Dynamic Antisymmetry and the Syntax of Noun Incorporation

DYNAMIC ANTISYMMETRY AND THE SYNTAX OF NOUN INCORPORATION

Studies in Natural Language and Linguistic Theory VOLUME 84

Managing Editors Marcel den Dikken, City University of New York Liliane Haegeman, University of Ghent, Belgium Joan Maling, Brandeis University Editorial Board Guglielmo Cinque, University of Venice Carol Georgopoulos, University of Utah Jane Grimshaw, Rutgers University Michael Kenstowicz, Massachusetts Institute of Technology Hilda Koopman, University of California, Los Angeles Howard Lasnik, University of Maryland Alec Marantz, Massachusetts Institute of Technology John J. McCarthy, University of Massachusetts, Amherst Ian Roberts, University of Cambridge

For further volumes: http://www.springer.com/series/6559

DYNAMIC ANTISYMMETRY AND THE SYNTAX OF NOUN INCORPORATION

by

Michael Barrie University of Ottawa, Ottawa, ON, Canada

123

Michael Barrie University of Ottawa Department of Linguistics Arts Hall 70 Laurier Avenue East K1N 6N5 Ottawa Ontario Canada [email protected]

ISSN 0924-4670 ISBN 978-94-007-1569-1 e-ISBN 978-94-007-1570-7 DOI 10.1007/978-94-007-1570-7 Springer Dordrecht Heidelberg London New York Library of Congress Control Number: 2011929131 © Springer Science+Business Media B.V. 2011 No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

This monograph is a revised version of my 2006 dissertation at the University of Toronto. I have included more discussion on various issues in noun incorporation in Northern Iroquoian languages based on changes in my understanding of the material since 2006. I have also added additional material gathered during postdocs at the University of British Columbia and the University of Ottawa. Much of the data on Northern Iroquoian languages comes from my own fieldwork and from several published sources. The system of interlinear glosses varies significantly from one source to the next. For consistency, I have employed a uniform set of abbreviations for the interlinear glosses throughout. I wish to thank my consultants Daisy Elijah (Oneida), Nora Carrier and Gloria Williams (Onondaga), Elizabeth Herrling (Halkomelem), and Beatrice Bullshields (Blackfoot). Without their assistance, I would not have been able to undertake this research. In addition, I wish to thank several colleagues of mine for their assistance with the relevant literature and/or data including Diane Massam (Niuean), Arsalan Khanemuyiour, Jila Ghomeshi, and Nick Pendar (Persian), Bettina Spreng and Martina Wiltschko (German), Will Seto, Lisa Seto and Max Mak (Cantonese), Yosuke Sato (Japanese), and Martina Wiltschko (Halkomelem). Parts of this monograph have been presented at numerous conferences including NELS35 and 36, the Linearization workshop at DGfS in 2010, the annual meetings of the Canadian Linguistic Association and the Linguistic Society of America, and the Workshop on Structure and Constituency in Languages of the Americas. I wish to thank the audiences there for insightful questions and suggestions that led me to revise and, hopefully, improve upon the details. Numerous people have listened to me talk about my ideas presented here or have collaborated with me on related projects, and have given useful and substantial feedback, or have simply asked vital, thought-provoking questions that allowed me to improve my analysis. This list includes (in no particular order) Elizabeth Cowper, Diane Massam, Alana Johns, Carrie Dyck, Martina Wiltschko, Gabriela Alboiu, Éric Mathieu, Joel Dunham, Bettina Spreng, Heather Bliss, Arsalan Khanemuyipour, Jila Ghomeshi, Kenji Oda, Yosuke Sato, plus others I have surely forgotten. Finally, I wish to acknowledge the financial support of SSHRC doctoral and postdoctoral fellowships and a Killam post-doctoral research fellowship. v

Abbreviations

A ABS ACC AG AGR APPL ART BEN BPS C CAUS CIS CL CMP COLL CONT CORE CUC DEG DEM DR DU DUC DUR EPEN ERG EXCL EZ F FACT FUT

absential absolutive accusative agent agreement applicative article benefactive Bare Phrase Structure common noun causative cislocative classifier completive collectivizer continuous determiner for core arguments (as opposed to oblique arguments) Canonical Use Constraint degree demonstrative bivalent direct (indicates that the object is lower on the thematic hierarchy than the subject) dual dualic durative epenthetic ergative exclusive ezafe vowel in Persian feminine factual mood future

vii

viii

GEN GER HAB INANIM INC INCL IND INF INSTR INTR INVIS ITER I.TH JOIN LCA LF LNK LOC M M/A N.CL NE NEG N/F NFS NI NOM NT NZLR OBJCL OBL OBV OPT P PART PAT PF PL POSS PRED PRF PRFV PROG PROX PSR

Abbreviations

genitive gerund habitual inanimate incompletive inclusive indicative infinitive instrumental intransitive invisible iterative intransitive theme joiner vowel, an epenthetic vowel in Northern Iroquoian languages Linear Correspondence Axiom logical form linker vowel local masculine mode-aspect noun class a particle in Northern Iroquoian languages of unclear function negative neuter/feminine noun forming suffix noun incorporation nominative neuter nominalizer object clitic oblique obviative optative proper noun participle patient phonetic form plural possessive predicative marker perfect perfective progressive proximate possessor

Abbreviations

PST PUNC PURP Q REAL REFL REL SBJV SG SRFL STAT TH TNS TR UG VCL VTA

ix

past tense punctual aspect purposive question particle realis reflexive relativizer subjunctive singular semi-reflexive stative theme vowel (indicates argument hierarchy relations in Algonquian languages) tense transitive Universal Grammar verbal classifier verb transitive inanimate

This is Blank Page Integra

x

Contents

1 Introduction . . . . . . . . . . . . . . . . 1.1 Outlook and Goals . . . . . . . . . . 1.2 Noun Incorporation . . . . . . . . . . 1.2.1 Properties of NI . . . . . . . . 1.2.2 On the Syntactic Nature of NI 1.3 Conclusion and Outline of Book . . . References . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

1 2 5 5 8 18 19

2 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . 2.1 A Bit of History of Phrase Structure . . . . . . . . . . . . . 2.2 Bare Phrase Structure . . . . . . . . . . . . . . . . . . . . . 2.3 Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Head Movement . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Alternative Accounts of Linearization . . . . . . . . . . . . 2.5.1 Demerge and Concatenate: SOV as Underlying Order 2.5.2 Parameterized Order . . . . . . . . . . . . . . . . . 2.5.3 Departures from the LCA . . . . . . . . . . . . . . . 2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

23 23 26 30 34 40 41 42 47 48 48

3 Unifying Antisymmetry and Bare Phrase Structure . . . 3.1 The Problem of Mutual C-Command . . . . . . . . . . 3.2 Previous Accounts . . . . . . . . . . . . . . . . . . . 3.2.1 Guimarães (2000) and Self-Merge . . . . . . . 3.2.2 Nunes and Uriagereka (2000) . . . . . . . . . . 3.2.3 Richards (2001) . . . . . . . . . . . . . . . . . 3.2.4 Conclusion . . . . . . . . . . . . . . . . . . . 3.3 The LCA and BPS . . . . . . . . . . . . . . . . . . . 3.4 Proposal . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Alternatives to Complement-to-Spec Roll-Up . . . . . 3.5.1 Spec-To-Spec Movement and Romance Clitics . 3.5.2 Avoidance of the Initial Merger Problem . . . .

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

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

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

53 53 54 57 60 61 61 61 71 77 78 81

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

. . . . . . .

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

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

xi

xii

Contents

3.6 Linearization and Late Insertion . . . . . . . . . . . . . . . . . . 3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Noun Incorporation in Northern Iroquoian . . . . 4.1 Northern Iroquoian Languages . . . . . . . . . 4.1.1 Clause Structure of Northern Iroquoian . 4.1.2 Nominal Structure in Northern Iroquoian 4.2 Patterns of NI in Northern Iroquoian . . . . . . 4.2.1 Productivity of NI . . . . . . . . . . . . 4.2.2 Nominal Roots . . . . . . . . . . . . . 4.3 Analysis . . . . . . . . . . . . . . . . . . . . . 4.3.1 Previous Analyses . . . . . . . . . . . . 4.3.2 NI as Phrasal Movement . . . . . . . . 4.4 Properties of Iroquoian NI . . . . . . . . . . . 4.4.1 NI in Ditransitives . . . . . . . . . . . . 4.4.2 NI and Overt DPs . . . . . . . . . . . . 4.5 Conclusion . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . .

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

85 88 88

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

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

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

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

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

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

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

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

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

93 93 94 97 100 101 106 107 107 112 118 118 121 123 123

5 Noun Incorporation and Its Kind in Other Languages 5.1 The Structure of Nominals and V/IN Order . . . . 5.2 NI in Sierra Popoluca . . . . . . . . . . . . . . . . 5.3 English Gerunds . . . . . . . . . . . . . . . . . . . 5.3.1 Description of NI in English Gerunds . . . 5.3.2 Analysis . . . . . . . . . . . . . . . . . . . 5.4 German Progressives . . . . . . . . . . . . . . . . 5.5 Persian ‘Long Infinitive’ Constructions . . . . . . . 5.6 Tamil Noun Incorporation and Coordination . . . . 5.7 Adverb Incorporation in Blackfoot . . . . . . . . . 5.8 Conclusion . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . .

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

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

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

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

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

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

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

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

127 127 130 132 133 135 138 143 147 150 154 155

6 V+N Order . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Polynesian Pseudo Noun Incorporation . . . . . . . . 6.1.1 The Structure of Niuean Nominals . . . . . . 6.1.2 Pseudo Noun Incorporation . . . . . . . . . . 6.1.3 Postlude on Tukang Besi Object Incorporation 6.2 Romance Compounds . . . . . . . . . . . . . . . . . 6.2.1 Description of V+N Compounds . . . . . . . 6.2.2 Analysis of Romance Compounds . . . . . . 6.3 Mapudungun . . . . . . . . . . . . . . . . . . . . . 6.4 Upriver Halkomelem . . . . . . . . . . . . . . . . . 6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . .

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

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

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

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

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

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

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

159 160 161 163 165 167 168 170 172 176 181 183

Contents

7 Conclusion . . . . . . . . . . . . 7.1 Summary . . . . . . . . . . 7.2 Limitations . . . . . . . . . 7.3 Conclusions and Implications References . . . . . . . . . . . . .

xiii

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

185 185 186 190 194

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

197

Chapter 1

Introduction

This monograph is a study of Noun Incorporation (NI) within the framework of Dynamic Antisymmetry. It deals primarily with NI in Northern Iroquoian (represented principally by Oneida and Onondaga, with some discussion on Mohawk); however, I also consider a number of other cases of NI and related phenomena in a variety of languages. I propose that NI is the direct result of the need to satisfy the Linear Correspondence Axiom (LCA). NI arises when the verb selects a bare nominal root as a complement, thereby creating a point of symmetric c-command. This point of symmetry is resolved by the nominal root raising to the specifier of VP. My account is couched within two recent, well-received proposals on phrase structure, which I meld into a single cohesive framework. Specifically, I examine Bare Phrase Structure (BPS) (Chomsky 1994) and Antisymmetry (Kayne 1994) and propose that the insights of both proposals can be maintained if we adopt a Dynamic Antisymmetric view as proposed by Moro (2000; 2004). Moro proposes that movement is driven by the need to satisfy the LCA, and furthermore that this movement takes place overtly. Thus, if two terms cannot be linearized because of their noncompliance with the LCA, something must move so that the LCA is satisfied. In particular, I examine the case of two symmetrically c-commanding heads; a situation that arises upon the initial merger of two heads (the Initial Merger Problem). This is precisely the situation in the case of NI as just described above. Thus, NI is purely a result of the phrase structural properties of the VP. This eliminates the need for a macroparametric mechanism, such as the Morphological Visibility Condition (Baker 1996). I will also show that not only does an LCA violation (in particular, symmetric c-command) trigger movement, it can also trigger a cascade of movements often referred to as Complement-to-spec roll-up in the literature.1 This approach predicts that NI should typically appear with N+V order. Indeed, this order is significantly more common than V+N order (Caballero et al. 2008); however, it is by no means universal, as there are numerous counter-examples. I discuss how NI with V+N order arises in later chapters.

1 These will come up later in the discussion where appropriate. Representative examples include snowballing (Aboh 2004a) and intraposition movement (Rackowski and Travis 2000).

M. Barrie, Dynamic Antisymmetry and the Syntax of Noun Incorporation, Studies in Natural Language and Linguistic Theory 84, DOI 10.1007/978-94-007-1570-7_1,  C Springer Science+Business Media B.V. 2011

1

2

1

Introduction

The remainder of this chapter is structured as follows. Section 1.1 presents the major goals of this monograph and gives a brief overview of how these goals will be accomplished. Section 1.2 introduces the phenomenon of NI and presents its properties that are pertinent to the discussion. Section 1.3 is an outline of the remainder of the monograph.

1.1 Outlook and Goals I identify the following goals of this monograph: • • • •

to present a thorough analysis of NI in Northern Iroquoian, to recast (Dynamic) Antisymmetry in a BPS framework, to argue for a Dynamic Antisymmetric analysis of NI, and to illustrate the Dynamic Antisymmetric analysis of NI with data from a variety of other languages.

In regards to the first goal I describe in detail the various properties of NI that are particular to Northern Iroquoian and show how the proposed analysis for NI coupled with an innovative structure for the doubling construction accounts for the specific properties of NI in this language group. A general description of NI is found in the following section; however, I also discuss several additional properties of NI found in Northern Iroquoian. Specifically, I discuss the kinds of nominals that can undergo NI and their morphological properties. I also discuss how NI in Northern Iroquoian interacts with wh-movement. In short, an in depth description of NI in Northern Iroquoian is presented along with the general properties of NI cross-linguistically. The facts presented here are based on a vast descriptive (Lounsbury 1949, 1953; Froman et al. 2002; Woodbury 1975b, 1975a, 2003; Michelson and Doxtator 2002; Bonvillain 1972) and theoretical (Michelson 1991; Baker 1988, 1996; Postal 1979) literature on this topic, as well as various field trips conducted from 2003 to 2005 on Oneida and 2005 to the present on Onondaga. The discussion on the general properties of NI is based on various cross-linguistic surveys on the topic (Mithun 1984; Caballero et al. 2008; Gerdts 1998; Massam 2009). Accommodating Antisymmetry and BPS in a unified framework is not a straightforward task. I review and build on previous work on this topic (Guimarães 2000; Oishi 2003; Richards 2001a; Uriagereka 1999; Kayne 2009a) and adapt a version of the LCA that is compatible with the machinery of BPS. This comprises the bulk of Chapter 3. Briefly, BPS requires the initial merger to be between two heads, giving rise to a violation of the LCA as mentioned above. To resolve this violation, the selected head raises to the specifier of the selector. The next head that this complex merges with, however, will be in a symmetric c-command relation with the head in specifier from the previous step, triggering another round of raising to satisfy the LCA. This process sets off a cascade of roll-up type movement referred to

1.1

Outlook and Goals

3

above. Of course, the derivation does not continue this way indefinitely. I resolve this situation by making the LCA sensitive only to heads with phonological content, as suggested by Moro (2000). Thus, a phonologically empty head halts this type of movement. I suggest that NI in particular highlights this fact about BPS and initial merge since it typically involves merger of two lexical heads, hence two heads with phonological content. Non-NI constructions involve initial merge between a lexical head and a functional head or between two functional heads – both situations of which involve the possibility of a phonologically null head, thus obscuring the Initial Merger Problem. I now discuss briefly how the current proposal captures the fact that a full DP complement to a verb and a bare noun2 complement to a verb often exhibit different linearization properties. The proposed analysis accounts for this asymmetry. When the verbal root, V, selects a full DP as a complement, V asymmetrically c-commands the material inside the DP in compliance with the LCA. When the V selects a bare nominal root as a complement, the two elements c-command each other, in violation of the LCA. Following Moro’s Dynamic Antisymmetry (2000), the nominal root raises to SpecVP to satisfy the LCA. By contrast, when V selects a full DP as a complement, the verb asymmetrically c-commands the material inside the DP. In the extended nominal projection alone a number of functional projections have been identified with syntactic and semantic effects (Ritter 1992, 1993; Szabolcsi 1983; Abney 1987; Cheng and Sybesma 1999; Ghomeshi 2003; Megerdoomian 2008); however, whether the head is overtly realized with a particular morpheme is subject to a great deal of cross-linguistic variation, and many heads are phonologically empty. This gives rise to a great deal of variation in the surface realization of the morpheme order of DPs. In any event, when the verb selects a full DP complement, in the vast majority of cases, the verb will asymmetrically c-command all the material inside the DP. As already stated, the empirical foundation for this study consists of NI and related phenomena. I offer an extensive discussion and analysis of NI in Northern Iroquoian in Chapter 4, and I discuss similar facts in a variety of other languages in Chapters 5 and 6. In the following examples, the verbs are shown in boldface and the complement nouns are italicized to show their relative positions. The (a) examples contain full DP complements and the (b) examples contain nominal complements that have undergone NI or a closely related phenomenon, which I later argue to be essentially the same as NI. Again, notice that the NI forms appear with the order N+V (V+N is discussed in Chapter 6), while the order of V and DP varies.

2 Throughout

this monograph, I use the term bare noun to mean functionally bare (i.e., a bare root or a bare nP), rather than morphologically bare in the sense of Giorgi and Longobardi (1991). This distinction is important because a morphologically bare noun in this other sense could contain a large number of functional projections with phonologically empty heads.

4

1

Introduction

1. NI in Onondaga (Northern Iroquoian) a. wahahní:no˛ T neT ganakdaT waT -ha-hnino˛ -T FACT -3. SG . M . AG -buy-make- PUNC ‘I’m making a house for him.’

neT NE

b. wahanakdahní:no˛ T waT - hanakt- ahnino˛ FACT - 3.SG . M . AG bed JOIN buy ‘He bought a bed.’

ka-nakt-aT AGR -bed- NFS

T

PUNC

2. NI in English Gerunds a. Alice enjoys collecting stamps. b. Alice enjoys stamp-collecting. 3. German Progressives3,4 a. Ich esse die the I eat.1.SG ‘I’m eating the apples.’

Äpfel. apple.PL

b. Ich bin beim Äpfelessen. I am at.the apple.PL- eat.INF ‘I’m eating apples.’ / ‘I’m busy apple-eating.’ 4. Persian Long Infinitives5 a. sima æz xundæn-e in ketab xoš-eš mi-yad Sima from reading-EZ this book good-3 SG CONT-come.3 SG ‘Sima likes reading this book.’ b. sima æz ketab xundæn xoš-eš mi-yad Sima from book reading good-3 SG CONT -come.3 SG ‘Sima likes reading books.’ In English, example 2, full DP complements appear to the right of the verb, and what is demonstrably a bare noun appears to the left of the gerundive form of the verb. Gerunds are taken up in more detail in Chapter 5. These data highlight the asymmetry between full DP objects and bare noun objects. Specifically, whether 3 Of course OV word order is possible with full DP objects, as German is an SOV language. The point here is that full DP objects can appear in either VO or OV word order, depending on whether the verb appears in 2nd position or sentence-finally. With bare nominal objects as in the progressive beim construction, only OV word order is found. 4 Unless otherwise stated, all German data are from Bettina Spreng and Martina Wiltschko (personal communication). 5 Unless otherwise stated, all Persian data are provided by Jila Ghomeshi, Arsalan Kahnemuyipour, and Nick Pendar (personal communication).

