A Minimalist Approach to Scrambling: Evidence from Persian 9783110199796, 9783110182965

Citation preview

A Minimalist Approach to Scrambling

≥

Studies in Generative Grammar 76

Editors

Henk van Riemsdijk Harry van der Hulst Jan Koster

Mouton de Gruyter Berlin · New York

A Minimalist Approach to Scrambling Evidence from Persian by

Simin Karimi

Mouton de Gruyter Berlin · New York

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

The series Studies in Generative Grammar was formerly published by Foris Publications Holland.

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

Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at ⬍http://dnb.ddb.de⬎.

ISBN 3-11-018296-3 쑔 Copyright 2005 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Berlin. Typesetting: OLD-Media, Neckarsteinach. Printed in Germany.

I dedicate this work to Professor Mohammad Reza Bateni, who introduced me to Generative Syntax

Acknowledgements

This volume is the result of a long journey that took me through various forms of thoughts, reasonings, doubts, and discoveries. Along this path, I greatly profited from discussions with Andrew Barss, Andrew Carnie, Sheila Dooley Collberg, Raffaella Folli, Heidi Harley, Eloise Jelinek, Terry Langendoen, Anne Lobeck, Massimo Piattelli-Palmarini, and Rudy Troike. Several portions of this work were presented at different seminars. I am grateful to our students in those seminars: William Alexander, Lika Chtareva, Jelina Li, Jan Mohammad, Meg O’Donnell, and Azita Taleghani. During my sabbatical at MIT in the spring of 1999, I had the opportunity to discuss some of the issues in this volume with the following individuals, for which I am thankful: Noam Chomsky, Ken Hale, Irene Heim, Anders Holmberg, Howard Lasnik, Alec Marantz, Shigeru Miyagawa, and David Pesetsky. Several portions of this work were presented at different talks, conferences and workshops in the US, Canada, Germany, and Iran. I am grateful to those audiences. I especially thank John Bailyn, Mohammad DabirMoghaddam, Joe Emonds, Julia Horvath, Jila Ghomeshi, Arsalan Kahnemuyipour, Katalin Kiss, Jacklin Kornfilt, Anoop Mahajan, Diane Massam, Karine Megerdoomian, Shigeru Miyagawa, and Vida Samiian. Without several research grants awarded to me by the office of the Social and Behavioral Science Research Institute at the University of Arizona, and travel grants awarded to me by the Foreign Travel Grant Committee at the University of Arizona, I could not have been able to finish this work. I am grateful for those grants. My appreciation to Robert Kennedy who carefully read and edited this manuscript. My mother, Dr. Malakeh Taleghani, has always provided me with great support , both morally and academically. I am infinitely endebted to her. My last, but certainly not least thanks go to Kimea, Rahjue, (Nicholas) Parviz, and (Ashley) Arezu, for bringing sunshine into my life, and for letting me use their names in the data presented in this volume.

Table of Contents

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

Chapter 1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

1.

The goal of the project . . . . . . . . . . . . . . . . . . . . . . . . . .

1

2. 2.1. 2.2. 2.3. 2.3.1. 2.3.2. 2.3.3.

Persian syntax: an overview . . . . . . . . . . . . . . . . . . . . . . Some data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Persian underlying word order: SOV or SVO? . . . . . . The internal structure of clauses . . . . . . . . . . . . . . . . . . The position of T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sentential argument of the verb . . . . . . . . . . . . . . . . . . Complex predicates. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 4 4 7 8 10 11

3. 3.1. 3.2. 3.3.

Scrambling in Persian . . . . . . . . . . . . . . . . . . . . . . . . . . . What elements scramble?. . . . . . . . . . . . . . . . . . . . . . . . What elements scramble in a limited fashion? . . . . . . What elements do not scramble? . . . . . . . . . . . . . . . . .

16 16 18 20

4. 4.1. 4.1.1. 4.1.2. 4.2. 4.3.

Theoretical assumptions . . . . . . . . . . . . . . . . . . . . . . . . . The minimalist program . . . . . . . . . . . . . . . . . . . . . . . . . Basic assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Phase theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Distributed morphology . . . . . . . . . . . . . . . . . . . . . . . . . Specificity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

20 21 21 24 26 26

5.

The outline of the monograph . . . . . . . . . . . . . . . . . . . .

28

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

Chapter 2

Literature on Scrambling . . . . . . . . . . . . . . . . . . . . . . . .

33

1.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33

2. 2.1.

The Base-generation approach . . . . . . . . . . . . . . . . . . . Base-generation approach: pre-MP views . . . . . . . . . .

34 34

x

Table of Contents

2.2. 2.3. 2.3.1. 2.3.2.

Base-generation-approach: MP views . . . . . . . . . . . . . . How does Persian fit in? . . . . . . . . . . . . . . . . . . . . . . . . . Empirical evidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

37 39 39 40

3. 3.1. 3.1.1. 3.1.2. 3.1.3. 3.1.4. 3.2. 3.2.1. 3.2.2. 3.2.3. 3.3. 3.3.1. 3.3.2.

Scrambling as syntactic movement . . . . . . . . . . . . . . . . The A-movement approach . . . . . . . . . . . . . . . . . . . . . . Clause-boundedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . Binding and reconstruction. . . . . . . . . . . . . . . . . . . . . . . WCO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clause bound scrambling and EPP . . . . . . . . . . . . . . . . The A’-movement approach . . . . . . . . . . . . . . . . . . . . . . Parasitic Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Binding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Non-argument, non-operator approach . . . . . . . . . . . . Webelshuth’s dichotomy . . . . . . . . . . . . . . . . . . . . . . . . . Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45 45 46 47 48 49 51 52 54 55 58 58 60

4. 4.1. 4.2. 4.3. 4.4.

Why does Lx, but not Ly, allow scrambling? . . . . . . . . . Verb position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adjunction sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bošković and Takahashi . . . . . . . . . . . . . . . . . . . . . . . . . Recent literature: EPP and discourse . . . . . . . . . . . . . .

61 61 63 65 65

5.

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

66

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

67

Chapter 3

Local Scrambling and A-Movement . . . . . . . . . . . . . . .

69

1.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

69

2. 2.1. 2.2. 2.3. 2.4. 2.4.1. 2.4.2. 2.4.3. 2.5. 2.6.

Subject in Persian. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Unaccusatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Passives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Raising constructions. . . . . . . . . . . . . . . . . . . . . . . . . . . . Subjectless constructions. . . . . . . . . . . . . . . . . . . . . . . . . Inalienable possessor constructions. . . . . . . . . . . . . . . . Inalienable pseudo-possessor constructions . . . . . . . . . Short infinitives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ECM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tough-constructions. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

71 72 74 76 78 78 83 84 86 88

Table of contents

xi

2.7. 2.7.1. 2.7.2. 2.7.3. 2.8. 2.8.1. 2.8.2. 2.8.3. 2.9.

Expletives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Weather constructions . . . . . . . . . . . . . . . . . . . . . . . . . . Overt expletives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Covert expletives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nominative Case, Agreement, and subject shift . . . . . EPP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . pro and PRO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

89 90 91 93 94 95 98 102 104

3. 3.1. 3.2. 3.3. 3.4. 3.5.

Objects in Persian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Specific versus nonspecific . . . . . . . . . . . . . . . . . . . . . . . TOPH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . TOPH revisited: object shift . . . . . . . . . . . . . . . . . . . . . Accusative Case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

104 105 106 108 110 110

4.

Persian as a topic-prominent language . . . . . . . . . . . . .

111

5.

The role of T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

113

6.

Local scrambling: A-movement? . . . . . . . . . . . . . . . . .

115

7.

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

116

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

117

Chapter 4

Operator/Discourse Domain and A’-Scrambling . . . .

122

1.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

122

2.

Adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

124

3. 3.1. 3.2. 3.3.

Topic positions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shifted Topic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Background Topic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-Focus] feature versus EPPs . . . . . . . . . . . . . . . . . . . . .

128 129 129 130

4. 4.1. 4.1.1. 4.1.2. 4.2. 4.2.1.

Focus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kiss’s definition of focus. . . . . . . . . . . . . . . . . . . . . . . . . Identificational focus versus information focus. . . . . . Inherent focus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Focus-movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Non-wh-phrases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

131 132 132 133 134 135

xii

Table of Contents

4.2.2. 4.2.3. 4.3. 4.4. 4.4.1. 4.4.2.

Wh-phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Focus-movement and EPPs . . . . . . . . . . . . . . . . . . . . . . Optionality of focus movement . . . . . . . . . . . . . . . . . . . Contrasted non-wh-phrases . . . . . . . . . . . . . . . . . . . . . . Contrastive focus in other languages. . . . . . . . . . . . . . .

136 149 150 153 154 157

5.

Restrictions on scrambling . . . . . . . . . . . . . . . . . . . . . . .

158

6.

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

161

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

162

Chapter 5

Scrambling, Scope, and Binding . . . . . . . . . . . . . . . . . . .

164

1.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

164

2. 2.1. 2.2. 2.3. 2.4. 2.5. 2.6.

Scope marking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . QP/QP interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wh/QP interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neg and ∃P/∀P Interaction . . . . . . . . . . . . . . . . . . . . . . . Scope of adjuncts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

165 165 167 168 169 170 172

3. 3.1. 3.2. 3.3. 3.4. 3.5. 3.6. 3.6.1. 3.6.2. 3.6.3. 3.6.4. 3.6.5. 3.7.

Binding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Anaphors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Free-standing pronouns. . . . . . . . . . . . . . . . . . . . . . . . . . R-expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clitic pronominals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-linguist analysis of binding . . . . . . . . . . . . . . . . . . German . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hindi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Korean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Japanese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Russian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Binding and phase theory . . . . . . . . . . . . . . . . . . . . . . . .

173 174 178 179 180 182 183 183 185 186 189 189 190

4.

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

191

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

192

Table of contents

xiii

Chapter 6

Long Distance Scrambling and Island Constraints . . .

195

1.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

195

2.

Island conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

196

3. 3.1. 3.2. 3.3.

Scrambling and island effects. . . . . . . . . . . . . . . . . . . . . The data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

201 201 205 209

4.

Processing effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

209

5.

Argument/adjunct structure effects . . . . . . . . . . . . . . .

213

6.

Representation versus derivation . . . . . . . . . . . . . . . . .

219

7.

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

223

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

224

Chapter 7

Theoretical Consequences . . . . . . . . . . . . . . . . . . . . . . .

225

1.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

225

2. 2.1. 2.2. 2.3.

Is typology of movement a myth? . . . . . . . . . . . . . . . . . Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Anti-WCO effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Floating quantifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

226 226 228 229

3. 3.1. 3.1.1. 3.1.2. 3.2. 3.3.

Syntactic positions and interpretation . . . . . . . . . . . . . Improper interactions of Move . . . . . . . . . . . . . . . . . . . Economy of Derivation . . . . . . . . . . . . . . . . . . . . . . . . . The Principle of Unambiguous Binding . . . . . . . . . . . . An alternative account . . . . . . . . . . . . . . . . . . . . . . . . . . Supporting evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . .

231 231 233 235 237 239

4.

Derivational versus representational syntax . . . . . . . .

241

5.

Further research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

243

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

245

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

246

Subject Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

259

Author Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

263

Chapter 1 Introduction

1. The goal of the project The term scrambling was coined in the 60s by John Robert Ross, who devised the first formulation of it, and stated that it was to be distinguished from the phenomenon called free word order. The distinction resided in the idea that scrambling involved movement. He further stated that scrambling was of stylistic nature (Ross 1967). Standard analyses of scrambling go back to Kerstens (1975), de Haan (1979), and Hoekstra (1984), who argue that adverbs have a fixed position, and that scrambling is an optional movement to the left of the adverb. Theoretical work on Japanese in the 80’s (Saito 1985 and 1989, Hoji 1985, among others) made it possible to clearly demonstrate the nature of scrambling as a movement operation with identifiable properties and effects. Literature on scrambling in the 90s and in the first few years of the third millennium reveals interesting syntactic and semantic properties that were not previously addressed1. In the context of the most recent theoretical ideas, scrambling raises at least three important questions: (a) what motivates this movement?, (b) is it an optional operation as it has been considered in the literature (e.g. Saito 1985 and work thereafter, among others)?, and (c) is it semantically vacuous as previously reported (e.g. Saito 1985 and work thereafter) or does it contribute to the semantic component of grammar? This study is an attempt to address these issues by analyzing the interaction of semantic and discourse-functional factors with syntactic properties of word order in a scrambling language, and by discussing the implications of this interaction for Universal Grammar (UG). Three interrelated goals are carefully followed in this work. The first is to analyze the syntactic structure of Persian, a language which exhibits freedom of constituent order. With this analysis, we account for the relative order of categorized expressions, the motivation for their possible rearrangements, and the grammatical results of those reorderings. In this respect, a broad range of major syntactic phenomena are examined, including subject shift and object shift, Case, passive, raising, tough, and the Exception-

2

Chapter 1: Introduction

al Case Marking (ECM) constructions, the Extended Projection Principle (EPP), binding, and scope interpretation of quantifiers, interrogative phrases, adverbial phrases, and negative elements. The second is to connect these insights to similar linguistic properties in languages in which scrambling occurs (e.g. German, Dutch, Hindi, Russian, Hungarian, Japanese, and Korean), and to provide a deeper understanding of this group of typologically related languages. The final and principal goal, however, is to situate the results of this work within the framework of the Minimalist Program (Chomsky 1993, 1994, 1995, 2000, 2001a&b, Lasnik 1993, 1999a&b, and related work), specifically phase theory (Chomsky 2001b). The application of the Minimalist Program (henceforth MP) to Persian syntax raises many questions: can the syntactic properties of this language be descriptively analyzed within the framework of MP? What semantic impact does the reordering of phrases in this language have? Can those rearrangements be correctly accounted for within this theory? How are syntactic processes such as Case determination, anaphoric relations, and scope interpretations analyzed with respect to the fairly free word order in this language? What aspects of these syntactic properties are language-specific and which ones are shared by other languages that allow reordering of categorial expressions? In other words, what aspects of these reorderings can be ascribed to Universal Grammar? Are the parametric properties of Persian of a syntactic nature, or outside the syntax? Are there alternative proposals to previous suggestions in the literature? Is there any need to modify aspects of MP in order to justify the existence of these properties? The investigation in this study indicates that scrambling is not an optional rule in Persian, and that its primary role is to serve semantic and discourse functional outputs. The analysis further shows that certain principles of MP, such as the Minimal Link Condition (MLC), are only seemingly violated by the rules of scrambling in this language. Moreover, the peculiarities of the nature of scrambling and its landing sites has lead some authors to the conclusion that scrambling, as a discourse motivated movement, is best considered as a third type of movement in natural language, next to the familiar A-A’ movements (e.g. Webelhuth 1992). Thus its peculiar differences from the other two types of movements would be accounted for in terms of its nature, which in turn justifies the peculiarity of its landing site discussed in the literature (Webelhuth 1992, Déprez 1994, Mahajan 1994, Dayal 1994). However, careful analysis of scrambling, as well as a reanalysis of the properties of A and A’ movements, cast some doubts on the relevance of a

Persian syntax: an overview

3

three or even two-way distinction with respect to syntactic movements in natural language. It seems, therefore, that what phrasal (XP) dislocations have in common is that they all are triggered by a strong feature that serves the semantic and discourse-functional output of the derivation. Thus binding relations and scope determination follow from different sets of universal properties that are not necessarily related to a specific movement type. In this introductory chapter, we present an overview of Persian syntax in section 2, followed by a brief discussion of what scrambles in this language in section 3. The major theoretical assumptions underlying this investigation are briefly summarized in section 4, followed by the outline of the content of the monograph in section 5.

2. Persian syntax: an overview Modern Persian belongs to the Iranian language family, which itself is a branch of Indo-Iranian. Indo-Iranian is an Eastern branch of the IndoEuropean language family. Modern Persian is spoken in Iran, Afghanistan, and Tajikestan where it is called Farsi, Dari, and Tajiki, respectively. The dialect we are examining in this monograph is a variant of Farsi, called Tehrani, the standard colloquial dialect spoken in Iran. We are utilizing the term Persian to refer to this variant throughout this monograph. In this section, we outline the basic properties of Persian syntax that are relevant to this study. Greenberg (1963) classifies Persian as a type III language. These languages are typically verb-final and postpositional. The genitive case and the modifying adjective precede the head noun, providing genitive- noun (GN) and adjective-noun (AN) orders, respectively. Although the unmarked Persian word order is SOV, all phrasal categories other than VP are head-initial. Therefore, Persian is not a typical type III language. Written Persian exhibits a rigid SOV order, except sentential arguments of the verb systematically appear in post-verbal position. The colloquial language, however, allows a great degree of rearrangements. A sample of data is presented in 2.1. In 2.2., the canonical position of Persian verb is discussed by resorting to Kayne’s (1994) theory of Antisymmetry of Syntax, a theory that suggests an underlying Specifier-HeadComplement (S-H-C) order for all categories in all languages. An analysis of the internal structure of Persian clauses with respect to the head of the Tense Phrase (TP), sentential arguments of the verb, and the status of complex predicates follows in 2.3.

4

Chapter 1: Introduction

2.1. Some data Although sentential arguments of the verb appear in the post-verbal position, all phrasal arguments of the verb precede it in a discourse-neutral statement. Thus, the basic word order with respect to the verb and its phrasal arguments is presented in (1): (1)

a. (S) b. (S)

(O+râ) (PP)

(PP) (O)

V V

In (1), the subject precedes the internal arguments of the verb. (1a) shows a sentence with a direct object marked by râ2, the Case marker for specific objects3. PP represents the position occupied by any subcategorized prepositional phrase. O in (1b) stands for the nonspecific object. Persian allows only one thematic object. Thus O+râ and the bare O cannot both appear in a given sentence. Examples are provided in (2) and (3)4. (2)

Parviz pirhan-o barâ Kimea xarid P shirt -râ for K bought ‘Parviz bought the shirt for Kimea.’

(3)

Parviz barâ Kimea pirhan xarid P for K shirt bought ‘Papar bought shirts for Kimea.’

Sentential arguments of the verb appear in the post-verbal position, as in (4). (4)

Kimea goft ke Parviz xune nist K said that P home not-be ‘Kimea said that Parviz was not home.’

See section 2.3.2. for a brief analysis of sentential arguments of the verb.

2.2. Persian underlying word order: SOV or SVO? The data presented in the previous section suggest that Persian verb appears in the final position underlyingly5 with the exception of sentential arguments of the verb (cf. (4)). The issue of such parametric differences with respect to the underlying word order is challenged by Kayne (1994).

Persian syntax: an overview

5

Kayne advances a radical alternative to the typological standards regarding word order by proposing that word order universally reflects structural hierarchy in terms of the Linear Correspondence Axiom (LCA) (Kayne 1994: 6). The LCA is restated in (5). (5)

d(A) is a linear ordering of T.

(Kayne 1994: 6)

In (5), d represents the nonterminal-to-terminal dominance relation, A is a set consisting of ordered pairs such that for each j, Xj asymmetrically c-commands Yj, and T is the set of terminals. The consequences of the LCA are stated in (6): (6)

Consequences of the LCA

a. There is a universal order consisting of Specifier-Head-Complement (S-H-C), and therefore, any order other than S-H-C is a derived one, b. For every moved phrase, there must be a distinct head to whose projection the moved element can adjoin as its Specifier. Thus all SOV languages must be derived from an underlying SVO order. In this regard, Kayne states that … SOV (and more generally S-C-H) is strictly impossible, in any language, if taken to indicate a phrase marker in which the sister phrase to the head (i.e., the complement position) precedes that head. On the other hand, SOV (S-C-H) is perfectly allowable if taken to indicate a phrase marker in which the complement has raised up to some Specifier position to the left of the head. (Kayne 1994: 35)

On the basis of Kayne’s theory, we must think of all those word order variations that violate the LCA in terms of different combinations of movements. For example, an OV order is derived from a VO order by the movement of O into a Specifier position higher than V. In a footnote, Kayne specifically suggests just that: … in the case of DP complements, the movement that positions the O to the left of V in an OV language could be thought of as significantly akin to “scrambling.” This might contribute to the understanding of why what has been called scrambling in the literature is more generally found in OV languages than in VO languages. (Kayne 1994: 140)

6

Chapter 1: Introduction

Kayne finally states that since Specifier-Head-Complement order is the only underlying order made available by UG, there can consequently be no directionality parameter for basic word order. Returning to Persian, let us first consider the implication of LCA stated in (6a). At first glance, this part of Kayne’s theory appears to be able to account for the surface order in this language. The following tree represents the SOV order, derived from an SVO order by movement. The structure of the tree is based on X-bar theory, where TP stands for the Tense Phrase, FP for a Functional Phrase, ‘ts’ for the trace of the subject, and ‘to’ for the trace of the object. (7)

TP Subject

T' T

FP

Object

F' VP

FO ts

V' V

to

The subject moves into the Specifier (Spec, henceforth) of TP, while the object moves into the Spec of a new head, Fo. The verb has the option to stay in-situ or adjoin the head Fo, providing an SOV order (see section 2.3.1. for a discussion of the position of T). Kayne’s theory faces the following problems: a. If SOV is derived from SVO, why is the former semantically and discourse functionally unmarked while the latter receives additional interpretation representing contrastive focus or topic? b. Suppose we assume that the object moves to the preverbal position to receive Case, a purely morphological movement that does not need to have an impact on the semantic content of the sentence. How could we account for the movement of PP if such a movement cannot be motivated by Case?

Persian syntax: an overview

7

c. Deriving the unmarked SOV order from the underlying SVO order requires unnecessary and unmotivated movements. This fact clearly violates the Principle of Economy that is the backbone underlying the Minimalist Program (see section 4 in this chapter). Based on these arguments, I follow the traditional trend that considers Persian a SOV language underlyingly6.

The second implication of Kayne’s theory in (6b), which states that every XP movement has to place that element into a Spec position of a head, provides us with an excellent tool to account for the landing site of scrambled elements in Persian. We will see in this volume that each head may project a Spec if this has some effect on the output of the derivation. Thus, each movement has the property of having an impact on the semantic and discourse functional output of the sentence.

2.3. The internal structure of clauses Assuming Persian to be an SOV language (unlike Kayne), we suggest the phrasal structure in (8) for this language (see section 4 in this chapter, and chapter 4 for revisions)7. (8)

CP C' Spec

TP C

T' Spec

vP T

v' Spec

v

VP PP

VP

Object [+/- Specific] V In (8), the specific and nonspecific objects are base-generated in the same position, although they surface in two different positions in an unmarked word order (cf. (1)). This issue is discussed in section 3 of chapter 3 in this volume.

8

Chapter 1: Introduction

Another issue of interest is the position of the functional heads C and T in (8): they both precede their complements. The presence of the complimentizer in the initial position of the clause (e.g. ke in (4)) justifies the position of C in this configuration. However, the position of T in a verb-final language is debatable. This issue is discussed in the following subsection. 2.3.1. The position of T If Persian is a SOV language, as proposed in this monograph and throughout the literature, the position of the functional head T in (8) calls for some explanation, since it should be to the right of the verb, and thus in the final position. Are there any theoretical and/or empirical pieces of evidence which would require T to be in an initial position in this verb final language? The answer is yes for the following empirical and theory-internal reasons. Empirical reason: The sentential argument of the verb follows its head in Persian, as illustrated by (4), repeated below. (4)

Kimea goft ke Parviz xune nist K said that P home not-be ‘Kimea said that Parviz was not home.’

If T were to be in a final position, the main verb would have to end up in the rightmost position of the sentence whenever it had to move to T (e.g. in a topic construction (cf. chapter 4)). The result would be an ill-formed construction as in (9) where the verb follows the sentential argument. (9)

*man tv [CP ke

Kimea in kâr-ro

mi-gir-e]

goft-am

I that K this job-râ dur-get-3sg said-3sg Intended meaning: I said that Kimea will get this job. Theory internal reasons 1. Within the theory I am assuming in this monograph, an XP can only move into the Spec of a functional head. Consequently, movements are always leftward if Spec precedes head. In a derived SVO order, therefore, the object cannot have moved to the right of the verb in Persian, where the Spec is always on the left side of head. This restriction leaves us with a verb movement that adjoins the verb to a functional head such as T, C,

Persian syntax: an overview

9

or some other functional heads (see chapter 4). Thus, we will have the construction in (10) for a derived SVO order at Spell-out. (10)

TP S

T' T V

vP T ts

v' v

VP

O

V' tv

In (10), the verb is adjoined to T. We will see in chapter 4 that verb movement in Persian is triggered by topic or contrastive focus. The configuration in (10) will be elaborated and revised in that chapter. 2. Verbs are base-generated with all their inflectional affixes within the theory we are assuming (Chomsky 1995). Therefore, V-movement does not apply for the sake of inflectional adjunction. Consequently, T does not have to be in a final position in this verb final language8. 3. Since compliment phrases (CP) and determiner phrases (DP) are headinitial in Persian, we arrive at a nice generalization by assuming that all functional heads appear in an initial position in this language9. As shown in chapter 4. the functional projections representing focus and topic are head-initial as well. The sentential argument of the verb always appears in a final position, similar to German and English. The discussion of the position of T raises the question whether or not this element is base generated in the final position, or its surface position represents a derived structure. This is the topic of the next section.

10

Chapter 1: Introduction

2.3.2. Sentential argument of the verb Sentential arguments of Persian verbs appear in the post-verbal position, as illustrated by (4) as well as the following examples10,11: (11) Kimea az man xâhesh kard [CP ke nâma-sh-o tayp kon-am.] K of me ask did-3sg that letter-her-râ type do-1sg ‘Kimea asked me to type her letter.’ (12) Kimea Parviz-o vâdâr kard [CP ke nâma-sh-o tâyp kon-e.] K P-râ force did-3sg that letter-her-râ type do-3sg ‘Kimea forced Parviz to type her letter.’ One piece of evidence supporting the claim that the clausal argument of the verb is base generated in the post-verbal position comes from the fact that it can never appear in the pre-verbal position, as the ill-formedness of (13) and (14) indicate: (13) *Kimea az man [CP ke nâsma-sh-o tâyp kon-am] xâhesh K of me that letter-her-râ type do-1sg request kard. do-3sg. (14) *Kimea Parviz-o [CP ke nâma-sh-o tâyp kon-e] vâdâr kard. K P-râ that letter-her-râ type do-3sg force do-3sg Note that if (11) and (12) were the result of V-to-T movement, this movement would have to be obligatory in all other cases, contrary to facts. Further support for the claim that sentential arguments of verbs are base-generated in post-verbal position comes from the fact that extraction out of these CPs is possible, as shown by Karimi (2001). (15) [un ketâb- â-ro]i man mi- dun- am [CP ke Kimea ti xaride.] that book-pl-râ I dur-know-1sg that K bought-have-3sg ‘As for those books, I know that Kimea has bought (them).’ The phrase un ketab-â-ro ‘those books’ is extracted out of the sentential argument in (15), suggesting that the CP is base-generated in its surface position12. If it were extraposed to its surface position, it would become an island to extraction, contrary to facts.

Persian syntax: an overview

11

The examples in (13) and (14) contain the complex predicates xâhesh kard ‘request do’ and vâdâr kard ‘force do’. Since verbal concepts overwhelmingly appear as complex predicates in Persian, and present interesting properties in this language, a brief discussion of these elements is presented in the following section. 2.3.3. Complex predicates Persian complex verbs, consisting of a non-verbal element (NV) and a light verb (LV), have been gradually replacing simple verbs in this language since the thirteenth century. The tendency to form complex verbs has resulted in the existence of two sets of verbs, simple and complex, for a number of verbal concepts. In many cases, the application of the simple verb is restricted to the written and elevated language. Sample examples of simple/complex pairs appear in (16) (See Karimi, 1997a&b, and Folli, Harley, and Karimi (in Press) for more data). The productivity of Persian complex predicate formation is such that it has completely replaced the former morphological rule of simple verb formation in this language (Bateni, 1989). (16) Simple âgâhânidan âghâzidan

Complex âgâh kardan âghâz kardan

(informed making) ‘to inform’ (start doing) ‘to start’

The light verb (LV) of Persian complex predicates (CPr) ranges over a number of simple verbs, as shown by Karimi (1997a&b). A sample of LVs employed in CPr constructions is provided in (17) (see Folli, Harley, and Karimi (in Press) for more data). (17) a. b. c. d. e. f. g. h. i.

kardan shodan xordan zadan dâdan dâshtan âmadan andâxtan âvardan

‘to do’ ‘to become’ ‘to collide’ ‘to hit’ ‘to give’ ‘to have’ ‘to come’ ‘to throw’ ‘to bring’

12

Chapter 1: Introduction

j. bastan k. bordan

‘to tie’ ‘to carry’

The light verb kardan ‘to do/make’ has almost entirely lost its heavy interpretation, and is the most productive LV. The LV shodan ‘to become’ is systematically used in so-called passive constructions (see section 2.2. in chapter 3 for a discussion on passive constructions in Persian). Another characteristics of Persian CPr is that its nonverbal (NV) element ranges over a number of phrasal categories, as exemplified by (18) (see Karimi 1997a&b, and Folli, Harley, Karimi (in Press) for additional data). (18) a. N+ LV kotak zadan/xordan dust dâshtan b. A+LV pahn kardan/shodan

derâz keshidan c. Particle+LV bâlâ âvardan bâlâ keshidan d. PP+V be donyâ âmadan be bâd dâdan

(beating hitting/colliding) ‘to beat, to get beaten’ (friend having) ‘to love’ (wide doing/becoming) ‘to spread, to widen’ (tr & intr.) (long pulling) ‘to lie down, to take a nap’ (up bringing) ‘to vomit’ (up pulling) ‘to steal’ (‘to world coming’) ‘to be born’ (to wind giving) ‘to waste’

Persian CPr cannot be considered a lexical unit since its NV element and LV may be separated by a number of elements, including (a) negative and inflectional affixes, (b) the auxiliary verb for future tense, and (c) emphatic elements (Mohammad and Karimi 1992). Furthermore, the NV element of Persian CPr allows limited modification. Finally, Gapping is allowed in the case of these elements (Karimi 1997a&b). These findings

Persian syntax: an overview

13

suggest that the LV and the NV element in Persian CPr are separately generated and combined in syntax, and become semantically fused at a different level. The two parts of CPr enjoy syntactic freedom to a certain degree; nonetheless, their semantic properties are the same as those of single word elements elsewhere in Persian and in the grammars of languages like English. Folli, Harley, and Karimi (henceforth FHK) argue that the conflicting properties of Persian CPr can be easily accommodated in a non-Lexicalist theory such as Distributed Morphology or other radical constructionalist theories like that proposed by Borer (2002), where all interpretation occurs post-syntactically. These authors further state that Persian CPr poses a serious problem for lexicalist accounts, which would essentially need to claim that Persian complex predicates are instances of ‘idioms’, receiving a separate entry in the lexicon complete with their syntactic structure. Employing Hale and Keyser’s (1993, 2002) model, FHK show that the structures provided by this model translate naturally to Persian CPr. That is, unergatives are formed when a nominal element is incorporated into a light verb which selects for an external argument. The examples in (19), taken from FHK, illustrate this idea. The English example appears in (19a) while its Persian counterpart is represented by (19b). (19)

a.

vP

b.

DP John

vP

DP

v’ v

N

DO

cry

‘John cried’

Kimea

v’ N gerye ‘cry

v kard ‘did’

‘Kimea cried’

The difference between (a) and (b) in (19) is that there is a head movement involved in (a) while the Persian example in (b) does not need this movement since the LV is lexical in this language. Note that the Persian verb appears in the final position due to the basic SOV order in this language. Similarly, the syntax of a CPr that translates as a typical inchoative, like bidâr shodan ‘awake becoming’, receives a similar structure as its English counterpart.

14

Chapter 1: Introduction

(20) a.

b.

vP

vP

AP

v BECOME

A

DP

awake

v

AP DP

A

John Kimea

‘John awoke’

bidâr shod ‘awake’ ‘became’ ‘Kimea awoke’

FHK further show that the alternation between the inchoative and the causative of awake in Persian is accomplished by changing the light verb from the equivalent of ‘become’ (shodan) to the causative ‘make’ (kardan), as hypothesized by Hale and Keyser for the English causative/inchoative alternation. (21) a.

vP

b.

DP Sue

vP DP

v' v

AP

CAUSE

A

v'

Parviz DP

awake John

AP D

v A

Kimea-ro bidâr kard awake make

The data in (20b) and (21b) show that altering the particular light verb in a Persian CPr can affect the appearance or absence of an Agent argument, as expected on a vP-shell theory of argument structure. FHK further show that this is a general property of the LV in Persian CPr. Moreover, they demonstrate the tight relationship between event type and the category of the NV element in the CPr. That is, the category of the complement to v determines the event type of the CPr, when the LV itself is not inherently telic. (22) presents a summary of the role of Persian LV, and (23) summarizes the role of Persian NV element in CPr. Both tables are borrowed from FHK.

Persian syntax: an overview

15

(22) The role of LV in CPr 1. Agentivity/ Causativity 2. Eventiveness 3. Duration

(23) Telic

Atelic

PP + LV Ex: be donyâ âmadan (to world coming) ‘to be born’ be âtash keshidan (to fire pulling) ‘to put on fire’

N + LV Ex: dast xordan (hand colliding) ‘to get touched’ dâd zadan (scream hitting) ‘to yell’ gush dâdan (ear giving) ‘to listen’ dast andâxtan (hand throwing) ‘to mock’

Particle + LV Ex: kenâr âmadan (side coming) ‘to get along, agree’ dar gozashtan (away passing) ‘to pass away’ A + LV Ex: derâz keshidan (long pulling) ‘to take a nap’ Eventive Nominal + LV Ex: shekast xordan (defeat colliding) ‘to be defeated’ shekast dâdan (defeat giving) ‘to defeat’ farib dâdan (deceit giving) ‘to deceive’

Table (22) shows that Persian LV determines the agentivity/causativity, eventiveness, and duration of the whole CPr, while table (23) shows that the NV element determines the telicity of the whole CPr: CPr is atelic if the NV is a (noneventive) N. Otherwise, it is telic. However, if the LV is inherently telic, such as shodan ‘become’, the NV element will not have an effect on the telecity of the whole CPr, as in (24) (borrowed from FHK): (24) a. xorshid barf-ro âb kard sun snow-râ water made ‘The sun melted the snow.’ b. barf âb shod snow water became ‘The snow melted.’

16

Chapter 1: Introduction

In chapter 3, we extend the structure of Persian CPr discussed in this section to passive constructions, and show that Persian passive constructions constitute a subset of CPr in this language.

3. Scrambling in Persian In this section, we offer a descriptive discussion of elements that undergo scrambling (section 3.1.), are subject to limited scrambling (section 3.2.), and do not undergo scrambling at all (3.3.). Theoretical analysis of scrambled elements, including their interpretation, is discussed in chapters 3–5.

3.1. What elements scramble? All phrasal arguments and adjuncts are subject to scrambling in this language. The only exception is the nonspecific subject and object that may undergo scrambling in a limited fashion (see section 3.2.). In addition, multiple scrambling in a clause is also allowed. The following examples exhibit scrambling of arguments in the main and subordinate clauses. Depending on the stress of the scrambled element, it may receive a topic or a contrastive focus interpretation. (25) Scrambling of the specific object over the subject a. Pirhan-o Parviz barâ Kimea xarid shirt -râ P for K bought ‘As for the shirt, Parviz bought (it) for Kimea.’ Or ‘It was the SHIRT that Parviz bought for Kimea.’ b. Parviz goft ke Pirhan-o Rahjue barâ Kimea xarid P said that shirt -râ R for K bought Lit. Parviz said that, as for the shirt, Rahjue bought (it) for Kimea. Or ‘Parviz said that it was the SHIRT that Rahjue bought for Kimea.’ (26) Scrambling of the Indirect Object over the Subject a. be Sasan hame mi-xand-an to S everyone dur-laugh-3pl ‘As for Sasan, everyone laughs at (him).’ Or Lit: It is at SASAN that everyone laughs.

Scrambling in Persian

17

b. Arezu goft ke be Sasan hame mi-xand-an A said that to S everyone dur-laugh-3pl Lit. Arezu said that as for Sasan everyone laughs at (him). Or Arezu said that it is at SASAN that everyone laughs. Persian allows long-distance scrambling as well. The following examples exhibit the movement of the embedded subject, specific object, and indirect object into the matrix clause. Note that Persian is a Null Subject language, and therefore, pro appears in the following data. All scrambled elements can be interpreted as focus or topic, depending on the stress they carry. Only the topic interpretation is provided in the English translation here. (27) Long distance scrambling of the embedded subject Kimea pro mi-dun-am ke in film-ro did-e K dur-know-1sg that this movie-râ saw-3sg ‘As for Kimea, I know that (she) has seen this movie.’ (28) Long distance scrambling of the embedded specific direct object in film-ro pro mi-dun-am ke Kimea did-e this movie-râ dur-know-1sg that K saw-3sg ‘As for this movie, I know that Kimea has seen (it).’ (29) Long distance scrambling of the embedded indirect object be Kimea man fekr mi-kon-am ke Arezu un ketâb-ro to K I thought dur-do-1sg that A that book-râ dâd-e gave-3sg ‘To Kimea I think that Arezu has given that book.’ Furthermore, scrambling of multiple arguments is also possible, as exemplified by the following sentence. (30) [be Sepide] [ketâb-â -ro] man dâd – am to S book -pl – râ I gave- 1sg ‘As for the books, TO SEPIDE I gave (them).’ or ‘As for Sepide, THE BOOKS I gave her.’ Adjuncts undergo scrambling as well:

18

Chapter 1: Introduction

(31) tu sinamâ pro fekr mi-kon-am pro Kimea-ro did in movie theater thought dur-do-1sg K -râ saw ‘It was in the movie theater that I think she saw Kimea’ Persian does not exhibit structural wh-movement. However, wh-arguments and wh-adjuncts are both subject to scrambling: (32) a. Ki goft-i xuna-ro xarid? who said-2sg house-râ bought-3sg ‘Who was it you said bought the house?’ b. chetori fekr mi-kon-i pro mâshin-ro dorost kard how thought dur-do-2sg car-râ fix did ‘How do you think she fixed the car?’ Scrambling of two wh-phrases is also possible. (33) [ki ] [chi-ro] fekr mi-kon-i be-xar-e who what-râ thought dur-do-2Sg subj-buy-3sg ‘Who is it you think will buy what?’ We will see in chapter 4 that the order of two scrambled wh-phrases is crucial for reasons discussed in that chapter.

3.2. What elements scramble in a limited fashion? Three types of elements undergo scrambling in a limited fashion: nonspecific subjects and objects, NV elements of CPr, and VPs. Although the specific object followed by râ may appear in different positions, as we saw in (25) and (28), the nonspecific object appears adjacent to the verb. It may be separated from the verb only if it bears a contrastive stress as in (34) and (35). That is, a topic interpretation is not possible in these cases (see chapter 4 for explanation). (34) Local scrambling of the nonspecific object a. Parviz PIRHAN barâ Kimea xarid P shirt for K bought ‘It was SHIRTS that Parviz bought for Kimea (not dresses).’

Scrambling in Persian

19

b. PIRHAN Parviz barâ Kimea xarid shirt P for K bought ‘It was SHIRTS that Parviz bought for Kimea (not dresses).’ (35) Long distance scrambling of the nonspecific object PIRHAN Parviz goft Kimea barâ dust-esh xarid shirt P said K for friend-her bought ‘It was SHIRTS (that) Parviz said Kimea bought for her friend.’ (not dresses) The same restrictions hold for nonspecific subjects of the so-called passive and unaccusative constructions (see chapter 3 for discussion). Another type of element that undergoes scrambling in a limited fashion is the NV element in Persian CPrs. These elements cannot be scrambled out of V’ unless they contain a quantificational element and receive heavy stress (Karimi 2003c), as attested by the contrast in (36). (36) a. Kimea [che zamin-e saxti ]i diruz [CPr ti xord ] Kimea what earth-Ez hard yesterday collided ‘What a hard fall Kimea had yesterday.’ Lit. Kimea what a hard earth yesterday collided. b. *Kimea zamin-e xaxti diruz xord Kimea earth-Ez hard yesterday collided In (36a), che ‘what’ provides a quantificational reading for the nonverbal element, and allows it to scramble. This element is missing in (36b), resulting in the ill-formedness of this sentence. The third type of elements that are subject to limited scrambling are verb phrases. These elements might locally undergo scrambling over the subject as in (37). (37) [un film-ro did-an] una that movie-râ saw-3pl they ‘They saw that movie.’ However, long distance scrambling of VP is not possible: (38) *[un film-ro did-an] Kimea goft una that movie-râ saw-3pl K said they

20

Chapter 1: Introduction

An explanation for the limited scrambling of the elements discussed in this section is provided in chapter 4.

3.3. What elements do not scramble? We saw in section 2.3.2. that sentential arguments of the verb may not appear in a preverbal position, as in (13) and (14), repeated below. (13) *Kimea az man [CP ke nâsma-sh-o tâyp kon-am] xâhesh K of me that letter-her-râ type do-1sg request kard. do-3sg. (14) *Kimea Parviz-o [CP ke nâma-sh-o tâyp kon-e] vâdâr kard. K P-râ that letter-her-râ type do-3sg force do-3sg These elements cannot be scrambled to the clause initial position either. (39) *[CP ke nâma-sh-o tâyp kon-am] Kimea az man xâhesh kard. that letter-her-râ type do-1sg K of me request did (40) *[CP ke nâma-sh-o tâyp kon-e] Kimea Parviz-o vâdâr kard. that letter-her-râ type do-1sg K P-râ force did The lack of scrambling with respect to sentential arguments of the verb, and the reasons behind it, are discussed in chapter 4.

4. Theoretical assumptions The theoretical framework employed in this project is the Minimalist Program (MP), specifically the most recent version of this model as introduced by Chomsky (2000, 2001a, 2001b). The basic assumptions underlying MP are presented in 4.1. Distributed Morphology is adopted where ever it becomes relevant in this work. The very basic properties of this theoretical framework are introduced in 4.2. Due to the importance of the semantic notion specificity in this work, a brief discussion of this term is provided in 4.3. More specific theoretical properties are introduced where ever they are relevant to the discussions in the text.

Theoretical assumptions

21

4.1. The Minimalist Program In this project, we employ the syntactic model known as the Minimalist Program as laid out by Chomsky (1993, 1994, 1995, 2000, 2001a) and specifically Chomsky’s Derivation by Phase (2001b)13. Below, we briefly outline the basic assumptions underlying the Minimalist Program as far as they are relevant to the theoretical analyses in this project. 4.1.1. Basic assumption The basic assumptions of MP employed in this work are described below14: Merge and Move: the computational system consists of two recursive operations: Merge and Move. Merge is the simplest such operation which takes a pair of syntactic objects (α and β), and forms a more complex one K constructed from α and β. In other words, Merge combines H(ead) with (C)omplement, and this combined unit is then merged with Spec. (41)

K

Spec α

β

Move takes an existing element in the structure, and places it in a c-commanding position. (42)

β Spec

K α

t

22

Chapter 1: Introduction

Features: features are either interpretable or uninterpretable15. That is, they either have a semantic content or are semantically conentless. For convergence, an uninterpretable feature F of the H(ead) (= Probe) must be deleted when it is placed in a local relation with another feature F’ of the head of XP (= Goal). Suppose, for example, that F is an uninterpretable feature such as a wh-feature of C (Complementizer) (C = Probe). F must find the feature F’ in a wh-phrase (H of wh-phrase = Goal), such that F and F’ agree, and F is deleted. Unlike uninterpretable features, interpretable features, such as agreement feature (φ-feature) of a noun, will not delete and are accessible at Logical Form (LF)16. In general, categorial features (the category an element belongs to, such as noun and verb) and the φ-features of the determiner (D) are interpretable, while the Case feature of D, v (the head of vP, the larger verbal projection), and tense (T) is unnterpretable. In the earlier version of MP, the matching requirement of uninterpretable features could trigger overt dislocation of an XP in one language, but not in others, thus dividing functional features into two groups: strong features (that trigger overt XP movement) and weak features (that do not trigger overt XP movement). The wh-feature, for example, triggers movement in English and German, but not in Persian. Therefore, this feature is considered to be strong in English and German, and weak in Persian. However, as we saw in section 3, Persian wh-phrases may undergo scrambling, representing contrastive focus. Thus, languages differ in their choice and association of features. These differences are captured in a more unified fashion in the most recent version of MP: movement is triggered only when a specific feature (EPP) is chosen from the lexicon, eliminating the strong/weak distinction (Chomsky 2000) (see specifically the discussion in chapters 3 and 4). Locality and Agree: The uninterpretable feature F of H (Probe) must be locally in an Agree relation with its identical counterpart F’ in the head (Goal) of a local XP. Locality, therefore, comes into play since F’ must be easily recognizable in the domain of F17. The Minimal Link Condition (MLC) is a manifestation of locality in this framework. This condition is stated in (43). (43) Minimal Link Condition (MLC) K attracts α only if there is no β, β closer to K than α such that K attracts β. (Chomsky 1995: 311)

Theoretical assumptions

23

Where ‘close’ and ‘c-command’ are defined as in (44i) and (44ii), respectively. (44) (i) β is closer to the target K than α if β c-commands α. (ii) X c-commands Y if a. every Z that dominates X dominates Y, and b. X and Y are disconnected. If, for example, X in (45) contains the feature F, and both WP and ZP contain the feature F’, MLC allows Agree between X and WP, but not X and ZP. (45)

XP X' Spec YP X Y' WP ZP Y

Economy: the derivation of a sentence is subject to general conditions of the Principle of Economy, restated in (46). (46) Principle of Economy (Chomsky 1995: 367) a. Add optional α to the numeration only if it has an effect at the interface. b. At each stage of a derivation, apply the most economical operation that leads to convergence. The Principle of Economy in (46) requires that the Principle of Full Interpretation 18 be satisfied (46a), and elements are minimally linked (46b). Thus MLC is a representation of (46b) as well. An implication of the Principle of Economy is that Merge is more economical than dislocation since the former requires less operations than the latter. Another implication of this principle is that syntactic operations are NOT optional. Since scrambling has been considered to be optional (Saito 1985 and work thereafter, Hoji 1985, among others), the optionality of

24

Chapter 1: Introduction

movement from a Minimalist point of view is naturally a central issue in this monograph. The question is whether optionality of movement is a syntactic phenomenon or it is of a different nature. 4.1.2. Phase theory The derivational aspect of the analysis advanced in this work is based on phase theory (Chomsky 2001b). A summary of this model is outlined below: a. Phases are propositional. vP and CP are the strong phases b. vP is the lexical phase, containing VP c. CP is the functional phase, containing TP d. v and C may be assigned a feature representing the Extended Projection Principle (EPP) to trigger movement19. e. The domain of H (Head) is not accessible to operations outside of HP, but only H and its edge: at every phase, the history of derivation within the previous phase is lost. f. An element that needs to be raised has to move to the edge of the phase. g. Edge is the residue outside of H-bar, either Spec or elements adjoined to HP. h. Interpretation/evaluation takes place at the phase level. Therefore, at the end of each phase the structure is sent to Logical Form (LF) and Phonological Form (PF) for interpretation. i. Therefore, there is no overt/covert distinction with two independent cycles, j. Spell-out is cyclic at the phase level. k. Once the operations are finished in one phase, there is no return to that phase. The two phases suggested by Chomsky are illustrated in (47). (47) [CP [C’ [TP [T’ [vP [v’ [VP [V’ … ]]]]]]]] Functional Phase Lexical Phase An example of a syntactic movement, wh-movement in this case, is provided by (48).

Theoretical assumptions

25

(48) Who [ did [vP t [ you say [CP t [ Mary [vP t [ thought [CP t [ John EPP EPP [vP t [ kissed t ]]]]]]]]]]]

EPP

EPP

EPP

EPP The feature representing the Extended Projection Principle (EPP) is basically a strong feature that triggers movement. This feature can be checked many times since it serves the requirement that certain (functional) heads project a Spec. In (48), for example, the wh-word is moved to satisfy the EPP features of v and C. Thus there is no ‘looking ahead’ phenomenon: that is, the movement of the wh-phrase is not triggered by a higher head (the matrix C) that does not yet exist. It is rather triggered by the EPP feature of v (and the intermediate C if there is a long-distance movement involved). The following points will become crucial in our analysis in the text: a. Pred(icate)P (the XP complement of v) contains the internal arguments and is the domain of existential closure (Diesing 1992, Kratzer 1995, Diesing and Jelinek 1995). Specific DPs move out of this phrase to receive interpretation. b. vP is the lexical phase, containing the PredP and the agent (in the case of agentive predicates). c. The strong phase above vP (= CP) represents operator/discourse functional elements such as contrastive focus, topic, and wh-operator. This phase consists of TP (whose optional Spec is reserved for some type of topicalized elements), optional FP (reserved for contrastively focused elements), optional TopP (reserved for topicchanging elements) and CP (reserved for clause type features and wh-operators). The configuration in (47) represents (a)–(c) above. (49) [CP [TopP [FP [TP [T’ [ [vP [v’ [XP [X’ ]] v ]]]]]]]] Operator/Discourse Phase

PredP

Lexical Phase The relevance of (49) becomes apparent in the body of the text. This phrase structure is revised in chapter 4 to include negation.

26

Chapter 1: Introduction

4.2. Distributed Morphology Following the basic ideas of Distributed Morphology (Halle and Marantz 1993), we assume that Lexical items consist of a bundle of semantic, phonological, and formal features, and the phonological insertion of these bundles appears post-syntactically. Morphemes display two distinct descriptions: the Morphosyntactic description, consisting of the root plus anything else attached to it, and the Morphophonological description, representing the Spell-out version of the lexical item. Thus a word such as cats will have the following descriptions post-syntactically: (50) cats a. Morphosyntactic description: [ Root [+plural]] b. Morphophonological description: [ kæt + s ] Similarly, ketâb-râ has the descriptions in (51): (51) ketâb-râ a. Morphosyntactic description: [ Root [+Specific/Acc]] b. Morphophonological description: [ ketâb + râ] ] See section 2.3. in chapter 2 where this description becomes relevant.

4.3. Specificity Specificity plays a significant role when interacting with scrambling. This is specially apparent with respect to direct objects: there are syntactic, semantic, and morphological differences between the specific objects and their nonspecific counterparts. Similar differences are observed with respect to specific and nonspecific subjects of unaccusative and (so-called) passive constructions. Due to the relevance of specificity to the analysis advanced in this work, we present a brief description of this phenomenon in this section. Enç (1991) defines specificity in terms of strong antecedent and weak antecedent. She states that a definite DP requires a strong antecedent. That is, there is an identity relation between this type of DP and its previously established discourse referent. Therefore, definite DPs are always specific. Proper names, pronouns, and noun phrases modified by a demonstrative or a definite article are definite, and thus specific. An indefinite DP

Theoretical assumptions

27

is specific if it denotes an inclusion relation to previously established discourse. In this case, it represents a weak antecedent. A nonspecific DP lacks an antecedent in the discourse altogether. Thus indefinites are two ways ambiguous: specific indefinite and nonspecific indefinite. In Karimi (2003c), a revised version of Enç’s (1991) definition of specificity was proposed. According to this revised version, the specific indefinite either denotes inclusion or is singled out in discourse (e.g. by a relative clause). A nonspecific DP either lacks a referent (= kind-level) or denotes existence (= existential). This is summarized in (52). (52)

DP Specific Definite

Nonspecific

Indefinite Existential Kind-level

Partitive

Modified (Karimi 2003c: 97)

The data in (53)–(57) represent different types of specific and nonspecific DPs. The relevant DP appears in italic. (53) Kimea un ketâb-ro xund K that book-râ read ‘Kimea read that book.’

definite

(54) Kimea yeki az ketâb-â-ro xund K one of book-pl-râ read ‘Kimea read one of the books.’

Partitive

(55) Kimea ye dâstân-i -ro goft ke az to shenide bud K a story-rel-râ told that fromyou heard was ‘Kimea told a story that (she) had heard from you.’

Modiied

(56) Kimea har ruz ketâb mi-xun-e K every day book hab-read-3sg ‘Kimea reads books every day.’

Kind-level

28

Chapter 1: Introduction

(57) Kimea emruz ye ketâb xund K today a book read ‘Kimea read a book today.’

Existential

In chapter 3 we will see that the specific DP, definite or indefinte, moves out of the PredP (VP or any complement XP of v), while the nonspecific elements remain in situ unless movement represents contrastive focus.

5. The outline of the monograph Chapter 2 provides a review of literature regarding the nature of scrambling and languages that reveal this property, and thus sets the stage for an in-depth analysis of scrambling in general, and Persian scrambling in particular, in the subsequent chapters. The survey starts with a discussion of the theories that consider scrambling as a base-generation phenomenon. In this regard, pre-MP literature (Neeleman 1994, Bayer and Kornfilt 1994, among others) as well as post-MP works (Bošković and Takahashi 1998 and Saito and Fukui 1998) are briefly discussed. It is shown that scrambling in Persian cannot be considered as a base-generation phenomenon. The survey moves on to include an analysis of Scrambling as a syntactic movement (Mahajan 1990, 1994; Déprez 1994; Saito 1989, 1992, 1994a, 1994b, 1994c; van Riemsdijk 1989; Webelhuth 1992; Vikner 1994; Müller and Sternefeld 1993, 1996, among others). Three different approaches to this movement are critically discussed. Finally, the question of why some languages allow scrambling and some do not, is addressed, literature on this issue is surveyed, and problems are pointed out. The discussions advanced in the following chapters, however, show that scrambling is just a myth. That is, movements attributed to scrambling are driven by the same syntactic tool that is responsible for movements observed in non-scrambling languages. If this line of argumentation is on the right track, ‘scrambling’ becomes a meaningless term, representing an era when our understanding of so-called scrambling languages was insufficient. Chapter 3 is devoted to an analysis of local scrambling within Persian vP. A number of syntactic structures that involve subjects, including unaccusatives, passives, tough, raising and subjectless constructions are examined. It is argued that Persian lacks typical A-movement involving subjects. It is also shown that this language has no overt expletives, and there is no evidence pointing to the existence of covert expletives. Furthermore,

The outline of the monograph

29

two types of EPP are discussed in this chapter, and one of them is argued to apply morphologically in this language, and the other to be responsible for XP movements in general. The two types of direct objects, specific and nonspecific, are discussed, and is shown that there are intriguing similarities between subject and object DPs. For example, this language exhibits both subject shift and object shift of specific DPs for primarily interpretive reasons. Nonspecific subjects and objects undergo scrambling only to receive a contrastive interpretation. It is concluded that Persian is a topic prominent language that has no A-movement in a traditional sense, and that all movements into the operator/discourse domain (= CP) are triggered by the second type of EPP, a purely syntactic device serving the semantic and discourse functional component of language. Chapter 4 is an examination of movement into the discourse/functional domain of the clause. The discussion in this chapter shows that sentential and vP adverbs occupy the vP boundary in an orderly fashion. Movements to the left of these elements represent discourse-functional operations, or involve movement of quantificational elements. These movements place the extracted XP in the Spec of a functional/quantifier head in the discourse/functional domain. The Spec of TP in this domain is suggested to be a topic position, representing the background topic. Any element, regardless of its grammatical function, may be placed in that position. Another topic position, representing the shifted topic, is suggested to occupy a higher position between TP and CP. FocP, a maximal projection representing contrastive/ identificational focus, is proposed to be sandwiched between the two topic positions. It is argued that movement into the functional/discourse domain is triggered by the optional selection of one type of EPP, and has an effect on the semantic output of the derivation. Furthermore, the claim that movement into the functional/discourse domain is feature-driven is shown to be supported by the fact that the movement of two elements belonging to the same category is subject to MLC. Finally, the discussion in chapter 4 provides an explanation for restricted scrambling in all those cases introduced in this introductory chapter. The semantic effects of scrambling is further examined in chapter 5. The interaction of different types of scope-bearing elements, as well as blocking effects created or rescued by scrambling, are examined in this chapter. Moreover, the interaction of scrambling with binding relations is discussed, and is shown that scrambling may feed Principles B and C of the Binding Theory, but may not block them. One exception to this gen-

30

Chapter 1: Introduction

eralization, the clitic pronominal, is also discussed in this chapter. Principle A is shown to be unaffected by scrambling, although some parametric differences are observed and analyzed. A cross-linguistic examination of the binding relations is provided which reveals striking similarities between Persian and other scrambling languages such as German, Hindi, Japanese, Korean, and Russian. Chapter 6 concentrates on Long Distance Scrambling (LDS) and constraints that govern this type of operation. A variety of interesting data that exhibit island effects are introduced, and two types of explanations are proposed: processing effects and syntactic effects. The syntactic constraint is shown to be responsible for cases where scrambling is blocked. This constraint is then extended to account for operator movements as well. It is argued that the only distinction between the two types of movement is the domain of the application of this constraint: while CP is the island domain for operator movement, vP serves as the island domain for LDS. The discussion in this chapter sheds some additional doubt on the merit of the typology of movement. It also provides further evidence that interpretation cannot solely rely on a phase-based derivational syntax. These issues are picked up in the final chapter. The final chapter offers two major theoretical consequences. The bulk of this chapter is devoted to an examination of the typology of movement, and raises some questions with regard to the relevance of the existence of an A/A’ distinction as a fundamental property of UG. The second issue concerns the problems that a purely derivational syntax based on the phase theory may pose for UG. Unsolved problems are introduced and reserved for future research. Notes 1. Recent literature on scrambling includes Saito (1992, 1994a,b), Grewendorf and Sternefeld (1990), Webelhuth (1992), Zwart (1993, 1997), Corver and van Riemsdijk (1994), Müller and Sternefeld (1993, 1996), Miyagawa (1997, 2001, 2003), Bošković and Takahashi (1998), Saito and Fukui (1998), Grewendorf and Sabel (1999), Bailyn (1995, 2001, 2003), Karimi (1999c, 2003a), among others. 2. There is no definite article in Persian. Thus O+râ receives a definite interpretation in the absence of an indefinite determiner. 3. Râ appears as -o and -ro in colloquial language. See Karimi (1990, 1996, and 2003c) and Ghomeshi (1997b) for analyses of this element. 4. Grammatical Glosses Pl = plural, ind = indefinite, dur = duration, sg = singular rel = relative, neg = negation, part = particle, subj = subjunctive

Notes

5.

