Wh-Movement: Moving On (Current Studies in Linguistics)

Cheng_pb.qxd 5/9/06 9:03 AM Page 1 L I N G U I S T I C S Moving On EDITED BY Lisa Lai-Shen Cheng and Norbert Corve...

Author: Noam Chomsky | Lisa Lai-Shen Cheng | Norbert Corver

25 downloads 932 Views 2MB Size Report

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

Report copyright / DMCA form

DOWNLOAD PDF

Cheng_pb.qxd

5/9/06

9:03 AM

Page 1

L I N G U I S T I C S

Moving On EDITED BY

Lisa Lai-Shen Cheng and Norbert Corver

FOREWORD BY

Noam Chomsky

Lisa Lai-Shen Cheng is Professor of Linguistics at Leiden University. Norbert Corver is Professor of Dutch Linguistics at Utrecht University.

CONTRIBUTORS

Brian Agbayani Lisa Lai-Shen Cheng Sandra Chung Norbert Corver Caterina Donati Kleanthes K. Grohmann Toru Ishii Heejeong Ko Howard Lasnik Philip LeSourd Chris H. Reintges Luigi Rizzi Balázs Surányi Akira Watanabe Henrietta Yang “A topic of everlasting importance and interest, kept at the forefront of the field by this fine collection. Some of the articles, by a good mix of eminent and rising scholars, have already attracted widespread attention. Not to be missed.” — C.-T. James Huang, Professor of Linguistics, Harvard University

Current Studies in Linguistics 42

0-262-53279-4

978-0-262-53279-2

The MIT Press Massachusetts Institute of Technology Cambridge, Massachusetts 02142 http://mitpress.mit.edu

Cheng and Corver, editors

Wh-movement — the phenomenon by which interrogative words appear at the beginning of interrogative sentences — is one of the central displacement operations of human language. Noam Chomsky’s 1977 paper “On Wh-Movement,” a landmark in the study of wh-movement (and movement in general), showed that this computational operation is the basis of a variety of syntactic constructions that had previously been described in terms of constructionspecific rules. Taking Chomsky’s paper as a starting point, the contributors to this collection reconsider a number of the issues raised in “On Wh-Movement” from the perspective of contemporary Minimalist syntactic theory (which explores the thesis that human language is a system optimally designed to meet certain interface conditions imposed by other cognitive systems with which the language faculty interacts). They discuss such wh-movement issues as wh-phrases and pied-piping, the formation of A-bar chains and the copy theory of movement, cyclicity and locality of wh-movement, and the typology of wh-constructions. By reconsidering core characteristics of the wh-movement operation first systematically discussed by Chomsky from the Minimalist perspective, this volume contributes to the further development of the theory of wh-movement and to the general theory of movement.

WH -Movement: Moving On

WH -Movement

WHO WHAT WHERE -Movement Moving On WHEN WHY EDITED BY

Lisa Lai-Shen Cheng and Norbert Corver FOREWORD BY

Noam Chomsky

Wh-Movement: Moving On

Wh-Movement: Moving On

edited by Lisa Lai-Shen Cheng and Norbert Corver

The MIT Press Cambridge, Massachusetts London, England

6 2006 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. MIT Press books may be purchased at special quantity discounts for business or sales promotional use. For information, please e-mail [email protected] .edu or write to Special Sales Department, The MIT Press, 55 Hayward Street, Cambridge, MA 02142. This book was set in Times New Roman on 3B2 by Asco Typesetters, Hong Kong, and printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Wh-movement : moving on / edited by Lisa Lai-Shen Cheng and Norbert Corver. p. cm. — (Current studies in linguistics ; 42) Includes bibliographical references and index. ISBN 0-262-03346-1 (alk. paper) — ISBN 0-262-53279-4 (pbk. : alk. paper) 1. Grammar, Comparative and general—Interrogative. 2. Grammar, Comparative and general—Word order. 3. Grammar, Comparative and general—Syntax. 4. Minimalist theory (Linguistics). I. Cheng, Lisa Lai-Shen. II. Corver, Norbert, 1963– III. Current studies in linguistics series ; 42. P299.I57W47 2006 415—dc22 2005058016 10 9 8 7

6 5 4 3 2

1

Contents

Contributors

vii

Foreword ix Noam Chomsky Acknowledgments

xiii

Chapter 1 Lines of Inquiry into Wh-Movement

1

Lisa Lai-Shen Cheng and Norbert Corver

PART I Wh-Phrases and Pied-Piping

19

Chapter 2 21

On Wh-Head Movement

Caterina Donati

Chapter 3 The Pied-Piper Feature

47

Akira Watanabe

Chapter 4 Pied-Piping, Feature Movement, and Wh-Subjects 71 PART II ¯ -Chains and Copy Theory A

Brian Agbayani

95

Chapter 5 On the Form of Chains: Criterial Positions and ECP E¤ects 97

Luigi Rizzi

Chapter 6 On Overt and Covert Wh- and Relative Movement in Hindi and Punjabi 135

Henrietta Yang

vi

Contents

Chapter 7 Movement, Wh-Agreement, and Apparent Wh-in-situ 165

Chris H. Reintges, Philip LeSourd, and Sandra Chung

PART III Cyclicity and Locality

195

Chapter 8 Conceptions of the Cycle

197

Howard Lasnik

Chapter 9 On the Relaxation of Intervention E¤ects 217

Toru Ishii

PART IV Wh-Constructions

247

Chapter 10 Top Issues in Questions: Topics– Topicalization–Topicalizability 249

Kleanthes K. Grohmann

Chapter 11 Mechanisms of Wh-Saturation and Interpretation in Multiple WhMovement 289

Bala´zs Sura´nyi

Chapter 12 On the Structural Height of Reason Wh-Adverbials: Acquisition and Consequences 319 Index

351

Heejeong Ko

Contributors

Brian Agbayani California State University, Fresno

Bala´zs Sura´nyi Hungarian Academy of Sciences

Lisa Lai-Shen Cheng Leiden University

Akira Watanabe University of Tokyo

Sandra Chung University of California, Santa Cruz

Henrietta Yang The University of Texas at Austin

Norbert Corver Utrecht University Caterina Donati University of Urbino Kleanthes K. Grohmann University of Cyprus Toru Ishii Meiji University Heejeong Ko Stony Brook University Howard Lasnik University of Maryland Philip LeSourd Indiana University Chris H. Reintges Leiden University/ULCL Luigi Rizzi University of Siena

Foreword Noam Chomsky

This outstanding collection of papers is an impressive testimonial to the great advances in empirical knowledge and theoretical understanding in recent years. Many of the questions addressed here could not even have been posed, let alone answered, not very long ago. Nevertheless, many of the earliest questions remain, though in quite new forms, as attested by a number of the papers here: questions about the nature of cyclic/ compositional operations; locality; what moves (heads? features?); the nature of the obscure issue of pied-piping; projection under movement; varieties of marking of intermediate positions; whether movement is implicated at all where that has often been assumed, rather than other relations with base generation; and many other questions that have a familiar ring. Old chestnuts often have hard shells. That should hardly surprise us in a ﬁeld that barely existed 50 years ago, despite a tradition going back thousands of years. The basic issue addressed in these papers is the ubiquitous phenomenon of displacement: basically, the fact that expressions are produced/heard in one position but interpreted as well in other positions—positions in which, furthermore, similar expressions (or even the same ones) appear and are interpreted in terms of natural local conﬁgurations. That much is uncontroversial. How to deal with the phenomenon, however, is anything but uncontroversial. Within generative grammar in the modern sense, the earliest work that tried to account for displacement, 50 years ago, did not have rules of movement at all. Rather, such phenomena as wh-displacement were formulated in terms of permutation. Thus, sentences (1) and (2) were in e¤ect derived by permutation of X and a in the syntactic object X-a-Y, where a is the element displaced to the front; (2) was derived from the counterpart of (1) with who replacing Mary (technically, the form underlying it).

x

Foreword

(1) Did John see Mary? (2) Who did John see? The reasoning was that the derivation of (2) from the yes/no construction accounted for its interrogative character. Such forms as Who saw Bill? were assigned what seems to be the appropriate form and interpretation in the same way. It very soon became clear that this approach, despite some attractive features, was hopelessly inadequate, for reasons too well understood today to review. In the following years, the approach was reformulated in terms of rules of movement, each of which leaves a gap and in this respect resembles the permutation rules of the initial work. By the early 1960s, it was becoming clear that these rules of movement were subject to various constraints: among those suggested were the A-over-A Condition, restrictions against reapplying an operation within its domain, what later came to be known as the Wh-Island Constraint, and others. The classic formulation of such ideas was, of course, Ross’s theory of islands in 1967, which remains a treasure trove of insights and empirical wealth. By the late ’60s it was also becoming clear that movement leaving a gap was inadequate: it is necessary to deal somehow with a fundamental duality of semantics, distinguishing argument structure from discourse-related and scopal phenomena. Several approaches developed to this array of problems. Within the Extended Standard Theory (Y-model) that was crystallizing at the time, the duality of semantics was expressed in terms of ‘‘deep and surface structure’’ e¤ects. The issues remain, though much revised, and with a somewhat di¤erent categorization of phenomena. By the late 1960s, there were also attempts to account for the taxonomy of islands in terms of more fundamental principles of locality and others, and in those terms, to reduce the array of possible movement rules to a much more restrictive ¯ -movement, later just Move a (or perhaps even Lasnik typology: A- and A and Saito’s simpliﬁcation to A¤ect a). One of the oldest chestnuts is why language should have such rules at all. To reframe the question in the terms of Luigi Rizzi’s elegant reformulation here, why should the basic elements of language be chains, each of which has a criterial position in addition to the position in which argument structure is determined within a local conﬁguration? Part of the answer might be that the conceptual-intentional interface imposes a condition of semantic duality, one of the interface conditions that the computational system of language must satisfy in some manner to be usable at all. But there are many devices that might meet the semantic du-

Foreword

xi

ality condition, so why does language typically resort to chains with criterial positions? Not always, at least on the surface, though the correlation is close enough to suggest that if we understood enough, it would be perfect. It had always seemed (to me in particular) that resort to movement operations to form such chains is a curious ‘‘imperfection’’ of language that had to be explained somehow, and that other approaches, with different devices to account for the phenomena, might turn out to be virtual notational equivalents, perhaps distinguishable only by subtle empirical tests. A few years ago it became clear that this is a misunderstanding. The conclusion follows from a closer analysis of the core operation on which language is based, as is any system of hierarchic discrete inﬁnity: the operation Merge, iterable without limit, which takes syntactic objects already formed—say, X and Y—and forms from them a new object Z. An elementary principle of computational e‰ciency, presumably available for language without speciﬁc Universal Grammar design, is the ‘‘No-Tampering Condition’’ NTC: in forming Z from (X; Y), X and Y should remain unmodiﬁed. In the simplest case, then, Z ¼ {X; Y}. Suppose that for expository convenience we think of Y as merged to X to form Z. There are, then, two logically possible cases: Y is not part of X (external Merge, as when (V; DP) merge to form {V; DP} ¼ VP), or Y is part of X (internal Merge, as when a wh-phrase is raised from CP to Spec,C, forming an extended CP). The NTC has always been assumed without question for external Merge. Extending it to internal Merge, in accord with elementary requirements of computational e‰ciency, automatically yields the so-called copy theory of movement, the null hypothesis. It follows, then, that the absence of movement operations would be an imperfection that requires explanation. Such operations ‘‘come free,’’ given the property of discrete inﬁnity. Any device other than movement with copies bears an empirical burden of proof, in fact a double burden: to justify the banning of internal Merge, and to justify the additional device. That line of reasoning seems to me to provide a substantial basis for proceeding along the general lines assumed in the papers collected in this volume. These papers largely trace back to the stage of research in which ¯ -movement, in particular wh-movement, the basic concepts of A- and A were emerging, along with the general conditions that they appear to meet. The papers incorporate a rich variety of advances since, and carry inquiry forward in original and highly provocative ways, opening many new and stimulating questions as well. There is, I think, little doubt that this volume will be an indispensable basis for what lies ahead.

Acknowledgments

This collection of papers grew out of a workshop entitled ‘‘On WhMovement,’’ which took place at Leiden University (ULCL) and Utrecht University (UiL-OTS) in December 2002 to mark the appearance of Noam Chomsky’s paper ‘‘On Wh-Movement’’ 25 years earlier. We gratefully acknowledge the ﬁnancial support of the Netherlands Organization of Pure Scientiﬁc Research (NWO), ULCL, and UiL-OTS. We would also like to thank our anonymous reviewers.

Wh-Movement: Moving On

Chapter 1 Lines of Inquiry into WhMovement

1.1

Lisa Lai-Shen Cheng and Norbert Corver

‘‘On Wh-Movement’’ and Beyond

In Language and Problems of Knowledge, Chomsky (1988, 60) characterizes linguistic inquiry as a process that typically begins with the study of structured expressions of a particular language, then passes into a grammatical description of (larger fragments of ) a particular language (i.e., a theory of competence), and ﬁnally addresses the explanatory task of answering why the facts of human language are the way they are. This last task leads to research into the language faculty and into the construction of a theory of Universal Grammar (UG). As Chomsky notes, these three tasks of linguistic inquiry are mutually interdependent but, in practice, do not necessarily evolve in strict succession. Linguistic inquiry may take the route from studying linguistic expressions, to constructing a language-particular grammar, and ﬁnally to developing a theory of UG, but it may just as well operate in reverse: a particular conception of UG may, for example, inﬂuence the way the syntactician perceives linguistic expressions, and this in turn will a¤ect his or her description of a particular language. One of the ﬁnest illustrations of the latter route is Chomsky’s (1977) paper ‘‘On Wh-Movement’’ (henceforth: OWM), the paper we commemorate in this volume. 1.1.1 On Wh-Movement OWM adopts and builds on the 1973 ‘‘Conditions on Transformations’’ framework, in which Chomsky initiated a research program that would lead to, among other things, a general theory of movement. The transformational rule system of this theoretical framework consists of general movement rules such as ‘‘Move wh-phrase’’ and ‘‘Move NP,’’ which essentially apply blindly. It should be noted that in OWM (p. 85, (43)), the formulation of the wh-movement rule speciﬁes the target of the movement operation: ‘‘Move wh-phrase into COMP.’’

2

Lisa Lai-Shen Cheng and Norbert Corver

The generative force of the movement rules is constrained by a set of conditions. One such condition is the condition of the (strict) cycle, which governs the order in which transformational processes apply.1 Another, the Subjacency Condition, states that movement is local, where locality is deﬁned in terms of the number of cyclic nodes (also called bounding nodes) that the moved constituent crosses (a single cyclic node being the maximum). More speciﬁcally, as stated in OWM (p. 73, (6)):2 (1) A cyclic rule cannot move a phrase from position Y to position X (or conversely) in: . . .X. . . [a . . . [b . . .Y. . . ] . . . ] . . .X. . . , where a and b are cyclic nodes. Phenomena of long-distance movement are analyzed as movement dependencies created by the successive-cyclic application of local movement steps (COMP-to-COMP movement in the case of wh-movement). A third type of condition governing (wh-)movement concerns the constituent structure derived by the application of movement. The movementderived structures, involving a relationship between a moved element and its coindexed trace, are taken to be subject to conditions such as the Propositional Island Condition (PIC) and the Speciﬁed Subject Condition (SSC) (see OWM, 73, (4)–(5)), which also govern anaphoric relations that are deﬁned by so-called rules of construal. The dependency between the wh-phrase and its trace is interpreted as a quantiﬁer-variable relationship at Logical Form (LF) by a rule of interpretation for wh-phrases. Importantly, as Chomsky points out (OWM, 83), the trace left behind by the moved phrase cannot always be identiﬁed as the variable within the scope of the wh-quantiﬁer. This holds speciﬁcally in contexts where pied-piping is involved in the wh-movement process (e.g., [Whose book]i did Mary read ti ?). In those cases, pied-piped material (here, book) must be placed back—reconstructed—into the extraction site. With the ‘‘big picture’’ of movement theory presented in ‘‘Conditions’’ as a theoretical background, in OWM Chomsky analyzes a variety of clausal constructions from English, such as comparatives (2a), topicalization (2b), clefts (2c), (ﬁnite and inﬁnite) indirect questions (2d), (ﬁnite and inﬁnitival) relatives (2e), inﬁnitival complements within the adjective phrase (2f ), and easy-to-please constructions (2g). (2) a. They have more enemies than we have b. This book, I really like c. It is this book that I really like

Lines of Inquiry into Wh-Movement

d. e. f. g.

3

I wonder who to see I found a book for you to read John is tall enough for us to see John is easy (for us) to please

Until the appearance of OWM, generative analyses of these constructions mostly took the route of linguistic inquiry that goes from studying structured expressions of a language to constructing a language-particular grammar, and then to inquiring into the language faculty. An automatic consequence of this route was that syntactic analyses often did not go beyond the construction-speciﬁc and language-particular description of human language. Each clausal construction in (2) was derived by its own transformational rule (Comparative Deletion, Topicalization, Clefting, Indirect Question Formation, Relativization, a rule for adjective-qualiﬁer complements, Tough-Movement), each stated in a rather languagespeciﬁc way.3 The ﬁnal result of this type of inquiry was not greater insight into the nature of the language faculty, but a bigger grammatical apparatus (e.g., a large class of possible transformational rules) for describing a particular language. OWM turned generative linguistic inquiry in a new direction. It showed that a particular conception of UG (speciﬁcally, the subtheories that concern the operation Move wh) inﬂuences the way one assigns structures to linguistic expressions of a speciﬁc language and how it inﬂuences the form of language-particular grammars. More speciﬁcally, Chomsky showed convincingly how the descriptively rich transformational rule system for English (e.g., for the constructions in (2)) could be eliminated, and how a varied set of English constructions like those in (2) could be explained in terms of a general (i.e., universal) movement rule, Move wh, in combination with general principles of language, each accounting for some general (universal) property of the wh-movement process. This integration of general principles of grammar with compelling, detailed analyses of particular phenomena in English makes OWM an exemplary publication in generative syntax. Importantly, this reduction operation not only led to dismantling the grammatical apparatus for the description of English, it also limited the class of possible movement rules, as deﬁned by UG (see OWM, 111). Movement, being one of the computational operations deﬁned by UG, is no longer deﬁned in terms of an information-laden structural description and a structural change. The class of movement rules (e.g., Move NP, Move wh) just states that something is moved. Movement ingredients

4

Lisa Lai-Shen Cheng and Norbert Corver

such as ‘‘What is moved?’’, ‘‘Where is it moved to?’’, ‘‘How local is the movement operation?’’, and ‘‘How is the movement operation ordered with respect to other movement rules?’’ are no longer part of the transformational rule itself; instead, they are ‘‘reallocated’’ as general, conﬁgurational properties to di¤erent components of the grammar (UG). The importance of OWM also lies in the contention that these general conﬁgurational movement properties that have been isolated can serve as a kind of diagnostic tool for identifying the application of wh-movement in some syntactic construction. The key properties that Chomsky considers to be diagnostic of wh-movement are these (see OWM, 86, (49)): (3) a. It leaves a gap. b. Where there is a bridge, there is an apparent violation of Subjacency, the PIC, and the SSC. c. It observes the Complex NP Constraint (CNPC). d. It observes the Wh-Island Constraint. Property (3a) follows from the assumption that wh-movement moves a phrase, which in turn implies that a gap (i.e., a category voided of phonological content) is left behind in the extraction site. Property (3b) refers to the fact that wh-movement can circumvent a violation of Subjacency, the PIC, and the SSC in structural contexts involving so-called bridge verbs. An important presupposition here is that wh-movement can proceed successive-cyclically (i.e., stepwise on successive cycles, making use of all the intermediate COMP positions that separate the extraction site from the ultimate landing site). The island properties (3c) and (3d) are implied by Subjacency.4 Chomsky formulates the issue about the diagnostic function of the whproperties in (3) as the central research question of OWM (see p. 86, (50)). (4) ‘‘Where we ﬁnd the conﬁguration [(3)] in some system of data, can we explain it on the assumption that the conﬁguration results from wh-movement?’’ The aim of OWM is thus to investigate whether or not the conﬁguration in (3) can serve as a diagnostic for wh-movement on the basis of the constructions in (2). Much of OWM then consists of demonstrating that each of the clausal constructions in (2) has (fully or to a great extent) the properties in (3). To give an example: the following sentences, most drawn from OWM, illustrate that an inﬁnitival indirect question construction like (2d) is based on wh-movement:

Lines of Inquiry into Wh-Movement

5

(5) a. I wonder [who to see] b. I wonder [who to order Mary [to promise [to visit]]] b 0 . I wonder [who to persuade Mary [that she should promise [to visit]]] c. *I wonder [who to insist on [ NP the principle [ S 0 that Bill should visit]]] d. *I wonder [what i to ask her [whenj to buy t i t j ]] (5a) illustrates that movement of who leaves behind a gap in the direct object position after to see. (5b) and (5b 0 ) show that who can be moved from the complement position of visit to the COMP position of the highest embedded clause thanks to the bridge properties of the verbs to promise, to persuade, and to order. (5c) and (5d) show, respectively, that inﬁnitival indirect questions obey the CNPC and the Wh-Island Constraint. In an inﬁnitival indirect question like (2d), a wh-word (who) is overtly present at the surface, which, of course, is very helpful for identifying whmovement. Identifying wh-movement in constructions lacking an overt wh-phrase is arguably more exciting, since there is no direct (overt) element signaling the presence of a displaced wh-phrase. Take, for example, a comparative construction like (2a). Even though no overt wh-expression is present in the comparative clause,5 the wh-diagnostics lead us to analyze this construction in terms of wh-movement: that is, a wh-phrase is moved to COMP (Spec,CP in present-day terms). Consider in this regard the following examples, where t indicates the original extraction site:6 (6) a. b.

