Contrast and Representations in Syntax 0198817924, 9780198817925

This book explores how grammatical oppositions - for instance, the contrast between present and past tense - are represe

562 69 2MB

English Pages 336 [331] Year 2020

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Contrast and Representations in Syntax
 0198817924, 9780198817925

Table of contents :
Cover
Title Pages
Contents
General preface
Acknowledgements
List of figures and tables
List of abbreviations
The contributors
1. Contrast and representations in syntax: Introduction
PART I FEATURES IN THE INFLECTIONAL SPINE
2. A feature-geometric approach to verbal inflection in Onondaga
3. Restricted and reversed aspectual contrasts
4. Sentience-based event structure: Evidence from Blackfoot
PART II CONTRAST IN THE ARGUMENT DOMAIN
5. Definite expression and degrees of definiteness
6. Cross-linguistic contrasts in the structure of causatives in clausal nominalizations
7. The T ł˛ıch o˛ syntactic causative and non-nominal CPs
PART III ARCHITECTURAL QUESTIONS
8. Against some approaches to long-distance agreement without Agree
9. Contrast in syntax and contrast in phonology: Same difference?
References
Index
Oxford_Studies_in_Theoretical_Linguistics

Citation preview

Contrast and Representations in Syntax

OX F OR D STU DIE S IN THE ORETICAL LINGUISTICS General Editors David Adger and Hagit Borer, Queen Mary University of London Advisory Editors Stephen Anderson, Yale University; Daniel Büring, University of Vienna; Nomi Erteschik-Shir, Ben-Gurion University; Donka Farkas, University of California, Santa Cruz; Angelika Kratzer, University of Massachusetts, Amherst; Andrew Nevins, University College London; Christopher Potts, Stanford University; Barry Schein, University of Southern California; Peter Svenonius, University of Tromsø; Moira Yip, University College London recent titles 62 The Morphosyntax of Transitions A Case Study in Latin and Other Languages by Víctor Acedo-Matellán 63 Modality Across Syntactic Categories edited by Ana Arregui, María Luisa Rivero, and Andrés Salanova 64 The Verbal Domain edited by Roberta D’Alessandro, Irene Franco, and Ángel J. Gallego 65 Concealed Questions by Ilaria Frana 66 Parts of a Whole Distributivity as a Bridge between Aspect and Measurement by Lucas Champollion 67 Semantics and Morphosyntactic Variation Qualities and the Grammar of Property Concepts by Itamar Francez and Andrew Koontz-Garboden 68 The Structure of Words at the Interfaces edited by Heather Newell, Máire Noonan, Glyne Piggott, and Lisa deMena Travis 69 Pragmatic Aspects of Scalar Modifiers The Semantics–Pragmatics Interface by Osamu Sawada 70 Encoding Events Functional Structure and Variation by Xuhui Hu 71 Gender and Noun Classification edited by Éric Mathieu, Myriam Dali, and Gita Zareikar 72 The Grammar of Expressivity by Daniel Gutzmann 73 The Grammar of Copulas Across Languages edited by María J. Arche, Antonio Fábregas, and Rafael Marín 74 The Roots of Verbal Meaning by John Beavers and Andrew Koontz-Garboden 75 Contrast and Representations in Syntax edited by Bronwyn M. Bjorkman and Daniel Currie Hall 76 Nominalization 50 Years on from Chomsky’s Remarks edited by Artemis Alexiadou and Hagit Borer For a complete list of titles published and in preparation for the series, see pp. 308–9.

Contrast and Representations in Syntax Edited by

BRONWYN M. BJORKMAN and DA N I E L C U R R I E HA L L

1

3 Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Bronwyn M. Bjorkman and Daniel Currie Hall 2020 © the chapters their several authors 2020 The moral rights of the authors have been asserted First Edition published in 2020 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2020935583 ISBN 978–0–19–881792–5 (hbk.) ISBN 978–0–19–881793–2 (pbk.) Printed and bound in Great Britain by Clays Ltd, Elcograf S.p.A. Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

OUP CORRECTED PROOF – FINAL, 28/8/2020, SPi

Contents General preface Acknowledgements List of figures and tables List of abbreviations The contributors

1. Contrast and representations in syntax: Introduction Bronwyn M. Bjorkman and Daniel Currie Hall

ix xi xiii xv xix

1

PA RT I . F E AT U R E S I N T H E I N F L E C T IO NA L SP I N E 2. A feature-geometric approach to verbal inflection in Onondaga Gabriela Alboiu and Michael Barrie 2.1 2.2 2.3 2.4

Introduction Mood, tense, and aspect in Onondaga Analysis Conclusions

3. Restricted and reversed aspectual contrasts Andrew Carnie and Sylvia L. R. Schreiner 3.1 3.2 3.3 3.4 3.5 3.6

Introduction Scottish Gaelic aspect morphology A single Asp head: Schreiner and Carnie (2016) The featural representation of aspect contrasts in Scottish Gaelic Predictions Conclusions

4. Sentience-based event structure: Evidence from Blackfoot Elizabeth Ritter 4.1 4.2 4.3 4.4 4.5 4.6

Sentience, argument structure, and event structure in Blackfoot Causers, agents, and intention Boundedness and event delimitation Sentience, boundedness, and entity delimitation Dynamicity, sentience, and eventiveness Conclusion

17 17 18 31 38

39 39 41 45 50 53 57

58 58 63 68 83 92 95

OUP CORRECTED PROOF – FINAL, 28/8/2020, SPi

vi contents

PA RT I I . C O N T R A ST I N T H E A R G UM E N T D OM A I N 5. Definite expression and degrees of definiteness Maria Kyriakaki 5.1 5.2 5.3 5.4 5.5 5.6

Introduction The expletive account The system of definiteness and its syntactic manifestations Definite determiners in English and Greek The three types of definite determiners Conclusions and directions for further research

6. Cross-linguistic contrasts in the structure of causatives in clausal nominalizations Martha McGinnis 6.1 6.2 6.3 6.4

Masdars are nominal Masdars are v-selecting Causees merged outside causative vP Conclusion

7. The Tłı  ˛cho  ˛ syntactic causative and non-nominal CPs Leslie Saxon 7.1 7.2 7.3 7.4

The wider context of this study Introduction and findings Methodology Introduction to Tłı  ˛cho  ˛ syntax, verb morphology, and the complement structure of ats’ele ‘cause’ 7.5 Properties of the caused event complement 7.6 Conclusion: The causative complement is an adverbial clause

99 99 103 109 121 133 135

138 141 145 172 178

179 179 180 182 183 193 210

PA RT I I I . A R C H I T E C T U R A L Q U E S T IO N S 8. Against some approaches to long-distance agreement without Agree Carson T. Schütze 8.1 8.2 8.3 8.4

Introduction Approaches to long-distance agreement Predicate inversion Long-distance object agreement, or How do you solve a problem like Icelandic? 8.5 Is Agree redundant? 8.6 Concluding remarks

9. Contrast in syntax and contrast in phonology: Same difference? Daniel Currie Hall 9.1 Why contrast matters

215 215 219 221 228 235 241

247 247

OUP CORRECTED PROOF – FINAL, 28/8/2020, SPi

contents vii 9.2 9.3 9.4 9.5 9.6

Contrast and features in phonology Inventories in phonology and syntax What do we need to distinguish? Consequences: Person and number in Mi’gmaq Conclusions

References Index

251 255 261 264 270

273 297

General preface The theoretical focus of this series is on the interfaces between subcomponents of the human grammatical system and the closely related area of the interfaces between the different subdisciplines of linguistics. The notion of ‘interface’ has become central in grammatical theory (for instance, in Chomsky’s Minimalist Program) and in linguistic practice: work on the interfaces between syntax and semantics, syntax and morphology, phonology and phonetics, etc. has led to a deeper understanding of particular linguistic phenomena and of the architecture of the linguistic component of the mind/brain. The series covers interfaces between core components of grammar, including syntax/morphology, syntax/semantics, syntax/phonology, syntax/pragmatics, morphology/phonology, phonology/phonetics, phonetics/speech processing, semantics/pragmatics, and intonation/discourse structure, as well as issues in the way that the systems of grammar involving these interface areas are acquired and deployed in use (including language acquisition, language dysfunction, and language processing). It demonstrates, we hope, that proper understandings of particular linguistic phenomena, languages, language groups, or inter-language variations all require reference to interfaces. The series is open to work by linguists of all theoretical persuasions and schools of thought. A main requirement is that authors should write so as to be understood by colleagues in related subfields of linguistics and by scholars in cognate disciplines. The role of features as atomic elements which the syntax operates on has become more central in syntactic theory over the past decade or so. This focus on the syntagmatic properties of features has meant that less attention has been paid to their paradigmatic role: their function in negotiating relationships of contrast within syntax. Feature geometries, or the underlying semantics that generates these, are one approach to feature contrast, but the necessity of these geometries and how they are represented in the syntax remain open questions.

x general preface The current volume highlights the central question of how the contrastive properties of features are connected to their role in simultaneously building meaning and form in larger syntactic structures. The chapters collected here bring new light to bear on this, drawing on a range of under-explored phenomena in a typologically diverse range of languages. David Adger Hagit Borer

Acknowledgements This volume came into being as the result of a workshop on contrast in syntax in April 2015, held on the occasion of Elizabeth Cowper’s retirement after nearly forty years at the University of Toronto. The workshop brought together linguists who had been supervised by Elizabeth at the undergraduate, graduate, or postdoctoral level, many of whom have contributed chapters to this volume. We take this opportunity to recognize those speakers from the conference who were unable to contribute to this volume: Lisa Cheng, Jila Ghomeshi, Päivi Koskinen, Julie Legate, Diane Massam, Kenji Oda, and Nick Pendar. Thanks especially to Diane, who co-organized the workshop (with Bronwyn Bjorkman) and who gave early advice on the development of this volume. Thanks as well to all the workshop participants, and to the Department of Linguistics at the University of Toronto, without whose support the workshop would not have been possible. We particularly acknowledge the assistance and expertise of Mary Hsu, the Departmental Officer, for her help at all stages of planning and logistics. Shortly after the workshop concluded, we embarked on the project of editing this volume. Beyond the individual contributors, we thank the entire editorial team at Oxford University Press, including the series editors David Adger and Hagit Borer, as well as the commissioning editors Julia Steer and Vicki Sunter for their guidance and assistance in bringing this volume together. We also gratefully acknowledge the contribution of the colleagues who reviewed this volume and its individual chapters, whose comments and suggestions have been very helpful. Finally, and most significantly, we gratefully acknowledge the tremendous contributions made by Elizabeth Cowper, without whom none of this work would have been possible. Of course, by providing the occasion for the workshop and inspiring its theme, she contributed the initial impetus that has led to this volume on contrast in syntax. More fundamentally, however, Elizabeth’s research over the course of her career has inspired many of the approaches to grammatical contrast developed in the chapters of this volume,

xii acknowledgements and many of the contributors can credit her teaching with helping to make them the linguists they are today. The editors both owe a great intellectual debt to Elizabeth, and will always regard her with great admiration as an excellent linguist, valued mentor, and dear friend. BMB & DCH

List of figures and tables Figures 5.1 Differing sets of contextually salient individuals

113

9.1 Language as an arch connecting forms to meanings

248

9.2 Hawaiian feature specifications (Morén 2006b) as a semilattice

254

9.3 Allophony versus syncretism

261

Tables 3.1 Summary of Scottish Gaelic aspect morphology

45

3.2 Regular Scottish Gaelic verbs (exemplified by fuirich ‘wait’)

55

3.3 The verb bi ‘be’

55

4.1 Blackfoot nouns

69

4.2 Blackfoot verb classes

69

4.3 Characteristics of D-languages and I-languages

74

4.4 Nominal inflectional suffixes

90

6.1 Masdar and verb stems in different TAM categories

141

7.1 Sample of forms of the Tłı  ˛cho  ˛ verb ats’ele

184

7.2 Categories of the Tłı  ˛cho  ˛ extended CP (to be revised in Table 7.3)

206

7.3 Revised categories of the Tłı  ˛cho  ˛ extended CP

206

9.1 Phonemic consonant inventory of Hawaiian

255

9.2 Suffixes marking number, gender, and case on nouns and indefinite adjectives in Old Church Slavonic

258

9.3 Another interpretation of the Hawaiian consonant inventory

259

9.4 Singular/plural vs. minimal/augmented

261

9.5 Mi’gmaq personal pronouns

265

List of abbreviations 1 2 3 acc accm add adv ag ai ai+o an aor appl area ben caus CEN cj cloc comp compr cond conj cont contr cop dat def dem dir dis DM du EPP erg et evid

first person second person third person accusative accompaniment final additive adverbial suffix agent animate intransitive animate intransitive verb with indefinite object animate aorist applicative areal prefix benefactive causative complex event nominalization conjunctive cislocative complementizer comparative conditional conjunction continuative contrastive copula dative definite demonstrative direct dislocative Distributed Morphology dual Extended Projection Principle ergative event time evidential

xvi list of abbreviations excl f fact foc fut gen ger hab ic ii in incl indef inf inst inv invis ipfv join LF LI loc m mod n neg nmlz nom obj obv OCS opt part past pat PCC perf PF pfv pl pns poss pres prog

exclusive feminine factual focus future genitive gerund habitual initial change inanimate intransitive inanimate inclusive indefinite infinitive instrumental inverse invisible imperfective joiner vowel Logical Form (interface level) lexical item locative masculine modalizer neuter negation nominalizer nominative object obviative Old Church Slavonic optative participant past tense patient Person–Case Constraint perfect Phonetic Form (interface level) perfective plural possessed noun suffix possessive present tense progressive

list of abbreviations xvii pron prosp prox prt PSM pssd ptcp punc purp pv qu re refl rel rep rest rl RMN RN rt sbjv SDA sg srfl stat ta TAM th ti tloc top tran ts UG vce VI vn

pronominal clitic prospective proximate particle Parallel Structures Model possessed participle punctual purpos(iv)e preverb question referring expression reflexive relative clause repetitive restricted relative root restrictive modification by nominals result nominalization reference time subjunctive Successive Division Algorithm singular semi-reflexive stative transitive animate tense/aspect/mood theme suffix transitive inanimate translocative topic transitive thematic suffix Universal Grammar voice vocabulary item verbal noun

The contributors Gabriela Alboiu is Associate Professor at York University, Toronto. She focuses primarily on clause structure, verb movement, left peripheries, structural Case, and obligatory control in Romanian. Her publications include book chapters (e.g. John Benjamins, De Gruyter Mouton, Springer, Elsevier, OUP) and journal articles (e.g. Syntax, Natural Language & Linguistic Theory, The Linguistic Review, Journal of Historical Syntax, Isogloss, Revue roumaine de linguistique, Canadian Journal of Linguistics). She is the author of The Features of Movement in Romanian (Bucharest University Press, 2002), and the co-author (with Virginia Hill) of Verb Movement and Clause Structure in Old Romanian (OUP, 2016). Michael Barrie is a syntactician at Sogang University in Seoul whose work also includes the syntax–semantics interface and the syntax–prosody interface. Much of his work is based on extensive fieldwork on Northern Iroquoian and on Algonquian languages. Outside of North America, Michael Barrie has also worked on Cantonese, Korean, and Romance. Bronwyn M. Bjorkman is Assistant Professor of Linguistics at Queen’s University in Kingston, Ontario. She completed her PhD in Linguistics at MIT in 2011, and prior to arriving at Queen’s was a Banting Postdoctoral Fellow at the University of Toronto. Her research focuses on the morphosyntax of tense and aspect, and in particular auxiliary verb constructions, as well as the representation and manipulation of features in syntax. This work has appeared in journals including Linguistic Inquiry, Glossa, and Natural Language & Linguistic Theory, and in several edited volumes. Andrew Carnie is Professor of Linguistics, Vice Provost for Graduate Education, and Dean of the Graduate College at the University of Arizona. His research centres on the syntax and phonology of the Celtic languages. His work in syntax focuses on phrase structure and Case. He has published thirteen books and numerous articles and chapters. His textbook Syntax: A Generative Introduction is in its third edition. He is currently pursuing an NSF documentary project developing an audiovisual corpus of Scottish Gaelic. Daniel Currie Hall is an Associate Professor and Coordinator of the Program in Linguistics at Saint Mary’s University in Halifax, Nova Scotia. Before taking up his current position, he completed a PhD at the University of Toronto in 2007 and worked as a postdoctoral researcher at the Meertens Instituut in Amsterdam. His research deals with features and contrasts in both phonology and morphosyntax, the latter primarily in a long-standing collaboration with Elizabeth Cowper, and has appeared in journals

xx the contributors such as Linguistic Variation, Glossa, Nordlyd, Lingue e linguaggio, and Phonology. He is an associate editor of the Canadian Journal of Linguistics. Maria Kyriakaki is Adjunct Faculty Professor at Deree, The American College of Greece. She completed her doctorate in 2011 at the University of Toronto. Her research focuses on syntax and its interface with morphology and semantics in languages such as English, German, Greek, and Ojibwe. Her published work appears in the Journal of Greek Linguistics and in proceedings of the Berkeley Linguistics Society and the Canadian Linguistic Association, and she has presented at various international conferences. Martha McGinnis is an Associate Professor at the University of Victoria. Her research explores the architecture of the grammar, focusing on argument structure, phi-features, syntactic movement, and binding. Her work has appeared in Linguistic Inquiry, Language, and Linguistic Variation Yearbook. Elizabeth Ritter is Professor of Linguistics in the School of Languages, Linguistics, Literatures and Cultures at the University of Calgary. Her research explores the morphosyntactic properties of functional categories and their contribution to semantic and pragmatic interpretation, with particular focus on nominal and clausal functors. Her current research, conducted in collaboration with Dr. Martina Wiltschko (ICREA), explores the functional superstructure in nominals that encodes information about the speech act that determines how we refer to individuals. Leslie Saxon (B.A., M.A. Toronto 1979; PhD University of California, San Diego 1986) retired in 2020 as a professor in the Department of Linguistics, University of Victoria, Victoria, BC, Canada. In addition to her syntactic research, she he is an editor of the Tłı  ˛cho  ˛ Yatıì Multimedia Dictionary online with Mary Siemens and others, and a collaborator in writing the Tłı  ˛cho  ˛ literacy manual, with Aliki Marinakis, Mary Koyina Richardson, and Mary Siemens. Dr. Saxon led her department at a time of major developments in Indigenous language revitalization at UVic, and served her community as the president of the Society for the Study of the Indigenous Languages of the Americas (2007) and of the Canadian Linguistic Association / Association canadienne de linguistique (2013–15). Sylvia L. R. Schreiner is an Assistant Professor in the Linguistics Program at George Mason University. Her research to date has focused on the morphosyntax and semantics of aspect, tense, and modality; nominal predication; the syntax–semantics interface; and Distributed Morphology. She has published articles in Lingua, Journal of Celtic Linguistics, and Morphology; and chapters in Formal Approaches to Celtic Linguistics and the Routledge Handbook of Syntax. She is currently principal investigator on an NSF-funded multi-year collaborative project on the documentation and maintenance of St. Lawrence Island Yupik. Carson T. Schütze is Professor of Linguistics at the University of California, Los Angeles, where he has taught since 1997. His research spans topics in syntax, morphology,

the contributors xxi first language acquisition, language processing, and linguistic methodology, often focusing on Germanic languages. His articles have appeared in Linguistic Inquiry, Lingua, Language, Syntax, The Linguistic Review, Journal of Child Language, Language Acquisition, Journal of Memory and Language, Memory & Cognition, and Zeitschrift für Sprachwissenschaft. His monograph The Empirical Base of Linguistics (1996, reprinted 2016) is often cited as a catalyst for the recent eruption of empirical and philosophical work on acceptability judgements.

1 Contrast and representations in syntax Introduction Bronwyn M. Bjorkman and Daniel Currie Hall

The focus of this volume is the dual role of features in syntax. On the one hand, features serve to define paradigmatic contrasts, delineating a language’s morphological and syntactic system of distinctions in grammatical person, number, tense, and so on. At the same time, features are also the basic building blocks of syntactic structure and the drivers of syntactic movement. This second role for features has received additional attention as the functional lexicon—the inventory of features that is active in a language, and the ways in which they are combined on heads—has increasingly been seen as the locus of parametric variation. Yet the connection between the paradigmatic and syntagmatic roles of features has not received comparable attention, though the identification of parameters with features has opened up new possibilities for exploring connections between the morphological system of a language and its syntax, and suggests a new role for featural contrast in syntactic theory. The role of features in syntax has changed a great deal in the development of generative syntactic theory. In the early days of the field, as set out in the Standard Theory of Chomsky (1965), differences between one language and another, or between one construction and another within a single language, were primarily derivational rather than representational: differences arose through the application of different phrase structure rules and different transformations, rather than from differences in the elements manipulated by syntax, or in the base structure of a given language. This is summarized in the following passage from Cowper (1992: 9): What was universal about this model was not any specific phrase structure rules or transformations, but rather the structure of the model and the various categories (NP, V, etc.) that it made use of. The rules in each of the components were seen as language-particular and therefore to be learned by the native speaker in the course of language acquisition. Bronwyn M. Bjorkman and Daniel Currie Hall, Contrast and representations in syntax: Introduction In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Bronwyn M. Bjorkman and Daniel Currie Hall. DOI: 10.1093/oso/9780198817925.003.0001

2 bronwyn m. bjorkman and daniel currie hall Beginning with the Principles and Parameters framework set out by Chomsky (1981), however, subsequent work in syntactic theory has sought both to constrain variability and to shift its locus. An early example of this can be seen in Lasnik and Saito (1992), who generalized transformations to a single operation, Move α (or, still more radically, Affect α), with crosslinguistic differences arising from parametric variation in constraints on the application of this operation, and in the filters on its output, as well as from the fact that different languages have different lexical items for it to operate on. The move towards grounding linguistic variation in representations, rather than in the operations that apply to those representations, has proceeded further within the Minimalist Program (Chomsky 1995 and subsequent work). This framework proposes to limit the syntactic component to a single structure-building operation Merge, combined with an abstract featurechecking operation Agree. All other constraints on language are attributed to interpretability requirements of the cognitive modules with which the grammar must communicate (the sensory-motor system and the conceptualintentional system). Assuming that these language-external cognitive modules are constant across speakers of different languages (excepting, e.g. the difference in modality between signed and vocal languages, which will entail different constraints from the sensory-motor system), this raises again the question of where cross-linguistic difference is located in grammar. Minimalist syntacticians have typically, if often implicitly, adopted the view of Borer (1984: 3) that variability resides in ‘the grammatical component that is idiosyncratic and learned in every language: the vocabulary and its properties’. This is the position that Baker (2008: 156) has called the Borer–Chomsky Conjecture (after Borer 1984 and Chomsky 1981, 1995). While the earlier Principles and Parameters framework conceived of parameters as UG-provided toggle switches with language-particular settings, the Borerian approach, as pursued in feature-driven Minimalist syntax, instead understands parameters in terms of the inventory of features active in a particular language (and perhaps also the structural positions in which those features occur). This both imposes constraints on what a possible parameter can be, and more clearly defines the task of parameter setting for learners. While some cartographic theories, such as that of Cinque and Rizzi (2008), posit an invariant universal ‘spine’ of functional projections, with differences between languages arising from whether particular features (and the heads in which they are located) are ‘active’ in a given language, much other work has argued that the sequence of functional projections is not fixed, and that

contrast and representations in syntax: introduction 3 the regularity that languages do exhibit in their functional sequences follows instead from requirements of semantic composition and from the ubiquity of functions such as clause typing and deictic anchoring. This latter view is described as ‘neoparametric’ by Cowper and Hall (2017), and has been pursued by a number of authors, including Bobaljik and Thráinsson (1998), Ramchand and Svenonius (2008, 2014), and Ritter and Wiltschko (2009). A central tenet of the neoparametric view is that variation in syntactic structure (i.e. in selection, checking, valuation, etc.) arises from the same formal features that define morphological paradigms and inflectional contrasts (with the proviso that structural variation may be further sensitive to metafeatures such as strength, interpretability, or ‘EPP’ properties that are invisible to morphological realization). Claims about a language’s functional syntax can—indeed, should—thus be grounded in the observation of its system of inflectional morphology. In a sense, the neoparametric approach applies to syntax a function that is analogous to Jakobson’s (1960: 358) famous definition of the poetic function, namely that it ‘projects the principle of equivalence from the axis of selection into the axis of combination’ (italics in the original). For Jakobson, this means that poetic language involves the syntagmatic combination of words that are paradigmatically connected by affinities of form, producing patterns such as metre, rhyme, and alliteration. In neoparametric syntax, the paradigmatic system of contrasts in the functional lexicon determines the formal materials (features) from which syntagmatic structures are built, and which are available for operations such as Agree. We might say, then, that the neoparametric approach projects the principle of contrast from the axis of selection into the axis of combination—i.e. from the lexicon into the syntax. Contrasts between features are perhaps rarely discussed in syntactic work, but they conceptually underlie much of our understanding of grammatical representations, across different modules of language. Contrastive features were first used in generative linguistics in the domain of phonology, following the structuralist tradition, and subsequently extended to syntax by Katz and Fodor (1963), Chomsky (1965, 1970), and Jackendoff (1977), among others.1 In phonology, features have traditionally been understood to divide the available phonetic space along articulatory or acoustic lines, for example dividing voiced sounds from voiceless ones, although some phonological

1 See Adger and Svenonius (2011) for further discussion of the history of features in generative syntax, and Dresher and Hall (forthcoming) on how structuralist conceptions of distinctive features were adapted into generative phonology.

4 bronwyn m. bjorkman and daniel currie hall features may be more abstract.2 In syntax, many features have consequences for semantic interpretation, morphological realization, or both. For example, tense and number features in many languages are both semantically interpretable and morphologically expressed, whereas Case and gender features frequently have morphological expression without necessarily having clearly identifiable semantic content. Other features may be wholly internal to syntax, such as Chomsky’s (1970) [±N, ±V] features defining a four-member set of lexical syntactic categories, or ‘EPP features’ driving overt syntactic movement (Chomsky 2000). This notion of grammatical contrast is distinct from whether a feature is interpretable or valued, a question that has received considerably more attention in Minimalist syntax. Chomsky (2000) proposes that syntactic operations are motivated in order to prevent unvalued—and thus uninterpretable—features from ‘crashing’ the derivation at LF. On this view, the syntactic activity of a feature is linked to whether it occurs in both valued and unvalued forms. Subsequent work has developed this idea further: Zeijlstra (2008) argues that evidence for uninterpretable instances of a feature is a precondition for its acquisition as a formal feature in a given language, while Pesetsky and Torrego (2007) argue for a further division between the interpretability of features and whether they are valued, allowing for the possibility that the value of a feature might be introduced in a different position than the feature is interpreted in. Though whether a feature is interpretable or valued is potentially orthogonal to whether it is grammatically contrastive, Wiltschko’s (2008) claim that only contrastive features are grammatically active implies a connection between these properties. Chapter 8 of this volume addresses the nature of the syntactic operation Agree, and is thus concerned with the representation of features in this sense. Beyond the simple idea of grammatical contrast, much recent work on feature systems has pursued the further idea that some cross-linguistic variation might be attributed not merely to which features are active in a language, but to how those features are related to one another in hierarchies or geometries. Within Minimalism, this has been pursued perhaps most notably in the domain of ϕ-features, accounting for both the organization of 2 Major class features such as [sonorant] and [consonantal] are difficult to define precisely in phonetic terms, and radically ‘substance-free’ approaches to phonology (e.g. Blaho 2008: 41) posit that phonological features in general need not have phonetic correlates at all.

contrast and representations in syntax: introduction 5 pronominal paradigms and the existence of person hierarchies in alignment and agreement, as in, for example, Harley (1994), Harley and Ritter (2002a,b), Cowper (2005a,b), Lochbihler (2008), Cowper and Hall (2012), Béjar and Rezac (2009), and McGinnis (2013). The hierarchical representation in (1), for example, has been proposed by Béjar and Rezac (2009: 42–3) to account for both the Person–Case Constraint (Bonet 1991) and the Direct–Inverse agreement system typical of Algonquian languages. The core idea is that a third-person argument is specified only for the root node [π]; second person is further specified with the dependent feature [participant]; and first person is further specified with [speaker]. A probe that encounters a goal that bears a subset of these features can continue probing in search of a goal that is more fully specified. (1)

π [participant] [speaker]

An alternative organization of person features for Algonquian is proposed by Lochbihler (2012), and Oxford (2017) recasts the hierarchy altogether, identifying grammatical person contrasts with different amounts of syntactic structure (projecting the paradigmatic contrasts into the syntagmatic domain). In Oxford’s account, first and second persons (speech act participants) take precedence for agreement not because they have more feature structure, but because their person features are not obscured by additional layers of syntactic projections. Some of Oxford’s representations are shown in (2). (2) a. Speech act participants (1 and 2) ϕP [+participant]

b. Animate impersonals (X) DefP Def

ϕP [−participant]

6 bronwyn m. bjorkman and daniel currie hall c. Proximate animate third persons (3) DP D

K

DefP Def

d. Obviative animate third persons (3′ ) KP

ϕP [−participant]

DP D

DefP Def

ϕP [−participant]

Hierarchical structures have also been advanced for case features, as in Bonet (1991, 1995), Béjar and Hall (2000), and Heap (2002), and for the organization of features on D, as in Cowper and Hall (2009) and Kyriakaki (2011). Beyond the nominal domain, some authors have pursued a featuregeometric approach to the structure of tense, aspect, and mood, accounting for the organization of inflectional systems in different languages, as in Cowper (1999, 2005a), Cowper and Hall (2005, 2017), and Clarke (2009, 2013). While this is not always recognized in syntactic work making use of feature hierarchies (particularly in the literature on hierarchies in ϕ-agreement), dependency relations expressed in feature geometries are a way of formalizing the notion of contrastive scope. In phonology, for example, Clements and Hume (1995) make the feature [anterior] a dependent of [coronal], indicating that anterior and posterior places of articulation contrast only within the coronal range. If [anterior] has scope only within the class of coronal segments, then [anterior] can identify dental and alveolar consonants as a natural class contrasting with retroflexes and postalveolars, but it cannot group dentals and alveolars with labials, or retroflexes and postalveolars with dorsals, as it does in feature systems in which it cross-classifies with [coronal] (e.g. Chomsky and Halle 1968). Similarly, in Cowper’s (1999; 2005a) geometry for English viewpoint aspect, Interval (which is spelled out by progressive -ing) is a dependent of Event, indicating that there is a contrast between perfective (momentaneous) and imperfective events, but that no such contrast exists among states: (3) a. State:

b. Perfective event: ∅

Event

c. Imperfective event: Event | Interval

contrast and representations in syntax: introduction 7 Thus a sentence like (4a) describes two perfective events (most naturally understood as occurring in sequence), while in (4b) the imperfective playing event spans an interval (most naturally understood as containing the perfective arrival); the state denoted by we were happy in (4c), on the other hand, is neither perfective nor imperfective (and can be felicitously construed as starting either before or immediately upon the arrival event), and adding overt imperfective morphology to it, as in (4d), degrades its grammaticality.3 (4) a. b. c. d.

When Christina arrived, we played a freilach. When Christina arrived, we were playing a freilach. When Christina arrived, we were happy. ∗? When Christina arrived, we were being happy.

It is worthwhile to note that the use of feature geometries to encode the scope of contrasts is logically independent of their use to structure feature representations on a single head. The dependency among person features proposed by Béjar and Rezac (2009) shown in (1) is a claim about the representation of person features. What it means, in this model, for a head to bear a first person feature is that it bears the full feature complex in (1); the feature [speaker] does not occur independently of this hierarchy. The feature dependencies proposed in Cowper’s (2005a) analysis of tense, mood, and aspect, by contrast, do not require the entailed and entailing features to occur on the same head. As Cowper and Hall (2013: 129) point out, a semantically dependent feature may appear on a separate syntactic head that selects a complement specified for the entailed feature. For example, Cowper and Hall (2012: 35), building on work by Cheng and Sybesma (1999), propose that in Mandarin the classification feature cl is semantically dependent on the individuation feature #, and that cl heads a projection that selects an NP complement that is specified with #. In this view, then, feature geometries primarily represent semantic entailments between features that occur within a domain of projections, but not necessarily on a single syntactic head within that domain.⁴ Like Oxford’s (2017) account of Algonquian person systems, this approach allows paradigmatic contrasts to be expressed syntagmatically.

3 Adding progressive -ing to a stative predicate can also force an eventive reading; compare She is sick with She is being sick (Cowper 2005a: 14). To the extent that (4d) is acceptable at all, it requires being happy to be construed as an activity rather than a state. ⁴ Although most of the dependency relations in Cowper’s (2005a) geometry correspond to semantic entailments, the presence of the purely syntactic feature finite (which has no semantically interpretable content) as both an entailed and an entailing feature is an exception.

8 bronwyn m. bjorkman and daniel currie hall Harbour (2011) and Harbour and Elsholtz (2012) question the explanatory usefulness of feature geometries, arguing that few if any of the dependency relations that have been posited in the literature can be independently motivated—and that for those that can be, a geometric treatment is unnecessary. In their view, any feature geometry must be either stipulative or redundant. However, as noted by Cowper and Hall (2014b), geometric arrangements of features remain a useful expository device for diagramming semantic entailments between features and for describing language-specific systems of contrasts. The neoparametric perspective in general predicts that languages should be able to vary in the formal features they use (Chomsky 2000: 100), but that the combinations into which these features can enter are constrained by the requirements of interpretability. Geometries are merely one of a range of theoretical tools for representing differences in the relative scope of features. Similar insights are also captured by contrastive hierarchies in phonology (Dresher 2009), and by various other morphosyntactic formalisms. In nanosyntax (Starke 2009; Caha 2009; Pantcheva 2011, among others), feature dependencies are syntactically articulated, with dependent features selecting projections of the features they entail as complements. For Harbour (2013, 2016), who formalizes features as functions applying in a particular order, the variation in order of application between one language and another generates different systems of semantic and morphological contrasts. While the identification of the featural lexicon as the locus of syntactic variation is a relatively recent development within generative syntax, the grammatical relevance of paradigmatic contrast has been recognized at least since Saussure (1916), who observed that the semantic value of any grammatical entity depends on the oppositions it enters into. For example: Le protogermanique n’a pas de forme propre pour le futur; quand on dit qu’il le rend par le présent, on s’exprime improprement, car la valeur d’un présent n’est pas la même en germanique que dans les langues pourvues d’un futur à côté du présent (Saussure 1916: 161).

Proto-Germanic has no special form for the future; to say that the future is expressed by the present is wrong, for the value of the present is not the same in Germanic as in languages that have a future along with the present (Saussure 1959: 117).

Expressing this in (somewhat naïve) binary features, we might say that while the present tense in a language like French is [−past, −future], the so-called present tense in Proto-Germanic (and many of its descendants) is

contrast and representations in syntax: introduction 9 merely [−past], and thus compatible with future reference. In phonology, there is a long tradition, stemming from Saussure (1916) and Trubetzkoy (1939), of using the presence or absence of features to encode the presence or absence of contrasts.⁵ In morphosyntax, distinctive features and the possibility of underspecifying them have come to prominence more recently, particularly within the framework of Distributed Morphology (DM; Halle and Marantz 1993). Underspecification of vocabulary items (VIs) is the standard DM mechanism for generating syncretisms in morphological paradigms; the geometric organization of features, a device adapted from phonology, has been used to delineate the underlying systems of distinctions that those VIs spell out (Harley 1994; Harley and Ritter 2002b; Cowper 2005a; Cowper and Hall 2014b). Similarly, feature representations are subject to Impoverishment, a mechanism of deleting syntactic features in certain configurations to account for other instances of syncretism (similar to mechanisms of delinking in phonological theory). The second person paradigm in English, which lacks a contrast between singular and plural, is standardly explained as the application of an Impoverishment rule that deletes any number features that occur on a head that also bears a second person feature. In both these cases— underspecification of VIs and Impoverishment—the underspecification assumed in DM is realizational rather than representational; it is located in the morphology, in the rules that associate abstract syntactic information with phonological content, rather than in the syntactic representation proper. Syntactic representations themselves remain fully specified in DM, however, though work in this framework often argues that certain features do not cooccur in syntax for principled reasons. Combining contrastive specification of distinctive morphosyntactic features with the neoparametric approach to functional projections offers new insights into connections between structure and meaning. For example, Cowper and Hall (2017) argue that the historical development of the English modals from verbs into auxiliaries (Lightfoot 1979) involved the addition of a feature Modality to the English inflectional system. As verbs, the (pre-)modal vocabulary items realized lexical roots that moved to T from the verb phrase; as inflectional elements, the Present-Day English modal VIs spell out a feature of T itself. The feature Modality contributes to a clause the information that the state or event described therein is a possibility or a necessity, rather than something simply asserted to be true; this is the same

⁵ See Dresher (2009) for a history of contrast in phonology, and Chapter 9 of this volume for a more detailed discussion of the parallels between phonological and syntactic contrast.

10 bronwyn m. bjorkman and daniel currie hall feature that characterizes the future and conditional tenses in languages such as French and Spanish.⁶ If this feature was indeed absent from the Old English system of tense and mood contrasts, then its absence from any individual clause was non-contrastive. Like the Proto-Germanic ‘present’ tense alluded to by Saussure, the Old English ‘present’ tense was not contrastively nonmodal or non-futurate, and was used with a much wider range of futurate meanings than are possible in Modern English (Cowper et al. 2015). In this way, the contrastive neoparametric approach to formal features offers insight not only into the synchronic organization of grammatical systems, but also into how they might change: the rise of new surface patterns in a language will force learners to postulate contrastive features that underly the change.

Organization of the volume The inspiration for this volume came from the Contrast in Syntax workshop held at the University of Toronto in April 2015. The workshop was held in honour of Elizabeth Cowper, who had recently retired after nearly forty years at the university, and the invited speakers included many of her past students and advisees. The individual chapters of this volume, in most cases revised and expanded from talks and posters presented at the workshop, represent a diversity of topics, perspectives, and concerns. They are united, however, by an interest in morphosyntactic representations, and in the formal encoding of syntactic contrasts. By syntactic contrasts we mean the systems of grammatical oppositions that exist within individual languages and, in a broader sense, how these systems can and do differ from one language to another. This theme is tied to a question that has been fundamental throughout the development of generative syntactic theory: What is universal in syntax (and in language more generally), and what is variable? The chapters of this volume address in varying ways, and from different perspectives, the distribution of syntactic features and their twofold role in defining paradigmatic contrasts and shaping syntactic structures. The first section of the volume addresses the role of contrastive features in defining the inflectional spine of the clause. Tense in particular has played a central role throughout modern generative syntax as the head of the clause

⁶ See Cowper (2005a), in which this feature is referred to as Irrealis; it was renamed Modality by Cowper and Hall (2007).

contrast and representations in syntax: introduction 11 and as a locus of argument licensing. In mainstream work tense was originally located in the generalized head Infl0 , and later distinguished in a dedicated head T0 (distinct from Agr0 , the locus of clausal agreement in some theories). Subsequent syntactic and semantic work has suggested that clausal inflection involves yet finer-grained structure, distinguishing further heads responsible for modality and aspect, among other categories. In recent work more radical variation has been proposed, suggesting that the content of individual heads— the features that define their semantic and syntactic contrasts—may vary, not only in the organization of features (Cowper 2005a), but in the substantive content they describe (e.g. a feature ±Coincidence may relate times in a tense-based language, but person or location in other languages: Ritter and Wiltschko 2014; Wiltschko 2014). The chapters in the first section vary in the degree to which they assume languages differ in their inventory of formal features, but all are concerned with features relating to tense, modality, and aspect. Chapter 2, by Alboiu and Barrie, proposes a feature geometry for the Onondaga inflectional system. They adapt proposals by Cowper and Hall (1999) and Cowper (2005a), but argue that in Onondaga it is modality, rather than tense, that anchors clauses to the utterance context, just as Ritter and Wiltschko (2014) propose that location and person anchor clauses in Halkomelem and Blackfoot, respectively (see also Ritter, this volume). Likewise in the inflectional domain of the clause, Carnie and Schreiner, in Chapter 3, analyse temporal aspectual contrasts in Scottish Gaelic, extending Cowper’s (2005a) geometry of Infl in another direction. Cowper proposes that two languages with the same set of inflectional features can nonetheless differ in whether those features are distributed across multiple heads, or bundled within a head. When bundled, features can stand in a dependancy relation: one feature may be possible only in the presence of another. Carnie and Schreiner propose that the system of prospective aspect in Scottish Gaelic aspectual system involves two further distinctive dependents of the past-tense feature Precedence: in particular, they argue that prospective perfects in Scottish Gaelic are not Irrealis (i.e. they are not modal), but instead involve a feature Reversed that results in a temporal relationship that is not backward-looking (as in perfects) but instead forward-looking. The second feature they propose is Restricted, which results in both the restricted (recent) perfect and the restricted prospective. In Chapter 4, Ritter looks at the system of lexical aspect in Blackfoot, arguing that event structure in this Algonquian language is based on contrasts in animacy and sentience, unlike the contrasts of dynamicity and boundedness

12 bronwyn m. bjorkman and daniel currie hall found in many other languages. Rather than distinguishing between states, telic events, and atelic events, Blackfoot differentiates eventualities according to the number of sentient arguments: inert eventualities have no sentient arguments; active and reactive eventualities have one; and interactive eventualities involve two. This gives rise to the traditional division of Algonquian verbs into four categories based on transitivity and the animacy of participants. At an abstract structural level, the system of contrasts in Blackfoot shows parallels with other aspectual systems, but the different substantive content of the formal features involved causes these contrasts to manifest themselves in different ways—for example, in causative constructions, it is the presence of multiple sentient arguments that gives rise to structural complexity, rather than the presence of multiple events (as in, for example, the approach to complex causatives in Ritter and Rosen 1993). Ritter’s chapter offers a bridge to the second section of the volume, which investigates the role of contrast in the argument structural domain of the clause. This domain has seen considerable expansion in recent years, beginning with the introduction of v0 or Voice0 (Chomsky 1995; Harley 1995; ?), and more recently with a wider proliferation of argument-introducing heads (e.g. the applicative heads of Pylkkänen 2008). The chapters in this section probe the representation of such heads, and the features that define them. Kyriakaki, in Chapter 5, examines cross-linguistic variation in the encoding of definiteness in the nominal domain. Focusing initially on a contrast between English and Greek in the representation of DPs, Kyriakaki argues that definiteness can be semantically and syntactically divided into two distinct components, familiarity and uniqueness, and extends this analysis to definiteness systems in a variety of other languages. Chapter 6, by McGinnis, examines Georgian gerundive nominalizations (masdars), arguing that they demonstrate that it is possible for causees to be merged outside of the causative v head, unlike causative structures in other languages such as English (Bjorkman and Cowper 2013). This expands the scope of potential variation in causative constructions, accounting for differences in their structure across languages. Saxon, in Chapter 7, also looks at causatives and cross-linguistic variation, presenting an analysis of Tłı  ˛cho  ˛ ats’ele (‘cause’), a verb that takes one DP and one adverbial complement, the latter of which can be an embedded clause headed by an adverbial complementizer. Like the previous chapter, this widens the typological picture of how we expect argument structures to be encoded in natural language.

contrast and representations in syntax: introduction 13 The final section of the volume addresses the representation and manipulation of features from a broader architectural perspective, with chapters discussing both how features are manipulated in the narrow syntax, and the yet more general question of how our theory of possible features and possible feature systems applies across different modules of grammar. In Chapter 8, Schütze turns his attention to the derivational mechanism of feature-checking and the structures within which it applies, contributing to current debates on this topic. He argues that feature-checking cannot always be accomplished in local (specifier–head) configurations, and that the (potentially long-distance) operation Agree cannot be eliminated from our model of syntax. In particular, he shows that a strictly local theory of agreement fails to account for several kinds of contrast in well-formedness, drawing primarily on examples from English and Icelandic. Chapter 9, by Hall, compares the applications of distinctive features and underspecification in syntax versus phonology, looking at the similarities and differences in how these concepts apply in the two domains. This chapter builds on earlier work by Cowper and Hall (2014c) to propose a new way of looking at the morphosyntactic ‘inventories’ in which features may be contrastive. Hall ultimately suggests that different dimensions of contrast are independent of one another: the dependencies expressed in a contrastive hierarchy cannot involve features of fundamentally different types, such as person and number. This makes predictions for the typology of different types of paradigmatic systems; Hall discusses a number of examples from nominal paradigms. From the collection as a whole emerges a view of syntax in which representations in general, and features in particular, are taken seriously as having identifiable morphological, syntactic, and semantic consequences, and in which each novel piece of machinery added to the grammar is something that must be argued for, not merely assumed. If the features that define morphological systems of contrast are the same ones that form and transform syntactic structures, this serious attentiveness to representations is essential to understanding their explanatory potential.

OUP CORRECTED PROOF – FINAL, 25/8/2020, SPi

PART I

FEATU R E S I N T H E I NF L E C T IONA L SPI N E

2 A feature-geometric approach to verbal inflection in Onondaga Gabriela Alboiu and Michael Barrie

2.1 Introduction Onondaga is a Northern Iroquoian language spoken in southern Ontario, Canada, and in New York State in the U.S. (Barrie 2015). The Onondagas make up part of the Iroquois Confederacy, or Haudenosaunee, which also includes the Senecas, Mohawks, Cayugas, Oneidas, and the Tuscaroras (who joined after the original formation of the Confederacy). The origins of Onondaga society can be traced back to at least the eleventh century in northern New York (Bradley 2005). The Onondagas, along with the other members of the Confederacy, continue to live in the same areas they have occupied for about a millennium. Iroquoian inflectional verbal morphology is well documented in the descriptive literature (Lounsbury 1949, 1953; Chafe 1961; Michelson and Doxtator 2002), but has received less attention from a generative perspective (Baker and Travis 1997, 1998). The goal of this chapter is to provide a formal account of the Onondaga verbal inflectional system, a language where tense is not the crucial ingredient of INFL, but rather where the realis versus irrealis distinction, alongside aspectual distinctions seem paramount. In order to do so, we employ a feature geometric account following work by Cowper (2005a), Cowper and Hall (1999), Kyriakaki (2006), and Slavin (2008). However, prior to proposing the theoretical account, we first discuss the empirical properties of the Tense, Aspect, Mood (TAM) system in Onondaga and show that this language appears to be, at least in some cases, tenseless. The analysis has interesting consequences for syntactic theory more generally, especially under views which take finiteness to be equivalent to tense (e.g. Carnie 2012, inter alia).

Gabriela Alboiu and Michael Barrie, A feature-geometric approach to verbal inflection in Onondaga In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Gabriela Alboiu and Michael Barrie. DOI: 10.1093/oso/9780198817925.003.0002

18 gabriela alboiu and michael barrie

2.2 Mood, tense, and aspect in Onondaga The template in (1) (based on Lounsbury 1949, 1953) lists the order of morphemes on the verbal complex for all Northern Iroquoian languages, starting at the left edge. The boldfaced elements are required in all verbal constructions. Crucially, the one constant of the Northern Iroquoian, and implicitly Onondaga, Infl domain is the presence of aspect and not of tense. (1) a. b. c. d. e. f. g. h.

pre-pronominal prefixes (mood, location, other semantic properties) pronominal prefixes (subject and object agreement) reflexive or semi-reflexive incorporated noun verbal root derivational suffixes (benefactive, causative, inchoative, etc.) aspect suffixes expanded aspect suffixes

There are four aspects in Onondaga. The first is the ‘habitual’ aspect (hab). This aspect is used to refer to an event that takes place repeatedly or on an ongoing basis. The second is the ‘punctual’ aspect (punc). This aspect refers to an entire event in its completeness. When used on a verb with a past tense reading, the event is described as ‘over and done with’. The punc cannot describe an event that is interrupted or incomplete. The third aspect is the ‘stative’ (stat) aspect, which refers to an event that is ongoing or incomplete or, if it occurs in the past tense, an event that has some bearing on the present (like the English present perfect). Finally, there is the ‘purposive’ aspect (purp), which refers to imminent action, and usually implies intent or volition on the part of the subject. Active verbs can appear with any of the first three aspects. Motion verbs can appear with any of all four aspects. Lexically specified stative verbs can only appear with the stative aspect (Woodbury 2003: 38). Since of the four aspect types, only the hab, punc, and stat aspects are used productively, we focus our discussion on them.1 1 In addition, it is unclear whether ‘purposive aspect’ is indeed aspectual in the same way as the hab, punc, and stat aspects. Not only is intent/volition implied, but the purp can be used together with the stat, as in (i), where it appears to correspond to a higher head in the Infl domain. (i) sahohde  ˛dyo  ˛ha:dyeʔ swaʔ- hoahte  ˛tyo  ˛ -h -atye -ʔ rep- fact- 3sg.m.pat- travel -stat -cont -purp ‘He is going home.’ / ‘He is on his way home.’

verbal inflection in onondaga 19

2.2.1 Simplex aspect constructions Simplex aspect constructions refer to verb forms that contain one of three aspect markers introduced above: habitual (hab), stative (stat), and punctual (punc), standard terms from the Iroquoianist literature. Complex aspect constructions, discussed below, contain one of these three aspect markers plus an expanded aspect marker. In this section, we discuss each type of aspect in turn. Iroquoian descriptive literature gives the following descriptions for the hab. Lounsbury (1953: 85) states that ‘Forms in the serial [hab] aspect represent actions which take place at repeated points in time; for the most part these correspond to the simple “present tense” form in English’, and Woodbury (2003: 30) mentions that ‘[The hab forms] are usually glossed in the simple present, e.g. I sew, or in the progressive, e.g., I am sewing, or both’. Our data are consistent with the standard description of the habitual as indicating an ongoing activity or state of affairs, as illustrated in (2). As in (2d), it often appears in nominalized forms and in professions. (2) a. haye  ˛thwas haye  ˛thw -as 3sg.m.ag- plant -hab ‘He plants / he is planting.’ b. khenowe  ˛khwa’ ne’ Meri khenowe  ˛hkw -ha’ ne’ Meri. 1sg.ag:3sg.f.pat- love -hab ne Mary ‘I love Mary.’ c. heno  ˛hsho  ˛nye  ˛nik heno  ˛hsh -o  ˛ny -e  ˛ni -k 1.sg.ag:3.sg.m.pat- house -make -ben -hab ‘I’m making a house for him.’ d.  ˛de o  ˛iho  ˛:nye  ˛nik  ˛o ate  ˛- Rih ˛ny o -e  ˛ni -k 3.sg.f.pat- srfl- matter- make -ben -hab ‘She’s a teacher.’ / ‘She is teaching.’ In sum, the habitual covers both states and imperfective events, so is durative and unbounded, hence an instance of imperfective viewpoint (in the sense of Comrie 1976 etc.).

20 gabriela alboiu and michael barrie With respect to the stative aspect, there are different descriptions in the Iroquoianist literature. Lounsbury (1953: 85) claims that ‘Forms in the perfective [stative] aspect represent states; some of these are the results of actions’, while Woodbury (2003: 30) mentions that ‘In English they [statives] are usually glossed with the perfect, e.g., I have sewn it, or, depending on the meaning of the base, with the progressive, e.g., I am sewing (Chafe 1980)’.2 Our findings align with these observations (see also Abbott 2006 for Oneida). For lexically stative roots, typically translated as adjectival non-verbal predication in English, statives are semantically a state, as in (3): (3)

hoda’gaide’ hoata’kait -e’ 3sg.m.pat- be.healthy -stat ‘He is healthy/feels good.’

Otherwise, use of stat signals a state that is the result of some previous action, and in this case it is translated with the present perfect in English, as shown in (4): (4) a. (onihsih) dyagone  ˛haye  ˛thwih (onihsih) tya- kone  ˛haye  ˛thw -ih (long.time.ago ) cloc- 3sg.f.pat- corn-join- plant -stat ‘She has planted corn (a long time ago).’ b. hohdo  ˛h hoahto  ˛ -h 3.sg.m.pat- disappear -stat ‘He has disappeared.’ The data in (5)–(6) suggest that the progressive reading of stat is available only when hab cannot have a progressive reading; compare (5a) with (5b) and (6a) with (6b): (5) a. hoda’ks hoita’k -s 3sg.m.pat- sleep -hab ‘He is sleepy.’ (not ‘He is sleeping.’) 2 stat aspect forces patient marking on subjects of intransitives of active verbs; we do not focus on this issue here.

verbal inflection in onondaga 21 b. hoda’wih hoita’w -ih 3sg.m.pat- sleep -stat ‘He is asleep/sleeping.’ (6) a. ekho  ˛nyaha’ ekhw-o  ˛ny -aha’ 3sg.f.ag- foodmake -hab ‘She is a cook.’/‘She cooks (habitually).’ b. hokho  ˛nih hokhw-o  ˛ny -ih 3sg.m.pat- foodmake -stat ‘He is cooking.’ In sum, use of the stative aspect may: (i) be resultative, in which case it also contains perfectivity (hence its alternative name), in that there is some previous finished event which results in some state; or (ii) indicate a lexically ‘stative’ root; or (iii) idiosyncratically replace the habitual. Crucially, in all its meanings, the stative instantiates imperfective aspect, and its prototypical use is to encode result state, since the last two uses are lexically determined and thus not part of the core syntactic derivation. Following Lounsbury (1953: 85), ‘Forms in the punctual aspect represent actions which take place at some particular point in time; this point in time may be past, future, or indefinite, depending on the pre-pronominal […] prefix’. For Woodbury (2003: 30), on the other hand, ‘The punctual aspect requires one of three modal prefixes, the factual waʔ-, the future  ˛-, e or the optative a·−. The factual is glossed with the simple English past, e.g., she sewed it; the future is glossed with the English future, e.g., I will sew it; and the optative is glossed variously in English as I may sew, I should sew, I would sew, etc.’ . (7) illustrates the punctual with some of our own elicited data. (7) a. wa’haye  ˛thwa’ wa’- haye  ˛thw fact- 3sg.m.ag- plant ‘He planted it.’ b. a’sehde’ wa’khyado  ˛’ a’sehte’ wa’- kyesterday fact- 1sg.ag‘I wrote it yesterday.’

-a’ -punc

hyato  ˛ -’ write -punc

22 gabriela alboiu and michael barrie c.  ˛haye e  ˛thwa’  ˛e haye  ˛thw -a’ fut- 3sg.m.ag- plant -punc ‘He will plant (it).’ d.  ˛gek e  ˛e kek -∅ fut- 1sg.ag- eat -punc ‘I will eat it.’ e. a:gek akek -∅ opt- 1sg.ag- eat -punc ‘I would eat it.’ The one thing to note is that all verbal forms with the punctual aspect have a pre-pronominal prefix in (7). While these pre-pronominal prefixes are ruled out with the habitual and the stative aspects in simplex constructions, they are obligatory with the punctual. This is a well-known fact of Iroquoian (Lounsbury 1949, 1953; Chafe 1960a,b,c,d, 1961; Froman et al. 2002; Michelson and Doxtator 2002; Woodbury 2003; Abbott 2006). Examples are shown in (8). (8) a.

ahse:deh *(wa’)hayenawa’s ahseteh wa’- hayenawa’s -∅ yesterday fact- 3sg.m.ag- help -punc ‘He helped yesterday.’ b. ∗ wa’hayenawa’seh wa’- hayenawa’s -eh fact- 3sg.m.ag- help -stat Intended: ‘He has/had helped.’ c.  ˛’hayenawa’s/ e *hayenawa’s ∗ (e  ˛’)- hayenawa’s -∅ fut- 3sg.m.ag- help -punc ‘He will help.’ d. ∗  ˛’hayenawa’sek/eh e  ˛’e hayenawa’s -ek/eh fut- 3sg.m.ag- help -hab/stat Intended: ‘He will have helped/will be helping.’

verbal inflection in onondaga 23 e.



wa’haye  ˛thwas wa’- haye  ˛thw -as fact- 3sg.m.ag- plant -hab Intended: ‘He was planting (and has finished).’ f. ∗ haye  ˛thwa’ haye  ˛thw -a’ 3sg.m.ag- plant -punc Intended: ‘He planted/will plant it.’

Compare (8e) to (7a) and (8f) to (7c). These properties are summarized in (9): (9) a. Punctual aspect requires a mood pre-pronominal prefix (future, factual, or optative). b. Habitual and stative aspect prohibit a mood pre-pronominal prefix. At this point, the following questions arise: i. What is the role of the punctual aspect and how does it differ from the stative and habitual? ii. Why does the punctual require mood pre-pronominal prefixes, while the stative and habitual cannot combine with these? iii. We refer to these pre-pronominal prefixes as ‘mood’ above in line with traditional descriptions in the Iroquoianist literature, but do these prefixes actually instantiate tense or mood/modality? For (i), the preliminary answer is that, unlike the habitual and the stative, which we have seen are instances of imperfective aspect, punctual denotes a situation viewed in its entirety, so, instantiates perfective viewpoint aspect. As for (ii), the data in (10) provide some insight: (10) a.

wa’eye  ˛ne  ˛da’nha’ gokho  ˛nih wa’eye  ˛ne  ˛ta’ fact- 3sg.f.ag- finish -ih make -stat ‘She finished cooking.’

-nha’ -punc

ko3sg.f.pat-

khw-o  ˛ny food-

24 gabriela alboiu and michael barrie b.



wa’eye  ˛ne  ˛da’nha’ wa’ekho  ˛nya’ wa’eye  ˛ne  ˛ta’ -nha’ kwa’- ekhwfact- 3sg.f.ag- finish -punc fact- 3.sg.f.pat- food ˛ny o -a’ make -punc Intended: ‘She finished cooking.’

(10) shows that, while the punctual appears with ‘finish’, it cannot appear on the verb referring to cooking. The stative is required instead. While the event of finishing is punctual, the event of making food cannot be momentary, so must be durative. Since, following Smith (1991), non-durative/punctual situations present a ‘closed structure which appears at a point in time’, we can assume they need some manner of temporal anchoring.3 In the next section, we argue that in Onondaga, temporal anchoring is realized via modality in conjunction with (im)perfectivity, rather than via tense.⁴

2.2.2 Modality as temporal anchoring There are three pre-pronominal prefixes in Onondaga, all of which are illustrated in (7). At first glance the data in (7) might be taken to indicate that the factual morpheme expresses past tense, the future morpheme future tense, and the optative morpheme irrealis mood. Since both tense and mood seem to be involved, one could think of Onondaga as having a high Infl head in which tense and mood features merge (i.e. similar to English T, which hosts both tense and mood, typically in complementary distribution). It is perhaps unsurprising, then, that labelling varies between ‘tense’ (see Lounsbury 1953 for Oneida) and ‘mood’ (Baker and Travis 1997 for Mohawk; Chafe 1960a,b,c,d, 1961 for Seneca; Foster, 1985, 1986). The facts discussed here suggest that a modal account is more likely to be on the right track given the division between realis (factual) and irrealis (optative and future). 3 A reviewer asks whether the punctual can appear on ‘cook’ at all. It can, but then it refers to a completed event of cooking something, unlike in (10), where completion is encoded by ‘finish’ rather than by ‘cook’. See (i) for an example. (i) gwaso  ˛hde’ wa’hakho  ˛nya’ gwaso  ˛hde’ wa’-ha-khw-o  ˛ny-a’ last.night fact-ag-food-make-punc ‘Last night he cooked.’ ⁴ See Jaszczolt (2009) on conceptualizations of temporal distinctions in terms of possibility and necessity.

verbal inflection in onondaga 25 A more careful investigation reveals that the factual modal prefix is not a past tense marker, but rather indicates that the speaker knows that the event has happened for a fact. Since we are normally only sure about events that happened in the past, it typically has a past tense reading. However, it is not always the case that anteriority to the moment of speech is denoted. Also possible are a root/dynamic modal value, as in (11), a performative value, as in (12), or a factual/indubitable present tense value, as in (13). (11) a. wa’sgé  ˛ khe  ˛h ne’ jíhah? wa’- ske  ˛ -’ khe  ˛h ne’ jihah fact- 2sg.ag- see -punc qu ne dog ‘Did you see the dog?’ / ‘Can you/are you able to see the dog?’ b. wa’hge  ˛’ ne’ sano  ˛hsa’ wa’- kke  ˛ -’ ne’ sano  ˛hsa’ fact- sg.ag- see -punc ne your.house ‘I saw your house.’ / ‘I’m able to/can see your house.’ (right now) (12)

wa’gnihsé  ˛:no  ˛h ne’ shagoksde  ˛’tshä:’ de  ˛hse’ ho  ˛wahksde  ˛’tshä:’ wa’- knihse  ˛n-  ˛o -h … fact- 1sg.ag:2du.pat- name- give -punc ‘I name you husband and wife.’

(13)

gae  ˛ no  ˛:h hwa’he’ kae  ˛ no  ˛:h hwa’- hae -’ which place tloc- fact- 3sg.m.ag- go/be -punc ‘Where is he going?’

In addition, forms with the factual mood prefix and the punctual aspect suffix are ruled out with negation, as seen in (14). (14) a.

hiya de’agokho  ˛nih hiya te’- akokhwno neg- 3sg.f.pat- food‘She didn’t cook.’ ∗ b. hiya de’wa’ekho  ˛nya’ hiya te’- wa’- eno neg- fact- 3sg.f.ag-

 ˛ny o -ih make -stat

khw-  ˛ny o -a’ food- make -punc

26 gabriela alboiu and michael barrie An obvious question is whether it is the punctual or the factual that is incompatible with the negative. However, the example in (15) shows that complementarity of distribution is between the factual modal prefix and negation and not between negation and punctual aspect.⁵ (15)

hiya tha:yekho  ˛nya’ hiya tha:yekhw -o  ˛ny -a’ no contr.opt- 3sg.f.ag- food -make -punc ‘She might not cook.’

So, while temporal distinctions are clearly available in Onondaga (see also Baker and Travis 1997 for Mohawk), the factual does not represent an instance of past tense, but rather denotes a necessarily true proposition, that is, one which is true in any circumstance and cannot be false. The default past tense interpretation of the factual and punctual combination follows in a straightforward manner once we acknowledge that Onondaga does not mark tense in these constructions. Following Smith (1991), perfective events are by default interpreted as past, while imperfective events are interpreted as present in ‘tenseless’ languages. The perfectivity of punctual aspect will therefore automatically trigger a past tense interpretation, while the imperfectivity of the stative and the habitual will render the situation in the present. We now need to account for the future and the optative morphemes. Foster (1985) argues that the future is more about probability and calls it a ‘predictive’ mood, while the optative expresses what the speaker thinks is desirable or possible, and so has a more indefinite flavour (but see Baker and Travis 1997 for some counterarguments). Neither is factive, and so they can each combine with negation. Our findings support the claim that both the future and the optative denote irrealis modality and seem to support the notion that future is predictive. Consider the data in (16). The future and optative refer to unrealized events and differ in the degree of certainty. The factual refers to realized or actual events only.⁶

⁵ The hab, on a par with the stat in (14a) and punc in (15), can also occur with negation, as in (i); this is unsurprising. (i) hiya de’ekho  ˛nyaha’ hiya te’ekhw-  ˛ny o -aha’ no neg- 3sg.f.ag- food- make -hab ‘She never cooks.’ ⁶ The symbol R in the morpheme ‘want’ in (16b) represents ‘ancient r’, on which see Barrie (2015).

verbal inflection in onondaga 27 (16) a.  ˛sa:hahde e  ˛:dya’  ˛e sa- haahte  ˛tyo  ˛ -a’ fut- rep- 3sg.m.ag- leave -punc ‘He will be going home.’ (‘predictive’ irrealis) b. he:he’  ˛sahahde o  ˛:dya’ haeR -he’  ˛o sa- haahte  ˛tyo  ˛ -a’ 3sg.m.ag- want -hab opt- rep- 3sg.m.ag- travel -punc ‘he want’ + ‘he might/may/could go home.’ ‘He is planning on going home.’ (‘indefinite’ irrealis) c. sahohde  ˛dyo  ˛há:dye’ sa- wa’- hoahte  ˛tyo  ˛ -h -atye -’ rep- fact- 3sg.m.pat- travel -stat -cont -purp ‘He is going home.’ / ‘He is on his way home.’ (indubitable) In sum, while sometimes labelled as tense, these pre-pronominal prefixes are best viewed as marking a realis–irrealis or factual–non-factual distinction, and so they are modal elements rather than tense elements. This is reinforced by the linear order of the verbal morphology. Assuming that suffixation is a result of head movement, we can assume that the prefixes represent functional heads in the clausal spine that are higher than the position reached by verb movement. Thus, the modal prefixes are higher than the aspect and extended aspect suffixes. Also, modality is typically associated with the CP layer rather than with the IP layer (D’Alessandro and Ledgeway 2010), while aspect is associated with the IP layer. Thus, at least in simplex aspectual constructions, temporal anchoring in Onondaga is realized either through the interaction of mood/modality and aspect (for perfectivity) or through the default present tense interpretation for imperfective aspect. Ultimately, what seems to be crucial for finiteness is the Comp–Infl relationship, which is often argued to guarantee temporal independence (e.g. Hill and Alboiu 2016).

2.2.3 Complex aspect constructions In addition to the simplex aspect constructions, Onondaga exhibits constructions where the habitual and stative can be augmented by suffixes referred to as the ‘habitual past’ (hab.past) and the ‘stative past’ (stat.past) in the Iroquoianist literature. These are traditionally described as past tense morphemes. Semantically, the data reveal that the habitual past is an imperfective past, as

28 gabriela alboiu and michael barrie illustrated in (17), while the stative past is either a perfective past or a past perfect progressive, as in (18).⁷ (17) a. ho  ˛wakho  ˛nye  ˛níhgwa’ ho  ˛wakhw-  ˛ny o -e  ˛ni -k -kwa’ 3sg.f.ag:3sg.m.pat- food- make -ben -hab -hab.past ‘She used to cook for him.’ b. khenowe  ˛khwahgwa’ khenowe  ˛hkw -ha’ -kwa’ 1sg.ag:3sg.f.pat- love -hab -hab.past ‘I used to love her.’ c. thaye  ˛thwasgwa’ thaye  ˛thw -as -kwa’ cloc- 3sg.m.ag- plant -hab -hab.past ‘He used to plant.’ d. hada’gwa’ hat -a’ -kwa’ 3sg.m.ag- stand -stat -hab.past ‘He was standing.’ (18) a. ho  ˛wakho  ˛nye  ˛nihna’ ho  ˛wakhw-  ˛ny o -e  ˛ni -h -na’ 3sg.f.ag:3sg.m.pat- food- make -ben -stat -stat.past ‘She had cooked for him.’ b. agida’wihna’ wakita’w -ih -na’ 1.sg.pat- sleep -stat -stat.past ‘I had been sleeping.’

⁷ The stat.past can appear alongside the purposive too, as long as past perfectivity is implied, as in (i): (i) hadowäthe’na’ haatowät -h -e’ -na’ 3sg.m.ag- hunt -dis -purp -stat.past ‘He had intended to hunt.’ (Woodbury 2003: 38) The dislocative (dis) is a future suffix meaning ‘going to’ and showing intention here (or movement, elsewhere); it is used with purposive aspect.

verbal inflection in onondaga 29 c. shagoye  ˛thwe  ˛ní:hna’ shakoye  ˛thw -e  ˛ni -ih -na’ 3.sg.m.ag:3.pat- plant -ben -stat -stat.past ‘He had planted it for her.’ As was shown in §2.2.1, the habitual and stative cannot appear with modal prefixes. We suggest that this is due to their semantics. Imperfectives, as ongoing eventualities, do not focus on the event as a whole unit, so they cannot denote a necessarily actualized proposition. Recall that the factual indicates an actualized event. Consequently, both fact . . . stat and fact … hab sequences are semantically impossible, and thus ruled out in the morphosyntax. Unlike with the factive, imperfectives can appear with the future and the optative, provided their default realis connotation has been ‘annihilated’. This can occur in one of two ways: either by adding the modalizer (mod), as in (19), or by using a complex aspect construction, i.e. expanding the habitual and the stative to the habitual past and the stative past, respectively, as in (20)–(22). The examples in (19) further show that, unlike with the complex aspect constructions seen in (17)–(18), modal prefixes are obligatorily required when the modalizer is present, as in (19e), though of course the factual prefix is impossible, as in (19b). (19) a.

b.

c.

d.

e.

dye  ˛gwa’ akhenowe  ˛khwak tye  ˛kwa’ akhenowe  ˛hkw -ha’ -ek maybe opt- 1sg.ag:3sg.f.pat- love -hab -mod ‘If I had loved her.’ ∗ wa’khenowe  ˛khwak wa’- khenowe  ˛hkw -ha’ -ek fact- 1sg.ag:3sg.f.pat- love -hab -mod  ˛gaye e  ˛thwik  ˛e kaye  ˛thw -ih -ek fut- 3sg.f.ag- plant -stat -mod ‘It will have been planted.’  ˛haye e  ˛thwasek  ˛e haye  ˛thw -as -ek fut- 3sg.m.ag- plant -hab -mod ‘He will plant.’ ∗ hokho  ˛nihek hokhw-  ˛ny o -ih -ek 3sg.m.pat- food- make -stat -mod

30 gabriela alboiu and michael barrie f.

(20)

aho  ˛wakho  ˛nye  ˛nihek aho  ˛wakhw-  ˛ny o -e  ˛ni -h -ek opt- 3sg.f.ag:3sg.m.pat- food- make -ben -stat -mod ‘She would be cooking for him.’

Future habitual  ˛hesninowe e  ˛khwahgwa’  ˛e hesninowe  ˛hkw -ha’ -kwa’ fut- 2du.ag:3.sg.m.pat- love -hab -hab.past ‘You two will be acting kindly towards him.’ (Woodbury 2003: 716)

(21)

(22)

Future stative a.  ˛wagegihna’ e  ˛e wakek -ih fut- 1.sg.pat- eat -stat ‘I might have eaten it.’ b.  ˛wagegik e  ˛e wakek -ih fut- 1sg.pat- eat -stat ‘I will have eaten it.’

-na’ -stat.past

-ek -mod

Complex aspect with opt modal prefix a. aho  ˛wakho  ˛nye  ˛nihna’ aho  ˛wakhw-  ˛ny o -e  ˛ni -h -na’ opt- 3sg.f.ag:3sg.m.pat- food- make -ben -stat -stat.past ‘She should have cooked for him.’ b. ahesninowe  ˛khwahgwa’ ahesninowe  ˛khw -ha’ -kwa’ opt- 2du.ag:3.sg.m.pat- love -hab -hab.past ‘You two should have loved him.’

2.2.4 Taking stock To summarize, the following generalizations hold with regard to the distribution of modal and aspectual affixes. The bare habitual and stative cannot appear with modal prefixes, while the punctual requires presence of a modal prefix. A modalizer stacked onto the habitual and stative enables these imperfective

verbal inflection in onondaga 31 aspects to occur with the future and the optative. On the other hand, the habitual past and the stative past morphemes allow either the future or the optative to be present. Lastly, only the punctual may appear with the factual. We are now in a position to provide a full summary: (23) a. no expanded aspect suffix aspect suffix mood prefix punctual obligatory (factual, future, or optative) habitual none stative none b. modalizer expanded suffix aspect suffix mood prefix punctual: incompatible with modalizer habitual obligatory (future or optative) stative obligatory (future or optative) c. habitual past or stative past expanded suffix aspect suffix mood prefix punctual: incompatible with habitual past and stative past habitual optional (future or optative) stative optional (future or optative) We provide an analysis of these facts in the next section.

2.3 Analysis We have argued that temporal anchoring in Onondaga is realized through modality in conjunction with (im)perfectivity, rather than through tense as is typical in Indo-European languages.⁸ Since in this language tense does not seem to be the crucial ingredient of Infl, we propose a feature-geometric account, following Cowper (2005a), in order to better capture the Onondaga verbal inflectional system.

2.3.1 Feature geometries Cowper (2005a) proposes the schema in (24) as the maximal dependency structure for Infl. In (24), each label represents a verbal inflectional feature ⁸ See in particular Wiltschko (2014) for an in-depth discussion of cross-linguistic variation in anchoring the event.

32 gabriela alboiu and michael barrie made available by UG and connected by entailment bottom-up, such that each node entails the presence of those above it in the tree structure. A certain feature is present in a given language if, and only if, there is a binary opposition for that property (i.e. a contrast). It is also assumed that the absence of a feature triggers a default interpretation of the node dominating that feature. [Proposition], [Finite], temporal and personal [Deixis], and [Irrealis] are mood features; [Entirety] and [Precedence] are tense features; and [Interval] and [Event] are Aspect features. (24)

Infl Proposition

Precedence

Event

Finite

Entirety

Interval

T-Deixis P-Deixis Irrealis For Cowper (2005a), following Cowper and Hall (1999), the components of (24) are monovalent features with syntactic or semantic content. [Event] distinguishes events (where this feature is present) from states (where this feature is absent). The feature [Interval], a dependent of [Event], encodes imperfective viewpoint aspect. A perfective event is a temporal point with no internal structure. These are the two aspectual features of Infl. The tense features are [Precedence], whose presence signifies that at least one moment of the event or state precedes the temporal anchor of the clause, and [Entirety], present when all moments of the event or state precede the temporal anchor. The other features in (24) determine the clausal properties traditionally associated with mood. A proposition is taken to be a cognitive manifestation of a state or event. The feature [Proposition] takes an event or state and ‘transforms it into its cognitive manifestation by linking it to a consciousness’ (Hall 2001: 20). A clause without [Proposition] denotes a bare state or event, as in We saw [the light flash] (Hall 2001: 19). Following Cowper (2005a), [Finite],

verbal inflection in onondaga 33 [Deixis], and [Irrealis] are dependents of [Proposition].⁹ The feature [Finite] is purely syntactic; it licenses nominative Case and subject phi-agreement. The feature [Deixis] sets the temporal and/or personal anchor of the clause to the deictic centre of the utterance/discourse. The temporal ([T-Deixis]) and personal ([P-Deixis]) indexing can be kept apart (as in Spanish) or bundled together (as in English). Simplifying somewhat, the feature [T-Deixis] sets the temporal anchor to the Utterance Time/Time of Speech, while the feature [P-Deixis] connects the proposition denoted by the clause to the set of propositions believed by the speaker (i.e. the consciousness at the deictic centre). Lastly, the feature [Irrealis] establishes a marked relation between the proposition and the consciousness at the deictic centre.

2.3.2 Features of Infl in Onondaga In this section, we discuss the featural properties of the various Onondaga Infl morphemes introduced in §2.2 with a view to sketching a feature-geometric analysis for this Iroquoian language. We also assume Distributed Morphology (?), in which the syntactic component manipulates only features, and Vocabulary Items are inserted cyclically, post Spell-Out, with a Vocabulary Item specified for a larger set of features winning out over a less specified Vocabulary Item at the insertion site. We begin with a brief discussion of Onondaga clausal structure before moving on to a detailed discussion of the tense, aspect, and mood morphemes. We assume the following clausal structure based on the verbal template:

(25)

Mood – Agr – incorporated noun – verb – ben/caus – Asp – expanded aspect CP > MoodP > TP > AspP > vP > VP

For this chapter we do not concern ourselves with the location of agreement. Given the morpheme order in (25), we assume head movement takes place as high as T (expanded aspect, see discussion below). This leaves the mood morpheme as a prefix. We also leave to future research exactly how prefixes morphologically fuse with the verbal complex. We now move on to the details of the aspect morphemes.

⁹ Though see Kyriakaki (2006) who proposes a different ordering of [Proposition] and [Finite] for Greek.

34 gabriela alboiu and michael barrie The first thing to note is that Onondaga disallows non-finite clauses, and so even non-propositional events are finite. This is illustrated in (26), which shows that the complement clause to a verb of sensory perception (i.e. a bare event) comes equipped with ϕ-features and nominative Case. Note that (26) means that the subject (‘he’) directly visually perceived the event of a man sitting there. (26)

hwa’hoge  ˛’ he  ˛gweh hatgoda’ hwa’hoke  ˛ -’ he  ˛kweh hatkod tloc- fact- 3.sg.m.pat- see -punc man 3sg.m.ag- sit -a’ -stat ‘He saw a man sitting there.’ (Woodbury 2003: 1328)

Thus, since there is no contrast, we assume that the feature [Finite] is absent in this language (rather than being a dependent of [Proposition] as in (24)). The habitual aspect appears with both states and (imperfective) events. It therefore cannot be specified for the feature [Event]. In addition, within events it covers both homogenous events (‘she’s teaching’) and non-homogenous events (‘she teaches’) in the sense of Slavin (2008). Therefore, it also cannot be specified for the feature [Interval]. We suggest the habitual has no specific Infl features at all, but rather spells out v. In other words, it is the default aspect marker. Thus, the Asp, Tense, and Mood heads will not have any features. The absence of [Precedence] is shown in (27), as temporal adverbs are not compatible with the habitual morpheme. Accordingly, all three TAM domains will automatically have default interpretations (i.e. realis).1⁰ (27)



ahse:deh hayenawa’sek ahseteh hayenawa’s -ek yesterday 3.sg.m.ag- help -hab

Recall that stative aspect prototypically indicates a result state. Following Ramchand (2008), result events instantiate an R head and are deeply embedded within the VP—crucially, lower than viewpoint aspect. Consequently, R competes for insertion with v, and wins because it is more specified. Just as

1⁰ Note that Cowper (2005a) argues the present tense in Spanish, which essentially covers the same imperfective meanings as the Onondaga hab, has the feature P-Deixis. This cannot be the case for Onondaga, as the hab is not in complementary distribution with irrealis (i.e. fut and opt). Assuming Distributed Morphology, the more specified fut/opt should block insertion of hab, contrary to fact.

verbal inflection in onondaga 35 with habitual aspect, this imperfective also has no specific Infl features. Given the perfective viewpoint aspect realized by the punctual, we assume it has the feature [Event], whose default reading is moment (Cowper 2005a). After head movement of v to Asp [Event], punc will be inserted in preference over hab as it is more specified. Moving beyond aspectual features, we now address tense, namely the stative past and the habitual past. Recall that in the habitual past at least one point of the event must precede the utterance time, and in the stative past the entire event must precede the utterance time. We suggest, then, that the habitual past, which denotes imperfective in the past, is specified for the feature [Precedence], while the stative past is specified for the feature [Entirety]. In both these cases, since tense is deictic, the feature [T-Deixis] must be present in Mood, but we assume it piggybacks on [P-Deixis] as it can never occur on its own. Regarding the mood node, we observed that all modal prefixes instantiate [P-Deixis] in Onondaga, as follows. The factual corresponds to a marked version of realis in that certainty is implied, while both the optative and the future instantiate irrealis, with future being more specific in that it is predictive. Consequently, we propose that a specified feature [P-Deixis] is always marked, either by the feature [Certainty], entailing [Realis]; by [Predictive], entailing [Irrealis]; or simply by [Irrealis]. This yields a split as in (28), rather than an entailment relationship between realis and irrealis. (28)

P-Deixis opt

Irrealis

fut Predictive

Realis Certainty

∅ (default, with stat and hab) fact

Since this feature split is not possible under a single node, we suggest that perhaps the feature [Irrealis] is realized in C, while the feature [Realis] is realized in Infl. Some support for a syntactic asymmetry between these two features is found when looking at interaction of mood markers with other prepronominal prefixes. Consider (29), repeated from (16): (29) a.  ˛sa:hahde e  ˛:dya’  ˛e sa- haahte  ˛tyo  ˛ -a’ fut- rep- 3sg.m.ag- leave -punc ‘He will be going home.’ (‘predictive’ irrealis)

36 gabriela alboiu and michael barrie b. sahohde  ˛dyo  ˛há:dye’ sa- wa’- hoahte  ˛tyo  ˛ -h -atye -’ rep- fact- 3sg.m.pat- travel -stat -cont -purp ‘He is going home.’ / ‘He is on his way home.’ (indubitable) Notice that the repetitive morpheme sa- follows the irrealis future mood marker in (29a) but, crucially, precedes the realis factual mood marker in (29b), suggesting that irrealis is higher than realis/factual.11 In conclusion, with respect to the Mood domain, we assume the following: (i) there is no Finite node (as mentioned above); (ii) there is [P-Deixis] with both realis and irrealis mood; and (iii) there may be [T-Deixis] complementing [P-Deixis]. Crucially, the deixis entailment is reversed from what Cowper (2005a) observes for languages like English and Spanish, hence the ‘tenseless’ flavour of Onondaga. That is, since [T-Deixis] is not always present in finite clauses, tense is not required. This contrasts with English, where finite clauses are always specified for tense. Recall further that the punctual, which is perfective, requires an anchor (traditionally a temporal anchor). We argued that in Onondaga this anchoring is realized by modality, which necessitates [P-Deixis]. Thus, perfectives are situated with respect to the speaker’s observations rather than to the utterance time. On the other hand, the stative and habitual, as imperfectives, do not need this anchoring, and so do not necessitate [P-Deixis]. However, recall that there are situations in which the stative and habitual appear with irrealis modal prefixes in one of two contexts: (i) with the habitual past and with the stative past; and (ii) with the modalizer /-ek/. In situations involving (i), we assume that the feature [Precedence] removes the situation from the ‘now’ so that [T-Deixis] becomes available. In situations involving the modalizer, we propose that this morpheme encodes the feature [General Tense]. By this we mean that it encodes the existence of a set of points related to the situation. We posit this feature on analogy with General Number (Corbett 2000; Rullmann and You 2006). General Number does not distinguish between singular and plural. Crucially, it is not the absence of number, but rather an expression of number that simply is not specified for plurality or singularity. See the works cited for more information. We propose that General Tense works much the same 11 Note that Irrealis Mood is higher that Repetitive (Aspect) in Cinque’s (1999) hierarchy of projections too.

verbal inflection in onondaga 37 way, indicating the existence of tense without specifying a value. We further assume that feature [General Tense] is inserted in the derivation to remove the situation from the default, hence its obligatory need for specific features of [T-Deixis], namely the [Irrealis] feature or the feature [Predictive].12 In sum, this yields the correspondences between overt morphology and abstract morphosyntactic structure shown in (30) and the feature geometry for the Onondaga Comp-Infl shown in (31). By positing [T-Deixis] as the most marked feature of [Irrealis], (31) correctly captures the empirical observation that no tense feature specifications are possible with the factual mood (i.e. [Certainty]) in Onondaga. (30) a. b. c. d. e. f.

hab: stat: punc: hab.past: stat.past: fact:

g. opt: h. fut: i. mod: (31)

no Infl features, v no Infl features, R [Event] [Precedence] ([General Tense]) [Entirety] ([Precedence], [General Tense]) [Certainty/Indubitable] ([Realis], [P-Deixis], Proposition]) [Irrealis] ([P-Deixis], [Proposition]) [Predictive] ([Irrealis], [P-Deixis], [Proposition]) [General Tense] Onondaga Comp-Infl

Proposition

General Tense

Event Precedence

P-Deixis Irrealis

Realis

Predictive

Certainty/Indubitable

Entirety

T-Deixis 12 One difference between General Number and General Tense is that General Number can remain unspecified for singular or plural. General Tense, as we have described it here for Onondaga, must be specified by [T-Deixis]. We leave this fact to future research.