6.

7.

8. 9.

10. 11. 12.

13.

14. 15. 16.

17.

18.

31

Ez = Ezafe particle. Ezafe construction is a DP consisting of the head (an element with the feature [+N] such as N or A), its modifier(s), an optional possessor DP, and the Ezafe particle e that is structurally utilized as a link between the head and its modifier. For different analyses of Ezafe Constructions see Samiian (1983, 1994), Karimi and Brame (1986), and Ghomeshi (1996, 1997a) Studies by traditional grammarians have considered Modern Persian a verb final language (Foroughi (1944), Boyle 1966; Khanlari 1976; Vazinpour 1977; Lazard 1992). Works by linguists have confirmed an SOV order for Persian (Jazayery and Paper 1961; Bateni 1969, 1977; Moyne 1970; Soheili Isfahani 1976; Farrokhpey (1979), DabirMoghaddam 1982; Samiian 1983; Darzi 1996; and Ghomeshi 1996). Marashi (1970) is the only linguist who has suggested an SVO order for this language. Haider (2000) states that Kayne’s LCA-theory is not acceptable since this theory is not tested in SOV languages. Our analysis in this work is compatible with Haider’s statement. The development of the phrasal structure within the theory of Principles and Parameters is made by contributions by Chomsky (1986b) with respect to CP, Stowell (1981) and Pesetsky (1982) regarding IP, Pollock 1989 for split IP, Larson (1988) for layered VP, and Brame (1981) and Abney (1987) for proposing DP as the functional domain of noun phrases. Zwart (1993: 86) argues along the same lines for Dutch. Ghomeshi (1997a) argues that D (in an Ezafe construction) appears in a final position within its projection. However, she has to consider two types of D since demonstratives appear in a DP initial position. Persian does not make gender distinctions, not even in its pronominal system. We use ‘she’ and ‘her’ in the English translations wherever the referent is arbitrary. For a more elaborated analysis of sentential arguments of verbs see Karimi (2001). This is the analysis advanced by Koster (1987) for German and Dutch. Also Zwart (1997) argues that sentential arguments of verbs are base-generated in the post-verbal position in Dutch, a language that is verb final underlyingly. The earlier stages of the Minimalist Program are the Standard Theory (Chomsky 1965), the Extended Standard Theory (Chomsky 1970, Jackendoff 1977), the Revised Extended Standard Theory (Chomsky 1973, 1977, 1980, Chomsky and Lasnik 1977), Principles and Parameters Theory (or Government and Binding Theory) (Chomsky 1981, 1982, 1986a, 1986b, 1991, and Chomsky and Lasnik 1993). For simplicity of presentation, we ignore differences among Chomsky (2000), (2001a) and (2001b) and details that are not relevant to the arguments presented in this work. The existence of uninterpretable features raises a number of empirical and conceptual problems. For a critical view see Karimi (2003d). Within this theory, phrases such as ArgP that consist of categories with only uninterpretable features cannot exist. The reason is that they remain with no label once their uninterpretable features are matched and deleted. One departure from the earlier version of MP is that feature checking is not required to apply in a Spec-head relation. All that is required is a c-command relation in a local domain. The principle of Full Interpretation was first introduced in Chomsky (1986b: 98) with respect to the Visibility Condition which requires an NP to be assigned Case at S-structure in order to be visible at LF for the sake of theta assignment. That is, an NP is interpretable at LF only if it is assigned a theta role, and the assignment of a theta role is possible only if an NP is assigned Case. This concept is changed within the MP in the sense that Case features, by virtue of being uninterpretable, must be eliminated after Agree. An

32

Chapter 1: Introduction

undeleted Case feature is uninterpretable at LF, and thus violates the Principle of Full Interpretation. Therefore, Visibility Condition is reduced to convergence at LF within this model (Chomsky 2001b: 21). 19. The EPP has gone through several interpretations since 1981: Chomsky (1981: 27) defines it as ‘… the structural requirement that certain configurations must have subjects.’ In Chomsky (1982) EPP becomes independent of the Projection Principle: it is the requirement that each sentence has a subject. Chomsky (1995) considers the EPP to be the D feature of T that triggers the subject to move into the Spec of TP. In Chomsky (2000: 109) the EPP becomes a feature assigned to functional heads. See Lasnik (2001), who suggests that the EPP feature forces the head to have a Spec resulting in object shift and subject shift. This analysis is compatible with our discussions in chapters 3 and 4 where, respectively, the subject/object shift and topic/focus are examined.

Chapter 2 Literature on Scrambling

1. Introduction This chapter is devoted to a review of literature on scrambling in the last two decades. It offers an overview of the most general trends in this research area, and sets the stage for an in-depth analysis of scrambling in general, and scrambling in Persian in particular, in the subsequent chapters. Generative studies on scrambling have taken two major approaches to this phenomenon in the last two decades: I. Base-generation approach: a. From a non-transformational point of view, scrambling is considered to be free word order at the level of D-structure. b. From an MP point of view, scrambling is considered to be the result of Merge. II. Movement approach: Scrambling is the result of Move α (or simply Move). From the base-generation point of view stated in Ia, the German sentences in (1a) and (1b) reflect the same order that exists at the level of Dstructure. The same assumption holds for (2a) and (2b). (1)

a. … weil ich wahrscheinlich diesen Mann gesehen habe because I probably this man seen have ‘Because I have probably seen this man.’ b. … weil ich diesen Mann wahrscheinlich gesehen habe

(2)

a. … weil niemand diesen Film sehen will because nobody this movie see wants ‘Because nobody wants to see this movie.’ b. … weil diesen Film niemand sehen will

34

Chapter 2: Literature on Scrambling

Advocates of Ib suggest that these sentences are generated by Merge, followed by some LF movement that places the arguments in their theta positions. For example, Saito and Fukui (1998) and Bošković and Takahashi (1998) argue that the word order variations in Japanese are the result of Merge (See section 2.2. for discussion). From the movement point of view stated in II, (1b) and (2b) are derived from (1a) and (2a), respectively. Thus, I and II make different predictions with respect to the syntactic role of scrambled constituents. Authors who consider scrambling the result of movement have offered two distinct proposals. One group considers scrambling as an optional, semantically vacuous movement (Saito 1985 and subsequent work, Hoji 1985, Fukui 1993, among others). The other group argues that scrambling has an impact on semantic interpretation such as scope identification (Miyagawa 1997, 2001; Beck and Kim 1997, Karimi 1999c, among others). The optionality of scrambling and the semantic effect of this operation form the center of discussion in the next chapters of this volume. We start the discussion in this chapter by providing an analysis of scrambling as a base-generation phenomenon. This discussion includes a review of recent works within the MP framework. We then turn to an analysis of scrambling as a syntactic movement by taking into account the similarities and differences between scrambling, on the one hand, and A and A’ movements, on the other. The organization of this chapter is as follows. In section 2, the base-generation approach is discussed. Scrambling as syntactic movement is addressed in section 3. Scrambling and typology of movement is the subject of section 4. Concluding remarks follow in section 5.

2. Base-generation approach In this section, we briefly review the literature with respect to configurationality and nonconfigurationality of scrambling languages and the Dstructure approach (in 2.1.), the base-generation approach from an MP point of view (in 2.2.), and the situation of Persian (in 2.3.).

2.1. Base-generation approach: pre-MP views From a base-generation point of view, some authors consider scrambling languages to be nonconfigurational. That is, the structure of the clause is as-

Base-generation approach

35

sumed to be flat in these languages (Hale 1980 and Farmer 1980 for Japanese; Haider 1988 for German; T. Mohanan 1990 for Hindi and Urdu; Kiss 1994, 2003 for postverbal elements in Hungarian, among other authors). Others have considered them configurational in the sense that there is a hierarchical, asymmetric relation between the constituents of a sentence (Saito and Hoji 1983, Hoji 1985, Saito 1985 for Japanese; Bayer and Kornfilt 1994 for German; and Neeleman 1994 for Dutch, among others). Assuming a nonconfigurational status for Japanese, Hale (1980) and Farmer (1980) propose the following phrase structure rule for this language20. (3)

X’ → X’* X (*means that there can be any number of X’ phrases, including none.)

However, Whitman (1982) and Saito (1985) have shown that Japanese phrase structure cannot be flat. Their analyses are based on syntactic asymmetries between the external and internal arguments (Williams 1981, Marantz 1981). Saito discusses the following syntactic asymmetries between the subject and the object in Japanese: a.

There are idioms consisting of a transitive verb + object, but not verb + subject,

b.

The semantic role of the subject often depends on the choice of the object, but the semantic role of the object depends on the lexical properties of the verb.

c.

Pronominal co-reference reveals structural asymmetry.

Regarding Hungarian, Kiss (2003) states that scrambling exists only in the post-verbal position in this language. She further argues that there are two domains in the preverbal position: an operator domain (consisting of quantifier, focus, and aspect positions) and a topic domain. The following configuration shows the specific positions discussed by Kiss (2003). TopP represents the topic phrase, DistP stands for the quantifier phrase, FP exhibits the focus phrase, and AspP represents the aspect phrase.

36

Chapter 2: Literature on Scrambling

(4) T o p XP P X P

TopP TopP DistP XP DistP

T o p P X P D i

XP FP XP AspP XP VP XP V

XP XP (Kiss 2003: 23)

In the post-verbal position, the order is free, and has no effect on binding relations. Kiss (1994), therefore, suggests that post-verbal elements are base-generated randomly in a flat word order in Hungarian. She concludes that

(5)

Scrambling may not effect the semantic interpretation of the sentence21. (Kiss 1994: 228)

Saito (1985 and in subsequent work) agrees with the statement in (5), although he considers scrambling as the result of syntactic movement. He suggests that the moved element is reconstructed into its original position at LF (see section 2.2. where Saito’s recent views are discussed). Assuming a configurational status for Dutch and German, Riemsdijk (1989) suggests that free word order in these languages arises from the base-generation of hierarchical structures in which the arguments and adjuncts are arbitrarily distributed. Neeleman (1994) makes a distinction between ordinary scrambling and focus in Dutch, and states that only the latter is the result of movement. Bayer and Kornfilt (1994) argue along the same lines. They suggest that non-focal scrambling is not the result of movement, but rather is base generated in its overt position. They further suggest that Long Distance Scrambling (LDS) is possible out of an infinitival zu/te complement clause (third construction) in German and Dutch

Base-generation approach

37

only if the matrix verb belongs to a limited lexical class of control verbs, as in (6). LDS is not possible out of a tensed complement clause altogether, as in (7)22. (6)

weil Heinrich den Wageni versprochen hat [PRO ei zu washen] because H. the car promised has to wash ‘Because Heinrich has promised to wash the car.’

(7)

*weil Heinrich [die Zebras]i gesagt hat [ei [dass [er dem Kind ei because H the zebra said has that he the child-Dat zeigen wird]]] show will ‘Because Heinrich has said that he will show the Zebras to the child.’ (Bayer and Kornfilt 1994: 30)

Bayer and Kornfilt suggest that LDS is allowed only in infinitive clauses in German since these types of clauses are IPs, rather than CPs23. Their arguments rest on the assumption that in German, unlike English, I(NFL) is the sister of V at the morphological structure, and therefore, V and I are jointly visible at the mother node24. This complex verbal projection, they suggest, continues beyond IP, since IP is a defective category and not an inherent barrier to movement. Thus, the Accusative feature of the embedded verb in (6) can be assigned to den Wagen, the object of the embedded clause that is base-generated in the higher clause. This process is blocked by the embedded C in (7) since this element lacks the raising property observed in I. In summary, base-generation theories consider languages with free word order as either having a flat or a hierarchical structure. The majority of authors who express the second view consider the D-structure identical to the S-structure, except for those cases that receive a focus interpretation. Focus is considered to be the result of movement. Thus they make a distinction between scrambling and focus.

2.2. Base-generation approach: MP views Analyzing scrambling in Japanese, Bošković and Takahashi (1998) take a base-generation approach, suggesting that “… scrambled elements are directly base-generated in their surface positions and undergo LF movement (lowering in most cases) to the positions where they receive theta

38

Chapter 2: Literature on Scrambling

roles.” Theta roles, they suggest, are formal features, and therefore capable of deriving movement. Their analysis is based on the assumption that scrambling is semantically vacuous. They suggest that in those cases where scrambling creates scope ambiguity, the verb moves to I at LF, and can theta mark the scrambled element. Consider the example in (8). (8)

Daremoi -ni dareka-ga ti atta everyone-Dat someone-Nom met ‘Everyone, someone met.’

According to Bošković and Takahashi, the scrambled quantifier phrase (QP) in (8) can take scope over the subject since the verb moves to I at LF, and can theta mark the object from that position. In contrast, the scrambled QP cannot take scope over the matrix subject in (9) since V to I is not possible across clause boundaries. (9)

Daremoi -ni dareka-ga [Mary-ga ti atta to] omotteiru everyone-Dat someone-Nom Mary-Nom met that thinks = for some x, x a person, x thinks that for every y, y a person, Mary met y. # for every y, y a person, there is some x, x a person, such that x thinks that Mary met y.

Saito and Fukui (1998) provide a similar proposal by suggesting that scrambling is an optional operation, and like any such operations should be considered a special case of Merge. They further maintain that scrambling in Japanese is semantically vacuous, and is undone at LF. They suggest that this type of Merge is possible only if it is compatible with the head parameter property of the language. Thus, they provide (10), incorporating the Head Parameter into the Merge operation. (10) K = {γ, (α,β)}, where γ ε {α,β} a. γ = α: head-initial, left-headed b. γ = β: head-final, right headed (Saito & Fukui 1998: 439) On the basis of (10), Merge produces the following configurations in English and Japanese:

Base-generation approach

(11) English: Japanese:

39

X’ = X/X’ X’’ X’ = X’’ X/X’

According to these authors, Heavy NP Shift in English is also generated by Merge in accordance with (10) and (11). Saito and Fukui’s proposal would be attractive if scrambling were a truly optional movement. The discussions in chapters 4 and 5 of this work indicates that this is not the case, at least in Persian. We will also see pieces of counter evidence with respect to Bošković and Takahashi’s theory of scrambling in 2.3.2.

2.3. How does Persian fit in? We now turn to Persian. This language is configurational since the c-command relation of the phrasal elements has an impact on binding and scope relations (see chapter 5). The question to address in this section is whether scrambling is the result of Merge or Move. That is, is there evidence indicating that scrambling in Persian is the result of Merge. 2.3.1. Empirical evidence The following example suggests that the separation of the quantifier and the head noun is best analyzed by taking a base-generation approach. (12) a. Kimea be bachche-hâ goft ke pro [hame gilâs-â-ro] K to child-pl said that all cherry-pl-râ bo-xor-an subj-eat-3pl ‘Kimea told the children to eat all the cherries.’ b. [gilâs-â ro]i Kimea be bachche-hâ goft ke pro [hama ti -ro] bo-xor-an c. [gilâs-â-ro]i Kimea [hama ti -ro]k be bachche-hâ goft ke pro tk bo-xor-an Persian allows the interaction of LDS and Floating Quantifiers (FQ). In (12b), the quantifier is stranded in its base-position. The sentence in (12c) illustrates an instance of LDS, followed by another movement whereby the quantifier is stranded in intermediate position.

40

Chapter 2: Literature on Scrambling

At first glance, the combination of *hama-ro gilâs-â-ro cannot be a base form since râ takes scope over the entire DP (Karimi 2001). Thus, the appearance of gilâs-â-ro in the initial position of the matrix clause in (12b) and (12c) does not seem to be the result of a syntactic movement. The ungrammaticality of this string is reminiscent of the following German example, borrowed from Bayer and Kornfilt (1994). (13) a. dass die Männer im Gasthaus beid-e Bier bestellen. that the Men-Nom in-the pub both-Nom beer order b. dass sie den Männer-n gestern beid-en den Ausgang zeigte that she the men-Dat yesterday both-Dat the exit showed Bayer and Kornfilt suggest that since *beide die Männer and *beiden den Männern are ill-formed quantifier phrases that cannot be taken as base sources, the scrambled DPs are in fact base-generated in the higher clause in these examples. Given this line of reasoning, the Persian examples in (12b) and (12c) seem to support the base-generation approach. 2.3.2. Problems Persian provides a number of problems for a theory that proposes a basegeneration approach to scrambling. First, we will see in chapter 4 that scrambling in Persian, local or long distance, represents either focus or topic, as we already saw in some cases in chapter 1. Additional examples are provided in (14). A scrambled element may be interpreted as topic or focus based on their stress. (14) a. Kimea goft [ke Rahjue ketâb-â-ro az Parviz xaride] K said that R book-pl râ from P bought is ‘Kimea said that Rahjue had bought the books from Parviz.’ b. [ketâb-â-ro]i Kimea goft [ke Rahjue ti az Parviz xaride] ‘As for the books, Kimea said that Rahjue has bought (them) from Parviz.’ or ‘It was the BOOKS that Kimea said that Rahjue had bought (them) from Parviz.’ c. [az Parviz]i Kimea goft [ke Rahjue ketâb-â ro ti xaride] ‘From Parviz, Kimea said that Rahjue had bought the books.’ ‘It was from PARVIZ that Kimea said that Rahjue had bought the books.’

Base-generation approach

41

The sentence in (14a) represents an unmarked word order. In (14b) and (14c), the object and the PP, respectively, have scrambled to the front of the matrix clause. Second, the trace of a scrambled quantifier plays a role in the scope interpretation of that quantifier. The following examples provide evidence. (15) a. har dâneshjui bâyad ye mas’ala-ro hal bo-kon-e each student must a problem-râ solution subj-do-3sg ‘Each student must solve a problem.’ ∀ > ∃ ; *∃ > ∀ b. [ye mas’ala-ro]i har dâneshjui bâyad ti hal-bo-kon-e ∃>∀;∀>∃ The sentence in (15a) receives a distributive reading while the one in (15b) is ambiguous between a collective and a distributive reading (although the collective interpretation is the primary one in this case). That is, for each student in (15a) there is a different problem that he or she has to solve. The primary reading of (15b) reveals that there is one single problem so that all students have to solve it. The ambiguous reading of (15b) clearly suggests that the copy of the scrambled element plays a role in the interpretation of this sentence. The examples in (15) provide counter evidence to the theory advanced by Saito and Fukui (1998), which considers scrambling a sub-case of Merge that is undone at LF. Furthermore, the data in (16) contradict Bošković and Takahashi’s account of scrambling. (16) a. har dâneshju-i fekr mi-kon-e [CP Kimea ye pesar-i-ro every student-ind thought dur-do-3sg K a boy-ind-râ dust dâr-e] ∀ > ∃ ; *∃ > ∀ friend have-3sg ‘Every student thinks that Kimea loves one boy.’ b. ye pesar-i-roi har dâneshju-i fekr-mi-kon-e [CP Kimea ti dust-dâr-e ] ∃>∀;∀>∃ Recall that Bošković and Takahashi claim that the ambiguity of quantifiers is available only within a simple clause due to V to I raising. Thus, the sentence in (16b), an instance of ambiguity created by LDS, contradicts their claim. Furthermore, their theory is based on the assumption that scrambled elements return to their argument position at LF to check their theta roles, implying that only arguments are subject to scrambling. The

42

Chapter 2: Literature on Scrambling

sentences in (17), however, show that adjuncts may undergo LDS, and create ambiguity as well. (17) a. cherâi fekr mi-kon-i [CP Kimea emruz bargasht ti]? why think pres-do-2sg K today returned ‘Why do you think Kimea returned today?’ b. keyi Kimea goft [CP ke Sepide xune xaride ti]? when K said that S house bought is ‘When did Kimea say that Sepide has bought a home?’ The adjuncts in (17) can be interpreted either in the matrix clause or in the embedded clause. A theory that is based on LF lowering of elements motivated by theta features, therefore, cannot account for these data25. Note that Kornfilt’s and Bayer’s theory of V to I raising fails to work as well in the case of LDS examples since C would block the movement. The third argument against a base-generation approach to scrambling comes from two sets of ill-formed data exhibiting scrambling out of different types of islands. It is well-known that movement out of syntactic islands is blocked (Ross 1968, Chomsky 1977, and work thereafter). If scrambling were the result of Merge, the ungrammaticality of these data would remain unexplained. The first set of data shows scrambling out of complex DP constructions.

(18) *Rahjuei [in vâghe’iyyat [CP ke ti bigonâh-e]]

bar hame rowshan-e

R this fact that innocent-3sg at all clear-3sg Intended meaning: ‘As for Rahjue, the fact that (he) is innocent is obvious to all.’ (19) *[ruy-e tappe]i Kimea [DP xune-i-ro [CP ke ti sâxte bud]] on -Ez hill K house-rel-râ that built was emruz foruxt today sold Intended meaning: ‘On the hill, Kimea sold the house that was built (on it).’ (20) *Kimeai [DP xânom-i [CP ke ruberu-ye mo’allem neshaste bud K

woman-rel that

infront-Ez teacher

sitting

was

Base-generation approach

bâ [DP mard-i] [CP ke ti mi-shenâxt ] ezdevâj

43

kard

with man-rel that dur-knew marriage did Intended meaning: ‘As for Kimea, the woman who was sitting in front of the teacher married the man (she = Kimea) knew.’ (21) *[bâ mardom]i man [DP se tâ

amrikâ’i -ro] tu xiyâbun

with people I three part American râ in street did-am [CP ke ti fârsi harf mi-zad-an] saw-1sg that Persian speech hab-hit-3sg Intended meaning: ‘With people, I saw three Americans who were talking (with them).’ The sentence in (18) represents scrambling out of the complement clause of a noun, while those in (19)–(21) exemplify scrambling out of relative clauses. All these sentences are sharply ungrammatical. The second set of examples illustrates scrambling out of adjuncts, as in (22b,c) and (23b,c). (22) a. [pro be dust-am telefon kard-am] [CPghablaz inke bachcha-ro to friend-my telefon did-1sg before that child -râ be-xâb-un-am] subj-sleep-cause-1sg ‘I called my friend before I put the child to sleep.’ b. * [pro be dust-am telefon kard-am] bachcha -roi [CP ghablaz inke ti be-xâb-un-am] c. *bachcha - roi [pro be dust-am telefon kard-am] [CP ghablaz inke ti be-xâb-un-am] (23) a. pro diruz be madrese raft-am [CP chonke bâ Arezu yesterday to school went-1sg because with A gharâr dâsht-am] appointment had-1sg ‘I went to school yesterday because I had an appointment with Arezu.’

44

Chapter 2: Literature on Scrambling

b. *diruz be madrese raft-am [bâ Arezu]i [CP chonke ti gharâr dâsht-am] c. *[bâ Arezu]i diruz be madrese raft-am [CP chonke ti gharâr dâsht-am] Note that (23b&c) are grammatical if the PP bâ Arezu ‘with Arezu’ is interpreted as part of the matrix clause: I went with Arezu to school yesterday, because I had an appointment. Again, the ungrammaticality of these examples remains unexplained given a base-generation approach. We saw that FQ is in general allowed in Persian, as in (12), repeated below in (24). (24) a. Kimea be bachche-hâ goft ke pro [hame gilâs-â-ro] K to child-pl said that all cherry-pl-râ bo-xor-an subj-eat-3pl ‘Kimea told the children to eat all the cherries.’ b. [gilâs-â ro]i Kimea be bachche-hâ goft ke pro [hama ti -ro] bo-xor-an c. [gilâs-â-ro]i Kimea [hama ti -ro]k be bachche-hâ goft ke pro tk bo-xor-an Neverthelss, the example in (25b) shows that FQ is not possible across an island. (25) a. Kimea bâ bachche-hâ da’vâ kard [CP chonke unâ hame K with child-pl quarrel did because they all sib-â-ro xorde bud-an] apple-pl-râ eaten were-3sg ‘Kimea quarreled with the children since they had eaten all apples.’ b. *[sib-â ro]i kimea bâ bachche-hâ da’vâ kard [CP chonke una hama-ro ti xorde bud-an] If scrambling were the result of base-generation, the ungrammaticality of these sentences would remain unexplained.

Scrambling as syntactic movement

45

One question remains to be explained: if scrambling is not a base-generation phenomenon, how can we account for the sentences in (12b) and (12c), repeated in (24b) and (24c), where *hama-ro gilâs-â-ro is not a source form, as mentioned before. This problem is easily solved if we assume, following the basic ideas of Distributed Morphology, that phonological insertion of lexical units applies post-syntactically (see section 4.2. in chapter 1): since the phrase hame sib-â ‘all the apples’ carries the features [Specific] and [Accusative], râ, representing Accusative Case for specific DPs shows up on both items. That is, once the Morphosyntactic description is turned into the Morphophonological description of the lexical items post-syntactically, râ shows up on the scrambled noun as well as the quantifier in-situ.

3. Scrambling as syntactic movement From a movement point of view, scrambling has been considered to be Amovement (Mahajan 1990, 1994b with respect to clause bound scrambling; Déprez 1994), or A’-movement (Mahajan 1990, 1994b for clause initial scrambling; Saito 1985 and in subsequent work up to 1998, among others). Clause-bound scrambling has been argued to be associated with the EPP in recent years (Miyagawa 1997, 2001, 2003 for Japanese, Bailyn 1999, 2003, and Lavine 1998 for Russian). Long Distance Scrambling, on the other hand, has been classified as an instance of A’-movement. Finally, based on German data, Webelhuth (1992) suggests a mixed landing site for scrambled elements, arguing that this position reveals both A and A’ properties. In this section, we address all three approaches. In 3.1., we discuss scrambling as an instance of A-movement. A brief analysis of clause-bound scrambling and its association with the EPP is included in this section. Scrambling as an A’-movement and problems related to this type of approach are discussed in section 3.2. This analysis is followed in section 3.3. by an examination of Webelhuth’s (1992) dichotomy regarding the mixed properties of the landing site of scrambled elements, and the problems that arise from this kind of analysis.

3.1. The A-movement approach A(rgument)-movement typically moves a DP from a [+Theta], [-Case] position to a [-Theta], [+Case] position. Thus, this movement is triggered

46

Chapter 2: Literature on Scrambling

by a morphological feature associated with Case. Passive and raising constructions are typical examples of A-movement. Passive: (26) Johni was arrested ti by the police last night. Raising: (27) Johni appears [ti to be very happy these days] A(rgument) positions are considered to have distinct syntactic properties as follows: First, movement to an A–position is subject to locality, thus A-movement is clause bound. Second, Weak Crossover (WCO) can be overridden if the movement is into an A–position. Third, binding is considered to be a property of A-positions. Fourth, reconstruction is considered a property of non-argument (A’-positions), and therefore, it is not possible from an A-position. That is, the copy of an A-moved element is not visible for semantic interpretation. This assumption has some consequences for binding as well as scope reading, as we see below. The syntactic properties of A-movement have been employed as diagnostic tools to examine the nature of scrambling. Based on these tests, local scrambling has been considered to be A-movement by a number of authors (Déprez 1994; Mahajan 1990, 1994b; Fanselow 1990; Santorini 1991, among others). These authors suggest that the landing site of the scrambled element is the Spec of a functional head, and the movement is Case-driven. Several reasons behind this assumption are discussed below. 3.1.1. Clause-boundedness Locality is one of the properties of A-movement, as mentioned before. We saw that scrambling is clause-bound, and therefore local, in German. The sentence in (28) is ill-formed since movement is not local. (28) *… weil Hans das Buchi versprochen hat [dass er ti kaufen wurde] because Hans the book promised has that he buy would Based on examples like (28), scrambling is considered an instance of Amovement.

Scrambling as syntactic movement

47

3.1.2. Binding and reconstruction Binding is another diagnostic tool employed to examine the nature of scrambling. The principles of the binding theory are stated in (29). (29) Binding Principles: Principle A: if α is an anaphor, interpret it as co-referential with a c-commanding phrase in D. Principle B: if α is a pronoun, interpret it as disjoint from every ccommanding phrase in D. Principle C: if α is an R-expression, interpret it as disjoint from every c-commanding phrase. Where D = the minimal domain. The minimal domain is defined in terms of Complete Functional Complex (CFC), as stated in (30): (30) The Domain of α is its minimal CFC, a projection that contains all grammatical functions compatible with its head. The copy of an A-moved element is subject to Principle A of the binding theory. That is, it must be bound by its antecedent within its local binding domain. The ill-formedness of (28) is due to the fact that the copy is not locally bound by its antecedent, thus violating Principle A of the binding theory. The same problem holds for the example in (31): the copy is not locally bound by its antecedent, the DP John, violating Principle A. (31) *Johni seems [CP that it appears [CP ti to be happy]] Scrambled elements are argued to create new binding relations, as in (32). (32) … weil

wir die Fraueni einanderi ti vorgestellt haben

because we the women each other introduced have ‘Because we have introduced the women to each other.’ If the object die Frauen were in an A’-position, it could not bind the anaphor, and the sentence would be ruled out by Principle A of the binding theory. The grammaticality of this sentence implies that the object must be in an A-position (see chapter 5 for an alternative analysis).

48

Chapter 2: Literature on Scrambling

Furthermore, reconstruction is suggested to be possible only from an A’ position (Mahajan (1990), among others). If scrambling is in fact Amovement, reconstruction cannot apply from the landing site of the scrambled element at LF. Consider the example in (32) once more. Scrambling feeds binding in this example, which in turn suggests that the scrambled DP is not reconstructed at LF, and thus must c-command the reciprocal from an A-position at that level. The following example, taken from Mahajan (1990), supports the same idea. (33) */???raam-ne ek duusre ki kitaabeNi baccoN-ko ti de dii Ram-subject each other’s books-DO children-IO give+past According to Mahajan, the direct object has scrambled into the Spec of AGRo in (33).26 This, is a position where the object receives Accusative Case. Thus it is an A-position, and therefore, reconstruction at LF cannot apply. This line of reasoning is employed to account for the ill-formedness of the sentence in (33) and similar cases where the anaphor remains unbound (see chapter 5 for an alternative analysis). 3.1.3. WCO Weak Crossover (WCO) is yet another test that has been employed to distinguish between an A-movement and an A’-movement. The WCO effect is obtained when an element in A’-position c-commands a bound variable inside a DP and its own trace at the same time at LF27. Thus a whtrace (A’-trace) triggers WCO-effects, while an NP-trace (A-trace) does not, as attested by the following examples28. (34) *Whoi does it seem to hisi father [that John loves ti ] A’-movement (35) Whoi ti seems to hisi father [ti to be happy] A-movement The wh-phrase c-commands the pronoun from an A’-position in (34), and is coindexed with it, so the pronoun is a bound variable in this case. The operator is therefore coindexed with a bound variable and its own trace at the same time, creating a WCO effect. In (35), the wh-phrase moves

Scrambling as syntactic movement

49

into the Spec of TP, an A-position, before moving into the Spec of CP, an A’-position. Thus, the pronominal in this example is not a bound variable, since it is bound by an element (the higher trace) in an A-position. It has been argued in the literature that clause-bound scrambling does not trigger WCO effects, and thus has to be considered A-movement. The following data illustrate this assumption. (36) weil Maria [jeden Gast]i [ohne seinemi Partner e vorzustellen] because M every guest without his partner to introduce allein ti lässt. alone leaves ‘Because Maria leaves each guest alone without introducing (them) to his partner.’ (37) weil Maria [jede Frau]i [ohne ihremi Partner e vorzustellen] because M every woman without her partner to introduce allein ti lässt alone leaves ‘Because Maria leaves each woman alone without introducing (them) to her partner.’ (Déprez 1994: 128) The argument is that the noun phrases jeden Gast and jede Frau must ccommand the pronoun from an A position in (36) and (37), since otherwise these sentences would be ungrammatical. In other words, (36) and (37) reveal an Anti-WCO effect, a property attributed to A-movements. I will come back to these data in chapter 7. 3.1.4. Clause bound scrambling and EPP In recent years, there has been some attempt to explain clause bound scrambling in terms of A-movement triggered by EPP. This is in fact in accordance with the original idea proposed by Mahajan (1990), who suggested that clause bound scrambling can be an instance of A-movement. Note that the discussion in this section is based on earlier assumptions regarding the EPP, when this phenomenon was considered to be a D feature specific to T (Chomsky 1995), triggering movement of XP into the Spec of TP. We will see in chapter 3 that EPP is now a feature that can be optionally assigned to any (functional) head when it has an effect on the output. Considering EPP as a D feature specific to T, Holmberg and Nikanne (2002) argue that different elements, including the subject, may move into

50

Chapter 2: Literature on Scrambling

the Spec of TP to check this feature in Finnish. Bailyn (2003) discusses clause bound scrambling in connection with EPP in Russian. On the basis of Japanese, Miyagawa (1997) suggests two types of scrambling: A-scrambling, associated with some feature on T, and A’-scrambling, associated with focus. Miyagawa (2001, 2003) then suggests that the feature triggering A-scrambling is EPP. He supports this claim by resorting to scope interaction of negation and quantifier phrases. Consider the following example. (38) zen’in-ga sono tesuto-o all-Nom that test-Acc ‘All did not take that test.’

uke-nakat-ta (yo/to omou) take-Neg-Past *not > all, all > not (Miyagawa 2001: 303)

The subject in (38) can only receive a wide scope with respect to negation, and therefore, must be out of the scope of negation. Since the subject is in an A-position, its copy does not enter the scope marking. In other words, there is no reconstruction in this case. Therefore, narrow scope is not available for the subject. Now compare (38) with (39). In this example, koto has replaced the exclamation yo/to omou. (39) zen’in-ga sono tesuto-o all-Nom that test-Acc ‘All did not take that test.’

uke-nakat-ta koto take-Neg-Past not > all, (all > not) (Miyagawa 2001: 303)

When the sentence ends with koto, the subject may receive either wide or narrow scope, as in (39). The structure of(39) is the one in (40). (40) [SUB [SpecT OBJ [ …. t ….. t …]]]

The contrast between (38) and (39), both exhibiting SOV order, is that Koto in the latter allows verb movement to T. This movement is considered to expand the domain of V, making the subject and the object equidistant from the Spec of TP (Chomsky 1993). Consequently, the object can move into the Spec position of TP to satisfy the EPP without violating the Minimal Link Condition (MLC) (see chapter 1). Then the subject undergoes A’-movement for focus. Since it is in an A’-position, reconstruc-

Scrambling as syntactic movement

51

tion is possible, and thus its copy may enter the process of scope identification29. Given Miyagawa’s reasoning, the subject in (41) must be in the Spec of TP, satisfying the EPP, while the object occupies an A’-position, representing focus. Thus the object can be reconstructed30. (41) Zibunzisini -o [Hanako-ga ti hihansita] (koto) self -Acc H -Nom criticized fact ‘Herself, Hanako criticized.’ Following the traditional assumption regarding LDS, Miyagawa states that this type of scrambling is an instance of A’-movement, and thus cannot be triggered by EPP. The ill-formedness of the following sentence, borrowed from Saito (1992), is thus explained since the scrambled DP will be reconstructed at LF, leaving the anaphor unbound. (42) *?karerai-o they

[Masao-ga

- Acc M

[otagaii-no sensei]-ni [CP [IP Hanako-ga ti

-Nom each other-Gen teacher-to H-Nom

hihansita ] to ] itta ] (koto) criticized Comp said fact] ‘*Them, Masao said to each otheri ’s teachers that Hanako criticized’ The sentence in (42) is thus a violation of Principle A of the binding theory, since the anaphor remains unbound in its minimal domain. (See chapter 5 for an alternative analysis.) We will see in chapter 3 that Persian lacks typical A-movement into Spec of TP. The local scrambling within the vP will be shown to occur primarily for the sake of interpretation, although Case and Agreement are also checked in that position by Agree. All movements into the discourse domain (CP), including the movement into the Spec of TP, represent discourse/operator movement, and are thus instances of A’-movement.

3.2. The A’-movement approach A’-movement is traditionally assumed to create a chain whose tail is [+Case] and its head is [-Case]. Thus, A’-movement is not motivated by

52

Chapter 2: Literature on Scrambling

Case, but rather by other kinds of features such as a wh-feature. A’-positions reveal specific syntactic properties as follows: First, parasitic gaps are licensed by elements in A’-positions. Second, copies left by A’-movement are variables, subject to Principle C of the binding theory. Third, copies left by A’-movement enter semantic interpretation at LF. That is, A’-movement is subject to reconstruction (Mahajan 1990, adopted by Chomsky thereafter). Fourth, A’-movement does not allow FQ. Based on the specific properties of A’-movement when employed as diagnostic tests, scrambling has been considered A’-movement by a number of authors (Mahajan 1990, 1994b with respect to LDS, and sometimes also within a clause; Saito 1985 and subsequent work up to 1998; Dayal 1994; Müller and Sternefeld 1994; Vikner 1994; Miyagawa 1997 with respect to focus, among others). In 3.2.1. and 3.2.2., we review two cases that have lead these authors to adopt the A’-approach. Counter evidence to a theory that considers scrambling an A’-movement is provided in 3.2.3. 3.2.1. Parasitic Gaps As mentioned before, a few syntactic tools have been utilized in the literature to distinguish between A and A’ movements, and to understand the nature of scrambling. A parasitic gap is one of those tools. Chomsky (1982) states that a parasitic gap is a variable directly bound by an element outside the adjunct containing it. As variables must be A’bound, only A’-movement would generate the antecedent for a parasitic gap. Chomsky (1986a) suggests, however, that parasitic gaps are not directly bound from outside the adjunct, but rather they are generated by the movement of an empty operator from the position of the parasitic gap to the initial position of the adjunct. Therefore, it is the chain headed by this operator that is identified with a DP outside the clause. Consequently, the adjunct in (43a) would have the structure in (43b): (43) a. What did you file [CP before reading]? b. [CP Oi [before [reading ei]] Since the operator is in an A’-position, there is no need for the antecedent of the gap to be in an A’-position. This claim does not seem to hold, however, given the following example:

Scrambling as syntactic movement

53

(44) *[Those stories]i seem [ti to have been filed ti [Ok without reading ek]] The ungrammaticality of (44) shows that parasitic gaps cannot be identified with elements in A-positions. It has been argued in the literature that scrambling licenses parasitic gaps (Webelhuth 1992, Vikner 1994, among others), and therefore, it must be A’-movement. Consider the following example. (45) … weil er den Patienteni [ohne PRO vorher ei zu untersuchen] ti because he the patient without first to examine operierte operated ‘Because he operated on the patient without first to examine (him).’ The scrambled element den Patienten licenses the gap (illustrated as ‘e’), and therefore, must be in an A’-position. The following Persian example reveals the same pattern: the scrambled wh-phrase licenses the gap. (46) kodum ketâb-roi Kimea fekr mi-kon-e Rahjue [bedune-inke which book-râ K think dur-do-3sg R without that pro ei be-xun-e] ti be ketâbxune pas-dâd subj-read-3sg to library returned ‘Which book does Kimea think Rahjue returned to the library without reading.’ Since the wh-word licenses the gap within the adjunct clause in (46), it must be in an A’-position. The following example represents the case of a scrambled object within a simple clause: (47) Kimea [bachche-hâ-ro]i [qhablaz-inke [pro e i be kelâs be-frest-e]] K child-pl-râ before that to class subj-send-3sg be hamdige t i mo’arrefi kard to each other introduction did ‘Kimea introduced the children to each other before sending (them) to class.’ Again, the scrambled object in (47) licenses the gap, and thus must be in an A’-position.

54

Chapter 2: Literature on Scrambling

3.2.2. Binding Binding relations have led some authors to claim that LDS is an instance of A’-movement (Mahajan 1990; Saito 1989, 1992; among others) since this type of movement does not allow binding from the landing site of the moved element. Consider the sentence in (42), repeated below in (48). [Masao-ga [otagaii -no sensei] -ni [CP [IP (48) *Karerai -o they -Acc M -Nom each other-Gen teacher-to itta] (koto) Hanako -ga ti hihansita] to] H - Nom criticized Comp said fact ‘*Themi, Masao said to each otheri’s teachers that Hanako criticized’ The argument is that the scrambled element must have been reconstructed in (48), leaving the anaphor unbound. Thus, reconstruction is considered to be responsible for the ill-formedness of this sentence. Kim (1992) makes a distinction between clause-internal and clause-initial scrambling. He suggests that clause-internal scrambling is an instance of A-movement, while clause-initial scrambling is A’-movement. (49) kutul-uli Yonghi-nun [sero-uyi kyoswu-eke] sokayhayssta they-Acc Y-Nom each other-Gen professor-Dat introduced ‘Them, Yonghi introduced to each other’s professors.’ (Kim 1992: 159) Kim, who allows the direct object to be generated in a position adjacent to the verb, must assume an intermediate Case position for this element in order to account for the well-formedness of the sentence in (49). That is, the object must have moved into an A-position (Spec of AGRo), and then further up into the initial position of the sentence. The first movement is Case-driven, and thus an A-movement which does not allow reconstruction. The second movement is an A’-movement. This is illustrated by (50). (50) kutul-uli Yonghi-nun [AGRoP ti ] [sero-uyi kyoswu-eke ] ti sokayhayssta

Scrambling as syntactic movement

55

Thus, the object moves back into the Spec of AGRo, c-commanding and binding the anaphoric indirect object from that position. Similarly, the well-formedness of the sentence in (51) is accounted for if the clause initial scrambling is considered an A’-movement. (51) sero-uyi kyoswu-lul, kutul-ii Yonghi-eke sokayhassta each other-Gen professor-Acc they-Nom Y - Dat introduced ‘Each other’s professors, they introduced to Yongi.’ (Kim 1992: 161) The direct object is reconstructed into the intermediate position at LF, and thus is properly bound by the c-commanding subject at that level. Another example of this type is provided by (52), taken from Hindi. In this sentence, the reflexive can be bound by the subject, but not by the indirect object. Mahajan justifies the lack of a binding relation between the indirect object and the reflexive by suggesting that reconstruction stops in the intermediate position (Spec of AGRo] which is the result of an Amovement. The second movement is A’-movement which allows reconstruction. Therefore, the reflexive can be bound only by the subject. (52) [apniii/*j kitaab ]k raami-ne t’k mohanj -ko tk dii self’s

book

Ram-Erg

Mohan-Dat

gave

We will come back to these data in chapters 5 and 7. We will see that all these data can be accounted for by taking into consideration the model introduced in chapter 3. 3.2.3. Problems Dayal (1994) points out that scrambling is an atypical A’-movement. Her argument is based on the interaction of Floating Quantifiers (FQ) with scrambling in Hindi. FQ seems to manifest certain properties31. First, it must be bound by a c-commanding antecedent within its local domain. (53) a. *A picture of the children has all been torn. b. *The children believe that Mary will all come. (Déprez 1994: 104)

56

Chapter 2: Literature on Scrambling

The ill-formedness of (53a) is due to the fact that the floating quantifier is not c-commanded by an antecedent. The sentence in (53b) is ungrammatical since the quantifier is not bound by its antecedent in its local domain (the lower TP). Second, A-movement allows FQ while A’-movement does not. (54) a. [The drug dealers]i have all / each ti been arrested. b. *[These drug dealers]i, the major said that the police will all ti arrest.

(Déprez 1994: 104)

(55) a. [The boys]i appear [all ti to have left]. b. *[The children]i, whoi I will have [all ti ] met before the end of this week,…

(Dayal 1994: 257)

If LDS is a typical A’-movement, it should not allow FQ. However, data from Persian indicate that this prediction is not borne out, as we saw in (12) and (24), repeated below in (56). (56) a. Kimea be bachche-hâ goft ke pro [hame gilâs-â - ro] K to child-pl said that all cherry-pl-râ bo-xor-an subj-eat-3pl ‘Kimea told the children to eat the cherries all up.’ b. [gilâs-â ro]i Kimea be bachche-hâ goft ke pro [hama ti -ro] bo-xor-an c. [gilâs-â-ro]i Kimea [hama ti -ro]k be bachche-hâ goft ke pro tk bo-xor-an In (56b), the quantifier is stranded in its base-position by LDS. The sentence in (56c) illustrates an instance of LDS and then quantifier stranding in the derived, intermediate position. Thus, these examples show that scrambling cannot be considered typical A’-movement. There is a more basic problem with the analyses that suggest that scrambling is A’-movement. That is, all these theories rely on the assumption that only A’-movement is subject to reconstruction. This claim, how-

Scrambling as syntactic movement

57

ever, is controversial. May (1977, 1985) shows that subjects of raising predicates such as seem can be interpreted as if they have been lowered at LF. Thus the example in (57), and similar cases, indicate that reconstruction is possible from an A-position as well. (57) A unicorni seems [TP ti to be in the garden]

The sentence in (57) has a reading that does not presuppose the existence of unicorns. This reading requires that the moved DP be reconstructed into the lower TP. The following example is another piece of evidence indicating that reconstruction is possible from an A-position. (58) [Each otheri’s husbands]k seemed to the womeni [IP tk to be awfully nice] Furthermore, psyche-verbs seem to allow reconstruction at LF. Belletti and Rizzi (1988) suggest that the D-structure configuration of a psycheverb like ‘worry’ is the one in (59). (59)

VP V' V

NP NP

worry THEME EXPERIENCER (Belletti and Rizzi 1988) According to these authors, the THEME in these constructions moves to the Spec of TP, and receives Nom Case in that position. This movement is, therefore, an A-movement. However, reconstruction seems to be possible from that position, as argued by Barss (1986, 1996)32. Consider the following example. (60) [Pictures of themselvesi ] annoy/?*hit the meni.

58

Chapter 2: Literature on Scrambling

The following example combines a psyche-verb and a raising verb. (61) [Each otheri’s parents]k seem [IP tk to have annoyed tk the womeni ] The sentences in (57), (58), (60), and (61) are pieces of counter-evidence to the claim that A-movement does not allow reconstruction33.

3.3. Non-argument, non-operator approach In this section, we discuss Webelhuth’s (1992) dichotomy and problems related to his proposal. Data representing this dichotomy are revisited in chapter 5, where cross-linguistic data are discussed. 3.3.1. Webelhuth’s dichotomy There are some cases in which the landing site of a scrambled element reveals both A and A’ properties simultaneously. The Persian example in (47), repeated below in (62), represents this problem. (62) Kimea [bachche-hâ-ro]i [qhablaz-inke [pro e be kelâs be-frest-e]] K child-pl-râ before that be hamdige ti mo’arrefi kard

to class subj-send-3sg

to each other introduction did ‘Kimea introduced the children to each other before sending (them) to class.’ The direct object bachche-hâ ‘children’ licenses the parasitic gap in (62), a property of A’ movement. At the same time, it binds the reciprocal hamdige, revealing a property of A-movement. The following German example exhibits the same properties: the scrambled object binds the reciprocal einander and licenses the gap simultaneously. (63) Peter hat die Gästei [ohne ei anzuschauen] einander ti P has the guests without looking-at each other vorgestellt introduced

Scrambling as syntactic movement

59

‘Peter introduced the guests to each other without looking (at them).’ A similar situation holds for the data in (36) and (37), repeated in (64) and (64), respectively. (64) weil Maria [jeden Gastk]i [ohne seinemk Partner e vorzustellen] because M every guest without his partner to introduce allein ti lässt. alone leaves ‘Because Maria leaves each guest alone without introducing to his partner.’ (65) weil Maria [jede Frauk]i [ohne ihremk Partner e vorzustellen] because M every woman without her partner to introduce allein ti lässt. alone leaves ‘Because Maria leaves each woman alone without introducing to her partner.’ (Déprez 1994: 128) The scrambled object licenses the gap, an A’ property. At the same time, it reveals an anti-WCO property, since it is c-commanding the co-indexed pronoun in the adjunct and its own trace at the same time. This fact implies that the object must be in an A-position. These peculiarities lead Webelhuth (1992) to propose a dichotomy which suggests that the landing site of scrambled elements exhibits mixed properties. Thus, he makes a distinction between three types of positions based on three types of movements: (66) Summary of Webelhuth’s proposals a. Spec/CP A’ (Operator) position Only A’ binding b. Spec/IP A (Argument) position Only A binding c. Adjoined non-A/non-Operator position A and A’ binding Webelhuth proposes that scrambling is the third type of movement in (66).

60

Chapter 2: Literature on Scrambling

3.3.2. Problems Discussing Webelhuth’s dichotomy, Saito (1992) argues that there are two problems with this account of scrambling. First, Webelhuth’s hypothesis suggests that an element in a non-operator position can A-bind an element that it c-commands and is co-indexed with. Thus, zibunzisin ‘self’ in (41), repeated below in (67), counts as an A-binder, and therefore, this sentence should be ruled out as a violation of condition C of the binding theory. (67) Zibunzisini -o [Hanako-ga ti hihansita] (koto) self -Acc H -Nom criticized fact ‘Herself, Hanako criticized.’ We saw that Miyagawa’s theory of scrambling, motivated by EPP and focus, accounts for the grammaticality of this sentence. That is, the object is focused in (67), while the subject satisfies EPP. Thus, the focused element, in an A’-position, may reconstruct. As a result, Principle C is not violated. Saito further states that the second problem with Webelhuth’s proposal is that LDS is also a non-operator movement, and therefore, his proposal incorrectly predicts that the example in (42), repeated below in (68), should be as good as the one in (69): the former is an instance of long distance scrambling, and the latter is an instance of clause-initial scrambling. (68) *Karerai -o

[Masao-ga

[otagaii-no

sensei]-ni [CP [IP

they -Acc M -Nom each other-Gen teacher-to itta] (koto) Hanako -ga ti hihansita] to] H -Nom criticized Comp said fact ‘Themi, Masao said to each otheri ’s teachers that Hanako criticized.’ (69) Karerai-o

[Masao-ga

[[otagaii-no

sensei]-ni ti

they -Acc M -Nom each other-Gen teacher-to syookaisita]] (koto) introduced fact ‘Themi, Masao introduced to each otheri’s teacher.’ In other words, Saito’s criticism is based on the argument that if a position has two properties simultaneously, the sentences in (68) and (69) should

Why does Lx , but not Ly , allow scrambling?

61

be able to take advantage of this dichotomy, and thus we should be able to obtain the same result in both sentences. Based on a revised version of Webelhuth’s hypothesis and Tada’s (1990) suggestions, Saito (1992) provides his own solutions to the problems discussed in this section. He proposes that the position of a scrambled element can be interpreted as a non-operator position at S-structure (in the sense of Webelhuth). When it arrives at LF, it can no longer be a non-operator position, and must be interpreted either as an operator position or an A-position. Otherwise, it disappears. In (68), this position is interpreted as an operator position at LF, subject to reconstruction, while the one in (69) is interpreted as an Aposition, thus interpreted in its surface position. In summary, considering scrambling as an A-movement, an A’-movement, or a non-A/non-operator movement faces some problems. We come back to these issues in chapters 5, where the cross-linguistic data are discussed.

4. Why does Lx, but not Ly, allow scrambling? The peculiarities of scrambling have led authors to raise questions about the nature of languages that allow this phenomenon: why is scrambling allowed in languages such as Persian, German, Hindi, Japanese, and Korean, but not in English and French? Is this movement related to specific properties of these languages? In this section, we take a look at the important factors proposed in the literature to be responsible for the differences between scrambling and non-scrambling languages. In 4.1., scrambling with respect to the position of the verb is discussed. Language typology in connection with possible adjunction sites is examined in 4.2. Bošković and Takahashi’s (1998) proposal is reviewed in 4.3. Finally, recent findings regarding scrambling and the factors that motivate it are briefly discussed in 4.4. 4.1. Verb position One interesting fact about scrambling is that it seems to be related to the position of the verb, since the majority of scrambling languages are verb final. For example, of all Germanic languages, four are of SOV type (Dutch, German, Frisian, and Old English), and they all exhibit scrambling (Neeleman 1994). Those languages that lack scrambling (English,

62

Chapter 2: Literature on Scrambling

Danish, Norwegian, Swedish, and Icelandic) are of SVO type. Neeleman suggests that scrambling applies in a theta domain. If this domain coincides with the domain of adjuncts, scrambling is possible; otherwise, it is not. In an SVO language like English, objects are base generated to the right of the verb, whereas in SOV languages, they are base-generated to the left of the verb, in the same domain as adjuncts. Therefore, scrambling is possible in the latter, but not the former34. Fukui (1993) explains the word order differences between English and Japanese on the basis of the notion cost (Chomsky 1991) and on the parametric difference with respect to the position of the verb in these languages. One interpretation of ‘cost’ is that there is no optional rule in the grammar. That is, the application of each rule will require some ‘effort’, and hence some ‘cost’. Therefore, each movement has to be motivated by some factor. Combining the notion of ‘cost’ with parametric values, Fukui suggests that “a grammatical operation is costless as long as it creates a structure consistent with the parameter value of a given language.” This proposal has two implications. First, a grammatical operation that destroys the parameter value of a language is costly, and hence has to be motivated by some deriving force (like Case). Second, if a grammatical operation is costless, it can optionally apply. However, if the result of an operation is consistent with the parameter value of a language, some other principles might ‘force’ it. A rule that is not consistent with the parameter value, however, must have a deriving force. One possible parameter Fukui suggests is the head-parameter35. Since the verb is in the final position in Japanese, leftward movements are costless in this language. Thus, the movement of the object does not have to be motivated by a morphological feature. In contrast, rightward movement would be inconsistent with the head-parameter, unless the movement is motivated by some sort of morphological force. Since Case is assigned leftward in Japanese, and there is no Spec/Head Agreement in this language (Fukui, 1988), there is no grammatical factor to force the rightward movement. In English, an SVO language, leftward movement is not consistent with the head-parameter, and thus this type of movement has to be morphologically motivated. This language exhibits only two types of rightward movements, namely extraposition and heavy NP shift, and neither one of them is motivated by a morphological force. The reason is that these movements are consistent with the head-parameter, and therefore, are costless.

Why does Lx , but not Ly , allow scrambling?

63

Saito and Fukui (1998) argue along the same lines, as we saw in section 2.2. in this chapter. They suggest that adjunction in the traditional sense is compatible with the value of the head parameter. English is head-initial, and therefore, allows heavy NP-shift. Japanese is head-final, and thus scrambling in the pre-verbal position is allowed in this language. There are at least two problems with the proposals presented in this subsection. First, even though the majority of scrambling languages exhibit an SOV order, there are also SVO languages that allow scrambling, such as Polish (Haegeman 1995) and Russian (Müller and Sternefeld 1993, Bailyn 1995, 2003). Moreover, this hypothesis does not say anything about languages that exhibit a mixed structure with respect to the head. That is, there are SOV languages that are head-initial in phrases other than VP (German and Persian), and allow scrambling out of PP and DP, two head-initial phrases.

4.2. Adjunction sites Müller and Sternefeld (1993) provide a different type of answer to the question raised in this section. They argue that the type of adjunction that is allowed in a given language determines the existence or lack of scrambling, and the scope of this movement, in that language. English, for example, does not allow adjunction at all, thus it does not exhibit any kind of scrambling. German, Korean, and Japanese allow VP and IP adjunction, and Russian has all three types of adjunction: VP, IP, and CP. According to these authors, lack of CP adjunction in German, Korean, and Japanese explains lack of long distance scrambling of adjuncts in those languages, since the trace of the adjunct would be subject to ECP. LDS of objects is possible in Korean and Japanese since the trace of the object is properly governed by the verb. In German, they suggest, all traces have to be antecedent-governed, and consequently, there is no LDS, including LDS of objects in this language. Russian allows scrambling of arguments and adjuncts since it allows all kinds of adjunction. An analysis of scrambling based on adjunction sites faces some problems. For example, Persian should allow CP adjunction, similar to Russian, since it allows LDS of arguments and adjuncts. Such a prediction, however, will have the following problems. First, if adjunction to CP is allowed in this language, extraction out of adjunct clauses should be permissible, contrary to facts. Consider the following contrasts.

64

Chapter 2: Literature on Scrambling

(70) a. [pro be Kimea telefon kard-am] [CP chonke Rahjue to K telefon did-1sg because R ketâb-â-ro be un dâde bud] book-pl-râ to him given was ‘I called Kimea because Rahjue had given her the books.’ b. *[pro be Kimea telefon kard-am] [CP Rahjuei [CP chonke ti ketâb-â-ro be un dâde bud] (71) a. pro diruz be madrese raft-am [CP chonke bâyad yesterday to school went-1sg because must Kimea-ro mi-did-am] K-râ dur-saw-1sg ‘I went to school yesterday because I had to see Kimea.’ b. *pro diruz be madrese raft-am [CP Kimea-roi [CP chonke bâyad ti mi-did-am] Similarly, if LDS is due to CP adjunction in this language, the (b) and (c) sentences in (72) should be grammatical, a prediction that proves to be wrong on empirical grounds: dâde] (72) a. Kimea porsid [CP ke Rahjue ketâb-â-ro be ki K asked that R book-pl-râ to whom given is ‘Kimea asked whom Rahjue has given the books to.’ b. **Kimea porsid [CP be kii [CP ke Rahjue ketâb-â-ro ti dâde] c. **Kimea porsid [CP ketâb-â-roi [CP ke Rahjue ti be ki dâde] The fact that the presence of ke ‘that’ is not the reason for the ungrammaticality of (72b) and (72c) is evident in the following example. Even a subject may be extracted in the presence of ke, violating the that-trace effect. (73)

Rahjuei pro fekr R dâde]

mi-kon-am [CP ke ti ketâb-â-ro be Kimea

tought dur-do-1sg that

book-pl-râ to K

Why does Lx , but not Ly , allow scrambling?

65

given-3sg ‘As for Rahjue, I think he has given the books to Kimea.’ The embedded subject appears in the matrix clause in this sentence. The result is grammatical. The discussion in this section shows that adjunction sites cannot be parametrized to account for the type of scrambling allowed in a language.