John met more linguists [than you met t] John met more linguists [than we thought [you said [Bill believed [Sue met t]]]] c. *John met more linguists than I believed [ NP the claim [ S 0 that Sue had met t]] d. *John met more linguists than I wondered [ S 0 when Sue had met t]

(6a) illustrates the gap property. (6b) shows that an apparently unbounded dependency can be created if the necessary bridge requirements are fulﬁlled. (6c) exempliﬁes the CNPC e¤ect, and (6d) shows that comparative formation is subject to the Wh-Island Constraint. The presence of the conﬁguration of wh-properties suggests that a rule of wh-movement is also involved in the derivation of comparatives. According to OWM, this rule moves an (overt) wh-phrase to COMP, where it subsequently gets deleted by a (local) wh-phrase deletion rule (see Chomsky’s deletion rule (47), p. 86).

6

Lisa Lai-Shen Cheng and Norbert Corver

Thus, OWM made a crucial contribution to the abandonment of a construction-based view of syntactic rules. Furthermore, it introduced the idea of a diagnostic tool in generative syntax—an instrument that can be used to ﬁnd out whether some computational operation is involved in the derivation of a linguistic expression. Ever since its introduction, this tool has been recognized as an important instrument in syntactic research and it has been employed by numerous linguists in studying a variety of constructions (see, e.g., Van Riemsdijk 1978; Groos and Van Riemsdijk 1981; Cinque 1982, 1990). Besides making these contributions to linguistic description and theorizing, the approach taken in OWM led to a drastic reduction in the number of available rules of grammar, and therefore in the class of available grammars, which Chomsky considered to be ‘‘the major goal of linguistic theory’’ (p. 125). OWM also led to further reﬁnement of the universal conditions that restrict the expressive power of the general transformational rules (Move wh, Move NP). Regarding the reduction of the available rules, OWM contributed in an important and compelling way to the debate over whether or not syntactic theory should allow unbounded transformations. OWM defends the position that there is no distinction between bounded and unbounded rules. All (cyclic) movement rules are subject to the Subjacency Condition and are therefore bounded. Apparently unbounded movement dependencies are the result of successive-cyclic movement. 1.1.2

Speciﬁc Issues

1.1.2.1 Comparatives The discussion between Bresnan (1975, 1976a,b, 1977) and Chomsky (OWM) about the existence of unbounded rules may rightly be called exemplary as regards the integration of theoretical and empirical argumentation. The phenomenon of comparative formation in English played a central role in this debate. As shown in (6b), the compared constituent in the comparative clause can be arbitrarily far from the head of the clause in comparative deletion constructions (Ross 1967). Bresnan proposes to account for this unbounded dependency between the two compared constituents by means of a transformation termed Comparative Deletion. This rule deletes the compared constituent of the comparative clause under identity with the head of the clause and can apply over an unbounded domain (so-called deletion over a variable). An important assumption in Bresnan’s analysis is that comparative deletion (see (6a)) and comparative subdeletion (see (7)) are one and the same phenomenon, and as such involve the same computational operation.

Lines of Inquiry into Wh-Movement

(7) John met more linguists than you met [

7

biologists]

As Bresnan (1977) points out, the missing constituent in the subcomparative clause does not correspond to anything that could have been moved. As shown in (8), for example, a speciﬁer like how many cannot be removed from a noun phrase. (8) *[How many] i did John meet [t i biologists]? For Bresnan, the well-formedness of (7) and the ill-formedness of (8) show that movement cannot be involved in subdeletion. And if subdeletion and comparative deletion are one and the same process, then comparative deletion should not be derived by movement either. In view of facts like (6a–d), this in turn leads to the conclusion that deletion must be long-distance deletion that is subject to island constraints, implying that islandhood is not a diagnostic property of movement processes. Chomsky goes into the matter at length in OWM. He presents evidence calling into question a uniform treatment of comparative deletion and comparative subdeletion. If comparative deletion involves wh-movement and hence displays the wh-diagnostics, then subdeletion should not be accounted for in terms of this movement rule. 1.1.2.2 Cylicity and Cyclic Nodes As pointed out above, OWM also led to further reﬁnement of some of the universal conditions that restrict the expressive power of the general transformational rules. One reﬁnement we would like to mention in particular concerns the class of cyclic/ bounding nodes that are relevant for Subjacency. In OWM, Chomsky proposes that, besides NP and S 0 , the category S should be added to the inventory of cyclic nodes.7 This has the beneﬁcial e¤ect that the impossibility of extracting from a subject NP—so far captured by the ad hoc Subject Condition—can be explained in terms of a general condition, Subjacency (see OWM, 111). In (9), for example, whom crosses two boundaries, NP and S. (9) *[Whom i did [ S [ NP your interest in t i ] seem to me rather strange]]? As Chomsky notes (OWM, 112), wh-movement cannot extract a phrase from a nonsubject NP either. In certain cases, this prediction turns out to be correct. (10) *Who i did [ S John destroy [ NP a book [about t i ]]]? In others, extraction from NP appears to be possible.

8

Lisa Lai-Shen Cheng and Norbert Corver

(11) a. Who did John write a book about? b. Who did you see a picture of ? For (11a), Chomsky adopts Bach and Horn’s (1976) proposal that the PP about who is in fact not part of the NP at D-Structure but rather a constituent within VP (see (12a)). Evidence for this is the possibility of pronominalizing the sequence a book, as in (12b). (12) a. John wrote [a book] [about who]? b. John wrote it [about Nixon] As Pronominalization cannot be applied to (11b)—*John saw it of Nixon—Chomsky concludes that (11b) is derived by ﬁrst applying PPExtraposition to of who and then applying Wh-Movement to the (NPexternal) PP. Schematically: (13) Who k did you see [a picture t i ] [ PP of t k ]i ? Although we cannot do full justice to the many interesting facets of OWM in this brief introduction, we believe that the issues we have touched upon su‰ciently indicate that OWM forms one of the landmarks in the development of generative syntax, and especially in the development of a general theory of movement. Perhaps most important, it set the stage for a radical shift from a construction-oriented and languagespeciﬁc approach to language to a more principle-oriented approach. Furthermore, it showed that this shift did not have to be at the expense of empirical coverage. Quite the contrary, we would say. This integration of principle-based syntax and empirical coverage also characterized the next stage in the history of generative grammar, the Principles-andParameters (P&P) approach (Chomsky 1980, 1981, 1986b), which uniﬁes a great many seemingly disparate grammatical phenomena in a conceptually simple and elegant overall framework of principles. 1.1.3 Minimalist Perspectives In this section, we make a few remarks about the theory of (wh-) movement (also called ‘‘internal Merge’’) in the context of the Minimalist Program (Chomsky 1995), which explores the thesis that human language may be a ‘‘perfect system’’—that is, a system optimally designed to meet certain interface conditions imposed by other cognitive systems that the language faculty (UG) interacts with. When we apply the minimalist approach to the phenomenon of wh-movement, the question arises how its properties can be explained in terms of principles of the interface systems (i.e., representational economy) and what principles of e‰cient com-

Lines of Inquiry into Wh-Movement

9

putation underlie the derivation of (clausal) expressions featuring a whphrase (i.e., derivational economy). Not unexpectedly, the properties or ingredients that are to be studied from this minimalist perspective are quite similar to those that were central in the ‘‘Conditions’’ framework (Chomsky 1973, OWM): ‘‘What is moved?’’, ‘‘Where is it moved to?’’, ‘‘How local is the movement operation?’’, ‘‘How is the movement operation ordered with respect to other operations of the grammar?’’ To this list, we should add the question, ‘‘What causes movement? That is, why move?’’ This question ﬁgures prominently in current minimalist research on movement, but was also addressed, though less extensively, in research on wh-movement within the ‘‘Conditions’’ framework.8 1.1.3.1 Optionality versus Last Resort In OWM (fn. 19), Chomsky states, ‘‘My assumption is that all rules of the ‘core grammar’ . . . are optional, the apparent obligatoriness deriving from ﬁlters and principles of interpretation, along the lines discussed in Chomsky (1973).’’ In The Minimalist Program (Chomsky 1995), he takes a somewhat di¤erent view, proposing a principle of derivational economy according to which movement is an operation of last resort. Thus, there is no free, truly optional movement. Movement must take place in order to satisfy some interface requirement. This property is formally implemented by taking the movement process to be triggered by the goal of eliminating uninterpretable properties/features of the moved constituent (e.g., the whphrase) and of the target of movement (some functional head—e.g., C). 1.1.3.2 Traces versus Copies An important departure from the ‘‘Conditions’’ framework concerns the structure derived by movement—in particular, the nature of the chain created by movement. At the time OWM was written, it was assumed that movement left behind a trace that was coindexed with the fronted (wh-) phrase. From the minimalist perspective, however, no new grammatical objects (e.g., categories, features) may be introduced in the course of the derivation. The only material available for constructing a linguistic expression is that provided by the lexical items that are part of the numeration (the so-called Inclusiveness Condition: no new features are introduced by the computational system (see Chomsky 2000)). As a consequence of this, the copy theory of movement was reintroduced (see also Chomsky 1993 and in fact Chomsky 1955): moving a phrase involves copying the phrase into a higher position, the original occurrence subsequently being deleted (or left unpronounced).

10

Lisa Lai-Shen Cheng and Norbert Corver

This theory has the advantage that no complex theory of reconstruction is needed to account for various interpretive phenomena; see Chomsky 1993 for discussion. 1.1.3.3 Locality Conditions and Phases Research on locality is also an important line of inquiry in minimalist theorizing. According to the minimalist conception of locality, movement must be short; that is, it can operate only within a limited amount of structure. This shortness of movement was captured by the Subjacency Condition in terms of the limitation that a moved phrase can cross at most one cyclic node. This locality constraint on movement operations, which ‘‘in e¤ect limits the ‘memory’ available to transformational rules’’ (OWM, 111), is now captured by the Phase Impenetrability Condition (Chomsky 2000, 2004a), which essentially states that only the edge (the speciﬁer and head) of a phase (e.g., CP and vP) is accessible to syntactic operations in higher phases. This condition has the e¤ect that a computational operation (e.g., the Agree relation, as part of the movement operation) cannot look too deeply inside a lower phase. According to this conception of locality, apparently unbounded movement results from successive-cyclic movement through the edges of di¤erent phases. Another type of locality condition, based on Rizzi’s (1990) Relativized Minimality, is the Defective Intervention Constraint (Chomsky 2000, 2004a; see also Chomsky’s (1995) Minimal Link Condition). According to this principle, the probe (i.e., target of movement) always tries to enter into a matching relation (Agree) with the closest potential matching feature (i.e., dependencies must be satisﬁed in the smallest structure where they can be satisﬁed). An intervention e¤ect (e.g., a violation of the WhIsland Constraint) is obtained if the probe a matches an inactive category b that is closer to a than a matching g. In such a conﬁguration, an Agree relation between a and g is barred. 1.1.3.4 Cyclicity The principle of cyclic rule application, present in OWM, is also central to the minimalist conception of the computational system of human language. In Chomsky 1993, structure is said to be built ‘‘bottom up,’’ and this derivational process is said to be constrained by the Extension Condition: roughly speaking, application of (external) Merge or Move (i.e., internal Merge) to some structure K always leads to an extension of this structure, K*, which includes K as a proper part. As Chomsky notes, the Extension Condition yields a version of the strict cycle. Cyclicity also shows up in the context of the ‘‘timing’’ between

Lines of Inquiry into Wh-Movement

11

syntax, on the one hand, and semantics and phonology, on the other. The idea is that in the course of a derivation, syntactic objects (so-called phases) can be transferred to the interface systems for spell-out and interpretation. Once this is done, the information that was part of the transferred objects is ‘‘forgotten’’ and no longer accessible to the computational system (cf. the Phase Impenetrability Condition), which should considerably reduce the load on memory (see Chomsky 2000, 2001). All this means that information is only locally available during the derivation of a linguistic expression. 1.2

Wh-Movement Moving On

With Chomsky’s OWM as a starting point of reﬂection, the present volume aims to further develop the theory of wh-movement (and movement in general). We have organized the chapters around four themes: ¯ -chains and the copy (part I) wh-phrases and pied-piping, (part II) A theory, (part III) cyclicity and locality, and (part IV) wh-constructions. In what follows, we summarize the chapters in each part. 1.2.1 Wh-Phrases and Pied-Piping The starting point of Donati’s chapter is the assumption made in OWM that wh-movement operates on a phrase (see OWM, rule (1b): ‘‘Move whphrase’’). Donati observes that this approach is incompatible with the minimalist feature-based approach to movement, whereby only features are visible to the computational procedure, and phrasal properties of whmovement need to be derived. Donati then argues that pure feature movement should be excluded as a syntactic option under the assumption that features cannot be merged. Thus, pied-piping is always involved. She proposes that head movement (i.e., pied-piping of X) is the unmarked pied-piping operation, since it is minimal (i.e., drags along the least material). The question then arises how to derive phrasal movement (piedpiping of XP—e.g., a wh-phrase) as an option of grammar. Donati proposes to relate the choice between head movement and phrasal movement to the phrase structure status of heads and phrases, in combination with the Condition on Uniformity of Chains (Chomsky 1995) and an economy condition on Merge Copy. An implication of this approach is that there is no principled reason for wh-movement to be restricted to phrasal constituents (e.g., a wh-phrase) in any context. To illustrate this point, Donati discusses two cases of wh-head movement: free relatives and comparatives.

12

Lisa Lai-Shen Cheng and Norbert Corver

Watanabe’s chapter explores the existence of an uninterpretable formal feature whose role is to mark a constituent that must be pied-piped. This so-called pied-piper feature is needed to solve a problem that arises under a movement theory in which movement consists of three (successive) suboperations: Agree, pied-piping, and Merge. More speciﬁcally, if Agree renders the goal inactive, then both the goal category and some category that contains the goal become inaccessible for further computation in narrow syntax. Obviously, pied-piping cannot be ordered before Agree, because Agree must ﬁrst locate the goal. Watanabe argues that the introduction of the pied-piper feature solves this problem: the pied-piper feature marks the category to be copied by pied-piping. He presents different types of empirical evidence in support of this feature: (1) the change of (Old) Japanese, a language with overt phrasal wh-movement and concomitant marking of the wh-phrase by the particle ka, into a whin-situ language without ka-marking on the wh-phrase; (2) the appearance of a focus particle on an overtly moved wh-phrase in languages with optional wh-movement (e.g., Kikuyu). Agbayani’s chapter discusses the nature of speciﬁers formed by movement by looking in particular at the peculiar behavior of wh-subjects in English, which have been argued to not undergo movement (George 1980; Chomsky 1986a), and which seem to display paradoxical behavior with respect to the wh-diagnostics of OWM (e.g., they obey Subjacency but apparently do not leave a gap). Agbayani o¤ers an alternative perspective on the requirement that a speciﬁer be created by overt (wh-)movement (i.e., the Extended Projection Principle), by closely examining the behavior of wh-elements as speciﬁers. In line with Chomsky (1995), Agbayani proposes that if a feature is extracted out of a category in overt syntax for checking, then PF requirements force subsequent piedpiping of the category, unless economy blocks the pied-piping, as he argues is the case with wh-subjects in English. He argues that such an analysis of speciﬁer formation via movement is expressed naturally in a ‘‘two-chain’’ theory where bare feature movement (Move F) and subsequent category pied-piping take place in overt syntax. Evidence for such an approach to movement comes from the peculiar behavior of whsubjects in English under the Vacuous Movement Hypothesis. ¯ -Chains and the Copy Theory 1.2.2 A Rizzi’s chapter starts by discussing the functional motivations of movement (i.e., movement as a last resort) and then connects this level of anal¯ -chains—and ysis to the study of the form of chains—in particular, A

Lines of Inquiry into Wh-Movement

13

the formal principles that constrain possible chain conﬁgurations. Rizzi ¯ -chains connect two semantic interface positions: an sproposes that A selection position and a criterial position, which expresses some scopediscourse property. He further tries to determine whether and under what conditions other positions are allowed to occur in well-formed chains, besides these semantic interface positions. He provides evidence for a principle of Criterial Freezing: a phrase meeting a criterion is frozen in place, and its chain cannot extend further. Rizzi further investigates subject chains, especially the EPP property (the obligatory appearance of subjects in clauses) and the Empty Category Principle property (the impossibility of subject extraction). He reduces the EPP to a particular criterion, the Subject Criterion, and proposes that the immobility of subjects can be analyzed as a consequence of Criterial Freezing. In her chapter, Yang argues that two apparently related constructions, relative clauses and questions, do not necessarily involve the same movement operation. She argues that relative clauses in both Hindi and ¯ -movement, while questions in both Punjabi involve successive-cyclic A languages involve scrambling. Her argument is based on the observation that in Punjabi, an intermediate copy of a relative phrase can be overtly realized in Spec,CP. Furthermore, more than one copy in a relative movement chain can be pronounced. She argues that the multiple realization of relative phrases in a movement chain can only be accounted for under a theory of multiple Spell-Out (such as that proposed in Bobaljik 2002). She further shows that the relevant facts concerning multiple realization fall out from the CP-Phase Constraint, according to which only one copy of a movement chain can be overtly realized in a given CP phase. Reintges, LeSourd, and Chung begin their chapter by contending that the copy theory of movement allows for the possibility that the lowest copy is pronounced. They argue that certain types of wh-in-situ patterns instantiate ‘‘hidden’’ movement conﬁgurations that are created by whmovement in overt syntax with subsequent pronunciation of only the lowest copy. Their evidence comes from the phenomenon of whagreement: internally headed relative clauses in Passamaquoddy, an Eastern Algonquian language, and wh-in-situ interrogatives in Coptic Egyptian, an Afroasiatic language spoken in the fourth through the fourteenth centuries ad. Under the assumption that this special wh-agreement morphology records the history of wh-movement before the derivation is shipped to PF, its occurrence in certain wh-in-situ constructions reveals that in these cases as well, movement has taken place in overt syntax. The only di¤erence from overt wh-movement is that the lower rather