1.2

Noun Incorporation

5

the object appears before or after the verb depends on whether the object is a full DP or bare noun. Of course, other factors may affect VO versus OV word order (see footnote 3), but the generalization still holds. The theory of phrase structure I advocate here accounts for this asymmetry in a straightforward way. There are, of course, instances of NI with the order V+N. I examine several particular cases in detail in Chapter 6: pseudo noun incorporation (PNI) in Niuean (Massam 2001), V+N compounds in Romance (such as Spanish toca-discos ‘record-player’ and French tranche-oeufs ‘egg slicer’), lexical suffixation in Upriver Halkomelem, and NI in Mapudungun. I argue that V+N order results from the presence of additional functional material in the IN that permits the LCA to linearize the verbal root with the elements inside the IN. Thus, no movement needs to take place and the order V+N surfaces. The major scientific contribution of this proposal is twofold. First, it accounts for the robust cross-linguistic generalization that when a verb takes a bare noun as a complement, we observe the order N+V more frequently than V+N (Caballero et al. 2008), and when the verb takes a full DP complement, we get roughly a 50–50 split between the orders V+DP and DP+V (Ruhlen 1975; Dryer 2008). In Chapter 2, I address the lack of a universal N+V order in NI constructions. Second, NI and a wide range of related phenomena fall out naturally from the theory of linearization proposed here. I do not need to posit any new syntactic mechanisms, so NI comes for free, in effect. As a consequence, we do not need a variety of different mechanisms to account for the phenomena in 2–4. Thus, rather than having to posit different mechanisms to account for NI and related phenomena, I propose that existing mechanisms accomplish this task.

1.2 Noun Incorporation NI has been the subject of study for well over 100 years now (Kleinschmidt 1852; Cuoq 1866; Kroeber 1909; Sapir 1911). It in fact continues to be the subject of much heated debate and controversy, especially concerning its status as a syntactic, morphological or lexical phenomenon (Baker 1996; Baker et al. 2005; Mithun 1984, 1986b; Sadock 1980, 1986; Rosen 1989). The works of several authors (principally those in the references just mentioned) have converged on a common set of prototypical properties of NI, which I describe below. I also discuss the typological work of Caballero et al. (2008), which describes in detail those properties of NI concerning the order between the IN and the verbal root.

1.2.1 Properties of NI Northern Iroquoian languages illustrate many prototypical properties of NI (Rice 1991; Baker 1988, 1996; Mithun 1984; Woodbury 1975b), which are described in depth in Chapter 2 and are analyzed in Chapter 3. These include: (i) the ability to incorporate a direct object, instrument, path or subject of an unaccusative, but not an

6

1

Introduction

indirect object, goal, source, comitative or subject (example 5, Mohawk, and example 6, Onondaga),6 (ii) the ability for the incorporated noun to appear with a full DP double or modifier (examples 7–8, Onondaga), (iii) the surface order N+V (observed throughout, though see the forthcoming discussion on order in the following paragraphs), (iv) the optionality of NI (example 9, Oneida), (v) frozen scope (observed in 10), and (vi) idiomatic meanings. I will show that properties (i) and (iii) boil down to a situation of symmetric c-command triggering NI via movement. Properties (ii) and (iv) are related to the structure for NI I propose in Chapter 3. These are illustrated in the following examples. Note that the properties illustrated below apply generally to all Northern Iroquoian languages (Baker 1988, 1996; Bonvillain 1972; Woodbury 1975b, 1975a; Mithun 1984; Abbott 2000; Froman et al. 2002). 5.

NI in Mohawk Ditransitives (Baker 1996: 207) a. t-a’- khey- athvni- tsher- u- ’ ne owira’a CIS - FACT- 1. SG . AG :3. SG . F. PAT-ball- NZLR - give- PUNC NE baby ‘I gave the ball to the baby.’ b. # t- a’- ke- wir- u- ’ ne athvno CIS - FACT- 1. SG . AG - baby- give- PUNC NE ball ‘I gave the baby to the ball.’ (NOT ‘I gave the ball to the baby.’)

6.

a.

b.

c.

7.

honathahidákheT honathah3.PL . M . PATSRFL path‘They are walking on a path.’

(Woodbury, 2003: 282)

waT hageT nhyayé˛ hdaT hakwaT FACT3. SG . M . AG :1. SG . PAT ‘He hit me with a stick.’ ohahaná:we˛h oahah3.SG . NT. PATroad‘The road is wet.’

T

itakherun-

T

nhyastick-

PURP

(Woodbury, 2003: 928) aye˛hd- daT JOIN hitPUNC

(Nora Carrier, Gloria Williams, speakers) nawe˛h JOINwetSTAT

a-

waT gnasgwohaeT neT gwísgwis knaskwohaewaT FACT- 1. SG . AG - animalwash‘I washed the pig.’

T

neT

PUNC

NE

(ibid.) kwiskwis pig

6 See Öztürk (2009) and Cagri (2009), however, for discussions on so-called subject incorporation in Turkish as well as Polinsky (1990) for subject incorporation in Chukchi.

1.2

Noun Incorporation

8.

a.

waT khní:no˛ T neT ne˛geT ganakdaT khnino˛ waT FACT- 1. SG . AG buy‘I bought this bed.’ waT gnakdahní:no˛ T neT ne˛geT waT knakt- aFACT- 1. SG . AG - bedJOIN ‘I bought this bed.’

b.

9.

7

NI in Oneida a. waPkhni:nú: ká:sleht waP-k-hninu-´: FACT-1. SG . AG -buy- PUNC ‘I bought a car.’ b.

waPkePslehtahni:nú: waPkFACT1. SG . AG ‘I bought a car.’

T

neT

ne˛keT

PUNC

NE

DEM

(ibid.) kanaktaT bed

hnino˛ buy-

T

neT

(ibid.) ne˛keT

PUNC

NE

DEM

(Daisy Elijah, speaker) ka-Psleht-Ø 3. SG . NT-car- NFS

eJOIN -

Pslehtcar-

aJOIN -

hninubuy-

´: PUNC

10.

Yah te-wake-nakt-a-hninu NEG -1. SG . PAT-bed- EPEN -buy. STAT No ‘I didn’t buy the bed.’ Or ‘I didn’t buy a(ny) bed.’ =‘There is a bed that I didn’t buy.’ (Baker, 2003: 116)

11.

(Onondaga, Gloria Williams and e˛dyo˛ dwe˛nage˛T sä:hgwaT e˛-d-yo˛ -ad-we˛n-a-ge˛T sR-hgw-aT Nora Carrier, speakers) FUT -DUC -3.SG . F. AG - SRFL-word- EPEN -rest/pillow-INSTR - PUNC ‘They put it under a pillow’ / ‘They tabled the matter.’

Example 5 shows a ditransitive, in which the direct object, but not the indirect object, can undergo NI. Example 6 shows NI of a path, an instrument and an unaccusative subject. Example 7 illustrates doubling of the IN (animal) with a hyponymous object (pig). Example 8 shows a stranded modifier (a demonstrative).7 Example 9 shows that NI is optional in Oneida. Example 10 shows that the IN in Mohawk exhibits frozen scope with respect to a scope bearing element outside the verbal complex. Finally, example 11 shows that NI often gives rise to idiomatic readings. Mithun (1984) also observes that NI constructions are often used in institutionalized activities and are sometimes restricted to particular kinds of uses. This is akin

7 The

use of the label modifier stranding to describe example 8 has theoretical implications that I discuss in greater detail in Chapter 4.

8

1

Introduction

to the Canonical Use Constraint (CUC) discussed by Kiparsky (1997). All of the properties just discussed hold in Northern Iroquoian generally. Perhaps unsurprisingly so, it will become clear throughout the discussion, especially in the later chapters, that not all languages with NI exhibit the exact same set of properties for this phenomenon. The landmark discussion on the types of NI, its discourse properties and effects on argument structure is Mithun (1984), a typology which maintains widespread currency in the literature today. I turn now to a discussion of the typological properties of NI as it relates to linear order. Based on Kayne’s (1994) principle of Antisymmetry, Baker (1996) proposed that NI is universally N+V – a claim he backed down on (Baker et al. 2005) in light of the V+N order found in Mapudungun. Caballero et al. (2008) conducted the first wide-scale cross-linguistic study on linear order in NI, where they show that N+V order is significantly more prevalent in NI constructions than V+N order. Note that OV and VO order (with full DP objects) is pretty much evenly distributed, (Ruhlen 1975; Dryer 2008). Caballero et al. report the following results to their study (adapted from source document). 12. Productive8 - OV - VO Non-productive Total

N+V

V+N

Total

26 (67%) 15 9 8 (100%) 36 (73%)

13 (33%) 2 10 0 (0%) 13 (27%)

39 17 19 8 49

As the reader can see, there is a distinct preference for N+V order; however, this difference is more pronounced because of the instances of non-productive NI (i.e., where such forms are lexicalized). Nevertheless, there is still a significant preference for N+V order in languages with productive NI.9

1.2.2 On the Syntactic Nature of NI I conclude this section with some arguments in favour of maintaining a syntactic analysis of NI, in spite of various counter-arguments. I discuss the range of elements that can be incorporated, idiomatic readings associated with NI constructions, and frozen scope. One argument against a syntactic treatment of NI lies in the fact that elements other than the direct object can be incorporated (Rosen 1989; Spencer 1995; 8 Not all languages with productive NI were classified as OV or VO languages, hence the discrepancy in totals. 9 Note that I have nothing to say about the correlation between VO/OV order and V+N, N+V order despite the fact that this is significant.

1.2

Noun Incorporation

9

Di Sciullo and Williams 1987). In particular, paths, locations, instruments and adverbs can undergo NI (or adverb incorporation in the case of the latter), in several languages, as discussed above. Excluded from NI constructions, however are goals, benefactives, sources and comitatives. This seemingly disparate set of elements (direct objects, paths, instruments, locations, and adverbs) is actually a natural class under a Larsonian (1988; 2004) analysis of VP-internal elements. Under Larson’s analysis, these elements are introduced in the specifier or complement of a VP shell. VP

13.

V’

XP

VP

V

V’

XP V

XP

Thus, when one of these elements is introduced as a bare root in complement position, it undergoes NI the same way the direct object does. The other elements, goals, sources, benefactives and comitatives, are introduced by functional projections of various kinds, or, in the case of comitatives, involves some other constructions altogether. The inability for indirect objects to undergo NI has been noted in the literature before (Mithun 1984; Baker 1996: 297ff.). The exact structure of ditransitives plays an important role here. Pylkkänen (2008), for instance, has argued that goal arguments are introduced by a low applicative head (see Larson 2010 for arguments against this approach, though).10 She also introduces other elements such as instruments by applicative heads; thus, this approach fails to capture the natural class described above. Larson’s (1988; 1990) original discussion of the double object construction placed the indirect object and goal directly inside a VP shell, which also fails to capture the natural class described above. Baker (1988; 1996) argues that the indirect object/goal/source is introduced in the VP shell by a phonologically null preposition. Consider the following example (Baker 1996: 298).

10 Larson

takes issue with the fact that Pylkkänen has divorced the goal argument from the event semantics. Thus, in a construction such as John baked Mary a cake, Mary is related to the cake by a to.the.possession.of predicate, but is not related to the event. Fatally, however, Larson argues that this approach fails to capture the fact that the subject, John, must be responsible for Mary’s receipt of the cake. Under Pylkkänen’s approach, the sentence above is compatible with a scenario in which John baked a cake and someone else brought it to Mary. Since I do not ultimately adopt Pylkkänen’s approach, I won’t pursue this line of reasoning further here.

10

1

14.

Introduction

VP V’

NP N

V

ball

give

PP P

NP

Ø

N baby

Baker argues that the null preposition provides the semantic distinctions necessary to encode direction in transfer of possession verbs (buy from vs. give to). Note that indirect objects in Northern Iroquoian languages can be either sources or goals, so the distinction must be encoded somewhere (see also Michelson 1991). Consider the following Onondaga examples (Woodbury 2003). 15. a. waT ho˛ wahyahní:no˛ T waT -ho˛ wa-ahy-a-hnino˛ -T FACT-3. AG :3. SG . M . PAT -fruit- JOIN-buy-PUNC ‘Someone bought fruit from him.’ b. waT khé:yo˛ T waT -khe-o˛ -T FACT -1.SG . AG :3. SG . F. PAT -give-PUNC ‘I gave it to her.’ It is not clear how Baker’s approach extends to paths, locations and instruments, however. Under the approach above, a null P is necessary to give rise to the correct semantics for the indirect object. Assumedly, a null P would be necessary to give rise to the correct reading (instrumental, locative, etc.) to the relevant nominal phrase. It would be unclear, then, what blocks incorporation of the indirect object, but allows it for paths, locations and instruments. Beck and Johnson (2004) also argue that indirect objects are introduced by a functional projection based on its behaviour with the adverb again. Crucially, double object constructions allow both a repetitive and restitutive reading with again. Consider the following example, with the repetitive and restitutive readings spelled out, respectively (Beck and Johnson 2004: 113). 16.

Thilo gave Satoshi the map again.

17.

a.

Thilo gave Satoshi the map, and that had happened before.

b.

Thilo gave Satoshi the map, and Satoshi had the map before.

1.2

Noun Incorporation

11

Beck and Johnson argue that this ambiguity arises as a result of an additional predicative head that relates the direct object to the indirect object (while still relating both to the main event, thereby avoiding the problems in Pylkkänen’s proposal, see note 10). Thus, Beck and Johnson propose the following structures for double object constructions and DP + PP constructions. The HaveP and PP serve as adjunction sites for again, accounting for the restitutive readings. 18. a. [vP Thilo [v’ v [VP give [BECOME[HaveP Satoshi [Have the map]]]]]] b. [vP Thilo [v’ v [VP the map1 give [BECOME[PP PRO1 to Satoshi]]]]] Crucially, the ambiguity noted by Beck and Johnson is absent with locations, instruments and adverbs. In the following examples, no restitutive reading is possible (although the adverb again may take scope over different parts of the VP).11 19.

a. John is cooking in the kitchen again. b. Joyce is trimming flowers with a knife again. c. Suzanne is typing very quickly again.

For the current purposes I will adopt the proposal of Beck and Johnson and assume the structures in 18 for ditransitive constructions, though I leave out the BECOME operator as I do not discuss the formal semantic properties here.12 The proposal that adverbs, direct objects, instruments and locations form a natural class also receives interesting empirical support from an unexpected source. Non-canonical objects in Mandarin Chinese also comprise the same set of elements, namely paths, locations and instruments, but exclude goals, sources, benefactives and comitatives (Li 2010). The aforementioned elements can appear in preverbal PPs (the canonical position for adjunct PPs), while paths, locations and instruments can optionally appear in the canonical object position in Mandarin (immediately to the right of the verb) without a preposition. Consider the following Mandarin 11 Note

that a restitutive reading is available with 19b; however, this is independent of the instrumental phrase. The same reading is available with the following sentence.

i. Joyce is trimming flowers again (and they used to be trimmed before). This reading is somewhat odd because cut flowers don’t typically re-grow their stems (thus requiring re-trimming). Nevertheless, it is available. Under Beck and Johnson’s approach, we would assume a stative BE predicate, which the adverb again could take scope over. ii. Joyce CAUSE flowers BE trim. Crucially, the instrumental phrase is not introduced by an additional predicative head since or else we would predict yet another restitutive reading. 12 Many of the arguments for the various proposals for the structure of ditransitives rely on a number of semantic distinctions. Further details of these semantic properties in Northern Iroquoian languages will have to wait for future research.

12

1

Introduction

Chinese data. Examples 20 and 21 illustrate non-canonical objects with a locative and instrumental, respectively. Examples 22 – 24 show that the non-canonical object construction is unavailable with benefactives, comitatives and sources. 20.

21.

22.

a.

ta zai canting he at restaurant ‘He eats at the restaurant.’

b.

ta chi canting he eat restaurant ‘He eats at the restaurant.’ / ‘He eats restaurant food.’

a.

ta xihuan yong zhe zhi maobi pen he like use this CL ‘He likes to write with this brush pen.’

b.

ta xihuan xie zhe zhi maobi pen he like write this CL ‘He likes to write with this brush pen.’

a.

wo gei ta zuo I for him make ‘I make clothes for him.’

b.

23.

a.

b.

24.

a.

b.

chi eat

(fan) (meal)

xie write

(zi) (word)

yifu clothes

∗ wo

zuo ta I make him (‘I make things for him.’) wo gen ta zuo I with him make ‘I make clothes with him.’

yifu clothes

∗ wo

zuo ta I make him (‘I make things with him.’)

wo xiang tushuguan I from library ‘I borrow books from libraries.’

jie borrow

shu book

∗ wo

jie tushuguan I borrow library (‘I borrow from libraries.’)

It would be a strange coincidence indeed if NI in the languages under discussion and non-canonical objects in Mandarin Chinese picked out the same set of concepts which are eligible to participate in these constructions. Thus, the non-canonical

1.2

Noun Incorporation

13

objects in Mandarin Chinese coupled with the Larsonian shell hypothesis actually argue in favour of a syntactic approach to NI of instruments, paths, locations and adverbs. NI constructions typically show idiomatic and unpredictable meanings. This was often cited as an argument in favour of a lexical analysis for NI. Consider the following example (Woodbury 2003: 304). 25. waT shagotshógwaks waT - shakoathsohkw-kØ FACT- 3. SG . M . AG :3. SG . F. PAT- SRFL - lipeat- PUNC ‘He kissed her.’ (lit: ‘He ate her lips.’) In 25, the idiomatic reading is available only in the NI construction.13 Marantz (1997) has argued against the notion of idiomatic meanings as an indicator of a lexical component to the grammar based on phrasal idioms in a wide variety of languages such as chew the fat, pull the wool over X’s eyes, make waves, and kick the bucket. These idioms have meanings which are not deducible from their syntax; however, they participate in syntactic processes (such as passivization: The wool was pulled over the customer’s eyes or focus fronting: Even the tiniest of waves I wouldn’t make with that administrator) and cannot be analyzed as syntactic atoms. One may object that these idioms are of a different nature or character than the idiomatic readings of IN constructions. Indeed the idioms in English (and many other languages) include determiners on the nouns rather than bare nominals. Given that I related the similarity of INs with bare nominals below, I consider yet additional data. Consider the following Cantonese data (Barrie 2008). 26.

a.

b.

27.

a.

b.

13 Note

caat3 haai4 polish shoe ‘flatter’ OR ‘polish shoes’ caat3 ni1 deoi3 haai4 CL shoe polish this ‘polish this pair of shoes’ #‘engage in this instance of flattery’ jam2 ca4 drink tea ‘have dim sum’ OR ‘drink tea’ jam2 ni1 di1 drink this CL ‘drink this tea’ #‘have this (bit of) dim sum’

ca4 tea

that this idiom is apparently found only in the Syracuse community in New York. My consultants in Six Nations did not recognize this idiom.

14

1

Introduction

With the bare nominal objects, both the idiomatic and compositional readings are available. When the object contains higher functional material, only the compositional reading is available, exactly like in the instances of NI shown so far. Note that the bare noun cannot be considered morphologically incorporated because it can be separated from the verb by aspectual morphology and can even be focused in some cases, while retaining the idiomatic reading. 28.

a.

jam2 zo2 ca4 PRFV tea drink ‘have had dim sum’ OR ‘have drunk tea’

b.

lei5 lin4 ca4 dou1 mei6 you even tea all not.yet ‘You’ve never even had dim sum before.’

jam2 drink

Another similar illustration comes from light verb constructions in Japanese. Again, consider the following data (Sato 2010). 29.