4.3. Bošković and Takahashi Bošković and Takahashi (1998) state that two conditions are necessary for the existence of scrambling (see also subsection 2.2.). First, the language must allow an argument to be base-generated in an IP adjoined position. Second, the language must allow arguments to move back into their theta positions at LF. As for the first requirement, these authors argue that Japanese allows multiple IP adjunction, evidenced by the existence of multiple subjects (see Kuroda 1988, Fukui and Saito 1992), whereas English does not exhibit this option. Regarding the second condition, they suggest that English, unlike Japanese, requires theta positions to be filled (either by the full XP or its copy) in overt syntax, a requirement that is absent in Japanese. Persian syntax provides several problems for this analysis. First, this language does not exhibit adjunction to IP. That is, it does not have multiple subjects. Nevertheless, it allows clause-bound as well as long-distance scrambling. Second, adjuncts may scramble into a higher clause in Persian (cf. (17) in section 2.3.2.). Since these elements do not have theta features, it is not obvious what would motivate their lowering at LF (see chapter 4 for a discussion of wh-phrases including wh-adjuncts).

4.4. Recent literature: EPP and discourse Research in recent years indicates that scrambling is triggered by different motivations. Focus is one factor that causes the dislocation of phrasal elements (Miyagawa 1997; Bailyn 1999, 2001, 2003; Karimi 1999c). The EPP is another motivation that triggers movement (Holmberg 2000; Holmberg and Nikanne 2002; Miyagawa 2001, 2003, Bailyn 2003). If this line of reasoning is on the right track, then scrambling can no longer be considered one single movement. Consequently, the question asked in

66

Chapter 2: Literature on Scrambling

this section becomes irrelevant. That is, we can no longer ask why some languages exhibit scrambling and some do not. The relevant question would be why some languages exhibit overt movements that represent discourse functions such as focus and some do not. In other words, the parametric difference between the languages that allow scrambling and those that do not boils down to similar parametric differences with respect to the presence and absence of structural wh-movement.

5. Conclusion In this chapter, the literature on scrambling and its nature, as well as the properties of languages that reveal this phenomenon were surveyed. We saw that one line of thought considers scrambling a base-generation phenomenon. Within the MP model, the base-generation approach is translated as an instance of Merge under certain parametric conditions. The Persian data show, however, that the base-generation approach cannot be maintained in this language. We further reviewed the literature on scrambling from a movement point of view. A brief survey of clause-bound scrambling into Case position was presented, revealing properties of A-movement such as locality, anti-WCO effects, FQ, anaphoric binding relations, and lack of reconstruction. Furthermore, scrambling triggered by EPP was briefly discussed. An examination of scrambling in Persian to determine whether this movement is an A-movement triggered by Case or EPP is provided in chapter 3. The discussion in this chapter also suggests that scrambling reveals some properties of A’-movement, such as licensing a parasitic gap and reconstruction. However, it was shown that A’-scrambling is different from a typical operator movement since it allows FQ, a property of A-movement. Furthermore, the data examined in section 3.2.3. show that reconstruction, a property considered to be specific to A’-movement, is also possible in the case of A-movement. These issues cast doubt on the typology of movement, specifically with respect to the nature of A and A’ movement as we understand them. We will come back to these issues in the concluding chapter in this volume, where the existence of the typology of movement in terms of A-A’ distinction is questioned. We further saw that analyzing scrambling as a non-Argument / non-operator movement, as suggested by Webelhuth, creates certain problems. The data discussed in this section are reanalyzed in chapter 5 where the

Notes

67

proposals advanced in chapters 3 and 4 are extended to account for crosslinguistic data. In the final section of this chapter, the properties of scrambling languages and their differences from non-scrambling languages were briefly surveyed. If the analyses advanced in the following chapters are on the right track, movement in scrambling languages is triggered by a strong feature into the Spec of a functional head, contributing to the semantics and discourse functional interpretation of the output. If scrambling is feature-driven just like other movements, it can no longer be considered an optional syntactic operation. Thus there will be no need for a parametric difference between the syntax of scrambling and non-scrambling languages, and all differences boil down to the choice of selecting a certain type of feature from the lexicon.

Notes 20. See Miyagawa (2001) for a fresh look at Hale’s analysis in light of recent findings regarding the syntactic properties of scrambling. 21. Kiss (1994) suggests that there is only one preferred order with respect to old/new information and [+/-Human] features in the post-verbal position in this language. 22. Bierwisch (1963) was the first to notice that scrambling out of a tensed clause is not possible in German. 23. We are using IP (INFL Phrase) where ever we cite authors who have employed this term. Otherwise TP (Tense Phrase) is used. 24. Bayer and Kornfilt (1994: 36-47) argue that this assumption is based on the fact that German allows VP-internal nominatives that are observed in the case of ergative verbs. 25. See Bailyn’s (2001) reply to Bošković and Takahashi for similar criticisms. 26. Vanden Wyngaerd (1989) was the first author to suggest that objects move into the Spec of AGRo. 27. An operator is an element in an A’-position. A wh-phrase sitting in the Spec of CP, for example, is an operator. A Bound variable is a pronoun that is bound by such an operator. 28. Webelhuth (1992) and Mahajan (1990) are the authors who originally suggested that Weak Crossover and reconstruction can be utilized to test an A and A’ movement, respectively. 29. Based on this analysis, Miyagawa suggests that Hale’s (1980) insight regarding the flat nature of Japanese sentences makes sense since verb movement creates a situation in which the subject and the object seem to lose their hierarchical status. 30. It is not clear whether verb movement is obligatory when koto appears in the final position. This issue is not relevant in the case of (41), however, since the object has moved into an A’-position. 31. Sportiche (1988) correctly suggests that floating quantifier is in fact quantifier stranding, since it is not the quantifier that moves, but rather the quantified noun element, leaving behind the quantifier. However, we will continue employing the term floating quantifier in this work, as has been the tradition in the literature.

68

Chapter 2: Literature on Scrambling

32. Psyche-verb cases were first noticed by Postal in “On Raising” in 1974. The raising analysis was invented by Belletti and Rizzi (1988). 33. Note that (58), (60), and (61) neither support nor contradict lack of reconstruction with respect to A-movement if we adopt the proposal that binding relations of anaphors are established by Merge, a position taken in this work (see chapter 5). 34. Neeleman, operating within a theory that presupposes base-generation for scrambling, further argues that overt Case is responsible for the type of scrambling we observe in different languages. This argument is based on the idea that theta roles are ordered, and that this order carries over to syntax (Grimshaw 1990). Dutch, for example, does not have overt Case. Therefore, the hierarchy of theta roles must be projected via c-commanding in this language. Scrambling with respect to adjuncts is allowed in this language since it does not change the order of the theta roles. In contrast to Dutch, theta roles correspond to overt Case in German. That is, the most prominent theta role is associated with the Nom Case, and the least one to Acc Case. In this language, therefore, the surface order of theta roles can be altered, since they are linked to Case. However, when a verb assigns Acc Case to its experiencer and theme simultaneously, the order will become fixed, since now the c-command relation will have to show the hierarchy of theta role prominency. This is illustrated by the contrast between (i) and (ii). (i)

Sie lehrte ihn eine Fremdsprache. she taught him(Acc) a foreign language (Acc) ‘She taught him a foreign language.’ (ii) *Sie lehrt eine Fremdsprache ihn. Given Neeleman’s system, Japanese and Korean are accounted for. However, his analysis does not account for Persian, since the bare object in this language can scramble, although in a limited manner, as we witnessed in chapter 1 (see also chapter 4). 35. A similar proposal is made by Déprez (1994) who states that sentence internal NPmovement to a Caseless position must be licensed under directional government (Déprez 1994: 126).

Chapter 3 Local Scrambling and A-Movement

1. Introduction The goal of this chapter is to analyze local scrambling within the Persian vP, and to determine the nature of this type of movement. In order to understand the properties of local phrasal rearrangements in Persian, we first analyze different syntactic properties related to the subject determiner phrase (DP) in this language. In this regard, we are interested in determining whether Persian is a subject prominent language, or if it is instead a topic prominent language. That is, is the element that moves out of vP the most prominent DP thematically (in an active clause), the one that agrees with the verb in number and person? Or can it be any kind of phrasal element that serves as the subject of the predicate? Related to this issue is the question of the EPP (Chomsky 1981, 1982, 1995), and how this property is satisfied in Persian. Yet another issue that goes to the heart of our analysis is the existence of overt and covert expletives: is there any evidence for their existence in Persian? Finally, the question of Nom Case and subject-verb agreement becomes relevant to our discussion. Our analysis of the subject DP shows that Persian lacks raising, tough, and exceptional case marking (ECM) constructions, and that there are no overt expletives in this language. It is also shown that there is no evidence for the existence of covert expletives. Next we examine the Persian object and its base and derived positions. The syntactic and semantic properties of specific and nonspecific objects are discussed, and the motivation for object shift as well as the relationship between the specific object and Acc(usative) Case are examined. We arrive at the conclusion that there are intriguing similarities between specific subjects and objects, on the one hand, and nonspecific subjects and objects, on the other. Our analysis in this chapter indicates that Persian exhibits subject shift in addition to object shift: specific objects as well as the specific themes of (so-called) unaccusatives and passives move out of the PredP, a move-

70

Chapter 3: Local Scrambling and A-Movement

ment whose primary purpose is to serve interpretation (Chomsky 2001b). Furthermore, we see that the argument outside of the lexical phase (vP) is the subject of predication (topic). This element can be any phrase, including the grammatical subject. However, there is no need for any XP to leave vP. That is, all lexical elements are inside vP when the entire clause represents only a proposition with no discourse information unit. Two different functions of the EPP are discussed in this chapter. One is argued to function in the sense of Chomsky (1981, 1982) to satisfy the Extended Projection Principle as a grammatical property: that is, every clause has to have a subject. We will coin the term “EPPg” to refer to this type of EPP, where g stands for grammatical. It is further argued that EPPg is satisfied morphologically in Persian. Crucially, this type of EPP is not a D (nominal) feature of T (tense) (Chomsky 1995) that triggers the movement of DP into the Spec. of TP. The other type of EPP is a purely syntactic device in the sense of Chomsky (2000). The EPP in this sense is utilized as a strong feature motivating overt movement to serve the semantic and discourse functional properties of language. In other words, this type of EPP is assigned to a functional head when it has an effect on the output of the derivation. We will coin the term “EPPs” for this type of EPP, where s stands for syntactic. This is also in line with Lasnik (1999a&b, 2001), who argues that EPP allows certain heads to have specifiers. We will see that EPPs is utilized to trigger the movement of an XP to the edge of vP. It is also used to move an XP out of vP, and to place it in the Spec of a discourse-functional head such as topic or focus. This last issue is picked up and discussed in detail in chapter 4. This chapter is organized as follows. Section 2 provides an in- depth analysis of Persian subjects. Different types of constructions, including unaccusatives and passives as well as raising, subjectless, ECM and tough constructions are discussed. Moreover, the question of whether or not overt and covert expletives exist in Persian is addressed in this section. An analytical discussion of the descriptive findings discussed in the previous subsections, and their summaries, conclude this section. Section 3 provides an examination of Persian specific and nonspecific direct objects, their semantic properties, and their syntactic status within the clause. Section 4 analyzes Persian as a topic-prominent language. The role of T is discussed in section 5, followed by an examination of the nature of movements internal to vP in section 6. Concluding remarks appear in section 7.

Subject in Persian

71

2. Subject in Persian The notion “subject” is traditionally deconstructed into three classes: (1)

a. The most prominent thematic argument b. The element receiving Nom Case and agreeing with the verb. c. The subject of predication (topic)

These three classes are translated as V-domain, I-domain, and C-domain in Platzack’s (2000) system. We discuss the first two components in this section, and the third one in section 4. The Projection Principle (Chomsky 1981: 38) maintains that direct and indirect θ-marking are properties of lexical items determined by the lexicon, and requires that these properties be observed throughout the syntactic derivation. Chomsky further links the Projection Principle to the requirement that all clauses have subjects (Chomsky 1981: 40), and coins the term Extended Projection Principle to represent this requirement (Chomsky 1982: 10). Chomsky (1995), within his general system of feature checking, suggests that the EPP is a D feature of T which has to be checked by an element in the Spec of TP. It is further assumed, since the early 1980s, that the EPP is a universal property of language. Some languages allow pro to satisfy the EPP in the subject position, and some do not, depending on parametric properties of the language. The issue of universality has prompted other parametric proposals in order to satisfy the EPP requirement. At least two parameters have been suggested in the literature as presented in (2). (2)

The EPP feature of T a. triggers XP movement to the Spec of TP b. triggers V movement to T (in most Null Subject Languages)

The statement in (2a) has been proposed for a number of languages (Chomsky 1995 for English; Holmberg 2000 for Finnish and Icelandic; Miyagawa 2001 for Japanese; Bailyn 2003 and Lavine 1998 for Russian). The one in (2b) is argued to apply basically in Null Subject Languages (Alexiadou and Anagnostopoulou 1998 for Greek and Spanish; Manzini and Savola 2002 for Northern Italian dialects, among others.). In this section, we examine different types of constructions to discover the status of Persian subjects with respect to (1a) and (1b). We show that

72

Chapter 3: Local Scrambling and A-Movement

there are no overt or covert expletives in this language, and that the Spec of TP is not obligatorily filled. The EPP, in the sense of Chomsky (1981, 1982), is suggested to be satisfied by the inflectional properties of the verb. We further argue that Nom(inative) Case and agreement are checked as part of the Agree process between the verb and the specific DP in Spec of vP. This is a departure of the general assumption that Nom Case and agreement are properties of T, and are checked by this element. Unaccusatives, passives and raising constructions are discussed in 2.1. through 2.3. Subjectless constructions, consisting of inalienable possessor, pseudo-inalienable possessor, and short infinitive (impersonal) constructions are the subjects of 2.4. A discussion of ECM and tough constructions follows in 2.5. and 2.6., respectively. The examination of the subject raises the question of whether there are overt and covert expletives in Persian. This topic is addressed in 2.7. Weather verbs are also discussed in this section. The outcome of the descriptions provided thus far is analyzed in 2.8. The summary of the discussion appears in 2.9.

2.1. Unaccusatives The theme of an unaccusative verb functions as the grammatical subject of the clause. If it is a singular noun, it remains within the PredP in a sentence with unmarked intonation, as shown by the following data. The grammatical subject appears in italics. (3)

barâ-mun mehmun âmad for us guest came-3sg ‘Some guest(s) came for us.’

(4)

Jelo dar sag neshaste in front door dog sitting-3sg ‘There is/are dog(s) sitting in front of the door.’

In (3) and (4), the grammatical subject follows the prepositional phrase. This element may appear in a higher position only if it bears heavy contrastive or emphatic stress, as in (5) and (6)36. The presence of the vP adverbials aghlab ‘often’ and hamishe ‘always’ shows that the moved theme is outside the vP (see chapter 4 for a detailed discussion of adverbs and their position).

Subject in Persian

(5)

MEHMUN aghlab barâ-shun mi-yâ-d guest often for-them dur-come-3sg ‘GUEST(S) often come for them.’

(6)

SAG hamishe jelo dar neshaste dog always in front door sitting-3sg ‘A DOG/DOGS is/are always sitting in front of the door.’

73

The subject precedes the vP adverb and the prepositional phrase in (5) and (6), receiving contrastive or emphatic interpretation. The specific grammatical subject, however, moves out of the PredP into a higher position, as in (7): (7)

in sag-e sefid jelo dar neshaste bud this dog-Ez white in front door sitting was ‘This white dog was sitting in front of the door.’

The under lying phrase structure of (3)–(7) is provided in (8). (8)

vP Spec

v' PP

v'

PredP DP

v

V

The subject in (3) and (4) remains in-situ, while it moves into the Spec of vP in (7). In (5) and (6), the subject has moved out of the vP through the Spec of vP into the functional domain. Within the framework we are adopting here, namely phase theory, movement out of a phase is only possible from the edge (specifier position) of that phase. Thus, all elements have to move into the Spec of vP before moving out into the higher phase. The question is whether the movement of the specific DP into the Spec of vP is the first step for it to move into the Spec of TP, or if it may remain in Spec of vP. In order to answer this question we need to examine the po-

74

Chapter 3: Local Scrambling and A-Movement

sition of subjects in other types of sentences (see sections 2.2.-2.6.), as well as the subject position with respect to adverbials (see section 2.7.3.). The important issue up to this point is that the nonspecific theme (= grammatical subject) of the unaccusative verb does not move out of the PredP. That is, at least one type of subject remains inside vP when there is no contrastive/emphatic stress involved.

2.2. Passive Whether or not there is syntactic passive construction in Persian has been highly controversial. Some linguists have argued that there is a structural passive construction in Persian, similar to that observed in English (Palmer 1971, Soheili Isfahani 1976, and Hajatti 1977). Moyne (1974), in contrast, suggests that Modern Persian lacks passive constructions, and all those cases that have been considered passive are in fact constructed with the inchoative verb shodan ‘become’. Dabir Moghaddam (1985) disagrees, suggesting that the inchoative shodan is not the same as the passive shodan, and joins the first group, arguing that Persian does exhibit structural passive constructions. Given the analysis of Persian complex predicates proposed by Folli, Harley, and Karimi (in Press), briefly discussed in chapter one of this volume, we argue that the passive construction is just an instance of CPr, with a past participle serving as its NV element. (9)

be Parviz gol dâde shod to P flower given became ‘Flowers were given to Parviz.’

(10) un gol-â be Parviz dâde shod that flower-pl to P given was ‘Those flowers were given to Parviz’ The past participle dâde has adjectival properties.37 The phrase structure of (9) and (10) is provided in (11). The complement of the verbal adjective moves into the Spec of vP if specific, as in (10) (illustrated by (11)). Otherwise it remains in-situ, as in (9).

Subject in Persian

(11)

75

vP v' …

v

AP PP

A'

be Papar

DP

'to Papar'

un gol-â

shod 'became' A

'those flowers'

dâde 'given'

The claim that there is no real passive construction in Persian is supported by the fact that certain agentive light verbs are replaced by inchoative or unaccusative light verbs in a CPr when a passive interpretation is intended. (12) a. Parviz xuna-ro xarâb P house-râ destroyed ‘Parviz destroyed the house.’ b. xune xarâb shod house destroyed became ‘The house was destroyed.’ c. *xune xarâb karde house destroyed made

kard did

shod became

(13) a. Parviz bachche-hâ-ro shekast dâd P child-pl-râ defeat gave ‘Parviz defeated the children.’ b. bachche-hâ shekast xord-an child-pl defeat collide-3pl ‘The children were defeated.’ c. *bachche-hâ shekast dâde shod-an The (a) sentences in (12) and (13) are agentive with the light verbs kardan ‘do/make’ and dâdan ‘give’, respectively. These light verbs are replaced by the inchoative shodan ‘become’ and the unaccusative xordan ‘collide’ in (12b) and (13b), respectively. The ill-formedness of the strings in (c) indicates that we have a CPr, with the nonverbal element as the head of the PredP. Thus, there is no room for another non-verbal element to serve as

76

Chapter 3: Local Scrambling and A-Movement

the head of CPr in these constructions, supporting the claim that passive constructions are just instances of CPrs.38 The structure of (13a) is provided in (14). (14)

vP v'

DP Parviz AP

v

DP

N

xuna-ro

xarâb

kard

The underlying structure of the corresponding inchoative CPr is presented in (15). (15)

vP AP

v

DP

N

xune

xarâb

shod

The analysis in this section is compatible with Moyne’s intuition that passives in Persian are just instances of inchoative constructions. What is relevant to our present discussion is that the nonspecific subject of the socalled passive construction, just like the grammatical subject of the unaccusative construction, does not move out of the predicate phrase into a higher position (cf. the example in (9))39. Thus there is no subject movement into the Spec of TP in these cases.

2.3. Raising constructions It has been argued in the literature that Persian lacks raising constructions (Hashemipour 1989, Karimi 1999c, and Ghomeshi 2001)40. This gen-

Subject in Persian

77

eral assumption is based on the following facts: (a) the embedded subject does not need to move into the matrix clause, (b) it agrees with the embedded verb, and (c) there is no agreement between the matrix verb and the moved embedded subject. These facts are illustrated by (16) and (17). (16) a. be nazar mi-yâ-d/*mi-yâ-n (ke) bachche-hâ to view dur-come-3sg/*dur-come-3pl that child-pl xaste bâsh-an tired be-3pl ‘It seems that the children are tired.’ b. bachche-hâ be-nazar mi-yâ-d/*mi-yâ-n (ke) t xaste bâsh-an (17) a. momken-e/*momken-an bachche-hâ possible-3sg/possible-3pl child-pl ‘It seems that children are tired.’ b. bachche-hâ momken-e/*momken-an child-pl possible-3sg/*possible-3pl ‘The children seem to be tired.’

xaste bâsh-an tired be-3pl xaste bâsh-an tired be-3pl

The examples in (16) and (17) show that there is no need for the embedded subject to be in the subject position of a raising verb, and that it does not agree with the matrix verb when it appears in the matrix clause41. Furthermore, any other phrasal element from the embedded clause may move into the matrix clause in these constructions: (18) a. be nazar mi-yâ-d/*mi-yâ -n (ke) bachche-hâ to view dur-come-3sg/*dur-come-3pl that child-pl ketâb-â-ro xunde bâsh-an book-pl-râ read be-3pl ‘It seems that the children have read the books.’ b. ketâb-â-ro be-nazar mi-yâ-d/*mi-yâ-n (ke) bachche-hâ xunde bâsh-an In (18b), the object has moved into the matrix clause while the embedded subject remains in-situ. As the inflection on the matrix verb indicates, there is no agreement between the verb and the extracted object. It could be argued that the matrix subject position is filled with a covert expletive. We will see in section 2.7. that there are no overt expletives in Persian, and that there is no evidence to assume the existence of covert expletives.

78

Chapter 3: Local Scrambling and A-Movement

2.4. Subjectless constructions There are interesting constructions in Persian with no overt subject in which the optional DP in the clause-initial position is co-indexed with a clitic that is attached to the NV-element within the CPr. This group of constructions consists of two subgroups which I call Inalienable Possessor constructions and Inalienable Pseudo-possessor constructions. They will be discussed in 2.4.1. and 2.4.2., respectively. There is yet another set of verbs that lack a subject, and are interpreted as impersonal. These verbs are known as masdar-e moraxxam ‘curtailed infinitive’ by Persian grammarians. We will call them short infinitives, and will discuss them in 2.4.3. These constructions differ from the first two sets in some ways, as we will see below. 2.4.1. Inalienable possessor constructions There is a class of CPr that does not allow an overt grammatical subject. The verb is always in the third person singular, and there is an optional DP in the clause initial position, co-indexed with the clitic attached to the NV-element inside the CPr. The semantics of this construction, representing experiencer-type construction analogous to Romance ‘J’ai froid’-type constructions, reveals a possessive interpretation. Consider the following examples: (19) (mani) gorosn-ami-e I hungry-me-is ‘I am hungry.’ (20) (mani) garm-ami-e I warm-me-is ‘I am warm’ (21) (Kimeai) sard-eshi-e K cold-her-is ‘Kimea is cold’ (22) (unai) xast-ashuni-e they tired-them-is ‘They are tired.’

Subject in Persian

79

The optional DP in the clause-initial position cannot be the grammatical subject since it does not agree with the verb morphologically. The fact that it must be co-indexed with the clitic is supported by the ungrammaticality of the following constructions: (23) *to gorosn-am-e you hungry-me-is (24) *to garm-am-e you warm-me-is What is the underlying structure of these sentences? What is the grammatical subject in these cases, and what is the role of the optional DP? We suggest that these sentences have an underlying possessor construction containing HAVE. The obligatory clitic is the possessor, co-indexed with the optional overt DP in the initial position. Kayne (1993) and Harley (1999) express the intuition that HAVE is in fact a prepositional element incorporated into a verbal be. Harley (1995: 208-9) suggests the phrase structure in (28) as the underlying structure for (25)–(27). (25) Calvin fears the weirdos from another planet (26) Calvin is afraid of the weirdos from another planet (27) Calvin has a deep-rooted fear of the weirdos from another planet.

80

Chapter 3: Local Scrambling and A-Movement

(28) EventP

Event' PP

BE

P'

Calvin HAVE

NP PP

(a) fear P

DP

weidos from another planet (Harley 1995: 209)

In order to derive the sentence in (25), the noun fear moves into HAVE, and the two of them move into BE, producing the verb fears. The sentence in (26) is derived by moving the noun fear into HAVE, producing the adjective afraid. The copula be will be spelled out in this case. The sentence in (27) is derived by moving HAVE into BE Following Harley (1995), we suggest that there is a preposition HAVE (‘predicate’ in our system) in the underlying structure of the sentences in (19) – (22), creating a possessor construction. Thus these sentences have an underlying structure similar to the one proposed by Harley for the sentence in (26). Furthermore, we suggest that the copy of the possessor DP man ‘I’, represented by the person/number features (φ-features) on the noun hunger, surface as a clitic pronoun. This structure is provided in (29).

Subject in Persian

(29)

81

vP v'

…

v

PredP DP mani (I)

BE

Pred' ROOT

Pred

HUNGER +φ features HAVE

Agree The incorporation of the root into the Pred (HAVE) gives us the adjective gorosne ‘hungry’ in this example, similar to the incorporation of fear into HAVE in (26) that gives us the adjective afraid. In both cases the copula BE is spelled out. Furthermore, there is an Agree relation between the possessor DP in the Spec of the PredP and its copy, the φ-features on the root. These features surface as the clitic pronoun, attached to the root, and co-indexed with the full DP post-syntactically. The fact that the clitic pronoun is a copy of the possessor DP explains the ungrammaticality of (23) and (24). What is the status of the possessor in those cases where the overt DP is missing? We suggest that pro is in the possessor position in the absence of an overt DP. This is compatible with the general property of the language that allows pro in sentences where the subject is not overt, as in (30) below. (30) a. to/pro in film-ro did-i? you this film-râ saw-2sg ‘Have you seen this movie?’ b. emruz man/pro ketâb-ro barâ-sh mi-frest-am today I book-râ for-her dur-send-1sg ‘I will send the book to her today.’

82

Chapter 3: Local Scrambling and A-Movement

We will see in section 4 that the optional overt DP serves as the topic of the clause. The copy of the possessor that appears as the clitic pronoun in (29) is reminiscent of a general double clitic construction in this language. Consider the sentence in (31a) whose underlying structure is provided in (31b). Irrelevant details are omitted. (31) a. (un ketâb-ro)i man ti be Kimea dâd-am-eshi that book-râ I to K gave-1sg-it ‘As for that book, I gave it to Kimea.’ b. TP

T' DP vP (un ketâb-ro)i T vP DP man

v'

ti

v

PredP

PP

be Kimea

Pred'

DP

Pred

ti

dâd-am-eshi Agree

The predicate dâd ‘gave’ subcategorizes for an object that is merged as its sister. The specific object moves into the Spec of vP for interpretation, as we will see in section 3. This object DP can be topicalized, and therefore, may appear in Spec of TP (see chapter 4 for discussion). The relevant

Subject in Persian

83

point here is that the copy of the object shows up as a clitic pronoun on the predicate. Thus, the relationship between the full object DP and the clitic pronoun on the Predicate in (31b) is the same as the full possessor DP and the clitic pronoun on the Predicate in (29)42. The next question to ask is this: what is the grammatical subject of the clause in these constructions? An expletive? This issue will be discussed in sections 2.7. and 2.8. where expletives are discussed, and an analysis of the data is provided. 2.4.2. Inalienable pseudo-possessor constructions The set of data we call inalienable pseudo-possessor constructions have similar properties as those discussed in the previous section. The difference is that the light verb is not BE in these examples. Consider the following examples: (32) (man) az in rang xosh-am mi-yâ-d I of this color pleasure-me dur-come-3sg ‘I like this color.’ (33) (man) az in rang bad-am I of this color bad-me ‘I dislike this color.’

mi-yâ-d dur-come-3sg

(34) (bachche-hâ) xeyli dard-eshun child-pl very pain-them ‘The children are hurting badly.’

mi-yâd dur-come-3sg

(35) (shomâ) dust-â-ye ghadimi farâmush-etun shode you friend-pl-Ez old forgetting-you become-3sg ‘You have forgotten old friends.’ Intuitively, these examples have a reading where HAVE is implied. In fact, the old fashioned version of (32) is the one in (36), where HAVE is present: (36) mâ in rang-râ xosh na-dâr-im we this color-râ pleasure neg-have-1pl ‘We do not like this color.’

84

Chapter 3: Local Scrambling and A-Movement

We suggest an underlying structure for these examples similar to that in (29). The difference is in the choice of light verb. In addition, the complement PredP of the light verb has a more complex internal structure. The phrase structure in (37) represents the sentences in (32) and (33). This phrase structure serves as the underlying structure for (34) and (35) as well (the embedded PP in (32) and (33) is missing in (34) and appears as a DP in (35)): (37)

vP v' … v

PredP

Pred' mi-yâd ‘comes’

DP NP mani

PP

ROOT

Pred HAVE

T az in rang xosh + φ-features ‘of this color’ Agree In this construction, similar to (29), the root xosh ‘pleasure’ is incorporated into HAVE, while the light verb is spelled out. In the absence of the optional DP man ‘I’, pro appears in the Spec of PredP. There is no overt grammatical subject in these constructions, and the optional DP serves as the topic of the sentence, once it moves out of the PredP. 2.4.3. Short infinitives The Persian infinitive consists of the past stem of the verb plus the infinitive suffix -an: [Past stem + an]. Subjectless verbs appear as short infinitives, known as masdar-e moraxxam ‘curtailed infinitive’. In these constructions, the bare past stem (lacking the infinitive suffix) is utilized. In most cases, the short infinitive forms are preceded by the impersonal modal bâyad ‘must’ or the auxiliary shodan ‘become’:

Subject in Persian

85

(38) kâr-ro bâyad zud anjâm dâd work-râ must soon result give-3sg ‘One must take care of the job fast.’ (39) bâyad xod-ro shenâxt-3sg must self-râ recognize ‘One must know herself.’ (40) besâdegi mi-sh-e vâred-e xune shod easily dur-become-3sg enter-Ez house become-3sg ‘One can easily enter the house.’ These constructions are different from the two subjectless cases discussed in 2.4.1. and 2.4.2. in that they must have a covert grammatical subject. This covert subject cannot be an expletive (see also section 2.7.), since the short infinitive can license Acc Case, as in (38) and (39). Thus, the existence of an expletive would violate Burzio’s Generalization, restated in (41): (41) Burzio’s Generalization i. A verb which lacks an external argument fails to assign Accusative Case. (Burzio 1986: 178-9) ii. A verb which fails to assign Accusative Case fails to theta mark an external argument. (Burzio 1986: 184) The existence of a covert subject other than an expletive is further supported by the fact that the covert subject binds the anaphor xod ‘self’ in formal Persian, as evidenced by the sentence in (39). Note that this anaphor is subject oriented in Persian. If our analysis is on the right track, the next question is this: what is the nature of this covert subject? Ghomeshi (2001) suggests that these verbs have an arbitrary subject. She offers the following contrast to support her claim: (42) mi-dun-am (ke) [bâyad raft] dur-know-1sg that must go ‘I know one must go.’

(Ghomeshi 2001: 20)

(43) *sa’y kard-am(ke)[bâyadraft] trydid-1sgthatmust go

(Ghomeshi 2001: 21)

86

Chapter 3: Local Scrambling and A-Movement

In (42), the short infinitive construction serves as the complement to the verb dânestan ‘to know’, a non-control verb. In contrast, in (43) it is the complement to the subject-control CPr sa’y kardan ‘to try’. The ungrammaticality of (43) is justified if the short infinitive subcategorizes for an arbitrary PRO as its external argument. The existence of the external argument is supported by the ability of the verb to license the Acc Case and for the anaphor to be bound by PRO, as mentioned above. Moreover, the arbitrary nature of PRO in these cases is evident in the English translation, that is, the arbitrary pronominal one. Does PRO remain inside vP or moves into the Spec of TP? This is part of the general question regarding the position of Persian subjects that is analyzed in sections 2.8. and 4. in this chapter.

2.5. ECM We have seen that Persian raising constructions are quite different from English since (a) the embedded verb is inflected for number and person, and agrees with its subject, and therefore, it is not necessary for the subject to move into the matrix clause for Case purposes, (b) there is no agreement between the matrix verb and the derived subject, and (c) any phrasal element other than the subject may also move into the matrix clause. In short, Persian lacks subject raising into a subject position. The question we would like to address in this section is this: does Persian exhibit subject raising into the object position? The answer is no, based on the following pieces of empirical evidence: (44) Kimea eteghâd dâr-e (ke) Rahjue-(*ro) mi-bar-e K belief have-3sg that R dur-win-3sg ‘Kimea believes that Rahjue will win.’ (45) Kimea tasavvor mi-kon-e (ke) Rahjue-(*ro) be-bar-e K imagine dur-do-3sg that R subj-win-3sg ‘Kimea considers Rahjue to win.’ In both cases, the thematic external argument of the embedded sentence is its grammatical subject as well, since the verb agrees with it in person and number. Furthermore, the embedded subject cannot take the particle -râ, an element that marks specific direct objects for Acc Case43. Clearly,

Subject in Persian

87

the embedded subject is not checked for Acc Case by the matrix verb, indicating that Persian lacks ECM constructions. The following data provide a counter example to the conclusion stated above. (46) pro ki-(ro) dust dâri (ke) bâzi-ro be-bar-e who-râ friend have that game-râ subj-win-3sg ‘Who is it you want to win the game?’ (47) pro ki-(ro) tarjih mi-d-i (ke) bâzi-ro be-bar-e who-râ preference dur-give-2sg that game-râ sub-win-3sg ‘Who is it you prefer to win the game.’ We will see in chapter 4 that Persian lacks structural wh-movement (see also Karimi 1999c). The wh-phrase may move, however, in which case it receives contrastive focus interpretation. If the matrix verb is transitive, as in the case of the main verbs in (46) and (47), the particle -râ optionally appears. That is, the subject of the embedded clause may optionally be checked by the matrix little v for Acc Case if the latter is transitive. This is not a typical ECM construction for the following reasons: (a) -râ is optional (as in (46) and (47), and (b) the subject may remain in its base position, as in (48) and (49), in which case, the presence of -râ renders the sentence ungrammatical. (48) dust dâri (ke) ki-(*ro) bâzi-ro be-bar-e friend have that who game-râ subj-win-3sg ‘Who do you want to win the game?’ Lit. who do you prefer that (they) win? (49) tarjih mi-d-i (ke) ki-(*ro) bâzi-ro be-bar-e preference dur-give-2sg that who game-râ sub-win-3sg ‘Who do you prefer to win the game.’ Lit. who do you prefer that (they) win? The underlying structure of the sentences in (46)–(49) is provided in (50). Irrelevant details are omitted.

88

Chapter 3: Local Scrambling and A-Movement

(50)

XP vP … v' Spec CP

v' PredP

v Ki

The wh-phrase remains inside the embedded CP in (48) and (49). It moves into the Spec of the matrix vP on its way to the focus position in the matrix clasue in (46) and (47). It is in Spec of vP that it is optionally interpreted as object of the matrix clause. In that case, it is followed by râ, the Case marker for specific objects. Kiss (2002) reports that a similar situation exists in Hungarian. That is, in spite of the fact that this language has no ECM constructions, there are cases where the embedded subject is extracted out of an object clause by an operator movement, and receives Acc Case44.

2.6. Tough-constructions Another type of construction that has to do with subjects is the toughconstruction. Examples are provided below. (51) Johni is tough to defeati. (52) Johni is easy to please ti. In these constructions, the overt subject originates as the object of the embedded verb, and moves into the subject position of the matrix clause. Thus, the subject position is considered to be a non-theta position. This is evident from the fact that an expletive may fill that position45. (53) It is easy to please John. Persian lacks this type of construction. Consider the following example:

Subject in Persian

89

(54) saxt-e (ke) pro Kimea-ro shekast be-d-i tough-3sg that K-râ defeat subj-give-2sg ‘It is tough/difficult (for you) to defeat Kimea’ The matrix clause lacks an overt DP in (54), and the object of the embedded clause is in its base position. This element can move into the matrix clause, as in (55). (55) Kimea-roi saxt-e (ke) pro ti shekast be-d-i The embedded object is clearly not in the subject position since it carries its Acc Case marker. Furthermore, the matrix verb does not agree with the extracted object, as in (56). (ke) ti shekast be-d-i (56) in bachche-hâ-roi saxt-e this child-pl-râ tough-3sg that defeat subj-give-2sg ‘As for these children, it is tough for you to defeat (them).’ The obvious conclusion is that there is no tough construction in Persian.

2.7. Expletives Lack of a subject in certain constructions, most importantly in inalienable possessor constructions and in raising constructions, prompts the question of whether the subject position is filled with an expletive in such cases. This section is devoted to a discussion of the existence of expletives in Persian. There are three types of expletives in English, exemplified by the following examples. (57) a. It gets cold in December b. It is obvious we were tricked into this war. c. There were several linguists on our trip to the Middle East last summer. The weather it appears in (57a), the extraposition it shows up in (57b), and the impersonal there is the grammatical subject in (57c). In addition, the expletive it is also used in cleft constructions, as in (58).

90

Chapter 3: Local Scrambling and A-Movement

(58) It was John I saw last night. In this section, we examine weather, extraposition, impersonal, and cleft constructions in Persian. Our discussion shows that there are no overt expletives in this language, and that there is no evidence to assume the existence of covert expletives. The weather constructions are discussed in 2.7.1., followed by a discussion of overt expletives in 2.7.2. and covert expletives in 2.7.3. 2.7.1. Weather constructions Chomsky (1981: 323-325) suggests that weather it is not an expletive, but rather a quasi-argument. His argument is based on the fact that weather it can bind PRO, as in (59). Chomsky states that α in this case must be PRO since it is not governed. (59) It sometimes rains after [α snowing]

(Chomsky 1981: 324)

Furthermore, weather predicates can take an object (Svenonius 2002), as in (60). Given Burzio’s Generalization restated in (41), weather it must be subcategorized by the verb. (60) It rained mackerel The idiomatic construction in (61) is another piece of evidence indicating that the weather verb takes an object, and thus must subcategorize for an external argument. (61) It’s raining cats and dogs. Persian data support the intuition that weather it must be a (quasi) argument of the verb. Consider the following examples. (62) diruz xeyli barf/bârun/tagarg âmad. Yesterday much snow/rain/hail came-3sg ‘Yesterday, it snowed/rained/hailed a lot.’ (63) emruz bâd mi-yâd today wind dur-come-3sg ‘It is windy today.’

Subject in Persian

91

(64) tu tâbestun havâ dir târik mi-sh-e in summer weather late dark dur-become-3sg ‘In summer, the weather gets dark late.’ The examples in (62)–(64) show that the weather it is represented by referential arguments barf ‘snow’, bârun ‘rain’, tagarg ‘hail’, bâd ‘wind’, and havâ ‘weather’ in Persian. Thus weather it is missing in this language. 2.7.2. Overt expletives It is well-known that richly agreeing Null-subject languages lack overt expletives (cf. Italian, Spanish, etc.) One exception to this general rule is Finnish. This language, although a richly agreeing Null-subject language, exhibits overt expletives (Holmberg and Nikanne 2002). Persian has a rich inflectional system and is a Null-subject language. The question we would like to answer in this section is this: is Persian similar to those Null-subject languages that lack overt expletives, or it is rather similar to Finnish? At first glance, it seems there is evidence for the existence of overt expletives in Persian. Consider the following examples that seem to pattern with extraposition constructions in English (except for the fact that in is optional in these examples): (65) (in) bar hame vâzeh-e [CP ke Kimea doxtar-e xubi-ye] this to all obvious-is that K girl-Ez good-3sg] ‘It is obvious to everyone that Kimea is a good girl.’ The element in ‘this’ may optionally appear in the subject position of the sentence in (65) and similar cases. In fact, Darzi (1996) has considered this element an expletive (see footnote 40 for details). Furthermore, in may optionally appear in cleft constructions: (66) (in) shomâ-hâ bud-in ke mamlekat-ro be in ruz andâxt-in this you-pl were-2pl that country-râ to this day brought-2pl ‘It was YOU who brought the country to this state.’ (67) (in) Kimea bud ke tunest bâ un be-sâz-e this K was that was able with him subj-put up-3sg ‘It was Kimea who was able to put up with him.’

92

Chapter 3: Local Scrambling and A-Movement

However, there are some problems with considering in an expletive. First, the postverbal clause in (65) may appear in the matrix clause, following in. The presence of in is obligatory in these cases. (68) in [CP ke Kimea doxtar-e xubi-ye] bar hame âshekâr-e this that K girl-Ez good-is] obvious-3sg In fact, several linguists have suggested that subordinate clauses, such as the one in (65), is an NP headed by in (Moyne and Carden 1974, SoheiliIsfahani 1976, Dabir-Moghaddam 1982). Furthermore, in can be attached to the complementizer ke in the object position of a preposition, as in (69)46. (69) man be in [CP ke Kimea doxtar-e xubi-ye] eteghâd dâr-am I to this that K good-Ez girl-is belief have-1sg ‘I believe that Kimea is a good girl.’ Lit: I have belief to this that …’ Finally, the usage of in in cleft constructions is very limited, as evidenced by (70)–(72). (70)

(*in) tu xiyâbun bud ke man un-o did-am this in street was that I he-râ saw-1sg ‘It was on the street that I saw him.’

(71) (*in) be Rahjue bud ke man ketâb-ro dâd-am this to R was that I book -râ gave-1sg ‘It was to Rahjue that I gave the book.’ (72) (*in) ruz-e shanbe bud ke man un-o did-am this day-Ez Saturday was that I he-râ saw-1sg ‘It was Saturday when I saw him.’ All these sentences are perfectly fine without in. Can in in (66) and (67) be considered a demonstrative rather than an expletive? Demonstratives can replace the whole DP in Persian, as in (73): (73) in injâ chekâr mi-kon-e? this here what dur-do-3sg ‘What is this (person/thing) doing here?’

Subject in Persian

93

If in is a demonstrative, the impossibility of its presence in (70) - (72) is explained: in cannot be associated with the PPs in (70) and (71) or the adverbial in (72). Thus, the data discussed in this section confirm that in is not a real expletive. The impersonal there is also absent in Persian. Persian infinitive hastan ‘be’ is the existential form of the copula budan ‘be’. The presence of this verb provides an existential reading, without the need for an expletive47. Consider the following contrast: (74) a. mive tu yaxchâl-e fruit in fridge-be-3sg ‘The fruit is in the fridge.’ b. mive tu yaxchâl hast fruit in fridge be-3sg ‘There is fruit in the fridge.’ The discussion in this section shows that Persian lacks overt expletives. 2.7.3. Covert expletives Rizzi (1982/1986) suggests that Null-subject languages do not have overt expletives since these languages, by virtue of being able to license pro, can license empty nonreferential subjects. In other words, he suggests that Null-subject languages have covert expletives. We have seen that Persian does not have overt expletives, which is compatible with Rizzi’s insight. However, there is no evidence that there are covert expletives in this language. Support for this claim comes from the fact that Persian is not subject to the Definiteness Effect. Languages that exhibit the existence of expletives cannot violate this effect, as in the case of English exemplified by (75). (75) *There is John in the room. We will see in chapter 4 that sentential adverbs appear in a vP initial position in Persian, as in (76). VP adverbs (including habitual and manner adverbials) also appear in a vP initial position, as in (77) and (78), and follow the sentential adverb, as in (79).

94

Chapter 3: Local Scrambling and A-Movement

(76) diruz Kimea bâ Rahjue da’vâ kard yesterday K with R fight did-3sg ‘Kimea fought with Rahjue yesterday.’ (77) ma’mulan Kimea tu ketâbxune dars mi-xun-e usually K in library lesson dur-read-3sg ‘Kimea usually studies in the library.’ (78) kâmelan Kimea be hush umade completely K to sense come-3sg ‘Kimea has completely regained her senses.’ (79) xoshbakhtâne kâmelan Kimea be hush umade fortunately completely K to sense come-3sg-3sg ‘Fortunately, Kimea has completely regained her senses.’ The sentences in (76) through (79) exemplify cases where the entire sentence is in focus (see chapter 4). The presence of the definite subject inside the vP shows that Persian is not subject to the Definiteness Effect, providing support for the claim that this language does not have covert expletives. We will see in section 2.8.1. that the lack of expletives is compatible with a general account of the way EPPg is satisfied in Persian.

2.8. Analysis We have seen thus far that Persian does not exhibit properties of a subject-prominent language. In other words, typical instances of A-movement do not exist in this language, based on the empirical facts examined in this section: (a) there are no raising or tough constructions, (b) the grammatical subject may stay inside the PredP in unaccusatives and the so-called passive constructions, (c) the specific subject, although out of PredP, may remain inside vP, and (d) there is no ECM. There are subjectless predicates in Persian that suggest the existence of expletives at first glance. We saw, however, that overt expletives do not exist, and no evidence can be found supporting the existence of their covert counterparts. Several questions arise: since the subject does not need to be extracted out of vP, how is Nom Case checked in Persian, and how is Agreement established between the overt subject DP and the verb? Furthermore, if the subject DP may remain inside vP, instead of moving into the Spec of TP,

Subject in Persian

95

how is the EPP satisfied? The discussion of the EPP raises another question: is there a need for the existence of pro and PRO in Persian? These questions are addressed in the following three sections. 2.8.1. Nominative Case, Agreement, and subject shift In this section, we discuss Nom Case and Agreement, and show that Persian exhibits local subject shift. That is, specific subjects escape existential closure by undergoing the rule of subject shift in this language. We have seen thus far that the theme of unaccusatives and so-called passives, which serve as the grammatical subjects of their predicates, are merged inside the PredP. Following Diesing (1992), Kratzer (1995), and Diesing and Jelinek (1995), we assume that the PredP is the domain of the existential closure. We further suggest that noun phrases inside this domain, by virtue of being part of the predicate, are neutral with respect to Case. That is, the subject inside the PredP lacks Nom Case. This is supported by the contrast between nonspecific objects and their specific counterparts: while specific objects are marked by -râ, nonspecific objects do not carry a Case marker (see section 3 in this chapter). This distinction can be extended to subjects, although Persian subjects, similar to subjects in many other languages (cf. Turkish, Urdu, Hindi, etc.) do not exhibit morphological Case. We further suggest that the same situation holds for Agreement. That is, there is no Agreement between the subject inside the PredP and the verb. This line of argumentation implies that only specific subjects are checked for Agreement. This is in fact borne out, as evidenced by the following examples. (80) a. jelo-ye dar se tâ sag neshaste bud/*bud-an fron-Ez door three part. dog sitting be-3sg/*be-3pl ‘Three dogs were sitting in front of the door.’ b. un sag-â-ye sefid jelo-ye dar neshaste bud-an/*bud that dog-pl-Ez white front-Ez door sitting be-3pl/be-3sg ‘Those white dogs were sitting in front of the door.’ The verb does not agree with the nonspecific subject in (80a). Agreement must hold between the specific subject and the verb in (80b). Recall that the nonspecific theme of the unaccusative verb must follow the prepositional phrase unless it is contrastively stressed (cf. examples in (3) and (4) versus (5) and (6)). This is not true in the case of the specific theme. In

96

Chapter 3: Local Scrambling and A-Movement

fact, the prepositional phrase must receive contrastive stress if it precedes the specific subject. This is illustrated by (81). (81) JELO-YE DAR un sag-â-ye sefid neshaste bud-an ‘It was in front of the door that those white dogs were sitting.’ We have established that there is no Agree relation between the verb and the nonspecific theme inside the PredP with respect to Case and φ-features. We have also seen that the subject does not have to move out of vP. Therefore, the only position where the specific subject and the verb can establish an Agree relation for the purpose of Nom Case and Agreement checking is the Spec of vP. Thus, we suggest the following configuration for Persian Nom Case and Subject-verb Agreement: (82) Nom Case and Subject -verb Agreement:

vP v' DP PredP

vAgreement

Agree It is uncontroversial that agents are merged in the Spec of vP (Kratze 1993), and thus must receive a specific interpretation. Agreement with the verb is obligatory in such cases. In summary, Agree is established between the verbal inflection and the specific subject in the Spec of vP, licensing Nom Case and Agreement. This analysis explains the fact that Persian lacks raising, tough, and ECM constructions. We further suggest that Nom Case is a property of Agreement (rich verbal inflection), instead of tensed T in Persian. This situation holds in Irish as well (Carnie, personal communication). The parametric situation of Nom Case is thus the one stated in (83). (83) Parametrizing Nom Case Nom Case is either a property of a. tensed T (as in English), or b. rich Agreement (as in Persian)

Subject in Persian

97

We will come back to this issue in section 2.8.3. There is evidence with respect to subject-verb agreement that supports the configuration in (82). This issue has to do with Persian subjects containing a [-ANIMATE] feature. These elements only optionally agree with the verb in number, as evidenced by the following data: (84)

a. ketâb-â ru miz-e/an book-pl on table-3sg/3pl ‘The books is/are on the table.’ b. deraxt-â sabz shod-e/an tree-pl green became-3sg/3pl ‘The trees has/have become green.’

Although the subjects are plural in (84), the verb can optionally be singular or plural. This type of agreement can be explained if the grammar allows the inanimate subject to optionally move out of the PredP to the Spec of vP in order to establish an Agree relation with the verb. In fact, it has been claimed in the literature that the plural suffix -hâ forces a specific reading (Samiian 1983). This is true with respect to animate subjects, but inanimate subjects do not necessarily seem to undergo this rule. That is, the inanimate subject carrying the plural suffix may maintain its nonspecific interpretation. In that case, it remains inside the PredP, and the verbal inflection shows the default third person singular. If the subject moves out of the PredP, it receive a specific reading, and the verb agrees with it. Is there any empirical evidence for this claim? As shown in Karimi (2003c), the nonspecific object, unlike its specific counterpart, cannot follow the verb. If our analysis is on the right track, the inanimate subject can appear in a post-verbal position only if the verb agrees with it. This prediction is in fact borne out, as evidenced by the following examples, in which the PredP has moved into a higher position to express emphasis. Only the specific subject, which is outside the PredP, can stay behind; the nonspecific subject, being part of the PredP, must move along with the verb. (85) a. [ru MIZ-an]i ketâb-â ti on table-be3pl book-pl ‘The books are on the TABLE.’ b. ??[ru MIZ-e]i ketâb-â ti -be3sg

98

Chapter 3: Local Scrambling and A-Movement

c. *[ru MIZ-e]i ketâb-â-ye to ti on table -be3sg book-pl-Ez you Intended meaning: Your books are on the TABLE (86) a. [SABZ shod-an]i deraxt-â green became-2pl tree-pl ‘The trees have become GREEN.’ b. ??[SABZ shode]i -3sg

deraxt-â

ti

ti

c. *[SABZ shode] i deraxt-â-ye mâ ti green became-3sg tree-pl-Ez our Intended meaning: Our trees have become GREEN. The ‘b’ sentences, where the verb is singular, are awkward. Those in ‘c’, where the subject has clearly become specific, are still worse. The analysis advanced in this section suggests that the process of subject shift exists in Persian, although the target and the outcome of this process is different from subject shift in English. Subject shift in Persian involves extraction of specific subjects from the PredP into the Spec of vP. The outcome of this operation has a semantic consequence, since the subject in the Spec of vP receives specific interpretation. This movement is necessary, otherwise the specific subject cannot be interpreted inside the PredP. Subject shift in English has a different outcome, since movement of the subject DP into the Spec of TP satisfies EPP, Nominative Case and Agreement48. 2.8.2. EPP The EPP is generally assumed to be universally satisfied by either the presence of the grammatical subject or an expletive in the Spec of TP49, as in English, or by V-to-T movement, as in a number of Null-subject languages. Considering Persian, we saw that raising predicates and the two types of subjectless constructions lack a subject. The first reaction would be that these constructions provide evidence for the existence of covert expletives. In the previous section, however, we discussed some evidence indicating that such an assumption cannot be maintained. We also saw that the grammatical subject may remain inside the vP. Thus, the EPP cannot

Subject in Persian

99

be satisfied by the presence of a subject DP or an expletive in the Spec of TP in this language. Alexiadou and Anagnostopoulou (1998) suggest that the EPP is satisfied by V-to-T movement in Null-subject languages. Since Persian is a Null-subject language, could we argue that V-to-T movement satisfies the universal requirement of checking EPP in this language? The answer is no due to the fact that Persian is a verb-final language, and that T is phrase-initial in this language. If there were V-to-T movement for the sake of the EPP in Persian, the SOV order could not be maintained. In contrast, if T were in final position and the EPP was satisfied by V-to-T movement, the sentential argument of the verb would have to appear preverbally, contrary to fact (see chapter 1 for discussions on the position of T and the sentential arguments of the verb). In the introductory section of this chapter we discussed two types of EPP: the first one is based on Chomsky’s (1982) original idea regarding the requirement that every sentence have a subject. We called this type of EPP the grammatical EPP, and coined the term EPPg for it. The other EPP, based on Chomsky (2000), suggests that the EPP is a purely syntactic device that is utilized to allow movement of phrasal categories. In other words, this type of EPP functions as a strong feature. The output of the application of this feature serves the semantic and discourse components of language. We called this type of EPP the syntactic EPP, and coined it with the term EPPs. If EPPg is not satisfied by the presence of the grammatical subject or an expletive in the Spec of TP, as in English, nor by V-to-T movement, as in some Null-subject languages, how is this requirement satisfied in Persian? We suggest that EPPg, similar to Nom Case, is satisfied by the rich morphological inflection on Persian verbs. That is, the inflection for number and person is rich enough to satisfy the universal requirement for the existence of a subject. This proposal provides a unified account not only for all those cases where there is an overt subject outside or inside vP, but also for those instances where the verb does not subcategorize for an external argument (raising predicates, inalianable possessor constructions, and inaliable pseudo-possessor constructions). If the analysis advanced in this section is on the right track, EPPg is universally satisfied by one of the following ways: a. XP in Spec of T (by movement or Merge as in English) b. v movement to T (as in Italian, Greek, etc.) c. Morphologically (as in Persian and Hungarian)

100

Chapter 3: Local Scrambling and A-Movement

It is conceivable to suggest that (b) and (c) should be reduced to (c): That is, in all languages that exhibit rich verbal morphology, EPPg is satisfied morphologically. Thus EPPg is parameterized as shown in (87). (87) Parameterizing EPPg EPPg is either a. a D feature of T, satisfied by the presence of XP in Spec of TP, or b. a property of the verbal morphology (overt agreement). A final point regarding Persian subjects deserves some attention. Persian does not display an asymmetry between the subject and object with respect to extraction, contrary to English. That is, extraction is possible out of the subject and object DP, and both processes are subject to the same restrictions. It is well-known that extraction out of subjects is blocked in English type languages. This is illustrated by (88). (88) *Whoi did [a book by ti ] win the prize? Consider now the following Persian examples. The contrast between (89a) and (89b) indicates that extraction is possible only out of nonspecific objects in Persian. (89) a. [az

kodum shâ’er]i Kimea [DP ye

she’r

ti]

xund

of which poet K a poet read-3sg ‘Of which poet did Kimea read a poem?’ b. *[az kodum shâ’er]i Kimea [DP in she’r ti] -ro xund-3sg * of which poet K this poem Lit: By which poet did Kimea read this poem?

-râ read

In (89a), the prepositional phrase az kodum shâ’er ‘from which poet’ is scrambled out of the nonspecific object. Extraction out of the specific DP is blocked in (89b). The same restriction holds for subjects, as illustrated by the contrast in (90).

Subject in Persian

(90)

a. [az

101

kodum nevisande]i emsâl [DP ye ketâb ti] montasher

of which writer this year a book published shod? became Lit: by which writer was a book published this year. b. *?[az kodum nevisande]i emsâl [DP in ketâb ti]] montasher * of which shod became

writer

this year this book

published

The crucial point for our analysis is that the same restriction holds for extraction out of objects and subjects50. What explains the parametric difference observed between Persian and English? Nom Case cannot be solely responsible for the difference, since the Spec of vP is the locus for Nom Case in Persian as Spec of TP is for English. The only difference we can see at this point is the way EPPg, in addition to Nom Case, is satisfied in these two languages: if our analysis is correct, EPPg and Nom Case are satisfied morphologically in Persian, but syntactically in English. The latter is triggered by a D feature on T that motivates DP movement into the Spec of TP, where the Nom Case is checked as well. This distinction might be the reason for the parametric difference between these two languages. Note that Hungarian, just like Persian, lacks subject-object asymmetry with respect to extraction (Kiss, Personal communication). Hungarian is similar to Persian in that Nom Case, Agreement, and EPPg are satisfied inside vP rather than in Spec of TP. This comparison suggests that Spec of TP, where EPPg and Nom Case is structurally checked in English, is an island, while Spec of vP, where EPPg is morphologically checked in Persian, is not51. The proposal regarding the morphological EPPg in Persian and Hungarian begs the following question: if the EPPg is satisfied by rich morphology, do we need pro? In other words, what is the status of pro? Is this element needed in UG? The same question would be raised with respect to the existence of PRO. These questions are addressed in the following section.

