82, 89, 92, 109, 112, 120, 122, 123, 124, 216 analogy 7, 63, 70, 71, 72, 105, 107, 161–8, 220 animacy 38, 45, 47, 50, 56, 57, 76, 81, 86, 89, 90, 109, 127, 145, 146, 147, 148, 181, 214–16, 217 Animacy Hierarchy 44–9, 56, 76, 81, 84, 92, 109 number as a component of 48–9 word class as a component of 47–8, 50 aspect see TAM attribute ordering 180, 182, 183, 184, 200, 201–2, 203 avoidance strategy 171, 175, 187, 197–8, 202, 217 basic form see defaults, morphological blocking 135, 176 canonical inflection 27 cardinality of features see features, number of values case 1–2, 13–16, 20, 28, 35, 37, 38–57, 111, 112, 113, 114, 116, 117, 118, 119, 123, 124, 134, 136, 137, 143, 144, 145, 146, 151, 152, 158, 169, 220, 223–7 core cases 40, 42, 47, 48, 49, 52, 53, 56, 124, 144, 158, 223–7 peripheral cases 40, 49, 51, 52, 56, 224–7 semantic cases 52 syntactic cases 52 case systems ergative-absolutive 40, 42, 47, 49 nominative-accusative 40, 42, 49 change, diachronic 52, 54, 55, 56, 57, 60, 61, 62, 63, 70–5, 82, 89, 91, 92, 95, 98, 105, 124, 125, 145, 152, 168–9, 169–70, 216

morphosyntactic 4–7 phonologically conditioned 4–7, 70, 161, 167, 170, 221 conjugation see inflectional class cumulative exponence 151, 155 DATR 177–80, 183, 236, 253 declension see inflectional class defaults 26, 45, 61, 71, 74, 88, 89, 100, 109, 127, 130, 132, 135, 139, 140, 142, 144, 146, 148, 159 default inheritance 173, 174–5, 177–80 morphological default 26, 45, 46, 64, 70, 96, 143, 148, 183 definiteness 38, 45, 54, 62, 84, 117, 142 differential agent marking 147, 158 differential object marking 147, 158 directionality 13, 23, 24–5, 70–5 directional effects 25, 57, 63–70, 74, 75, 134, 135, 145, 153, 155, 158–61, 164, 166–7, 183 directional rules 133, 139, 143, 144, 145, 147, 148, 149, 150, 221 directional syncretism 136–44, 145, 148, 149, 150, 160, 163, 174 see also take-overs disjunction 81, 126, 154–5, 157, 176 elsewhere condition 135, 150 elsewhere form 127, 129, 130, 157, 182, 183 see also Panini’s principle exceptionality 172, 176, 177 Feature Ranking Principle 114, 158–60, 164–6 feature structure 13, 103, 126, 127, 132, 133, 151, 161–2, 169, 170, 171, 181, 210–11 cross-classifying 126, 127, 130–1, 132, 133 flat 126, 127–9, 130, 155, 160, 162, 163, 166, 171 hierarchical 17, 35, 61, 64, 75, 126, 127–8, 129–30, 132, 133, 150, 160, 171

279

280

Subject Index

features as a context for syncretism 19–23, 20, 29, 30, 31, 32, 33, 34, 35, 38, 59, 61, 90, 91, 111, 112, 114, 116, 117, 118–19, 120, 121, 122, 136, 151 contextual 124 feature dependency 111, 113 feature ordering 35, 113, 114, 118, 119, 151, 172, 173, 174, 182 inherent 109, 112, 120, 123 nominal 113–19 non-autonomous values 15, 43, 54, 83 number of values 123 verbal 119–22 frequency 37, 75, 110, 117, 152 gender 21, 28–9, 31, 32, 33, 34, 35, 37, 53, 81–91, 106, 110, 111, 112, 113, 114, 116, 117, 118, 119, 120, 121, 122, 123, 124, 127, 128, 129, 130, 132, 135, 140, 151, 180–1, 216, 220 convergent systems 82–3, 84, 127 crossed systems 82–3 semantics of 81, 84–5, 86–7, 88, 89, 91 glossing conventions 11–12 harmonicity 188 homophony 5, 151, 152 accidental 9–10, 24, 61, 81, 141, 142, 148, 160–3, 166, 167, 168, 206 avoidance of 70, 157, 168 impoverishment 63, 160–3, 166 inclusive–exclusive distinction 57, 61, 62, 110, 111 indexes, morphomic 103, 135, 174, 182, 184, 221 individuation 44, 50, 109 inferential-realizational theories 171, 172, 175–7 inflectional classes 38, 47, 48, 49, 59, 91, 93, 97, 98, 99, 100, 108, 114, 115, 119, 134, 135, 139, 140, 141, 142, 145, 148, 155, 204–5, 207–11, 214 inflection–derivation distinction 3, 8 inverse 76, 78, 81, 94, 109 irregularity 176 KATR 173 Lexical-Functional Grammar 131 lexical knowledge representation 174 lexical semantics 3, 154–5