30.

a.

beknyoo-suru studying-do ‘to study’ OR ‘to discount’

b.

benkyoo-o suru do studying-ACC ‘to study’ NOT ‘to discount’

a.

ryoori-suru cooking -do ‘to cook’ OR ‘to handle well’

b.

ryoori-o suru do cooking - ACC ‘to cook’ NOT ‘to handle well’

Again, the idiomatic readings are available on the bare noun constructions only. Both the Cantonese and the Japanese data provide further illustrations that the availability of idiomatic readings is correlated with the bareness of the noun rather than with any concept of word-hood. While idiomatic readings are found nearly always exclusively with NI constructions, this is by no means universal. Consider the following Chipewyan data (Cook and Wilhelm 1998: 59). 31.

a.

na-jë´ th-the-Ø-Ø-da ITER-hook- M / A -3. SG - VCL -sit ‘S/he is fishing again.’ (lit: sitting with a hook)

b.

jë´ th gha˛ the-Ø-Ø-da hook with M / A -3. SG - VCL -sit ‘S/he is fishing again.’ (lit: sitting with a hook)

1.2

Noun Incorporation

15

Here, the idiomatic reading is found with both the incorporated and nonincorporated forms, again underscoring the lack of involvement of any notion of word-hood in establishing idiomatic readings. Related to the above is the fact that NI also subject to the CUC (Kiparsky 1997), whereby pragmatically or culturally determined application of the incorporated object is understood. This information typically cannot be gleaned from the syntax. For instance in the phrase butter the toast, we understand a particular way of spreading butter on the toast (with a knife, only on one flat surface, not on the edges or on only half of the surface, for example). The following Oneida example (Michelson and Doxtator 2002) illustrates this phenomenon for NI. 32.

wahathwistásheteP waPhaFACT3. SG . M . AG ‘He counted money.’

atSRFL -

hwistmoney-

ashetcount-

eP PUNC

The sentence in 26 cannot refer to counting kinds of money or numbers of bill or coins. It can only refer to counting the sum value of the money. Harley (2008), however, showed that this is not a property of the lexicon, but a property of bare nouns, based on the following English data. 33.

in bed bed at church

∗ on

in the bed on the bed at the church

Unless we wish to posit that forms such as in bed and at church (and, in North American dialects of English, in the hospital) are syntactic atoms spit out from a generative lexicon, the CUC cannot be used as an argument against a syntactic approach to NI. In fact, it shows that NI impinges on our understanding of the syntax of bare or reduced nominal expressions. Finally, I address the issue of frozen scope. NI constructions typically show frozen scope as shown in example 10 above. This has also been cited as a property of words, the idea being that words are syntactic atoms and thus morphemes inside words cannot take scope outside the word. I argue that this is more a property of bare nouns, however. Consider the following Blackfoot example (Glougie 2000). 34.

a.

ihkan-ano-yi-a om-yi piita PST .all-see.TR -3’- PL DEM-OBV eagle ‘They all saw the eagle.’ (∃ > ∀)

b. ihkan-iyapi-ya piita PST .all-see.INTR - PL eagle ‘They all saw an eagle.’ (∀ > ∃) c.

piita ihkan-iyapi-ya PST .all-see. INTR - PL eagle ‘They all saw an eagle.’ (∃ > ∀)

16

1

Introduction

The difference to note between these two examples is that the object in 34a is a full DP with a demonstrative, while the following two examples contain a bare nominal. Note also the intransitive marking on the b and c examples highlighting the fact that they contain bare nominal objects. The bare noun in 34b, c has frozen scope despite being syntactically free in the sentence. Again, it has been shown that a putative property of NI, namely frozen scope, intended to argue for its status as a lexical phenomenon, in fact is merely a property of bare nouns. In fact, van Geenhoven (1998) has shown that frozen scope in West Greenlandic and bare plurals in English pattern identically in terms of scopal relations, thus highlighting the similarity in properties between bare nouns and INs. van Geenhoven gives the following example from West Greenlandic (p. 3, ex 3, Bittner 1994: 118). 35.

Junna Kaali-mit allagar-si-nngi-l-a-q Kaali- ABL letter-get- NEG - IND - TR -3. SG Junna. ABS ‘It is not the case that Juuna got a letter/letters from Kaali.’ #‘There is/are a letter/letters from Kaali that Juuna didn’t get.’

As shown, the IN cannot take scope above negation. Bittner also gives the following English data showing that bare plurals in English also cannot take scope above sentential negation (p. 4, ex (5), citing Carlson 1977: 19). 36.

John didn’t see spots on the floor. =It is not the case that John saw spots on the floor. =There were spots on the floor such that John didn’t see them.

Related to the above discussion is Williams’ (2007) argument that word-internal morphemes cannot take their resolution outside the word. Consider the following data. 37.

a. b.

John described himself. John told self-destruction stories.

In 37a, the word himself finds its resolution higher up in the clause. In other words, himself must be coreferential with John. In 37b, on the other hand, the self in selfdestruction cannot look outside the word to find its resolution. Thus this sentence can refer only to generic self-destruction stories and not to self-destructions stories about John. There are, however, some situations in which word-internal morphemes sometimes do look outside the word for resolution. First, consider the following piece of data that complements 37. 38.

I witnessed John’s self-destruction.

Here, self must be coreferential with John, thereby taking its resolution outside the ‘word’ that contains it. Consider also reflexives in Onondaga.

1.2

39.

Noun Incorporation

17

John waT hadadaehsé˛ thwaT John waT haJohn FACT- 3. SG . M . AG‘John kicked himself.’

atatREFL -

aehse˛thwkick-

aT PUNC

Here, the reflexive morpheme atat must find its resolution outside the word that contains it. Thus, atat must corefer to John. One final example is found in a particular kind of Blackfoot wh-question. 40.

tsa ááni anná Mary anní John omaanistsíkkamiyoowatahpi aniskayi apastaminam? tsa how

aani say

ann-wa Mary DET- PROX Mary

ann-yi John DET John

om-aanist-ikkam-iy-oowat-a-hp-yi 3-DEG-fast-?-eat.VTA-EPEN-NZLR-INANIM. SG ann-yi-ahk-yi DET- OBV- INVIS - INANIM . SG

apastaminamm apple

‘How quickly did Mary say that John ate that apple?’ (embedded scope on ‘how’) The embedded verbal complex contains a degree variable, aanist, which is bound by the wh-element, tsa, in the left edge of the matrix clause. These three lines of evidence illustrate that the word does not act as a syntactic island for resolution of anaphora or variable binding. This section has described the basic properties of NI, which I list below in bullets. It is important to note that not all languages with NI instantiate all the properties below to the same extent. Thus, it is difficult to pin down a single set of defining properties for NI. • • • • •

NI acts on direct objects, locations, instruments and paths NI does not act on indirect objects, subjects, goals, sources or comitatives the IN can often appear with a full nominal double NI constructions typically appear with N+V order NI constructions typically alternate with non-NI constructions, in other words, NI is often optional • the IN often exhibits frozen scope • NI constructions often give rise to special or idiomatic readings This section has also given various arguments in favour of a syntactic approach to NI. These arguments are all based on blurring the distinction between autonomous syntactic and morphological/lexical modules of the grammar. Specifically, I showed that many of the so-called morphological or lexical properties of NI are also

18

1

Introduction

applicable to a variety of syntactic constructions. These included non-canonical objects, idiomatic readings, and frozen scope. In the case of non-canonical objects, I showed that the range of incorporable elements (direct objects, locations, paths, instruments and adverbs) constitutes a natural syntactic class based on theoretical considerations on Larson’s work on the VP shell and on empirical facts based on Li’s discussion of non-canonical objects in Mandarin Chinese. In the case of idiomatic readings, I reiterated Marantz’ observation that phrasal elements (kick the bucket) can have idiomatic readings, too, and discussed more facts specifically related to NI. Finally, I showed, based on van Geenhoven’s work, that frozen scope is a property of of INs and of bare nouns in many languages.

1.3 Conclusion and Outline of Book This chapter has given a brief outline of the goals of this monograph and described how these will be accomplished. To recapitulate, I offer an in depth description of NI in Northern Iroquoian, formulate a theory of phrase structure that maintains the core ideas of BPS and the LCA, relate this theory of phrase structure to NI in Northern Iroquoian, and finally show how this theory can account for linear or in NI constructions in a number of other languages. Section 1.2 also described the general properties of NI found cross-linguistically based on a number of prior works. It also discussed the linear order of NI constructions. What was found was the following: while VO and OV is quite evenly distributed cross-linguistically, N+V order prevails over V+N order. This order is most robust in non-productive or lexicalized constructions, but still holds in productive NI. The remainder of this monograph is structured as follows. Chapter 2 presents the theoretical background of this monograph. It begins with a brief discussion on the history of phrase structure within generative grammar and then presents the theories of BPS and Antisymmetry necessary for an understanding of the subsequent material. The chapter ends with a discussion on Head Movement and its absence from UG. Readers familiar with the basic tenets of Bare Phrase Structure and Antisymmetry may wish to skip some of this material. Chapter 3 starts off by outlining some of the theoretical problems inherent in bringing together BPS and Antisymmetry and discusses some earlier efforts in this direction. Section 3.2 discusses in more detail the problems raised at the beginning of the chapter, and explores the two logically possible ways of redefining the LCA in BPS terms, both of which have been previously discussed in the literature. This section also explores in detail the consequences that each of these approaches would have for the rest of the grammar. Section 3.3 discusses the core problem mentioned above, which is the initial merger of two heads, which we recall is an LCA violation. This section discusses the ramifications of adopting a dynamic view of Antisymmetry, which includes Complement-to-spec roll-up. Section 3.4 discusses some other possibilities that are compatible with the approach adopted here and shows how they might be implemented in particular circumstances, such as clitic

References

19

climbing. Finally, Section 3.5 discusses late vocabulary insertion and the Distributed Morphology framework, and how it bears on the current proposal. Chapter 4 is the main empirical illustration of the proposal put forth in the previous chapter, which is the analysis of NI in Northern Iroquoian. It begins with a brief defence of a syntactic approach to NI and illustrates the various patterns of NI found in Northern Iroquoian. The following section presents an analysis of NI in this language group and discusses some previous analyses of NI in Iroquoian. Section 4.4 discusses some other core properties of NI in Iroquoian, including NI in ditransitives and doubling. Finally, Section 4.5 discusses some properties of Iroquoian DPs. Chapter 5 discusses various other kinds of N+V incorporation including putative NI in gerunds in English, German and Persian. Although the process looks similar in all the constructions in these three languages, minor cross-linguistic differences manifest themselves in interesting ways. While all three cases have N+V order, Persian optionally allows V+N order. Also, both Persian and English allow only bare nouns in the NI constructions while German allows number marking, but no other kind of marking. The chapter also discusses incorporation of conjoined nominals in Tamil. Chapter 6 discusses various kinds of V+N incorporation, which on the face of it is problematic for the approach here. In particular, I discuss Pseudo NI in the sense of Massam (2001). Data from Niuean and Chamorro are discussed, along with the structure of nominals in these languages. The proposal made here will be shown not to be fully compatible with Massam’s approach. This chapter additionally discusses V+N compounds in Romance languages as well as NI in Mapudungun and lexical suffixation in Upriver Halkomelem. Chapter 7 is a conclusion. It includes a summary of the major empirical and theoretical findings of this volume. It also discusses some of the limitations of this proposal with the hopes of stimulating future research. Finally, it includes some brief comments on some extensions of the proposals laid out here.

References Abbott, Clifford. 2000. Oneida. Munich: Lincom Europa. Abney, Stephen. 1987. “The English Noun Phrase and its Sentential Aspect.” PhD diss., MIT press, Cambridge, MA. Aboh, Enoch. 2004a. The Morphosyntax of Complement-Head Sequences: Clause Structure and Word Order Patterns in Kwa. Oxford: Oxford University Press. Baker, Mark C. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago, IL: University of Chicago Press. Baker, Mark C. 1996. The Polysynthesis Parameter. Oxford: Oxford University Press. Baker, Mark C. 2003. Lexical Categories. Cambridge: Cambridge University Press. Baker, Mark C., Roberto Aranovich, and Lucía A. Golluscio. 2005. “Two Types of Syntactic Noun Incorporation: Noun Incorporation in Mapudungun and Its Typological Implications.” Language 81 (1):138–76. Barrie, Michael. 2008. “Bare Nouns in Cantonese.” Paper presented at the 12th Bilingual Workshop in Theoretical Linguistics, Ottawa, ON, December 5–6. Beck, Sigrid, and Kyle Johnson. 2004. “Double Objects Again.” Linguistic Inquiry 35 (1):97–123. doi:10.1162/002438904322793356.

20

1

Introduction

Bittner, Maria. 1994. Case, Scope, and Binding. Dordrecht: Kluwer Academic Publishers. Bonvillain, Nancy. 1972. “Noun Incorporation in Mohawk.” In Papers in Linguistics from the 1972 Conference on Iroquoian Research, edited by Michael K. Foster, 18–26. Mercury Series Ethnology Division. Ottawa, ON: National Museum of Man. Caballero, Gabriela, Michael J. Houser, Nicole Marcus, Teresa McFarland, Anne Phycha, Maziar Toosarvandani, Suzanne Wilhite, and Johanna Nichols. 2008. “Nonsyntactic Ordering Effects in Syntactic Noun Incorporation.” Linguistic Typology 12 (3):383–421. Cagri, Ilhan M. 2009. “Arguing Against Subject Incorporation in Turkish Relative Clauses.” Lingua 119 (2):359–73. Carlson, Gregory N. 1977. “Reference to Kinds in English.” PhD diss., University of Massachussetts, Amherst, MA. 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. Chomsky, Noam. 1994. Bare Phrase Structure. MIT Occasional Papers in Linguistics. Cambridge, MA: MIT Department of Linguistics and Philosophy, MITWPL. Cook, Eung-Do, and Andrea Wilhelm. 1998. “Noun Incorporation: New Evidence from Athapaskan.” Studies in Language 22 (1):49–81. Cuoq, Jean-André. 1866. Etudes philologiques sur quelques langues sauvages. Montreal, QC: Dawson Brothers. Di Sciullo, Anna Maria, and Edwin Williams. 1987. On the Definition of Word. Cambridge, MA: MIT Press. Dryer, Matthew S. 2008. Map 83: Order of Object and Verb. The World Atlas of Language Structures Online. http://wals.info/feature/83. Accessed August 13, 2010. Froman, Frances, Alfred J. Keye, Lottie Keye, and Carrie Dyck. 2002. English-Cayuga/CayugaEnglish Dictionary. Toronto, ON: University of Toronto Press. Gerdts, Donna B. 1998. “Incorporation.” In The Handbook of Morphology, edited by Andrew Spencer, and Arnold M. Zwicky, 84–100. Malden, MA: Blackwell. Ghomeshi, Jila. 2003. “Plural Marking, Indefiniteness, and the Noun Phrase.” Studia Linguistica 57 (2):47–74. Giorgi, Alessandra, and Giuseppe Longobardi. 1991. The Syntax of Noun Phrases. Cambridge MA: Cambridge University Press. Glougie, Jennifer. 2000. “Topics in the Syntax and Semantics of Blackfoot Quantifiers and Nominals.” MA Thesis, University of British Columbia, Vancouver, BC. Guimarães, Maximiliano. 2000. In Defense of Vacuous Projections in Bare Phrase Structure. University of Maryland Working Papers in Linguistics 9:90–115. Harley, Heidi. 2008. “Bare Roots, Conflation and the Canonical Use Constraint.” University of Lund, Lund, Sweden, February 5–6, 2008. Jiang-King, Ping. 1998. “Sonority Constraints on Tonal Patterns.” In The Proceedings of the Seventeenth West Coast Conference on Formal Linguistics, edited by Kimary Shahin, Susan Blake, and Eun-Sook Kim, 332–46. Stanford, CA: Center for the Study of Language and Information. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kayne, Richard. 2009a. Antisymmetry and the Lexicon. Linguistic Variation Yearbook 8 (1):1–31. Kiparsky, Paul. 1997. “Remarks on Denominal Verbs.” In Complex Predicates, edited by A. Alsina, Joan Bresnan, and Peter Sells, 247–88. Stanford, CA: Center for the Study of Language and Information. Kleinschmidt, Samuel. 1852. Grammatik der grönländischen Sprache mit theilweisem einschluss des Labradordialects. Berlin: Reimer. Kroeber, Alfred L. 1909. “Noun Incorporation in American Languages.” Paper presented at the XVI. Internationalen Amerikanisten-Kongress, 2d, Vienna and Leipzig. Larson, Richard K. 1988. “On the Double Object Construction.” Linguistic Inquiry 19 (3):335–91.

References

21

Larson, Richard K. 1990. “Double Objects Revisited: Reply to Jackendoff.” Linguistic Inquiry 21 (4):589–632. Larson, Richard K. 2004. “Sentence Final Adverbs and “Scope”.” In Proceedings of NELS 34, edited by Keir Moulton, and Matthew Wolf, 23’43. Amherst, MA: GLSA Publications. Larson, Richard K. 2010. “On Pylkkänen’s Semantics for Low Applicatives.” Linguistic Inquiry 41(4):701–04. doi:10.1162/LING_a_00020. Li, Yen-hui Audrey. 2010. “Case and Objects.” Paper presented at the GLOW in Asia VIII, Beijing Language and Culture University, August 12–14. Lounsbury, Floyd Glenn. 1949. “Iroquoian Morphology.” PhD diss., Yale University, New Haven, CT. Lounsbury, Floyd Glenn. 1953. Oneida Verb Morphology. New Haven, CT: Yale University. Marantz, Alec. 1997. “No Escape from Syntax: Don’t Try Morphological Analysis in the Privacy of Your Own Lexicon.” University of Pennsylvania Working Papers in Linguistics 4 (2):201–25. Marantz, Alec. 2001. “Words.” Paper presented at the 20th West Coast Conference on Formal Linguistics, University of Southern California, February 23–25, 2001. Massam, Diane. 2001. “Pseudo Noun Incorporation in Niuean.” Natural Language and Linguistic Theory 19 (1):153–97. Massam, Diane. 2009. Noun Incoporation: Essentials and Extensions. Language and Linguistics Compass 3 (4):1076–96. Megerdoomian, Karine. 2008. “Parallel Nominal and Verbal Projections.” In Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud, edited by Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizaretta, 73–103. Cambridge, MA: MIT Press. Michelson, Karin. 1991. “Possessor Stranding in Oneida.” Linguistic Inquiry 22 (4):756–61. Michelson, Karin, and Mercy Doxtator. 2002. Oneida-English/English-Oneida Dictionary. Toronto, ON: University of Toronto Press. Mithun, Marianne. 1984. “The Evolution of Noun Incorporation.” Language 60 (4):847–94. Moro, Andrea. 2000. Dynamic Antisymmetry. Cambridge, MA: MIT Press. Moro, Andrea. 2004. “Linear Compression as a Trigger for Movement.” In Triggers, edited by Anne Breitbarth, and Henk van Riemsdijk, 387–430. Berlin, New York, NY: Mouton de Gruyter. Oishi, Masayuki. 2003. “When Linearity Meets Bare Phrase Structure.” Current Issues in English Linguistics 2:18–41. Öztürk, BalkIz. 2009. “Incorporating Agents.” Lingua 119 (2):334–58. Polinsky, Maria. 1990. “Subject Incorporation: Evidence from Chukchee.” In Grammatical Relations: A Cross-Theoretical Perspective, edited by Katarzyna Dziwirek, Patrick Farrell, and Errapel Mejias-Bikandi, 349–64. Stanford, CA: The Center for the Study of Language and Information. Postal, Paul. 1979. On Some Syntactic Rules in Mohawk. New York, NY: Garland Publishing. Pylkkänen, Liina. 2008. Introducing Arguments. Cambridge, MA: MIT Press. Rackowski, Andrea, and Lisa Travis. 2000. “V-Initial Languages: X or XP Movement and Adverbial Placement.” In The Syntax of Verb Initial Languages, edited by Andrew Carnie, and Eithne Guilfoyle, 117–42. Oxford: Oxford University Press. Rice, Keren. 1991. “Intransitives in Slave (Northern Athapaskan): Arguments for Unaccusatives.” International Journal of American Linguistics 57 (1):51–69. Richards, Norvin. 2001a. “A Distinctness Condition on Linearization.” In 20th West Coast Conference on Formal Linguistics, edited by Karine Megerdoomian. Somerville, MA: Cascadilla Press. Ritter, Elizabeth. 1992. “Cross-Linguistic Evidence for Number Phrase.” Canadian Journal of Linguistics 37 (2):197–218. Ritter, Elizabeth. 1993. “Where’s Gender?” Linguistic Inquiry 24 (4):795–803. Rosen, Sara Thomas. 1989. “Two Types of Noun Incorporation: A Lexical Analysis.” Language 65 (2):294–317.