38 gabriela alboiu and michael barrie

2.4 Conclusions In this chapter, following Cowper (2005a) and Cowper and Hall (1999), we have laid out the foundation for a feature-geometric analysis of the mood/tense/aspect features in Onondaga. As with other non–Indo-European languages (see also Slavin 2008; Clarke 2009), a feature-geometric approach is better suited to handling Infl systems that are not ‘tense-centric’. Clearly, however, Onondaga is not really ‘tenseless’, but rather it does not employ tense as an obligatory feature of tensed clauses. Rather, aspect take the role of activating Infl. This is the result of [T-deixis] being located at the bottom of the feature geometry. It is to this reversed hierarchy that we attribute the perceived lack of tense in Onondaga.

Acknowledgements We are grateful to our Onondaga speakers, Nora Carrier and Gloria Williams, for sharing their expertise with us. All uncited Onondaga data in this chapter are from original fieldwork. We also wish to thank Carrie Dyck, Roronhiakehte Deer, Karin Michelson, and Marianne Mithun for insightful discussions on Iroquoian grammar in general as well as the participants of the Contrast Workshop at the University of Toronto for fruitful discussion. All errors are our own. This research was partially supported by a Sogang Research Grant of 2011 and 2012 and a SSHRC standard research grant (410–2011–2417).

3 Restricted and reversed aspectual contrasts Andrew Carnie and Sylvia L. R. Schreiner

3.1 Introduction Scottish Gaelic is well known for the range and diversity of the contrasts it expresses in its aspectual and tense systems (Ramchand 1993; Reed 2012). As such it stands as an important proving ground for theories that represent morphosyntactic contrasts in tense, mood, and aspect. One such theory (Cowper 1998, 1999, 2003, 2005a, 2010; Cowper and Hall 2007) identifies aspectual and tense contrasts using a feature geometry paired with a Distributed Morphology system of lexical insertion (for a contrasting view of feature relationships that does not use geometric dependencies see Harbour 2014a). Examining closely the contrasts expressed by the English auxiliary system (and those of other languages), Cowper and her colleagues propose a feature geometry. This feature structure is realized as lexicalized tense and aspect morphology. This geometry explains the range of possible semantic contrasts in the English system. Cowper (2005a: 12) proposes that the dependencies expressed in the following feature geometry articulate the set of contrasts expressed in the English inflectional system: (1)

Infl Proposition Finite/Deixis

Precedence

Event Interval

Irrealis The feature event distinguishes clauses expressing events from those that express states. The dependent feature interval then further distinguishes Andrew Carnie and Sylvia L. R. Schreiner, Restricted and reversed aspectual contrasts In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Andrew Carnie and Sylvia L. R. Schreiner. DOI: 10.1093/oso/9780198817925.003.0003

40 andrew carnie and sylvia l. r. schreiner among event types. The presence of this feature represents the marked imperfective viewpoint aspect; its absence indicates the unmarked perfective viewpoint aspect. In English, the feature interval is realized by the -ing suffix on verbs. The precedence feature is used to indicate that there is a temporal ordering relation of one time prior to another. This might be the relation between reference time and speech time, as in the English past tense, or it might be between event time and reference time, as found in English perfect aspect. The realization of this feature in the morphology will depend upon the other features it co-occurs with. proposition is the feature that marks the cognitive manifestation of an event. finite/deixis is a bundled feature which both checks nominative case in the syntax and expresses the speaker-centred identification of the discourse anchor. The combination of finite/deixis and precedence will trigger the insertion of a past tense marker, such as -ed or the auxiliary was. When linked with a second temporal anchor (such as a reference time), precedence will trigger the insertion of a past participle morpheme like -en. Finally, the irrealis feature is used to mark certain kinds of modal force, including may and will. The ‘future’ tense in English is taken to be a modal structure using irrealis. In this chapter, we argue for two modifications to this feature geometry, both of which involve new dependencies upon the precedence feature. The first modification is relatively uncomplicated; it simply allows for a restriction on the time between the two intervals ordered by the precedence feature. We call this feature restricted. We will show that this feature is used to encode the temporally restricted near or ‘after’ perfect and prospective aspects found in Scottish Gaelic. The second modification is more controversial, as it is explicitly rejected by Cowper (2003) and Cowper and Hall (2007).1 We claim that certain kinds of forward-looking (i.e. future-like) meanings in Scottish Gaelic are not expressed by modal features like irrealis, but instead by a feature reversed. This feature indicates that the temporal ordering of event times managed by precedence are reversed, so that they are forward-looking rather than backward-looking. In particular, we claim that this feature is used to mark unrestricted and restricted prospectives in the grammar. The proposed new dependencies are articulated in (2). (2)

precedence restricted

reversed

1 More precisely, they reject the idea that the future is merely the opposite ordering of event time and discourse anchor from that found in the past.

restricted and reversed aspectual contrasts 41 This chapter is structured as follows. In §3.2, we review the rich morphological system of Scottish Gaelic for expressing aspectual contrasts. We explore five aspectual particles that mark progressive, unrestricted perfect, restricted perfect, unrestricted prospective, and restricted prospective aspects. In §3.3, we take a brief detour to summarize the claim made by us in previous work that the seemingly multimorphemic sequence a’ dol aL has in fact been grammaticalized into a single Asp head. In §3.4, we show how the two revisions to Cowper’s feature system map onto Scottish Gaelic aspect morphology. Finally, in §3.5, we look at some predictions made by our proposals. We show that the reversed feature correctly predicts semantic differences in the way the future is represented. We also suggest that the restricted feature can be extended from the aspectual domain to the tense system of many languages.

3.2 Scottish Gaelic aspect morphology A quick review of the morphology and syntax marking the aspectual system in Scottish Gaelic will help set the scene for understanding the patterns we discuss in the rest of the chapter. See MacAulay (1996) for some foundational work that shaped our characterizations here. In (3) we see the progressive.2 This is marked with the auxiliary verb bi in one of its tensed forms with the verbal noun (vn) form of the main verb. An aspectual particle immediately precedes the vn. It takes the form ag in front of vowels and a’ in front of consonants. (3)

Progressive Tha mi ag ithe cèic. be.pres I prog eat.vn cake ‘I am eating cake.’

Scottish Gaelic has two perfect aspects. The first is an unrestricted perfect, roughly equivalent to the English have perfect. This aspect is also marked using the auxiliary bi with a verbal noun form of the main verb. The aspectual particle used is air. In this form the word order is slightly different, with the object preceding the vn but following the aspect marker, as seen in (4):3 2 All the Scottish Gaelic data in this chapter comes from our original fieldwork. Some of it may appear in other papers by us, including but not limited to Reed (2012) and Schreiner and Carnie (2016). 3 The transitive particle a, shown in parentheses in (4), is normally omitted before vowel-initial words, but is present in these same constructions when the verbal noun begins with a consonant.

42 andrew carnie and sylvia l. r. schreiner (4)

Unrestricted Perfect Tha mi air an cèic (a) ithe. be.pres I perf the cake tran eat.vn ‘I have eaten cake.’

The second perfect (5) is sometimes called the ‘after perfect’. It is called this because the particle that marks the aspect is homophonous with the preposition as dèidh/an dèidh, which means ‘after’ in other contexts. We will refer to this as the ‘restricted perfect’. This aspect has the same syntactic structure as the unrestricted perfect but differs from it in meaning. The time span between the event time and the reference time must be short or restricted (see Reed 2012 for a more formal and detailed analysis of the semantics of the restricted perfect. See also Ó Sé 2004). (5)

Restricted Perfect (After-perfect) Tha mi as dèidh an cèic (a) ithe. be.pres I rest.perf the cake tran eat.vn ‘I have just eaten the cake.’ (lit., ‘I’m after eating the cake.’)

In addition to the two backward-looking perfect aspects, Scottish Gaelic has two forward-looking aspects. The first is an unrestricted prospective, marked with a’ dol aL (literally ‘going to’; we will return later to what we believe the correct analysis of its morphology is). The particle is followed by the vn and then the object. This structure is used to indicate that an event will occur after a reference point. Effectively it is the semantic mirror image of a perfect. For more details on the semantics of this construction see Reed (2012) and Schreiner and Carnie (2016). (6)

Unrestricted Prospective Tha mi a’ dol aL dh’ithe cèic. be.pres I prosp eat.vn cake ‘I am going to eat cake.’

Schreiner and Carnie claim that evidence for the aspectual nature of the prospective comes from the behaviour of adverbials, among other things. The separation of event and reference times, regardless of what order they

restricted and reversed aspectual contrasts 43 are in, means that each can be separately referenced. One property of the perfect is that it can occur with adverbials in two positions, yielding distinct interpretations: Clause-final adverbials allow both event time (et) and reference time (rt) readings (7a), while clause-initial adverbials allow only a reference time reading (7b) (see also Hornstein 1993; McCoard 1978; Comrie 1976, 1985; Klein 1992; Michaelis 1994; Reed 2012). (7) a. Bha Calum air a’ bhùth fhàgail aig be.past Calum perf the.f.sg shop leave.vn at meadhan-latha. mid-day ‘Calum had left the store at noon.’ (rt or et reading—either ‘at noon’ refers to the time under discussion, and Calum has already left by then, or his leaving was at noon.) b. Aig meadhan-latha, bha Calum air a’ bhùth at mid-day be.past Calum perf the.f.sg shop fhàgail (mar tha). leave.vn (already) ‘At noon, Calum had left the store (already).’ (rt reading) The same set of interpretations holds for the prospective. With a clausefinal adverbial, either reading is available (8a), but the clause-initial adverbial is restricted to rt readings (8b). (8) a. Bha Calum a’ dol aL phòsadh Màiri aig meadhan-latha. be.past Calum prosp marry.vn Màiri at mid-day ‘Calum was going to marry Màiri at noon.’ (et or rt interpretation available—i.e. either the wedding will take place at noon, or we don’t know when the wedding will be, but we’re reporting about the state of affairs at noon.) b. Aig meadhan-latha bha Calum a’ dol aL phòsadh Màiri, at mid-day be.past Calum prosp marry.vn Màiri ach aig uair gabh e an t-eagal. but at hour take.past 3sg.m the fear ‘At noon, Calum was going to [i.e. planning to] marry Màiri, but at one he got scared.’ (rt interpretation only)

44 andrew carnie and sylvia l. r. schreiner We also have evidence that the prospective is an aspect rather than a future tense. It can co-occur with any tense auxiliary (9a), and it cannot co-occur with other aspectual particles without a second auxiliary (9b).⁴ Bha/tha/bithidh Calum a’ dol aL phòsadh Màiri. be.past/be.pres/be.fut Calum prosp marry.vn Màiri ‘Calum was/is/will be going to marry Màiri.’ b. ∗ Tha Iain air a’ dol aL sgrìobhadh. be.pres Iain perf prosp write.vn ∗ ‘[Iain has going to write.]’

(9) a.

So, following Reed (2012), we conclude that the a’ dol aL construction marks a true prospective aspect. The other forward-looking aspect parallels the restricted perfect. It is limited to contexts where the event time and the reference time are perceived to be close together (again, see Reed 2012 for a formalization of this restriction). This restricted (or near) prospective has the same syntax as air and as dèidh, but is marked with the particle gu (10):⁵ (10)

Restricted Prospective Tha mi gu an cèic (a) ithe be.pres I rest.prosp the cake tran eat.vn ‘I’m about to eat the cake.’

Table 3.1 summarizes the distinctions discussed.

⁴ A number of examples in this section are drawn from Reed [Schreiner] (2012). ⁵ Adger (1994) identifies gu as a marker of ‘E>R’ (essentially a prospectivity marker). Reed (2012) demonstrates that this morpheme is in fact limited to contextually restricted time spans, as seen in the following examples: (i) # Tha Iain gu taigh a thogail ann an còig bliadhna. be.pres Iain gu house tran build.vn in five year #‘Iain is about to build a house five years from now.’ (ii) Tha e gu ceumnachadh ann am mìos. be.pres 3sg.m gu pace/graduate.vn in month ‘He is about to graduate in a month.’ (iii) Tha Iain gu bhith trang a-màireach. be.pres Iain gu be.vn busy tomorrow ‘Iain is about to be/going to be busy tomorrow.’

restricted and reversed aspectual contrasts 45 Table 3.1 Summary of Scottish Gaelic aspect morphology Particle

Interpretation

Structure

a’ air as dèidh gu a’ dol aL

progressive unrestricted perfect restricted perfect unrestricted prospective restricted prospective

a’ + vn (+ obj) air (+ obj + tran) + vn as dèidh (+ obj + tran) + vn gu (+ obj + tran) + vn a’ dol aL + vn (+ obj)

3.3 A single Asp head: Schreiner and Carnie (2016) In order to understand how the feature structures proposed in this chapter are integrated into the syntax of the language, we need to explain our assumptions about the underlying structure of the particles that mark aspect. In particular, we need to demonstrate that what appears on the surface to be a complex multiword structure for the restricted prospective (a’ dol aL ), has in fact been grammaticalized as a single aspectual head. Schreiner and Carnie (2016), building on foundational work by Ramchand (1993, 1997) and Adger (1994, 1996), claim that the syntactic structure underlying the overtly marked periphrastic aspects of Scottish Gaelic is that in (11).⁶ The aspectual particle is the realization of an Asp (Aspect) head. The auxiliary bi is the realization of T (Cowper 2010). The subject is licensed in a position between T and Aspect. For convenience, we label this position as the specifier of an AgrP, but nothing critical relies on this categorization. We leave aside the issue of the alternation in position of the object and VN, and assume that it is a consequence of different mechanisms of case-licensing (see Adger 1994; Carnie 1995; Bobaljik and Carnie 1996, among many others, for a discussion of this alternation).⁷

⁶ We have simplified the tree here by putting tha in T. It probably originates in a v head, possibly located above Asp (Travis 2010): [TP T+tha [vP tV [AspP Asp [VP . . . ]]]]. We abstract away from this detail. ⁷ In particular we assume, following the authors listed here, that accusative case is licensed postverbally in VO orders and pre-verbally in OV orders by the tran particle. The typical analysis of tran is that it is either an Agr category or a v category, perhaps corresponding to Voice. It is unlikely to be an Asp head, because it is not correlated with any semantic distinction and shows up in contexts without any marked aspect, such as non-finite clauses.

46 andrew carnie and sylvia l. r. schreiner (11)

TP T tha

AgrP DP subject

Agr’ Agr

AspP

⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩

Asp a’/ag air as dèidh a’ dol aL gu

vP ⎫ ⎪ ⎪

v′

⎬ ⎪ ⎪ ⎭

v tran

vnP verbal noun

This analysis is relatively straightforward when it comes to the particles a’/ag, air, as dèidh, and gu but it is far from transparent for the prospective marker a’ dol aL . The prospective marker translates literally as ‘going to’.⁸ It is a legitimate question whether the aspectual particle analysis in (11) is appropriate or whether this particular sequence might be better analysed as a biclausal structure with an embedding verb ‘go’, taking an infinitival complement clause as in (12). (12)

AspP Asp a’/ag

VP V dol

AspP Asp ag/aL (dh) [prospective]

vnP verbal noun



⁸ The use of a’ dol as a progressive verb of motion and aL as a directional preposition can be seen in examples such as (i): (i) Tha mi a’ dol aL Ghlaschu. be.pres 1sg prog go to Glasgow ‘I’m going to Glasgow.’

restricted and reversed aspectual contrasts 47 Schreiner and Carnie (2016) argue against (12) and claim that the a’ dol aL sequence in Scottish Gaelic (but not Modern Irish) has been grammaticalized into an Asp head. In this section, we briefly summarize that argument. We will not recapitulate all the arguments from Schreiner and Carnie (2016) here, but give a few points to give a flavour of the evidence. In Modern Irish, a wide variety of embedding predicates are compatible with what McCloskey (1983) identifies as a prospective interpretation. For example, Irish allows other verbs such as stad ‘stop’, suigh ‘sit’, lean ‘follow’, and tosaigh ‘begin’ to take a prospective complement. McCloskey (p.c.) has suggested to us that this is evidence that in Irish at least the ‘going’ (ag gabhail) portion of the construction is not critical to the prospective meaning, and instead these are biclausal constructions like (12), where the aL (dh) functions as the sole prospective marker, albeit one that can also be used to mark purpose clauses.⁹ (13) a. Tá mé ag gabhail a dh’ithe greim bidh. be.pres 1sg prog go.vn prt eat bite food.gen ‘I’m going to eat a bite of food.’ b. Thosaigh mé a dh’ithe mo dhinnéar. begin.past 1sg prt eat.vn my dinner ‘I began eating/to eat my dinner.’ c. Stad mé a dh’ithe mo dhinnéar. stop.past 1sg prt eat.vn my dinner ‘I stopped eating/to eat my dinner.’ d. Lean mé orm a dh’ithe mo dhinnéar. follow.past 1sg on.1sg prt eat.vn my dinner ‘I went on to eat my dinner/I went on eating my dinner.’ e. Shuigh mé a dh’ithe mo dhinnéar. sit.past 1sg prt eat.vn my dinner ‘I sat down to eat my dinner.’

[Irish]

The facts of Scottish Gaelic, however, are quite different. Other embedding verbs are incompatible with prospective interpretations. To the extent that such ⁹ A word is in order here about this particle. In its most usual form aL (where the superscript L indicates that it triggers the lenition consonant mutation on the following verbal noun), it is ambiguous with the transitive particle found with inverted object order. The aL here, however, is distinct from that particle in two ways: (a) it doesn’t trigger object inversion, and (b) rather than deleting in front of verbal nouns beginning with a vowel, as the transitive marker does, this one prefixes a dh in front of vowelinitial words. We use (dh’) throughout this chapter to distinguish this aL particle from other aL particles found in the language. The same distinctions also exist in Irish.

48 andrew carnie and sylvia l. r. schreiner embedding verbs license any interpretation at all, they are restricted to purpose clause interpretations and exclude any separate aspectual meaning. (14) a.

b.

c.

d.

e.

f.

Stad sinn a dh’ithe cèic. [Scottish Gaelic] stop.past 1pl prt eat.vn cake ‘We stopped in order to eat cake.’ / ∗ ‘We stopped eating cake.’ (purpose clause only) Bha sinn a’ stad a dh’ithe cèic. be.past 1pl prog stop.vn prt eat.vn cake ‘We stopped in order to eat cake.’ / ∗ ‘We stopped eating cake.’ (purpose clause only) ∗ Thòisich mi a dhannsadh a-rithist. begin.past 1sg prt dance.vn again Intended: ‘I began to dance again.’ ∗ Bha sinn a’ tòiseachadh a dh’ithe cèic. be.past we prog start.vn prt eat.vn cake. Intended: ‘We were starting to eat the cake.’ ∗ Thòisich a’ mhuir a dh’fhàs dorcha. begin.past the sea prt grow.vn dark Intended: ‘The sea began to grow dark.’ ∗ Thòisich àireamh-shluaigh an Ath-Leitheann a dh’fhàs begin.past population the Broadford prt grow nas motha. compr bigger Intended: ‘The number of people in Broadford began to grow bigger.’

Further, and again unlike Irish, purpose clause interpretations are completely unavailable in Scottish Gaelic ‘going to’ constructions when they use the a’ dol aL form, as in (15): (15)

Tha mi a’ dol aL dh’ithe cèic. be.pres 1sg prosp eat.vn cake ‘There is cake-eating in my future.’ / ∗ ‘I’m travelling in order to do some cake eating.’

In order to get the motion + purpose interpretation, a locative must be used, as in (16):

restricted and reversed aspectual contrasts 49 (16)

Tha mi a’ dol ann a dh’ithe cèic. be.pres 1sg prog go.vn there prt eat.vn cake ‘I’m going there in order to eat cake.’ (purpose clause interpretation)

Finally, when the verb ‘go’ is not in its a’ dol form, but in a different form (e.g. the preterite), then the purpose clause interpretation becomes the only available one, and the prospective interpretation disappears, as in (17): (17)

Chaidh mi aL dh’ithe cèic. go.past I prt eat.vn cake. ‘I went in order to eat cake.’ / ∗ I was going to be eating cake at some future point.’ (There must be physical going; i.e. purpose clause)

Clearly there is something special about a’ dol aL in Scottish Gaelic. In Schreiner and Carnie (2016), we propose that this sequence has been grammaticalized into an Asp head marking prospective aspect and as a consequence does not allow biclausal purpose clause interpretations. A prediction of this claim is that a’ dol aL should behave as a single particle when it comes to, for example, constituency phenomena. In Schreiner and Carnie (2016) we present a number of such cases, but repeat only one here: If a’ dol aL is a grammaticalized Asp head, then it shouldn’t be allowed to be interrupted by adverbial material. The examples in (18) demonstrate that this position is available for embedding predicates like a’ stad ‘stopping’, but not for a’ dol aL ‘going to’. Tha mi a’ stad a-rithist aL dh’ithe cèic. be.pres 1sg prog stop.vn again prt eat cake. ‘I am stopping again to eat cake.’ b. Tha mi a’ suidhe a-rithist aL dh’ithe cèic. be.pres 1sg prog sit.vn again prt eat.vn cake ‘I am sitting again to eat cake.’ c. ∗ ? Tha mi a’ dol a-rithist aL dh’ithe cèic. be.pres 1sg prog go again prt eat.vn cake. ‘I’m going again to eat cake.’ (marginally acceptable with a purpose reading) d. Tha mi a’ dol aL dh’ithe cèic a-rithist. be.pres 1sg prosp eat.vn cake again ‘I’m going to eat cake again.’

(18) a.

50 andrew carnie and sylvia l. r. schreiner This, along with facts from clefting and ellipsis, lead Schreiner and Carnie (2016) to conclude that a’ dol aL , despite surface appearances, has been reanalysed as an Asp head. This analysis aligns a’ dol aL with the other aspect particles, and leads us to propose a featural representation for it in line with what we propose for the other particles, as detailed in §3.4.

3.4 The featural representation of aspect contrasts in Scottish Gaelic Two of the marked aspects in Scottish Gaelic are easily expressed in Cowper’s (2003; 2005a) system of contrastive features: the progressive and the unrestricted perfect. The progressive in Scottish Gaelic seems to have the same basic properties as the equivalent construction in English (Ramchand 1993; Reed 2012). So, we adopt the same feature structure as proposed by Cowper, and suggest the Vocabulary Item insertion (VI) rule in (19) for the a’/ag morpheme. The feature structure here is identical to that used for the VI rule for English -ing by Cowper. We assume that this feature structure is housed in the progressive Aspect head. (19)

Asp VI Rule:



event

a’/ag

interval The simple past and future forms (see e.g. (14), (27)), on the other hand, encode perfective aspect. Cowper (2003) proposes for English that eventive clauses without interval receive a perfective interpretation. We suggest the same is true in Scottish Gaelic. Just as in English, a feature structure of event without interval is lexicalized by a rule that yields a simple form, while event with a dependent feature results in a periphrastic form. We propose the VI rule in (20) for the simple forms: (20)

Asp VI Rule:

event





There is no aspectual marking in these cases; instead, the different forms are the realizations of the feature structures in Tense.

restricted and reversed aspectual contrasts 51 The unrestricted perfect is similarly straightforward. The precedence feature can be realized in either T (giving past tense) or Asp (giving the perfect). We propose the VI rule in (21) for air, the unrestricted perfect Asp head: (21)

Asp VI Rule:

precedence



air

The other three aspects (the restricted perfect, the unrestricted prospective, and the restricted prospective) require some modifications to Cowper’s system. We start with the restricted aspects. As the name implies, the time interval between the event and the reference time is limited (see Reed 2012 for discussion). So we propose that the temporal ordering feature precedence bears a dependent feature that marks this limitation: restricted. When this feature appears as the only dependent of precedence, we have a restricted perfect meaning. This is lexicalized with the rule in (22): (22)

Asp VI Rule:

precedence



as dèidh

restricted Next we turn to the unrestricted prospective aspect. First, recall from above that prospectives are like perfects in that they have a precedence relation between rt and et, but the precedence relation is reversed: while in perfects the event precedes the reference time, in prospectives, the event has not yet occurred at rt.1⁰ We suggest that this is best represented by a feature dependent upon precedence called reversed (23): (23)

precedence reversed

We use this dependent feature rather than proposing a subsequence feature as an alternate to the precedence feature, because of the way it can combine with the restricted feature discussed above. If we had a subsequence 1⁰ An anonymous reviewer notes that perfect events, given that they precede reference time, must have an endpoint, making them different from prospectives in more ways than just the ordering of event and reference times. We note, however, that not all perfects include such an endpoint; in particular, in the case of ‘universal’ perfects in languages such as Scottish Gaelic and English, the perfect event continues into the reference time. Thus we assume that it is not encoded in the semantics of the perfect that the endpoint of reference time must precede event time; instead, such an arrangement is merely one possibility for the times involved in perfect aspect.

52 andrew carnie and sylvia l. r. schreiner feature, restricted would have to be allowed to be a dependent of both precedence and subsequence.11 With a reversed feature, restricted can always be a dependent of precedence independent of the direction of the precedence relationship.12 Cross-linguistic motivation for such a feature is not limited to the temporal domain. Pantcheva (2011), for example, proposes a reversal operation in her discussion of the decomposition of the Path head. This operation acts on a Goal path to yield a Source path, meaning that in her system a Source is more complex than a Goal in its semantic makeup (a Goal is a Goal, but a Source is a Goal + operation). She suggests that this additional complexity is attributable to a general cognitive bias (the ‘Goal bias’): ‘children and adults prefer to linguistically express Goals over Sources when describing motion events, change of state events, change of possession events, etc.’ (Pantcheva 2011: 70). precedence without the reversed feature is interpreted with the event time before the reference time, as in a perfect (the usual case). precedence with the reversed feature has the event happening after the reference time, yielding a prospective. The VI rule for the grammaticalized a’ dol aL head is given in (24): (24)

Asp VI Rule:

precedence



a’ dol aL

reversed Finally, the reversed feature co-occurs with the restricted feature in the restricted prospective. This is realized by the VI rule in (25), which inserts gu in this context.

11 An anonymous reviewer wonders whether reversed could just occur under a subsequence feature, as well as under the precedence feature. Although nothing necessarily prohibits us from locating reversed under both features, listing these as separate features instead of making reversed dependent upon precedence makes the similarities between perfect and prospective (in contrast to other aspects) accidental. We suggest that Gaelic shows the two restricted aspects it does, and not more, because restricted is a feature of aspects that predicate a separation relation between et and rt, and not others (such as the progressive). The separation of these times results in an interval that may in principle be short or long; a language like Scottish Gaelic can grammaticalize information about this interval. 12 An anonymous reviewer asks what our proposal means for languages that don’t use the reversed feature—could they use precedence for both future and past? Cowper’s (2003: 5) description of this feature is ‘[Precedence] signifies that the IP in whose head it appears is located temporally prior to its temporal anchor.’ This temporal anchor can be speech time (in the case of Tense) or reference time (in the case of Aspect). Languages with a true future tense (rather than a modal) would presumably employ the reversed feature for tense, dependent on precedence.

restricted and reversed aspectual contrasts 53 (25)

Asp VI Rule:

precedence restricted



gu

reversed

The addition of reversed and restricted to the feature geometry thus captures exactly the additional aspectual contrasts found in Modern Scottish Gaelic.

3.5 Predictions The addition of the restricted and reversed features is motivated by the aspectual contrasts in Scottish Gaelic. But, of course, any complication of a theoretical claim requires confirmation from its interaction with other parts of the grammatical system. In this section, we consider two predictions of the proposal we made above. First, we look at the possibility that the reversed feature can also be used to distinguish past from future tenses, and the implications this has for tense morphology in Scottish Gaelic. Second, we look at the typological implications of the restricted feature and see if it has use beyond Scottish Gaelic and beyond restricted perfects and prospectives.13

3.5.1 Two pathways to the future? We have largely ignored the fact that for Cowper and her colleagues the feature precedence has two realizations in English and other languages: It is realized as past tense when anchored to the speech time, and as the past participle when anchored to reference time. If, as we propose, there is a reversed feature dependent upon precedence, then it follows that this feature should 13 An anonymous reviewer brings up a typological question of the precedence feature: Why is the default relation event time before reference time (as in a perfect), and not the other way around (as in a prospective)? It is difficult to determine without a large, targeted typological study what the frequency status of perfects vs. prospectives is in the world’s languages, as the existing data on prospectives is limited. As we are not proposing a cross-linguistic feature geometry here, we do not attempt to make a universal claim about a typological default for this feature. It seems possible that having the choice between a precedence feature and a subsequence feature might benefit analyses of any languages that might include a prospective but not a perfect. However, a subsequence feature would not improve the current analysis. Putting a subsequence feature in place of precedence would not change the analysis (unless it was to reflect a wider typological pattern). If precedence and subsequence features were both present, with restricted dependent on each, the generalization that restricted is a feature of perfects and prospectives to the exclusion of progressives is lost.

54 andrew carnie and sylvia l. r. schreiner be active in the T domain as well as in the Asp domain. So we might ask if the dependency precedence—reversed could be a pathway to future tense. Cowper (2005a) and Cowper and Hall (2007) explicitly exclude a ‘pure’ subsequence model of the future (i.e. using the opposite of precedence to mark the future), noting that futures in English almost always have some kind of modal force to them. They propose that future ‘tense’ is really one possible realization of the modality feature irrealis. This is consistent with the fact that English will behaves both syntactically and semantically like a modal auxiliary. We claim that the same is not true for the Scottish Gaelic future. We claim that the Scottish Gaelic future tense is not linked to irrealis but it is instead a representation of precedence—reversed, but realized in the T head rather than the Asp head. First note that future tense marking in Scottish Gaelic is not accomplished via the usual means for marking mood/modality. Most mood/modality in Scottish Gaelic is expressed by a structure using two verbs. The modality either is expressed as a copular construction with some kind of modal nominal, such as urrainn ‘able/ability’ in (26a), or is a verb itself as in (26b)–(26c) (Byrne 2002; Black 2006). There is no modal marking on the lexical verb. (26) a. B’ urrainn dhomh cèic ithe. cop.cond able to.1sg cake eat.vn ‘I can eat cake.’ b. Faodaidh mi cèic ithe. may.fut I cake eat.vn ‘I may eat the cake.’ c. Feumaidh mi cèic ithe. must.fut I cake eat.vn ‘I must eat the cake.’ The future tense, by contrast, is marked directly on the verb in Scottish Gaelic by the suffix -(a)idh, not by a separate modal verb or modal noun structure. Note that future tense morphology and modality are not incompatible with each other, as shown in (26b). Modal verbs in Scottish Gaelic may appear in the future or past tense like other verbs. So future tense and modals are grammatically distinct in the language. More revealing, however, is the fact that Scottish Gaelic, unlike its close sibling Modern Irish, completely lacks a present tense form of the verb.1⁴ 1⁴ Modern Irish has a present tense form for regular verbs, a special past habitual form, and a present habitual specifically for the verb ‘to be’. All of these are missing from the Modern Scottish Gaelic paradigms.

restricted and reversed aspectual contrasts 55 Table 3.2 Regular Scottish Gaelic verbs (exemplified by fuirich ‘wait’) Past Independent Dependent

Dh’fhuirich Cha do dh’fhuirich

Present — —

Future/Habitual Fuirichidh Chan fhuirich / Ma fhuiricheas

Table 3.3 The verb bi ‘be’

Independent Dependent

Past

Present

Future/Habitual

Bha Cha robh

Tha Chan eil / Am bheil

Bithidh Cha bhi / Ma bhitheas

The exception is the auxiliary verb bi, which expresses the traditional threeway tense distinction. The majority of verbs in Scottish Gaelic morphologically express a two-way distinction between a past and a future/habitual form. The past is typically formed by applying the initial consonant mutation of lenition to the first sound (or in some cases by prefixing dh’). The future is formed with a suffix, typically -(a)idh. There are also variant forms that appear in ‘dependent’ contexts in which the verb is preceded by a particle (such as the negative marker cha, the relative clause marker, or the conditional particle). Examples of these forms are shown in Table 3.2 using the verb fuirich ‘wait’. The verb bi ‘to be’ has a richer paradigm. It has a present tense in addition to the past and future/habitual, as shown in Table 3.3. An important point about the non-modal nature of future morphology in Scottish Gaelic is visible from the fact that many of the bleached uses of the English present tense (e.g. generic, habitual, etc.; see Cowper 1998) are expressed with the future tense in Scottish Gaelic (Gleasure 1986, 1990): (27) a. Bithidh Mairi ag òl. be.fut Mary prog drink.vn ‘Mary drinks.’ b. Bruidhidh Màiri Gàidhlig gu tric. speak.fut Mary Gaelic frequently ‘Mary speaks Gaelic frequently.’ Note that the forms in (27) lack modal force,1⁵ and are clearly not irrealis in meaning. This is explained if future tense in Scottish Gaelic is not a realization 1⁵ Of course, it is not the case that all future tense forms in Scottish Gaelic lack modal force, but we speculate that any such interpretations follow from conversational implicatures rather than grammatical modality.

56 andrew carnie and sylvia l. r. schreiner of the irrealis feature. This conclusion is consistent with the view expressed by Hayashi (2011: 85–110), who argues that the South Baffin Inuktitut -laaq also expresses a general future tense rather than modal future. Ó Sé (1990) argues that the realis/irrealis distinction is realized in Modern Irish as part of the preverbal particle system, rather than verbal/tense morphology. If his analysis transfers over to Scottish Gaelic, then it follows that we should not use irrealis to mark future tense in the language. These facts suggest to us that the future tense in Scottish Gaelic is a realization not of irrealis but of precedence, like the past tense. Like prospective aspect, though, the time relation in future tense is reversed with respect to its precedence counterpart. This is explained if it (like prospective aspect) realizes precedence—reversed.

3.5.2 Precedence and restricted tense systems A second prediction of the analysis presented above is that just as precedence—reversed might extend to both tense and aspect systems, precedence—restricted could do the same and be found in tense systems as well as aspect systems.1⁶ One such possibility is the French recent or immediate past (28): (28)

Je viens de dire … I come of say.inf

[French]

‘I just said …’ Another possible example is the hodiernal (for ‘today’ events whose runtimes are contained within the current day) and hesternal (for ‘yesterday’ events whose runtimes occurred in the day before the current one, or a bit farther in the past) tenses described in some detail by Dahl (1985). These tenses are restricted in distance from the time of speech. They are found in a wide 1⁶ One possibly surprising fact for us is that Scottish Gaelic, which uses the restricted feature in the aspectual system, lacks a restricted tense system. Under Cowper’s feature system we expect that featural distinctions can express themselves in either temporal or aspectual domains, as the domains differ only in what discourse anchors the features refer to. Cowper (2003: 11–2) debates the possibility of single vs. multiple projections for Tense, Aspect, and Mood, leaving it as a question open to further empirical support. The fact that Scottish Gaelic lacks restricted contrasts in tense (and e.g. that French has a restricted tense contrast but no concomitant contrast in the aspectual domain) suggests that the featural specifications for Tense and Aspect need to be able to be different within a given language. This would fall out of the analysis if the inflectional phrases of Tense and Aspect are projected separately and allow different features.

restricted and reversed aspectual contrasts 57 variety of languages and language families, including many Bantu languages, Inuktitut, and some Polynesian languages (see e.g. Givón 1972; Johnson 1977; Nurse 2003; Hayashi 2011; Cable 2013). In such languages, the Tense node rather than the Aspect node (or possibly in addition to it, if the language happened to employ restricted aspects as well) would contain precedence— restricted. Further feature(s) would be necessary for languages that employ more than one distinction of tense restriction.

3.6 Conclusions We have argued that Cowper’s (2003, 2005a) feature-geometric representation of temporal and aspectual contrasts should be modified to include two additional features: reversed, which is used to mark both prospective aspect and non-modal futures, and restricted, which accounts for restricted perfect and prospective aspects, and possibly hodiernal and recent tenses. Adding these features as dependents to the already well-motivated precedence feature is a minor adjustment that expands the explanatory power of Cowper’s system to a new range of aspectual and tense contrasts.

Acknowledgements Thanks to Muriel Fisher, Heidi Harley, Bridget Copley, Richard Compton, Seth Cable, Mary Willie, Claire Bowern, Michael Bauer, Jim McCloskey, Daniel Currie Hall, Bronwyn Bjorkman, and audiences at the Contrast in Syntax Workshop in 2015 and the Celtic Linguistics Conference in Rennes, France in 2012, as well as several anonymous reviewers. We especially thank Elizabeth Cowper for inspiring the approach we present here. This research was supported in part by a grant from the US National Science Foundation #BCS0639059.

4 Sentience-based event structure Evidence from Blackfoot Elizabeth Ritter

4.1 Sentience, argument structure, and event structure in Blackfoot It is a striking fact about Blackfoot that sentience constrains argument structure in a variety of ways.1 First, as observed by Frantz (2017), external arguments of transitive verbs must be sentient.2 For example, instruments cannot function as subjects (1), but must be realized as adjuncts licensed by the verbal prefix iiht- (2).3

1 Blackfoot is an Algonquian language of southern Alberta and northern Montana. Though numbers vary, Genee and Junker (2018: 278) estimate that there are very roughly 4,700 speakers, pointing out that this group includes individuals with different levels of fluency. 2 I use the terms sentient and sentience to mean notional animacy (also known as semantic animacy). This property characterizes humans and other rational beings (e.g. deities and domesticated animals) that are capable of perceiving and responding to the world around them. I reserve the terms animate and animacy for grammatical animacy, a morphosyntactic property reflected in the form of nouns, verbs, and determiners. 3 All data in this chapter is taken from the sources cited. For data from published sources, I include page numbers, for data from online sources and manuscripts, I include example numbers. In some cases, glosses have been modified for uniformity. The following changes have been made: (i) The theme marker that appears immediately after ti verb stems is glossed differently in each source. This morpheme, whose form depends on whether the subject is 1st/2nd person or 3rd person, indicates the object is grammatically inanimate. Since person agreement is marked elsewhere on the ti verb, I have glossed it simply as th throughout; (ii) The suffixes -wa and -yi that appear on demonstratives and nouns are glossed as 3rd person singular and 4th person singular by some authors, and as proximate (singular) and obviative (singular) by others. I consistently gloss them as 3rd singular and 4th singular; (iii) The verbal marker known as the direct/inverse suffix, which is used when both the subject and primary object are grammatically animate, indicates which of the two is higher on the person hierarchy. Some authors simply gloss this with two numerals separated by a colon to indicate the person of the subject and object, respectively. These authors use 3 for 3rd person proximate and 4 for 3rd person obviative (aka 4th person). Others simply use dir(ect) to indicate that the subject is higher on the hierarchy, and (inv)erse to indicate that the object is. I adopt the latter convention throughout; (iv) Some authors gloss the verbal prefix ii- as past. However, I gloss this as ic ‘initial change’ as this is the traditional Algonquian term for this element. It differs from other prefixes in that the first vowel in the verb complex (not including the person prefix) is replaced with ii- whenever there is no other tense/aspect prefix. See footnote 20, Bliss (2013), and references cited therein for discussion. Elizabeth Ritter, Sentience-based event structure: Evidence from Blackfoot In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Elizabeth Ritter. DOI: 10.1093/oso/9780198817925.003.0004

sentience-based event structure 59 (1)



oma isttoána ikahksínima annistsi ikkstsíksiistsi om-wa isttoan-wa ikahksini-m-wa ann-istsi ikkstsiksi-istsi dem-3sg knife-3sg cut.ti-th-3sg dem-in.pl branch-in.pl

‘That knife cut off those branches.’ (adapted from Frantz 2017: 49) (2)

oma isttoána om-wa isttoán-wa dem-3sg knife-3sg ikkstsíksiistsi ikkstsiksi-istsi branch-in.pl

iihtsikahksínii’pi iiht-ikahsíni-’p-yi means-cut.ti-th-in.pl

annistsi ann-istsi dem-in.pl

‘By means of the knife [somebody] cut off those branches.’ (adapted from Frantz 2017: 49) This sentience requirement on external arguments even extends to causer arguments of morphologically derived causative verbs, such as (3): (3)

kitsó’kááttsaayaawa kit-io’kaa-attsi-aː-yi-aawa 2-sleep.ai-caus.ta-dir-3pl-pron

‘You put them to sleep.’ (adapted from Frantz 2017: 112) Like instruments, non-sentient causers are expressed as adjuncts, and licensed by the means prefix iiht-: (4)

Iiyiksopoyi iihtomatsinaka’si niksi áínaka’si. iiyi-sopo-yi iiht-omat-inaka’si niksi ainaka’si be.hard-wind-in.sg means-start-roll.ai dem wagon

‘A strong wind caused this wagon to start rolling.’ (Meadows 2010: 19) Another set of arguments that must be sentient in Blackfoot are the noncore objects of morphologically derived transitive verbs. Bliss (2007, 2010) demonstrates that the applicative verbal suffix -omo licenses sentient goals (5); non-sentient goals are expressed as adjuncts, licensed by the purpose prefix iht- (6):

60 elizabeth ritter (5)

Nitááhkanomoawa ana issítsimaan amiksi si’káániksi. nit-(w)aahkan-omo-a-wa an-(w)a issitsimaan am-iksi si’kaan-iksi 1-sew-ta.ben-dir-3prox.sg dem-3sg baby dem-an.pl blanket-an.pl

‘I sewed those blankets for the baby.’ (adapted from Bliss 2010: (12a)) (6)

Nihtááhkanayi amiksi si’káániksi ani ákssin. n-iht-aahkan(i)-a-yi am-iksi si’kaan-iksi an-(y)i akssin 1-purp-sew.ta-dir-3pl dem-an.pl blanket-an.pl dem-in.sg bed.in

‘I sewed those blankets for the bed.’ (adapted from Bliss 2010: (12b)) Similarly, Meadows (2010) shows that the accompaniment suffix -m licenses sentient non-core objects that volitionally perform the action described by the verb alongside the external argument. For example, the sentence in (7) is acceptable if Pokaaki goes willingly with the giant, but not if he kidnaps her. (7)

Omahkmatapiwa iihpokiistapokska’siimiiwa omahk-matapi-wa ii-ohpok-miistap-okska’si-m-yii-wa big-person-3sg ic-with-away-run-ta.accm-dir-3sg ni Pokaaki. ann-yi Pokaaki dem-obv.sg Pokaaki

‘The giant ran away with Pokaaki.’ (adapted from Meadows 2010: 26) What is the source of the strict sentience requirement on these different arguments? And what is the source of contrast between Blackfoot and languages such as English where sentience plays a much more minor role? In this chapter, I argue that the answer to these questions can be found in the usual suspects, i.e. the functional syntactic categories and their formal feature content, cf. Cowper (2005a,b), Bjorkman and Cowper (2013), Cowper and Hall (2014a), Ritter (2014), Wiltschko (2014), among others, for arguments that functional categories and their feature content are subject to crosslinguistic variation. Following Wiltschko and Ritter (2015), I assume that the strict sentience requirements on argument structure are part of the grammar (i.e. part of the ‘narrow syntax’) of Blackfoot, and more specifically, that sentience is formalized as a feature that is subject to selection, a feature-checking operation, much like Agree. Wiltschko and Ritter propose that in Blackfoot all sentient DPs have a monovalent phrasal sentience feature ([m]), and that this feature

sentience-based event structure 61 is selectable by different functional heads, including v, Appl and Asp, as schematized in (8):⁴ (8)

FP DP [m] F u[m]

...

Given the widely-held view that syntactic arguments realize components of event structure (Tenny 1994; Levin and Rappoport-Hovav 1998; Ritter and Rosen 2000; Ramchand 2008, among others), this sentience requirement on syntactic arguments may also indicate that sentience is a primary determinant of event structure in Blackfoot. In this chapter, I explore the hypothesis that Blackfoot event structure is organized around the distinctive capacity of sentient arguments for intentional action and sensory/psychological response. I propose that sentient arguments define the ultimate initial and terminal bounds for events in this language, and develop a sentience-based typology of situations (i.e. states and events) that distinguishes the following types: (i) interactive situations, which involve two sentient arguments; (ii) active and reactive situations, which involve a single sentient argument; and (iii) inert situations, which involve no sentient arguments. I assume that each sentient argument bearing an event role is introduced by a distinct event-related functional category. Since there is no general consensus on the names of these categories, I shall simply label them F. It follows that interactive situations, which have two sentient arguments, are complex events with two event-related functional categories, active and reactive situations, which have only one sentient argument, have only one event-related functional category, and inert situations, which lack sentient arguments altogether, have no event-related functional categories. These structural differences are schematized in (9):

⁴ Wiltschko and Ritter (2015) represent this phrasal feature as [H] to indicate that it attaches to DPs that denote humans and humanoids, i.e. other individuals that share the sentience property of humans. However, in Ritter and Wiltschko (2016), [H] is used for impersonal pronouns that are necessarily human. Consequently, I represent the feature in question as [m] to capture the fact that sentient DPs denote individuals that Reinhart (2003) characterizes as mental state holders.

62 elizabeth ritter (9) a. Interactive situation FP1 DP [m] F1 u[m]

FP2 DP [m] F2 u[m]

vP

b. Active/reactive situation FP DP [m] F u[m] c. Inert situation vP

vP

DP v

Root

The hypothesis that sentience is an integral component of event structure roles in Blackfoot leads to the following predictions: Prediction 1: If event initiation is determined by sentience, then we should expect to find sentience restrictions on causers and agents alike. Prediction 2: If event delimitation is determined by sentience, then we should expect that the boundedness of the object should play no role in distinguishing between event types.

sentience-based event structure 63 Prediction 3: If event delimitation is determined by sentience, rather than boundedness, then sentience, rather than boundedness, may also play a role in distinguishing between nominal types. Prediction 4: If only sentient arguments may bear event roles, then inert situations, which have no sentient arguments, are not eventive in the relevant sense. We should expect, then, that eventiveness is not correlated with dynamicity. The remainder of this chapter will show that these predictions are borne out, and discuss broader consequences of the sentience-based event structure for the grammar of Blackfoot, and for theories of event structure.

4.2 Causers, agents, and intention Causers and agents are canonical external arguments. Treatments of causers and agents in well-studied languages, such as English, assume that they are distinguished by the fact that only agents must act intentionally, and as a result, only agents must be sentient (e.g. Doron 2003; Reinhart 2003; Folli and Harley 2007). (10) a. The wind/John opened the door. b. John ate the cake.

causer agent

Causers and agents participate in different kinds of events, as evidenced in part by the availability of an anti-causative construction (11a), but not an antiagentive one (11b): (11) a. b.



The door opened. The cake ate. (cf. The cake was eaten.)

The contrast in (11) is attributed to the fact that causatives denote complex events, consisting of a causing subevent and a result subevent, whereas agentive events are simplex events that involve both action and change. The anticausative is possible because the result subevent can be isolated from the causing subevent. The ungrammaticality of (11b) is due to the fact that the agentive event cannot be decomposed in the same fashion.

64 elizabeth ritter Further evidence of this difference in event structure comes from interpretive options for adverbial modifiers. As illustrated in (12a), when again is added to a causative event, the result is ambiguous—the adverb can either modify the causing subevent, indicating that the causer did some action again (the so-called repetititve reading), or it can modify the result subevent, indicating the result obtains again (the so-called restitutive reading). However, when again is added to an agentive event, no ambiguity arises. The adverb can only modify the simplex event, and consequently, indicates repetition of the action by the agent. (12) a. John opened the door again. i. repetitive: John opened the door before, and he has done so again. ii. restitutive: The door was open before, and now it is open again. b. John read the book again. i. repetitive: John read the book before, and he has done so again. ii. ∗ restitutive: The book was read before, and it has been read again. Summarizing, there are two sources of contrast between English agentive and causative events: (i) whether the external argument necessarily acts intentionally and (ii) whether the event consists of one or two subevents. I assume that the requirement for an external argument that acts with intention is formalized in the syntax as selection for a sentient argument, and that the complexity of the event is represented by the functional categories projected. Selection for a sentient argument is represented as selection for an argument specified with the feature [m], which indicates that the argument denotes a mental state holder; cf. Reinhart (2003). Causative events require two functional categories, and since intention is not necessary, the external-argument introducing head does not select for [m], as schematized in (13a).⁵ Agentive events, on the other hand, require only one eventive functional category, but this category selects for [m], as schematized in (13b): ⁵ Complexity of the event structure, and intentionality are independent of one another. For example, have-causatives such as (i) are complex events that involve intention on the part of the causer, and also on the part of the causee, cf. Folli and Harley (2007); Copley and Harley (2009); Bjorkman and Cowper (2013). (i) Elizabeth/*Fear of failure had the students read three articles. I assume that (i) is a complex event that requires two argument-selecting functional categories, and that the intentionality constraint on the causer and causee arguments is due to the fact that both functional heads selects for [m].

sentience-based event structure 65 (13) a. Causative event

b. Agentive event

FP1

FP

DP

DP [m] Fcaus

FP2

Fagent u[m]

XP

DP F2

XP

Turning to Blackfoot, we find that the structures in (13) fail to capture the facts of this language. As noted in the introduction, in Blackfoot morphological causative verbs and agentive (transitive) verbs alike require a sentient external argument. This is due to the fact that morphological causatives are realized as transitive verbs, and hence, must satisfy the same strict sentience requirement as other transitive verbs. Recall also that non-sentient causers are realized as adjuncts licensed by the verbal means prefix -iiht. The relevant examples are repeated here as (14) and (15). (14)

kitsó’kááttsaayaawa kit-io’kaa-attsi-aː-yi-aawa 2-sleep.ai-caus.ta-dir-3pl-pron

‘You put them to sleep.’ (adapted from Frantz 2017: 112) (15)

Iiyiksopoyi iihtomatsinaka’si niksi áínaka’si. iiyi-sopo-yi iiht-omat-inaka’si niksi ainaka’si be.hard-wind-in.sg means-start-roll.ai dem wagon

‘A strong wind caused this wagon to start rolling.’ (Meadows 2010: 19) In the remainder of this section I will show that Blackfoot causative clauses with a sentient causee have a complex (interactive) event structure, but causatives with a non-sentient causee have a simplex (active) event structure. Evidence that Blackfoot causative clauses with two sentient arguments have a complex event structure was first adduced by Johansson (2009). She explored the event structure of clauses containing causative verbs derived by the

66 elizabeth ritter addition of the suffix -attsi. Adapting tests developed by Pylkkänen (2008) to Blackfoot, she found that agent-oriented adverbs, such as sstsim- ‘reluctantly’ are ambiguous between modification of the causer sub-event and the causee sub-event, as illustrated in (16): (16)

Anna Mary áísstsimáíhpiyiattsi ni otsitapíím. ann-wa Mary á-sstsim-ihpiyi-attsi ann-yi ot-sitapíím dem-3sg Mary ipfv-reluctant-dance-caus.ta dem-4sg 3-doll

a. ‘Mary, reluctantly, made the doll dance.’ Context: Mary has been asked to demonstrate that her doll can dance, but she doesn’t want to. b. ‘Mary made the doll, reluctantly, dance.’ Context: In play, Mary is asking her doll, who she views as a [sentient] being to dance. Her doll doesn’t want to dance, but Mary begs the doll until finally the doll consents to dance, albeit reluctantly. (Johansson 2009: 35) What this shows is that -attsi causative verbs require an agentive causer and an agentive causee. Following Pylkkänen, Johansson interprets this as evidence that Blackfoot -attsi causatives have two external argument-introducing functional categories: voiceP and VoiceP. The causer is introduced in the specifier of the higher category, voiceP, and the causee is introduced in the specifier of the lower category, VoiceP. The causativizing suffix is Merged in the head of voice, and the verb-forming morpheme is Merged in Voice. Abstracting away from category labels, the structure Johansson proposes is essentially identical to the complex structure I ascribe to interactive events. Though Johansson does not address the observed sentience requirement on both the causer and causee, I attribute this to the fact that both voice and Voice select for a sentient arguments, formalized in (9a) as selection for [m]. All valence-increasing verbal suffixes in Blackfoot add a sentient argument to a verb that inherently has a single sentient external argument. The causative suffix -attsi is representative in that it adds a sentient causer to an intransitive verb with a sentient (causee) argument.⁶ Derived causative verbs with a

⁶ I adopt Johansson’s (2009) analysis of the causativizing morpheme -attsi as voice, a category that obligatorily introduces an external argument (the causer), and selects a VoiceP complement. Like voice, Voice is a functional category that introduces an external argument. When Voice is the complement of -attsi, its external argument is interpreted as a causee, i.e. an argument whose action is directed by the causer.

sentience-based event structure 67 non-sentient causer are ruled out by the strict sentience requirement on external arguments of transitive verbs. Causative verbs with a non-sentient causee exist, but these are all lexical causatives, i.e. basic transitive verbs with a sentient external argument. Strikingly, when the causative event involves only a single sentient argument, it has a simplex event structure in Blackfoot. Evidence for this is due to Louie (2010). She reports that, unlike its English translation, the Blackfoot sentence meaning ‘I opened the door again’ is not ambiguous. More specifically, in Blackfoot, the adverb ‘again’ can modify the action of the sentient causer, but not that of the non-sentient causee, giving rise to a repetitive interpretation, but not a restitutive one: (17)

Nimáttsistá’kowai’piksi’p. ‘I opened it again.’ a. Repetitive: omi kitsím nitsíkowai’piksíp nitáitsipissi omi kitsim nit-ikowai’pikisi-’p nit-á-it-ipi-ssi dem door 1-open.ti-th 1-ipfv-rl-enter-cj itsohkohpápoka ki nimatts(it)sistai’kowai’piksíp. it-yohkohpápoka ki ni-matt-(it)-ista’-ikowai’piksi-’p rl-blow.shut conj 1-add-(rl)-again-open.ti-th ‘I opened that door when I went in, but then it blew shut. And (then) I opened it again.’ b. Restitutive:

omi kitsím ííkaikowaistsii, nitáitsipissi omi kitsim iikaa-ikowaistsii nit-á-it-ipi-ssi dem door pfv-open.ii 1-ipfv-rl-enter-cj # itsohkohpápoka ki nimatts(it)sistai’kowai’piksí’p. it-yohkohpápoka ki ni-matt-(it)-ista’-ikowai’piksi-’p rl-blow.shut conj 1-add-(rl)-again-open.ti-th

‘That door was already open when I went in, then it blew shut. # And (then) I opened it again.’ (Louie 2010: 4) Together these data indicate that the complexity of event structure in Blackfoot causatives depends on the number of sentient arguments. Causative events that contain a sentient causer and a non-sentient causee are simple active events. Causative events that contain both a sentient causer and a sentient causee are complex interactive events. Moreover, the facts of morphological and lexical causatives discussed above bear out my first prediction— that causers and agents are subject to the same strict sentience restriction.

68 elizabeth ritter

4.3 Boundedness and event delimitation In this section, I turn my attention to the second prediction listed in §4.1: Prediction 2: If event delimitation is determined by sentience, then we should expect that the boundedness of the object should play no role in distinguishing between event types. The literature on Blackfoot has explored this issue from a variety of perspectives, and the evidence strongly suggests that sentience plays a role in event structure. What is less clear is whether boundedness also plays a role. In this section, I re-examine the evidence that bears on this question. Before turning to the question at hand, I define the notion boundedness and its role in event structure, and review some basic facts about the morphology of Blackfoot nouns and verbs that will inform the discussion below. Boundedness is a semantic property that characterizes both entities and events. Simplifying somewhat, an entity or event is unbounded if its subparts are entities or events of the same type (Mourelatos 1978; Moens 1987; Declerck 1989; Jackendoff 1991; Verkuyl 1993, among many others). Mass nouns and atelic events denote unbounded entities and events, respectively. For example, a cupful of water, and a half cupful of water both denote the same type of entity, i.e. water. Similarly, running for an hour and running for a minute both denote the same type of event, i.e. running. Bounded entities and events cannot be divided in this way. These are denoted by count nouns and telic events: A book and a page of that book denote different types of inanimate entities; similarly, writing a book and writing a page are different kinds of writing events. Building on seminal work by Travis (1992, 2010), Wiltschko (2012) proposes that boundedness is represented in the narrow syntax if the interpretable feature [±bounded] is associated with Inner Aspect, a functional category that is Merged in the lexical layer of both nominal and clausal syntactic spines. Before turning to the evidence that my second prediction is borne out, I briefly describe morphological properties of Blackfoot nouns and verbs that will be relevant to the discussion. Like all Algonquian languages, Blackfoot has an animacy-based noun classification system. Animate nouns are pluralized with the suffix -iksi; inanimate nouns with the suffix -istsi.⁷ The two classes are also distinguished semantically. The inanimate class contains only nouns that denote things, materials, and other non-sentient entities, such as forces ⁷ In most Algonquian languages, nouns can be classified as animate or inanimate based on the form of their stems and/or singular suffixes (e.g. Wolfart 1973; Goddard 2002; Valentine 2010). Blackfoot differs from its sister languages in that classification of a noun as animate or inanimate is based on the form of its plural suffix.

sentience-based event structure 69 of nature. The animate class includes all nouns that denote humans, animals, and gods, as well as a sizeable number of nouns that denote objects, notably objects that roll, objects made of metal, and an assortment of nouns that denote berries, items of clothes, body parts, and other inanimate objects. Representative examples are provided in Table 4.1. In Algonquian languages, the morphological classification of verbs is based on animacy and transitivity (Bloomfield 1946). These two properties give rise to four classes of verbs—two transitive classes, distinguished by the grammatical animacy of their object, and two intransitive classes, distinguished by the grammatical animacy of their subject. Bloomfield’s labels for each of these verb classes, and their properties are summarized in Table 4.2. The remainder of this section is organized as follows: In §4.3.1, I summarize arguments by Ritter and Rosen (2010) and Louie (2008) that Blackfoot does not grammaticalize boundedness in the verbal domain. In §4.3.2 I address counterarguments from Kim (2015, 2017a,b). Finally, in §4.3.3, I discuss the relationship between the animacy-based morphosyntactic verb classification that characterizes all Algonquian languages, and the sentience-based aspectual classification that I propose is necessary, at least for Blackfoot. Table 4.1 Blackfoot nouns Animate nouns

Inanimate nouns

saahkómaapi saahkómaapiiksi

‘boy’ ‘boys’

niítahtaa niítaistsi

‘river’ ‘rivers’

imitáá imítaiksi

‘dog’ ‘dogs’

i’ksisako i’ksisakoistsi

‘meat’ ‘meats’

áinaka’si áínaka’siiksi

‘wagon’ ‘wagons’

isttsikónistsi isttsikónistsiistsi

‘sled’ ‘sleds’

po’táa’tsis po’táa’tsiiksi

‘stove’ ‘stoves’

iitáísooyo’p iitáísooyo’pistsi

‘table’ ‘tables’

mi’ksiníttsiim mi’ksiníttsiimiksi

‘buffalo berry’ ‘buffalo berries’

okonok ókonokistsi

‘Saskatoon berry’ ‘Saskatoon berries’

Table 4.2 Blackfoot verb classes Verb class transitive animate transitive inanimate animate intransitive inanimate intransitive Source: Bloomfield (1946)

(ta) (ti) (ai) (ii)

Classifying argument

Animacy

object object subject subject

animate inanimate animate inanimate

70 elizabeth ritter

4.3.1 Boundedness is not grammaticalized in Blackfoot clauses Blackfoot verb stems are minimally bimorphemic, consisting of a root and verb class morpheme that specifies which morphological class the verb belongs to: ta, ti, ai, or ii. (Algonquianists call the verb class morpheme the final, because it is the last morpheme in the stem.) A given root will typically combine with two or three different finals, deriving a set of related verbs. One reason for the existence of sets of related verbs is that Blackfoot ta and ti verbs require a definite direct object. If the direct object is indefinite, a verb of the ai class is required. (Henceforth, I refer to these animate intransitive verbs that take indefinite objects as ai+o.) This is illustrated in (18) with the ta, ti, and ai(+o) verbs meaning ‘eat’.⁸ (18) a. Naowatsiw amo mamii. na-oowat-yii-wa amo mamii evid-eat.ta-th-3sg dem fish.an ‘S/he ate this fish.’ b. Naowatoom ani akoopis. na-oowatoo-m-wa ani akoopis evid-eat.ti-th-3sg dem soup.in ‘S/he ate that soup. c. Naoyiw (mamii / akoopis). na-ooyi-wa mamii akoopis evid-eat.ai-3sg fish soup ‘S/he ate (fish/soup).

ta

ti

ai(+o) (Ritter and Rosen 2010: 134)

Given the well-established link between boundedness of the object and boundedness (telicity) of the predicate, Ritter and Rosen (2010) investigated whether verb class alternations between ta/ti verbs and their ai or ai+o counterparts signals a shift between telic and atelic predicates in Blackfoot (e.g. Verkuyl 1972, 1989, 1993; Hinrichs 1985; Tenny 1987, 1994; Krifka 1992; Harley 2003). To this end, they employed Blackfoot versions of standard telicity tests, including (a) determining whether the aspectual preverb iksist‘finish’ can be felicitously combined with related ta/ti and ai/ai+o verbs; (b) determining whether addition of the adverbial prefix iimat- ‘almost’ gives ⁸ Ritter and Rosen (2010) gloss the verbal prefix na- as a past tense morpheme. However, I follow Bliss and Ritter (2015) in analysing it as an evidential prefix that signals that the knowledge holder has conclusive evidence that a past time situation has indeed occurred.

sentience-based event structure 71 rise to an ambiguity when combined with related ta/ti and ai/ai+o verbs; and (c) whether adding the imperfective prefix -a to related ta/ti and ai/ai+o gives rise to an imperfective paradox.⁹ What Ritter and Rosen found was that there is no difference between ta/ti verbs on the one hand and ai/ai+o verbs on the other: (a) All the verbs tested were compatible with iksist- ‘finish’. (b) Adding iimat- ‘almost’ never results in ambiguity. Rather, for both ta/ti and their ai/ai+o counterparts, adding this adverbial modifier signals only that the event almost started. (c) They replicated Dunham’s (2007) findings that the imperfective prefix acannot occur on verbs that express perfective or completed events. Strikingly, Dunham’s examples all use ai+o verbs. Louie (2008: 168) draws similar conclusions, noting that ‘[w]here English encodes relations between sub-events via temporal notions (inner aspect), Blackfoot encodes relations between sub-events via notions of person/ participancy.’1⁰ She observes that Blackfoot lacks alternations in the realization of internal arguments, such as the spray/load alternation or dative alternation. Following Tenny (1987, 1994), she assumes that the spray/load alternation is due to the fact that the direct object bears the event measure role because changes in the object reflect progress through the event. This is illustrated in (19), where the endpoint of the event depends on the choice of direct object. In (19a), the direct object is the wall: when it is completely covered with red paint, the event has reached its telos or endpoint. Similarly, in (19b), the direct object is the red paint: when all the paint has been applied to the wall, this event has reached its telos or endpoint. (19) a. Mary sprayed the wall with red paint. b. Mary sprayed the red paint on the wall.

⁹ Dowty (1977) coined the term imperfective paradox to describe the observation that the progressive form of an accomplishment verb does not entail completion of the event. For example, John was writing a mystery novel does not entail that John wrote a (complete) mystery novel. On the other hand, John was running entails that John ran. This is because accomplishments, such as writing a mystery novel, are telic (bounded) events—writing part of a novel is not the same as writing a novel. However, activities, such as running, are atelic (unbounded) events—any amount of running constitutes a running event. 1⁰ In fact, Louie proposes that the Bloomfield verb classes, which are based on transitivity and animacy, constitute the lexical aspectual verb classes. My proposal shares with hers the basic insight that Blackfoot lexical aspect is determined not by boundedness or telicity, but rather by some notion of animacy (sentience on my analysis, and person/participancy on Louie’s). However, I depart from Louie’s treatment in that I do not assume that the Bloomfield classes are the aspectual classes precisely because I hypothesize that initiating arguments must be both grammatically animate and sentient. It follows that ta verbs with grammatically animate non-sentient objects will belong to the same aspectual class as ti verbs, and similarly that ai verbs with animate non-sentient subjects will belong to the same aspectual class as ii verbs.

72 elizabeth ritter Strikingly, in Blackfoot, the inflection on the verb indicates that it is always the locatum—the object that is sprayed, loaded, stuffed, etc. with the material—that serves as the direct object. Louie interprets this as evidence that telicity is not a determinant of argument structure in Blackfoot. (Translations of the two variants are realized through differences in the order of the locatum and the material but are otherwise identical, as illustrated in (20).) (20)

Context: My sister has her head stuck in between the banister railings. a. Nitoohtohpúúna aná niisis nit-ooht-ohpoon-a an-wa n-iisis 1-means-grease.ta-dir dem-3sg 1-sister iihtaisstsikahkimo’pi. iihtáisstsikahkimo’p-yi11 butter-in ‘I rubbed my sister with butter.’ b. Nitoóhtohpúúna iihtaisstsikahkimo’pi aná nit-ooht-ohpoon-a iihtáisstsikahkimo’p-yi an-wa 1-means-grease.ta-dir butter-in.sg dem-3sg niisis. n-iisis 1-sister Target: ‘I rubbed butter on my sister.’ (Louie 2008: 71)

Louie also develops an argument based on the observation, due to Matthewson and Reis Silva (2007), that all Blackfoot verbs that lack overt markers of viewpoint aspect have a default perfective interpretation. In this respect, Blackfoot differs from languages such as German, Inuktitut, and Russian, where the default interpretation is determined by the telicity of the predicate (Bohnemeyer and Swift 2004). More specifically, Bohnemeyer and Swift observe that unaccusative verbs are typically telic, and interpreted as perfective, while unergative verbs are typically atelic, and interpreted as imperfective. This is not the case in Blackfoot, a fact that Louie interprets as

11 The word meaning ‘butter’ is actually a nominalized verb. For ease of exposition I have not provided the full morpheme breakdown in the examples above. For the interested reader, Louie’s morpheme breakdown and morpheme-by-morpheme gloss are provided here: (i) iiht-á-isstsikahkimo-’p-yi means-ipfv-spread.ti-th-in.sg ‘butter’

sentience-based event structure 73 further evidence that telicity (or boundedness) is not grammaticalized in this language.12 To summarize, both Ritter and Rosen (2010) and Louie (2008) provide a range of arguments in support of the proposal that Blackfoot does not grammaticalize boundedness distinctions. In the next subsection, I review Kim’s (2017b) arguments against this view.

4.3.2 Counter-evidence? Telicity dependent on sentience While the evidence adduced by both Ritter and Rosen and Louie (2008) shows that in many respects Blackfoot differs from languages in which telicity (boundedness) is a significant factor in determining argument structure and event structure, this is a rather surprising finding. Exploring some of the implications of this conclusion, Kim (2017b) argues that telicity is grammaticalized in Blackfoot, but that it is dependent on sentience. More specifically, she shows that if the external argument denotes a sentient being, then the event maybe delimited (bounded) by the direct object or some other delimiter. This seems to be the case with at least two classes of verbs: directed motion verbs and verbs of consumption or creation. I discuss each of these in turn. Kim explores properties of ai directed motion verbs formed with the root oo ‘go’. She shows that when the argument that moves is sentient, a goal DP may be added, but when it is non-sentient, a goal is not possible—even if the head noun of the DP argument that moves belongs to the animate class, as illustrated in (21):13 (21) a.

b.

Anna saahkomaapi itap-oo-wa oomi isspahkoyi. dem boy.an goal-go.ai-3sg dem hill.in ∗

‘The boy went to that hill.’ Anna ainaka’si itap-oo-wa oomi isspahkoyi. dem wagon.an goal-go.ai-3sg dem hill.in ‘The wagon moved to that hill.’

(Kim 2015: 127)

12 Louie (2008) also argues that all AI and II verbs are unaccusative, a consequence of her proposal that the Bloomfield classes are the lexical aspectual classes in Blackfoot. In §4.3.2, I review Kim’s (2017b) convincing evidence that Blackfoot has both unaccusative and unergative intransitives. Whether all verbs are unaccusative or not, the fact that the default viewpoint aspect is not sensitive to boundedness is consistent with the view defended here that this language does not grammaticalize boundedness. 13 Note that the verb root oo ‘go’ requires a prefixal element. Frantz and Russell (2017) characterize the prefix as an adjunct. Kim (2015 et seq.) analyses the required prefix as either a transitive preposition whose DP complement appears elsewhere in the clause, or an intransitive preposition. See §4.5.2 for further discussion.

74 elizabeth ritter Kim interprets this as evidence that boundedness is only possible if a sentient argument is present, and notes that this is a core property of the I-language type in the event-based typology of Ritter and Rosen (2000). This typology distinguishes two types of languages: D(elimitation)-languages, in which eventhood is primarily determined by the terminal bound, and I(nitiation)-languages, in which eventhood is primarily determined by the initial bound. Table 4.3, from Ritter and Rosen (2000: 195), summarizes key differences between these two types of languages. Kim’s claim that Blackfoot is an I-language shares with the current proposal the insight that sentience, rather than telicity, is the primary determinant of event structure. The two proposals differ in that Kim assumes that the Vendler– Dowty event classes (activities, accomplishments, achievements, and states) also apply to Blackfoot, whereas I propose that Blackfoot’s event classes are qualitatively different (active, reactive, interactive, and inert), due to the fact that they are distinguished by the presence or absence of both a sentient initiator and a sentient delimiter. In the introduction to this chapter I provided evidence for this proposal based on a review of a range of constructions in which sentient objects, rather than event measures, obligatorily function as direct objects. Table 4.3 Characteristics of D-languages and I-languages D-languages

I-languages

Accomplishments form a natural class with achievements [in that both have a terminal bound (ER)] Sensitive to semantic and syntactic properties of the object, including: • specificity or definiteness • case-marking

Accomplishments form a natural class with activities [in that both have an initial bound (ER)] Sensitive to semantic and syntactic properties of the subject, including: • agentivity • animacy [i.e. sentience (ER)] • person Make a grammatical distinction between topic and subject Ergative splits on the basis of properties of the subject Subject and object agreement specified for person features Quirky case subjects Animacy hierarchies

Accusative case may be restricted to delimiting objects Ergative splits on the basis of perfective aspect / past tense Object agreement not specified for person features

Source: Ritter and Rosen (2000)

sentience-based event structure 75 As Kim observes, in the context of directed motion verbs, only a sentient argument may serve as an agent. I interpret this evidence as indicating that only a sentient agent is capable of intentional control over the event, and consequently, only a sentient argument has the ability to set a goal. Thus, only a sentient agent argument is compatible with a linker on the verb that licenses a delimiting goal DP. Before turning to the evidence that sentient delimiters are required for bounded events, I briefly discuss Kim’s evidence that challenges the conclusions of Ritter and Rosen and Louie (2008) that the choice between a ta/ti and an ai or ai+o verb does not express a difference in telicity. Adapting a test initially developed by Matthewson (2004) and Bar-el, Davis, and Matthewson (2005), Kim argues that culmination of a ta/ti event is not cancellable, while culmination of an ai event is. This is exactly what one would expect if ta/ti verbs are constituents of telic (bounded) predicates, while their ai counterparts are not. The test that Kim uses can be illustrated with examples from English. As exemplified by the contrast between (22) and (23), when the direct object of a verb of consumption is a definite singular DP that denotes a bounded entity, culmination cannot be cancelled; however, if the object is a bare plural NP that denotes an unbounded entity, culmination can be cancelled: (22) a. John ate the muffin . . . ∗ but there is still some left. b. John ate the muffin . . . ∗ and he is still eating it. (23) a. John ate muffins . . . but there are still some left. b. John ate muffins . . . and he is still eating them. Kim shows that the test also works for Blackfoot. More specifically, she shows that culmination is not cancellable when the predicate consists of a ta/ti verb and definite object, but that it is cancellable when the predicate consists of an ai+o verb and a bare NP object. However, Kim also shows that quantified objects, such as ‘two (cups of) water’ are possible with both ta/ti and ai+o verbs, and that the same difference in cancellability of culmination is observable with these objects. In other words, when a quantified object appears with a ta/ti verb, culmination is not cancellable (24), but when a quantified object appears with an ai verb, culmination is cancellable (25):

76 elizabeth ritter (24) a.

Anna saahkomaapi ii-simato-m-yi anni-istsi dem boy.an ic-drink.ti-th-3pl dem-in.pl aohkii-istsi . . . water-in.pl ‘That boy drank those two cups of water . . . ’ b. # ki saaki-ohksisstaki-yi-aistsi. and still-have.leftover.ii-3pl-3pl.pron ‘ . . . but they are still left.’ c. # ki annohk saaki-a-iisimi-yi. and now still-ipfv-drink.ai-3pl ‘ . . . but he is still drinking them.’

naato’k two

(25) a.

Anna saahkomaapi ii-simi-wa (naato’k-istsi aohkii-istsi) . . . dem boy.an ic-drink.ai-3sg two-in.pl water-in.pl ‘That boy drank (two cups of water) . . . ’ ki saaki-ohksisstaki-yi-aistsi. and still-have.leftover.ii-3pl-3pl.pron ‘ . . . but they are still left.’ ki annohk saaki-a-iisimi-yi. and now still-ipfv-drink.ai-3pl ‘ . . . but he is still drinking (them).’ (Kim 2017b: (ex. 41–2))

b.

c.

At first glance, this is a very surprising result because a quantified object is bounded. However, on reflection, I believe that it points to a problem with the diagnostic used. Observe that in English, a verb of consumption with a quantified object passes standard telicity tests (26), but the results of the culmination cancellability test depend on whether the quantified object represents all of the available consumables, or just a portion of them (27). (26) a. John drank two cups of water {in five minutes / # for five minutes}. b. John finished drinking two cups of water. c. John almost drank two cups of water. i. ✓ John almost started drinking two cups of water. ii. ✓ John almost finished drinking two cups of water. d. John was drinking two cups of water. ≠ John drank two cups of water.

sentience-based event structure 77 (27) a. John drank two cups of water . . . but there are still {two cups of water / # those two cups of water}. b. John drank two cups of water . . . and he is still drinking {cups of water / # them}. A closer examination of the examples in (24) and (25) shows that the ti verb ‘drink’ requires a definite object, ‘those two (cups of) water’, but its ai counterpart requires an indefinite object, ‘two (cups of) water’. Arguably, in the first case the object represents all the available water, but in the second case it does not. Thus, the results of this test should not be construed as evidence that ta/ti verbs may be used to denote telic events, while ai verbs may not, or that boundedness of the object plays a role in event structure.

4.3.3 Sentience and argument structure alternations In the last subsection, I pointed out that the difference between the current proposal and the one defended in Kim (2017b) is in the semantic content of the object that serves as an event delimiter. In this subsection, I discuss evidence for my hypothesis that what matters is whether the delimiting object is sentient. As I observed in the introduction to this chapter, the claim that sentient objects are special in Blackfoot is well documented in the literature. Here I review additional evidence demonstrating that sentient objects are grammatically distinguished from other objects, and suggest that this distinctive behaviour is a reflex of their event role. Recall that verbs in Blackfoot agree with both the subject (external argument) and object (internal argument). Following Bliss (2007, 2010), I adopt Dryer’s (1986) term primary object for the object that triggers agreement on the verb. For transitive verbs, the primary object is always the direct object. For ditransitive verbs, the primary object is the indirect object just in case it is sentient, and the direct object otherwise. Since all sentient nominals are grammatically animate, all ditransitive verbs that agree with their sentient indirect object belong to the ta class. Bliss (2007, 2010) demonstrates that while ti and ai+o verbs are always transitive, their ta counterparts are alternatively transitive or ditransitive.1⁴ 1⁴ Louie’s (2008) observation that Frantz and Russell’s (2017) Blackfoot dictionary of stems, roots, and affixes lists many ditransitive TA verbs of transfer of possession but no TI verbs of this type provides additional support for this claim.

78 elizabeth ritter This is illustrated with the verb ‘borrow’ which has a ti form that is used with an inanimate direct object (28a), as well as a ta form which is alternatively used as a transitive verb with a grammatically animate direct object as its primary object (28b), and as a ditransitive verb with a grammatically animate—and sentient—indirect object as its primary object (28c):1⁵ (28) a. Nitaahkóma’tatoo’p amo isttokimaa’tsis. nit-waahkóma’t-atoo-o’p amo isttokimaa’tsis 1-borrow-ti-th dem drum.in ‘I borrowed this drum.’ b. Anahk Sam aahkom’atatsii nitsinaka’simiksi. anwa-hk Sam waahkom’at-at-yii nit-inaka’simiksi dem-invis Sam borrow-ta-dir 1-car.an ‘Sam borrowed my car.’ c. Nitaahkoma’takka ana Sam nitsinaka’simiksi. nit-waahkoma’t-at-ok-wa an-wa Sam nit-inaka’simiksi 1-borrow-ta-inv-sg dem-sg Sam 1-car.an ‘Sam borrowed from me my car.’ (Bliss 2010: 5–6) The primary object of a monotransitive ta verb is normally a selected object that bears the thematic role of theme or patient.1⁶ However, the primary object of a ditransitive ta verb bears either the thematic role of goal (29a), or a noncore thematic role, such as source (28c), or malefactive (29b).1⁷ Evidence that it is the goal and not the theme that functions as the primary object in (29a) can be gleaned from the fact that the verb bears a proximate suffix signaling that it agrees with the proximate-marked animate goal DP, ‘my mother’ (rather than the obviative-marked animate theme DP, ‘the baby’).