markedness 22–3, 46, 49, 56, 57, 63, 64, 94, 110, 132, 133, 158–60, 162, 163, 164–6, 168 see also unmarkedness mood see TAM morphological class 17–19, 204 morphomes see indexes, morphomic mother-in-law language 155, 175 natural classes 103, 108, 111, 126–31, 136, 146, 171 see also unnatural classes negation 22–3, 74, 120, 122 Network Morphology 11, 12, 113, 138, 171, 172–80, 188 neutralization 13, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 76, 83, 92, 114, 122, 173, 181, 182, 183, 194–6 number 21, 23, 25, 26, 31, 32, 35, 37, 38, 45, 47, 48, 49, 59, 60, 61, 68, 69, 72, 79, 82–90, 91, 96, 97, 98, 100, 101, 107, 108, 109, 110, 111, 112, 113, 114, 116, 117, 118, 119, 120, 121, 122, 123, 124, 132, 151, 152, 161–2, 163, 165, 169, 181, 188, 199, 200–1, 216, 220 Number Hierarchy 95 object 40, 42, 56, 57, 75, 76, 77, 78, 79–80, 81, 91, 92, 120, 128 Optimality Theory 45, 132, 137, 145–50 Panini’s principle 135, 138, 190, 192 see also elsewhere condition Paradigm Function Morphology 138, 171 person 11, 17–18, 19–22, 25, 26, 31, 32, 37, 45, 48, 57, 61, 62, 63, 75–81, 92, 93, 94, 95, 96, 97, 98, 100, 101, 104, 105, 106, 107, 109, 110, 111, 117, 120, 121, 122, 124, 134, 136, 151, 156, 159–60, 169, 197, 199, 200–1, 204, 220, 228, 235 polarity 16, 17, 35, 103–11, 122, 132–3, 151, 184 prepositions, case government of 55 prosody 98 regularity 23–4, 172, 183, 207, 212, 218 relevance hierarchy 152–3 rules rule ordering 138, 139, 140, 141, 145, 148 rules of exponence 135, 137 rules of referral 134, 135, 137, 160, 161, 164, 165, 167, 171, 172, 173–4, 182, 183, 187, 188, 197–9, 201, 202, 203, 204, 210, 218, 221, 222 symmetrical rules 133–50

Subject Index semantics 38–9, 54, 56, 57, 60, 61, 62, 63, 82, 89, 91, 92, 95, 98, 105, 124, 125, 152, 216 semantic naturalness 61, 171, 180–2 semantic values 55, 56 see also systematicity, semantic sound change; see change, phonologically conditioned split ergativity 42, 43, 44, 45, 48 stems 53, 97, 98–100, 106, 107, 139, 148 bare stems see defaults, morphological stress see prosody subject 40, 43, 57, 65, 75, 76–8, 79, 80, 81, 91, 92, 104, 105, 106, 109, 110, 120, 128 suppletion 176 Surrey Syncretisms Database 10, 111, 114–22, 124, 219 syncretism block 8 canonical 27, 33–5, 38 complete 59, 62 contrary 15–17, 35, 150, 151 convergent 145 history of 3–4 lexically determined 206–7 morphologically determined 207–12 nested 14, 17, 35, 129, 150 simple 13, 17, 35, 83 syntactically determined 213–16 partial 59, 61, 62, 63, 64, 81, 129, 146, 147, 157, 158, 163, 164, 171, 172, 173, 182, 183, 184, 187, 188, 194, 196, 198, 199, 200, 201, 202, 203, 211, 218, 221 syntactic class see word class Systematic Homonymy Claim 152–4, 182 systematicity 9–10, 22, 24, 139, 141, 148, 153, 157, 160–1, 166–7, 168–9, 182–3, 217

281

morphological 105–8, 111, 157, 166, 167, 168, 184, 186, 198, 217 semantic 108–11 take-overs 152–3, 155, 156, 162, 182 see also directionality TAM (tense-aspect-mood) 26, 37, 57, 59, 61, 62, 63, 64, 65, 66, 67, 70, 72, 73, 91, 93, 95, 110, 111, 120–2, 124, 156, 160, 164 tense see TAM tone see prosody transitivity 187, 194, 197, 204 underspecification 35, 59, 61, 62, 63, 64, 81, 129, 146, 147, 155, 157, 158, 163, 164, 171, 172, 173, 180–2, 183, 184, 187, 188, 191–4, 196, 198, 199, 200, 201, 202, 203, 204, 211, 218, 221 uninflectedness 13, 27, 30–3, 32, 33, 34, 35, 36, 173, 181, 182, 183 unmarkedness 23, 25–7, 44, 45, 56, 63, 128, 136, 145, 146, 155–6, 158, 161, 169 see also markedness unnatural classes 103, 108, 131–2 see also natural classes verbs, two-place 75–81, 197 voice 120 Watkins’ Law 63 word class 38, 45, 47, 48, 211 World Atlas of Language Structures 7, 11, 12, 37, 57, 75, 76, 79, 219 zero ending see defaults, morphological zero morphology see defaults, morphological