22

1

Introduction

Ruhlen, Merritt. 1975. A Guide to the Languages of the World. Stanford, CA: Language Universals Project, Stanford University. Sadock, Jerrold. 1980. “Noun Incorporation in Greenlandic: A Case of Syntactic Word Formation.” Language 56 (2):300–19. Sadock, Jerrold. 1986. “Some Notes on Noun Incorporation.” Language 62 (1):19–31. Sapir, Edward. 1911. “The Problem of Noun Incorporation in American Languages.” American Anthropologist 13:250–82. Sato, Yosuke. 2010. “Bare Verbal Nouns, Idiomatization and Icorporation in Japanese.” Paper presented at the Theoretical East Asian Linguistics, 6, Peking University: Beijing. Spencer, Andrew. 1995. “Incorporation in Chukchi.” Language: Journal of the Linguistic Society of America 71 (3):439–89. Szabolcsi, Anna. 1983. “The Possessor that Ran Away from Home.” The Linguistic Review 3 (1):89–102. Uriagereka, Juan. 1999. “Multiple Spell Out.” In Working Minimalism, edited by Samuel Epstein, and Norbert Hornstein, 251–82. Cambridge, MA: MIT Press. van Geenhoven, Veerle. 1998. “Semantic Incorporation and Indefinite Descriptions: Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic.” Dissertations in Linguistics. (DiLi). Stanford, CA: Center for the Study of Language and Information. Williams, Edwin. 2007. “Dumping Lexicalism.” In The Oxford Handbook of Linguistic Interfaces, edited by Gillian Ramchand, and Charles Reiss. Oxford: Oxford University Press. Woodbury, Hanni. 1975a. “Noun Incorporation in Onondaga.” PhD diss., Yale University, New Haven, CT. Woodbury, Hanni. 1975b. “Onondaga Noun Incorporation: Some Notes on the Interdependence of Syntax and Semantics.” International Journal of American Linguistics 41 (1):10–20. Woodbury, Hanni. 2003. Onondaga-English/English-Onondaga Dictionary. Toronto, ON: University of Toronto Press.

Chapter 2

Theoretical Background

This chapter establishes the theoretical background in which the forthcoming analysis is couched. In particular, I adopt the tenets of Bare Phrase Structure, Antisymmetry and Dynamic Antisymmetry. Readers familiar with these concepts may wish to skip to the next chapter, where the theoretical proposal is presented, or to skim the following sections. The analyses of NI are presented in Chapters 4, 5, and 6. However, before these analyses can be presented, I propose in Chapter 3 how BPS and the Linear Correspondence Axiom (LCA) can be made compatible with each other. In doing so, I will show that NI can be captured without appealing to head movement, which has fallen out of favour in recent years. The remainder of this chapter is organized as follows. Section 2.1 covers a brief history of phrase structure and how it relates to linearization. Section 2.2 outlines the theory of BPS, and Section 2.3 outlines the theory of Antisymmetry. Section 2.4 discusses the role of head movement and its elimination from UG. Section 2.5 covers various alternative theories of linearization. Finally, Section 2.6 concludes the chapter.

2.1 A Bit of History of Phrase Structure Chomsky (1957; 1965) proposes that the grammar must distinguish between principles that determine the structure of sentences in a given language and those that derive the different word orders found for sentences in that language. The former are known as phrase structure (PS) rules and the latter are known as transformations. Phrase structure rules of the Standard Theory and its descendents gave way to the Headedness Parameter (Chomsky 1981; Stowell 1981; Travis 1989 inter alia) in X-Bar Theory (Chomsky 1970; Jackendoff 1977). In its broadest form, the Headedness Parameter allows for the four settings in 1; however, many formulations allow only the first two, with the specifier obligatorily on the left (Oishi 2003; Richards 2008).

M. Barrie, Dynamic Antisymmetry and the Syntax of Noun Incorporation, Studies in Natural Language and Linguistic Theory 84, DOI 10.1007/978-94-007-1570-7_2,  C Springer Science+Business Media B.V. 2011

23

24

1.

2 Theoretical Background

Four Structural Types of XPs a. XP Spec

X’ X

b.

S-H-C order Compl

XP Spec

S-C-H order

X’ Compl

c.

X XP

X’ X

Spec

H-C-S order

Compl

d.

XP X’ Compl

Spec

C-H-S order

X

The prevailing view, introduced nearly simultaneously by Stowell (1981) and Chomsky (1981), was that the Headedness Parameter was a category-neutral specification of the linear order of the specifier, head and complementizer in a given language. The result of this approach was that clusters of word order properties (postpositions, post-nominal determiners, and OV order versus prepositions, prenominal determiners, and VO order; see Greenberg 1963) could be easily accounted for with one parametrically determined setting for headedness of all XPs. In contrast to this is the view that different syntactic categories can be assigned different parameter settings, a view that leads to severe over-generation. There are several problems with the Headedness Parameter, however, many of which have been previously discussed (Kayne 2003a, 2009; Kroch 2001). I discuss these only briefly here and leave the reader to consult the references indicated for a fuller discussion. Kayne discusses at length several typological gaps that the Headedness Parameter predicts should exist. Further, Kroch (2001) points out that all SOV languages (even Japanese according to Kayne) depart from the Greenbergian correspondences in some way. Nakajima (1999) also points out that one of the goals of BPS is to pursue a derivational rather than a representational approach to phrase structure. He suggests

2.1

A Bit of History of Phrase Structure

25

that the representational nature of the Headedness Parameter is inconsistent with the derivational approach of BPS. Specifically, Nakajima argues that the headedness parameter is a condition on the representation of a syntactic structure and that what is needed is a derivational approach to linearization.1,2 Additionally, Kayne (2009) points out that if the Headedness Parameter existed, it would be a very strange parameter. Ideally, parameters should be properties of functional elements in the Lexicon, and the Headedness Parameter is not tied to any such functional lexical item. Kayne draws the following analogy. Most models of generative syntax assume a bottom-up derivation; however, top-down models have also been proposed (Phillips 2003). No one, though, has ever proposed that languages vary parametrically between these two choices such that some languages are bottom-up and some are top-down. A Headedness Parameter, Kayne contends, would be no different than the bizarre parameter just described. Given the numerous problems with the Headedness Parameter, it comes as no surprise that one would resurrect the Universal Base Hypothesis.3 The form of the Universal Base Hypothesis adopted here, of course, is Antisymmetry (Kayne 1994), which is described in detail below. Before continuing, however, I address a common remark on Antisymmetry concerning the postulation of putatively unmotivated movements and functional projections to achieve the correct word order in this framework (see Richards 2008 for recent discussion). I note that this approach lacks no more explanatory adequacy than an approach that assumes a Headedness Parameter with both left and right adjunction. Headedness coupled with free left and right adjunction fails to prevent the same kinds of unattested structures that free generation of functional projections and movements do. One matter of ongoing research is to understand the kinds of movements and functional projections available in UG, in an attempt to constrain Antisymmetry. The next section outlines the theoretical approaches in which the proposal in Chapter 3 is framed: BPS and Antisymmetry. The end of this section has a short discussion of head movement and its proposed elimination from UG. Readers familiar with these concepts can skip to Section 2.5. 1 Nakajima

offers such an approach, which we argue against later. The current proposal does offer a strongly derivational approach to linearization. 2 Wojdak (2008), in fact, does propose a derivational approach to word order assuming an updated version of the Headedness Parameter. Again, however, such an approach suffers from many of the same problems as the traditional Headedness Parameter as described in this section, including over-generation. 3 The exact formulation of the Universal Base Hypothesis (Bach 1968) is still a matter of debate (Zwart 1997; Broekhuis 2006). On the basis of the prevalence of SVO and SOV patterns crosslinguistically, it comes as no surprise that the two contenders for this hypothesis are the S-H-C order (Kayne 1994, inter alia) and S-C-H order (Fukui and Takano 1998, inter alia). Still, others have proposed that the directionality parameter specifies only the order between the head and the complement, with the specifier universally on the left (Oishi 2003; Ernst 2003; Richards 2008). We address these issues at the end of this chapter.

26

2 Theoretical Background

2.2 Bare Phrase Structure Since early Minimalist efforts (Chomsky 1993, 1994, 1995b) it has been assumed that X-bar theory is an unnecessary enrichment of UG. Instead, syntactic structures are built by a simple pair-wise algorithm, Merge, under the rubric of BPS. Thus, rather than assuming a theory of phrase structure that constitutes a separate module of UG (X-bar Theory), BPS posits a single operations, Merge, which manipulates phrase structure.4 This operation takes as its input two elements, α and β, and gives as its output the unordered set {α, β}. Furthermore, it is assumes that this unordered pair must have a label of some sort in order to be accessed by subsequent instantiations of Merge. We call this label γ, and express the result as follows. 2.

{γ, {α, β}}

There are various possibilities as to the nature of γ, all of which are considered in turn in Chomsky (1995a). He concludes that the label is an identical copy of either α or β. Later, Chomsky (2008) offers a more precise algorithm for determining the label. I eschew this discussion here and assume simply that the label of the probe (i.e., the selector) projects. The operation Merge operates over lexical items and syntactic objects (the output of Merge). Also, in the absence of evidence to the contrary, only the top-most node in a tree is eligible to participate in Merge (i.e., Merge obeys the Extension Condition).5 Suppose we have the two phrase markers shown in 3a and b. They can participate in Merge to give the structure in 3c (assuming a projects) but there is no way to merge one of the phrase markers counter-cyclically with a subcomponent of the other. 3.

Merger of Phrase Markers a. K = {a, {a, b}}6 b. L = {c, {c, d}} c. Merge (K, L) = {a, {{a, {a, b}}, {c, {c, d}}}} (if a, the label of K, projects)

Before proceeding, a comment is in order about how phrase markers are represented here. For a phrase such as [DP the dog], I treat the three representations in 4 (where other functional projections such as NumP and nP have been omitted for the sake 4 In

its original formulation (Chomsky 1995a), Merge and Move were distinct operations (see also Nunes 2004). More recently (Chomsky 2004, appearing originally in 2001) has proposed that Move is not a distinct operation, but is simply a version of Merge referred to mnemonically as Internal Merge. 5 Citko (2005) argues for parallel merge, which is counter-cyclic, however. See also Richards (1998) on tucking in. Furthermore, one could argue that the null hypothesis is not to restrict Merge to operate over only the root node in absence of evidence to the contrary. See Citko and Richards cited above, as well as Chomsky (2008). 6 For consistency throughout, when lexical items are represented by Roman letters, I use standard font for labels and italic font for the actual lexical items.

2.2

Bare Phrase Structure

27

of discussion) as notational variants (except in the discussion on Antisymmetry below where X-Bar notations are the topic of discussion). I use phrase markers as in 4c throughout the rest of this work, since most readers are familiar with this type of notation. Strictly speaking, 4c is an X-Bar theoretic representation; however, the larger trees that are presented below would be rather unweildy in true BPS representations. 4.

Representation of Phrase Structure a. {the, {the, dog}} b.

the dog

the

c.

DP D

N

the

dog

Next I address the issue of adjunction in BPS, and whether or how it could carry over to an Antisymmetric approach. Prior to Beyond Explanatory Adequacy, (Chomsky 2004), Chomsky proposed two types of Merge: set Merge (substitution) and pair Merge (adjunction). Formally, the difference between the two is the structure of the label of the set formed by Merge. Set Merge works as illustrated in 3, while pair Merge differs in that the label is an ordered pair, , giving rise to the set {, {a, b}}. Chomsky (2004), alters the adjunction operation so that adjunction of a to b forms the ordered pair rather than an unordered set with a two-headed label, {, {a, b}}. Set Merge is unchanged, giving rise to {a, {a, b}}. Regardless of the set-theoretic composition of adjunction structures, Chomsky assumes that the adjoined XP (here a) does not participate in any new relations formed by the growing tree structure, and that b participates in the same relations that would have existed if adjunction had not taken place (agreement, binding, etc.). In other words, adjunction does not change the relation of the phrase it adjoins to with the rest of the sentence. Evidence for this view of adjunction comes from anti-reconstruction effects (Lebeaux 1988). 5.

[Which picture of herselfj that Johni likes] did hei ask Maryj to buy twh ?

In 5, the wh-phrase must reconstruct to the trace position indicated in order for the anaphor to be licensed. In reconstructing, however, we would expect a Condition C violation since John is c-commanded by a coreferential pronoun. The sentence, however, is grammatical, supporting the claim that the adjoined XP (underlined) does not participate in any new relations with the rest of the structure. The following sentence, now, is problematic for the picture drawn so far. 6.

∗ He i

knows which picture of herselfj that Johni likes Maryj bought.

28

2 Theoretical Background

In 6, regardless of whether the wh-phrase reconstructs, a condition C violation is created by the pronoun, he, in the matrix clause. To account for such data, Chomsky proposes an operation Simpl (=simplify) that converts an adjoined structure to a set Merged structure {a, b}, which participates in structural relations as any set merged phrase does. Another major proposal for adjunction holds that adjuncts are merged ‘late’ or post-cyclically (Speas 1990; Stepanov 2001; Lebeaux 1988, 1991).7 The idea behind these proposals is that the adjunct merges to its host phrase after all overt transformations have taken place. This is shown in 7. 7.

Postcyclic Merge of Adjuncts a. He did purchase [DP what photo] b. [DP what photo] did he purchase [DP what photo] c. [DP what photo [that John1 likes]] did he1 purchase [DP what photo] d.

[DP what photo [that John1 likes]] did he1 purchase [DP what photo]

base structure wh-movement add adjunct delete lower copy

Within an Antisymmetric framework (as explained below), there is no distinction between a specifier and an adjunct in terms of phrase structure. Any given XP may sport either one specifier or one adjunct. Although set theoretic notation may be devised to encode the difference between adjuncts and non-adjuncts, such differences seem theoretically superfluous unless the properties of adjuncts can be captured by these set theoretic concepts. It seems to me, however, that neither {, {a, b}} nor accomplishes this task, and no account appears forthcoming. In fact, recent work in this area points toward a move away from these original formalisms (Hornstein and Nunes 2008). Furthermore, stipulating order in this way is at odds with Antisymmetry and would require significant modification of the LCA. This, of course, is not a problem in and of itself, but in the absence of clear argumentation and vast empirical support, such far-reaching changes to the theory are to be avoided. For the purposes of this monograph, then, I assume that adjuncts compete with non-adjuncts for the unique specifier position within an XP, leaving aside the issue of the set theoretic implementation. This approach is compatible with the traditional view in which adjuncts adjoin to lexical or functional projections in the clause or to a Cinquean (1999) view in which many adverbial elements appear within their own functional projections. I now leave the issue of adjunction and end this section with a brief discussion on movement. 7 Not discussed here are two other proposals for adjunction structures. The first is Rubin (2003), who proposes a Mod(ifier)P shell to host adjunction. He proposes that it is a property of ModP that properties of the pre-existing structure (c-command, etc.) do not change. The other proposal is that of Safir (1999), who proposes that anti-reconstruction effects in adjunction structure are due to vehicle change, thus eliminating the need for any special structural apparatus for adjuncts.

2.2

Bare Phrase Structure

29

In order to account for the displacement property of language in a constrained and consistent way, constituents that fulfil the same semantic role (i.e., bear the same thematic role) are assumed to be merged into the same base position, regardless of where they appear on the surface. Thus, in the sentences, John stole the book, the book was stolen, and Which book do you think John stole the DP containing the word ‘book’ is merged in the same position in all three sentences.8 Furthermore, movement must be motivated in some way (cf. there seems to be someone at the door versus ∗ there seems someone to be at the door). In other words, constituents move, but in a constrained manner. Given the phrase structure in 8, suppose that AP must raise to SpecBP. 8.

BP B

CP C

AP

Movement is accomplished by remerger of AP with BP, a process often referred to as internal Merge. This contrasts with previous conceptions of Move, which assumed it was a composite of primitive operations: Copy + Merge (Chomsky 1995b; Nunes 2004).9 Upon merger of AP and BP, a copy of AP appears both in its remerged position and its original merged position. Under normal circumstances, the lower copy deletes at PF. 9.

BP AP

BP B

CP C

AP

This treatment permits the elimination of Move as a primitive operation. I reiterate that in 9 the category AP is either an adjunct or specifier, the choice between the two not being relevant to the forthcoming discussion. Thus, movement of AP (or internal Merge of AP) results in {B, {AP, {B, {B, CP}}}}. Finally, I address the trigger of movement. Moro (2000) pursues a research program in which symmetric c-command (discussed below) is the sole trigger for movement. In a nutshell, when two elements cannot be linearized by the LCA, 8 Notwithstanding to this are situations in which a DP is coreferential with a null element, thus giving the illusion that movement has taken place when none really has. Examples include PRO in traditional analyses of control and operators in some kinds of relative clauses. 9 Nunes’ (2004) conception of Move actually consists of a composite of four operations: Copy + Merge + Form Chain + Chain Reduction.

30

2 Theoretical Background

one of the two elements must raise to break the symmetry. Within the Minimalist Program movement has been argued to take place to remove strong uninterpretable features (Chomsky 1993, 1995b) or to satisfy a poorly understood EPP feature (Chomsky 2000, 2001). I assume that movement can take place for reasons other than the satisfaction of the LCA; however, I do not take a stance here on exactly how these triggers are to be understood.

2.3 Antisymmetry Kayne (1994) develops a theory of linearization in which c-command relations determine surface word order through the LCA. 10.

Linear Correspondence Axiom d(A) is a linear ordering of T.

(Kayne, 1994: 6)

A is the complete set of ordered pairs {<X1 , Y1 >, <X2 , Y2 >, <X3 , Y3 > . . . <Xn , Yn >} where Xi and Yi are any two non-terminal syntactic nodes such that Xi asymmetrically c-commands Yi . d(A) is the mapping from A to the set of ordered pairs of terminals such that ai is dominated by Xi and bi is dominated by Yi . T is the set of terminal nodes. In other words, when a syntactic node X asymmetrically c-commands a node Y, all the terminal nodes dominated by X precede all the terminal nodes dominated by Y. 11 gives the definition of c-command adopted in Kayne (1994). 11.

X c-commands Y iff X and Y are categories and X excludes Y and every category that dominates X dominates Y. (Kayne 1994: 18).

The following example illustrates how the LCA is implemented. In these examples, a, b, and c are terminals; A, B, and C are minimal projections dominating only terminals, and AP, BP, CP are maximal projections. Both minimal and maximal projections count as categories with respect to the definition in 11. Consider now example 12. 12.

BP AP

BP

A

B

CP

a

b

C c

2.3

Antisymmetry

31

The set A, upon which the LCA will be evaluated, consists of the set of ordered pairs in 13, which gives rise to d(A) in 14, the mapping from A to the set of pairs of ordered terminals as explained above. 13.

{<AP, B>, <AP, CP>, <AP, C>, }

14.

{, , }

In order for a derivation to be linearized, every terminal node must be ordered with respect to every other terminal node, either directly, or by transitivity. In other words, linear ordering must be total. Furthermore, linear ordering cannot be contradictory. That is, if a node A c-commands B, then no element inside B, including B itself, can c-command A or anything inside A. Such a configuration would result in contradictory linearizations and the derivation would crash at the PF interface. As a result of the LCA, tight restrictions are placed on phrase structure. Each maximal projection can contain only one specifier or one adjunct (the distinction between the two essentially being dissolved). Multiple adjunction or adjunction to a maximal projection that already has a specifier is prohibited. This is because the two adjuncts, or an adjunct and a specifier to the same maximal projection cannot be linearized with respect to each other.10 This is shown in the following example. 15.