1⁵ Examples from Bliss have been modified in two respects: (i) I have omitted the internal morphological structure of demonstratives where it is not required to determine which DP is the primary object and (ii) I have included the animacy specification of all object common nouns. 1⁶ As pointed out by an anonymous reviewer, cross-clausal agreement (CCA) constitutes an exception to this generalization. When CCA occurs, an argument (subject or primary object) triggers agreement both on the verb of the clause in which it is Merged, and on the verb of the immediately dominating clause. Bliss (2008) provides compelling evidence that CCA signals that the DP which triggers agreement in the higher clause is contrastively focused. Interestingly, all her examples involve DPs that denote human beings. Whether this is necessarily the case, or whether CCA affects the event structure of the higher clause are questions I leave for future research. 1⁷ Bliss (2007) refers to these as non-thematic roles. However, since source, benefactive, and malefactive are thematic roles that are not determined—or not required—by the verb’s meaning, I follow Bliss (2010) and refer to them as non-core thematic roles.

sentience-based event structure 79 (29) a. Niksissta nitóhkotawa n-iksisst-wa nit-ohkot-a-wa 1-mother.an-prox 1-give.ta-dir-prox ‘My mother, I gave her the baby.’ b. Nitsííkamo’satoka ana Rosie nit-ikamo’s-at-ok-wa an-wa Rosie 1-steal-ta-inv-prox dem-prox Rosie ‘Rosie stole from me that money.’

anni issitsímaan. ann-yi issitsimaan dem-obv baby.an

(Bliss 2007: 4) ihtophomo’pi. ihtophomo’pi money.in

(Bliss 2007: 5)

Note that in these examples the direct object differs from the primary object in person, number, obviation, and/or animacy, and the verb agreement indicates that it is the sentient indirect/non-core object, and not the direct object, that serves as the primary object. Frantz (2017) describes three different final suffixes that introduce non-core arguments and appear at the right edge of the verb stem (or root).1⁸ Strikingly, all three of these finals license the addition of a sentient nominal, and derive ta verbs. In §4.2, I discussed the first of these valence-increasing suffixes: the causative final -attsi, which adds a sentient causer. There I reviewed evidence originally adduced by Johansson (2009) that causative events involving both a sentient causer and a sentient causee are complex, while those with a nonsentient causee are not. I also argued that the sentient causee in -attsi derived causatives serves as an event-delimiting (primary) object. I now suggest that this is the case for all sentient primary objects. The second valence-increasing suffix that Frantz discusses is the benefactive suffix -omo. This suffix derives a ditransitive verb by adding a new primary object to a basic transitive verb. Bliss (2010) shows that, as with other ditransitive ta verbs, the thematic role of this primary object is variable; she provides examples with a benefactive, a goal and a source as the primary object of a ta verb with -omo. Some of her examples are reproduced here.1⁹ (30) a. Nitsskítomowa ana niksísst. nit-ihkit-omo-a an-wa n-iksisst 1-bake-ta.ben-dir dem-prox 1-mother ‘I baked for my mother.’

Benefactive

1⁸ In traditional Algonquianist literature these are called concrete finals. Frantz (2017) also discusses two concrete finals that suppress arguments: the reflexive and reciprocal finals. Significantly, the arguments that are suppressed must also be sentient. 1⁹ As noted in §4.1, Bliss (2007, 2010) also demonstrates that the primary object introduced by -omo must be sentient. As illustrated by the contrast between (5) and (6), if the sentence contains a nonsentient goal, it is the theme/patient that is realized as the primary object, and the goal is expressed as an optional adjunct. The same is true for non-sentient DPs bearing the thematic role of source. (The theta roles of benefactive and malefactive are not available to non-sentient DPs.)

80 elizabeth ritter b. Ana Leo nitsapóhtomoka. Goal an-wa Leo nit-sapoht-omo-ok-wa dem-sg Leo 1-give.tobacco-ta.ben-inv-3sg ‘Leo gave me some tobacco.’ c. Ana Rosie nito’tomoka nitsinika’simiks. Source an-wa Rosie nit-o’t-omo-ok-wa nit-inika’simiks dem-sg Rosie 1-take-ta.ben-inv-3sg 1-car.an ‘Rosie took my car from me.’ (Bliss 2010: 3) Bliss analyses the benefactive as an applied suffix, and proposes that the syntactic structure of a predicate that contains such a suffix is more complex than that of the related transitive verb due to the addition of ApplP. I analyse ApplP as an eventive functional category that licenses a sentient delimiter which participates in a complex (interactive) event. The third valence-increasing suffix that Frantz discusses is the accompaniment suffix -m. Like the benefactive suffix, -m derives ta verbs, and licenses a primary object that is both grammatically animate and sentient. In this case, the new primary object always bears the non-core thematic role of companion. Whenever the verb bears the accompaniment suffix, it also contains the prepositional prefix ohpok- ‘with’, as illustrated in (31):2⁰ (31)

Nitána iihpoká’po’takiimiiwa nohkóyi. n-itan-wa ii-ohpok-a’po’taki-m-ii-wa n-ohko-yi 1-daughter-3sg ic-with-work.ai-ta.accm-dir-3sg 1-son-4sg

‘My daughter worked with my son.’

(Frantz 2017: 116)

It has been observed that in languages where boundedness plays a role in event structure, unergative manner of motion verbs, such as dance cannot be transitivized by adding a companion object, unless a delimiting phrase is also added (e.g. Levin and Rappoport-Hovav 1995; Rosen 1996). This is illustrated

2⁰ Two modifications have been made to Frantz’s morphemic analysis and glossing: First, Frantz analyses the first segments of the verb as a portmanteau morpheme, iihpok-, and assumes that it comprises past tense marking and the accompaniment linker. I have divided this into two morphemes ii- and ohpok-. The first is labelled I(nitial) C(hange), an Algonquianist term that indicates that it triggers ablaut of the first vowel that follows it. Following Ritter and Wiltschko (2014), I assume that Blackfoot lacks grammatical tense marking. Given this assumption, I analyse initial change as an adverbial prefix that expresses past time reference. Second, Frantz labels the verbal suffix -m simply as ta, but I have labelled it as ta.accm to indicate that this morpheme is the valence-increasing morpheme that derives accompaniment stems.

sentience-based event structure 81 with an English example in (32). Strikingly, Blackfoot verbs like ihpiyi do not require a delimiting phrase. Rather, they may be transitivized by adding the accompaniment suffix (and prepositional prefix ohpok-), as shown in (33). The only constraint is that the primary object denote a sentient being. (32) a. b. (33)



John danced Mary. John danced Mary across the room.

Na Maanikapi áísstsimohpokihpiyiimiiwa ann-wa Maanikapi a-sstim-ohpok-ihpiyi-m-yii-wa dem-prox.sg Maanipaki ipfv-reluctantly-with-dance-ta.accm-dir-3sg ni Ikkstaakii. ann-yi Ikkstaakii dem-obv.sg Ikkstaakii

‘Maanikapi dances with Ikkstaakii reluctantly.’

(Meadows 2010: 46)

As noted in §4.1, Meadows (2010) has demonstrated that the companion always denotes a DP that willingly performs the same action as the agent. Companions are grammatically distinguished from associates, which are adjunct nominals that lack volitional control over their actions. Associates are licensed by a different prepositional prefix on the verb (omohp-/ohp-); cf. Frantz (2017) and Meadows (2010). Unlike non-core arguments, including companions, they may be either sentient or non-sentient, either grammatically animate or inanimate, associated with either the agent or patient/theme, and occur with verbs of any class. Moreover, they never trigger agreement on the verb or affect the choice of the verb class morphology. Consequently, following Meadows, I analyse associates as adjuncts, rather than non-core arguments of the verb. Moreover, I suggest that, unlike companions, they play no role in event structure. Some representative examples are given in (34): (34) a. Nomohpoowatoo’pa omi n-omohp-oowatoo-’p-wa om-yi 1-with-eat.ti-th-3sg dem-in.sg miistsikápa’siksi. miistsikapa’s-iksi carrot-an.pl ‘I ate that meat with those carrots.’

i’ksisakoyi i’ksisako-yi meat-in.sg

omikisi om-iksi dem-an.pl

(Meadows 2010: 82)

82 elizabeth ritter b. Nomohpinoyiiwa na Mianni n-omohp-ino-yii-wa ann-wa Mianni 1-with-see.ta-dir-3sg dem-3sg Mianni ni’siksi. n-i’s-iksi 1-brother-an.pl ‘I saw Mary with my brothers.’ c. Iihpiito’too ni otsitapíím. ii-ohp-iit-o’too ann-yi ot-atapíím ic-with-loc-arrive.ai dem-4sg 3-doll ‘She arrived with her doll.’

niksi ann-iksi dem-an.pl

(Meadows 2010: 86)

(Meadows 2010: 83)

Summarizing, I have shown that there is ample evidence that nominals that denote sentient beings are special in Blackfoot. Both causers and causees of morphologically derived causative verbs are subject to a strict sentience requirement. Similarly, sentient sources, goals, benefactives, malefactives, and companions are all realized as primary objects; nonsentient nominals bearing these roles are realized as adjuncts. There is ample evidence for their status as primary objects: Unlike associates and other adjuncts, (a) they only occur with ta verbs; (b) they trigger verb agreement; and (c) their animacy and person specification relative to that of the external argument determines whether the verb bears a direct or inverse theme marker. I submit that what makes these sentient primary objects special is that they serve as delimiting objects—not because they are event measures, but rather because they are cognizant of the action (and intention) of the external argument. In other words, sentient benefactives, malefactives, sources, goals, companions, and causees all knowingly participate in the event, and they all have the capacity to recognize when the event is over. I suggest that this is formalized with the addition of the formal feature [m] to the DPs that bear these roles, and that this feature is checked by an eventive functional category that selects for [m]. In all cases, these sentient primary objects co-occur with a sentient causer or agent. I propose that these arguments are participants in an interactive event, i.e. an event that requires two sentient arguments, each selected by a different eventive functional category, as schematized in (9a), reproduced as (35):

sentience-based event structure 83 (35)

Interactive situation FP1 DP [m] F1 u[m]

FP2 DP [m] F2 u[m]

vP

4.3.4 Summary This section has argued that the evidence bears out my second prediction—that boundedness of the object plays no role in distinguishing between event types. I reviewed the arguments from Ritter and Rosen (2010) and Louie (2008) in support of this position. I also argued that Kim’s apparent counter-evidence that event measures and non-sentient goal PPs serve as event delimiters is not compelling. Event measures serve as delimiters in languages like English where bounded events require an object that denotes a bounded entity (or a bounded path). The claim that boundedness plays no role in the classification of events raises the question of whether it also plays no role in the classification of entities. In the next section I explore this possibility.

4.4 Sentience, boundedness, and entity delimitation This section explores the third prediction listed in §4.1: Prediction 3: If event delimitation is determined by sentience, rather than boundedness, then sentience, rather than boundedness, may also play a role in distinguishing between nominal types. Although relatively little work has been undertaken on the event-related properties of Blackfoot nominals, the research that explores this question takes the position that animacy, rather than boundedness, underlies the aspectual

84 elizabeth ritter classification of entities. In this section, I review the arguments that led to this conclusion, first summarizing the evidence that Blackfoot nominals lack a grammaticalized boundedness distinction (§4.4.1), and then summarizing the evidence that Blackfoot nominals have a grammaticalized animacy distinction that serves the same interpretive function as the grammaticalized boundedness distinction in languages like English, i.e. this is the content of nominal Inner Aspect (§4.4.2). Finally, I discuss the relationship between sentience, animacy and event structure in the context of the current proposal (§4.4.3). Before turning to the discussion of boundedness and animacy in Blackfoot nominals, let me briefly explain the rationale behind this third prediction. It has been frequently observed that there is a close parallel between the countmass distinction and the telic–atelic distinction in English (e.g. Mourelatos 1978; Bach 1986; Jackendoff 1991). Jackendoff proposes that both singular count nouns and telic events are [+bounded], while bare plural or mass nouns and atelic events are [−bounded]. Wiltschko (2012) proposes that this parallel arises because in both cases the [±bounded] feature serves as the head of the functional category, Inner Aspect, which is part of both the nominal and verbal spines. Wiltschko’s proposal provides an elegant syntactic treatment of this semantic fact. However, the question arises as to why there should be this syntactic symmetry. In other words, why should the nominal and verbal spines share the category and content of Inner Aspect, particularly since other functional categories in the two spines differ both in their label and their feature content? The answer may lie in another widely-discussed connection between nominal and verbal inner aspect: the fact that a telic predicate consists of a verb of a particular lexical semantic class and a bounded object (e.g. build a house is telic, but build houses is not). A number of researchers have observed the special relationship between the temporal structure of the event and the bounded object of so-called incremental theme verbs (Hinrichs 1985; Tenny 1987, 1994; Verkuyl 1993; Krifka 1998), as well as other classes of accomplishment verbs, including change of state verbs and goal directed motion verbs (Ramchand 1997; Krifka 1998; Harley 2003). As Tenny and Pustejovsky (2000: 12) explain, ‘[i]ncremental theme verbs, which have direct objects that are consumed or created in increments over time, as in drink a glass of wine, can be represented as a homomorphism from objects to events’. The reason for this is that changes in the direct object are commensurate with progress through the event. For example, when half the glass of wine has been consumed, half of the event has been completed,

sentience-based event structure 85 when the entire glass of wine has been consumed, the entire event has been completed.21 Harley (2003) extends this object-to-event homomorphism to other classes of verbs. In particular, she proposes that it underlies lexical aspectual differences among denominal verbs, and supports her proposal with the following evidence: Bounded (count) noun roots, such as whelp, foal, and hop serve to derive telic verbs, while unbounded (mass) noun roots, such as spawn, drool, and sweat derive atelic verbs. Following Hale and Keyser (1993), she assumes that all these denominal verbs are derived by a syntactic incorporation process, and attributes the object-to-event homomorphism effect to the fact that prior to incorporation, these noun roots occupy the same syntactic position as incremental themes. Clearly if the object-to-event-homomorphism is due to the part structure of both the event and the object, we would not expect to find the same homomorphism in a language with sentience-based aspectual classes of verbs. My hypothesis is that sentient primary objects serve as event delimiters in Blackfoot because they have the cognitive capacity to recognize the intentional action of the initiating agent. My third prediction derives from the observation that a sentient DP object requires a nominal whose lexical head is a grammatically animate noun.22 This animacy requirement for sentient DPs is reminiscent of the requirement that bounded DP objects have a count noun as their lexical head.

4.4.1 Nominal I-Asp is not specified for [±bounded] Wiltschko (2012) argues that the nominal aspectual distinction grammaticalized in Blackfoot is animacy, rather than boundedness, based on the absence of a grammaticalized count–mass distinction in the language, and the presence of a grammaticalized animate/inanimate distinction. Her reasoning is as follows: A grammaticalized count–mass distinction arises when nominal Inner Aspect is specified for the feature [±bounded]; absence of a grammaticalized count–mass distinction may arise if Inner Aspect is either not present, or not 21 See Krifka (1992, 1998) for a formal semantic analysis of this object-to-event homomorphism in terms of part structure. 22 I am abstracting away from descriptions of fictional worlds in which normally non-sentient objects (e.g. teapots or flowers) are capable of thought, feeling, and action. In Blackfoot, this requires semantic type shifting of the nominal, but is not accompanied by a shift in morpho-syntactic class from inanimate to animate (cf. Johansson 2008).

86 elizabeth ritter associated with [±bounded]. She offers several arguments to support the claim that Blackfoot’s Inner Aspect is not associated with [±bounded].23 First, it appears that all Blackfoot nouns can be pluralized, regardless of whether they denote bounded individuals or unbounded substances. As noted in §4.3, it is the shape of the plural suffix that serves as the most reliable diagnostic of noun class in this language. Significantly, in the introduction to their Blackfoot dictionary of stems, roots, and affixes, Frantz and Russell (2017: xvi) describe the category of noun as ‘[stems] taking the plural suffix -iksi or -istsi’. Some of the entries for nouns include the plural form, others do not, but this is not determined by the meaning of the noun: some entries for nouns that denote substances include plural forms, and some entries for nouns that denote individuals do not. In some cases, Frantz and Russell add a parenthetical comment about how to interpret the plural form for substance nouns. For example, in the entry for aiksinoosak, an animate noun meaning bacon, Frantz and Russell (2017: 8) include its plural form, áiksinóósakiksi, followed by the English translation of the plural, which they list as ‘bacon (slabs or slices of)’. Second, as Wiltschko observes, there are no classifiers in Blackfoot. Rather, numerals are combined directly with nouns, regardless of whether they denote individuals or substances. In the case of inherently unbounded substance nouns, the basis for countability is contextually determined. For example, natokayi a’apannistsi ‘two bloods’ can refer to two drops of blood, two bags of blood for transfusion, two pools of blood, etc. Third, there are no determiners or quantifiers that occur only with count or mass nouns.2⁴ The language has no definite or indefinite articles, and all demonstrative stems may occur with any noun. As for quantifiers or quantifier-like elements, there are two types—adjectival quantifiers that are prefixed to the noun stem (36), and adverbial quantifiers that are prefixed to the verb stem (37).

23 See also Louie (2008: 164–5) for the suggestion that Blackfoot lacks a mass–count distinction, due to the absence of semantic distinctions of nominal quantity (Borer 2005) or quantization (Krifka 1998). Louie views this as a plausible hypothesis, given that there seem to be no semantic (or syntactic) distinctions of verbal quantity or verbal quantization in the language. Quantity, quantization, and boundedness are similar notions in that they all distinguish between events and entities whose subparts constitute events and entities of the same type and those that do not. The former are non-quantized, homogenous, and unbounded; the latter are quantized, non-homogenous, and bounded. 2⁴ Demonstratives are inflected for proximity to the utterance situation, number, and obviation (i.e. discourse salience of the nominal). Bliss (2013) argues that demonstratives are not determiners. If this is the case, then the language has no words that belong to the morphosyntactic category determiner.

sentience-based event structure 87 (36) a. i. i’náksikóónksko i’nak-koonssko small-snow ‘a little bit of snow’ b. i. i’náksipokaa i’nak-pokaa small-child ‘baby’

ii. i’nákaohkíí iii. i’nákónnikis i’nak-aohkii i’nak-onnikis small-water small-milk/breast ‘a little bit of water’ ‘a little bit of milk’ ii. i’nákánao’kssi iii. i’nákónnikis i’nak-ánao’kssi i’nak-onnikis small-half.dollar small-milk/breast ‘a quarter of a ‘a small breast’ dollar’ (adapted from Wiltschko 2012: 152)

(37) a. Nitáótahkoinammiksi ákaonnatohsimmi. nit-aotahkoinamm-iksi akaa-onnat-ohsi-m-yi 1-orange-an.pl pfv-less-be(come).ai-3pl ‘My oranges, they have become few.’ b. Nitáóhkiimi ákaonnatohsiiyi. nit-áóhkii-m-yi akaa-onnat-ohsiiyi-wa 1-water-pssd-in.sg pfv-less-be(come).ii-3sg ‘My water supply is low.’ (adapted from Frantz and Russell 2017: 198) As with numerals, the interpretation of quantifiers is contextually determined. For example, inák- means small in quantity when it is attached to a substance noun, but small in size or some other relevant property when it is attached to a count noun. This is illustrated in (36) by the contrast between the inanimate noun meaning milk and the animate noun meaning breast, both of which have the same stem, ónnikis. When inák- ‘small’ is added to the inanimate stem, the result is a small quantity of milk, but when it is added to the animate stem, the result is a body part that is small in size. When inákis attached to other substance nouns, e.g. ‘snow’ and ‘water’, the interpretation is consistently a small amount, but when it is attached to other nouns denoting bounded individuals, the interpretation is much more variable: when inák- is added to ‘child’, the result is a young child, i.e. a baby, and when it is added to half dollar, the result is a coin that is small(er) in value, i.e. a quarter. We see the same variability in interpretation with the verbal prefix onnat-, which Frantz and Russell (2017: 198) define as ‘few, less, low in number’. Their examples of related animate intransitive (ai) and inanimate intransitive (ii) verbs meaning ‘be [or become (ER)] few, low/diminish,’ which are reproduced in (37), indicate that they may be used to characterize lessening of the quantity of either substances or bounded individuals.

88 elizabeth ritter Finally, Wiltschko observes that in English, a typical mass–count language, bare NP arguments are possible if the noun is [−bounded], i.e. a mass or bare plural argument. Blackfoot also has restrictions on bare NP arguments, but they are based on a property of the verb, not the argument: transitive verbs, whether they belong to the transitive animate (ta) or transitive inanimate (ti) class, never occur with bare NP objects. However, ai verbs sometimes appear with an object, and when they do, their objects are bare NPs—either bare nouns that are unmarked for number, or bare plural nouns; cf. Bliss (2013); Kim et al. (2017).2⁵ As the examples in (38) illustrate, objects of ai verbs are interpreted as non-specific while those of ta and ti verbs are specific. (38) a. Nit-oowatoo-’p-yi ann-istsi aipasstaam-istsi. 1-eat.ti-th-3pl dem-in.pl apple-in.pl ‘I ate those apples.’ b. Nit-oowatoo-’p-wa ann-yi aipasstaam. 1-eat.ti-th-3sg dem-in.sg apple.in ‘I ate that apple.’ c. Nit-ooyi aipasstaam. 1-eat.ai apple ‘I ate (one or more) apples.’ (adapted from Kim et al. 2017: 5) Summarizing, the evidence adduced in this section provides strong support for Wiltschko’s (2012) proposal that the count–mass distinction is not grammaticalized in Blackfoot. Assuming that the count–mass distinction is due to the presence of [± bounded] in Inner Aspect, we can conclude that Blackfoot nominal Inner Aspect lacks this feature.

4.4.2 Nominal I-Asp is specified for [±animate] As noted in the last subsection, Wiltschko (2012) proposes that languages that lack a mass–count distinction may either lack Inner Aspect altogether, or they may associate a different formal feature with this function category. 2⁵ Frantz (2017: 12–4) uses the term non-particular to describe these bare NP objects that are either non-referential or non-specific in reference. He describes non-particular nominals as lacking either a demonstrative or possessive prefix, and as having a non-particular suffix -i. Neither Kim et al.’s consultants nor Bliss’s use the non-particular suffix, so I have not included it in the morphemic analysis in these examples.

sentience-based event structure 89 She proposes that Blackfoot belongs to the second type of language, and more specifically that Blackfoot Inner Aspect is associated with [±animate]. Below, I summarize two of the arguments adduced in support of this proposal.2⁶ The fact that only nouns that denote bounded entities may have both singular and plural forms in a count–mass language like English may be interpreted as evidence of an interaction between [±plural] in the functional head Num(ber) and [±bounded] in Inner Aspect (I-Asp). This is not the case in Blackfoot where virtually all nouns may be pluralized. However, there is evidence that [±animate] interacts with the contents of both Num and a higher functional head of the nominal spine. As noted above, the form of the plural marking depends on the animacy specification of the noun. Wiltschko suggests that this can be interpreted as an Agree relation between Num and I-Asp, which is specified for [±animate]. Similar facts obtain between the form of plural marking and gender in some languages (e.g. Hebrew, Greek), so this fact only tells us that [±animate] is a formal grammatical feature associated with a head that is in the local domain of Num. Ritter (2014) adds a compelling argument for Wiltschko’s hypothesis that Blackfoot animacy is not a kind of gender, based on the observation that the inflectional suffix of [+animate]—but not [−animate]—nouns is specified for obviation (in addition to number and animacy). Animate nominals that are discourse salient are [+proximate]; those that are not are [−proximate], i.e. obviative.2⁷ The full paradigm of nominal inflectional suffixes is given in Table 4.4. Observe that Blackfoot has both singular and plural suffixes, that the form of the plural suffix depends on the animacy specification, and that the form of the animate singular suffix depends on the obviation specification. The fact that animate nouns have a more complex inflectional paradigm is completely unexpected if [±animate] is a gender feature. I am not aware of any language where nouns of one gender are more highly specified than others, e.g. masculine nouns are inflected for case, but feminine nouns are not. On the other hand, this fact is unsurprising if [±animate] is an aspectual feature: Just as only [+bounded] nouns may be additionally specified for 2⁶ Some of Wiltschko’s arguments that animacy is an aspectual property are based on differences between German gender and Blackfoot animacy. However, not all the properties of German she considers generalize to all gender languages. For example, she notes that gender-shifting is used as a derivational strategy in Blackfoot, but not in German. This argument is significantly weakened by the fact that gender-shifting is used as a derivational strategy in Hebrew, another language with grammatical gender (Bat-El 1986; Ritter 1993). For this reason, I consider only arguments that are based on contrasts between Blackfoot animacy and most—if not all—gender languages. 2⁷ In possessed nominals obviation is structurally, rather than pragmatically, determined. If the possessor is a 3rd person nominal and the possessed is [+animate], then the possessed must be [−proximate]; cf. Frantz (2017: 15).

90 elizabeth ritter Table 4.4 Nominal inflectional suffixes Number

[−plural] [+plural]

Obviation [+proximate] [−proximate] —

Animacy [+animate] -wa -yi -iksi

[−animate] -yi -istsi

[±plural] in English, only [+animate] nouns may be additionally specified for [±proximate] in Blackfoot. Another striking fact about the feature [±animate] is that it underlies the classification of verbs as well as nouns. As discussed in the introduction to §3, the standard classification system for verbs in Blackfoot and other Algonquian languages, consists of four classes, based on transitivity and animacy of a designated argument (Bloomfield 1946). Importantly, the animacy status of verbs is not just a form of gender agreement (Louie 2008; Wiltschko 2012; Ritter 2014). There are at least three pieces of evidence that support this claim that animacy plays a role in verb classification. First, the verb’s animacy classification is determined by a morpheme that is a constituent of the verb stem, not an inflectional suffix. (Agreement suffixes are inflectional suffixes, and not part of the verb stem.) Second, the verb’s animacy classification contributes to the determination of its argument structure. More specifically, only animate intransitive (ai) and transitive animate (ta) sometimes select an additional argument, so they are notionally transitive and ditransitive, respectively, as discussed in §3. Recall that ai+o verbs are animate intransitive verbs that select an indefinite object (ai+o verbs). Inanimate intransitive (ii) verbs never select an indefinite object, i.e. there are no ii+o verbs. Recall also that transitive animate (ta) verbs differ from their transitive inanimate (ti) counterparts in that they may be either transitive or ditransitive. Third, in some Algonquian languages, including Blackfoot, ta verbs can select grammatically inanimate objects and ai verbs can select grammatically inanimate subjects just in case these nominals denote anthropomorphized individuals (Goddard 2002; Johansson 2008).2⁸ Therefore, [±animate] is an intrinsic morpho-syntactic feature of both nouns and verbs in Blackfoot. This is problematic for the characterization of 2⁸ The first two pieces of evidence mentioned here are discussed at greater length in Ritter (2014); the third piece of evidence has not been previously used to motivate the claim that [±animate] is an aspectual, rather than a gender feature in Blackfoot.

sentience-based event structure 91 [±animate] as gender, since gender is an intrinsic property of nouns but not verbs (Hockett 1958; Corbett 1991). However, it is consistent with the view that [±animate] constitutes the interpretable content of nominal (and verbal) Inner Aspect. In the next subsection I explore the relationship between animacybased aspectual classification and sentience-based event structure.

4.4.3 Sentience, animacy, and event structure In §§4.1–4.3, I argued that Blackfoot has sentience-based event structure. In other words, eventive situations involve participants that have the mental capacity for cognitive action and/or response. I propose that this is formalized in the syntactic representation by initiating and/or terminating eventive functional categories that select for sentient DPs, and that selection for sentience is formalized by means of [m], an uninterpretable sentience feature on the functional head. Then in §§4.4.1 and 4.4.2, I reviewed the evidence in support of the assumption that Blackfoot has a functional category, Inner Aspect, in both the nominal and verbal spine whose head has an interpretable animacy feature [±animate]. The goal of the current subsection is to articulate the relationship between [m] and [±animate]. The discussion above leaves no doubt that [±animate] is a morphosyntactic head feature in Blackfoot. In the nominal spine, it determines the form of plural marking and the availability of obviation marking (through Agree). In the verbal spine, it is reflected in the form of transitive and intransitive verbs, and determines the grammatical animacy of their objects and subjects, respectively.2⁹ Sentience, on the other hand, is a phrasal feature that is semantically and pragmatically determined. It is attached to any DP that denotes a sentient individual. In the unmarked case, these are DPs whose Inner Aspect is [+animate]. However, there are exceptions in both directions—DPs that denote non-sentient entities, including dead bodies, have [+animate] Inner Aspect, but do not bear [m]; on the other hand, DPs that denote anthropomorphized objects have either [+animate] or [−animate] Inner Aspect, and do bear [m]. In other words, animacy determines the classification of nouns, whereas sentience determines the classification of DPs. The proposal

2⁹ As noted in §4.4.2, there is one exception to this generalization: transitive animate verbs may select [−animate] primary objects in descriptions if they are used to denote anthropomorphized objects that can think, feel, and act.

92 elizabeth ritter defended in this chapter is that [m] is a selectable feature, and, in particular, that both initiating and terminating eventive functional categories select DPs specified for [m] as event initiators and terminators, respectively.

4.5 Dynamicity, sentience, and eventiveness In this section I briefly consider facts that bear on my fourth and final prediction, repeated below: Prediction 4: If only sentient arguments may bear event roles, then inert situations, which have no sentient arguments, are not eventive in the relevant sense. We should expect, then, that eventiveness is not correlated with dynamicity. In English and other languages whose aspectual structure has been widely discussed in the literature, there is a fundamental distinction between events, which are dynamic situations, and states, which are not. A dynamic situation is one that involves a process or change, or both. If Blackfoot event structure is defined in terms of intentional action and sentient response, then all that is required is at least one sentient argument. For example, verbs such as love, hate, envy, and ache, which require a sentient experiencer, would constitute s(entient)-events, even though they are non-dynamic. Similarly, we should expect to find that intransitive verbs with non-sentient arguments systematically have both stative and change-of-state (inchoative) uses. In order to determine whether this prediction is borne out, I explore the semantic content of inanimate intransitive (ii) and animate intransitive (ai) verbs. ii verbs have subjects that are both notionally and grammatically inanimate; ai verbs, on the other hand, have grammatically animate subjects, which may be either notionally animate, e.g. saahkómaapi ‘boy’, or notionally inanimate, e.g. po’táa’tsis ‘stove’. I expect that ii verbs will never denote s-events, but should have both dynamic and stative uses, and that ai verbs will have a split behaviour. They may denote (non-dynamic) s-events when they are predicated of a sentient, grammatically animate argument, and otherwise will pattern with ii verbs.

4.5.1 Verb classification and verb meaning Frantz and Russell’s (2017) Blackfoot dictionary of stems, roots, and affixes uses Bloomfield’s classification system for verbs. It contains both definitions and

sentience-based event structure 93 illustrative examples for each entry. An examination of the ii verbs listed, suggests that verbs of this class may be used to describe both stative situations and anti-causative dynamic situations involving a single inanimate (and nonsentient) entity, as exemplified with the verbs ihkitsi ‘dry’ and iitsskihkaa ‘be matted’ in (39) and (40), respectively: (39) a. stative: b. anti-causative:

ákaihkitsiwa kaapoksíínimaani áakihkitsiwa

‘the floor is dry’ ‘it will dry’

(Frantz and Russell 2017: 27) (40) a. stative:

áakiitsskihkaawa ‘it will be matted’ ko’tokááni ákaiitsskihkaawa ‘Your hair has become matted.’

b. anti-causative:

(Frantz and Russell 2017: 39) We should also expect to find a meaning shift between related ii and ai verb stems, but only if the ai verb is predicated of a sentient argument. Listings for pairs of ii and ai verbs in Frantz and Russell (2017) suggest that this prediction is also borne out: (41) a. iiyikoo b. iiyikoosi

ii ai

‘be difficult, hard (intensely)’ ‘have a difficult time / be difficult’ (Frantz and Russell 2017: 40)

(42) a. ipahtsa’pii b. ipahtsá’pssi

ii ai

‘be a mistake/error’ ‘err, make a mistake’ (Frantz and Russell 2017: 79)

Note that the ai verb in (41b) may be predicated of a sentient experiencer while the one in (42b) is predicated of a sentient, but non-intentional actor. Adapting Wiltschko’s (2014) analysis of limited control in Halkomelem, I suggest that iiyikoosi, when it means ‘have a difficult time’, and ipahtsá’pssi both have a single functional category that licenses the sentient argument, which is either an experiencer or non-intentional actor. In both cases the event is reactive—it denotes a result in the case of non-intentional action, or a sentient response in the case of experience.

94 elizabeth ritter

4.5.2 Sentience, paths, and events As discussed in §4.3.2, in a series of papers, Kim shows that ai positional and motion verbs are compatible with either a sentient or non-sentient theme argument: (43) a. Anna saahkomaapi wamis-oo-wa. dem boy up-go.ai-3sg ‘The boy went up.’ b. Anna ainak’asi wamis-oo-wa. dem wagon up-go.ai-3sg ‘That wagon moved up.’ (Kim 2017a: 129) But, surprisingly, only a sentient theme is compatible with a path or goal ‘PP’ object:3⁰ (44) a.

b.

Anna saahkomaapi itap-oo-wa oomi isspahkoyi. dem boy.an goal-go.ai-3sg dem hill.in ∗

‘The boy went to that hill.’ Anna ainak’asi itap-oo-wa oomi isspahkoyi. dem wagon.an goal-go.ai-3sg dem hill.in ‘The wagon moved to that hill.’ (Kim 2017b: ex. (12))

As Kim observes, these findings indicate that path and goal PPs are like other kinds of goals in that they require a sentient external argument in Blackfoot. Arguably, goal directed motion requires a sentient theme because only sentient themes are capable of intentional action towards a goal. I interpret this as 3⁰ According to Kim (2017a,b) Blackfoot prepositions are realized as verbal prefixes, called linkers in the Algonquianist literature, while their complements are realized as independent nominals (NPs or DPs). In (44), for example, itap- is a preposition/linker sometimes translated as ‘to’, and its selected complement is the DP oomi isspahkoyi ‘that hill’. Directional and locational adverbs are also realized as verbal prefixes. These are called non-linkers in the Algonquianist literature, and analysed by Kim (2015) as intransitive prepositions. The prefix wamis- in (43) is an example of an intransitive preposition/nonlinker. As illustrated in (i), it is incompatible with a complement realized as independent DP: (i) ∗ Waamis-oo-wa anni isspahkoyi. up-go.ai-3sg dem hill ‘S/he went up that hill.’ (adapted from Kim 2014: ex. (13a))

sentience-based event structure 95 evidence that non-sentient themes, whether they are predicated of ai or ii verbs, participate in inert situations, and as such cannot serve as event initiators. Summarizing the discussion in this section, I have argued that the fundamental distinction among situation types in Blackfoot is not between states and events, but rather between inert situations on the one hand, and active, reactive, and interactive situations on the other. The latter involve sentient agents with the capacity to act intentionally in order to initiate a dynamic situation, as well as sentient goals with the capacity to recognize that the agent’s intention has been realized, or sentient experiencers with the capacity to recognize that a non-dynamic situation obtains. An inert situation, on the other hand, is one that lacks such a participant. Though it may involve action on the part of an inanimate object, it cannot be goal oriented, because goals must be set by sentient agents.

4.6 Conclusion Argument structure realizes event structure. In many languages whose event structure has been investigated, arguments that undergo a change of state or location as a result of the action named by verb have a privileged status. But why should undergoers have a privileged status? The evidence from Blackfoot indicates that event-defining arguments are not universally fixed in this way. The distinctive properties of this language can best be understood if we take a step back: Event structure is determined by one or more arguments with a particular semantic property. The task before us is to determine which semantic property is relevant, and how this figures into the organization of events. On the surface, Blackfoot looks very different from many other languages, but I suggest that this difference is due to the fact that in Blackfoot, all eventive functional categories select for sentient arguments, formalized as DPs with the feature [m], whereas in other languages only a subset of functional categories do so. For example, it has been argued that animacy/sentience/person is necessary for external arguments (subjects) of transitive clauses in a number of genetically distinct languages, including Japanese (Kuno 1973), Jakaltek (Craig 1977), Halkomelem (Gerdts 1988), and Lakhota (Van Valin and LaPolla 1997). In other languages, a strict animacy/sentience/person restriction is imposed on external arguments in some, but not all clause types (e.g. English have causatives, on which see Ritter and Rosen 1993, 1997). Copley and Harley (2009) characterize the external argument of have causatives as a director—a

96 elizabeth ritter sentient being with the ability to bring about the event. Extending the analysis proposed for Blackfoot, in all these cases, the functional category that selects the external argument, selects for a DP specified for [m]. In Blackfoot, on the other hand, a sentience feature licenses all eventive arguments. It would seem then that the difference between Blackfoot and languages such as Japanese and English is not a categorical one, but rather one of degree.

Acknowledgements This chapter is dedicated to Elizabeth Cowper. Your work, and your response to mine has consistently challenged me to think about problems differently, and to address questions that I wouldn’t have otherwise considered. Though I never had the privilege of taking one of your classes, you were one of my best teachers. I am also grateful to the Blackfoot speakers and linguists who collaborated on the fieldwork that made their work and this chapter possible. The Blackfoot language-keepers acknowledged by my sources are Beatrice Bullshields, Noreen Breaker, Rachel Ermineskin, Sandra Manyfeathers, and Brent Prairie Chicken; the linguists whose work informed mine are Heather Bliss, Donald Frantz, Sara Johansson, Kyumin Kim, Meagan Louie, and Kim Meadows; Norma Jean Russell, co-author of the Blackfoot dictionary is both a language keeper and author of an important source of information. Thank you all.

OUP CORRECTED PROOF – FINAL, 28/8/2020, SPi

PART II

C ON TR AST I N T H E A RG UM E N T D OM A I N

5 Definite expression and degrees of definiteness Maria Kyriakaki

5.1 Introduction In the literature, definite articles that appear with proper names and generic nouns have often been treated as semantic expletives that fulfil a syntactic (Longobardi 1994) or morphological (Lekakou and Szendrői 2012) requirement. An example is the Greek definite article, which must appear with proper names, generic nouns, and possessive nominals, and which can also appear more than once in a single nominal:1 (1) a. ∗ (I) Anna sinandise ∗ (to) Janni sto parko. the.f Anna met.3sg the.m John in.the park ‘Anna met John in the park.’ (proper names in subject and object positions) b. ∗ (I) elefandes aγapun ∗ (ta) fistikja. the.pl elephants love.3pl the.n.pl peanut.n.pl ‘Elephants love peanuts.’ (generic subject and object) ∗ ∗ c. (to) ðiko mu vivlio / (to) vivlio mu / ∗ (to) vivlio the.n own my book the.n book my the.n book ∗ (tis) Stellas the.f.gen Stella.gen ‘my own book’ / ‘my book’ / ‘Stella’s book’

1 Greek possessives can be headed by indefinite articles or numerals, as discussed in §5.4.1. The indefinite counterpart of (1c) is provided below: (i) ena ðiko mu vivlio / ena vivlio mu / ena vivlio tis Stellas a.n own my book a.n book my a.n book the.f.gen Stella.gen ‘a book of mine’ / ‘a/some book of mine’ / ‘a/some book of Stella’s’ Maria Kyriakaki, Definite expression and degrees of definiteness In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Maria Kyriakaki. DOI: 10.1093/oso/9780198817925.003.0005

100 maria kyriakaki d. θa iθela (to kocino) to triandafilo (to kocino) would like.1sg.past the red the rose the red ‘I would like the red rose.’ (restrictive modification by nominals (RMN)) Definite expressions as in (1d), where the adjective is introduced by its own determiner, are traditionally called ‘determiner spreading’ (Androutsopoulou 1996) or ‘polydefinites’ (Kolliakou 2004). In Kyriakaki (2011) this phenomenon is described as restrictive modification by nominals, RMN for short, since the article followed by the bare adjective forms a nominal that restrictively modifies the head noun. Briefly, RMN exhibits a flexible syntactic distribution in that the article and adjective can be placed before or after noun, as in (1d), where the phrase to kocino (lit. ‘the red (one)’) can either precede or follow to triandafilo (‘the rose’). This is in contrast with bare adjectives, which are disallowed in postnominal positions in definite nominals (Alexiadou and Wilder 1998; Alexiadou 2001a; Campos and Stavrou 2004; Kolliakou 2004).2 Additionally, it has been further observed that the prenominal adjective headed by its own determiner is typically interpreted restrictively (see Kolliakou 2004; Kyriakaki 2011). To account for the flexible distribution of the Greek definite article, Lekakou and Szendrői (2012) have argued that the definite article is inserted to spell out morphological case, but that it is a semantic expletive otherwise. However, I show that the Greek definite article only appears in definite contexts and is completely incompatible with indefinite DPs. This fact alone suggests that it may contribute some aspect of definiteness to the DP. Moreover, a cross-linguistic examination reveals that other definite articles show similar properties. French, for instance, German, and especially the Southern German dialects, use their definite articles extensively (see examples in (2)), while Scottish English also presents a similar phenomenon to the Greek polydefinite construction:

2 For example, the postnominal bare adjective asimenja is disallowed in (ii): i asimenja pena the.f.nom silver.f pen.f ‘the silver pen’ (ii) ∗ i pena asimenja the.f.nom pen.f silver.f Intended: ‘the silver pen’ (i)

definite expression and degrees of definiteness 101 (2) a. dass die Insekten nicht aussterben können [German] that the insects not become.extinct can ‘that insects cannot go extinct’ (Brugger 1993: 5) ∗ ∗ b. (Les) / Des dodos sont éteints. [French] def.pl indef.pl dodos are extinct.m.pl ‘Dodos are extinct.’ (Krifka et al. 1995: 68) c. the friend the footballer [Scottish English] ‘my friend the footballer’ (Kyriakaki 2011: 6) None of these articles appear in indefinite DPs, and all of them are argued to contribute semantically to the DP. They are thus not semantic expletives. Rather, these examples provide support for the possibility that such determiners form a group of definite determiners with distinct properties from fully specified and non-specified definite determiners, the true expletives. If there are three types of definite determiners, i.e. fully specified definite determiners, true expletives, and those that appear to be partially specified for definiteness, the question arises as to what their syntactic representations should be. Following and adapting previous work on definiteness (Heim 1982; Heim and Kratzer 1998; Lyons 1999, among others), I propose that definiteness can be decomposed into two features, familiarity and uniqueness, and that semantically definite determiners may spell out either both of these features or only one. When only one is spelled out, it is familiarity. In contrast, expletive determiners spell out neither uniqueness nor familiarity. Definite determiners spelling out both familiarity and uniqueness are fully definite, and exhibit a restricted distribution. Definite determiners spelling out only familiarity are less specified and are predicted to have a wider distribution. In some cases they appear with proper names; in others they can be restrictively modified by other definite nominals; and they can also appear with kind-denoting generic plurals. Finally, definite articles that spell out neither feature are the true expletives; these may sometimes appear even in indefinite nominals. Hence, there are three types of determiners, corresponding to the three degrees of definiteness: full, partial, and zero specification.3 3 As Marika Lekakou (p.c.) points out, since there are two features assumed, four different types of definiteness are expected, including one type where only uniqueness is realized. I leave this question open for future cross-linguistic research.

102 maria kyriakaki Structurally, in the case of a partially specified definite determiner, the features of definiteness are mapped onto two syntactic projections, Familiarity Phrase (FamP) and for uniqueness, iota Phrase (ιP), as in (3a). FamPs are predicative and appear as the syntactic complement of ι. If, on the other hand, the determiner is fully definite, definiteness is mapped to a single projection as in (3b), e.g. an argumental DefP: (3) a.

ιPe ι

b.

DefP[ι,fam]

FamP⟨e,t⟩ Fam

Def[ι,fam]

Fully definite determiners spelling out both features of definiteness are Def heads. Partially specified definite determiners spelling out familiarity are Fam heads and project predicative FamPs. Such FamPs may combine with other types of predicative definite nominals, as in the case of restrictive modifying nominals, or they may be selected by a generic operator. ιPs, however, are argumental and may not combine with other nominals in the structure. Finally, true expletives are bare Ds and project typical DPs.⁴ Hence, by this mechanism, we can account for the different degrees of definiteness that exist cross-linguistically. This chapter is organized as follows: §5.2 presents some of the most influential research on expletive definite articles and outlines some of the implications raised in this work. Drawing from relevant previous analyses of definiteness, §5.3 lays out the proposed analysis of definiteness and shows how such an analysis can account for the properties of partially specified definite determiners, such as those of the Greek definite determiner, which allows RMN and appears with generic nominals. For both of these constructions, it will be argued that it is the mapping of the features of definiteness that makes such structures possible. An interesting parallel is also drawn between ιPs, in which a unique individual or set of individuals is selected by means of iota, and GenPs, definite

⁴ This work does not assume that there is a DP layer on top of ιP. DPs are taken to be nominals that may involve expletive definite determiners and other non-definite determiners. In support of this, Kyriakaki (2011) shows that indefinite DPs in Greek allow bare postnominal adjectives, while definite nominals do not. The claim that is put forward here (to be tested in future research) is that semantically definite determiners spell out a combination of definiteness features and therefore should involve only a semantically definite nominal.

definite expression and degrees of definiteness 103 generic plurals, which identify a set of entities with specific properties. §5.4 applies the account to the English definite article and compares its properties with Greek. As argued, the English definite article constitutes a case of a fully definite determiner, while the null determiner found in other types of definite expressions is a Fam exhibiting properties similar to those of the Greek definite determiner. §5.5 presents the three types of determiners, and §5.6 summarizes the proposal and finishes with a discussion that provides new insights and research directions on definiteness.

5.2 The expletive account Longobardi (1994) was among the first to distinguish between two types of definite articles in Romance and Germanic languages: a substantive one, i.e. a semantically definite article, and an expletive one. Longobardi (1994) argues that the expletive article can be found in proper names, generic nouns, and mass nouns: (4) a. (Die) Biber bauen Dämme. the.pl beavers build.pl dams ‘Beavers build dams.’ b. (Der) Hans ist angekommen. the.m.sg Hans is arrived ‘Hans has arrived.’ c. (Die) Milch ist weiss. the.f.sg milk is white ‘Milk is white.’

[German]

(5) a. (Il) Gianni mi ha telefonato. the Gianni me has called ‘Gianni called me up.’ b. ∗ (I) castori costruiscono dighe. the beavers build dams ‘Beavers build dams.’

[Italian]

(6)

(∗ The) Beavers are mammals.

[English]

To account for this pattern, Longobardi (1994) argues that the definite article in Italian and German languages is an expletive that may optionally

104 maria kyriakaki be present. For the cases where it is obligatory, as in Italian generic subjects like (5b), Longobardi argues that the definite article is inserted because bare generic subjects are not allowed in positions that are not lexically governed. Similar claims that the definite article can be an expletive have been made for Greek too. Roussou and Tsimpli (1994), for instance, claim that the Greek definite article is an expletive, and that Greek also does not allow bare generic subjects. The Greek determiner is inserted to satisfy the lexical government requirement and to allow the nominal to function as an argument: (7)



(I) manaviðes kserun ariθmitiki. the.m.pl greengrocer.m.pl know.3pl arithmetic ‘Greengrocers know arithmetic.’

[Greek]

Roussou and Tsimpli (1994: 73) argue that the interpretation of the subject of the individual-level predicate ‘know arithmetic’ is generic and that the subject is not interpreted as a definite description despite the presence of an overt D. The subject of a stage-level predicate, on the other hand, allows either a generic or a definite interpretation: (8)



(I) fitites piγenun se ðjaðilosis. the.m.pl student.m.pl go.3pl to demonstration.f.pl ‘Students go to demonstrations’ or ‘The students are going to demonstrations.’ (Roussou and Tsimpli 1994: 73)

Roussou and Tsimpli argue that two interpretations are possible in (8), a habitual one (‘Students go to demonstrations’) and a progressive one (‘The students are going to demonstrations’). In the first case they claim that the noun phrase is generic and indefinite, while in the second case the noun phrase is a definite description. They point out that similar facts arise with generic objects, as well. Note that here the determiner is also obligatory: (9)

∗ Adipatho (tis) apsimaxies. despise.1sg the.f.acc.pl skirmish.f.pl ‘I despise skirmishes.’ (Roussou and Tsimpli 1994: 75)

Here too, Roussou and Tsimpli argue that the definite article is an expletive. But they observe that Longobardi’s (1994) analysis cannot account for generic objects, since it incorrectly predicts that bare objects, being lexically governed,

definite expression and degrees of definiteness 105 should be well formed. They conclude that Longobardi’s analysis should be reformulated to account for the Greek facts, as well. In their account of RMN, or in their terminology ‘close appositives’, such as (1d), Lekakou and Szendrői (2012, 2013) also claim that the Greek definite article is semantically inert. Treating polydefinites as involving reference to a unique individual at the intersection of two (or more) sets, they claim that Greek DPs are predicates of type ⟨e,t⟩. In the case of RMN, the two DPs are structural sisters, since the adjective can agree with either of them, as in (10a) and (10b).⁵ This relation is represented in the structure in (10c). (10) a. O aetos to puli ine meγaloprepos / the.m eagle.m the.n bird.n be.3sg majestic.m meγaloprepo. majestic.n ‘The eagle that is a bird is majestic.’ b. To puli o aetos ine meγaloprepos / the.n bird.n the.m eagle.m be.3sg majestic.m meγaloprepo. majestic.n ‘The eagle that is a bird is majestic.’ (Lekakou and Szendrői 2012: 114) c. Lekakou and Szendrői’s (2012: 117) proposed structure DefP Def ∅

DP1,2 DP1 D o or: to

DP2 NP aetos puli

D to o

NP puli aetos

Based on their claim that the two DPs denote sets, and not individuals, Lekakou and Szendrői argue that the definite article in Greek cannot be the element that turns the predicate into an argument, and conclude that all ⁵ Note that in (10b) judgements vary: for a few speakers there cannot be agreement between the masculine noun aetos ‘eagle’ and the adjective meγaloprepos ‘majestic’. Rather, the adjective here can only agree with the noun puli, implying that there is an asymmetric relation between the two nominals, rather than a symmetric one. This is in fact what is argued for later in this section.

106 maria kyriakaki definite determiners in Greek are semantically expletive, and do not yield an individual. For them, the source of semantic definiteness, i.e. the uniqueness presupposition, is located in a distinct, phonologically null functional head, Def, which c-commands and takes scope over DP, as in (10c). Lekakou and Szendrői thus propose that in languages with expletive determiners there is a dissociation between Def and D. They attribute this split to the existence of morphological case in the language and argue that the syntactic K(ase) head is responsible for marking the nominal as an argument of the clausal predicate. They conclude that languages with morphological case marking will also have RMN constructions as in (1d), repeated as (11), where the determiner occurs more than once. (11)

θa iθela to kocino to triandafilo. would like.1sg.past the red the rose ‘I would like the red rose.’

With respect to proper names in Greek, Lekakou and Szendrői follow Kripke (1980) in treating proper names as arguments of type e, and thus the obligatory occurrence of the definite article with proper names is for them an argument in favour of the expletive nature of D. The accounts reviewed above (Longobardi 1994; Roussou and Tsimpli 1994; Lekakou and Szendrői 2012, 2013) commonly take definite articles that appear with proper names, generics, restrictive modification by nominals as in (11), and other definite expressions, to be expletives that do not contribute any semantic definiteness. However, this criterion does not seem to sufficiently explain the properties of such determiners. An important question that arises is that if the definite article, in the case of Greek, were truly an expletive, why does it never appear with indefinite nominals? This issue cannot be addressed by an expletive account. An account that takes the definite article to contribute definiteness, on the other hand, predicts that it should appear in all and only definite nominals. As I argue next, all the nominals in which the definite article occurs, namely proper names, RMN, definite generics, and possessives, are in fact types of definite expressions. Generic nominals, in particular, have been argued previously (Lyons 1999) to be definite, denoting a familiar set to both speaker and hearer (see §5.3 for more). Hence the criterion suggested here to determine whether the definite article is an expletive is whether it can appear with indefinite nominals. Let us consider the generic definite subject and object DPs from examples (8) and (9) again, repeated in (12).

definite expression and degrees of definiteness 107 (12) a. ∗ (I) fitites piγenun se ðjaðilosis. the.m.pl student.m.pl go.3pl to demonstration.f.pl ‘Students go to demonstrations.’ or ‘The students are going to demonstrations.’ (Roussou and Tsimpli 1994: 73) ∗ b. Adipatho (tis) apsimaxies. despise.1sg the.f.acc.pl skirmish.f.pl ‘I despise skirmishes.’ (Roussou and Tsimpli 1994: 75) In neither of these sentences is it possible to have an existential indefinite interpretation of the noun with the determiner. In (12a), the nominal i fitites ‘the students’ singles out the set of students from the larger set of entities in the universe of discourse. Although a specific group of students may not be picked out, the definite determiner identifies the familiar category of students, as opposed to non-students. Similarly, in (12b), the object DP tis apsimaxies ‘the skirmishes’ selects a certain type of interaction, the set that contains all types of skirmishes. For existential indefinite objects, bare nominals are in fact possible: (13)

Efera molivi / molivja ke stilo. brought.1sg pencil.n pencil.n.pl and pen.n.sg/pl ‘I brought {a pencil / pencils} and {a pen / pens}.’

Clearly, these objects are interpreted differently from the one in (12b). The difference is that in (13) an existential indefinite interpretation is available, while in (12b), where the definite article is present, the indefinite interpretation is absent. As we will see, the impossibility of interpreting nominals like those in (12b) as existential indefinites can be straightforwardly explained if the Greek definite determiner does in fact contribute some aspect of semantic definiteness. The question to be examined, then, is what exactly this contribution is. In sum, morphologically definite plurals in Greek are not interpreted as existential indefinites. Rather, they must be interpreted either as specific definites (i.e. there is a specific unique entity that is picked out) or as definite generics.⁶ True indefinite interpretations are not available. It thus seems that ⁶ As argued in §5.3, definite generics are considered to be definite in the sense that a single unique set is picked out. As a reviewer notes, in a sentence such as (i) there is a sense in which the generic is unique, since there will be just one typically-behaving population of wugs: (i) Wugs are hunted by cormorants in the daytime and so prefer to remain snug in their reed nests.

108 maria kyriakaki the determiner is not merely filling a syntactic requirement. Rather, it makes a semantic contribution. I propose that it marks a type of definiteness, which as it turns out can be of two types, generic or referential. With respect to Lekakou and Szendrői’s (2012, 2013) account, the argument that the two nominals involved in RMN are sisters does not appear to be irrefutable. Consider the following example for instance, with the two nouns leoparðali ‘leopard’ (grammatically feminine) and θilastiko ‘mammal’ (grammatically neuter). The adjective epikinðini ‘dangerous’ can agree only with leoparðali: (14) a. I leoparðali to θalasio θilastiko ine ekseretika the.f leopard.f the.n sea.n mammal.n be.3sg exceptionally epikinðin-i/∗ -o. dangerous-f/-n ‘The leopard that is a sea mammal is exceptionally dangerous.’ b. To θalasio θilastiko i leoparðali ine ekseretika the.n sea.n mammal.n the.f leopard.f be.3sg exceptionally epikinðin-i/∗? -o. dangerous-f/-n ‘The sea leopard is exceptionally dangerous.’ These examples show that the adjective ‘dangerous’ easily agrees with only one of the two nominals, namely the more specific one, leoparðali ‘leopard [seal]’, a subset of θalasio θilastiko ‘sea mammal’. In contrast, ungrammaticality may result if the adjective agrees with the more general noun. Hence the relation between the two nominals must be asymmetric, as is also argued in §5.4.2. Lekakou and Szendrői’s (2012) proposal that the two DPs are in a symmetrical relation and interpreted intersectively cannot be maintained here. Moreover, Lekakou and Szendrői predict a correlation between morphological case marking and the possibility of polydefinites, because in their analysis, languages with expletive Ds use K to mark nominals as arguments. However, Scottish English also exhibits similar constructions (e.g. the friend the footballer in (2c)), even though there is no overt case marking on the nominal. It therefore cannot be morphology that makes such constructions possible, but rather the semantics of the determiner. In conclusion, the analyses discussed here do not provide a fully explanatory account of the properties of the Greek definite determiner. However, an analysis that allows for partial definiteness can account for its syntactic distribution, its semantic contribution, and its inflectional properties, as well

definite expression and degrees of definiteness 109 as those of other definite nominals and adjectives. Accordingly, the standard view of determiners, that they serve to turn an otherwise predicative noun into an argument, is adopted here. The determiner occupies its own functional projection (Abney 1987), while the noun typically occupies the head of a Number projection (Ritter 1991; Borer 2005; Cowper 2005b; see also Kyriakaki 2011 on NumberP in Greek nominals). I also assume that pure expletives are in fact possible. However, determiners that appear uniformly with definite expressions such as proper names, generics, RMN, and definite possessives, are not taken to be completely unspecified for definiteness. Rather, these may well be partially specified, as argued below.

5.3 The system of definiteness and its syntactic manifestations 5.3.1 An overview of definiteness In this section, after a brief review of relevant analyses of definiteness, I propose a view of definiteness that takes into account specific aspects of those, and develop a system that can be further extended to other languages. Various semantic analyses have been proposed over the years to account for the cross-linguistic properties of definite determiners. Research into the meaning and function of the definite article has argued either that the definite article requires the referent of the DP to be familiar within the discourse (Heim 1982) or that the referent must be uniquely identifiable (Heim and Kratzer 1998).⁷ It has also been observed that the familiarity or uniqueness requirement alone does not suffice cross-linguistically to pick out a unique individual, and that contextual restrictions play an important role in determining a unique individual (see for instance Chung and Ladusaw 2004; Giannakidou 2004). Gillon (2006, 2009) argues in favour of domain restriction (C), which for her is an unpronounced element of type ⟨e,t⟩ that is interpreted by intersective predicate modification with the nominal. C is the characteristic function of the set of elements under discussion, and is associated only with determiners, and not quantifiers as argued in Giannakidou (2004). For Gillon (2009), all determiners are associated with domain restriction, while other features, such as uniqueness in English, may also be present. English definiteness, then,

⁷ The English determiner the, in particular, has been previously identified with features of definiteness (Christophersen 1939; Prince 1981; Heim 1988, among others), and uniqueness (Kadmon 1992; Frege 1997 [1892]; Russell 1998 [1905]).

110 maria kyriakaki results from the combination of domain restriction, i.e. the set of elements in the contexts, and the assertion of uniqueness. This combination generates familiarity effects. Gillon further makes the claim that domain restriction is a universal property of definiteness cross-linguistically, as in English and Skwxwú7mesh, while other features, such as uniqueness in English, can be language-specific. Lyons (1999), on the other hand, claims that it is not possible to provide a universal semantics for definiteness, since the determiner may encode different semantic functions in different languages. Instead, he treats definiteness as the grammaticalization of what he labels as ‘semantic/pragmatic definiteness’. He considers definiteness to be a grammatical category on a par with tense, mood, number, and gender, not grammatically active in every language, but only in those that show overt marking. Drawing on the different views of definiteness that these approaches have to offer, I propose a syntactic analysis of definiteness that accounts for the semantic and distributional properties of definite determiners previously simply dismissed as semantic expletives. I propose that definiteness can be grammatically active in languages as a syntactic category DefP.⁸ Somewhat in the spirit of Gillon (2006, 2009), I propose that definiteness involves two components, familiarity and uniqueness.⁹ Unlike Gillon’s (2009) unpronounced domain restriction, I assume familiarity can be spelled out by the determiner and can occupy a separate syntactic node, an assumption made also by Martí (2003) for domain restriction. Depending on the features spelled out by the definite determiner, uniqueness and familiarity can be mapped into two distinct syntactic projections—iota Phrase (ιP) and Familiarity Phrase (FamP), respectively. FamP is an ⟨e,t⟩ predicate, as argued next, which is also in accordance with Gillon’s proposal on domain restriction, and contains the set of salient entities. A definite determiner may spell out both uniqueness and familiarity. Such determiners are fully specified for definiteness, in which case definiteness is ⁸ Providing a syntactic approach that is more economical and does not have to employ Restrict and Specify choice functions (Chung and Ladusaw 2004, for indefinite nominals) to narrow down the domain, I choose familiarity as a feature of definiteness that may be present in languages as a grammatical category. This way, the familiarity effects mentioned by Gillon are a direct result of the feature spelled by the determiner, and not a result of the combination of uniqueness and domain restriction. Finally, familiarity, as opposed to domain restriction, is more clearly associated with the properties of the determiner, i.e. with the D position, than with a quantifier which has been previously argued by Giannakidou (2004). For argumentation on why quantifiers are not associated with contextually salient entities, see Gillon (2009). ⁹ As we will see, these components, especially uniqueness, might vary cross-linguistically depending on the type of DP. Definiteness may thus additionally involve other features, such as genericity or specificity, which can also map to separate syntactic projections, much like the expanded versions of what were previously identified as the single categories CP and IP.

definite expression and degrees of definiteness 111 realized in a single syntactic node, DefP, as in (15a). As we will see in the next section, English the is one example of a fully specified determiner spelling out Def. In contrast, definite determiners spelling out only one aspect of definiteness, i.e. familiarity, are underspecified for uniqueness.1⁰ Definiteness is in this case mapped to two separate projections, ιP and FamP. The Greek definite article is one such example. As shown in (15b), it spells out only the lower of these positions, and can thus appear in contexts where fully specified definite determiners do not. (15) a.

DefP

Def0 (English the)

b.

ιP

ι

FamP

Fam (Greek definite article) In (15b), Fam merges with the nominal, and ι merges with FamP. What is the semantic type of FamP—i.e. is it argumental or predicative? The fact that FamP is dominated by ιP suggests that FamP must be predicative. As I argue next, the determiner denoting only familiarity, Fam, selects a contextually salient or familiar set of entities. Since Fam gives us a set, and not a single entity, I propose that FamP is predicative, and so its semantic type should be ⟨e,t⟩. The resulting nominal, however, must be an argument of type e. Furthermore, a single entity must be picked out of the set of contextually salient entities. I propose that Heim and Kratzer’s ι merges next, and gives us an argument, a unique entity of type e. This gives us the following structure:11 1⁰ Interestingly, Gillon (2009) also observes that determiners in Skwxwú7mesh behave like some intermediate category with definite-like behaviour, and points out the need for an intermediate category of D. The analysis proposed here can capture the partially definite behaviour of D through partial specification of definiteness. 11 Following previous analyses (Ritter 1991; Borer 2005; Cowper 2005b, inter alia, and Kyriakaki 2011 for Greek), I assume there is a NumP in the structure where the noun moves to. Kyriakaki (2011) presents additional evidence from genitives that argues in favour of that position. A direct effect of the assumption that the noun moves to NumP is that prenominal adjectives in Greek should then adjoin to NumP.

112 maria kyriakaki (16)

ιPe

ι

FamP⟨e,t⟩ Fam

NumP …N

Definiteness can thus be decomposed into two syntactic nodes: ιP, whose head appears to be phonologically null, and FamP, whose head is spelled out in Greek by the definite article. As argued next, this structure gives a novel and revealing view of the properties of certain definite expressions that could not be fully accounted for before. These expressions are restrictive modification by nominals (RMN) and definite generics. As we will see, it is the predicative status of FamP that makes both of these possible.12

5.3.2 Underspecified definite expressions 5.3.2.1 Restrictively modified nouns Polydefinites—i.e. nouns that are restrictively modified by other nominals in Greek—are most often common nouns, possessive or not, and proper names.13 In most varieties of English, such nouns can only be possessed nouns and proper names. Generally, restrictively modified DPs are different from those that are uniquely established in the context. The latter are unique entities that are selected from a set of entities, the D set, while as we will now see, the uniqueness of the restricted DPs arises from the intersection of this set with another set.1⁴

12 In the case of definite generics, it will be also shown how a generic operator can apply to such structures in combination with Fam, as opposed to the uniqueness operator ι. 13 See §5.2 for a brief overview of previous work, or Kyriakaki (2011) for a complete analysis of the RMN phenomenon and the view followed in this work. 1⁴ As correctly observed by one of the reviewers, nominal modifiers can be purely descriptive as well as restrictive. For example, I might refer to someone as Claudio the baker even if he is the only person named Claudio known to the speaker and the listener, for example as a way of reminding my interlocutor of who I am referring to, or what Claudio’s occupation is. See Wiltschko (2013) for a semantic discussion of this type of structure.

definite expression and degrees of definiteness 113 Consider for instance the following examples: (17) a. John offered me some coffee. b. John the assistant offered me some coffee. Following Longobardi (1994), Massam and Ghomeshi (2009) argue from such examples that the proper DP John forms a definite phrase with a definite null D. They argue that this null D is specified with the feature [singular], though otherwise its meaning and function is similar to the. This means that the null D for them has a similar semantics to that given for the by Heim and Kratzer (1998).1⁵ Accordingly, the context in (17a) supplies us with a unique entity, John. This entity is selected by the null determiner from a set of entities with different names. That is, in (17a) we have a unique entity that is contextually salient, and this is John. In the context of (17b) however, John by itself does not give us a unique entity. Rather, we have instead a non-singleton set of entities named John, i.e. a set of contextually salient people with the same name. The selection of the unique entity is achieved by the intersection of this set with another set, which contains the assistant. Thus, in (17a) the contextually salient set C of people named John, which is a subset of D, is a singleton set containing a unique individual y. This is schematized in Figure 5.1a. In (17b), C does not contain a single individual but rather is a non-singleton set {x, y, z…}, as shown in Figure 5.1b. Hence, in Figure 5.1a a unique individual John is successfully selected. In Figure 5.1b, though, C is a non-singleton set and thus a unique individual cannot be selected. This is parallel to Greek. That is, proper names in English, and also possessives, as will be shown, are underspecified for definiteness in the same way as are Greek DPs containing the definite article. I propose that English proper names with restrictive modifiers (such as John the assistant

D

D C: {y}

C : {x, y, z}

Figure 5.1 Differing sets of contextually salient individuals 1⁵ In §5.4, it is argued that the semantics of the null determiner resembles that of the Greek definite article, and not that of the English determiner the. That is, null D also lacks the uniqueness specification, and it is for this reason that the context supplies us with the unique entity, here John.

114 maria kyriakaki in (17b)) and Greek polydefinites both involve sets and their intersection. In other words, the matrix nominal must not be a full argument (at this point in the derivation), but rather a predicative function of type ⟨e,t⟩.1⁶ The same, I claim, is true of the restrictive modifier. Regarding the restrictive modifier in particular, the idea that it is a function is not new. Restrictive modifiers are commonly argued to be of type ⟨e,t⟩ and to intersectively modify a noun which is also of type ⟨e,t⟩. For instance, Heim and Kratzer (1998) propose that restrictive modifying relative clauses are not propositions, but rather predicates. Relative clauses are just like other modifiers in the noun phrase, such as PPs and APs. In a sentence like The house which is empty is available, the restrictive relative clause which is empty has the same denotation as the adjective empty, and is thus of type ⟨e,t⟩. When the two are combined by Predicate Modification (as in (18a)), the resulting ⟨e,t⟩ predicate can then be the argument of ⟦the⟧ (type ⟨⟨e,t⟩,e⟩), giving the unique empty house. Heim and Kratzer (1998: 82–3) further argue that the restrictive modifier first intersects with the noun to the exclusion of the determiner. This is shown in (18a). A structure as in (18b), with the modifier adjoining higher, would yield the wrong semantic type. (18) a.

DPe D⟨et,e⟩

NP⟨e,t⟩ N⟨e,t⟩ ∗

b.

DPt

DPe D⟨et,e⟩

Mod⟨e,t⟩

Mod⟨e,t⟩ NP⟨e,t⟩

Structures like (18b) are ruled out because if the determiner combines first with the noun, the result will be an argument of type e. The adjunction of the predicative ⟨e,t⟩ modifier will therefore yield a proposition (type t) instead of 1⁶ In a complex nominal consisting of a modified and a modifying nominal (as in (14a) and (17b)), I will refer to the modified nominal as the matrix nominal, since this nominal, and not the modifying one, is shown in (14) to determine agreement on the adjective. The structure discussed here is along the lines of the structure presented in (18a), with the matrix nominal of type ⟨e,t⟩ intersecting with the modifying nominal which is also of the same type. See §5.4 for the details of this mechanism.

definite expression and degrees of definiteness 115 an individual (type e), and the DP will thus be unable to compose with a verb or verb phrase. Since higher adjunction of the restrictive modifier leads to uninterpretability, (18b) cannot be the structure of modified nominals. As Heim and Kratzer point out, this is the case even in languages that seem to exhibit the higher adjunction structure. As they argue, the modifier must be interpreted low at some other level of representation. I will also argue that modifying DPs (DPmod ) are ⟨e,t⟩ predicates that adjoin low, though they can be interpreted higher at LF. I now consider which types of Greek nominals allow restrictive modification, and examine how the grammar generates only the well-formed structures. Not all DPs can be modified by restrictive nominals. As shown in (19), a restrictive modifying DP with a definite article cannot modify an indefinite nominal, even if the indefinite is specific. This is true whether the modifier is prenominal or postnominal: (19) a.



ena triandafilo [DPmod to kocino] a/one rose the red ∗ ‘a rose the red one’ (unacceptable if restrictive) (or, in a felicitous English translation, ‘a rose that is red’) b. ∗ [DPmod to kocino] ena triandafilo the red a/one rose Intended: ‘the red one among roses’ c. ∗ [Kapjos/Enas tipos] o apenadi irθe apo ðo. some/a guy the across came.3sg from here ∗ ‘Some/A guy the one across the street dropped by.’

Additionally, restrictive DPs cannot be headed by overt indefinite determiners: (20) a. # ena triandafilo ena kocino a/one rose a/one red ∗ ‘a rose a red one’ (on the restrictive reading) ∗ b. ena kocino ena triandafilo a/one red a/one rose *‘a red one a rose’ (or, as a felicitous translation in English for both (20a) and (20b): ‘a rose that is red’)

116 maria kyriakaki Only definite DPs can be restrictively modified by another nominal. Since the only overt difference between definite and indefinite DPs is the determiner, the determiner may be what makes this type of restrictive modification possible. As I will argue, the fact that the definite article allows for this modification indicates that the definite determiner in Greek does not uniquely select an individual, and the restrictive modifying nominal can thus contribute to the selection of a unique individual.1⁷ This contrasts with English, where, as we will see in more detail, the definite determiner the does not allow for restrictive modification by other nominals. As we can observe in (21), nominals not headed by the, such as proper names, can be modified by restrictive nominals: (21)

John the professor is taller than John the doctor.

In Greek, proper names also allow modification by restrictive DPs, as shown in (22): But note that Greek proper names appear with the definite article: (22)

o Jannis {o kaθijitis / i iðiofiia / o psilos} the John the professor the genius the tall ‘John {the professor / the genius / the tall (guy)}’

In contrast, English definite DPs with the do not allow such modification, although Greek definite DPs with common nouns allow it: (23) a. b.

∗ ∗

the professor {the linguist / the genius / the athlete} The professor the linguist is taller than the professor the biologist.

(24) a. o kaθijitis {o γlosoloγos / i iðiofiia / o ðrastirios} the professor the linguist the genius the active ‘the professor who is a linguist / a genius / active’ b. O kaθijitis o γlosoloγos ine psiloteros apo ton the professor the linguist is taller than the kaθijiti to violoγo. professor the biologist ‘The professor who is the linguist is taller than the professor who is the biologist.’

1⁷ The idea that the restrictive modifying nominal contributes to the selection of a unique individuals is not novel in the literature on polydefinites (see Lekakou and Szendrői 2012 and references therein).

definite expression and degrees of definiteness 117 The unacceptability of the English examples suggests that the English determiner the obligatorily selects a unique individual, making further restriction impossible. In contrast, the grammaticality of (21) suggests that the null D in a proper DP may not uniquely select an individual. Interestingly, possessed nominals also allow for restrictive modification both in Greek, as shown below, and in English:1⁸ (25)

o jitonas mu {o jatros / i iðiofiia} the neighbour my the doctor the genius ‘my neighbour {the doctor / the genius}’

Since possessed nominals allow for further restriction in the same way proper DPs do, this implies that their determiners have similar properties. In Greek, examples like this are expected to be acceptable since the same definite determiner appears here as with proper names. In sum, Greek definite nominals can be restrictively modified by another nominal. The noun of such nominals can be count, generic, possessive, or a proper name. In English, definite nominals introduced with the definite determiner the may not be modified. However, proper names and pronominal possessors, which do not contain the, can be modified, and thus pattern similarly to Greek DPs.1⁹ Hence, Greek definite nominals introduced by the definite article and English definite nominals not introduced by the pattern alike, seeming to be partially specified in terms of definiteness: i.e. a unique entity is not necessarily selected, and thus the additional restrictive nominal is possible. In contrast, English definite nominals introduced by the pick out a unique referent and thus do not allow further restriction of the head noun. The difference between the two nominal classes, those that can be modified and those that cannot, is in the definite determiner: in the former case the determiner is only specified for one semantic component of definiteness, while in the latter it is fully definite. Hence, before presenting an analysis for English definite nominals, we will also look more closely at the properties of the determiner (see §5.4). 1⁸ English possessives and proper names are considered in more detail in §5.4. 1⁹ As an anonymous reviewer observes, the question arises as to why proper names can be restrictively modified by nominals in English, but common nouns cannot. Based on the structure in (16) and building on it further, proper names are FamPs of type ⟨e,t⟩, as opposed to common nouns with the definite determiner the, which are argumental DefPs. In the structure, the predicative matrix FamP can easily intersect with the also predicative modifying nominal, which adjoins low but is interpreted higher with FamP at LF. In contrast, in the case of a common noun phrase with the fully definite the, an argumental DefP of type e cannot combine with a modifying nominal of type ⟨e,t⟩ for interpretation effects. See §5.4.2 for more.

118 maria kyriakaki 5.3.2.2 Generics Greek generics constitute another type of nominals that involves a partially specified definite determiner. Here, I adopt the standard claim that the determiner is inserted to turn the predicative noun into an argument (Longobardi 1994, among others). However, as discussed in §5.2, existential indefinite interpretations are not possible. The definite determiner must thus contribute some type of definiteness. Let us consider the generic definite subject and object DPs again. In example (26), repeated from (8) and (12a), the nominal i fitites ‘the students’ singles out the category of students: (26)



(I) fitites pijenun se ðiaðilosis. the.m.pl student.m.pl go.3pl to demonstration.f.pl ‘Students go to demonstrations.’

Hence, although a specific group of students may not be picked out, the definite determiner picks out the set of students as opposed to non-students. The same holds for example (27), repeated from (7), where a single set is picked out, containing all the greengrocers: (27)



(I) manaviðes kserun ariθmitiki. the.m.pl greengrocer.m.pl know.3pl arithmetic ‘Greengrocers know arithmetic.’ (Roussou and Tsimpli 1994: 73)

Similarly for objects, in (28), repeated from (9) and (12b), the object DP tis apsimaxies ‘the skirmishes’ selects a certain type of interaction, the set that contains all types of skirmishes: (28)

∗ Adipaθo (tis) apsimaxies. despise.1sg the.f.acc.pl skirmish.f.pl ‘I despise skirmishes.’ (Roussou and Tsimpli 1994: 75)

To conclude, morphologically definite plurals must be interpreted either as existential definites (i.e. as picking out a specific unique entity) or as generics. Moreover, as argued in §5.2, truly indefinite interpretations are not possible with definite plurals. It thus seems that the determiner is playing more than a purely syntactic role. That is, it does in fact make a semantic contribution. Genericity in Greek is manifested semantically, where the reference of the nominal is to the whole ensemble (in (26) and (27) this is to students and

definite expression and degrees of definiteness 119 greengrocers, respectively), as well as morphologically, i.e. with the definite determiner and the plural marking on the nominal, unless it is a mass noun. In the case of a mass noun, the noun is in singular form, and so is the determiner, which is obligatory in generic contexts, though not in existential indefinite ones: (29) a. ∗ (To) nero ine politimo aγaθo. the.n water.n be.3sg precious.n commodity.n ‘Water is a precious commodity.’ b. Erikse (arketo) hioni sta vuna. fall.3sg.past a.lot.of.n snow.n on.the.n.pl mountain.n.pl ‘(Plenty of) snow fell on the mountains.’ All Greek generic nouns thus obligatorily combine with the definite determiner, so that names of kinds are denoted. As Chierchia (1998) argues, since kinds must be arguments, the determiner is projected to turn the predicate noun into an argument. Semantically, Lyons (1999) argues that the reference to a whole ensemble is what may characterize English generics as familiar. That is, a hearer may fail to identify the individual referent of a noun phrase like a/the pencil, but the ensemble denoted by generic pencils is familiar to us all. From a semantic point of view, English generics are familiar, and thus at least partially definite. This familiarity is not overtly marked in English. In Greek, though, genericity, and thus familiarity, is spelled out by a vocabulary item, the definite determiner itself. Thus, familiarity in Greek is manifested both semantically and morphosyntactically. Working from the bottom up, definite generics must therefore involve a syntactic projection responsible for the familiarity effects, i.e. FamP: (30)

Generics, step 1 FamP⟨e,t⟩ Fam i ‘the’

NumP fitites ‘students’

We have seen that the determiner selects a contextually salient or familiar set of entities, i.e. the set referred to by the nominal in question—in this

120 maria kyriakaki case, the NumP i fitites (‘students’). This property is now captured by the proposal that nominals containing the definite article have a FamP in their structure, a predicative FamP. The subsequent question is what type of phrase dominates FamP. Unlike other definite nominals, it cannot be an ιP, since a specific entity or group of entities is not picked out here. Instead, the whole class of entities sharing a property (such as ‘students’ in (26)) is singled out from sets of entities with different properties (non-students). In Carlson and Pelletier (1995), generic nouns are said to involve a generic operator (Gen), which singles out a set of entities with specific properties. I will thus posit that definite generics have a generic operator (Gen) that selects Fam and projects a GenP.2⁰ The relevant structure is shown in (31a), while the corresponding structure for a typical specific definite noun phrase is repeated in (31b): (31) a. Generics, step 2 GenPe Gen

FamP⟨e,t⟩ Fam i ‘the’

NumP

fitites ‘students’ b. Specific definites ιPe ι

FamP⟨e,t⟩ Fam i ‘the’

NumP fitites ‘students’

Definite plurals thus are of two types: specific definites dominated by ιP, where there is a unique individual or plurality that is selected by means of the 2⁰ In line with Chierchia (1998), I assume that kinds involve pluralities, while singular definites denote singularities. Structurally, this could mean that Gen has a plurality feature that enters into an Agree relation with Num, and genericity is thus possible only with plural or mass nouns.

definite expression and degrees of definiteness 121 iota function, and generic definites dominated by GenP, where a set of entities with specific properties are singled out. In both types the higher node Gen or ι derives an argument of type e.

5.4 Definite determiners in English and Greek 5.4.1 DPs and FamPs In §5.3.2.1, we saw that English proper names and possessive definite DPs allow restriction by other nominals in the same way as Greek definite nominals, as in the English translations of the Greek nominals in (32): (32) a. o Janis o γlossoloγos the.m.nom John.m.nom the.m.nom linguist.m.nom ‘John the linguist’ b. o jitonas mu {o jatros / i iðiofiia} the neighbour my the doctor the genius ‘my neighbour {the doctor / the genius}’ (repeated from (25)) This similarity with Greek is accounted for if these English nominals are headed by a determiner with the same syntactic and semantic properties as the Greek definite determiner. In contrast, English definite nominals introduced by the determiner the disallow further restriction on the noun. As shown in (24b), repeated below as (33), such constructions are well formed in Greek: (33)

O kaθijitis o γlosoloγos ine psiloteros apo ton kaθijiti the professor the linguist is taller than the professor to violoγo. the biologist ‘The professor who is the linguist is taller than the professor who is the biologist.’