102

Chapter 3: Local Scrambling and A-Movement

2.8.3. pro and PRO We have suggested thus far that EPPg is satisfied by the rich verbal morphology of Persian. Does this mean that there is no need for pro or PRO? We suggest that the existence of pro and PRO is required by two independent factors. First, binding of anaphors (as discussed in chapter 5) supports the claim that pro and PRO are required in the architecture of UG. The object anaphor must be bound by a c-commanding antecedent in Spec of vP, thus requiring the existence of the Null subject in the absence of an overt DP. It could be argued that φ-features are among the features hosted by T. In that case, the existence of pro/PRO becomes irrelevant. However, if agreement is established between the DP in Spec of vP and the φ-features on the verbal inflection, as suggested in this work, there is no need for T to host the φ-features, and thus the presence of pro/PRO becomes necessary. Second, pro is required by Case. That is, the inflection of the verb has a Nom Case feature that needs to be in an Agree relation with a DP in order to be checked and eliminated. Lack of pro causes the derivation to crash, since the Nom Case feature of the verb remains unchecked. In the case of the so-called raising verbs, there is no Nom Case on the verb, since there is no external argument. Therefore, pro does not exist in such constructions52. If Nom Case justifies the existence of pro, how can we account for the presence of PRO? Chomsky and Lasnik (1993) and Chomsky (1995) suggest that PRO has Null Case, and therefore, can only appear in infinitive clauses where the verb lacks Tense, and thus has no Nom Case. Infinitive clauses do not exist in Persian, and subjunctive forms are employed in control constructions and other cases where non-finite clauses are utilized in English. These forms have the same rich morphology as other forms. If the rich morphology is responsible for the Nom Case in Persian, as we suggested in section 2.8.1., PRO receives Nom Case in the same way as pro and overt DP, since it appears in the same positions as those two elements. This is illustrated in the following examples. (91) a. man dust dâr-am [CP I friend have-1sg ‘I like to go.’ b. man dust-dâr-am [CP I friend have-1sg ‘I like for you to go.’

ke pro be-r-am] that subj-go-1sg ke to be-r-i] that you subj-go-2sg

Subject in Persian

103

c. sa’y kard-am [CP ke PRO be-r-am] ‘I tried to go.’ PRO has the same distribution in (91) as pro and the overt DP, confirming the claim that PRO receives Nom Case. This claim is a clear departure from Chomsky and Lasnik (1993) and Chomsky (1993). PRO appears in Nom Case positions in some other languages as well. Sigurdsson (1991), for example, suggests that PRO receives Case in Icelandic. In Irish, the distribution of PRO is free. That is, it appears in finite and non-finite clauses alike, and thus must receive Case in that language as well53. Persian PRO is in complementary distribution with overt DP in control constructions. This is shown in (92). (92) a. Parviz sa’y kard [ke PRO/*Kimea be-r-e] P effort made-3sg that K subj-go-3sg ‘Parviz tried to go.’ b. Parviz mi-tun-e [ke PRO/*Kimea be-r-e] P dur-can-3sg that K subj-go-3sg ‘Parviz is able to go.’ This observation suggests that the presence of PRO is determined by control, rather than Nom Case. This fact explains why PRO cannot appear in an object position even though it is allowed in a Case position. (93) *man PRO did-am I saw-1sg Lack of control must be responsible for the ungrammaticality of (93): the verb didan ‘see’ is not a control verb, and therefore, does not allow the presence of PRO. This situation holds in English as well: PRO appears in control constructions, either in arbitrary control constructions or subject/ object control constructions. This seems to be the case in Irish as well (Stenson 1989). Thus we draw the following generalization. (94) PRO is a property of control constructions universally. Consequently, the presence of PRO becomes unrelated to Nom Case (and lack of EPP), and is determined by control.

104

Chapter 3: Local Scrambling and A-Movement

2.9. Summary We saw that certain predicates lack an external argument in Persian. Furthermore, the grammatical subject may remain in situ. It was argued that Persian lacks overt and covert expletives. Thus, EPPg was suggested to be satisfied by the rich morphological inflection on the verb. Moreover, it was shown that the specific subject, if originated within the PredP as in the case of unaccusatives and the so-called passive construction, must move into the Spec of vP to escape the domain of existential closure. This movement has the following consequences: The specific subject receives interpretation in the Spec of vP, where the Nom Case and the φ-features of the DP are also checked by the rich verbal Agreeement. EPPg was also argued to be satisfied by the rich inflection in this language. Since the EPPg, the Nom Case and the φ-features of the subject are checked inside vP, it is not surprising that Persian lacks raising, tough, and ECM constructions.

3. Objects in Persian In this section, we concentrate on Persian direct objects. We will see a number of similarities between the subject and the object in this language. For example, the nonspecific object remains inside the PredP, while its specific counterpart moves out of this phrase into the lower Spec of vP. Furthermore, the movement of the nonspecific object is constrained in the same way as the movement of the nonspecific subject: its extraction correlates with a contrastive reading. As we saw in previous section, neither the subject nor the object may appear in the post-verbal position if nonspecific. First we provide a brief review of Karimi (2003c) regarding the distinctive properties of the two types of objects in 3.1. The proposal, called the Two Object Position Hypothesis (TOPH) in Karimi (2003c), and supporting arguments for the TOPH are reviewed in 3.2. This proposal is revised in 3.3. Finally, we briefly discuss Acc Case in this language. A short summary concludes this section.

Objects in Persian

105

3.1. Specific versus nonspecific We saw in chapter 1 that the specific object and its nonspecific counterpart appear in two different positions at Spell-out. Here are some examples, taken from Karimi (2003c: 91). (94) a. Kimea aghlab barâ mâ she’r mi-xun-e K often for us poem dur-read-3sg ‘It is often the case that Kimea reads poetry for us’ b. Kimea aghlab barâ mâ ye she’r az Hafez mi-xun-e K often for us a poem from Hafez dur-read-3sg ‘It is often the case that Kimea reads a poem by Hafez for us’ (95) a. Kimea aghlab hame–ye she’r-â-ye tâza-sh-ro barâ mâ K often all-Ez poem-pl-Ez fresh-her-râ for us mi-xun-e dur-read-3sg ‘It is often the case that Kimea reads all her new poems for us’ b. Kimea aghlab ye she’r az Hafez-ro barâ mâ mi-xun-e K often a poem by Hafez-râ for us dur-read-3sg ‘It is often the case that Kimea reads a (particular) poem by Hafez for us’ The direct objects in (94) are nonspecific, follow the indirect object, and lack the particle -râ. Those in (95) are specific, precede the indirect object, and are followed by -râ. The distinctions between the two types of objects are summarized as follows (cf. Karimi 2003c: 104): 1.

Semantic asymmetries: a. Non–specific objects are part of the description of the predicate; specific objects are not. b. Non–specific objects are compatible with adverbs denoting process; specific objects are compatible with adverbs denoting accomplishment. c. Specific objects enter scope interaction when scrambled; non– specific objects do not.

106 2.

Chapter 3: Local Scrambling and A-Movement

Syntactic asymmetries: a. Specific objects can bind the indirect object; non–specific objects cannot. b. Non–specific objects have to precede the verb (or the NV element within a CPr), and be adjacent to it; specific objects are not subject to this adjacency restriction. c. Specific and non–specific objects cannot appear together in a coordination construction.

3. Morphological asymmetry: Non–specific objects allow a process of lexicalization (compounding) with the verb; specific objects do not. 3.2. TOPH Based on the syntactic, semantic, and morphological distinctions reviewed above, Karimi (2003c) suggests that the specific object is generated in a higher position than the nonspecific one: the former is in the Spec of VP while the latter is a sister to V. This distinction is stated in the following hypothesis: (96) Two Object Position Hypothesis (TOPH) a. [VP DP[+Specific] [V’ PP V]] b. [VP [V’ PP [V’ DP[–Specific] V]]]

(Karimi 2003c: 105)

The following empirical arguments are offered to support TOPH (Karimi 2003c: 111-116) Scope The fact that the nonspecific object is part of the predicate explains why this element cannot take scope over negation54. (97) Kimea (YE) ketâb na–xarid Kimea (a) book neg–bought ‘Kimea did not buy (ONE) book/books’ This sentence can have the interpretation in (98a), but not the one in (98b):

Objects in Persian

107

(98) a. It is not the case that Kimea bought one book/books b. *There is a book/are books such that Kimea did not buy it/them The nonspecific object cannot take scope over the negation, even when it is scrambled into a higher position. This is not true in the case of its specific counterpart. Lack of wide scope reading in the case of nonspecific objects is explained if these elements are syntactically and semantically part of the predicate, unlike the specific object. Anti–Weak Crossover Effect and Binding TOPH accounts for apparent Weak Crossover (WCO) violations. Consider the following contrast: (99) a. Kimea i [ye dâneshju]k be hamshâgerdi -hâ-shi/*k t K a student to classmate -pl-her mo’arrefi kard introduction did-3sg ‘Kimea introduced one student to her classmates’ b. Kimeai ye dâneshju-rok be hamshâgerdi-hâ-shi/ mo’arrefi kard The sentence in (99a) is well–formed only if the pronominal affix is co-indexed with the subject. In (99b), the pronominal affix can be co-indexed with either the subject or the specific object. The fact that the scrambled nonspecific object and the pronominal affix cannot be coindexed in (99a) seems to suggest an instance of WCO violation. Binding in Double Object Constructions There is evidence indicating that the indirect object has to be in the domain of the specific direct object in Persian. This is supported by the following examples. (100) man [se-tâ bachche-hâ-ro]i be hamdigei mo’arrefi I three-part child-pl-râ to each other introduction Kard-am did-1sg ‘I introduced the three children to each other’

108

Chapter 3: Local Scrambling and A-Movement

(101) *Kimea be bachche-hâi hamdigai -ro mo’arrefi kard K to child -pl each other-râ introduction did-3sg Assuming the TOPH, in which the specific object is in the Spec position of VP, the binding relations in (100) and (101) are accounted for in a simple way. That is, the specific direct object is not in the domain of the indirect object, and hence the latter cannot bind the former55. This also suggests that the indirect object must be in an A’-position in (101), where it is unable to bind the direct object. TOPH is revised in the following section in a way that it is compatible with the analysis advanced in this work.

3.3. TOPH revisited: object shift In chapter 1, we proposed the following tentative phrase structure for a double object construction. Note that PredP was introduced as VP (cf. (8), chapter 1). However, later in that chapter it was suggested that PredP is the complement to v, and ranges over a number of XPs, including VP. (102)

CP C'

Spec

TP

C

T'

Spec

vP

T

v'

Spec PredP PP

v

Pred' Object

Pred

[+/-Specific]

The specific object moves into the Spec of vP in order to receive interpretation.

Objects in Persian

(103)

109

vP vP DPs v' DPo PredP

v

to

The configuration in (102) differs from TOPH stated in (96): both objects are merged in the same position in (102), while they were in two different positions in (96). Is the new analysis compatible with the empirical facts that support the TOPH? The configuration in (103) shows that the specific object moves into the lower Spec of vP, where it receives its interpretation. Note that this is compatible with the following rule stated by Chomsky (2001b). (104)

At the phonological border of vP, XP is assigned INT’

Chomsky considers the rule in (104) as a special case applicable in object shift languages. We saw that this rule applies to the theme of the unaccusatives and the so-called passive constructions as well: the specific theme, as the grammatical subject of the predicate, moves into the Spec of vP in order to be interpreted. According to Chomsky, this movement is free. We suggested, however, that this movement, similar to all other movements, is triggered by EPPs. The analysis of specific objects in this work provides the same results as the TOPH does. That is, the WCO and binding relations discussed in this section (cf. (99)–(101)) are accounted for by this analysis: the specific object is interpreted in the Spec of vP. It is thus in that position that the binding relations are established. WCO is not violated since the specific object has no interpretation when it crosses the pronominal56. Note that the nonspecific object in (99a) lacks the particle -râ, the element that marks specificity for Acc Case. Thus, the semantics of this element does not allow it to bind the clitic pronoun. Finally, the current analysis shows that the spe-

110

Chapter 3: Local Scrambling and A-Movement

cific object undergoes object shift to escape the existential domain. The same situation holds for specific subjects. Thus, the analysis advanced in this work provides a more unified account of specific subjects and object.

3.4. Accusative Case As discussed in section 2.8.1., Nom Case is checked by Agree between the inflection of the verb and the specific subject in the Spec of vP. Acc Case is checked in a similar fashion: the Acc Case feature is locally checked by Agree between the specific object in the lower Spec of vP and v. The following configuration illustrates this relation57. (105)

vP vP DPs v' DPo PredP

v

to Agree 3.5. Summary In this section, we discussed the asymmetries between the specific and nonspecific objects, and suggested that the specific object moves out of the PredP in order to escape the domain of existential closure. This movement, as in the case of the specific grammatical subject of unaccusatives and so-called passives, is triggered by EPPs. The object receives its interpretation in the derived position, and thus is able to bind the indirect object in a lower position. It was suggested that Acc Case is checked by a local Agree relation between the specific object and v. This analysis provides a simple and unified system for specific subjects and objects: both are shifted into the Spec of vP primarily for interpretation while

Persian as a topic-prominent language

111

their Case is also checked in that position58. The agent is generated and checked for Case in that position by Merge and Agree, respectively. Obviously, this analysis allows v to project multiple Specs. We will see in chapter 4 that the existence of multiple Specs is motivated independently elsewhere in this language.

4. Persian as a topic-prominent language We have established thus far that Persian is not a subject-prominent language. That is, all phrasal elements may stay inside the vP, the proposition entity that lacks an information unit. They move into the information unit to be interpreted as topic or focus. Adverbs follow the information unit, and precede the proposition entity. Given this line of reasoning, Persian must be a topic-prominent language (see also Kiss 1995, 1997, 2002). That is, any specific element can be topicalized, serving as the subject of the predicate, and the rest of the sentence will be predicated of this specific entity59. There are certain differences between subject-prominent and topicprominent languages60. In a subject prominent language, the passive construction is utilized to extract a non-subject topic (Holmberg 2000). In a topic-prominent language, a phrasal constituent is extracted without the need to use a passive construction. Compare the following contrast between English and Persian: (106) a. This cake was baked by my mother b. in keyk-ro mâdar-am poxte this cake-râ mother-my cooked ‘This cake, my mother has baked (it).’ The Persian counterpart of (106a) is awkward at best, while the English counterpart of (106b), represented by (107a), is ungrammatical. (107) a. *This cake has my mother backed. (= 106b) b. ??in keyk bevasile-ye mâdar-am poxte shode (= 106a) in cake by-Ez mother-my cooked became ‘This cake was baked by my mother.’ In (107a), the object, rather than the subject, has moved into the Spec of TP, rendering the sentence ungrammatical, since only subjects may un-

112

Chapter 3: Local Scrambling and A-Movement

dergo this movement in active sentences in a subject-prominent language like English. The awkwardness of (107b) is due to the fact that this sentence is modeled based on the English passive construction, since the usage of the by phrase is highly unnatural in these constructions in Persian. Taking out this prepositional phrase, the result is an inchoative CPr, as discussed in section 2.2. The following data provide a similar contrast between English and Finnish, the latter is considered to be a topic prominent language. (108) a. Graham Greene has written this book. b. *This book has Graham Greene written (109) a. Graham Graham b. tämän this

Greene Greene kirjan book

on has on has

kirjoittanut tämän kirjan. written this book kirjoittanut Graham Greene written Graham Greene (Holmberg & Nikanne 2002: 78)

While (109b) is a natural sentence in Finnish, its English counterpart in (108b), where the object occupies the Spec of TP, is sharply ungrammatical. The contrast between Persian and Finnish, on the one hand, and English, on the other, is observed in other subject-prominent versus topicprominent languages as well: French is a subject-prominent language, while Russian is a topic-prominent language. (110) *Ce livre a écrit Graham this book has written Graham

Greene Greene

French

(111) Etu knigu napisal Graham Greene Russian This book has written Graham Greene (Holmberg & Nikanne 2002: 78) Returning to Persian, we saw that all arguments may remain in situ when the entire sentence remains in the propositional phase, even when the subject is specific. Examples were provided in (76)–(79), repeated below in (112)–(115). (112) diruz Kimea bâ Rahjue da’vâ kard yesterday K with R fight did-3sg ‘Kimea fought with Rahjue yesterday.’

The role of T

113

(113) ma’mulan Kimea tu ketâbxune dars mi-xun-e usually K in library lesson dur-read-3sg ‘Kimea usually studies in the library.’ (114) kâmelan Kimea be hush umade completely K to sense come- 3sg ‘Kimea has completely regained her senses.’ (115) xoshbakhtâne kâmelan Kimea be hush umade fortunately completely K to sense come-3sg-3sg ‘Fortunately, Kimea has completely regained her senses.’ The element that moves out of the vP can be any argument, including the subject: (116) a. Kimea emruz ketâb-o be Parviz mi-d-e. K today bool-râ to P dur-give-3sg ‘Kimea will give the book to Parviz today.’ b. ketâb-o emruz Kimea t be Parviz mi-d-e. c. be Parviz emruz Kimea ketâb-o t mi-d-e In (116a), the subject is topicalized, while the direct object and the indirect object are topicalized in (116b) and (116c), respectively. We see in chapter 4 that topicalized elements either move into the Spec of TopP or the Spec of TP, depending on their interpretation with respect to discourse. This movement is triggered by EPPs in order to provide the intended output. 5. The role of T If the line of argumentation advanced thus far is on the right track, Case and Agreement are satisfied inside vP, and EPPg is a morphological entity in this language, satisfied by the rich morphological inflection on the verb. The question that arises at this point is this: where is the tense of the verb checked, and what is the role of T? Persian verbal stems appear in two forms: present and past. Here are some examples: (117) Present a. rav b. xor

Past raf+t xor+d

‘go’ ‘eat’

114

Chapter 3: Local Scrambling and A-Movement

c. nevis d. shav e. kon

nevesh+t sho+d kar+d

‘write’ ‘become’ ‘do/make’

As seen in these examples, the past morpheme has the two allomorphs /t/ and /d/ in Persian. There is also a set of infinitives known as masdar-e ja’li ‘pseudo infinitive’. These infinitives are historically formed by adding /id/ to an existing noun. Examples are provided below. (118) noun/Present a. raghs b. xâb

Past raghsid xâbid

‘dance’ ‘sleep’

One major difference between the stems in (117) and those in (118) is that the relation between the present and past root in the former is not always regular (cf. nevis versus nevesh or shav versus sho), while it is always regular in the latter. The bottom line is that the bundle of features representing the present stem lacks the feature that is spelled out as /t/, /d/, and /id/, while the bundle of features representing the past tense includes it. We suggest that T must carry the same type of feature: [+/- Pres]. Thus, tense is locally checked by Agree between v and T in the following configuration. (119)

TP vP T …

v

Agree (for Tense) In chapter 4 we will see that TP can project a Spec as the landing site of an element marked [+Topic]. Thus T must be able to optionally select the feature [+Topic] in this language. Does T optionally select the feature [+Topic] universally? If so, all languages must be considered topic-prominent. The parametric difference boils down to the range of elements that may function as the subject of predication in a given language. Thus in a language like English (or French and Danish, among others), only grammatical subjects are marked [+Topic], while in Persian (or Hungarian and

Local scrambling: A-movement?

115

Finnish, among others), any specific element can be marked [+Topic]. The problem with this line of argumentation is that all subjects must be interpreted as specific in English, contra the fact. However, if T is allowed to optionally select the feature [+Topic], the problem disappears: only in the case of specific subjects is this feature selected. Otherwise, the derivation crashes.

6. Local scrambling: A-movement? Thus far, we have discussed one type of movement that applies to allow specific objects and specific grammatical subjects of unaccusatives and socalled passives to escape the domain of existential closure. We suggested that this movement represents instances of subject shift and object shift. The landing site of this movement is the Spec of vP, where Nom Case and Acc Case are checked by an Agree relation between the specific DP and v. Case positions have traditionally been considered to have the properties of an A-position. Furthermore, the subject of the agentive verb is generated in Spec of vP by Merge, a theta position. Thus, Spec of vP should reveal properties of A-positions. Does this mean that subject shift and object shift are instances of A-movement? In other words, is vP-internal scrambling a type of A-movement in this language? Below we address this question by considering binding relations. We saw that the specific object binds into the indirect object from the Spec of vP. A relevant example was provided in (100a), repeated as (120) below. (120) man [se-tâ bachche-hâ-ro]i be hamdigei mo’arrefi I three-part child -pl-râ to each other introduction kard-am did-1sg ‘I introduced the three children to each other’ The sentence in (121) shows that the grammatical subject of the so-called passive may bind a pronominal it c-commands: (121) Kimeai dar daftar-ashi/k dide shod K in office-her seen became-3sg ‘Kimea was seen in her office.’

116

Chapter 3: Local Scrambling and A-Movement

The binding relations in (120) and (121) suggest that the landing site of the vP-internal scrambling is an A-position. However, we also saw that the movement of the indirect object into the Spec of vP does not provide a binding relation (cf. 101). This issue suggests that Spec of vP is not a typical A-position. We come back to binding relations in chapter 5, and to AA’ distinctions in chapter 7, where a critical analysis of the typology of movement is discussed.

7. Conclusion We saw in this chapter that Persian does not exhibit obligatory subject movement into the Spec of TP. This conclusion was drawn on the basis of the fact that (a) subjects do not necessarily move out of the vP, (b) there are no raising, tough, and ECM constructions in this language, and (c) overt expletives do not exist, and there is evidence suggesting that covert expletives are absent in this language as well. In fact, no element needs to appear in the vP external position in a situation when the entire propositional phase is focused. We saw that specific grammatical subjects of unaccusatives and so-called passives have to move out of the PredP, the domain of existential closure. Similarly, specific objects move out of that domain to receive interpretation. Thus Persian exhibits both subject and object shift. It was further argued that Nom Case is checked by Agree between the verbal inflection and the specific DP in the Spec of vP. Similarly, φ-features of specific subjects were suggested to be checked in that position as well. The subject of an agentive verb is generated in this position by Merge, where it is checked for Nom Case and agreement. Thus Spec of vP could be considered an A-position, and subject shift would be an instance of A-movement. A similar situation holds for specific objects: they move into the lower Spec of vP for interpretation, where their Case is checked by Agree with v in that position. The suggestions advanced in this chapter are supported by the fact that the extracted object and the grammatical subject of unaccusatives and so-called passives can bind an anaphor from their derived position, namely the Spec of vP. The similarity between specific subjects and objects is further supported by extraction facts: extraction is possible out of subjects and objects, and both movements are restricted by the same constraint. It is conceivable to suggest, however, that object shift and subject shift do not exemplify instances of A-movement, since these two movements

Notes

117

apply primarily for the purpose of interpretation, and thus is different from a typical A-movement triggered by the need to satisfy EPP and Case. Furthermore, binding relations are not always established from that position (cf. 101) If this is the case, then Persian does not exhibit instances of A-movement. Moreover, EPPg is suggested to be satisfied morphologically in this language. This issue raises the question of whether the existence of pro and PRO is required. It was proposed that the existence of these Null elements is forced by Nom Case, and that the distribution of PRO is constrained by control. The presence of pro and PRO is supported by binding requirement of anaphors. The syntactic properties of Persian syntax discussed in this chapter are consistent with properties of topic-prominent languages. We will see in chapter 4 that the element extracted out of vP receives a discourse functional interpretation such as topic and focus. Thus, these movements should be considered as instances of A’-movement in a traditional sense. Furthermore, we suggested that T optionally selects the feature [+Topic], and thus the element in the Spec of TP is topic in this language. Whether or not T may optionally select the feature [+Topic] universally was briefly explored, and its possibility suggested.

Notes 36. This is comparable with Hungarian, a topic prominent language. Kiss (2002) shows that bare subjects remain in a VP-internal position in Hungarian. She further shows that they may appear VP-externally as contrastive topics in sentences that display a particular rise-fall intonation contour. 37. The claim that past participles have adjectival properties is evident in the fact that they modify nouns, as in (i) and (ii). (i) a. dar baste shod door closed became-3sg ‘the door became closed.’ b. dar-hâ-ye baste door-pl-Ez closed ‘The closed doors.’ (ii) a. sabzi poxte shod vegetable cooked became-3sg ‘The vegetable cooked.’ b. sabzi-ye poxte vegetable-Ez cooked ‘The cooked vegetable.’

118

Chapter 3: Local Scrambling and A-Movement

38. See Folli, Harley, and Karimi (in Press) for more detailed discussions of the internal structure of Persian CPr. See also chapter one in this volume for a brief discussion of these elements. 39. There is a subjectless construction in Persian that is utilized as passive. In these cases, the verb appears in the form of third person plural: (i) jom’e-hâ madrese-hâ-ro mi-band-an Friday-pl school-pl-râ dur-close-3pl ‘On Fridays, (they) close the schools.’ I assume that the subject is pro in these constructions, representing the impersonal one as in the English translation. Evidence comes from the fact that mardom ‘people’ can appear in the subject position in these constructions, replacing pro. (ii) mardom shab-â dar-â-ro mi-band-an people night-pl door-pl-râ dur-close-3pl ‘People close the doors at night.’ These constructions are similar to impersonal sentences in English: one closes the door at night. 40. Darzi (1996) suggests that Persian exhibits raising constructions. His analysis is based on the observation that the subject position of the matrix clause can be filled with the demonstrative in ‘this’ which he considers to be an expletive. In the presence of in, no embedded element can move into the matrix clause. (i) (in) lâzem ast [CP ke [Ali ketâb -râ be u be-dah-ad ]] this necessary is that A book-râ to him subj-give-3sg ‘It is necessary that Ali gives the book to him.’ (ii) (*in) Ali (*in) lazem ast [CP ke [ t ketâb-râ be u be-dah-ad (Darzi 1996: 93-94) Darzi suggests that the embedded subject has moved into the matrix subject position in (ii), causing the ungrammaticality of this sentence. See section 2.7.2. in this chapter where overt expletives are discussed. It is shown that these elements are absent in Persian. 41. The so-called raising verb may appear in a construction with no subordinate clause, as in (i). (i) bachche-hâ xaste be-nazar mi-yâ-n/*mi-yâ-d child-pl tired to view dur-come-3pl/*dur-come-3sg ‘The children seem tired.’ In this construction, bachche-hâ is the subject of the adjective phrase xaste which has moved into the Spec of vP, thus agreeing with the verb (see section 2.8. for analysis.) This type of construction is a true raising structure in English, as the translation of (i) indicates. 42. The clitic pronoun is optional in (31), while it is obligatory in possessor constructions. One explanation might be due to the general position of possessive pronouns in Persian: they are attached to the possessed, as in ketâb-am ‘my book’. This is not true in the case of direct objects since only in double clitic constructions the pronoun appears attached to the verb. 43. Note that the DP Rahjue in (44) and (45) can be interpreted as the object of the transitive version of the verb bordan ‘to win’. In that case, râ follows the object, and pro appears in the subject position. 44. Kiss (2002: 122) suggests that the matrix verb has an Acc Case in these situations. Since the Acc Case cannot be assigned to the embedded CP, it is assigned to the first available DP; that is, the extracted embedded subject which is passing through the Spec of the intermediate CP. This is somewhat different from the standard account suggested by

Notes

45.

46.

47.

48.

49.

50.

51.

119

Kayne (1980) and Chomsky (1981: 174) where it is claimed that the Acc Case is assigned by the matrix verb to the trace of the extracted subject in the Spec of the intermediate CP. The position taken in this work is different from both of those accounts since Acc Case is considered to be checked within the vP here (see section 3.4.). Thus the assumption is that the extracted embedded subject, on its way to the matrix clause, is optionally checked for Acc Case within the matrix vP. Tough constructions have been problematic for syntactic theories. Consider the following example where the subject position is filled by the external argument of the verb please (=John), suggesting that the subject position of this verb is a theta position. (i) Who did you say John pleases? The presence of the expletive in (53), however, contradicts this observation. A similar situation holds in Hungarian, as discussed by Kiss (2002): (i) Az nyilván való [CP hogy a tezst túl nehéz volt a diákok számára that-Nom obvious that the test too difficult was the students for ‘That is obvious that the test was too difficult for the students.’ (ii) Az [CP hogy a tezst túl nehéz volt a diákok számára] nyilván való that that the test too difficult was the students for obvious (Kiss 2002: 112) Kiss (2002: 112) states that “… the pronominal element associated with a clause is not a regular expletive, but represents the head of the complex NP.” This is the position taken in this work regarding Persian. The verb hastan is also used to imply emphasis. (i) man mo’allem hast-am I teacher be-1sg ‘I am a TEACHER.’ Kiss (2002) suggests that specific and nonspecific subjects occupy different positions in English as well. She shows that only the specific subject can be followed by sentence adverbials without a comma intonation. Nonspecific subjects must follow them. This is illustrated by the following examples: (i) *A baby boy was luckily born. (ii) Luckily, a baby boy was born. (iii) John was luckily born on time. (Kiss 2002: 117) Based on these data, the specific subject must be in a higher position than the Spec of TP. It has been argued that elements other than the subject can move into the Spec of TP to satisfy EPP. Considering Japanese, Miyagawa (2001, 2003) shows that objects may move into the Spec of TP to check the EPP feature of T. This movement is possible only in constructions where there is a V-to-T movement (see section 3.1.4. in chapter 2 for a brief discussion). Bailyn (2003) argues along the same lines. Based on binding relations, he suggests that local scrambling in Russian is triggered by EPP, which allows objects and prepositional phrases, in addition to subjects, to move to the Spec of TP. See section 1 in chapter 4 for some data. Karimi and Lobeck (1997) and Karimi (1999a) argue that specificity alone does not block extraction out of a DP. Karimi (1999a) shows that extraction out of a DP is blocked syntactically when the Spec of DP is filled by a determiner. Based on Persian and English, Karimi and Lobeck (1997) argue that extraction out of a specific DP is blocked syntactically by either a base-generated or derived lexical determiner, in a Spec-Head agreement relation. Thus, both of these works suggest a syntactic explanation for extraction, or lack of it, out of specific DPs. This proposal, if on the right track, suggests that extraction out of the subject is not possible in languages that have XP movement into the Spec of TP for the sake of EPPg.

120

52.

53.

54.

55.

Chapter 3: Local Scrambling and A-Movement This prediction is, in fact, borne out with respect to Finnish, since extraction out of the object is not possible in this language if it has moved into the Spec of TP to satisfy EPP (Anders Holmberg, personal communication). Japanese provides a problem for this assumption, since extraction out of the subject DP is possible in this language (Saito and Fukui 1998, among others), although EPP forces the subject (or the object) to move into the Spce of TP to satisfy EPP (Miyagawa 2001, 2003). The contrast between languages that allow extraction out of the subject and those that do not needs further research with respect to parametric differences between those languages, and a better understanding of the properties of A-movement, in general, and EPP, in particular. We leave this issue for future research. In the case of unaccusatives and so-called passives, where the grammatical subject is specific, Nom Case of the inflection is activated. That is, Nom Case is activated in the inflection when there is a subject DP in Spec of vP. Harley (2000) argues that the free distribution of PRO in Irish correlates with the absence of EPP in that language. Her analysis is based on McClosky’s (1996) claim that Irish lacks EPP for the following reasons: a. lack of expletives. b. the existence of the class of salient unaccusative verbs whose sole argument is a PP which remains within the VP. c. in non-finite clauses, the verb does not raise to T. Harley suggests that in an EPP-language, such as English and Icelandic, there are two types of EPP features: [Overt EPP] and [Null EPP]. The former appears on a finite T, and the overt subject must move into the Spec of TP in that case. The latter appears on a non-finite T. It is in this case that PRO appears in Spec of TP. In other words, Harley derives the presence of PRO in a EPP-language from Null EPP feature rather than Null Case. Irish, in contrast, lacks EPP. Therefore, PRO appears in tense and infinitive clauses in this language. Thus, Harley derives the free distribution of PRO in Irish from the lack of EPPg in this language. However, if EPPg is satisfied by Agreement in Irish, as suggested for Persian in this work, the distribution of PRO is explained without the need to assume a lack of EPPg in this language. In Karimi (2003c: 104-5), it is argued that the nonspecific object becomes part of the predicate by the rule of Syntactic Word Formation (SWF). Translated into the system developed in this volume, the nonspecific object is already part of the predicate by virtue of being generated inside the PredP, where it remains throughout the derivation (unless it is contrastively focused). It could be argued that the ungrammaticality of (101) is due to the fact that the indirect object is within a prepositional phrase, and hence cannot c-command the direct object. However, binding facts observed in German indicate that this is not the correct conclusion. First, consider the German example in (i) where the indirect object is bound by the direct object. (i) dass wir die Gâstei einanderi vorgestellt haben that we the guests-Acc each other introduced have ‘That we have introduced the guests to each other.’ (Müller and Sternefeld 1994: 351) German does not allow the direct object to be bound by the indirect object, although the latter is not embedded within a prepositional phrase in this language. (ii) *dass wir den Gâsteni einanderi vorgestellt haben that we the guests-Dat each other introduced have (Müller and Sternefeld 1994: 352) Thus (ii) shows the same effects we see in (101).

Notes

121

56. We will come back to this issue in chapter 7 where the properties of A and A’ movements are critically analyzed. 57. In some languages, both specific and nonspecific objects receive Case (Szabolcsi 1993, and Rado 1994 for Hungarian, Kim 1992 for Korean). It could be argued that Acc Case is morphological in those languages rather than structural. 58. Alexiadou and Anagnostopoulou (2001: 210) argue that subject shift and object shift are both driven by Case. This is somewhat different from the analysis advanced in this work, since the specific object and the theme (the grammatical subject) of unaccusatives and so-called passives are not interpreted in the position they are initially merged. Thus, subject shift and object shift are primarily for the sake of interpretation, rather than Case. 59. This definition of topic explains why a nonspecific object is not a natural candidate to occupy the topic position. See chapter 4 for discussion. 60. Topic prominent languages include Chinese (Li and Thompson 1976), Tagalog and related languages (Guilfoyle et al. 1992), Hungarian (E. Kiss 1995, 1997, 2002, 2003), Finnish (Holmberg 2000, Holmberg and Nikanne 2002), Japanese (Miyagawa 2001), and Greek and Spanish (Alexiadou and Anagnostopoulou 1998).

Chapter 4 Operator/Discourse Domain and A’-Scrambling

1. Introduction In this chapter we discuss the syntactic properties of a domain we called the operator/discourse domain in previous chapters. The tentative phrase structure provided in chapter 1 is repeated in (1) below. (1)

[CP [TopP [FP [TP [T’ [ [vP [v’ [XP [X’ ]] v ]]]]]]]] Operator/Discourse Phase

PredP

Lexical Phase The phrase structure in (1) suggests that all projections beyond vP represent operator or discourse phrases. This proposal includes the Spec of TP. The analysis advanced in chapter 3 suggests that Persian is a topic prominent language, and that the movement of the subject out of vP is not driven by Case or EPPg. It has been argued in the literature that Spec TP serves as an A-position for checking EPP (in the sense of Chomsky1995) in some topic-prominent languages, hosting not only subjects, but also objects and even prepositional phrases. This idea is proposed by Miyagawa (2001, 2003) for objects in Japanese and by Bailyn (2003) for objects and argument PPs in Russian. The examples in (2) are taken from Bailyn (2003). (2)

a. ???Svoji dom byl u Petrovyxi [self’s house]-Nom was at the Petrovs ‘The Petrovs had their own house’ b. U Petrovyx byl svoji dom at the Petrovs was [self’s house]-Nom ‘The Petrovs had their own house’ (Bailyn 2003: 160-1)

Introduction

123

The movement of the prepositional phrase into Spec of TP has provided the necessary requirement for anaphoric binding: the pre-posed element c-commands and binds the anaphor. Since binding is possible from an Aposition (see chapter 2), Bailyn suggests that Spec of TP must be an Aposition in Russian. Persian is not compatible with Russian in this respect. We argued in previous chapters that EPPg is satisfied by the verbal inflection in Persian, and that local scrambling is not a typical instance of A-movement. This conclusion includes movement into Spec of TP: this movement is not A-movement, evidenced by the anaphoric binding relations in the following data: (3)

a. Kimea [bachche-hâ-ro]i be hamdigei mo’arrefi kard K child-pl-râ to each other introduction did-3sg ‘Kimea introduced the children to each other.’ b. * Kimea [hamdigar-ro]i be [bachche-hâ]i mo’arrefi kard c. * be [bachche-hâ]i Kimea [hamdigar-ro]i mo’arrefi kard

We saw in chapter 3 that the specific direct object is interpreted at the edge of vP, and can bind into the indirect object from that position. This is evident by the example in (3a). The sentence in (3b) violates Principle A of the Binding Theory, which requires that an anaphor be locally bound by its antecedent. In this example, the reciprocal hamdige ‘each other’ is not bound by its antecedent bachche-hâ ‘children’. The example in (3c) shows that the movement of the indirect object into the Spec of TP does not establish the required situation for anaphoric binding, implying that Spec of TP is not an A-position61. In this chapter we suggest that Spec of TP is one of the two topic positions in Persian. We show that the CP domain, including the Spec of TP, is an operator/discourse domain, and that all XP-positions in this domain are interpreted as topic, focus, or quantified positions. Therefore, movement into these positions must be classified as A’-movement in a traditional sense (see chapter 2 on properties of A’-movement). We start our discussion with a brief analysis of adverb placement in section 2, followed by a proposal suggesting two topic positions in section 3. Distinct types of focus including properties of non-wh-focus, wh-arguments and wh-adjuncts are discussed in section 4. It is shown that the two types of wh-phrases have distinct internal structures, which lead to distinct syntactic behavior. Since the CP domain is an operator domain, we expect to see conflicting effects in this domain between scope bearing el-

124

Chapter 4: Operator/Discourse Domain and A’-Scrambling

ements such as wh-phrases and focused non-wh-phrases, and quantified phrases. This issue, and the optionality of focus movement, are discussed in section 4. We saw in chapter 1 that certain elements are either subject to limited scrambling or do not undergo scrambling at all. We will be in a position to explain the reasons for those restrictions in section 5. Concluding remarks follow in section 6.

2. Adverbs Adverb positions have been utilized for the purpose of identifying the location of arguments (Webelhuth 1984, 1992, Diesing 1992, among others). Jonas and Bobaljik (1993) and Bobaljik and Jonas (1996) use adverbs to identify the location of subjects in Scandinavian. These authors have assumed two subject positions in Icelandic, German, and one dialect of Faroese: one position precedes and one follows the adverb. Cinque (1999) divides adverbs into two major groups62. He suggests that “an ordered sequence of ‘higher’ (sentence) adverbs precedes an ordered sequence of ‘lower’ adverbs …” (Cinque: 1999: 16). That is, the order of one group cannot change with respect to the other: the higher adverb always precedes the lower one. Strict order for different types of adverbs has been also proposed by Holmberg (1993) for Finnish (per Svenonius 2002: 206). Consider the following examples. (4)

a. Pekka ei ehkä olisi aina valittanut siitä Pekka not perhaps would always complained about it ‘Perhaps Pekka wouldn’t have always complained about it.’ b. *Pekka ei aina olisi ehkä valittaut siitä Pekka not always would perhaps complained about it

In (4b) the sentential adverb perhaps follows the lower adverb always, rendering the sentence ungrammatical. Adopting Cinque’s terminology, the following data exemplify Persian High and Low adverbs. (5)

High (sentence) Adverbs amdan ‘intentionally’ xoshbaxtâne ‘fortunately’

Low (VP) Adverbs zirakâne ‘cleverly’ hamishe ‘always’

Adverbs

zâheran ‘apparently’ badbaxtâne ‘unfortunately’ ma’mulan ‘usually’ Ehtemalan ‘possibly’

125

âghelâne ‘wisely’ hanuz ‘still’, ‘yet’ hargez ‘never’ kâmelan ‘completely’ nâgahân ‘suddenly’ taghriban ‘almost’

The modal adverbials shâyad ‘may be’ and bâyad ‘must’, as well as temporal adverbs such as emruz ‘today’, are among the High adverbs. The combinations in (6) show that low adverbs may not precede the high ones, as suggested by Cinque: (6)

Ehtemâlan hamishe ‘possibly always’ xoshbaxtâne âghelâne ‘luckily wisely’ emruz kâmelan ‘today completely’

*hamishe ehtemâlan *âghelâne xoshbaxtâne *kâmelan emruz

Given the phrase structure proposed in (1), we suggest that the low adverbs are adjoined to vP. The following sentences exemplify low adverb placement in this language. (7)

Low adverb placement a. hanuz [vP [PredP jelo dar sag neshaste] still front door dog sitting Lit: still dogs are sitting in front of the door. b. [TP Jelo dar [ hanuz [vP [PredP t sag neshaste]]]] Lit: in front of the door, still dogs are sitting.

The sentence in (7a) is employed when the entire sentence is in focus. That is, the whole vP represents new information. In (7b), the prepositional phrase is topicalized. A similar situation holds for High adverb placement, as in (8). (8)

High adverb placement a. xoshbaxtâne [vP Kimea ketâb-â-ro [PredP be ketâbxune fortunately K book-pl-râ to library pas dâde]] return given-3sg ‘Fortunately, Kimea has returned the books to the library.’

126

Chapter 4: Operator/Discourse Domain and A’-Scrambling

b. [TP Kimea xoshbaxtâne [vP t ketâbâ-ro [PredP be ketâbxune pas dâde]]]] ‘As for Kimea, luckily (she) has returned the books to the library.’ The data in (9) show the combination of the two types of adverbs within a sentence: (9)

a. [TP Kimea xoshbaxtâne ma’mulan [vP movaffagh-e]] K luckily usually successful-be3sg ‘Luckily Kimea is usually successful.’ b. *Kimea ma’mulan xoshbaxtâne movaffagh-e

The High adverb may be placed in a higher position, as in (10). (10) a. xoshbaxtâne [TP Kimea âghelâne [vP kâr-ro anjâm dâd]] luckily K wisely job-râ end did-3sg ‘Lickily, Kimea wisely finished the job.’ b. xoshbaxtâne [TopP Kimea [TP kâr-ro âghelâne [vP anjâm dâd]]] ‘Luckily Kimea finished the job wisely.’ The High adverb takes scope over the part of the sentence that it c-commands. This is compatible with Jackendoff (1972), McConnell-Ginet (1982), and Ernst (1984), who suggest that adverb placement is driven by interpretation. Given the phrase structure in (1), the subject Kimea is in Spec-TP in (10a). In (10b), the subject is in Spec-TopP while the object is in Spec TP. Thus, the High adverb can be placed as high as in a position preceding the higher topic. Indefinites are not quantificational when inside vP, unless they are bound by a quantificational element such as an adverb (Diesing 1992, among others). They become quantificational when they move out of vP into the operator domain. In this case, they take scope over the elements they c-command. This is shown in (11). (11) a. hargez [vP ye dâneshju na-yumad] never one student neg-come-3sg ‘Not one student has ever come.’ b. ye dâneshju hargez [vP t na-yumad] one student never not-come-3sg ‘One (specific) student was such so that (she) never came.’

Adverbs

127

The subject is inside the vP in (11a), but it has moved into the operator domain in (11b). In this example, it takes scope not only over the adverb, but also over the negation63. Given the data thus far, we suggest the following positions for Persian adverbs. Hadv stands for High adverb and Ladv for Low adverb. (12) [CP (Hadv) [TopP (Hadv) [TP (Hadv) (Ladv) [ vP

]]]]

Before concluding this section, there is yet another type of adverb that needs to be introduced. These are the so-called ‘focus’ adverbs (Cinque 1999, Ernst 2002). The set of Persian focus adverbs consist of Faghat ‘only’, hattâ ‘even’ or hattâ … ham ‘even … too’. These elements are adjoined to all kinds of XPs (DPs, vPs, VPs (PredP), APs, PPs, and Adverb phrases). Examples with faghat ‘only’ are provided in (13)–(16). As we can see from these data, faghat takes scope either over all elements or only the elements it immediately c-commands64. (13) faghat Kimea bastani xarid only K ice-cream bought-3sg ‘Only Kimea bought ice-cream’ (no-one else did) or ‘What happened was that Kimea bought an ice-cream.’ (nothing else happened) (14) Kimea faghat bastani xarid K only ice-cream bought-3sg ‘Kimea bought only ice-cream’ (and nothing else) or ‘Kimea only bought ice-cream’ (that was all she did: buying ice-cram)65 (15) Kimea faghat barâ man bastani xarid K only for me ice-cream bought-3sg ‘Kimea bought ice-cream only for me.’ (and for nobody else) or ‘What Kimea did was buying ice-cream for me.’ (and she did nothing else) (16) Kimea faghat emruz bastani xarid K only today ice-cream bought-3sg ‘Only today Kimea bought ice-cream.’ Or ‘What Kimea did today was eating ice-cream.’ (and she did nothing else)

128

Chapter 4: Operator/Discourse Domain and A’-Scrambling

The only-phrase faghat will be discussed in section 4 (specifically in 4.1.2.) as part of the group called inherent identificational focus (in terms of Kiss 1998) 66. We continue with a discussion of two types of topic in the following section.

3. Topic positions A variety of positions have been suggested to represent the topic position in different languages. Müller and Sternefeld (1993), for example, suggested a type of substitution into a fixed clausal position (between CP and IP) for German and English. Baltin (1982), Johnson (1988), and Lasnik and Saito (1992) propose adjunction to IP for English. As for long-distance topicalization, Chomsky (1977) argues that this movement obeys the same island conditions as wh-movement. This seems to indicate that topicalization uses the Spec of CP as an escape hatch (See also Koster 1975). This restriction does not hold in Persian, since more than one topic is allowed to undergo long distance movement in this language, as in (17). (17) Ketâb-â-roi be Kimeak man fekr mi-kon-am pro ti tk dâd book-pl-râ to K I thought dur-do-1sg gave-3sg Lit: As for the books, to Kimea, I think she gave (them). The embedded direct and indirect objects have been topicalized in the main clause. The sentence is perfectly grammatical. The topic has also been discussed not only with respect to its position, but its semantic properties compared to other discourse elements such as focus. It is well-known that the topic represents old information or familiarity while focus reveals new information. Furthermore, it is generally assumed that the topic is not quantificational, and thus does not enter into scope interaction with quantified elements (Kiss 2003, among others). Moreover, it has been suggested that topic position, as opposed to focus position, is not unique in a sentence. That is, more than one topic may appear in the same clause (Kiss 1995, Svenonius 2002, among others). In this section we discuss two types of topics in Persian: Shifted Topic in the Spec of TopP, and Background Topic in a lower position, namely the Spec of TP. It has been argued that this type of distinction has also been found in the grammar of Mayan (Aissen 1992) and Russian (King 1995).

Topic positions

129

3.1. Shifted Topic Consider the following sentence: (18) ketâb-â chi shod? book-pl what became-3sg ‘What happend to the books.’ A natural answer to this question would be the following: (19) Kimea unâ-ro diruz be man dâd K they-râ yesterday to me gave-3sg ‘Kimea gave them to me yesterday.’ In this sentence, Kimea indicates a change of attention in discourse: the attention has shifted from books to Kimea. We employ the term Shifted Topic for this type of topicalization, and place the moved element in Spec-TopP.

3.2. Background Topic Consider the sentences in (18) and (19) once more. We have already established that Kimea is the Shifted Topic in (19). What is the role of unâro ‘them’ in this example? It refers to ketâb-â, and thus has been already in the discourse, and does not represent a shift of attention in discourse. We call this type of topic the Background Topic. That is, it has been in the discourse, and thus qualifies as a topic, but remains in the background. Additional examples are provided in (20a) (question) and (20b) (answer). (20) a. Kimea diruz ketâb-â-ro be Parviz dâd? K yesterday book-pl-râ to P gave-3sg ‘Did Kimea give the books to Parviz yesterday?’ b. na, RAHJUE unâ-ro diruz be Parviz dâd. No, R them-râ yesterday to P gave-3sg ‘No, RAHJUE gave them to Parviz yesterday.’ There are no structural yes/no questions in Persian. The interrogative sentence is identified by rising intonation (specified by a question mark in the written language). The sentence in (20a), therefore, has the same

130

Chapter 4: Operator/Discourse Domain and A’-Scrambling

syntactic structure as an indicative sentence. The DP Rahjue in (20b) represents contrastive focus (see section 4 of this chapter for discussion). That is, Rahjue in (20b) is contrasted with Kimea in (20a). Thus, it cannot be a topic. The DP ketâb-â was mentioned in the discourse before, and therefore, it is a Background Topic. We suggest that this type of topic is placed in Spec-TP. In the absence of a Shifted Topic, the Background Topic may appear in the initial position (Spec-TopP): (21) unâ-ro RAHJUE diruz be man dâd they-râ R yesterday to me gave-3sg ‘As for them, it was RAHJUE who gave them to me.’ In this example, the DP unâ, although a Background Topic with respect to the question in (20a), appears in Spec-TopP in the absence of a Shifted Topic.

3.3. [-Focus] feature versus EPPs Holmberg and Nikanne (2002) suggest that all arguments have the feature [+/-Focus]. The semantic effect of [+Focus] is information focus, and the semantic effect of [-Focus] is part of the presupposition or ‘ground’ in the sense of Vallduvi and Engdahl (1996). They further show that an element with the feature [-Focus] can be contrastively focused. This is true in Persian, as we will see in the next section. These authors argue that elements with the feature [-Focus] must move out of the predicate phrase (e.g. vP in our system). In fact, they claim that movement to Spec of TopP in Hungarian is triggered by the [-Focus] feature. Kiss (2002) disagrees with this assumption, since presupposed elements may remain in situ in Hungarian. This is true of Persian as well (cf. the discussions in chapter 3). Thus, elements with the feature [-Focus] do not have to be topicalized, as shown in (22). (22) hanuz Kimea xaste be nazar mi-yâd still K tired to view dur-come-3sg ‘Kimea still looks tired.’ In (22), the entire clause reveals new information. Thus, the DP Kimea, although [-Focus] inherently by virtue of being a name, remains inside

Focus

131

vP. This means that although presupposition is a required property for topic, it is not sufficient. Additional discourse-based properties, such as shift in topic or backgrounding, are necessary for a presupposed element to be topicalized. We have suggested thus far that movement in general is triggered by EPPs. This includes the movement of a DP into a topic position (SpecTopP or Spec-TP). The output of the syntactic derivation will be the input to the semantic component in the CP-phase, providing the topic interpretation.

4. Focus What is focus? Following Chomsky (1971, 1976) and Jackendoff (1973), Zubizarreta (1998) defines focus as the nonpresuppositional part of the sentence. She uses the traditional question/answer test to determine how a statement is partitioned in terms of focus and presuppostion. The examples in (23) can be paraphrased as in (24), and the latter can be represented in terms of existential quantification, presented in (25): (23) a. b. c. d.

What happened? What did John do? What did John eat? Who ate the pie?

(24) a. b. c. d.

Something happened. John did something. John ate something. Someone ate the pie.

(25) a. b. c. d.

There is an x, such that x happened. There is an x, such that John did x. There is an x, such that John ate that x. There is an x, such that x ate the pie.

In these examples, focus substitutes the wh-phrase while the rest of the sentence is presupposed.

132

Chapter 4: Operator/Discourse Domain and A’-Scrambling

4.1. Kiss’s definition of focus Kiss (1998) provides an elaborated definition of focus. For her, the questions and answers in (23) and (24) represent a type of focus she calls information focus. She distinguishes this type of focus from something she calls identificational focus. This is the type of focus that is also interpreted as contrastive focus in some languages. 4.1.1. Identificational focus versus information focus Kiss (1998) suggests that a. Identificational focus expresses exhaustive identification; information focus marks the nonpresupposed nature of the information it carries. b. Identificational focus is moved into the specifier position of a functional projection; information focus does not involve movement. c. Every sentence must have an information focus, but the presence of an identificational focus is not necessary in the sentence. Szabolcsi (1997) argues along the same lines for Hungarian. She states that a DP in a (derived) focus position receives an exclusion-by-identification interpretation, while a DP in situ does not. Turning to Persian, identificational focus is interpreted as contrastive focus in this language67. That is, the contrasted element is chosen out of a closed set of entities. The definition of contrastive focus intended in this work is provided in (26). (26) Contrastive focus operates on a closed set of entities whose members are known to the participants in the discourse. The example in (27) illustrates a case of contrastive focus in Persian. (27) halâ barâ KIMEA pro nâme mi-nevis-an now for K letter hab-write-3sg ‘Now it is for KIMEA they are writing letters .’ (as opposed to someone else.) The DP Kimea bears heavy stress, and receives contrastive reading. In the absence of a heavy stress, the DP in situ receives an interpretation revealing information focus. In that case, the translation of (27) would be something like now they are writing a letter for Kimea.

Focus

133

The DP denoting contrastive focus does not have to be specific, as the following example illustrates: (28) pro barâ Kimea ye KETAB xarid-am for K a BOOK bought-1sg ‘It was a book that I bought for Kimea.’ (as opposed to a shirt) The indefinite noun phrase ye KETAB ‘a book’ is contrastively focused in (28). Even predicative bare nominals can function as contrastive focus in Persian. The reason is that nonspecific elements, when in focus, are interpreted as names of properties. In these cases, the inherently nonspecific element denotes a referent that is a member of a relevant set, present in the domain of discourse68. (29) pro barâ Kimea KETAB xarid-am for K BOOK bought-1sg ‘It was books that I bought for Kimea.’ (not pants) The predicative bare nominal is a member of a closed set in this example, and is contrasted against the other members of that set. 4.1.2. Inherent focus Kiss (1998) suggests that certain elements carry an inherent focus feature: only-phrases and wh-phrases belong to this group. We saw the distribution of faghat ‘only’ in section 2, where we discussed the focus adverbs. These phrases receive exhaustive focus interpretation: (30) a. faghat Kimea se tâ film dide Only K three part film seen is ‘It is only KIMEA (not anyone else) who has seen three movies.’ b. Kimea faghat se ta film dide ‘Kimea has seen only THREE movies (not four or five).’ Persian allows two elements bearing contrastive focus in the same sentence only if at least one of them has an inherent focus feature: (31) a. KIMEA mâh-e gozashte faghat se-tâ film dide K month-Ez previous only three-part movie saw-3sg

134

Chapter 4: Operator/Discourse Domain and A’-Scrambling

‘It was KIMEA who saw only three movies last month.’ (everyone else has seen more movies.) b. *KIMEA se tâ FILM did K three -part FILM saw-3sg One of the focal elements in (31a) has an inherent focus feature represented by faghat ‘only’. In contrast, the two focused elements in (31b) lack an inherent feature, and thus the sentence is ill-formed. The ungrammaticality of (31b) supports Kiss (1998), who suggests that identificational focus is exhaustive (unless inherent). Wh-phrases solicit new information, and therefore, they must have an inherent focus feature. Persian does not exhibit obligatory wh-movement, although these phrases are subject to scrambling. If scrambled, they force an answer that denotes a contrastive focus. We discuss wh-constructions and their movement in section 4.2.2.

4.2. Focus-movement Movement that represents discourse function is not a new idea, and goes back to the early seventies (Jackendoff 1973, Culicover and Rochemont 1983, and Rochemont 1986). Some authors have considered focus movement as an instance of Move triggered by a discourse-functional feature (Miyagawa 1997; Karimi 1999c, 2003d; Bailyn 2001; Kiss 2003, among others). It has been argued that focus movement places the focused element in the Spec of a focus phrase (Kiss 1994, 1998, 2003; Szabolcsi 1997; Rizzi 1997). This is illustrated in (32). (32)

FocP

Foc' Foc

XP

This position is argued to be unique, implying that there is only one focus phrase per sentence (Kiss 1998, Rizzi 1997). We saw that contrastive focus must be unique unless there is an inherent focus. We will see in section

Focus

135

4.2.2. that more than one contrastive wh-phrase is possible, as long as they are both in Spec of FocP. We start with an examination of focus movement of non-wh-phrases in 4.2.1., and extend the discussion to focus movement of wh-phrases in 4.2.2. The summary of this section appears in 4.2.3. 4.2.1. Non-wh-phrases In Persian, the contrasted element bears a heavy stress, and optionally moves into the Spec of FocP, as in (33). (33) a. man DIVAN-E HAFEZ-o barâ Kimea xarid-am I divân-Ez Hafez-râ for K bought-2sg b. [FocP DIVAN-E HAFEZ-oi [man ti barâ Kimea xarid-am]] ‘It was DIVAN-E HAFEZ that I bought for Kimea.’ (not something else) (34) a. man be KIMEA ye ketâb dâd-am I to K a book gave-1g b. [FocP be KIMEAi [man ti ye ketâb dâd-am ]] ‘It was to KIMEA that I gave a book.’ (not to someone else) DIVAN-E HAFEZ in (33a) and (33b) receives a contrastive reading since it denotes contrast with respect to other members of a set of relevant objects. Similarly, KIMEA in (34a) and (34b) contrasts with someone else, or a set of relevant people. As we see, however, the movement of the contrasted element is optional. The examples in (33) and (34) are natural reactions to the statements in (35a) and (35b), respectively. (35) a. pro shenid-am to divân-e Saadi-ro barâ Kimea xarid-i heard-1sg you divân-Ez Saadi -râ for K bought-2sg ‘I heard you bought Saadi’s Divân for Kimea.’ b. pro shenid-am to be Sasan ye ketâb dâd-i heard-1sg you to S a book gave-2sg ‘I heard you gave a book to Sasan.’