14

Lisa Lai-Shen Cheng and Norbert Corver

than the higher copy is pronounced. The authors also show that despite superﬁcial similarities, wh-in-situ constructions in Palauan, an Austronesian language, contrast sharply with Passamaquoddy relatives and Coptic wh-in-situ interrogatives. While wh-agreement is found in Palauan questions formed with overt wh-movement, this agreement is not triggered by wh-in-situ constructions, which suggests that no narrow syntactic movement is involved in such constructions. The authors propose that in those cases, the scope of the interrogative phrase results from unselective binding. 1.2.3 Cyclicity and Locality Lasnik’s chapter examines the motivations and consequences of di¤erent conceptions of the cyclic principle that have appeared in generative syntactic theorizing since the mid-1960s. Lasnik starts with the relation between the cycle and generalized transformations, and with how the latter have been resurrected in minimalism. He then discusses the necessity of the principle of strict cyclicity within minimalism. In particular, he points out that (1) empirical arguments for cyclicity, as Freidin (1978) notes, are based on derivations that can be independently excluded under plausible assumptions; and (2) the theoretical argument based on preservation of existing structure (Chomsky 2000) cannot be sustained under a settheoretic conception of phrase structure. Lasnik further examines arguments for and against successive cyclicity for A-movement, concluding that A-movement is indeed successive cyclic. Ishii’s chapter investigates the weakening of intervention e¤ects. In particular, he shows that indirect questions do not constitute wh-islands for extraction when the following two conditions are met: (1) indirect questions are inﬁnitival or subjunctive, and (2) extracted wh-phrases are Dlinked. He argues that this hitherto unexplained anomaly concerning the Wh-Island Constraint is given a principled minimalist account by an operation of covert restructuring and Chomsky’s (2000, 2001, 2004a) Phase Impenetrability Condition (PIC). He shows that contrary to the widespread belief that the Wh-Island Constraint can be subsumed under the Defective Intervention Constraint, we still need a traditional Subjacencytype account based on the notion of the phase to fully accommodate the wh-island e¤ect. 1.2.4 Wh-Constructions Grohmann’s chapter is concerned with monoclausal question constructions from German that involve two wh-phrases between which a topical-

Lines of Inquiry into Wh-Movement

15

ized constituent may or may not be sandwiched. He argues that what may look like a multiple wh-question involving a fronted topic is in fact a case of multiple topicalization and does not serve as a proper information question. Grohmann starts by claiming that the so-called intervention e¤ect, also dubbed the ‘‘Beck e¤ect,’’ does not exist in German multiple wh-questions. That is, any attempt to capture the Beck e¤ect in terms of some notion of intervention undergenerates and consequently fails. Grohmann then develops an alternative that capitalizes on a property other than quantiﬁcation that illegitimate interveners share: they are all nontopicalizable. Thus, wh-movement is permitted only across a fronted topic in German. Building on this topicalizability-based account of intervention e¤ects, Grohmann proposes that multiple wh-questions involving a fronted topic are derived by movement to a left-peripheral topic position of both wh-phrases (as well as the intervening topic). In other words, these multiple wh-interrogative constructions are actually multiple wh-topic constructions. Finally, Grohmann further connects these multiple wh-topic constructions to instances of clitic-doubling whphrases in (colloquial) Bulgarian. Sura´nyi’s chapter deals with multiple wh-constructions in Hungarian. Sura´nyi argues that Hungarian exhibits (at least) three syntactic patterns of multiple wh-questions, which he relates in a principled way to di¤erent focus structures. Modulo distinctions in focus structure, he relates the three multiple wh-constructions to three distinct sets of answerhood conditions they are matched with, arguing that multiple (overt or covert) movement to the same projection results in a pair-list interpretation, while otherwise only a single-pair reading is available. In closing, Sura´nyi o¤ers a typological perspective on languages that permit multiple wh-fronting and languages that do not. He shows that the account of answerhood conditions he has proposed on the basis of Hungarian can be maintained in the context of a more general typology of interpretations available in multiple wh-constructions. Ko’s chapter discusses the structural position of the reason whadverbial ‘why’ in three typical wh-in-situ languages: Korean, Japanese, and Mandarin Chinese. Ko proposes that in these languages, ‘why’ is externally merged in Spec,CP as a CP modiﬁer and that this high baseposition of ‘why’ plays a central role in determining its unique syntax in wh-in-situ languages. Evidence for base generation in Spec,CP comes from the interaction between ‘why’ and scope-bearing elements in Korean and Japanese. Additional support comes from acquisition data: children learning Korean make systematic distinctions, signaled by placement,

16

Lisa Lai-Shen Cheng and Norbert Corver

between the wh-expression way ‘why’ and other wh-adjuncts such as eti ‘where’. Ko further explores the implications of the ﬁne structure of CP for the syntax of ‘why’ in wh-in-situ languages. Examining the antisuperiority e¤ect, she argues that there are two distinct C heads in wh-in-situ languages for licensing wh-interrogatives. In particular, ‘why’ is licensed by a lower C (CInt ), whereas the other wh-phrases are licensed by a higher C (CFoc ). This approach, in conjunction with other independent proper¯ties of wh-in-situ languages—in particular, the presence or absence of A scrambling—explains the crosslinguistic variations between Korean and Japanese, on the one hand, and Chinese, on the other hand, regarding multiple wh-questions. 1.3

Conclusion

In various respects, Chomsky’s paper ‘‘On Wh-Movement’’ is a landmark in generative-syntactic theorizing about wh-movement. First, it represents a crucial step in the development from a construction-oriented and language-speciﬁc approach to human grammar to an approach in terms of general rules and general principles. Second, it illustrates clearly how in-depth linguistic theorizing and detailed empirical analysis of linguistic phenomena in a particular language can go together. Finally, it shows the great beneﬁt practicing syntacticians derive from the availability of good diagnostic tools. In this introduction, we have shown that OWM was also an important starting point for further theorizing on wh-movement. Many of the ‘‘wh-issues’’ that are central in OWM recur in later conceptions of the theory of grammar. At the same time, it is clear that new thoughts on wh-movement are still developing in current syntactic research. Notes 1. In OWM (p. 73, (3)): ‘‘Cycle: transformational rules, e.g., [Move wh-phrase and Move NP], meet the condition of the (strict) cycle; the subjacency condition is a property of cyclic rules, i.e., part of the deﬁnition of the cycle.’’ 2. In OWM, NP, S, and S 0 are taken to be the cyclic nodes (see further discussion in section 1.1.2). 3. As a matter of fact, even the universal principles of grammar that were postulated often had a construction-speciﬁc ﬂavor: the Tensed-S Condition, the Speciﬁed Subject Condition, and island conditions such as the Complex NP Constraint. 4. A ﬁfth diagnostic that is not on Chomsky’s list but is present in the discussion of topicalization and used as a wh-diagnostic thereafter is this: ‘‘It creates whislands.’’

Lines of Inquiry into Wh-Movement

17

5. As noted by Bresnan (1972), many dialects of American English allow the presence of wh-words in comparatives, as in John is taller than what Mary is. 6. In later work, wh-diagnostics are also used to identify null operator movement. For example, Chomsky (1986a, 55) shows that parasitic gap constructions display island e¤ects, obeying for instance the Wh-Island Constraint and the Complex NP Constraint. 7. There is brief mention of getting rid of S 0 . Also, in the discussion about topicalization S 00 is mentioned as a potential cyclic node. 8. As Chomsky stated in a recent interview regarding minimalism, ‘‘The question we’re now thinking about is the question ‘why?’ Why these principles and not some other principles? These ‘why’ questions, that’s the Minimalist Program’’ (2004b, 151). References Bach, Emmon, and Laurence Horn. 1976. Remarks on ‘‘Conditions on transformations.’’ Linguistic Inquiry 7, 265–299. Bobaljik, Jonathan David. 2002. A-chains at the PF-interface: Copies and ‘‘covert’’ movement. Natural Language and Linguistic Theory 20, 197–267. Bresnan, Joan. 1972. Theory of complementation in English syntax. Doctoral dissertation, MIT. Published, New York: Garland, 1979. Bresnan, Joan. 1975. Comparative deletion and constraints on transformations. Linguistic Analysis 1, 25–74. Bresnan, Joan. 1976a. Evidence for a theory of unbounded transformations. Linguistic Analysis 2, 353–399. Bresnan, Joan. 1976b. On the form and functioning of transformations. Linguistic Inquiry 7, 3–40. Bresnan, Joan. 1977. Variables in the theory of transformations. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 157–196. New York: Academic Press. Chomsky, Noam. 1955. The logical structure of linguistic theory. Ms., Harvard University. Revised 1956 version published in part by Plenum Press, New York, 1975; University of Chicago Press, Chicago, 1985. Chomsky, Noam. 1973. Conditions on transformations. In Stephen Anderson and Paul Kiparsky, eds., A festschrift for Morris Halle, 232–286. New York: Holt, Rinehart and Winston. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1980. On binding. Linguistic Inquiry 11, 1–46. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. Chomsky, Noam. 1986a. Barriers. Cambridge, Mass.: MIT Press.

18

Lisa Lai-Shen Cheng and Norbert Corver

Chomsky, Noam. 1986b. Knowledge of language: Its nature, origin, and use. New York: Praeger. Chomsky, Noam. 1988. Language and problems of knowledge: The Managua lectures. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2004a. Beyond explanatory adequacy. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Chomsky, Noam. 2004b. The generative enterprise revisited: Discussions with Riny Huybregts, Henk van Riemsdijk, Naoki Fukui, and Mihoko Zushi. Berlin: Mouton de Gruyter. Cinque, Guglielmo. 1982. On the theory of relative clauses and markedness. The Linguistic Review 1, 247–294. Cinque, Guglielmo. 1990. Types of A¯-dependencies. Cambridge, Mass.: MIT Press. Freidin, Robert. 1978. Cyclicity and the theory of grammar. Linguistic Inquiry 9, 519–549. George, Leland. 1980. Analogical generalization in natural language syntax. Doctoral dissertation, MIT. Groos, Anneke, and Henk van Riemsdijk. 1981. Matching e¤ects in free relatives: A parameter of core grammar. In Adriana Belletti, Luciana Brandi, and Luigi Rizzi, eds., Theory of markedness in generative grammar, 171–216. Pisa: Scuola Normale Superiore. Riemsdijk, Henk van. 1978. On the diagnosis of wh movement. In Samuel Jay Keyser, ed., Recent transformational studies in European languages, 189–206. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1996. Residual verb second and the Wh-Criterion. In Adriana Belletti and Luigi Rizzi, eds., Parameters and functional heads, 63–90. Oxford: Oxford University Press. Ross, John Robert. 1967. Constraints on variables in syntax. Doctoral dissertation, MIT. Published as Inﬁnite syntax! Norwood, N.J.: Ablex, 1986.

PART I Wh-Phrases and Pied-Piping

Chapter 2 On Wh-Head Movement

2.1

Caterina Donati

Introduction

The merit of ‘‘On Wh-Movement’’ (Chomsky 1977) was that it explained in a principled way and uniﬁed under a single theory—movement— a number of properties previously unobserved, or unrelated, or unaccounted for. I will start by considering with a critical and ‘‘minimalist’’ eye another property of wh-movement that was simply assumed as a primitive in the theory, and left unexplained: the fact that wh-movement operates on a phrase. More generally, the theory of movement as it developed in the 1970s and 1980s took for granted the existence of two conﬁgurationally di¤erent movements, head movement and phrasal movement, which would never overlap and never compete in their domain of application, in what can really be seen as a remnant of a ‘‘construction-driven’’ approach to syntax. This approach can no longer be maintained if the minimalist attitude is to be taken seriously. More precisely, an approach assuming as a primitive the existence of two kinds of movement is incompatible with a feature-driven approach. As is well known, in Chomsky 1995, 2000, 2001, movement is reanalyzed as a complex operation crucially triggered and governed by features: an uninterpretable feature a, the probe, searches locally for a matching feature b, the goal (Agree), a copy of which is merged in a local conﬁguration for checking (Merge Copy or Internal Merge). What the computational procedure really sees is a feature, and the phrasal properties of (standard) wh-movement need to be derived. The chapter is organized as follows. After ﬁrst excluding simple feature movement as a syntactic option in section 2.2, in section 2.3 I address the issue of the status of head movement, reviewing with a critical eye Chomsky’s recent attempt to exclude it from narrow syntax. I conclude that

22

Caterina Donati

head movement does exist as a syntactic option and that it is indeed the unmarked pied-piping operation, being minimal. Next, I systematically review the conditions on movement spelled out in the theory starting from ‘‘On Wh-Movement,’’ searching for a property of head movement that might make it unavailable in standard wh-movement contexts, hence triggering the less minimal, more costly operation of phrasal movement (section 2.4). Finding no such property, I propose in section 2.5 that the choice between phrase and head as the target of the movement operation is simply due to the phrase structure status of heads and phrases, together with the Condition on Uniformity of Chains (Chomsky 1995). In section 2.6, I provide two case studies, free relatives and comparatives, showing that wh-movement does not necessarily involve phrases and that the fact that it does in the classical case is just a consequence of the deﬁnition of what a head/phrase is, together with the interpretation of these constructions. Finally, in section 2.7 I examine some cases that look like residuals of speciﬁc constraints on head movement, showing that they are e¤ects of a phonological requirement that can interfere with any movement operation. In section 2.8, I o¤er some conclusions. 2.2

No Feature Movement

A preliminary question, explicitly addressed in Chomsky 1995, is why the complex operation described above as movement (Agree þ Internal Merge) cannot simply occur at the level of features, and some kind of (generalized) pied-piping is necessarily involved—in other words, why features cannot move. Chomsky’s ﬁrst tentative answer is that features cannot move owing to a phonological constraint: For the most part—perhaps completely—it is properties of the phonological component that require such pied-piping. Isolated features and other scattered parts of words may not be subject to its rules. (Chomsky 1995, 262–263)

The problem with this proposal is that it implies that post-Spell-Out movement, being free from any phonological constraint, should display no pied-piping. There is good evidence, however, of the existence of pied-piping—that is, of phrasal movement—at LF. A simple case providing evidence for covert phrasal movement is antecedent-contained deletion: in (1), where the deleted VP is contained within its antecedent, the only way to recover ellipsis without descending

On Wh-Head Movement

23

into inﬁnite regress is to allow quantiﬁer raising of the phrase every city Bill does.1 (1) John will visit [every city Bill does] [every city Bill does] John will visit [t] A more convincing alternative is to identify the constraint banning feature movement with a property of Merge. When it applies to the numeration, Merge indeed is restricted to operate on lexical items, on bundles of features. It is thus natural and minimal to assume that this restriction holds as well when Merge is part of a movement operation. Let us then assume that there is a single operation Merge in the syntax, which applies freely and is constrained to operate on words (Matushansky 2006).2 Assuming that features cannot move (or better, be merged), some sort of pied-piping is required. The minimal amount of material a feature should pied-pipe to form a mergeable object is a word, that is, a head. This derives head movement. What about XP-movement? Being antiminimal, why should it exist at all? 2.3

The Status of Head Movement

The easiest way to solve this puzzle is to proceed as we did for feature movement, banning it from syntax. This is what Chomsky (2001) does: he assumes that head movement is not a syntactic operation but part of the ‘‘phonological branch of the computation.’’ If this is true, then the minimal amount of material a feature can pied-pipe becomes a phrase, and the phrasal properties of (standard) wh-movement end up being derived for free. This move is untenable, however, for several reasons: it is based on wrong assumptions, both theoretical and empirical; it is incompatible with other aspects of Chomsky’s theory (e.g., phases); and it brings undesirable consequences. First, this proposal is based on a wrong theoretical assumption, namely, that head movement violates crucial syntactic requirements—and in particular, that head movement does not extend the target (hence being countercyclic). This is true for the standard head adjunction conﬁguration given in (2).

24

Caterina Donati

(2)

The traditional reason for this conﬁguration is related to the standard cases of head movement usually considered in the literature: in V-to-T movement, or in V-to-C movement, the two heads conﬂate and behave like a single constituent, hence the assumption that they form a sort of ‘‘derived lexical item’’ represented in the head adjunction conﬁguration in (2). But this is not the only conﬁguration head movement can in principle produce. Suppose we have a head X endowed with an uninterpretable feature a that needs to be deleted. In principle, nothing prevents a copy of a head Y endowed with the matching feature b from merging with the root of the structure, as in (3). (3)

The conﬁguration in (3) is obtained by merging a new item to the root of the tree, hence complying with the Extension Condition. Given (3), the head conﬂation e¤ect correlated in many cases with this conﬁguration can be the result of an independent process, perhaps phonological, which in principle has nothing to do with head movement: this process—call it ‘‘a‰xation’’—is something that can happen to two adjacent heads independently of how and why they ended up being adjacent. Separating head movement from a‰xation is a good idea in light of several empirical facts: ﬁrst, there are cases of a‰xation that do not correlate with head movement, but with phrasal movement (see below); second, there are cases where head movement does not yield any e¤ect of this kind (section 2.6); third, there are cases where a‰xation does not involve moved heads, but merged ones (section 2.7). As an illustration of the ﬁrst cases, consider (4) as a possible derivation for a clitic construction in Italian.

On Wh-Head Movement

25

(4)

Clitic pronouns in Romance have always been problematic given standard assumptions about phrasal and head movement: clitics are clearly maximal projections in the position where they are generated, but they end up being a‰xed to an inﬂectional head. This tension is at the root of standard and inﬂuential analyses such as Kayne’s (1989), where the clitic starts out as a phrase but moves and adjoins as a head, changing its phrase structure status in the course of the derivation (violating a condition like that on the uniformity of chains; Chomsky 1995). The derivation in (4) is by far more minimal: here the phrase structure status of the clitic never changes during the derivation, and the clitic moves as a phrase. The D-to-T incorporation is an independent process akin to phonology.3 Summarizing so far, nothing about head movement prevents it from being cyclic, and assuming the existence of cyclic head movement appears to be empirically convenient. But also inherent in banning head movement from syntax is a troublesome implication, which appears to be too strong: being phonological, head movement should never feed LF, hence never have any interpretive e¤ect. This may well be true if we consider movement of predicative heads, which need to reconstruct for interpretability (Heim and Kratzer 1998). But even in these cases, the issue depends crucially on the status we assign to the notion of extended projection (Grimshaw 1991): the extended projections of V, or of N, are typically deﬁned as head movement domains, and they do look like interpretable objects as a whole.4 On the other hand, the LF e¤ects of head movement become simply unquestionable once we consider movement of quantiﬁcational heads, which do acquire their scope through this mechanism; we will consider a case of this type in section 2.6. Returning once again to Chomsky’s proposal, there is also a technical problem with its implementation if it is combined with other aspects of

26

Caterina Donati

the more recent framework: the idea that head movement is phonological clashes with the details of the derivation-by-phases approach to computation. Suppose the system is building a simple structure like (5). (5)

Suppose moreover that V 0 must move to T 0 to check some uninterpretable feature. At the level of the ﬁrst strong phase, vP, the sister of v (VP) is sent to Spell-Out. But the edge of the current phase, v 0 , is not yet accessible to phonology. In other words, the launching site and the landing site of this head movement operation are not accessible to phonology in the same phase. A solution could be that V-to-T movement happens when the next phase (CP) is sent to PF; but then the derivation must deal with two phases at once, hence raising the computational burden, contrary to the whole spirit of phase theory (Matushansky 2006). Finally, pushing head movement onto a phonological branch of the computation has a broader undesirable consequence for the theory: it complicates the grammar by positing a ‘‘second syntax’’ in the PF branch. It also raises at least two questions. For one thing, if both phrasal movement and head movement are driven by features, it seems at least strange to locate them on two di¤erent computations. For another, consider that Chomsky (2000) assumes that PF also involves some kind of phrasal movement (e.g., extraposition; see also Lebeaux 1990), so that the problem of the duplicity of movement is reduplicated within this second syntax. The conclusion is that there is no principled way to exclude head movement from narrow syntax, and a simple stipulation opens a number of problems. Head movement is indeed an operation available to narrow syntax; moreover, it is the default movement option, being the minimal pied-piping strategy. The next step, then, is to ﬁnd some property of head movement that makes it unavailable in standard wh-movement contexts, hence triggering the less minimal, more costly operation: phrasal movement.