In this section, I determine the syntactic and semantic properties of the heads of English definite nominals and compare them with those of Greek. Starting with English proper names, I assume that there is a null D in these nominals, as was first suggested by Longobardi (1994). The question then is whether this null D is the same as the overt the. According to Ghomeshi and

122 maria kyriakaki Massam (2009), the null determiner of proper names has the same semantic features as the, but is additionally specified with [+singular] and [+proper]. In terms of definiteness, they argue that both Ds make the nominal phrase definite. The account proposed here is similar to Ghomeshi and Massam’s proposal in taking proper names to involve a null definite determiner. However, the null D and the must differ in the type of definiteness that they contribute. Although the null D may be specified with the additional features [+singular] and [+proper], as Ghomeshi and Massam propose, it appears to be underspecified in terms of definiteness. This is why expressions like John the teacher (not the accountant) are allowed, but expressions such as ∗ the woman the doctor (not the manager) are ungrammatical. The fact that the null D allows these restrictive DPs suggests that it is similar to the Greek determiner: unlike the, the English null D does not always pick out a unique individual. Proper names can thus be used in contexts where more than one entity with that name is established in the context. If the English null D does not pick out a unique individual but rather one or more familiar individuals, it is underspecified in the same way the Greek determiner is. This suggests that definiteness in English, as in Greek, is divided between ιP and FamP. Like the Greek definite article, the English definite null D that appears with proper names and possessed definite nominals is a Fam head.21 Uniqueness arises as in Greek, from the higher iota head, as shown in (34): (34) a. English definite null D ιP ι

FamP Fam ∅

b. Greek definite D ιP ι

FamP Fam o/i/to, etc.

Consequently, the English null definite determiner (D∅ ) can allow nominal modification as the Greek determiner does. D∅ is also only partially specified for definiteness, suggesting that definiteness is split between two syntactic heads. This provides evidence against the claim that restrictive nominal 21 As an anonymous reviewer points out, null D also appears with mass nouns in English, which, like plurals, disallows RMN unless possessed, e.g. ∗ (our) beer Hop Zombie. Clearly, such examples must involve a partially definite null D. It will be interesting to further examine how definiteness is manifested exactly with mass nouns. I leave this question open for future research.

definite expression and degrees of definiteness 123 modification of the sort found in polydefinites is unique to languages with rich inflection. Instead, it depends on the definiteness specification of the determiner. This result is not surprising, since, as Lyons (1999) points out, definiteness is encoded in different ways not only cross-linguistically, but also within individual languages. Turning to English possessors, restrictive DPs can appear not only with proper names, but also with possessed nominals, as in my cousin the writer (not my cousin the florist). Expressions like John’s sister the writer (not the florist) are also possible. If possessed nominals have a null D similar to the one that occurs with proper names, then these constructions are not unexpected.22 Based on what we have seen in the previous section then, restrictive modifying nominals are predicted to be possible in possessive constructions.23 Let us consider first DPs with a pronominal possessor, as in my cousin. Here, D may not always select a unique entity. Instead, it selects an entity from a set of possessed entities, which might not be fully definite. In Greek for instance, the possessive pronoun combines with either the definite or the indefinite article: (35)

to / ena vivlio mu the.n a/one.n book.n my.gen ‘my book’ / ‘a/some book of mine’

In English, the possessive D also seems to also be underspecified for definiteness. That is, the uniqueness of the referent is not due to the possessive D, but rather arises from the context. A phrase like my cousin does not necessarily presuppose that there is only one cousin. In a context where both the speaker and the hearer know that there is more than one salient entity, such a phrase would be infelicitous, as in (36a). This is because the null iota head cannot pick out a unique entity. Contrast this with (36b), where the restrictive DP is present and iota can now pick out a unique entity: (36) a. # I saw my cousin this morning. (Infelicitous if more than one salient entity is available, e.g. Abigail, Shanna, and Mariabella.) b. I saw my cousin Abigail this morning (not Shanna or Mariabella). On the other hand, in contexts where there are no other salient entities, the phrase her cousin is felicitous in the same way a proper name would be. As 22 See (38) for an example of the structure of possessives such as John’s sister. 23 Note that the discussion of possessive constructions here is limited to those with a definite interpretation.

124 maria kyriakaki with proper names, uniqueness in possessive DPs comes from the context and is not imposed or presupposed by the determiner itself. Hence, the determiner in the possessive DP is not like the definite determiner the, but rather is similar to the null D; in other words, it is a Fam head. Similarly, John’s sister does not necessarily entail that John has only one sister. If, for instance, more than one sister is salient, the phrase John’s sister is infelicitous. Again, the set is not in the domain of the iota operator. Thus, here too, the presence of the restrictive DP is necessary: (37) a. # We met John’s sister. (equally salient sisters: Abigail and Shanna) b. We met John’s sister Shanna (and not Abigail). Similar to DPs with a possessive pronoun, in a context where only one entity is salient, a unique entity is successfully picked out. In a context where there is more than one salient entity, the selection of the unique entity will come from the intersection of the sets of the possessive nominal and the modifying DP. Thus, these DPs also pattern like those with a possessive pronoun, and like those headed by a null D. Hence, we can conclude that the possessive D is a Fam head, as well. That the uniqueness does not arise from the determiner becomes more evident if we contrast a phrase Mary’s teacher with the teacher of Mary. The latter requires that there be a unique teacher. As the main difference is the determiner, it follows that the uniqueness must arise from the. Since in the possessive DPs, this uniqueness is not presupposed, the possessor does not necessarily select a unique entity either. Rather, like the null D, it selects a set of familiar entities. Like English proper DPs, then, possessive DPs also involve FamPs. Following Ritter (1991) and Adger (2003), I assume that independent genitives occupy the specifier of the determiner phrase as a result of movement; that is, they move to this projection to satisfy a strong genitive feature on D. This D is null. I thus assume, here too, that possessive DPs involve a null determiner. We can now explain why these DPs may also allow restrictive modification. As in our treatment of English proper names, I propose that English possessed DPs have a null Fam head. Its specifier is occupied by the independent genitive, which has moved to that position to satisfy Fam’s strong genitive feature (see (38)). English pronominal possessors, too, are phrasal. Bernstein and Tortora (2005) argue that they are NumPs which move to the specifier of a functional projection FP, to check a definiteness feature in D. Similarly, I assume here that

definite expression and degrees of definiteness 125 pronominal possessors are phrases that move to the specifier of a functional projection, in this case FamP. As with independent genitives, I assume that this movement is triggered so that the pronominal possessor phrase will check a genitive feature on Fam. We can now capture the properties of English possessives and proper names with a single structure, in which features of definiteness are mapped onto the two separate syntactic projections ιP and FamP: (38) a. John’s/my sister Abigail (possessors) b. John the accountant (proper name) c. ιP ι

FamP Possessor John’s/my

FamP Fam ∅

Possessum (sister Abigail) (John the accountant)

According to this structure, if the English nominal contains a possessor (such as my in my sister Abigail or John’s in John’s sister Abigail), it occupies the specifier of FamP. In the case of a proper nominal, following Longobardi (1994), I assume that the proper name raises to the head of the determiner phrase, in this framework Fam.2⁴ In (38b), then, John the accountant, John raises to Fam to satisfy a strong definiteness feature, such as the [proper] feature mentioned by Ghomeshi and Massam (2009). Both proper and possessive nominals are FamPs with a null D, which is why restrictive nominals are possible. As argued also in §5.4.2, restrictive nominals and FamPs are predicates that can intersect, and are converted to type e only by a higher head such as ι or Gen. In contrast, when the nominal is headed by the definite determiner the, there is no split of definiteness between ι and Fam. The determiner the encodes both and thus no (DP) restriction is possible. Definiteness is spelled out by a single vocabulary item, the, and thus DefP, being of type e, cannot combine with a predicative modifying nominal. 2⁴ Alternatively, we could assume that the proper name moves to the specifier of FamP since the modified nominal can also be phrasal. For example: Max and Leo your cousins (not Max and Leo at your school). Many thanks to an anonymous reviewer for this observation.

126 maria kyriakaki Hence, the definiteness underspecification that the Greek article exhibits is not unique. Rather, through a comparison with English, it seems not unusual that a definite determiner can be a Fam head without ι. In summary, the definite determiners we have examined thus far are of two types: fully specified and underspecified. The English definite determiner the is an example of a fully specified definite determiner. As we have seen, the does not appear with restrictive modification by nominals, with proper names, or with generic plurals.2⁵ Semantically, it spells out both familiarity and uniqueness. Examples of underspecified definite determiners are English null D and Greek D. These definite determiners encode only familiarity. They can thus be modified by other definite nominals. When they are modified, uniqueness arises from the intersection of the matrix nominal and the modifying DP. When these underspecified definite DPs are not modified, uniqueness arises from contextual restrictions.2⁶ Turning to the structure, definiteness is decomposed in the case of underspecified English and Greek Ds into FamP and ιP. FamP contains a set of salient entities, while ιP gives us a unique entity, as in (39a). If the determiner is fully definite, definiteness is mapped to a single projection, as in (39b). The fully specified English determiner the is associated with a single Def head that encodes both familiarity and uniqueness.

(39) a.

ιP

b.

ι

FamP

DefP

Def the

Fam English:∅ Greek: o/i/to, etc. 2⁵ Note that English also has a form of definite generic in the singular, as in The beaver has a flat tail (Carlson 1977). These singular definite generics are less than fully productive, being restricted to ‘wellestablished’ kinds (in Lyons’s (1999) terms, the ‘familiar to us all’ kinds); common nouns which are too general do not have a definite generic counterpart. (E.g. The airport is a busy place is infelicitous on a generic reading, in contrast with the generic bare plural Airports are busy places.) These facts might be explained by positing another type of the. The question would be what its specification is. It cannot be Fam, since restriction by a nominal is not possible. However, it could be a generic counterpart of the, which identifies a unique entity, here a prototypical member of a set. More research on this is needed. 2⁶ In §5.3, we have also seen that the generic operator in Greek definite generics singles out a set of entities with specific properties that is familiar to us all. In such cases, a GenP is projected above FamP.

definite expression and degrees of definiteness 127

5.4.2 A compositional note on determiners of modifying nouns In this section I now consider determiners in modifying nominals, and specifically English modifying nominals introduced by the. It is argued that English modifying nominals are predicative and adjoin low in the same way that relative clauses do, but are interpreted higher with the matrix Fam, which is what makes such structures possible.2⁷ In the case of a matrix DefP, which is headed by the, the modifying nominal cannot be interpreted with it, since the matrix DefP is of type e and cannot compose with the predicative modifier. To start with, we have seen that English modified proper names and possessives are predicative FamPs. But what about modifying nominals themselves? Do they involve a Fam head, and thus project a predicative FamP, or are they argumental nominals containing a fully definite head? A first intuition would be that modifying nouns should be predicative. Compositionally, for the head noun and the modifying noun to combine, they must both be predicates to ultimately combine with ι: (40)

ιPe ι

FamP⟨e,t⟩ Fam

NumP⟨e,t⟩

Head noun⟨e,t⟩

Mod. noun⟨e,t⟩

If the modifying nominal is of type ⟨e,t⟩, it will be able to compose with modified nominals by Predicate Modification. In Greek, modifying nominals are predicative FamPs headed by the underspecified Fam. They can thus easily combine with the predicative head nominal. In English, though, modifying

2⁷ Somewhat in spirit of Lekakou and Szendrői (2012), the two nominals involved denote intersecting sets. In contrast to Lekakou and Szendrői’s analysis, though, these nominals are shown to be in an asymmetric syntactic relation to each other. The modifying nominal originally adjoins low and carries an index in the sense of Heim and Kratzer (1998), as discussed below.

128 maria kyriakaki nominals can be headed either by the underspecified null Fam, as in (41a), or by the fully specified determiner the, as in (41b): (41) a. John [my friend]mod (not my colleague) b. John [the linguist]mod (not the mathematician) In (41a), my friend is a possessed nominal, which, as we have seen, is headed by the null Fam. Its semantic type is ⟨e,t⟩, and it can thus compose with the modified noun by Predicate Modification. Hence, in the case of modifying possessors, the head nominal combines with them straightforwardly. In expressions like in (41b), though, the modifier the linguist is not a FamP: it is headed by the definite article the, which we have said is a fully specified definite determiner. Compositionally, the predicative head noun should not be able to combine with the fully definite nominal. Otherwise, the combination of the two would wrongly yield an object of type t, and the determiner would not be able to apply to it by Functional Application. Setting aside a detailed derivation for now, for phrases such as John the linguist we would have roughly the following structure:2⁸ (42) a. John the linguist b. ιPe ι

FamP Fam ∅



NumPt

John⟨e,t⟩

the linguiste

In all cases, modifiers should not be treated as arguments, but rather as predicates. Heim and Kratzer (1998) also argue that modifiers, like relative clauses or DPs, carry an index and should thus be predicative. Hence, modifying DPs can be treated as predicative in the same way other modifiers are. To treat modifying DPs as ⟨e,t⟩ we can adopt a parallel view to Heim and Kratzer’s analysis of restrictive relative clauses, in which Predicate 2⁸ A more detailed structure is presented towards the end of this section.

definite expression and degrees of definiteness 129 Abstraction applies, turning the argument into a predicate.2⁹ The restrictive modifying DP (in our example the linguist) is now of type ⟨e,t⟩, and it can now compose with the ⟨e,t⟩ predicate of the matrix nominal by predicate modification.3⁰ One might propose that the modifier headed by the article the in these constructions is not a fully definite phrase, but rather a FamP, similar to the Greek modifying FamP. However, this assumption does not seem to be on the right track. The English modifier headed by the seems to be more readily acceptable in contexts where the property in question is ascribed to one individual in the relevant context:31 (43) a. #/?? John the doctor and Mary the doctor b. John the doctor and Mary the other doctor Examples such as (43b) are perfectly fine, which is exactly what is expected, since one distinct property is assigned to a unique individual.32 Further evidence that the English modifier headed by the must exhaustively identify a unique person in the context comes from examples like those in (44). In the following examples, the modifier is introduced by the determiner the, which in regular DPs has been argued to presuppose uniqueness. If this is what the always contributes, it should also fully identify a unique entity in these modifiers too. This prediction is borne out. The presence of more than one definite modifier results in ungrammaticality. This is true of both proper and possessive DPs as shown in (44a) and (44b), respectively:33 (44) a. b.

∗ ∗

John the doctor the cardiologist my friend the doctor the cardiologist

2⁹ Predicate Abstraction is defined by Heim and Kratzer (1998: 96) as follows: ‘If α is a branching node whose daughters are a relative pronoun and β, then ⟦α⟧ = λx ∈ D . ⟦β⟧x .’ 3⁰ Alternatively, another type-shifting operation can be used, changing the semantic type of the modifying DP from e to ⟨e,t⟩, giving a singleton set containing the unique individual denoted by the e-type DP (see Kyriakaki 2011 for more on this alternative). 31 Note that in a very specific context, the phrase John the doctor and Mary the doctor could be semantically acceptable, e.g. in a context where there is more than one person named John, and more than one person named Mary, and exactly one John is a doctor and exactly one Mary is a doctor. Many thanks to Daniel C. Hall for pointing out this possibility. 32 Many thanks to Elizabeth Cowper for bringing this example to my attention. 33 Ileana Paul (p.c.) observes that multiple possessive modifiers do not seem possible either: ∗ John my brother your coach. The modifier thus seems to involve a more complex semantics that remains to be further investigated.

130 maria kyriakaki This is not the case for Greek, as also predicted in this analysis. The Greek determiner does not require a unique entity, and thus not only should it allow for nominal modification, but also it should not block the presence of more than one nominal modifier.3⁴ This prediction is borne out: (45)

o filos mu o jatros o karðioloγos the friend.m.nom my the doctor the cardiologist ‘my friend the doctor the cardiologist’

Hence, in English the determiner the consistently identifies a unique entity. This is true when it heads a matrix nominal, and thus does not allow modification by another DP, but also when it heads a modifier, and thus no other modifier is allowed. This shows that the definiteness in modifying and modified DPs headed by the is of the same type, i.e. they are fully definite phrases and not FamPs. The difference is that modifying the-nominals carry an index that makes them predicates, while modified nominals do not carry such an index, and are thus argumental. Turning to the syntactic position of English modifying nominals, I argue that they attach lower, to NP, and not to nP. If they attached to nP, as in (46), we would make two false predictions: first, that restrictive DPs should occur prenominally, and second, that they should also co-occur with adjectives which adjoin to nP.3⁵ (46)

ιPe ι

FamP⟨e,t⟩ Fam ∅

John

NumP Num

nP⟨e,t⟩ DP⟨e,t⟩ the linguist

nP⟨e,t⟩ ⟨John⟩

3⁴ See Kyriakaki (2011) for a full analysis of the structure of the Greek modifying nominal. 3⁵ Following Haegeman (1991) and Adger (2003), I assume that adjectives are adjuncts to nP, and not specifiers of nP as argued by Cinque (1993). See Alexiadou (2001a) for a discussion of the second perspective.

definite expression and degrees of definiteness 131 Since proper names raise higher, we cannot test these predictions with proper nouns. However, other types of nominals that allow restrictive DPs, such as possessive DPs with pronominal determiners, do not involve any movement to the DP.3⁶ As shown in (47), a phrase like my friend the accountant (not the linguist) becomes ungrammatical if the restrictive DP is moved to a prenominal position (such as the specifier of FocP, as in (47b)). Moreover, if we assume that the modifier is an adjunct to nP, as in (47c), where the matrix noun stays in n, the ungrammatical ∗ my the accountant friend is derived: (47) a. b. c. d.

my friend [the accountant] (not the linguist) [FocP the accountant]i my friend ti (not the linguist) ∗ my [the accountant] friend ∗ ιP ∗

ι

FamP⟨e,t⟩ NumP my

FamP Fam ∅

NumP Num

nP⟨e,t⟩ DP⟨e,t⟩ the accountant

nP⟨e,t⟩ friend

If, however, we assume that the modifying DP adjoins lower than nP, to NP, the correct structure is derived:

3⁶ Recall that it is only proper nominals that move higher to satisfy a strong [proper] feature on Fam. Given that possessed common nominals lack such a feature, there can be no movement to Fam.

132 maria kyriakaki (48) a. my friend [the accountant] (not the linguist) b.

ιPe ι

FamP⟨e,t⟩ NumP my

FamP⟨e,t⟩ Fam ∅

NumP⟨e,t⟩ Num

nP⟨e,t⟩ n⟨e,t⟩ n

N friend

NP⟨e,t⟩ DP⟨e,t⟩ the accountant

NP ⟨friend⟩

More evidence that the restrictive DP should not be an adjunct to nP comes from the fact that restrictive DPs and adjectives do not co-occur in the same position, which entails that they cannot co-occur as adjuncts to the same functional projection. Thus, a phrase such as my friend cannot be modified by an adjective immediately followed by a restrictive DP such as the accountant, as in (49a). The only way to formulate such a modification would be to have the accountant follow the noun, as in (49b), which is the order derived by the structure in (48). (49) a. b.



my good, the accountant friend my good friend the accountant

This possibility would capture an interesting parallel between Greek and English. In Greek, adjectives adjoin higher up, to NumP, and restrictive DPs adjoin to nP.3⁷ In English, adjectives adjoin to nP and restrictive DPs to NP.3⁸ 3⁷ In Kyriakaki (2011) it is argued that the noun must move to NumP based on evidence provided by genitives, and that adjectives should therefore adjoin to NumP, contra Alexiadou (2003), who argues that it is the adjective that moves past the noun, while the noun stays in situ. Following Pollock (1989), I take the more common view that the noun moves, while the adjective as an adjunct stays in situ. Therefore, the noun moves so that nominal inflection takes place in Num, and adjectives adjoin above it. In effect, the postnominal modifying nominal is derived by its adjunction to nP. 3⁸ Other possibilities would be to assume either that the restrictive nominal adjoins to the right (deriving the order Noun–DPrestr ), or that the noun moves to Num. In the first case, such an assumption would be problematic since the position taken in this work is that adjuncts consistently left-adjoin (or are in the specifier position, which is to the left of heads; see Adger 2003 or Kayne 1994 for

definite expression and degrees of definiteness 133 When the noun moves to n, it moves past the restrictive DP, giving the correct order as shown in (48). With this structure, we can now easily derive the correct word order of English modified nominals. Semantically, too, this structure does not cause any problems for the interpretation of the nominal as an argument. The ⟨e,t⟩ modifier intersects with the ⟨e,t⟩ NP. It is interpreted higher at LF, possibly in the specifier of Fam, i.e. above Fam but below ι, since such modifying nominals are only possible with FamPs. In this way, Fam can select the set containing the familiar entity ‘my friend the accountant’, and the ι operator will select the unique individual in that set. Also, this structure does not derive ungrammatical constructions of the form ∗ the teacher the chemist because of interpretation effects. The modifier, here the chemist, which adjoins low, cannot be interpreted higher, since the higher the, of type ⟨⟨e,t⟩,e⟩, will compose with its noun complement of type ⟨e,t⟩ and return an argument DP, i.e. a DefP of type e. If the ⟨e,t⟩ modifier is raised higher at LF, it cannot compose with DefP at this point, as opposed to the predicative FamP, and the derivation will crash. Therefore, a correct result is that nominals of the form ∗ the teacher the chemist cannot be derived. In sum, English modifying nominals introduced by the are ⟨e,t⟩ predicates that intersect with modified nominals by Predicate Modification. This type of predication is thus not a result of the lack of definiteness, since in all cases uniqueness is presupposed. The restrictive modifying nominal is not a simple predicate, though, since it can refer to only one entity. Rather, it is a referring expression that denotes a unique property. We have seen this to be a one-to-one relation; that is, a unique property is assigned to a unique entity, and a unique entity cannot be assigned more than one property. In contrast, we have seen for Greek that this relationship is one to many; that is, in Greek a unique entity can be assigned more than one property, since the matrix nominal may be modified by more than one restrictive nominal. In this analysis, this variation is nicely accounted for by the lexical difference between the English and Greek definite determiners.

5.5 The three types of definite determiners So far, definite expressions have been shown to involve various aspects of definiteness depending on the specification of the definite determiner.

assumptions on base structure). In the second case, we would have to assume also that adjectives adjoin to NumP and not nP. In the absence of compelling evidence in favour of either of these possibilities, I pursue here the possibility that DP modifiers adjoin to NP.

134 maria kyriakaki Focusing on definite expressions, such as restrictive nominals, proper names, generics, and possessors, it has been argued that in addition to fully definite and expletive determiners, there is a third type of determiners, those that are partially specified for definiteness. This in turn gives rise to three degrees of definiteness, full, partial, and zero definiteness. A closer look at Greek nominals shows that the definite determiner of this language exhibits a systematic and flexible, but not entirely free distribution: it systematically appears with proper names, definite restrictive nominals, definite possessors, and definite kind-denoting generics, but it cannot appear in purely indefinite or existential contexts. These properties strongly suggest that such determiners are not pure expletives, but rather that they do in fact contribute some aspect of definiteness. Similar properties are also exemplified by the English null definite determiner, providing further support for the proposal that determiners can be partially definite. Null D may also appear with restrictive nominals, generic plurals, and possessives. In contrast to the Greek definite determiner and English null D, it is shown that the English definite determiner the shows a fixed, limited distribution. It does not combine with restrictive nominals, generic plurals, or possessive nominals of the form discussed in this work. Unlike the Greek definite article and English null D, English the disallows any additional definiteness restriction made on the nominal. The English definite determiner the exemplifies a fully specified definite determiner. True expletive determiners, in contrast, are not specified for definiteness and show a freer distribution, since they can be optional (Longobardi 1994), as illustrated in (50), repeated from (5a).3⁹ (50)

(Il) Gianni mi ha telefonato. the Gianni me has called ‘Gianni called me up.’

[Italian]

In addition to their optionality, truly expletive determiners can also appear in indefinite contexts. As Zamparelli (2002: 8) shows, the Italian definite article can appear in existential indefinite contexts:

3⁹ It should be noted however, that there is variation in the use of the Italian definite article in Italian dialects, and even within Standard Italian (see Longobardi 1994 for more). Future research into this language can determine how definiteness of the sort discussed here is manifested in the various regions.

definite expression and degrees of definiteness 135 (51)

C’ é {Gianni / il mio cane} in giardino. there is Gianni the my dog in garden ‘{Gianni / My dog} is in the garden.’ Cf. English: ∗ There is {Gianni / my dog} in the garden.

(52)

In cantina ci sono [i topi] e sotto il lavello in basement there are the mice and under the sink vivono [gli scarafaggi]. live the cockroaches ‘There are mice in the basement and cockroaches living under the sink.’

(53)

Che fai per mestiere? Fotografo [gli uccelli]. what do for living photograph.1sg the birds ‘What do you do for a living? I photograph birds.’

These examples clearly show that the Italian definite article is a true expletive, since it can be optional and can appear with existential indefinite nominals. The Italian definite determiner is therefore fully unspecified for the semantics of definiteness. In conclusion, there are three types of morphologically definite determiners that can appear with definite expressions: fully, partially, or non-specified for definiteness. In turn, these give rise to three degrees of definiteness: full, partial, and zero definiteness.⁴⁰

5.6 Conclusions and directions for further research This work proposes a view of definiteness that takes into consideration the semantically definite properties of determiners previously taken to be expletives, and argues that such determiners do in fact contribute definite semantics to the nominal. The degree of their definiteness contribution depends on the type of specification of each determiner. That is, determiners may be fully definite (specified for both familiarity and uniqueness), partially definite (specified for familiarity only), or expletive (unspecified for definiteness). Determiners that lack a specification for uniqueness may still appear in nominals that receive a unique interpretation. In the case of predicative FamP, ⁴⁰ For a detailed view of the three types of definite determiners and the criteria used to identify expletives, see also Kyriakaki (2014).

136 maria kyriakaki uniqueness arises from contextual restrictions, or in the case of restrictive nominals, from intersection with the matrix nominal. In the case of fully definite nominals, uniqueness arises from the definite determiner spelling out both uniqueness and familiarity. Both ιPs and FamPs appear in definite contexts only. However, predicative Fams can be modified by other definite nominals, they are possible with proper names, and they can introduce generic kind-denoting only nouns. Finally, true expletives, which are non-specified for definiteness, can appear even in non-definite contexts. The proposed model allows for the various degrees of definiteness to arise in a simple and straightforward manner. Additionally, the proposed model can also predict the syntactic-semantic properties of definite determiners. Fully specified definite determiners are correctly predicted to show a more limited distribution. An example of this type of definite determiner is the English determiner the. It picks out a unique, familiar individual, and further restriction is therefore impossible. Underspecified Fam determiners select a familiar set, which can be further restricted. They are thus predicted to show a more flexible distribution. Examples of such determiners are the Greek definite article and English null D. Finally, true expletives are completely unspecified for definiteness. The Italian determiner is such an example. Its zero specification for definiteness enables it to appear in various contexts, including existential indefinites, as well as to be optional as well. Hence, determiners spelling out all features of definiteness are correctly predicted to exhibit a limited distribution, while those spelling out fewer or no features predicted to exhibit a more flexible distribution. This work also offers new insights about the essence and mapping of definiteness, while it also provides us with some new intuitions about the notion of expletives. By assuming that specification of definiteness can come in three degrees, interesting cross-linguistic applications may arise in languages where definiteness is morphologically marked with a determiner or even an affix. Additionally, the syntactic manifestations of definiteness may differ crosslinguistically. By recognizing that definiteness consists of semantic features that can be mapped onto the syntax in different ways, various structures can be accounted for, as in the case of the definite expressions considered in this work. Hence, by this proposed mechanism, a new syntactic-semantic view of definiteness is offered that allows for further cross-linguistic research in the field.

definite expression and degrees of definiteness 137

Acknowledgements My deepest thanks go to Elizabeth Cowper to whom this work is dedicated, as well as to Daniel Currie Hall and Bronwyn Bjorkman, the editors of this volume, for their help and support, and the anonymous reviewers for all their invaluable feedback. I am particularly indebted to members of the Syntax Project at the University of Toronto, and to my current affiliation, the American College of Greece.

6 Cross-linguistic contrasts in the structure of causatives in clausal nominalizations Martha McGinnis

It is traditionally assumed that the causee ‘subject’ of a causativized transitive predicate falls within the syntactic scope of the causativizer, as in (1), where X represents the causer, Y the causee, and Z the object of an embedded transitive predicate. (1)

[X cause [Y verb Z]]

However, I will argue that gerundive nominalizations in Georgian (masdars) provide evidence for a different structure, schematized in (2): the causee is projected by a type of applicative phrase, which merges between the causative vP and a higher Voice projection that introduces the causer external argument. (2)

[X voice [Y appl [cause [verb Z]]]]

Such an analysis is fairly novel, though by no means unprecedented. Pylkkänen (2008: 120) proposes a similar structure for causativized unergatives in Japanese, with the causee merging outside the causative v. Kim’s (2011) analysis of causativized transitives in Korean has the causee merging in the specifier of a high instrumental applicative head, though within the scope of the causative element (see also Legate 2014 for Acehnese). Georgian masdars provide evidence for both aspects of these structures. A key piece of evidence involves nominalizations (masdars) in which causative morphology can be expressed, while a causee cannot. Specifically, the causee of a causativized transitive predicate can be expressed in the verbal context (3a), but not in the masdar (3b).1 1 Orthographic transcriptions and glosses have been regularized, and morpheme boundaries added where straightforward. The Georgian alphabet makes no distinction between upper and lower case; Martha McGinnis, Cross-linguistic contrasts in the structure of causatives in clausal nominalizations In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Martha McGinnis. DOI: 10.1093/oso/9780198817925.003.0006

the structure of causatives in clausal nominalizations 139 (3) a. redakTorma KiTas es Ceril-i tavis žurnalši editor.erg Kita.dat this letter-nom self.gen magazine.in gamoakveqnebina. publish.caus.aor.3sg ‘The editor had Kita publish this letter in his magazine.’ (Aronson 1990: 307) b. Ceril-is gamokveqnebineba (∗ KiTas{-tvis/mier}) letter-gen publish.caus.n Kita.dat/gen(-for/by) moulodnelia.2 be.unexpected.pres/fut.3sg ‘The causing of the letter to be published (∗ by Kita) is unexpected.’ I argue below that the Georgian masdar excludes causees because they merge high in the verb phrase, outside the vP selected by the nominalizer that derives the masdar. Aronson (1990) refers to the masdar as a ‘verbal noun’, a term also used for English gerunds. Morphologically, the masdar is based on a stem closely resembling that of a future tense verb in Georgian, as shown in (4). (4) a. da=i-Cq-eb-s. pv=vce-begin-ts-pres/fut.3sg ‘She/he will begin it/them.’ b. da=Cq-eb-a pv=begin-ts-n ‘beginningn ’ upper-case letters in the transcriptions represent ejective obstruents. Abbreviations are as follows: 1: first person; 3: third person; acc: accusative; aor: aorist; appl: applicative; ben: benefactive; caus: causative; dat: dative; dem: demonstrative; erg: ergative; gen: genitive; ger: gerund; ipfv: imperfective; n: nominalizing head; nom: nominative; opt: optative; pres/fut: present/future; part: participant (1/2 person); perf: perfect; pl: plural; poss: possessive; pfv: perfective; ptcp: participle; pv: preverb; refl: reflexive; rel: relative clause; sg: singular; TAM: tense/aspect/mood; ts: thematic suffix; v: verbalizing head; vce: Voice. 2 Many Georgian speakers I have consulted reject masdars containing causative -in. Likewise, Harris (1981: 152) contrasts a finite clause containing causative -in (i), which her consultants accept, and the corresponding masdar (ii), which they do not. This example is considered ill-formed regardless of whether the causer and causee are expressed (Rusudan Asatiani, personal communication). I leave this matter for further research. (i) Mepe-m ga=a-Cmend-in-a sasaxle king-erg pv=vce-clean-caus-aor.3sg palace.nom ‘The king had the palace cleaned.’ [or ‘The king had him/her clean the palace.’] (ii) ∗ sasaxl-is ga-Cmend-in-eb-a mep-is mier palace-gen pv-clean-caus-ts-n.nom king-gen by ‘the king’s having the palace cleaned’ [or ‘the king’s having him/her clean the palace’]

140 martha mcginnis The future verb form (4a) includes some affixes not found in the masdar (4b). For example, the future form includes a tense/aspect/mood (TAM) suffix, here -s, which indicates a clause in present or future tense, with a thirdperson singular subject; the distinction between present and future tense is usually signalled by the presence of the aspectual preverb (pv) in future. Note that preverbs can be complex: the preverb mo-, indicating action directed toward a speech participant (a speaker or addressee), can occur on its own or immediately following other preverbs (Aronson 1990: 42). The future form can also include a prefix associated with transitivity alternations, which I analyse as a realization of the Voice (vce) head (here, i-). The remaining morphology can be found in both the tensed verb form and the masdar. This includes the bound lexical root of the verb (here, Cq- ‘begin’), followed by a thematic suffix (ts; here, -eb), which varies depending on the root; the form of this thematic suffix sometimes alters in the masdar. Both forms can also bear an aspectual preverb (here, da=), indicating perfective aspect. Preverbs appear in the future, conditional, perfect, and pluperfect TAM categories, as well as in perfective uses of the conjunctive and aorist. The preverb also indicates perfectivity in the masdar (Aronson 1990: 50, fn. 6). Thus, a masdar bearing a preverb has a perfective interpretation (5a), while a masdar without one has an imperfective interpretation (5b). (5) a. da=Cer-∅-a pv=write-ts-n ‘writingn.perf ’

b. Cer-∅-a write-ts-n ‘writingn.ipfv ’

Table 6.1 compares the morphological forms of the masdar and verb stems in the various TAM categories. These categories are grouped into three series, associated with characteristic case-marking patterns. For example, the subject and direct object of a transitive clause show nom-acc marking in Series I, ergnom in Series II, and dat-nom in Series III. Masdars and verbs can be marked as perfective (pfv) or imperfective (ipfv). As noted above, while masdars permit thematic suffixes and aspectual preverbs, they lack TAM suffixes, as well as a set of prefixes associated with transitivity alternations. One subset of these—including u- and e- in Series III (Table 6.1)—is associated with a dative argument, such as an applied (indirect) object, or the dative subject of the perfect and pluperfect. Another set (including a- and i-) is associated with the presence or absence of an external argument referentially distinct from the object. These prefixes, which I will analyse as Voice prefixes, will be discussed further in §6.2. As (6a) illustrates,

the structure of causatives in clausal nominalizations 141 Table 6.1 Masdar and verb stems in different TAM categories Masdar

(pfv) (ipfv)

da=

Series I

Present (ipfv) Future (pfv) Imperfect Conditional Conjunctive (ipfv) Conjunctive (pfv) Aorist (pfv) Aorist (ipfv) Optative Perfect Pluperfect

xaT -av -s xaT -av -s xaT -av -d -a da= xaT -av -d -a xaT -av -de -s da= xaT -av -de -s da= xaT -a xaT -a da= xaT -o da= u- xaT -av -s da= e- xaT -a

Series II

Series III

da=

xaT -v xaT -v

-a -a

‘the act of painting (result)’ ‘the act of painting (activity)’ ‘(s)he is painting it’ ‘(s)he will paint it’ ‘(s)he was painting it’ ‘(s)he would paint it’ ‘if (s)he were painting it’ ‘if (s)he will paint it’ ‘(s)he painted it (result)’ ‘(s)he painted it (activity)’ ‘that (s)he paint it’ ‘(s)he has painted it’ ‘(s)he had painted it’

the masdar also lacks person and number agreement, such as the first-person v- and plural -t seen in (6b). (6) a. a=šen-eb-a pv-build-ts-n ‘buildingn ’

b. a=v-a-šen-eb-t. pv=1-vce-build-ts-pl ‘We will build it.’

In the following sections, I argue that the Georgian masdar involves a nominalizing head that selects a vP complement, and that both causers and causees are excluded from masdars because they merge outside the causative vP. §6.1 demonstrates that masdars are nominal. §6.2 argues that the nominalizing head in a masdar is v-selecting, based on evidence from morphosyntax, argument structure, and causatives. §6.3 provides evidence that causees merge outside vP.

6.1 Masdars are nominal Masdars are structurally and distributionally nominal (Harris 1981; Vamling 1989; Wier 2014). Like simple nouns, masdars allow a genitive modifier to precede the nominal (7). The fixed ordering of this modifier, relative to the nominal, contrasts with the freer ordering of verbal arguments. (7) a. givi-s mo=Tqu-eb-a Givi-gen pv=lie-ts-n ‘Givi’s lie’

(Wier 2014)

142 martha mcginnis b. karteli gramaTiK-is sCavl-a Georgian grammar-gen study.ts-n ‘the studying of Georgian grammar’

(Aronson 1990: 76, 83)

Like English nominalizations, masdars can be either complex event nominalizations (CENs), with an eventive (process) reading as in (8a) and (9a), or result nominalizations (RNs), with an entity reading as in (8b) and (9b) (Léa Nash, personal communication). CEN and RN masdars are morphologically identical. (8) a. am siTqvebis gamotkvma čatarda laboratoria-ši dem words.gen pronunciation.nom occur.aor.3sg laboratory-in ivnis-ši. June-in ‘The pronunciation of these words took place in the laboratory in June.’ b. am siTqvas ori gamotkvma akvs. dem word.dat two pronunciation.nom have.pres/fut.3sg ‘This word has two pronunciations.’ (9) a. am pilmis damtavreba sami dghe gagrdzelda. dem film.gen ending.nom three days continue.aor.3sg ‘The ending (finishing) of this film continued for three days.’ b. ver viTan am pilmis damtavreba-s. can’t 1sg.stand dem film.gen ending-acc ‘I can’t stand the ending of this film.’ Some masdars allow only an eventive reading, for example xaTva ‘act of painting’ or daCqeba ‘act of beginning’, comparable to English legalization. Others allow only a result reading, such as šecdoma ‘error’ or šexvedra ‘meeting’, comparable to English decision. It can be argued that the distinction between CENs and RNs is not primarily a structural one. For example, either result n or eventive n can attach directly to a lexical root. This possibility can be indicated by the choice of a special allomorph for the nominalizing head (Marantz 1997; Embick 2010), like -th in growth. When the nominalizing head attaches directly to the lexical root, no causative reading is available for grow, and the root-conditioned -th allomorph of the nominalizer appears (10a). An intervening verbalizing head makes a causative reading available for grow, but requires the use of the default

the structure of causatives in clausal nominalizations 143 nominalizer -ing (10b). Nevertheless, the root nominalization growth allows either an eventive reading (11a) or a result reading (11b). (10) a. {the / ∗ John’s} [[√grow ] -thn ] of the tomatoes b. {the / John’s} [[[√grow ] vcaus ] -ingn ] of the tomatoes (11) a. the constant growth of the tomatoes all summer b. the three small growths on his left index finger This observation poses difficulties for the view that CENs contain both vP and AspP, while RNs either lack both (Borer 2003), or contain only vP (Alexiadou 2001b). Both CENs and RNs can also contain overt verbal suffixes (Borer 2003; Harley 2009, inter alia), as shown in (12). Indeed, such a verbalizer need not even be root-attached, as illustrated in (13). (12) a. The organ-iz-ation of the workers took place only gradually. b. The workers eventually formed a strong organ-iz-ation.

CEN RN

(13) a. The fiction-al-iz-ation of Shakespeare’s life took place after his death. CEN b. Historical novels are, by definition, fiction-al-iz-ations. RN I conclude that the semantic distinction between RNs and CENs arises not from a structural difference, but from a difference in the semantics of the nominalizing head n (Moulton 2014). It appears that both heads can combine with a root, as in (11), a vP, as in (12), or even a vP containing an aP, as in (13). However, a distributional distinction does arise in causative masdars, which can only be interpreted as CENs, as discussed in §6.2.3. Result masdars share quantification properties with simple count nouns (Harris 1981; Wier 2014). For example, the result reading of (14a) is available for the plural (14b). Result masdars also allow numeral quantifiers (14c). (14) a. mo=gon-eb-a pv=remember-ts-n.nom ‘remembering, memory’ b. mo=gon-eb-eb-i pv=remember-ts-pl-n.nom ‘memories’ c. ert-i txovn-a one-nom request-n.nom ‘one request’

144 martha mcginnis Also like nouns, masdars allow modification by adjectives (15a) and relative clauses (15b) (Harris 1981; Wier 2014). (15) a. sašinel-i mo=Tqu-eb-a terrible-nom pv=lie-ts-n.nom ‘the terrible lie’ b. čxub-i, romel-i-c v-nax-e fighting-n.nom which-nom-rel 1-see-aor.part ‘the fight which I saw’ As regards their external distribution, masdars—like other inanimate nominal expressions in Georgian—can function as subjects, objects, or objects of postpositions, and are case-marked accordingly. There is also an adjunct (‘during/while’) use of masdars suffixed with -isa-s (Wier 2014). Masdars with an eventive reading share some distributional properties with clauses, in that they can serve as complements of verbs such as aPireb- ‘intend’ (16a), which otherwise take finite clausal complements (16b). (16) a. v-aPireb-∅ [Ceril-is da=Cer-a]-s. 1-intend-pres/fut letter-gen pv=write-n-dat ‘I intend to write a letter.’ (lit., ‘I intend the writing of a letter.’) b. v-aPireb-∅ [da=v-Cer-o Ceril-i]. 1-intend-pres/fut pv=1-write-opt.part letter-nom ‘I intend to write a letter.’ (Wier 2014) Unlike finite clauses, however, masdars can undergo scrambling, and are themselves opaque to scrambling (Wier 2014). Thus the genitive internal argument of the masdar cannot scramble to the left edge of the matrix clause in (17a), unlike the nominative object of the embedded optative verb in (17b). (17) a.

b.



Ceril-isi v-aPireb-∅ [ti da=Cer-a]-s. letter-gen 1-intend-pres/fut pv=write-n-dat ‘I intend to write a letter.’ (lit., ‘I intend the writing of a letter.’) Ceril-ii v-aPireb-∅ [ti da=v-Cer-o]. letter-nom 1-intend-pres/fut pv=1-write-opt ‘I intend to write a letter.’

In short, masdars resemble nominal expressions in a variety of ways, both internally and externally, though their distribution also overlaps with that

the structure of causatives in clausal nominalizations 145 of full clauses. I propose that the masdar is a nominalization, in which a verbal phrase combines with a noun-forming head, n, instead of with the inflectional heads underlying the Voice, TAM, and agreement affixes. In the following section, I argue that the nominalizing head in a masdar selects vP, based on evidence from morphosyntax and argument structure, particularly in causative clauses.

6.2 Masdars are v-selecting Both cross-linguistically and within a given language, gerundive nominalizations can be divided into two main categories: (i) those that license direct objects via structural Case, like English PRO-ing and poss-ing gerunds (18a– 18b), and (ii) those whose internal arguments have the appearance of the arguments of nominals, like English ing-of gerunds (18c), also known as mixed nominalizations (Abney 1987; Cowper 1995; Harley and Noyer 1998; Bjorkman, Cowper, and Siddiqi 2018, inter alia). (18) a. Alexi avoided [PROi dropping the ball]. b. We resented [Bob’s losing the race]. c. [The selling of essays] annoyed the professor.

PRO-ing poss-ing ing-of

This distinction in object licensing can be attributed to the presence or absence of a quasi-functional head, Voice, which is responsible for introducing the external argument and licensing structural Case on the object (Kratzer 1996). Kratzer proposes that ing-of gerunds lack Voice, but contain a VP projection, allowing adverbial modifiers (19) (see also Moulton 2004). (19) a. The [VP shutting of the gates regularly at ten o’clock] had rendered our residence very irksome to me. b. The [VP running of marathons too quickly] causes exhaustion. c. The [VP playing of cards frequently] improves a person’s memory. The verb can moreover be analysed as syntactically complex, consisting of a category-neutral root in combination with a verb-forming head, v (Marantz 1997; Arad 2003). By contrast, PRO-ing and poss-ing gerunds can be analysed as containing Voice. On the other hand, these gerunds lack finite T, associated with tensemarking, agreement, and an overt nominative subject. Similarly, Turkish

146 martha mcginnis gerunds contain Voice but lack finite T, so the external argument is genitive, while the object is accusative, as in (20).3 (20)

(ben) [kız-ım-ın I daughter-my-gen razı ol-du-m. consent-past-1sg

viski-yi whiskey-acc

ic  ¸-me-sin]-e drink-ger-3sg-dat

‘I consented to my daughter’s drinking the whiskey.’ (George and Kornfilt 1981) In Georgian, by contrast, masdars cannot take an accusative object; the object must be genitive (21). This suggests that masdars lack VoiceP, as Kratzer proposes for English ing-of gerunds. (21)

v-irčev [ami-s ga=Ket-eb-a]-s. 1-prefer.pres/fut this-gen pv=do-ts-n-dat ‘I prefer to do this.’

(Vamling 1989: 99)

As we will see below, Georgian masdars provide new evidence for the separation of VoiceP and vP (see also Pylkkänen 2008; Harley 2013). I propose that they involve a nominalizing head that selects, not a VoiceP or the lexical root of a verb, but a phrasal vP, as in (22). (22)

nP n

vP √

v √

(DP)

In what follows, I present evidence for this analysis from morphosyntax, argument structure, and masdars based on causativized transitive predicates.

3 An anonymous reviewer notes that a Turkish gerund can also contain a passive-marked verb; this is impossible in a Georgian masdar.

the structure of causatives in clausal nominalizations 147

6.2.1 Masdars are v-selecting: Evidence from morphosyntax The morphosyntax of the masdar provides evidence that it contains vP, but not VoiceP. I analyse the verbal category head v as separate from the lexical root, which I assume is category-neutral (√). Arad (2003) argues (citing Léa Nash, personal communication) that Georgian nouns, verbs, and adjectives can share a morphologically identifiable root, as illustrated by √cx in (23), which is semantically associated with the notion of heat: (23) a. b. c. d. e.

acxobs namcxvari sicxe cxeli acxelebs

‘to bake’, v. ‘cake’; literally, participle of ‘bake’ ‘fever’, n. ‘hot’, adj. ‘to heat’, v., from (23d)

In a verbal context, a lexical root in Georgian is typically followed by a thematic suffix. This suffix shows contextual allomorphy that depends on the choice of lexical root, properties of Voice, and even the TAM category. An adjective like lamaz- (24a), a noun like inTeres- ‘interest’ (24b), or an adverb like mzad- (24c), with the same lexical root as a verb, lacks this thematic suffix (Aronson 1990: 71–2), suggesting that the suffix is associated with v. (24)

Class I transitive a. ga=a-lamaz-eb-s pv=caus-beautiful-ts-pres/fut.3sg ‘beautify’

Class II intransitive ga=lamaz-d-eb-a pv=beautiful-v-ts-pres/fut.3sg ‘become beautiful’

b. da=a-inTeres-eb-s da=inTeres-d-eb-a pv=caus-interest-ts-pres/fut.3sg pv=interest-v-ts-pres/fut.3sg ‘interest someone’ ‘become interested’ c. da=a-mzad-eb-s pv=caus-ready-ts-pres/fut.3sg ‘prepare’

da=mzad-d-eb-a pv=ready.v-ts-pres/fut.3sg ‘be(come) prepared’

In some cases, the event structure of the vP is also reflected in overt suffixes, including the causative suffix -in (see §6.2.3), or the -d suffix that is found in certain inchoative Class II intransitives (such as those in (24)), but not in their causative transitive counterparts. Such suffixes are most straightforwardly analysed as realizations of v. Thus, I conclude that the thematic suffix is not itself a realization of v. I propose instead that it is a ‘dissociated’ mor-

148 martha mcginnis pheme, adjoined to v in a post-syntactic morphological component (see Embick 1997). The distribution of the thematic suffix is sensitive to clausal TAM features. It is found in the future (25a), present, imperfect, perfect, and other TAM categories that arguably express the event as having temporal extension. By contrast, it is absent from the aorist (25b), optative, and pluperfect, which arguably express a single point in (realis or irrealis) time. TAM categories that lack the thematic suffix (aorist and optative) are also distinguished by the presence of ergative case-marking on agents.⁴ According to Dixon (1994), ergativity tends to be associated with perfectivity and/or past tense. As a first approximation, it appears that ergative case in Georgian is associated with the conjunction of these: for example, ergative is found with the aorist (25b), but not with the (perfective) future (25a) or the imperfective past.⁵ (25) a. Nino vano-s surat-s Nino.nom Vano-dat picture-acc ‘Nino will draw Vano a picture.’ b. Nino-m vano-s surat-i Nino-erg Vano-dat picture-nom ‘Nino drew Vano a picture.’

da=u-xaT-av-s. pv=vce-draw-ts-pres/fut.3sg da=u-xaT-a. pv=vce-draw-aor.3sg

Note that the ‘perfectivity’ expressed by the aspectual preverb differs from the aspectual properties associated with the thematic suffix. An overt thematic suffix can either co-occur with the perfective preverb, as in the future tense, or not, as in the present tense, as shown in Table 6.1. Moreover, an aorist verb may or may not bear a perfective preverb (26), but in either case the thematic suffix is absent, and an overt agent DP (if any) will be ergative. The interpretations in (26) suggest that the preverb expresses completion (or perhaps inception, in future tense; see Mezhevich’s (2005) analysis of a similar distribution for Russian aspectual prefixes).

⁴ Note that despite its past perfective semantics, the pluperfect does not show ergative case-marking. However, it does lack an overt thematic suffix. According to Marantz’s (1991) disjunctive case hierarchy, ergative case could be absent because it is blocked by dative case on the subjects of agentive verbs in the pluperfect, as it is by inherent dative case on experiencer subjects in the aorist. McGinnis (2008) argues that, in the perfect (and pluperfect) series, the subject of an agentive verb is projected below Voice, as an applied argument bearing inherent dative case. ⁵ While the optative does not literally express past tense, it is noteworthy that other languages can use a past tense form to express optative meaning, as in the English subjunctive Wish you were here.

the structure of causatives in clausal nominalizations 149 (26) a. da=xaT-a. pv=paint-aor.3sg ‘(S)he painted it (and it was completely painted).’ b. xaT-a. paint-aor.3sg ‘(S)he painted it (but it wasn’t completely painted).’ (Léa Nash, p.c.) The aspectual properties influencing the distribution of the thematic suffix are also distinct from those of the English progressive. The suffix can appear with stative experiencer verbs, which are incompatible with the progressive in English: (27) a. m-γvidz-av-s. 1.dat-be_awake-ts-pres/fut.3sg ‘I am awake.’ b. m-e-γvidz-eb-a.⁶ 1.dat-vce-be_awake-ts-pres/fut.3sg ‘I will be awake.’

(Aronson 1990: 337)

I hypothesize that overt thematic suffixes appear in contexts with temporal extension, while a zero allomorph occurs in the aorist, which expresses an event as taking place at a past non-extended point in time. The thematic suffix can appear in the masdar, providing evidence for a vP projection within these nominalizations. In fact, the suffix (underlined in (28)) can appear in the context of both event and result readings for the masdar, suggesting that a vP projection is present in both CENs and RNs: (28) a. b. c. d.

da=Cq-eb-a gamo=tkv-m-a da=mtavr-eb-a cxovt-eb-a

‘act of beginning’ ‘act of pronouncing/a pronunciation’ ‘act of finishing/an end(ing)’ ‘life’

CEN CEN/RN CEN/RN RN

⁶ Note that in this class of verbs, future tense is indicated by a Voice prefix and the thematic suffix, not by an aspectual preverb.

150 martha mcginnis As noted above, the form of the thematic suffix is conditioned by the root. As the examples in (29) show, some thematic suffixes (-av, -am, -i) also undergo morphophonological changes in the masdar, while others (-eb, -ob) do not. (29) a. b. c. d. e.

Finite verb i-Cq-eb-s muša-ob-s xaT-av-s gamo=tkv-am-s targmn-i-s

Masdar Cq-eb-a muša-ob-a xaT-v-a gamo=tkv-m-a targmn-∅-a

Verb meaning ‘begin’ ‘work’ ‘paint’ ‘pronounce’ ‘translate’

In short, the presence of the thematic suffix constitutes evidence for a vP projection in the masdar. On the other hand, masdars lack prefixes signalling the presence or absence of an external argument—plausibly, exponents of the Voice head. For example, the masdar cannot include the prefix a- (30a), which in the verbal context (30b) is associated with the addition of an external argument to the unaccusative in (30c). (30) a. a=(∗ a-)šen-eb-a pv=(∗ vce-)be_built-ts-nom ‘buildingvn ’ b. a=a-šen-eb-s. pv=vce-be_built-ts-pres/fut.3sg ‘(S)he will build it/them.’ c. a=šen-d-eb-a. pv=be_built-v-ts-pres/fut.3sg ‘It will be built.’ Similarly, the masdar cannot include the prefix i- (31a), which in (31b) marks the (unaccusative) intransitive variant of the transitive clause in (31c). (31) a. da=(∗ i-)xaT-v-a pv=(∗ vce-)paint-ts-n ‘paintingn ’ b. da=i-xaT-eb-a. pv=vce-paint-ts-pres/fut.3sg ‘(S)he/it will be painted.’

the structure of causatives in clausal nominalizations 151 c. da=xaT-av-s. pv=paint-ts-pres/fut.3sg ‘(S)he will paint it/them.’ I conclude that the masdar includes vP, but not VoiceP. The structure of the Georgian masdar in (32a) can thus be represented as in (32b). (32) a. am siTqvebis gamotkvma dem words.gen pronunciation.nom ‘the pronunciation of these words’ b.

DP D am

nP vP

n -a

√P

DP siTqvebis

v v -∅

pv gamo=

ts -m

√ tkv

The preverb, if any (here, gamo=), merges with the category-neutral lexical root (here, tkv-), modifying its aspectual properties. A direct object merges as an argument of this root, and in the masdar context bears genitive case associated with n (or D). I propose that the thematic suffix (here, -m) realizes a head adjoined to v in the post-syntactic morphological component, while the masdar suffix (here, -a) realizes a higher nominalizing head n. Although Georgian clause structure is generally taken to be right-headed, I assume the DP is left-headed, since the demonstrative precedes the noun.⁷ The case of the masdar DP is of course determined by its position within the larger clause structure.

⁷ I leave a detailed investigation of the internal word order of masdars for further research.

152 martha mcginnis In the verbal context, additional projections are present, including Voice—potentially with an external argument in its specifier—and one or more functional heads associated with the TAM and number suffixes, which I represent simply as T. An example is given in (33). (33) a. Nino siTqveb-s gamo=tkv-am-s. Nino.nom words-acc pv=pronounce-ts-pres/fut.3sg ‘Nino will pronounce the words.’ b.

TP T′

DP Nino VoiceP

T -s Voice′

tDP vP

Voice -∅

√P

v √′

DP siTqvebs

pv gamo=

v -∅

ts -am

√ tkv

A number-marking suffix can reflect the number of the subject (or of a firstor second-person plural object). Given its morphological interactions with the TAM suffix, I analyse it as a fissioned Vocabulary item realizing a number agreement feature on the T head (Béjar 2003; McGinnis 2013; see Blix 2016 for a different analysis). Word structure and morpheme order are illustrated for finite verbs (34) and masdars (35). (34) a. da=i-Cq-eb-s. pv=vce-begin-ts-3sg.pres/fut ‘(S)he will begin it/them.’

the structure of causatives in clausal nominalizations 153 b.

TP PreV da=

T Voice Cl+Voice ∅-i-

T+# -∅-s v

√ Cq

v v -∅

(35) a. da=Cq-eb-a pv=bagin-ts-n ‘act of beginning’ b.

n

v

n -a

√ PreV da=

ts -eb

v √ Cq

v -∅

ts -eb

Although, in the verbal context, the aspectual preverb appears outside inflectional heads that are excluded from the masdar (34a), it appears to merge low in the structure, given that it can be included in the masdar (35a). Further evidence that the preverb merges low in the structure is that it is subject to an alternation conditioned by the number of the direct object, with da= indicating a plural object (Schmidt 1957; Harris 1981: 200). I propose that the preverb merges with the lexical root, which introduces the direct object. In the verbal context, then, the preverb undergoes movement, attaching to the left of the Voice head—as well as the person-marking clitic, if any. Such movement may be motivated by a licensing relation between Tense and the aspectual preverb. While different preverbs convey distinct idiosyncratic

154 martha mcginnis semantics, and the choice of preverbs depends in part on the lexical root, the presence or absence of the preverb is also related to TAM features, as noted above. This relationship could be expressed by means of a syntactic agreement relation in which (in some cases, optional) licensing selectional features on T in certain TAM categories probe for the preverb, triggering overt movement of the preverb to adjoin to the T head. The preverb sits outside a clitic that reflects person features of the subject or object, depending on the merged position of these arguments; I analyse this clitic as a fissioned Vocabulary item, realizing a person agreement feature on the Voice head (see Béjar 2003; McGinnis 2008; but see also Foley 2016). In the nominal context, T and Voice are absent, and the preverb appears immediately to the left of the root. I assume it does not undergo movement in this case, but simply combines with the root that selects it, in the formation of a complex head structure. To summarize, Georgian morphosyntax provides evidence for a vP layer in the masdar, along with a category-neutral root, an aspectual preverb, and a thematic suffix adjoined to v—all contained within the nominalizing nP and DP projections. On the other hand, heads associated with Voice and TAM morphology are absent from the masdar. In the next section we will see further evidence for this analysis, based on the argument structure of the masdar.

6.2.2 Masdars are v-selecting: Evidence from argument structure Another source of evidence for the proposed analysis of the Georgian masdar comes from its argument structure—in particular, the syntactic expression of the participants corresponding to the subject, direct object, and indirect object of a verbal clause. Only the direct object corresponds to a DP argument in the masdar; the subject and indirect object correspond instead to PPs. I propose that this is because agentive DP subjects are introduced by Voice, which is absent in the masdar, while applied DP arguments are Case-licensed by Voice. In a simple transitive clause, a direct object is Case-licensed by Voice as well; however, in the masdar it can be licensed by Case from n instead, yielding genitive case-marking and a fixed prenominal position. Let us begin by considering the direct object in more detail. In English, a gerund can take an object with accusative case (36a), while a derived nominalization cannot (36b). Kratzer (1996) argues that accusative case is associated with active Voice, which introduces an external argument. Since derived nominalizations in English lack Voice, they permit neither structural Case for the internal argument, nor the addition of an external argument to a non-agentive root (36c) (Chomsky 1970; Marantz 1997).

the structure of causatives in clausal nominalizations 155 (36) a. b. c.

[PRO growing tomatoes] is easiest in a warm climate. the growth ∗ (of) tomatoes ∗ John’s growth of tomatoes

Georgian masdars resemble English derived nominalizations in this respect. The case array in Georgian varies with person, verb conjugation class, and clausal TAM series. In Series I TAM categories with a nominative subject, such as present tense, a direct object is marked with accusative case (37a), which in Georgian is morphologically identical to dative case—but which, unlike dative case, alternates with nominative case in Series II TAM categories with an ergative subject, such as the aorist (37b). A direct object is also nominative in clauses with a dative subject, including those with a subject-experiencer verb or a Series III TAM category such as the perfect or pluperfect. (37) a. kartul Cign-s Kitxul-ob-s. Georgian.acc⁸ book-acc read-ts-3sg.pres/fut ‘(S)he is reading a Georgian book.’ b. kartul-i Cign-i Ca=i-Kitx-a. Georgian-nom book-nom pv=vce-read-3sg.aor ‘(S)he read a Georgian book.’ (Aronson 1990: 51, 58) By contrast, in a masdar, the direct object must be a genitive prenominal argument (38), as expected if it is Case-licensed in the same way as a typical nominal possessor. (38)

roca rusuli liTeraTur-is Klas-is Kreba when Russian literature.gen class.gen meeting.nom da-i-Cq-eb-a, [ “boris godunov-is” Kitx-v-a ] pv=vce-begin-3sg.pres/fut Boris Godunov-gen read-ts-n ga=grjel-d-eb-a. pv=continue-v-ts-3sg.pres/fut ‘When the Russian literature class meeting begins, the reading of Boris Godunov will continue.’ (Aronson 1990: 78–9, 85)

⁸ Nominative or genitive concord on Georgian adjectives is indicated by the nominative suffix (-i), and ergative concord by the ergative suffix (-ma). Accusative/dative concord is not indicated by any overt suffix.

156 martha mcginnis Further evidence for the absence of Voice is that masdars lack an external argument DP. In Georgian masdars, the prenominal genitive argument must be thematically licensed as an internal argument—I assume by merging directly with the lexical root. External arguments of transitive clauses are usually omitted. Along with the absence of structural object Case, the absence of such external arguments from the masdar provides further evidence that Voice is absent.⁹ There is some evidence that genitive external arguments are ruled out in the masdar, not only because Voice is absent, but also because, in Georgian, genitive is a realization of Case licensed by n, parallel to the Case licensing indicated by of in English nominal complements. By contrast, the genitive case on English prenominal DPs reflects Case licensed by D. This difference leads to differences in the distribution of genitive arguments in English and Georgian. In English, as noted above, non-agentive roots like grow can take an external argument in a verbal context (39a), but not in a derived nominalization (39b). This suggests that derived nominalizations in English lack Voice. However, a root implying external causation, like the root of destroy, does allow for an agentive possessor in a derived nominalization, as in (39c). Marantz (1997) proposes that in this case, the agentive interpretation is licensed by the semantics of the root, not by a functional head like Voice. (39) a. John grows tomatoes. b. {the / ∗ John’s} growth of tomatoes c. the army’s destruction of the city In a Georgian masdar, however, the possibility in (39c) is unavailable.1⁰ I propose that this contrast arises because an agentive possessor in English can be licensed by D, while a Georgian nominals only permit Case licensing by n. Thus, in Georgian, genitive case cannot appear on an external argument that merges outside n—nor can it be assigned in the verbal context, which lacks n. This analysis correctly predicts that genitive external arguments are ruled out even when they would be semantically interpretable in the absence of Voice.

⁹ Unergatives do not behave as expected: the genitive argument of a masdar can correspond to the subject of an unergative predicate, as in Givi-s mo=t’qu-eb-a ‘Givi’s lie’ in (7a). I leave this matter for further research. 1⁰ Hewitt (1991: 15–16) presents a literary counterexample to this generalization, and notes that Mediaeval Georgian allowed genitive subjects in transitive masdars. I analyse these as instances of Case licensing by D in the masdar, which is no longer possible in modern Georgian (except in a literary register).

the structure of causatives in clausal nominalizations 157 By contrast, English D can license Case, and thus permits an agentive DP in the context of an externally caused change-of-state root, as in (39c). In an even closer comparison to the Georgian masdar, English nominalizations also allow an agentive possessor in the context of a causative vP projection. Even in the absence of accusative case (40), English ing-of gerunds permit a prenominal possessor. This possessor allows a wide range of meanings, depending on the root (Marantz 1997; Moulton 2014) can also be interpreted as a non-agentive possessor—for example, someone who owned a recording of a reading of Pride and Prejudice, or someone who liked one. Kratzer (1996) argues that the possibility of non-agentive readings for prenominal possessors in ing-of gerunds provides evidence that the Voice projection is absent from these gerunds. (40) a. Maria’s reading of Pride and Prejudice was less criticized than Anna’s. b. Maria’s eating of the apple pie was less criticized than Anna’s. There is evidence that a vP layer is present in ing-of gerunds, at least in some cases. For example, the derived nominalization in (41a) disallows an agentive possessor, while the corresponding ing-of gerund allows one (41b). Nevertheless, accusative case is absent in both contexts, indicating the absence of Voice. I propose that a vP layer is absent from (41a) but present in (41b), between the root and the nominal head nP. The root-conditioned nominalizing suffix -th is restricted to the local domain of the root, so it cannot appear outside vP. Thus, the n head in (41b) can only be realized by the default nominalizing suffix -ing (see Embick and Marantz 2008). However, despite the absence of Voice, the v head can add a causative interpretation that semantically licenses the prenominal genitive agent in (41b). By hypothesis, this DP is Case-licensed by D. (41) a. {the / ∗ John’s} growth of tomatoes b. John’s growing of tomatoes is becoming obsessive. Although Georgian masdars also contain a vP projection, they disallow a genitive agent, as predicted if genitive arguments in Georgian are licensed by n, not D. On the other hand, a masdar based on a transitive clause can express an external argument as a PP headed by the postposition mier ‘by’ (42a). Mierphrases are also used in the passive, as shown in (42b).

158 martha mcginnis (42) a. [

monadiris mier datvis moKvl-a ] hunter.gen by bear.gen killing-n.nom aKrzaluli-a. be_forbidden-pres/fut.3sg ‘[The killing of bears by hunters] is forbidden.’ (Harris 1981: 157) b. bavšvi da=Kben-ili=a zaγlis mier. child.nom pv=be_bitten-ptcp=be.pres/fut.3sg dog.gen by ‘The child is bitten by a dog.’ (Harris 1981: 103)

To be more specific, a mier-phrase can occur with what Harris (1981) refers to as the analytic passive. In this type of passive, the main verb is a participle, bearing an enclitic form of the copula verb qopna. In traditional Georgian grammars, the term ‘passive’ is used for what Harris calls the synthetic passive. This is an intransitive (unaccusative) clause with an active verb form, which cannot combine with a mier-phrase. The examples below show the active (43a), the analytic passive (43b), and the synthetic ‘passive’ (43c). (43) a. deda xaČapur-s a-cx-ob-s. mother.nom cheese_bread-acc vce=bake-ts-pres/fut.3sg ‘Mother is baking cheese-bread.’ (Harris 1981: 193) b. xaČapur-i gamomcxvaria ded-is mier. cheese_bread-nom pv.bake.ptcp.be.3sg mother-gen by ‘The cheese-bread {is baked / has been baked} by Mother.’ (Harris 1981: 193) ∗ c. xa-Čapur-i cxv-eb-a ( ded-is mier). cheese_bread-nom bake-ts-pres/fut.3sg mother-gen by ‘The cheese-bread is baking (∗ by mother).’ (Harris 1981: 197) I propose that the mier-phrase in (43b) adjoins to the VoiceP of the analytic passive—which, like the English passive, implies an external argument. In a synthetic passive, there is no implied external argument, and thus no question of adjoining a mier-phrase to VoiceP. Some sources (e.g. Hewitt 1970: 262; Wier 2014) describe masdars with an external argument mierphrase as marked or impossible. Thus, I postulate that VoiceP is absent from the masdar, and that the mier-phrase instead adjoins exceptionally to vP, a dispreferred position. It is worth noting that, unlike English by-phrases, but like Spanish por-phrases (Grimshaw 1990, citing Esther Torrego, personal communication), Georgian mier-phrases can occur in the passive, but cannot

the structure of causatives in clausal nominalizations 159 modify a nominal that lacks a vP layer (44a)—though they can modify an uninflected passive participle (44b) (Léa Nash, personal communication). I conclude that the mier-phrase does not adjoin to nP. (44) a.

b.



Chomsky-s mier Cigni Chomsky-gen by book.nom ‘a book by Chomsky’ Chomsky-s mier da=Cer-ili Cigni Chomsky-gen by pv=write-ptcp book.nom ‘a book written by Chomsky’

Further evidence for the absence of Voice from the masdar comes from the interpretation of silent (PRO and implicit) external arguments. English PROing and ing-of gerunds contrast with respect to control by a higher subject. In the context of a higher subject-control verb, the (PRO) subject of a PRO-ing gerund is obligatorily controlled by the higher subject. Thus (45a) expresses a contradiction: the first clause asserts that Nancy sent a letter, while the second clause instead presupposes that the speaker sent the letter. By contrast, an ingof gerund in the same syntactic position does not require control of its implicit subject; thus, (45b) allows a non-contradictory reading in which Nancy sent the letter. (45) a. # Nancy sent a letter, and I regret [PRO sending this letter]. b. Nancy sent a letter, and I regret [the sending of this letter]. Like ing-of gerunds, complement masdars in Georgian need not involve control (Vamling 1989: 112): (46)

nino-m gaagzavn-a Ceril-i, da v-nanob [am Ceril-is Nino-erg send-aor letter-nom and 1-regret this letter-gen gagzavna]-s. sending-dat ‘Nino sent a letter, and I regret the sending of this letter.’

These contrasts can be captured if VoiceP and its PRO specifier are present in PRO-ing gerunds, but absent from ing-of gerunds and Georgian masdars.

160 martha mcginnis The claim that masdars contain vP, but not VoiceP, makes rather complex predictions regarding applied arguments. According to Hewitt (1991), masdars based on applicatives are possible (examples based on Cer- ‘write’): (47) a. Cer-a b. da=Cer-a c. mi/mo=Cer-a

imperfective (applicative or not) perfective (non-applicative) perfective (applicative)

However, further investigation suggests that, while masdars permit relational expressions, they disallow true applicatives. In the verbal context, an indirect object is marked as dative, but dative case is banned in masdars (Nash n.d.); instead, a postpositional phrase must be used: (48) a. vano-m lali-s Cign-i cxr-a. Vano-erg Lali-dat book-nom give-aor.3sg ‘Vano gave Lali the book.’ b. Cign-is mi=cem-a { lalis-tvis / lali-ze / book-gen pv=give-n Lali-for Lali-on ‘giving of the book to Lali’



lali-s } Lali-dat

It has been argued that applicatives fall into two general classes, known as low and high applicatives (Pylkkänen 2008; Harley 2013). A high applicative relates an applied DP specifier to a vP complement, typically as a beneficiary or maleficiary of the event associated with vP. Such applicatives place no transitivity restrictions on the complement vP. By contrast, a low applicative merges below vP, typically expressing possession, or a transfer of possession, and directly relating its indirect-object specifier (a source, goal, or possessor) to its direct-object complement (a theme). Such applicatives typically require the expression of a vP-internal argument, since they combine with this argument. A causative that selects vP, rather than VoiceP, is predicted to rule out an embedded high applicative, since these merge outside vP (Pylkkänen 2008; Harley 2013). This is illustrated for ChiBemba in (49). ChiBemba has a high benefactive applicative, which can combine with an unergative predicate (49a). However, this applicative cannot be embedded within the complement of the v-selecting causative (49b): (49) a. Mwape aa-boomb-ela Mutumba. Mwape past-work-ben Mutumba ‘Mwape worked for Mutumba.’

the structure of causatives in clausal nominalizations 161 b.



Naa-tem-en-eshya Mwape Mutumba iciimuti. 1sg.past-cut-ben-caus Mwape Mutumba stick ‘I made Mwape [cut Mutumba a stick].’ (Givón 1976)

Likewise, if the Georgian masdar contains vP but not VoiceP, it too is predicted not to embed a high applicative phrase. However, it is unclear whether high applicatives exist in Georgian. Applied arguments in Georgian reportedly can combine with some unergative verbs (50), a characteristic of high applicatives. On the other hand, the dative argument in such examples is interpreted as a goal rather than a beneficiary, suggesting that it may not be a true high applicative. This structure could be analysed as a low applicative, on the view that unergative verbs in Georgian are underlyingly transitive (Hale and Keyser 1993). Marantz (1991) proposes a similar analysis to account for the observation that both transitive and unergative verbs in Georgian allow ergative subjects. (50) a. deda-m kaliCvil-s u-mGer-a. mother-erg daughter-dat appl-sing-aor.3sg ‘Mother sang to (her) daughter.’ b. mezobl-is jaGli xSirad m-i-qep-s. neighbour-gen dog.nom often 1.dat-vce-bark-pres/fut.3sg ‘The neighbour’s dog often barks at me.’ (Aronson 1990: 207–8) Georgian also has more typical instances of low applicatives, such as goal (51a) and possessor applicatives (51b) in transitive clauses. (51) a. izane-m ekim-s Ceril-i ga=u-gzavn-a. John-erg doctor-dat letter.nom pv=vce-send.ts-aor.3sg ‘John sent the doctor a letter.’ (Aronson 1990: 169) b. Svil-s tma da=u-varcxn-a. child-dat hair.nom pv=vce-comb-aor.3sg ‘(S)he combed the child’s hair.’ (Aronson 1990: 173) In phrase-structural terms alone, a v-selecting nominalization would be expected to allow low applicatives, since these merge below vP; yet these too are banned from the masdar. A similar restriction also exists in English ingof gerunds (Kayne 1984; Abney 1987; Sichel 2010). Like derived nominalizations, which arguably lack a vP projection (52a), ing-of gerunds disallow the

162 martha mcginnis double object construction, a low applicative (52b). By contrast, gerunds with accusative objects, such as poss-ing gerunds, allow an applicative (52c). (52) a. b. c.

∗ ∗

his/the gift (of) Mary of a Fiat his/the giving (of) Mary of a Fiat his giving Mary a Fiat

As these authors note, ECM (53) and control clauses (54) are also incompatible with both derived nominalizations and ing-of gerunds, though acceptable with poss-ing gerunds. (53) a. b. c.



(54) a. b. c.







his belief of/in [Bill to be Caesar] his believing of [Bill to be Caeasr] his believing [Bill to be Caesar] his persuasion of Mary [PRO to stay] his persuading of Mary [PRO to stay] his persuading Mary [PRO to stay]

On the other hand, as they also point out, causatives (55) and unseparated verb-particle constructions (56) are possible in ing-of gerunds, though not in derived nominalizations. (55) a. b. c.



his growth of tomatoes his growing of tomatoes his growing tomatoes

(56) a. b. c.



his explanation (away) of the problem (away) his explaining (away) of the problem (∗ away) his explaining (away) the problem (away)

The contrasts in (55) can be taken to support the view that ing-of gerunds contain a vP projection that is absent in derived nominals. If a causative head introduces a vP projection, it will be unable to occur in a derived nominalization lacking vP, as in (55a), but will be able to occur in an ing-of gerund containing vP, as in (55b). We can speculate that the particle in (56) enters an obligatory relation with v, perhaps because it expresses aspectual information. If so, then it too will be ruled out in a derived nominalization. The ill-formedness of the ing-of gerunds in (53)–(54) suggests that the absence of Voice leads not only to the absence of accusative case for the

the structure of causatives in clausal nominalizations 163 direct object, but also to a total inability to license non-core DPs, such as applied objects, or subjects of embedded non-finite clauses (see also Chomsky 1986). Given the analysis above that the genitive DP in a Georgian nominal is Case-licensed by n, it follows that such Case-licensing is available only for a thematic argument of the root. At present I have no account of this restriction. Thematically related (inherent) Case is usually analysed as being assigned via Merge. Thus, an alternative analysis of the genitive object in a masdar would be to treat it as merged with n instead of the lexical root, receiving inherent Case via Merge. This would account for the restriction to thematic arguments of the nominal. However, Georgian root suppletion provides evidence against this analysis. Certain roots in Georgian show suppletion based on the number or animacy of an internal argument. For example, a singular object appears with one allomorph of the verb root for ‘throw’, as in (57a), while a plural object appears with a different allomorph, as in (57b) (Harris 1981: 19). (57) a. bavšv-ma gada=agd-o child-erg pv=throw.sg-aor.3sg ‘The child threw the stone.’ b. bavšv-ma gada=qar-a child-erg pv=throw.pl-aor.3sg ‘The child threw the stones.’

kva. stone.nom kv-eb-i. stone-pl-nom

The same phenomenon is seen in the masdar context, as illustrated by the singular allomorph of the root in (58a) and its plural counterpart in (58b) (Harris 1981: 160). (58) a. kvi-s gada=gd-eb-a aKrzaluli-a. stone-gen pv=throw.sg-ts-n forbidden.ptcp-be.3sg ‘Throwing a stone is forbidden.’ b. kv-eb-is gada=qr-a aKrzaluli-a. stone-pl-gen pv=throw.pl-n forbidden.ptcp-be.3sg ‘Throwing stones is forbidden.’ Locality restrictions on contextual allomorphy (Embick and Marantz 2008; Embick 2010) generally prevent a category-forming head (such as v or n) from intervening between a suppletive node and its conditioning context—in this case, the object DP. I therefore conclude that the object merges directly with the root, not with nP. Thus I maintain that the Case licensed by n in Georgian is

164 martha mcginnis structural, not inherent. Note that it would be challenging to argue that the root of the nominal is responsible for licensing inherent Case on its complement, given that this complement is marked with object case in the presence of Voice, not with genitive case. It is unclear why a root would license inherent Case on its complement only in the absence of Voice. Thus, I maintain that genitive case is a realization of structural Case licensed by n in Georgian, with the stipulation that this Case is (somehow) restricted to the internal argument of the root. By contrast, I postulate that Case on applied arguments in Georgian is licensed structurally by Voice. When Voice is absent, as in a masdar, no applied argument can occur, as illustrated in (59a). By contrast, in Czech, a nominalization with a genitive object can also include a dative applied object (59b). Dvořák (2011) proposes that an applied dative in Czech has inherent Case; thus, it is compatible with nominalizations lacking Voice. (59) a. Cign-is mi-cem-a { lalis-tvis / lali-ze / ∗ lali-s } book-gen giving Lali-for Lali-on Lali-dat ‘giving of the book to Lali’ b. Darování knih-y Mari-i (se Karl-ovi giving.nom book-gen Mary-dat refl Charles-dat ne-vyplatilo). not-pay_off.past ‘Giving a book to Mary (didn’t pay off for Charles).’ The structural Case analysis predicts that an applied dative in Georgian can sometimes become the subject of a clause lacking an external argument. This prediction is borne out to some degree. For example, there is evidence that a type of malefactive argument can serve as a clausal subject (Aronson 1990: 345; McGinnis 1998: 85–6). While these malefactives have invariant dative case marking, they can be regarded as bearing quirky structural Case, like dative subjects in Icelandic. In Georgian, dative subjects trigger third-person plural subject agreement, indicated by a plural suffix (-t) on the verb. In (60), a malefactive argument is added to an unaccusative structure that is usually marked by the prefix i-, a realization of the Voice head.11 When unaccusative Voice is accompanied by a dative argument, it is spelled out instead as e-, supporting the view that an Agree relation exists between Voice and the dative DP. 11 Nash (1995) analyses i- as a realization of v, which she regards as responsible for the introduction (or non-introduction) of an external argument, parallel to Voice in the current analysis.

the structure of causatives in clausal nominalizations 165 (60) a. deid-eb-s nino da-e-Karg-a-t. aunt-pl-dat Nino.nom pv-vce-lose-aor-pl ‘The aunts had Nino lost on them.’ b. ded-eb-s švil-eb-i da-e-Čr-a-t. mother-pl-dat son-pl-nom pv-vce-cut-aor-pl ‘The mothers had the sons wounded on them.’ A first- or second-person pronoun can trigger plural agreement whether it is a subject or an object, but a third-person argument triggers plural agreement only if it is a subject. Certain clauses in Georgian take a dative subject (61a), including clauses in the perfect or pluperfect with a Class 1 (transitive) or Class 3 (unergative) verb, as well as clauses with a Class 4 verb, usually involving an experiencer subject. Word order in Georgian is flexible, but the plural suffix cannot be triggered by a dative third-person object, regardless of word order (61b). (61) a. deideb-s gela u-qvar-t. aunts-dat Gela.nom vce-love.pres/fut-pl ‘The aunts love Gela.’ b. deideb-s gela e-čxub-eb-a-(∗ t). aunts-dat Gela.nom vce-fight-ts-pres/fut-(∗ pl) ‘Gela will fight with the aunts.’ Evidence that the dative argument in (61a) is the subject comes from binding asymmetries (Harris 1981; Marantz 1989; McGinnis 1998). A dative experiencer subject can bind a nominative object anaphor (62a), while the reverse binding relation is ruled out (62b) Harris (1981: 143). On the other hand, when the subject is nominative, as in (63a), it can bind a dative object anaphor, and the opposite binding relation is impossible (63b) (Marantz 1989). The judgements in (62) and (63) hold regardless of word order. (62) a.

temur-s u-qvar-s tav-isi tav-i. Temur-dat vce-love-pres/fut.3sg self-gen.NOM self-nom ‘Temur loves himself.’ b. ∗ tav-is tav-s u-qvar-s temur-i. self-gen self-dat vce-love-pres/fut.3sg Temur-nom (‘Temur loves himself.’)