BP AP

BP

A

DP

a

D

B

d

b

BP CP

Here BP has either two adjuncts/specifiers (the distinction between the two having evaporated). There is no asymmetric c-command relation between AP and DP in this structure under Kayne’s definition of c-command. The reason for this is that the three instances of BP in example 15 constitute a single category. Thus, B is dominated by BP, but AP and DP are not. They are thus not dominated by any category in this structure. In order for a category β to dominate an element α every segment of β must dominate α. Since the lowest segment of BP does not dominate 10 Guimarães (2008) shows that some constructions with multiple specifiers are actually compliant with the LCA. This is true only if one of the specifiers is a head and if there is no higher head that selects the XP with multiple specifiers as a complement. Since multiple specifiers are permitted in Dynamic Antisymmetry (as long as any point of symmetry that is formed is subsequently removed), I will not concern myself with the wrinkles Guimarães raises for strict versions of Antisymmetry.

32

2 Theoretical Background

either AP or DP, the category BP does not dominate either of these elements. Even if another head, E, merges with BP, the lack of asymmetric c-command between AP and DP remains, and their linear order remains unresolved. 16.

EP BP

E e

AP

BP BP

A

DP

a

D

B

d

b

CP

In 16, E c-commands both AP and DP.11 Also, AP and DP both c-command B; however, there is still no asymmetric c-command relation between AP and DP. As a result, no linear order is established between these two elements and the derivation crashes at PF. Consider now a contradictory ordering.12 In 17, BP asymmetrically c-commands C, and CP asymmetrically c-commands B. This gives rise to the set A in 18, and its image under the function d. We see that d(A) contains the contradictory ordering and , and the derivation crashes. 17.

DP AP

DP

A

BP

CP

a

B

C

b

c

18.

{<AP, B>,<AP, CP>, <AP, C>, , }

19.

{, , , }

11 Note

that E is in a symmetric c-command relation with both AP and DP since E, AP and DP are all dominated by the same set of maximal projections (only EP), but that E asymmetrically c-commands both A and D. Thus, e can be ordered with respect to a and d. The crucial point here is that there is no way that a and d can be ordered with respect to each other. 12 Note that the DP in 17 does not have a head. The inadmissibility of this kind of structure can be used to derive endocentricity. Thus, just as Kayne removes X’-Theory as a primitive from UG, deriving it instead from the LCA, endocentricity can also be removed as a primitive from UG. See Moro (2000), however, who uses the point of symmetry in the structure in 17 for small clauses. See Guimarães (2008) for related discussion, though.

2.3

Antisymmetry

33

More recently, Antisymmetry has been shown to capture a number of typological gaps that would otherwise remain mysterious (Kayne 2003a, 2009; Carstens 2002; Kayne 2010). Capturing these typological facts within an Antisymmetric framework has emerged as a significant research program. For instance, it is quite common for VO languages to have a constraint ruling out V-Adv-O; however, OV languages typically never have such a constraint. This is surprising under a symmetric view of syntax, where both left- and right-headed VPs are available but falls out naturally from an Antisymmetric view of syntax. I leave the reader to consult the references above for further consequences of Antisymmetry along these lines. Importantly for us, Dynamic Antisymmetry also predicts that NI will predominantly be of the N+V order, a prediction that largely holds to be true, according to Caballero et al. (2008). As mentioned elsewhere in this monograph, not all NI is of the order N+V. Chapter 5 is devoted to a discussion of V+N order. Moro (2000; 2004; 2009) develops a research program, Dynamic Antisymmetry, in which he proposes that the LCA is a PF constraint deriving from bare output conditions required for the articulo-perceptual interface.13 The bare output condition is simply that a linear order must be derived from the output PF receives from the syntax. Following Moro (2000), I call this the weak antisymmetric view. This contrasts with Kayne (1994), who assumes that the LCA holds throughout the derivation. Kayne’s original motivation for assuming that the LCA holds at every level of grammar and throughout the derivation (the strong antisymmetric view) has to do with the role of X-bar theory. Kayne’s original proposal sought to derive the properties of X-bar theory from Antisymmetry, thus eliminating X-bar theory as a primitive from UG. Since it was assumed that the principles of X-bar theory hold at every stage of the derivation, Kayne assumed that the LCA should also hold throughout the derivation. Since X-bar theory has been abandoned in the Minimalist Program, there is no compelling reason to assume that the LCA holds throughout the derivation. As Moro points out, since the LCA is concerned with linear order – a phonological aspect of language – there is no reason to assume that it holds anywhere but at PF. Under Moro’s Dynamic Antisymmetry approach, symmetric c-command serves as a trigger for movement. Movement, in this case, is the search for asymmetric c-command so that the LCA is satisfied. Moro discusses three specific contexts in which symmetric c-command holds, shown below. 20.

Symmetric C-Command a.

XP ZP

XP YP

13 Chomsky

XP

(1995b: 337) actually suggested this in a brief discussion of Romance clitics.

34

2 Theoretical Background

b.

XP X

c.

Y XP

ZP

YP

In the examples above, the elements in the ovals c-command each other. Moro proposes that this symmetry acts as a trigger for movement. In fact, Moro pursues a research program in which symmetry, not morphology, serves as the sole trigger for movement.14 In this monograph, I will be concerned only with the type of symmetry illustrated in 20b. To summarize, the LCA requires that all terminal heads enter into a noncontradictory linear ordering. The LCA is evaluated at the PF interface (following Moro 2000 but contra Kayne 1994). Movement can eliminate instances of symmetry, and Moro takes this to be its prime motivation. He does ask whether there are other motivations for movement, but leaves the question unanswered.15 If there is a choice as to what to move to eliminate symmetry, other properties of the grammar, in principle, may come into play. The preceding paragraphs have discussed the principles of Antisymmetry that are required for the proposal for phrase structure in Chapter 2. I now turn to head movement and its status in UG.

2.4 Head Movement Although Kayne (1994) states that head-adjunction is possible (but only to the left), it has been argued more recently that head-movement is not available to UG, or is restricted in some way (Harley 2004; Fanselow 2003; Chomsky 2000; Kayne 2003b; Koopman and Szabolcsi 2000; Mahajan 2003). I discuss here some of the properties of head movement and the problems involved in restricting or eliminating it entirely. First, however, I begin with a brief discussion of head-movement within an early minimalist framework.16 In his discussion of head movement, Chomsky (1995b) invokes uninterpretable [V-] features and [N-] features to account for verb-movement and NP-movement in language. In French, for instance, T hosts a strong [V-] feature that must be checked 14 See, for example, (Baauw 1998; Guasti and Moro 2001; Koncar 2005) for developments in this direction. 15 Moro acknowledges that it may be difficult or even impossible to treat all cases of movement as the resolution of symmetrical constructions. He admits that other possible triggers for movement such as checking Case in passives or EPP may have to be admitted in UG. 16 For a general discussion of head movement, see Roberts (2001).

2.4

Head Movement

35

by overt verb raising. In both English and French, T also hosts a strong [N-] feature (the EPP), which attracts the subject NP to SpecTP. The question that arises here is why the [V-] feature is satisfied by head-movement while the [N-] feature is satisfied by XP-movement. As it stood, this difference had to be stipulated.17 The elimination of head-movement from UG solves this problem by leaving XPmovement as the only type of movement. This results in a simplification of the grammar, since it reduces the number of options available to the learner to account for displacement. Consider the following scenario, with the following underlying and derived order of elements. 21.

Underlying versus derived order in a hypothetical language a. A. . .B. . .C (underlying) b. B. . .A. . .C (derived)

We observe that B has raised from its base position a position to the left of A. This movement could in principle be accomplished either by head-movement or by XP-movement as shown in the following phrase markers. In the following examples, X and Y are phonologically null. 22.

Phrase markers for hypothetical language a. AP [underlying order] A

XP X

BP B

YP Y

CP …

C

b.

AP A Bi

[derived order – head movement] XP

A

X

BP ti

YP Y

CP C

17 Massam

…

(2000b) proposes that VOS word order in Niuean arises when the VP moves to satisfy EPP in SpecTP. Thus we see here an example of a strong [V-] feature being satisfied by XP-movement. Other examples include Oda (2003; 2005) who proposes a VP-raising analysis to account for certain word order facts in Irish, and Aboh (2004) also discusses cases of VP movement and N movement.

36

2 Theoretical Background

c.

AP

[derived order – XP-movement] AP

BP B

A

YP

XP

tCP

Y

CP C

XP …

X

tBP

In 22c, CP moves to SpecXP, then the remnant BP moves to SpecAP. The result is that the same word order is derived in 22b and 22c. If the heads X and Y are phonetically null, there is no a priori way to distinguish between these two analyses. From the child’s perspective, then, having two possible transformations available to account for the same overt displacement severely complicates language acquisition in a way that a model of grammar that holds that head movement does not exist does not. Another problem with head-movement is the lack of a clear understanding of how it works in BPS. In X-Bar Theory, phrasal movement can be either adjunction or substitution, but it was never clear which of these two types head movement should be.18 Given that the distinction between specifiers and adjuncts (i.e., substitution and adjunction) has vanished, I do not dwell on this point. Furthermore, In X-Bar Theory, nodes are assigned bar levels derivationally as in 23a, but this option is not available in BPS, which distinguishes only between categories and heads. Thus, when a head, X, adjoins to another head, Y, the structure cannot project to Y as in 23a since Y in this structure has no theoretical status in BPS (see also Harley 2004). Rather, when these two heads merge, forming the projection Y? in 23b, the category cannot be a Y since it is not a terminal/lexical item. In fact, it is not clear how Y? would differ from YP, unless BPS is endowed with additional machinery. Recall that in BPS, Merge of X and Y gives rise to {Y, {X, Y}}, which is indistinguishable from [YP X Y]. 23.

a. X

Y?

b.

Y Y−1

X

Y

The fact that when X adjoins to Y, the resulting category cannot be a head, leads into the next problem with head-movement. As Mahajan (2003) points out, the moved head, X, does not c-command its trace under standard definitions of c-command. Although Kayne (1994: 17–18) argues that a c-command relation holds between the moved head and the elements inside the complement of the host head. The moved head a fortiori c-commands its trace, then. In Chapter 3, however, a revised definition of c-command will be presented that precludes a c-command relation between a 18 See

Harley (2004) for discussion of this point.

2.4

Head Movement

37

moved head and its trace. Baker (1988: 54–55) adopts Aoun and Sportiche’s (1983) definition of c-command, which is better known as m-command. Here’s is Baker’s formulation: 24.

A c-commands B iff A does not dominate B, and for every maximal projection C, if C dominates A then C dominates B.

This version of c-command (really m-command) was developed to accommodate the notion of government in the GB era (Chomsky 1986). While it is quite correct that under this definition a raised head c-commands its trace, it is also true that a head c-commands its specifier (as per the traditional notion of Government). Under an Antisymmetric framework this result is unwelcome as it would create a most unconventional ordering relating between the head of a phrase its specifier. Consider the following scenario. 25.

BP AP

BP

A

B

CP

a

b

C c

Here, B asymmetrically c-commands A, giving rise to the order , and ultimately giving rise to the order Head-Specifier-Complement, in violation of all standard notions of constituency. The elimination of head-movement from UG also solves the problem of countercyclicity. The Extension Condition requires that moved or external elements merge with the root projection only. Since head-movement requires the head to merge into a position lower than the root, cyclicity is violated.19 As the arrow indicates in 26, head-movement does not target the root of the phrase structure (XP), but rather something below it (X). 26.

XP X

YP Y

ZP

19 See Fanselow (2003) and Mahajan (2003) for an extended discussion of this and other problems

with head-movement.

38

2 Theoretical Background

27.

XP X Y

YP X

tY°

ZP

Holding to cyclicity, the possibility remains that the head Y in 26 could raise and merge with the root XP (as in Matushansky 2006; Toyoshima 2000). Under X-bar Theory, it was stipulated that such a movement would be ruled out, since only phrases, and not heads, could occupy the specifier position.20 It has also been noted (Harley 2004; Chomsky 1995b: 321; Toyoshima 2000: 44) that head-movement violates Uniformity of Chains (Chomsky 1994) since, prior to movement, the head is a minimal projection and after movement, the moved element is both a minimal and a maximal projection. Although Kayne (1994) also states that Head-to-Spec movement is unavailable, nothing within Antisymmetry rules it out.21 Let us return to the problem of how the derivation knows whether to move a head or an XP. Under the previous assumption that overt displacement is triggered by the need to satisfy an uninterpretable feature, [uF], this feature must probe a target with a matching feature [F]. Thus, the computation searches downward until it finds the first instance of a matching feature. If this feature is contained within a head that has not projected, then the head is copied and internally merged. If the feature is found on the label of a projection, then the computation targets the node containing that label, which is then copied and internally merged. Under this approach, a head will never be available for movement, unless it is also a maximal projection. This is illustrated in the following structures. Here, the feature [F] is being probed and targeted for movement. Recall, following BPS, that any features that appear on a head, X, also appear on XP, since XP has X as its label. 28.

Probing a feature [F] a. YP Y

XP = {X[F], {X, ZP}} X

ZP

[F] 20 Jackendoff

(1977) actually notes that specifiers, but not complements can be heads, but for different reasons than we are assuming here. Jackendoff’s statement was made under much older assumptions where what are now treated as functional heads were thought to occupy specifier positions (e.g., determiners). Cinque (1996) notes that a bare head cannot appear in specifier position within an Antisymmetric framework. I will return to this point in Section 2.5.2 below and in Chapter 3. 21 Dobrovie-Sorin (1994), for instance, argues that pronominal clitics in Romanian raise to SpecIP, rather than to I, a move made possible due to the fact that clitics, following Chomsky (1995b), are both maximal and minimal projections. Boškovi´c (2002) also argues that clitics are heads that occupy the specifier of a functional projection. Romance clitics are touched on very briefly in Chapter 2.

2.4

Head Movement

b.

39

YP Y

X [F]

In 28a, if the feature [F] is probed by a higher element, only the XP is visible since it is closer to the probe than its head X. Thus, X will never be targeted for movement. In 28b, of course, the head X is the only element bearing the feature [F], so it, of course, can undergo movement. Bobaljik and Brown (1997) propose a solution to the problem that head movement creates for cyclicity. They suggest that head movement proceeds by interarboreal movement (sideward movement in Nunes’ (2004) terms). Their approach assumes that the derivation proceeds in parallel. Thus, in 29, there are two phrase markers being built up simultaneously. 29.

YP + Y

X

ZP

Next, they propose that Y is copied and merges with X. 30.

YP

X +

Y

ZP

Y

X

Finally, they propose that X merges with YP and the lower copy of Y is deleted (presumably at PF). If we assume that the raised head Y must c-command the lower copy, we must assume that X does not dominate Y since not every segment of X dominates Y. 31.

XP X Y

YP X

Y

ZP

Note that the resulting structure in 31 is isomorphic to the one in 27. This approach solves the cyclicity problem since in all cases only the root of a phrase marker participates in Merge, but it still does not answer the question of how the system knows whether to raise a head or a phrase. In the first stage in this derivation shown in 29, the computation could just as easily have chosen to copy YP instead of Y and merge it with X, giving a structure without any head movement. The problem of two heads merging to form a complex head in 30 is problematic for the very reasons discussed above. Namely, once the two heads undergo Merge, the structure formed is a phrase, not a head since complex heads are unformulable in BPS.

40

2 Theoretical Background

Why then, do we get a complex head [X◦ Y X] in example 30, rather than an XP [XP X Y]? Bobalijk and Brown argue that when the two heads are merged, they undergo Pair Merge rather than Set Merge; in other words, the moved head is adjoined to the host head. Since adjunction does not project a new category, but rather just extends the category being adjoined to, then X simply projects another segment of X when Y adjoins to it. Regardless of Bobaljik and Brown’s arguments, the structures they posit are still unformulable within the set-theoretic constraints of BPS. I conclude that head movement as traditionally conceived is incompatible with BPS. Before concluding this section, I point out that head movement has also been analyzed as a purely PF phenomenon based on the putative lack of semantic effects (Boeckx and Stjepanovi´c 2001; Chomsky 2000). The evidence for the semantic import of NI is quite strong, however, ruling out a strictly PF analysis of this phenomenon. The actual properties involved vary from language to language, naturally, however, many common semantic side effects of NI include frozen scope (van Geenhoven 1998), definiteness and specificity (Mithun 1984; Sadock 1980 inter alia) many of which have been discussed above. The problem now becomes how to deal with cases previously thought of as head-movement. An XP-movement analysis must be provided for these cases.22 Generally, when it appears that a head has raised without its complement, a remnant XP-movement analysis is pursued. There is, of course, a vast literature exploring this approach, which cannot be summarized here (Takano 2000 inter alia; Lee 2000; Ambar and Pollock 2002; Baltin 2001; Cummings 2002; Kandybowicz and Baker 2003; Aldridge 2003; Massam 2000; Haegeman 2001, 2000; Müller 1996, 1997, 2004). The current study extends the remnant XP-movement approach to the domain of NI – a process which has been previously treated as involving head movement (Baker 1988).23

2.5 Alternative Accounts of Linearization Before concluding this chapter, I review alternative theories of linearization. As I showed above, the overwhelming majority of the world’s languages are either SVO or SOV. It comes as no surprise, then, that S-C-H order has also been proposed as the universal underlying order, from which SVO languages must be derived. Such approaches do not rely on asymmetric c-command as the basis of linear order, but rather rely on other properties. I will review three such proposals. The first proposal

22 Or,

as Elizabeth Cowper has pointed out to me, an analysis in which no movement takes place can be pursued; that is ‘moved’ heads are initially merged in the higher position rather than raised there from a lower position. 23 . . .at least as far as syntactic analyses are concerned. There are, of course, many analyses of NI which treat this phenomenon as lexical rather than syntactic. We address this issue more at the beginning of Chapter 3.

2.5

Alternative Accounts of Linearization

41

takes S-C-H to be the universal underlying order, while the other two propose that the order of specifiers, heads and complements is set parametrically.

2.5.1 Demerge and Concatenate: SOV as Underlying Order Fukui and Takano (1998) propose a theory of linearization based on BPS that postulates two operations in addition to Merge. Starting with the notion of Merge as a symmetric pair-wise operation that operates bottom-up, they propose a second operation, Demerge, that operates top-down. Demerge acts only on maximal projections. Finally, a third operation, Concatenate, linearizes elements as they are demerged. Thus, starting with the root node, XP in example 32, Demerge acts on the specifier, YP, and demerges it from XP. X’ is not available to Demerge, since it is not a maximal projection. Since only YP was demerged, Concatenate acts on it first and places it before the material of its sister, X’, which has yet to be concatenated. In other words, Concatenate places YP at the left edge of the phonological output (represented by the open < bracket). 32.

XP XP YP

X’

Aspect). More generally, if we assume a clausal hierarchy as in 43, we expect the possible orders in 44, but not those in 45. 43.

CP > MoodP > TP > AspP > vP > VP

46

2 Theoretical Background

44.

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

45.

a. b c.

C-Mood-T-Asp-v-V C-Mood-T-Asp-V-v C-Mood-T-V-v -Asp C-Mood-V-v -Asp -T C-V-v -Asp -T -Mood V-v -Asp -T -Mood-C ∗ C-Mood-T-v-V-Asp

∗ C-Mood-Asp-V-v-T ∗ Mood-Asp

-V-v-T-C

The forms in 44 are predicted by Baker’s Mirror Principle. Traditionally, this would be captured by head movement of the verbal head up to progressively higher and higher functional heads, but are also fully compatible with a Koopman and Szabolcsi (2000) approach, where head movement is abandoned in favour of pure phrasal movement. The forms in 45 are predicted as possible under Wojdak’s proposal but are rare or non-existent (see Julien 2002 for in depth discussion). Another smaller problem involves compounding with lexical roots. Wojdak discusses NI in Nuu-chah-nulth, a language with affixal verbal predicates. In Northern Iroquoian and English, it is unclear how linear ordering works. English is clearly SVO and Northern Iroquoian prefers SVO in discourse neutral utterances, although word order is significantly freer in Northern Iroquoian. This suggests that both English and Northern Iroquoian have Merge (V, XP)  V-XP as a directionality convention. It is unclear how the following contrast in English compounds and Northern Iroquoian NI can be achieved in Wojdak’s system. 46.

a. b.