On Wh-Head Movement

2.4

27

Head Movement versus Phrasal Movement

We are now looking for the property that triggers phrasal movement instead of head movement in the classical wh-constructions. The ﬁrst place to look is in the deﬁnition of movement, among the conditions that constrain its various components. Let us review what these constraints on movement look like, given minimalist assumptions and the feature approach, to see if there are any that might be responsible for the choice between the two movements. The traditional account relies on conditions concerning Agree, the search procedure—call them locality or closeness. Head movement, the default option, is limited by the Head Movement Constraint (HMC; Travis 1984). Whenever the goal is not (a feature of ) the head of the complement of the probe, the nondefault option is chosen, and the goal piedpipes its phrase; the HMC has recently been reframed in these terms by Pesetsky and Torrego (2001, 363), as follows: (6) Head Movement Generalization Suppose a head H attracts a feature of XP as part of a movement operation. a. If XP is the complement of H, copy the head of XP into the local domain of H. b. Otherwise, copy XP into the local domain of H. This is not a viable solution for both theoretical and empirical reasons. On the theoretical side, there is no room in a feature approach for such a constraint. Notice, interestingly, that Pesetsky and Torrego arrive at (6) by assuming that phrasal movement is the default option, blocked in case (a) because it results in remerging. But the system cannot work this way. If movement is triggered by a search procedure of a feature (Agree), locality must be deﬁned in terms of features with no reference to the amount of material that gets moved with the feature. Such is the case with the Minimal Link Condition (7) and its more recent version (8). (7) Minimal Link Condition K attracts a only if there is no b, b closer to K than a, such that K attracts b. (Chomsky 1995, 311) (8) Locality conditions yield an intervention e¤ect if probe a matches inactive b that is closer to a than matching g, barring Agree (a, g). (Chomsky 2001, 4)

28

Caterina Donati

In other words, locality is a condition on the search procedure, which is a component of movement deﬁned at the feature level. At that level, the distinction between phrase and head plays no role and indeed does not exist at all. Therefore, the criterion we seek for choosing between head movement and phrasal movement cannot be found among the locality conditions.5 On the empirical side, let us look at the well-known though controversial ‘‘long head movement’’ phenomena recorded in the literature (e.g., Lema and Rivero 1990; Manzini 1994; Roberts 1994; Carnie 1995; Borsley, Rivero, and Stephens 1996). For example, the simple Breton sentence in (9) appears to instantiate the derivation in (10), where the verb raises to a left-periphery head, skipping the head I—hence violating the HMC, or the generalization (6). (9) Lennet en deus Yann al levr read 3m.sg has Yann the book ‘Yann has read the book.’ (Roberts 1998)

(Breton)

(10)

The same derivation yields an ungrammatical result if the sentence contains a di¤erent kind of intervener, such as negation. (11) *Lennet n’en deus ket Tom al levr read neg 3m.sg has neg Tom the book The explanation proposed by Roberts (1994) and others is that Relativized Minimality–type constraints such as (7) and (8) apply ‘‘across the board,’’ to any movement dependency, no matter whether it involves a

On Wh-Head Movement

29

head or a phrase. In other words, what counts for locality is the class of interveners and the class of attractees, not the amount of moved material. More data of this sort are given in (12) through (14). (12) Citao sam knjigu read I-have book ‘I have read the book.’ ´ avar 1994) (Wilder and C

(Croatian)

(13) Procel sum knigata read I-have the-book ‘I have read the book.’ (Borsley, Rivero, and Stephens 1996)

(Bulgarian)

(14) Dar-te he un exemplo give-you I-will an example ‘I will give you an example.’ (Borsley, Rivero, and Stephens 1996)

(Old Spanish)

The interpretation of such data in terms of long head movement is controversial and still being discussed, but the point is that these are exactly the data we predict given a feature approach to movement: if locality is a condition on the search procedure component of movement, then it is to be deﬁned in terms of features, or classes of features. We thus predict that the same kind of constraints will govern both head movement and phrasal movement. But if locality is not responsible for the choice between head and phrasal movement, we must look at other movement constraints.6 A second class of constraints on movement relates to the goal, which must be active (Chomsky 2000) and not embedded in an island. But these constraints do not seem to be relevant for present purposes, since they apply to any movement operation. Finally, a third class of constraints is conceivable, relating to the other operation involved in Move, namely, Internal Merge. The Condition on Uniformity of Chains (CUC) can be deﬁned as such a constraint. (15) The Condition on Uniformity of Chains A chain is uniform with regard to phrase structure status. (Chomsky 1995, 253) (I will show in the next section that this constraint is only apparently irrelevant for the head/phrase distinction we are seeking.) Also conceivable is an economy condition on Internal Merge, for which what is to be merged is the minimal amount of material necessary for convergence: (16).

30

Caterina Donati

(16) Merge just enough material for convergence. This is indeed the notion of minimality I referred to earlier, when I said that the minimal amount of pied-piped material is a head, giving head movement as the unmarked option. The only conclusion we can draw from what we have seen so far is that there is no di¤erence whatsoever between head movement and phrasal movement, except that the former is more minimal than the latter. To ﬁnd a way to derive phrasal movement as an option and its selection in standard wh-constructions, we must turn to some other component of the grammar. 2.5

Head versus Phrase

If the di¤erence between head and phrasal movement does not reside in the movement module, the only alternative is to rely on phrase structure theory. There is indeed a primitive di¤erence between heads and phrases, which we will capitalize on to derive the complementarity we need: in a nutshell, a head projects; a phrase is a projection. Together with the independent conditions on Internal Merge spelled out earlier (the CUC (15) and the economy condition (16)), this primitive di¤erence is enough to derive in a principled way (1) the complementarity between head movement and phrasal movement, (2) the choice of phrasal movement in standard wh-constructions, and (3) the existence of minimally di¤erent whconstructions involving head movement. But let us proceed step by step, trying to be really minimal. Suppose a probe a on a head A at the root attracts a goal b. Given that a feature cannot be merged, some extra material needs to be merged (call it B), and the o¤ending feature is deleted (17). The operation Merge (A, B) is asymmetric in essence, so that given the conﬁguration in (17) one of two things must happen: either A or B must project, yielding the two conﬁgurations (18) or (19). The element that projects is the head of the resulting phrase. (17)

(18)

(19)

Let us combine these trivial assumptions with what we know about movement. Suppose a moved item and its copy must be the same with re-

On Wh-Head Movement

31

spect to their phrase structure status (the CUC). This means that when a head, which by deﬁnition projects, gets moved, it must project; but when a phrase, itself by deﬁnition a projection, gets moved, it does not project. As a result, whenever a feature moves as a head, all the features it is associated with (and notably the categorial feature) project. Therefore, head movement changes the feature composition of the target. When the very same feature moves as a phrase, this does not happen, and the target remains unchanged. The economy condition on Internal Merge ensures that the two movements never overlap: if the grammar always copies enough material for convergence, it will select head movement unless convergence at LF (interpretation) chooses di¤erently.7 To see how this account works, let us look at (standard) whconstructions. In interrogatives, the complementizer head selecting the structure contains a feature, call it [wh], that needs to be checked, acting as a probe. The goal corresponds to a wh-element embedded in the clause, (a copy of ) which needs to be merged in a local conﬁguration with the probe. Since features cannot be merged (see section 2.2), the minimal option is for the wh-feature attracted by C to pied-pipe the wh-word alone (head movement). But this minimal option does not yield a convergent derivation: moving the wh-feature as a head means projecting all the features associated with it, and notably its categorial feature (D). This would turn the interrogative clause into a complex DP as in (20), which is not interpretable as an interrogative clause at the interface.8 This is why the more costly derivation (21) is selected. (20)

(21)

Summarizing so far, there is only one operation, Move, which is triggered by a feature and deﬁned at that level, and which merges just enough material for convergence. Then, convergence at LF decides whether ‘‘enough material’’ is a head (which retains its projection property throughout the derivation, given the CUC) or a phrase (which remains a projection throughout the derivation). In standard wh-constructions, such as interrogatives, LF convergence selects the less minimal option, that of moving the entire phrase, preserving the simple CP categorial status of the clause.

32

Caterina Donati

2.6

Free Relatives and Comparatives

Good evidence that the phrasal status of wh-movement is only a byproduct of the environments where it occurs (interrogative CPs) would then be an instance of wh-movement embedded in a di¤erent environment, notably in a context compatible with DP selection and showing no phrasal pied-piping.9 This is exactly what we ﬁnd with free relatives. Consider the minimal pairs in (22a–b) and (23a–b) in Italian and English, respectively. (22) a. *Ho mangiato quanti biscotti hai preparato [t] have-1sg eaten how-many cookies have-2sg prepared ‘*I have eaten what cookies you have prepared.’ b. Mi chiedo quanti biscotti hai preparato [t] me wonder how-many cookies have-2sg prepared ‘I wonder how many cookies you have prepared.’ c. Ho mangiato quanto hai preparato [t] have-1sg eaten what have-2sg prepared ‘I have eaten what you have prepared.’ (23) a. *I shall visit [what town] you will visit [t] b. I wonder [what town] you will visit [t] c. I shall visit [what] you will visit [t] At ﬁrst sight, free relatives (a,c) and interrogatives (b) di¤er in that the wh-element needs to be bare in the former but not in the latter.10 This anti-pied-piping restriction correlates with a di¤erent interpretation of the two constructions: as is well known, free relatives are complex nominal structures, while interrogatives are simple clauses. This is shown straightforwardly by the selection possibilities of the two structures: free relatives like (22a,c) or (23a,c) are compatible with a DP-selecting verb like mangiare (24a) or visit (25a); interrogatives like (22b) or (23b) are compatible only with non-DP-selecting verbs like chiedersi (24b) or wonder (25b). Mangero` i biscotti I-will-eat the cookies b. *Mi chiedo i biscotti I wonder the cookies

(24) a.

(25) a. I shall visit Rome b. *I wonder Rome This correlation can be expressed through a generalization like (26).

On Wh-Head Movement

33

(26) A simple wh-structure excludes pied-piping exactly in those cases in which it occurs in a nominal position. Given what we already know about heads and phrases, we can suggest a principled explanation for (26). What happens in free relatives is that the wh-element moves as a head. This means that, besides checking the whfeature on C as usual, it endows the clause with the D-feature required for interpretation, as in (20). Consider (27) for a more detailed comparison of free relatives (27a), full relatives (27b), and interrogatives (27c). The basic structural di¤erences are minimal, but they correlate with important interpretive and syntactic di¤erences. Relative clauses are characterized by the fact that they are complex nominal clauses (a DP embedding a CP: Kayne 1994; Bianchi 1999), the di¤erence between free and full relatives being that D is moved in the former and merged in the latter. Interrogatives and free relatives involve the same uninterpretable feature [wh] on C, which is checked through phrasal movement in the former and through head movement in the latter. The crossed arrows in (27a) and (27c) correspond to instances of movement violating the CUC: a phrase projecting (27a);11 a head not projecting (27c). (27) a.

b.

c.

A second piece of evidence that we are on the right track comes from a di¤erent instance of the generalization in (26), represented by comparative clauses. As we will see, comparative clauses are wh-constructions displaying all the typical properties associated with wh-movement, together with clear complex nominal properties. We thus predict that comparatives should disallow phrasal pied-piping and involve wh-head movement. This prediction is conﬁrmed, as the following review of the comparative construction’s properties will show. First, comparatives are wh-constructions, displaying all the typical properties associated with movement since Chomsky 1977. For example, they are sensitive to island e¤ects, as (28) and (29) show for Italian and

34

Caterina Donati

English. In Italian, the comparative is moreover introduced by an overt wh-element. (28) a. Complex Noun Phrase Constraint *I ate more cookies than I met a man who ate [t] b. Wh-Island Constraint *I ate more cookies than I wonder who ate [t] (29) a. Complex Noun Phrase Constraint *Ho mangiato piu` biscotti have-1sg eaten more cookies un uomo che ha mangiato [t] a man who has eaten b. Wh-Island Constraint *Ho mangiato piu` biscotti have-1sg eaten more cookies ne abbia mangiati [t] of-them has eaten

di quanti ho incontrato of wh have-1sg met

di quanti mi chiedo chi of wh me wonder who

Second, comparatives display clear complex nominal properties, as has been observed primarily on the semantic side. In particular, they give rise to the kind of scope ambiguities typically associated with nominals, as the following classic example from Russell 1905 shows: (30) I thought your yacht was larger than it is a. [[than it is] [I thought your yacht was larger]] b. I thought [[than it is] [your yacht was larger]]

(consistent) (inconsistent)

The easiest and most standard way to account for the ambiguity of (30) is to derive it from the scope possibilities of the comparative clause, so as to include or exclude the belief verb from the comparison, as informally represented in (30a–b) (von Stechow 1984). The complex nominal character of comparatives is also visible in their interpretation, which is that of amount or degree descriptions such as (free) relatives; in this way, they di¤er from corresponding whinterrogatives, which are simple amount or degree predicates (e.g., Heim 1985; Donati 2000). Again, this is represented informally in (31). (31) Mary ate more cookies than Paul did by [[y > ix [Paul ate x-many cookies]5[Mary ate y-many cookies]] On the syntactic side, comparatives are strong islands for extraction, just like all complex nominal clauses, but unlike standard whconstructions.

On Wh-Head Movement

35

(32) a. *When do you eat more cookies in the morning than Paul does [t] [t]? b. *What do you eat the soup more quickly than Paul does [t] [t]? Given that comparative clauses are wh-structures and they display these nominal properties, the prediction is clear: comparatives should involve wh-head movement. This is exactly what the data in (33) and (34) show, from Romanian and Bulgarian, respectively. (33) a. *Maria e cu mult mai des¸teapta˘ decıˆt de frumoasa˘ e Maria is with much more clever of-wh of beautiful is Zamﬁra [t] Zamﬁra ‘*Maria is much more clever than beautiful Zamﬁra is.’ b. Maria e cu mult mai des¸teapta˘ decıˆt e Zamﬁra [t] de Maria is with much more clever of-wh is Zamﬁra of frumoasa˘ beautiful ‘Maria is much more clever than Zamﬁra is beautiful.’ c. Cıˆt de frumoasa˘ e Zamﬁra [t]? wh of beautiful is Zamﬁra ‘How beautiful is Zamﬁra?’ (Grosu 1994)12 (34) a. *Ivan izpi povece vino ot-kolkoto bira Maria izpi [t] Ivan drank more wine of-how-much beer Maria drank ‘*Ivan drank more wine than beer Maria drank.’ b. Ivan izpi povece vino ot-kolkoto Maria izpi [t] bira Ivan drank more wine of-how-much Maria drank beer ‘Ivan drank more wine than Maria drank beer.’ c. Kolko bira izpi Maria [t]? how-much beer drank Maria ‘How much beer did Maria drink?’ d. *Kolko izpi Maria [t] bira? how-much drank Maria beer ‘*How much did Maria drink beer?’ (Izvorski 1995) (33) and (34) show that the very same wh-element (cıˆt in Romanian, kolko in Bulgarian13) moves as a head in comparatives (a–b), but as a phrase in interrogatives (c). This provides good evidence that the wh-element

36

Caterina Donati

involved in the embedded comparative is a Q head, which endows the clause with the Q-feature required for interpretation. The same evidence holds also in English and Italian.14 In English, however, the facts are partially obscured since the wh-element is null and the contrast with interrogatives is not minimal. (35) a. *Mary ate more cookies than candies she ate b. Mary ate more cookies than she ate candies c. How many candies did she eat? d. *How many did she eat candies? In Italian, as we will see in the next section, the movement pattern illustrated so clearly by the Romanian and Bulgarian data is complicated by a gap in the paradigm that requires an explanation. All the data discussed in this section provide strong evidence for the following analysis: Comparatives and free relatives, on the one hand, and interrogatives, on the other hand, display a minimal syntactic contrast that correlates with a semantic di¤erence. While all involve whmovement, triggered by the same feature and constrained by the same restrictions, free relatives and comparatives move a bare head, which changes the CP conﬁguration domain into a DP/QP category, turning the clause into a complex nominal; interrogatives move a phrase, which leaves the CP conﬁguration domain unchanged as a simple (interrogative) clause. This is what we expect if movement is an operation merging just enough material for convergence, and if heads and phrases di¤er only in their projection ability. 2.7

The Freezing E¤ect of Agreement

As I mentioned, Italian comparatives require a ﬁnal glance. In Italian, while there is clear evidence that the wh-element quanto moves as a head (see (36)), the situation is complicated by some gaps in the paradigm: in particular, quanto never strands its complement (37a), except when it is cliticized with ne (37c) or when it shows no agreement with its adjectival complement (37b). (36) a. *Maria ha mangiato piu` biscotti di quante caramelle Paolo Maria has eaten more cookies of wh-f.pl candies Paolo abbia mangiato [t] has eaten ‘*Maria has eaten more cookies than candy Paolo ate.’

On Wh-Head Movement

37

b. *Maria e` piu` intelligente di quanto bella sia [t] Maria is more intelligent of wh beautiful is ‘*Maria is more intelligent than beautiful she is.’ (37) a. *Maria ha mangiato piu` biscotti di quante Paolo abbia Maria has eaten more cookies of wh-f.pl Paolo has mangiato [t] caramelle eaten candies ‘Maria has eaten more cookies than Paolo has eaten candies.’ b. Maria e` piu` intelligente di quanto sia [t] bella Maria is more intelligent of wh is beautiful ‘Maria is more intelligent than she is beautiful.’ c. Maria ha mangiato piu` biscotti di quanti ne abbia Maria has eaten more cookies of wh-m.pl of-them has mangiati [t] Paolo eaten Paolo ‘Maria has eaten more cookies than Paolo has.’ If we compare these data with those of Bulgarian, the generalization seems to be that agreement within the QP has a sort of a freezing e¤ect, blocking head movement of the Q alone and the corresponding stranding of the remnant of the phrase. The same e¤ect seems at work in the free relative cases discussed above: both in English and in Italian (38), the D head can never strand its nominal complement.15 (38) a. *Mangero` quanti vorrai [t] biscotti I-will-eat how-many you-will-want cookies b. *I shall visit what you will visit [t] town This kind of gap in the paradigm is very important for the hypothesis under discussion here, because it seems like a residue of a constructionspeciﬁc property of head movement, which would contradict our attempt at uniﬁcation. We will see, however, that this constraint has nothing to do with head movement per se and that the same kind of e¤ect occurs with phrasal movement. We will therefore end up conﬁrming the claim that movement is a unitary operation. But let us proceed step by step. Consider in (39) the (simpliﬁed) structure of a comparative construction containing an agreeing QP, like quante caramelle in (37a).

38

Caterina Donati

(39)

Suppose Q-N agreement means Q has an uninterpretable feature that needs to be checked by a feature of N. Head movement is not triggered here because it is useless: the head-head checking conﬁguration is already there, merged in the base. Since agreement feeds pronunciation, this conﬁguration must be conserved until Spell-Out. This means that at the level of the ﬁrst strong phase (vP), Q and NP must be adjacent. But at that point, VP, the sister of v, becomes inaccessible to the computation (Phase Impenetrability Condition16). There is therefore no way of moving Q without N: if it moves before Spell-Out, agreement cannot feed pronunciation; after that, it violates the PIC. The restriction on movement at play in (37) is due not to a peculiarity of head movement, but to the e¤ect of a phonological adjacency requirement. Crucially, a similar e¤ect is attested with phrasal movement, conﬁrming that it has nothing to do with head movement per se. As is well known, preposition stranding is prohibited in those languages, like French and Italian, where P and D tend to conﬂate in a PþD suppletive form (Law 1998; Salles 1997). In (40), this restriction against preposition stranding is shown at play in an Italian interrogative. (40) *Cosa hai parlato [di t]? what you-have talked of Consider the corresponding simpliﬁed structure in (41).