166 martha mcginnis (63) a.

vano tav-is tav-s e-laParaK-eb-a. Vano.nom self-gen self-dat vce-talk-ts-pres/fut.3sg ‘Vano is talking to himself.’ ∗ b. vano-s tav-isi tav-i e-laParaK-eb-a. Vano-dat self-gen.nom self-nom vce-talk-ts-pres/fut.3sg (‘Vano is talking to himself.’)

Further evidence that dative arguments can be subjects comes from suppletion of the verb root. As noted above, suppletion in Georgian verb roots can be conditioned by the animacy of the internal argument (Harris 1981: 19–20). In the Class 4 verb in (64), the nominative argument triggers the inanimate suppletive form of the root (miakvs), rather than the animate form (miqavs), suggesting that this nominative argument is the direct object, not the subject. (64)

gela-s Cign-eb-i miakv-s samKitxvelo-ši. Gela-dat book-pl-nom take-pres/fut.3sg reading_room-in ‘Gela is taking the books into the reading room.’

Moreover, nominative objects in Georgian cannot trigger number agreement, which distinguishes them from nominative subjects. Consider the contrast between (65a) and (65b). In (65a), a lone nominative argument raises to the subject position, triggering plural subject agreement on the verb. Agreement with ergative and nominative subjects is realized, not as the ‘default’ plural -t, but as a TAM suffix agreeing with the person and number features of the subject (here, 3pl -nen). In (65b), however, a dative malefactive argument raises to subject, controlling number agreement on the verb—here, plural -t— and blocking the nominative object from doing so. (65) a. bavšv-eb-i da=Karg-nen. children-pl-nom pv=vce-be_lost-pl-aor.3pl ‘The children are lost.’ b. deid-eb-s bavšv-eb-i da=e-Karg-a-t. aunts-pl-dat children-pl-nom pv=vce-lost-aor-pl ‘The aunts had the children lost on them.’ (Léa Nash, p.c.) The dative malefactive also behaves like an ordinary subject with respect to focus and word order. In general, the immediately preverbal position in Georgian contains existentially asserted material (Nash 1995); for example,

the structure of causatives in clausal nominalizations 167 the object is interpreted as unmarked new information in the SOV order. However, a subject in the immediately preverbal position behaves differently: it need not be existentially focused. In OSV order, for example, the subject has contrastive rather than existential focus. In dat-nom-V examples with a malefactive argument, like (65b), the nominative argument is interpreted as unmarked new information, like an object, rather than contrastively focused, like a subject (Léa Nash, personal communication). Nevertheless, aside from malefactives and experiencers, dative applied arguments generally do not become subjects. For example, in the periphrastic passive of a ditransitive clause, only the direct object can become the subject (66). On the other hand, it is worth noting that a dative applied object is also completely ruled out in this context. For some speakers, the participant corresponding to the dative can be expressed instead as a PP headed by the postposition -tvis (Harris 1981: 111–12): (66)

surat-i iqo nacvenebi {%vano-s-tvis / ∗ vano-s}. picture-nom be.aor.3sg show.ptcp Vano-gen-for Vano-dat ‘The picture was shown to Vano.’

What accounts for the difference between this passive, which disallows a dative subject, and the malefactive unaccusatives in (60), which allow one? I postulate that quirky Case is licensed on a DP both by the theta-assigning head and by a structural Case-checking head (McGinnis 1998); moreover, the theta-assigning head can specify the type of head that can check the structural Case of a quirky Case-marked DP. In Georgian, applicative heads introducing malefactives or experiencers allow the quirky DP they introduce to check structural Case on T, while other applicative heads require it to check structural Case on Voice. Supposing that, as is usually the case, the structural Case feature of Voice is suppressed in the Georgian periphrastic passive, then applied arguments that require Case-checking by Voice cannot be Case-licensed, and only a postpositional counterpart can be used instead. The proposal that the masdar lacks VoiceP correctly predicts a variety of its characteristics. For example, it predicts that a transitive external argument cannot be realized as a DP in a masdar, but only as an optional, marginal mier-PP in a dispreferred vP-adjoined position. It also predicts the unavailability of a PRO external argument, accounting for the fact that implicit external arguments are not subject to obligatory control in the masdar. The absence of the Case-licensing associated with transitive Voice accounts for the

168 martha mcginnis ungrammaticality of an applied DP argument, whether high or low, and for the absence of object case on the internal argument of the masdar, which is licensed instead by n. In the following section, I argue that causative masdars provide additional evidence for vP without VoiceP in Georgian masdars.

6.2.3 Masdars are v-selecting: Evidence from causative masdars The Georgian masdar can contain a causative predicate. If causative meaning is expressed by a v head (Marantz 1997; Pylkkänen 2008), then causative masdars must contain vP. In a causative masdar, the causer can only be introduced as an optional mier-phrase, parallel to the external argument of a masdar based on a simple transitive vP. In a causativized transitive clause, the causee (the subject of the embedded transitive) is marked as dative; in a masdar based on such a clause, no causee can be introduced at all. These observations can be captured on the view that Voice not only introduces a (DP) causer, but also licenses the dative causee. If Voice is absent, as in a masdar, then neither is possible. In the verbal context, Georgian causatives are indicated by a prefix a-, with or without a suffix -in(-eb). A causative can be based on a transitive clause, suggesting that the causative head in Georgian can embed at least a vP projection—unlike a root-selecting causative, such as that found in English transitive grow, bake, and so forth (Pylkkänen 2008). According to Lomashvili (2011: 59), Georgian causatives are bieventive, allowing adverbial modification of either a higher causative vP or a lower embedded vP. The word order in (67a) reportedly facilitates the high reading for the adverb isev ‘again’, in which the causing event is repeated; the order in (67b) facilitates the low reading, in which the caused event is repeated. (67) a. isev dato sTudenTeb-s a-tviniereb-in-eb-s again Dato.nom students-dat vce-tame-caus-ts-pres/fut.3sg maimunebs. monkeys ‘Dato again makes students tame monkeys.’ b. dato sTudenTeb-s a-tviniereb-in-eb-s Dato.nom students-dat vce-tame-caus-ts-pres/fut.3sg maimunebs isev. monkeys again ‘Dato makes students tame monkeys again.’

the structure of causatives in clausal nominalizations 169 On the other hand, only the higher, causing event can be modified by an agent-oriented modifier, such as siamovnebit ‘with pleasure’ or ganzrax ‘intentionally’ (68a) (Nash 2017). According to Pylkkänen (2008), the illformedness of agent-oriented modification for a caused event, as in (68b), provides evidence that the causative involved does not embed a full VoiceP. Thus, we can conclude that the causative v selects a vP complement but not a VoiceP. On the other hand, the causative vP itself is dominated by VoiceP, permitting agent-oriented modification of the higher event. (68)

keti-m gogo-s leks-i ganzrax Keti-erg girl-dat poem-nom intentionally gada=a-targmn-in-a. pv=vce-translate-caus-aor.3sg a. ‘[Keti made the girl translate the poem] intentionally.’ b. ∗ ‘Keti made [the girl translate the poem intentionally].’

For some speakers, a masdar can include a causative bearing the -in(-eb) suffix (Léa Nash, p.c.); Hewitt (1991) refers to the causative masdar (da-/mi-/mo-)Cer-in-eb-a, based on Cer ‘write.’ The causative prefix a- disappears in the masdar, but its causative interpretation remains available, and allows adverbial modification of either the causing or the caused event; thus, like the verbal causative in (69a), the masdar causative in (69b) allows both high and low readings for the manner adverb male ‘quickly’ (Léa Nash, p.c.)—that is, indicating either that the causing event (the building) occurred in a quick manner, or that the resulting state (being built) was soon achieved. es saxli male a=a-šen-a. (69) a. vano-m Vano-erg this house.nom quickly pv=vce-be_built-aor.3sg ‘Vano quickly built this house.’ b. saxl-is male a=šen-eb-a house.gen quickly pv=be_built-ts-n ‘the building of the house quickly’ The occurrence of the adverb in (69b), particularly on the high reading, provides further evidence that the nominalizing head of a masdar selects a phrasal, potentially causative, vP category—not a projection of a lexical root (√P), and not a lexical unit (V0 ), either of which would disallow manner modification.

170 martha mcginnis Causative masdars also reveal a potential structural difference between result nominalizations (RNs) and complex event nominalizations (CENs). As discussed in §6.1, both nominalizing heads can combine with a root, a vP, or even a vP containing an aP. However, it appears that the RN head cannot combine with a vP containing a vP. Georgian masdars based on -in causatives, such as those in (70), permit only an event reading, not a result reading (Léa Nash, p.c.). (70) a. gamotkmevineba b. ashenebineba c. dasdzevineba d. gameorebineba e. mosmenineba f. daCerineba g. gagzavnineba h. daxaTvineba i. gagebineba j. gaatbobineba k. moKvlevineba l. dadgmevineba m. moČrevineba

‘making someone pronounce X’ ‘making someone build X’ ‘making someone punish X’ ‘making someone repeat X’ ‘making someone listen to X’ ‘making someone write X’ ‘making someone send X’ ‘making someone paint X’ ‘making someone understand X’ ‘making someone warm X up’ ‘making someone kill X’ ‘making someone stand X up’ ‘making someone cut X off ’

If this restriction holds generally, as appears to be the case, this means that selection of the result reading can take place only if n is in a local domain of the root, while a non-local n head yields only an event reading for the nominalization. This local domain is apparently not the ‘inner domain’ defined by the category head immediately c-commanding the root (Embick and Marantz 2008; Embick 2010), given that RNs can contain other categories, such as an aP. Instead, it appears to be the larger ‘domain of special meaning’ proposed by Marantz (1997), defined by causative/agentive v. (71) boundary of domain for special meaning agent v An idiomatic reading is possible for a causative of a non-agentive predicate, like those in (72), but not for a causative of an agentive verb, as in make

the structure of causatives in clausal nominalizations 171 X swim/fly a kite. This parallels the observation that Georgian RNs cannot include causatives of agentive verbs. (72) a. make oneself scarce b. make ends meet c. Marie a laissé tomber Luc. Marie has let fall Luc ‘Marie dropped Luc like a hot potato.’ d. On lui fera passer le goût du pain. they him.dat make.fut pass the taste of.the bread ‘They’ll kill him.’ (lit. ‘They will make the taste of bread leave him.’) Thus, it appears that the result nominalizer is restricted to the domain of special meaning for idiomatic expressions (McGinnis 2016). This analysis predicts that CENs, but not RNs, can contain an agent, since agents are merged outside the domain of special meaning. Idioms exclude agents; as Marantz (1984) points out, idioms that do include a subject are non-agentive. For example, (73a) is a stative passive; active examples also have non-agentive subjects, such as a theme (73b–73c), a possessor (73d), or an inanimate causer (73e). (73) a. b. c. d. e.

X’s goose is cooked the roof caved in on X the shit hit the fan the cat’s got X’s tongue What’s eating X?

As predicted, although CENs can lack a syntactic agent (74a) or can contain a PRO agent (74b), RNs cannot contain an agent (75). (74) a. The recording of Bach’s cello concertos took longer than we expected. b. PRO Recording Bach’s cello concertos took longer than we expected. (75) a.

The recording of Bach’s cello concertos is on display in the music library. ∗ b. PRO Recording Bach’s cello concertos is on display in the music library.

In short, while both CENs and RNs can attach to either roots or vPs, evidence from both English and Georgian suggests that RNs are restricted to

172 martha mcginnis the domain of special meaning. This curious restriction raises the possibility that the result n is not a true alloseme of the nominalizing head, but rather a semantically empty categorizing head (Marantz 2013), licensed only within the domain of special meaning. To recap, we have seen a broad array of evidence that Georgian masdars contain vP, but not VoiceP. The presence of thematic and valency-changing suffixes suggests that vP is present in the masdar. Further evidence for the presence of vP arises comes from causative masdars. For some speakers, masdars can include the causative suffix -in, and masdars based on causative verbs lacking this suffix also allow causative readings; adverbial modification is possible for both the causing and resulting events. On the other hand, the absence of Voice is supported by the omission of Voice prefixes from the masdar, as well as by the argument structure characteristics of the masdar. These include the expression of logical objects as genitive DPs; the omission of transitive external arguments, including PRO, or else their expression as mier-PPs; and the ungrammaticality of applied DP arguments. In the following section, I argue that causative masdars also provide evidence for the unexpected conclusion that transitive causees can be merged in the specifier of a high applicative merged outside causative vP.

6.3 Causees merged outside causative vP The argument structure of causative masdars provides evidence that a transitive causee can merge outside the causative head. Analyses of causatives typically locate the causee within the scope of the causative head (Harley 1995; Richards 2001; Kim 2011; Bjorkman and Cowper 2013; Legate 2014, inter alia). For example, in Harley’s analysis of Mary gave Susan the boot (76), the causer Mary merges outside the causative v head, but this head c-commands the causee Susan. (76)

vP

‘Mary gave Susan the boot.’ v′

Mary vcaus

PP P′

Susan Phave

the boot

the structure of causatives in clausal nominalizations 173 On the standard approach just described, the causee argument merges below the causative v. Some have even proposed that a causer argument can merge below causative v. For example, Lomashvili (2011: 58) proposes that in a ‘double’ causative—that is, a causative of a transitivized unaccusative predicate—the lower causer, or transitive causee (Y in (77a)), merges inside both causative heads (77b), even though it is not a semantic argument of the unaccusative predicate, ‘Z be beautiful’: (77) a. ga=a-lamaz-eb-in-a. pv=vce-be.beautiful-ts-caus-aor.3sg ‘X caused Y to make Z beautiful.’ b. VoiceP Voice′

X (causer)

vP vP

Voice vcaus

vP

vcaus v′

Y (causee) VP Z (theme)

v V



V

By contrast, Pylkkänen (2008) proposes that in Japanese root-causativized unergatives such as (78a), the causee merges outside the causative head, as shown in (78b). (78) a. John-ga kodomo-o nak-asi-ta. John-nom child-acc cry-caus-past ‘John made the child cry.’

174 martha mcginnis b. causer Voice causee vcaus



Georgian masdars provide evidence for a structure similar to that in (78b), in which the dative causee of a causativized transitive predicate is in a specifier above vcaus , which I analyse as the specifier of a high applicative phrase (Kim 2011; Legate 2014). In the verbal context, it is possible for the causer, causee, and theme of a causativized transitive in Georgian to merge as DP arguments. For example, in the present-tense clause in (79), the causer Soso is nominative, the (optional) causee Dato is dative, and the theme c’igns ‘book’ is also morphologically dative—though the case of the theme is regarded here as accusative, since it alternates with nominative case in clauses with an ergative or dative subject. For example, in the aorist clause in (80), the causer is ergative, the causee is once again an optional dative DP, and the theme is nominative. (79)

soso dato-s nino-s-tvis c’ign-s Soso.nom Dato-dat Nino-gen-for boot-dat a-čuk-eb-in-eb-s. vce-donate-ts-caus-ts-pres/fut.3sg ‘Soso makes Dato give Nino a book (as a present).’ (Harris 1981: 132)

(80)

soso-m dato-s nino-s-tvis c’ign-i Soso-erg Dato-dat Nino-gen-for book-nom a-čuk-eb-in-a. vce-donate-ts-caus-aor.3sg ‘Soso made Dato give Nino a book (as a present).’ (Harris 1981: 132)

Causative clauses with a dative subject appear to have a different argument structure. For the most part, a Georgian clause permits only one ‘true’ (nonalternating) dative, which may be an applied argument, a causee, an experiencer subject, or the subject of a clause with a Series III TAM category

the structure of causatives in clausal nominalizations 175 (perfect or pluperfect). For example, in the perfect, which has an evidential interpretation, a causativized transitive has a dative causer and a nominative theme, while the causee can only be expressed as a postpositional phrase using -tvis ‘to/for’.12 (81)

turme mePe-s šen-tvis gauCmendinebia sasaxle. apparently king-dat you-for clean.caus.perf.3sg palace.nom ‘Apparently the king (has) made you clean the palace.’ (Harris 1981: 132)

In principle, then, the causee of a causativized transitive can be expressed as a PP when it cannot be expressed as a dative argument. The same is true for applied arguments, which cannot be dative in a series III clause (82) or a passive clause (recall (66)), but can be expressed as a -tvis PP. (82)

nino-s surati dauxatia Nino-dat picture.nom draw.perf.3sg ∗ vano-s }. Vano-dat ‘Nino has drawn a picture for Vano.’

{

vano-s-tvis / Vano-gen-for

(Nash n.d.)

However, a different pattern is found in the masdar. Another example of a causativized transitive clause is given in (83), which is structurally similar to (79)–(80). (83)

redakTorma KiTas es Cerili tavis žurnalši editor.erg Kita.dat this letter.nom self.gen magazine.in gamoakveqnebina. publish.caus.aor.3sg ‘The editor had Kita publish this letter in his magazine.’ (Aronson 1990: 307)

As discussed in §6.2.2, some speakers allow a masdar containing causative -in; this inter-speaker variability is indicated by the per cent symbol (%). In this context, a direct object (here Ceril-is ‘letter’) is expressed as a genitive DP in (84a) (Léa Nash, p.c.). The causer of a causative masdar may optionally 12 The literature does include examples with two dative applicatives, where one is a goal/recipient applicative and the other is a possessor (Harris 1981: 115) or a benefactive applicative (Lomashvili 2011). I leave the analysis of such ‘multiple applicatives’ to further research.

176 martha mcginnis be expressed using a mier-phrase (84a), while a goal or beneficiary can be expressed using a postpositional phrase (84b). A causee, however, cannot be expressed at all (84c) (Léa Nash, p.c.). This contrasts with the situation in the perfect or pluperfect, where, as we saw in (81), the causee can be expressed as a -tvis PP. (84) a. % redakTor-is mier Ceril-is gamo=kveqn-eb-in-eb-a. editor-gen by letter-gen pv=publish-ts-caus-ts-n ‘the causing (by the editor) of the letter to be published’ b. Cign-is mi-cem-a { lalis-tvis / lali-ze / ∗ lali-s } book-gen giving Lali-for Lali-on Lali-dat ‘giving of the book to Lali’ (Nash n.d.) c. % Ceril-is gamo=kveqn-eb-in-eb-a (∗ KiTas (-tvis / letter-gen pv=publish-ts-caus-ts-n Kita.gen -for -mier)) -by ‘the causing of the letter to be [published (∗ by Kita)]’ I propose that what accounts for the impossibility of expressing the causee is the absence of the functional category that introduces this argument in the verbal context. This would be the case if, in a causativized transitive, the causee merges outside the causative vP selected by the nominalizing head of the masdar (85). I postulate that it merges in the specifier of an applicative head ApplCausee , whose semantic function is similar to that of an instrumental applicative. That is, the causer effects a causing event—which in turn brings about the caused event—while the causee, like an instrument, plays a mediating role in effecting this causing event. (85)

VoiceP Voice′

causer

ApplP causee selected by masdar n head

Voice

Appl′ ApplCausee

vP vP √P theme

vcaus v



the structure of causatives in clausal nominalizations 177 As noted in §6.2.2, a mier-phrase can be used only to express the causer, not the causee. This restriction, which holds in both verbal and nominal contexts, follows from the analysis above, in that the causee is not associated with causative v at all. Structurally, the causer is the external argument of both vP events, while the causee plays only a mediating role in bringing these events about. On the assumption (from 6.2.2) that the mier-phrase adjoins to vP, in a double causative like (85) it could adjoin to either vP with little or no difference in meaning, since both vPs would be semantically associated with the same external argument. Why is it impossible to use a -tvis phrase to express the causee in a masdar? I hypothesize that this is because the causee interpretation of the -tvis phrase is semantically licensed by the same applicative head that introduces a dative causee. If a causee cannot be Case-licensed as a dative DP, a -tvis PP can be used instead, provided that the ApplCausee head is still merged (without a DP specifier). If the masdar does not contain this head, this analysis predicts that the -tvis PP cannot be used to express a causee. The same licensing requirement may hold for -tvis PPs that express an indirect object, on the view that indirect objects are introduced by a lower applicative head, which, unlike ApplCausee , can be contained in the masdar. By contrast, I have proposed that the mierphrase that expresses an external argument can exceptionally adjoin to vP, which is included in the masdar. The above analysis of the -tvis PP is reminiscent of an analysis that has been proposed for certain DP/PP alternations in English. In the usual case, it has been argued that DP indirect objects and to-PPs in English ditransitive clauses are associated with separate structures, and hence yield contrasts like the one in (86) (Green 1974; Oehrle 1976; Gropen et al. 1989, inter alia). Nevertheless, Bruening (2010) argues that in cases like (87), the PP is introduced by the same syntactic head as the indirect object DP in (86a). (86) a. b. (87)



The count gives me the creeps. The count gives the creeps to me.

This story is designed to give the creeps to people who hates spiders. (Bresnan et al. 2007: 72, (6b))

A similar analysis can be applied to the -tvis PPs that replace dative DPs in Georgian perfect and pluperfect clauses: they are semantically licensed by the Appl heads that introduce dative DPs in other clause types. The absence

178 martha mcginnis of the ApplCausee head from the masdar thus accounts for the impossibility of expressing a transitive causee as either a dative DP or a -tvis PP in this context.

6.4 Conclusion I have argued here that Georgian masdar nominalizations provide evidence that causees can be merged outside the causative vP. The Georgian masdar has a v-selecting nominal head that can attach outside vP, but not outside either VoiceP or the ApplP that introduces the causee of an embedded transitive predicate. The ApplCausee head semantically licenses either a dative DP or, in a perfect or pluperfect clause, a -tvis PP. Since this head cannot be contained in the masdar, the causee cannot be expressed, even as a PP. Since vP can be contained in the masdar, it can license a PP to express the external argument. An ApplP below vP can also license a PP to express an indirect object in the masdar. Structurally, then, the causee is not a lower causer; instead it resembles a high instrumental applicative, which mediates the causing event initiated by the causer. The causer itself is the external argument of both the causing and caused events. The observations described here constitute evidence for a new understanding of the relation between argument and event structure. Just as the causative vP does not directly introduce the causer (Pylkkänen 2008), the caused event does not necessarily include the causee.

7 The Tłı  ˛cho  ˛ syntactic causative and non-nominal CPs Leslie Saxon

7.1 The wider context of this study My interest in the topic of this chapter arose out of collaboration with Tłı  ˛cho  ˛ linguist and educator Mary Siemens. We collaborated in editing a dictionary (Saxon and Siemens 1996), multimedia online dictionary (Tłı  ˛cho  ˛ Community Services Agency 2005–), literacy manual (Marinakis, Richardson, Saxon, and Siemens 2010 [2007]), and other publications in Tłı  ˛cho  ˛ (Dogrib), a northern Dene (Athabaskan) language spoken in communities between Great Slave Lake and Great Bear Lake of the Northwest Territories, Canada. Mary taught me to carefully observe contrasts between suffixed and unsuffixed forms of verbs, and noted, for example, the triplet nezı  ˛ ‘it is good’, nezı  ˛ı  ˛ ‘(the) one that is good’, and nezı  ˛ ˛ı ̀ ‘in a good manner; well’.1 She identified the complement clauses in causative and inchoative sentences as a context for suffixation by a marked-tone mora. In what follows I call this suffix the adverbial complementizer (C). The presence of this complementizer yields phonologically long vowels, which are not always easy to hear or spell.2 In support of our efforts for the development of the literacy manual I followed up 1 Examples are written using the Tłı  ˛cho  ˛ practical orthography, which is not fully standardized. Vowels a, e, o roughly represent IPA values; ı = /i/. Vowels with an ogonek are nasal and grave accent marks low tone. C’ = glottalized C. Voiced obstruent symbols represent plain voiceless obstruents, and voiceless obstruent symbols represent voiceless aspirated obstruents. ł = /ɬ/, sh = /ʃ/, zh = /ʒ/, gh = /γ/, x = /x/, j = /ʤ/, tł = /tɬ/. VV = two-mora (phonologically long) vowel. The representation of long vowels is variable in the Tłı  ˛cho  ˛ orthography and may be represented or not according to the writer. Here, except when quoting another source, I choose to represent vowel length in spellings. Abbreviations used are: 1, 2, 3, 4 = first, second, third, ‘fourth’ person; acc = accusative; adv = adverbial suffix (Japanese); area = areal prefix; C or comp = complementizer; cop = copula; foc = focus; fut = future; ipfv = imperfective; L = low tone; neg = negative; nmlz = nominalizer; O = object; opt = optative; pl = plural; past = past tense; pfv = perfective aspect; pns = possessed noun suffix; Q = question morpheme; refl = reflexive; sg = singular; S = subject; top = topic. 2 Preliminary phonetic studies by Martel and Saxon (2005) and Hucklebridge (2015) observed neutralization of the vowel length contrast word-finally. Hucklebridge’s study further suggests that tonal contrasts in vowels are maintained despite neutralizations in vowel length. Leslie Saxon, The Tłı ˛cho  ˛ syntactic causative and non-nominal CPs In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Leslie Saxon. DOI: 10.1093/oso/ 9780198817925.003.0007

180 leslie saxon on Mary’s observations. This chapter is one result. I am responsible for errors of description and interpretation.

7.2 Introduction and findings In this chapter my goal is to develop an analysis of the syntactic properties of the Tłı  ˛cho  ˛ verb ats’ele,3 a verb of broad usage which can be glossed ‘cause, let, do to’. I show that ats’ele requires two complements in all of its uses: a DP and an ‘adverbial’ phrase in a sense to be made clear. Since the causative construction built on this verb includes a clausal complement, this means that the complement in the causative usage is adverbial.⁴ Thus, the results of this study inform our understanding of the complement system of the language, and challenge a commonly held view that complementizers invariably derive nominal categories, as discussed in Rosenbaum (1967) and many subsequent authors. (For discussion see, among others, Stowell 1981; Grimshaw 1990; works in Kornfilt and Whitman 2011; and Paul 2014.) This view of the complement to ats’ele is in line with work by ?Kratzer (2016) and Moulton (2009, 2015), who identify classes of clausal complements as modifiers or restrictors in the sense of Chung and Ladusaw (2004). In the conclusion to this chapter I speculate that the semantic type of the adverbial complement is predicate. This stance mirrors ?’s (?) proposal which is at the centre of Moulton’s (2009; 2015) analysis of the CPs of CP-taking verbs like believe, admit, or regret: ‘that that-clauses are predicates of propositional content’ (Moulton 2015: 308 [emphasis added]). While English, among other languages, doesn’t have distinct complementizers for CPs interpreted as predicates contrasting with other semantic types, Tłı  ˛cho  ˛ does. In Tłı  ˛cho  ˛, the causative complement clause is explicitly marked by the adverbial suffix.⁵

3 The verb in all of its forms can also be spelled with an initial glottal stop: ʔats’ele, reflecting Tłı  ˛cho  ˛ orthographic practice. ⁴ As noted in §7.1, the clausal complement of the inchoative verb ats’ede ‘become’ is also adverbial, as in (i). I leave study of inchoatives to future research. naowo ̀ nezı  ˛ı  ˛ eyıts’ ˛o naowo ̀ nezı  ˛-le k’ehoahso ̀  ˛ ˛ ò aahde ha. [ naowo ̀ nezı  ˛ı  ˛ eyıts’ ˛o naowo ̀ nezı  ˛-le k’ehoahso ̀  ˛- o˛̀ ] aahde ha. law good and law good-neg 2plS.ipfv.know-C 2plS.ipfv.become fut lit. ‘you will come to know good laws and bad laws’ ‘you will [be like God,] knowing good and evil’ (Genesis 3:5) ⁵ A critical difference between the CPs under discussion here is that they are not predicates of propositional content but rather of caused events. The similarities I see between this work and Kratzer’s and Moulton’s lie in our shared claims that the CPs are not nominal, and that they function as predicates.

(i)

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 181 This study of causative clauses in Tłı  ˛cho  ˛ complements semantic work by Bogal-Allbritten (2016) on the verb in Navajo nisin ‘think, want’. For sentences containing this verb, Bogal-Allbritten (2016: 189, fn. 39) hints that the complements are modifiers and argues that interpretations are ‘determined by particular functional material within the embedded clause’ (Bogal-Allbritten 2016: 194). I elaborate on this idea for Tłı  ˛cho  ˛ in the conclusion.⁶ A syntactic question about Tłı  ˛cho  ˛ causatives concerns the constituency of the clausal complement: I will show that the embedded clause has aspectual, tense, and polarity properties independent of the matrix clause, a circumstance not possible when the complement to a causative verb is a verbal projection.⁷ Tłı  ˛cho  ˛ is a northern Dene (Athabaskan) language spoken in communities between Great Slave Lake and Great Bear Lake of the Northwest Territories, Canada. Rice (1989) discusses periphrastic causatives, complement structures, and complementizers in the neighbouring and closely related language Slave (also known as Slavey or Dene). Hargus (2009) treats periphrastic causatives in Tse’khene, a Dene language of northern British Columbia; see also Cook (1984: 114 ff., 2004: 354 f., 2013: 425 ff.) on periphrastic causatives in Tsúūt’ína,̀ Dënesu  ˛łıné, and Tsilhqot’in respectively. Concerning the adverbial complementizer, Saxon (2003a,b,c) provides surveys of the use of this morpheme in Tłı  ˛cho  ˛ and across the languages of the Dene family (supporting the conclusion that it existed in Proto-Athabaskan). See also Scollon (1985), Fernald (2000), Jung (2002), Mithun (2008), and Cox (2012). As I will show briefly below, following Saxon (2003b,c), the adverbial complementizer contrasts with a nominalizer, represented in the example above, nezı  ˛ı  ˛ ‘(the) one that is good’. This item also goes back to ProtoAthabaskan. Its usage, widely described in the literature (for example, in the grammars cited above), is analysed syntactically and semantically by Wilhelm (2014) for Dënesu  ˛łıné. We see below that the causative verb ats’ele selects a DP⁸ and a CP headed by a complementizer of the form V̀ (that is, a vowel mora with marked tone). The causative construction is characterized by control: the DP complement of ats’ele controls into the CP complement expressing the caused event. Although it has clause-related properties independent of the matrix, this CP does ⁶ Bogal-Allbritten (2016: 198) argues more generally in support of conclusions by ? and Moulton that embedded clauses or phrases contribute greatly to the interpretations of the main clause, allowing fully compositional semantics. My unified treatment across motion and causative interpretations of ats’ele aligns well with their approach. ⁷ An important question for Moulton (2015) is the precise structural position and movement history of the CP complement in sentences like English I decided that he was a fraud. Though this matter is important, I will not pursue it for Tłı  ˛cho  ˛ beyond brief consideration in §7.4.3. ⁸ I will assume without argument that ‘noun phrases’ in Tłı  ˛cho  ˛ are of category DP. Nothing in this chapter depends on this designation.

182 leslie saxon not include any modal or discourse-related categories, and so I identify the complement as TypeP, a category larger than IP but not as large as matrix clauses, following Ceong (2011) and other works examining an articulated CP (Rizzi 1997, 2001, among others). In addition to detailing the clausal complement in causative sentences, this chapter seeks to raise—if not fully answer—the question of the function of adverbial complements, in Dene languages and generally. Though the semantics of causative constructions is not a focus of this chapter, I will outline the hypothesis that the adverbial suffix derives predicates (cf. Wilhelm 2014 on nominalizations in Dene and their semantic status as arguments).

7.3 Methodology Most of the Tłı  ˛cho  ˛ example sentences in this chapter come from authoritative Tłı  ˛cho  ˛ translations of the New Testament and the book of Genesis (Dogrib Translation Committee 2003, 2008). This source is truly exceptional for present purposes as it is full of stories of God and others causing a great many events. Biblical examples are supplemented by examples from elicitation with Rosa Mantla. The Tłı  ˛cho  ˛ translations were produced by a translation team headed by Mary Siemens and Jaap Feenstra, who based their work on the original languages of biblical writings. The examples presented are rendered as in the published form of the book, available online.⁹ Each example is provided in a form with ordinary punctuation and spacing as well as with spacing and punctuation suitable for glossing. The English glosses, morphological analyses, and literal translations found in this chapter are due to the author, as is my selection of the English biblical translation, cited from the New Revised Standard Version, accessed through http://www.biblegateway.com/. I am responsible for all errors of interpretation. The next section of this chapter serves as a brief introduction to Tłı  ˛cho  ˛ syntax, the morphology associated with the causative verb ats’ele, and the complement structure of this verb. §7.5 is a study of the internal syntax of the clausal complements to ats’ele, done with a view to establishing that the syntactic category of the complement is larger than vP (cf. Ritter and Rosen 1993; Folli and Harley 2007; Bjorkman and Cowper 2013; Chung 2017, among others, on causatives in English, Romance, and other languages). §7.6 concludes. ⁹ The text of the Tłı  ˛cho  ˛ Bible is available online at http://www.bible.com/versions/800-dogntnohtsi-nihtle.

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 183

7.4 Introduction to Tłı ˛cho  ˛ syntax, verb morphology, and the complement structure of ats’ele ‘cause’ Tłı  ˛cho  ˛ syntax shows many of the expected typological features of a headfinal language, including the use of postpositions and of post-verbal functional items identifiable as heads of the IP and CP domains (see Saxon 1986, 2014; Rice and Saxon 2005; Welch 2012, 2015a,c). The verb is final in its domain, as we see in the examples that follow. The verb under study can be translated into English in many ways, as ‘cause’, ‘let’, ‘do to’, ‘send’, and ‘take’, according to syntactic and discourse context. In the interlinearization I have glossed it as ‘cause’. Invariably it requires a DP direct object plus one other (non-nominal) complement. The second complement can be a PP, an adverb, or a clause marked with the complementizer of the form V.̀ A generalization that we can draw is that all of these are ‘adverbial’, in the way that the second complement to the English verb word is ‘adverbial’: We worded it carefully / in a round-about way / thus. In the subsections below I provide a few examples of the verb’s use with different complement types and conclude with a justification for this generalization. Verbs in Tłı  ˛cho  ˛, as in other Dene languages, have no bare or infinitival forms. The word ats’ele is a citation form that can be translated as ‘to cause’ or ‘one causes’. Many verbs in Tłı  ˛cho  ˛ show ablaut in the verb stem: for ats’ele, the imperfective and optative stem is -le; the perfective stem is -la.̀ The passive stems include a d- prefix. In Table 7.1 I provide examples of some of the forms of this verb to prepare the reader for their variety. In these examples the verb stem is bolded.1⁰ In the examples below, as here, morphological categories expressed within the verb word are listed but not individually identified. In the subsections that follow, I show the complement structure options for ats’ele.

7.4.1 [DP] [PP] ats’ele With DP and PP complements ats’ele is typically translated as a transitive verb of caused motion, ‘send’, ‘bring’, or ‘take’. The PP is strictly required and usually 1⁰ The stem of this verb is different in its progressive and habitual forms (ats’ehʔı  ˛ and ats’ehʔ ˛ı ̀ respectively). This chapter does not include examples illustrating these variants but they show all the same subcategorization patterns. In the glossing, object inflection is listed before subject inflection to reflect morpheme order; this decision has no theoretical significance.

184 leslie saxon Table 7.1 Sample of forms of the Tłı  ˛cho  ˛ verb ats’ele Aspect and voice

Arguments

Form

Verbal gloss

imperfective active imperfective reflexive imperfective passive perfective active perfective reflexive perfective passive optative

2sO.1sS refl.1pS 3O 3O.1pS refl.3pS 2pO 2pO.3pS

anehłe adets’ele awedle awets’ ˛ ˛l ıı̀ ̀ a ̀ adeg ˛ ˛l ıı̀ ̀ a ̀ anaxııdl ̀ ̀ à anaxegııle

‘I cause you (sg.) …’ ‘we cause ourselves …’ ‘she/he is caused …’ ‘we caused him …’ ‘they caused themselves …’ ‘you (pl.) were caused …’ ‘may they cause you (pl.) …’

Source: Dogrib Translation Committee (2003, 2008)

immediately precedes the verb. The PP is bolded alongside the verb in the examples below. (1)

N o˛htsı ̀  ˛ cheekee ˛ł ̀ ı e ̀ wexe ̀ ats’ele ha. [ N o˛htsı ̀  ˛ cheekee ̀  ˛ł ı e ̀ ] [wexe]̀ ats’ele ha. God follower.pns one 3.with 1plS.ipfv.cause fut lit. ‘we are going to send one disciple (=God’s follower) with him’ ‘With him we are sending the brother’ (2 Corinthians 8:18)

(2)

do  ˛-eyagı  ˛ı  ˛lı  ˛ı  ˛ hazo  ˛ o˛̀ Zezı ̀ ts’ o˛̀ agog ˛ı ˛l ı̀ ̀ a ̀ [ do  ˛-eyagı  ˛ı  ˛lı  ˛-ı  ˛ hazo  ˛ o˛̀ ] [ Zezı ̀ ts’ o˛̀ ] person-3plS.be_sick-nmlz all Jesus to agog ˛ı ˛l ı̀ ̀ a ̀ 3plO.3plS.pfv.cause lit. ‘they brought all the people who were sick to Jesus’ ‘they […] brought all who were sick to [Jesus]’ (Matthew 14:35)

(3)

m o˛hdaa ̀ eyıı-le ̀ nek’ ̀ e ts’ o˛̀ agogele ha [m o˛hdaa] ̀ [ eyıı-le ̀ nek’e ̀ ts’ o˛̀ ] agogele ha some that-neg country to 3plO.3plS.ipfv.cause fut lit. ‘they are going to send some of them to other (lit. “not that”) countries’ ‘[they will] be taken away as captives among all nations’ (Luke 21:24)

Example (4) contains the featured structure in the main clause (‘I am going to send you to them’) as well as in the bracketed embedded purpose clause (‘for you to take the ones sitting in darkness into the light’).

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 185 (4)

togootł’ ̀ oò geehkw’ ̀ ee sıı̀ dze  ˛h ts’ o˛̀ agı  ˛ı  ˛le gha gıts’ o˛̀ anehłe ha [ [ togootł’ ̀ oò geehkw’ ̀ e-e sıı̀ ] [ dze  ˛h darkness 3plS.ipfv.sit-nmlz foc light agı  ˛ı  ˛le gha ] [gıts’ o˛]̀ anehłe 3plO.2sgS.ipfv.cause for 3pl.to 2sgO.1sgS.ipfv.cause

ts’ o˛̀ ] to ha fut

lit. ‘for you to take the ones sitting in darkness into the light, I am going to send you to them’ ‘to whom I am sending you […] so that they may turn from darkness to light’ (Acts 26:17–18)

7.4.2 [DP] [AdvP] ats’ele With DP and AdvP complements, the verb translates as a causative or as a transitive verb of caused motion, as also seen in §7.4.1. Both complements are strictly required. In the following examples the AdvP and V are bolded. (5)

Israel got’ı  ˛ ˛ı ̀ edını ̀ ładı  ˛ ˛ı ̀ agı  ˛ ˛l ı ̀ a ̀ nı  ˛de ̀ Israel got’ı  ˛ ˛ı ̀ [edını]̀ [ładı  ˛ ˛] ı̀ agı  ˛ ˛l ı ̀ à nı  ˛de ̀ Israel inhabitant refl.mind differently 3O.3plS.pfv.cause if lit. ‘if the people of Israel made their minds different(ly)’ ‘even those of Israel, if they do not persist in unbelief ’ (Romans 11:23)

The adverb ładı  ˛ ˛ı ̀ ‘differently’ in (5) bears a suffix of the same shape as the adverbial complementizer to be seen in §7.4.3, as do the adverbs in (6) and (7). Note in contrast ładı  ˛ı  ˛ ‘different’, as in n o˛htsı ̀  ˛ ładı  ˛ı  ˛‘‘foreign divinities’ (Acts 17:18), lit. ‘different gods’. The tone marking on the last vowel of ładı  ˛ ˛ı ̀ makes the word an adverb rather than an adjective. These facts are the source of my use of the term ‘adverbial complementizer’.11 In (6) and (7) the direct object of ats’ele is expressed overtly only through object inflection in the verb: s- ‘first person singular object’ in (6) and go- ‘first person plural object’ in (7).

(i)

11 Japanese shows a pattern similar to what we see in (5), as observed by Junko Shimoyama: Hana-wa hashigo-o taka-ku shi-ta. Hana-top ladder-acc high-adv do-past ‘Hana made the ladder high(er)’ (Junko Shimoyama, p.c.)

This pattern is found with adjectives in Japanese but not verbs.

186 leslie saxon (6)

(7)

“Zezı ̀ k’aat’ı ̀ ı ̀ as ˛ı ˛l ı̀ ̀ a,̀ ” gohdı. ̀ “Zezı ̀ [k’aat’ı ̀ ı]̀ as ˛ı ˛l ı̀ ̀ a,̀ ” gohdı. ̀ Jesus healthily 1sgO.3S.pfv.cause 3plO.3S.ipfv.tell lit. ‘“Jesus made me well,” he told them.’ ‘it was Jesus who had made him well’

(John 5:15)

N o˛htsı ̀  ˛ daǹ ı ̀ gogho  ˛neeto ̀  ˛o  ˛ sıı̀ ch’a ˛łak’a ̀ı a ̀ agogele ha nııle. ̀ N o˛htsı ̀  ˛ daǹ ı ̀ gogho  ˛neeto ̀  ˛-o  ˛ sıı̀ ch’a ̀ [ı  ˛łak’aa]̀ God how 1plO.3S.ipfv.love-nmlz foc from separately agogele ha nııle. ̀ 1plO.3plS.ipfv.cause fut not lit. ‘[they] will not keep us separately from how God loves us’ ‘[nothing] will be able to separate us from the love of God […]’ (Romans 8:38)

These examples have a causative sense parallel to the literal English translations. In (8) the adverb expresses directionality, and the sentence reads as involving motion. (8)

Do  ˛ kwe ˛ʔ ı  o˛ ̀ o˛̀ agı  ˛ ˛l ı ̀ a.̀ Do  ˛ [kwe] [ı  ˛ʔ o˛ ̀ o˛]̀ agı  ˛ ˛l ı ̀ a.̀ person stone away 3sgO.3plS.pfv.cause ‘[people] took away the stone’

(John 11:41)

Note that the DP and Adv precede the verb in that order. In other contexts direct objects usually immediately precede the verb in Tłı  ˛cho  ˛, with adjuncts appearing earlier, as in (9). (9)  ˛wha I  ˛ ˛a ̀ goht’ ˛o denahk’e nezı  ˛ı  ˛ sıı̀ nıahwha ̀ [I  ˛wha  ˛ ˛] à [ goht’ ˛o denahk’e nezı  ˛ı  ˛ quickly clothes most good-nmlz nıahwha ̀ 2plS.ipfv.put_down.pl lit. ‘quickly put down the best clothes’ ‘Quickly, bring out a robe—the best one’

sıı̀ foc

]

(Luke 15:22)

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 187 The predication relation between the DP and PP or AdvP in constructions with ats’ele appears to dictate subject-predicate word order, and also marks the PP or AdvP as a complement to ats’ele.12 Dene languages, including Tłı  ˛cho  ˛, are known to have extensive sets of adverbial prefixes for verbs. Just as the direct object of ats’ele may be expressed by a verbal prefix, so may the adverbial phrase. Forms of the verbs hats’ele ‘cause thus; do thus to’ and dats’ ̀ ele ‘cause what; do what to’ occur in (10). Haand da-̀ (underlined) are adverbial prefixes glossed ‘thus’ and ‘Q’ respectively and occur productively in the language with a limited number of verb stems; compare hats’ı  ˛ı  ˛wo  ˛ ‘to want/think thus’, hats’edı ‘to say thus’; dats’ı ̀  ˛ı  ˛wo ̀ ‘to want/think what’, dats’ ̀ edı ‘to say what’ (Tłı  ˛cho  ˛ Community Services Agency 2005–) . (10)

Dawı ̀  ˛ı  ˛le ha newo  ˛o  ˛ sıı̀ hawı  ˛ı  ˛le […]. [ Dawı ̀  ˛ı  ˛le ha newo  ˛-o  ˛ Q.3O.2sgS.ipfv.cause fut 2sgS.ipfv.want-nmlz hawı  ˛ı  ˛le […]. thus.3O.2sgS.ipfv.cause lit. ‘what you wish to do to her, do thus to her’ ‘do to her as you please’

sıı̀ foc

]

(Genesis 16:6)

Each of the two complements of ats’ele ‘cause’ is expressed by a verbal prefix.

7.4.3 [DP] [CP] ats’ele Ats’ele also allows a clausal complement to appear with the direct object, and this combination yields a causative translation for ats’ele. The CP is strictly required and is headed by a complementizer of the form V.̀ This form is the same as the form of the adverbial suffix observed in ładı  ˛ ˛ı ̀ ‘differently’ in (5).13 Concerning this suffix, in the biblical sources there is a convention such that the complementizer is represented orthographically only when the result is a 12 Many thanks to an anonymous reviewer, who points out a striking similarity with comparative aspect-marked verbs in Navajo. Like ats’ele, they obligatorily select non-nominal complements of the same types described here (with the exception of clauses) and likewise exhibit restricted word order options (Bogal-Allbritten 2013). 13 Tłı  ˛cho  ˛ also has a synthetic causative, as in, for example, wenats’ehdı ̀ ‘remind’ and ts’eeht ̀ ı ̀ ‘make expensive; charge a lot for’, derived from wenats’edı ̀ ‘remember’ and det̀ ı ̀ ‘be expensive’. It will not be discussed here, but descriptions of the patterns can be found in many grammars of Dene languages, sometimes called the ‘classifier causative’ for the morpheme involved.

188 leslie saxon long vowel with a contour tone. When it creates merely a long vowel (when the preceding syllable has a marked tone lexically), the complementizer is not represented. The lengthy example in (11) involves two conjoined causatives, literally ‘he will cause us too to be resurrected with Jesus’ and ‘he will cause us to be standing in his presence with you’. Compare the spellings nats’ıd̀ à ‘living again’ in the first conjunct with nats’ ̀ eehza ̀ à ‘standing’ in the second to see this orthographic convention in operation. (11)

goxı  ˛ sı Zezı ̀ xe ̀ nats’ıd̀ a ̀ agole ha ho  ˛t’e, eyits’ ˛o naxıxe ̀ wenada  ˛ ˛a ̀ nats’ ̀ eehza ̀ à agole ha wek’ets’ ̀ eezo  ˛ [ [ goxı  ˛ sı ] [ Zezı ̀ xe ̀ nats’ıd̀ a ̀ ] 1pl also Jesus with 1plS.ipfv.live_again.C agole ha ho  ˛t’e ], eyits’ ˛o [ [ naxıxe ̀ 1plO.3S.ipfv.cause fut 3S.ipfv.be and 2pl.with wenada  ˛ ˛a ̀ nats’e ̀ ehzà à ] agole ha ] 3.in_presence 1plS.ipfv.stand-C 1plO.3S.ipfv.cause fut wek’ets’ ̀ eezo  ˛ 3O.1plS.ipfv.know lit. ‘we know that he will cause us too to be resurrected with Jesus and that he will cause us to be standing in his presence with you’ ‘we know that [he] will raise us also with Jesus, and will bring us with you into his presence’ (2 Corinthians 4:14)

This convention is in operation throughout the corpus, and is responsible for the spellings observed.1⁴ In some other cases too the adverbial complementizer is not seen in spellings.1⁵ The following examples provide further exemplification of the periphrastic causative. (12)

do  ˛zh ˛ı ˛ı̀ ̀ t’a ̀ ehdaa ̀ aseneel ̀ à [ do  ˛zh ˛ı ˛ı̀ ̀ t’a ̀ ehda-a ̀ ] aseneel ̀ à person.body with 1sgS.ipfv.live-C 1sgO.2sgS.pfv.cause lit. ‘you have made me live with/using a human body’ ‘a body you have prepared for me’ (Hebrews 10:5)

1⁴ This orthographic practice matches Hucklebridge’s (2015) preliminary findings that even when Tłı  ˛cho  ˛ vowel length is neutralized tonal contours are maintained. 1⁵ For example, the suffix is not spelled out when it attaches to a clitic such as ha ‘fut’ or -le ‘neg’. For the purposes of this chapter I include a gloss ‘C’ regardless of whether the morpheme is explicitly represented in orthography.

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 189 (13)

sı  ˛ whatso  ˛ o˛̀ whıhda aseahłe ha [sı  ˛] [ whatso  ˛ o˛̀ whıhda ] aseahłe ha 1sg alone 1sgS.ipfv.be.C 1sgO.2plS.ipfv.cause fut lit. ‘you will make me be alone’ ‘you will leave me alone’ (John 16:31)

(14) shows a passive form of ats’ele, with the passive verb stem -dle. (14)

hazo  ˛ o˛̀ naxık o˛̀ ts’ o˛̀ naahde ̀ anaxedle ha [hazo  ˛ o˛]̀ [ naxık o˛̀ ts’ o˛̀ naahde ̀ all 2pl.house to 2plS.ipfv.return.pl.C anaxedle ha 2plO.ipfv.be_caused fut lit. ‘you will all be made to return to your homes’ ‘you will all be scattered, each one to his home’

]

(John 16:31)

Ats’ele in (15) takes a first person singular reflexive form with the use of the reflexive prefix de-. The perfective stem is -la.̀ (15)

do  ˛ hazo  ˛ o˛̀ gıgha eghalaehdaa ̀ do  ˛ o˛̀ ehłı  ˛ ˛ı ̀ adehł ̀ à [ do  ˛ hazo  ˛ o˛̀ gıgha eghalaehda-a ̀ everyone 3pl.for 1sgS.ipfv.work-nmlz ehłı  ˛- ˛ı ̀ ] adehł ̀ à 1sg.ipfv.be-C refl.1sgS.pfv.cause

do  ˛- o˛̀ person-pns

lit. ‘I have made myself be a person who works for everyone’ ‘I have made myself a slave to all’ (1 Corinthians 9:19) Let’s now consider the constituency of the DP and CP preceding the verb. Are they separate complements to ats’ele, as in (16), or do they form a small clause, as in (17)?

190 leslie saxon (16)

Ats’ele: Structure with two complements (argued for) vP v′

DP VP

v V′

DP CP (17)

V

Ats’ele: Structure with a small clause complement (argued against) vP v′

DP VP Small Clause DP

v V

CP

The assumption that ats’ele has just a single subcategorization in all of its senses, as a verb of motion or a causative verb, might argue in favour of the first option. The verbal forms in (10) also argue for the first option, as verbal prefixes mark each complement separately. In connection with this argument, an anonymous reviewer notes that small clauses in English allow separate pronominalization of the subject and predicate, as in I consider that student a phenomenal genius, and Fred considers her one too. I note that English allows a small clause to be pronominalized as a whole: They believed that student to be a phenomenal genius, and Fred believed so too. This latter possibility does not exist with the Tłı  ˛cho  ˛ causative.

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 191 The question arises anew, though, with examples like (18). (18)

Nexe ̀ sıgh ̀ a ̀ hoʔo ̀  ˛ o˛̀ agohłe ha […]. [ Nexe ̀ sıgh ̀ a ̀ hoʔo ̀  ˛- o˛̀ ] agohłe 2sg.with peace area.ipfv.be-C 1sgS.area.ipfv.cause ha […]. fut lit. ‘I will cause there to be peace with you […].’ ‘I […] will bless you […]’

(Genesis 26:24)

The literal translation, ‘I will cause there to be peace with you’, suggests object raising and therefore a small clause analysis. The causative verb here includes the prefix go-, an object agreement form marking a location, situation, or state of affairs, glossed ‘areal’, parallel to ‘there’ in the English literal translation. We see the same areal prefix together with the adverbial prefix ha‘thus’ in hagohłe, the main verb in (19). (19)

Secheekee ̀ Abraham wets’ıhʔ o˛̀ negha hagohłe ha. Secheekee ̀ Abraham wets’ıhʔ o˛̀ negha hagohłe 1sg.follower Abraham 3.because 2sg.for thus.1sgS.area.ipfv.cause ha. fut lit. ‘I will cause [the situation] [to be] thus for you because of my follower Abraham’ ‘(I […] will bless you […]) for my servant Abraham’s sake’ (Genesis 26:24)

I have glossed the verb, ‘I cause a situation [to be] thus’. On the timehonoured assumption that pro-forms represent individual constituents, harepresents the CP and go- the causee. From such examples, I conclude that not only people but also states of affairs are legitimate causees. The particular state of affairs, when spelled out in a sentence, is identified by a distinct CP complement. Thus I conclude that the clausal complement of ats’ele is not contained in a small clause; that is, that (16) sketches the vP structure of causative sentences.

192 leslie saxon

7.4.4 Summary: An adverbial complement not a nominalized clause The discussion above shows that the causative verb ats’ele requires two complements: one a DP and the other a PP, Adv, or CP-V.̀ Considering this fact and the fact that the complementizer marking the CP complement is identical to a suffix observed on adverbs, it is possible to generalize and say that the second complement must be adverbial.1⁶ Further supporting this conclusion is the contrast between the adverbial suffix or complementizer -V̀ and the nominalizing suffix -V, glossed ‘nmlz’ in this chapter. This suffix, a mora with unmarked tone, marks some Tłı  ˛cho  ˛ nominals and nominalized clauses. It appears in examples (2), (4), (7), (9), (15), and others elsewhere in this chapter. See Wilhelm (2014) on the nature of nominalization in the closely related language Dënesu  ˛łiné. Compare the nominalizations bracketed in (20) and (21) with the CPs representing caused events in examples (12) and (15), containing the same embedded verbs. (20)

[…] daǹ ı ̀ ehkw’ı geedaa gho  ˛ hoghagoehto ̀  ˛ ha. [ […] daǹ ı ̀ ehkw’ı geeda-a how right 3plS.ipfv.live-nmlz hoghagoehto ̀  ˛ ha. 3plO.3S.ipfv.teach fut lit. ‘he will teach them about how to live right’ ‘he will proclaim justice to [them]’

]

gho  ˛ about

(Matthew 12:18)

1⁶ John Lyon (p.c.) asks about Tłı  ˛cho  ˛ translations of examples like ‘they made me the president’, with a nominal expressing the caused state of affairs. I have observed two options in Tłı  ˛cho  ˛. Future study will consider the semantic and pragmatic differences between them. One option is to use the copula ‘…be the president’, as is seen in (15). The other is to use a different verb, ts’ehtsı  ˛ ‘make; construct’, as in (i)–(ii): (i) N ˛htsı ò  ˛ Egypt nek’ ̀ e gha k’aowo ̀ sehtsı ̀  ˛. N ˛htsı ò  ˛ [ Egypt nek’ ̀ e gha k’aowo ̀ ] sehtsı ̀  ˛. God Egypt country for boss 1sgO.3S.pfv.make ‘God has made me lord of all Egypt’ (Genesis 45:9) (ii) Eyı do  ˛ wegoahʔo ̀  ˛ nı  ˛de ̀ naxı-nahk’e wehłı  ˛za weahtsı  ˛. Eyı do  ˛ wegoahʔo ̀  ˛ nı  ˛de ̀ naxınahk’e wehłı  ˛za weahtsı  ˛. that person 3O.2plS.pfv.find if 2pl-more_than 3.devil_child 3O.2plS.ipfv.make lit. ‘when you have found that person, you make him a devil’s child more than you’ ‘you make the new convert twice as much a child of hell as yourselves’ (Matthew 23:15) Such examples are notable because ts’ehtsı  ˛, the ordinary transitive verb ‘make’, appears to be morphologically unchanged with the extra complement.

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 193 (21)

Eta ̀ weza  ˛hda o a ̀ ehłı  ˛ı  ˛ sıı̀ segha ayıı̀ wele?̀ [ Eta ̀ weza  ˛hda o a ̀ ehłı  ˛-ı  ˛ ] sıı̀ segha ayıı̀ father 3.son elder 1sgS.ipfv.be-nmlz foc 1sg.for what wele?̀ 3S.opt.be lit. ‘what would it be for me, being my father’s oldest son?’ ‘of what use is a birthright to me?’ (Genesis 25:32)

The two suffixes have different tones and clearly are not the same. Also, compare (19) with (22). (22)

do  ˛ wı  ˛ı  ˛zıı̀ yegho  ˛ hoghase ̀ ehto ̀  ˛ nııle ̀ do  ˛ wı  ˛ı  ˛zıı̀ yegho  ˛ hoghase ̀ ehto ̀  ˛ nııle ̀ person at_all 3.about 3S.1sgO.pfv.teach neg lit. ‘nobody at all taught me about it’ ‘nor was I taught it’

(Galatians 1:12)

In (22) a nominalized clause is replaced by the pronominal prefix ye- ‘it’, and not by the adverbial prefix ha- ‘thus’, as in (19).

7.5 Properties of the caused event complement The goal of this section is to answer the question of the syntactic category of the clause headed by the adverbial complementizer. I assume the clausal structure in (23) for Tłı  ˛cho  ˛, due to Welch (2015b). Welch provides evidence for this structure based on the co-occurrences and sequencing of the heads instantiating phrasal categories.

194 leslie saxon (23)

Tłı  ˛cho  ˛ clause structure CP C

FocP Foc

NegP Neg

ModP Mod

IP AspP Asp

vP VP

I

v (Welch 2015b)

From observing evidence relating to AspP, IP, and NegP in the clausal complements to ats’ele I suggest a slight revision to the hierarchy shown in (23) and conclude that ats’ele selects TypeP, a category in the expanded CP domain proposed in Ceong and Saxon (2013), building on proposals of Cheng (1997) and Rizzi (1997, 2001). In §§7.5.1 to 7.5.4 I make this case.

7.5.1 Independent aspect in the complement clause Aspectual properties of the complement to ats’ele are not dependent on the aspectual properties of ats’ele, arguing that the complement is of category AspP or larger. Elicitation shows that indeed aspectual marking in the complement and matrix clauses is semantically based. 7.5.1.1 Complements of perfective ats’ele In (24)–(29), we see examples involving ats’ele in the perfective aspect. Examples (24)–(26) show perfective complements. The complement verbs

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 195 xage ̀ ede ̀ ‘they exited’, agı  ˛ ˛l ı ̀ à ‘they caused’, and t’asawee  ˛la-le ̀ ‘you did not harm her’ show distinctive perfective stem forms. (24)

Do  ˛ hazo  ˛ o˛̀ xage ̀ ede ̀ e ̀ agog ˛ı ˛l ı̀ ̀ a ̀ tł’axo  ˛ o˛,̀ […] [ Do  ˛ hazo  ˛ o˛̀ ] [xage ̀ edè e]̀ person all 3plS.pfv.exit.pl-C tł’axo  ˛ o˛,̀ […] after lit. ‘after they made all the people go out, […]’ ‘when the crowd had been put outside […]’

(25)

agog ˛ı ˛l ı̀ ̀ a ̀ 3plO.3plS.pfv.cause

(Matthew 9:25)

Yedayeh ̀ Nezı  ˛ı  ˛ Asıa nek’ ̀ e godı nezı  ˛ı  ˛ do  ˛ ts’ o˛̀ agı  ˛ ˛l ı ̀ a ̀ ag o˛ ̀ o˛l̀ a-le. ̀ Yedayeh ̀ Nezı  ˛ı  ˛ [ Asıa nek’ ̀ e godı nezı  ˛ı  ˛ do  ˛ ts’ o˛̀ Holy Spirit Asia country story good person to agı  ˛ ˛l ı ̀ à ] ag o˛ ̀ o˛l̀ a-le. ̀ 3O.3plS.pfv.cause.C 3plO.3S.pfv.cause-neg lit. ‘the Holy Spirit didn’t let them take the gospel to the people in Asia’ ‘[they had] been forbidden by the Holy Spirit to speak the word in Asia’ (Acts 16:6)

(26)

t’asawee  ˛la-le ̀ anehł ̀ à [t’asawee  ˛la-le] ̀ anehł ̀ à 3O.2sgS.pfv.harm-neg.C 2sgO.1sgS.pfv.cause lit. ‘I made you not harm her’ ‘I didn’t let you touch her’

(Genesis 20:6)

The examples in (27)–(29) show caused events expressed using imperfective forms of verbs. Note again that in these examples, as in (24)–(26), the causative verb ats’ele is perfective. (27)

[…] m o˛hdaa ̀ sı  ˛laeno ̀  ˛ ełexe ̀ geehkw’ ̀ ee ̀ agog ˛ı ˛l ı̀ ̀ a.̀ […] [m o˛hdaa] ̀ [ sı  ˛laeno ̀  ˛ ełexe ̀ geehkw’è è some fifty together 3plS.ipfv.sit-C agog ˛ı ˛l ı̀ ̀ a.̀ 3plO.3plS.pfv.cause lit. ‘they made some sit fifty together’ ‘they sat down in groups […] of fifties’

]

(Mark 6:40)

196 leslie saxon (28)

N o˛htsı ̀  ˛ wexe ̀ ts’ewhı ̀  ˛ ˛ı ̀ ts’eedaa ̀ ag o˛ ̀ o˛l̀ a.̀ N o˛htsı ̀  ˛ [ wexe ̀ ts’ewhı ̀  ˛ ˛ı ̀ ts’eeda-a ̀ ] ag o˛ ̀ o˛l̀ a.̀ God 3.with peacefully 1plS.ipfv.live-C 1plO.3S.pfv.cause lit. ‘God made us live peacefully with him’ ‘[…] will we be saved by his life’ (Romans 5:10)

(29)

Laban wets’ ˛o tıts’aad̀ ıı̀ ta whelaa ̀ ay ˛ı ˛l ı̀ ̀ a-le. ̀ [ Laban wets’ ˛o tıts’aad̀ ıı̀ ta whela-a ̀ ] Laban 3.from animal among 3S.ipfv.be-C ay ˛ı ˛l ı̀ ̀ a-le. ̀ 3O.3S.pfv.cause-neg lit. ‘he didn’t let them be among Laban’s animals’ ‘[He] did not put them with Laban’s flock’ (Genesis 30:40)

Elicitation confirms that minimal pairs are available, as we see in (30) and (31), due to Rosa Mantla. (30)

Semo  ˛ goht’ ˛o hazo  ˛ o˛̀ k’enaıhtsoò as ˛l ı ̀ a.̀ Semo  ˛ [ goht’ ˛o hazo  ˛ o˛̀ k’enaıhtso-ò ] as ˛l ı ̀ a.̀ 1sg.mother clothes all 1sgS.pfv.wash-C 1sgO.3S.pfv.cause ‘My mother made me wash all the clothes’ (Rose Mantla, p.c.) a. b. c. d. e. f.



 ˛ła I a ̀ aht’ı  ˛.  ˛ła I a ̀ wek’enaehtse. ∗  ˛ła I a ̀ wek’e eghalaehda. ̀ Hot’a ̀ wegho  ˛ naht’e.  ˛ła I a ̀ sınıehłe. ̀  ˛ła I a ̀ deyıınahłe. ̀ ∗

intended: ‘I’m still doing it’ intended: ‘I’m still washing them’ intended: ‘I’m still working on it’ ‘I’ve already finished it’ ‘I’m still piling them up’ ‘I’m still putting them back away’

The context clause in (30) features a causative in the perfective aspect with a perfectively marked caused event (verb: k’enaıhtso ‘I have washed O’). The use of the embedded perfective entails that the caused event is completed: the only continuations that are pragmatically acceptable are those which presuppose the completion of the clothes-washing. In (31) the caused event is marked instead with imperfective aspect (verb: k’enaehtse ‘I wash O’): (31)

Semo  ˛ goht’ ˛o hazo  ˛ o˛̀ k’enaehtsee ̀ as ˛l ı ̀ a.̀ Semo  ˛ [ goht’ ˛o hazo  ˛ o˛̀ k’enaehtse-e ̀ ] as ˛l ı ̀ a.̀ 1sg.mother clothes all 1sgS.ipfv.wash-C 1sgO.3S.pfv.cause ‘My mother made me wash all the clothes’ (Rosa Mantla, p.c.)

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 197 a.  ˛ła I a ̀ aht’ı  ˛. b.  ˛ła I a ̀ wek’enaehtse. c.  ˛ła I a ̀ wek’e eghalaehda. ̀ d. Hot’a ̀ wegho  ˛ naht’e. e.  ˛ła I a ̀ sınıehłe. ̀ f.  ˛ła I a ̀ deyıınahłe. ̀

‘I’m still doing it’ ‘I’m still washing them’ ‘I’m still working on it’ ‘I’ve already finished it’ ‘I’m still piling them up’ ‘I’m still putting them away’

In this case the use of the embedded imperfective in (31) is compatible with continuations where the washing is still under way or already completed. Thus we see that the embedded clause has aspect determined semantically, not dependent on aspect in the matrix. A similar generalization emerges with corpus examples featuring the complement verb weghats’ ̀ eeda ‘look at’. (32) involves the embedded imperfective verb weghats’ ̀ eeda ‘we look at it’, and (33) the embedded perfective verb ghagı ̀  ˛ı  ˛dà ‘they looked at [it]’. (32)

Sets’ o˛̀ K’aowo, ̀ Neta ̀ weghats’ ̀ eedaa ̀ agonel ̀ a ̀ nı  ˛de ̀ gogha hot’a ̀ ha. Sets’ o˛̀ K’aowo, ̀ [ Neta ̀ weghats’eedà à 1sg.to lord 2sg.father 3O.1plS.ipfv.look_at-C agonel ̀ à nı  ˛de ̀ gogha hot’a ̀ ha. 1plO.2sgS.pfv.cause if 1pl.for enough fut

]

lit. ‘My Lord, if you have made us look at your father, it will be enough for us.’ ‘Lord, show us the Father, and we will be satisfied’ (John 14:8) (33)

Zezı-Krı ̀ edecheekee ̀ ayıı̀ ghagı ̀  ˛ı  ˛da ̀ ag o˛ ̀ o˛l̀ aà [ Zezı-Krı ̀ edecheekee ̀ [ ayıı̀ Jesus-Christ refl.follower what ag o˛ ̀ o˛l̀ a-a ̀ ]DP 3plO.3S.pfv.cause-nmlz

ghagı ̀  ˛ı  ˛da ̀ 3plS.pfv.look_at.C

lit. ‘what Jesus Christ made his disciples look at’ ‘the revelation of Jesus Christ’

]

(Revelation 1:1)

The English verb show is usually translated as ‘cause to look at’ in Tłı  ˛cho  ˛, as in (32). In this example the caused event of looking is not necessarily limited in time. On the other hand, (33) is the New Testament translation of the word revelation. Here, the caused event, ‘his disciples look at [it]’, is expressed with a perfective verb. Weghats’ ̀ eeda ‘look at’, though atelic in imperfective forms,

198 leslie saxon receives a telic (completive) interpretation in perfective forms, as observed by Wilhelm (2003; 2007: 51ff.) for Dënesųłiné. She claims that perfective viewpoint in this language includes a post-time, which effectively closes off the event. Her observations ring true for Tłı  ˛cho  ˛ too. Thus the translation of ‘revelation’ featuring an embedded perfective form rightly highlights revelations as telic caused events.1⁷ 7.5.1.2 Complements of imperfective ats’ele While perfective ‘cause’ permits perfective and imperfective complements, with imperfective ‘cause’ we find imperfective but not perfective complements. Many examples above show the same pattern highlighted in (34)–(35). (34)

“Do  ˛ hazo  ˛ o˛̀ ł ˛a a ˛ ̀ tł’oh k’e geehkw’ ̀ ee ̀ agıahłe,” gohdi. ̀ “[ Do  ˛ hazo  ˛ o˛̀ ] [ ł ˛a a ˛ ̀ tł’oh k’e geehkw’è è ] person all together grass on 3plS.ipfv.sit-C agıahłe,” gohdi. ̀ 3plO.2plS.ipfv.cause 3plO.3S.ipfv.tell lit. ‘He told them, “Make all the people sit on the grass together.”’ (cf. (14)) ‘he ordered them to get all the people to sit down in groups on the […] grass’ (Mark 6:39)

The imperfective verbs geehkw’ ̀ e ‘they sit’ and ełaawı-le ̀ ‘they do not die’ identify the caused events in (34) and (35). (35)

“[…] Lazarus ełaawı-le ̀ ayele ha ˛l ı e,̀ ” gedı. “[…] [Lazarus] [ełaawı-le] ̀ ayele ha  ˛l ı e,̀ ” Lazarus 3S.ipfv.die-neg.C 3O.3S.ipfv.cause fut past gedı. 3pl.ipfv.say

1⁷ In contrast with this permitted aspectual contrast in Tłı  ˛cho  ˛, Ritter and Rosen (1993: 537, fn. 13) observe that perfect have is disallowed in the complements of English causatives. I cite their example here. (i) ∗ John had/made Bill have finished his assignment by 5:00. As noted by McCawley (1971), have in non-finite contexts may be interpreted as past or perfect, judging from its grammatical co-occurrence with the classes of adverbials compatible with simple past (They wanted to have finished yesterday) or perfect (They wanted to have finished by now). We observe that neither type of interpretation is possible in English causatives, considering (i) and (ii). (ii) ∗ John had/made Bill have finished his assignment yesterday.

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 199 lit. ‘They said, “He should have caused Lazarus not to die.”’ ‘[they] said, “Could not he […] have kept [Lazarus] from dying?”’ (John 11:37) These examples and those earlier in the chapter show that both atelic (‘sit’, ‘be’, ‘not die’) and telic (‘resurrect’, ‘return’) caused events pattern alike, showing imperfective morphology, when the main verb is imperfective ats’ele. The absence of perfective verbs in the complements to imperfective causative verbs recalls Ritter and Rosen’s (1993: 537, fn. 13) observation that perfectives do not appear under English causative verbs have or make. Perhaps this limitation reflects a real-world timing conflict between imperfective causing and perfective caused events. (See also footnote 17.)

7.5.2 Independent tense in the complement clause Welch (2015a,b) argues that IP (a projection for tense) dominates AspP in Tłı  ˛cho  ˛. Our examples in this section show that ha ‘future’ may occur in the complements to ats’ele, arguing that the complement to ats’ele is of category IP or larger. The sense of such sentences is that the caused event or situation is not realized immediately. The examples (36)–(39) feature caused states of affairs whose potential is in the future or whose realization is expected to take some time.1⁸

1⁸ Note that in (37) and (i) the causee controls not the grammatical subject of the complement but a non-subject. In (37), the controlled DP is the object of the postposition gıgha ‘for them’ and, in (i), the controllee is the possessive ‘your’ of naxıkw ˛o ̀ ‘your flesh’. (i) naxıkw ˛ ò nat’̀ a ̀ anaxegele ha hogeehdz ̀ a ̀ ho  ˛t’e [ naxıkw o˛̀ nat’̀ a ̀ ] anaxegele ha hogeehdz ̀ à ho  ˛t’e 2pl.flesh 3S.ipfv.be_cut.C 2plO.3plS.ipfv.cause fut 3plS.ipfv.try 3S.ipfv.be lit. ‘they try to make you so that your flesh is cut’ ‘[they] try to compel you to be circumcised’ (Galatians 6:12) The possibility for control of non-subjects has been documented in Tłı  ˛cho  ˛ (Saxon 1984, 1986), Slave (Rice 1989), other Dene languages, and in other languages of the world (such as Tagalog; Landau 2013, following Schachter 1976). This is an important property of control in these languages, but not a focus of this study. I speculate that this possibility may be related to the fact that CP in Tłı  ˛cho  ˛ may function as a predicate and thus that either subject or object, or another participant, may be controlled from the matrix clause.

200 leslie saxon (36)

Do  ˛zhıı̀ hoono ̀  ˛-daats’ o˛-n ̀ ake ̀ go ˛hch ı̀ ̀ ı,̀ wecheekeede ̀ e ̀ gı  ˛ı  ˛lı  ˛ ha ag o˛ ̀ o˛l̀ a.̀ Do  ˛zhıı̀ hoono ̀  ˛-daats’ o˛-n ̀ ake ̀ go ˛hch ı̀ ̀ ı,̀ [ wecheekeede ̀ è man ten-more_than-two 3plO.3S.pfv.choose 3.disciple gı  ˛ı  ˛lı  ˛ ha ] ag o˛ ̀ o˛l̀ a.̀ 3plS.ipfv.be fut.C 3plO.3S.pfv.cause lit. ‘he chose twelve men, he made them so that they would be his disciples’ ‘And he appointed twelve, whom he also named apostles, to be with him’ (Mark 3:14)

The embedded clause in (36) shows the future tense marker ha. I take the interpretation here to be that Jesus chose the disciples, but they will become his followers only later. Similar interpretations are suggested with the embedded future-marked events in (37)–(39): ‘they will believe’, ‘they will understand it easily’, and ‘we will live like him’. (37)

[…] gıgha ehkw’ı-ahodı ha agole goehdz ̀ a.̀ […] [ gıgha ehkw’ı-ahodı ha ] agole 3pl.for truly-be_spoken fut.C 3plO.3S.ipfv.cause goehdz ̀ a.̀ 3plO.3S.ipfv.try lit. ‘he was trying to make it be truly spoken for them [he was trying to make them believe]’ ‘[he was] trying to convince them […]’ (Acts 28:23)

(38)

N o˛htsı ̀  ˛ nahot ̀ ı-le ̀ gınıedı ̀ ha ag o˛ ̀ o˛l̀ a ̀ ne t’a.̀ N o˛htsı ̀  ˛ [ nahot ̀ ı-le ̀ gınıedı ̀ ha ] God hard-neg 3O.3plS.ipfv.understand fut.C ag o˛ ̀ o˛l̀ a ̀ ne t’a.̀ 3plO.3S.pfv.cause cop because lit. ‘because God has made them so that they will understand it easily’ ‘because God has shown it to them’ (Romans 1:19)

(39)

Ededı  ˛ hanı ̀ gogha daı  ˛ʔaà sıı̀ wek’ ˛e ˛̀e ̀ ts’eeda ha ag o˛ ̀ o˛l̀ a.̀ Ededı  ˛ hanı ̀ gogha daı  ˛ʔa-a ̀ sıı̀ 3 thus 1pl.for 3S.pfv.suffer-nmlz foc ts’eeda ha ] ag o˛ ̀ o˛l̀ a.̀ 1plS.ipfv.live fut.C 1plO.3S.pfv.cause

[

wek’ ˛e ˛̀e ̀ 3.like

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 201 lit. ‘he who suffered thus for us caused us so that we would live like him’ ‘Christ also suffered for you, leaving you an example, so that you would follow in his steps’ (1 Peter 2:21) In supporting delayed or potential causation as seen in these examples, ats’ele contrasts with English cause and make, which, in past tense, prohibit interpretations in which the caused event or situation has independent tense.1⁹ (40) a. We caused the sink to overflow. b. They made me sing. With these English examples, the causing event and the caused event are both in the past. The contrast between the two languages can be taken to be related to the selection properties of the causative verbs—and the fact that Tłı  ˛cho  ˛ does not have infinitives.2⁰ Common in the corpus are examples of the causative verb marked with future tense, as in (41) and (42), repeated from (13)–(14), and (43)–(44). (41)

sı  ˛ whatso  ˛ o˛̀ whıhda aseahłe ha [sı  ˛] [ whatso  ˛ o˛̀ whıhda ] aseahłe ha 1sg alone 1sgS.ipfv.be.C 1sgO.2plS.ipfv.cause fut lit. ‘you will make me be alone’ ‘you will leave me alone’

(42)

hazo  ˛ o˛̀ naxık o˛̀ ts’ o˛̀ naahde ̀ anaxedle ha [hazo  ˛ o˛]̀ [ naxık o˛̀ ts’ o˛̀ naahde ̀ all 2pl.house to 2plS.ipfv.return.pl.C anaxedle ha 2plO.ipfv.be_caused fut lit. ‘you will all be made to return to your homes’ ‘you will all be scattered, each one to his home’

(John 16:31) ]

(John 16:31)

1⁹ With English let my intuitions are not so clear: the example The manager let me sing tomorrow seems possibly grammatical to me, in contrast with the clearly ungrammatical *The manager made me sing tomorrow. 2⁰ The ‘past’ adverbial ˛l ı e ̀ is not found embedded under ats’ele in the corpus, further supporting Welch’s (2015a; 2015c) conclusion that it and the future particle ha are unalike in category and function.

202 leslie saxon (43)

[…] do  ˛ edek o˛̀ gıxe ̀ nahd ̀ e ̀ asegele ha, […] […] do  ˛ [ edek o˛̀ gıxe ̀ nahd ̀ è ] person refl.house 3pl.with 1sgS.ipfv.live.C asegele ha, […] 1sgO.3plS.ipfv.cause fut lit. ‘people will let me live with them in their homes’ ‘people may welcome me into their homes’ (Luke 16:4)

(44)

daǹ ıgho  ˛ nets’eʔ ˛ı ̀ agoı  ˛le ha? ̀  ˛ goxı daǹ ıgho ̀  ˛ [goxı  ˛] [nets’eʔ ˛] ı̀ agoı  ˛le ha? why 1pl 2sgO.1plS.ipfv.see.C 1plO.2sgS.ipfv.cause fut lit. ‘why will you let us see you?’ ‘how is it that you will reveal yourself to us?’ (John 14:22)

All of the examples observed in the corpus with a future causative verb feature a complement with imperfective aspect. Examples with the causative verb modified by the past tense adverbial appear in the corpus also. In (45), the past tense (plus perfective aspect) in the main clause seems to highlight the fact that the writer changed his ways, relegating the persecution to the past. (45)

N o˛htsı ̀  ˛ wecheekee ̀ dagı  ˛ı  ˛ʔaa ̀ agehł ̀ a ˛l ̀ ı e ̀ t’a ̀ [ N o˛htsı ̀  ˛ wecheekee ̀ ] [dagı  ˛ı  ˛ʔa-a]̀ agehł ̀ à God 3.follower 3plS.ipfv.suffer-C 3plO.1sgS.pfv.cause  ˛l ı e ̀ t’a ̀ past because lit. ‘because I caused God’s followers to suffer’ ‘because I persecuted the church of God’ (1 Corinthians 15:9)

In (46), the passage emphasizes the situation in the past regarding permission for divorce, in contrast with the time of speaking. (46)

Moses, naxıts’eke ̀ e ̀ ʔ o˛ahd ̀ e ̀ ha anax ˛ı ˛l ı̀ ̀ a ˛l ̀ ı e.̀ Moses, [ naxıts’eke ̀ e ̀ ʔ o˛ahd ̀ è ha ] anax ˛ı ˛l ı̀ ̀ a ̀ Moses 2pl.wife 2plS.ipfv.divorce fut.C 2plO.3S.pfv.cause  ˛l ı e.̀ past lit. ‘Moses allowed you that you will divorce your wives.’ ‘Moses allowed you to divorce your wives’ (Matthew 19:8)

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 203 The use of the future in the complement highlights the potential for divorce.21 This example has explicit expression of independent tenses in matrix and complement clauses.