47.

NI in Oneida a. waPkhni:nú: ká:sleht waP-k-hninu-´: FACT -1.SG . AG -buy-PUNC ‘I bought a car.’ b.

John enjoys collecting watches. John enjoys watch-collecting.

waPkePslehtahni:nú: waPkeFACT 1.SG . AG - JOIN ‘I bought a car.’

[Daisy Elijah, speaker] ka-Psleht-Ø 3.SG . NT-car- NFS

Pslehtcar-

aJOIN-

hninubuy-

´: PUNC

It is unlikely that one would want to argue that Psleht is affixal in Oneida and even more so for watch in English. Thus, one has recourse only to the directionality conventions to determine order. Thus, one would have to posit two sets of directionality conventions to account for the data above. Alternatively, one could argue that some movement takes place for these examples, but this is unsatisfying given that

2.5

Alternative Accounts of Linearization

47

Wojdak’s proposal was originally intended to capture ordering effects in NI-type constructions. Given the problems associated with the various proposals above, I do not adopt the proposals of Oishi (2003), Richards (2007) or Wojdak (2008). I go through one final proposal for linearization, which departs significantly from those discussed so far.

2.5.3 Departures from the LCA Nakajima (1999) develops a derivational approach to word order, in which linear order is determined as the derivation proceeds. He adopts Spec-Head-Compl as the base order, but uses a theoretical apparatus substantially different from the LCA. Instead, he appeals to the concepts of ‘label’ and ‘adjunct’ rather than to c-command to formulate his theory of linearization, shown in 48 (Nakajima 1999: 64). 48.

Derivational Linear Precedence Principle (DLPP) When Merge combines items α and β and creates a new term K, either α or β is the leaf of K, and the leaf precedes the other item in temporal order in K.

49.

Leaf When Merge combines items α and β and creates a new term K, α is the leaf of K iff: a. α is the label of K, OR b. α adjoins to K.

50.

Adjunction When Merge combines items α and β and creates a new term K, α is the adjunct of K iff: a. b.

α does not become the label of K, AND β is an already projected term.

Like Kayne’s LCA, these principles establish S-H-C as the base word order. The difference here is that two heads can exist in a configuration of mutual c-command and multiple left-adjunctions to the same functional projection are permitted. One of Nakajima’s motivations for this approach is directly related to the initial merger problem. He notes that under any implementation of the LCA under BPS, a head and its complement symmetrically c-command each other and cannot be linearized.28 He argues that the LCA forces raising of one of two symmetrically c-commanding 28 This

is not quite accurate. Under Uriagereka’s (1999) approach, no point of symmetry arises between a head and a full XP complement. This approach is discussed in full detail in Chapter 3.

48

2 Theoretical Background

heads, and gives English verb phrases such as see it and kick him as evidence that raising is, in fact, not required and therefore that the LCA is not tenable. He adopts Chomsky’s (1995b) assumption that pronouns are bare D heads. There is strong evidence, however, that English pronouns are not simple heads, but rather XPs with internal structure (Rose-Marie Déchaine and Martina Wiltschko 2002). Thus, the English examples that Nakajima offers do not provide clear evidence against the LCA. Clearer examples of pronominals that are simple heads are Romance clitics, which typically must raise from post-verbal position to the IP domain, further calling into question Nakajima’s proposal.

2.6 Conclusion This chapter has presented a brief historical sketch of phrase structure and linearization, and has laid out the properties of BPS and Antisymmetry. In the present work, I make the following assumptions. I adopt the theory of BPS virtually unaltered Chomsky (1994). I also assume the basic ideas of Antisymmetry as proposed by Kayne (1994) and of Dynamic Antisymmetry as proposed by Moro (2000). However, as I have indicated, the implementation of these ideas must be modified in order to bring it into line with BPS. This will be the main order of business in Chapter 2. I have also discussed other proposals for linearization that depart from the basic assumptions of Antisymmetry and have shown that these proposals fail to capture the different behaviour of heads and phrases when they appear in complement position. At the beginning of this chapter I showed that this difference consistently gives rise to word order differences. The theory of phrase structure proposed in the next chapter captures this fact about natural language. This chapter has also outlined various alternative approaches to linearization that do not rely on the LCA or on Kayne’s notion of Antisymmetry. What these proposals all have in common is that they fail to predict the word order effects that arise when a head, as opposed to a phrase, appears in complement position. Namely, they fail to account for the asymmetry in which N+V is more frequent than V+N, while VO and OV are of relatively equal frequency. Oishi’s proposal retains all the original problems of the Headedness Parameter, and Fukui and Takano’s proposal fails to capture the cross-linguistic gaps related to Antisymmetry (discussed above and in Kayne 2003a). Nakajima’s proposal avoids both of these problems, but at the cost of introducing significant additional machinery.

References Aboh, Enoch. 2004. “Snowballing Movement and Generalized Pied-Piping.” In Triggers, edited by Anne Breitbarth, and Henk Van Riemsdijk, 15–48. The Hague: Mouton de Gruyter. Aldridge, Edith. 2003. “Remnant Movement in Tagalog Relative Clause Formation.” Linguistic Inquiry 34 (4):631–40. Ambar, Manuela, and Jean-Yves Pollock. 2002. “Topic vs. Comment in Some Subject Inversion Sentences in French and Portuguese.” Journal of Portuguese Linguistics 1 (1):119–38.

References

49

Aoun, Joseph, and Dominique Sportiche. 1983. “On the Formal Theory of Government.” The Linguistic Review 2:211–36. Baauw, Sergio. 1998. “Subject-Verb Inversion in Spanish: Movement as Symmetry Breaker.” Linguistics in the Netherlands 15:1–12. Bach, Emmon. 1968. “Nouns and Noun Phrases.” In Universals in Linguistic Theory, edited by Robert T. Harms Emmon Bach, 91–124. New York, NY: Holt, Rinehart, and Winston. Baker, Mark C. 1985. “The Mirror Principle and Morphosyntactic Explanation.” Linguistic Inquiry 16 (3):373–416. Baker, Mark C. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago, IL: University of Chicago Press. Baltin, Mark. 2001. “Movement to the Higher V is Remnant Movement.” Linguistic Inquiry 33 (4):653–59. Bobaljik, Jonathan, and Jason Brown. 1997. “Interarborial Operations: Head Movement and the Extension Requirement.” Linguistic Inquiry 28 (2):345–56. Boeckx, Cedric, and Sandra Stjepanovi´c. 2001. “Head-ing Toward PF.” Linguistic Inquiry 32 (2):345–55. Boškovi´c, Željko. 2002. “Clitics as Nonbranching Elements and the Linear Correspondence Axiom.” Linguistic Inquiry 33 (2):329–40. Broekhuis, Hans. 2006. “The Universal Base Hypothesis: VO or OV?” In Linguistics in the Netherlands 2006, edited by Jeroen van de Weijer, and Battelou Los, 28–39. Amsterdam: John Benjamins. Caballero, Gabriela, Michael J. Houser, Nicole Marcus, Teresa McFarland, Anne Phycha, Maziar Toosarvandani, Suzanne Wilhite, and Johanna Nichols. 2008. “Nonsyntactic Ordering Effects in Syntactic Noun Incorporation.” Linguistic Typology 12 (3):383–421. Carstens, Vicki. 2002. “Antisymmetry and Word Order in Serial Constructions.” Language 78 (1):3–50. Chomsky, Noam. 1957. Syntactic Structures. Janua Linguarum. The Hague: Mouton. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam. 1970. “Remarks on Nominalization.” In Readings in English Transformational Grammar, edited by Roderick Jacobs, and Peter Rosenbaum, 184–221. Washington DC: Georgetown UP. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press. Chomsky, Noam. 1993. “A Minimalist Program for Linguistic Theory.” In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, edited by Ken Hale, and Samuel J. Keyser, 1–52. Cambridge, MA: MIT Press. Chomsky, Noam. 1994. Bare Phrase Structure. MIT Occasional Papers in Linguistics. Cambridge, MA: MIT Department of Linguistics and Philosophy, MITWPL. Chomsky, Noam. 1995a. “Bare Phrase Structure.” In Government and Binding Theory and the Minimalist Program: Principles and Parameters in Syntactic Theory, edited by Gert Webelhuth, 383–439. Generative Syntax. Oxford; 1. Oxford: Blackwell. Chomsky, Noam. 1995b. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. “Minimalist Inquiries: The Framework.” In Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, edited by Roger Martin, D. Michaels, and Juan Uriagereka, 89–156. Cambridge, MA: MIT Press. Chomsky, Noam. 2001. “Derivation by Phase.” In Ken Hale: A Life in Language, edited by Michael Kenstowicz, 1–52. Cambridge, MA: MIT Press. Chomsky, Noam. 2004. “Beyond Explanatory Adequacy.” In The Cartography of Syntactic Structures, vol. 3, Structures and Beyond, edited by Adriana Belletti. Oxford: Oxford University Press. Chomsky, Noam. 2008. “On Phases.” In Foundational Issues in Linguistic Theory, edited by P. Oltero, 133–66. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1996. “The ‘Antisymmetric’ Programme: Theoretical and Typological Implications.” Journal of Linguistics 32 (2):447–64.

50

2 Theoretical Background

Cinque, Guglielmo. 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. New York, NY: Oxford University Press. Citko, Barbara. 2005. “On the Nature of Merge: External Merge, Internal Merge, and Parallel Merge.” Linguistic Inquiry 36 (4):475–96. Cummings, Constance. 2002. “XP (Remnant) Movement and Verb Serialization in Yoruba.” PhD diss., New York University. Déchaine, Rose Marie, and Martina Wiltschko. 2002. “Decomposing Pronouns.” Linguistic Inquiry 33 (3):409–42. Dobrovie-Sorin, Carmen. 1994. The Syntax of Romanian. The Hague: Mouton de Gruyter. Epstein, Samuel, Erich Groat, Ruriko Kawashima, and Hisatsugo Kitahara. 1998. A Derivational Approach to Syntactic Relations. Oxford: Oxford University Press. Ernst, Thomas. 2003. “Adjuncts and Word Order Asymmetries.” In Asymmetry in Grammar: Volume I: Syntax and Semantics, edited by Anna Maria di Sciullo, 187–208. Amsterdam: John Benjamins. Fanselow, Gisbert. 2003. “Münchhausen-Style Head Movement and the Analysis of Verb Second.” UCLA Working Papers in Linguistics 13:40–76. Frank, Robert, and K. Vijay-Shanker. 2001. “Primitive C-Command.” Syntax 4 (3):164–204. Fukui, Naoki, and Yuji Takano. 1998. “Symmetry in Syntax: Merge and Demerge.” Journal of East Asian Linguistics 7:27–68. Greenberg, Joseph. 1963. Universals of Language. Cambridge, MA: MIT Press. Guasti, Maria-Teresa, and Andrea Moro. 2001. “Romance Causatives and Dynamic Antisymmetry: Essays Offered to Lorenzo Renzi.” In Current Studies in Italian Syntax, edited by Guglielmo Cinque, and Giampaolo Salvi, 173–88. Amsterdam: North Holland. Guimarães, Maximiliano. 2008. “A Note on the Strong Generative Capacity of Standard Antisymmetry Theory.” Snippets 18:5–7. Haegeman, Liliane. 2000. “Remnant Movement and OV Order.” In The Derivation of VO and OV, edited by Peter Svenonius, 69–96. Amsterdam: John Benjamins. Haegeman, Liliane. 2001. Antisymmetry and Verb-Final Order in West Flemish. The Journal of Comparative Germanic Linguistics 3 (3):207–32. Harley, Heidi. 2004. “Merge, Conflation and Head Movement: The First Sister Principle Revisited” In NELS 34, edited by Keir Moulton, and Matthew Wolf, 239–54. Amherst, MA: GLSA. Hornstein, Norbert, and Jairo Nunes. 2008. “Adjunction, Labelling, and Bare Phrase Structure.” Biolinguistics 2 (1):57–86. Jackendoff, Ray. 1977. X-bar Syntax: A Study of Phrase Structure. Linguistic Inquiry Monographs. Cambridge, MA: MIT Press. Julien, Marit. 2002. Syntactic Heads and Word Formation. Oxford: Oxford University Press. Kandybowicz, Jason, and Mark C. Baker. 2003. “On Directionality and the Structure of the Verb Phrase: Evidence from Nupe.” Syntax 6 (2):115–55. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kayne, Richard. 2003a. “Antisymmetry and Japanese.” English Linguistics 20:1–40. Kayne, Richard. 2003b. “Antisymmetry, Adpositions and Remnant Movement.” Paper presented at the workshop on Antisymmetry and Remnant Movement, New York University, October 31 – November 1, 2003. Kayne, Richard. 2009. “Why is Syntax Antisymmetric.” Paper presented at the Theoretical Approaches to Disharmonic Word Orders, Newcastle University, New Castle, DE. Kayne, Richard. 2010. “Why Are There No Directionality Parameters?” Paper presented at the WCCFL 28, University of Southern California, Los Angeles, CA. Koncar, Katarina. 2005. “Living on the edge: Wh-movement in Serbo-Croatian.” MA Thesis, University of Toronto, Toronto, ON. Koopman, Hilda, and Anna Szabolcsi. 2000. Verbal Complexes. Cambridge, MA: MIT Press. Kroch, Anthony. 2001. “Syntactic Change.” In The Handbook of Contemporary Syntactic Theory, edited by Mark Baltin, and Chris Collins, 699–729. Malden, MA: Blackwell. Lebeaux, David. 1988. “Language Acquisition and the Form of the Grammar.” PhD diss., University of Massachusetts, Amhers, MA.

References

51

Lebeaux, David. 1991. “Relative Clauses, Licensing and the Nature of the Derivation.” In Syntax and Semantics 25: Perspective on Phrase Structure: Heads and Licensing, edited by Susan Rothstein, 209–39. San Diego, CA: Academic Press. Lebeaux, David. 2009. Where Does Binding Theory Apply? Cambridge, MA: MIT Press. Lee, Felicia Ann. 2000. “Antisymmetry and the Syntax of San Lucas Quiavini Zapotec.” PhD diss., University of California, Los Angeles. Mahajan, Anoop. 2003. “Word Order and (Remnant) VP Movement.” In Word Order and Scrambling, edited by Simin Karimi, 217–37. Malden, MA: Blackwell Publishers. Massam, Diane. 2000. “VSO and VOS: Aspects of Niuean Word Order.” In The Syntax of Verb Initial Languages, edited by Andrew Carnie, and Eithne Guilfoyle, 97–116. Oxford: Oxford University Press. Matushansky, Ora. 2006. “Head Movement in Linguistic Theory.” Linguistic Inquiry 37 (1): 69–109. doi:10.1162/002438906775321184%U http://www.mitpressjournals.org/doi/abs/10. 1162/002438906775321184. Mithun, Marianne. 1984. “The Evolution of Noun Incorporation.” Language 60 (4):847–94. Moro, Andrea. 2000. Dynamic Antisymmetry. Cambridge, MA: MIT Press. Moro, Andrea. 2004. “Linear Compression as a Trigger for Movement.” In Triggers, edited by Anne Breitbarth, and Henk van Riemsdijk, 387–430. Berlin, New York, NY: Mouton de Gruyter. Moro, Andrea. 2009. “Rethinking Symmetry: A Note on Labelling and the EPP.” Snippets 19:17–8. Müller, Gereon. 1996. “A Constraint on Remnant Movement.” Natural Language & Linguistic Theory 14 (2):355–407. Müller, Gereon. 1997. “Extraposition as Remnant Movement.” In Rightward Movement, edited by Dorothee Deerman, David LeBlanc, and Henk van Riemsdijk, 215–46. Amsterdam/Philadelphia, PA: John Benjamins Publishing. Müller, Gereon. 2004. “Verb-Second as vP-First.” Journal of Comparative Germanic Linguistics 7 (3):179–234. Nakajima, Takashi. 1999. “Word Order in the Minimalist Program: A Derivational Approach.” PhD diss., Cornell University, Ithaca, NY. Nunes, Jairo. 2004. Linearization of Chains and Sideward Movement. Cambridge, MA: MIT Press. Oda, Kenji. 2003. “Irish as a VP Fronting Language.” In Proceedings of Canadian Linguistic Association 2002. Université de Montréal à Québec. Oda, Kenji. 2005. “V1 and Wh-questions: A typology.” In Verb First: On the Syntax of VerbInitial Languages, edited by Andrew Carnie, Heidi Harley, and Sheila Ann Dooley, 107–33. Amsterdam/Philadelphia, PA: John Benjamins. Oishi, Masayuki. 2003. “When Linearity Meets Bare Phrase Structure.” Current Issues in English Linguistics 2:18–41. Phillips, Colin. 2003. “Linear Order and Constituency.” Linguistic Inquiry 34 (1):37–90. Richards, Marc D. 2007. “On Feature Inheritance: An Argument from the Phase Impenetrability Condition.” Linguistic Inquiry 38 (3):563–72. Richards, Marc D. 2008. “Desymmetrization: Parametric Variation at the PF Interface.” Canadian Journal of Linguistics 53 (2/3):275–300. Richards, Norvin. 1998. “The Principle of Minimal Compliance.” Linguistic Inquiry 29 (4):599–629. Roberts, Ian. 2001. “Head Movement.” In The Handbook of Contemporary Syntactic Theory edited by Mark Baltin, and Chris Collins, 113–47 Oxford: Blackwell. Rubin, Edward J. 2003. “Determining Pair-Merge.” Linguistic Inquiry 34 (4):660–68. Sadock, Jerrold. 1980. “Noun Incorporation in Greenlandic: A Case of Syntactic Word Formation.” Language 56 (2):300–19. Safir, Ken. 1999. “Vehicle Change and Reconstruction in A Chains.” Linguistic Inquiry 30 (4):587–620. Speas, Margaret. 1990. Phrase Structure in Natural Language. Dordrecht: Kluwer Academic Publishers.

52

2 Theoretical Background

Stepanov, Arthur. 2001. “Cyclic Domains in Syntactic Theory.” PhD diss., University of Connecticut, Storrs, CT. Stowell, Tim. 1981. “Origins of Phrase Structure.” PhD diss., MIT, Cambridge, MA. Takano, Yuji. 2000. “Illicit Remnant Movement: An Argument for Feature-Driven Movement.” Linguistic Inquiry 31 (1):141–56. Toyoshima, Takashi. 2000. “Head-To-Spec Movement and Dynamic Economy.” PhD diss., Cornell University, Ithaca, NY. Travis, Lisa de Mena. 1989. “Parameters of Phrase Structure.” In Alternative Conceptions of Phrase Structure, edited by Mark Baltin, and Anthony Kroch, 263–79. Chicago, IL: Chicago University Press. Uriagereka, Juan. 1999. “Multiple Spell Out.” In Working Minimalism, edited by Samuel Epstein, and Norbert Hornstein, 251–82. Cambridge, MA: MIT Press. van Geenhoven, Veerle. 1998. Semantic Incorporation and Indefinite Descriptions: Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic. Dissertations in Linguistics. (DiLi). Stanford, CA: Center for the Study of Language and Information. Wojdak, Rachel. 2008. The Linearization of Affixes: Evidence from Nuu-Chah-Nulth. New York, NY/Berlin: Springer. Zwart, C. Jan-Wouter. 1997. “The Germanic SOV Languages and the Universal Base Hypothesis.” In The New Comparative Syntax, edited by Liliane Haegeman, 246–67. London, New York: Longman Linguistics Library.