136

Chapter 4: Operator/Discourse Domain and A’-Scrambling

DIVAN-E HAFEZ ‘Havez’s Divan’ in (33) is contrasted with divan-e Sa’di ‘Sa’di’s Divan’ in (35a). Similarly, be KIMEA ‘to Kimea’ in (34) is contrasted with be Sasan ‘to Sasan’ in (35b). We saw in this section that contrasted elements are heavily stressed in Persian, and optionally move into the Spec of the FocP. The optionality of focus movement is discussed in section 4.4. in this chapter. 4.2.2. Wh-phrases Persian does not exhibit obligatory single wh-movement comparable to English, nor obligatory multiple wh-movement observed in Bulgarian. Nevertheless, wh-arguments may undergo optional movement individually or multiply in this language. Similarly, the sentential wh-adjunct cherâ ‘why’ and the wh-adjunct chetor(i), similar to other kinds of wh-adjuncts such as kojâ ‘where’ and key ‘when’, may appear in different positions in Persian clauses. We start our discussion with some data, followed by a brief review of previous literature on this issue. Karimi and Taleghani (2003) (K&T henceforth) have examined wh-constructions in Persian. In agreement with their analysis, we review the essence of their discussion, followed by supporting evidence they present. The summary of this discussion concludes this subsection. Data The wh-argument can remain in situ, as in (36a), and is optionally moved, as in (36b) and (36c). (36) a. Kimea diruz [vP ketâb-ro be ki dâd]? K yesterday book-râ to who gave ‘Who did Kimea give the book to yesterday?’ b. [TopP Kimea [FocP be ki [TP diruz [vP ketâb-ro t dâd]]]] c. [FocP be ki [TP Kimea [diruz [vP ketâb-ro t dâd]]]] ‘Who was it that Kimea gave the book to yesterday?’ As observed by Karimi (1999c), the wh-phrase in (36a) receives an information focus interpretation. The answer to this question is something like ketâb-ro be Parviz dâd ‘she gave the book to Parviz.’ The wh-phrase in (36b) and (36c) has moved into the Spec of FocP, and receives a contrastive interpretation (Karimi 1999c, 2003d). This question occurs when the speaker has a set of people in mind, and wonders which one was the

Focus

137

receiver of the book. The topic phrase Kimea precedes the FocP in (36b), while it follows it in (36c). Given the phrase structure in (1), the subject DP is in Spec-TopP in (36b) and in Spec-TP in (36c). The distinction we observed in (36) is more apparent in the following contrast. (37) a. pro fekr mi-kon-i [(ke) emshab [vP ki be restorân thought dur-do-2sg that tonight who to restaurant bi-yâd] subj-come-3sg ‘Who do you think will come to the restaurant tonight?’ b. [FocP ki pro fekr mi-kon-i [(ke) emshab t barâ shâm who thought dur-do-2sg that tonight for dinner bi-yâd] subj-come-3sg ‘Who is it you think will come for dinner tonight?’ According to K&T’s informants, the first sentence is an appropriate utterance when the speaker and hearer are going to a restaurant, and they wonder who they might see there without having a specific set of restaurant-goers in mind. The one in (37b) is appropriate when the speaker and the hearer have invited a set of guests, and they wonder which people out of that set will show up. Relevant to our discussion is the fact that the wh-phrase follows the complementizer ke ‘that’ in (38), indicating that the wh-phrase is not in the Spec of CP. (38) pro fekr mi-kon-i [ke ki-ro Kimea tu mehmuni thought dur-do-2sg that who-râ K in party be-bin-e subj-see-3sg Lit: you think who is it that Kimea will see at the party? More than one argument wh-phrase can be fronted, as in (39b). (39) a. to fekr mi-kon-i (ke) ki bâ ki be-raghs-e you thought dur-do-2sg that who with who subj-dance-3sg ‘Who do you think will dance with whom?’ b. [TopP kii bâ kij [TP to [vP fekr-mi-kon-i [CP (ke) ti tj who with who you thought-dur-do-2sg that

138

Chapter 4: Operator/Discourse Domain and A’-Scrambling

be-raghs-e] subj-dance-3sg Lit: ‘Who with whom is it that you think will dance.’ Again, the fronted wh-phrases receive a contrastive interpretation. This occurs in a context when the speaker has a specific set of people in mind. The answer to this question is something like: it will be the case that Kimea will dance with Rahjue, and Papar with Arezu, as opposed to other possible combinations of the members of the set. The two fronted wh-phrases cannot be separated by another element, as in (40). (40) b. ??kii emruz bâ kij pro fekr-mi-kon-i [CP ti tj who today with who thought-dur-do-2sg be-raghs-e] subj-dance-3sg (Karimi 2003d) The sentence in (40) is awkward at best. The fact that the two fronted whphrases cannot be separated supports Richards’ (1997, 2001) claim that there are multiple Spec positions in multiple wh-fronting languages, such as Bulgarian69. Turning to wh-adjuncts, these elements appear in different positions as well. (41) a. Kimea in mâshin-ro chetori dorost karde K this house-Acc how right did-3sg b. Kimea chetori in mâshin-ro dorost karde c. chetori Kimea in mâshin-ro dorost karde ‘How has Kimea fixed this car?’ The wh-adjunct chetori ‘how’ can be placed in different positions. In each case, it takes scope over that part of the sentence it c-commands. As mentioned before, this is compatible with Jackendoff (1972), McConnell-Ginet (1982), and Ernst (1984), who suggest that adverb placement is driven by interpretation. The following sentences show that cherâ ‘why’ may also appear in different positions. (42) a. pro fekr mi-kon-i (ke) Kimea cherâ in xuna-ro dar thought dur-do-2sg that K why this house-râ in

Focus

139

Tusân sâxte Tusân built-3sg b. pro fekr mi-kon-i (ke) cherâ Kimea in xuna-ro dar Tusân sâxte c. cherâ fekr mi-kon-i (ke) Kimea in xuna-ro dar Tusân sâxte ‘Why do you think Kimea built this house in Tucson?’ The adjunct cherâ ‘why’ appears in the embedded clause in (42a&b), and has scope only in that clause. The sentence (42c) is ambiguous, indicating that the wh-adjunct must have originated in the embedded clause, and has moved into the matrix clause. This wh-adjunct is subject to obligatory movement from its base position, as noted by Kahnemuyipour (2001). (43) a. Ali bâ Maryam ezdevâj kard [chon dust-esh dâsht] A with M marry did because friend-her had-3sg ‘Ali married Maryam because he loved her.’ b. *Ali bâ Maryam ezdevâj kard [cherâ] A woth M marry did3sg why c. Ali cherâ bâ Maryam ezdevâj kard A why with M marry did-3sg ‘Why did Ali marry Maryam?’ (Kahnemuyipour 2001) The wh-adjunct originates in the postverbal position, and moves into the matrix position in order to take scope over that sentence. Previous accounts Raghibdust (1994) suggests that wh-phrases in Persian undergo topicalization. Cheng (1991) argues along the same lines with respect to Egyptian Arabic. However, wh-movement has been vastly considered as an instance of focus movement in the literature. This analysis can be traced back to the observation that a wh-element is inherently focused (Rochmont 1978, 1986; Horvath 1986; Bresnan and Mchombo 1987; Kiss 1998, among others). If wh-phrases are focal inherently, then there is some problem with an analysis that considers wh-movement as an instance of topicalization for the following reasons: – Topic implies familiarity/old information, while focus represents new information. – Topic lacks quantificational reading, and thus can optionally appear in a Clitic Left Dislocation (CLLD) construction, focus cannot.

140

Chapter 4: Operator/Discourse Domain and A’-Scrambling

(44) ketâb-ro diruz Kimea xarid-(esh) book-râ yesterday K bought-it ‘The book, Kimea bought it yesterday.’ (45) ki-ro to diruz did-i-*sh who-râ you yesterday saw-2sg-her ‘Who was it you saw yesterday?’ (46) KETAB-ro diruz Kimea xarid-*esh ‘It was the book that K bought yesterday.’ The contrast we observe in these examples is due to the fact that the topicalized element in (44) is not quantificational, while the focal elements in (45) and (46) are70. The following example supports the idea that quantificational elements are not compatible with CLLD constructions: (47) har filmi-ro yekbâr pro did-am-*esh each movie-râ once saw-1sg-it ‘I saw each movie once.’ The quantified element har filmi ‘each movie’ cannot bind the pronoun clitic -esh in (47). Analysis We saw that wh-phrases do not undergo movement to Spec of CP and are not topicalized. The questions K&T address at this point are these: – what is the nature of the two types of wh-phrases, and – how does the sentence receive a question interpretation in the absence of a wh-phrase in the Spec of CP? In order to answer these questions, K&T first provide a proposal regarding the internal structures of the two types of wh-phrases. This proposal shows that wh-arguments have a D-head, while wh-adjuncts lack this nominal head, and are purely quantificational. Furthermore, the head of wh-arguments has an inherent focus feature. Wh-adjuncts lack this feature. Finllay, both heads have a wh-feature. These proposals are presented in (48).

Focus

(48)

wh-argument a.

wh-adjunct

DP

b.

QP

D' D+Foc/+wh

141

Q' wh-word

Q+wh

wh-word

K&T observe the following similarities between the two types of whphrases: – There is a wh-operator in the Spec of CP in a wh-interrogative sentence (Aoun and Li 1993)71. This is illustrated in (49). (49)

[CP OP [

….

wh ….. ]]

– The wh-feature moves up to C to be in an Agree relation with the operator72. This proposal suggests that the sentence becomes ungrammatical if the feature movement is blocked. This prediction is borne out, as we will see later in this section. Furthermore, K&T suggest the following properties for wh-arguments: – It is the structural position that gives the wh-argument a contrastive reading. – The empty D-head can be filled by a quantified element in which case it looses its nominal property, and becomes purely quantificational. This is shown in (50): (50) Ki ‘who’ yeki ‘someone’ Chi ‘what’ yechizi ‘something’ Kojâ ‘where’ yejâi/yekjâi ‘somewhere’

hichki ‘no-one’ hichchi ‘nothing’ hijjâ ‘nowhere’

– In the case of D-linked wh-arguments such as kodum ketâb ‘which book’, che-kasi ‘which person’, and che chizi ‘which thing’, K&T suggest that the D-head is filled by a wh-operator representing contrast in these cases. Thus, D-linked wh-arguments are inherently contrastive. This type of wh-argument, similar to only-phrases, does not have to move into the Spec of FocP. Its appearance in different positions is de-

142

Chapter 4: Operator/Discourse Domain and A’-Scrambling

pendent on its scope over the rest of the clause, as shown in (51a) and (51b). Similar to other types of quantified phrases, they cannot appear in a CLLD construction, as in (51c). (51) a. Kimea kodum ketâb-ro emruz be to dâd? K which book-râ today to you gave ‘As for Kimea, Which book was it that Kimea gave you today?’ b. kodum ketâb-ro Kimea emruz be to dâd? ‘Which book was it that Kimea gave to you today?’ c. kodum ketâb-ro to emruz xarid-i-*sh? Which book-râ you today bought-2sg-it ‘Which book did you buy today?’ As (48b) suggests, wh-adjuncts have an empty quantifier head, and lack a D-head. Lack of a nominal head in these phrases is evident by the fact that they cannot appear in a cleft construction. (52) a. un ki bud ke har ruz injâ mi-âmad that who was that every day here dur-came-3sg ‘Who was it that used to come here every morning?’ b. *un chetori/cherâ bud ke Kimea mashin-ro dorost kard that how was that K car-Acc fix did ‘*how/why was it that he fixed the car?’ English makes the same distinction with respect to wh-arguments and wh-adjuncts. (53) a. Who was it that she loved so much? b. What was it that she bought yesterday? (54) a. *How was it that he fixed the car? b. *Why was it that he left his job? Obviously, the sentence in (54b) is ill-formed as a cleft construction, but fine otherwise. That is, ‘why’ cannot be interpreted as being extracted out of the embedded phrase. The summary of the properties of the wh-phrases is shown in (55).

Focus

(55) Wh-adjunct Wh-argument in situ Spec/FocP D-linked wh-phrase

Q + + + +

D – + + +

143

Informational contrastive – – + – – + – +

The element in Spec of FocP and the D-linked wh-phrases have the same properties. All wh-phrases are quantificational by virtue of the movement of their wh-feature to C. Supporting evidence Thus far the following claims have been made in this section: A. The extracted wh-argument is in Spec of FocP, bearing contrastive focus interpretation. B. There is a wh-operator in the Spec of CP, and C. The wh-feature moves to C for a local Agree relation with the whoperator K&T offer the following pieces of evidence in support of their analysis. We start with the claim in A. A. Evidence for Spec of FocP as the landing site for wh-arguments Sabel (2000) suggests that wh-movement is triggered by the need to check not only the wh-feature, but also the focus feature. The following examples, taken from Kikuyu, indicate that the wh-phrase may remain in situ, as in (56a), or move to the clause initial position, as in (56b). (56) a. kama..ú a- ɔ´ n-! ´irε´ o? Kamau SP-see-T who? (SP = Subject prefix, T = Tense/Aspect affix) ‘Who did Kamau see?’ b. nóo o-t εm- ´irε´ mo-te? FP-who PP-cut-T CP-tree (FP = Focus particle, PP = pronominal prefix) ‘Who cut a tree?’ (Sabel 2000: 429) The focus particle appears on the extracted wh-feature, but not on the one in situ. Within the system proposed here, the focus particle must be in Spec of FocP, and is attached to the wh-argument after it moves into that position in (56b).

144

Chapter 4: Operator/Discourse Domain and A’-Scrambling

B. Evidence for the existence of a wh-operator in the Spec of CP In Dari, a variant of Persian spoken in Afghanistan, the wh-operator may surface overtly as a scope marker. This is shown in (57)–(59)73. The overt wh-phrase in the matrix clause marks the scope of the wh-phrase in the embedded clause. (57) chi fekr mi-kon-i [u ki-râ did] ? what thought dur-do-2sg she who-râ saw ‘Who do you think she saw?’ Lit: what do you think who she saw? (58) chi feker mi-kon-i [ u cherâ raft]? what thought dur-do-2sg he why went ‘Why do you think he left?’ Lit: what do you think why she left? (59) chi feker mi-kon-i [u chetor az zendaan faraar kard]? What thought dur-do-2sg he how from jail escape did-3sg ‘How do you think he escaped from prison?’ Lit: what do you think how he escaped from prison. Similar cases have been observed in other languages. Turano (1995) observes that wh-phrases move into the focus position in Albanian. The following examples, exhibiting this property of Albanian, are taken from Sabel (2000). The wh-phrase has moved into the Spec of FocP in its own clause in (60a), while it has moved farther up into the Spec of FocP in the higher clause in (60b). In both cases, there is a scope marker in the Spec of CP in the matrix clause, allowing the wh-phrase to receive a wide scope. (60) a. [CP A mendon [CP se Maria thotë [CP se çfarei ka Q you-think that M says that what has sjelle burri ti ]]] brought her-husband ‘What do you think that Mary says her husband has brought?’ b. [CP A mendon [CP se çfarei Maria thotë [CP se ka Q you-think that what M says that has sjelle burri ti]]] brought her-husband (Sabel 2000: 430) These examples are compatible with the analysis advanced here, which states that there is a wh-operator in Spec of CP: the empty operator shows up as a lexical element in Albanian.

Focus

145

Further evidence for the presence of a scope marker comes from German and Romani. McDaniel (1989) argues that the wh-phrase moves in these two languages, but not necessarily all the way to Spec of CP. However, the Spec/CP must be filled by the scope marker was in German and so in Romani when there is partial wh-movement. The ‘a’ sentences in (61) and (62) represent full wh-movement in German and Romani, while those in ‘b’ exemplify partial wh-movement with a scope marker in the Spec of CP. The scope marker in these languages is the overt realization of the wh-operator in Persian. (61) a. [mit wem]i glaubt [IP Hans [CP ti dass [IP Jakob jetzt ti with who believes Hans that Jakob now spricht]]]? speaks ‘With whom does Hans think that Jakob is now talking?’ b. wasi glaubt [IP Hans [CP [mit wem]i [IP Jakob jetzt ti what believes Hans with whom Jakob now spricht ]]]? talks  (62) a. kasi [IP o Demiri mislinol [CP ti so [IP i Ar´ifa dikhl a ti ]]]? ‘Whom does Demir think that Arifa saw?’  b. soi [IP o Demiri mislinol [CP kasi [IP i Ar´ifa dikhl a ti ]]]? (McDaniel 1989: 568) Note that so is homophonous with the complementizer in Romani. C. Evidence for the movement of the wh-feature to C As mentioned above, K&T propose that the wh-feature must move to C to be in a local Agree relation with the operator. This proposal is supported by three pieces of evidence. First, in a multiple wh-construction, we always have a pair-listing interpretation, as noted by Lotfi (2003): (63)

Ki chi-ro xarid who what bought-3sg ‘Who bought what?’

(64) ki bâ ki bâzi kard who with whom play did-3sg ‘Who played with whom?’

146

Chapter 4: Operator/Discourse Domain and A’-Scrambling

The answer to (63) is something like: ‘Kimea bought a dress, Rahjue a hat, and I bought Chomsky’s book on 9–11’. A single-listing interpretation is not an option, indicating that all wh-features must move to C. The second piece of evidence is provided by extraction out of adjuncts. In the following example, the movement of the wh-feature is blocked in C. This feature cannot move out of the adjunct CP. (65) *[CP Op [ C [ Parviz na-raghsid [CP [ chonke C [ ki unjâ bud]]] *

P

neg-dance-3sg

because

who there was

Note that (65) is grammatical if the embedded clause is interpreted as an echo question. Finally, it is well-known that wh phrases cannot stay in the domain of certain scope-bearing elements, including Negative Polarity Items (NPI). The following contrast shows that the movement of the wh-phrase out of the domain of the NPI rescues the ungrammatical sentence: (66) a. *[CP OP [hichkas chi-ro na-xarid] nobody what-râ neg-bought b. [CP OP [FocP chi-roi hichkas ti na-xarid ‘What was it that no one bought?’ The NPI intervenes between the wh-operator and the wh-phrase in (66a), blocking the interrogative interpretation. Note that this sentence is fine as an echo question. The wh-argument has moved into the Spec of FocP in (66b), rescuing the ungrammatical sentence. The same blocking effects are observed when an adjunct is in the scope of negation, as in (67). Again, the movement of the wh-adjunct to a position outside the scope of negation rescues the sentence. (67) a. *pro fekr mi-kon-i hichkas cherâ na-yâmad thought dur-do-2sg no one why neg-came-3sg b. pro fekr mi-kon-i cherâ hichkas na-yâmad ‘Why do you think no one came?’ In order to explain the ungrammaticality of (66a) and (67a), the phrase structure in (1) needs to be revised to include the Neg phrase.

Focus

147

(68) [CP [TopP [FocP [NegP Neg [TP [vP PredP ]]]]]] Quantifier phrases are merged in different positions in order to have scope over the c-commanded domain. The quantifier phrases in (66a) and (67a) are in a higher position than the wh-phrase. This is illustrated in (69). Irrelevant details are omitted. (69) [CP OP C [NegP Neg [hichkas [TP wh …….. In order for the wh-feature to be in an Agree relation with the operator in the Spec of CP, it needs to move to C. In (69), the head of the negative quantifier phrase is intervening. The licensing of the NPI by Neg blocks the movement of the wh-feature through this head, as illustrated in (70). (70) [CP OP C [NegP Neg [ hichkas Q [TP [vP wh …….. *

As suggested by K&T, the configuration in (70) violates MLC, stated in chapter 1, and restated in (71) 74: (71) Minimal Link Condition (MLC) K attracts α only if there is no β, β closer to K than α such that K attracts β. (Chomsky 1995: 311) That is, the head feature of the NPI and the wh-feature of the wh-phrase, two scope-bearing elements, compete for the same position. The NPI head is closer, and therefore, the movement of the wh-feature is blocked. The movement of the wh-phrase into a higher position allows the wh-feature to move to C. In these cases, the wh-argument is in the Spec of FocP. (72) [CP OP C [FocP wh [NegP Neg [QP hichkas The wh-adjunct in (67b) is in a higher position than the NegP as well, as shown in (73)75. (73) [CP OP C [ wh [NegP Neg [QP hichkas

148

Chapter 4: Operator/Discourse Domain and A’-Scrambling

The restriction discussed thus far is observed with respect to LDS: (74) a. kii pro fekr

mi-kon-i ti hich film-i-ro

na-dide

who thought dur-do-2sg no movie-ind-râ neg-seen-3sg ‘Who do you think has not seen any movie?’ b. *hich film-i-roi fekr mi-kon-i ki ti na-dide

(75) a. *pro fekr mi-kon-i hichkas kodum film-ro thought dur-do-2sg nobody which movie-râ na-dide neg-seen-3sg b. kodum film-roi pro fekr mi-kon-i hichkas ti na-dide ‘Which movie do you think nobody has seen?’ The contrast in (74) represents a case where the application of scrambling to the NPI destroys the structure. The one in (75) represents a case where the application of scrambling to the wh-phrase rescues the sentence. Similar blocking effects have been observed in other languages. If this analysis is on the right track, it holds for those languages as well. The following data are taken from Korean and Japanese. Korean (76) a. *Amuto muos-ul sa-chi anh-ass-ni anyone what-Acc buy-CHI not do-Past-Q b. muos-ul Amuto sa-chi anh-ass-ni ‘What did no one buy?’ (Beck & Kim 1997: 339) (77) a. *Hanako-sika nani-o yoma-nai no? Hanako-only what-Acc read-neg Q b. Nani-o Hanako-sika yoma-nai no? (Tanaka 1997, per Pesetsky 2000) Hindi and Turkish reveal similar properties.

Focus

149

Hindi (78) a. ??Koi nahiiN kyaa paRhaa anyone not what read-Perf.M b. kyaa koi nahiiN paRhaa what anyone not read-Perf.M ‘What did no one read?’ (Beck and Kim 1997: 377) Turkish (79) a. *kimse kimi görmedi? anyone who-Acc see-Neg-Past b. kimi kimse görmedi? who-Acc anyone see-Neg-Past ‘Whom did nobody see?’

(Beck and Kim 1997: 380)

German has full and partial wh-movement, as we saw before. However, in the case of multiple wh-constructions, only the first one is fronted. The presence of the NPI in the higher position than the wh-in-situ renders the sentence ungrammatical. This shows once again that the NPI must be responsible for blocking the movement of the wh-feature. Consider the contrast in (80). German (80) a. *Wer hat niemanden wo angetroffen? who has no one where met b. Wer hat wo niemanden angetroffen? ‘Who didn’t meet anybody where?’ (Beck and Kim 1997: 340) 4.2.3. Summary K&T’s analysis shows that wh-arguments and wh-adjuncts do not have the same internal structure, and therefore, reveal different syntactic behavior. While the wh-argument has a D-head, the wh-adjunct lacks a nominal head, and is purely quantificational. Furthermore, the wh-argument may stay in-situ and receive a reading representing information focus. Alternatively, it may move into the Spec of FocP, where it receives contrastive reading. Wh-adjuncts move as well, but their land-

150

Chapter 4: Operator/Discourse Domain and A’-Scrambling

ing position is not the FocP. Due to their purely quantificational property, their derived position must be the same as sentential adverbs. This analysis shows that wh-adjuncts lack a focus feature, and thus are neutral with respect to information/contrastive focus interpretation. It was also shown that the two types of wh-phrases share one important property: the wh-feature must be in a local Agree relation with an operator in Spec of CP, thus this feature must move to C. This proposal suggests that intervening elements must block the movement of the wh-feature to C, a prediction that was shown to be borne out.

4.3. Focus-movement and EPPs Karimi (1999c) suggests that it is the focus feature that triggers the movement of the wh-phrase as well as the non-wh-phrase in a scrambling language such as Persian76. This assumption is not correct for the following reasons. First, as we saw in the previous section, wh-adjuncts lack a focus feature. However, they also move due to the fact that their trace enters scope marking (cf. 41). Second, if the focus feature were responsible for movement, we would need two different types of focus with respect to wh-arguments: one representing information focus and one representing contrastive focus which is not an economical solution. In Karimi (2003d), a three-way feature selection was suggested for contrastive focus: F, F’, and Fs. F represented the feature that triggers movement, and resides in the head of FocP, F’ was the semantic feature representing contrastive focus, and Fs represented stress. The combination of F’ and Fs provided the contrastive focus. The selection of F triggered movement. This system suffers from problems similar to those mentioned in the previous paragraph: It cannot account for the movement of wh-adjuncts, and predicts the existence of two different focus features in the case of wh-arguments: the informational focus feature and the contrastive focus feature77. Furthermore, this proposal cannot account for the fact that movement is obligatory in the case of contrasted wharguments. That is, it is the structural position that provides the contrastive reading for these phrases, while this is not so in the case of contrasted non-wh-phrases. Contrasted non-wh-phrases are heavily stressed, and their movement is optional (see section 4.4.), although they receive a contrastive interpretation in situ. This contradictory outcome is obviously not welcome in grammar.

Focus

151

We suggest, instead, that the syntactic tool responsible for all movements, the EPPs, is responsible for focus movement. Only if this feature is selected, movement applies. In the case of wh-phrases, the movement to the Spec of FocP, triggered by EPPs, provides the contrastive reading. In the case of non-wh-phrases, it is the stress that provides the contrastive reading rather than movement (however, see section 4.4.1., where the movement of the focus feature of contrasted non-wh-phrases is discussed). Given the proposal in this section, the focus feature will not be responsible for XP movement in either case. Reducing the feature responsible for all kinds of movements to EPPs is economically and empirically a welcome result. If focus movement is triggered by EPPs, it must be subject to syntactic principles and conditions. This is in fact borne out: Superiority, as an instance of MLC, must be observed in multiple wh-constructions. Violation of superiority renders the sentence ungrammatical, as in (81b). That is, of two elements bearing the wh-feature, EPPs may move only the higher one78. (81) a. Ki chi-ro xarid who what-râ bought-3sg ‘Who bought what?’ b. *chi-ro ki xarid what -râ who bought The sentence in (81b) is grammatical if the wh-phrase in situ bears no stress at all. In that case, it is interpreted as an indefinite DP (similar to someone in English): what is it that someone will buy? How do wh-arguments interact with wh-adjuncts? The movement of the wh-adjunct over the wh-argument renders the sentence ungrammatical, as in (82). Stressing the wh-argument does not save the structure. (82) a. b. c. d.

*cherâ ki/KI raft *chetori ki/KI raft *kojâ ki/KI raft *key ki/KI raft

The data in (82) support the claim that wh-movement is feature-driven, and therefore, subject to the Superiority Condition (subsumed under MLC).

152

Chapter 4: Operator/Discourse Domain and A’-Scrambling

The interaction of wh-arguments with wh-adjuncts reveals some other interesting properties. Consider the data in (83). (83) a. ki cherâ who why b. ki chetori who how c. ki kojâ who where d. ki key who when ‘Who left?’

raft went-3sg raft went-3sg raft went-3sg raft went-3sg

In all these sentences, the argument is interpreted as a wh-phrase, and receives an answer. The wh-adjunct must remain unstressed and receives no wh-interpretation. These sentences are natural reactions to the following questions, employed for the purpose of clarification. (84) a. un cherâ raft? she why went ‘Why did she leave?’ b. un chetori raft? she how went ‘How did she go?’ c. un kojâ raft? she where went ‘Where did she go?’ d. un key raft? she when went ‘When did she go?’ The sentences in (83) are ungrammatical in any other context. These observations indicate that although both kinds of wh-phrases have wh-features, their inherent differences do not allow them to interact: the wh-argument has an inherent focus feature, while the wh-adjunct lacks this element. However, the Superiority Condition is still relevant as the data in (82) attest. The following contrast shows that the Superiority Condition (cf. MLC) must be observed in the case of non-wh-phrases as well. Of the two nonwh-elements bearing the focus feature, the higher one may be moved by

Focus

153

selecting the EPPs. The movement of the lower one causes the derivation to crash due to the MLC. (85) a. [FocP [be KIMEA]i man ti faghat se-â ketâb dâd-am] to K I only three-part book gave-1sg ‘It was to KIMEA that I gave only three books.’ b. *[FocP [se-tâ KETAB]i man faghat be Kimea ti dâd-am] Intended meaning: It was three BOOKS that I gave only to Kimea.’ In (85a), the contrasted indirect object has moved into the Spec of FocP, while the only-phrase, carrying an inherent focus feature, stays in situ. In (85b), the nonspecific direct object moves into the Spec of FocP, crossing the inherently contrastive indirect object. The result is a violation of the MLC, and the sentence is ungrammatical. The sentence in (85b) is grammatical with an interpretation where faghat ‘only’ modifies the moved direct object: it was only three books that I gave to Kimea. The explanation for this interpretation is straight forward: the direct object faghat se tâ KETAB moves into the edge of vP in order to be able to move out (see also footnote 5). The focus adverb faghat remains in that position while the noun phrase moves further up into the Spec of FocP. This is illustrated in (86). (86) [FocP [se-tâ KETAB] k [FocP man [FocP [faghat tk]i [FocP be Kimea ti dâd-am ]]]] To sum up, we have shown in this section that focus movement is triggered by the selection of the EPPs feature from the lexicon. This analysis is supported by the fact that this type of movement is subject to the MLC when more than one element carrying the same type of features appears in the same clause.

4.4. Optionality of focus movement We saw that wh-arguments must move into the Spec of FocP in order to receive contrastive reading. We also saw that this movement is due to the selection of an EPPs feature from the lexicon. The movement of contrasted non-wh-phrases, however, is optional, since these elements may re-

154

Chapter 4: Operator/Discourse Domain and A’-Scrambling

ceive contrastive interpretation in situ as long as they are heavily stressed. How can the optionality of movement be explained in those cases? We address this issue in section 4.4.1. Focus movement in other languages is discussed in 4.4.2. 4.4.1. Contrasted non-wh-phrases We suggest that UG allows the focus feature to be interpreted as contrastive in three different ways: – Syntactically, by overt movement of XP into the Spec of FocP (whphrases), – Morphologically, by being realized as a morpheme, and – Phonologically, by being interpreted as heavy stress in PF (non-whphrases). We discussed the first option in section 4.2. The second option is realized as a morpheme such as only in only-phrases. Furthermore, this option shows up as an affix in some other languages. For example, the suffix -su marks focus in Yaqui (Jelinek 1994). In Navajo, contrastive focus can be marked by the particle -ga (Young and Morgan 1987, Barss, Hale, Perkins, Speas 1991, Willie 1991, 2000, Willie and Jelinek 2000, Hale, Jelinek, Willie 2003)79. In these languages, the morphological marker represents the contrastive feature. In languages that lack a contrastive affix, the focus feature of non-whphrases is interpreted as heavy stress in PF. The cyclic LF and PF interpretation in each phase allows the stress (PF interpretation) to receive the contrastive reading (LF interpretation). Note that contrasted elements that are marked by a focus morpheme, as in Navajo, do not bear focus intonation (Hale et al. 2001). That is, contrastive focus in non-wh-phrases is realized either morphologically or phonologically. We see in this section that the focus feature of contrasted non-wh-phrases must move to Foc, similar to the movement of the wh-feature to C. We saw in section 4.2.2. that wh-phrases may only receive an echo interpretation in the domain of a Negative Polarity Item (NPI). This is because the interaction of Neg and the NPI blocks the movement of the whfeature to C (cf. (66a) and (67a)). We also saw that the structure can be rescued by the movement of the wh-phrase to Spec of FocP, thus out of the domain of the NPI (cf. (66b) and (67b)). A similar situation seems to

Focus

155

hold for contrasted non-wh-phrases, suggesting that the focus feature of the contrasted element must move to Foc. Consider the following contrast: (87) a. ??hichkas KIMEA-ro na-did nobody KIMEA-râ neg-saw b. [KIMEA-ro]i hichkas ti na-did ‘It was Kimea who nobody saw.’ In (87a), the NPI blocks the movement of the contrastive feature to Foc. The same situation holds for only-phrases: (88) a. *hichkas faghat Kimea-ro na-did nobody only K-Acc neg-saw b. [faghat Kimea-ro]i hichkas ti na-did ‘It was only Kimea that nobody saw.’ The movement of the contrastive feature of the only-phrase to Foc in (88a) is blocked by the licensing of NPI by Neg. This is shown in (89). (89) [FocP Foc [hichkas [NegP Neg [TP Only-phrase … *

Note that the movement of a wh-feature can also be blocked by an onlyphrase, as in (90a). The structure is rescued by the movement of the whphrase into the Spec of FocP. (90) a. *faghat Kimea ki-ro did (OK as an echo question) only K who-Acc saw-3sg b. ki-ro faghat Kimea did ‘Who was it that only Kimea saw?’ The reason is that the focus feature of the only-phrase in (90a) is closer to the head of FocP than the wh-phrase is, blocking the movement of the whfeature to C through the head of FocP. This is shown in (91). (91) [CP OP C [FocP Foc [Only-phrase [TP [vP wh … *

156

Chapter 4: Operator/Discourse Domain and A’-Scrambling

The discussion in this section supports the claim that the movement of a contrasted element is not triggered by the focus feature. In the case of non-wh-phrases, the focus feature moves by itself to adjoin to Foc, without forcing the entire XP to move. We suggest that the movement of the non-wh-phrase is triggered by EPPs, similar to the movement of whphrases. Thus, the movement applies only when EPPs is chosen from the lexicon. Consequently, optionality is reduced once again to an optional selection of EPPs from the lexicon. The sentence in (92a) presents an interesting property with respect to the movement of focus features. This sentence is ungrammatical on a reading where the focus adverb faghat modifies the subject. However, (92b) is grammatical where faghat modifies the direct object Kimea. (92) a. *[faghat dâneshju-hâ] KIMEA-ro dust dâr-an only student-pl KIMEA-Acc friend have-3pl b. faghat dâneshju-hâ KIMEA-ro dust dâr-an How can our analysis explain these facts? If faghat modifies the subject in (92a), its focus feature must move to Foc, blocking the movement of the focus feature of the object to Foc. Thus, the derivation crashes. However, if faghat modifies the indirect object, the subject will no longer be contrastive, and the focus feature of the object can move to adjoin to Foc, providing the grammatical reading. Note that the sentence in (85a), repeated below in (93), is grammatical. In this sentence, the stressed element revealing contrastive reading, is in Spec of FocP, while the only-phrase is in a lower position. (93) [FocP [be KIMEA ]i man ti faghat se-â ketâb dâd-am ] to K I only three-part book gave-1sg ‘It was to KIMEA that I gave only three books. (not to Parviz).’ The presence of the contrasted element in a higher position blocks the contrastive interpretation of the lower element in (92a). This is not true in (93). What is the explanation for this contrast? We suggest that the only-phrase, by virtue of being inherently focused, has a stronger feature, and thus blocks the movement of the focus feature in the lower phrase in (92). In other words, it seems that inherent focus

Focus

157

reveals some sort of superiority: it blocks the movement of the non-inherent focus feature. The reverse situation does not hold, as in (93). In summary, contrastive reading of non-wh-phrases is available by either stress or morphology (cf. only-phrases) in Persian. In both cases, the focus feature must adjoin to Foc. The focus feature of only-phrases, however, reveals superiority with respect to the focus feature of stressed elements, as we saw in the last section. The optional selection of EPPs seems to violate the principle of economy in the case of non-wh-phrases, since contrastive reading does not depend on XP movement. However, the possibility of movement may rescue the structure in cases where feature movement is blocked (cf. (66a), (67a), (87a), and (88a)). We suggest that EPPs is vacuously chosen in other cases, since the system needs its selection in those instances where it has an effect on the output of the derivation (cf. (66b), (67b), (87b), and (88b)). 4.4.2. Contrastive focus in other languages Contrastive/identificational focus movement is obligatory in some languages and optional in others. In a language like Hungarian, the element carrying the identificational/contrastive focus must move into the preverbal position, as in (94). (94) Mari egy kalapot neztt ki maganak Mary a hat-Acc picked out herself-Acc ‘It was a hat that Mary picked for herself.’

(Kiss 1998: 249)

Compare this sentence with the following, in which ‘a hat’ represents information focus. (95) Mari neztt ki maganak egy kalapot Mary picked out herself a hat ‘Mary picked for herself a HAT.’

(Kiss 1998: 249)

Even Only-phrases have to undergo movement in Hungarian. (96) Janos csak Marit hivta meg John only Mary-Acc invited PERF ‘It was only Mary that John invited.’

(Kiss 1998: 265)

158

Chapter 4: Operator/Discourse Domain and A’-Scrambling

In addition to Hungarian, Greek and Finnish are two other languages in which contrastive focus involves obligatory movement. In English, Identificational focus appears in a cleft construction, as in (97a), or stays in situ, bearing heavy stress, as in (97b). That is, focus movement is optional in this language. (97) a. It was the war that we were afraid of. b. We were afraid of theWAR. (not something else) The optionality of movement is true of only-phrases as well. (98) a. John invited only Mary. b. It was only Mary than John invited. In Rumanian, Italian, and Catalan, focus movement is optional as well. The examples in (99) are from Italian: (99) a. Chi di voi due ha rotto il vaso? which of you two has broken the vase ‘Which one of you two has broken the vase?’ b. Maria ha rotto il vaso. ‘It is Maria who has broken the vase.’ c. L’ha [VP rotto Maria] il vaso ‘it is Maria who has broken the vase.’

(Kiss 1998: 269)

Within the system we are proposing in this section, the selection of EPPs is obligatory in Hungarian, Greek, and Finnish which places the contrasted element in Spec of FocP. Rumanian, Italian, Catalan, and English, similar to Persian, have two choices: either the contrastive feature moves to Foc, or EPPs triggers the movement of XP into the Spec of FocP.

5. Restrictions on scrambling We saw in chapter 1 that certain elements are either subject to restricted scrambling, or do not undergo this operation at all. We are now in a position to explain those restrictions. We start with nonspecific objects, and move on to NV-elements, VPs, and clauses.

Restrictions on scrambling

159

Nonspecific objects We saw in this chapter that movement into the operator domain is due to topic, contrastive focus, and quantifier raising (adverbial elements in particular). This observation explains why nonspecific objects move only when they are contrastively focused. This is so because contrastive reading maintains the novelty of the nonspecific object while placing it within a discourse background. It cannot be topicalized, however, since it bears new information. This fact explains why nonspecific objects are more restricted in terms of movement than their specific counterparts. The latter are qualified to represent topic in addition to contrastive focus. NV-elements NV-elements are not subject to movement unless they become quantificational, as the following contrast, taken from chapter 1, attests. (100) a. *Kimea zamin-e xaxti diruz xord Kimea earth-Ez hard yesterday collided b. Kimea [che zamin-e saxti ]i diruz [CPr ti xord ] Kimea what earth-Ez hard yesterday collided ‘What a hard fall Kimea had yesterday.’ Lit. Kimea what a hard earth yesterday collided. The NV-element in (100a) is an X’, a defective projection, and thus is not subject to movement (Samiian 1983). In (100b), che turns this defective projection into a maximal quantificational phrase. Therefore, its movement, as an instance of quantifier raising, becomes possible. Verb Phrases Verb phrases can be fronted in their own clause, as in (101). They are not subject to LDS, as in (102). Both examples are taken from chapter 1. (101) [un film-ro did-an] una that movie-râ saw-3pl they ‘They saw that movie.’ (102) *[un film-ro did-an] Kimea goft una that movie-râ saw-3pl K said-3pl they

160

Chapter 4: Operator/Discourse Domain and A’-Scrambling

The sentence in (101) is an instance of topicalization: the direct object has moved into the Spec of TP, while the verb has moved to T. Thus, what we have in this sentence is not really a vP movement. Rather, it represents a regular topic movement, followed by verb movement. Focus movement may trigger the movement of the verb as well. (103) KIMEA-ro did-am tu xiyâbun K-râ saw-1sg in street ‘It was Kimea that I saw on the street.’ (104) KIMEA dâd ketab-ro be dust-esh K gave-3sg book -râ to friend-her “It was Kimea who gave the book to her friend.” In these examples, the verb has adjoined the head of the FocP. Topicalization and focus trigger verb movement in other languages as well. Consider the following example from German (taken from Haider 1984). (105) Ich glaube [CP den Fritzi magk [IP jeder ti tk]] I believe Art-Acc F. likes everyone The fact that we have a verb movement in (101), (103), and (104) explains the ungrammaticality of (102): since V-movement is subject to locality, LDS may not apply to verbs. Scrambling of Clauses If the analysis in this chapter is on the right track, and the movement into the operator domain represents topic, contrastive focus, and movement of quantificational elements in Persian, then the inability of clauses to scramble becomes evident: these elements cannot be topicalized, since they do not denote old information. Furthermore, they are not subject to contrastive focus due to the fact that clauses lack a D-head. Finally, they are not subject to quantifier raising, simply because they are not quantified.

Conclusion

161

6. Conclusion We saw in this chapter that movement into the operator/discourse domain is triggered by EPPs, which places the XP in the Spec of FocP or one of the two topic positions. In the case of the former, the XP receives a contrastive focus interpretation. The higher topic position, Spec of TopP, is reserved for the switched topic, and the lower one, Spec of TP, represents the background topic. We also saw that sentential adverbs and whadjuncts are subject to quantifier raising, and may be placed in different position inside the operator/discourse domain. The claim that all these movements are feature-based is supported by the fact that the movement of two elements belonging to the same category is subject to MLC. The discussions in this chapter suggest that UG provides three different options for contrastive focus interpretation:

– The syntactic option, realized by the movement of the XP into the Spec of FocP (wh-arguments in Persian and contrastive XPs in Hungarian, Greek, and Finnish), – The morphological option, realized by the presence of a morpheme (only in Persian, Hungarian, and English only-phrases, affixes in Navajo, Yaqui, etc.), and – The phonological option, realized by interpreting the focus feature as stress (Persian non-wh-phrases). It was further argued that the focus feature of non-wh-phrases moves to be adjoined to Foc. Similarly, the wh-feature of wh-phrases was shown to undergo movement to C. These claims were based on the fact that the movement of a Foc-feature and a wh-feature is blocked when the XP carrying one of them is in the domain of another scope bearing element (cf. NPI). The discussion in this chapter provides an explanation for restricted scrambling in the case of nonspecific objects and the NV-element of CPr. It also explains the lack of scrambling in the case of long distance movement of vP and scrambling of clauses. All these restrictions are explained once the nature of scrambling is shown to represent instances of topic and contrastive focus.

162

Chapter 4: Operator/Discourse Domain and A’-Scrambling

Notes 61. See chapter 5 for an in-depth analysis of anaphoric relations in Persian. 62. Cinque also discusses post-VP adverbs. Those adverbs, however, are not relevant to Persian, and therefore, will not be pursued here. 63. The extracted subject must be in a higher position than negation in (11b). The most natural way of expressing this sentence is the one in (i), where the subject is clearly out of the domain of negation. (i) ye dâneshju hast ke hargez na-yumade one student is that never neg-come-3sg ‘There is one student who has never come.’ 64. But also see the example in (92) and the discussion related to it.in section 4.4.1 65. There is a third reading for (14): faghat may modify the subject Kimea. The structure underlying this reading is the one in (i) where the DP Kimea has moved out of the adverbial phrase. (i) Kimea [faghat t] bastani xarid K only ice cream bought-3sg The same reading is available in the case of the sentences in (15) and (16). 66. There is also the disjoint negative focus adverb hattâ … ham ‘even … too’ (= not even). The element that this focus adverb modifies appears between its two parts. Examples are provided in (i) - (iii). (i) hattâ Kimea ham ketâb ne-mi-xun-e even K too book neg-dur-read-3sg ‘Not even Kimea reads books.’ (ii) Kimea hattâ ketâb ham ne-mi-xun-e ‘Kimea does not even read books.’ (iii) Kimea hattâ barâ man ham ketâb ne-mi-xun-e K even for me too book neg-dur-read-3sg ‘Kimea does not read books even for me.’ 67. Kiss (1998) argues that identificational focus does not have to be [+contrastive]. She shows that while this kind of focus is [+contrastive] in Rumanian, Italian, Catalan, Modern Standard Arabic, and Greek, it is [+/- contrastive] in Hungarian and English. She further suggests that the feature [+exhaustive] is part of the properties of the identificational focus. Identificational focus in Persian, similar to Rumanian, Italian, Catalan, Greek, and Modern Standard Arabic, is marked as [+Contrastive], as stated in text. 68. Thanks to Katalin Kiss for clarifying this issue (personal communication). See also Szabolcsi (1983) for discussion. 69. Analyzing multiple wh-constructions in Bulgarian and some other languages, Richards (1997) suggests that in addition to MLC, Shortest Move (ShM) is required to adequately account for the data in those languages. Consider the following data: (i) a. koj kogo vizda ? who whom sees ‘Who sees whom?’ b. *kogo koj vizda (Rudin 1988: 472–473, per Richards 1997: 63) Furthermore, the examples in (40) also supports Grewendoff’s (2001) analysis where he suggests that multiple wh-fronting is a case of wh-cluster fronting where all wh-phrases are attached to each other, and move as a single cluster. 70. The same situation has been observed by Rizzi (1997) for Italian and by Kiss (2003) for Hungarian, among others.

Notes

163

71. See also Megerdoomian and Ganjavi (2001) and Taleghani (2003) who have made the same suggestion for Persian. 72. See Pesetsky (2000) who explains the existence of a Superiority violation in German by suggesting a one-Spec-CP system in this language combined with the movement of the wh-feature to C. According to Pesetsky, English cannot violate the Superiority Condition, since this language has a multiple-Spec-CP system, and thus all wh-phrases must move into the Spec of CP (either overtly or covertly). 73. The data are from Jan Mohammad, a native speaker of Dari (p.c.) 74. This is in the spirit of Hasegawa (1994). But also see Beck (1996). 75. Contrastive non-wh elements and inherent contrastive elements reveal similar properties. See section 4.4.1. for discussion. 76. There have been other attempts to analyze scrambling as a movement motivated by focus feature. See, for example, Miyagawa (1994, 1997). 77. The proposal advanced in Karimi (2003d) is based on the assumption that the contrasted wh-argument is heavily stressed. Further investigation shows that even though the extracted wh-argument receives a slightly heavier stress, it does not bear the same kind of intonation that the contrasted non-wh-phrase does. 78. A similar constraint is observed in Japanese. The following data are from Takahashi (1993), cited by Richards (1997). (i) a. John-ga [Bill-ga dare-ni [Mary-ga nani-o tabeta John-Nom Bill-Nom who-Dat Mary-Nom what-Acc ate to itta to ] omotteriu no? that said that thinks Q ‘Who does John think that Bill told t that Mary ate what?’ b. Dare-ni John-ga [Bill-ga t [Mary-ga nani-o tabeta to] itta to] omotteriu no c.*Nani-o John-ga [Bill-ga dare-ni [Mary-ga t tabeta to] itta to] omotteriu no 79. Persian does not have an overt focus marker, but EMPHATIC elements are marked by ke, as in (i) - (ii). (i) to-ke hanuz injâ-i you-EMPH still here-2sg ‘You are still here.’ (ii) ketâb-ro ke na-xund-i book-râ EMPH neg-read-2sg ‘The book, you haven’t read.’

Chapter 5 Scrambling, Scope, and Binding

1. Introduction It has been suggested in the literature that scrambling, particularly long distance scrambling (LDS), is semantically vacuous and is undone at LF (Saito 1985 and work thereafter, Kuroda 1988, Fukui 1993, among others). We saw in previous chapters that scrambling in Persian represents topic and focus, and has an effect on the semantic outcome of the derivation. In this chapter, we concentrate on the semantic impact of scrambling with respect to scope and binding relations. As for scope, we have already seen in chapter 4 that scrambling creates blocking effects when whphrases interact with negation. In this chapter, we examine other types of scope-bearing elements. We will see further evidence indicating that scrambling changes the truth value of the sentence, and creates or resolves blocking effects. We will specifically arrive at the conclusion that scope is read off the surface structure, as stated in (1)80. (1)

Scope is determined in overt syntax.

The statement in (1) implies that scope is determined only by overt movement, taking into account the moved element and its trace. Thus scope ambiguity resulting from scrambling is resolved in the following ways: (2)

Scope ambiguity is resolved on the basis of a. Deleting the head of the chain, and interpreting the copy at the tail (scope reconstruction), b. Interpreting the head of the chain, and ignoring the copy at the tail81.

Next, we examine binding relations prior to and after movement to determine the affects of scrambling on binding interpretation. Our analysis leads us to the following basic conclusions:

Scope marking

(3)

165

a. The anaphoric relation of α and β is established as soon as α is merged with the category γ that contains β. b. Scrambling neither feeds nor bleeds Principle A. c. Scrambling may feed Principles B and C, but does not bleed them.

We provide evidence from other scrambling languages to support the discussion advanced in this chapter. The statements in (3) will be refined in the course of our analysis of Persian and other languages. Scope marking is discussed in section 2. Binding relations, including a cross-linguistic analysis, follows in section 3. A brief analysis of Phase Theory with respect to binding relations is also offered in that section. Concluding remarks appear in section 4. 2. Scope marking Given the phrasal architecture of Persian clauses provided in previous chapters, quantified elements must be in the operator domain of the clause, namely within the CP phase. The arrangement of arguments and adjuncts in this phase has an impact on their scope marking, as we saw in previous chapters. In this section, we concentrate on the scope alterations of quantified elements created by scrambling. The interaction of two quantifiers is discussed in 2.1., followed by a discussion of wh-phrase/ quantifier interaction in 2.2. The scope of universal and existential quantifiers with respect to negation is discussed in section 2.3., followed by a brief review of the scope of adjuncts in 2.4. The general outcome of these examinations are discussed in 2.5. Section 2.6. summarizes and highlights the main issues concerning scope marking in this language. 2.1. QP / QP interaction Movement of quantificational elements alters their scope. Consider the sentences in (4). In these examples, a universal QP and a specific indefinite DP interact with each other. (4)

a. har dâneshju-i tu in kelâs ye ketâb-i-ro mi-xun-e every student-ind in this class a book-ind-râ dur-read-3sg ‘Every student in this class reads one poem.’ ∀ > ∃ ; *∃ > ∀

166

Chapter 5: Scrambling, Scope, and Binding

b. [ye ketâb-i-ro]i har dâneshju-i tu in kelâs ti mi-xun-e ∃>∀;∀>∃ The sentence in (4a) has only a distributive interpretation: each student reads a different book. (4b) is ambiguous. The primary reading of this sentence is a collective one: there is one specific book that every student reads. The following examples provide additional pieces of evidence. (5)

a. man [ye she’r-i-ro] [barâ har shâgerd-i ] mi-xun-am I a poem-ind-râ for each student-ind dur-read-1sg ‘I read a (specific) poem for every student.’ *∀ > ∃ ; ∃ > ∀ b. [barâ har shâgerd-i]j man [ye she’r-i-ro] tj mi-xun-am ‘For each student I read one poem.’ ∃>∀;∀>∃ The sentence in (5a) has only one interpretation: there is one specific poem such that I read it for every student. (5b) is ambiguous, since scrambling has altered the c-command relation of the overt direct and indirect objects. The primary reading is a distributive one in this case: for each student, I read a different poem.

Contrary to what has been suggested in the literature, LDS reveals the same effect. (6)

a. har dâneshju-i tu in kelâs fekr mi-kon-e [CP Kimea ye every student-ind in this class thought dur-do-3sg K a pesar-i-ro dust dâr-e] boy-ind-râ friend have-3sg ‘Every student in this class thinks that Kimea loves one boy.’ ∀ > ∃ ; *∃ > ∀ b. ye pesar-i-roi har dâneshju-i tu in kelâs fekr-mi-kon-e [CP Kimea ti dust-dâr-e] ∃>∀;∀>∃ In (6a), the universal quantifier has wide scope over the embedded indefinite DP: for every student there is a different boy that she thinks Kimea loves. (6b) is ambiguous. The reading in which the existential quantifier receives wide scope is the primary one in this case: there is one specific boy such that every student thinks Kimea loves that boy.

Scope marking

167

The statements in (1) and (2) account for the scope interaction discussed thus far. The one in (2) confirms that both copies (head and tail) may receive interpretation.

2.2. WhP / QP interaction Similar to cases discussed in the previous section, scrambling alters the scope between a quantified element and a wh-phrase. (7)

a. har kârmand-i be kodum kândidâ ra’y mi-d-e ? each employee-ind to which candidate vote dur-give-3sg] Lit. ‘Each employee votes for which candidate?’ ∀ > whP ; *whP > ∀ b. [be kodum kândidâ]i har kârmand-i ti ra’y mi-d-e ‘Which candidate is such that each employee votes for her?’ whP > ∀ ; ∀ > whP

(7a) has a distributive reading. The answer to this question is that A votes for (1), B for (2), C for (1), and so on. In (7b), the wh-phrase has scrambled to the front of the sentence, creating ambiguity. The primary reading for this sentence is a collective one: which specific candidate is such that every employee votes for that person? LDS exhibits the same pattern: (8)

a. pro fekr-mi-kon-i [CP har dâneshju be kodum ostâd thought dur-do-2sg each student-ind to which professor ra’y be-d-e] vote subj-give-3sg ‘Which professor do you think each student will vote for?’ ∀ > whP ; *whP > ∀ b. be kodum ostâdi pro fekr-mi-kon-i har dâneshju ti ra’y be-d-e whP > ∀ ; ∀ > whP

(8a) receives a distributive reading. The answer to this question is something like A will vote for Smith, B for Miller, and C for Jones. (8b) is ambiguous, indicating that the scrambled element as well as its copy may enter the scope relation. The primary interpretation of this sentence is a collective one. Therefore, the most plausible response to this question would be something like they will vote for Jones. Once again, the state-

168

Chapter 5: Scrambling, Scope, and Binding

ment in (2) accounts for the ambiguity of the ‘b’ sentences in this section. Furthermore, the ambiguity of (8b) contradicts the assumption that LDS is subject to radical reconstruction.

2.3. Neg and ∃P / ∀P Interaction Now consider the following data in which the verb is negated: (9)

a. har dâneshju-i ye ketâb-i-ro ne-mi-xune each student-ind one book-ind-râ neg-dur-read-3sg ‘For each student there is one book that she does not read.’ b. *ye ketâb-i-ro har dâneshju-i ne-mi-xune

(9a) has the interpretation in (10). (10)

For every x, student (x), ∃ y, book (y), & ¬ [ read (x, y) ]

The existential quantifier cannot take wide scope over the universal quantifier har when interacting with negation. Thus, scrambling destroys the structure in (9a), producing the ill-formed string in (9b). The existential quantifier takes scope over the universal quantifier only when it is clearly out of the negative clause, as in (11). (11) ye ketâb-i hast ke har dâneshju-i (un-o) ne-mi-xune a book-ind is that each student-ind (that-râ) neg-dur-read-3sg ‘There is a book such that not every student reads it.’ The contrast in (12) provides yet another example along the same lines. (12) a. *ye dâneshju har doxtar-i-ro ne-mi-buse one student each girl-ind-râ neg-kiss-3sg b. har doxtar-i-roi ye dâneshju ti ne-mi-bus-e ‘Every girl is such that one student does not kiss her (all other students do).’ The interpretation of (12b) is the one in (13): (13)

For every x, girl (x), ∃ y, student (y), & ¬ [ kiss (y, x) ]

Scope marking

169

Note that (12a) is grammatical if the existential quantifier is interpreted as generic, thus losing its quantificational force. On the generic reading, the interpretation of the sentence would be something like a student in general does not kiss every girl (this is not a good behavior!). The fact that it is the c-commanding
P that is responsible for the ill-formedness of (12a) is supported by the grammaticality of the following example in which the universal quantifier, although in a negative clause, is not c-commanded by an existential quantifier. Compare (14) with the one in (12a)82. (14) in dâneshju har ketâbi-ro ne-mi-xun-e this student every book-râ neg-dur-read-3sg ‘This student does not read every book.’ (he reads some books.) The sentence in (14), where the subject is a specific DP rather than an existential QP, is perfectly well-formed.

2.4. Scope of adjuncts We saw in chapter 4 that the LDS of wh-adjuncts creates ambiguity. That is, the copy of a scrambled adjunct plays a role in its scope reading, as evident by the following examples: (15) a. cherâ fekr mi-kon-i [CP Kimea emruz bargasht]? why think dur-do-2sg K today returned ‘Why do you think Kimea returned today?’ b. key Kimea goft [CP ke Sepide xune xaride?] when K said that S house bought is ‘When did Kimea say that Sepide has bought a house?’ The wh-adjuncts in (15) can be interpreted as having scope over the matrix clause or just the embedded clause. The ambiguity created by the extraction of adjuncts is not restricted to wh-adjuncts. In the following examples, the sentential adjunct has moved to the focus position in the matrix clause, creating ambiguity. (16) a. EMRUZ pro goft [CP ke Kimea ketâb-â-ro mi-yâr-e] TODAY said that K book-pl-râ dur-bring-3sg ‘TODAY she said Kimea will bring the books.’

170

Chapter 5: Scrambling, Scope, and Binding

b. bedune hich DALILI Kimea fekr-mi-kon-e [CP ke Sasan without any reason K thought dur-do-3sg that S in kâr-ro karde ] this work-râ done-3sg ‘Without any reason Kimea thinks that Sasan has done this.’ The sentential adverbs EMRUZ ‘ today’ and bedune hich DALILI ‘without any reason’ in (16) can take scope over either the matrix clause or the embedded clause. Again, these ambiguities contradict the assumption that LDS is subject to radical reconstruction.

2.5. Discussion The data examined in this section suggest that scope is determined based on the surface structure. Is there any evidence that this is a universal property of grammar? The discussion in this section is an attempt to offer some answers, although they might not be conclusive. The following contrast, discussed by Lasnik (1999a), indicates that scope is not affected after Spell-out. (17) a. A man is likely to be here. b. There is likely to be a man here. While the DP a man in (17a) can either take wide or narrow scope with respect to likely, it can only take narrow scope in (17b). The same contrast was observed by Chomsky (1991) with respect to there constructions and negation: (18) a. Many linguistics students aren’t here. b. There aren’t many linguistics students here. These examples show that scope is determined in overt syntax, and that there is no covert movement involving scope. The following contrast, borrowed from Fox (1999), confirms the same generalization: while (19b) is ambiguous, (19a) is not. This contrast is explained if scope reading applies strictly in syntax based on the chain created by overt movement. (19) a. John seems to a (#different) teacher [t to be likely to solve every one of these problems]. ∃ > ∀; ∗∀ > ∃

Scope marking

171

b. [Every one of these problems]i seems to a teacher [ti to be likely to be solved t by John.] ∃ > ∀; ∀ > ∃ (Fox 1999: 159) However, there are cases that seem to contradict this generalization. Consider (20), also borrowed from Fox (1999). (20) a. [At least one soldier]i seems (to Napoleon) [ti to be likely to die in every battle] b. [At least one soldier]i seems to himself [ti to be likely to die in every battle] (Fox 1999: 160) (20b) receives only one interpretation that is compatible with its surface structure: the existential quantifier receives a wide scope in an A position. Fox states that (20a) can be ambiguous between (∃ > ∀) and (∀ > ∃). However, the first interpretation is implausible since the same person cannot die in every battle. Fox suggests that the wide scope of the universal quantifier is obtained by reconstructing the existential quantifier and raising the universal quantifier to a position higher than the existential phrase. This proposal presupposes the existence of covert XP movement, and contradicts the position taken in this work which suggests that scope is determined in overt syntax. There can be an alternative analysis for (20a) and similar cases. That is, the wide scope reading of the existential quantifier is not compatible with the semantic properties of die. Thus, this interpretation does not hold, based on the meaningless output of the composition of the semantics of die and the wide scope reading of the existential quantifier. As a result, the quantifier loses its quantificational force, and fails to enter the scope race. Thus the universal quantifier, in the absence of a competitor, receives wide scope in situ without the need for the covert XP movement. This approach is in fact supported by comparable examples from Persian (see below). As we have seen thus far, one of the by-products of scrambling is that it has an affect on the scope in Persian. Thus, scope marking is accomplished by overt movement. The sentence in (21), the Persian counterpart of (20a), shows that the universal quantifier receives a wide scope in situ, without the need to be scrambled. (21) hadde-aghal ye sarbâz be nazar mi-yâ-d ehtemâl at least one soldier to view dur-come-3sg possibility

172

Chapter 5: Scrambling, Scope, and Binding

dâshte bâsh-e dar har jang be-mir-e have subj be-3sg in every battle subj-die-3sg ‘At least one soldier seems to be likely to die in every battle.’ Wide scope reading of the existential quantifier is not even available in this sentence. The universal QP can be scrambled over the existential QP only to express focus or emphasis, but the movement is not necessary for scope interpretation. That means the semantics of ‘die’ cancels out the wide scope interpretation of the existential quantifier. Now compare (21) with (22), where the verb allows the wide scope of the existential phrase. (22) hadde-aghal ye sarbâz be nazar mi-yâ-d ehtemâl at least one soldier to view hab-come-3sg possibility dâshte bâsh-e dar har mosâbeghe barande be-sh-e have subj be-3sg in every competition winner subj-become-3sg ‘At least one soldier (#different) seems to be likely to become the winner in every competition.’ The existential quantifier receives unambiguous scope over the universal quantifier. Conversely, (23) shows that scrambling provides wide scope for the universal quantifier. (23) [dar har mosâbeghe]i hadde-aghal ye sarbâz be nazar mi-yâ-d ehtemâl dâshte bâsh-e ti barande be-sh-e ‘In every competition, at least one (different/#different) soldier seems to be likely to become the winner.’ As the translation of (23) reveals, this sentence is ambiguous, although the primary reading is the one in which the universal quantifier receives wide scope.