7.5.3 Independent negation in the complement clause NegP in Tłı  ˛cho  ˛ dominates IP (Welch 2015b); see (23) above. Since the complement to the causative verb can be negative, I take this to show that the complement can be at least as big as NegP. The examples from above repeated here show that the causative clause and the embedded clause may be independently negated. In (47)–(48), repeated from (25) and (29), the causative clause is negative. (47)

Yedayeh ̀ Nezı  ˛ı  ˛ Asıa nek’ ̀ e godı nezı  ˛ı  ˛ do  ˛ ts’ o˛̀ agı  ˛ ˛l ı ̀ a ̀ ag o˛ ̀ o˛l̀ a-le. ̀ Yedayeh ̀ Nezı  ˛ı  ˛ [ Asıa nek’ ̀ e godı nezı  ˛ı  ˛ do  ˛ Holy Spirit Asia country story good person agı  ˛ ˛l ı ̀ à ] ag o˛ ̀ o˛l̀ a-le. ̀ 3O.3plS.pfv.cause.C 3plO.3S.pfv.cause-neg

ts’ o˛̀ to

lit. ‘the Holy Spirit didn’t let them take the gospel to the people in Asia’ ‘[they had] been forbidden by the Holy Spirit to speak the word in Asia’ (Acts 16:6) (48)

̀ Laban wets’ ˛o tıts’aad̀ ıı̀ ta whelaa ̀ ay ˛ı ˛l ı̀ ̀ a-le. [ Laban wets’ ˛o tıts’aad̀ ıı̀ ta Laban 3.from animal among ay ˛ı ˛l ı̀ ̀ a-le. ̀ 3O.3S.pfv.cause-neg

whela-a ̀ 3S.ipfv.be-C

lit. ‘he didn’t let them be among Laban’s animals’ ‘[He] did not put them with Laban’s flock’

]

(Genesis 30:40)

21 On the meaning of ha ‘future’ see Anisman (2017). She argues contra Welch (2015b) for the modal status of the so-called future in Tłı  ˛cho  ˛, a position supported also by this example and others in this section. Although she disagrees with Welch on the semantic nature of this morpheme, she does not counter his proposed hierarchical structure.

204 leslie saxon In (49)–(50), repeated from (26) and (35), the embedded clause is negated. (49)

(50)

t’asawee  ˛la-le ̀ anehł ̀ à [t’asawee  ˛la-le] ̀ anehł ̀ à 3O.2sgS.pfv.harm-neg.C 2sgO.1sgS.pfv.cause lit. ‘I made you not harm her’ ‘I didn’t let you touch her’

(Genesis 20:6)

“[…] Lazarus ełaawı-le ̀ ayele ha ˛l ı e,̀ ” gedı. “[…] [Lazarus] [ełaawı-le] ̀ ayele ha  ˛l ı e,̀ ” Lazarus 3S.ipfv.die-neg.C 3O.3S.ipfv.cause fut past gedı. 3pl.ipfv.say lit. ‘They said, “He should have caused Lazarus not to die.”’ ‘[they] said, “Could not he […] have kept [Lazarus] from dying?”’ (John 11:37)

There is more to say about how these causatives are interpreted, but I put that question aside to continue my focus on the syntactic properties of ats’ele.22

7.5.4 Ats’ele selects a clause, not a vP or VP Consider the syntactic model of Tłı  ˛cho  ˛ clause structure under consideration, which is repeated from (23) in (51).

22 The verb ats’ele in combination with a direct object and the postposition ch’à ‘against’ and its clausal complement translates as ‘prevent’. I hope to investigate the patterns with this collocation in future work. (i) Amıı̀ […] naowo ̀ ehkw’ıı k’eaht’ ̀ e ch’a ̀ anax ˛ ˛l ıı̀ ̀ a?̀ Amıı̀ [ […] naowo ̀ ehkw’ıı k’eaht’ ̀ e ] ch’a ̀ ] anax ˛ ˛l ıı̀ ̀ a?̀ who law true 2plS.ipfv.obey against 2plO.3S.pfv.cause lit. ‘who prevented you from obeying the true law?’ ‘who prevented you from obeying the truth?’ (Galatians 5:7)

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 205 (51)

Tłı  ˛cho  ˛ clause structure CP C

FocP Foc

NegP Neg

ModP Mod

IP AspP

I Asp

vP VP

v

(Welch 2015b)

We have observed that the complement of ats’ele ‘cause’ is independent of the matrix clause in aspect, tense, and polarity. This finding leads to the conclusion that ats’ele indeed selects a CP with an adverbial head, and not VP or vP or another type of intermediate category. Unexplained by this conclusion is the impossibility of modal and other categories of an expanded CP in the causative complement. Welch (2015b) identifies a projection to house a class of modals outside of IP; Saxon (2014) situates other particles as the heads of categories within an expanded CP domain as in (52), following ideas of Cheng (1997); Rizzi (1997, 2001); Speas (2004); Ceong (2011, 2019); Ceong and Saxon (2013). (52)

Tłı  ˛cho  ˛ expanded CP domain ForceP CP TypeP IP

Type

Force C

(Saxon 2014, after Ceong and Saxon 2013)

206 leslie saxon The particles instantiating each of these additional categories of Welch (2015b) and Saxon (2014) are presented in Table 7.2 (to be revised). (Saxon 2014 does not include counterparts to Welch’s ModP, NegP, and FocP in her treatment.) It will be possible to clarify the relation between ModP and the categories of Saxon (2014) through further research. In the meantime, let’s pursue one hypothesis. The adverbial suffix takes scope over negation; see examples (49)– (50). The same is true of the yes-no interrogative marker as ˛ı ˛: ı̀ ̀ both positive and negative yes-no questions are found in Tłı  ˛cho  ˛. Because as ˛ı ˛ı̀ ̀ is incompatible with the adverbial suffix, they are potentially both expressions of the category TypeP. I therefore revise Table 7.2 as in Table 7.3. Adopting this model, we analyse causative ats’ele as selecting TypeP, specifically the adverbial suffix. Under this analysis none of the functional items of Table 7.2 Categories of the Tłı  ˛cho  ˛ extended CP (to be revised in Table 7.3) Item

Meaning

Saxon (2014)

nı ̀ s ˛o ˛̀ò so  ˛ nı no  ˛ ˛ò ho  ˛ t’e -le welı ̀ wele ̀ as ˛ ˛ıı̀ ̀ ha

focus question (matrix scope only) warning; prohibitive uncertainty evidential clausal focus negative epistemic modal: asserting possibilities jussive modal yes–no interrogation future

ForceP }

CP:Evaluation CP:Epistemic — —

}

Category Welch (2015b) ⎫ ⎬ ⎭

CP FocP NegP



ModP

TypeP IP (assumed)

— IP

Table 7.3 Revised categories of the Tłı  ˛cho  ˛ extended CP Item

Meaning

Category

nı ̀ s ˛o ˛̀ò so  ˛ nı no  ˛ ˛ò ho  ˛ t’e as ˛ ˛ıı̀ ̀ -V̀ -le welı ̀ wele ̀ ha

focus question (matrix scope only) warning; prohibitive uncertainty evidential’ clausal focus yes–no interrogation adverbial suffix negative epistemic modal: asserting possibilities jussive modal future

ForceP }

CP:Evaluation CP:Epistemic FocP

}

TypeP NegP

}

ModP IP

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 207 higher categories in Table 7.3 will co-occur with the adverbial suffix, which is our finding from study of the biblical sources.23 Further exploring the notion of clause typing, I note that elements of the causative complement in Tłı  ˛cho  ˛ can be questioned. In examples (53)–(55), due to Rosa Mantla, the matrix causative verb and question phrase are shown in bold.2⁴ (53)

Sechı behtsı  ˛ ˛ı ̀ daot’ı ̀ ̀  ˛ k’eged ̀ ıı̀ ̀ agogele ha? Sechı [ behtsı  ˛ ˛ı ̀ daot’ı ̀ ̀  ˛ k’eged ̀ ı-̀ ı ̀ ] 1sg.younger_brother vehicle what_kind 3plS.ipfv.keep-C ha? agogele 3plO.3plS.ipfv.cause fut ‘What kind of vehicles are they going to make my brothers drive?’ (Rosa Mantla, p.c.)

(54)

Joanne amıı̀ gha edzı  ˛ ˛ı ̀ aneela?̀  ˛- ˛ı ̀ ] aneela?̀ Joanne [ amıı̀ gha edzı Joanne who for 3S.ipfv.sing-C 2sgS.pfv.cause ‘Who did you make Joanne sing for?’

(Rosa Mantla, p.c.)

23 There is a significant question that arises from Table 7.3. In the biblical sources, elements of TypeP do not co-occur in the same clause with elements of ModP, their complementarity suggesting that all four belong to a single category. However, we see from Table 7.3 that NegP intervenes between TypeP and ModP. Keeping with the hypothesis expressed in Table 7.3, we will be led to say that the incompatibility of TypeP and ModP is due to semantic rather than syntactic factors, plausible given the senses of these terms. There is additional diachronic motivation for distinguishing ModP from TypeP: the Mod heads welı ̀ and wele ̀ derive from an optative form of the copula ts’ı  ˛lı  ˛ ‘be’ (Welch 2015b: 270), but the diachronic sources of the proposed elements of TypeP are quite separate. 2⁴ I made the case in §7.4 that the adverbial clause is subcategorized by ats’ele, and therefore an argument. Its argument status is likely not relevant to the possibility for questioning, however, since adjuncts are not islands for wh-phrases in situ in Tłı  ˛cho  ˛. Example (i) in this note is a direct question: ayıı̀ ‘what’ appears in a conditional clause (bracketed), which is an adjunct. (See Anisman 2019 on conditionals in Tłı  ˛cho  ˛.) (i) ayıı̀ dahł ̀ a ̀ nı  ˛de ̀ welo  ˛ whıle ̀ ts’ ˛ ò ehdaa ̀ asedle ha? [ ayıı̀ dahł ̀ à nı  ˛de ̀ ] welo  ˛ whıle ̀ ts’ ˛ ò ehda-a ̀ what Q.1sgS.pfv.do if 3.end 3.be_missing to 1sgS.ipfv.live-C asedle ha? 1sgO.ipfv.be_caused fut lit. ‘if I do what will I be made to live until there is no end?’ ‘what must I do to inherit eternal life?’ (Mark 10:17)

208 leslie saxon (55)

Neba amıı̀ k’exeetsıı ̀ agı  ˛ ˛l ı ̀ a?̀ Neba [ amıı̀ 2sg.older_sister who agı  ˛ ˛l ı ̀ a?̀ 3O.3plS.pfv.cause

k’exeetsı-ı ̀ 3sgS.ipfv.carry_by_pack-C

‘Who did they make your sister pack around?’

]

(Rosa Mantla, p.c.)

The morphology of the verb or postposition in the complement clause, namely lack of overt agreement with the questioned DP, ensures that the question phrases are in situ, despite the fact that these are direct questions with the question phrase having matrix scope (see Saxon 1986, 1998 for discussion).2⁵ A search of the biblical sources for all question words reveals no examples of direct questions like these. However, nominalizations of causative clauses with embedded in situ question phrases appear in the corpus; see (56)–(58). In these examples the question phrase and causative verb are bolded, and the relevant part of the translation is underlined. (56)

Amıı̀ do  ˛ ts’adıı ̀ sıı̀ N o˛htsı ̀  ˛ ayıı̀ t’a ̀ natso ̀ ayı  ˛ ˛l ı ̀ aà sıı̀ t’a-at’ı ̀  ˛ ha ho  ˛t’e. Amıı̀ do  ˛ ts’adı-ı ̀ sıı̀ [ N o˛htsı ̀  ˛ [ ayıı̀ t’a ̀ who person 3S.ipfv.help-nmlz foc God what with natso ̀ ]CP ayı  ˛ ˛l ı ̀ a-a ̀ sıı̀ ]DP t’a-at’ı ̀  ˛ 3S.ipfv.strong.C 4O.3S.pfv.cause-nmlz foc 3S.ipfv.use ha ho  ˛t’e. fut area.3S.ipfv.be lit. ‘whoever helps people must use what God makes themsg strong with’ ‘whoever serves must do so with the strength that God supplies’ (1 Peter 4:11)

In this example, both the subject and object of wet’a-ats’ ̀ et’ı  ˛ ‘use’ are ‘free relative clauses’, headed by the question phrases amıı̀ ‘who’ and ayıı̀ ‘what’ 2⁵ In situ questioning is the usual pattern in Tłı  ˛cho  ˛ (Saxon 1979). Adjuncts embedded in the complement clause of the periphrastic causative can be questioned, as we see with the direct question in (i) asking ‘where?’. (i) Edı  ˛ ˛ı ̀ chekaa geetee ̀ agıts’ele ha? Edı  ˛ ˛ı ̀ chekaa geete-e ̀ agıts’ele ha? where child 3plS.ipfv.sleep-C 3plO.1plS.ipfv.cause fut ‘Where are we going to let the kids sleep?’ (Rosa Mantla, p.c.)

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 209 respectively. The object translates literally as ‘what God makes themsg strong with’: the question phrase originates in the complement of the causative verb. In the next example the free relative, an appositive DP, translates literally as ‘what you will make God’s followers’ minds strong with’. (57)

N o˛htsı ̀  ˛ wecheekee ̀ ayıı̀ t’a ̀ gını ̀ natso ̀ ò aahłe ha sıı̀ eyı gha naho ̀  ˛hoahde.̀ [ [ N o˛htsı ̀  ˛ wecheekee ̀ ayıı̀ t’a ̀ gını ̀ natsò ò God 3.follower what with 3pl.mind 3S.ipfv.strong-C ]CP aahłe ha sıı̀ ]DP eyı gha naho ̀  ˛hoahde.̀ 2plS.ipfv.cause fut foc that for 2plS.ipfv.strive lit. ‘strive for that, what you will make God’s followers’ minds strong with’ ‘strive to excel in [spiritual gifts] for building up the church’ (1 Corinthians 14:12)

In the free relative in (58), an appositive again, ayıı̀ ‘what’ is the object of ‘know’ within the complement of ats’ele ‘cause’. (58)

ayıı̀ k’eehso ̀  ˛ o˛̀ as ˛ı ˛l ı̀ ̀ aà sıı̀ wegho  ˛ gohde ha. [ [ ayıı̀ k’eehso ̀  ˛- o˛̀ ]CP as ˛ı ˛l ı̀ ̀ a-a ̀ sıı̀ what 1sgS.ipfv.know-C 1sgO.3S.pfv.cause-nmlz foc ]DP wegho  ˛ gohde ha. 3.about 1sgS.ipfv.talk fut lit. ‘what he made me know, I will talk about it’ ‘I will [boast about] … revelations [of the Lord]’ (2 Corinthians 12:1)

See also in (33) the translation for ‘revelation’, literally ‘what Jesus Christ made his disciples look at’. The translations of these sentences show that the scope of the question phrase is not the causative complement but always the higher domain. Concerning TypeP, assuming that there is a wh-interrogative clause Type, the adverbial suffix is as incompatible with a wh-operator as it is with the yes– no operator. Although it goes beyond the scope of this chapter to fully explore the analysis of this type of example, they have not been observed before, to my knowledge, and so I include them for future study.

210 leslie saxon

7.6 Conclusion: The causative complement is an adverbial clause I have shown in this work that the periphrastic causative in Tłı  ˛cho  ˛ involves a clausal complement, using evidence that the complement clause is syntactically and semantically independent of the main clause in terms of aspect, tense, and polarity. This situation shows notable contrasts with what is found in languages in which the causative complement is a type of verb phrase. My point is emphasizing the adverbial form of the complement is to offer a different kind of support for the findings of Moulton (2009, 2015), BogalAllbritten (2016), Kratzer (2016), and others, that embedded clauses serve as predicates, showing properties therefore of modifiers. I can point to causatives in other languages with similar qualities: Davison (2013) and Butt (2013) analyse oblique infinitives in Hindi-Urdu causatives; Junko Shimoyama (p.c.; see footnote 11) notes adverbial complements to a class of causatives in Japanese; Hailey Hyekyeong Ceong (p.c.) and an anonymous reviewer observe that the clausal complements in Korean periphrastic causatives are marked by adverbializers (see Yoon and Brown 2011 for a description). Like the well-described Dene nominalizing suffix in its own right, the Tłı  ˛cho  ˛ adverbializer under study in this chapter can be traced back to the proto-language as *qu’.2⁶ Cognates of the Tłı  ˛cho  ˛ adverbializer in four other Dene languages are exemplified in causative constructions in (59)–(62). In these examples the causative verb (also cognate with the Tłı  ˛cho  ˛ verb) and the complementizer are shown in bold. (59)

Mary yek’agoli gú ʔayí  ˛lá [Dene (Slavey)] [ Mary yek’agoli gú ] ʔayí  ˛lá Mary 3.taste.4 comp 3.caused.4 ‘He let Mary taste it’ (Fort Good Hope dialect; Rice 1989: 1304)

(60)

k’āda ̀ tínáts’ínáh-gù ágūlāh k’āda ̀ tínáts’ínáh-gù ágūlāh back one.moves_camp-comp you.make.them ‘Make them move their camps back!’

[Tsúut’ín ̀ a]̀

(Cook 1984: 117)

In Dene (Slavey), Tsúut’ín ̀ a,̀ and Tsilhqot’in, the complementizer has the same basic syllabic and tonal qualities as in the proto-language. Note that 2⁶ In other work I have proposed *Gu’, but the existence of reflexes in daughter languages with initial q suggests *qu’.

the tłı  ˛cho   ˛ syntactic causative and non-nominal cps 211 Cook (2013) represents it orthographically in Tsilhqot’in as a proclitic on the causative verb. (61)

yaŵestɨg qé-ʔasinlágh [yaŵestɨg] qé=ʔasinlágh 1.opt.talk for=she.made.me ‘She made me talk.’

[Tsilhqot’in]

(Cook 2013: 428)

In Tse’khene, as in Tłı  ˛cho  ˛, the segmental forms of the complementizer have given way to its realization as marked tone (L in the gloss stands for a low tone morpheme). (62)

Ditsııgh ̀ ̀ a’̀ dut’elal̀ ah. ̀ Ditsııgh ̀ ̀ a’̀ dut’el L ulah. ̀ her.hair red comp she.made ‘She dyed her hair red’

[Tse’khene]

(Hargus 2009)

The periphrastic causatives in all of these languages show identical patterns, with cognate complementizer and verb. In addition, a simple adverbializing function for the ‘adverbial complementizer’ occurs in all of the Dene languages in which the morpheme is attested, including Ahtna, Dakelh, Dënesųłiné, Hupa, Koyukon, Mattole, Navajo, Tsúut’ín ̀ a,̀ and Western Apache (Saxon 2003a,b). Although Moulton (2009, 2015) does not consider causative complements as a locus for non-nominal CPs, I believe that the facts in Tłı  ˛cho  ˛ make a compelling case. In this light, let us consider again the three-way contrast with which I opened this chapter. (63) a. nezı  ˛ nezı  ˛ 3S.ipfv.be_good ‘It is good’ b. nezı  ˛ı  ˛ nezı  ˛-ı  ˛ 3S.ipfv.be_good-nmlz ‘(the) one that is good’ c. nezı  ˛ ˛ı ̀ nezı  ˛- ˛ı ̀ 3S.ipfv.be_good-C ‘in a good manner; well’

212 leslie saxon In future work I would like to thoroughly investigate the semantic type of each. I will make the initial hypothesis that the sentence in (63a) represents a proposition: this expression may be true or false. The nominalization pattern in (63b) is convincingly argued to be a semantic argument for the neighbouring Dënesu  ˛łiné language (Wilhelm 2014), and I will take this as my initial hypothesis for Tłı  ˛cho  ˛ too. My initial findings suggest the hypothesis that the adverbial in (63c) represents the semantic type predicate. This semantic type is expressed by VP in many languages, but in a language like Tłı  ˛cho  ˛ where there is no realization of a VP separate from clausal inflection, semantic predicates are likely to be expressed as clauses. In this language, there is a distinct complementizer serving to identify such clauses.

Acknowledgements I acknowledge and thank Mary Siemens for her expertise and friendship, and for pointing out to me what I am calling the adverbial complementizer. Thank you to my colleague and friend Rosa Mantla for her many contributions to my understanding. For their contributions to the present chapter, many thanks to Elizabeth Cowper, Hailey Hyekyeong Ceong, and two anonymous reviewers, as well as Adar Anisman, Sonya Bird, Heather Bliss, Elizabeth Bogal-Allbritten, Bridget Copley, Alana Johns, John Lyon, Rosa Mantla, Éric Mathieu, Keren Rice, Junko Shimoyama, Guillaume Thomas, Suzanne Urbanczyk, Nicholas Welch, Martina Wiltschko, and the participants and organizers of the University of Toronto Linguistics Department’s Dog Days Summer Workshop 2015 and a meeting of the Research in Indigenous Languages and Linguistics network in Victoria, May 2017. I extend my profound thanks to the many people who have patiently taught me, an outsider, about the Tłı  ˛cho  ˛ language and culture over the past decades since my introduction to it by the late Vital Thomas and Elizabeth Mackenzie, and Alice Mackenzie and Therese Mantla. Since starting my experience in the NWT as a research assistant to Keren Rice, I have learned from Keren and benefited from her knowledge and advice. Thank you! I extend my deep gratitude to Elizabeth Cowper, who directed my master’s work on Tłı  ˛cho  ˛ syntax (Saxon 1979).

OUP CORRECTED PROOF – FINAL, 28/8/2020, SPi

PART III

ARC HI TE C T U R A L QU E ST ION S

8 Against some approaches to long-distance agreement without Agree Carson T. Schütze

8.1 Introduction 8.1.1 Motivation Within Chomsky’s work under the heading of the Minimalist Program there have been two proposals for the structural configuration under which a relationship of agreement can be established: in early Minimalism that configuration was the Spec(ifier)–Head configuration; beginning with Chomsky (2000), that was abandoned in favour of Agree, arguably for empirical reasons: instances of long-distance agreement seem to demand it.1 But Agree is obviously less constrained than Spec–Head, allowing the distance between probe and goal to be unbounded as long as no strong phase boundary intervenes,2 so it is worth asking whether this weakening of the theory is unavoidable. Since its introduction, several authors have attempted to argue that long-distance agreement can be captured without appeal to Agree. Whether that ultimately turns out to be possible will obviously depend, among other things, on exactly what other syntactic mechanisms the theory makes available and what the constraints thereon are. These are matters of ongoing development and debate, so I do not believe the question can be definitively settled at this point. What is currently both possible and worthwhile, I suggest,

1 Like most ‘new’ ideas in syntax, Agree had long been anticipated in the literature. For instance, Sigurðsson (1990–1) proposed that ‘long-distance’ Case assignment to nom objects in Icelandic is accomplished by ‘chain-government’, whereby in (i) Infl governs have which governs been which governs sold which governs some boats. Cf. also Raposo and Uriagereka (1990). (i) Það hafa verið seldir nokkrir bátar. [Icelandic] there have.3pl been sold some boats.nom.pl ‘There have been some boats sold.’ 2 On the question of whether this is a problem for the Phase-Impenetrability Condition and what might be done about it, see Richards (2012). See also §8.5.4. Carson T. Schütze, Against some approaches to long-distance agreement without AGREE In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Carson T. Schütze. DOI: 10.1093/oso/9780198817925.003.0008

216 carson t. schütze is to scrutinize the details of the specific proposals of this type that have been made, in order to assess whether they are empirically viable and theoretically desirable.

8.1.2 Definitions and debates I take the Spec–Head relation to require no further elaboration. As for Agree, the definition in (1) captures the relevant properties introduced by Chomsky (2000): (1)

Agree (α, β) can be established, where α is a probe and β is a goal, if the uninterpretable features of α and the interpretable features of β ‘match’ (i.e. are nondistinct: ‘same feature, independently of value’), α c-commands β, there is no closer matching goal γ (unless β and γ are equidistant), and β is ‘active’ (i.e. has an (undeleted) uninterpretable feature).

Subsequent to Chomsky’s introduction of Agree, some authors, e.g. Béjar and Rezac (2009), have suggested that under some circumstances the features of a probe could find themselves on (what we would traditionally call) an X-bar, whence they could probe Spec-XP under c-command as a valid instance of Agree. However, Chomsky himself has repeatedly excluded this possibility explicitly, as demonstrated in the following passages: Continuing with Agree, (52c) requires that the matched goal G must be easily located. We want to identify a domain D(P) of the probe P, such that G is within D(P). There are two candidates for D(P): the smallest and the largest of the categories labeled by the label containing P. The former includes only the complement of P; the latter its specifiers as well. Search space is more limited if D(P) is the smallest category, as so far assumed […]. [T]here is substantial evidence that G must be in the complement of the probe P, not its specifiers. (Chomsky 2000: 135) Apparent SPEC–H relations are in reality head–head relations involving minimal search (local c-command). […] Call H a probe P, which seeks a goal G within XP; P = H c-commands G […]. [T]here is a relation Agree holding between probe P and goal G, which deletes uninterpretable features if P and G are appropriately related. […] [T]he SPEC–head relation does not exist. (Chomsky 2004: 113–14)

against long-distance agreement without agree 217 It is impossible […] for an agreement relation to be established between [a probe] W and an element within its SPEC. (Chomsky 2007: 9) For minimal computation, the probe should search the smallest domain to find the goal: its c-command domain. It follows that there should be no mcommand, hence no Spec-head relations. (Chomsky 2008: 146)

Minimalist textbooks have likewise been unequivocal on this point: ‘A probe is a head with [−interpretable] features […]. [A] given probe peruses its c-command domain in search of a goal’ (Hornstein, Nunes, and Grohmann 2005: 317); ‘Agree […] hold[s] between (elements in) a head and some other element down its c-command path’ (Lasnik, Uriagereka, and Boeckx 2005: 233). Therefore, throughout this chapter, I will follow the original definition of Agree, under which it and Spec–Head are mutually exclusive.3 The purpose of this chapter is most definitely not to endeavour to settle any of the active debates that the introduction of Agree and subsequent attempts to apply and refine it have prompted, including but not limited to the question of whether agreement in natural language actually operates upwards, downwards, in either direction (as a parametric choice), in both directions (depending on the feature/phenomenon involved, or simultaneously), etc., etc. These debates involve a wide range of empirical phenomena spanning a large number of typologically diverse languages and often hinge on highly technical assumptions about how the mechanical operations of the syntax should be defined and implemented—see, for example, Baker (2008); Wurmbrand (2014); Preminger and Polinsky (2015); Schneider-Zioga (2017); Bjorkman and Zeijlstra (2019), among many others. This chapter will have none of those properties. Its goals are narrower and, in a sense, more concrete. It will accept, for the sake of discussion, the conceptual point that Agree is a relatively powerful, relatively unconstrained mechanism (by comparison with Spec–Head) the need for which ought therefore to be challenged. It will then examine in depth some of the most carefully worked-out proposals in the literature for re-analysing phenomena that motivated the introduction of Agree and ask whether those proposals are successful.

3 It is worth noting that there are empirical results in the literature that have relied crucially on this narrow definition, for example, McCloskey’s (2002) analysis of the interactions between A-bar extraction, resumption, and complementizers in Irish: ‘This [incorrect outcome], however, will not be possible if we maintain that the agreement relation can only hold between a head and an element within its c-command domain’ (McCloskey 2002: 204).

218 carson t. schütze

8.1.3 Empirical domains: Long-distance agreement I focus on two phenomena where Agree has frequently been employed: English existential (and unaccusative) there-constructions (2), and Icelandic long-distance number agreement in the presence of a quirky subject (3); in both cases, the number of intervening nonfinite clauses between the Probe (the agreeing head) and the Goal (the agreement trigger) is in principle unbounded (Sigurðsson 2000). For perspicuity I will identify the Probe and Goal with single and double underscoring, respectively. (2) a. There seem3pl [to have been many horses in the stable].⁴ b. There seem3pl [to have been believed [to be many horses in the stable]]. (3) a. Jóni virðast [hafa likað þessir sokkar].⁵ John.dat seem.3pl to.have liked these socks.nom.pl ‘John seems to have liked these socks.’ b. Jóni virðast [vera taldir [líka hestarnir ]]. John.dat seem.3pl to.be thought to.like the.horses.nom.pl ‘John seems to be thought to like the horses.’ I will structure the discussion around one prominent analysis of each of these phenomena, in order to have as much concrete detail to work with as possible: Moro’s analysis of existentials (§8.3) and Koopman’s analysis of Icelandic (§8.4). These choices are not arbitrary: recent work arguing against Chomsky’s version of Agree has appealed to them as providing viable counter-analyses. For example, Zeijlstra (2012) cites Moro (1997) and Koopman (2006) with

⁴ In the survey reported by Koopman (2004), fifteen out of seventeen English speakers preferred a plural over a singular verb in such long-distance environments without an experiencer. When a singular experiencer intervenes (e.g. There seem(s) to Bill to be three men in the room, considerable variation emerged, with some speakers preferring the singular, some the plural, and some expressing no preference. Thus, while more data collection is certainly warranted, Boeckx (2008: 36) is almost certainly wrong when he presents as ‘facts which […] have gone unnoticed in the literature’ that the plural form is ‘∗ ?’ and the singular receives no mark. ⁵ Sigurðsson (1990–1, 1996) reports considerable interspeaker variation as to whether agreement here is preferred, optional, or dispreferred, versus default (3sg) inflection. Obviously what is relevant here is that agreement is possible. He notes that cross-clausal agreement is ‘weaker’ than clause-internal agreement with nom objects, as in (33) below, which is obligatory for most speakers with most verbs. (For the most part, first- and second-person nom objects cannot trigger any agreement, but see the references in this note for exceptions.)

against long-distance agreement without agree 219 approval.⁶, ⁷ (For arguments that Agree is genuinely necessary for some other long-distance agreement phenomena, see Boeckx 2009, among others.) Before delving into those analyses, I suggest a way of classifying approaches to longdistance agreement in general (§8.2). Afterwards I turn to consideration of a more general claim by Hornstein that counter-analyses avoiding Agree are always guaranteed to be available as a matter of principle (§8.5). I conclude with a brief discussion of some issues concerning variation (§8.6).

8.2 Approaches to long-distance agreement While I do not attempt an exhaustive survey, I would like to suggest that the myriad approaches to the problem of long-distance agreement can be taxonomized into four broad categories, characterized in terms of the Agree nomenclature.⁸ A) Contrary to appearances, the probe and the goal were actually local, at some (non-surface) stage of the derivation. This is what I call the ‘clandestine meeting’ approach. B) The probe and the goal were never local, but some element agrees locally with the goal, thereby acquiring its features, then moves to become local to the probe and agrees with it, thereby offloading its/the goal’s features onto the probe. This is what I call the ‘secret courier’ approach. C) One cannot establish that A or B has happened, so one invents a sui generis technology that connects the probe’s and goal’s features by fiat. This is what I call the ‘teleportation’ approach. D) One bites the bullet and admits that the probe and the goal can be far apart, hence Agree. (This may be regarded as less magical/stipulative than ‘teleportation’ to the extent that Agree captures phenomena beyond long-distance agreement sensu stricto.)

⁶ Zeijlstra (2012) sketches a variation of Koopman’s analysis (cf. §8.4) for one Icelandic example in which he proposes to do away with Koopman’s null expletive but retains, inter alia, the two TPs. ⁷ Zeijlstra (2012: 529) is in error when he states that ‘after Schütze (2003), Koopman assumes that both T heads are “compressed” into one finite verb establishing both singular and plural agreement, which is realized as plural agreement.’ Schütze (2003) crucially did not posit two T heads, but rather that the single T head of the clause attempts to agree both with the subject (which fails when the subject is quirky, yielding default 3sg features) and with the object (when it is nom), with results depending on the relationship between the phonological forms that would express those two sets of phi-features. ⁸ An excellent historical tour of approaches to expletive constructions up through its time of writing can be found in Hartmann (2008).

220 carson t. schütze Some examples from the literature on English there-constructions should illustrate what I have in mind with this classification. First, imagine a transformational rule—an example of the ‘clandestine meeting’ approach: (4b) is derived from (4a) by rightward movement of the bracketed associate, followed by either insertion of there (Burt 1971; Milsark 1974) or pronunciation of the trace of the associate as there (Kuno 1971; Dresher and Hornstein 1979); cf. also Harris (1957); Stowell (1978). (4) a. [Many horses] seem to have been in the stable. ⇓ b. There seem to have been [many horses] in the stable. Second, consider appeals to linking, coindexing, or a chain⁹—these are instances of the ‘teleportation’ approach: there must be in a particular (representational) relationship with the associate, but the properties of this relationship are not identical to other instances of that relationship (e.g. in (5), there does not refer to the same entity as the associate, despite being coindexed with it). (5)

Therei seem to have been [many horses]i in the stable.

Third, consider Expletive Replacement—another instance of the ‘clandestine meeting’ approach: this operation used movement to capture the locality of the expletive–associate relationship, but was quickly rejected because it wrongly predicted possible high scope for the associate.1⁰ Fourth, consider Expletive Adjunction/Affixation—another instance of the ‘teleportation’ approach: the unique LF configuration (6) is assumed to result in phi-feature sharing/matching without high scope-taking; but one must wonder if the notion ‘LF affix’ (Chomsky 1991) is even coherent. (6)

[[Many horses]i [there]] seem to have been ti in the stable.

Fifth, consider a novel application of Affix Hopping—another ‘teleportation’ solution: Bošković (1997) also adopts the proposal that expletive there is an LF ⁹ As in Chomsky (1986). 1⁰ That is, (i) does not have a reading in which many men scopes over negation, whereas (ii) does. (i) There are not many men in the garden. (ii) Many men are not in the garden.

against long-distance agreement without agree 221 affix, but rather than the associate adjoining to it, it adjoins to the associate via LF lowering ‘affix hopping’.11 There agrees in a Spec–Head configuration with Infl, then it covertly hops down to affix to the associate, and this affixation requires the affix’s number value to match that of its host (for reasons unclear). Sixth, consider Bobaljik’s (2002) single output approach: the associate moves to subject position, where it triggers agreement locally, but for both phonology and semantics it is the lower copy of the associate that is interpreted (see also Preminger 2009); this has superficial attributes of a ‘clandestine meeting’ approach, but when the broader implications are considered I believe it deserves to be seen as a case of ‘teleportation’. (See §8.5 for further discussion.) Seventh and finally, consider the family of there-raising proposals: one of these, predicate inversion, will occupy us in §8.3. Others have been put forward by, inter alios, Richards and Biberauer (2005), Deal (2009), and Alexiadou and Schäfer (2010) (though the latter do not deal with existentials); they involve there starting out low in the structure, but they employ Agree, biting the bullet. By contrast, Kayne’s (2008) there-raising analysis is a ‘secret courier’ approach, because what actually raises is a structure containing a trace of the associate, and hence the relevant features, which then agrees locally with Infl.

8.3 Predicate inversion Moro (1997, 2006, 2017) proposes a ‘secret courier’ solution: there is the predicate (pred) of a small clause (SC) and originates in a local relationship with the goal, the subject of predication, which it agrees with (conceivably under Spec–Head),12 then there undergoes predicate inversion to get to subject position, where it triggers agreement with the probe (under Spec–Head), as in (7): (7)

[IP [IP There areV [VP tV [SC [many copies of the book] tthere ]]] [in the studio]].

11 Since this version of affix hopping, unlike its traditional namesake, can cross an unbounded amount of intervening material, ‘affix long-jumping’ seems a more appropriate moniker. 12 Moro (2006: fn. 22) himself notes that getting there to agree (in number) with its ‘subject’ is not trivial, since a predicate need not match its subject in number: (i) [The childrenpl ]i are [SC ti the problemsg ]. Perhaps this issue could be circumvented by positing that a predicate must match its subject when its own number is unspecified, as there’s might be.

222 carson t. schütze The purported appeal of this approach is that the D-structure position of there and its raising to subject position are supposed to be independently motivated on the basis of inverse copular sentences, as in (8): (8)

[IP [The cause of the riot]pred wasV [VP tV [SC [a picture of the wall] tpred ]]].

But does that parallel empirically pan out? The following four subsections pursue that question.

8.3.1 Extraction: Contrasts not as expected Five sets of extraction contrasts raise problems for the predicate inversion approach. First, extraction of the notional subject is possible in canonical copular sentences, in contrast with inverse copular sentences,13 suggesting predicate inversion is responsible for the badness of (9b): (9) a. b.



What provision do you believe t was the objection to the bill? What provision do you believe the objection to the bill was t?

Moro claims that existentials pattern with inverse copular sentences, giving the following ungrammatical examples: (10) a. b.

∗ ∗

Which wall do you think there was t? Which girls do you think that there are t in the garden?

From these, Moro (2006: 213) concludes that ‘there can be no wh-movement of the [associate] if the [associate] is headed by which.’ However, the cases in (10) do not appear to generalize; if we use D-linked sets that are easier to accommodate, which/what-NP extractions are fine:1⁴

13 As Moro notes, this restriction holds only for D-linked wh-phrases. (i) is fine, as is the existential (ii): (i) What do you believe the objection to the bill was t? (ii) What did you see that there was t in the bathroom? See Heycock (1995) for a possible explanation. 1⁴ The presence of a PP ‘coda’ following the trace position in these examples seems to be an ameliorating factor. Hartmann (2008) found this made a significant difference in a magnitude estimation experiment.

against long-distance agreement without agree 223 (11) a. Which magazines did you say there were t in the waiting room? b. What options do you think there will be t on next year’s Prius? c. What kind of cake do you expect there will be t at the wedding reception? In (11), which/what seems to be ranging over kinds or types rather than individuals or tokens. This might explain why they sound better than (10). But the same kind/type readings should be available in the inverse copular examples in (12), yet they are still completely out. (12) a. b. c.



Which magazines did you say the target of the lawsuit was t? What options do you think the cause of the Toyota recall was t? ∗ What kind of cake do you expect the biggest challenge for the chef will be t? ∗

Thus, the real contrast is between inverse copulars on the one hand and existentials and canonical copulars on the other. Second, the associate of an existential can be extracted as a how many DP (13) (e.g. Hoekstra and Mulder 1990), as can the subject of a canonical copular (14a); these contrast with the inverse copular (14b): (13)

How many men do you think there were t in the room?

(14) a.

How many nude pictures do you think t were the cause of the scandal? ∗ b. How many nude pictures do you think the cause of the scandal was/were t?

Third, Safir (1985) observes that relativization of ‘total amounts’ is possible in existentials (15); subjects of canonical copulars also allow this (16); these contrast with inverse copulars (17): (15) a. The very few books that there are t in the house are all mysteries. b. All the children that there were t in the orphanage have finally been adopted. (16) a. The very few books that t are in the house are all mysteries. b. All the children that t were in the orphanage have finally been adopted.

224 carson t. schütze (17) a.



The very few books that the only source of entertainment is t are all mysteries. b. ∗ All the children that the focus of our concern was t have finally been adopted.

Fourth, den Dikken (2006) observes other flavours of wh-movement that fail on the notional subject of inverse copulas (18). Again, these contrast with associates of existential there (19), though the latter may not be perfect; canonical copulars are fine (20a) and (20c), unless the unmoved counterpart is itself ungrammatical, as (21) is for (20b). (18) a.



Whose oversized comforter do you suspect the problem for the washing machine was t? b. ∗ They fixed whatever problem the cause of the first shuttle disaster was believed to be t. ∗ c. How big of a poster would you say that gift to your brother was t?

(19) a. ? Whose clothes do you suspect there were t on your roommate’s bed? b. ? They fixed whatever problems there were believed to be t with the shuttle design. c. How big of a poster would you say there was t on your brother’s wall? (20) a. b. c. (21)





Whose clothes do you suspect t were on your roommate’s bed? They fixed whatever problems t were believed to be with the shuttle design. How big of a poster would you say t was on your brother’s wall?

Several problems were believed to be with the shuttle design.

Fifth, extraction from a post-copular DP is possible in canonical copular sentences (22a), but not in inverse copular sentences (22b): (22) a.

Which riot do you think a picture of the wall was [the cause of t]? b. ∗ Which wall do you think the cause of the riot was [a picture of t]?

As Moro himself notes, inverse copulars contrast with existentials (23).

against long-distance agreement without agree 225 (23) a. Which book do you think there were [many copies of t] in the library? b. Which models do you think there are [pictures of t] in the album? To summarize this subsection: with respect to extraction, existentials pattern with canonical copulars, but not with inverse copulars.

8.3.2 Quantifier Raising out of the post-copular DP: Contrast unexpected Quantifier Raising (QR) out of the postcopular DP is possible in canonical copulars (24a), but impossible in inverse copulars ((24b); judgements from Moro 1997): (24) a. A picture of the wall wasn’t [the cause of many riots]. ✓ many > not = Many riots are such that a picture of the wall wasn’t the cause of them. b. The cause of the riots wasn’t [a picture of many walls]. ∗ many > not ≠ Many walls are such that the cause of the riots wasn’t a picture of them. Moro asserts (contra his theory’s prediction) that existentials work differently from inverse copulars; QRing out is possible: (25) a. There isn’t [a picture of many walls] in the album. ✓ many > not = Many walls are such that there isn’t a picture of them in the album. b. There weren’t [copies of many books] in the library. ✓ many > not = Many books are such that there weren’t copies of them in the library. Assuming the facts are as described, existentials do not track inverse copulars with respect to scope.

8.3.3 Small clause environments: Contrasts not fully expected Inside small clauses, canonical copulars are possible, but inverse copulars are impossible, in contrast to full (infinitival) clauses.1⁵ 1⁵ A potential exception, depending on the details of the analysis, could be the following example from Heycock and Kroch (1999): (i) If Bill has an alibi for 6 p.m., that makes the murderer John.

226 carson t. schütze (26) a. Mary considers a picture of Stalin (to be) the cause of the riot. b. Mary considers the cause of the riot ∗ (to be) a picture of Stalin.1⁶ Here Moro is correct that there existentials pattern with inverse copulars: (27)

The inspector considers there ∗ (to be) too many people in the elevator.

Moro proposes that what is wrong with the copulaless versions of (26b) and (27) is that a small clause has no ‘extra space’ that would allow the predicate to get above the subject, an explanation based on the idea that the predicate (which is there in (27) on his analysis) must originate low. However, Heycock (1995) observes that predicate inversion across raising verbs is not possible either, as in (28b) and (29b),1⁷ and here the issue cannot be the lack of a landing site because the upstairs nonthematic subject position is available for that purpose. (28) a. % John seems the best man for the job. b. ∗ The best man for the job seems John. (29) a. b.



John is considered the best man for the job. The best man for the job is considered John.

Nonetheless, existentials are good (for many, especially British, speakers) in this environment.1⁸

To the extent that one can cook up anything vaguely similar with an existential, the copula is certainly not omissible, but the parallel is not very compelling: (ii) If Bill had an alibi for 6 p.m., that would make there ∗ (be) reasonable doubt. 1⁶ Contra the standard judgement in the literature, many people seem to accept the small clause version of this sentence, particularly with focal accent on the final DP. Similar accenting does not seem to help (27), however, which only creates another circumstance where existentials and inverse copulars are not parallel. 1⁷ Heycock (1995) points out the following exceptions: (i) The best solution became/remained instant retreat. den Dikken (2006) intriguingly suggests these are good because become means ‘come to be’ and remain means ‘continue to be’, so that be is actually present in these sentences, contra appearances. By contrast, seem and (be) considered in (28) and (29) obviously cannot be semantically decomposed into aspectual material plus be. 1⁸ Strictly speaking, the empirical issue is not how good (a) sounds relative to (b) in (30) and (31). Rather, it is how degraded each of them sounds relative to its counterpart in which to be has been included, e.g. as in (i): (i) a. There was considered to be insufficient evidence to prosecute the officer. b. The reason for the officer’s acquittal was considered to be the lack of witnesses.

against long-distance agreement without agree 227 (30) a. % There was considered insufficient evidence to prosecute the officer. b. ?∗ The reason for the officer’s acquittal was considered the lack of witnesses. (31) a. % There is considered inadequate justification for building a hotel on campus. b. ?∗ The best justification for building a hotel on campus is considered the demands of the parents. The conclusion is that, with respect to the full range of small clause environments, existentials do not track inverse copulars.

8.3.4 Optionality: Contrast unexpected As a final simple but perhaps not trivial point, predicate inversion is optional, but ‘there-inversion’ is obligatory:1⁹ (32) a. b. c.

There was a book missing t from the shelf. A book was t there missing from the shelf. ∗ A book was missing t there from the shelf. ∗

8.3.5 Summary There is no independent evidence that there originates in a local ‘predication’ relation with the associate or that there undergoes predicate inversion. Indeed, the evidence overwhelmingly demonstrates that there does not have this derivational history, since it does not pattern with inverted predicates. Therefore, a ‘secret courier’ solution that would explain long-distance agreement in English existentials in terms of predicate inversion is no better than any other stipulated solution to the long-distance agreement problem. In an acceptability rating task using a 7-point Likert scale (1 = worst, 7 = best) I conducted with 50 American speakers using Amazon Mechanical Turk, the mean ratings for sentences like (ia) and (ib) were almost identical: 5.56 vs. 5.50, respectively. However, sentences like (30a) and (30b) differed substantially: 4.76 vs. 3.72. The difference in the effect of ‘removing’ to be from the two sentences types (0.80 vs. 1.78) was significant (F(1, 98) = 4.70, p < .02). Thus, predicate inversion fares much worse than existential there in small clauses. 1⁹ It is not relevant that the strings in (32b) and (32c) are (relatively) acceptable if there is stressed and read as a locative.

228 carson t. schütze This is not to say that Moro’s idea might not be right for other languages. In particular, McCloskey (2014: 356) argues that in Irish existentials the superficial counterpart to there, ann, is actually crucially different from it: ‘[ann] has semantic content, [there] does not.’ He notes two further properties of ann not shared by there: ann may be focused and may appear as the (in situ) predicate of a small clause.2⁰ McCloskey speculates about other languages that may have existential constructions of the Irish type, e.g. those formed in German with da. On rather different grounds, Zwart (1992) suggests that Dutch ‘expletive’ er is sometimes a raised predicate and sometimes a purely syntactic placeholder.

8.4 Long-distance object agreement, or How do you solve a problem like Icelandic? 8.4.1 The data Icelandic has achieved notoriety for its quirky subjects, such as the datives (dat) in (33) and (34), which—except for not triggering agreement—behave entirely like subjects; in quirky-subject clauses, a nominative (nom), if there is one, behaves like an object.21 (33)

Henni líkuðu hestarnir. her.dat liked.3pl the.horses.nom.pl ‘She liked the horses.’

(34) a. Jónii virðast [ti hafa likað þessir sokkar ]. John.dat seem.3pl to.have liked these socks.nom.pl ‘John seems to have liked these socks.’

2⁰ Thanks to Andrew Carnie for bringing these facts to my attention. 21 Evidence abounds in the literature, including (for subjecthood of dat) binding of subject-oriented reflexives, being and controlling PRO, deleting and licensing deletion under conjunction reduction, licensing object NPIs, restrictions on extraction, etc.; (for objecthood of nom) failing subjecthood tests, undergoing Object Shift, etc. See Jónsson (1996) for review.

against long-distance agreement without agree 229 b. Jónii virðast [ti vera taldir [ti líka hestarnir ]]. John.dat seem.3pl to.be thought to.like the.horses.nom.pl ‘John seems to be thought to like the horses.’ In (34) the same VP is embedded under one and then two raising predicates, and the dat raises to become their subject. The finite verb (the probe) at the top of the raising structure cannot agree with the dat; instead it agrees (in number) with the nom object (the goal) in the lowest clause, as expected under Agree. How could we get this pattern without Agree?

8.4.2 A Spec–Head analysis Koopman (2003, 2006) explicitly argues that Spec–Head agreement is up to the task. Since she provides a very detailed proposal, it is worth considering at some length. She advocates for the Strong Agreement Hypothesis: ‘the Spechead agreement configuration […] is the only configuration leading to the spellout of agreement’ (Koopman 2003: 207). ‘Under the strong agreement hypothesis all cases of long distance agreement then are to be reanalyzed as arising at some early point in the derivation under a local Spec-head relationship’ (Koopman 2003: 211). This is a ‘clandestine meeting’ solution. More technically, ‘If XP agrees with Y, XP is merged with YP, or XP is merged with ZP which is merged with YP (or XP is merged with WP which is merged with ZP which is merged with YP, etc.)’ (Koopman 2006: 162, ex. (10)). That is, a head agrees with its specifier, or the specifier of its specifier, or the specifier of the specifier of its specifier, etc. (This allows ‘long-distance’ agreement to the extent that the distant Goal can pied-pipe structure above it into the specifier of the Probe.) This is illustrated in the following diagrams from Chandra (2011: 117–18):

(35) a.

YP XP

YP Y



230 carson t. schütze

b.

YP ZP XP

YP ZP

Y

c.

… YP

ZP WP XP

YP ZP

Y



WP W



Koopman’s structure for a monoclausal Icelandic sentence like (33) is as given in (36) (which abstracts away from the crucial intermediate steps of V-raising). In this structure, Subj1 P hosts the EPP feature that derives the surface position of the subject (independent of its case marking); Nom1 P checks subject Nominative case (cf. Cardinaletti 2004 for split subject positions); T1 P hosts canonical tense features; DatP checks Dative case; vP introduces the experiencer argument and projects DatP to provide it with inherent Dative case; VoiceP, when it contains the morpheme -st, selects a complement containing a NomP;22, 23 Subj2 P hosts the EPP feature that derives the surface position of the object (independent of its case marking); Nom2 P checks object Nominative case; T2 P has no tense semantics (cf. Collins and Thráinsson 1996); and finally, VP introduces the theme (internal argument).

22 For dat-nom verbs that do not have the -st suffix, like the one in (33), Koopman assumes a silent counterpart. 23 Note the apparent non-locality of selection: Subj2 P intervenes between VoiceP and NomP.

against long-distance agreement without agree 231 (36) Henni her.dat

Subj1P

Nom1P Øexpl Nom1

T1 líkuðu liked.3pl

T1P

DatP t

Dat

vPexp t v

VoiceP

Voice “-st”

Subj2P

hestarnir the.horses.nom

Nom2P t

Nom2

T2P T2

VP t

V

Five properties of this account merit attention. First, all cases are assigned/ checked in a Spec–Head relationship with a case head (curved double-headed arrows in (36)): quirky cases (here dat) are made available by vPs; nom cases are made available by TPs. Second, independent of those case positions, both the subject and the object (bearing any case) have their surface position determined by an EPP feature of a Subj head. Third, the nom case associated with subjects of canonical nom-acc clauses (Nom1 ) is not the same nom case associated with Nominative objects (Nom2 ) of quirky-subject clauses—it

232 carson t. schütze cannot be, under the Strong Agreement Hypothesis,2⁴ because a nom object never gets as high as T1 ; consequently, since subject nom is assumed to be part of every clause, it must undergo checking with a null expletive (∅expl ) in Spec-Nom1 P when the subject is quirky. Fourth, it is stipulated that only a Dative-experiencer-assigning v can take a VoiceP complement that contains a NomP—this is necessary because otherwise, predicates with two nom arguments would counterfactually be predicted to be possible in Icelandic;2⁵ this loses the intuition that nom becomes assignable to the object just when it is not assigned to the subject (as in Schütze 1997, among others). Fifth, to get longdistance agreement as in (37) (repeated from (3)), the account requires that ‘a second source for agreement must be assumed, [ . . . ] “agreement climbing” i.e. agreement with a plural predicate triggered under complex verb formation’ (Koopman 2006: 190). (37) a. Jóni virðast [ti hafa likað þessir sokkar ]. John.dat seem.3pl to.have liked these socks.nom.pl ‘John seems to have liked these socks.’ b. Jónii virðast [ti vera taldir [ti líka hestarnir ]]. John.dat seem.3pl to.be thought to.like the.horses.nom.pl ‘John seems to be thought to like the horses.’ Koopman (2006: 187) elaborates on this last point: ‘[A] nominative object DP will trigger agreement on all the verbs in a restructuring domain, because of two different processes: regular nominative agreement triggered low in the structure, and copying of agreement features as a byproduct of complex verb formation’ [emphasis added]. ‘[C]omplex verb formation always requires a local Spec-head relation between the restructuring predicate and the predicate of its complement’ (Koopman 2006: 187; emphasis added). I believe this is alluding to structures like (35c), where W was originally the lowest predicate and XP its local nom argument, then ZP the next higher predicate, and YP the highest predicate, assuming the lower ZP and lower YP each contain a trace.

2⁴ But see Schütze (1997: §4.1.1) for extensive argumentation that it should be. 2⁵ Consistent with this generalization, there is one situation in which a clause can contain two nom nominals in Icelandic: a copular construction, which as in English can occur in either canonical or inverse word order. The ensuing agreement patterns are complex and show considerable interspeaker variation—see Hartmann and Heycock (2016, 2017).

against long-distance agreement without agree 233 Thus, in (37b) the complex predicate ‘to like to be thought seem’ must be formed by moving each predicate into the spec of the next higher one. Subsequent movements, which Koopman does not discuss, must somehow restore the underlying word order of the predicates, which has been reversed, as well as the placement of adverbial material, which (38) shows can intervene between the predicates in such configurations (presumably excluding a head movement account of complex predicate formation):2⁶ (38) a. Jóni virðast einfaldlega vera taldir líla John.dat seem.3pl simply to.be thought to.like hestarnir. the.horses.nom.pl ‘John simply seems to be thought to like the horses.’ b. Mér mundu þá virðast þeir vera hérna. me.dat would.3pl then seem they.nom.pl to.be here ‘It would then seem to me that they are here.’

8.4.3 Is there independent evidence for the necessary Restructuring? Hróarsdóttir (2000) shows that Old(er) Icelandic had three surface patterns typical of (Germanic) Restructuring: 1. Short leftward object movement within VP; 2. Object scrambling out of VP to the left of nonfinite auxiliaries and sentential adverbs; 3. V-raising of non-finite main verbs to the left of non-finite auxiliaries. These all departed the language at the same time: they were stable from the fourteenth through the seventeenth centuries, but declined rapidly in the eighteenth century and all but disappeared by the nineteenth. Furthermore, those verbs that were attested with OV order in Older Icelandic fit very well 2⁶ My most sincere gratitude to Ásgrímur Angantýsson, Haldór Ármann Sigurðsson, Jóhannes Gísli Jónsson, Þórhallur Eyþórsson, and Höskuldur Þráinsson for these Icelandic judgements and discussion; standard disclaimers apply.

234 carson t. schütze into the semantic classes of verbs that license Restructuring in modern West Germanic (cf. Wurmbrand 1998), with only a half-dozen exceptions, for which Hróarsdóttir has plausible explanations. So, we know what Icelandic-cumRestructuring would look like: it used to be attested, but Modern Icelandic does not have it. The one climbing-like phenomenon that Modern Icelandic still has is quite plausibly the same that is found in Modern French, which has also lost Restructuring, viz. ‘L-Tous’ (Kayne 1975); importantly, such quantifier climbing, unlike clitic climbing, is possible in both languages not only from infinitival clauses but also from subjunctive clauses, at least marginally (Roberts 1997). (39) a. Jón hefur ekkert viljað lesa. John has nothing wanted to.read ‘John hasn’t wanted to read anything.’ b. Hann hefur ekkert viljað að ég segði. he has nothing wanted that I say.sbjv ‘He didn’t want me to say anything.’ (Hróarsdóttir 2000: 334) (40) a. Marie a tout voulu faire. [French] Mary has all wanted to.do ‘Mary wanted to do everything.’ b. Il n’a rien voulu que je dise. he NE=has nothing wanted that I say.sbjv ‘He didn’t want me to say anything.’ (Roberts 1997: 441)

8.4.4 Summary Under the Strong Agreement Hypothesis, the following elements are needed but lack independent motivation: a source for nom other than canonical Tense/C; a stipulation to prevent that second nom source from overgenerating (to block two nom arguments in one clause); Restructuring (complex verb formation); operations to undo the word order consequences of complex verb formation, which is what allows agreement to climb long distance; null expletives, whose distribution presumably must be restricted so that they are part of the numeration only when nom cannot be assigned to a contentful

against long-distance agreement without agree 235 argument;2⁷ a second configuration for agreement, i.e. Spec(-of-Spec)∗ -Head, apparently violating the Strong Agreement Hypothesis.2⁸

8.5 Is Agree redundant? 8.5.1 The general idea Hornstein (2009: ch. 6) offers analyses of long-distance agreement in English existentials and Icelandic nom object constructions as illustrations of a more general claim, namely that Agree should not be part of Minimalist syntactic machinery, essentially because it is redundant with other components of the theory. He argues that anything that can be achieved with Agree can be achieved without Agree, and thus by Occam’s Razor Agree should be done away with. Hornstein adumbrates three potential schemes for mimicking the effects of Agree (assuming there is some mechanism for highly local feature checking/valuation, the exact structural configuration for which is not his primary concern—it is the long-distance applicability of Agree that he argues is superfluous and hence undesirable). I enumerate them before considering their specific application to the constructions discussed in §§8.3 and 8.4. The first mechanism is covert movement—in particular, a second cycle of post–Spell-Out movement, as in early Minimalism and before, as opposed to simply spelling out lower copies/instances of moved elements in a single-cycle theory. Hornstein suggests that there may well be independent need for an LF cycle separate from considerations of implementing long-distance agreement, and thus it may be wrong to think that adding Agree to the theory buys us 2⁷ Opinions obviously vary widely on whether null expletives have any place in Minimalist syntax. I am inclined to agree with Jane Grimshaw, who said ‘Null expletives are an abomination’ (uttered during the presentation of Grimshaw and Samek-Lodovici 1995). In my view, null expletives are simply an unfalsifiable mechanical device used to force a derivation to proceed differently from how it normally would, typically by blocking otherwise-obligatory A-movement. A more compelling theory ought to explain why these operations are blocked in just the situations where they are. Beyond conceptual considerations, there is an empirical worry: the overt expletive in Icelandic triggers indefiniteness effects on the subject, much like English there, but Koopman’s null subject expletive obviously does not, since structure (36) has no semantic restrictions on its subject. Unless the featural content of the expletives somehow encodes this difference, it is hard to see how it can be derived. 2⁸ Koopman intends (35b) to be built by succesive Spec-Head agreement, but this could work only if the phrasal projection of the agreeing head (e.g. XP = Nom2P) inherits the φ-features of its head X, although those are uninterpretable and have been checked against an argument, so XP in turn can check the uninterpretable features of W, which must be inherited by WP, etc.

236 carson t. schütze the simplification of a single cycle theory. I take no position on this general architectural point. Hornstein does not propose covert movement in this sense as an analysis for either English there-constructions or Icelandic nom object agreement, with good reason: as the literature on both constructions has established, the DPs that trigger agreement show no general evidence of being higher than their surface positions at LF.2⁹ The second mechanism is what Hornstein calls ‘Sportiche-doubling’, by which he means a structure in which the agreed-with DP underlyingly consists of two parts, one of which can be split off and moved to a position local to the head (e.g. verb) manifesting the agreement. (This is another example of what I dubbed the ‘secret courier’ approach in §8.2.) The inspiration for this moniker is Sportiche’s (1988) classic analysis of quantifier float, schematized in (41). (41) a. D-Structure: have [VP [DP all [DP the children]] left] b. S-Structure: [DP the children]i have [VP [DP all ti ] left] Whatever the merits or challenges of this analysis for quantifier float in particular may be, Hornstein (2009: 131) suggests that the general idea can be taken much further, much more abstractly: [O]ne can even imagine ‘pure’ cases of Sportiche-doubling in which there isn’t any surface hint of an overt mover. Consider what things would look like if a DP were paired with a null double […]. Say, for example, a null pronominal clitic pro agreed with DP (on externally merging with it) and moved to the specifier of α agreeing with it there. (2)

[…[pro1 α0 …[t1 DP]…]]

On the surface this would look like an instance of long distance agreement. This proposal raises numerous questions for which Hornstein supplies no general answers (though he does make some suggestions for the particular application of this idea to English there-constructions; see §8.5.2). What independent evidence is there for the existence of constituents consisting of an arbitrary DP plus a clitic in the relevant languages?3⁰ What is the structure 2⁹ More precisely, in Icelandic the scope-taking options for nom DPs triggering agreement on finite verbs above them are no different from those for non-nom DPs in the same surface positions but not triggering agreement. 3⁰ This question would be particularly relevant if one were to apply the analysis to Icelandic: there are no restrictions on the semantics of nom objects that trigger agreement; in particular, they are not limited to the kinds of DPs that can be clitic doubled in languages that have (audible) clitic doubling.

against long-distance agreement without agree 237 of a DP that contains these two pieces, and how does that structure allow extraction of pro without violating constraints on extraction, e.g. the Left Branch Condition (in languages that enforce it, such as English and Icelandic)?31 What mechanism guarantees that their phi-features will match?32 How does the Case Filter apply to these DPs, or restated in current terminology, what unvalued/uninterpretable features do they bear? Does pro satisfy the EPP requirement of α (if it has one) and block other elements from moving to Spec-α, or if not, why not? Do the answers to these questions follow from any general principles or independently established properties of overtly studiable phenomena? The third mechanism Hornstein suggests for implementing long-distance agreement is simply movement (a.k.a. Internal Merge) coupled with deletion/nonpronunciation of the higher copy/occurrence. (In most if not all cases, this would additionally require stipulating that the higher copy/occurrence not be interpreted—e.g. as noted above for there-constructions, the associate takes scope in its surface position, not at the height of the expletive.) This is what Bobaljik (2002) refers to as the Lower Right Corner in the matrix of possibilities for privileging copies: movement occurs, but with no consequences for either PF or LF. I believe it is worth pondering rather deeply whether it makes sense to entertain a theory that espouses an operation that says at one time that an element has been dislocated and at the same time that this dislocation has no consequences for the structure at the phonological interface and no consequences for the structure at the semantic interface. What truly is the content of claiming dislocation has taken place under such circumstances? No more or less than that operations that could otherwise not have happened because the ‘dislocated’ element in question was not local to the landing site did in fact take place. (Hence my classifying this approach as ‘teleportation’ in §8.2.) That certainly seems tantamount to stating that those operations can happen long distance. So I might agree with Hornstein in one sense: if one truly wants to countenance movement that has no consequences at either interface then it indeed seems redundant to have a distinct Agree operation in the grammar, because both mechanisms achieve the same outcome: features can be checked at a distance. However, this is not the end of the discussion: there are both conceptual and empirical issues that demand further consideration.

31 It is worth noting the dissimilarity between the post-movement structures in (41) and Hornstein’s (2). 32 As pointed out in note 12, a small clause consisting of two DPs does not enforce phi-feature agreement between the DPs, for example.

238 carson t. schütze Conceptually, it seems to me that there is an issue of transparency in the choice of implementation. If we concur that the theory needs to allow features to be checked long-distance with no further consequences, is it better for the theory to include an operation that explicitly checks features at a distance with no further consequences, or instead for it to include an operation Move (whether or not that is any more than an informal cover term for Internal Merge, however implemented), which in the typical case has the property of actually moving/displacing an element (for pronunciation or interpretation or both, i.e. changing one or both of the LF or PF representations), but which can be stipulated in extraordinary circumstances to do neither, thereby in effect encoding a completely different operation that involves no movement/displacement but simply happens to implement long-distance feature checking? One conceivable tack would be to declare that there can be no such thing as vacuous Move, i.e. Move that has no effect on any interface representation. This would mean Bobaljik’s Lower Right Corner is excluded, and Agree is the only way to achieve ‘pure’ long-distance agreement. The relative appeal of this solution to Hornstein’s redundancy problem compared to his own (barring Agree) will surely vary depending on one’s taste. But I have so far taken for granted Hornstein’s empirical claim, namely that any facts capturable using Agree are also capturable by one of the three alternative strategies. Establishing this is far from trivial. All three alternatives crucially involve movement, which means that those alternatives will have to meet two conditions that their Agree-based counterpart analyses do not: they must obey the known constraints on the relevant kind(s) of movement in the language in question, and they must rely on the availability of landing sites appropriate for the moving element in that language. As we shall see in §§8.5.2 and 8.5.3, when Hornstein attempts to use Move to avoid Agree, his analyses fail to meet these desiderata.

8.5.2 Proposal for there-constructions Hornstein proposes a ‘Sportiche-doubling’ account of English thereconstructions, but does not lay out the structure of the underlying DPs in question. The greatest detail he puts forward comes from the following passage: ‘We can take it to be similar to genitive DPs like John’s book. The principal difference between the postulated there + associate DP and genitive

against long-distance agreement without agree 239 DPs is that there cannot carry genitive case’ (Hornstein 2009: 143).33 The best additional hint at the intended structure comes from Hornstein’s discussion of how he hopes to ensure that the features of there will match the features of the associate: ‘In many languages, Ds (or Spec Ds) agree with their nominal complements. […] If there is a dummy version of these, then it too could “agree” with its complement’ (Hornstein 2009: 140). In the elided portion of this quote he alludes to the fact that determiners can inflect for number and gender according to the features of the NPs they combine with. But there cannot possibly occupy the D position since the associate can contain its own determiner. It is not clear what he has in mind with regard to ‘Spec Ds’ agreeing with their ‘complements’, but if we are to think of Spec-D as the possessor position, as the previous quote suggests, then it is obviously false that possessors match the phi-features of their possessa: witness Johnsg ’s bookspl . So Hornstein’s proposal faces at least three rather grave obstacles: there is no serious proposal for the underlying doubling structure, no empirical evidence that DPs of the form [there many horses] are attested,3⁴ and no mechanism for ensuring phi-feature agreement between the two pieces of such structures, without which long-distance agreement cannot be implemented.

8.5.3 Proposal for Icelandic Hornstein (2009: 148) prefaces his discussion of Icelandic (and by implication similar long-distance agreement phenomena) with the following disclaimer: ‘In contrast to [the discussion of there-constructions], my aim here is not to provide alternative analyses of these constructions but to show that a movement account can do as well as an AGREE-based one regardless of the data involved’ [emphasis in original]. Nevertheless, he does then sketch an analysis for an example analogous to (33) in which the nom object has ‘tucked in’ as a second, lower specifier of TP in order to trigger agreement on the finite 33 As noted in §8.5.1, this raises an obvious problem: English does not allow extraction of possessors or other left branches from DPs. While Hornstein does not explicitly address this issue, hints in footnotes suggest he may hope to get around it by stipulating that there needs Case and (presumably unlike a possessor) cannot get it within the containing DP, though the latter claim in turn seems problematic in light of the fact, acknowledged by Hornstein in the same context, that there is licensed as the subject of acc-ing gerunds in English (e.g. [There being little time left] meant we had to skip lunch). 3⁴ Hornstein seems to commit himself to there being the leftmost element in the DP. In this respect he may be slightly worse off than Kayne (2008), who appeals to the existence, in nonstandard English, of DPs like that there book, whose order he claims is consistent with a potential analysis he alludes to.

240 carson t. schütze verb, while its lower copy is pronounced (and, presumably, interpreted). In a footnote he remarks, ‘[I]t is a standard assumption concerning Icelandic that it has multiple subject positions. Whether this is coded as allowing multiple specifiers of T or another FP with a specifier […] does not matter here’ (Hornstein 2009: 149, fn. 64). In other words, his proposal is that the landing site for the movement of the object is whatever the lower subject position of Icelandic transitive expletive constructions is. But that analysis simply cannot work, because that position is not guaranteed to be available: as has long been noted in the literature, nom object agreement is possible in transitive expletive constructions, where the notional subject occupies that position:3⁵ (42) a. Það there

líkuðu liked.3pl

mörgum many.dat

þessir these.nom.pl

tómatar. tomatoes.nom.pl [Icelandic]

‘Many liked these tomatoes.’ (Koopman 2006: 178, citing Halldór Sigurðsson, p.c.) b. Það leiddust sumum þessar ræður. there found_boring.3pl some.dat these.nom.pl speeches.nom.pl ‘Some people found these speeches boring.’ (ibid.) c. Það voru konungi gefnar ambáttir í vetur. there were.3pl king.dat given.pl slaves.nom.pl in winter ‘There was a king given slaves this winter.’ (Zaenen, Maling, and Thráinsson 1985: 462)

8.5.4 Summary This discussion illustrates, contra the emphasized portion of the quotation in §8.5.3, that the details of the data do matter, because movement requires a landing site, while Agree does not. Movement must also obey all the known constraints on movement in the language under discussion, while Agree need not (though certain minimality effects are built into it).3⁶ Thus, the 3⁵ Superficially one might wonder why there is no ‘defective intervention’ effect in these examples: the dat subject might be expected to block Agree between V+T and the nom object. In reality, Icelandic það is a ‘topic’ (Spec-CP) expletive, not a subject expletive (Vikner 1995), these are V2 structures, and the relevant Agree relationship was established prior to T-to-C. Icelandic’s two subject positions are Spec-CP and Spec-TP. The problem for Hornstein remains the same. 3⁶ In fact, Bošković (2007) argues that Agree, in contrast to Move, is not subject to the PhaseImpenetrability Condition at all.

against long-distance agreement without agree 241 following claim by Hornstein (2009: 149) is false: ‘As virtually every current minimalist theory includes both the Copy Theory of Movement and the option of pronouncing lower copies, AGREE accounts cannot empirically cover cases that movement accounts cannot. […] Thus, as a point of logic, the empirical coverage of AGREE-based accounts cannot be superior to Move-based ones.’ On the following page the audacity of this program is made more explicit: ‘Every AGREE-based analysis can be mapped into one involving movement. Here’s the recipe: in cases of I[nverse]-agreement assume that the “goal” has moved to a specifier of the “probe” in overt syntax and that the lower copy of the goal is pronounced (rather than the higher copy in the specifier position of the Probe)’ (Hornstein 2009: 150; emphasis added).

8.6 Concluding remarks 8.6.1 Variation This chapter has focused on cases where Agree seems to be required in order to explain the presence of long-distance agreement. Even if the line of argumentation embarked upon here continues to hold up and Agree turns out to be the best—even only—way to capture the phenomena discussed, more will need to be said in order to deal with cross-linguistic and crossdialectal variation. One dimension of variation mentioned in §8.3.5 involved the possibly different status of words superficially corresponding to there in, for example, Irish versus English. A complete theory of how natural language agreement works will also need to explain why, alongside the varieties of English described above, there are English dialects all over the world that have invariant singular agreement with there across the board (Rupp 2005), in contrast to the pattern in (2), for example: (43)

There was pits everywhere.

[West Midlands English]

Similarly, variation with regard to the Icelandic object agreement facts was alluded to in note 5. And in Faroese, which shares a great many properties with Icelandic, the predicates that take nom objects in Icelandic often take acc objects, and these objects never trigger agreement (44a); Thráinsson et al. 2004), nor does the dat subject, as (44b) shows:

242 carson t. schütze (44) a. Honum nýtist fleiri bókahillar heima hjá him.dat needs.3sg more bookshelves.acc.pl at.home with sær. [Faroese] self. ‘He needs more bookshelves in his home.’ (Barnes 1986: 19) b. Mær dámar børn. me.dat likes.3sg children.acc ‘I like children.’ (Höskuldur Þráinsson, p.c.) On many people’s assumptions, case is a reflex of agreement; if this is correct, the reason that these objects are not nom is because they are not agreed with, rather than vice versa.3⁷ Thus, it appears that just because a syntactic configuration arises in which Agree could in principle apply, that does not necessarily mean it will apply. (One tack that has been suggested for dealing with such variation is to parameterize the direction of the probe–goal relation, as alluded to in §8.1.2.) One subcase of this problem was remarked upon by Chomsky (2001) himself soon after the introduction of Agree: as Kayne (1985, 1989) had noted long before with regard to Romance participle agreement, it is not uncommon to find that a predicate realizes morphophonological agreement with its argument only if that argument has (overtly) moved to a position higher than the predicate. Such observations were among the motivations for claiming that the Spec–Head configuration was required for establishing an agreement relation. Kayne showed that in French, perfect participles never agree with following direct objects; rather, these objects only (can) trigger agreement when they have moved leftward across the participle (in his analysis, to/through the specifier position of an Agr projection that contains the participle): (45a) shows lack of agreement with an in situ object, while (45b) shows agreement when the object is cliticized and (46) when it is questioned or relativized.

3⁷ Icelandic speakers show no propensity to turn nom objects into acc ones even when the verb fails to agree with the object. This suggests that there must be a difference between a language making a parametric choice to exclude the possibility of agreement with the object versus a speaker of a language that allows such agreement failing to exercise that option. (Perhaps, as suggested below, we need to entertain the possibility that syntactically valued features need not always be morphophonologically expressed.)

against long-distance agreement without agree 243 (45) a. Ils ont {repeint / ∗ repeinte } la table. they have repainted.m / repainted.f the.f table.f ‘They have repainted the table.’ b. Ils l’ont repeinte. they it.f=have repainted.f ‘They have repainted it.’

[French]

(46) a. Je sais combien de tables ils ont repeintes. I know how.many of tables.f.pl they have repainted.f.pl ‘I know how many tables they have repainted.’ b. les chaises que Paul a repeintes the chairs.f.pl that Paul has repainted.f.pl ‘the chairs that Paul has repainted’ (47a) shows agreement when the underlying object is passivized and moved to subject position; (47b) shows an impersonal passive where agreement is blocked if the underlying object remains in situ and subject position is occupied by an expletive (which in French, perhaps importantly different from English there, is a pronoun fully specified for phi-features). (47) a. Cette table a été repeinte par Marie. this.f table.f has been repainted.f by Marie ‘This table has been repainted by Marie.’ b. Il a été repeint(∗ es) quelques tables. ∗ it.m.sg has been repainted.m.sg( f.pl) some tables.f.pl ‘There have been some tables repainted.’ Finally, (48) shows agreement in an ECM small clause of which the underlying object has become the subject. (48)

Je croyais ces tables repeintes. I believed these tables.f.pl repainted.f.pl ‘I believed these tables to (have) be(en) repainted.’

The situation in Italian is somewhat different: as in French, the in situ direct object does not trigger agreement (49a) while cliticization thereof does (49b), but wh-movement does not (49c):

244 carson t. schütze (49) a. Paolo ha {visto / ∗ viste } le ragazze. [Italian] Paolo has seen.m.sg / seen.f.pl the.f.pl girls.f.pl ‘Paolo has seen the girls.’ b. Paolo le ha viste. Paolo them.f.pl has seen.f.pl ‘Paolo has seen them.’ c. le ragazze che Paolo ha {visto / ∗ viste } the.f.pl girls.f.pl that Paolo has seen.m.sg / seen.f.pl ‘the girls that Paolo has seen’ Furthermore, unlike in French, there is at least one counterexample to the generalization that participles do not agree with objects that follow them: this occurs in an absolute construction described by Belletti (1981):3⁸ (50)

Conosciuta Maria, Gianni è cambiato del tutto. met.f.sg Maria.f.sg Gianni.m.sg is changed.m.sg completely ‘Having met Maria, Gianni has completely changed.’

Alluding primarily to the French pattern, Chomsky (2001: 46, note 39) remarks: Among unresolved questions are the reasons for Romance-style participial agreement contingent on movement, as discussed in Kayne 1989 and subsequent work. It is simple enough to state the parameter (‘Spell out ϕ-features of Prt-goal only if probe induces Move’), but there should be a more principled account, perhaps related to the manifestation-move correlations discussed by Guasti and Rizzi ([2002]).

Of course, the Italian facts show that it would not suffice even descriptively to state a language-wide parameter of the sort suggested by Chomsky, however stipulative that might be—different constructions must be independently parameterizable. Moreover, while stating the parameter might be simple, implementing it would appear to require that some diacritic be attached to the phi-features of the participle indicating that they are not to be spelled out, and then some additional mechanism must be posited such that if a higher probe with an EPP feature subsequently Agrees with the same goal that the 3⁸ Of course, one might hope to determine whether the object ever raised across the participle during the derivation of (50).

against long-distance agreement without agree 245 participle previously Agreed with, the diacritic will be erased. (No doubt more modern technologies could offer alternative implementations.) The facts mentioned in this subsection are of course just the tip of a vast descriptive iceberg. While there are broad tendencies both within and across languages for DPs to be more likely to trigger agreement and to trigger richer agreement when they are higher, exceptionless generalizations in this domain seem hard to come by. There is no doubt that Chomsky’s version of Agree faces serious empirical challenges within the narrow domain of predicate–argument agreement, let alone when it comes to the much broader range of phenomena that researchers have suggested it could/should be applied to.

8.6.2 The bottom line Nonetheless, I believe that the serious, arguably fatal problems identified in the particular analyses of long-distance agreement in English existentials and Icelandic nom object constructions scrutinized above cannot be easily dismissed. The question I take to be open is whether there are any Minimalist analyses of these phenomena that do not rely on Agree that fare better when examined in similar detail. Until such analyses are promulgated, Agree must remain as part of the Minimalist arsenal.

Acknowledgements It would be hard to overstate my intellectual debt to Elizabeth Cowper. She taught me my first real syntax class and got me hooked, using the manuscript of what became Cowper (1992), and my 25+-year love affair with Icelandic syntax began when I read Cowper (1988). Her advice was invaluable when I was writing my M.A. forum paper and later turning it into a book (Schütze 1996), when I was deciding where to go to pursue my PhD, and on innumerable other occasions. I cannot thank her enough. Thanks to Diane Massam and Bronwyn Bjorkman for organizing the fantastic workshop/reunion that yielded this volume, and to Bronwyn and Daniel Currie Hall for putting the book together and for their patience. Some of the ideas in this chapter were presented in an earlier form in Schütze (2011a; 2011b), so I acknowledge the respective audience and reviewers for their input. The work also benefited greatly from discussions in my graduate seminar ‘How and why subject position gets filled’ in Fall 2010, for which I am grateful to the attendees: Byron Ahn, Laura Kalin,

246 carson t. schütze Robyn Orfitelli, Craig Sailor, Jacopo Torregrossa, Kaeli Ward, and Pam Munro. For helpful comments and suggestions on previous drafts of this article I would like to thank two anonymous reviewers, Andrew Carnie, and Hedde Zeijlstra, as well as Caroline Heycock, Iara Mantenuto, Patricia Schneider-Zioga, and Susi Wurmbrand. This work was supported by a grant from the UCLA Academic Senate Council on Research.

9 Contrast in syntax and contrast in phonology Same difference? Daniel Currie Hall

The purpose of this chapter is to explore potential parallels between contrastive feature specification in phonology, particularly the Modified Contrastive Specification of the Toronto School (Dresher, Piggott, and Rice 1994; Hall 2007; Dresher 2009, 2015; etc.), and the role of contrast in morphosyntactic representations (following lines suggested by Harley and Ritter 2002a; Cowper 2005a; Wiltschko 2008; Ritter and Wiltschko 2014; Cowper and Hall 2014c; etc.). In more personal terms, it brings together the two main areas of my research agenda, and ideas from the two scholars who have had the greatest influence on me as a linguist: Elizabeth Cowper and Elan Dresher. The structure of the chapter is as follows. §9.1 provides background on the importance of contrast in linguistic representations generally. §9.2 describes approaches to contrastive feature specification in phonology, in which contrastive features are identified in the context of phonemic inventories. §9.3 addresses the question of what constitutes the morphosyntactic analogue of a phonemic inventory. §9.4 presents a specific proposal, whose consequences are explored in §9.5 with reference to person and number in Mi’gmaq.