Chapter 3

Unifying Antisymmetry and Bare Phrase Structure

In this chapter, I discuss the challenges involved in unifying Antisymmetry and Bare Phrase Structure (BPS) and develop a proposal that captures the insights of both theories. There are two problems in formulating a theory of phrase structure which retains the core properties of both Antisymmetry and BPS. The first has to do with the theoretical framework in which Antisymmetry was developed, namely X-Bar Theory. Kayne’s original formulation does not translate into BPS in a straightforward way. As will become clear, choices have to be made about how to reformulate the LCA under BPS. I will consider various proposals for this as I proceed. The second problem deals with a particular aspect of BPS that seems irreconcilable with Antisymmetry, namely the initial merger of two heads. When two heads are merged at the outset of a derivation, they c-command each other, in violation of the LCA. I refer to this as the Initial Merger Problem. This chapter is organized as follows. In the first section, I outline the problem of mutual c-command inherent to BPS. In the second section I review some previous attempts to bring Antisymmetry and BPS to line with each other. In the third section, I discuss how the LCA can be restated within BPS. In the fourth section, I develop the core proposal of this chapter, in which the principles of Dynamic Antisymmetry come into play. The fifth section discusses some alternative proposals for linearization that are independent of the LCA. The sixth section discusses how the proposal works under a Late Insertion framework. The last section is a summary and conclusion.

3.1 The Problem of Mutual C-Command Antisymmetry was originally formulated to derive the effects of X-bar theory. With the elimination of X-bar theory from UG in favour of BPS (Chomsky 1994), certain aspects of Antisymmetry must be re-examined. Consider first the very outset of a derivation. There is nothing in the work space, so Merge must select two lexical items (i.e., two heads) from the Lexicon to undergo Merge. The following example shows the merger of two heads, α and β.

M. Barrie, Dynamic Antisymmetry and the Syntax of Noun Incorporation, Studies in Natural Language and Linguistic Theory 84, DOI 10.1007/978-94-007-1570-7_3,  C Springer Science+Business Media B.V. 2011

53

54

3 Unifying Antisymmetry and Bare Phrase Structure

1.

Note that this representation was originally ruled out in Kayne (1994). Kayne’s original reason for prohibiting structures like 1 was his attempt to derive the principles of X-bar Theory, which did not permit heads in specifier1 or complement position. This type of structure, however, is permissible, and indeed unavoidable in BPS at the outset of a derivation. The problem, of course, is that the two heads c-command each other, in violation of the LCA. Thus, no linear order can be established between these two heads. This is what I will refer to as the Initial Merger Problem. This, of course, is not the first time this problem has been noted. As we will see below Chomsky (1995b) noted the problem and Guimarães (2000) discussed it in more detail in a short paper. Symmetric structures such as this were the foundation of Moro’s (2000) Dynamic Antisymmetry. What I argue here is that the Initial Merger Problem and BPS in general necessitates a re-thinking of Antisymmetry. This is the topic of the remainder of this chapter. First, however, I discuss some previous accounts.

3.2 Previous Accounts Chomsky (1995b) and Alexiadou (1997) both consider the Initial Merger Problem briefly and propose that it serves as a triggers movement. I discuss their contributions here, noting that said contributions are limited to the domains in which they were developed. The goal here is to come up with a general solution to the problem and, indeed, to come up with a way in which both the LCA and BPS can be maintained. Following the discussion on Chomsky’s and Alexiadou’s proposals, I discuss the use of Self-Merge (Guimarães 2000; Kayne 2009) to overcome the Initial Merger Problem and point out several problems with this approach. The section ends with a brief introduction to Nunes & Uriagereka’s and Richards’ proposals to bringing these two theories in line with each other. They will be evaluated against each other in Section 3.3, so my remarks on them here are brief. Chomsky (1995b: 337) noticed the Initial Merger Problem early on, in a brief discussion of Romance clitics. Chomsky had already suggested that the LCA is a PF constraint, and that violations of it can be tolerated in the syntactic component, as long as they are repaired before PF. The particular situation Chomsky was discussing involved Romance clitics, which are generally believed to raise to the IP

1 Other

researchers have actively relied on non-branching specifiers and complements. Boškovi´c (2002), for instance, proposes that syntactic clitics are obligatorily non-branching elements that appear as either the specifier or complement of a functional projection.

3.2

Previous Accounts

55

domain (Kayne 1989, 1991).2 A typical scenario is shown in the Spanish example below, where the verbal form quiero (‘I want’) selects a bare D clitic te (2. SG). In example 2, the verb and the clitic are in a mutual c-command relation, and thus cannot be linearized. Chomsky suggests that the clitic te raises to the IP domain and escapes symmetric c-command. Example 3 shows a typical analysis of leftward head-adjunction of the clitic to Infl, which was fairly standard at the time and remains so to this day (although see Roberts 2010 for another up-to-date view). 2.

VP V

D

quiero

te

3.

‘I love you.’

IP pro

I’

Imin Dj te

VP I

Vi

I

V

D

ti

tj

quiero

Under the framework being developed here, however, head movement is unavailable. The preceding discussion shows how the problem of the linearization of Romance clitics might be solved, but does not provide a general solution to the problem of the initial merger of two heads since the two heads undergoing initial merger will not always be a verb and a pronominal clitic.3 Alexiadou (1997: 31), addressing the Initial Merger Problem, proposes to remedy it by any kind of the leftward movement of the non-projecting element. She considers the case of a bare verbal root symmetrically c-commanding a bare adverbial head versus a branching adverbial phrase and proposes that the structure of the adverbial 2 See Uriagereka (1995a, 1995b) for arguments for a functional projection, FP, which hosts clitics in western Romance. There is also a large literature that suggests that pronominal clitics in Romance are base-generated in their surface position (Burzio 1986; Jaeggli 1986; Strozer 1976; Roberge 1990), thus obviating the problem. 3 Note that the facts are not as simple as they’re made out to be here. There are different kinds of pronouns in Romance, some of which behave as clitics as mentioned here, others of which can function independently. Both kinds have different distributions. Obviously, a full-scale study of the linearization of pronominal elements in Romance is beyond the scope of this monograph. See Moro (2000) for discussion.

56

3 Unifying Antisymmetry and Bare Phrase Structure

is responsible for the linearization properties of the verb and adverbial. Consider the contrast in the following Greek examples.4 4.

a.

b.

c.

d.

O Janis DET John ‘John behaves well.’ O Janis DET John ‘John behaves well.’ ferete poli behaves very ‘He behaves well.’ ∗ kalo-ferete well-behaves (‘He behaves well.’)

ferete behaves

kal-a well-ADV

kalo-ferete well-behaves kala well poli very

Alexiadou argues that bare adverbs are selected directly by the verbal root, giving rise to the following structure. 5.

VP V

A

ferete

kalo

She then argues that the adverb undergoes head movement and adjoins to V to satisfy the LCA. She likens adverb incorporation to NI and proposes that both take place to satisfy a similar set of licensing conditions, of which satisfying the LCA is only one part. She also appeals to the defective nature of the IN and the adverb that undergoes incorporation to drive this process. Ultimately, my proposal here is clearly in line with much of Alexiadou’s discussion; however, there are additional complications involved with both the LCA and BPS and how they can be brought in line with each other. This will form much of the discussion in the following section, but first I turn to some additional previous work on this topic. I now turn to three specific accounts of Antisymmetry which specifically acknowledge its incompatibility with BPS. The first is that of Guimarães (2000), which looks specifically at the Initial Merger Problem. Guimarães’ solution is to admit unary branching (i.e., non-branching structures) in order to avoid LCA violations. The second is that of Uriagereka (1999), and Nunes and Uriagereka (2000). They propose that only terminal elements (essentially, heads) participate in the 4 Alexiadou (p. 3) notes that the -a morpheme is an adverbial marker in Greek (akin to -ly in English). She does not explicitly mention the function of the -o marker in the incorporated forms.

3.2

Previous Accounts

57

LCA. As a result, all specifiers must undergo Spell-Out, thereby becoming islands for movement. I will show that this proposal suffers from problems as there are undeniable cases of movement out of some specifiers. Finally, Richards’ (2001) proposal, which most closely approximates the proposal put forth here, reformulates the LCA so that it retains the core tenets of BPS, but with fewer problems than that of Uriagereka’s and Nunes and Uriagereka’s proposal. He proposes that all nodes participate in the evaluation of the LCA. I now turn to the discussion these three proposals.

3.2.1 Guimarães (2000) and Self-Merge Guimarães (2000) acknowledges the paradox described in the second chapter concerning the apparent incompatibility of BPS and Antisymmetry. His solution is to allow for non-branching or vacuous projections in order to avoid symmetric c-command. He suggests that vacuous projections are fully compatible with BPS, thus allowing the computational component to retain the essence of that theory, while still satisfying Antisymmetry. His argument proceeds as follows. The operation, Merge, takes two objects, α and β, and places them in a single complex structure, K, with label γ such that K = {γ, {α, β}}. Nothing in this definition precludes α and β from being identical, however. In other words, an element, α, can merge with itself (Self-Merge).5 Guimarães, in fact, argues that in the absence of evidence against Self-Merge, to exclude it from UG would be an unwarranted stipulation. Once α undergoes Self-Merge we obtain the following set: K = {γ, {α, α}}. Note that the set {α, α} is identical to the set {α}, following the Extensionality Axiom of Set Theory (Partee et al. 1993).6 Consequently, when α merges with itself, it in fact forms the set K={γ, {α}}. The following example illustrates Self-Merge with a hypothetical verb selecting a bare noun. In this case, the noun has undergone Self-Merge. 6.

VP V

NP N

In order to prevent Self-Merge from taking place all over the derivation, Guimarães assumes, following Collins (1997) and Chomsky (2000), that Merge is costly.7 Thus, 5 Note that Self-Merge is just a label of convenience. Guimarães does not intend for the term to refer to an operation distinct from Merge. 6 This axiom states that if a given element appears more than once in the same set, then this set is equivalent to an otherwise identical set in which the element in question appears only once. 7 Recently, however, Chomsky (2008) assumes that not only is Merge free, but that Move is free, too. If so, some other restriction on Self-Merge would have to exist. Furthermore, Chomsky

58

3 Unifying Antisymmetry and Bare Phrase Structure

Self-Merge (and Merge in general) can take place only when required by the derivation. Since Self-Merge is required only to ensure that the LCA can linearize that part of the derivation created by the initial merger of two heads, it is predicted that it will take place only in this situation, that is, only at the very beginning of the derivation, when two heads would otherwise be merged (although see below). Related to Guimarães’ proposal is Kayne’s (2009), which proposes that SelfMerge is constrained by a different property of grammar. Specifically, only lexical categories, which constitute an open class, can undergo Self-Merge. Kayne adopts Collins’ (2005) proposal that only uninterpretable/unvalued features can be the locus of parametric variation. Furthermore, Kayne adopts Chomsky’s (2001) proposal that unvalued features must be valued immediately upon entering the derivaiton. If this is true, then items which undergo Self-Merge cannot bear uninterpretable features, which means that items that undergo Self-Merge cannot be the locus of parametric variation. Assuming that the set of parameters is finite, then the set of items that cannot undergo Self-Merge is a closed set and the set of items that can is an open set (i.e., lexical categories). Kayne further proposes that only nouns are lexical categories. Verbs are either true light verbs (assumedly verbs like do or make) or are derived from nouns (such as laugh). Thus, Kayne essentially argues for the structure in example 6 as well; however, he argues that Self-Merge happens whenever there is a noun in the derivation. The Initial Merger Problem is taken care of by assuming that a noun must be one of the two elements that kick-starts the derivation. I make note of the following observations on these two related proposals first. Note that Guimarães’ approach requires substantial look-ahead on the part of the grammar. He invokes the notion of invasive interfaces (Boeckx 1999), which ‘look down’ into the derivation and effect changes to satisfy constraints at the interface. Thus, as the LCA is a PF constraint, PF can effect changes at an earlier stage of the derivation so that when the derivation reaches PF, it does not violate LCA (or any other PF constraint). Invasive interfaces, then, is the antithesis of Procrastinate (Chomsky 1993). I show below, however, that Guimarães’ implementation of the principle of invasive interfaces is problematic in several regards. Turning to Kayne’s proposal, I observe that it drastically alters our understanding of lexical categories and the noun-verb distinction. He notes that under his proposal the categorizing head n can be dispensed with. Although this proposal prevents the Initial Merger Problem from taking hold with respect to a verb taking a nominal complement, it fails to prevent the same problem when the verb takes and adverbial complement (in the sense of Larson 2004). These are attested in Greek, as shown above (Alexiadou 1997; Rivero 1992), Chukchi (Spencer 1995) and in Blackfoot, based on the following example (Frantz 1991).8

suggests that Merge does not take place to satisfy unvalued features, since these can be valued long-distance by Agree. Rather, Merge takes place to satisfy a poorly understood Edge Feature (EF). Assumedly, EF cannot be satisfied by Self-Merge, however. See Kayne (2009a) for further discussion. 8 I discuss the Blackfoot facts in more detail in Chapter 5.

3.2

7.

Previous Accounts

59

nitsikkamiyowatsi nitIkkamiy1fastØ‘I ate the apple quickly.’

owateat.TR-

tsi MOOD

One would have to amend Kayne’s proposal to include adverbs with nouns to account for what is essentially the same problem and same set of empirical facts. Self-Merge, itself, also presents an array of problems. First, Guimarães assumes that the LCA is a PF phenomenon, violations of which are repaired in the syntax as per invasive interfaces. There is a look-ahead problem with this approach, however. If the LCA is not computed until PF, violations of it should be tolerable in the narrow syntax, as long as they are taken care of promptly (i.e., before PF). This is not Guimarães’ solution, however. The PF violation (namely the LCA violation) never exists in the syntax. Steps are taken (namely Self-Merge) to ensure that the violation never happen to begin with. It is unclear, then, exactly how the LCA is a PF constraint under Guimarães’ approach. Guimarães is concerned only with the Initial Merger Problem; however, there is another situation in which two heads appear in a situation of symmetric c-command other than at the start of the derivation. Consider the following hypothetical structure, in which the specifier of XP is a head, Z. There is no violation of the LCA here as Z asymmetrically c-commands X, and X asymmetrically c-commands the material inside YP.9 Thus, no Self-Merge can take place at this point following either Guimarães’ or Kayne’s proposals (assuming Z is not nominal). 8.

XP Z

XP X

YP

Now, if another head, W, merges with XP, a problem arises as W and X are in a symmetric c-command relation. 9.

WP W

XP Z

XP X

YP

9 That Z asymmetrically c-commands X is not entirely obvious. First, it is clear that X does not c-command Z since it is not the case that every category that dominates X also dominates Z. Specifically, XP dominates X but does not dominate Z since not every segment of XP dominates Z. Now, in order to see how Z c-commands X we must realize that the condition that every category that dominates Z also dominate X is vacuously satisfied since there are no categories that dominate Z. The discussion below elaborates this point.

60

3 Unifying Antisymmetry and Bare Phrase Structure

At this point, one could simply stipulate that the specifier and complement positions are available only to XPs (be it formed in the traditional manner or by Self-Merge). This, however, is simply a notational variant of X-Bar Theory and lacks any independent motivation. If we push Gruimarães’ proposal, though, one might presume that W must undergo Self-Merge before undergoing Merge with XP since the grammar would detect an impending LCA violation after simple merger of a bare W. Consider the structure resulting from such Self-Merge in example 10. 10.

?P WP W

XP Z

XP X

YP

It is unclear what element would project in this phrase marker, but let us consider both possibilities in turn. If W projects (?P = WP), then XP and Z both asymmetrically c-command W, giving rise to the incorrect linear order. If X projects (?P = XP), then WP asymmetrically c-commands Z, giving rise to <W, Z> and Z asymmetrically c-commands W (vacuously as described above), giving rise to a contradiction. Thus, Self-Merge of W does not salvage the situation. Since Z is the cause of the problem, a solution along the lines of what Guimarães proposes would involve Self-Merge of Z. However, this would require even more look-ahead than described above. Z would have to know that the XP it is about to become a specifier of will itself eventually merge with another head, W here, that has phonological content. This requires more foresight than one would wish to endow UG with. In other words, the decision to undergo Self-Merge should be made locally (Collins 1997). 11.

WP W

XP ZP Z

XP X

YP

Having dispensed with Guimarães’ mechanism of Self-Merge as a solution to the Initial Merger Problem (and a fortiori Kayne’s proposal on Antisymmetry and the noun-verb distinction), I now turn to Nunes and Uriagereka’s treatment of the LCA within a BPS framework.

3.2.2 Nunes and Uriagereka (2000) Nunes and Uriagereka (2000) and Uriagereka (1999) analyze CED effects (Huang 1982) within an Antisymmetric and Minimalist framework and develop a theory of

3.3

The LCA and BPS

61

Multiple Spell-Out to account for them. They, too, note the incompatibility of the LCA and BPS and propose the following reformulation of the LCA. 12.

LCA = A lexical item α precedes a lexical item β iff α asymmetrically c-commands β. (Nunes and Uriagereka 2000: 23)

The crucial difference between the version of the LCA in example 12 and the original version in Kayne (1994) is that the LCA in 12 applies to heads rather than to categories. I will compare the empirical effects of this and Richards’ proposal below in Section 3.3 as both proposals will be evaluated against the same set of data.

3.2.3 Richards (2001) Richards (2001) assumes that the LCA is evaluated over all nodes in the tree (or syntactic objects in the BPS sense), in contrast to Kayne’s original proposal, where the LCA was evaluated among all non-terminal nodes (and Nunes and Uriagereka’s proposal, where the LCA is evaluated over all terminal nodes, or heads). Richards’ assumptions about linearization are as follows (Richards 2001: 2). 13.

Spell-Out considers the set A of pairs of asymmetrically c-commanding XPs and Xs in the tree which the syntax gives it, and generates from this a set of instructions for linearization; if is in A, then the image of α (that is, the terminals dominated by α) precedes the image of β.

Again, this has significant empirical consequences that I outline in Section 3.3.

3.2.4 Conclusion In the previous two sections, I considered three possible ways to solve the problem of mutually c-commanding heads. Guimarães (2000) and Kayne (2009) proposed to permit vacuous projections in phrase structure, essentially modifying BPS but leaving the basic tenets of Antisymmetry unchanged. I have shown, however, that this move is unsatisfactory in several regards. I also very briefly introduced Nunes and Uriagereka’s and Richards’ proposals. The difference between these two proposals is that Nunes and Uriagereka version of the LCA is evaluated only among heads, whereas Richards’ version of the LCA is evaluated among all syntactic objects. The details will be spelled out in the forthcoming sections. I will show that Richards’ version of the LCA, in which all syntactic objects are considered in the evaluation of the LCA, captures the empirical facts in a more straightforward way.

3.3 The LCA and BPS There have been several attempts to recast the LCA within BPS, given that the LCA was originally formulated in an X-bar theoretic framework (Epstein et al. 1998; Guimarães 2000; Nunes and Uriagereka 2000; Oishi 2003; Uriagereka 1999;

62

3 Unifying Antisymmetry and Bare Phrase Structure

Richards 2001; Richards 2008). I discuss here Kayne’s original formulation of the LCA within the context of BPS. First, it is clear that the structure in example 14 is not a valid phrase marker in BPS, as there is no inherent distinction between a minimal projection and a terminal in this framework. For instance, there is no distinction between C and c, for example. This structure also contains unary branching, which is ruled out in BPS (see the discussion in Section 3.2.1 above). 14.

DP D d

BP AP

BP

A

B

CP

a

b

C c

Instead, let us consider the phrase marker in 15, where I assume, for now, that AP and CP have internal structure. 15.

DP d

BP BP

AP …a…

b

CP …c…

If A is defined in terms of non-terminals (that is, where A consists of a set of ordered pairs , α and β non-terminals, and α asymmetrically c-commands β) we run into serious problems very quickly. In 15, AP asymmetrically c-commands CP, but BP does not enter into a c-command relation with anything.10 Thus, the only ordering of non-terminals is {<AP, CP>}, which gives rise to the non-total ordering {}. Furthermore, if another head is merged with DP, as in 16, no new orderings are computable. 16.