2.6. Summary The analysis in this section shows that scrambling affects the scope of quantified elements. Thus, it is not subject to radical reconstruction. Furthermore, the ambiguity created by scrambling is determined by either interpreting the head of the chain, or its copy at the tail of the chain. Therefore, scope is suggested to be determined in overt syntax in each phase.

Binding

173

Consequently, there is no need for covert XP movement, as suggested by Chomsky (1995).

3. Binding In this section, we discuss binding relations and the effect of scrambling on those relations. Binding is discussed in the light of the principles stated in section 3.1.2. in chapter 2, and repeated below in (24). (24) Binding Theory (BT) a. If α is an anaphor, interpret it as co-referential with a ccommanding phrase in D. b. If α is a pronoun, interpret it as disjoint from every ccommanding phrase in D. c. If α is an R-expression, interpret it as disjoint from every ccommanding phrase. Where D = the minimal domain. We show in this section that anaphoric binding relations are in essence established by Merge, and scrambling does not have any impact on the relation between anaphors and their antecedents. In contrast, clitic pronominals allow binding on the basis of the base and derived positions of the antecedent. Furthermore, scrambling may feed binding principles in the case of free standing pronouns and R-expressions, but may not bleed them. The summary of these issues was stated in (3), and is repeated below. (3)

a. The anaphoric relation of α and β is established as soon as α is merged with the category γ that contains β. b. Scrambling neither feeds nor bleeds Principle A. c. Scrambling may feed Principles B and C, but does not bleed them.

Finally, we see that pronominal binding relations cannot be established by merely derivational operations, and must be interpreted representationally if phase theory is to be maintained. Similarly, Principle C can be maintained only if the binding relation is calculated representationally. Anaphors are addressed in section 3.1., followed by an examination of free-standing pronominals and R-expressions in 3.2. and 3.3., respectively.

174

Chapter 5: Scrambling, Scope, and Binding

The status of clitic pronominals is discussed in 3.4. A summary of this discussion is provided in 3.5. Section 3.6. contains a cross-linguistic analysis of binding relations in other scrambling languages. The problems that binding relations create for phase theory is the subject of section 3.7.

3.1. Anaphors There are two anaphoric elements in Persian: the reflexive xod ‘self’ and the reciprocal hamdige ‘each other’. As in many other languages, the Persian reflexive xod is subject-oriented83. This element is utilized only in the written and formal language today. In the colloquial language, a clitic pronoun is attached to xod, and thus its anaphoric property is lost. Since word order is far more rigid in the formal and written language than in the spoken language, xod cannot appropriately serve as a diagnostic element to show the effect of scrambling on binding. Thus we concentrate on the reciprocal hamdige in this work84. Consider first the following data. Irrelevant details are not included. (25) a. [vP unâj bachche-hâ-roi [PredP be hamdigej/i mo’arrefi they child-pl-râ to each other introduction kard-an]]] did-3pl ‘They introduced the children to each other.’ b. [TopP unâj [TP [be hamdigej/i]k [vP bachche-hâj-ro [PredP tk mo’arrefi kard-an ]]] c. [TopP [be hamdigej/i]k [TP unâj [vP bachche-hâ-roi [PredP tk mo’arrefi kard-an] The specific direct object receives its interpretation in Spec of vP, as discussed in chapter 3. Therefore, it c-commands the reciprocal from that position. The binding relation between the direct object and the reciprocal is established as soon as the former moves into the Spec of vP in (25a). The subject is then merged in the same binding domain. Thus, the reciprocal may also be bound by the subject. The sentence in (25b) shows that the movement of the reciprocal into a position higher than the direct object does not change the binding relation. This is also true when the anaphor is moved to the highest position of the sentence, the Spec of

Binding

175

TopP. Note that the phrase containing the anaphor must move to the edge of vP to be able to move out. The structures of vP in (25b) and (25c) are shown in (26). Again, details are omitted. (26) [vP tS tAnaphor bachche-hâj-ro [PredP tAnophor mo’arrefi kard-an]] The lowest copy of the anaphor is c-commanded by the direct object, while the higher one is only c-commanded by the copy of the subject in Spec of vP. Nevertheless, the binding relation between the anaphor, on the one hand, and the subject and the object, on the other, remains untouched. Now consider the following sentences. (27) a. [vP bachche-hâi hamdiga-roi [PredP busid-an]] child-pl each other-râ kissed-3pl ‘The children kissed each other.’ b. [TP [hamdiga-roi]k [vP bachche-hâi tk [PredP busid-an]]]

The reciprocal is bound by the subject within the vP in (27a). Its movement into the Spec of TP does not render the sentence ungrammatical. LDS shows the same effect. That is, the surface position of the scrambled element does not enter the binding relation, as in (28). (28) a. mo’allem-âj fekr mi-kon-an [CP ke [tP [vP bachche-hâi teacher-pl thought dur-do-3pl that child – pl aks-â-ye hamdiga-roi/*j [PredP be modir picture-pl-Ez each other-râ to principal neshun dâd-an]]]] sign gave-3pl ‘The teachers think that the children showed each other’s pictures to the principal.’ b. mo’allem-âj [TP [aks-â-ye hamdiga-roi/*j] k fekr mi-kon-an [CP ke [TP’ [vP [bachche-hâi tk ] [PredP be modir neshun dâd-an ]]]]] c. [aks-â-ye hamdiga- roi/*j ] k mo’allem-â fekr mi-kon-an

176

Chapter 5: Scrambling, Scope, and Binding

[CP ke [TP [vP bachche-hâi tk [PredP be modir neshun dâd an]]]] The reciprocal in (28) can only be bound by the embedded subject. Its appearance in the higher position does not affect the binding relation. Note that the reciprocal moves first into the Spec of the matrix vP in order to be able to move into the higher phase, the matrix discourse domain, in (28b) and (28c). In the Spec of the matrix vP, the copy of the reciprocal is c-commanded by the copy of the matrix subject that is base generated in the higher Spec of vP, as shown in (29). (29) [Subject [Reciprocal [vP tS tAnaphor … Nevertheless, a binding relation between the matrix subject and the reciprocal is not possible. The data thus far show that Principle A is satisfied by the availability of an antecedent by Merge within its own lexical phase, and that scrambling does not affect this relation. The following data support this idea by clearly indicating that scrambling does not feed binding of anaphors. (30) *[vP [be bachche-hâi]k hamdiga-roi [PredP tk mo’arrefi to child-pl

each other-râ

kard]

introduction did

The indirect object has moved out of the PredP into the Spec of vP on its way to move out of this phase in (30). In that position, it c-commands the direct object. Nevertheless, the binding relation between this element and the direct object is not established. It could be argued that Rizzi’s (1990) Chain Condition is responsible for the ill-formedness of (30). This condition states that an anaphor cannot c-command the trace of its antecedent. However, the trace in (30) represents the whole PP containing the antecedent, not the antecedent itself. Furthermore, (31) indicates that the ungrammaticality of this sentence must be independent of the Chain Condition, since the reciprocal, embedded inside a DP, does not c-command the copy of the phrase containing its antecedent in this example, and therefore, the sentence cannot be subject to Rizzi’s Chain Condition85.

Binding

(31) *[vP [be bachche-hâi]k dust-â-ye to child-pl moarrefi kard introduction did

177

hamdiga-roi [PredP tk

friend-pl-Ez each other-râ

If (30) and (31) are not ruled out by Rizzi’s Chain Condition, what is responsible for their ungrammaticality? The examples introduced in this section suggest that an anaphoric relation is established when the anaphor and its antecedent are in positions where they receive their interpretation. Thus the following generalization: (32) An anaphoric relation between α and β is established where α and β are interpreted. Thus, (3a) is revised as in (33) (33) The anaphoric relation of α and β is established as soon as α is merged with the category γ that contains β, where α and β are interpreted. Anaphoric binding in English supports the generalization in (33). One of the well-known properties of an operator movement, such as wh-movement in this language, is that the surface position and the intermediate copies may enter the anaphoric binding relations. Consider the following example, borrowed from Barss (1996). (34) John wonders which picture of himself Max said Sam saw where t yesterday. The reflexive in (34) can be construed with John, Max, and Sam. The same situation holds in the case of the following examples. (35) a. John asked which pictures of himself Bill bought. b. John asked which pictures of himself Bill said Dan bought. This type of binding is not allowed in Persian, as we observed in (28). How can we explain the distinction between Persian and English in this regard? The answer might have to do with the intermediate landing

178

Chapter 5: Scrambling, Scope, and Binding

sites. That is, wh-phrases take their scope position in the Spec of CP in English. Thus, Spec of CP is the position where the wh-phrase receives its interpretation in this language. Therefore, the intermediate Spec of CP serves for binding relations (see also section 5 in chapter 6 and sections 3.2. and 3.3. in chapter 7 for further discussion). The Spec of vP, in contrast, is the position where the local agent and specific subject and object are interpreted (see chapter 3), and therefore, that position serves as a binding position for those elements. All other elements move into Spec of vP purely for the sake of movement into a higher phase. Thus, they cannot enter an interpretive relation from that position (cf (30) and (31)). If this suggestion is on the right track, the anaphoric binding relations in English are predicted by (33).

3.2. Free-standing pronouns Free-standing pronominals are subject to Principle B of the binding theory. This is illustrated by the following example: (36) *[vP Kimeak un-rok [PredP dust dâr-e]] K her-râ friend have-3sg (36) violates Principle B, which states that a pronoun must be free in its domain. If the lexical phase is the domain for pronominals, the ungrammaticality of this sentence is explained, since the pronoun is bound by the subject in its binding domain. Scrambling feeds Principle B, as evident by the following contrast. (37) a. Rahjuei goft [CP ke [TP [vP Kimeak [DP pedar-e un-roi/*k] R said that K father-Ez his/her-râ dust dâr-e ]]]] friend have-3sg ‘Rahjue said that Kimea loves his/*her father.’ b. [TopP [Rahjuei]m [TP [pedar-e un-ro*i/*k]j [vP tm tj goft [CP ke [TP [vP Kimeak [VP tj dust dâr-e]]]] In (37a), the free-standing pronoun inside the possessive DP can only refer to the matrix subject. In (37b), the DP containing the pronoun has

Binding

179

moved long-distance into the matrix discourse domain via the matrix vP, where it is c-commanded by the copy of the matrix subject inside the vP. The fact that the matrix subject cannot be considered the antecedent of the pronoun clearly shows that scrambling feeds Principle B. Furthermore, scrambling does not bleed Principle B, as shown by the following example. (38) a. *Kimea goft [CP ke [TP [vP Parviz Ashley-roi [PredP be dust-e K said-3sg P A-râ to friend-Ez kard ] uni mo’arrefi her introduction did-3sg b. *[TP [Ashley-roi]k [vP Kimea goft [CP ke [TP [vP Parviz tk [PredP be dust-e uni mo’arrefi kard ] The free-standing pronoun is bound by the object DP in (38a), rendering the sentence ungrammatical. In (38b), the embedded direct object has moved out of its own phase. Nevertheless, the free-standing pronoun cannot be co-indexed with it. The data in this subsection support the statement in (3c). That is, scrambling may feed Principle B, but does not bleed it.

3.3. R-expressions R-expressions are subject to Principle C of the binding theory. This is illustrated by the following example. (39) *[vP unk Kimea-rok [PredP dust dâr-e]] she K-râ friend have-3sg The sentence in (39) violates Principles C, since the R-expression has to be free everywhere, while it is bound by the pronoun in this sentence. The following data indicate that scrambling does not bleed Principle C. (40) a. *Sepide fekr mi-kon-e [CP ke [TP [vP uni Kimea-roi [PredP S thought hab-do-3sg that she K-râ dust dâr-e]]]] friend have-3sg Intended meaning: Sepide thinks that shei loves Kimeai.’

180

Chapter 5: Scrambling, Scope, and Binding

b. *[TP [Kimea-roi]j Sepide fekr mi-kon-e [CP ke [TP [vP uni [PredP tj dust-dâr-e]]]]] The R-expression in the embedded clause in (40a) cannot be bound by the local subject, since this binding relation would violate Principle C. Dislocating this DP to a c-commanding position does not bleed Principle C, as evidenced by (40b). The following examples show that scrambling feeds Principle C. (41) a. Rahjue goft [CP ke [TP [vP [DP pedar-e Kimeak] un-ok R said that father-Ez K her-râ [PredP dust dâr-e]]]] friend have-3sg ‘Rahjue said that Kimea’s father loves her.’ b. *Rahjue goft [CP ke [TP [un-ok]i [vP [DP pedar-e Kimeak] ti [PredP dust dâr-e]]]] The pre-scrambling sentence in (41a) is fine. The string in (41b) is subject to Principle C, since scrambling has placed the pronominal in a position where it c-commands the R-expression. Clearly, scrambling has created a context where Principle C is violated. Furthermore, LDS feeds Principle C, as evident in the following example. (42) *[TP [un-ok]i Rahjue goft [CP ke [TP [vP [DP pedar-e Kimeak] ti [PredP dust dâr-e]]]]] The pronoun has moved long distance into the discourse domain of the matrix clause in (42), rendering the sentence ill-formed on a reading where the pronoun is co-indexed with the R-expression it c-commands. The data in this subsection once again support the statement in (3c), where it is suggested that scrambling may feed Principle C, but may not bleed it.

3.4. Clitic pronominals Clitic pronominals differ from the free-standing pronominals discusssed in 3.2. in that DP, rather than vP, is their binding domain, similar to pos-

Binding

181

sessive pronominals in English. That is, they must be free in the DP that contains them. Scrambling, however, does not destroy their binding relation, similar to reciprocals discussed in 3.1. Consider the following data. (43) a. [vP Kimeai pedar-esh-roi/k [PredP dust dâr-e]] K father-her-râ friend have-3sg ‘Kimea loves her father.’ b. [TP [pedar-esh-roi]k [vP Kimeai tk [PredP dust dâr-e]]]

(44) a. [vP Kimeai kif-esh-roi/*k [PredP be Sepidek dâd K purse-her-râ to S gave-3sg ‘Kimea gave her purse to Sepide.’ b. [TP kif-eshu/*k -ro]j [vP Kimeai tj [PredP be Sepidek dâd]]]

The binding relation between the clitic pronoun and its antecedent is established by Merge, and is not affected by scrambling in (43) and (44). Unlike reciprocals, however, scrambling feeds binding in the case of clitic pronouns. That is, it creates a new binding relation on the basis of the derived c-command relation between the antecedent and the clitic pronoun. (45) Kimeai [vP [be Sepidek]j kif-esh-roi/k [PredP tj dâd

The binding relation between the indirect object and the clitic pronominal is established by the movement of the former into the Spec (edge) of vP, a position that is higher than the clitic pronominal. Thus, clitic pronominals are bound purely on the basis of a c-command relation between them (or their copy) and their antecedent. Therefore, these elements are not subject to the generalization stated in (32) with respect to anaphors. The sentence in (45) is an instance of anti-WCO. Data similar to (45) have led some authors to suggest that clause-bound scrambling is an Amovement (Mahajan 1990, 1994). However, (46) shows that LDS has the same effect. (46) a. [vP Kimeak be dust-eshk/*j goft [CP ke [vP pro Sepida-roj K to friend-her said that S-râ

182

Chapter 5: Scrambling, Scope, and Binding

[PredP da’vat karde]]]] invitation did-3sg ‘Kimea told her friend that she has invited Sepide.’ b. [TP [Sepida-roj]i [vP ti Kimeak [PredP be dust-eshk/j goft [CP ke [vP pro ti [PredP da’vat karde]]]]]] In (46b), the embedded direct object has moved into the matrix discourse domain via the matrix vP. Scrambling into the edge of the matrix vP allows this element to bind the clitic pronominal. Unless we assume that LDS is also A-movement, the grammaticality of (46b) is not accounted for86. (See chapter 7 where the relevance of typology of movement is discussed). The properties of clitic pronouns are summarized in (47). (47) a. The binding relation of a clitic pronoun is determined by Merge. b. Scrambling feeds the binding of clitic pronouns, bleeding Principle B. The property in (47b) contradicts the statement in (3c) which, suggests that scrambling does not bleed Principle B. This means that clitic pronouns are special cases that differ from free-standing pronouns. In the case of clitic pronouns, it is the surface c-command relation that feeds binding, as mentioned before. The distinction between free-standing pronominals and their clitic counterparts might be due to their distinct binding domain: while the lexical phase is the domain for the former, DP serves as the binding domain for the latter. We have no explanation for why the domain distinction provides the difference we observed in this section.

3.5. Summary The generalizations in (3) are restated in (48), where (3a) is replaced by (33), in (48a). (48) a. The anaphoric relation of α and β is established as soon as α is merged with the category γ that contains β, where α and β are interpreted. b. Scrambling neither feeds nor bleeds Principle A.

Binding

183

c. Scrambling may feed Principles B and C, but does not bleed them. The statement in (48b) supports the radical reconstruction theory proposed for scrambling (Saito 1985 and work thereafter, Saito and Fukui 1998, among others). The second part in (48c) is in accordance with Lebeaux (1998), who argues that Principles B and C must hold everywhere. However, the first part of this statement rejects the radical reconstruction theory, as well as Lebeaux’s insight, since it states that scrambled elements may be subject to Principles B and C in their surface positions. Note that even LDS does not support the radical reconstruction theory. The statement in (48c) further suggests that Principles B and C have specific properties that differ from Principle A. That is, while movement does not feed Principle A, it does so with respect to Principles B and C. Finally, the discussion of clitic pronouns shows that these elements differ from free-standing pronouns in that scrambling creates new binding relations, thus bleeding Principle B of the binding theory in their case. This shows that clitic pronominals are not subject to principle B in the same way that free-standing pronominals are.

3.6. Cross-linguistic analysis of binding In the previous section we discussed binding relations in Persian. In this section, we consider similar facts in other (scrambling) languages that reveal the same properties, and thus support the conclusions presented in (48). We will see, however, that (48b) may partially be parameterized. 3.6.1. German The specific object binds an anaphor in the indirect object position in German. If the direct object receives its interpretation in Spec of vP, as suggested for Persian in chapter 3, the anaphoric binding relation is established by (48a) in German as well. This is illustrated by (49a). Furthermore, (49b) confirms (48b) which suggests that scrambling does not bleed Principle A. (49) a. … dass [wir die Fraueni einanderi vorgestellt haben] that we Art-Acc women each other introduced have ‘… that we have introduced the women to each other.’

184

Chapter 5: Scrambling, Scope, and Binding

b. *… dass [wir den Fraueni einanderi ti vorgestellt haben] that we Art-Dat women each other introduced have The antecedent precedes and c-commands the anaphor in (49a). The illformedness of (49b) indicates that scrambling does not bleed binding relation in German, similar to Persian (but see footnote 85). Another piece of evidence supporting the previous observation is provided by the following contrast. These data are borrowed from Moltmann (1991). (50) a. [was für information über sich selbsti] gibt what sort of information about himself (Acc) gives [IP dieser Arzt nie [VP einem patienteni ]] this doctor never a patient (Dat) ‘What sort of information about himselfi (does) this doctor never give a patienti?’ b. *[welche informationen über sich selbsti ] gibt [IP dieser which information about himself (Acc) gives this Arzt nie [VP einem patienteni ]] doctor never a patient (Dat).’ The wh-word was für ‘what sort of’ in (50a) is nonspecific, while welche ‘which’ in (50b) is specific. Both wh-phrases contain a reflexive pronoun. Although the nonspecific object allows the reflexive to be bound by the dative DP, the specific object precludes this option. Our clausal architecture discussed in chapter 3 and the anaphoric relations proposed in (48) in this chapter account for the contrast in (50): the specific object in (50b) receives its interpretation in Spec of vP. Thus, the anaphor contained in this DP remains unbound within the phase that contains it. The nonspecific object, in contrast, is merged within the VP (our PredP) and is interpreted within that domain. Thus, the anaphoric binding relation is established by Merge in a position where the elements involved receive interpretation, as suggested in (48a). The binding relations in the two sentences in (50) are presented in (51). (51) a. [was für informationen über sich selbsti ]j gibt [TP dieser Arzt nie [VP einem patienteni tj]]

Binding

185

b. *[welche informationen über sich selbsti ]j gibt [TP dieser Arzt nie [VP tj einem patienteni ]] *

Unlike Persian, however, the anaphor may not c-command the antecedent at Spell-out in German: (52) a. *Gestern haben wir sichi die Katzei ti im Spiegel yesterday have we itself-Dat Art-Acc cat in mirror gezeigt shown (Santorini 1991) b. *… dass [wir einanderi die Fraueni ti vorgestellt haben] that we each other Art-Acc women introduced have We saw in previous sections that scrambling feeds Principles B and C. The data in (52) suggest that this movement may feed Principle A as well, at least in some languages. Thus we revise (48b) by including this parametric distinction, as in (53b)87. (53) a. The anaphoric relation of α and β is established as soon as α is merged with the category γ that contains β, where α and β are interpreted. b. Scrambling does not bleed Principle A. c. Scrambling does not feed Principle A (subject to parametric differences). d. Scrambling may feed Principles B and C, but does not bleed them. 3.6.2. Hindi Consider the following data in Hindi. (54) a. raam-nei apniii/*j kitaab mohan-koj loTaaii Ram-subject self book Mohan-IO returned-perf-f ‘Rami returned self’si/*j book to Mohanj.’ b. apniii/*j kitaab raam-nei mohan-koj loTaaii (Mahajan 1990: 34-5) The direct object ‘self’s book’ cannot be bound by the indirect object in (54a), but is bound by the subject. Once again, this is because the specific

186

Chapter 5: Scrambling, Scope, and Binding

direct object receives its interpretation in Spec of vP, a position higher than the indirect object. (54b) supports the argument that binding relations are established by Merge (cf. 53a). It also suggests that scrambling does not necessarily feed Principle A, similar to Persian and unlike German (cf. 53c). The grammaticality of the following data is explained along the same lines. (55) a. raajaa-ne kon sii daasiii uskei pitaa-ko LOTAA dii king-subject which maid-DO her father-IO return givepst-f ‘Which maidi did the king return to heri father.’ b. raajaa-ne sab daasiyaaNi unkei pitaa-ko LOTAA diiN king-subject all maids-DO her father-IO return givepst-f-p ‘The king returned all the maidsi to theiri father.’ (Mahajan 1990: 27-28) Once again, the well-formedness of these examples is predicted by (53a), as the indirect object is c-commanded by the direct object in Spec of vP. The following contrast is also accounted for by (53a)88: (56) a. *mE-ne ek du:sre-ki tasvi:ro-ko bacco-ko di-yaa I-Erg each-other-Gen pictures-Acc children-Dat gave-perf b. mE-ne bacco-ko ek du:sre-se milaa-yaa I-Erg children-Acc each other-with introduced-perf ‘I introduced the children to each other.’ The anaphor is not bound in (56a), explaining the ungrammaticality of the sentence. 3.6.3. Korean Similar data are observed in Korean. (57) *Yonghi-nun [sero-uyi kyoswu-lul] kutul-ekei Y-Nom each other-Gen professor-Acc they-Dat sokayha-ss-ta introduce-pst-Ind ‘Yonghi introduced each other’s professors to them.’ (Kim 1992: 158)

Binding

187

Kim follows the traditional tendency by base-generating the direct object in a position adjacent to the verb. He rules out the sentence in (57) by suggesting that the direct object moves to the left of the indirect object. He further argues that clause-bound scrambling is an instance of A-movement that does not allow reconstruction, and thus the anaphor cannot be bound by its antecedent at LF. Within our system, this sentence is straightforwardly ruled out by (53a): the specific direct object is interpreted in the Spec of vP, a c-commanding position with respect to the indirect object. The anaphor contained in this DP cannot be bound by the indirect object, rendering the sentence ungrammatical. The following sentences support (53a), and show that Korean does not reveal a parametric property with respect to (53c). (58)

sero-uyi kyoswu-lul, kutul-ii Yonghi-eke each other-Gen professor-Acc they-Nom Yonghi-Dat sokayhassta introduced ‘Each other’s professors, they introduced to Yonghi.’ b. *sero-uyi kyoswu-lul, Yonghi-nun kutul-ekei *each other-Gen professors-Acc Yonghi-Nom they-Dat sokayhassta introduced ‘*Each other’s professors, Yonghi introduced to them.’ (Kim 1992: 161)

The structures of (58a) and (58b) are provided in (59a) and (59b), respectively. (59) a. [sero-uyi kyoswu-lul]k, kutul-ii [vP tk Yonghi-eke b. *[sero-uyi kyoswu-lul]k, Yonghi-nun [vP tk kutul-ekei *

sokayhassta] sokayhassta The binding relation between the subject and the direct object is established by the time the former is merged with the category containing the latter in (59a). The movement of the anaphor to a c-commanding position does not feed Principle A, supporting (53c). In (59b), the anaphor re-

188

Chapter 5: Scrambling, Scope, and Binding

ceives its interpretation in Spec of vP, and thus cannot be bound by the indirect object in a lower position. Both sentences confirm (53a). The example in (60) was introduced in chapter 2 (cf. (49)). The grammaticality of this sentence is explained within the system we are proposing here. (60) kutul-uli Yonghi-nun [sero-uyi kyoswu-eke] sokayhayssta they-Acc Y-Nom each other-Gen professor-Dat introduced ‘Them, Yonghi introduced to each other’s professors.’ (Kim 1992: 159) The structure of this sentence is the one in (61). Given (53a), there is no need to propose an intermediate A-position (cf. (50) in chapter 2) to explain the grammaticality of this sentence. (61) kutul-uli Yonghi-nun [vP ti ] [sero-uyi kyoswu-eke] sokayhayssta Here are some additional data borrowed from Cho (1994): (62) Mary-ka [John-uy sensayng]k-ul ku-eykey sokayhyessta M-Nom Ji-Gen teacher-Acc hei-Dat introduced ‘Mary introduced John’s teacher to him.’ (Cho 1994: 177) (63) *nay-ka [caki-uyi sensayng ]-ul John-eykeyi sokayhayessta I-Nom self-Gen teacher-Acc J-Dat introduced ‘*I introduced self’s teacher to John.’ (Cho 1994: 180) Cho, following a tradition established in Mahajan (1990), suggests that the object has moved into the Spec of AGRo in these examples, an instance of A-movement, and thus there is no reconstruction. The movement of the object in (62), therefore, bleeds Principle C, since the R-expression is no longer c-commanded by the pronoun. Cho explains the ungrammaticality of (63) by suggesting that the movement of the direct object into the Spec of AGRo leaves the anaphor unbound. The contrast between (62) and (63) is straightforwardly explained based on (53a) with no need to stipulate the existence of AGRo. Furthermore, if the object is interpreted in Spec of vP, as suggested in this work, there is no need to assume that scrambling may bleed Principle C, contrary to what Cho’s argumentation suggests.

Binding

189

3.6.4. Japanese The grammaticality of (64) is accounted for, once again, by (53a). (64) [Karerai-o]j [Masao-ga [vP tj [otagaii-no sensei]-ni they-Acc M-Nom each other-Gen teacher-to [syookaisita]]] (koto) introduced fact ‘Themi, Masao introduced to each otheri’s teacher.’ The object is interpreted in the Spec of vP where it c-commands the indirect object within its domain. Thus, the binding relation is established where the antecedent receives its interpretation, confirming (53a). The following example is borrowed from Miyagawa (1997). (65) John-ga gakusei-tatii-o otagaii-ni syookaisita John-Nom students-Acc each other-Dat introduced ‘John introduced the students to each other.’ (Miyagawa 1997: 7) Once again, the example in (65) is consistent with (53a)89. Finally, the following example, taken from chapter 2, was ruled out by the assumption that LDS is subject to radical reconstruction. (66) *?karerai-o [Masao-ga [otagaii-no sensei]-ni [CP [IP Hanako-ga ti they-Acc M-Nom each other-Gen teacher-to H-Nom hihansita] to] itta] (koto) criticized Comp said fact] ‘Themi, Masao said to each otheri ’s teachers that Hanako criticized.’ This sentence, however, is explained by (53a), similar to other cases discussed in this section: the anaphor is unbound in the position in which it receives interpretation. 3.6.5. Russian The following data, taken from Bailyn (1999), support the proposals advanced thus far. (67) a. Ivan predstavil Petrovyx drug drugu Ivan introduced Petrovs-Acc each other-Dat ‘Ivan introduced the Petrovs to each other.’

190

Chapter 5: Scrambling, Scope, and Binding

b. Ivan predstavil drug drugu Petrovyx Ivan introduced each other-Dat Petrovs-Acc ‘Ivan introduced the Petrovs to each other.’ c. ?Ivan predstavil Petrovym drug druga Ivan introduced Petrovs-Dat each other-Acc d. *Ivan predstavil drug druga Petrovym Ivan introduced each other-Acc Petrovs-Dat The sentences in (67a) and (67d) confirm (53a). The one in (67b) shows that scrambling does not feed Principle A, as suggested by (53c). The sentence in (67c) is marked as awkward by Bailyn. This sentence shows that scrambling does not completely bleed Principle A in this language, as suggested by (53b).

3.7. Binding and phase theory Binding relations of pronouns seem to provide some problems for a derivational theory based on phase. Consider the sentence in (37a), repeated below in (68). In this example, the pronominal in the embedded clause is co-indexed with the subject in the matrix clause. (68) Rahjuei goft [CP ke [TP [vP Kimeak [DP pedar-e un-roi/*k ] R said that K father-Ez his/her-râ dust dâr-e]]]] friend have-3sg ‘Rahjue said that Kimea loves his/*her father.’ If interpretation applies cyclically within a phase, then by the time the derivation reaches the matrix clause, there is no access to the lower phases any more. How can a binding relation between the matrix subject and the embedded object be established within phase theory? It seems that co-reference between the subject in the higher clause and the embedded object can only be established representationally, rather than derivationally90. This suggestion can be extended to the interpretation of R-expressions, as confirmed by the following examples. (69) a. *Hei told me that Johni loves Mary. b. *uni goft ke pro Kimeai barande mi-sh-e she said that K winner dur-become-3sg

Conclusion

191

The violation of Principle C in (69a) and (69b) requires interpretation (or rather lack of it) to be established on a representational basis. (70) provides yet another example. (70) a. *uni fekr mi-kon-e [CP ke [TP [vP hame Kimea-roi [PredP she thought hab-do-3sg that all K-râ dust dâr-an ]]]] friend have-3pl In this example, the two co-indexed elements are in different phases. Thus, Principle C functions on the basis of the complete sentence, rather than phase-by-phase interpretation. Thus, we conclude that binding interpretations are not established by derivation, but rather by representation (see also section 6 in chapter 6 and section 4 in chapter 7 on this issue).

4. Conclusion The main goal of this chapter was to provide further evidence showing that scrambling is not semantically vacuous. We saw that phrasal rearrangements in a language like Persian in fact have an impact on the semantic interpretation of the output. This conclusion is based on the interaction of quantified elements, on the one hand, and anaphoric binding relations, on the other. With respect to scope, it was shown that scrambling may create ambiguity when two or more quantified elements interact with each other. We also discussed cases where this operation may create or rescue scope blocking effects. As for binding relation, the conclusions in (53), repeated below, were established. (53) a. The anaphoric relation of α and β is established as soon as α is merged with the category γ that contains β, where α and β are interpreted. b. Scrambling does not bleed Principle A. c. Scrambling does not feed Principle A (subject to parametric differences). d. Scrambling may feed Principles B and C, but does not bleed them.

192

Chapter 5: Scrambling, Scope, and Binding

The statement in (53d), where it suggests that scrambling may feed Principles B and C, contradicts the traditional assumption that scrambled elements are radically reconstructed. The same conclusion is true of (53c) in those languages that allow scrambling to feed Principle A (cf. German). We saw that an anaphor contained in a wh-phrase can be bound by a DP that c-commands the wh-phrase in the intermediate CP in a language like English, which exhibits structural wh-movement. In Persian, by contrast, the anaphor can be bound only in its original phase, where it receives interpretation. This distinction is captured by (53a): if the Spec of CP is the position where wh-phrases receive interpretation, a DP merged in a c-commanding position can bind the anaphor in Spec of CP in the course of the derivation (see also section 5 in chapter 6 and sections 3.2. and 3.3. in chapter 7). The discussions in this chapter have additional consequences. First, they show that Principle A differs in certain ways from Principles B and C. That is, while the former is not fed by movement in most scrambling languages, the latter are. Second, we saw that German, unlike other scrambling languages, allows scrambling to feed Principle A. If the first movement out of vP is into the Spec of this projection, as assumed within phase theory, we can no longer account for this parametric distinction in terms of an A-A’ distinction, since Spec of vP would be considered an A position in German and an A’ position in the rest of scrambling languages discussed in this chapter. Finally, we saw that a movement theory based on phase cannot account for the binding relation between a pronominal and its antecednet, nor for violations of Principle C. This issue will be further discussed in chapter 7. Notes 80. This issue was first addressed by Huang (1982) with respect to Chinese. See also Saito (1994c) regarding Japanese. 81. See Fox (1999) for similar analysis with respect to A’-chains. 82. We have no explanation for the contrast between (12a) and (14) at this point. These data are simply intended to show that scope is determined by the surface structure, and that scrambling has an effect on the interpretation of the clause and its truth value. 83. Hindi is another language in which the reflexive is subject-oriented (Dayal 1994). 84. The reflexive xod ‘self’ seems to be object oriented in certain contexts: (i) a. Sepidej Kimea-râk nesbat be vaz’-e-xodj/k bie’tenâdid S K-râ with respect to situation-Ez herself indifferent saw-3sg ‘Sepide found Kimea indifferent toward her own situation’ b. Kimeak Sepida-râj be âyande-ye xodj/k omidvâr kard K S-râ to future-Ez self hopeful did-3sg ‘Kimea made Sepide hopeful toward her own future.’

Notes

193

c. Kimeak Sepida-râj az kâr-e xodk/j bâz dâsht K S-râ from work-Ez self back kept-3sg ‘Kimea hold Sepide back from her own work.’ In all these examples, the reflexive seems to refer to either the subject or the object. The direct object, however, is the subject of a small clause in these cases. The structure of (ib), therefore, would be the one in (ii). (ii) [vP Kimeak [PredP [SC Sepida-roj [PP be âyande-ye xodk/j]] omidvâr kard]] Thus, the reflexive is still subject-oriented in these examples. The phrase containing the reflexive may scramble over the object without rendering the sentence completely ill-formed. (iii) ?Kimeak [be âyande-ye xodi/k ]i [PredP [SC Sepidâ-râk ti] omidvâr kard] The awkwardness of this sentence is due to the fact that the written language does not allow scrambling as easily as the spoken language does. 85. German shows a similar restriction observed in Persian with respect to reciprocals (see also section 3.6.1. in this chapter). (i) a. … dass wir die Gäste einander vorgestellt haben. that we the guests-Acc each other-Dat introduced have ‘That we have introduced the guests to each other.’ b. *… dass wir den Gästeni einander ti vorgestellt haben. Unlike Persian, however, German allows scrambling to feed binding when the reciprocal is embedded inside the phrase. (ii) a. *… dass wir Bilder von einander den Gästen gezeigt haben that we pictures of each other-Acc the guests-Dat shown have b. … dass wir den Gästeni Bilder von einander ti gezeigt haben ‘That we have shown the guests pictures of each other.’ The contrast between (ib) and (iib) seems to support Rizzi’s Chain Condition. However, the validity of this condition is questionable considering the discussion in the text. 86. In fact, Browning and E. Karimi (1994) have argued that long-distance scrambling in Persian must also invoke A-movement since it is not subject to WCO. The following example, borrowed from those authors, is similar to the one in (40b) in text. (i) mehry [ali-ro]i be mâdar-eshi goft ke hassan ti zade Mehry Ali-râ to mother-his said that Hassan beaten has ‘Mehry told his mother that Hassan beat Ali.’ (Browning & E. Karimi 1994: 80) 87. It is interesting to note that surface c-command of the antecedent is not required in the case of subjects in German: (i) a. … dass sichi der Fritzi ti schlau vorkommt that himself Art-Nom Fritz sly appears ‘… that Fritz appears sly to himself.’ b. …dass sich einanderi die Kinderi ti geküsst haben that each other Art-Nom children kissed have ‘… that the children have kissed each other.’ It could be argued that the subject is still within the vP in (ia&b), and that the scrambled object occupies the Spec of vP, thus lower than T. In that case, the anaphor will still be c-commanded by T which contains the Agreement, and can be considered as the antecedent of the anaphor. The structure of (ia) is presented in (ii). The anaphor sich is in Spec of vP, thus accessible to T. (ii) [dass [TP [ T [vP sichi der Fritzi [VP ti schlau vorkommt]]] That the subject may stay within the vP by Spell-out in German is supported by psycheverb and unaccusative constructions: the unmarked word order in these cases is Dat – Nom – V.

194

Chapter 5: Scrambling, Scope, and Binding (iii)

… dass ihm das neue Haus gefallen hat that him Art-Nom new house pleased has ‘That the new house has pleased him.’ (iv) … dass dem Taucher die Luft ausgegangen ist. that Art-Dat diver Art-Nom air ran-out is. ‘That the diver ran out of air.’ Furthermore, subjects of stage-level predicates may also remain within the verb phrase, as in (vb) (Diesing 1992). (v) a. … weil Professoren ja doch verfügbar sind. since professors Prt Prt available are ‘Since professors are available.’ (generic reading) b. … weil ja doch Professoren verfügbar sind ‘Since there are professors available.’ (existential reading) (Diesing 1992: 370) The following sentences provide further support for the structure in (ii): the anaphor has moved up to the Spec of CP, rendering the sentence ungrammatical. (vi) a. *Sichi ist der Fritzi schlau vorgekommen. himself is Art-Nom Fritz sly appeared b. *Sichi sind die Fraueni schlau vorgekommen themselves are Art-Nom women sly appeared In (vi), the anaphor is in the Spec of CP, and thus is not c-commanded by Agreement in T. 88. The data in (56) are from Rajish Bhutt (personal communication). 89. Miyagawa (1997) suggests that Japanese allows both ID-DO and DO-ID order. Based on scope relations, he suggests that the direct object and the indirect object are not in a c-command relation with respect to each other. If this is in fact the case, then Japanese is different from the other scrambling languages discussed in this work. However, the binding relation in (65) is still consistent with (53a) if either order is base-generated. 90. This idea is supported independently by resumptive pronouns in Lebanese Arabic, as discussed by Aoun and Li (2003). A short summary of their analysis is presented in chapter 7.

Chapter 6 Long Distance Scrambling and Island Constraints

1. Introduction In this chapter, we concentrate on Long Distance Scrambling (LDS) and its interaction with island conditions. In light of the ongoing discussions in this monograph, we examine the differences between scrambling, on the one hand, and typical operator movements such as structural wh-movement and topicalization, on the other. We see that scrambling is subject to a constraint that blocks LDS, which is different from island constraints observed with regard to long distance operator movements. A closer examination of the two types of movements shows, however, that they are both subject to the same constraint, although each type has its own domain for the application of this constraint. It is suggested that CP is the island domain for operator movements, while vP serves as the island domain for movements into the optional discourse positions (cf. Spec of TopP, FocP and TP). This chapter is structured as follows. In section 2, LDS is compared with long distance operator movement. This comparison shows that the former is immune to the type of constraint that blocks the latter. In section 3, a set of novel and interesting LDS data are discussed. This section is devoted to an analysis of a ‘strange’ constraint that blocks LDS in Persian. In section 4, two possible explanations based on processing effects are examined for this constraint. A syntactic alternative is proposed and defended in section 5. This proposal is then extended to the constraint responsible for blocking operator movements. The discussion in this section sheds further doubt on the typology of movement that provides a distinct division between A and A’ movements, an issue that is picked up for further discussion in chapter 7. The ill-formed examples are revisited in section 6, where it is shown that they become grammatical if the dislocated XP is treated as being generated by Merge rather than Move. This analysis once again raises the question whether the grammar needs an interpretation device based on representation, in addition to the one based on derivation (see also section 3.7. in chapter 5 and section 4 in chapter 7). Section 7 concludes this chapter.

196

Chapter 6: Long Distance Scrambling and Island Constraints

2. Island conditions In this section, we compare LDS with long distance operator movements observed in English and German. We will see that LDS reveals a greater freedom than a typical operator movement with respect to familiar island conditions. First, consider the examples in (1) and (2). (1b) and (2b) clearly violate the MLC, and therefore, are ill-formed. (1)

a. Whoi did you say [CP ti [TP ti saw who]]? b. *Whoi did you say [CP ti [TP who saw ti ]]?

(2)

a. Whoi [TP ti persuaded whom [TP PRO to leave town]]? b. *Whomi did [TP who persuade ti [TP PRO to leave]] town]]?

Saito and Fukui (1998) provide the data in (3) to suggest that scrambling does not violate the MLC. (3)

a. [IP Mary-ga John-ni sono hon-o watasita] M-Nom J to that book-Acc handed ‘Mary handed that book to John.’ b. [IP sono hon-oi [John-nik [Mary-ga tk ti watasita]]] that book-Acc J to M-Nom handed c. [IP John-nik [sono hon-oi [Mary-ga tk ti watasita]]] (Saito & Fukui 1998: 443)

The well-formedness of the examples in (3b) and (3c) suggests that the order of scrambled elements is not crucial, and thus this operation is not subject to the MLC. We saw in chapter 4, however, that scrambling in Persian does obey the MLC, as long as two elements bear the same discourse features and compete for the same position. This claim is attested by the following contrast. (4)

ki bâ ki mi-raghs-e] who with who dur-dance-3sg Who will dance with whom?’ Lit: who with whom is it that will dance?

(5)

*bâ kij ki tj mi-raghs-e]

Island conditions

197

As discussed in chapter 4, Persian lacks structural wh-movement. These elements, however, are subject to movement, and end up in Spec of FocP. The ill-formedness of (5) suggests that the lower wh-phrase may not move by crossing the higher one. Furthermore, the surface order of two scrambled wh-phrases is subject not only to MLC, but also the condition on Shortest Move or Tuck in proposed by Richards (1997, 2001). The following data support this claim. (6)

[FocP ki-ro]i [FocP be ki]j Kimea ti tj mo’arrefi

kard

who-râ to who K introduction did Lit: Who to whom was it that Kimea introduced? (7)

*[FocP be ki]j [FocP ki-ro]i Kimea ti tj mo’arrefi kard

In (7), the highest wh-phrase is extracted first, satisfying the MLC. Then the lower wh-phrase moves into a lower Spec, satisfying Richards’ Shortest Move or Tuck in conditions. This analysis explains the ill-formedness of (7), since the condition on Shortest Move is violated in this example. Consider the following contrast. (8)

kii

[bâ

kij ] pro fekr

mi-kon-i [CP ti tj be-raghs-e]

who with who thought dur-do-2sg Lit. Who with who is it you think will dance? (9)

subj -dance-3sg

* bâ kij kii pro fekr-mi-kon-i [CP ti tj be-raghs-e]

MLC and Shortest Move are obeyed in (8). In (9), the condition on Shortest Move is violated, causing the derivation to crash. Nevertheless, scrambling seems to violate island conditions, as illustrated by the well-formedness of (8). In this example, two elements have been extracted out of the embedded clause, suggesting that LDS does not create

198

Chapter 6: Long Distance Scrambling and Island Constraints

an island. That is, scrambling triggered by focus has dislocated both whphrases in (8), without rendering the sentence ungrammatical. The example in (10) shows that LDS of two elements into two distinct discourse positions does not create a syntactic island either. (10) [TopP ketâb-â-roj [FocP KIMEAi momken-e [CP ti tj be Sepide book-pl-râ K possible-3sg to Sepide dâde bâsh-e]]] given be-3sg Lit. As for the books, it is KIMEA who is possible to have given (them) to Sepide. The DP ketâb-â-ro ‘books’ is in the Spec of TopP, while the contrastively focused DP Kimea is in Spec of FocP. Their extraction out of the embedded clause does not render the sentence ungrammatical. That the property observed in (10) is not specific to LDS in Persian is attested by the Japanese examples in (3b) and (3c), as well as the one in (11b). (11) a. [IP Bill-ga [CP [IP Mary-ga John-ni sono hon-o B-Nom M-Nom J to that book-Acc watasita] to] itta] (kotto) handed that said fact ‘Bill said that Mary handed that book to John.’ b. [sono hon-oi [John-nik [Bill-ga [CP [IP Mary-ga tk ti that book-Acc J to B -Nom M-Nom watasita] to] itta]]] (kotto) handed that said (fact) (Saito & Fukui 1998: 443) Now compare the data in (10) and (11b), instances of LDS in Persian and Japanese, respectively, with those in (12) and (13), instances of operator movement in English. (12) *[What]i [in the living room]j did Mary say Bill found ti tj?

Island conditions

199

(13) *Whati [to Ben]j do you think [CP [IP Mary will give ti tj ]]]?

The strings in (12) and (13) show instances of wh-movement and topicalization. Both examples are ill-formed, indicating that the moved elements must have gone through the same intermediate positions. This is illustrated in (14) and (15). (14) *[Wh]i [PP]j did Mary say [CP ti tj [IP Bill found ti tj]]

(15) *[Wh]i [PP]j do you think [CP ti tj [IP Mary will give ti tj ]]]

Furthermore, Persian allows LDS and clause bound scrambling in the same complex sentence. This is illustrated in (16). (16) [FocP [be ki]j ] momken-e [CP ke [TP [TopP [gol-â-ro]i] un ti tj to who possible-is that flower-pl-râ she dâde bâsh-e]] given subj-3sg Lit. To whom is it possible that, as for the flowers, she has given (them)? The wh-phrase has moved into the matrix clause, whereas the object is topicalized in the embedded sentence. This freedom is not available in English, as attested by the ill-formedness of (17). (17) *Whati do you think [CP ti [to Benj [IP Mary will give ti tj ]]]?

The same situation holds in German.

200

Chapter 6: Long Distance Scrambling and Island Constraints

(18) *Ich weiss weni

du

sagtest [CP ti Edej habek [IP tj ti

I know who-Acc you said getroffen tk]] met

Ede has-subj

The wh-phrase has moved through the Spec of CP in (17) and (18). Topicalization of embedded elements is not allowed in these cases, as the illformedness of these data indicates. Note that topicalization in the embedded clause triggers the verb to move to the second position in German, revealing a V2 phenomenon normally observed in the matrix clause in this language. In addition to structural wh-movement and topicalization, German displays instances of scrambling. The following example shows that scrambling does not have a blocking effect on wh-fronting. (19) Ich weiss nicht [CP wasi er sagte [dem Fritzj

diese Frau tj ti ]]

I know not what he said Art-Dat Fritz Art-Nom woman geschenkt hat given has Lit: I don’t know what he said this woman gave to Fritz. Obviously, the scrambled element in (19) does not interfere with whmovement. We have suggested thus far in this work that scrambling is triggered by EPPs, and that the moved element ends up in the functional/operator domain. We also showed that when more than one element represents contrastive focus, they occupy multiple specifiers of FocP. Therefore, LDS is not blocked in those cases either. This contrasts with structural operator movements in English and German. That is, wh-phrases move cyclically through the Spec of CP. Although some authors have argued that the landing site of a structural topic is a position other than the Spec of CP, the ill-formedness of the data in this section indicates that they must move through the Spec of CP anyway91. Within the phase theory of MP, the extracted element must move to the edge of each strong Phase (vP and CP) in order to move into the higher phase. That means that LDS

Scrambling and island effects

201

moves each element through the Specs of vP and CP into a higher clause. Operator movement undergoes the same process. If this is universally the case, then why is it that we observe such a sharp distinction between Persian (and Japanese), on the one hand, and English (and German), on the other? We will return to this issue in section 5. The observations discussed in this section raise additional questions. Is LDS immune to island conditions in general? That is, is there any kind of constraint that would block this operation? If so, what would such a constraint be like? How would it fit into UG? We address these issues in the remainder of this chapter.

3. Scrambling and island effects We have suggested in this monograph that scrambling is triggered by EPPs. We have further suggested that the extracted element ends up in the Spec of a discourse-functional head, and that this movement is subject to the MLC in those cases where two elements compete for the same position. We have also seen that LDS does not obey the same type of island conditions as operator movements do. A closer examination of Persian syntax indicates that although LDS does not obey familiar island conditions, there is a constraint that blocks this movement. This section is devoted to a discussion of this constraint. The relevant data are presented in 3.1., followed by a descriptive generalization in 3.2. The summary of this section appears in 3.3.

3.1. The data92 LDS of a subject DP is blocked if there is another subject in the target clause. This is shown in (20)–(23). In the ‘b’ sentences, the blocking element in the matrix clause appears in bold. Intermediate phases are not shown in these data. (20) a. Kimea bâvar mi-kon-e [CP ke bachche-hâ ketâb-â-ro K belief dur-does-3sg that child-pl book-pl râ az Parviz xarid-an] from P bought-3pl ‘Kimea believes that the children have bought the books from Parviz.’

202

Chapter 6: Long Distance Scrambling and Island Constraints

b. *bachche-hâi Kimea bâvar mi-kon-e [CP ke ti ketâb-â-ro

az Parviz xarid-an]

(21) a. Kimea goft [CP ke Rahju ketâb-â-ro az Parviz xaride] K said that R book-pl râ from P bought is ‘Kimea said that Rahju has bought the books from Parviz.’ b. *Rahjuei Kimea goft [CP ke ti ketâb-â-ro az Parviz xaride]

in kâr-ro karde ] (22) a. Parviz fekr mi-kon-e [CP ke ki P thought dur-do-3sg that who this work-râ done is ‘Who does Parviz think has done this job?’ b. *kii Parviz fekr mi-kon-e [CP ke ti in kâr-ro karde]

(23) a. hame mi-dun-an [CP ke un xâbide] all dur-know-3pl that she sleep is ‘Everyone knows that she is sleeping.’ b. *uni hame mi-dun-an [CP ke ti xâbide ]

The embedded subjects in (20b)–(23b) have moved into the matrix clause. In addition to the scrambled subject DP, there is also the matrix subject in that clause that appears in bold. All four sentences are illformed. At first glance, the ungrammaticality of these sentences seems to be the result of a classic that-trace effect. The following data show, however, that the ungrammaticality of (20b)–(23b) is independent of this condition. (24) a. Parvizi be-nazar mi-yâd [CP ke ti xeyli bâhush bâshe ] P to view dur-come that very smart is Lit: Parviz seems that (he) is very smart. b. kii be-nazar mi-yâd [CP ke ti xeyli bâhush bâshe ] who to view hab-come that very smart Lit: Who seems that (he) is very smart.’

is

Scrambling and island effects

203

The embedded subjects have been extracted out of a CP that contains ke ‘that’ in (24). Nevertheless, both sentences are fine. Note that the matrix verb in (24), a so-called raising verb, does not subcategorize for an external argument, as we saw in chapter 3. Thus, this sentence contrasts with those in (20b)–(23b), where two DPs with the same grammatical function appear in the matrix clause93. The contrast between the ‘a’ and ‘b’ sentences in (25) and (26) below shows that the same constraint we observed with respect to the subject DP in (20b)–(23b) holds for scrambled indirect objects as well. Again, the bold elements are the XPs in the matrix clause that share the same grammatical function with the moved YPs, and thus block LDS. Intermediate phases are not shown. (25) a. Parviz be Kimea goft [CP ke pro ketâb-ro be Rahju dâde] P to K said that book râ to R given is ‘Parviz told Kimea that he has given the book to Rahju.’ b. *[be Rahju]i Paviz be Kimea goft [CP ke pro ketâb-ro ti dâde]

(26) a. Parviz az Kimea porsid [CP ke pro ketâb-ro be ki dâde] P of K asked-3sg that book râ to who given is ‘Whom did Parviz ask Kimea that she has given the book to?’ b. *[be ki]i Parviz be Kimea goft [CP ke pro ketâb-ro ti dâde]

The LDS of the embedded indirect object is blocked as long as there is another XP with the same grammatical function in the target clause, illustrated by (25b) and (26b). The same restriction holds in the case of direct objects, as in (27) and (28). (27) a. Kimea pesar-â-ro tashvigh kard [CP ke pro K boy- pl - râ encouragement did that doxtar-â-ro be-bus-an] girl-pl- râ subj-kiss-3pl ‘Kimea encouraged the boys to kiss the girls.’ b. *[doxtar-â- ro]i Kimea pesar-â-ro tashvigh kard [CP ke pro ti be-bus-an]

204

Chapter 6: Long Distance Scrambling and Island Constraints

(28) a. Kimea pesar-â-ro tashvigh kard [CP ke pro K boy-pl-râ encouragement did that ki-ro be-bus-an] who- râ subj-kiss-3pl ‘Who did Kimea encourage the boys to kiss?’ b. *[ki-ro]i Kimea pesar-â-ro tashvigh kard [CP ke pro ti be-bus-an] The data in the ‘b’ sentences in (20)–(23) and (25)–(28) clearly show that an element cannot scramble into a higher clause if another element with the same grammatical function already exists in that clause. The grammatical sentences in (29) are additional pieces of evidence supporting the claim that scrambling is possible only if the moved element does not have a grammatical counterpart in the target clause. (29) a. bachche-hâi be nazar mi-yâd [CP

ke ti xaste bâsh-an ]

child-pl to view dur-come-3sg that tired be-3pl ‘The children seem to be tired.’ b. [ketâb-â-ro]i fekr mi-kon-am [CP ke Kimea ti xaride ] book-pl-râ thought hab-do-3sg that K bought is Lit. As for the books, I think that Kimea has bought. c. [be Arezu]i fekr-mi-kon-am [CP ke Kimea mozu-ro ti gofte] to A thought hab-do-1sg that K ubject-râ told is Lit: As for Arezu, I think that Kimea has told the matter. In (29a), the embedded subject has moved into a clause whose verb does not subcategorize for an external argument. Similarly, in (29b) and (29c), an embedded YP has moved into a matrix clause that does not contain an XP with the same grammatical function. In summary, we saw in this subsection that the presence of a scrambled YP in the target clause is licit as long as there is no XP with the same grammatical function in that clause. Recall that LDS moves an element cyclically through the edges of strong phases. Thus, a scrambled XP in the matrix clause must have moved through the edge of the embedded vP and CP, as well as the edge of the matrix vP. Thus, I propose the following informal

Scrambling and island effects

205

condition. The configuration in (30) represents the strong phase where the two grammatically identical elements appear together. (30) Condition on LDS LDS is blocked in the following configuration *[Phase YPiα XPα ……[ ti ] Where α represents a specific grammatical function (e.g. subject). The condition in (30) does not hold in the case of a typical operator movement, such as wh-movement, as illustrated by the example in (31). (31) Whoi does John think [ti will go to Europe tomorrow?] The grammaticality of (31), on the one hand, and the ill-formedness of the sentences discussed in this subsection, on the other, support the long standing claim that scrambling is different from the typical operator movement.

3.2. Discussion The contrast between the ill-formed and well-formed sentences in the previous subsection might support the idea that LDS is not the result of movement, but rather is generated by the operation of Merge94. That is, the illformed sentences are ruled out since two argument XPs of the same kind cannot be base-generated in a single position. However, scrambling across two clauses provides some evidence against this assumption. That is, the intermediate clause is crucial to the output of the derivation. To illustrate this point, consider first the data in (32) and (33). These sentences show that scrambling across an intervening clause is possible. (32) ketâb-â -roi ehtemâl dâr-e [CP ke Kimea book-pl râ possibility have-3sg that K fekr kon-e [CP man ti be Rahjue dâd-am]] thought do-3sg I to R gave-1sg ‘As for books, it is possible that Kimea thinks that I have given (them) to Rahjue.’