9.1 Why contrast matters Language is a phenomenon of human cognition that connects forms (sounds or visual signs) with meanings. Neither the forms nor the concepts they express are peculiar to language: ideation, vocalization, and gesticulation all exist outside of linguistic expressions. What is peculiar to language is the formal system in the middle—an arched bridge that is strikingly different in some ways from the two banks on which its ends rest, to adopt an image proposed Daniel Currie Hall, Contrast in syntax and contrast in phonology: Same difference? In: Contrast and Representations in Syntax. First edition. Edited by: Bronwyn M. Bjorkman and Daniel Currie Hall, Oxford University Press (2020). © Daniel Currie Hall. DOI: 10.1093/oso/9780198817925.003.0009

248 daniel currie hall abstract, discrete, language-internal

syntax phonology

semantics

phonetics forms

meanings

concrete, continuous, interfacial

Figure 9.1 Language as an arch connecting forms to meanings

by Elan Dresher (p.c., though see also Dresher 2014 for a distinct but related arch metaphor), schematized in Figure 9.1. Forms and meanings are apprehensible, whereas the system that links them is abstract. Phonetic forms are inherently gradient; although some articulatory differences have much greater acoustic effects than others, and thus give rise to natural boundaries between categories of sounds (on which see Stevens and Keyser 2010 and references cited therein), the vocal tract is capable of producing much subtler distinctions than appear to be used phonologically. (To the extent that they are perceptible, some of these distinctions may be used to communicate social or affective information without being grammaticalized.) Similarly, at the opposite end of the arch, meanings can include categories with fuzzy boundaries and arbitrarily fine-grained nuances. The middle of the arch, on the other hand, deals in discrete units. It is also generally held to include categories that do not directly align with either forms or meanings: for example, morphosyntactic cases, whose mapping to forms is often complicated by allomorphy and syncretism, and whose correspondence with meanings is contingent on the lexical semantics of verbs and prepositions. Similarly, phonemic categories are subject to allophony and contextual neutralization on the form side, and their associations with meanings are largely arbitrary (in the sense of Saussure 1916). To the extent that the abstract formal computational systems of language deal with identifiable pieces of form and meaning (i.e. with material that is interpretable at the PF or LF interface in the model of Chomsky 1995), they do not deal with all of the ways forms or meanings can potentially differ from one another. For example, while some grammatical number systems distinguish plural from paucal (‘more than a few’ from ‘a few’), none distinguish specific higher numbers, for example thirty-nine from thirty-eight. Likewise,

contrast in syntax and contrast in phonology 249 phonological tone systems may distinguish multiple pitch levels, but none distinguish as many as the chromatic scale of Western tonal music. Formal representations distill the arbitrarily fine-grained worlds of sound and meaning into discrete primitives, generally represented as features. The representational richness of the features partially determines the computational power and complexity of the grammar. In principle, the linguistic computation might have access to much more detailed information about form and meaning than is demonstrably grammatically contrastive (as suggested in various phonetically grounded theories of phonology; see e.g. Hayes, Kirchner, and Steriade 2004), but contrast sets a lower bound on how much information is featurally encoded. Methodologically, it is useful to start from this lower bound, enriching our representations only when it turns out to be empirically necessary to do so, an approach elegantly described by Elizabeth Cowper (p.c.) as ‘retreat from perfection in the face of data’. If we begin by assuming that features represent only enough information to distinguish grammatically contrasting categories, then it should be relatively easy to discover if those representations are inadequate; if, on the other hand, we start with richer, more detailed representations, empirical evidence will not necessarily make it obvious whether they can be simplified. Feature systems based on minimal representation of grammatical contrasts share two important properties. First, in such a framework the presence of complex structures in a system implies the presence of simpler ones. For example, Harley and Ritter (2002a: 509) write: Assuming a model of feature acquisition along the lines of Rice and Avery (1995) or Brown (1997), we expect that if a language has a pronoun with a complex geometry, the simpler geometries that form the subconstituents of the complex geometry are also available in the language.

Second, the interpretation of any representation depends not only on its own content but also on what it contrasts with. For example, in the feature system of Harley and Ritter (2002a), a referring expression (re) with the features in (1) is interpreted as ‘first person singular’ if it stands in opposition to the representations in (2) and (3): (1)

re participant

individuation

250 daniel currie hall (2)

re participant

individuation

Addressee (3)

re participant

individuation Group

That is, if the feature [Addressee] is grammatically active in the system, then a Participant node lacking [Addressee] will be interpreted as first person, and if the feature [Group] is grammatically active, then its absence is contrastively interpreted as singular number. However, if [Group] and [Addressee] are simply not part of the system at all, as in the Winnebago (Hocąk) pronoun paradigm shown in (4), then the representation in (1), corresponding to the pronoun nee in (4a), means ‘any number of discourse participants’. (See also Wiltschko 2008 and Cowper and Hall 2014a,b for further discussion and examples.) (4) Winnebago pronouns a. nee ‘I’/‘you’/‘we’ b. ʔee ‘she’/‘he’/‘it’/‘they’ (Noyer 1992: 151; Harley and Ritter 2002a: 512; Harbour 2014b: 134–5) Similarly, in a typical phonemic consonant inventory, the absence of the feature [Voice] from an obstruent might contrastively indicate voicelessness, while the absence of [Voice] on a sonorant might not—if all sonorant phonemes are voiced, then none need be specified as voiced.1 The essential principle of contrastive specification, then, is that no property is specified as a feature unless it contrasts with its absence—or, if features are

1 However, [Voice] could still be specified contrastively on sonorants as well as obstruents if the voicing contrast takes sufficiently wide scope; see §9.2.1 for further discussion.

contrast in syntax and contrast in phonology 251 treated as binary rather than privative, with its opposite value. The next section discusses ways this principle has been implemented in phonology.

9.2 Contrast and features in phonology In phonology, the basic unit of contrast is the phoneme:2 minimally, a set of phonological feature specifications for any given language must be sufficient to differentiate the phonemes of the underlying inventory.3 However, while the shape of the inventory constrains what features can be assigned, it does not necessarily fully determine them, particularly in cases where two (sets of) phonemes differ in multiple properties. This section describes two phonological feature-specification procedures, neither of which is deterministic.⁴

9.2.1 Division Modified Contrastive Specification (Dresher, Piggott, and Rice 1994; so named by Paradis and Prunet 1991 to distinguish it from versions of Contrastive Specification proposed by Steriade 1987 and Clements 1988) proceeds by using features to divide the inventory, as in (5): (5)

Successive Division Algorithm (SDA) a. Begin with no feature specifications: assume all sounds are allophones of a single undifferentiated phoneme. b. If the set is found to consist of more than one contrasting member, select a feature and divide the set into as many subsets as the feature allows for. c. Repeat step (b) in each subset: keep dividing up the inventory into sets, applying successive features in turn, until every set has only one member. (Dresher 2009: 16)

2 See Dresher (2011) for an overview of the history and current status of the notion of the phoneme in phonology, and Hall (2013) for discussion of some complications to that notion. 3 There may also be other contrastive phonological information that distinguishes the underlying representation of one lexical item from that of another, such as moras or prosodic boundaries. ⁴ One procedure that is deterministic, based on identifying minimal pairs of phonemes, does not reliably produce specifications that are adequate to distinguish the members of the inventory to which it is applied; see Archangeli (1988), Dresher (2009: §2.5), Nevins (2015), Hall (2017: §2.1).

252 daniel currie hall In this approach, no feature is assigned unless it serves to mark some phonemic contrast, although some features may appear to be predictable from others in retrospect. The relative scope of contrasts, expressed in the sequence of divisions, is crucial, and can be represented by hierarchical branching diagrams of the sort used by Cherry, Halle, and Jakobson (1953) to measure the information content of phonemes. To take an example from Mackenzie (2013), suppose that the algorithm is applied to an inventory that includes (voiced) implosives (/ɗ/) and both voiced and voiceless plain stops (/d, t/), but no voiceless stop with glottal constriction such as ejective /t’/. This three-way contrast can be represented using two binary features, [±constricted glottis] and [±voice]. Because the two features do not fully cross-classify, whichever feature takes narrower scope will be unspecified on one class of stops, as illustrated in (6): (6)

Two orderings of laryngeal features a. [±constricted glottis] ≫ [±voice] tdɗ [+c.g.] ɗ

[−c.g.]

[+voice] [−voice] d t b. [±voice] ≫ [±constricted glottis] tdɗ [+voice] [+c.g.] ɗ

[−voice] t

[−c.g.] d (Mackenzie 2013: 300)

The hierarchy and specifications in (6a) capture the intuition that if we know that a segment is [+constricted glottis], its voicing is predictable; (6b) captures the intuition that if we know that a segment is [−voice], it is predictably not implosive. We might further reduce the number of specified features by taking advantage of both of these redundancies at once, specifying implosive /ɗ/ only

contrast in syntax and contrast in phonology 253 as [+constricted glottis] and voiceless /t/ only as [−voice]. This approach, however, would leave no explicit indication of how these two segments differ from each other (cf. Halle 1959: 32, quoted below in (10)). And, as Archangeli (1988) and Dresher (2009: §2.5) have shown, there are many circumstances in which omitting every feature value that can be predicted from some other feature value results in representations from which the missing features cannot be recovered at all. As Mackenzie (2013) notes, each of the two hierarchies in (6) appears to be attested: in Ngizim voicing harmony, implosives behave as if they are unspecified for [±voice], consistent with (6a), while in Hausa, voiceless consonants are ignored by [±constricted glottis] co-occurrence restrictions, consistent with (6b).

9.2.2 Addition The Parallel Structures Model of feature geometry (PSM; Morén 2003, 2006a,b) also incorporates the idea that complexity of representations is driven by contrast, but in a way that suggests a different approach to assigning specifications. In the PSM, for every feature used in a given system, there is one segment specified only with that feature. This principle is similar to the Autonomous Interpretation Hypothesis in Element Theory (Harris and Lindsey 1995). Hall (2012) proposes an algorithm for building representations based on this principle; unlike the SDA, it starts with an empty inventory and proceeds by adding segments and features, as in (7): (7)

Building parallel structures a. Start with no segments, and no features. b. To add a segment to the inventory, do either of the following: i. Create a new primitive segment by introducing a new feature F (or by placing an existing feature F under a different node), or ii. Create a new complex segment by combining the features of any existing segment with the sole feature F of any existing primitive segment. (Hall 2012)

In this approach, as with the SDA in (5), no feature is specified unless it marks a contrast. For example, it is not possible to introduce a segment with

254 daniel currie hall the feature specification [F, G, H] unless there are already segments specified as [F], [G], and [H], plus at least one of [F, G], [G, H], or [H, F].⁵ The SDA lends itself to contrastive hierarchies like the ones in (6), in which inventories branch into progressively smaller natural classes identified by progressively larger sets of features. The procedure in (7) suggests instead a semilattice like Figure 9.2, in which simple properties combine to produce more marked segments and representations. Figure 9.2 shows Morén’s (2006b) feature specifications for the phoneme inventory of Hawaiian. At the bottom of the diagram are the individual features, below their organizing nodes in the geometry. Directly above each combination of feature and organizing node is the corresponding ‘primitive’ segment specified only with that feature. (For example, both /h/ and /a/ have only the feature [Open], which is under the C-Manner node in the case of /h/ and under V-Manner in /a/.) Segments characterized by multiple features are shown higher up in the diagram, connected to the simpler segments that have subsets of their features. (For example, /n/ has both [C-Manner: Open] and [C-Manner: Closed]; /p/ has [C-Manner: Closed] and [C-Place: Labial]; and /m/ has all three of the features that occur on /n/ and /p/.) Iosad (2012, 2017) proposes a synthesis of sorts between the SDA and the PSM, using monovalent features and successive division of the inventory. As Hall (2007) discusses, applying the SDA as formulated in (5) to monovalent features will always leave exactly one ‘empty’ segment with no features in any given inventory; in Iosad’s adaptation, languages may assign a single feature to this otherwise unspecified segment. The resulting specifications, in which there are multiple segments with the minimum specification of one feature, o

m n

p

h C-Manner

ʔ C-Manner

Open

Closed

k

u

e

w l i C-Place C-Place V-Manner Labial

Tongue

a V-Manner

Closed

Open

Figure 9.2 Hawaiian feature specifications (Morén 2006b) as a semilattice

⁵ In principle, this procedure allows for an inventory to consist entirely of primitive segments, in which case each segment is specified with only one feature and each feature occurs on only one segment. The fact that such inventories do not appear to exist suggests a preference for step (7b-ii) over step (7b-i), which would be one possible formal implementation of Clements’s (2003, 2009) principle of Feature Economy.

contrast in syntax and contrast in phonology 255 resemble those assigned by the procedure in (7), except that there can be features that do not occur as the sole specification of any segment. In the SDA, more complex feature specifications are created as the inventory is divided into smaller and smaller subsets; in the procedure in (7), they are created through the combination of simpler representations. In each case, however, the complexity of representations is constrained, though not wholly determined, by the size and shape of the underlying inventory. The principal differences between the two follow from differences in the conception of features—where the PSM treats features as element-like primitives that make specific substantive contributions to the content of phonemes, the SDA treats them as serving primarily to differentiate classes of phonemes from one another.

9.3 Inventories in phonology and syntax 9.3.1 Phonemic inventories Under either of the two approaches described in the preceding section, contrast in phonology is determined relative to an inventory of phonemes. For example, the specifications in Figure 9.2 serve to differentiate the phoneme inventory of Hawaiian, which comprises the typical five-vowel inventory /i e a o u/ and the consonant inventory in Table 9.1. The grammar must be able to encode the difference between one phoneme in another at least for the sake of representing contrasting lexical underlying forms, since the phonemic content of lexical items is (by definition) unpredictable. The Contrastivist Hypothesis (Hall 2007; Dresher 2009) posits that the same features that serve this minimal function are also all that are needed to account for their phonological behaviour. Phonological inventories can be seen as collections of points (or, more realistically, regions) in multidimensional phonetic space. Inventories frequently Table 9.1 Phonemic consonant inventory of Hawaiian Labial stop fricative nasal approximant

Coronal

p m w

n l

Dorsal

Glottal

k

ʔ h

256 daniel currie hall have idiosyncratic gaps. In Hawaiian, for example, there is no phonemic coronal stop /t/, nor a phonemic dorsal nasal /ŋ/, but even in much larger inventories, the phonetic space is seldom uniformly filled with phonemes. Such gaps, both systematic (e.g. the absence of voiced obstruents in Hawaiian) and seemingly accidental (e.g. the absence of /t/ or /ŋ/), are crucial to crosslinguistic variation in the contrastiveness of features. For example, if a series of ejectives (represented by /t’/) were added to the systems in (6), then there would be full cross-classification of the features [±voice] and [±constricted glottis], and neither would be unspecified, regardless of the order of divisions. In the case of Hawaiian, if the inventory did not include /ʔ/, then the representation [C-Manner: Closed], assigned to /ʔ/ in Figure 9.2, might be used for /k/ instead. Or, if Hawaiian had a phonemic contrast between /k/ and /t/, then some additional place feature would need to enter the system to distinguish them. Phonemes can also be distinct phonologically without being distinct phonetically. For example, Compton and Dresher (2011) argue that Northern Alaskan Iñupiaq has a phonemic contrast between a ‘strong [i]’ that triggers palatalization of a following consonant and a ‘weak [i]’ that does not. The examples in (8) illustrate the pattern. Stems ending in vowels other than [i], such as ‘house’ in (8a), do not trigger palatalization. Of stems ending in [i], there are some, such as ‘wound’ in (8b) that cause palatalization, and others, like ‘place’ in (8c), that do not. Since it cannot be predicted from any other property of an [i]-final stem whether it will trigger palatalization, this difference between (8b) and (8c) must be stored in the lexicon in some form. Compton and Dresher (2011) show that positing an underlying contrast between two phonemes that both surface as [i] offers an elegant synchronic account with a clear diachronic origin, as their ‘weak [i]’ is the reflex of an earlier /ə/. (8)

Strong and weak [i] in Northern Alaskan Iñupiaq a. i. [iɣlu] ‘house’ ii. [iɣlu-lu] ‘and a house’ iii. [iɣlu-nik] ‘houses’ (oblique) b. i. [iki] ‘wound’ ii. [iki-ʎu] ‘and a wound’ iii. [iki-ŋik] ‘wounds’ (oblique) c. i. [ini] ‘place’ ii. [ini-lu] ‘and a place’ iii. [ini-nik] ‘places’ (oblique) (Compton and Dresher 2011: 206, citing Kaplan 1981)

contrast in syntax and contrast in phonology 257 Abstractness of the sort entailed in Compton and Dresher’s (2011) analysis of Iñupiaq has been controversial in phonology. For example, Kiparsky’s ([1968] 1973) Alternation Condition prohibits absolute neutralization of underlying phonological features (see also Hyman 1970, 1973 and Harms 1973 for further discussion). However, one way or another, it must be possible for linguistic objects to be grammatically distinct without being phonetically distinct. (Accidental homophony is perhaps the most obvious and least interesting type of case.) For Iñupiaq, the Alternation Condition would require that there be no underlying phonological distinction between strong /i/ and weak /ə/, since both surface as [i] in all contexts, but Kiparsky’s approach would instead posit diacritic ‘rule features’ to distinguish the /i/-final stems that trigger palatalization from those that do not. In either analysis, a feature contrast on one item (either a rule feature on the stem or a phonological feature on its final vowel) is audible only on an adjacent item (the immediately following consonant). Abstract or displaced contrasts of this sort exist in syntax as well. For example, a C head that identifies a clause as a wh-question can do so either by being itself audible or by having a wh-phrase move to its specifier (Cheng 1991). The phenomenon in (8) is essentially the same thing: formally distinct items can signal differences in meaning either by sounding different themselves or by having audibly different effects on their surroundings.

9.3.2 Morphosyntactic inventories? Contrastive specification depends on identifying the domain of items to be contrasted; to determine whether a given set of features assigned to a linguistic object is sufficient or necessary, one must know what other objects it needs to be differentiated from. In phonology, the domain of contrast is the phonological inventory, and the objects to be distinguished are phonemes. Given that there is not a traditional notion of ‘morphological inventories’ immediately analogous to phonological inventories, what is the domain of contrastive specification in morphosyntax? If morphosyntactic representations can be represented as feature bundles at all, and if the morphosyntactic component of any given language uses a finite set of features, it should in principle be possible to identify the set of objects these features distinguish. One obvious—though ultimately spurious, as discussed below—candidate would be inflectional paradigms, which bear at least a superficial resemblance to phonological inventories; both are sets of linguistic objects arrayed in a

258 daniel currie hall multidimensional space, often with idiosyncratic asymmetries. For example, consider the suffixes of the Old Church Slavonic (OCS) ‘twofold’ noun declension, shown in Table 9.2 (based on Lunt 1959: 54).⁶ In this declension paradigm, number, gender, and case define a threedimensional grammatical space, analogous to the phonetic space defined by place and manner in the Hawaiian consonant inventory in Table 9.1. However, whereas the absence of a contrast in Table 9.1 appears as a gap, absences of contrast in Table 9.2 show up as syncretisms. For example, in the dual number, there is no contrast between nominative and accusative, or between genitive and locative, or between dative and instrumental. There are also less tidy syncretisms—for instance, the form -a is highly ambiguous, variously representing feminine nominative singular, masculine or neuter genitive singular, masculine nominative or accusative dual, and neuter nominative or accusative plural. Some of these syncretisms seem quite systematic, as in the reduction from six cases to three in the dual; others have a more haphazard appearance.⁷ OCS has no distinct form for, e.g. feminine genitive dual. This does not mean, though, that there is no feminine genitive dual in OCS; it merely means that the feminine genitive dual shares its form with various other combinations of gender, number, and case, such as masculine dative singular. While there are fewer forms in Table 9.2 than there are slots, the table has no empty slots, unlike Table 9.1. Upon closer examination, however, this difference between morphological paradigms and phonological inventories is less categorical than it at first Table 9.2 Suffixes marking number, gender, and case on nouns and indefinite adjectives in Old Church Slavonic

nom. acc. gen. loc. dat. inst.

singular dual neut. fem. masc. neut. fem. -o, -e -a -ě, -i -o  ¸ -a -y, -ę -u -ě, -i -u -oma, -ema -ama -omь, -emь -ojo  ˛, -ejo  ˛

masc. -ъ, -ь

masc. -i -y, -ę

plural neut. -a

-ъ, -ь -ěxъ, -ixъ -omъ, -emъ -y, -i

fem. -y, -ę -axъ -amъ -ami

Source: Lunt (1959: 54)

⁶ Paired forms occupying the same cell are in phonologically predictable distribution, the form on the right appearing on stems ending in ‘soft’ (palatal) consonants and the form on the left elsewhere. ⁷ See Béjar and Hall (2000) for an analysis of OCS inflectional syncretisms in the framework of Distributed Morphology.

contrast in syntax and contrast in phonology 259 appears. To some extent, it is an artefact of the way phoneme inventories are conventionally represented, with single symbols occupying individual table cells and often no explicit indication of allophonic variation. For example, while Hawaiian does not have a phonemic /t/ that contrasts with /k/, that does not mean that there is no phonetic [t]. Rather, [t] occurs as an allophone of /k/ in something approximating free variation, as illustrated in (9). (9)

Interchangeability of [t] and [k] in Hawaiian a. [kanaka] ∼ [tanata] ‘people’ b. [ko] ∼ [to] ‘sugar cane’ c. [tabetee] ∼ [kabekee] ‘cabbage’

(Schütz 1995: 77–8)

To state the allophony in terms usually used for morphological paradigms, Hawaiian has a syncretism between coronal and dorsal place in the stop series. Similarly, Hawaiian /n/ encompasses [n] and [ŋ]; /w/ encompasses [w] and [v]; and /l/ encompasses [l] and [r]. Given the sparseness of the inventory, it is possible to eliminate several apparent gaps by identifying the relevant classes of consonants in more general terms than those used in Table 9.1: the coronal and dorsal places can be conflated into a single lingual category (cf. Hockett 1955: 98), and among the class of continuants there is no phonemic basis for discriminating among fricatives, liquids, and glides. Even when these more general categories are adopted, as in Table 9.3, there is still one true gap (the anatomically impossible glottal nasal), and of course other phonological systems have gaps that are not phonetically inevitable (such as the absence of /t’/ in Ngizim and Hausa, reflected in (6)). Given that phoneme inventories can have syncretisms as well as gaps, then, we might next ask whether morphological paradigms can have gaps as well as syncretisms. In various languages, we find phenomena such as Person– Case Constraint (PCC) effects and the Amn’t Ban (on both of which see e.g. Table 9.3 Another interpretation of the Hawaiian consonant inventory

stop nasal continuant

Labial

Lingual

Glottal

p m w/v

t/k n/ŋ l/r

ʔ h

260 daniel currie hall Nevins 2008). These restrictions, however, involve particular combinations of morphemes rather than combinations of features. It is not obvious that it would be constructive to posit an inventory of clitic sequences or an inventory of contractions, with underspecification of attested members based on the absence of contrast with unattested ones. Rather, restrictions of this sort seem amenable to mechanisms such as impoverishment, or relativized probes, or (in the case of ∗ amn’t) perhaps even phonotactic filters. Yang (2016, 2017) argues that children acquiring language posit productive morpholexical rules if and only if it is computationally more efficient to do so than it would be to list each form that the rule would generate separately. This criterion is met when the proportion of exceptional forms is sufficiently low. Gaps in a morphological paradigm will be possible only when the paradigm itself is too irregular to make a rule computationally worthwhile; otherwise, the rule will fill in the gaps with the regular form. In English, for example, there are too many irregularities to justify a general rule for contracting negation to auxiliaries ending in sonorants (e.g. will+not yields /woʊnt/ rather than ∗ /wɪlnt/), and so each such contraction is treated as an independent item. Most varieties of English happen to lack a form ∗ amn’t. Under Yang’s approach, a gap of this sort can only ever be the absence of one idiosyncratic form among several such forms that are not unified into any more general grammatical pattern; if there were enough of a pattern to support a rule, the gaps would all be filled in by it. If Yang is correct in predicting that gaps like ∗ amn’t can exist only where there are no rules to apply, then it seems unlikely that these kinds of gaps have interesting representational consequences for the grammar in the same way that gaps in phonological inventories do. A potentially more promising, if unsurprising gap is the general absence of singular inclusive pronouns and agreement, which is perhaps analogous to the absence of a glottal nasal in Table 9.1. It is logically impossible for a singular pronoun to include both the speaker and the addressee (given that languages do not have dedicated morphological resources for soliloquy), much as it is physically impossible to close off the flow of air at the vocal folds while sustaining airflow through the nasal cavity. In an agreement system like that of Upper Svan (Table 9.4a), this gap might be a basis for underspecifying inclusive forms for number, or singular forms for clusivity. (Interestingly, some languages have a first-person inclusive dual that occupies a paradigm slot parallel to the other singular forms, as in the Winnebago paradigm in Table 9.4b; Harbour (2008) uses this to argue for the feature [±augmented].) I will return to this possibility in more detail in the next section.

contrast in syntax and contrast in phonology 261 Table 9.4 Singular/plural vs. minimal/augmented a. Upper Svan agreement

1 excl. 1 incl. 2 3

b. Winnebago agreement

Singular

Plural

xw— x(various)

xw- … -d l- … -d x- … -d -x

1 excl. 1 incl. 2 3

Minimal

Augmented

hahinra∅

ha- … -wi hin- … -wi ra- … -wi -ire

Source: Harbour (2008: 4)

9.4 What do we need to distinguish? Two things that have no basis for comparison, that do not have a single property in common, as, for example, an inkpot and free will, do not form an opposition. Trubetzkoy (1969: 68)⁸

The problem with treating paradigms as inventories is that they are merely inventories of vocabulary items—the surface realizations of abstract grammatical categories, not the grammatical categories themselves. In the case of phonological inventories, the phonemes are the grammatical categories, and their surface realizations are allophones. Morphological syncretisms are not the analogue of allophony, in which there is an absence of contrast at the more abstract level, but its inverse, in which an underlying contrast is neutralized at the surface level. This difference is illustrated schematically in Figure 9.3, with reference to the arch in Figure 9.1. ↑

TOWARDS THE TOP OF THE ARCH

/k/

[t]

[k]



m./n.

m./n./f.

dat. sg.

gen./loc. dual

-u

↓ TOWARDS THE PHONETIC END ↓

Figure 9.3 Allophony versus syncretism

⁸ In the original German: ‘Zwei Dinge, die gar keine Vergleichungsgrundlage, d. i. keine einzige gemeinsame Eigenschaft besitzen (z. B. ein Tintenfaß und die Willensfreiheit) bilden keinen Gegensatz’ (Trubetzkoy 1939: 60–1).

262 daniel currie hall In Distributed Morphology, underspecification of vocabulary items is routinely used to account for syncretisms; if a single vocabulary item is inserted in some contexts that are [−F] and others that are [+F], then it cannot have any specification for [±F] (Halle and Marantz 1993). This kind of underspecification, however, is very different from the underspecification of phonemes described above in §9.2. Unlike phonemes, vocabulary items compete for insertion, and need not be featurally distinct from one another. In phonology, Halle (1959: 32) has proposed the following definition of distinctness:⁹ (10)

Definition of the different-from relation Segment-type {A} will be said to be different from segment-type {B}, if and only if at least one feature which is phonemic in both, has a different value in {A} than in {B}; i.e., plus in the former and minus in the latter, or vice versa. (Halle 1959: 32)

In morphology, or at least in Distributed Morphology, on the other hand, it is entirely unexceptionable for one vocabulary item to be specified with a proper subset of the features of another. In such a case, the more highly specified item blocks the more general one in contexts where all of its features are matched, and the less specified one is treated as a default, in accordance with the Elsewhere Principle (Kiparsky 1973; Halle and Marantz 1993). The features specified on vocabulary items, unlike the features specified on phonemes, are not properties of the vocabulary items themselves; rather, they are specifications of contexts in which the vocabulary items can be inserted. Similarly, allophones, the phonological counterparts of vocabulary items in Figure 9.3, do not necessarily require grammatically distinct representations. Some allophones clearly are distinguished by the same kinds of phonological features that distinguish phonemes. For example, Czech has a non-contrastive voiceless postalveolar fricative trill [r]̝ ̊ that is predictably derived from /r/̝ by final or assimilatory devoicing. Because the same processes also produce alternations between contrasting pairs such as /d/ and /t/, the simplest hypothesis is that the allophonic difference between [r]̝ ̊ and [r]̝ is represented in the same way as the phonemic difference between /t/ and /d/.1⁰ Other allophonic differences, however, need not be expressed in phonological features at all, ⁹ This definition presupposes binary features; one consequence of positing monovalent features is that the difference between the contrastive and non-contrastive absence of a positive value is no longer visible in the feature values themselves (though this information is recoverable from the hierarchical organization of the features). 1⁰ See Hall (2004, 2007) for further discussion of Czech voicing patterns, and compare Halle’s (1959) and Chomsky’s (1964) similar argument about voicing in Russian.

contrast in syntax and contrast in phonology 263 particularly if they have no phonological consequences and no obvious basis in feature spreading. For example, Campbell and Campbell (2009) report that in Jamamadí /l/ is typically realized as [ɾ] before back vowels and as [l] before front vowels. An allophonic difference in manner on the consonant is conditioned by a phonemic difference in place on the vowels, and does not appear to be connected with any other allophonic or phonemic alternations; there are no obvious insights to be gained by attempting to formalize this allophony in distinctive features. In both phonology and morphology, then, it is the contrasting items on the more abstract side of Figure 9.3 that consistently need to be assigned featurally distinct representations. Just as phonemes need to be distinguished underlyingly even if they are sometimes neutralized at the surface, morphosyntactic objects need to be distinguished even when they are realized syncretically. For example, consider the case of an Old Church Slavonic noun that surfaces with the suffix -u, syncretic between masculine or neuter dative singular and any gender genitive or locative dual. Although the suffix does not indicate the gender of the noun, the same noun in other contexts will appear with different suffixes according to its gender. Although the suffix does not tell us whether the number is singular or dual, a coindexed pronoun in a different syntactic position may need to mark this distinction. And although the suffix does not tell us the case (unless we independently know that the noun is singular), different cases are assigned by different syntactic heads—the grammar cannot simply be agnostic as to whether this form is dative or genitive or locative. To the extent that there are morphosyntactic inventories parallel to phonemic ones, then, those inventories are closer to the top of the arch; they are not made up of vocabulary items, but of the more abstract entities that those vocabulary items spell out. Thus Cowper and Hall (2014c) propose that the relevant inventories in morphosyntax consist of ‘functional lexical items (LIs), considering only the Lexicon in the narrow sense and not the Encyclopedia (Marantz 1996)’. Here, I would like to propose a minor amendment to this formulation. Rather than inventories of lexical items, it may be more appropriate to think in terms of separate morphosyntactic inventories for substantively different dimensions. As argued by Bobaljik and Thráinsson (1998) and Cowper and Hall (2013, 2017), contra Cinque and Rizzi (2008), a single functional head may contain multiple features. Even in such cases, though, features marking different kinds of morphosyntactic/semantic properties generally cross-classify fully. For example, Cowper and Hall (2013) posit that Early Modern English had a single Voice/Aspect head (VAsp). On this VAsp head, voice could be either Passive or unmarked (active); aspect could be Process, Result, or unmarked. Any of

264 daniel currie hall the three aspects could combine with either of the two voices; the fact that they were bundled on a single head accounts for (among other things) the fact that when both Process and Passive were present, they were spelled out by a form that was not distinct from the corresponding active form, as in the (relatively late) example in (11a), whereas Present-Day English spells out voice and aspect on separate heads, as in (11b). (11) a. Our Garden is putting in order, by a Man who bears a remarkably good Character…. (Jane Austen, letter of 8 Feb. 1807, quoted in Denison 1998: 148) b. Our garden is being put in order. The surface conflation of active and passive, though, is partly due to the absence of any specifically passive vocabulary items spelling out this head. There are no gaps or syncretisms in the set of combinations of features that can appear on the head itself. The connection between voice and aspect here is simply that they occupy the same syntactic projection, rather than each appearing on its own head. There is no sense in which one of them is predictable given the other, or in which one of them is contrastive only in the context of a specific value of the other. Accordingly, there is no need to consider, for example, the [Process, Passive] VAsp head as an item in our inventory that needs to be distinguished from each other possible VAsp head, let alone from pronouns or tenses or whatever other functional heads there are. Instead, there is an inventory of voices and an inventory of aspects, and items from these two inventories occupied the same head in Early Modern English. Within each of the separate morphosyntactic inventories posited by this view, we find asymmetries and gaps that can be explained in terms of contrasts and their relative scope. For example, in the number inventory, languages can differ in the relative scope of the features that distinguish mass, singular, and plural, or in the D inventory, Deictic and Distal may be contrastive only in the context of Definite (Cowper and Hall 2014b).

9.5 Consequences: Person and number in Mi’gmaq What are the consequences of separating dimensions of morphosyntactic contrast, as opposed to recognizing just a single inventory of LIs distinguished

contrast in syntax and contrast in phonology 265 Table 9.5 Mi’gmaq personal pronouns

1 excl. 1 incl. 2 3

Singular

Dual/Plural

ni’n — gi’l negm

ninen ginu gilew negmow

Source: McClay (2012: 18)

by a range of different kinds of features? There are some clear and potentially testable differences between the two approaches, even taking into account the fact that the underlying system of contrasts on LIs may be obscured by syncretisms in how those LIs are spelled out by vocabulary items. Consider, for example, a pronoun system that includes a two-way number distinction and an inclusive–exclusive contrast in the first person plural, like that of Mi’gmaq, shown in Table 9.5. Although a contrast between dual and plural is marked elsewhere in the Mi’gmaq grammar (specifically in intransitive verb agreement paradigms; Little 2018), there is only one set of non-singular pronouns. Following Harbour (2014a), I assume that the singular–non-singular contrast in Mi’gmaq is marked by [±atomic], and that the dual–plural contrast, where it exists, by [±minimal] subdividing the [−atomic] category. This means that, as in the Winnebago paradigm in Table 9.4b, there is no first person inclusive singular, and that this is a gap in the inventory of lexical items, not merely in the inventory of vocabulary items: a [+atomic] individual cannot include both the speaker and the addressee. If number and person are treated as distinct morphosyntactic dimensions, then their features are organized into two separate contrastive hierarchies, as in (12). (12)

Number and person as separate hierarchies in Mi’gmaq

a. Number feature hierarchy # [+atomic] sg

[−atomic]

[+minimal] du

[−minimal] pl

b. Person feature hierarchy π [+author] [−part] 1 excl

[+part] 1 incl

[−author] [+part] 2

[−part] 3

266 daniel currie hall The person features [±author] and [±participant] in (12b) are adapted from Harbour (2016). For Harbour, these features are not first-order predicates representing properties that could be used to divide a set of LIs, but rather operations on lattices: [+author] adds the speaker to all members of a lattice; [−author] removes the speaker; [+participant] adds the speaker and the hearer both jointly and separately; and [−participant] removes all discourse participants. Hall and Cowper (2016) and Cowper and Hall (2019), however, argue that first-order predicate versions of these features can achieve the same empirical coverage as Harbour’s (2016) versions if they are organized into a contrastive hierarchy. Under this view, [±author] simply means ‘{includes, does not include} the speaker’, and [±participant] means ‘{includes, does not include} a(t least one) discourse participant’. The scope ordering of the features takes on a role analogous to that of the order of application of operations in Harbour’s approach. In four-way person systems (first inclusive vs. first exclusive vs. second vs. third), [±author] takes scope over [±participant], as in (12b), and the denotation of [±participant] narrows in accordance with the narrowness of its scope: when [±author] has already been specified, [±participant] means ‘{includes, does not include} a discourse participant other than the speaker’. Three-way systems, which lack an inclusive–exclusive contrast, are derived by giving [±participant] wider scope, as in (13). The [−participant] branch cannot be subdivided by [±author], because there is no way of narrowing the meaning of [+author] that would make it meaningful in a context where discourse participants have already been excluded. (13)

Person feature hierarchy for languages with no clusivity contrast π [+participant] [+author] 1

[−participant] 3

[−author] 2

Both in Harbour’s (2016) original proposal and in Cowper and Hall’s adaptation of it, the difference between three- and four-way person systems depends on the ordering of the features; for Cowper and Hall, this ordering represents scope in a contrastive hierarchy. If Mi’gmaq has the separate number and person hierarchies shown in (12), then there is not much underspecification of lexical items: [±minimal] is

contrast in syntax and contrast in phonology 267 unspecified on singulars, but all other features are specified. In the vocabulary items that spell out the pronoun paradigm, [±minimal] is entirely unspecified, so that the dual–plural contrast is not marked. If, on the other hand, person and number were combined into a single contrastive hierarchy in Mi’gmaq, the gap created by the logical incompatibility of [+atomic] with the inclusive could lead to further underspecification in the lexicon. (14) shows what would happen if person features were to take scope over number in a combined hierarchy. (To save space, I omit [±minimal] from this figure, as the contrast between dual and plural is not crucial to the empirical phenomena or theoretical questions at issue here.) (14)

Combined contrastive hierarchy with person ≫ number φ [−author]

[+author] [−part] [+atomic] 1 excl sg

[−atomic] 1 excl du/pl

[+part] 1 incl

[+part] [+atomic] 2 sg

[−atomic] 2 du/pl

[−part] [+atomic] 3 sg

[−atomic] 3 du/pl

Given this hierarchy, the first person inclusive would be unspecified for number. My proposal that separate dimensions of contrast constitute separate contrastive hierarchies predicts that such a system should not exist. The Mi’gmaq pronoun paradigm is equally compatible with the specifications in (12) and (14), as the vocabulary items appear to be largely synthetic, rather than spelling out person and number features in a transparently separate way. However, if first person inclusive truly lacked a number specification, this could be apparent in agreement patterns; for example, we might expect to find a system in which the analogue of Mi’gmaq ginu triggered default singular agreement on verbs, while all other non-singular pronouns triggered nonsingular (plural, dual, or syncretic dual/plural) agreement. I predict that this pattern should not be attested, and I am not aware of any language in which it occurs. In the specific case of Mi’gmaq, Coon and Bale’s (2015) analysis of transitive agreement patterns provides clear empirical evidence against the combined hierarchy in (14), and (as discussed in more detail below) relies theoretically on treating person and number as distinct dimensions even when they probe together. In transitive verbs with a third-person subject and a first- or second-

268 daniel currie hall person non-singular object, the verb is marked with the suffix -ugsi (followed by object agreement), as illustrated in (15).11 (15)

Mi’gmaq verbs with third-person subject and non-singular first- or second-person object a. Mu nem-ugsi-w-gw. neg see-3>part.pl-neg-1incl ‘He doesn’t see usincl .’ b. Mu nem-ugsi-w-eg. neg see-3>part.pl-neg-1excl ‘He doesn’t see usexcl .’ c. Mu nem-ugsi-w-oq. neg see-3>part.pl-neg-2pl ‘He doesn’t see youpl .’ (Coon and Bale 2015: 90, 92)

If the object is singular, -ugsi does not appear, and its slot is instead occupied by a suffix that indicates the person of the object, as in (16) (Coon and Bale 2015: 89, 91, 92). (16)

Mu nem-i’li-w-g. neg see-1obj-neg-3 ‘She doesn’t see me.’

The presence of -ugsi with first-person inclusive objects, as in (15a), indicates that [−atomic] is indeed specified on the inclusive plural, and thus rules out (14) as a possible contrastive hierarchy for Mi’gmaq person and number. However, a combined hierarchy would not have to have the scope shown in (14). If number features were given scope over person features, the result would be as shown in (17).

11 Following Coon and Bale (2015), I gloss ugsi as 3>part.pl for ‘third-person subject acting on plural discourse-participant object’. Their examples all include negation because the corresponding affirmative forms have additional complications resulting from phonological or allomorphic interactions between the two agreement suffix positions. I have quoted their translations unaltered, including the use of gendered English third-person pronouns; as Mi’gmaq marks animacy but not gender, instances of ‘he’ and ‘she’ in the translations might better be understood as roughly equivalent to English singular they.

contrast in syntax and contrast in phonology 269 (17)

Combined contrastive hierarchy with number ≫ person φ [+atomic]

[+author] 1 sg

[−atomic]

[−author]

[+part] 2 sg

[−part] 3 sg

[+author] [−part] 1 excl pl

[+part] 1 incl pl

[−author] [+part] 2 pl

[−part] 3 pl

In this hierarchy, [−participant] would be unspecified on first person singular. Whether the absence of this specification would have any visible effect would, of course, depend on whether any vocabulary items made crucial reference to the negative value of this feature. (17) is thus less susceptible to empirical refutation than (14), and harder to distinguish from the specifications assigned by the separate hierarchies in (12). However, Coon and Bale’s (2015) analysis of Mi’gmaq agreement patterns provides an independent rationale for treating person and number as separate dimensions. When the subject is a discourse participant and the object is another (necessarily distinct) discourse participant, the first agreement suffix position (the one occupied by -ugsi in (15)) reflects the person of the object. The second slot agrees with either the subject or the object, depending on their features. It preferentially agrees with a first-person (exclusive12) plural argument in either position; failing that, with a second-person plural; failing that, with the subject. (18)

Mi’gmaq agreement with discourse participants a. Mu nem-i’li-w-eg. neg see-1obj-neg-1excl ‘You don’t see usexcl .’ b. Mu nem-u’ln-u-oq. neg see-2obj-neg-2pl ‘I don’t see youpl .’ c. Mu nem-u’ln-u-eg. neg see-2obj-neg-1excl ‘Weexcl don’t see yousg/pl .’ (Coon and Bale 2015: 92)

12 First-person inclusives cannot participate in this hierarchy at all, because the referents of the subject and object cannot overlap; an inclusive can only occur with a third-person argument.

270 daniel currie hall In (18a), the first-person plural object wins out over the second-person subject, while in (18b), the second-person plural object wins out over the firstperson singular subject; as a result, neither of these has overt subject agreement at all. In (18c), the subject is first-person plural, and so it takes precedence over a second-person object of either number. Coon and Bale account for this pattern by positing that the second agreement slot realizes a syntactic head in which person and number features have fused into a single probe. For the person part of the probe, first person counts as a better match than second person. Coon and Bale represent this using a system of monovalent features in which the probe is specified for [speaker, participant]; first persons are a better match because they have both of these features, whereas second persons have only one of them.13 However, in order to win out, a possible goal in object position must also be specified as plural. Coon and Bale express this numerically, according to what proportion of the probe’s features the potential goal matches: on the person side, [speaker, participant] scores as 1, [participant] alone as 0.5, and third person as 0; on the number side, non-singular scores as 1 and singular as 0. The two scores are then multiplied together. Crucially, then, a first-person singular ends up with a score of 0, and thus does not take precedence over a second-person plural, which scores 0.5. If person and number features were treated as a single bundle, then a first-person singular and a second-person plural might be expected to tie, with each bearing two of the three specifications being sought by the probe ([speaker, participant] versus [participant, > 1] in Coon and Bale’s feature system). Treating person and number as separate categories enables Coon and Bale to account for the fact that the person asymmetry applies only in the context of plural number. Though this does not speak directly to the question of how morphosyntactic objects are organized into inventories for the purposes of contrastive specification, it offers an independent motivation for saying that the grammar treats different dimensions of contrast separately.

9.6 Conclusions This preliminary proposal about how morphosyntactic features are assigned suggests both a particular view of the architecture of the grammar and some new questions to ask about how that architecture works. In the theoretical

13 I set aside for future work exactly how this mechanism might be adapted to use the binary person feature system of Harbour (2016) and Cowper and Hall (2019).

contrast in syntax and contrast in phonology 271 framework pursued here, the abstract, discrete computational system at and near the top of the arch in Figure 9.1 operates on inventories of contrasting combinations of distinctive features. The phonemic inventory is one such inventory, built out of features that are abstractions of phonetic properties. The morphosyntax, on the other hand, comprises several such inventories, some built with features that have identifiable interpretations on the conceptual end of the arch (such as person or number), and some with features that seem to be purely internal to the grammar (such as case or noun class). Each of these morphosyntactic inventories represents a single dimension of contrasts, rather than a multidimensional space like the phonetic space occupied by phonemes. Sometimes, though, it is not entirely obvious how many dimensions there are, or which dimension is responsible for the contrast between two items. For example, in languages that use the same morphology to express both number and honorification, as in the French tu/vous distinction or the Czech ty/vy, is the politeness distinction separately encoded in the grammar, or is there a single formal feature with two different pragmatic interpretation, as in Ritter and Wiltschko’s (2019) recycling analysis? Is vocative case really a case, or is it a form of second-person marking? Approaches to contrast developed in phonology offer promising ways of analysing patterns in morphology and syntax, but first one must identify what is to be contrasted.

Acknowledgements I am grateful to Bronwyn Bjorkman, two anonymous reviewers, and participants in the Contrast in Syntax workshop at the University of Toronto for their helpful comments and questions on earlier versions of this work. Like all my work in morphosyntax (and outside of it, for that matter), this chapter also owes a great intellectual debt to my long-standing collaboration and friendship with Elizabeth Cowper. Thank you!

References Abbott, Clifford. 2006. Oneida. In Keith Brown (ed.), Encyclopedia of language & linguistics, 30–3. Amsterdam: Elsevier, 2nd edn. doi:10.1016/B0-08-044854-2/05020-3 Abney, Steven Paul. 1987. The English noun phrase in its sentential aspect. PhD thesis, Massachusetts Institute of Technology. Adger, David. 1994. Functional heads and interpretation. PhD thesis, University of Edinburgh. Adger, David. 1996. Aspect, agreement and measure phrases in Scottish Gaelic. In Robert D. Borsley and Ian Roberts (eds.), The syntax of the Celtic languages, 200–22. Cambridge: Cambridge University Press. doi:10.1017/CBO9780511586279.007 Adger, David. 2003. Core syntax: A minimalist approach. Oxford: Oxford University Press. Adger, David, and Peter Svenonius. 2011. Features in minimalist syntax. In Cedric Boeckx (ed.), The Oxford handbook of linguistic minimalism, 27–51. Oxford: Oxford University Press. Alexiadou, Artemis. 2001a. Adjective syntax and noun raising: Word order asymmetries in the DP as the result of adjective distribution. Studia Linguistica 55(3): 217–48. Alexiadou, Artemis. 2001b. Functional structure in nominals: Nominalization and ergativity. Amsterdam: John Benjamins. Alexiadou, Artemis. 2003. Adjective syntax and (the absence of) noun raising in the DP. In Anoop Mahajan (ed.), Syntax at sunset: Head movement and syntactic theory, 1–39. Los Angeles: UCLA. Alexiadou, Artemis, and Florian Schäfer. 2010. There-insertion: An unaccusativity mismatch at the syntax–semantics interface. In Mary Byram Washburn, Sarah Ouwayda, Chuoying Ouyang, Bin Yin, Canan Ipek, Lisa Marston, and Aaron Walker (eds.), WCCFL 28 online proceedings, 9 pp. Los Angeles: University of Southern California. Published online at http://sites.google.com/site/wccfl28pro/alexiadou-schaefer. Alexiadou, Artemis, and Chris Wilder. 1998. Adjectival modification and multiple determiners. In Artemis Alexiadou and Chris Wilder (eds.), Possessors, predicates and movement in the determiner phrase, 303–32. Amsterdam: John Benjamins. Androutsopoulou, Antonia. 1996. The licensing of adjectival modification. In José Camacho, Lina Choueiri, and Maki Watanabe (eds.), WCCFL 14: The proceedings of the fourteenth West Coast Conference on Formal Linguistics, 17–31. Stanford, CA: CSLI Publications. Anisman, Adar. 2017. A modal analysis of the Tłı  ˛cho  ˛ future. Presented at the NorthWest Linguistic Conference, University of British Columbia. Anisman, Adar. 2019. When if is when and when is then: The particle nı  ˛dè in Tłı  ˛cho  ˛. PhD thesis, University of Victoria. Arad, Maya. 2003. Locality constraints on the interpretation of roots: The case of Hebrew denominal verbs. Natural Language & Linguistic Theory 21(4): 737–78. Archangeli, Diana. 1988. Underspecification in phonology. Phonology 5(2): 183–207. Aronson, Howard I. 1990. Georgian: A reading grammar. Columbus, OH: Slavica, corrected edn. Bach, Emmon. 1986. The algebra of events. Linguistics and Philosophy 9(1): 5–16.

274 references Baker, Mark, and Lisa Travis. 1997. Mood as verbal definiteness in a ‘tenseless’ language. Natural Language Semantics 5(3): 213–69. Baker, Mark, and Lisa Demena Travis. 1998. Events, times, and Mohawk verbal inflection. Canadian Journal of Linguistics/Revue canadienne de linguistique 43(2): 149–203. Baker, Mark, 2008. The syntax of agreement and concord. Cambridge: Cambridge University Press. Bar-el, Leora, Henry Davis, and Lisa Matthewson. 2005. On non-culminating accomplishments. In Leah Bateman and Cherlon Ussery (eds.), NELS 35: Proceedings of the thirtyfifth annual meeting of the North East Linguistic Society, 87–102. Amherst, MA: GLSA. Barnes, Michael P. 1986. Subject, nominative and oblique case in Faroese. Scripta Islandica 37: 13–46. Barrie, Michael. 2015. A grammar of Onondaga. Munich: LINCOM Europa. Bat-El, Outi. 1986. Extraction in Modern Hebrew morphology. Master’s thesis, University of California, Los Angeles. Béjar, Susana. 2003. Phi-syntax: A theory of agreement. PhD thesis, University of Toronto. Béjar, Susana, and Daniel Currie Hall. 2000. Marking markedness: The underlying order of diagonal syncretisms. In Rebecca Daly and Anastasia Riehl (eds.), Proceedings of the Eastern States Conference on Linguistics (ESCOL) 1999, 1–12. Ithaca, NY: Cornell Linguistics Circle Publications. Béjar, Susana, and Milan Rezac. 2009. Cyclic agree. Linguistic Inquiry 40(1): 35–73. Belletti, Adriana. 1981. Frasi ridotte assolute. Visita di grammatica generativa 6: 3–32. Bernstein, Judy B., and Christina Tortora. 2005. Two types of possessive forms in English. Lingua 115(9): 1221–42. Bjorkman, Bronwyn M., and Elizabeth Cowper. 2013. Inflectional shells and the syntax of causative have. In Shan Luo (ed.), Proceedings of the 2013 annual conference of the Canadian Linguistic Association. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2013-proceedings/ Bjorkman, Bronwyn M., Elizabeth Cowper, and Daniel Siddiqi. 2018. Unifying -ing without participles. Presented at GLOW 41, Budapest. Handout available online at http://www. dropbox.com/s/9keggagzktgaa2c/Bjorkman-Cowper-Siddiqi-handout.pdf. Bjorkman, Bronwyn M., and Hedde Zeijlstra. 2019. Checking up on (ϕ)-Agree. Linguistic Inquiry 50(3): 527–69. doi:10.1162/ling_a_00319 Black, Ronald. 2006. Cothrom ionnsachaidh: Scottish Gaelic grammar and exercises. Peebles, UK: Ronald Black, 9th edn. Blaho, Sylvia. 2008. The syntax of phonology: A radically substance-free approach. PhD thesis, Universitetet i Tromsø. Bliss, Heather. 2007. Object agreement in Blackfoot: Sentient and non-sentient controllers. In H. C. Wolfart (ed.), Papers of the 38th Algonquian Conference, 11–28. Winnipeg: University of Manitoba. ′ Bliss, Heather. 2008. Structuring information in Blackfoot: Against the A -agreement analysis of cross-clausal agreement. In Susie Jones (ed.), Proceedings of the 2008 annual conference of the Canadian Linguistic Association. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2008-proceedings/ Bliss, Heather. 2010. Argument structure, applicatives and animacy in Blackfoot. In Heather Bliss and Raphael Girard (eds.), Proceedings of the 13th and 14th Workshop on the Structure and Constituency of Languages of the Americas. Vancouver: UBC Department of Linguistics. Bliss, Heather. 2013. The Blackfoot configurationality conspiracy: Parallels and differences in clausal and nominal structures. PhD thesis, University of British Columbia.

references 275 Bliss, Heather, and Elizabeth Ritter. 2015. A direct evidential in Siksiká Blackfoot. In Jérémy Pasquereau (ed.), SULA 8: Proceedings of the 8th meeting on the Semantics of UnderRepresented Languages in the Americas, 1–16. Amherst, MA: GLSA. Blix, Hagen. 2016. South Caucasian agreement: A Spanning account. Master’s thesis, University of Vienna. Bloomfield, Leonard. 1946. Algonquin. In Harry Hoijer (ed.), Linguistic structures of Native America, 85–129. New York: Viking Fund. Bobaljik, Jonathan David, and Andrew Carnie. 1996. A Minimalist approach to some problems of Irish word order. In R. D. Borsley and I. Roberts (eds.), The syntax of the Celtic languages, 223–40. Cambridge: Cambridge University Press. Bobaljik, Jonathan David. 2002. A-chains at the PF interface: Copies and ‘covert’ movement. Natural Language & Linguistic Theory 20(2): 197–267. Bobaljik, Jonathan David, and Höskuldur Thráinsson. 1998. Two heads aren’t always better than one. Syntax 1(1): 37–71. Boeckx, Cedric. 2008. Aspects of the syntax of agreement. New York: Routledge. Boeckx, Cedric. 2009. On long-distance Agree. Iberia 1: 1–32. Bogal-Allbritten, Elizabeth. 2013. Decomposing notions of adjectival transitivity in Navajo. Natural Language Semantics 21(3): 277–314. Bogal-Allbritten, Elizabeth. 2016. Building meaning in Navajo. PhD thesis, University of Massachusetts, Amherst. Bohnemeyer, Jürgen, and Mary Swift. 2004. Event realization and default aspect. Linguistics and Philosophy 27(3): 263–96. Bonet, Eulàlia. 1991. Morphology after syntax: Pronominal clitics in Romance. PhD thesis, Massachusetts Institute of Technology. Bonet, Eulàlia. 1995. Feature structure of romance clitics. Natural Language and Linguistic Theory 13(4): 607–47. Borer, Hagit. 1984. Parametric syntax: Case studies in Semitic and Romance languages. Studies in Generative Grammar. Dordrecht: Foris. Borer, Hagit. 2003. Exo-skeletal vs. endo-skeletal explanations: Syntactic projections and the lexicon. In John Moore and Maria Polinsky (eds.), The nature of explanation in linguistic theory, 32–67. Stanford, CA: CSLI Publications. Borer, Hagit. 2005. In name only. Oxford: Oxford University Press. Bošković, Željko. 1997. The syntax of nonfinite complementation: An economy approach. Cambridge, MA: MIT Press. Bošković, Željko. 2007. On the locality and motivation of Move and Agree: An even more minimal theory. Linguistic Inquiry 38(4): 589–644. doi:10.1162/ling.2007.38.4.589 Bradley, James W. 2005. Evolution of the Onondaga Iroquois: Accommodating change, 1500–1655. Lincoln, NE: University of Nebraska Press. Bresnan, Joan, Anna Cueni, Tatiana Nikitina, and Harald Baayen. 2007. Predicting the dative alternation. In Gerlof Bourne, Irene Krämer, and Joost Zwarts (eds.), Cognitive foundations of interpretation, 69–94. Amsterdam: Royal Netherlands Academy of Science. Brown, Cynthia A. 1997. Acquisition of segmental structure: Consequences for speech perception and second language acquisition. PhD thesis, McGill University. Bruening, Benjamin. 2010. Double object constructions disguised as prepositional datives. Linguistic Inquiry 41(2): 287–305. Brugger, Gerhard. 1993. Generic interpretations and expletive determiner. University of Venice Working Papers in Linguistics 3(1): 1–30.

276 references Burt, Marina K. 1971. From deep to surface structure: An introduction to transformational syntax. New York: Harper and Row. Butt, Miriam. 2013. Control vs. complex predication: Identifying non-finite complements. Natural Language & Linguistic Theory 32(1): 165–90. Byrne, Michel. 2002. Gràmar na gàidhlig. Isle of Lewis: Stòrlann-Acair. Cable, Seth. 2013. Beyond the past, present, and future: Towards the semantics of ‘graded tense’ in Gĩkũyũ. Natural Language Semantics 21(3): 219–76. doi:10.1007/ s11050-012-9092-3 Caha, Pavel. 2009. The nanosyntax of case. PhD thesis, Universitetet i Tromsø. Campbell, Robert, and Barbara Campbell. 2009. Jamamadí preliminary phonemic statement. Anápolis, GO: Associac  ¸ão Internacional de Linguística SIL. Campos, Héctor, and Melita Stavrou. 2004. Polydefinite constructions in Modern Greek and in Aromanian. In Olga Mišeska Tomić (ed.), Balkan syntax and semantics, no. 67 in Linguistik Aktuell, 136–73. Amsterdam: John Benjamins. Cardinaletti, Anna. 2004. Toward a cartography of subject positions. In Luigi Rizzi (ed.), The structure of CP and IP, vol. 2 of The Cartography of Syntactic Structures, 116–65. Oxford: Oxford University Press. Carlson, Gregory N. 1977. Reference to kinds in English. PhD thesis, University of Massachusetts, Amherst. Carlson, Gregory N., and Francis Jeffry Pelletier (eds.). 1995. The generic book. Chicago: University of Chicago Press. Carnie, Andrew. 1995. Non-verbal predication and head-movement. PhD thesis, Massachusetts Institute of Technology. Carnie, Andrew. 2012. Syntax: A generative introduction. Oxford: Wiley-Blackwell, 3rd edn. Ceong, Hailey H. 2011. The syntax of Korean polar alternative questions: A-not-A. Master’s thesis, University of Victoria. Ceong, Hailey H. 2019. The morphosyntax of clause typing: single, double, periphrastic, and multifunctional complementizers in Korean. PhD thesis, University of Victoria. Ceong, Hailey H., and Leslie Saxon. 2013. Features of questions and interrogatives. In Shan Luo (ed.), Proceedings of the 2013 annual conference of the Canadian Linguistic Association, 15 pp. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2013-proceedings/ Chafe, Wallace L. 1960a. Seneca morphology I: Introduction. International Journal of American Linguistics 26(1): 11–22. Chafe, Wallace L. 1960b. Seneca morphology II: Irreducible verbs and nouns; the modal verb. International Journal of American Linguistics 26(2): 123–9. Chafe, Wallace L. 1960c. Seneca morphology III: Expanded pronominal prefixes. International Journal of American Linguistics 26(3): 224–33. Chafe, Wallace L. 1960d. Seneca morphology IV: Expanded bases. International Journal of American Linguistics 26(4): 283–9. Chafe, Wallace L. 1961. Seneca morphology V: Expanded aspect suffixes; attributive suffixes. International Journal of American Linguistics 27(1): 42–5. Chafe, Wallace L. 1980. Consequential verbs in the Northern Iroquoian languages and elsewhere. In Kathryn Klar, Margaret Langdon, and Shirley Silver (eds.), American Indian and Indoeuropean studies: Papers in honor of Madison S. Beeler, no. 16 in Trends in Linguistics: Studies and Monographs, 43–9. The Hague: Mouton. Chandra, Pritha. 2011. (Dis)agree: Exploring agreement mechanisms. Newcastle upon Tyne: Cambridge Scholars Publishing.

references 277 Cheng, Lisa Lai-Shen. 1991. On the typology of Wh-questions. PhD thesis, Massachusetts Institute of Technology. Cheng, Lisa Lai-Shen. 1997. On the typology of Wh-questions. New York: Garland. Cheng, Lisa Lai-Shen, and Rint Sybesma. 1999. Bare and not-so-bare nouns and the structure of NP. Linguistic Inquiry 30(4): 509–42. Cherry, E. Colin, Morris Halle, and Roman Jakobson. 1953. Toward the logical description of languages in their phonemic aspect. Language 29(1): 34–46. Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6(4): 339–405. Chomsky, Noam. 1964. Current issues in linguistic theory. The Hague: Mouton. Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press. Chomsky, Noam. 1970. Remarks on nominalization. In Roderick Jacobs and Peter Rosenbaum (eds.), Readings in English transformational grammar, 184–221. Waltham, MA: Blaisdell. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. Chomsky, Noam. 1986. Knowledge of language: Its nature, origin, and use. New York: Praeger. Chomsky, Noam. 1991. Some notes on economy of derivation and representation. In Robert Freidin (ed.), Principles and parameters in comparative syntax, 417–54. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka (eds.), Step by step: Essays on Minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, MA: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz (ed.), Ken Hale: A life in language, 1–52. Cambridge, MA: MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Adriana Belletti (ed.), Structures and beyond, vol. 3 of The Cartography of Syntactic Structures, 104–31. Oxford: Oxford University Press. Chomsky, Noam. 2007. Approaching UG from below. In Uli Sauerland and Hans-Martin Gärtner (eds.), Interfaces + recursion = language? Chomsky’s Minimalism and the view from syntax–semantics, 1–29. Berlin: Mouton de Gruyter. Chomsky, Noam. 2008. On phases. In Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizarreta (eds.), Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud, 133–66. Cambridge, MA: MIT Press. Chomsky, Noam, and Morris Halle. 1968. The sound pattern of English. New York: Harper and Row. Christophersen, Paul. 1939. The articles: A study of their theory and use in English. Copenhagen: Munksgaard. Chung, Sandra. 2017. Another way around causatives in Chamorro. In Claire Bowern, Laurence Horn, and Raffaella Zanuttini (eds.), On looking into words (and beyond): Structures, relations, analyses, 263–88. Berlin: Language Science Press. Chung, Sandra, and William A. Ladusaw. 2004. Restriction and saturation. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1993. A null theory of phrasal and compound stress. Linguistic Inquiry 24(2): 239–97. Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford: Oxford University Press. Cinque, Guglielmo, and Luigi Rizzi. 2008. The cartography of syntactic structures. CISCL Working Papers 2: 42–58.

278 references Clarke, Sarah. 2009. The manifestation of viewpoint aspect in Inuktitut. Studia Linguistica 63(3): 292–322. Clarke, Sarah. 2013. Aspectual scope and contrast in English and Japanese. PhD thesis, University of Toronto. Clements, G. N. 1988. Towards a substantive theory of feature specification. NELS 18: 79–93. Clements, G. N. 2003. Feature economy in sound systems. Phonology 20(3): 287–333. Clements, G. N. 2009. The role of features in phonological inventories. In Eric Raimy and Charles E. Cairns (eds.), Contemporary views on architecture and representations in phonology, 19–68. Cambridge, MA: MIT Press. Clements, G. N., and Elizabeth Hume. 1995. The internal organization of speech sounds. In John Goldsmith (ed.), The handbook of phonological theory, 245–306. Oxford: Blackwell. Collins, Chris, and Höskuldur Thráinsson. 1996. VP-internal structure and object shift in Icelandic. Linguistic Inquiry 27(3): 391–444. Compton, Richard, and B. Elan Dresher. 2011. Palatalization and ‘strong i’ across Inuit dialects. Canadian Journal of Linguistics 56(2): 203–28. Comrie, Bernard. 1976. Aspect: An introduction to the study of verbal aspect and related problems. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press. Comrie, Bernard. 1985. Tense. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press. doi:10.1017/CBO9781139165815 Cook, Eung-Do. 1984. A Sarcee grammar. Vancouver: UBC Press. Cook, Eung-Do. 2004. A grammar of Dëne Su  ˛’liné (Chipewyan). Winnipeg: Algonquian and Iroquoian Linguistics. Cook, Eung-Do. 2013. A Tsilhqut’in grammar. Vancouver: UBC Press. Coon, Jessica, and Alan Bale. 2015. The interaction of person and number in Mi’gmaq. Nordlyd 40(1): 85–101. Copley, Bridget, and Heidi Harley. 2009. Futurates, directors and have-causatives. Snippets 19(1): 5–6. Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press. Corbett, Greville G. 2000. Number. Cambridge: Cambridge University Press. Cowper, Elizabeth. 1988. What is a subject? Non-nominative subjects in Icelandic. In James Blevins and Juli Carter (eds.), Proceedings of NELS 18, 94–108. Amherst, MA: GLSA. Cowper, Elizabeth. 1992. A concise introduction to syntactic theory: The Government– Binding approach. Chicago: University of Chicago Press. Cowper, Elizabeth. 1995. English participle constructions. Canadian Journal of Linguistics 40(1): 1–38. Cowper, Elizabeth. 1998. The simple present tense in English: A unified treatment. Studia Linguistica 52(1): 1–18. Cowper, Elizabeth. 1999. Feature geometry and verbal inflection. Toronto Working Papers in Linguistics 17: 79–96. Cowper, Elizabeth. 2003. Tense, mood and aspect: A feature-geometric approach. Ms. University of Toronto. Cowper, Elizabeth. 2005a. The geometry of interpretable features: Infl in English and Spanish. Language 81(1): 10–46. Cowper, Elizabeth. 2005b. A note on number. Linguistic Inquiry 36(3): 441–55. Cowper, Elizabeth. 2010. Where auxiliary verbs come from. In Melinda Heijl (ed.), Proceedings of the 2010 annual conference of the Canadian Linguistic Association. Toronto. Published online at http://cla-acl.ca/actes-2010-proceedings/

references 279 Cowper, Elizabeth, and Daniel Currie Hall. 1999. Semantic composition and syntactic structure: English inflection. Presented at the 1999 annual meeting of the Canadian Linguistic Association, Université de Sherbrooke. Cowper, Elizabeth, and Daniel Currie Hall. 2005. The pieces of π. In Marie-Odile Junker, Martha McGinnis, and Yves Roberge (eds.), Proceedings of the 2004 annual conference of the Canadian Linguistic Association. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2004-proceedings/ Cowper, Elizabeth, and Daniel Currie Hall. 2007. The morphosyntactic manifestations of modality. In Milica Radišić (ed.), Proceedings of the 2007 annual conference of the Canadian Linguistic Association. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2007-proceedings/ Cowper, Elizabeth, and Daniel Currie Hall. 2009. Argumenthood, pronouns, and nominal feature geometry. In Jila Ghomeshi, Ileana Paul, and Martina Wiltschko (eds.), Determiners: Universals and variation, 97–120. Amsterdam: John Benjamins. Cowper, Elizabeth, and Daniel Currie Hall. 2012. Aspects of individuation. In Diane Massam (ed.), Count and mass across languages, 27–53. Oxford: Oxford University Press. Cowper, Elizabeth, and Daniel Currie Hall. 2013. Syntactic change and the cartography of syntactic structures. In Stefan Keine and Shayne Sloggett (eds.), NELS 42: Proceedings of the forty-second annual meeting of the North East Linguistic Society, vol. 1, 129–40. Amherst, MA: GLSA. Cowper, Elizabeth, and Daniel Currie Hall. 2014a. Reductiō ad discrīmen: Where features come from. Nordlyd 41(2): 145–64. Published online at http://septentrio.uit.no/index. php/nordlyd/article/view/3411 Cowper, Elizabeth, and Daniel Currie Hall. 2014b. The features and exponence of nominal number. Lingue e Linguaggio 13(1): 63–82. Cowper, Elizaeth, and Daniel Currie Hall. 2014c. Morphosyntactic features and the scope of contrast. Presented at the 38th meeting of the Atlantic Provinces Linguistic Association (APLA 38), University of New Brunswick, Fredericton, November 2014. Cowper, Elizabeth, and Daniel Currie Hall. 2017. The rise of contrastive modality in English: A neoparametric account. Linguistic Variation 17(1): 67–96. doi:10.1075/lv.17.1.04cow Cowper, Elizabeth, and Daniel Currie Hall. 2019. Scope variation in contrastive hierarchies of morphosyntactic features. In David W. Lightfoot and Jon Havenhill (eds.), Variable properties in language: Their nature and acquisition, 27–41. Washington: Georgetown University Press. Cowper, Elizabeth, Daniel Currie Hall, Bronwyn Bjorkman, Rebecca Tollan, and Neil Banerjee. 2015. There’s no future in Old English. In Santa Vīnerte (ed.), Proceedings of the 2015 annual conference of the Canadian Linguistic Association, 15 pp. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2015-proceedings/ Cox, Christopher. 2012. Structuring stories: Personal and traditional narrative styles in Tsuut’ina. In James A. Crippen, Christopher Cox, Richard Dauenhauer, and Nora Marks Dauenhauer (eds.), Working papers in Athabaskan languages 2011, chap. 1. Fairbanks, AK: Alaska Native Language Center. Craig, Collette. 1977. The structure of Jacaltec. Austin: University of Texas Press. Dahl, Östen. 1985. Tense and aspect systems. Oxford: Basil Blackwell. D’Alessandro, Roberta, and Adam Ledgeway. 2010. At the C-T boundary: Investigating Abruzzese complementation. Lingua 120(8): 2040–60. Davison, Alice. 2013. Non-finite complements and modality in de-na ‘allow’ in Hindi-Urdu. Natural Language & Linguistic Theory 32(1): 137–64.

280 references Deal, Amy Rose. 2009. The origin and content of expletives: Evidence from ‘selection’. Syntax 12(4): 285–323. doi:10.1111/j.1467-9612.2009.00127.x Declerck, Renaat. 1989. Boundedness and the structure of situations. Leuvense Bijdragen 78: 275–308. Denison, David. 1998. Syntax. In Suzanne Romaine (ed.), The Cambridge history of the English language, vol. 4: 1776–1997, 92–329. Cambridge: Cambridge University Press. den Dikken, Marcel. 2006. Relators and linkers: The syntax of predication, predicate inversion, and copulas. Cambridge, MA: MIT Press. Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press. Dogrib Translation Committee. 2003, 2008. Nò  ˛htsı  ˛ Nı  ˛htł’è, Zezì wegò  ˛hłı  ˛ tł’axo  ˛ò  ˛. Toronto: Canadian Bible Society. Available online at http://www.bible.com/versions/ 800-dognt-nohtsi-nihtle. Accessed 19 June 2017. Doron, Edit. 2003. Agency and voice: The semantics of the Semitic templates. Natural Language Semantics 11(1): 1–67. Dowty, David R. 1977. Toward a semantic analysis of verb aspect and the English ‘imperfective’ progressive. Linguistics and Philosophy 1(1): 45–77. Dresher, B. Elan. 2009. The contrastive hierarchy in phonology. Cambridge: Cambridge University Press. Dresher, B. Elan. 2011. The phoneme. In Marc van Oostendorp, Colin J. Ewen, Elizabeth Hume, and Keren Rice (eds.), The Blackwell companion to phonology, vol. 1, 241–66. Oxford: Wiley-Blackwell. Dresher, B. Elan. 2014. The arch not the stones: Universal feature theory without universal features. Nordlyd 41(2): 165–81. Dresher, B. Elan. 2015. The motivation for contrastive feature hierarchies in phonology. Linguistic Variation 15(1): 1–40. Dresher, B. Elan, and Daniel Currie Hall. forthcoming. Developments leading toward generative phonology. In B. Elan Dresher and Harry van der Hulst (eds.), The Oxford history of phonology, chap. 17. Oxford: Oxford University Press. Dresher, B. Elan, and Norbert Hornstein. 1979. Trace theory and NP movement rules. Linguistic Inquiry 10(1): 65–82. Dresher, B. Elan, Glyne L. Piggott, and Keren D. Rice. 1994. Contrast in phonology: Overview. Toronto Working Papers in Linguistics 13: iii–xvii. Dryer, Matthew S. 1986. Primary objects, secondary objects, and antidative. Language 62(4): 808–45. Dunham, Joel. 2007. The ‘durative’ in Blackfoot: Understanding imperfectivity. In Amy Rose Deal (ed.), Proceedings of SULA 4: Semantics of under-represented languages in the Americas, no. 35 in University of Massachusetts Occasional Papers in Linguistics, 49–64. Amherst, MA: GLSA. Dvořák, Věra. 2011. Inherent case and locality requirement: Evidence from ditransitives and their nominalizations. University of Pennsylvania Working Papers in Linguistics 17(1): 95–104. Embick, David. 1997. Voice and the interfaces of syntax. PhD thesis, University of Pennsylvania. Embick, David. 2010. Localism versus globalism in morphology and phonology. Cambridge, MA: MIT Press. Embick, David, and Alec Marantz. 2008. Architecture and blocking. Linguistic Inquiry 39(1): 1–53. Fernald, Theodore B. 2000. Generalizations in Navajo. In Theodore B. Fernald and Paul R. Platero (eds.), The Athabaskan languages: Perspectives on a Native American language family, 51–72. Oxford: Oxford University Press.

references 281 Foley, Steven. 2016. Case–agreement mismatches in Georgian. Poster presented at the South Caucasian Chalk Circle (SCCC), Paris. Folli, Rafaella, and Heidi Harley. 2007. Causation, obligation, and argument structure: On the nature of little v. Linguistic Inquiry 38(2): 197–238. Foster, Michael K. 1985. The language of tense, mood, and aspect in Northern Iroquoian descriptions. International Journal of American Linguistics 51(4): 403–5. Foster, Michael K. 1986. Updating the terminology of tense, mood, and aspect in Northern Iroquoian descriptions. International Journal of American Linguistics 52(1): 65–72. Frantz, Donald G. 2017. Blackfoot grammar. Toronto: University of Toronto Press, 3rd edn. Frantz, Donald G., and Norma Jean Russell. 2017. Blackfoot dictionary of stems, roots, and affixes. Toronto: University of Toronto Press, 3rd edn. Frege, Gottlob. 1997 [1892]. On sense and reference. In Peter Ludlow (ed.), Readings in the philosophy of language, 563–83. Cambridge, MA: MIT Press. Froman, Frances, Alfred Keye, Carrie Joan Dyck, and Lottie Keye. 2002. English– Cayuga/Cayuga–English dictionary. Toronto: University of Toronto Press. Genee, Inge, and Marie-Odile Junker. 2018. The Blackfoot language resources and digital dictionary project: Creating integrated web resources for language documentation and revitalization. Language Documentation and Conservation 12: 274–314. George, Leland, and Jaklin Kornfilt. 1981. Finiteness and boundedness in Turkish. In Frank Heny (ed.), Binding and filtering, 105–27. London: Croom Helm. Gerdts, Donna. 1988. A nominal hierarchy in Halkomelem clausal organization. Anthropological Linguistics 30(1): 20–36. Ghomeshi, Jila, and Diane Massam. 2009. The proper D connection. In Jila Ghomeshi, Ileana Paul, and Martina Wiltschko (eds.), Determiners: Universals and variation, 67–96. Amsterdam: John Benjamins. Giannakidou, Anastasia. 2004. Domain restriction and the arguments of quantificational determiners. In Robert B. Young (ed.), Proceedings of the 14th Semantics and Linguistic Theory conference, 110–26. Ithaca, NY: Cornell University. doi:10.3765/salt.v14i0.2902 Gillon, Carrie. 2006. The semantics of determiners: Domain restriction in Skwxwú7mesh. PhD thesis, University of British Columbia. Gillon, Carrie. 2009. The semantic core of determiners: Evidence from Skwxwú7mesh. In Jila Ghomeshi, Ileana Paul, and Martina Wiltschko (eds.), Determiners: Universals and variation, 177–214. Amsterdam: John Benjamins. Givón, Talmy. 1972. Studies in ChiBemba and Bantu grammar. Los Angeles: Department of Linguistics and African Studies Center, University of California. Supplement #3 to Studies in African Linguistics. Givón, Talmy. 1976. Some constraints on Bantu causativization. In Masayoshi Shibatani (ed.), Grammar of causative constructions, no. 6 in Syntax and Semantics, 325–51. New York: Academic Press. Gleasure, James. 1986. Synthetic and analytic: Some comments on the Irish / Gaelic present / future. Scottish Gaelic Studies 14: 94–101. Gleasure, James. 1990. The evolution of the present / future tense in Scottish Gaelic. Scottish Gaelic Studies 16: 181–9. Goddard, Ives. 2002. Grammatical gender in Algonquian. In H. C. Wolfart (ed.), Papers of the 33rd Algonquian Conference, 195–231. Winnipeg: University of Manitoba. Green, Georgia M. 1974. Semantics and syntactic regularity. Bloomington, IN: Indiana University Press. Grimshaw, Jane. 1990. Argument structure. Cambridge, MA: MIT Press.