On Wh-Head Movement

39

(41)

If P and D need to conﬂate, they need to be adjacent at Spell-Out. But at the ﬁrst strong phase, when vP is sent to Spell-Out, VP becomes no longer accessible to the computation: the DP cosa cannot make a long movement alone, and preposition stranding is barred. 2.8

Conclusions

The theory of movement as it emerged from ‘‘On Wh-Movement’’ is still the core of the current theory of movement. This chapter has focused on what appears to be a stipulation of such a theory in light of the minimalist framework—the limitation of wh-movement to phrases—and has critically reviewed the more general division of movement into two distinct subtypes: head movement and phrasal movement. Having critiqued Chomsky’s recent attempt to ban head movement from syntax and reviewed the di¤erent constraints on movement and their legitimacy in a minimalist framework, I conclude that there is only one movement, triggered by features and deﬁned at that level. The di¤erence in the amount of material pied-piped in di¤erent environments is due to an economy principle on Merge that declares, ‘‘Merge just enough material for convergence.’’ Given that principle, the existence of two distinct movements can be derived from the irreducible di¤erence that holds by deﬁnition between heads and phrases: heads project, phrases are projections. A natural implication of this conclusion is that there is no principled reason for wh-movement to be restricted to phrases in any context, as the two cases of wh-head movement discussed in this chapter illustrate.

40

Caterina Donati

Notes Di¤erent versions of this chapter have been presented at Going Romance 2002, the workshop ‘‘On Wh-Movement,’’ and GLOW 2003. Many thanks to the audiences for their important comments, questions, and suggestions. 1. There is a highly detailed and well-motivated recent literature dedicated to covert phrasal movement: see Sauerland 1998, Fox 1999, Cecchetto 2000, Nissenbaum 2000, and Pesetsky 2000, among others. See Brody 1995 for a di¤erent view also discussing antecedent-contained deletion cases. 2. See Pesetsky 2000 for a di¤erent view, where both feature movement and phrasal movement are available to syntax as di¤erent operations. 3. The phenomena related to Romance cliticization are outrageously simpliﬁed in the text, which is meant only to show that the standard analysis may be complicated by a mistaken theoretical assumption (i.e., that a‰xation is head movement) that we might want to eliminate. (4) is not meant to be a full-ﬂedged alternative analysis of cliticization, of course, an analysis that should for example take into account the di¤erences among clitics standardly reduced to the distinction between a class of syntactic clitics and a class of phonological ones. 4. See also Zwart 2001 for a discussion of some clear interpretive e¤ects of instances of verb movement. 5. In fact, there is a sense in which a head is indeed always more remote than its phrase: fewer nodes separate the probe from the phrase bearing the goal than separate it from the corresponding head. But this notion of locality is too restrictive: it amounts to excluding head movement as a whole, which is not what we want. 6. There is an alternative I will not explore here, as it is less minimal than the one I choose to pursue. Suppose head movement and phrasal movement are systematically triggered by two intrinsically di¤erent classes of features; then it is conceivable that they are sensitive to di¤erent interveners and display di¤erent locality constraints. See Matushansky 2006 for a proposal along these lines. 7. The economy condition in (16) also makes an interesting prediction concerning movement of a feature associated with a given category to a target associated with the same feature. Since in this case the projection property of heads does not a¤ect the categorial status of the target, the minimal option should be compulsory, and phrasal movement systematically excluded. More precisely, we predict derivations moving a DP into Spec,DP, or a PP into Spec,PP, or a CP into Spec,CP to be ruled out as violations of the economy condition in (16). This means that standard analyses of constructions like the following need to be revised: (i) Possessives John’s wife (DP in Spec,DP; see, e.g., Kayne 1994) (ii) Exclamatives Into the dungeon with the traitor! (PP in Spec,DP; see Barbiers 1995)

On Wh-Head Movement

41

(iii) Clitic left-dislocations Che fosse un assassino, l’ho sempre saputo that was an assassin it(cl)-have-1sg always known (CP in Spec,CP; see, e.g., Cecchetto 1999) ‘That he was an assassin, I have always known it.’ Addressing the details of such a revision goes far beyond the limits of this chapter. Notice that some idiosyncratic properties of these constructions would beneﬁt from an analysis in which the dislocated constituent is adjoined to some extra functional head: in particular, this would explain the impossibility of recursion in both exclamative PPs and clitic left-dislocations, and the exclamative mood of examples like (ii). Thanks to Peter Svenonius for pointing out these facts to me. 8. As an anonymous reviewer points out, it is not true that a nominal cannot be interpreted as an interrogative. This is exactly what happens with so-called concealed questions, as discussed in Grimshaw 1979 and illustrated in (i). (i) a. I b. I c. I d. *I

know what answer he gave wonder what answer he gave know the answer he gave wonder the answer he gave

Under certain predicates (e.g., know but not wonder), a simple DP can replace a full-ﬂedged interrogative. The existence of such a class of nominals does not a¤ect the conclusions drawn here: without entering into the debate about what the structure of concealed questions is, it will be enough to say that the complex DPs resulting from head movement discussed here are not of the appropriate form— that is, are not concealed questions. 9. Again, the fact that some environments selecting for interrogative CPs also admit special DPs as concealed questions is not addressed here. I will simply concentrate on standard nominals. See note 8. 10. There is an exception to this generalization that readily comes to mind: both English and Italian have a class of free relatives that appear to allow some amount of pied-piping. (i) Mangero` qualunque biscotto preparerai [t] will-eat-1sg whatever cookie will-prepare-2sg ‘I will eat whatever cookie you will prepare.’ (ii) I shall visit [whatever town] you will visit [t] The fact that this movement pattern correlates with the presence of some extra material in the head of the clause, the su‰x -ever/-unque, suggests an obvious solution: namely, that these relatives are only apparently ‘‘free’’ (i.e., defective) and in fact correspond to full relative clauses. As such, they are generated as the complement of an external determiner (Kayne 1994; Bianchi 1999), the universal quantiﬁer -ever/-unque. (iii) I shall visit [DP [D ever] [CP [DP what town] [ . . . ]]] For some reason perhaps related to its universal value (Larson 1988), -ever/-unque triggers the raising and head adjunction of the wh-determiner, yielding (iv).

42

Caterina Donati

(iv) [DP [D whati [D ever]] [CP [DP ti town] [ . . . ]]] See Battye 1989 for a series of empirical arguments demonstrating that these relatives ought to be treated as ‘‘pseudo-free relatives.’’ See also Kayne 1994:154 for a similar analysis proposed on totally di¤erent grounds. 11. There is more than I have said in the text, of course: while the deﬁnition of heads as projecting features is certainly correct, that of phrases as nonprojecting features leaves adjunction out of the picture. This could be a problem for the analysis proposed here if a derivation such as (i) were allowed. (i)

Here, the phrase XP is moved and it projects in its landing site, turning the root into a right adjunct. There are, however, good reasons to exclude such a derivation on independent grounds. One concerns the legitimacy of the chain formed in (i), which violates a strict version of the CUC, since XP is maximal in its base position, but nonmaximal in its landing site (see Chomsky 1995, 257–259, for discussion). Another reason concerns the structure in (i), which is a case of rightadjunction and questionable as such (Kayne 1994). Thanks to Howard Lasnik for pointing this potential problem out to me. 12. Notice that Grosu (1994) proposes a di¤erent analysis of the movement pattern illustrated in these data. 13. Notice that the wh-element kolko is introduced by a special preposition ot in comparatives; other such prepositions are found in Italian (di; see (36)–(37)) and Spanish (de; see note 14, (iii)–(iv)). A similar analysis can be given of de in decıˆt in Romanian. In Donati 2000, I discuss the nature of this kind of particle in detail, showing that it does not form a constituent together with the wh-element it precedes, but instead governs the entire comparative clause. 14. As an anonymous reviewer points out, this crosslinguistic generalization does not hold in Spanish, at least if we consider the construction discussed by Price (1990, 43) and illustrated in (i) and (ii). Mi padre vende ma´s libros que discos compra mi madre my father sells more books than records buys my mother b. *Mi padre vende ma´s libros que compra mi madre discos my father sells more books than buys my mother records ‘My father sells more books than my mother buys records.’

(i) a.

La mesa es ma´s larga que ancha es la puerta the table is more large than wide is the door b. *La mesa es ma´s larga que es la puerta ancha the table is more large than is the door wide ‘The table is larger than the door is wide.’

(ii) a.

On Wh-Head Movement

43

In these examples, a null operator appears to pied-pipe its complement, contrary to what my analysis of head movement predicts. I have no proposal to make about this construction (but see Kennedy 2002), although its strongly emphatic ﬂavor may be invoked to justify a more complex structure for this type of comparative. Notice moreover that the preferred construction for comparatives in Spanish, illustrated in (iii) and (iv), indeed exhibits the movement pattern predicted by the analysis, with a bare element lo moving alone, stranding its complement. Mi padre vende ma´s libros de lo que mi madre compra (en) discos my father sells more books of Op that my mother buys (as) records b. *Mi padre vende ma´s libros de lo discos que (mi madre) compra my father sells more books of Op records that (my mother) buys (mi madre) (my mother) ‘My father sells more books than my mother buys records.’

(iii) a.

La mesa es ma´s larga de lo que la puerta es ancha the table is more large of Op that the door is wide b. *La mesa es ma´s larga de lo ancha que (la puerta) es (la puerta) the table is more large of Op wide that (the door) is (the door) ‘The table is larger than the door is wide.’

(iv) a.

Thanks to Daniela Capra for these data. 15. This possibility appears to be realized in other languages. For example, socalled internally headed relative clauses in Japanese (e.g., Shimoyama 1999) and nominalized PE-clauses in Italian Sign Language (LIS; Cecchetto, Geraci, and Zucchi, to appear) can be analyzed as involving the movement of a D head alone, stranding its nominal complement. Interestingly, no agreement surfaces in the DP either in Japanese or in LIS. See Branchini and Donati 2005 for details. 16. Chomsky (2001, 13) states the Phase Impenetrability Condition as follows (emphasis original): (i) [In a phase a with head H,] the domain of H is not accessible to operations outside a; only H and its edge are accessible to such operations. References Barbiers, Sjef. 1995. Another case of scrambling. Linguistics in the Netherlands 1995, 13–24. Battye, Adrian. 1989. Free relatives, pseudo-free relatives and the syntax of CP in Italian. Rivista di Linguistica 1, 219–250. Bianchi, Valentina. 1999. Consequences of antisymmetry: Headed relative clauses. Berlin: Mouton de Gruyter. Borsley, Richard, Maria Luisa Rivero, and John Stephens. 1996. Long head movement in Breton. In Richard Borsley and Ian Roberts, eds., The syntax of the Celtic languages: A comparative perspective, 53–74. Cambridge: Cambridge University Press.

44

Caterina Donati

Branchini, Chiara, and Caterina Donati. 2005. A snippet for the typology of relativization strategies: Relatives in Italian Sign Language. Paper presented at GLOW in Asia 2005, Nehru University, Delhi. Brody, Michael. 1995. Lexico-Logical Form. Cambridge, Mass.: MIT Press. Carnie, Andrew. 1995. Nonverbal predication and head movement. Doctoral dissertation, MIT. Cecchetto, Carlo. 1999. A comparative analysis of left and right dislocation in Romance. Studia Linguistica 53, 40–67. Cecchetto, Carlo. 2000. Doubling structures and reconstruction. Probus 12, 1–34. Cecchetto, Carlo, Carlo Geraci, and Alessandro Zucchi. To appear. Strategies of relativization in LIS. Natural Language and Linguistic Theory. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Donati, Caterina. 2000. La sintassi della comparazione. Padova: Unipress. Fox, Danny. 1995. Condition C e¤ects in ACD. In Rob Pensalﬁni and Hiroyuki Ura, eds., Papers on minimalist syntax, 105–120. MIT Working Papers in Linguistics 27. Cambridge, Mass.: MIT, Department of Linguistics and Philosophy, MITWPL. Fox, Danny. 1999. Reconstruction, binding theory, and the interpretation of chains. Linguistic Inquiry 30, 157–196. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10, 279–326. Grimshaw, Jane. 1991. Extended projection. Ms., Brandeis University. Grosu, Alexander. 1994. Three studies in locality and Case. London: Routledge. Heim, Irene. 1985. Notes on comparatives and related matters. Ms., University of Texas. Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell. Izvorski, Roumyana. 1995. A DP-shell for comparatives. In Antonietta Bisetto, Laura Bruge`, Joa˜o Costa, Rob Goedemans, Nicola Munaro, and Ruben van de Vijver, eds., Proceedings of Console III, 99–121. The Hague: Holland Academic Graphics.

On Wh-Head Movement

45

Kayne, Richard. 1989. Facets of Romance past participle agreement. In Paola Beninca`, ed., Dialect variation and the theory of grammar, 85–103. Dordrecht: Foris. Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press. Kennedy, Chris. 2002. Comparative deletion and optimality in syntax. Natural Language and Linguistic Theory 20, 553–621. Larson, Richard. 1988. Scope and comparatives. Linguistics and Philosophy 11, 1–26. Law, Paul. 1998. A uniﬁed analysis of preposition stranding in Romance and Germanic. In Pius Tamanji and Kiyomi Kusumoto, eds., Proceedings of NELS 28, 219–234. Amherst: University of Massachusetts, GLSA. Lebeaux, David. 1990. Relative clauses, licensing and the nature of the derivation. In Juli Carter, Rose-Marie De´chaine, Bill Philip, and Tim Sherer, eds., Proceedings of NELS 20, 318–332. Amherst: University of Massachusetts, GLSA. Lema, Jose´, and Maria Luisa Rivero. 1990. Long head movement: ECP vs. HMC. In Juli Carter, Rose-Marie De´chaine, Bill Philip, and Tim Sherer, eds., Proceedings of NELS 20, 333–347. Amherst: University of Massachusetts, GLSA. Manzini, M. Rita. 1994. Locality, minimalism, and parasitic gaps. Linguistic Inquiry 25, 481–508. Matushansky, Ora. 2006. Head movement in linguistic theory. Linguistic Inquiry 37.1. Nissenbaum, Jonathan. 2000. Investigations of covert phrase movement. Doctoral dissertation, MIT. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: MIT Press. Pesetsky, David, and Esther Torrego. 2001. T-to-C movement: Causes and consequences. In Michael Kenstowicz, ed., Ken Hale: A life in language, 355–426. Cambridge, Mass.: MIT Press. Price, Susan. 1990. Comparative constructions in Spanish and French. London: Routledge. Rivero, Maria Luisa. 1991. Long head movement and negation. The Linguistic Review 8, 319–351. Roberts, Ian. 1994. Two types of head-movement in Romance. In David Lightfoot and Norbert Hornstein, eds., Verb movement, 207–242. Cambridge: Cambridge University Press. Roberts, Ian. 1998. Have/Be raising, Move F, and Procrastinate. Linguistic Inquiry 29, 113–125. Russell, Bertrand. 1905. On denoting. Mind 14, 479–493. Salles, Heloisa. 1997. Prepositions and the syntax of complementation. Doctoral dissertation, University of Wales, Bangor. Sauerland, Uli. 1998. The meaning of chains. Doctoral dissertation, MIT.

46

Caterina Donati

Shimoyama, Junko. 1999. Internally headed relative clauses in Japanese and E-type anaphora. Journal of East Asian Linguistics 8, 147–182. Stechow, Arnim von. 1984. Comparing semantic theories of comparison. Journal of Semantics 3, 1–77. Travis, Lisa. 1984. Parameters and e¤ects of word order variation. Doctoral dissertation, MIT. ´ avar. 1994. Long head movement? Verb movement Wilder, Chris, and Damir C and cliticization in Croatian. Lingua 93, 1–58. Zwart, C. Jan-Wouter. 2001. Syntactic and phonological verb movement. Syntax 4, 34–62.

Chapter 3 The Pied-Piper Feature

3.1

Akira Watanabe

Theory of Pied-Piping

In recent theorizing (Chomsky 2000, 2001, 2004), the movement operation is decomposed into three parts, consisting of Agree, pied-piping, and Merge. Pied-piping is the process that creates a copy of some category that contains the goal of the Agree operation. The copy is then merged to become a speciﬁer of the projection whose head contains the probe and the EPP feature. In this chapter, I would like to propose that pied-piping also makes use of a formal feature, which I call the pied-piper feature. The pied-piper feature resolves the technical problem that accompanies the decomposition of Move into three suboperations. Pied-piping must copy some phrase that contains the goal for Agree. But the Agree operation renders the goal inactive, making it inaccessible for further computation in narrow syntax including pied-piping. Pied-piping cannot take place before Agree, however, because the goal has to be located by Agree. The pied-piper feature saves this situation. It marks the category to be copied by pied-piping. The pied-piper feature also helps reduce the workload of the PF computation. Suppose that the pied-piper is hooked onto the phonological features of the category to be copied. This is plausible if Chomsky (2001) is right in claiming that pied-piping requires phonological content—in other words, that the category to be copied must have phonological content. (See also Takahashi 1997, 2000.) Suppose further that the pied-piper feature is retained in the copy, but not in the original, as in (1), where YP is raised. F(PP) is the shorthand for the pied-piper feature.

48

Akira Watanabe

(1)

It follows that the PF materials are retained only in the newly created copy. In other words, the pied-piper feature is the device that determines pronunciation of a chain before Spell-Out without forcing the computational system to handle phonological features directly in narrow syntax. The PF computation is not called upon to make the decision for pronunciation anymore, even though we assume that deletion of phonological content of the original takes place during the PF computation.1 The idea here is that pronunciation of a chain and the need for phonological content in pied-piping are tightly connected. Let us consider what happens if the decision for pronunciation is not made in narrow syntax in the model of cyclic Spell-Out proposed by Chomsky (2000, 2001, 2004). Suppose HP is a strong phase. The complement of H is sent to the phonological computation after Spell-Out. Examination of the complement of H alone does not tell us whether there is a copy remaining in the next higher phase. Take the case where H is an interrogative C that agrees with YP, as in (2). (2) [CP YP C [ . . . YP . . . ]] In that case, the original YP within the complement of C no longer has an active uninterpretable feature at the point of Spell-Out. Thus, it is indistinguishable for the PF computation from a phrase ZP that is to be pronounced within the complement of H after being rendered inactive by agreement with a head below H, as in (3), where H is v in (3a) and C in (3b). (3) a. v [ . . . [CP ZP C . . . ]] b. C [TP ZP T . . . ] To make sure that YP within the complement of C will not be pronounced in (2), it would be necessary to look at the material to be sent to PF at the next strong phase (namely, the copy in Spec,HP ¼ CP) as well. Thus, cyclic Spell-Out itself does not alleviate the pronunciation problem. It is the pied-piper feature that allows the decision for pronunciation to be made in a very local fashion, as part of pied-piping. In this

The Pied-Piper Feature

49

sense, the use of the pied-piper feature is a desirable design speciﬁcation for the human language faculty. I will demonstrate that straightforward empirical evidence for the piedpiper feature is found in some cases of wh-movement. The theory of whmovement is outlined in section 3.2. Section 3.3 addresses the problems left open in my earlier treatment of the loss of wh-movement in Old Japanese (Watanabe 2002, 2004c). The solution to one of these problems points to a morphological piece that can only be analyzed as the piedpiper feature. Section 3.4 extends the domain of inquiry to verb-second in order to explore the nature of the pied-piper feature. Section 3.5 concludes the discussion. 3.2

Theory of Wh-Movement

Insights into the pied-piping mechanism should come from the domain of wh-movement, where a variety of categories can be displaced. Amovement is essentially limited to DP. Whatever principle of Universal Grammar (UG) lies behind this restriction on A-movement would make an additional feature superﬂuous, so that if some feature is involved in pied-piping in general, morphophonological realization of that feature is likely to be null in the domain of A-movement. In wh-movement, on the other hand, the working of the pied-piper feature will not be superﬂuous, since deciding which category to copy for Merge is not a trivial question. One major goal of this chapter is to present evidence that there are indeed wh-movement cases in which the pied-piper feature is morphophonologically realized. As the background for discussion, let us consider the features that take part in wh-dependencies. On the basis of the analysis of f-feature agreement, Chomsky (2000, 2001, 2004) proposes that for the Agree operation to apply, the goal as well as the probe must be active in the sense of containing uninterpretable features. In the A-movement system, the probe is a set of uninterpretable f-features, and the goal a set of interpretable ffeatures and structural Case. Chomsky (2000, 128) suggests that the whmovement system will mimic the A-movement system rather closely if the probe is an uninterpretable Q-feature of C and the wh-phrase contains an interpretable Q-feature and an uninterpretable wh-feature, as in (4). (4) Chomsky 2000 probe ¼ uninterpretable Q-feature goal ¼ interpretable Q-feature & uninterpretable wh-feature

50

Akira Watanabe

It should be mentioned, however, that Chomsky (2001, 48n57) leaves completely open what kind of features are involved in wh-movement. In Watanabe 2002, I propose, given the analysis of loss of wh-movement in Old Japanese, that what renders the goal active is an uninterpretable focus feature, which exhibits focus morphology but has no inherent semantic content. (See Watanabe 2004a on the morphological realization of this feature in negative concord.) In Watanabe 2004c, I argue further that the Q-feature of C is interpretable. It follows that for Agree to apply, only the goal needs to be active. (5) Watanabe 2002, 2004c probe ¼ interpretable Q-feature goal ¼ interpretable Q-feature & uninterpretable focus feature This is a natural hypothesis in the bottom-up derivational system. The idea is that once an uninterpretable feature is thrown in, the computational system gets alerted. In this sense, the requirement that the probe be active is superﬂuous. To review the part of the arguments in Watanabe 2004c about the Qfeature of C that will become relevant later in this chapter, there are whin-situ languages such as Chinese and Lakhota in which only wh-adjuncts are sensitive to wh-islands, as in (6). (6) a. Ni xiang-zhidao [wo weishenme mai shenme]? you wonder I why buy what i. ‘What do you wonder why I bought?’ ii. *‘Why do you wonder what I bought?’ b. [Tuwa takuwe cheya ha4 ki] Mary inu4gˇa he? who why cry dur C 0 Mary you-ask Q i. ‘Who did you ask Mary why (he) was crying?’ ii. *‘Why did you ask Mary who was crying?’