206

Chapter 6: Long Distance Scrambling and Island Constraints

(33) be RAHJUEi ehtemâl dâr-e [CP ke Kimea fekr kon-e [CP man ketâb-â ro ti dâd-am]] ‘It is to RAHJUE that it is possible that Kimea thinks I have given the books.’ However, the movement of a YP is blocked if the intermediate clause contains an XP with the same grammatical function. The relevant examples are provided in (34) and (35). The bold DP in (35) is the blocking element. (34) be-nazar mi-yâd [CP ke Kimea be Arezu gofte to-view dur-come that K to A said is [CP ke Parviz emruz kâr ne-mi-kon-e]]] that P today work neg-dur-do-3sg ‘It seems that Kimea has told Arezu that Parviz does not work today.’ (35) *Parvizi be-nazar mi-yâd [CP ke Kimea be Arezu gofte [CP ke ti emruz kâr ne-mi-kon-e]]]

In (35), the subject of the lowest sentence is scrambled into the highest clause. Recall that the main verb in this sentence does not subcategorize for an external argument. Thus, nothing in the matrix clause prevents the intermediate subject from moving into the matrix clause, as in (36). (36) Kimeai be-nazar mi-yâd [CP ke ti be Arezu gofte [CP ke K to view dur-come that to A told is that Parviz emruz kâr ne-mi-kon-e]]] P today work neg-dur-do-3sg ‘As for Kimea, it seems that (she) has told Arezu that Parviz will not work today.’ The grammaticality of (36) is due to the fact that the embedded subject moves into a subjectless clause, without crossing another subject DP in an intermediate clause. Thus it must be the intermediate subject in (35) that

Scrambling and island effects

207

blocks the movement. This suggestion is in fact supported by the illformedness of the following sentence. (37) *be-nazar mi-yâd [CP ke Parvizi Kimea be Arezu gofte [CP ke ti emruz kâr ne-mi-kon-e]]] In (37), the subject of the most embedded clause has moved into the intermediate clause, where there is another subject. Thus, this sentence is subject to the condition in (30). The implication of this discussion is twofold: first, we have an instance of Move rather than pure Merge in these cases, and second, scrambling applies cyclically through the intermediate phases95. Indirect and direct objects reveal the same restriction we observe in (35). This is illustrated by (38) and (39), respectively. Intermediate traces are ignored. Indirect Object (38) a. *[FocP be RAHJOUEi [vP man fekr

mi-kon-am [CP ke

to R I thought dur-do-1sg [vP Kimea be Parviz gofte [CP ke [vP pro K to P said is ketâb-â-ro ti be-d-e]]]]]]

that

that K

book-pl-râ subj-give-3sg Intended meaning: it is to RAHJOUE, that I think that Kimea has told Parviz to give the books. b. *[ FocP be kii [vP to fekr mi-kon-i [CP ke [vP Kimea to who you thought dur-do-2sg that K be Parviz gofte [C ke [vP pro ketâb-â-ro ti be-d-e]]]]]] to P said is that book-pl-râ subj-give-3sg Intended meaning: it is to whom, that you think that Kimea has told Parviz to give the books.

208

Chapter 6: Long Distance Scrambling and Island Constraints

Direct Object (39) a. *[FocP DOXTAR-A-roi [vP man fekr

mi-kon-am [CP ke

girl-pl-râ I thought dur-do-1sg that [vP Kimea pesar-â-ro tashvigh karde [CP ke [vP pro ti K boy-pl-râ encouragement done is that be-bus-an]]]]]] subj-kiss-3pl Intended meaning: It is the GIRLS, I think that Kimea has encouraged the boys to kiss. b. *[ FocP ki-roi [vP to fekr mi-kon-i [CP ke [vP Kimea who-râ you thought dur-do-2sg that K pesar-â-ro tashvigh karde [CP ke [vP pro ti be-bus-an]]]]]] boy-pl-râ encouragement done is that subj-kiss-3sg Intended meaning: Who is it, you think that Kimea has encouraged the boys to kiss (them). The indirect objects in (38) and the direct objects in (39) cannot scramble into the matrix clause due to the existence of XPs with the same grammatical function in the intervening intermediate clauses. Therefore, the XPs in bold are the ones blocking the movement. Note that the scrambled elements in (38) and (39) bear heavy stress, representing focus. Without the stress, these sentences would become even worse. The restriction we observed in this section is similar to a parallel constraint on Across the Board movement (ATB) in English, as the contrast in (40) and (41) indicates. (40) Who did [Mary kiss t], [Stephanie date t], and [Anne marry t]? (41) *Who did [Mary kiss t], [Stephanie date Bill] and [Anne marry t]? The difference between these two sentences is that the chain is blocked by an overt XP in (41), but not in (40). We will come back to this issue in section 6.

Processing effects

209

3.3. Summary In this section, we saw a number of data that illustrate an interesting constraint governing the application of LDS in Persian. That is, scrambling is blocked when it involves the movement of a YP into a clause that contains an XP with the same grammatical function. For example, an embedded direct object may not move into the target clause that contains another direct object. This constraint holds also in those cases when an XP bearing the same grammatical function as the scrambled YP is present in the intermediate clause. In other words, LDS can apply to a YP as long as there is no competing XP bearing the same grammatical function in the target or the intermediate clause. How can these constraints be accounted for? This issue is addressed in the following two sections.

4. Processing effects The data discussed in the previous section suggest that the presence of two XPs bearing the same function might be difficult to process. In this section, we address this issue by resorting to two processing theories. Pickering and Barry (1991) suggest that the length of time that the processor has to hold the relevant element in memory, before it can be linked to the subcategorizer, is crucial in order to process that element. The contrast in the following data, they suggest, corresponds to the time needed to process the wh-phrases they contain. (42) a. In which box did you put the very large and beautifully decorated wedding cake bought from the expensive bakery? b. Which box did you put the very large and beautifully decorated wedding cake bought from the expensive bakery in? (42a) is less awkward that (42b). According to Pickering and Barry, the time needed to process the element in which box with respect to its subcategorizer put in (42a) is shorter than between the time needed to process which box with respect to its subcategorizer in in (42b). Therefore, the former is easier to process than the latter96. Gibson (1991) suggests that arguments which require thematic roles, but have not yet received such roles (parsing left to right), are associated with memory cost. On the basis of this theory, Gibson and Hickok (1993)

210

Chapter 6: Long Distance Scrambling and Island Constraints

suggest that the difference between the ‘a’ and ‘b’ sentences in (42) is the distance between the wh-phrase and the attachment point of its trace. In (42a), the trace is attached to the verb phrase, whereas in (42b) it is attached to the PP inside the VP. This difference is illustrated by (43). The two relevant nodes are in bold. (43) a. [In which box ] did you [VP put [NP ] t ] b. [Which box ] did you [VP put [NP ] [PP in t ] In other words, the structure in (43b) is more complex than the one in (43a) with respect to the relationship between the wh-phrase and its trace. How can the ill-formed data in the previous section be analyzed in the framework of these two theories? Let us start with Pickering and Barry’s account. Given their theory, the sentence in (38b), repeated below in (44), is difficult to process (from left to right) since we need to cross several nodes to get to the verb that subcategorizes this element. Therefore, it is the length of the time that makes processing and interpreting this and similar sentences difficult. (44) *[ FocP be kii [vP to

fekr

mi-kon-i [CP ke [vP Kimea

to who you thought dur-do-2sg that K be Parviz gofte [C ke [vP pro ketâb-â-ro ti be-d-e]]]]]] to P said is that book-pl-râ subj-give-3sg Intended meaning: it is to whom, that you think that Kimea has told Parviz to give the books. One piece of counter evidence to this proposal is that the same difficulty holds with respect to adjuncts, as in (45)97. (45) *keyi Kimea goft [CP ke [vP pro diruz

shenide

when K said that yesterday heard is [CP ke [vP Arezu ti xune xaride]] that

A

house bought is

Note that this sentence is grammatical if the wh-phrase is interpreted as modifying the matrix verb, or if it is understood as part of the adverbial

Processing effects

211

phrase in the intermediate clause: when yesterday. Since adjuncts are not subcategorized by the verb, the ‘linking’ problem between the fronted element and the subcategorizer does not hold in their case. Thus, the sentence in (45) cannot be ruled out by Pickering’s and Barry’s theory. Now we turn to Gibson’s (1991) and Gibson’s and Hickok’s (1993) theory, and its application to the sentence in (44). The wh-phrase in this sentence requires a theta role. The verb in the intermediate clause could assign this role. But there is another element that needs the same theta role. Thus, the problem with this sentence is that the trace of the whphrase is attached to the lowest VP, although the intermediate clause is the closest domain where it could get its theta role assigned. In other words, the processor sees too many candidates for one theta role, before she gets to the trace of the wh-phrase in the lowest clause. Thus, the complexity of this clause is a burden on memory which in turn renders it unacceptable. Adjuncts provide a problem for this theory as they did with regard to the first one. The reason is that these elements do not require a theta role. Thus the ungrammaticality of (45) cannot be attributed to the complexity of the sentence with respect to theta assignment. The most compelling argument against an explanation based on processing comes from a comparison between the ill-formed sentences discussed in this chapter and their counterparts in a language like English. The following examples are perfectly well-formed, even though they posit the same problem as the scrambling data discussed in this chapter. (46) a. Whoi do you think [ti will fix my car next week?]] b. Whoi did you say [Mary thinks [ti will fix my car next week?]] (47) a. Whoi did you persuade Mary [PRO to kiss ti?]] b. Whoi did you say [Bill persuaded Mary [PRO to kiss ti ]] It takes a long time for the wh-phrase to be linked to its subcategorizer in (46) and (47). The same problem holds with respect to theta marking and the attachment of the trace in these cases. In (47b), for example, there is a transitive verb in the intermediate clause that could serve as the theta assignor for the wh-phrase. Its theta role, however, is used up by Mary. The trace of the wh-phrase is attached to a category lower than the one containing the intermediate transitive verb. Nevertheless, this sentence is perfectly well-formed. Finally, the following dialogue shows that familiarity with the subject matter does not improve the relevant examples.

212

Chapter 6: Long Distance Scrambling and Island Constraints

(48) Speaker A: Kimeai be Arezu goft [CP ke proi fekr mi-kon-e K to A told that thought dur-do-3sg [CP ke proi ketâb-â-ro be Parviz dâde]] that book-pl-râ to P given is ‘Kimea told Arezu that she thinks that she has given the books to Parviz.’ Speaker B: *be RAHJUEk to fekr

RAHJUE

uni be Arezu goft [CP ke proi she to A

told

that

mi-kon-e [CP ke proi ketâb-â-ro tk dâde ]]

thought dur-do-3sg that book-pl-râ given is Intended meaning: It was to RAHJUE that she told Arezu that it was possible that she had given the books. Almost all the information needed to understand the second utterance is provided by Speaker A in the first utterance. Speaker B disagrees only with one portion of the first utterance, and highlights that portion by stressing and scrambling the DP into the matrix clause. In fact, the second utterance is understandable, especially in the context of the first one. However, it is still ungrammatical. The following contexts provide additional pieces of evidence indicating that the ill-formedness of the data discussed in section 3 is not a matter of processing. (49) Speaker A: momken-e [CP Arezui be-xâd [CP ke Parviz uni-ro possible-3sg A subj-want that P her-râ da’vat kon-e]] invitation do-3sg ‘It is possible that Arezu wants that Parviz invites her.’ Speaker B: *KIMEAk momken-e [CP Arezui be-xâd [CP ke tk K

possible-3sg

A

subj-want that

uni-ro da’vat kon-e]] her-râ invitation do-3sg Intended meaning: It is KIMEA that Arezu possibly wants to invite her.

Argument/adjunct structure effects

213

(50) Speaker A: pro fekr mi-kon-i [CP Kimeai be-xâd [CP ke ki thought dur-do-2sg K subj-want that who da’vat-eshi kon-e]] invite-her do-3sg ‘Who do you think Kimea wants to invite her?’ Speaker B: *RAHJUEk pro fekr mi-kon-am [CP Kimeai R thought hab-do-1sg be-xâd [CP ke tk da’vat-eshi kon-e]]

K

subj-want invite-her do-3sg Intended meaning: It is RAHJUE that I think Kimea wants to invite her. The second utterances in (49) and (50), produced by Speaker B, are not difficult to process in the context of the first utterances produced by Speaker A. Nevertheless, they are ill-formed. Scrambling is blocked by the existence of the bold element in the intervening clause in each case.

5. Argument/adjunct structure effects The ungrammatical data in sections 3 and 4 give the impression that LDS is a cyclic movement through the argument positions since scrambling of YP is blocked by the XP bearing the same functional properties. Thus, LDS might be considered an instance of A-movement. This analysis, however, faces some problems. First, it is not compatible with properties of a typical A-movement (e.g., it is not triggered by Case, or EPP in the sense of Chomsky (1995), and creates scope ambiguity (see chapter 5)). The second problem has to do with pro in subject position. If the condition in (30) is a restriction on A-movement into the argument position, we should expect the existence of pro to block such movement. The following data show that this prediction is not borne out, and that the condition in (30) can be maintained only if the grammatical counterpart of the scrambled element is an overt DP. (51) kii pro fekr mi-kon-i [CP ti fardâ

bi-yâ-d]

who thought-dur-do-2sg tomorrow subj-come-2sg ‘Who is it that you think will come tomorrow?’

214

Chapter 6: Long Distance Scrambling and Island Constraints

(52) Kimeai pro fekr

dur-kon-an [CP ti emruz

K thought dur-do-3pl today tu madrese bâsh-e] in school be-3sg ‘As for Kimea, they think (she) is at school today.’ (53) KIMEAi pro fekr mi-kon-am [CP pro mi-xâ-n [CP ti KIMEA thought-dur-do-1sg dur-want-3pl entexâb be-sh-e]] selected subj-become-3sg ‘It is KIMEA that I think they want to be selected.’ In (51) and (52), the embedded subject appears in the matrix clause. The matrix subject is not lexical, and the sentence is fine. In (53), the lowest subject moves into the matrix clause, crossing the intermediate subject. The intermediate subject, as well as the matrix subject, are null referential pronouns, and the result is grammatical. Thus, the condition in (30) needs to be revised, as in (54). (54) Condition on LDS (Revised) LDS is blocked in the following configuration *[Phase YPiα XPα ……[ ti ] Where α represents a specific grammatical function, and XP is Lexical. The final criticism regarding an analysis that suggests that the condition in (54) represents instances of A-movement into an argument position is provided by adjuncts: LDS of adjuncts exhibits the same problem, as evidenced by the sentence in (45), repeated in (55). (55) *keyi Kimea goft [CP ke [vP pro diruz

shenide

when K said that yesterday heard is [CP ke [vP Arezu ti xune xaride ]] that

A

house bought is

Argument/adjunct structure effects

215

The example in (55) shows that LDS of an adjunct is not possible if a similar adjunct exists in the intermediate clause. As mentioned before, this sentence is grammatical if the wh-phrase is interpreted as modifying the matrix verb, or if it is understood as part of the adverbial phrase in the intermediate clause: when yesterday. The contrast in (56) reveals the same point. (56) a. keyi Kimea goft ke

Arezu ti xune

xaride?

when K said that A house bought is ‘‘When did Kimea say that Arezu has bought a house?’ b. *keyi Kimea emruz goft ke Arezu ti xune xaride ‘When K

today said that A

house bought is

The sentence in (56b) is ill-formed since there already is a temporal adverb in the matrix clause, and thus the embedded adverb cannot be scrambled into the higher clause, as predicted by the condition in (54). Note again that this sentence is grammatical if we interpret the wh-phrase as part of the temporal adverb: when today. The problems discussed thus far suggest that (54) is not a constraint on A-movement, and thus YP does not move through the argument position. If the argument structure effect is not responsible for the constraint governing (54), how can this condition be explained? In Karimi (1999c), a discourse functional feature D is suggested to be a property of C, and is assumed to be responsible for scrambling. It is further suggested that this feature would look for the most local element in order to license its movement into the Spec of a discourse-functional head. Therefore, the lower element would be rejected as a possible candidate as long as there is a higher candidate with the same grammatical function. This stipulation would account for the fact that only lexical elements are possible blocking candidates, since null pronouns may not function as focus or topic, and thus are not subject to movement into TopP, TP, or FocP. Although this solution might solve the problem, it is undesirable, since the existence of the discourse feature D in C is not independently motivated. In the remaining portion of this section, we argue that the condition in (54) is in fact a restriction on movement into potential argument/adjunct positions. That is, the ungrammaticality of the sentences we have seen thus far is decided within vP: once an XP moves into vP (matrix or intermediate), it cannot move into a position that is already taken by another

216

Chapter 6: Long Distance Scrambling and Island Constraints

element that has the same grammatical features. Consequently the derivation crashes. The claim that this movement is not a typical A-movement is not only supported by the fact that it is not driven by Case or EPP in the sense of Chomsky (1995), but also by the clear observation that it applies to adjuncts as well as arguments. Two problems remain unexplained. First, how can we account for the fact that the existence of pro does not block scrambling in the same way that the overt DP does? The solution might lie in the fact that the extracted element is competing with a phonologically null element, thus an uneven competition allows the derivation to survive. The second problem has to do with adjuncts. How can movement go through adjunct positions? We saw in chapter 4 that adverbials occupy a position adjoined to vP. Thus, they are part of the vP. Once that position is occupied by a certain adverb, another adverb cannot move there, as we saw in (55) and (56b). Recall that operator movement is not subject to (54), as we observed in (46) and (47), repeated below in (57) and (58). (57) a. Whoi do you think [ti will fix my car next week?]] b. Whoi did you say [Mary thinks [ti will fix my car next week?]] (58) a. Whoi did you persuade Mary [PRO to kiss ti?]] b. Whoi did you say [Bill persuaded Mary [PRO to kiss ti ]] Why does (54) apply to scrambing, but not to operator movement such as wh-movement (and structural topicalization in English and German, as we saw before)? Our response is that elements that undergo operator movement are interpreted in Spec of CP, and therefore, CP becomes an island when two or more wh-phrases undergo structural wh-movement. If structural topicalization is subject to the same island conditions as whphrases, as suggested by Chomsky (1977), and confirmed by the data in (14), (15), (17), and (18) in section 2, CP must be the island domain for those elements as well. In contrast, when two or more elements undergo scrambling, they are interpreted in Spec of the relevant functional phrase: TopP, TP, or FocP. However, since TopP, FocP, and the Spec of TP are optional, vP becomes the relevant phase where island effects surface. This is presented in the following statement:

Argument/adjunct structure effects

217

(59) Domains of Island effects a. CP is the island domain for operator movement. b. vP is the island domain for scrambling What are the justifications for (59)? Scrambled elements are interpreted as topic only when they arrive in Spec of TP or TopP. Thus, the XP that moves into those positions does not bear a specific feature. Therefore, YPα and XPα in (54) are identical in terms of grammatical features when the former moves into the higher vP. This is not true of wh-phrases since those elements have an additional feature, namely the wh-feature. Thus YPα and XPα are not exactly identical when the former, but not the latter, is a wh-phrase in a language that has structural wh-movement. This analysis could be extended to topics in a language where topicalization is a structural movement into the Spec of CP. Scrambled elements that move into the Spec of FocP have additional properties: they either have the wh-feature (in the case of wh-phrases) or stress (in the case of non-wh-elements). We saw in chapter 4, however, that movement of these elements is not obligatory. Thus, there must be a distinction between a strong wh-feature, as in English, and a weak one, as in Persian. While the former is visible in vP, the latter is not. Consequently, the scrambled wh-phrase is treated like any other phrase in vP. Similarly, stress is not visible as a syntactic feature. Therefore, YPα bearing stress is interpreted just like XPα even when it is stressed. Consequently, any scrambled argument is interpreted in terms of its grammatical features within vP, competing with the existing counterpart in that domain. The same situation holds for the scrambled adverbial, since this element is adjoined to vP, competing with the adverb that is already in that position. Based on these discussions, (54) is revised to include not only LDS, but operator movement as well. Thus the Condition on Long Distance Movement (CLDM) stated in (60) represents island conditions for both types of movements. (60) Constraint on Long Distance Movement (CLDM) LDM is blocked in the following configuration *[Phase YPiα XPα ……[ ti ] Where α represents relevant features, and XP is lexical. The constraint in (60) accounts for both scrambling and operator movement. The only difference is the island domain of the two types of move-

218

Chapter 6: Long Distance Scrambling and Island Constraints

ments: while vP is the island domain of the former, CP is the island domain for the latter (see also sections 3.2. and 3.3. in chapter 7). Does (60) account for other scrambling languages? It is well-known that scrambling of subjects is not possible in Japanese. Consider the following sentences, borrowed from Saito (1985). (61) *sonoi okasi-ga

John-ga [ti oisii to]

omotteiru

this candy-Nom J-Nom tasty that thinks ‘This candy, John thinks is tasty.’ (62) *konoi giron-ga

Mary-ga John-ni [ti okasii to]

this argument-Nom M-Nom J-Dat ‘This argument, Mary told John is strange.’

itta

strange that said (Saito 1985: 185, 193)

Saito notes that the ill-formedness of (61) and (62) could be considered to be a processing strategy which assumes the least amount of scrambling possible. After discussing other types of constructions that can improve by pragmatic considerations, he concludes that these sentences must be ill-formed due to a syntactic constraint. Thus, it is conceivable to suggest that this type of scrambling must be subject to (60). Now consider the following sentences representing colloquial Russian. They seem to be compatible with the CLDM stated in (60)98. (63) ty kogo dumajes on ljubit ? you whom think-2sg he loves ‘Who do you think he loves?’ (64) *on kto dumajet pridet pervyj ? he who thinks will come first ‘Who does he think will come first?’ (65) **ty komu skazal jemu ja peredal knigu ? you who-Dat said him-Dat I delivered book ‘Who did you say to him I delivered the book to?’ (Sinicyn 1983)

Representation versus derivation

219

The wh-phrase in (63) is the object of the lower clause, and thus this sentence is not subject to (60). In (64), the moved wh-phrase is a subject, and therefore, there are two phrases bearing the same grammatical function in the matrix clause. A similar situation holds in (65) with respect to the indirect object. Thus, the ill-formedness of (64) and (65) suggests that (60) applies in Russian as well. The ill-formed sentences discussed in this and previous sections improve and become grammatical if the extracted XP is treated as a left-dislocated element, thus not subject to movement. We turn to this issue in the next section.

6. Representation versus derivation We saw that Russian seems to be subject to the constraint in (60). The following sentence, however, contradicts that assumption, since there are two subjects in the same clause99. (66) ??tyi doktor videl kogda ti pod”ezzal? you doctor see when came ‘Did you see when the doctor came?’

(Bailyn 1999)

Bailyn (1999) states that his informants did not like (66), and found it “…marginal only with a certain intonational pattern indicative of Left-Dislocation.” Thus, Bailyn suggests that the embedded subject might be basegenerated in its surface position, coindexed with a null resumptive pronoun in the embedded clause in these cases. This suggestion is in fact supported by Persian facts. That is, the ill-formed sentences discussed in previous sections improve radically if the fronted XP is interpreted as a left dislocated element, coindexed with an overt pronoun, or a null referential pronoun. First consider the ill-formed sentence in (21b), repeated below in (67). (67) *Rahjuei Kimea goft [CP ke ti ketâb-â-ro az Parviz xaride] R

K

said-3sg that book-râ

of P

bought is

Now compare (67) with (68) and (69). In the well-formed sentences, the subject is left-dislocated, followed by a pause.

220

Chapter 6: Long Distance Scrambling and Island Constraints

(68) Rahjue -roi, Kimea goft [CP ke proi /uni fardâ

mi-r-e

R-râ K said that he tomorrow dur-go-3sg orupâ] Europe ‘As for Rahjue, Kimea said that (he) goes to Europe tomorrow.’ (69) Rahjue-roi, Kimea bâvar mi-kon-e [CP ke proi/uni ketâb-â-ro R-râ K belief dur-do-3sg that he book-pl-râ az Arezu xaride] from A bought is ‘As for Rahjue, Kimea believes that (he) has bought the books from Arezu.’ The dislocated DPs in (68) and (69) are followed by the particle râ, and are co-indexed with the (null or overt) subject pronoun in the embedded clause. As we saw in chapter 1, râ marks specific DPs for Accusative Case. Left-dislocated elements are usually marked by this element as well. Thus, we have a case of base-generated DPs in both sentences, and therefore, neither of them is subject to the condition in (60) since the left-dislocated elements in these sentences are not the result of Move, and reside outside of the clause. The relation between the left-dislocated XP and the overt/resumptive pronoun can only be established representationally rather than derivationally, since by the time the left dislocated element is inserted into the structure, there is no access to the lower phases derivationally. These examples provide additional evidence to the claim in chapter 5 that binding relations cannot be solely based on derivational syntax (see also section 4 in chapter 7). Now consider the case of a scrambled subject that crosses an intermediate subject. The ill-formed sentence in (35) is repeated below in (70). (70) *Parvizi be-nazar mi-yâd [CP

ke

Kimea be Arezu

P to-view dur-come-3sg that K gofte [CP ke ti emruz kâr ne-mi-kon-e]]] said-is

to A

that today work neg-dur-do-3sg

Compare (70) with its left-dislocated version in (71).

Representation versus derivation

(71) Parviz-oi, be-nazar mi-yâd [CP ke

221

Kimea be Arezu gofte

P-râ to-view dur-come that K to A [CP ke proi /uni emruz kâr ne-mi-kon-e]]

said is

that he today work neg-dur-do-3sg ‘As for Parviz, it seems that Kimea has told Arezu that (he) does not work today.’ A similar situation holds for left-dislocated objects. First consider the illformed sentence in (27b), repeated below in (72). (72) *[doxtar-â- ro]i Kimea pesar-â-ro tashvigh girl-pl-râ be-bus-an] subj-kiss-3pl

K

kard [CP ke pro ti

boy-pl-râ encourage did

that

This sentence becomes well-formed when the object is left-dislocated, as in (73). (73) un doxtar-â-roi, Kimea pesar-â-ro tashvigh kard [CP ke pro un-â-roi that girl-pl-râ K boy-pl-râ encouragement did that be-bus-an] subj-kiss-3pl ‘As for those girls, Kimea encouraged the boys to kiss them.’ The overt pronoun can be replaced by a clitic pronoun attached to the verb, as in (74). (74) un

doxtar-â-roi, Kimea pesar-â-ro tashvigh kard [CP

that girl-pl-râ K be-bus-an-eshuni

ke

boy-pl-râ encouragement did that

subj-kiss-3pl-them ‘As for those girls, Kimea encouraged the boys to kiss them.’ An additional example is provided in (75).

222

Chapter 6: Long Distance Scrambling and Island Constraints

(75) ketâb-â-roi, Kimea bachche-hâ-ro tashvigh kard [CP book-pl-râ K child-pl-râ hama-shun -roi bexun-an

ke

encouragement did that

all-them-râ read-3pl ‘As for the books, Kimea encouraged the students to read all of them.’ In (75), the left-dislocated DP is co-indexed with the clitic pronoun attached to the quantificational object in the embedded clause. We saw that a scrambled object that crosses an intermediate object causes the derivation to crash. Left-dislocation saves the structure in these cases, similar to those we observed with respect to subjects. Consider first the ill-formed example in (39a), repeated below in (76). (76) *[FocP DOXTAR-A-roi [vP man fekr

mi-kon-am [CP ke

girl-pl-râ I thought dur-do-1sg that [vP Kimea pesar-â-ro tashvigh karde [CP ke [vP pro ti K boy-pl-râ encouragement done is that be-bus-an ]]]]]] subj-kiss-3pl Intended meaning: It is the GIRLS, I think that Kimea has encouraged the boys to kiss. Compare (76) with its left-dislocated version in (77). (77) doxtar-â-roi, [TP man fekr

mi-kon-am [CP ke

Kimea

girl-pl-râ I thought dur-do-1sg that K pesar-â-ro tashvigh karde [CP ke pro be-bus-an-eshuni]] boy-pl-râ encouragement done is that subj-kiss-3pl ‘As for the girls, I think that Kimea has encouraged the boys to kiss them.’ One final set of data calls for discussion. In section 3, we saw instances of Across the Board (ATB) data exemplified in (40) and (41), repeated below in (78) and (79).

Conclusion

223

(78) Who did [Mary kiss t], [Stephanie date t], and [Anne marry t]? (79) *Who did [Mary kiss t], [Stephanie date Bill] and [Anne marry t]? These data are different from those discussed in sections 3 and 5 since they cannot be considered the result of movement, and thus are not subject to the condition in (60). This is so since there is only one wh-phrase in each sentence, and thus no violation of island condition can be responsible for the ungrammaticality of (79). Furthermore, the wh-phrase is not interpreted within vP in a language with structural wh-movement. Therefore, (79) cannot have been blocked inside the intermediate vP by virtue of the existence of the overt object. Consequently, the connection between the object positions is representationally blocked by the intermediate overt object. Thus, ATB is fundamentally different from the scrambling cases we have observed. The ungrammaticality of (79) provides additional evidence indicating that representational syntax is needed in addition to derivational syntax to account for interpretation.

7. Conclusion In this chapter, a set of novel data was introduced that exhibits a certain type of island condition distinct from those observed with respect to typical operator movements. In order to explain this condition, two types of processing strategies were introduced, discussed, and rejected. A syntactic constraint was then proposed and defended based on empirical evidence. This constraint was then extended to the island condition that governs the operator movement, and it was suggested that the condition constraining LDS accounts for operator movement as well. The only distinction between the two types of movement is the domain of the application of this constraint: while CP is the island domain for operator movement, vP serves as the island domain for LDS. The discussion in this chapter, if on the right track, sheds some additional doubt on the merit of the typology of movement. The CLDM stated in (60) covers a range of A-movement (movement to an argument position) and A’-movement (movement to an adjunct position and Spec of CP). However, both movements are subject to the same condition. The bottom line seems to be the position where an element is interpreted, not the type of movement it undergoes. This issue is further discussed in section 3.2. and 3.3. in chapter 7.

224

Chapter 6: Long Distance Scrambling and Island Constraints

Finally, the left-dislocated constructions discussed in section 6, together with instances of ATB examples where the connection between the antecedent and one of the empty positions is disrupted, supports the idea that interpretation cannot be solely based on cyclic derivations within a phase, but need to be applied representationally in certain cases.

Notes 91. Müller and Sternefeld (1993) suggest that the Spec of TP is the landing site for topic in English and German. Baltin (1982), Johnson (1988), and Lasnik and Saito (1992) argue that topicalization in English is an IP adjunction. Chomsky (1977) suggests, however, that topicalization obeys the same island conditions as wh-movement. See also Koster (1975) on this issue. 92. A previous version of the data discussed in this section appeared in Karimi (1999c). 93. The following examples show that the same pattern holds of underlying objects of unaccusative verbs that serve as grammatical subjects of the embedded clause. (i) *chi un fekr mi-kon-e [vP t [VP t xorde shod]] what she thought dur-do-3sg eaten became Intended meaning: what does she think was eaten? (ii) *ki Parviz fekr mi-kon-e [vP t [VP t entekhâb shode ]] who P thought dur-do-3sg chosen become is ‘Intended meaning: who does Parviz think has been chosen? These examples show that structural subjects compete with each other for the same position. 94. See section 2.2. in chapter 2 for a discussion of recent views that consider scrambling as the result of the operation Merge. 95. The fact that scrambling is an instance of Move is supported by scope ambiguity that is created when two quantified elements interact with each other. See chapters 4 and 5 in this monograph for discussion. See also Miyagawa (1997, 2001) and Beck and Kim (1997), among others. 96. Pickering and Barry (1991) basically argue against the existence of empty categories and their role with respect to processing. 97. As we saw in chapter 4, adjuncts may scramble, creating ambiguity. The following examples are borrowed from that chapter. (i) cherâi fekr mi-kon-i [CP Kimea emruz ti bargasht]? why think pres-do-2sg K today returned ‘Why do you think Kimea returned today?’ (ii) keyi Kimea goft [CP ke Sepide ti xune xaride] when K said that S house bought has ‘When did Kimea say that Sepide has bought a home?’ Both sentences are ambiguous with respect to the scope interpretation of the adjuncts. 98. Thanks to David Pesetsky for making these data available to me. 99. The example in (66) is presented by Müller and Sternefeld (1993) as a grammatical sentence.

Chapter 7 Theoretical Consequences

1. Introduction The discussion in previous chapters suggests that scrambling is triggered by EPP (Chomsky 2000), represented as EPPs in this volume, and has a semantic and/or discourse functional effect on the output of the derivation. That is, it represents Topic or Focus, creates scope ambiguity, and feeds at least two Principles of the binding theory100. Furthermore, it was suggested that the optionality of scrambling boils down to the choice of selecting an EPPs feature from the lexicon. Thus, this operation is not optional syntactically. It has been suggested in the literature that LDS, although A’-movement, is not typical operator movement (Webelhuth 1992, Dayal 1994). This proposal suggests a-three way distinction in terms of the typology of Move: A-movement, A’-movement, and Scrambling. However, if movement is triggered by EPPs, and the output of the movement has an effect on the semantic output of the derivation, the typology of movement we have been accustomed with will become rather suspicious. In this final chapter, we discuss issues that enforce this suspicion. We will arrive at the conclusion that the A and A’ distinction is not a basic property of UG, and that the effects attributed to these two types of movements must be derived from other properties of grammar. Crucially, we suggest that this distinction is determined by the position where XP receives an interpretation. Finally, by reviewing some of the discussions in previous chapters, we will highlight the problems that phase theory faces as long as it purely relies on phase by phase derivation and interpretation. This chapter is structured as follows. The typology of movement is criticized in section 2, by discussions concentrating on the properties of reconstruction (2.1.), Anti-WCO effects (2.2.), and floating quantifiers (2.3.). Section 3 discusses the improper interaction of different types of movement, and suggests that this restriction is determined by the position where

226

Chapter 7: Theoretical Consequences

XP receives interpretation rather than by the typology of movement. Derivational versus representational syntax is briefly examined in section 4. A discussion of unresolved issues concludes this chapter in section 5.

2. Is typology of movement a myth? In this section, we examine different types of syntactic properties that have been employed as diagnostic tools to distinguish the A and A’ distinction. We will see that neither of these tools can be maintained as a clear cut device to differentiate the two types of movement. Reconstruction, for example, has been considered to be a property of A’-movement. This issue, including pieces of counter-evidence to this assumption, are discussed in section 2.1. Anti-WCO and Floating Quantifiers (FQ) are regarded as properties specific to A-movement. These two topics are examined in sections 2.2. and 2.3., respectively.

2.1. Reconstruction Reconstruction is considered to be a property of A’-movement. That is, the copy of an A’-moved element is assumed to be available for LF interpretation, whereas the copy of an A-moved element is supposed to lack this property. However, this distinction does not exist in certain types of raising and psyche verb construction, which are clear cases of A-movement. Consider the following example. (1)

A unicorni seems [TP ti to be in the garden]

The moved DP in (1) may have a narrow reading, a fact indicative of reconstruction. Additional examples are provided in (2) and (3). (2)

[A first year student]i seems to David ti to be at the party.

(3)

[Someone from New York]i is very likely ti to win the lottery. (Fox 1999: 178)

Both of these sentences are ambiguous with respect to the scope of the subject DPs. This fact indicates that reconstruction is possible in these cases, all of which are instances of A-movement.

Is typology of movement a myth?

227

As we saw in chapter 2, psyche-verbs seem to allow reconstruction at LF. The following configuration has been suggested by Belletti and Rizzi (1988) to represent the D-structure of a psyche-verb like ‘worry’. (4)

VP V' V

NPExperiencer NPTheme

worry According to Belletti and Rizzi, the theme in these constructions moves to the Spec of TP, and receives Nom Case in that position. This movement is, therefore, an A-movement. As we saw in chapter 2, however, Barss (1986) has shown that reconstruction seems to be possible from that position. The following examples are taken from that chapter. (5)

[Pictures of themselvesi ] annoy/?*hit the meni.

(6)

[Each otheri’s parents]k seem [IP tk to have annoyed tk the womeni ]

In (5), the copy of the phrase containing the reflexive is bound by the experiencer where the predicate is a psych verb like worry. In (6), we have a combination of psych verb construction and raising. The well-formedness of this sentence suggests that the lowest copy of the phrase containing the reciprocal enters the interpretation. Both sentences clearly show that reconstruction is possible from an A-position. In contrast, reconstruction does not seem to apply in all cases of A’movement. Consider the following data. (7)

a. *Anyone read none of the books b. [None of the books]i did anyone read ti. (Takano 2003: 522)

(8)

(Laka 1990) a. *[Buy any records]i she didn’t ti b. *[Whose theory about anything]i does no one like ti? (Philips 1996)

228

Chapter 7: Theoretical Consequences

c. *[Anything]i has nobody done ti

(Kayne 1998)

The A’-movement of the negative phrase rescues the sentence in (7). The grammatical sentence in (7b) shows that there is no reconstruction in this case. The ungrammatical sentences in (8) would have been well-formed if the A’-moved element had been reconstructed. The data in this section show that there is no clear cut distinction between A and A’ movement based on reconstruction.

2.2. Anti-WCO effects Weak Crossover (WCO) is also utilized as a diagnostic factor to distinguish the type of movement involved in a derivation. It is considered to be allowed when interacting with an A-movement, but not an A’-movement, as illustrated by the following contrast. (9)

a. *Whoi does it seem to hisi father [that John loves ti ] b. Whoi ti seems to hisi father [ti to be happy]

However, LDS, generally considered to be an instance of A’-movement, feeds binding in the case of clitic pronominals, as discussed in chapter 5. The following data are taken from that chapter, where LDS reveals an anti-WCO property. (10) a. [vP Kimeak be dust-eshk/*j goft [CP ke [vP pro Sepida-roj K to friend-her said that S-râ [PredP da’vat karde]]]] invitation did-3sg ‘Kimea told her friend that she has invited Sepide.’ b. [TP [Sepida-roj]i [vP Kimeak ti [PredP be dust-eshk/j goft [CP ke [vP pro ti [PredP da’vat karde]]]]]]

The sentence in (10b) is an instance of LDS, an A’-movement. Nevertheless, it clearly shows anti-WCO and binding effects. Similar to Persian, LDS of a wh-phrase may remedy WCO effects in Japanese:

Is typology of movement a myth?

229

(11) a. ?darei-o soitui-no hahaoya-ga ti aisiteiru no who-Acc guy-Gen mother-Nom love Q ‘Whoi does hisi mother love?’ b. darei-o soitui-no hahaoya-ga [ti Hanako-ga ti aisiteiru who guy-Gen mother-Nom H-Nom love omotteiru no think Q ‘Who does his mother think that Hnako loves?’ (Saito 1992: 108-109) (11b) is a clear anti-WCO case since the wh-phrase has crossed the pronominal it binds, and has undergone long distance scrambling101. The well-formedness of the grammatical data in this subsection indicates that drawing the line between A and A’ movement on the basis of Anti-WCO and binding does not prove to be reliable.

2.3. Floating quantifiers Similar to Anti-WCO, Floating Quantifier (FQ) has been considered to be possible only when A-movement is involved. This is as attested by the following contrasts. (12) a. [The drug dealers]i have all / each ti been arrested. b. *[These drug dealers]i, the major said that the police will all ti arrest.

(Déprez 1994: 104)

(13) a. [The boys]i appear [all ti to have left]. b. *[The children]i, whoi I will have [all ti ] met before the end of this week, …

(Dayal 1994: 257)

LDS, however, allows FQ in Persian, as we saw in chapter 4. Some data are repeated below. Details of the derivations are not included.

230

Chapter 7: Theoretical Consequences

(14) a. Kimea be bachche-hâ goft ke pro [hame gilâs-â-ro] Kimea to child-pl said that all cherry-pl-râ bo-xor-an subj-eat-3pl ‘Kimea told the children to eat all the cherries.’ b. [gilâs-â ro]i Kimea be bachche-hâ goft ke pro [hama ti-ro] bo-xor-an c. [gilâs-â-ro]i Kimea [hama ti-ro]k be bachche-hâ goft ke pro tk bo-xor-an The quantifier is left behind in its main clause in (14b), and in its intermediate position in (14c). In the case of (14b), we have a clear case of long distance scrambling which is suggested to be an instance of A’-movement throughout the literature (see chapters 2 and 4 for details). The example in (14c) cannot be considered a case of A-movement either, since the scrambled element has moved into a topic position, traditionally considered to be A’-movement. Finally, Holmberg and Nikanne (2002) provide data from Finnish representing cases where FQ is allowed to be controlled by an antecedent in a topic position. (15) Ilmeisesti kriitikot ovat (kaikki) ylistäneet tätä kirjaa Evidently the critics have all praised this book (16) Ilmeisesti tätä kirjaa on (kaikki) kirjoittanut Graham Greene Evidently these books has all written Graham Greene (Holmber & Nikanne 2002: 88) Holmberg and Nikanne consider the Spec of FP, a position roughly similar to the Spec of AGRsP, a Case position for the subject. It is also considered to be a topic position. In (15), the subject topic controls the preverbal FQ from the Spec of FP, a Nominative Case position. Thus, this sentence is predicted to be grammatical on the assumption that FQ is allowed to be licensed from an A-position. In (16), the object topic controls the preverbal FQ from the same position. This position, however, is only a topic position for the object in Finnish since Accusative Case is checked inside vP. Thus, the well-formedness of (16) is surprising on the assumption that FQ can only be licensed from an A-position.

Syntactic positions and interpretation

231

The data discussed in this section shed additional doubt on the distinction between A’-movement and A-movement. In the following section, we will see cases where one type of A’-movement cannot interact with another type. We will argue that all those cases may have a different solution which will also account for some of the ill-formed sentences discussed in this section.

3. Syntactic positions and interpretation In this section we examine the interaction of some types of movement which seem to be prohibited. We discuss previous claims and argue for a new solution. We start with the discussion of the interaction of wh-movement with scrambling and topicalization in 3.1. Two different proposals that have been offered to account for the impossibility of those interactions are presented in 3.1.1. and 3.1.2. An alternative proposal, comparable with the one discussed in 3.1.2., is offered in 3.2. Supporting evidence for this proposal follows in 3.3.

3.1. Improper interactions of Move As we saw in chapters 2 and 4, it has been suggested in the literature that LDS and some of the clause initial scrambling cases are instances of A’movement. The question we address in this section is whether two types of A’-movement may be combined in a given language. That is, can an XP first undergo scrambling or topicalization and then wh-movement and vice versa? Or is this type of combination barred as an instance of improper movement? Crucial to our analysis is the fact that wh-scrambling seems to be prohibited in a language like English that exhibits structural wh-movement, as evidenced by the following contrast. (17) a. He wants to know [CP whoi [TP ti [vP ate what]] b. *He wants to know [CP whoi [TP ti whatk [vP ate tk]]] The wh-phrase is scrambled out of vP in (17b), and the result is ill-formed. A similar situation holds in German, a language that exhibits structural wh-movement as well.

232

Chapter 7: Theoretical Consequences

(18) a. Wiei hat [IP der Fritz ti was repariert]? how has Art F what fixed b. *wiei hat [IP wask [IP der Fritz ti tk repariert] (Müller and Sternefeld 1996: 483) In (18b), the wh-phrase has scrambled into a higher position. Since German allows scrambling of DPs and PPs, lack of wh-scrambling seems to require some explanation. Consider now the following example, taken from Japanese. (19) John-ga

[[nani-oi [Mary-ga ti katta]] to]

itta no

John-Nom what-Acc M-Nom bought that said Q ‘John said that [what Mary bought] (Saito ad Fukui 1998: 445) The example in (19) shows that scrambling of wh-phrases is possible in Japanese. Similar to Japanese, Persian allows the wh-phrase to undergo scrambling. We saw in previous chapters that wh-phrases in this language undergo movement, and land in Spec of FocP. The following sentence illustrates this claim: (20) Kimea ki-roi

mi-xâd [CP pro fardâ ti

be-bin-e]

K who-râ hab-want tomorrow subj-see-3sg ‘Who is it that Kimea wants to see tomorrow?’. Furthermore, it is well-known that the combination of wh-movement and topicalization is not possible even in English and German, two languages that exhibit structural movements for both types of constructions (Lasnik and Saito 1992, Rizzi 1991). This is illustrated by the following contrasts in English and German. (21) a. Whoi ti said [CP that [IP John likes who]] b. *Whoi ti said [CP that whok [IP John likes tk]] (Müller and Sternefeld 1996: 482)

Syntactic positions and interpretation

233

(22) a. Weri sagte ti [CP dass [IP Fritz wenk gesehen habe]] who said that F whom seen hassubj b. *Weri sagte ti [CP wenk habe [IP Fritz tk gesehen]] who

said

whom hassubj F seen (Müller and Sternefeld 1996: 482)

The wh-phrase in the German example in (22b) has been topicalized, evidenced by the fact that the auxiliary habe ‘have’ has moved into a higher head position in the embedded clause, similar to the V2- phenomenon in matrix clauses in this language. Although German allows structural topicalization of DPs and PPs, the ill-formedness of (22b) seems surprising. In this section, we discuss the data in (17)–(22) by presenting two solutions previously offered in the literature. We proceed as follows. In 3.1.1., Epstein’s (1992) account based on the Economy of Derivation (Chomsky 1991, 1993) is examined. Müller’s and Sternefeld’s arguments based on their Principle of Unambiguous Binding (PUB) (1993, 1996) is discussed in 3.1.2. 3.1.1. Economy of Derivation Epstein rules out ungrammatical data such as those discussed in the previous section on the basis of Economy of Derivation (Chomsky 1991, 1993), stated below in (23). (23) Economy of Derivation If Derivations D1 and D2 are in the same reference set, the one that involves fewer operations is to be preferred. where the reference set is defined as in (24). (24) D1 and D2 are in the same reference set iff they start with the same numeration, and yield the same LF and PF output. According to Epstein, the ill-formedness of (21b) and (22b) follows from the Economy of Derivation. That is, the topicalized element must move one more time at LF. The LF representation of these two sentences, provided in (25) and (26), respectively, illustrates this point; two applications of Move are involved in these examples.

234

Chapter 7: Theoretical Consequences

(25) *[ CP [SPEC [whoi ] whok ] ti ] said [CP that [IP tk John likes tk ]]

(26) *[CP [SPEC [weri ] wenk ] sagte ti [CP tk habe [IP Fritz tk gesehen]] who

whom said

hassubj

F

seen

Now compare (25) and (26) with the LF representations of (21a) and (22a), provided in (27) and (28), respectively. (27) Whoi whok ti said [CP that [IP John likes tk ]]

(28) Weri wenk

sagte ti [CP dass [IP Fritz tk gesehen habe]]

who whom said

that

F

seen

hassubj

There is only one operation involved in (27) and (28). Thus, according to Epstein, the derivation of these sentences is more economical than that of (25) and (26). The same situation holds with respect to wh-scrambling. The LF representations of (17b) and (18b) are illustrated by (29) and (30), respectively. (29) *He wants to know [CP whoi whatk [TP ti [VP tk [VP ate tk ]]]

(30) *wiei wask hat [IP tk [IP der Fritz ti tk repariert]

Again, two instances of Move are involved in these cases. The ‘a’ sentences in (17) and (18), where wh-scrambling has not applied, involve only one operation, similar to (27) and (28). Therefore, they are more economical. As Müller and Sternefeld (1996) observed, there is a basic problem with this analysis. That is, the sentence in (31b) is ill-formed, even though the scrambled element is not a wh-phrase. (31) a. I wonder [CP whoi [IP ti saw John] b. *I wonder [CP whoi [IP ti [VP Johnk saw tk ]]

Syntactic positions and interpretation

235

The DP John does not move at LF. Therefore, there is only one movement involved in (31b). Nevertheless, the sentence is ill-formed. Thus given Epstein’s Economy of Derivation, the contrast between (31b), on the one hand, and scrambled cases in Persian and Japanese, on the other, remains a mystery. 3.1.2. The Principle of Unambiguous Binding The ill-formed sentences discussed in this section could be argued to illustrate instances of improper movement (Chomsky 1986a: 22). That is, a variable must be A-free in the domain of the head of its chain. In other words, an XP cannot undergo A’-movement followed by A-movement, since the copy left by the first movement will be bound by the head of its chain in an A-position. The following sentences clearly show this type of violation. (32) a. *which students are believed [CP t [TP it was notified t]] A-movement A’-movement b. *which child seems [CP t [TP there was killed t]] A-movement

A’-movement

The situation of the copies and their antecedents in (32) is simplified in (33) for clear illustration. (33) *XP t’ t A

A’

The problem with (33) is that the first movement creates a variable, whereas the second one creates an anaphor. Therefore, the intermediate trace is in a strange situation. That is, it is in an A’ position, and at the same time it receives an anaphoric interpretation, since it is bound by an element in an A position. Along the same lines, Müller and Sternefeld (1993, 1996) suggest that t’ (the intermediate trace) in (33) is in violation of a constraint they call Principle of Unambiguous Binding (PUB). This principle is stated in (34). (34) Principle of Unambiguous Binding (PUB) A variable that is α-bound must be β-free in the domain of the head

236

Chapter 7: Theoretical Consequences

of its chain (where α and β refer to different types of positions). (Müller and Sternefeld 1996: 496) Müller and Sternefeld suggest the following landing sites for different types of movements. (35) Landing Sites of Movement a. Wh-movement at S-structure is substitution in [Spec, C] b. Wh-movement at LF is either substitution in [Spec, C] or, if [Spec, C] is already filled, it is (right-) adjoined to [Spec, C]. c. Topicalization in Germanic languages is substitution in [Spec, T] (where TP is a maximal projection intervening between CP and IP, and T is the landing site of the verbsecond (V2) movement). d. Scrambling is left-adjunction to XP (VP or IP in German and Korean). (Müller and Sternefeld 1996: 496) Based on Müller and Sternefeld’s analysis, the sentences in (25) and (26), repeated below in (36) and (37), respectively, are ill-formed, since two types of movements associated with two types of landing sites are involved in the derivation of these sentences. (36) *[ CP [SPEC [whoi ] whok ] …… [CP [TP tk ] [IP tk ]] β

α (Müller and Sternefeld 1996: 498)

(37) *[CP [SPEC [weri ] wenk ] …. [CP [TP t’k ] ….. tk ]] β

α

The overt movement, represented by α, is into the Spec of TP, while the LF movement, represented by β, is into the Spec of CP, violating PUB. A similar situation holds with respect to scrambling. Consider the LF representation in (30), illustrated by (38). (38) b. *[CP [SPEC wiei ] wask ] …. [IP tk [IP …… tk ] β

α (Müller and Sternefeld 1996: 498)

Syntactic positions and interpretation

237

The analysis based on PUB faces the same problem we discussed with respect to the sentence in (31b), repeated below in (39). That is, there is no LF movement involved, and thus there is no PUB violation in this case102. (39) *I wonder [CP whoi [IP ti [VP Johnk saw tk ]] The explanations discussed in this and previous subsections face yet another problem. That is, LF movement of XP has been controversial in recent literature. In fact, in the last several years the tendency has been to eliminate LF movement of XPs (Chomsky 1995), an issue that is compatible with economy considerations. In that case, neither Economy of Derivation nor PUB can account for the data presented in this section. Thus, we still need to find an explanation for the ill-formedness of the ‘b’ sentences in (17), (18), (21) and (22). This issue is discussed in the next section.

3.2. An alternative account An alternative interpretation of Müller and Sternfeld’s PUB can in fact solve the problem. That is, if XP moves into the Spec of a functional head such as TopP or FocP, it receives its interpretation in that position, and thus cannot move into a new position to receive a different reading. In other words, in a language that exhibits structural wh-movement, whphrases receive their interpretation in the Spec of CP. In contrast, in a language that displays scrambling to represent discourse functions, the landing site of a wh-phrase is the Spec of FocP rather than CP, as we saw in chapter 4. In that position, the wh-phrase receives a focus interpretation103. Thus, wh-movement in Persian results in a different interpretation than in English. Similarly, an XP in a topic position receives a reading compatible with its position, and thus is barred from moving into a different position to receive a distinct reading. On the basis of the discussion thus far, we suggest the following constraint. (40) Constraint on Interpretation (CI) Within the functional domain, If XP receives interpretation in α, it cannot be interpreted in β. The ill-formed data in (17b) and (18b), repeated below in (41a-b), are explained by CI in (40).

238

Chapter 7: Theoretical Consequences

(41) a. *He wants to know [CP whoi ti [TP whatk [vP ate tk ]]] b. *wiei hat [IP wask [IP der Fritz ti tk repariert ] (40) suggests that the ill-formedness of the sentences in (41) is the result of improper movement into the wrong landing site, which prevents interpretation: wh-phrases in English and German are interpreted only in Spec of CP (or by the movement of the wh-feature to C in the case of whin-situ constructions in German). Thus, the problem with the data in (41) is not the economy of derivation, but rather the fact that a wh-phrase cannot be interpreted in a position other than the Spec of CP if it undergoes movement. Similarly, (40) explains the ill-formedness of (21b) repeated in (42a). (22b), repeated in (42b), receives a different explanation. (42) a. *Whoi ti said [CP that whok [IP John likes tk ]] b. *Weri sagte ti [CP wenk who

said

habe [IP Fritz tk gesehen]]

whom hassubj

F

seen

In (42a), the wh-phrase has moved into an improper position, and thus remains with no interpretation. In (42b), the wh-phrase is in Spec of CP, and thus this sentence should be grammatical. However, the matrix verb does not subcategorize for a CP with a wh-feature on its head. Consequently, even though the wh-phrase may receive interpretation in Spec of CP, the interrogative reading that is required when a wh-phrase is in Spec/CP at Spell-out cannot be obtained. The ill-formedness of (39), repeated in (43), is also accounted for by CI in (40). In (43), the DP John cannot receive an interpretation in its surface position due to the fact that the topic is interpreted in Spec of CP in English, and that position is occupied in this example. Thus, its movement into an improper position leaves it with no interpretation. (43) *I wonder [CP whoi [IP ti [VP Johnk saw tk ]] Given this line of reasoning, we do not need to stipulate an LF movement to account for the ungrammatical data discussed in this section.

Syntactic positions and interpretation

239

Let us now return to the cases that represent instances of classic improper movement. Examples were provided in (32), repeated below in (44). (44) a. *which students are believed [CP t [TP it was notified t]] A-movement A’-movement b. *which child seems [CP t [TP there was killed t]] A-movement

A’-movement

If there is no A/A’ distinction, and improper movement in the sense of Chomsky (1986a) is thus not responsible for the ill-formedness of these sentence, how can we rule them out? The response is simple: both sentences are subject to Superraising, subsumed under MLC. The extracted wh-phrase must move to the closes EPPg position. That position is not available, and thus the derivation crashes.

3.3. Supporting evidence Further evidence that is based on interpretation sites, rather than an A/A’ distinction, is provided by a comparative analysis of binding relations in Persian and English. We saw in chapter 5 that one of the well-known properties of operator movement, such as wh-movement in English, is that the surface position and the intermediate copies may enter into anaphoric binding relations. We saw that this is not true in the case of Persian. The examples provided in that chapter are repeated below. (45) a. John asked which pictures of himself Bill bought. b. John asked which pictures of himself Bill said Dan bought. The anaphor himself can have John and Bill as its antecedent in (45a) and John, Bill, and Dan in (45b). This situation does not hold in Persian, as evidenced by (46). (46) mo’allem-âj [TP [aks-â-ye hamdigai/*j-ro]k fekr mi-kon-an [CP ke teacher-pl picture-pl each other-râ thought-dur-do-3sg that [TP’ [vP [bachche-hâi tk ] [PredP be modir neshun dâd-an]]]]] child-pl to principal show did-3sg

240

Chapter 7: Theoretical Consequences

‘The teachers think that it was the pictures of each other that the children showed the principal.’ In this example, the reciprocal can only be bound by the subject of the embedded clause. As discussed in chapter 5, the distinction between (45) and (46) has to do with the interpretation sites in the two languages: wh-phrases take their scope position in the Spec of CP. In other words, Spec of CP is the position where the wh-phrase receives its interpretation. Thus the intermediate Spec of CP serves as a potential interpretation site for wh-phrases. The Spec of vP is a position where the local subject and object are interpreted. All other elements move into that position purely for the sake of being at the edge of vP for further movement. Thus, they cannot enter an interpretive relation from that position. This suggestion explains why the copy in the Spec of CP enters into anaphoric binding relations, but the copy left in the Spec of vP by LDS does not, an issue that is predicted by (40). Further support for (40) is provided by the constraint we observed in the previous chapter in a scrambling language such as Persian: the operation crashes if the scrambled element has a grammatical counterpart in the target clause. This was shown to be true of the subject, object, indirect object, and even adverbials. The following contrast, taken from that chapter, clearly shows that the movement of the subject into the matrix clause, where an overt subject is present, renders the sentence ungrammatical. (47) a. Parviz fekr mi-kon-e [CP ke ki in kâr-ro karde] P thought dur-do-3sg that who this work-râ done is ‘Who does Papar think has done this job?’ b. *kii Parviz fekr mi-kon-e [CP ke ti in kâr-ro karde]

We also saw that this restriction does not hold in a language like English that exhibits structural wh-movement. The following data represent this fact. (48) a. Whoi do you think [ti will fix my car next week?]] b. Whoi did you say [Mary thinks [ti will fix my car next week?]] Based on the distinctions between island conditions in English and Persian, the following constraint was proposed.

Derivational versus representational syntax

241

(49) Constraint on Long Distance Movement (CLDM) LDM is blocked in the following configuration *[Phase YPiα XPα ……[ ti ] Where α represents relevant features, and XP is lexical. As discussed in chapter 6, CP is the island phase for languages with structural wh/topic movement, while vP is the phase for scrambling languages. This claim is crucially based on interpretation sites, which is compatible with CI in (40). Some of the FQ data discussed in section 2.3. pose counter- evidence to an analysis that rejects a structural distinction between A/A’. We will return to those data in the final section of this chapter.

4. Derivational versus representational syntax Another issue that was brought up in this work was the interaction of phase theory with interpretation, particularly of pronominal binding relations. As we saw in chapter 5, this interaction seems to provide some problems for a derivational theory based on phase. The following example, taken from chapter 5, illustrates this problem. (50) Rahjuei goft [CP ke [TP [vP Kimeak [DP pedar-e un-roi/*k] R said that K father-Ez his/her-râ dust dâr-e]]]] friend have-3sg ‘Rahjue said that Kimea loves his/*her father.’ We noticed that if interpretation applies cyclically within a phase, by the time the derivation reaches the matrix clause, access to the lower phases is impossible. It was suggested that co-reference between the subject in the higher clause and the embedded object be established representationally, rather than derivationally. This idea is supported independently by resumptive pronouns in Lebanese Arabic, as discussed by Aoun and Li (2003). Aoun and Li argue that chains can be generated derivationally by Move or Agree, or representationally. The former is constrained by MLC, while the latter is controlled by a constraint they call the Minimal Match Condition (MMC). This condition is restated in (51).