The Syntax-Morphology Interface: A Study of Syncretism (Cambridge Studies in Linguistics)

Ergativity (Cambridge Studies in Linguistics)

The Syntax of Welsh: A Transformational Study of the Passive (Cambridge Studies in Linguistics (No. 18))

Derivations in Minimalism (Cambridge Studies in Linguistics)

A Notional Theory of Syntactic Categories (Cambridge Studies in Linguistics)

The Acquisition of Two Languages from Birth: A Case Study (Cambridge Studies in Linguistics)

The Linguistics of Laughter: A corpus-assisted Study of Laughter-talk (Routledge Studies in Linguistics)

Morphological Productivity (Cambridge Studies in Linguistics)

Syntax and Parsing (Cambridge Studies in Linguistics)

Morphological Productivity (Cambridge Studies in Linguistics)

Functional Categories (Cambridge Studies in Linguistics)

Fluent Aphasia (Cambridge Studies in Linguistics)

Intonational Phonology (Cambridge Studies in Linguistics)

Morphological Productivity (Cambridge Studies in Linguistics)

The Syntax of Agreement and Concord (Cambridge Studies in Linguistics)

The Syntax of Argument Structure (Cambridge Studies in Linguistics)

The Syntax of Adjuncts (Cambridge Studies in Linguistics)

The Syntax of Negation (Cambridge Studies in Linguistics)

Relevance Relations in Discourse: A Study with Special Reference to Sissala (Cambridge Studies in Linguistics)

The Contrastive Hierarchy in Phonology (Cambridge Studies in Linguistics, 121)

A-Morphous Morphology (Cambridge Studies in Linguistics (No. 62))

How to Study Linguistics: A Guide to Study Linguistics

Cognitive Linguistics (Cambridge Textbooks in Linguistics)

On Definiteness: A Study with Special Reference to English and Finnish (Cambridge Studies in Linguistics)

Principles of Dependency Phonology (Cambridge Studies in Linguistics (No. 47))

Principles of English Stress (Cambridge Studies in Linguistics)

Predication Theory: A Case Study for Indexing Theory (Cambridge Studies in Linguistics)

Velimir Khlebnikov: A Critical Study (Cambridge Studies in Russian Literature)

Velimir Khlebnikov: A Critical Study (Cambridge Studies in Russian Literature)

The Syntax-Morphology Interface: A Study of Syncretism (Cambridge Studies in Linguistics)

Ergativity (Cambridge Studies in Linguistics)

The Syntax of Welsh: A Transformational Study of the Passive (Cambridge Studies in Linguistics (No. 18))

Derivations in Minimalism (Cambridge Studies in Linguistics)

A Notional Theory of Syntactic Categories (Cambridge Studies in Linguistics)

The Acquisition of Two Languages from Birth: A Case Study (Cambridge Studies in Linguistics)

The Linguistics of Laughter: A corpus-assisted Study of Laughter-talk (Routledge Studies in Linguistics)

Morphological Productivity (Cambridge Studies in Linguistics)

Syntax and Parsing (Cambridge Studies in Linguistics)

Morphological Productivity (Cambridge Studies in Linguistics)

Functional Categories (Cambridge Studies in Linguistics)

Fluent Aphasia (Cambridge Studies in Linguistics)

Intonational Phonology (Cambridge Studies in Linguistics)

Morphological Productivity (Cambridge Studies in Linguistics)

The Syntax of Agreement and Concord (Cambridge Studies in Linguistics)

The Syntax of Argument Structure (Cambridge Studies in Linguistics)

The Syntax of Adjuncts (Cambridge Studies in Linguistics)

The Syntax of Negation (Cambridge Studies in Linguistics)

Relevance Relations in Discourse: A Study with Special Reference to Sissala (Cambridge Studies in Linguistics)

The Contrastive Hierarchy in Phonology (Cambridge Studies in Linguistics, 121)

A-Morphous Morphology (Cambridge Studies in Linguistics (No. 62))

How to Study Linguistics: A Guide to Study Linguistics

Cognitive Linguistics (Cambridge Textbooks in Linguistics)

On Definiteness: A Study with Special Reference to English and Finnish (Cambridge Studies in Linguistics)

Principles of Dependency Phonology (Cambridge Studies in Linguistics (No. 47))

Principles of English Stress (Cambridge Studies in Linguistics)

Predication Theory: A Case Study for Indexing Theory (Cambridge Studies in Linguistics)

Velimir Khlebnikov: A Critical Study (Cambridge Studies in Russian Literature)

Velimir Khlebnikov: A Critical Study (Cambridge Studies in Russian Literature)

Recommend Documents