(Chinese)

(Lakhota)

Tsai (1994, 1999) proposes that Chinese should be analyzed as employing unselective binding for argument wh-in-situ and LF movement for adjunct wh-in-situ. Lakhota receives essentially the same analysis, as argued by Bonneau (1992). In the framework proposed by Chomsky (2000, 2001, 2004), this means that only a wh-adjunct undergoes Agree, driven by an uninterpretable feature of the wh-phrase. (Also see Cheng and Rooryck 2002.) At the same time, there is no evidence that the interpretability of the Q-feature of the interrogative C di¤ers depending on whether it is associated with an argument wh-phrase or an adjunct wh-phrase. Furthermore, wh-in-situ in general must always undergo Agree in languages like

The Pied-Piper Feature

51

Japanese where the wh-island e¤ect is detected even for argument whphrases, as in (7). (7) ??John-wa [Mary-ga nani-o katta kadooka] Tom-ni John-top Mary-nom what-acc bought whether Tom-dat tazuneta no? asked Q ‘What did John ask Tom whether Mary bought?’ Again, there is no evidence that the status of the Q-feature varies crosslinguistically. On the other hand, it is not implausible to assume the presence of an additional uninterpretable feature on wh-adjuncts in Chinese and Lakhota or on wh-phrases in general in Japanese. The morphosyntax of wh-phrases di¤ers between Chinese and Lakhota on the one hand and Japanese on the other, as discussed by Tsai and reviewed later on. Thus, (5) should be chosen as the system of agreement for wh-movement. 3.3

Old Japanese

3.3.1 Loss of Wh-Movement There are two open questions the analysis of the loss of wh-movement in Old Japanese presented in Watanabe 2002. Let us start by reviewing this analysis. The form of wh-questions in the relevant periods is as follows: (8) a. 7th–8th centuries: b. 10th–11th centuries:

[wh-ka [TP subj-nom . . . t . . . ] j] [[TP . . . wh-j . . . ] (particle)]

Representative examples are given in (9). (9) a. Kasugano-no fuji-ha chiri-ni-te nani-wo-ka-mo Kasugano-gen wisteria-top fall-perf-conj what-acc-ka-mo mikari-no hito-no ori-te kazasa-mu? hike-gen person-nom pick-conj wear.on.the.hair-will ‘Since the wisteria ﬂowers at Kasugano are gone, what should hikers pick and wear in their hair?’ (Man’youshuu, no. 1974) b. Ito ayashiki mi-kokoro-no, geni, ikade very curious hon-character-nom indeed how naraha-se-tamahi-kemu? develop-hon-hon-would.have ‘How has his curious character developed indeed?’ (Tales of Genji, ‘‘Ukifune’’)

52

Akira Watanabe

In the seventh and eighth centuries, the wh-phrase precedes the nominative subject marked by no or ga, as in (9a). This word order restriction is no longer found in the tenth and eleventh centuries, as illustrated in (9b).2 In other words, Old Japanese used to have overt phrasal wh-movement, but it was lost. Other changes, summarized in (10), took place at about the same time (see Watanabe 2002). (10) Changes related to the loss of wh-movement a. Decrease in the use of the particle ka b. Tendency of ka to appear in rhetorical questions c. Ka no longer possible in yes/no questions Notice that the wh-phrase in (9a) is accompanied by the particle ka, a regular pattern in the seventh and eighth centuries, whereas no such particle is used in (9b). The remaining use of ka in the tenth and eleventh centuries, to which we turn below, is typically found in rhetorical questions. Furthermore, though ka used to be able to mark the focus of yes/no questions, too, as in (11), where the focused phrase is preposed over the nominative subject as in the case of wh-questions, this use became obsolete after wh-movement was lost. (11) Hatsuse-no kaha-ha ura na-mi-ka fune-no Hatsuse-gen river-top shore absent-ness-ka boat-nom yori-ko-nu? approach-come-neg ‘Is it because Hatsuse River has no shore that no boat comes near?’ (Man’youshuu, no. 3225) As an explanation of these correlated changes, I argue in Watanabe 2002 that ka was the realization of the uninterpretable focus feature in the seventh and eighth centuries and that it was reanalyzed as the polarity item specialized for rhetorical wh-questions when wh-movement was lost. One problem left open in Watanabe 2002 is why the particle ka continued to be used in genuine wh-questions, albeit less frequently, even after the loss of wh-movement. (10b) is only a tendency. In other words, ka was not restricted to rhetorical wh-questions. Let us look at some details. On the basis of data collected in Isobe 1990, (12) gives the breakdown of all the uses of ka in Tales of Genji, a literary work attributed to the beginning of the eleventh century. (12) Wh-questions with ka in Tales of Genji (Isobe 1990) Genuine questions: 84 (47%) Rhetorical questions: 95 (53%)

The Pied-Piper Feature

53

(13a) is an example of a genuine, nonrhetorical wh-question, and (13b) an example of a rhetorical wh-question. (13) a. Toshi-ha ikutu-ni-ka monoshi-tamahi-shi? age-top how.many-loc-ka be-hon-past ‘How old was she?’ (Tales of Genji, ‘‘Yufugaho’’) b. Dare-ka-ha yoroshiu omohi-kikoe-n? who-ka-top not.bad think-hon-will ‘Who would think it is not bad?’ (Tales of Genji, ‘‘Suma’’) The question is what the status of the particle ka in genuine wh-questions is after the loss of wh-movement. A clue to the answer is found in Isobe’s (1990) data on the relation between the clause-ﬁnal form and the possibility of the rhetorical question reading, summarized in table 3.1. Notice that the rhetorical wh-question reading is concentrated in the cases where the particle ka is combined with the modal auxiliary mu. Now, Den Dikken and Giannakidou (2002) observe that the antiD(iscourse)-linking marker the hell or its equivalent forces the rhetorical question reading in various languages including English, French, Spanish, Table 3.1 Choice of modal auxiliary in wh-questions with ka in Tales of Genji (from Isobe 1990) Ending

Genuine question

Rhetorical question

mu kemu ramu mashi mu to sura-mu bekara-mu mu to su beshi beki-zo zo (particle) j

21 14 9 1 1 3 0 3 2 0 15

70 0 0 2 0 0 1 9 0 1 2

Total number of full sentences

69

85

Ellipsis

15

10

54

Akira Watanabe

Hungarian, and Greek when combined with a modal. This observation is illustrated for English in (14). (14) a. Who would buy that book? b. Who the hell bought that book? c. Who the hell would buy that book? In contrast to (14a–b), (14c) cannot be interpreted as a genuine information-seeking question. It is then reasonable to analyze the tenthand eleventh-century Old Japanese phenomenon in a similar way. The particle ka was the anti-D-linking marker equivalent to the hell at that time and induced the rhetorical question reading with the help of the modal auxiliary mu. Note that the changes listed in (10) are explained straightforwardly under this analysis. 3.3.2 Appearance of Head-Internal Relative Clauses Let us now turn to the other problem left open in Watanabe 2002. This problem is related to the fact, discussed in Watanabe 2004c, that headinternal relative clauses (HIRCs) started to appear at about the same time as wh-movement was lost in Old Japanese. As mentioned in section 3.2, there are at least two types of wh-in-situ languages. The type represented by Chinese and Lakhota mainly makes use of unselective binding.3 The Japanese type, on the other hand, forms the dependency by means of Agree. After overt wh-movement was lost in Old Japanese, the question is which type of wh-in-situ language emerged. Watanabe 2002 ends with this question, which is impossible to answer conclusively in diachronic syntax as long as we limit our attention to questions of locality. However, the di¤erence between the two types is not limited to the mode of dependency. In Watanabe 2004c, I claim that the properties of wh-in-situ and the HIRC are closely connected. The type of wh-in-situ in Old Japanese can be identiﬁed by focusing on the HIRC and related properties. Among the Chinese-type wh-in-situ languages, Chinese and Lakhota diverge in the availability of HIRCs. Given the structure in (15) for the HIRC (cf. Kayne 1994), its absence in Chinese and presence in Lakhota can be attributed to the absence (Chinese) or presence (Lakhota) of D. (15) [DP CP D 0 ] The Lakhota HIRC is illustrated in (16a–b). (16) a. [[Mary owi4zˇa wa4 kagˇe] ki ] he ophewathu4 Mary quilt a make the dem I-buy ‘I bought the quilt that Mary made.’

The Pied-Piper Feature

55

b. [[Wichota wowapi wa4 yawa pi cha] ob wo"u4glaka pi many-people paper a read pl ind with we-speak pl ki ] he L.A. Times e the that L.A. Times be ‘The newspaper that we talk to many people who read (it) is the L.A. Times.’ c. [Wicˇha´sˇa ki ] [matho´ wa4] kte´ man the bear a kill ‘The man killed a bear.’ Note the deﬁnite article ki at the end of the (outer) HIRC in (16a–b). (16c) is an example of an ordinary clause, where the nouns are accompanied by an article. It follows from the presence of HIRCs in Modern Japanese (see (17a)) that this language also has D. The properties of the HIRC in Modern Japanese di¤er from those of the Lakhota HIRC, however. While Lakhota can embed one HIRC inside another as in (16b) (Williamson 1987), Japanese cannot. Compare (16b) with (17b). (17) a.

Mary-ga [John-ga [zibun-no gakusei-ga juuyouna Mary-nom John-nom self-gen student-nom important kasetsu-o teianshita to] jimanshite-ita no]-no kekkan-o hypothesis-acc proposed C 0 boasted-had C 0 -gen defect-acc shitekishita pointed. out ‘Mary pointed out a defect of the important hypothesis which John had boasted that his student proposed.’ b. *[John-ga [MIT-no gakusei-ga subarashii ronbun-o kaita John-nom MIT-gen student-nom excellent paper-acc wrote no]-o posuto-doku-toshite saiyoushite-ita no]-no C 0 -acc post-doc-as adopted-had C 0 -gen shuppan-ga okureta publication-nom was.delayed ‘Publication of an excellent paper which John had hired as a postdoc an MIT student who wrote (it) was delayed.’

The head-D relation for the outer HIRC in (16b) and (17b), shown schematically in (18), is blocked in (17b). (18) [DP [CP . . . [DP [CP . . . head . . . ] D 0 ] . . . ] D 0 ] # " X This is an instance of a locality violation found with the Agree relation, but not with unselective binding. We are thus led to the conclusion that

56

Akira Watanabe

the dependency in HIRCs employs agreement in Modern Japanese and unselective binding in Lakhota. This di¤erence in HIRCs can be linked further to the type of indeterminate system found in each language (see Watanabe 2004c). As discussed in detail by Aoun and Li (1993) and Tsai (1994, 1999), the indeterminate system also has two varieties. Modern Japanese adds a particle to a whphrase to form a non-wh quantiﬁcational expression, as in (19). (19) a. Dare-ga ringo-o tabeta no? who-nom apple-acc ate Q ‘Who ate an apple?’ b. Daremo-ga ringo-o tabeta everyone-nom apple-acc ate ‘Everyone ate an apple.’ c. Daremo ringo-o tabe-nak-atta anyone apple-acc eat-neg-past ‘No one ate an apple.’ d. Dareka-ga ringo-o tabeta someone-nom apple-acc ate ‘Someone ate an apple.’ Chinese, on the other hand, does not add such a particle, as illustrated in (20). (20) a. Ni xiang mai shenme (ne)? you want buy what Q ‘What do you want to buy?’ b. Wo shenme dou mai I what all buy ‘I will buy everything.’ c. Wo bu xiang mai shenme I not want buy what ‘I don’t want to buy anything.’ Aoun and Li (1993) and Tsai (1994, 1999) indeed claim that this parametric variation of the indeterminate system is responsible for the di¤erence in locality of wh-in-situ. Lakhota belongs to the Chinese type, as the following examples from Williamson 1984 show: (21) a. Charlotte taku kagˇa he? Charlotte what make Q ‘What did Charlotte make?’ ‘Did Charlotte make something?’

The Pied-Piper Feature

57

b. Charlotte taku kagˇe Charlotte what make ‘Charlotte made something.’ Recall that Lakhota patterns with Chinese in terms of the locality of whin-situ. How is the parametric variation of the indeterminate system related to the typology of HIRCs? The ﬁrst step in answering this question is the idea, proposed in Watanabe 1992, that the quantiﬁcational particle attached to a wh-phrase is an instance of D as shown in (22) and that the wh-phrase in its wh-use is accompanied by a null D. (22)

The common part is called an indeterminate. The absence of D in Chinese is consistent with the fact that the Chinese indeterminate system does not use particles. Now suppose that the quantiﬁcational particle attached to an indeterminate always requires agreement.4 The indeterminate-particle relation indeed exhibits the locality property typical of agreement, as observed by Takahashi (2002), who points to the following contrast: (23) a.

[[Dare-o hihanshita hito]-o taihoshita keikan]-mo who-acc criticized person-acc arrested policeman-mo basserareta was-punished ‘For every person x, the policeman who arrested a person who criticized x was punished.’ b. *[[Dare-o hihanshita dare-ka]-o taihoshita keikan]-mo who-acc criticized who-ka-acc arrested policeman-mo basserareta was-punished ‘For every person x, the policeman who arrested someone who criticized x was punished.’

Note that the conﬁguration in (18) is found in (23), except that D agrees with an indeterminate in (23). (23b) is ruled out because the two D heads are of the same type. Suppose then that D in Modern Japanese is

58

Akira Watanabe

uniformly of the agreeing type. It follows that the head-D dependency in the Japanese HIRC involves agreement. The indeterminate system of the Chinese type, on the other hand, does not a¤ect the nature of D since it is incompatible with D. Lakhota employs binding D outside the indeterminate system as a default. Chinese does not use D at all. Parameterizing D in this way explains the properties of wh-in-situ as well. Notice that D is involved both in wh-in-situ (as the goal) and in HIRCs (as the probe) in languages with the Japanese-type indeterminate system, as in (24). (24) a. [DP [CP . . . head . . . ] D 0 ] "

(HIRC)

b. [CP [TP . . . [DP QP D 0 ] . . . ] C 0 ] "

(wh-in-situ)

Since D is uniformly of the agreeing type, both wh-in-situ and HIRCs are sensitive to minimality. In languages with the Chinese-type indeterminate system, wh-in-situ does not make use of D and falls back on unselective binding by C, again as a default. Table 3.2 summarizes the various properties of the three languages discussed here. D plays a central role in this typology.5 The key assumption is the uniformity of the D type in a given language. Now we are ready to go back to Old Japanese. As observed by Kondo (1981), the HIRC started to appear in the tenth century, by which time overt wh-movement had already been lost. One of the earliest examples of the HIRC is given in (25). (25) [Kano Joukyau-den-no-mae-no-matsu-ni yuki-no that Joukyau-den-gen-front-gen-pine.tree-loc snow-nom furikakari-keru]-wo orite, . . . fall-perf-acc snap-conj ‘She snapped o¤ a branch of the pine tree covered with snow in front of that Joukyau-den, and . . .’ (Tales of Yamato, 139th tale) Table 3.2 Typology of wh-in-situ languages with an indeterminate system

Chinese Lakhota Japanese

Indeterminate

Wh

HIRC

D

no particle no particle particle

binding binding checking

— binding checking

— binding checking

The Pied-Piper Feature

59

For more examples, see Kuroda 1974. As shown in Watanabe 2004c, the correlation between the loss of overt wh-movement and the appearance of HIRCs is strong evidence that the mechanism of wh-in-situ makes HIRCs possible.6 As in the case of wh-in-situ, it is practically impossible to determine the type of the Old Japanese HIRC directly in terms of locality. If we look at indeterminates, however, it seems justiﬁed to conclude that wh-in-situ and the HIRC in Old Japanese are basically of the same type as their Modern Japanese counterparts, since Old Japanese also has an indeterminate system of the Modern Japanese type, as shown by the examples in (26). (26) a. Ima-ha tare-mo tare-mo e nikumi-tamaha-ji now-top who-mo who-mo neg.possible hate-hon-will.not ‘Now, it would be impossible for anyone to hate him.’ (Tales of Genji, ‘‘Kiritsubo’’) b. Idureno on-kata-mo . . . toridorini ito medetakere-do . . . which hon-person-mo variously very beautiful-though ‘Though every one of them is very beautiful in various ways, . . .’ (Tales of Genji, ‘‘Kiritsubo’’) In this period, mo is the only particle attached to an indeterminate. Reduplication as in (26a) was also productive at that time, judging from the quantitative data in Yamanishi 1987. The conclusion that wh-in-situ in Old Japanese is the same as wh-in-situ in Modern Japanese now forces us to reconsider the analysis of the particle ka presented in Watanabe 2002, according to which it was the realization of an uninterpretable focus feature when the wh-phrase was overtly displaced. It turns out that wh-in-situ participates in feature checking after the loss of overt phrasal wh-movement in Old Japanese after all. There would be no reason why ka ceased to be used obligatorily after the loss of wh-movement, if ka had been the realization of an uninterpretable focus feature, the driving force behind agreement. What role, then, did the particle ka play in the seventh and eighth centuries? 3.3.3 Pied-Piping The range of options for analysis provided by the theory of UG is in fact quite narrow. Recall that the movement operation consists of three parts as in (27). (27) Move ¼ Agree þ pied-piping þ Merge Among these, Agree remains after the loss of wh-movement, by hypothesis. So it is pied-piping and Merge that were lost. Merge is driven by the