282 references Grimshaw, Jane, and Vieri Samek-Lodovici. 1995. Optimal subjects and subject universals. Presented at the workshop ‘Is the Best Good Enough? Optimality and Competition in Syntax’, Massachusetts Institute of Technology. Subsequently published as Grimshaw and Samek-Lodovici (1998). Grimshaw, Jane, and Vieri Samek-Lodovici. 1998. Optimal subjects and subject universals. In Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky (eds.), Is the best good enough? Optimality and competition in syntax, 193–219. Cambridge, MA: MIT Press. Gropen, Jess, Steven Pinker, Michelle Hollander, Richard Goldberg, and Ronald Wilson. 1989. The learnability and acquisition of the dative alternation in English. Language 65(2): 203–57. Guasti, Maria Teresa, and Luigi Rizzi. 2002. Agreement and tense as distinct syntactic positions: Evidence from acquisition. In Guglielmo Cinque (ed.), Functional structure in DP and IP, vol. 1 of The Cartography of Syntactic Structures, 167–94. Oxford: Oxford University Press. Haegeman, Liliane. 1991. Introduction to Government and Binding Theory. Oxford: Blackwell. Hale, Kenneth L., and S. Jay Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In Kenneth L. Hale and S. Jay Keyser (eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, 53–109. Cambridge, MA: MIT Press. Hall, Daniel Currie. 2001. The featural semantics of English modal verbs. Ms., University of Toronto. Hall, Daniel Currie. 2004. A formal approach to /v/: Evidence from Czech and Slovak. In Olga Arnaudova, Wayles Browne, María Luisa Rivero, and Danijela Stojanović (eds.), Formal approaches to Slavic linguistics: The Ottawa meeting 2003, 187–205. Ann Arbor, MI: Michigan Slavic Publications. Hall, Daniel Currie. 2007. The role and representation of contrast in phonological theory. PhD thesis, University of Toronto. Hall, Daniel Currie. 2012. Building parallel structures. Presented at the CASTL Decennium, Universitetet i Tromsø, September 2012. Hall, Daniel Currie. 2017. Contrastive specification in phonology. In Mark Aronoff (ed.), Oxford research encyclopedia of linguistics. Oxford: Oxford University Press. DOI: 10.1093/acrefore/9780199384655.013.26 Hall, Daniel Currie, and Elizabeth Cowper. 2016. Morphosyntactic feature hierarchies are contrastive hierarchies. Presented at Hierarchical Structures in Phonology, Morphology and Syntax, Universitetet i Tromsø, October 2016. Hall, Kathleen Currie. 2013. A typology of intermediate phonological relationships. The Linguistic Review 30(2): 215–75. Halle, Morris. 1959. The sound pattern of Russian: A linguistic and acoustical investigation. The Hague: Mouton. Halle, Morris, and Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In Kenneth Hale and Samuel Jay Keyser (eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, 111–76. Cambridge, MA: MIT Press. Harbour, Daniel. 2008. Mass, non-singularity, and augmentation. MIT Working Papers in Linguistics 49: 239–66. Harbour, Daniel. 2011. Descriptive and explanatory markedness. Morphology 21(2): 223–45.

references 283 Harbour, Daniel. 2013. ‘Not plus’ isn’t ‘not there’: Bivalence in person, number, and gender. In Ora Matushansky and Alec Marantz (eds.), Distributed morphology today: Morphemes for Morris Halle, chap. 8, 135–50. Cambridge, MA: MIT Press. Harbour, Daniel. 2014a. Paucity, abundance, and the theory of number. Language 90(1): 185–229. Harbour, Daniel. 2014b. Poor pronoun systems and what they teach us. Nordlyd 41(1): 125–43. Harbour, Daniel. 2016. Impossible persons. Cambridge, MA: MIT Press. Harbour, Daniel, and Christian Elsholtz. 2012. Feature geometry: Self-destructed. Ms., Queen Mary University of London and Technische Universität Graz. Hargus, Sharon. 2009. Causatives and transitionals in Kwadacha Tsek’ene. Presented at the Athabaskan Languages Conference, Berkeley, CA. Harley, Heidi. 1994. Hug a tree: Deriving the morphosyntactic feature hierarchy. MIT Working Papers in Linguistics 21: 289–320. Harley, Heidi. 1995. Subjects, events, and licensing. PhD thesis, Massachusetts Institute of Technology. Harley, Heidi. 2003. How do verbs get their names? Denominal verbs, manner incorporation, and the ontology of verb roots in English. In Nomi Erteschik-Shir and Tova Rapoport (eds.), The syntax of aspect: Deriving thematic aspectual interpretation, 42–64. Oxford: Oxford University Press. doi:10.1093/acprof:oso/9780199280445.003.0003 Harley, Heidi. 2009. The morphology of nominalizations and the syntax of vP. In Anastasia Giannakidou and Monika Rathert (eds.), Quantification, definiteness, and nominalization, 320–42. Oxford: Oxford University Press. Harley, Heidi. 2013. External arguments and the Mirror Principle: On the distinctness of Voice and v. Lingua 125(1): 34–57. Harley, Heidi, and Rolf Noyer. 1998. Mixed nominalizations, short verb movement and object shift in English. In Pius N. Tamanji and Kiyomi Kusumoto (eds.), Proceedings of NELS 28, vol. 1, 143–57. Amherst, MA: GLSA. Harley, Heidi, and Elizabeth Ritter. 2002a. Person and number in pronouns: A featuregeometric analysis. Language 78(3): 482–526. Harley, Heidi, and Elizabeth Ritter. 2002b. Structuring the bundle: A universal morphosyntactic feature geometry. In Horst J. Simon and Heike Wiese (eds.), Pronouns: Grammar and representation, 23–39. Amsterdam: John Benjamins. Harms, Robert T. 1973. How abstract is Nupe? Language 49(2): 439–46. Harris, Alice C. 1981. Georgian syntax: A study in Relational Grammar. Cambridge: Cambridge University Press. Harris, John, and Geoff Lindsey. 1995. The elements of phonological representation. In Jacques Durand and Francis Katamba (eds.), Frontiers of phonology, 34–79. Harlow: Longman. Harris, Zellig. 1957. Co-occurrence and transformation in linguistic structure. Language 33(3): 283–340. Hartmann, Jutta. 2008. Expletives in existentials: English there and German da. PhD thesis, Tilburg University. Hartmann, Jutta, and Caroline Heycock. 2016. Evading agreement: A new perspective on low nominative agreement in Icelandic. In Christopher Hammerly and Brandon Prickett (eds.), NELS 46: Proceedings of the forty-sixth annual meeting of the North East Linguistic Society, vol. 2, 67–80. Amherst, MA: GLSA. Hartmann, Jutta, and Caroline Heycock. 2017. Variation in copular agreement in Insular Scandinavian. In Höskuldur Thráinsson, Caroline Heycock, Hjalmar P. Petersen, and

284 references Zakaris Svabo Hansen (eds.), Syntactic variation in Insular Scandinavian, 233–76. Amsterdam: John Benjamins. Hayashi, Midori. 2011. The structure of multiple tenses in Inuktitut. PhD thesis, University of Toronto. Hayes, Bruce, Robert Kirchner, and Donca Steriade (eds.). 2004. Phonetically based phonology. Cambridge: Cambridge University Press. Heap, David. 2002. Split subject pronoun paradigms: Feature geometry and underspecification. In Teresa Satterfield, Christina Tortora, and Diana Cresti (eds.), Current issues in Romance languages: Selected papers from the 29th Linguistic Symposium on Romance Languages (LSRL), Ann Arbor, 8–11 April 1999, 129–44. Amsterdam: John Benjamins. Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. PhD thesis, University of Massachusetts, Amherst. Heim, Irene. 1988. The semantics of definite and indefinite noun phrases. New York: Garland. Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell. Hewitt, Brian George. 1970. The typology of subordination in Georgian and Abkhaz. No. 5 in Empirical Approaches to Language Typology. Berlin: Mouton de Gruyter. Hewitt, Brian George. 1991. Lexicography of the Caucasian languages I: Georgian and Kartvelian. In Rufus Gouws, Ulrich Heid, Wolfgang Schweickard, and Herbert Ernst Wiegand (eds.), Wörterbücher: Ein internationales Handbuch zur Lexikographie, vol. 3, no. 5 in Handbücher zur Sprach- und Kommunikationswissenschaft, 2415–7. Berlin: De Gruyter. Heycock, Caroline. 1995. The internal structure of small clauses: New evidence from inversion. In Jill Beckman (ed.), NELS 25: Proceedings of the twenty-fifth annual meeting of the North East Linguistic Society, vol. 1, 223–38. Amherst, MA: GLSA. Heycock, Caroline, and Anthony Kroch. 1999. Pseudocleft connectedness: Implications for the LF interface level. Linguistic Inquiry 30(3): 365–98. doi:10.1162/002438999554110 Hill, Virginia, and Gabriela Alboiu. 2016. Verb movement and clause structure in Old Romanian. Oxford Studies in Diachronic and Historical Linguistics. Oxford: Oxford University Press. Hinrichs, Erhard Walter. 1985. A compositional semantics for Aktionsarten and NP reference in English. PhD thesis, Ohio State University. Hockett, Charles F. 1955. A manual of phonology. Baltimore: Waverly Press. Hockett, Charles F. 1958. A course in modern linguistics. New York: Macmillan. Hoekstra, Teun, and René Mulder. 1990. Unergatives as copular verbs: Locational and existential predication. The Linguistic Review 7(1): 1–79. Hornstein, Norbert. 1993. As time goes by: Tense and Universal Grammar. Cambridge, MA: MIT Press. Hornstein, Norbert. 2009. A theory of syntax: Minimal operations and Universal Grammar. Cambridge: Cambridge University Press. Hornstein, Norbert, Jairo Nunes, and Kleanthes K. Grohmann. 2005. Understanding Minimalism. Cambridge: Cambridge University Press. Hróarsdóttir, Þorbjörg. 2000. Word order change in Icelandic: From OV to VO. Amsterdam: John Benjamins. Hucklebridge, Shay. 2015. The great Tłı  ˛cho  ˛ Yatıì vowel length mystery of 2015. Ms., University of Toronto. Hyman, Larry M. 1970. How concrete is phonology? Language 46(1): 58–76. Hyman, Larry M. 1973. Nupe three years later. Language 49(2): 447–52.

references 285 Iosad, Pavel. 2012. Representation and variation in substance-free phonology. PhD thesis, Universitetet i Tromsø. Iosad, Pavel. 2017. The phonologisation of redundancy: Length and quality in Welsh vowels. Phonology 34(1): 121–62. doi:10.1017/S0952675717000057 Jackendoff, Ray. 1977. X-bar syntax: A study of phrase structure. Cambridge, MA: MIT Press. Jackendoff, Ray. 1991. Parts and boundaries. Cognition 41(1): 9–45. Jakobson, Roman. 1960. Closing statement: Linguistics and poetics. In Thomas A. Sebeok (ed.), Style in language, 350–77. Cambridge, MA: MIT Press. Jaszczolt, Katarzyna M. 2009. Representing time: An essay on temporality as modality. Oxford: Oxford University Press. Johansson, Sara. 2008. The status of nominal gender in Algonquian: Evidence from psych verbs. In Susie Jones (ed.), Proceedings of the 2008 annual conference of the Canadian Linguistic Association. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2008-proceedings/ Johansson, Sara. 2009. A syntactic analysis of Blackfoot -attsi causatives. B.A. honours thesis, University of Calgary. Johnson, Marion Rose. 1977. A semantic analysis of Kikuyu tense and aspect. PhD thesis, The Ohio State University. Jónsson, Jóhannes Gísli. 1996. Clausal architecture and case in Icelandic. PhD thesis, University of Massachusetts, Amherst. Jung, Dagmar. 2002. Clause combining in Apache. In Nicholas Evans and Hans-Jürgen Sasse (eds.), Problems of polysynthesis, 167–83. Berlin: Akademie Verlag. Kadmon, Nirit. 1992. On unique and non-unique reference and asymmetric quantification. New York: Garland. Kaplan, Lawrence D. 1981. Phonological issues in North Alaskan Inupiaq. Fairbanks: Native Language Center. Katz, Jerrold J., and Jerry A. Fodor. 1963. The structure of a semantic theory. Language 39(2): 170–210. Kayne, Richard S. 1975. French syntax: The transformational cycle. Cambridge, MA: MIT Press. Kayne, Richard S. 1984. Connectedness and binary branching. Dordrecht: Foris. Kayne, Richard S. 1985. L’accord du participe passé en franc  ¸ais et en italien. Modèles linguistiques 8: 73–89. Reprinted in English translation as Kayne (2000b). Kayne, Richard S. 1989. Facets of Romance past participle agreement. In Paula Benincà (ed.), Dialect variation and the theory of grammar, 85–103. Dordrecht: Foris. Reprinted in Kayne (2000a: 25–39). Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Kayne, Richard S. 2000a. Parameters and universals. Oxford: Oxford University Press. Kayne, Richard S. 2000b. Past participle agreement in French and Italian. In Kayne (2000a), 10–24. Translated from Kayne (1985). Kayne, Richard S. 2008. Expletives, datives, and the tension between morphology and syntax. In Theresa Biberauer (ed.), The limits of syntactic variation, 175–217. Amsterdam: John Benjamins. Kim, Kyumin. 2011. High applicatives in Korean causatives and passives. Lingua 121(3): 487–510. Kim, Kyumin. 2014. Argument structure theory and non-core arguments. Ms., University of Calgary. Kim, Kyumin. 2015. Spatial PPs and the structure of motion verbs in Blackfoot. UBC Working Papers in Linguistics 39: 125–39.

286 references Kim, Kyumin. 2017a. Animacy and transitivity alternations in Blackfoot. In Monica Macaulay and Margaret Noodin (eds.), Papers of the 46th Algonqiuan Conference, 123–40. East Lansing, MI: Michigan State University Press. Kim, Kyumin. 2017b. Sentience, initiation, and delimitedness in (lexical) aspect. Ms., Cheongju University. Kim, Kyumin, Elizabeth Ritter, Martina Wiltschko, and Hotze Rullmann. 2017. 2 + 2 = 3: Number contrasts in Blackfoot. Glossa 2(1): 96. 1–15. Kiparsky, Paul. [1968] 1973. How abstract is phonology? In Osamu Fujimura (ed.), Three dimensions in linguistic theory, 5–136. Tokyo: TEC Company. Originally published by the Indiana University Linguistics Club. Kiparsky, Paul. 1973. ‘Elsewhere’ in phonology. In Stephen Anderson and Paul Kiparsky (eds.), A festschrift for Morris Halle, 93–106. New York: Holt, Rinehart, & Winston. Klein, Wolfgang. 1992. The present perfect puzzle. Language 68(3): 525–52. doi:10.2307/ 415793 Kolliakou, Dimitra. 2004. Monadic definites and polydefinites: Their form, meaning and use. Journal of Linguistics 40(2): 263–333. Koopman, Hilda. 2003. The locality of agreement and the structure of the DP in Maasai. In William Earl Griffin (ed.), The role of agreement in natural language: Proceedings of the fifth annual Texas Linguistics Society Conference, 207–27. Austin: Texas Linguistic Forum. Koopman, Hilda. 2004. Data from 17 speakers on English agreement in raising constructions with there. Ms., University of California, Los Angeles. Available online at http://linguistics.ucla.edu/people/koopman/papers/KoopmanAgreement-Lite.pdf. Koopman, Hilda. 2006. Agreement configurations: In defense of ‘Spec head’. In Cedric Boeckx (ed.), Agreement systems, 159–99. Amsterdam: John Benjamins. Kornfilt, Jaklin, and John Whitman (eds.). 2011. Nominalizations in linguistic theory. Amsterdam: Elsevier. Special issue of Lingua (vol. 121, no. 7). Kratzer, Angelika. 1996. Severing the external argument from its verb. In Johan Rooryck and Laurie Zaring (eds.), Phrase structure and the lexicon, 109–37. Dordrecht: Kluwer. Kratzer, Angelika. 2016. A combinatorics for attitude ascriptions. Video recording. Available online at http://works.bepress.com/angelika_kratzer/35/ Krifka, Manfred. 1992. Thematic relations as links between nominal reference and temporal constitution. In Ivan Sag and Anna Szabolcsi (eds.), Lexical matters, 29–53. Stanford, CA: CSLI Publications. Krifka, Manfred. 1998. The origins of telicity. In Susan Rothstein (ed.), Events and grammar, 197–236. Dordrecht: Kluwer. Krifka, Manfred, Francis Jeffry Pelletier, Gregory N. Carlson, Alice Ter Meulen, Godehard Link, and Gennaro Chierchia. 1995. Genericity: An introduction. In Greg N. Carlson and Francis Jeffry Pelletier (eds.), The generic book, 1–124. Chicago: University of Chicago Press. Kripke, Saul. 1980. Naming and necessity. Oxford: Blackwell. Kuno, Susumu. 1971. The position of locatives in existential sentences. Linguistic Inquiry 2(3): 333–78. Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge, MA: MIT Press. Kyriakaki, Maria. 2006. The geometry of tense, mood and aspect in Greek. Master’s thesis, University of Toronto. Kyriakaki, Maria. 2011. DETs in the functional syntax of Greek nominals. PhD thesis, University of Toronto. Kyriakaki, Maria. 2014. The three degrees of definiteness. In Herman Leung, Zachary O’Hagan, Sarah Bakst, Auburn Lutzross, Jonathan Manker, Nicholas Rolle, and Katie

references 287 Sardinha (eds.), Proceedings of the fortieth annual meeting of the Berkeley Linguistics Society, 253–69. Berkeley, CA: Berkeley Linguistics Society. doi:10.3765/bls.v40i0.3143 Landau, Idan. 2013. Control in generative grammar: A research companion. Cambridge: Cambridge University Press. Lasnik, Howard, and Mamoru Saito. 1992. Move α: Conditions on its application and output. Current Studies in Linguistics. Cambridge, MA: MIT Press. Lasnik, Howard, Juan Uriagereka, and Cedric Boeckx. 2005. A course in Minimalist syntax: Foundations and prospects. Oxford: Blackwell. Legate, Julie. 2014. Voice and v: Lessons from Acehnese. Cambridge, MA: MIT Press. Lekakou, Marika, and Kriszta Szendrői. 2012. Polydefinites in Greek: Ellipsis, close apposition, and expletive determiners. Journal of Linguistics 48(1): 107–49. doi:10.1017/ S0022226711000326 Lekakou, Marika, and Kriszta Szendrői. 2013. The cross-linguistic distribution of polydefinites: Case and expletive determiners. In Seda Kan, Claire Moore-Cantwell, and Robert Staubs (eds.), NELS 40: Proceedings of the fortieth annual meeting of the North East Linguistic Society, vol. 2, 43–54. Amherst, MA: GLSA. Levin, Beth, and Malka Rappoport-Hovav. 1995. Unaccusativity: At the syntax–lexical semantics interface. Cambridge, MA: MIT Press. Levin, Beth, and Malka Rappoport-Hovav. 1998. Building verb meanings. In Miriam Butt and Wilhelm Geuder (eds.), The projection of arguments: Lexical and compositional factors, 97–134. Stanford, CA: CSLI Publications. Lightfoot, David. 1979. Principles of diachronic syntax. Cambridge: Cambridge University Press. Little, Carol-Rose. 2018. A binary feature analysis of Mi’gmaq number agreement. In William G. Bennett, Lindsay Hracs, and Dennis Ryan Storoshenko (eds.), Proceedings of the 35th West Coast Conference on Formal Linguistics, 242–50. Somerville, MA: Cascadilla Proceedings Project. Lochbihler, Bethany. 2008. Person encoding in the Ojibwe inverse system. In Marc Richards and Andrej L. Malchukov (eds.), Scales, vol. 86 of Linguistische Arbeits Berichte, 295–315. Leipzig: Institut für Linguistik, Universität Leipzig. Lochbihler, Bethany. 2012. Aspects of argument licensing. PhD thesis, McGill University. Lomashvili, Leila. 2011. Complex predicates: The syntax–morphology interface. Amsterdam: John Benjamins. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25(4): 609–65. Louie, Megan. 2008. Atemporal anchoring of individuals, events and subevents in Blackfoot: Consequences for the syntax–semantics interface. Master’s thesis, University of Toronto. Louie, Megan. 2010. Deriving full repetition in Blackfoot with additive focus. In Nan Li and David Lutz (eds.), Proceedings of SALT 20, 1–18. Ithaca, NY: CLC Publications. Lounsbury, Floyd G. 1949. Iroquoian morphology. PhD thesis, Yale University. Lounsbury, Floyd G. 1953. Oneida verb morphology. Ms. Yale University. Lunt, Horace Gray. 1959. Old Church Slavonic grammar. The Hague: Mouton. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. MacAulay, Donald. 1996. Some thoughts on time, tense and mode, and on aspect in Scottish Gaelic. Scottish Gaelic Studies 14: 193–204. Mackenzie, Sara. 2013. Laryngeal co-occurrence restrictions in Aymara: Contrastive representations and constraint interaction. Phonology 30(2): 297–345. Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge, MA: MIT Press.

288 references Marantz, Alec. 1989. Relations and configurations in Georgian. Ms., University of North Carolina at Chapel Hill. Marantz, Alec. 1991. Case and licensing. In Germán F. Westphal, Benjamin Ao, and Hee-Rahk Chae (eds.), ESCOL ’91: Proceedings of the eighth Eastern States Conference on Linguistics, 234–53. Columbus, OH: ESCOL Publications Committee. Reprinted in Reuland (2000), 11–30. Marantz, Alec. 1996. ‘Cat’ as a phrasal idiom: Consequences of late insertion in distributed morphology. Ms., MIT. Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. Penn Working Papers in Linguistics 4(2): 201–25. Marantz, Alec. 2013. Locality domains for contextual allomorphy across the interfaces. In Ora Matushansky and Alec Marantz (eds.), Distributed Morphology today: Morphemes for Morris Halle, 95–115. Cambridge, MA: MIT Press. Marinakis, Aliki, Mary Koyina Richardson, Leslie Saxon, and Mary Siemens (eds.). 2010 [2007]. Tłı  ˛cho  ˛ k’ ˛è è ˛ ets’eetł’èe xè enı  ˛htł’è k’e yats’ehtıı / Reading and writing in Tłı  ˛cho  ˛ Yatıì. Behchokò  ˛, NWT: Tłı  ˛cho  ˛ Community Services Agency. Martel, Joseph, and Leslie Saxon. 2005. Vowel length neutralization in Dogrib stems: An acoustic study. Presented at the annual meeting of the Society for the Study of the Indigenous Languages of the Americas, Oakland, CA. Martí, Luisa. 2003. Contextual variables. PhD thesis, University of Connecticut. Matthewson, Lisa. 2004. On the absence of telic accomplishments in St’át’imsets. UBC Working Papers in Linguistics 15: 65–78. Matthewson, Lisa, and Amelia Reis Silva. 2007. An instantaneous present tense in Blackfoot. Presented at SULA 4: Semantics of Under-Represented Languages in the Americas, University of São Paulo. McCawley, James D. 1971. Tense and time reference in English. In Charles J. Fillmore and D. Terence Langendoen (eds.), Studies in linguistic semantics, 97–113. New York: Holt, Rinehart, & Winston. McClay, Elise. 2012. Possession in Mi’gmaq. Undergraduate honours thesis, McGill University. McCloskey, James. 1983. A VP in a VSO language. In Gerald Gazdar, Ewan Klein, and Geoffrey Pullum (eds.), Order, concord and constituency, 9–55. Dordrecht: Foris. McCloskey, James. 2002. Resumption, successive cyclicity, and the locality of operations. In Samuel David Epstein and T. Daniel Seely (eds.), Derivation and explanation in the Minimalist Program. Oxford: Blackwell. McCloskey, James. 2014. Irish existentials in context. Syntax 17(4): 343–84. doi:10.1111/ synt.12020 McCoard, Robert W. 1978. The English perfect: Tense-choice and pragmatic inferences. Amsterdam: North-Holland Publishing. McGinnis, Martha. 1998. Locality in A-movement. PhD thesis, Massachusetts Institute of Technology. McGinnis, Martha. 2008. Phi-feature competition in morphology and syntax. In Daniel Harbour, David Adger, and Susana Béjar (eds.), Phi-theory: Phi-features across modules and interfaces, 155–84. Oxford: Oxford University Press. McGinnis, Martha. 2013. Agree and fission in Georgian plurals. In Ora Matushansky and Alec Marantz (eds.), Distributed Morphology today: Morphemes for Morris Halle, 39–58. Cambridge, MA: MIT Press. McGinnis, Martha. 2016. A vP layer in result and complex event nominals. Presented at the 2016 annual meeting of the Canadian Linguistic Association, University of Calgary.

references 289 Meadows, Kim. 2010. On the role of sentience in Blackfoot: Evidence from the accompaniment and associative constructions. Master’s thesis, University of Calgary. Mezhevich, Ilana. 2005. Grammatical and structural relations in Russian word order. PhD thesis, University of Calgary. Michaelis, Laura A. 1994. The ambiguity of the English present perfect. Journal of Linguistics 30(1): 111–57. Michelson, Karin, and Mercy A. Doxtator. 2002. Oneida–English/English–Oneida dictionary. Toronto: University of Toronto Press. Milsark, Gary Lee. 1974. Existential sentences in English. PhD thesis, Massachusetts Institute of Technology. Mithun, Marianne. 2008. The extension of dependency beyond the sentence. Language 84(1): 69–119. Moens, Marc. 1987. Tense, aspect and temporal reference. PhD thesis, University of Edinburgh. Morén, Bruce. 2003. The Parallel Structures Model of feature geometry. Working Papers of the Cornell Phonetics Laboratory 15: 194–270. Morén, Bruce. 2006a. Consonant–vowel interactions in Serbian: Features, representations and constraint interactions. Lingua 116(8): 1198–244. Morén, Bruce. 2006b. The division of labor between segment-internal structure and violable constraints. In Sylvia Blaho, Patrik Bye, and Martin Krämer (eds.), Freedom of analysis?, 313–44. The Hague: Mouton. Moro, Andrea. 1997. The raising of predicates: Predicative noun phrases and the theory of clause structure. Cambridge: Cambridge University Press. Moro, Andrea. 2006. Existential sentences and expletive there. In Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell companion to syntax, vol. 3, 1483–1508. Oxford: Blackwell, 1st edn. doi:10.1002/9780470996591.ch24 Moro, Andrea. 2017. Existential sentences and expletive there. In Martin Everaert and Henk van Riemsdijk (eds.), The Wiley Blackwell companion to syntax, vol. 2, 210–36. Hoboken, NJ: John Wiley & Sons, 2nd edn. doi:10.1002/9781118358733.wbsyncom044 Moulton, Keir. 2004. External arguments and gerunds. Toronto Working Papers in Linguistics 22: 121–36. Moulton, Keir. 2009. Natural selection and the syntax of clausal complementation. PhD thesis, University of Massachusetts, Amherst. Moulton, Keir. 2014. Simple event nominalizations: Roots and their interpretation. In Ileana Paul (ed.), Cross-linguistic investigations of nominalization patterns, 119–44. Amsterdam: John Benjamins. Moulton, Keir. 2015. CPs: Copies and compositionality. Linguistic Inquiry 46(2): 305–42. Mourelatos, Alexander P. D. 1978. Events, processes, and states. Linguistics and Philosophy 2(3): 415–34. Nash, Léa. 1995. Portée argumentale et marquage casuel dans les langues SOV et dans les langues ergatives: L’exemple du géorgien. [Argument scope and case-marking in SOV and in ergative languages: The case of Georgian]. PhD thesis, Université Paris 8. Nash, Léa. 2017. Causees are not agents. Presented at Linguistic Perspectives on Causation, Hebrew University of Jerusalem. Nash, Léa. n.d. Licensing dative arguments. Ms., Université Paris 8. Nevins, Andrew. 2008. Cross-modular parallels in the study of phon and phi. In Daniel Harbour, David Adger, and Susana Béjar (eds.), Phi-theory: Phi-features across modules and interfaces, 329–67. Oxford: Oxford University Press.

290 references Nevins, Andrew. 2015. Triumphs and limits of the contrastivity-only hypothesis. Linguistic Variation 15(1): 41–68. Noyer, Robert Rolph. 1992. Features, positions and affixes in autonomous morphological structure. PhD thesis, Massachusetts Institute of Technology. Nurse, Derek. 2003. Aspect and tense in Bantu languages. In Derek Nurse and Gérard Philippson (eds.), The Bantu languages, chap. 6. London: Routledge. Ó Sé, Diarmuid. 1990. Tense and mood in Irish copula sentences. Ériu 41: 61–75. Ó Sé, Diarmuid. 2004. The ‘after’ perfect and related constructions in Gaelic dialects. Ériu 54: 179–248. Oehrle, Richard. 1976. The grammatical status of the English dative alternation. PhD thesis, Massachusetts Institute of Technology. Oxford, Will. 2017. Size matters: Deriving person hierarchies from nominal structure. Ms., University of Manitoba. Pantcheva, Marina Blagoeva. 2011. Decomposing path: The nanosyntax of directional pronouns. PhD thesis, Universitetet i Tromsø. Paradis, Carole, and Jean-Franc  ¸ois Prunet. 1991. Introduction: Asymmetry and visibility in consonant articulations. In Carole Paradis and Jean-Franc  ¸ois Prunet (eds.), The special status of coronals: Internal and external evidence, no. 2 in Phonetics and Phonology, 1–28. San Diego: Academic Press. Paul, Ileana (ed.). 2014. Cross-linguistic investigations of nominalization patterns. Amsterdam: John Benjamins. Pesetsky, David, and Esther Torrego. 2007. The syntax of valuation and the interpretability of features. In S. Karimi, V. Samiian, and W. Wilkins (eds.), Phrasal and clausal architecture: Syntactic derivation and interpretation. in honor of Joseph E. Emonds, 262–94. Amsterdam: John Benjamins. Pollock, Jean-Yves. 1989. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20(3): 365–424. Preminger, Omer. 2009. Long-distance agreement without probe–goal relations. In Michael T. Putnam (ed.), Towards a derivational syntax: Survive-Minimalism, 41–56. Amsterdam: John Benjamins. Preminger, Omer, and Maria Polinsky. 2015. Agreement and semantic concord: A spurious unification. Ms., University of Maryland. Available online at http://ling.auf.net/lingbuzz/ 002363. Prince, Ellen F. 1981. On the inferencing of indefinite this NPs. In Aravind K. Joshi, Bonnie Lynn Webber, and Ivan A. Sag (eds.), Elements of discourse understanding, 231–50. Cambridge: Cambridge University Press. Pylkkänen, Liina. 2008. Introducing arguments. Cambridge, MA: MIT Press. Ramchand, Gillian. 1993. Aspect and argument structure in Modern Scottish Gaelic. PhD thesis, Sanford University. Ramchand, Gillian. 1997. Aspect and predication: The semantics of argument structure. Oxford: Oxford University Press. Ramchand, Gillian. 2008. Verb meaning and the lexicon: A first phase syntax. Cambridge: Cambridge University Press. Ramchand, Gillian, and Peter Svenonius. 2008. Mapping a parochial lexicon onto a universal semantics. In Theresa Biberauer (ed.), The limits of syntactic variation, 219–45. Amsterdam: John Benjamins. Ramchand, Gillian, and Peter Svenonius. 2014. Deriving the functional hierarchy. Language Sciences 46(B): 152–74. Raposo, Eduardo, and Juan Uriagereka. 1990. Long-distance case assignment. Linguistic Inquiry 21(4): 505–37.

references 291 Reed, Sylvia L. 2012. The semantics of grammatical aspect: Evidence from Scottish Gaelic. PhD thesis, University of Arizona. Reinhart, Tanya. 2003. The theta system—an overview. Theoretical Linguistics 28(3): 229–90. Reuland, Eric (ed.). 2000. Explaining Burzio’s Generalization. Amsterdam: John Benjamins. Rice, Keren. 1989. A grammar of Slave. Berlin: Mouton de Gruyter. Rice, Keren, and J. Peter Avery. 1995. Variability in a deterministic model of language acquisition: A theory of segmental elaboration. In John Archibald (ed.), Phonological acquisition and phonological theory, 23–42. Hillsdale, N.J.: Lawrence Erlbaum. Rice, Keren, and Leslie Saxon. 2005. Comparative Athapaskan syntax: Arguments and projections. In Guglielmo Cinque and Richard S. Kayne (eds.), The Oxford handbook of comparative syntax, 698–774. Oxford: Oxford University Press. Richards, Marc. 2012. Probing the past: On reconciling long-distance agreement with the PIC. In Artemis Alexiadou, Tibor Kiss, and Gereon Müller (eds.), Local modelling of nonlocal dependencies in syntax, 135–54. Berlin: Walter de Gruyter. Richards, Marc, and Theresa Biberauer. 2005. Explaining Expl. In Marcel den Dikken and Christina Tortora (eds.), The function of function words and functional categories, 115–53. Amsterdam: John Benjamins. Richards, Norvin. 2001. An idiomatic argument for lexical decomposition. Linguistic Inquiry 32(1): 183–92. Ritter, Elizabeth. 1991. Two functional categories in noun phrases: Evidence from Modern Hebrew. In Susan D. Rothstein (ed.), Perspectives on phrase structure: Heads and licensing, no. 25 in Syntax and Semantics, 37–62. San Diego: Academic Press. Ritter, Elizabeth. 1993. Where’s gender? Linguistic Inquiry 24(4): 795–803. Ritter, Elizabeth. 2014. Featuring animacy. Nordlyd 41(1): 103–24. Ritter, Elizabeth, and Sara Thomas Rosen. 1993. Deriving causation. Natural Language & Linguistic Theory 11(3): 519–55. Ritter, Elizabeth, and Sara Thomas Rosen. 1997. The function of have. Lingua 101(3–4): 295–321. Ritter, Elizabeth, and Sara Thomas Rosen. 2000. Event structure and ergativity. In Carol Tenny and James Pustejovsky (eds.), Events as grammatical objects, 187–238. Stanford, CA: CSLI Publications. Ritter, Elizabeth, and Sara Thomas Rosen. 2010. Animacy in Blackfoot: Implications for event structure and clause structure. In Malka Rappoport-Hovav, Edit Doron, and Ivy Sichel (eds.), Lexical semantics, syntax, and event structure, 124–52. Oxford: Oxford University Press. Ritter, Elizabeth, and Martina Wiltschko. 2009. Varieties of INFL: Tense, location, and person. In Jeroen van Craenenbroeck (ed.), Alternatives to cartography, 153–202. Berlin: Walter de Gruyter. Ritter, Elizabeth, and Martina Wiltschko. 2014. The composition of INFL: An exploration of tense, tenseless languages, and tenseless constructions. Natural Language & Linguistic Theory 32(4): 1331–86. doi:10.1007/s11049-014-9248-6 Ritter, Elizabeth, and Martina Wiltschko. 2016. Humanness as an alternative to case licensing. In Lindsay Hracs (ed.), Proceedings of the 2016 annual conference of the Canadian Linguistic Association. Toronto: Canadian Linguistic Association. Published online at http://cla-acl.ca/actes-2016-proceedings/ Ritter, Elizabeth, and Martina Wiltschko. 2019. The syntax of formality: Universals and variation. Presented at the 2019 annual meeting of the Canadian Linguistic Association, University of British Columbia.

292 references Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar: A handbook of generative syntax, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2001. On the position ‘Int(errogative)’ in the left periphery of the clause. In Guglielmo Cinque and (ed.), Current studies in Italian syntax: Essays offered to Lorenzo Renzi, 287–96. Leiden: Brill. Roberts, Ian. 1997. Restructuring, head movement, and locality. Linguistic Inquiry 28(3): 423–60. Rosen, Sara Thomas. 1996. Events and verb classification. Linguistics 34(2): 191–223. Rosenbaum, Peter. 1967. The grammar of English predicate complement constructions. Cambridge, MA: MIT Press. Roussou, Anna, and Ianthi-Maria Tsimpli. 1994. On the interaction of case and definiteness in Modern Greek. In Irene Philippaki-Warburton, Katerina Nicolaidis, and Maria Sifianou (eds.), Themes in Greek linguistics, 69–76. Amsterdam: John Benjamins. Rullmann, Hotze, and Aili You. 2006. General number and the semantics and pragmatics of indefinite bare nouns in Mandarin Chinese. In Klaus von Heusinger and Ken Turner (eds.), Where semantics meets pragmatics, 175–96. Amsterdam: Elsevier. Rupp, Laura. 2005. Constraints on nonstandard -s in expletive there sentences: A generativevariationist perspective. English Language and Linguistics 9(2): 255–88. doi:10.1017/ S1360674305001656 Russell, Bertrand. 1998 [1905]. On denoting. In Gary Ostertag (ed.), Definite descriptions: A reader, 35–49. Cambridge, MA: MIT Press. Safir, Kenneth. 1985. Syntactic chains. Cambridge: Cambridge University Press. Saussure, Ferdinand de. 1916. Cours de linguistique générale. Paris: Payot. Saussure, Ferdinand de. 1959. Course in general linguistics. New York: Philosophical Library. Translated from Saussure (1916) by Wade Baskin. Saxon, Leslie. 1979. Dogrib questions and topicalization. Master’s thesis, University of Toronto. Saxon, Leslie. 1984. Dogrib pronouns. Tech. rep., National Museum of Man. Final report on contract no. 1630-1-469. Saxon, Leslie. 1986. The syntax of pronouns in Dogrib (Athapaskan): Some theoretical consequences. PhD thesis, University of California, San Diego. Saxon, Leslie. 1998. Complement clauses in Dogrib. In Leanne Hinton and Pamela Munro (eds.), Studies in American Indian languages: Description and theory, 204–11. Berkeley, CA: University of California Press. Saxon, Leslie. 2003a. The complementizer *Gu’ in Athapaskan: Its reflex in Dogrib. In Yunhee Chung, Carrie Gillon, and Rachel Wojdak (eds.), Proceedings of WSCLA VIII: The eighth Workshop on Structure and Constituency in Languages of the Americas, no. 12 in UBC Working Papers in Linguistics, 59–74. Vancouver: UBC Department of Linguistics. Saxon, Leslie. 2003b. The nominalizing and adverbial suffixes in Proto-Athabascan. Presented at the Athabascan Languages Conference, Arcata, CA. Saxon, Leslie. 2003c. Two complementizers in Proto-Athapaskan. Presented at the International Conference on Historical Linguistics, Copenhagen. Saxon, Leslie. 2014. ‘Discourse particles’ and selection by higher predicates. Presented at WSCLA 19, Memorial University of Newfoundland, St. John’s. Saxon, Leslie, and Mary Siemens (eds.). 1996. Tłı  ˛cho  ˛ Yatıì enı  ˛htł’è / A Dogrib dictionary. RaeEdzo [Behchokò  ˛], NWT: Dogrib Divisional Board of Education. Schachter, Paul. 1976. The subject in Philippine languages: Topic, actor, actor-topic, or none of the above. In Charles Li (ed.), Subject and topic, 491–518. New York: Academic Press. Schmidt, Karl Horst. 1957. Eine südkaukasische Aktionsart? Münchner Studien zur Sprachwissenschaft 10: 9–24.

references 293 Schneider-Zioga, Patricia. 2017. That’s not how you agree, either: A contribution to the discussion on directionality of agreement. Ms., California State University, Fullerton. Schreiner, Sylvia L. R., and Andrew Carnie. 2016. The syntax and semantics of Scottish Gaelic a’ dol aL. Journal of Celtic Linguistics 17(1): 1–30. Schütz, Albert J. 1995. The voices of Eden: A history of Hawaiian language studies. Honolulu: University of Hawai‘i Press. Schütze, Carson T. 1996. The empirical base of linguistics: Grammaticality judgments and linguistic methodology. Chicago: University of Chicago Press. Reprinted by Language Science Press, Berlin, 2016. Schütze, Carson T. 1997. INFL in child and adult language: Agreement, case and licensing. PhD thesis, Massachusetts Institute of Technology. Schütze, Carson T. 2003. Syncretism and double agreement with Icelandic nominative objects. In Lars-Olof Delsing, Cecilia Falk, Gunlög Josefsson, and Halldór Á. Sigurðsson (eds.), Grammatik i fokus / Grammar in focus: Festschrift for Christer Platzack 18 November 2003, vol. 2, 295–303. Lund: Department of Scandinavian Languages, Lund University. Schütze, Carson T. 2011a. Agree cannot be reduced to spec–head agreement (without a lot of stipulations). Presented at Parallel Domains: Locality in Syntax/Phonology and the Representation of Constituency (A Workshop in Honor of the Work of Jean-Roger Vergnaud), Los Angeles. Schütze, Carson T. 2011b. There does not undergo predicate inversion. Snippets 24: 16–7. Scollon, Ronald. 1985. The sequencing of clauses in Chipewyan narrative. In Johanna Nichols and Anthony C. Woodbury (eds.), Grammar inside and outside the clause, 113–31. Cambridge: Cambridge University Press. Sichel, Ivy. 2010. Event-structure constraints on nominalization. In Artemis Alexiadou and Monika Rathert (eds.), The syntax of nominalizations across languages and frameworks, 159–98. Berlin: Mouton de Gruyter. Sigurðsson, Halldór Ármann. 1990–1991. Beygingarsamræmi [agreement]. Íslenskt mál og almenn málfræði 12–13: 31–77. Sigurðsson, Halldór Ármann. 1996. Icelandic finite verb agreement. Working Papers in Scandinavian Syntax 57: 1–46. Sigurðsson, Halldór Ármann. 2000. The locus of case and agreement. Working Papers in Scandinavian Syntax 65: 65–108. Slavin, Tanya. 2008. Towards a feature geometric account of Infl in Yucatec Mayan. Toronto Working Papers in Linguistics 29. Smith, Carlota S. 1991. The parameter of aspect. Dordrecht: Kluwer. doi:10.1007/ 978-94-011-5606-6 Speas, Margaret. 2004. Evidentiality, logophoricity and the syntactic representation of pragmatic features. Lingua 114(3): 255–76. Sportiche, Dominique. 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19(3): 425–49. Starke, Michal. 2009. Nanosyntax: A short primer to a new approach to language. Nordlyd 36(1): 1–6. Steriade, Donca. 1987. Redundant values. CLS 23(2): 339–62. Stevens, Kenneth Noble, and Samuel Jay Keyser. 2010. Quantal theory, enhancement and overlap. Journal of Phonetics 38(1): 10–19. doi:10.1016/j.wocn.2008.10.004 Stowell, Tim. 1978. What was there before there was there? In Donka Farkas, Wesley M. Jacobsen, and Karol W. Todrys (eds.), Papers from the fourteenth regional meeting of the Chicago Linguistic Society, 458–71. Chicago: Chicago Linguistic Society.

294 references Stowell, Tim. 1981. Origins of phrase structure. PhD thesis, Massachusetts Institute of Technology. Tenny, Carol. 1987. Grammaticalizing aspect and affectedness. PhD thesis, Massachusetts Institute of Technology. Tenny, Carol. 1994. Aspectual roles and the syntax–semantics interface. Dordrecht: Kluwer. Tenny, Carol, and James Pustejovsky. 2000. A history of events in linguistic theory. In Carol Tenny and James Pustejovsky (eds.), Events as grammatical objects, 3–37. Stanford, CA: CSLI Publications. Thráinsson, Höskuldur, Hjalmar P. Petersen, Jógvan í Lon Jacobsen, and Zakaris Svabo Hansen. 2004. Faroese: An overview and reference grammar. Tórshavn: Føroya Fróðskaparfelag. Tłı  ˛cho  ˛ Community Services Agency. 2005–. Tłı  ˛cho  ˛ Yatıì multimedia dictionary. Published online at http://tlicho.ling.uvic.ca/. Travis, Lisa deMena. 1992. Inner aspect and the structure of VP. In Mariette Champagne and Sylvie Ratté (eds.), Proceedings from the second annual Lexical-Syntactic Relations Workshop, 130–46. Montreal: Cahiers linguistiques de l’UQAM. Travis, Lisa deMena. 2010. Inner aspect: The articulation of VP. Dordrecht: Springer. Trubetzkoy, Nikolai S. 1939. Grundzüge der Phonologie. Travaux du cercle linguistique de Prague 7. Trubetzkoy, Nikolai S. 1969. Principles of phonology. Berkeley and Los Angeles: University of California Press. Translated by Christiane A. M. Baltaxe from Trubetzkoy (1939). Valentine, J. Randolph. 2010. Nishnaabemwin reference grammar. Toronto: University of Toronto Press. Vamling, Karina. 1989. Complementation in Georgian. No. 23 in Travaux de l’Institut de linguistique de Lund. Lund: Lund University Press. Van Valin, Robert D., and Randy J. LaPolla. 1997. Syntax: Structure, meaning and function. Cambridge: Cambridge University Press. Verkuyl, Henk J. 1972. On the compositional nature of the aspects. Dordrecht: Reidel. Verkuyl, Henk J. 1989. Aspectual classes and aspectual composition. Linguistics and Philosophy 12(1): 39–94. Verkuyl, Henk J. 1993. A theory of aspectuality: The interaction between temporal and atemporal structure. Cambridge: Cambridge University Press. Vikner, Sten. 1995. Verb movement and expletive subjects in the Germanic languages. Oxford: Oxford University Press. Welch, Nicholas. 2012. The bearable lightness of being: The encoding of coincidence in twocopula languages. PhD thesis, University of Calgary. Welch, Nicholas. 2015a. Deducing clause structure from the right periphery in Tłı  ˛cho  ˛ Yatıì. International Journal of American Linguistics 81(2): 261–91. Welch, Nicholas. 2015b. Mapping the frontier: Discourse particles and the cartography of the Dene clause. Presented at the annual meeting of the Linguistic Society of America, Portland, OR. Welch, Nicholas. 2015c. Obligatory future in a Dene language. International Journal of American Linguistics 81(1): 41–81. Wier, Thomas R. 2014. Nonconfigurationality and argumenthood in Georgian. Lingua 145(1): 36–64. Wilhelm, Andrea. 2003. Quasi-telic perfective aspect in Denë Su  ˛liné (Chipewyan). In Robert B. Young and Yuping Zhou (eds.), Proceedings from Semantics and Linguistic Theory XIII, 310–27. Ithaca, NY: CLC Publications.

references 295 Wilhelm, Andrea. 2007. Telicity and durativity: A study of aspect in Denë Su  ˛liné (Chipewyan) and German. New York: Routledge. Wilhelm, Andrea. 2014. Nominalization instead of modification. In Ileana Paul (ed.), Crosslinguistic investigations of nominalization patterns, 51–82. Amsterdam: John Benjamins. Wiltschko, Martina. 2008. The syntax of non-inflectional plural marking. Natural Language & Linguistic Theory 26(3): 639–94. Wiltschko, Martina. 2012. Decomposing the count–mass distinction: Evidence from languages that lack it. In Diane Massam (ed.), Count and mass across languages, 146–71. Oxford: Oxford University Press. Wiltschko, Martina. 2013. Descriptive relative clauses in Austro-Bavarian German. Canadian Journal of Linguistics 58(2): 157–89. Wiltschko, Martina. 2014. The universal structure of categories: Towards a formal typology. Cambridge: Cambridge University Press. Wiltschko, Martina, and Elizabeth Ritter. 2015. Animating the narrow syntax. The Linguistic Review 32(4): 869–908. Wolfart, H. Christoph. 1973. Plains Cree: A grammatical study. Transactions of the American Philosophical Society 63(1): 1–90. Woodbury, Hanni. 2003. Onondaga-English/English-Onondaga dictionary. Toronto: University of Toronto Press. Wurmbrand, Susanne. 1998. Infinitives. PhD thesis, Massachusetts Institute of Technology. Wurmbrand, Susi. 2014. The Merge Condition: A syntactic approach to selection. In Peter Kosta, Steven L. Franks, Teodora Radeva-Bork, and Lilia Schürcks (eds.), Minimalism and beyond: Radicalizing the interfaces, 139–77. Amsterdam: John Benjamins. Yang, Charles. 2016. The price of productivity: How children learn to break the rules of language. Cambridge, MA: MIT Press. Yang, Charles. 2017. How to wake up irregular (and speechless). In Claire Bowern, Laurence Horn, and Raffaella Zanuttini (eds.), On looking into words (and beyond): Structures, relations, analyses, no. 3 in Empirically Oriented Theoretical Morphology and Syntax, 211–33. Berlin: Language Science Press. doi:10.5281/zenodo.495447 Yoon, Jaehoon, and Lucien Brown. 2011. Korean: A comprehensive grammar. New York: Routledge. Zaenen, Annie, Joan Maling, and Höskuldur Thráinsson. 1985. Case and grammatical functions: The Icelandic passive. Natural Language & Linguistic Theory 3(4): 441–83. doi: 10.1007/BF00133285. Reprinted as Zaenen, Maling, and Thráinsson (1990). Zaenen, Annie, Joan Maling, and Höskuldur Thráinsson. 1990. Case and grammatical functions: The Icelandic passive. In Joan Maling and Annie Zaenen (eds.), Modern Icelandic syntax, no. 24 in Syntax and Semantics, 95–136. San Diego: Academic Press. Reprinted from Zaenen, Maling, and Thráinsson (1985). Zamparelli, Roberto. 2002. Definite and bare kind-denoting noun phrases. In Claire Beyssade, Reineke Bok-Bennema, Frank Drijkoningen, and Paola Monachesi (eds.), Romance languages and linguistic theory 2000, no. 232 in Current Issues in Linguistic Theory, 305–43. Amsterdam: John Benjamins. doi:10.1075/cilt.232.17zam Zeijlstra, Hedde. 2008. On the syntactic flexibility of formal features. In Theresa Biberauer (ed.), The limits of syntactic variation, vol. 132, 143–73. Amsterdam: John Benjamins. Zeijlstra, Hedde. 2012. There is only one way to agree. The Linguistic Review 29(3): 491–539. doi:10.1515/tlr-2012-0017 Zwart, C. Jan-Wouter. 1992. Dutch expletives and small clause predicate raising. In Kimberly Broderick (ed.), NELS 22: Proceedings of the twenty-second annual meeting of the North East Linguistic Society, 477–91. Amherst, MA: GLSA.

Index Abbott, C. 20, 22 Abney, S. P. 109, 145, 161 accompaniment suffixes 80–81 addition 253–255 Adger, D. 44n, 45, 124, 130n, 132n Adger, D. and P. Svenonius 3n adjectives, modifying nominals 132–133 adverbial clauses, causative complements as 210–212 adverbial complementizers 181, 185–193, 211 Affect α 2 Affix Hopping 220–221 agents 63–67 agreement long-distance 218n, 219–221, 235–245 speaker preference 218n Agree operation 2, 215, 216–217, 242–245 Alexiadou, A. 100, 130n, 132n, 143 Alexiadou, A. and F. Schäfer 221 Alexiadou, A. and C. Wilder 100 Algonquian languages (see also Blackfoot; Mi’gmaq) animacy 90 classification systems 68–69 Direct–Inverse agreement system 5 allophony 261, 262–263 Alternation Condition 257 Amazon Mechanical Turk 227n Androutsopoulou, A. 100 Angantýsson, Á. 233n animacy, and Inner Aspect 89–91 animacy-based classification systems 68–69 Anisman, A. 203n, 207n applicatives 167 high and low 160–161 Arad, M. 145, 147

Archangeli, D. 251n, 253 argument structure, evidence for masdars as v-selecting 154–168 Aronson, H. I. 139–140, 142, 147, 149, 155, 161, 164, 175 Asatiani, R. 139n aspect (see also habitual aspect; Inner Aspect, perfective events vs. imperfective events; prospective aspect; punctual aspect; stative aspect; tense-aspect-mood (TAM)) in complement clauses 194–199 contrasts in Scottish Gaelic 50–53 in Onondaga 18–24 Scottish Gaelic 41–45 viewpoint aspect 72 aspectual preverbs 140, 153 associates 81 Athabaskan languages see Dene languages Autonomous Interpretation Hypothesis 253 Bach, E. 84 Baker, M. 2, 217 Baker, M. and L. D. Travis 17, 24, 26 Bar-el, L., H. Davis, and L. Matthewson 75 bare NP arguments 88 Barnes, M. P. 242 Barrie, M. 17, 26n Bat-El, O. 89n Béjar, S. 152, 154 Béjar, S. and D. C. Hall 6, 258n Béjar, S. and M. Rezac 5, 7, 216 Belletti, A. 244 benefactive suffixes 79–80 Bernstein, J. B. and C. Tortora 124 Bjorkman, B. M. and E. Cowper 60, 64n, 172, 182 Bjorkman, B. M., E. Cowper, and D. Siddiqi 145

298 index Bjorkman, B. M. and H. Zeijlstra 217 Black, R. 54 Blackfoot animacy 89–91 boundedness 68–77 causers and agents 65–67 dynamicity and eventiveness 92–95 event delimitation 68–69 as an I-language 74 nominal Inner Aspect 85–91 noun classifications 69, 85–91 sentience 58–63, 77–83, 91–95 verb classifications 69–77, 92–93 Bliss, H. 58n, 59–60, 77–80, 86n, 88 Bliss, H. and E. Ritter 70n Blix, H. 152 Bloomfield, L. 69, 92–93 Bobaljik, J. D. 221, 237, 238 Bobaljik, J. D. and A. Carnie 45 Bobaljik, J. D. and H. Thráinsson 3, 263 Boeckx, C. 218n Bogal-Allbritten, E. 181, 187n, 210 Bohnemeyer, J. and M. Swift 72 Bonet, E. 5, 6 Borer, H. 2, 86n, 109, 111n, 143 Borer–Chomsky Conjecture 2 Bošković, Ž. 220–221, 240n boundedness (telicity) 68–77, 85 Bradley, J. W. 17 Bresnan, J. et al. 177 Brown, C. A. 249 Bruening, B. 177 Brugger, G. 101 Burt, M. K. 220 Butt, M. 210 Byrne, M. 54 Cable, S. 57 Caha, P. 8 Campbell, R. and B. Campbell 263 Campos, H. and M. Stavrou 100 canonical copulars 222–227 Cardinaletti, A. 230 Carlson, G. N. 126n Carlson, G. N. and F. J. Pelletier 120 Carnie, A. 45, 228n cartographic theories 2 Case features hierarchical representations 6 morphological interpretation 4

Case Filter 237 Case licensing 156, 163–164, 167 causative complements, as adverbial clauses 210–212 causatives 172–178 bieventive causatives 168 causative unergatives 138 evidence for masdars as v-selecting 168–172 Tłı  ¸cho  ¸ 181, 183–193, 201 causers 63–67 Ceong, H. H. 182, 205, 210 Ceong, H. H. and L. Saxon 194, 205 Chafe, W. L. 17, 20, 22, 24 Chain function 220 Chandra, P. 229–230 Cheng, L. L.-S. 194, 205, 257 Cheng, L. L.-S. and R. Sybesma 7 Cherry, E. C., M. Halle, and R. Jakobson 252 ChiBemba 160–161 Chierchia, G. 119, 120n Chomsky, N. 154, 163 Agree operation 216–217, 242, 244–245 contrastive features 3 EPP features 4 form and meaning 248 LF affix 220 Minimalist Program 2, 4, 215 neoparametric view of syntax 8 Principle and Parameters framework 2 Russian 262n Standard Theory 1 Christophersen, P. 109n Chung, S. 182 Chung, S. and W. A. Ladusaw 109, 110n, 180 Cinque, G. 130n Cinque, G. and L. Rizzi 2, 263 clandestine meeting approach to long-distance agreement 220, 221, 229 Clarke, S. 6 clausal structure, Onondaga 33–34 clause-final adverbials, Scottish Gaelic 43 Clements, G. N. 251, 254n Clements, G. N. and E. Hume 6 cognitive modules 2 coindexing 220 Collins, C. and H. Thráinsson 230

index 299 companions 81 complement clauses aspect in 194–199 independent negation 203–204 independent tense 199–203 complementizers (C) 179, 181, 185–193, 210–211 complex aspect constructions 19, 27–30 complex event nominalizations (CENs) 142–143, 170–171 Compton, R. and B. E. Dresher 256–257 Comrie, B. 19, 43 concrete finals 79n constraints 2 contrastive features 3–4 contrastive hierarchies 8, 13, 252, 254, 265–269 contrastive scope 6 Contrastive Specification 251 Contrastivist Hypothesis 255 contrasts definition 10 morphosyntactic inventories 257–261, 263, 271 phonemic inventories 255–257, 259, 271 Cook, E.-D. 181, 210–211 Coon, J. and A. Bale 267–270 Copley, B. and H. Harley 64n, 95–96 copulars, canonical and inverse 222–227 Copy Theory of Movement 241 Corbett, G. G. 36, 91 count nouns 68, 84, 86 covert movement 235–236 Cowper, E. 1, 5, 6, 7, 9, 17, 31–33, 34n, 36, 39, 45, 50, 54, 55, 56n, 60, 109, 111n, 129n, 145, 247, 249 Cowper, E. and D. C. Hall 3, 5, 6, 7, 8, 9, 17, 32, 39, 54, 60, 247, 250, 263, 264, 266, 270n Cowper, E. et al. 10 Cox, C. 181 Craig, C. 95 cross-clausal agreement (CCA) 78n cross-linguistic differences 2 definite articles 100 Czech allophony 262 nominalizations 164 politeness distinction 271

Dahl, Ö. 56 D’Alessandro, R. and A. Ledgeway 27 dative subjects 164–167 in causative clauses 174 Davison, A. 210 Deal, A. R. 221 Declerck, R. 68 definite determiners 99–101, 121–133 expletive 103–109 familiarity and uniqueness 101–102, 109, 110–112 full, partial and zero 133–135 generic 118–121, 126n specified and underspecified 126 definiteness 109–112 D(elimitation)-languages 74 demonstratives, Blackfoot 86 den Dikken, M. 224, 226n Dene languages 179, 181, 187 (see also Dënesųłıné; Navajo; Slave (Slavey); Tłı  ¸cho  ¸ adverbializers 210 complementizers (C) 211 Dënesu  ¸łıné 181, 192, 198, 212 Denison, D. 264 dependency relations 7–8 derived nominalizations 154–155 determiners definite 99–101, 121–133 of modifying nominals 127–133 determiner spreading 100 different-from relation 262 directed motion verbs 73, 75 Direct–Inverse agreement system 5 distinctive features 9 Distributed Morphology (DM) 9, 34n, 39, 258n, 262 division 251–253 Dixon, R. M. W. 148 domain restriction 109–110 domains 170 Doron, E. 63 Dowty, D. R. 71n Dresher, B. E. 8, 9n, 247–248, 251, 253, 255 Dresher, B. E. and D. C. Hall 3n Dresher, B. E. and N. Hornstein 220 Dresher, B. E., G. L. Piggott, and K. D. Rice 247, 251 Dryer, M. S. 77

300 index Dunham, J. 71 Dutch, expletives 228 Dvořák, V. 164 dynamicity 92–95 Early Modern English 263–264 Elsewhere Principle 262 Embick, D. 142, 148, 163, 170 Embick, D. and A. Marantz 157, 163, 170 entailments 7–8, 71n EPP features 4 ergativity 148 event classes 74 event-defining arguments 95 event delimitation 68–69 eventiveness 92–95 eventive readings 7n event structure 95 active/reactive situation 62 atelic events 68, 84 boundedness in 68–69 interactive situation 62, 83 nominal Inner Aspect 85–91 existentials 222, 226–227 Expletive Adjunction/Affixation 220 expletive definite articles 103–109, 134 Expletive Replacement 220 expletives 228 extraction 222–225 EyÞórsson, Þ. 233n factual modal prefixation 25–26, 35 familiarity 101–102, 109, 110–112 in Greek 119 Familiarity Phrase (FamP) 102, 110–112, 119–120 Greek 121–126 modifying nominals as 127–129 Faroese 241–242 see also Icelandic Feature Economy 254n feature geometries 7–8, 9, 31–33, 39–40, 249, 253–255 features in syntax, role of 1 Feenstra, J. 182 Fernald, T. B. 181 finite/deixis feature 40 Foley, S. 154 Folli, R. and H. Harley 63, 64n, 182 form and meaning 247–248

Foster, M. K. 24, 26 Frantz, D. G. 58–59, 65, 79–81, 88n, 89n Frantz, D. G. and N. J. Russell 73n, 77n, 86–87, 92–93 Frege, G. 109n French agreement 243 definite articles 100, 101 modality 10 politeness distinction 271 recent/immediate past tense 56 restructuring 234 Froman, F. et al. 22 Functional Application 128 functional lexicon 1 future tense 8–9, 26, 53–56 Georgian 140 habitual and stative 30 gender features, morphological interpretation 4 Genee, I. and M.-O. Junker 58n General Number 36 General Tense 36–37 generics 118–121 definite determiners 126n genitive agents 157 George, L. and J. Kornfilt 146 Georgian applicatives 161 argument structure 154–168 causatives 174–178 dative subjects 164–167 gerundive nominalizations 138–141 gerunds 146 masdars as nominalizing heads 141–145 nominative objects 166 passivization 158–159 root suppletion 163 thematic suffixes 147–148, 149–151 word order 165–167 Gerdts, D. 95 German animacy 89n definite articles 100–101 expletive definite articles 103–104 restructuring 234

index 301 Germanic, future tense 8–9 gerundive nominalizations 138–141 gerunds 146, 157, 159 ing-of 161–162 Ghomeshi, J. and D. Massam 121–122, 125 Giannakidou, A. 109, 110n Gillon, C. 109–110, 111n Givón, T. 57, 161 Gleasure, J. 55 Goal path 52, 80, 218 goal PP objects 94 Goddard, I. 68n, 90 Greek definite articles 99–100, 111–112 definite determiners 121–133, 134 expletive definite articles 104–109 Familiarity Phrase (FamP) 121–126 generics 118–121 modifying nominals 130, 132–133 polydefinites 112, 113–114 possessives 99n restrictive modification by nominals (RMN) 115–117 Green, G. M. 177 Grimshaw, J. 158, 180, 235n Grimshaw, J. and V. Samek-Lodovici 235n Gropen, J. et al. 177 Guasti, M. T. and L. Rizzi 244 habitual aspect 18, 19, 20–21, 23, 34, 35 habitual past 27–30, 35 Haegeman, L. 130n Hale, K. L. and S. J. Keyser 85, 161 Halkomelem 93, 95 Hall, D. C. 32, 129n, 247, 251n, 253, 254, 255, 262n Hall, D. C. and E. Cowper 266 Hall, K. C. 251n Halle, M. 253, 262 Halle, M. and A. Marantz 9, 262 Harbour, D. 8, 39, 250, 260–261, 265–266, 270n Harbour, D. and C. Elsholtz 8 Hargus, S. 181, 211 Harley, H. 5, 9, 70, 84, 85, 143, 146, 160, 172 Harley, H. and R. Noyer 145 Harley, H. and E. Ritter 5, 9, 247, 249

Harms, R. T. 257 Harris, A. C. 139n, 141, 143–144, 153, 158, 163, 165–166, 167, 174–175 Harris, J. and G. Lindsey 253 Harris, Z. 220 Hartmann, J. 219n, 222n Hartmann, J. and C. Heycock 232n Hausa 253, 259 Hawaiian 254, 255–256, 259 Hayashi, M. 56, 57 Hayes, B., R. Kirchner, and D. Steriade 249 Heap, D. 6 Hebrew, animacy 89n Heim, I. 101, 109 Heim, I. and A. Kratzer 101, 109, 111, 113, 114–115, 127n, 128–129 Hewitt, B. G. 156n, 158, 160, 169 Heycock, C. 222n, 226 Heycock, C. and A. Kroch 225n hierarchical representations 4–5 Hindi–Urdu, causatives 210 Hinrichs, E. W. 70, 84 Hockett, C. F. 91, 259 Hoekstra, T. and R. Mulder 223 Hornstein, N. 43, 235–237, 238–241 Hornstein, N., J. Nunes, and K. K. Grohmann 217 Hróarsdóttir, Þ. 233, 234 Hucklebridge, S. 179n, 188n Hyman, L. M. 257 Icelandic 215n (see also Faroese) long-distance agreement 239–240 quirky subjects 228–229, 231 restructuring 233–234 Spec–Head agreement 229–233 imperfective events 29 vs. perfective events 6–7, 26 imperfective paradox 71n Impoverishment 9 incremental theme verbs 84–85 inert situation 62, 95 inflectional morphology 3 inflectional paradigms 257–258 initial change morphemes 80n I(nitiation)-languages 74 Inner Aspect 68, 84 nominal Inner Aspect 85–91

302 index Internal Merge 237 interval feature 40 Iñupiaq 256–257 inventories 261 morphosyntactic 257–261, 263, 271 phonemic 255–257, 259, 271 inverse copulars 222–227 Iosad, P. 254 iota Phrase (𝚤P) 102, 110–112, 120, 122 IP layer 27 Irish (Modern) 47 Iroquoian languages 17 (see also Onondaga) order of morphemes 18 punctual aspect 22 irrealis 35–37 irrealis feature 40, 54, 56 Italian agreement 243–244 definite determiners 134–135 expletive definite articles 103–104 Jackendoff, R. 3, 68, 84 Jakaltek 95 Jakobson, R. 3 Jamamadí, 263 Japanese 95 adverbial complementizers 185n causatives 210 causative unergatives 138, 173 Jaszczolt, K. M. 24n Johansson, S. 65–66, 79, 85n, 90 Johnson, M. R. 57 Jónsson, J. G. 228n, 233n Jung, D. 181 Kadmon, N. 109n Kaplan, L. D. 256 Katz, J. J. and J. A. Fodor 3 Kayne, R. S. 132n, 161, 221, 234, 239n, 242 Kim, K. 73–77, 94, 138, 172, 174 Kim, K. et al. 88 Kiparsky, P. 257, 262 Klein, W. 43 Kolliakou, D. 100 Koopman, H. 218n, 219n, 229–233, 235n, 240

Korean causatives 210 causativized transitives 138 Kornfilt, J. and J. Whitman 180 Kratzer, A. 145–146, 154, 157, 180, 210 Krifka, M. 70, 84, 85n, 86n Krifka, M. et al. 101 Kripke, S. 106 Kuno, S. 95, 220 Kyriakaki, M. 6, 17, 33n, 100, 101, 102n, 109, 111n, 112n, 129n, 130n, 132n Lakhota 95 Landau, I. 199n language, as an arch 248 language acquisition 260 Lasnik, H. and M. Saito 2 Lasnik, H., J. Uriagereka, and C. Boeckx 217 Left Branch Condition 237 Legate, J. 172, 174 Lekakou, M. 101n Lekakou, M. and K. Szendrői 99, 100, 105–106, 108, 116n, 127n Levin, B. and M. Rappoport-Hovav 61, 80 lexical items (LIs) 263, 265–266 LF affix 220–221 Lightfoot, D. 9 linking 220 Little, C.-R. 265 Lochbihler, B. 5 Lomashvili, L. 168, 173, 175n long-distance agreement 218n, 219–221, 235–245 Longobardi, G. 99, 103–104, 106, 113, 118, 121, 125, 134 Louie, M. 67, 71–73, 77n, 86n, 90 Lounsbury, F. G. 17, 18, 19–20, 21–22, 24 Lower Right Corner 237, 238 Lunt, H. G. 258 Lyon, J. 192n Lyons, C. 101, 106, 110, 119, 123 MacAulay, D. 41 Mackenzie, S. 252–253 malefactive arguments 164, 166–167 Mandarin Chinese, dependency relations 7 Mantla, R. 182, 196, 207–208

index 303 Marantz, A. 142, 145, 148n, 154, 156, 157, 161, 165, 168, 170–171, 263 Marinakis, A., M. K. Richardson, L. Saxon, and M. Siemens 179 Martel, J. and L. Saxon 179n Martí, L. 110 masdars 138–141 argument structure 154–168 causatives 174–178 nominalizing heads 141–145 thematic suffixes 149–151 v-selecting 145–172 Massam, D. and J. Ghomeshi 113 mass nouns 68, 84, 86 Matthewson, L. 75 Matthewson, L. and A. Reis Silva 72 McCawley, J. D. 198n McClay, E. 265 McCloskey, J. 47, 217n, 228 McCoard, R. W. 43 McGinnis, M. 5, 148n, 152, 154, 164, 165, 167, 171 Meadows, K. 59, 60, 65, 81–82 meaning and form 247–248 Merge operation 2, 163 Mezhevich, I. 149 Michaelis, L. A. 43 Michelson, K. and M. A. Doxtator 17, 22 Mi’gmaq 265–270 Milsark, G. L. 220 Minimalist Program 2, 4, 215, 235 ϕ-features 4–5 Mithun, M. 181 modality and habitual and stative pasts 29 as temporal anchoring 24–27 verbs to auxiliaries 9 modalizers 29 Modified Contrastive Specification 247, 251 modifying nominals 127–133 Moens, M. 68 mood future tense as predictive 26 modality as temporal anchoring 24–25 Morén, B. 253, 254 Moro, A. 221, 222, 224–226 morphemes order 18, 33 pre-pronominal prefixes 24–25, 27

morphosyntactic inventories 263, 271 morphosyntax distinctive features 9 evidence for masdars as v-selecting 147–154 inventories 257–261 Moulton, K. 143, 145, 157, 180, 181n, 210, 211 Mourelatos, A. P. D. 68, 84 movement 240 and long-distance agreement 237 Move operation 238 Move α 2 nanosyntax 8 Nash, L. 142, 147, 149, 164n, 166–167, 169–170, 175 Navajo 187n negation, and factual modal prefixation 25–26 neoparametric view of syntax 3, 8, 9–10 Nevins, A. 251n, 260 Ngizim 253, 259 nominal inflectional suffixes 90 nominalizations 164 derived 154–155 nominalizing heads, masdars as 141–145 nominative objects 166 Northern Alaskan Iñupiaq 256–257 nouns classification systems 68–69, 85–91 modifying nominals 127–133 nominal 85–91 Noyer, R. R. 250 null determiners 121–124 null expletives 235n number feature hierarchy 265, 267–270 semantic and morphological interpretation 4 number suffixes 152 Nurse, D. 57 object-to-event homomorphism 85 Oehrle, R. 177 Old Church Slavonic (OCS) 258, 263 Old English, modality 10

304 index Onondaga aspect 18–24 background 17 clausal structure 33–34 complex aspect constructions 19, 27–30 Infl morphemes 33–37 temporal anchoring 24–27 optativity 26, 148n optionality 227 organizing nodes 254 Ó Sé, D. 42, 56 Oxford, W. 5–6 palatalization 256–257 Pantcheva, M. B. 8, 52 Paradis, C. and J.-F. Prunet 251 Parallel Structures Model (PSM) 253–255 passivization 158–159 Path heads 52 path PP objects 94 Paul, I. 129n, 180 perfect aspect, Scottish Gaelic 41 restricted perfect aspect (after perfect) 42, 44, 51 restricted prospective aspect 44 unrestricted perfect aspect 41–42, 51 perfective events vs. imperfective events 6–7, 26 Person–Case Constraint (PCC) 5, 259–260 person feature hierarchies 265–270 person features, dependency relations 7 person hierarchies 5–6 second person paradigm 9 Pesetsky, D. and E. Torrego 4 Phase-Impenetrability Condition 215n, 240n ϕ-features 4–5, 235n phonemic inventories 255–257, 259, 271 phonetic forms 248 phonetics, class features 4n phonology 249 allophony 262–263 contrastive features 3–4 contrastive scope 6 features encoding contrasts 9, 251–255 inventories 255–257, 259 substance-free approaches 4n

phrase structure rules 1 poetic function 3 politeness distinction 271 Pollock J.-Y. 132n polydefinites 100, 112, 113–114 possessors 157 precedence feature 40, 51–52, 53–57 Predicate Abstraction 128–129 predicate inversion 221–228 Predicate Modification 114–115, 127–128, 133 predictive mood 26 Preminger, O. 221 Preminger, O. and M. Polinsky 217 pre-pronominal prefixes 24–25, 27 preverbs 153–154 primary object 77–79, 81–82 Prince, E. F. 109n Principle and Parameters framework 2 Probe 217, 218, 241 progressive aspect 50 Scottish Gaelic 41 pronominal possessors 124–125 pronouns 250 Mi’gmaq 265–270 proper names expletive definite articles 106 modifying nominals 131 null determiners 121–124 pronominal possessors 125 restrictive modification by nominals (RMN) 117n proposition feature 40 prospective aspect 42, 43–44, 46–47 unrestricted prospective aspect 42, 51 Proto-Germanic, future tense 8–9 punctual aspect 18, 21–24, 25–26 purposive aspect 18 Pylkkänen, L. 66, 138, 146, 160, 168–169, 173, 178 Quantifier Raising (QR) 225 quantifiers 86–87 Ramchand, G. 34, 39, 45, 50, 61, 84 Ramchand, G. and P. Svenonius 3 Raposo, E. and J. Uriagereka 215n realis 35 Reed, S. L. 39, 41n, 42, 43, 44, 50

index 305 referring expressions (REs) 249–250 Reinhart, T. 61n, 63, 64 restricted feature 40, 56–57 restrictive modification by nominals (RMN) 100, 105–106, 108, 112–117 restructuring 232 Icelandic 233–234 result nominalizations (RNs) 142–143, 170–171 reversed feature 40, 51, 52, 53–56 Rice, K. 181, 199n, 210 Rice, K. and J. P. Avery 249 Richards, M. 215n Richards, M. and T. Biberauer 221 Richards, N. 172 Ritter, E. 60, 89, 90, 109, 111n, 124 Ritter, E. and S. T. Rosen 61, 70–71, 73, 74, 182, 198n, 199 Ritter, E. and M. Wiltschko 3, 61n, 80n, 247, 271 Rizzi, L. 182, 194, 205 Roberts, I. 234 roots semantic 147, 170 suppletion 163 Rosen, S. T. 80 Rosenbaum, P. 180 Roussou, A. and I.-M. Tsimpli 104, 106–107, 118 Rullmann, H. and A. You 36 Rupp, L. 241 Russell, B. 109n Russian 262n Safir, K. 223 Saussure, F. de 8–9, 248 Saxon, L. 181, 205–206, 208 Saxon, L. and M. Siemens 179 Schachter, P. 199n Schmidt, K. H. 153 Schneider-Zioga, P. 217 Schreiner, S. L. R. and A. Carnie 41n, 42–43, 45–50 Schütz, A. J. 259 Schütze, C. T. 219n, 231n, 232 Scollon, R. 181 Scottish English definite articles 100, 101 expletive definite articles 108

Scottish Gaelic 39 aspect contrasts 50–53 aspect morphology 41–45 reversed feature 53–56 single aspectual heads 45–50 tense-aspect-mood (TAM) 56n scrambling 144 second person paradigm 9 secret courier approach to long-distance agreement 221, 227, 236 semantic roots 147, 170 semilattices 254 sentience 58–63, 75 Blackfoot 77–83, 91–95 Shimoyama, J. 185n, 210 Sichel, I. 161 Siemens, M. 179, 182 Sigurðsson, H. Á. 215n, 218, 233n, 240 simplex aspect constructions 19 single aspectual heads 45–50 single output approach 221 Skwxwú7mesh 110, 111n Slave (Slavey) 181, 199n, 210 Slavin, T. 17, 34 small clauses 225–227 Smith, C. 24, 26 Source path 52, 80 South Baffin Inuktitut 56 Spanish, modality 10 speaker preference 218n Speas, M. 205 Spec–Head agreement 229–233, 242 Spell-Out movement 235 Sportiche, D. 236 Sportiche-doubling 236, 238 spray/load alternation 71–72 Standard Theory (Chomsky) 1 Starke, M. 8 stative aspect 18, 20–21, 34–35 stative past 27–30, 35 Steriade, D. 251 Stevens, K. N. and S. J. Keyser 248 Stowell, T. 180, 220 Strong Agreement Hypothesis 229, 231–232, 234–235 sub-events 71 subsequence feature 51–52, 53n Successive Division Algorithm (SDA) 251–252, 253–255

306 index suffixes accompaniment 80–81 benefactive 79–80 nominal inflectional 90 number-marking 152 Old Church Slavonic (OCS) 258, 263 thematic 147–151 verb forms 179 syncretism 261–262 Tagalog 199n teleportation approach to long-distance agreement 220–221, 237 telicity (boundedness) 68–77, 84–85 temporal anchoring 24–27 Tenny, C. 61, 70, 71, 84 Tenny, C. and J. Pustejovsky 84–85 tense future 8–9, 26, 53–56 (see also future tense) modality as temporal anchoring 24–25 perfective vs. imperfective events 26 semantic and morphological interpretation 4 tense-aspect-mood (TAM) 34 Georgian 140, 155 Scottish Gaelic 56n and thematic suffixes 148 thematic suffixes 147–151 there-constructions 219–220, 221–222, 224, 227–228, 236, 238–239, 241 there-raising 221 Thráinsson, H. 233n, 242 Thráinsson, H. et al. 241 Tłı  ¸cho  ¸ 179–182 adverbializers 210 aspect in complement clauses 194–199 causative ats’ele 183–193 causative complements 210–212 clause structure 193–194, 204–209 extended CP 206 independent negation in complement clauses 203–204 independent tense in complement clauses 199–203 orthography 179n–180n question phrases 207–209 syntax 183 translations 182

Torrego, E. 158 transformational rules 220 transformations 1–2 transitive agreement 267–268 transitivity 77–78 Travis, L. de M. 68 Trubetzkoy, N. S. 9, 261 Tse’khene 181, 211 Tsilhqot’in 210–211 Tsúùt’ínà 210 Turkish, gerunds 146 underspecification 262 of vocabulary items (VIs) 9 unergatives 156n uniqueness 101–102, 109, 110–112, 122, 124 Upper Svan 260–261 Valentine, J. R. 68n valued and unvalued forms 4 Vamling, K. 141, 146, 159 Van Valin, R. D. and R. J. LaPolla 95 Vendler–Dowty event classes 74 verbal inflectional system, Onondaga 17 verbal nouns see masdars verbs animacy 90–91 classification systems 69–77, 84–85, 90–91, 92–93 sentience 77–83 suffixation 179 Verkuyl, H. J. 68, 70, 84 vocabulary items (VIs) 153–154, 267 insertion rule 50–51 underspecification 9 Voice 145–146, 152, 154, 156, 164 passive vs. active 263–264 v-selecting masdars 145–172 Welch, N. 193–194, 199, 201n, 203n, 205–206, 207n wh-movement 222, 224, 243–244 Wier, T. R. 141, 143–144, 158 Wilhelm, A. 181, 182, 192, 198, 212 Wiltschko, M. 4, 31n, 60, 68, 84, 85, 86–89, 90, 93, 112n, 247, 250 Wiltschko, M. and E. Ritter 61

index 307 Winnebago 250, 260–261, 265 Wolfart, H. C. 68n Woodbury, H. 18, 19–20, 21, 22, 28n, 34 word order 165–167 Wurmbrand, S. 217, 234 X-bar 216

Yang, C. 260 Yoon, J. and L. Brown 210 Zaenen, A. J. Maling, and H. Thráinsson 240 Zamparelli, R. 134 Zeijlstra, H. 4, 219n Zwart, C. J.-W. 228

OX F OR D STU DIE S IN THE ORETICAL LINGUISTICS General Editor David Adger and Hagit Borer, Queen Mary University of London Advisory Editors Stephen Anderson, Yale University; Daniel Büring, University of Vienna; Nomi Erteschik-Shir, Ben-Gurion University; Donka Farkas, University of California, Santa Cruz; Angelika Kratzer, University of Massachusetts, Amherst; Andrew Nevins, University College London; Christopher Potts, Stanford University; Barry Schein, University of Southern California; Peter Svenonius, University of Tromsø; Moira Yip, University College London published 1 The Syntax of Silence Sluicing, Islands, and the Theory of Ellipsis by Jason Merchant 2 Questions and Answers in Embedded Contexts by Utpal Lahiri 3 Phonetics, Phonology, and Cognition edited by Jacques Durand and Bernard Laks 4 At the Syntax-Pragmatics Interface Concept Formation and Verbal Underspecification in Dynamic Syntax by Lutz Marten 5 The Unaccusativity Puzzle Explorations of the Syntax-Lexicon Interface edited by Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert

13 Aspect and Reference Time by Olga Borik 14 Direct Compositionality edited by Chris Barker and Pauline Jacobson 15 A Natural History of Infixation by Alan C. L. Yu 16 Phi-Theory Phi-Features Across Interfaces and Modules edited by Daniel Harbour, David Adger, and Susana Béjar 17 French Dislocation Interpretation, Syntax, Acquisition by Cécile De Cat 18 Inflectional Identity edited by Asaf Bachrach and Andrew Nevins 19 Lexical Plurals by Paolo Acquaviva 20 Adjectives and Adverbs Syntax, Semantics, and Discourse edited by Louise McNally and Christopher Kennedy 21 InterPhases Phase-Theoretic Investigations of Linguistic Interfaces edited by Kleanthes Grohmann 22 Negation in Gapping by Sophie Repp 23 A Derivational Syntax for Information Structure by Luis López

6 Beyond Morphology Interface Conditions on Word Formation by Peter Ackema and Ad Neeleman

24 Quantification, Definiteness, and Nominalization edited by Anastasia Giannakidou and Monika Rathert

7 The Logic of Conventional Implicatures by Christopher Potts

25 The Syntax of Sentential Stress by Arsalan Kahnemuyipour

8 Paradigms of Phonological Theory edited by Laura Downing, T. Alan Hall, and Renate Raffelsiefen

26 Tense, Aspect, and Indexicality by James Higginbotham

9 The Verbal Complex in Romance by Paola Monachesi 10 The Syntax of Aspect Deriving Thematic and Aspectual Interpretation edited by Nomi Erteschik-Shir and Tova Rapoport

27 Lexical Semantics, Syntax, and Event Structure edited by Malka Rappaport Hovav, Edit Doron, and Ivy Sichel 28 About the Speaker Towards a Syntax of Indexicality by Alessandra Giorgi

11 Aspects of the Theory of Clitics by Stephen Anderson

29 The Sound Patterns of Syntax edited by Nomi Erteschik-Shir and Lisa Rochman

12 Canonical Forms in Prosodic Morphology by Laura J. Downing

30 The Complementizer Phase edited by Phoevos Panagiotidis

31 Interfaces in Linguistics New Research Perspectives edited by Raffaella Folli and Christiane Ulbrich 32 Negative Indefinites by Doris Penka

47 Pseudogapping and Ellipsis by Kirsten Gengel 48 Syntax and its Limits edited by Raffaella Folli, Christina Sevdali, and Robert Truswell

33 Events, Phrases, and Questions by Robert Truswell

49 Phrase Structure and Argument Structure A Case Study of the Syntax-Semantics Interface by Terje Lohndal

34 Dissolving Binding Theory by Johan Rooryck and Guido Vanden Wyngaerd

50 Edges in Syntax Scrambling and Cyclic Linearization by Heejeong Ko

35 The Logic of Pronominal Resumption by Ash Asudeh

51 The Syntax of Roots and the Roots of Syntax edited by Artemis Alexiadou, Hagit Borer, and Florian Schäfer

36 Modals and Conditionals by Angelika Kratzer 37 The Theta System Argument Structure at the Interface edited by Martin Everaert, Marijana Marelj, and Tal Siloni 38 Sluicing Cross-Linguistic Perspectives edited by Jason Merchant and Andrew Simpson 39 Telicity, Change, and State A Cross-Categorial View of Event Structure edited by Violeta Demonte and Louise McNally 40 Ways of Structure Building edited by Myriam Uribe-Etxebarria and Vidal Valmala 41 The Morphology and Phonology of Exponence edited by Jochen Trommer 42 Count and Mass Across Languages edited by Diane Massam 43 Genericity edited by Alda Mari, Claire Beyssade, and Fabio Del Prete 44 Strategies of Quantification edited by Kook-Hee Gil, Steve Harlow, and George Tsoulas 45 Nonverbal Predication Copular Sentences at the Syntax-Semantics Interface by Isabelle Roy 46 Diagnosing Syntax edited by Lisa Lai-Shen Cheng and Norbert Corver

52 Causation in Grammatical Structures edited by Bridget Copley and Fabienne Martin 53 Continuations and Natural Language by Chris Barker and Chung-chieh Shan 54 The Semantics of Evaluativity by Jessica Rett 55 External Arguments in Transitivity Alternations A Layering Approach by Artemis Alexiadou, Elena Anagnostopoulou, and Florian Schäfer 56 Control and Restructuring by Thomas Grano 57 The Interaction of Focus, Givenness, and Prosody A Study of Italian Clause Structure by Vieri Samek-Lodovici 58 The Morphosyntax of Gender by Ruth Kramer 59 The Morphosyntax of Imperatives by Daniela Isac 60 Sentence and Discourse edited by Jacqueline Guéron 61 Optimality-Theoretic Syntax, Semantics, and Pragmatics From Uni- to Bidirectional Optimization edited by Géraldine Legendre, Michael T. Putnam, Henriëtte de Swart, and Erin Zaroukian