EP e

DP d

BP AP …a…

BP b

CP … c…

10 Recall that for Kayne (1994), c-command is a relation that holds only among non-terminals, not

among the lexical heads themselves.

3.3

The LCA and BPS

63

Since a formulation of the LCA in terms of non-terminals does not result in a usable system, that is, it does not give a total ordering of heads, let us consider various alternatives. The LCA can be reformulated in terms of terminals, or heads, which is the approach taken by Uriagereka (1999) and Nunes and Uriagereka (2000). There is another logical possibility, namely, that in which the LCA is reformulated in terms of all nodes (or syntactic objects) in the tree (Richards 2001). I will consider both of these approaches, and, ultimately, will adopt the all-nodes approach of Richards (2001). Let us consider first a reformulation of the LCA in which A is determined by the set of syntactic objects in the tree, and d(A) is reformulated accordingly. I use the term syntactic object as defined by Chomsky (1995a). A syntactic object is any lexical item (LI) drawn from the lexicon or any set, {α, {α, β}} such that α and β are syntactic objects (i.e., the merger of two syntactic objects). 17.

A = the set of ordered pairs {<X, Y>} such that X and Y are syntactic objects and X asymmetrically c-commands Y.

18.

d(A) is the mapping from A to the set of ordered heads {<x, y>}, such that x is dominated by X and y is dominated by Y.

Note, however, that Kayne’s (1994) definition of c-command, repeated in 19, cannot be used here, since it was defined for categories (XPs) only. 19.

X c-commands Y iff X and Y are categories, and X excludes Y and every category that dominates X dominates Y. (Kayne, 1994: 18).

I therefore adopt the revised definition for c-command in 20, which does not limit the relation to XPs, but rather is defined for all syntactic objects in the tree. 20.

x c-commands y iff x and y are syntactic objects, and x excludes y and every category that dominates x dominates y.

If we reconsider the tree in 16, we observe the following results for A and d(A), where d(A) reduces to the linear order <e, d, a, b, c>. This gives us the desired results, namely a total, irreflexive, non-contradictory ordering of the terminals of the phrase marker. 21.

A = {(e, d), (e, BP), (e, AP), (e, a), (e, b), (e, CP), (e, c), (d, a), (d, b), (d, CP), (d, c), (AP, b), (AP, CP), (AP, c), (b, c)}

22.

d(A) = {(e, d), (e, a), (e, b), (e, c), (d, a), (d, b), (d, c), (a, b), (a, c), (b, c)}

We now turn to the other logical possibility in which A is defined solely in terms of heads. First, note that stating the LCA in terms of heads alone has a certain appeal since it reduces the number of ordered pairs that must be considered. Also, PF must ultimately produce a linear order of heads, not phrases; it therefore seems reasonable to state the LCA in terms of heads. Consider the following definition for the set A.

64

23.

3 Unifying Antisymmetry and Bare Phrase Structure

A is the set {<X, Y>} of all ordered pairs of heads such that X asymmetrically c-commands Y.

Let us first consider a simpler version of example 15, without any specifiers. This is shown in 24. 24.

DP d

AP a

BP b

CP …c…

Under this definition of the LCA, the set A consists of the following set of ordered pairs, {, , , , , } which give the linear order . Note that under this approach there is no need to make reference to the image of A, d(A). The LCA can therefore be re-stated as in 25, which is essentially Nunes and Uriagereka’s (2000) definition for the LCA. 25.

A is a linear ordering of T, where T is the set of all heads.

Let us now turn to complex specifiers, as in 26. I consider atomic specifiers (specifiers consisting only of a head) later. 26.

DP d

BP AP a

BP FP

…f…

b

CP …c…

The structure of AP itself is identical to that of BP in 24 above. The linearization of AP thus occurs in the same manner, giving the order . There is, however, no c-command relation between either of the heads a and f inside AP and any of the other heads in the phrase structure below the specifier position (i.e., b or lower). This is because nothing inside the specifier c-commands out of the specifier, since only heads can enter into a c-command relation. Thus, linear order cannot be established between the heads a and f of AP, and the rest of the structure. Although d asymmetrically c-commands the material inside AP, this is not sufficient to allow for a total ordering. Based on the mechanism developed so far, the set A for the structure in 26 would appear as in 27, which would then reduce to the six possible orderings shown

3.3

The LCA and BPS

65

in 28. Since A does not give rise to a total, irreflexive, non-contradictory ordering, the phrase structure in 26 cannot be linearized, and the derivation crashes at the PF interface. 27.

A = {, , , , , }

28.



Uriagereka (1999) recognizes this problem and proposes that complex XPs must undergo Spell-Out before merging into the specifier position of another phrase. Thus, the derivation must permit Multiple Spell-Out, with Spell-Out taking place each time a complex XP is merged into specifier position. Thus, the AF node in example 26 undergoes Spell-Out before it is merged in the specifier position of BP. typeface in the forthcoming examples. When d This is represented by the merges with BP the result is the structure in 29, where AP has undergone Spell-Out and is internally linearized. 29.

DP d

BP BP b

CP …c…

At this point, Uriagereka does not decide how AP is linearized with the rest of the structure. He offers two choices. Either AP is somehow linearized before b and after d, as the phrase marker suggests, or the linearization of the Spelled-Out element is not handled by syntax, but rather by some performance system. Uriagereka pursues the latter option, but does not give a definitive account.11 Multiple Spell-Out as conceived above has empirical consequences as discussed by Uriagereka (1999), Nunes and Hornstein (2000) and Nunes (2004). They argue that it captures CED effects (Huang 1982) as shown in the pair of sentences in 30. The explanation goes as follows. The DP [a picture of who] is the complement of the verb in 30a. When wh-movement takes place, this DP has not yet been Spelled Out since it is not in a specifier position. Thus the wh-phrase who is free to raise. In 30b, on the other hand, the DP [a picture of who] is a derived subject in SpecTP. Before the DP moves into specifier position it undergoes Spell-Out. This DP, {a, }, is then effectively frozen for further operations (but see below).

11 Note that we cannot simply treat AP as an atomic element (i.e., treat it as though it were a head)

because it’s in a symmetric c-command relation with d. The resolution of symmetric c-command is discussed below.

66

30.

3 Unifying Antisymmetry and Bare Phrase Structure

CED Effects in English a. b.

Whoi ∗ Who i

did Ashleigh put a picture of ti on the wall? was [a picture of ti ]j put tj on the wall?

Multiple Spell-Out, as developed by Uriagereka (1999), Nunes and Uriagereka (2000), and Nunes (2004), seems to account nicely for subject island effects. Following Cinque (1999), adjuncts are taken to be specifiers of separate functional projections in the main structure of the clause. Thus, this approach classifies adjuncts and specifiers together and predicts that they block extraction (see the papers cited above for more discussion on the full range of CED effects). Thus, if we were to accept Uriagereka and Nunes’s proposal, we would have a viable reformulation of the LCA within BPS and capture CED effects for free. We will see below, however, that this approach undergenerates. The formulation of A in terms of heads alone, aside from being theoretically more motivated, appears to account elegantly for CED effects. However, the account proves to be overly simplistic as follows: Under this approach, extraction from a specifier should never be permitted. This clearly cannot be the case however if we consider the following data.12 31.

Extraction from a Specifier a. b.

Which room did John talk to his brother in t? Who did John talk to t in the kitchen?

Both sentences in 31 are grammatical. Let us investigate standard proposals on these structures in light of Nunes and Uriagereka’s approach. Under an analysis that assumes that the locative PP is in the specifier of a functional projection, along the lines of Cinque (1999; 2010), we would predict that the first sentence is ungrammatical, while the second sentence is grammatical, contrary to fact since both are grammatical. Consider the structure in 32, in which the locative PP appears in the specifier of a LocP.13 I assume for simplicity that the PP complement of the verb raises to some licensing position (Koizumi 1995; Lasnik 1995). Koizumi and Lasnik, of course, dealt mainly with DP complements, however. Since I ultimately reject the Multiple Spell-Out approach taken here I do not delve into the details of the placement of DP (and PP) complements any further.

12 Another approach to these adjuncts which is commonly (and tacitly) assumed is that they are not

subject to the LCA and are right-adjoined in the traditional manner. Under Nunes and Uriagereka’s approach, such adjuncts are still not linearizable with the rest of the structure, so the following discussion still holds. 13 Note that I have followed a traditional analysis that uses head movement since a remnant movement or PF head movement account of these facts is beyond the scope of this monograph.

3.3

The LCA and BPS

32.

67

CP DPi

CP

which C room didj

IP DPk

IP

John

I

vP

tj tk

vP

v

XP

talkl

PPm

XP

to his X brother

LocP PP

LocP

in ti Loc

VP tm

V tl

In this structure, both the locative PP and the direct object appear in specifier positions, and thus are required to undergo Multiple Spell-Out, and will be unavailable for subextraction. Crucially for us, however, the locative PP should not allow subextraction and the Multiple Spell-Out analysis, coupled with a Cinquean view of locative adjuncts, incorrectly predicts that this sentence should be ungrammatical. Alternatively, we could adopt a Larsonian VP shell for structures such as these, in which the ‘adjunct’ is the complement of the lower V head. Now we predict that the first sentence is grammatical and that the second sentence is ungrammatical, again contrary to fact. Consider this time the structure in 33. 33.

CP DPi

CP

which room C

IP

did DP John

IP I

vP v talk

VP PP

to his brother V

VP PP in

tj

68

3 Unifying Antisymmetry and Bare Phrase Structure

Here, the locative PP can be freely extracted from since it is not a specifier; however, the complement PP, to his brother, which is in a specifier position, is not available for subextraction, contrary to fact. Curiously, if the locative PP is absent, then subextraction from the complement PP is predicted to be grammatical, since it would then appear in a complement position, rather than in a specifier. Thus, even if I explain this conundrum, the Multiple Spell-Out approach outlined above predicts the existence of a language that differs in the extraction possibilities of the object depending on the presence or absence of VP level adjuncts in the clause as discussed for 33 above. So far as I know, no such language exists. Thus, under any view of the structure of clauses with locative PP adjuncts, Nunes and Uriagereka’s approach is untenable.14, 15 To be sure, consider the following example that drives the nail in the coffin for this approach. Here, both PPs have been extracted from, which should be impossible under the Multiple Spell-Out approach. Which roomi was Johnj talked to tj in ti ?

34.

Stepanov (2001a; 2001b; 2007) argues that CED effects are not a unitary phenomenon. Instead, he argues that the impossibility of extracting from subjects and from adjuncts requires two different explanations. Part of the empirical support for this approach lies in the vast array of cross-linguistic variation with respect to extraction from subjects, while extraction from adjuncts is by and large uniformly disallowed in natural language (although see Truswell 2007 for a discussion on the possibility of extraction from adjuncts). Stepanov argues that the impossibility of extracting from subjects in English is due to the fact that subjects are moved constituents (see below for details).16 What I would like to concentrate on here is the impossibility of extracting from subjects. As Stepanov points out, there is a great deal of cross-linguistic variation here. The generalization that he derives is that extraction is possible from subjects only in those languages where the subject remains in situ. Consider the following examples (Stepanov 2007, (1a) English and (23a) Turkish, respectively).

14 See

also Sheehan (2009) for additional arguments against Nunes and Uriagereka.

15 One could conceivably argue that both analyses above are possible and that somehow the choice

is made that will give rise to a convergent derivation. It is possible to show that extraction from both positions can happen simultaneously. Consider the following sentence. i.

Which roomj was Johni spoken to ti in tj ?

This sentence, while slightly odd, is still acceptable. Ultimately, one of the two PPs from which the nominal is sub-extracted must appear in a specifier position. 16 Stepanov also argues that adjuncts are added to the derivation post-cyclically (following Lebeaux 1988; Speas 1990). By the time the adjunct is added post-cyclically, it is too late for one of its subconstituents to be extracted and raised to the specifier of a higher probe. By then, the probe and goal are in separate phases.

3.3

35.

The LCA and BPS

69

Variation in sub-extraction from subjects ∗ Who does [a picture of t ] [ t hang on the wall? a. j j i vP i b. [Opi [Ahmet-in ti git-me-si]-nin ben-i go-INF - AGR - GEN I-ACC Ahmet- GEN üz-dü-˘g-ü ev house sadden- PST- COMP - AGR Lit. ‘the house [which [that Ahmet went to __ ] saddened me]’

While the extraction of who in the English example in 35a is ungrammatical, as expected, the extraction of relativization in the Turkish example in 35b is wellformed (which the English counterpart is ungrammatical). Stepanov shows that the subject remains in situ in the Turkish data under consideration. These data support Stepanov’s claim that extraction out of a moved constituent is barred. This makes an interesting prediction about extraction from the subjects of stage- and individuallevel predicates. Kratzer (1995) argues that subjects of individual-level predicates are merged in the IP domain, not predicate-internally. If Stepanov’s proposal is correct, extraction from the subject of an individual-level predicate should be possible in principle, although other factors might result in degraded judgments. The data in 36 support Stepanov’s proposal.17 36.

Extraction out of Subjects of Individual-Level Predicates a. Who would naked pictures of be funny to look at? b. Which continent are the animals on generally quite ferocious? c. The people from that city are cold. d. Which city are the people from cold?

In 36a and b, the subject has been extracted from with individual-level predicates. The predicate in c is ambiguous between a stage-level reading (where ‘cold’ refers to body temperature) and an individual-level reading (where ‘cold’ refers to temperament and means roughly ‘unfeeling’). However, once the subject is extracted from as in d, only the individual-level reading is available. Stepanov accounts for these facts as follows. Assuming the copy theory of movement, PF must delete lower identical copies of moved constituents. Stepanov also 17 The judgments aren’t as clear cut as one would hope. Many examples are significantly degraded,

although still not as ungrammatical as the stage-level counterparts. i. ?∗ What are books about interesting? ii. ?∗ What is a book about interesting? iii. ∗ What is a book about on the table? iv. What is there a book about on the table? I am not sure what accounts for the degraded status of i. and ii., perhaps the type of DP plays a role; however, the generalization still holds that extraction from subjects of individual-level predicates is less degraded than extraction from subjects of stage-level predicates.

70

3 Unifying Antisymmetry and Bare Phrase Structure

assumes an operation, Scan, which evaluates chains for identity at PF. Extracting an element from a moved constituent destroys the identity relation in the original movement, as shown below. 37. [who] was [a picture of who] taken [a picture of who]

In 37, Stepanov argues that deletion under identity at PF can happen in either of two orders: first the lower copy of a picture of who can delete, then the lower copy of who can delete (38 and 39, respectively). 38.

[who] was [a picture of who] taken [a picture of who]

39.

[who] was [a picture of who] taken [a picture of who]

Alternatively, the lower copy of who can delete first, but then we run into a problem, because the lower copy of a picture of who is now distinct from the higher copy, which is now, a picture of who. 40.

[who] was [a picture of who] taken [a picture of who]

41.

[who] was [a picture of who] taken [a picture of who]  crash

Stepanov argues that the operation Scan takes place all at once. Under this view, Stepanov claims, Scan is unable to evaluate the identity or lack thereof of the two copies of a picture of who because the two copies are at once identical and nonidentical, pending deletion of who in the higher copy. Since Scan cannot operate on this derivation, it crashes. There is a problem with this approach, however, with respect to chain formation at PF for deletion purposes. Consider again the input to PF before deletion. Who actually forms a three-member chain. It is true that one of the chains is part of a larger chain since the lower two instances of who are located inside links of the chain formed by the two instances of a picture of who, but if we assume that Scan looks through the derivation for identical copies, it will find all three copies of who and mark the lower two for deletion. 42.

[who] was [a picture of who] taken [a picture of who]

43.

[who] was [a picture of who] taken [a picture of who]

There is no problem, then, in forming a chain from the two instance of a picture of who, since the two copies are now identical. There is, of course, no problem if Scan evaluates the chain for a picture of who first, as described above. Thus, it does not in fact matter which chain is evaluated first. The output is the same in both cases, and this sentence is incorrectly predicted to be grammatical.

3.4

Proposal

71

Finally, I note the following data, which shows that sub-extraction from nominals moved a short distance is possible (and, in reference to the discussion above that sub-extraction from specifiers is possible). In the following example, the passivized object is raised to a low position (Rezac 2006) but, nonetheless, still allows subextraction. 44.

Whoj were there [XP [naked pictures of tj ]i [VP plastered ti all over town]?

Despite the problems with Stepanov’s analysis, he clearly shows that CED effects cannot be treated as a unitary phenomenon as in Nunes and Uriagereka (2000). The next section proposes a unification of BPS and Antisymmetry, which contains a solution to the Initial Merger Problem. The proposal adopts the results of the discussion above, namely that the LCA must be reformulated to be computed over all nodes.

3.4 Proposal The previous sections have pointed out that the initial merger of two heads creates a configuration of symmetric c-command, which must be resolved in order for the LCA to linearize the structure. The point of symmetry in 45 can be eliminated by raising the complement to specifier position in AP:18 45.

AP a

46.

b AP

bi

AP a

ti

At this point, b, strictly speaking, c-commands a, since every category that dominates b (none, in fact) also dominates a. In other words, there is no category that dominates b that does not dominate a. Crucially, the category AP dominates a but 18 It has often been argued that complement-to-specifier raising within the same projection is ruled

out (Abels 2003; Grohmann 2003; Pesetsky and Torrego 2001). Kayne (2004), in particular, rules this type of movement out based on feature checking as follows. Since phrases move to check features, it follows that movement can take place only to a new feature checking position. Thus, any phrase that checks its features in the complement of a head, H, will have no further features to check against H in SpecHP. Since I propose that b in 46 moves to SpecAP not to check a feature, but to eliminate a point of symmetry, I assume this movement is not problematic. This proposal is consistent with a view of syntax in which complement-to-specifier movement for feature checking is ruled out. See also Matushansky (2006) and Toyoshima (2000).

72

3 Unifying Antisymmetry and Bare Phrase Structure

does not dominate b. Thus, b vacuously, and asymmetrically, c-commands a since b is undominated, and b and a can be linearized (see Kayne 1994: 133, fn 3). Before continuing, I would like to remark that, although not crucial to the forthcoming analysis, this proposal is compatible with the Copy Theory of movement (Chomsky 2000 et seq.) whereby Copy Deletion (specifically the choice of which copy to delete) takes place along the lines suggested by Nunes (2004). Thus, traces are used here as a shorthand for copies that are deleted at PF. Note that this does not rule out the possibility of pronouncing lower copies altogether; however, lower copies that violate the LCA (such as ti in 46) cannot be realized. The formulation of the LCA and c-command assumed thus far captures the fact that by the end of the derivation only a single tree (i.e., a single root node) can remain. Assume the following subtrees have been formed in parallel derivations. 47.

AP bi

XP yj

AP a

XP

ti

x

tj

Considering the pair {b, y} and taking the definition of c-command literally, b c-commands y, since every maximal projection that dominates b (which, again, is none) also c-commands y. Thus, the ‘structure’ in 47 is unlinearizable and crashes at the PF interface. This is a welcome result as it captures that notion that a derivation does not converge unless it forms a single phrase marker without further stipulation. Now, once a new head, c, is merged with AP as in 48, b and c are in a relation of symmetric c-command, in violation of the LCA. 48.

CP c

AP bi

AP a

ti

In 48, b and c are dominated by the same set of maximal projections, namely CP. b is not dominated by AP since not every segment of AP dominates b. Every category that dominates b also dominates c (thus, b c-commands c) and every category that dominates c also dominates b (thus, c c-commands b). Asymmetric c-command can be established between c and b the same way as between a and b above. That is, AP can raise to the specifier of CP. This gives us the tree structure shown in 49 and the linear order of heads shown in 50.19 19 Another

option, of course, is to raise b to SpecCP. I discuss this option below.

3.4

Proposal

73

49.

CP APj bi

CP

a

50.

c

AP

tj

ti

b