60

Akira Watanabe

EPP feature of the head that contains the probe. In principle, it should not a¤ect the shape of the wh-phrase. The only remaining possibility is that the particle ka was related to the mechanism of pied-piping in the seventh and eighth centuries. I would like to propose that there is an uninterpretable feature, called the pied-piper feature, that plays an important role in pied-piping, as outlined in section 3.1. Ka was the realization of that feature. The Inclusiveness Condition, which says that no new features are introduced by the computational system during the derivation (Chomsky 2000, 113), dictates that the pied-piper feature should be assigned at the time of forming a numeration. At this point, the computational system cannot foresee which category to copy for pied-piping. The selection of the category to be copied, therefore, must be made when the computational system builds the structure that contains the goal for Agree as well as the pied-piper feature. If the pied-piper feature is placed in a wrong position, computation cannot proceed any further. Exactly how this is implemented is a question for future research. 3.3.4 Crosslinguistic Connections Evidence for the pied-piper feature is not limited to the diachronic change in Old Japanese. It is also found in optional wh-movement languages. Clements (1984) observes that a preposed wh-phrase is accompanied by a focus particle in Kikuyu. Compare the examples in (28). (28) a. kama.u´ a-O´n-!´ırE´ o? Kamau sp-see-t who ‘Who did Kamau see?’ b. no´.o o´-ew-eciı´ri-a ngoee a-u´e-ı´rE a´te kama.u´ a-O´n-! ´ırE´? fp.who sp-t-think-t Ngugi sp-say-t that Kamau sp-see-t ‘Who do you think Ngugi said that Kamau saw?’ The focus particle ne´ is attached to the fronted wh-phrase o to form no´.o in (28b). The same focus particle is used for the fronted non-wh focus, too, as in (29). (29) ne´ kaana´kE´ ka´ma´.! u´ a-O´n-!´ı! rE´ fp Kanake Kamau sp-see-t ‘It is Kanake that Kamau saw.’ Sabel (2000) takes this focus particle to be the realization of the focus feature as in the analysis of Old Japanese in Watanabe 2002, though Sabel assumes that this feature is interpretable.7 Given the perspective taken

The Pied-Piper Feature

61

here, however, this Kikuyu particle is also a good candidate for the realization of the pied-piper feature. Bergvall (1983) obseves that wh-in-situ in Kikuyu is sensitive to the Complex NP Constraint, as illustrated in (30) (tone omitted).8 (30) *matiO:e mo:do orea otiniriE ke? they.not.know person dem cut what ‘What did they not know anyone who cut?’ As suggested in Watanabe 2004b, the Complex NP Constraint e¤ect can be reduced to the locality of Agree, once it is shown that the D head of the complex NP contains a feature that is probed by the interrogative C. Recall also the earlier discussion of locality in HIRCs and the indeterminate-particle association. I speculate that the feature in question is related to quantiﬁcation, though exploring the idea goes well beyond the scope of this chapter. The important point here is that the Kikuyu wh-in-situ makes use of agreement, so we are led to assume that it contains an uninterpretable focus feature. The focus particle itself, then, should be analyzed as the realization of the pied-piper feature. Thus, if a particle is attached to the preposed wh-phrase, it is a candidate for realizing the pied-piper feature. A particle attached to wh-in-situ, on the other hand, cannot be a realization of the pied-piper feature. Sugahara (1996) and Miyara (2003) discuss the particle that accompanies whin-situ in Okinawan, illustrated in (31).9 (31) Wan-ya [Taru-ga nuu-ga kam-yi-ra] shiribusaN I-top Taru-nom what-wh eat-pres-Q want.to.know ‘I want to know what Taru eats.’ Sugahara and Miyara share the view that this particle hosts the whfeature, but Miyara characterizes the di¤erence between (32a) and (32b) by saying that (32a) is emphatic, indicated by the addition of ‘on earth’ in the translation. (32) a. Taa-ga-ga ich-u-ra? who-nom-wh go-pres-Q ‘Who on earth is going?’ b. Taa-ga-j ich-u-ga? who-nom-wh go-pres-Q ‘Who is going?’ It then seems reasonable to treat the particle ga on a par with the particle ka in Old Japanese after the loss of wh-movement, namely, as an anti-Dlinking marker.

62

Akira Watanabe

According to Kishimoto (1992), Sinhala has a particle similar to the one in Okinawan. Even though Kishimoto does not comment on its semantic e¤ect, it is likely that it is also an anti-D-linking marker. 3.4

Verb-Second

Overt displacement of a wh-phrase is often accompanied by head movement, resulting in the verb-second conﬁguration. In this section, I would like to explore the possibility that the pied-piper feature is involved in this well-known phenomenon. The schematic conﬁguration of verb-second is shown in (33), where XP is raised from within TP as an operator. (33) [CP XP [[C 0 [V 0 þ T 0 ] þ C 0 ] TP]] It is true that verb movement to C is not limited to cases where an overt operator is raised to Spec,CP. It is also found in yes/no questions, for example, leading some researchers to posit a null operator in Spec,CP even for yes/no questions. There is a sense, however, in which overt phrasal movement induces verb movement to C, given the rather systematic pattern displayed by wh-questions in Germanic.10 The question is how. The idea I would like to pursue is that the pied-piper feature that accompanies overt phrasal movement to Spec,CP is responsible for verb movement to C. Note ﬁrst that the raised ﬁnite verb marks the right edge of the phrase in Spec,CP, since verb-second arises in the head-initial structural representation. Compare the verb-second conﬁguration in (34a) with the structure in (34b), where the pied-piper feature (abbreviated as F(PP)) is realized as part of the raised operator. (34) a.

b.

The Pied-Piper Feature

63

When the pied-piper feature is realized on the right edge of the displaced operator, its positioning is strikingly similar to that of the raised verb. It is, therefore, not unreasonable to assume that the pied-piper feature of the fronted operator induces verb movement as well. To implement the idea that verb-second is due to the pied-piper feature, we need to have a theory of head movement in general. Chomsky (2000, 2001) has suggested that head movement is not part of the narrow syntactic computation but belongs to morphophonology, on the grounds that it di¤ers from phrasal movement in signiﬁcant respects. The properties listed in (35) certainly set head movement apart from phrasal movement. (35) Peculiarities of head movement a. Failure to extend the syntactic object b. Failure to pied-pipe the maximal projection c. Absence of interpretive e¤ects Zwart (2001), on the other hand, argues that the agreement system lies behind head movement, while recognizing that it has morphophonological aspects as well. My claim is that head movement is morphophonological but has a syntactic origin (namely, the pied-piper feature), at least in some cases. Recall from section 3.1 that the pied-piper feature must be hooked to the phonological material. Now suppose that when Agree takes place, the pied-piper feature is copied onto the head that contains the probe. This is an extension of the idea suggested by Chomsky (1998) that Agree involves feature copying. Feature copying converts the structure in (36a) into the one in (36b) after the head H agrees with the feature F. (36) a. [H 0 [ . . . [XP . . . F. . . ] . . . ]] b. [H 0 þ F [ . . . [XP . . . F. . . ] . . . ]] For empirical support from switch reference and negative concord, see Watanabe 2000, 2004a. My proposal is that the pied-piper feature is also copied onto the probing head when Agree takes place. It thus looks as if the pied-piper feature itself entered into the checking relation with the EPP feature. Once it becomes part of the head that contains the probe, the pied-piper feature asks for phonological material. It copies the phonological content of the closest head when the head where the copied piedpiper feature is located does not have phonological content itself. Here, the closest head means the head of its sister. This copying of phonological material results in head movement. Since it takes the form of copying the

64

Akira Watanabe

phonological content of a head onto the next higher head, the properties in (35) follow straightforwardly. Notice that this theory also predicts V-to-T movement when the subject is raised to Spec,TP. Obviously, this is a wrong result for English sentences that do not contain an auxiliary. Essentially the same problem arises in embedded questions in Germanic in general. As illustrated by the Danish examples in (37), there is no verb-second e¤ect in indirect questions. (37) a. Hvad har børnene set? what have children-the seen ‘What have the children seen?’ b. Jeg ved ikke hvilken ﬁlm børnene har set I know not which ﬁlm children-the have seen ‘I don’t know which ﬁlm the children have seen.’ (Vikner 1995, 48–49) But the theory I am proposing incorrectly requires verb raising to C in (37b), unless it is modiﬁed in some way. Now I would like to claim that head movement needs a catalyst in addition to the copied pied-piper feature. More generally, I propose that head movement is induced when the request for phonological material is combined with a catalyst. It does not take place otherwise. We should in fact allow something other than the copied pied-piper feature to ask for phonological material, because the verb raising that creates verb-initial sentences (including yes/no questions) takes place without an accompanying phrasal movement. The copied pied-piper feature cannot be held responsible for verb raising in such cases. Now, the role of a catalyst is to fulﬁll the request for phonological material only in its presence. The failure of verb raising in (37b), then, can be accounted for quite easily. In contrast to the matrix C, the embedded interrogative C does not contain a catalyst and thus fails to induce head movement, even though the piedpiper feature is copied.11 In other words, the request for phonological material is aborted in the absence of a catalyst. The catalyst replaces Chomsky’s (1993) strong V-feature. The di¤erence is that while the strong V-feature is assumed to trigger overt head movement by itself, a catalyst works in collaboration with a request for phonological material such as the pied-piper feature. Although at ﬁrst sight the theory looks as stipulative as the theory of strong V-features, it actually succeeds in capturing the empirical properties of head movement in an optimal way. The notion of a catalyst

The Pied-Piper Feature

65

accounts for the robust generalization (e.g., Roberts 1993; Vikner 1997; Rohrbacher 1999) that rich verbal morphology is linked to V-to-T movement. In other words, the catalyst for V-to-T movement is the property marked on the T head that is realized as rich inﬂection. At the same time, Jonas (2002) observes that Shetland English retained V-to-T movement until quite recently despite the absence of rich inﬂection. For these cases, the brute force speciﬁcation of the catalyst is needed, which is triggered by syntactic cues in parameter setting, as Jonas suggests. The same parameter setting triggered by syntactic cues probably applies to the catalyst on the matrix interrogative C in Germanic in general as well. Thus, empirical evidence points to the conclusion that the presence of a catalyst is triggered in parameter setting by inﬂectional morphology or by syntactic cues. In this sense, the category of catalysts is a mixed bag, indicative of morphological arbitrariness, which is incompatible with the notion of strong features. The notion of catalysts is intended to capture the case-by-case nature of head movement as copying of phonological material. The idea that head movement is copying of phonological material onto a head captures the peculiarities listed in (35). Copying of phonological material takes place in phrasal movement as well. The pied-piper feature connects phrasal movement to head movement in terms of the request for phonological material. 3.5

Conclusion

This chapter has explored the possibility that there is an uninterpretable feature that plays a signiﬁcant role in generalized pied-piping. This feature is sometimes realized as a focus particle, at either the left or the right edge of the fronted phrase. It also determines which element of a chain is to be pronounced. Signiﬁcantly, the particle in question is absent in the in-situ agreement dependency, which does not make use of categorial pied-piping. Strong empirical support comes from the loss of whmovement in Old Japanese and from optional wh-movement in Kikuyu. The catalyst theory of head movement developed here rests on the intuition that verb-second arises as a by-product of pied-piping. It says that head movement arises when the request for phonological material is combined with a second factor, which is called a catalyst for the sake of convenience. This theory is intended to capture the generalization that there are no cases where wh-movement suppresses head movement that is otherwise available. The pied-piper feature copied onto the probing head

66

Akira Watanabe

is one instantiation of the request for phonological material. It is a task for future research to determine an exhaustive list. The crucial steps in linking phrasal movement to head movement are copying of the pied-piper feature on the probing head during Agree and the characterization of the pied-piper feature as the request for phonological material. Phrasal movement and head movement, of course, di¤er in signiﬁcant respects. In the case of phrasal movement, more than the phonological content of the pied-piped category is copied, whereas head movement is simply copying of the phonological material of the closest head. This di¤erence can be understood as arising from the way the pied-piper feature works in narrow syntax and in morphophonological computation. Copying of phonological material for head movement takes place after Spell-Out, at which point only PF features remain. Since this chapter is programmatic, many lines of inquiry remain to be pursued. One intriguing question is whether the pied-piper feature can be used independently of Agree. One candidate for such a case is scrambling of the Japanese type, which does not seem to be driven by agreement (see Saito 2003 and the references cited there). One possibility is that scrambling arises when free assignment of the pied-piper feature automatically creates a copy, which must be merged with another syntactic object. Notes I would like to thank the audience at the workshop ‘‘On Wh-Movement’’ as well as Hironobu Kasai, Masakazu Kuno, and a reviewer for helpful discussions and/ or constructive comments. 1. Note that this hypothesis makes Nunes’s (1999) account of PF realization of a chain superﬂuous. If there are cases in which the original is pronounced under certain circumstances (see Bosˇkovic´ 2002 and the references cited there), or if covert phrasal movement after Spell-Out is an option allowed by UG (Chomsky 2004), things are not so straightforward. Note that except for a one-step movement operation, covert phrasal movement is incompatible with the idea that pied-piping requires phonological content. A reviewer points to antecedent-contained deletion (ACD) as robust evidence for covert phrasal movement. But Merchant (2001) has shown convincingly (see also Watanabe 2004a for an additional argument) that ellipsis is PF deletion subject to the semantic identity condition. If so, ACD is not empirical evidence for covert QR as a syntactic operation. In general, semantic considerations do not provide a decisive argument for covert phrasal movement just as a matter of logic. It goes beyond the scope of this chapter to discuss complications of this sort. 2. It is impossible to clarify the situation in the ninth century, owing to the lack of remaining documents.

The Pied-Piper Feature

67

3. The exception is the ‘why’-type adjunct. 4. See Watanabe 2004b for detailed discussion. 5. See Cheng and Rooryck 2000, 2002 regarding non-indeterminate wh-in-situ in French and European Portuguese. 6. Imbabura Quechua has HIRCs, even though it is not a wh-in-situ language (Cole and Hermon 1994). See Watanabe 2004c for a solution to this problem. 7. Interpretable features do not go to PF. Therefore, they should not be morphophonologically realized. 8. The Complex NP Constraint e¤ect is masked in Japanese wh-in-situ because of the option of large-scale pied-piping, which should not be confused with the categorial pied-piping that takes place in phrasal displacement in general. For largescale pied-piping, see Nishigauchi 1990 and the subsequent literature. 9. This particle can appear clause-ﬁnally, too, as in (i) and (32b). (i) Taru-ya nuu kam-yi-ga? Taru-top what eat-pres-Q ‘What does Taru eat?’ This may be one indication that it does not realize the pied-piper feature. 10. And in Romance as well according to Rizzi (1996), though the analysis of Romance is controversial. 11. In Watanabe 1996, I suggest that Lightfoot’s (1991) degree-0 learnability ensures the default speciﬁcation for the embedded interrogative C, as long as the matrix C and the embedded C are speciﬁed independently. The default in the theory proposed here means the absence of a catalyst. References Aoun, Joseph, and Yen-hui Audrey Li. 1993. On some di¤erences between Chinese and Japanese wh-elements. Linguistic Inquiry 24, 365–372. Bergvall, Victoria L. 1983. Wh-questions and island constraints in Kikuyu: A reanalysis. In Jonathan Kaye, Hilda Koopman, Dominique Sportiche, and Andre´ Duga, eds., Current approaches to African linguistics, vol. 2, 245–260. Dordrecht: Foris. Bonneau, Jose´. 1992. The structure of internally headed relative clauses. Doctoral dissertation, McGill University. Bosˇkovic´, Zˇeljko. 2002. On multiple wh-fronting. Linguistic Inquiry 33, 351– 383. Cheng, Lisa L.-S., and Johan Rooryck. 2000. Licensing wh-in-situ. Syntax 3, 1– 19. Cheng, Lisa L.-S., and Johan Rooryck. 2002. Types of wh-in-situ. Ms., Leiden University. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20, 1–52. Cambridge, Mass.: MIT Press.

68

Akira Watanabe

Chomsky, Noam. 1998. Some observations on economy in generative grammar. In Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky, eds., Is the best good enough?, 115–127. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Clements, G. N. 1984. Binding domains in Kikuyu. Studies in the Linguistic Sciences 14, 37–56. Cole, Peter, and Gabriella Hermon. 1994. Is there LF wh-movement? Linguistic Inquiry 25, 239–262. Dikken, Marcel den, and Anastasia Giannakidou. 2002. From hell to polarity: ‘‘Aggressively non-D-linked’’ wh-phrases as polarity items. Linguistic Inquiry 33, 31–61. Isobe, Yoshihiro. 1990. Chuuko-wabun no yoosetumei gimon-hyoogen. NihonBungaku Kenkyuu 26, 165–176. Baiko Jo Gakuin College. Jonas, Dianne. 2002. Residual V-to-I. In David Lightfoot, ed., Syntactic e¤ects of morphological change, 251–270. Oxford: Oxford University Press. Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press. Kishimoto, Hideki. 1992. LF pied-piping: Evidence from Sinhala. Gengo Kenkyuu 102, 46–87. Kondo, Yasuhiro. 1981. Chuukogo-no juntai-kouzou-nituite. Kokugo to Kokubungaku 58(5), 18–31. Kuroda, S.-Y. 1974. Pivot-independent relativization in Japanese (I). Papers in Japanese Linguistics 3, 9–93. Reprinted in Japanese syntax and semantics. Dordrecht: Kluwer, 1992. Lightfoot, David. 1991. How to set parameters. Cambridge, Mass.: MIT Press. Merchant, Jason. 2001. The syntax of silence. Oxford: Oxford University Press. Miyara, Shinshou. 2003. Wh-particle computation. Paper presented at the Workshop on Classical Japanese and Linguistic Theory, Sophia University. Nishigauchi, Taisuke. 1990. Quantiﬁcation in the theory of grammar. Dordrecht: Kluwer. Nunes, Jairo. 1999. Linearization of chains and phonetic realization of chain links. In Samuel David Epstein and Norbert Hornstein, eds., Working minimalism, 217–249. Cambridge, Mass.: MIT Press.

The Pied-Piper Feature

69

Rizzi, Luigi. 1996. Residual verb second and the Wh-Criterion. In Adriana Belletti and Luigi Rizzi, eds., Parameters and functional heads, 63–90. Oxford: Oxford University Press. Roberts, Ian. 1993. Verbs and diachronic syntax. Dordrecht: Kluwer. Rohrbacher, Bernhard. 1999. Morphology-driven syntax: A theory of V to I raising and pro drop. Amsterdam: John Benjamins. Sabel, Joachim. 2000. Partial wh-movement and the typology of wh-questions. In Uli Lutz, Gereon Mu¨ller, and Arnim von Stechow, eds., Wh-scope marking, 409– 446. Amsterdam: John Benjamins. Saito, Mamoru. 2003. A derivational approach to the interpretation of scrambling chains. Lingua 113, 481–518. Sugahara, Mariko. 1996. Shuri Okinawan kakari musubi and movement. In Masatoshi Koizumi, Masayuki Oishi, and Uli Sauerland, eds., Formal Approaches to Japanese Linguistics 2, 235–254. MIT Working Papers in Linguistics 29. Cambridge, Mass.: MIT, Department of Linguistics and Philosophy, MITWPL. Takahashi, Daiko. 1997. Move-F and null operator movement. The Linguistic Review 14, 181–196. Takahashi, Daiko. 2000. Move F and raising of lexical and empty DPs. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 297–317. Cambridge, Mass.: MIT Press. Takahashi, Daiko. 2002. Determiner raising and scope shift. Linguistic Inquiry 33, 575–615. Tsai, Wei-Tien Dylan. 1994. On economizing the theory of A-bar dependencies. Doctoral dissertation, MIT. Tsai, Wei-Tien Dylan. 1999. On lexical courtesy. Journal of East Asian Linguistics 8, 39–73. Vikner, Sten. 1995. Verb movement and expletive subjects in the Germanic languages. Oxford: Oxford University Press. Vikner, Sten. 1997. V 0 -to-I 0 movement and inﬂection for person in all tenses. In Liliane Haegeman, ed., The new comparative syntax, 189–213. London: Longman. Watanabe, Akira. 1992. Wh-in-situ, subjacency, and chain formation. MIT Occasional Papers in Linguistics 2. Cambridge, Mass.: MIT, Department of Linguistics and Philosophy, MITWPL. Watanabe, Akira. 1996. Case absorption and wh-agreement. Dordrecht: Kluwer. Watanabe, Akira. 2000. Feature copying and binding: Evidence from complementizer agreement and switch reference. Syntax 3, 159–181. Watanabe, Akira. 2002. Loss of overt wh-movement in Old Japanese. In David Lightfoot, ed., Syntactic e¤ects of morphological change, 179–195. Oxford: Oxford University Press.