242

Chapter 7: Theoretical Consequences

(51) Minimal Match Condition (MMC) An operator must form a chain with the closest XP it c-commands that contains the same relevant features. (Aoun & Li 2003: 29) Aoun and Li’s argument is based on true resumptive pronouns in island constructions, where movement is not an option. Consider the following contrast taken from their work (Aoun & Li 2003: 36). (52) a. [CP wh1 [IP [Island RP1 wh1 ] …]] b. *[CP wh2 [IP [Island wh1 RP2 ] …]] The wh-phrase in Spec of CP is an operator and needs to bind a variable in order to receive proper interpretation. In (52a), the resumptive pronoun (RP1) is bound by the closest XP (wh1) that c-commands it and shares the same relevant feature with it, namely the wh-feature. Thus, semantic matching successfully applies, and the operator wh1 receives an interpretation via the resumptive pronoun RP1. In (52b), by contrast, RP2 is c-commanded by wh1 which shares the wh-feature with it. Therefore, it cannot be bound by wh2. Thus, the operator wh2 remains uninterpreted, and the sentence becomes ungrammatical. The fact that the operator wh1 in (52a) receives its interpretation by MMC across the island indicates that binding in this case is not possible by movement, and must be achieved representationally. The discussion of left-dislocation cases discussed in chapter 6 suggests the same difficulty with respect to a derivational grammar purely based on phase theory. Consider first the following example taken from that chapter. (53) Rahjue-roi, Kimea bâvar mi-kon-e [CP ke proi/uni ketâb-â-ro R-râ K belief dur-do-3sg that he book-pl-râ az Arezu xaride] from A bought is ‘As for Rahjue, Kimea believes that (he) has bought the books from Arezu.’ The pronoun un ‘he’ or its alternative, pro, receives its interpretation from the left-dislocated DP Rahjue-ro. This co-reference is not possible within a purely derivational model based on phase, since the left-dislocat-

Further research

243

ed element is generated by Merge in its surface position, outside of the phase containing the pronoun/pro. An additional problem is provided by Across The Board cases, especially those where interpretation is blocked by an intervening element. The following examples, taken from chapter 6, represent this fact. (54) Who did [Mary kiss t], [Stephanie date t], and [Anne marry t]? (55) *Who did [Mary kiss t], [Stephanie date Bill] and [Anne marry t]? Since there is no movement involved in (54), the relation between the whphrase and the copies in the embedded clauses has to be established representationally. The same logic holds for (55), where the link is blocked by the intervening object. In this section, we reviewed cases discussed in previous chapters, and arrived at the conclusion that UG must allow interpretation based on representation, in addition to derivation, if phase theory is to be maintained.

5. Further research Abandoning the A/A’ distinction provides some problems, and several issues remain unresolved. We will discuss some of them in this section, and will leave them for future research. Let us first return to FQ, discussed in this chapter. It was shown that LDS in Persian licenses FQ, contrary to the assumption that FQ can only be licensed from an A-position. The well-formed sentences in (14b) and (14c) are repeated in (56a) and (56b), respectively. (56) a. [gilâs-â ro]i Kimea be bachche-hâ goft ke pro [hama ti-ro] cherry-pl-râ Kimea to child-pl said that all -râ bo-xoran subj-eat-3pl b. [gilâs-â-ro]i Kimea [hama ti-ro]k be bachche-hâ goft ke pro tk bo-xor-an Given CI in (40), (56a) receives a straight forward analysis. In this example, the moved element receives its topic interpretation within the dis-

244

Chapter 7: Theoretical Consequences

course domain in the matrix clause, separated from the quantifier in the lower phase (cf. vP). In (56b), in contrast, the quantifier is floated in the same discourse phase. Thus, this sentence seems to violate IC in (40), since the antecedent and the quantifier are in two different interpretation sites within the same phase. We suggest, however, that the lower position is what we called the background topic position in chapter 4. The DP gilâs-â-ro ‘cherries-Acc’ is in a position we called shifted topic. Since both elements are in topic positions, there is no clash. Thus, the FQ in the intermediate position is licensed by the DP in the higher position. Now consider the English examples in (12) and (13), repeated in (57) and (58). (57) a. [The drug dealers]i have all/each ti been arrested. b. *[These drug dealers]i, the major said that the police will all ti arrest.

(Déprez 1994: 104)

(58) a. [The boys]i appear [all ti to have left]. b. *[The children]i, whoi I will have [all ti ] met before the end of this week,…

(Dayal 1994: 257)

The Persian examples in (56) clearly show that locality is not a requirement for licensing a FQ. The data in (57b) and (58b), however, pose a problem for such generalization. In these two sentences, the A/A’ distinction seems to hold unless the ungrammaticality of these sentences is explained in a different way. We have no explanation for their ill-formedness at this point, and leave this issue for further research. Some general problems remain to be explained. For example, how can we explain the locality condition that governs anaphoric binding relations (anaphors and NP-traces), instances of A-movement, but is absent in the case of different instances of A’ movements (including LDS)? A second type of general question is this: why do languages differ with respect to the type of interpretation sites they allow? That is, why is the wh-phrase interpreted in Spec/CP in English and German, but in Spec/ FocP in Persian and Hungarian? Furthermore, why do languages have different types of movements? That is, why do Persian, Hindi, Japanese,

Notes

245

and Korean allow long distance movement of their phrasal categories, but not German (except for wh-phrases)? This goes back to some basic questions we have faced all along: why do some languages have obligatory whmovement and some do not? Although different types of explanations have been offered in the literature (see section 4 in chapter 2), an indepth analysis of these parametric distinctions requires further research.

Notes 100. We saw in chapter 5 that scrambling feeds Principle A in German, although it does not feed it in Persian. 101. See Richards (1997) for an alternative analysis. 102. Müller and Sternefeld rule (39) out by suggesting that English does not allow scrambling since it does not have VP, IP, or CP adjunction. See section 4.2. in chapter 2, where the issue of adjunction sites, and the problems related to it, were critically discussed. 103. The interrogative interpretation of the wh-phrase in Persian is accomplished by the movement of the wh-feature to C, as we saw in chapter 4.

References

Abney, Steven 1987 The English noun phrase in its sentential aspect. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Aissen, Judith 1992 Topic and focus in Mayan. Language 68 (1): 43-80. Alexiadou, Artemis, and Elena Anagnostopoulou 1998 Parametrizing AGR: word order, V-movement and EPP checking. Natural Language and Linguistic Theory 16 (3): 491- 539. 2001 The subject-in-situ generalization and the role of Case in deriving computations. Linguistic Inquiry 32 (2): 193-232 Aoun, Joseph, and Yen-hui Audrey Li 1993 wh-elements in-situ: syntax of LF? Linguistic Inquiry 24 (2): 199-238. 2003 Essays on the Representational and Derivational Nature of Grammar. Cambridge/London: The MIT Press. Bailyn, John 1995 A configurational approach to Russian “free” word order. Ph.D. diss., Department of Linguistics, Cornell University. 2001 On scrambling: A reply to Bošković and Takahashi. Linguistic Inquiry 32 (4): 635-657. 1999 Eliminating optional movement in Russian. Ms., Department of Linguistics, SUNY at Stony Brook. 2003 Does Russian scrambling exist? In Word Order and Scrambling, S. Karimi (ed.), 156-176. Oxford/Berlin: Blackwell Publishers. Baltin, Mark 1982 A landing site theory of movement rules. Linguistic Inquiry 13 (1): 1-38. Barss, Andrew 1986 Chains and anaphoric dependence. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. 1996 Derivations and Reconstruction. Studies in the Linguistic Sciences 24 (2): 19-38. Barss, Andrew, Kenneth Hale, Margaret Speas, and E. Perkins 1991 Movement at logical form and Navajo syntax,. In Logical Structure and Linguistic Structure: Cross-Linguistic Perspectives, J. Huang, and R. May (eds) 2547. Dordrecht/Boston: Kluwer. Bateni, Mohammad Reza 1969 Towsif-e sâxtemân-e dasturi-ye zabân-e farsi. [Description of the Grammatical Structure of Farsi Language]. Tehran, Iran: Amir Kabir Publishing. 1977 Negâhi Tâze be Dastur-e Zabân [New Approach to the Grammar of Language]. Tehran, Iran: Agah Publishing. Bayer, Josef, and Jaklin Kornfilt 1994 Against scrambling as an instance of move-alpha. In Studies on Scrambling, N. Cover and H. van Riemsdijk (eds.), 17-60. Berlin/New York: Mouton de Gruyter.

References

247

Beck, Sigrid 1996 Quantified structures as barriers for LF movement. Natural Language Semantics 4: 1-56. Beck, Sigrid, and Shin-Sook Kim 1997 On wh-and operator scope in Korean. Journal of East Asian Linguistics 6: 339-384. Belletti, Adriana, and Luigi Rizzi 1988 Psyche verbs and theta theory. Natural Language and Linguistic Theory 6: 291-352. Bierwisch, Manfred 1963 Grammatik des Deutschen Verbs.. Berlin: Akademic Verlag. Bobaljik, Jonathan, and Diane Jona 1996 Subject positions and the role of TP. Linguistic Inquiry 27 (2): 195-236. Bošković, Zeljko, and Daiko Takahashi Scrambling and last resort. Linguistic Inquiry 29 (2): 347-366. Boyle, J. A. 1966 Grammar of Modern Persian, Otto Harrassowitz. Brame, Michael 1981 The head-selector theory of lexical specifications and the nonexistence of coarse categories. Linguistic Analysis 10 (4): 321-325 Borer, Hagit 2002 The syntax and semantics of quanitity. Ms. University of Southern California. Bresnan, Joan, and S. Mchombo 1987 Topic, pronoun, and agreement in Chichewa. Language 63 (4): 741-782. Browning, Marguerite, and Ezzat Karimi 1994 Scrambling to object position in Persian. In Studies on Scrambling, N. Cover and H. van Riemsdijk (eds), 61-100, Berlin/New York: Mouton de Gruyter. Burzio, Luigi 1986 Italian Syntax. Dordrecht: Reidel. Cheng, Lisa 1991 On the typology of wh-questions. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Cho, Jai-Hyoung 1994 Scrambling in Korean: crossover, reconstruction, and binding theory. Ph. D. diss., Department of Linguistics, University of Connecticut. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge: The MIT Press. 1970 Remarks on nominalization. In English Transformational Grammar, R. Jacobs and P.S. Rosenbaum (eds.), 184-221. Walthm, MA: Ginn. 1971 Problems of Knowledge and Language. New York: Pantheon Books. 1973 Conditions on transformations. In A Festschrift for Morris Halle, S. Anderson and P. Kiparsky (eds.), 232-286. New York: Holt, Reinhart, and Winston. 1976 Conditions on rules of grammar. Linguistic Analysis 2 (3): 303-351. Reprinted in (1977) Essays on Form and Interpretation, New York: North Holland. 1977 On wh-movement. In Formal Syntax, P. Culicover, T. Wasow, and A. Akmajian (eds.), 71-132. New York: Academic Press. 1980 On binding. Linguistic Inquiry 11 (1): 1-46. 1981 Lectures on Government and Binding. Dordrecht: Foris. 1982 Some Concepts and Consequences of the Theory of Government and Binding. Cambridge/London: The MIT Press.

248

References

1986a 1986b 1991

Barriers. Cambridge/London: The MIT Press. Knowledge of Language. New York: Praeger. Some notes on economy of derivation and representation. In Principles and Parameters in Comparative Grammar, R.Freiden (ed.), 417-454. Cambridge/ London: The MIT Press. 1993 A minimalist program for linguistic theory. In The View from Building 20, K. Hale and S. J. Keyser (eds.), 1-52. Cambridge/London: The MIT Press. 1994 Bare phrase structure. MIT Occasional Papers in Linguistics 5: 1-51. 1995 The Minimalist Program. Cambridge/London: The MIT Press. 1998 Some observations on economy in generative grammar. In Is the Best Good Enough?, P. Barbosa, D. Fox, P. Hagstrom, M. McGinnis, and D. Pesetsky (eds.), 115- 127. Cambridge/London: The MIT Press. 2000 Minimalist inquiries: the framework. In Step by Step: Essays on minimalist syntax in honor of Howard Lasnik, R. Martin, D. Michaels, and J. Uriagereka (eds.), 89-155, Cambridge/London: The MIT Press. 2001a Beyond explanatory adequacy. MIT Occasional Papers in Linguistics 20. 2001b Derivation by phase. In Ken Hale: A life in language, M. Kenstowicz (ed.), 152. Cambridge/London: The MIT Press. Chomsky, Noam, and Howard Lasnik 1977 Filters and control. Linguistic Inquiry 8 (3): 425-504. Chomsky, Noam, and Howard Lasnik 1993 Principles and parameters theory. In Syntax: An International Handbook of Contemporary Research, J. Jacobs, A. von Stechow, W. Stemefeld, and T. Vennemann (eds.), 506-569. Berlin/New York: Mouton de Gruyter. Cinque, Guglielmo 1999 Adverbs and Functional Heads: A Cross-Linguistic Perspective. New York/ Oxford: Oxford University Press. Corver, Norbert, and Hank van Riemsdijk (eds.) 1994 Studies on Scrambling. Berlin/New York: Mouton de Gruyter. Culicover, Peter, and Michael Rochemont 1983 Stress and focus in English. Language 59 (1): 123-165. Dabir-Moghaddam, Mohammad 1982 Syntax and semantics of causative constructions in Persian. Ph.D. diss., Department of Linguistics, University of Illinois at Urbana. 1985 Majhul dar zabân-e farsi [Passive in Farsi language]. Iranian Journal of Linguistics 2 (1): 31-46. 1995 Compound Verbs in Persian. Journal of Iranian Linguistics 12 (1): 2-46. Darzi, Ali 1996 Word order, NP movement, and opacity conditions in Persian. Ph.D. diss., Department of Linguistics, University of Illinois at Urbana-Champaign. Dayal, Veneeta Srivastav 1994 Binding facts in Hindi and the scrambling phenomenon. In Theoretical Perspectives on Word Order in South Asian Languages, M. Butt, T. Holloway King, and G. Ramchand (eds.), 237-261. Stanford: CSLI Publications. Déprez, Viviane 1989 On the typology of positions and chains. Ph.D diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. 1994 Parameters of object movement. In Studies on Scrambling, N. Cover and H. van Riemsdijk (eds.), 101-152. Berlin/New York: Mouton de Gruyter.

References

249

Diesing, Molly 1992 Indefinites. Cambridge/London: The MIT Press. Diesing, Molly, and Eloise Jelinek 1995 Distributing arguments. Natural Language Semantics 3 (1): 123-176. É. Kiss, Katalin 1994 Scrambling as the base-generation of random complement order. In Studies on Scrambling, N.Corver, and H. van Riemsdijk (eds.), 221-300. Berlin/New York: Mouton de Gruyter. 1995 Introduction. In Discourse Configurational Languages, K. É. Kiss (ed.), 3-27. New York/Oxford: Oxford University Press. 1997 Discourse configurationality in the languages of Europe. In Constituent Order in the Languages of Europe, A. Siewierska (ed.). Berlin/New York: Mouton de Gruyter. 1998 Identificational focus versus information focus. Language 74 (2): 245-273. 2002 The EPP in a topic-prominent language. In Subjects, Topic, and EPP, P. Svenonius (ed.), 107-124. New York/Oxford: Oxford University Press. 2003 Argument scrambling, operator movement, topic movement in Hungarian. In Word Order and Scrambling, S. Karimi (ed.), 22-43. Oxford/Berlin: Blackwell Publishers. Enç, Murvet 1991 The semantics of specificity. Linguistic Inquiry 22 (1): 1-25. Epstein, David 1992 Derivational constraints on A’-chain formation. Linguistic Inquiry 23 (2): 235-259. Ernst, Thomas 1984 Towards an integrated theory of adverb position in English. Ph. D. diss., Department of Linguistics, Indiana University. 2002 The Syntax of Adjuncts. Cambridge Studies in Linguistics. Cambridge: Cambridge University Press. Fanselow, Gilbert 1990) Scrambling as NP movement. In Scrambling and Barriers, G. Grewendorf and W. Sternefeld (eds.), 113-140. Amsterdam/Philadelphia: John Benjamins. Farmer, Anne 1980 On the interaction of morphology and syntax. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Farrokhpey, Mahmoud 1979 A syntactic and semantic study of auxiliaries and modals in Modern Persian. Ph.D. diss., University of Colorado at Boulder. Folli, Raffaella, Heidi Harley, Simin Karimi (in Press) Determinants of event type in Persian complex predicates. To appear in Lingua. Foroughi, A. 1944 A Concise Persian Grammar. New York: Orientalia, Inc. Fox, Danny 1999 Reconstruction, binding theory, and the interpretation of chains. Linguistic Inquiry 30 (1): 157-196. Fukui, Naoki 1993 Parameters and optionality. Linguistic Inquiry 24 (3): 399-420. 1988 Deriving the differences between English and Japanese: a case study in parametric syntax. English Linguistics 5: 249- 270.

250

References

Fukui, Naoki, and Mamuro Saito 1992 Spec-head agreement, X’-compatibility, and optional movement. Paper presented at MIT, Cambridge. Ghomeshi, Jila 1996 Projection and inflection: a study of Persian phrase structure. Ph.D diss., Department of Linguistics, University of Toronto. 1997a Non-projecting nouns and the ezafe construction in Persian, Natural Language and Linguistic Theory 15 (4): 729-788. 1997b Topics in Persian VPs, Lingua 102: 133-167. 2001 Control and thematic agreement. Canadian Journal of Linguistics 46 (1/2): 9-40. Gibson, Edward 1991 A computational theory of linguistic processing: memory limitations and processing breakdown. Ph.D. diss.,, Carnegie Mellon University. Gibson, Edward and Gregory Hickok 1993 Sentence processing with empty categories. Language and Cognitive Processes 8: 147-161. Greenberg, Joseph H. 1963 Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Language, J. H. Greenberg, (ed.), 73-112. Cambridge/London: The MIT Press. Grimshaw, Jane 1990 Argument Structure. Cambridge/London, The MIT Press. Grewendorf, Günter 2001 Multiple wh-fronting,. Linguistic Inquiry 32 (1): 87-122. Grewendorf, Günter, and Wolfgang Sternefeld 1990 Scrambling and Barriers. Amsterdam/Philadelphia, John Benjamins. Grewendorf, Günter, and Joachim Sabel 1999 Scrambling in German and Japanese: adjunction versus multiple specifiers. Natural Language and Linguistic Theory 17 (1): 1-65. Guilfoyle, Eithne, Henrietta Hung, and Lisa Travis 1992 Spec of IP and spec of VP: two subjects in Austronesian languages. Natural Language and Linguistic Theory 10 (3): 375-411 Haan, G. de 1979 Conditions on Rules. Dordrecht: Foris. Haegeman, Liliane 1995 The Syntax of Negation. Cambridge: Cambridge University Press. Haider, Hubert 1988 θ-tracking systems; evidence from German. In Configurationality, L. Maracz, and P. Muysken (eds.), 85-206. Dordrecht: Foris. 2000 OV is more basic than VO. In The derivation of VO and OV, P. Svenonius (ed.), 45-67. Amsterdam/Philadelphia: John Benjamin. Hajatti, Abdolkhalil 1977 Ke constructions in Persian: descriptive and theoretical aspects. Ph.D. diss., Department of Linguistics, University of Illinois at Urbana Hale, Kenneth 1980 Remarks on Japane se phrase structure; comments on the papers on Japanese syntax. In Theoretical Issues in Japanese Linguistics, Y. Otsu and A. Farmer (eds.), 85-203. MIT Working Papers 2, MIT, Cambridge.

References

251

Hale, Kenneth, and Samuel Jay Keyser 1993 On argument structure and the lexical expression of syntactic relation. In The View From Building 20, K. Hale and S. J. Keyser (eds.), 53-109. Cambridge/ London: The MIT Press. Hale, Kenneth, and Samuel Jay Keyser 2002 Prolegomenon to a Theory of Argument Structure. Cambridge/London: The MIT Press. Hale, Kenneth, Eloise Jelinek, MaryAnne Willie 2003 Topic and focus positions in Navajo. In Word Order and Scrambling, S. Karimi (ed.), 1-21. Oxford/Berlin: Blackwell Publishers. Halle, Morris, and Alec Marantz 1993 Distributed morphology and the pieces of inflection. In The View from Building 20, K. Hale and S. Keyser (eds.), 111- 176. Cambridge/London: The MIT Press. Harley, Heidi 1995 Subjects, events, and licensing. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. 2000 You’re having me on: Aspects of have. In La Grammaire de la Possession, J. Guéron and A. Zribi- Hertz (eds), 195- 226. Paris: Université Paris X – Nanterre. 2001 Irish, the EPP and PRO. Ms. Department of Linguistics, University of Arizona. Hasegawa, Nobuko 1994 Economy of derivation and A’-movement in Japanese. In Current Topics in English and Japanese, M. Nakamura (ed.), 1-25. Tokyo: Hituzi Syobo. Hashemipour, Margaret 1989 Pronominalization and control in Modern Persian. Ph. D. diss. Department of Linguistics, University of California at San Diego. Hoekstra, Teun 1984 Transitivity: Grammatical Relations in Government Binding Theory. Dordrecht: Foris. Hoji, Hajime 1985 Logical form constraints and configurational structures in Japanese. Ph.D diss., Department of Linguistics, University of Washington. Holmberg, Anders 1986 Word order and syntactic features in the Scandinavian languages. Ph.D. diss., University of Stockholm. 1993 Two subject positions in IP in Mainland Scandinavian. Working Papers in Scandinavian Syntax 52: 29-41. 2000 Deriving the OV order in Finnish. In The Derivation of VO and OV Orders, P. Svenonius (ed.), 123-152. Amsterdam/Philadelphia: John Benjamins. Holmberg, Anders and Urpo Nikanne 2002 Expletives, subjects, and topics in Finnish. In Subjects, Expletives, and the EPP, P. Svenonius (ed.), 71-105. New York/Oxford: Oxford University Press. Horvath, Julia 1986 Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Huang, James 1982 Logical relations in Chinese and the theory of grammar. Ph.D. Diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology.

252

References

Jackendoff, Ray 1972 Semantic Interpretation in Generative Grammar. Cambridge/London: The MIT Press. 1973 The base rules for prepositional phrases. In A Festschrift for Morris Halle, S. Anderson and P. Kiparsky (eds.), 345- 356. New York: Holt, Rinehart and Winston. 1977 X’-Syntax: A Study of Phrase Structure. Cambridge/London: The MIT Press. Jazayery, Mohammad Ali, and H. H. Paper 1961 A Reference Grammar of Modern Persian. Anne Arbor: The University of Michigan Press. Jelinek, Eloise 1994 Quantification in Straits Salish. In Quantification in Natural Language, E. Bach, A. Kratzer, E. Jelinek, and B. Partee (eds.), 487-540. Dordrecht/Boston: Kluwer. Johnson, Kyle 1988 Clausal gerunds, ECP and government. Linguistic Inquiry 19 (3): 583-609. Jonas, Diane, and Jonathan Bobaljik 1993 Specs for subjects: the role of TP in Icelandic. MIT Working Papers in Linguistics 16: 99-148. Kahnemuyipour, Arsalan 2001 On wh-questions in Persian. Canadian Journal of Linguistics 46 (1-2): 41-61. Karimi, Simin 1990 Obliqueness, specificity, and discourse functions. Linguistic Analysis 20 (3/4): 139-191. 1996 Case and specificity: Persian râ revisited. Linguistic Analysis 26 (3/4): 173-194. 1997a Persian complex predicates and LF incorporation. Proceedings of Chicago Linguistic Society (CLS) 33: 215-229. 1997b Persian complex verbs: idiomatic or compositional. Lexicology 3 (2): 273-318. 1999a Specificity effect: evidence from Persian. Linguistic Review 16: 155-141. 1999b A note on parasitic gaps and specificity. Linguistic Inquiry 30 (4): 704-713. 1999c Is scrambling as strange as we think it is?’ MIT Working Papers in Linguistics 33: 159-190. 2001 Persian complex DPs: how mystrious are they? Canadian Journal of Linguistics 46 (1/2): 63-96. 2003a (ed.), Word Order and Scrambling. Oxford/Berlin: Blackwell Publishers. 2003b Introduction. In Word Order and Scrambling, S. Karimi (ed.), viii-xx. Oxford/ Berlin: Blackwell Publishers. 2003c On object positions, specificity, and scrambling in Persian. In Word Order and Scrambling, S. Karimi (ed.), 91-124. Oxford/Berlin: Blackwell Publishers. 2003d Focus movement and the nature of uninterpretable features. In Formal Approaches to Function in Grammar, A. Carnie, H. Harley and M. Willie (eds.), 297-306. Amsterdam/Philadelphia: John Benjamins. Karimi, Simin & Michael Brame 1986 A generalization concerning the ezafe constructions in Persian. Paper presented at the annual meeting of Western Conference on Linguistics (WECOL), University of British Columbia, Canada. Karimi, Simin and Anne Lobeck 1997 Specificity effects in English and Persian. In the Proceedings of North East Linguistics Society (NELS) 28: 175-186.

References

253

Karimi, Simin, and Azita Taleghani 2003 Looking alike is not being alike a story of argument and adjunct wh-constructions in Persian. Paper presented at the Western Conference on Linguistics (WECOL), Tucson, Arizona. Kayne, Richard 1980 Extensions of binding and Case-marking. Linguistic Inquiry 11 (1): 75-96. 1993 Toward a modular theory of auxiliary selection. Studia Linguistica 47: 3-31. 1994 The Antisymmetry of Syntax. Cambridge/London: The MIT Press. 1998 Overt versus covert movement. Syntax 1 (1): 128-191. Kerstens, J. 1975 Over afgeleide structuur en de interpretatie van Zinnen,’ ms., University of Amsterdam, Amsterdam. Khanlari, Parviz 1976 Dastur-e Zabân-e Farsi [The Grammar of Farsi Language]. Tehran, Iran: Bonyad-e Farhang. Kim, Dae-Bin 1992 The specificity/non-specificity distinction and scrambling theory. Ph.D. diss., Department of Linguistics, University of Wisconsin- Madison. King, Tracy 1995 Configuring Topic and Focus in Russian. Stanford: CSLI. Koster, Jan 1975 Dutch as an SOV Language. Linguistic Analysis 1 (2): 111-136. Kratzer, Angelika 1993 On external arguments. In University of Massachusetts Occasional Papers 17. Amherst: University of Massachusetts. 1995 Stage-level and individual-level predicates. In The Generic Book, G. Carlson and F. J. Pelletier (eds.), 75-125. Chicago: The University of Chicago Press. Kuroda, S-Y 1988) Whether we agree or not: a comparative syntax of English and Japanese. In Papers from the Second International Workshop on Japanese Syntax, W. Poser, (ed.), 103-143. Center for the Study of Language and Information, Stanford University, Stanford. Laka, Itziar 1990 Negation in syntax: on the nature of functional categories and projections. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Larson, Richard 1988 On the double object construction. Linguistic Inquiry 19 (2): 335-391. Lasnik, Howard 1993 The minimalist theory of syntax: motivations and prospects. Paper presented at the 2nd Seoul International Conference on Generative Grammar. 1999a Minimalist Analysis. Oxford/Berlin: Blackwell Publishers. 1999b On feature strength: three minimalist approaches to overt movement. Linguistic Inquiry 30 (1): 197-218. 2001 A note on the EPP. Linguistic Inquiry 32 (2): 356-362. Lasnik, Howard, and Mamuro Saito 1992 Move α: Conditions on Its Application and Output. Cambridge/London: The MIT Press.

254

References

Lavine, James 1998 The syntax of feature-checking, focus structure, and sentential stress. Paper presented at the annual conference of AATSEEL, San Francisco. Lazard, Gilbert 1992 A Grammar of Contemporary Persian, Los Angeles, California: Mazda Publishers. Lebeaux, David 1998 Where does the binding theory apply. Ms. NEC Research Institute. Li, Charles, and Sandra Thompson 1976 Subject and topic: A new typology of language. In Subject and Topic, C. Li (ed.), 457-490. New York: Academic Press. Lotfi, Ahmad 2003 Persian wh-riddle. In Multiple wh-fronting, C. Boecks and K. Grohmann (eds). Amsterdam/Philadelphis: John Benjamins. Mahajan, Anoop 1990 The A/A’ distinction and movement theory. Ph.D. diss., Department of Linguistics and philosophy, Massachusetts Institute of Technology. 1994 Toward a unified theory of scrambling. In Studies on Scrambling, N. Corver, and H. van Riemsdijk (eds.), 301-330. Berlin/New York: Mouton de Gruyter. Manzini, Rita, and Leonardo Savola 2002 Parameters of subject inflection in Italian dialects. In Subjects, Topic, and EPP, P. Svenonius (ed.), 157-199. New York/Oxford: Oxford University Press. Marantz, Alec 1981 A Theory of Grammatical Relations. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Marashi, Mehdi 1970 The Persian verb: a partial description for pedagogical purposes. Ph.D. diss., The University of Texas at Austin. May, Robert 1977 The Grammar of Quantification, Ph. D. Diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. MIT. [Published (1990), Garland, New York] 1985 Logical Form: Its Structure and Derivation. Cambridge/London: The MIT Press. McClosky, James 1996 On the scope of verb movement in Irish. Natural Langauge and Linguistic Theory 14 (1): 241-283. McConnell-Ginet, Sally 1982 Adverbs and logical form. Language 58 (1): 144-184. McDaniel, Dana 1989 Partial and multiple wh-movement. Natural Language and Linguistic Theory 7 (3): 565-604. Megerdoomian, Karine and Shadi Ganjavi 2002 Against optional wh-movement. Ms., Department of Linguistics, University of Southern California. Miyagawa, Shigeru 1994 Scrambling as an obligatory movement. In the Proceedings of Nanzan University International Symposium on Japanese Language Education and Japanese Language Studies, 81-92. Nanzan University, Nagoya. 1997 Against optional scrambling. Linguistic Inquiry 28 (1): 1-25.

References 2001

255

The EPP, scrambling, and wh-in-situ. In Ken Hale: A Life in Language, M. Kenstowicz, (ed.), 293-338. Cambridge/London: The MIT Press. 2003 A-movement scrambling and options without optionality. In Word Order and Scrambling, S. Karimi (ed.), 177-200. Oxford/Berlin: Blackwell Publishers. Mohammad, Jan and Simin Karimi 1992 Light verbs are taking over: complex verbs in Persian. In The Proceedings of Western Conference on Linguistics 5: 195-212. Mohanan, Tara 1990 Arguments in Hindi. Ph.D. diss., Department of Linguistics, Stanford University. Moltmann, Friederike 1991 Scrambling in German and the specificity effect. Ms., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Moyne, John 1970 The structure of verbal constructions in Persian. Ph.D. diss., Department of Linguistics, Harvard University. 1974 The so-called passive in Persian. Foundations of Language 12: 249-267. Moyne, John, and Guy Carden 1974 Subject reduplication in Persian. Linguistic Inquiry 5: 205- 249. Müller, Gereon and Wolfgang Sternefeld, 1993 Improper movement and unambiguous binding. Linguistic Inquiry 24 (3): 461-507. 1994 Scrambling as A-bar movemen. In Studies on Scrambling, N. Corver and H. van Riemsdijk (eds.), 331-385. Berlin/New York: Mouton de Gruyter. 1996 A’-chain formation and economy of derivation. Linguistic Inquiry 27 (3): 480-511. Neeleman, Ad 1994 Scrambling as a D-structure phenomenon. In Studies on Scrambling, N. Corver and H. van Riemsdijk (eds.), 387- 429. Berlin/New York, Mouton de Gruyter. Pesetsky, David 1982 Paths and categories. Ph.D. Diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. 2000 Phrasal movement and its kin. Cambridge/London: The MIT Press. Philips, Colin 1996 Order and structure. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Pickering, M. and G. Barry 1991 Sentence processing without empty categories. Language and Cognitive Processes 6: 229-259. Platzack, Christer 2000 Multiple interfaces. In Cognitive Interfaces: Constraints on Linking Cognitive Information, Urpo Nikanne and Emile van der Zee (eds.), 21-53. New York/ Oxford: Oxford University Press. Pollock, Jean-Yves 1989 Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20 (3): 365-424. Postal, Paul 1974 On Raising. Cambridge/London, The MIT Press.

256

References

Rado, Janina 1994 Object positions, interpretation, and agreement in Hungarian. In Functional Projections, Graduate Linguistic Students Association, E. Benedicto and J. Runner (eds.), 131-152. Department of Linguistics, South College, University of Massachusetts. Raghibdust, Shahla 1994 Multiple wh-fronting in Persian. Cahiers de Languistique 21: 27-58. Richards, Norvin 1997 What Moves Where When in What Language? Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. 2001 Movement in Language: Interactions and Architectures. New York/Oxford, Oxford University Press. Riemsdijk, Henk van 1978 A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Dordrecht, Foris. 1989 Movement and regeneration. In Dialect Variation and the Theory of Grammar, P. Beninca (ed.), 105-135. Dordrecht, Foris. Rizzi, Luigi 1977 The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (ed.), 281-337. Dordrecht/Boston: Kluwer. 1982 Issues in Italian Syntax. Dordrecht, Foris. 1986 Null objects in Italian and the theory of Pro. Linguistic Inquiry 17 (3): 501-557. 1990 Relativized Minimality. Cambridge/London: The MIT Press. 1991 Residual verb second and the wh-criterion. Ms. Universite de Geneve. Rochemont, Michael 1978 A Theory of Stylistic Rules in English. New York: Garland. 1986 Focus in Generative Grammar. Amsterdam/Philadelphia: John Benjamins. Ross, John 1967 Constraints on Variables in Syntax. Ph.D. Diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Rudin, Cathrine 1988 On multiple questions and multiple wh-fronting. Natural Language and Linguistic Theory 6 (4): 445-501. Sabel, Joachim 2000 Partial wh-movement and the typology of wh-questions. In wh-scope marking, U. Lutz, G. Müller, and A. von Stechow. (eds.), 409-447. Amsterdam/Philadelphis, John Benjamins. Saito, Mamoru 1985 Some asymmetries in Japanese and their theoretical implications. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. 1989 Scrambling as semantically vacuous A’ movement. In Alternative Conceptions of Phrase Structure, M. Baltin and A. Kroch (eds.), 182-200. Chicago: University of Chicago Press. 1992 Long distance scrambling in Japanese. Journal of East Asian Linguistics 1: 69-118. 1994a Additional wh-effects and the adjunction site theory. Journal of East Asian Linguistics 3: 195-240. 1994b Improper adjunction. Ms., Department of Linguistics, University of Connecticut.

References 1994c

257

Scrambling and the functional interpretation of wh-phrases. In Explorations in Generative Grammar: A Festschrift for Dong-Whee Yang, K. Young-Sun (ed.), 571-588. Seoul, Korea: Hankuk Publishing Company. Saito, Mamoru, and Hajime Hoji 1983 Weak crossover and move alpha in Japanese. Natural Language and Linguistic Theory 1 (2): 245-259. Saito, Mamoru, and Naoki Fukui 1998 Order in phrase structure and movement. Linguistic Inquiry 29 (3): 439-474. Samiian, Vida 1983 Origins of phrasal categories in Persian: an X-Bar analysis. Ph. D diss., Department of Linguistics, University of California, Los Angeles. 1994 The ezafe construction: some implications for the theory of X-bar syntax. In Persian Studies in North America: Studies in Honor of Mohammad Ali Jazayery, M. Marashi (ed.), 17-42. Maryland: Iranbooks. Santorini, Beatrice 1991 Scrambling and INFL in German. Ms., Department of Linguistics, University of Pennsylvania. Sigurdsson, Halldór Árman 1991 Icelandic Case-marked PRO and the licensing of lexical arguments. Natural Language and Linguistic Theory 9 (2): 327-364. Sinicyn, Michael 1983 Where can we put the filter? Paper presented at GLOW. Sportiche, Dominique 1988 A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19 (3): 425-449 Soheili-Isfahani, Abolghasem 1976 Noun phrase complementation in Persian. Ph.D. diss., Department of Linguistics, University of Illinois at Urbana. Stenson, Nancy 1989 Irish autonomous impersonals. Natural Language and Linguistic Theory 7 (3): 373-406. Stowell, Tim 1981 Origins of phrase structure. Ph.D. diss., Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Svenonius, Peter 2002 Subject positions and the placement of adverbials. In Subjects, Expletives, And the EPP, P. Svenonius (ed.), 201- 242. New York, Oxford University Press Szabolcsi, Anna 1983 Focusing properties or the trap of first order,’ Theoretical Linguistics 10 (1): 125-146. 1993 The Noun Phrase: Syntax and Semantics series. New York: Academic Press. 1997 Strategies in scope taking. In Ways of Scope Taking, A. Szabolcsi (ed.), 109155. Dordrecht/Boston: Kluwer. Tada, Hiroaki 1990 Scramblings. Paper presented at the Workshop on Japanese Syntax, Ohio State University. Takahashi, Daiko 1993 Movement of wh-phrases in Japanes. Natural Language and Linguistic Theory 11 (4): 665-678.

258

References

Takano, Yuji 2003 How antisymmetric is syntax? Linguistic Inquiry 34 (3): 516-526. Taleghani, Azita 2003 Persian wh-construction: a case of F-selection versus C- selection. Ms., Department of Linguistics, University of Arizona. Turano, Guiseppina 1995 On indefinite elements in Albanian. Ms., Universitá di Firenze. Vallduvi, Enric, and Elizabet Engdahl 1996 The linguistic realization of information packaging. Linguistics 34: 459-519. Vazinpour, Nader 1977 Dastur-e Zabân-e Farsi [The Grammar of Farsi Language], Tehran, Iran: Sepahiyan-e Enghelab University. Vikner, Sten 1994 Scandinavian object shift and West Germanic scrambling. In Studies on Scrambling, N. Cover and H. van Riemsdijk (eds.), 487-517. Berlin/New York: Mouton de Gruyter. Webelhuth, Gert 1984 German is configurational. The Linguistic Review 4: 203-246. 1992 Principles and Parameters of Syntactic Saturation. New York/Oxford: Oxford University Press. Whitman, J 1982 Configurationality parameters. Ms., Department of Linguistics, Harvard University. Williams, Edwin 1981 Argument structure and morphology. Linguistic Review 1: 81-114. Willie, Mary Anne 1991 Pronouns and obviation in Navajo. Ph. D. diss., Department of Linguistics, University of Arizona. 2000 The inverse voice and possessive yi-lbi- in Navajo. International Journal of American Linguistics 66 (3): 360-381. Willie, Mary Anne and Eloise Jelinek 2000 Navajo as a discourse configurational language. In The Athabaskan Languages: Perspectives on a Native American Language Family, T.M. Fernald and P. Platero (eds), 252-287. New York/Oxford: Oxford University Press. Wyngaerd, Guido vanden 1989 Object shift as an A-movement rule. MIT Working papers in Linguistics 11: 256-271. Young, R. and W. Morgan 1987 The Navajo Language: A Colloquial Dictionary. Albuquerque: University of New Mexico Press. Zubizarreta, Maria Luisa 1998 Prosody, Focus, and Word Order. Cambridge/London: The MIT Press. Zwart, C. Jan-Wouter 1993 Dutch syntax: A minimalist approach. Ph.D. diss., Department of Linguistics, University of Groningen. 1997 Morphosyntax of Verb Movement: A Minimalist Approach to Dutch Syntax. Dordrechet/Boston: Kluwer.

Subject Index

adverbs High 124-127 Low 124,-125, 127 focus see focus Agree 22-23, 32, 41, 62, 72, 81, 84, 96-97, 102, 110-111, 114-116, 141, 143, 145, 147, 150, 241 antecedent strong 26 weak 26, 27 base generation 28, 33-37, 39, 40, 42, 44, 66, 68 position 39, 56, 89, 139 binding bound variables 48-49 Principle A 47, 51, 123, 165, 173, 176, 182-183, 185-187, 190-192, 245 Principle B 47, 178-179, 182-183 Principle C 47, 52, 60, 173, 179-180, 188, 191-192 Principle of Unambiguous Binding (PUB) 233, 235-237 Burzio’s Generalization 85, 90 Case Accusative 37, 45, 48, 85, 110, 230 Exceptional Case Marking (ECM) 69-70, 72, 86-88 Nominative 95, 98, 230 c-command 31, 48, 68, 120, 126, 166, 176, 181 Chain Condition 176-177, 193 cleft construction 142, 158 Complete Functional Complex (CFC) 47 Complex Predicates (CPr) 11-19, 74-76, 78, 86, 106, 112, 118, 161 agentivity 15 causativity 15 Light Verb (LV) 11-14, 75, 83-84 Non-verbal (NV) element 11-15, 75 Constraint on Interpretation (CI) 237

Constraint on Long Distance Movement (CLDM) 217, 251 constructions ECM 70, 72, 86-88, 94, 96, 104, 116 passive 12, 16, 26, 46, 74-76, 94, 104, 109, 111-112, 115, 118 raising see raising subjectless 70, 78, 84-85, 98, 118, 206 possessive 78, 118 inalienable possessor 78, 89 inalienable pseudo-possessor 83, 99 short infinitives 78, 84 tough 69-70, 72, 88-89, 94, 96, 104, 116, 119 unaccusative 19, 26, 72, 94, 95, 104, 115116, 120-121, 193 control arbitrary 103 object 103 subject 86 copy 41, 46-47, 50-51, 80, 164, 167, 169, 172, 175-176, 179, 227, 236, 240 crossover anti-weak crossover (anti-WCO) 49, 59, 66, 181, 225, 228-229 weak crossover (WCO) 46, 48, 67, 107, 109, 228 definiteness definites 26-28, 30, 94 indefinites 26-27, 30, 133, 151, 166 D-head 140-142, 149, 160 discourse referent 26 Distributed Morphology 13, 20, 26, 45 morphophonological 26 morphosyntactic 26 root 26 D-linked 141, 143 Economy of Derivation 233, 235, 237-238 effects argument structure 213, 215, 217 blocking 146, 148, 164, 191

260

Subject Index

island 30, 201, 216-217 processing 30, 195, 209 expletives covert 28, 69, 70, 72, 90, 93-94, 104, 116 extraposition 89-91 impersonal 89-90, 93 overt 28, 69, 77, 90-91, 93-94, 116, 118 weather 89-91 Extended Projection Principle 25, 71 EPP 29, 32, 45, 49-51, 60, 65-66, 69, 72, 95, 98-99, 117, 119-120, 213-216, 225 EPPg 70, 94, 99-102, 104, 113, 117, 120, 122-123, 239 EPPs 70, 99, 109-110, 113, 130-131, 150151, 153, 156-158, 161, 200-201, 225 features EPP see Extended Projection Principle EPPg see Extended Projection Principle EPPs see Extended Projection Principle focus 130, 133-134, 140, 143, 150-157, 161 interpretable 22 uninterpretable 22, 31-32 wh 22, 52, 140-141, 143, 145-147, 149151, 154-155, 161, 163, 217, 238, 242, 245 Floating Quantifiers (FQ) 39, 44, 52, 5556, 66, 226, 229-230, 241, 243-244 focus adverb 127, 153, 156, 162 contrastive 6, 9, 16, 22, 25, 28, 87, 130, 132-134, 143, 150, 154, 157-161 inherent 133-134, 140, 152-153, 156 movement see Move/movement position 88, 128, 132, 144, 169 functional head 8-9, 32, 46, 49, 67, 70, 201, 215, 237 phase see phase language Albanian 144 Arabic Egyptian 139 Lebanese 194, 241 Modern Standard 162 Bulgarian 136, 138, 162 Catalan 158, 162 Danish 62, 114 Dutch 2, 31, 35-36, 61, 68

English 9, 13-14, 22, 37, 39, 61-63, 65, 71, 74, 86, 89, 91-103, 111-112, 114-115, 118-120, 128, 136, 142, 158, 162-163, 177, 181, 192, 196, 198-201, 208, 216217, 224, 231-232, 237-239, 241, 245 Faroese 124 Finnish 50, 71, 91, 112, 115, 120-121, 158, 161, 231 French 61, 112, 114 German 2, 9, 22, 30, 33, 35-37, 45-46, 61, 63, 67-68, 120, 124, 145, 149, 160163, 183-185, 192-193, 196, 200, 216, 231-233, 236, 238, 244-245 Germanic 61, 236 Greek 71, 99, 121, 158, 161-162 Hindi 2, 30, 55, 61, 89, 95, 148-149, 185, 192, 244 Hungarian 2, 35-36, 99, 101, 114, 117, 119, 121, 130, 132, 157-158, 161-162, 244 Icelandic 62, 71, 120, 124 Irish 96, 103, 120 Italian 91, 99, 158, 162 Northern dialects 81 Japanese 1-2, 34-35, 37-39, 50, 61-63, 65, 67-68, 119-122, 148, 163, 189, 192, 194, 198, 201, 218, 228, 232, 235, 244 Kikuyu 143 Korean 2, 30, 63, 68, 148, 186-187, 237 Mayan 128 Navajo 154, 161 Persian Dari 3, 144, 163 Farsi 3 Tajiki 3 Polish 63 Romani 145 Rumanian 158, 162 Russian 30, 50, 63, 71, 112, 119, 122123, 128, 189, 218-219 Spanish 71, 91, 121 Turkish 95, 148 Urdu 35, 95 Yaqui 154, 161 Linear Correspondence Axiom (LCA) 56, 31 Logical Form (LF) 22, 24, 31-32, 34, 36-38, 41-42, 48, 51, 55, 57, 61, 65, 154, 164, 187, 226-227, 233-238

Subject Index Merge 3, 21, 23, 33-34, 38-39, 41-42, 66, 68, 99, 101, 115-116, 173, 176, 181-182, 184185, 195, 205, 207, 224, 243 Minimal Link Condition (MLC) 2, 22-23, 29, 50, 147, 151-153, 161-162, 196-197, 201, 239, 241 Minimal Match Condition (MMC)

242

Minimalist Program (MP) 2, 7, 20, 22, 28, 31, 33-34, 66, 200 Move/movement 21, 33, 39, 134, 195, 197, 207, 220, 234, 241 A-, 2, 28-29, 45-49, 51, 54, 58, 61, 66, 6869, 115-117, 120, 123, 181, 188, 193, 213-215, 224, 226, 228, 235, 239, 244 A’-, 2, 45, 51, 54-56, 61, 67, 117, 123, 196, 223, 225, 228-231, 235 Across the Board (ATB) 208, 222, 223 dislocation 3, 22-23, 65 clitic left dislocation (CLLD) 139, 142 left-dislocation 222, 242 focus 124, 134-136, 139, 150-151, 153154, 157-158, 160 improper 226, 231, 235-239 non-operator 58-61 object shift 1, 29, 32, 69, 108-110, 115116, 121 operator 30, 51, 61, 66, 88, 177, 195, 198, 200-201, 205, 216-217, 223, 225, 239 optionality of 23, 34, 124, 136, 153-154, 156, 158, 225 Shortest Move (ShM) 162, 197 subject shift 29, 32, 69, 95, 98, 115-116, 121 typology of 30, 34, 116, 182, 195, 223, 225-226 verb/v 8, 9, 50, 67, 160 wh 18, 24, 66, 77, 128, 134, 136, 139, 143, 145, 149, 151, 192, 197, 200, 205, 216-217, 223-224, 231-232, 237, 239240, 245, 247 negation 25, 50, 106-107, 127, 146, 162, 164-165, 168, 170 Negative Polarity Item (NPI) 146-149, 154-155, 161

261

operator domain 35, 123, 126-127, 159-160, 165 movement see Move/movement parametric 2, 4, 30, 62, 66, 71, 96, 101, 114, 120, 185-187, 191-192, 245 parasitic gap (PG) 52-53, 58, 66 passives see constructions Phase 24-25, 30, 73, 154, 172, 176, 178-179, 184, 190-191, 200-201, 203, 207, 216, 220, 224-225, 241-244 CP 131, 165 edge 24, 70, 73, 123, 153, 175, 182, 200, 204, 240 functional/operator/discourse 24-25, 122 lexical 24-25, 70, 122, 176, 182 propositional 24, 112 strong 24-25, 200, 204-205 phase theory 2, 24, 30, 73, 165, 173-174, 190, 192, 200, 225, 241-243 phrase structure 25, 35, 73-74, 79, 108, 122 possessive constructions see constructions Principle of Economy 7, 23, 157 raising 1, 28, 37, 41-42, 46, 68, 86, 94, 104, 116, 226-227 construction 46, 76, 86, 98, 118 quantifier- 159-161, 171, 266 predicate/verb 57-58, 77, 102, 118, 203 reading 19, 41, 57, 83, 97, 139, 149, 156, 162, 166, 180, 226, 237-238 collective 41, 166-167 contrastive 104, 132, 135, 141, 149, 151, 153-154, 157 distributive 41, 166-167 existential 93, 194 generic 169, 194 scope see scope reconstruction 46-48, 50, 52, 54-58, 61, 6768, 164, 168, 170, 172, 183, 187, 189, 226228 relative clause 43 scope 34, 39-40, 50-51, 105-107, 123, 126128, 138-139, 142, 144-147, 164, 167, 172, 178, 191-192, 194, 224, 226, 240 ambiguity 38, 164, 213, 224-225 marker/marking 50, 144-145, 150, 165, 171

262

Subject Index

narrow 50, 170 reading 46, 107, 169-172 wide 50, 107, 144, 166, 168, 171-172 scrambling local 18, 28, 51, 69, 115, 119, 123 long distance (LDS) 17, 19, 30, 36-37, 39, 41-42, 45, 51-52, 56, 60, 63-64, 148, 159-160, 164, 166-170, 175, 181-182, 189, 193, 195-196, 198-201, 203-205, 209, 213-215, 217, 223, 225, 228-231, 240, 243-244 specificity 20, 26, 119 non-specific 4, 7, 16, 18-19, 26-29, 69, 76, 95-97, 100, 104-107, 109-110, 119121, 133, 153, 158-159, 161, 184 specific 4, 7, 16-18, 25-29, 45, 69-70, 72, 74, 86, 88, 94-96, 98, 100, 105-110, 112, 115-116, 119-121, 133, 159, 165, 169, 174, 178, 183-185, 187, 220 subject null-subject 91, 93, 98-99 prominent 69, 111-112 subjectless see constructions Superiority Condition 151-152 Syntactic Word Formation (SWF) 120 syntax derivational 30, 173, 190, 220, 223, 241242 representational 173, 190-191, 220, 223-224, 226, 241-242

telicity 15 atelic 15 telic 14-15 topic background 29, 128-130, 161, 244 prominent 29, 69-70, 111-112, 117, 121122 shifted 29, 128-130, 244 Two Object Position Hypothesis (TOPH) 104, 106-109 Universal Grammar (UG) 1, 2, 6, 30, 101102, 154, 161, 201, 225, 243 verbs causative 14 inchoative 13-14, 74-76, 112 transitive 35, 87, 211 unaccusative 72, 74-75, 95, 120 Webelhuth’s dichotomy 45, 58-61 word order post-verbal 3-4, 10, 31, 35-36, 67, 97, 104, 139 pre-verbal 6, 10, 20, 35, 63, 99 230 SOV 3-8, 13, 31, 50, 61-63, 99 SVO 4-9, 31, 62-63

Author Index

Abney, Steven 31 Aissen, Judith 128 Alexiadou, Artemis 71, 99, 121 Anagnostopoulou, Elena 71, 99, 121 Aoun, Joseph 141, 194, 241 Bailyn, John 30, 45, 50, 63, 65, 67, 71, 119, 122-123, 134, 189-190, 219 Baltin, Mark 128, 224 Barry, Guy 209-211, 224 Barss, Andrew 57, 154, 177, 228 Bateni, Mohammad Reza 11, 31 Bayer, Josef 28, 35-37, 40, 52, 67 Beck, Sigrid 34, 148-149, 163, 224 Belletti, Adriana 57, 68, 227 Bierwisch, Manfred 67 Bobaljik, Jonathan 124 Borer, Hagit 13 Boškovič, Zeljko 30, 37-39, 41, 61, 65, 67 Boyle, J. A. 31 Brame, Michael 31 Bresnan, Joan 139 Browning, Marguerite 193 Burzio, Luigi 85, 90 Carden, Guy 92 Cheng, Lisa 139 Cho, Jai-Hyoung 188 Chomsky, Noam 2, 9, 20-24, 31-32, 42, 4950, 52, 62, 69-72, 90, 99, 102-103, 109, 119, 122, 128, 131, 148, 170, 173, 213, 216, 224225, 233, 235, 237 Cinque, Guglielmo 124, 127, 162 Corver, Norbert 30 Culicover, Peter 134 Dabir-Moghaddam, Mohammad 31, 74, 92 Darzi, Ali 31, 91, 118 Dayal, Veneeta Srivastav 2, 52, 55-56, 192, 225, 229, 244 Déprez, Viviane 2, 28, 45-46, 49, 55, 56, 59, 68, 229, 244 Diesing, Molly 25, 95, 124, 126, 194

É. Kiss, Katalin 35-36, 67, 88, 101, 111, 117-119, 121, 128, 130, 132-134, 139, 157158, 162 Enç, Murvet 26-27 Engdahl, Elizabet 130 Epstein, David 233-234 Ernst, Thomas 126-127, 138 Fanselow, Gilbert 46 Farmer, Anne 35 Farrokhpey, Mahmoud 31 Folli, Raffaella 11-15, 74, 118 Foroughi, A. 31 Fox, Danny 170-171, 192, 226 Fukui, Naoki 28, 30, 34, 38, 39, 41, 62-63, 65, 120, 164, 183, 196, 198, 232 Ganjavi, Shadi 163 Gibson, Edward 209, 211 Ghomeshi, Jila 30-31, 76, 85 Greenberg, Joseph H. 3 Grimshaw, Jane 68 Grewendorf, Günter 30 Guilfoyle, Eithne 121 Haan, G. de 1 Haegeman, Liliane 63 Haider, Hubert 31, 35, 160 Hajatti, Abdol-khalil 74 Hale, Kenneth 13-14, 35, 67, 154 Halle, Morris 26 Harley, Heidi 11-15, 74, 79-80, 118, 120, Hasegawa, Nobuko 163 Hashemipour, Margaret 76 Hickok, Gregory 209, 211 Hoekstra, Teun 1 Hoji, Hajime 1, 23, 34-35 Holmberg, Anders 49, 65, 71, 91, 111-112, 120-121, 124, 130, 23o Horvath, Julia 139 Huang, James 192 Jackendoff, Ray 31, 126, 131, 134, 138 Jazayery, Mohammad Ali 31

264

Author Index

Jelinek, Eloise 25, 95, 154 Johnson, Kyle 128, 224 Jonas, Diane 124

Moyne, John 31, 74, 76, 92 Müller, Gereon 28, 30, 52, 63, 120, 128, 224, 233-238, 245

Kahnemuyipour, Arsalan 139 Karimi, Ezat 193 Karimi, Simin 10-15, 19, 27, 30-31, 34, 40, 65, 74, 76, 87, 97, 104-106, 118-120, 136, 138, 150, 163, 193, 215, 224, Kayne, Richard 3-7, 31, 79, 119, 228 Kerstens, Johan 1 Keyser, Samuel Jay 13-14 Khanlari, Parviz 31 Kim, Dae-Bin 54-55, 121, 186-188 Kim, Shin-Sook 34, 148-149, 224 King, Tracy 128 Kornfilt, Jaklin 28, 35-37, 40, 42, 67 Koster, Jan 31, 128, 224 Kratze, Angelika 25, 95 Kuroda, S.-Y. 65, 164

Neeleman, Ad 28, 35-36, 61-62, 68 Nikanne, Urpo 49, 65, 91, 112, 121, 130, 230

Laka, Itziar 227 Larson, Richard 31 Lasnik, Howard 2, 31, 70, 102-103, 128, 170, 224, 232 Lavine, James 45, 71 Lazard, Gilbert 31 Lebeaux, David 183 Li, Charles 121 Li, Yen-hui Audrey 141, 194, 242 Lotfi, Ahmad 145 Mahajan, Anoop 2, 28, 45-46, 48-49, 52, 54-55, 67, 181, 185-186, 188 Manzini, Rita 71 Marantz, Alec 26, 35 Marashi, Mehdi 31 May, Robert 57 McClosky, James 120 McConnell-Ginet, Sally 126 McDaniel, Dana 145 Mchombo, Sam 139 Megerdoomian, Karine 163 Miyagawa, Shigeru 30, 34, 45, 50-52, 60, 65, 67, 71, 119-122, 134, 163, 189, 194, 224 Mohammad, Jan 12, 163 Mohanan, Tara 35 Moltmann, Friederike 184 Morgan, William 154

Paper, H. H. 31 Perkins, E. 154 Pesetsky, David 31, 148, 163, 224 Philips, Colin 227 Pickering, Martin 209-211, 224 Platzack, Christer 71 Pollock, Jean-Yves 31 Postal, Paul 68 Rado, Janina 121 Raghibdust, Shahla 139 Richards, Norvin 138, 162-163, 197, 245 Riemsdijk, Henk van 28, 30, 36 Rizzi, Luigi 57, 68, 93, 134, 162, 176-177, 193, 227, 232 Rochemont, Michael 134 Ross, John Robert 1, 42 Rudin, Cathrine 162 Sabel, Joachim 30, 143-144 Saito, Mamoru 1, 28, 30, 34-36, 38-39, 41, 45, 51-52, 54, 60-61, 63, 65, 120, 128, 164, 183, 192, 196, 198, 218, 224, 229, 232 Samiian, Vida 31, 97, 159 Santorini, Beatrice 46, 185 Savola, Leonardo 71 SigurDsson, Halldór Árman 103 Sinicyn, Michael 218 Sportiche, Dominique 67 Soheili-Isfahani, Abolghasem 31, 74, 92 Speas, Margaret 154 Stenson, Nancy 103 Sternefeld, Wolfgang 28, 30, 52, 63, 120121, 128, 224, 228-232, 245 Stowell, Tim 31 Svenonius, Peter 124, 128 Szabolcsi, Anna 121, 132-133 Tada, Hiroaki 61 Takahashi, Daiko 30, 37-39, 41, 61, 65, 67, 163 Takano, Yuji 227 Taleghani, Azita 136, 163

Author Index Thompson, Sandra 121 Turano, Giuseppina 144 Vallduvi, Enric 130 Vazinpour, Nader 31 Vikner, Sten 28, 52-53 Webelhuth, Gert 2, 28, 30, 45, 53, 58-61, 66-67, 124, 225

Whitman, John 35 Williams, Edwin 35 Willie, MaryAnn 154 Wyngaerd, Guido vanden

67

Young, Robert 154 Zubizarreta, Maria Luisa 131 Zwart, C. Jan-Wouter 30-31

265