70

Akira Watanabe

Watanabe, Akira. 2004a. The genesis of negative concord: Syntax and morphology of negative doubling. Linguistic Inquiry 35, 559–612. Watanabe, Akira. 2004b. Indeterminates and determiners. In Aniko´ Csirmaz, Youngjoo Lee, and Mary Ann Walter, eds., Proceedings of the Workshop on Altaic Formal Linguistics I, 390–405. Cambridge, Mass.: MIT, Department of Linguistics and Philosophy, MITWPL. Watanabe, Akira. 2004c. Parametrization of quantiﬁcational determiners and head-internal relatives. Language and Linguistics 5, 59–97. Williamson, Janis. 1984. Studies in Lakhota grammar. Doctoral dissertation, University of California, San Diego. Williamson, Janis. 1987. An indeﬁniteness restriction for relative clauses in Lakhota. In Eric Reuland and Alice G. B. ter Meulen, eds., The representation of (in)deﬁniteness, 168–190. Cambridge, Mass.: MIT Press. Yamanishi, Masako. 1987. Daremo kou. Kokugo to Kokubungaku 64(7), 47–66. Zwart, C. Jan-Wouter. 2001. Syntactic and phonological verb movement. Syntax 4, 34–62.

Chapter 4 Pied-Piping, Feature Movement, and Wh-Subjects

4.1

Brian Agbayani

Introduction

This chapter considers the nature of speciﬁers formed by movement, particularly from the perspective of wh-movement. The empirical foundation of this study is the peculiar behavior of wh-subjects in English, which have been argued to not undergo movement (George 1980; Chomsky 1986) and which seem to paradoxically both exhibit and not exhibit the classic diagnostics for wh-movement (Chomsky 1977). At the same time, the chapter addresses the broader question of why speciﬁers are formed at all within a syntactic theory that involves a mechanism of feature checking or Agree at a distance, since these involve relations between features, not whole categories. Minimalist approaches toward overt movement to a speciﬁer position (Spec) have taken it to be a consequence of (1) a requirement that a Spec be created by overt movement to check a strong feature (Chomsky 1995), or (2) simply the property that certain heads require Specs (Chomsky 2000; Lasnik 2001). These approaches (particularly (2)) have the e¤ect of generalizing what is known as the Extended Projection Principle or EPP to all (functional) heads that require a Spec. The goal of the chapter is to o¤er an alternative perspective on the Spec requirement by closely examining the behavior of wh-elements as speciﬁers. I will propose that if a feature is extracted out of a category in overt syntax for checking, then PF requirements force subsequent pied-piping of the category (essentially following a suggestion of Chomsky 1995), unless economy blocks the pied-piping, as I will argue is the case with wh-subjects in English. I will argue that this particular driving force for ‘‘generalized pied-piping’’ of XP to Spec has nothing to do with properties of the functional head targeted by movement, unlike the driving force for feature checking/Agree. Such an analysis of Spec formation via movement is expressed naturally in a

72

Brian Agbayani

‘‘two-chain’’ theory where bare feature movement and subsequent category pied-piping take place in overt syntax. I will argue that there is signiﬁcant evidence for such an approach to movement from the peculiar behavior of wh-subjects in English under the Vacuous Movement Hypothesis, addressed in George 1980 and considered in some detail in Chomsky 1986. Section 4.2 discusses this evidence in detail, and sections 4.3 and 4.4 argue that the facts are captured most naturally in a two-chain theory of overt movement.1 Section 4.5 shows that the two-chain approach naturally captures some otherwise mysterious e¤ects on semantic interpretation. Finally, section 4.6 considers some implications and questions raised by the analysis. 4.2

Wh-Subjects

4.2.1 The Vacuous Movement Hypothesis In contrast to nonsubject wh-questions in English, subject questions have been notoriously ambiguous as to whether they involve overt whmovement or not. An analysis either with movement (2a) or without movement (2b) would yield the right linear order of elements. (1) Who left? (2) a. [CP who i [IP t i left]] b. [CP [IP who left]] Option (2b) has been explored in the form of the Vacuous Movement Hypothesis (VMH), ﬁrst proposed by George (1980) and adopted by Chomsky (1986). The basic intuition of the VMH is stated in (3). (3) Vacuous Movement Hypothesis A wh-subject does not move locally to Spec,CP. The surface order appears to verify such an analysis. In nonsubject questions, the wh-element is clearly displaced, and auxiliary inversion is observed (4a–b). Subject questions lack both a clear indication that the wh-element has been displaced and a clear indication that inversion has occurred (4c). Moreover, nonemphatic subject questions conspicuously lack do-support (4d). (4) a. b. c. d.

[CP What j has i [IP Kim t i bought t j ]]? [CP Howj has i [IP Robin t i ﬁxed the car t j ]]? [CP [IP Who has ﬁxed the car]]? [CP [IP Who ﬁxed the car]]?

Pied-Piping, Feature Movement, and Wh-Subjects

73

Given the linear pattern, the logic of the VMH becomes clear in light of language acquisition: if a child acquiring the language does not have physically detectable evidence for the application of a transformation in a particular structure, then it is plausible that the child does not posit the application of the transformation in that structure. The following subsections consider the VMH in (3) in relation to facts that are independent of the linear order of elements. 4.2.1.1 Clefts and Relatives The option of either an overt or a null whoperator is generally available for nonsubject clefts (5a) and relative clauses (5b). (5) a. It is Kim [CP who/Op [IP Pat likes t]] b. Kim met a student [CP who/Op [IP Robin knows t]] Subject clefts and relatives typically do not allow the null operator variant, however. (6) a. *It is Kim [Op likes Pat] b. *Kim met a student [Op knows Robin] If a wh-operator lands in Spec,CP in both nonsubject and subject clefts and relatives, then there would be no straightforward way to capture the asymmetry between (5) and (6). However, if we adopt the VMH, the absence of the null operator option in subject clefts and relatives has a possible independent source. (7) a. It is Kim [CP [IP who/*Op likes Pat]] b. Kim met a student [CP [IP who/*Op knows Robin]] As (7) illustrates, the wh-operator does not move to Spec,CP. Op is forced to remain in the subject position. It is plausible that the absence of the null operator option is linked to the absence of phonetically null pro in Spec,IP. The unavailability of the null operator option in English subject clefts/relatives may be linked to the impossibility of null subjects in ﬁnite clauses.2 4.2.1.2 Ellipsis Ellipsis with an auxiliary as a remnant is impossible in main clause nonsubject questions in English. (8) Pat will insult someone a. Who [Pat will insult t] ? b. *Who will [Pat insult t] ?

74

Brian Agbayani

(8a–b) involve sluicing/IP-ellipsis. The contrast shows that auxiliary inversion is excluded in sluicing contexts (see also Boeckx and Stjepanovic´ 2001; Lasnik 2001). An unexpected asymmetry arises, however, with respect to main clause subject questions. (9) Someone will arrive early a. Who [will arrive early] ? b. Who will [arrive early] ? (9b) is ﬁne, in contrast to (8b). Notice that both (8a) and (9a) involve sluicing, the auxiliary being deleted along with other material embedded in IP. We run into problems if (9b) is analyzed as a case of sluicing too, since the presence of the auxiliary as a remnant after sluicing is generally prohibited. The problem does not arise if we take (9b) to be a case of VPellipsis, and—following the VMH in (3)—the wh-subject does not move and there is no inversion. (10) [CP [IP Who will [ VP arrive early] ]]? The VP-ellipsis represented in (10) is licensed under sisterhood with a nonnull functional head (Lobeck 1990, 1995). This is possible only if the auxiliary does not undergo inversion. The availability of VP-ellipsis with main clause subject questions, but not with nonsubject questions, is consistent with the VMH in (3). If the wh-subject remains in situ, inversion does not apply.3 4.2.1.3 Local Topicalization Lasnik and Saito (1992) point out that subjects do not undergo local topicalization. The contrasts in (11) illustrate the inappropriateness of the typical intonational break following a topic that is the subject. (11) a. Kim, I like t b. *Kim, t left c. Kim thinks that Pat, Robin likes t d. *Kim thinks that Robin, t likes Pat If topicalization is analyzed as movement to Spec,CP, then it too is subject to the VMH, yielding the surface representation (12b) for (11b). (12) a. [CP Kim [IP I like t]] b. [CP [IP Kim left]]

Pied-Piping, Feature Movement, and Wh-Subjects

75

4.2.2 Potential Counterevidence Paradoxically, there appears to be substantial counterevidence to the VMH in (3). All of the following evidence suggests that the wh-subject occupies Spec,CP on the surface. 4.2.2.1 Clausal Typing Cheng (1991) argues that in languages with overt wh-movement, Spec,CP must contain a wh-element (where whelements include wh-subjects). This follows from the Clausal Typing Hypothesis. (13) Clausal Typing Hypothesis (Cheng 1991) A wh-element or a particle must be present in CP to type a whquestion. According to Cheng, the Clausal Typing Hypothesis (CTH) takes e¤ect at S-Structure. It predicts that if a language lacks overt wh-movement, it will employ a particle in the CP system in interrogative sentences. It follows that the wh-subject must move to Spec,CP in English—otherwise, subject questions would not be properly typed as interrogative.4 In the contemporary minimalist framework, where the level of SStructure plays no part, one could argue that the CTH takes e¤ect only at LF, since clausal typing would seem to be principally a matter of semantic interpretation. Such an approach would allow the wh-subject to remain in the subject position in overt syntax, then raise covertly to satisfy the CTH. This works if the VMH in (3) is interpreted as relevant for overt wh-movement only. However, if the CTH is relevant for LF only, we lose the generalization captured by Cheng: languages that lack overt wh-movement exhibit an interrogative particle in CP, and languages that lack a particle exhibit wh-movement. This is so because the CTH (relevant for LF only) predicts an apparently nonexistent language with obligatory wh-in-situ and covert insertion of a particle in (both yes/no and wh-) interrogatives, which would have no phonological e¤ect.5 This argues that the CTH is relevant for PF as well as LF. Thus, clausal typing remains a problem for the VMH as stated in (3).6 4.2.2.2 Wh-Islands Like nonsubject wh-elements, wh-subjects induce island e¤ects, as the examples in (14) illustrate. (14) a. *What i does Kim wonder [CP who bought t i ]? b. *Which country i did Pat meet a student [CP who visited t i ]? Under a traditional Subjacency account of island e¤ects, the wh-element in the matrix CP needs to have raised successive-cyclically to the

76

Brian Agbayani

embedded CP, then to the matrix clause. Successive-cyclic movement should be possible if the wh-subject in the embedded clause is not in Spec,CP. However, the relevant interpretations are blocked. The whsubject who appears to be in Spec,CP, thereby blocking successive-cyclic movement of what/which country.7 Thus, the existence of these island e¤ects is problematic for the VMH in (3). 4.2.2.3 Sluicing Like nonsubjects, wh-subjects are sluicing remnants, as shown in (15b). (15) a. I heard Pat insulted someone. Tell me who [Pat insulted t] b. I heard someone left early. Tell me who [left early] c. I heard someone left early. *Tell me [who left early] This is problematic for the VMH in (3) if sluicing is wh-movement followed by deletion of IP. The VMH predicts that (15c) should be the only possibility, since the wh-subject remains in Spec,IP. However, the wh-subject behaves like a nonsubject with respect to being a sluicing remnant. 4.2.3 Summary This section has explored evidence for and against the VMH as stated in (3). The linear characteristics of subject questions, the unavailability of null operators in subject clefts and subject relatives, and the inability of subjects to undergo local topicalization all argue for the VMH. Arguments against this hypothesis include the issue of clausal typing at PF, the inducement of island e¤ects by wh-subjects, and the behavior of whsubjects under sluicing. This paradox is problematic for any ‘‘strong’’ VMH approach like that proposed in (3). Consider in this respect Radford’s (1997, 293–294) hypothesis that the deﬁning characteristic of an interrogative clause is that it must contain an interrogative speciﬁer in order to be interpreted as a question at LF. Under Radford’s analysis, subject wh-questions in English consist of just a TP. The wh-subject occupies Spec,TP, and TP satisﬁes the requirement that a question contain an interrogative speciﬁer. Radford claims that in this case, there is no need to project the structure any further into CP and, by economy considerations, no possibility of doing so. However, considerations of economy may in fact preclude such an analysis. Under the interrogative speciﬁer analysis, the numeration/ lexical array from which a subject question is built must not contain C (i.e., C must not be selected in the formation of the numeration/array).

Pied-Piping, Feature Movement, and Wh-Subjects

77

If we assume that economy conditions operate locally (Collins 1997; Chomsky 1998), then a global property such as ‘‘look-ahead’’ is to be avoided in favor of local determination of economy. It follows that the computational system makes use of locally determined solutions rather than comparing the numerations/arrays of alternative convergent derivations. With respect to subject questions, the choice not to select C must be based on the eventual output of the derivation (i.e., LF convergence as an interrogative structure), involving ‘‘look-ahead.’’ This argues in favor of positing that C is present and projects in both subject and nonsubject wh-questions. Moreover, the fact that whsubjects are sluicing remnants suggests that in such cases, the wh-subject occupies a position external to IP (which is subject to ellipsis) and provides further evidence against the interrogative speciﬁer approach. Note that these issues also arise for any approach that simply requires that a wh-element be in a clause edge position, rather than the stronger condition that it occupy Spec,CP (as would be the case in an approach that eschews ‘‘absolute’’ structural positions; see, for example, Epstein 1999). The link between the conception of the EPP—the notion that some head requires a speciﬁer—and the behavior of wh-subjects becomes apparent when we consider the nature of (XP-)movement in syntax from the perspective of feature movement and PF convergence. A resolution of the ‘‘VMH paradox’’ will entail an alternative conception of phrasal movement, which we explore in the next section. 4.3

In Spec,CP or Not?

This section will explore three conceptions of phrasal movement in syntax. The ﬁrst is the traditional notion of Move a (where a, an entire phrasal category, raises to Spec). I will argue that such an approach crucially does not resolve the VMH paradox mentioned in the previous section. The second approach is Chomsky’s (2000) notion of Agree at a distance followed by ‘‘second Merge’’ (which also raises an entire phrasal category to Spec). I will argue that the ‘‘second Merge’’ portion of the operation also crucially does not resolve the paradox. I will argue instead for an approach to overt movement that involves two separate operations: Move F, which moves the feature to be checked to the domain of the attracting head H, and Pied-Pipe, which raises, to Spec,H, the category of which the checked feature is a proper subpart. This third approach resolves the VMH paradox in a straightforward way.

78

Brian Agbayani

4.3.1 Move a The traditional conception of XP-movement, illustrated in (16), is that it is movement of a phrasal category a to the Spec of a c-commanding functional head H. a—a category containing (under standard assumptions) formal, semantic, and phonological features—raises to Spec,H to enter into speciﬁer-head agreement with H. (16)

Such an ‘‘all-or-nothing’’ approach to XP-movement gives rise to the VMH paradox. Under this approach, the wh-subject is forced to occupy either Spec,IP or Spec,CP, and the paradoxical facts discussed in section 4.2 suggest that it needs to be located in both the IP and CP systems simultaneously. 4.3.2 Agree at a Distance B ‘‘Second Merge’’ For Chomsky (2000), XP-movement involves a two-step process, illustrated in (17). Feature checking is mediated by the operation Agree, which applies at a distance between a feature of a c-commanding head H (a probe) and a feature (e.g., wh, f) of an XP category a that is ccommanded by H (a goal). a is then merged as a speciﬁer of H to eliminate the EPP feature of H. Thus, movement to Spec (or ‘‘second Merge’’) is independent of the checking of formal features under Agree. (17)

The second Merge step of the derivation is problematic from the point of view of the VMH. It faces the same problem as the speciﬁer-head agreement approach in that the category is forced to occupy Spec,HP.

Pied-Piping, Feature Movement, and Wh-Subjects

79

One could attempt to derive the e¤ect of the VMH by simply saying that the EPP property is inactive in the case of wh-subjects, but this raises the question of why the EPP property would be allowed to be inactive only in such constructions. It could be argued that C is simply not inserted in overt syntax in questions with wh-subjects, so that all relevant movements are done in covert syntax and the clause at Spell-Out is simply TP, with no CP structure projected above it. However, this gives rise to the problems mentioned at the end of section 4.2 with respect to the interrogative speciﬁer analysis. We are thus left with an unresolved paradox under this approach as well. It is important to note that in this context, Chomsky (2000) rejects the notions of feature movement and feature chains advocated in previous work (e.g., Chomsky 1995). He states: In the absence of clear evidence to the contrary, I will assume that feature chains do not exist, hence that features cannot move or be attracted. (2000, 119) In MP [Chomsky 1995] Agree is analyzed in terms of feature movement (Attract) and a concept of matching that is left unclear. Here we take matching to be identity and dispense with Attract, with complications it induces about extended MLIs [modiﬁed lexical items], feature chains, and other matters. (2000, 123)

In what follows, I revisit the notion of feature movement and suggest that the VMH paradox discussed in section 4.2 constitutes evidence in favor of the feature movement approach. The following section ﬁrst outlines the two-chain theory, which incorporates feature movement, and then applies it to the observed behavior of wh-subjects. 4.4

The Two-Chain Theory and the Vacuous Movement Hypothesis

4.4.1 Displacement in Overt Syntax I explore the notion that overt displacement is the product of a two-step procedure: (18) a. Move F The feature F (to be checked) of category a is extracted out of a and moves to the domain of a functional head H; F enters into a checking relation with an uninterpretable feature of H. b. Pied-Pipe Category a is pied-piped to Spec,H. Economy requires only Move F for the purpose of feature checking. If checking requires only the feature F (e.g., wh- or f-features) of category

80

Brian Agbayani

a, then derivational economy should favor an operation (which mediates checking) that a¤ects only F and not its entire category a. Thus, if there are no other considerations outside of feature checking, only Move F should apply, yielding no physically detectable e¤ects of displacement. Pied-Pipe—which I argue follows the application of Move F—must be forced by some PF requirement(s); covert movement, for which PF considerations do not hold, therefore involves only Move F. The two-step procedure of Move F followed by Pied-Pipe is illustrated in (19). (19) a.

b.

Step (a) of (19) represents the application of Move F. H with an uninterpretable feature induces the matching feature F to raise out of category a once H is merged. I assume that when F is checked against the uninterpretable feature of H, the uninterpretable feature of H undergoes erasure, whereas F does not. F is generally interpretable at LF for the category out of which it was extracted (e.g., f-features of an NP, such as person or number features, or an interrogative/wh-feature of a wh-word, which functions for clause typing and yields the scope of the wh-word), so F is not subject to erasure since it yields certain semantic e¤ects. Note that a ‘‘category’’ is understood as a set containing phonological, semantic, and formal feature matrices. I assume that lexical items enter the derivation from the lexicon with all of their features intact (as opposed to what happens under Distributed Morphology, where phonological features of lexical items are introduced later in the derivation (Halle and Marantz 1993)). I use the term category to refer both to a lexical item present in the structure and to a constituent that is composed of distinct lexical items. Note also that (19a) represents feature movement as targeting the root, whereby it creates a Spec of HP, as opposed to adjoining to the head H (cf. Chomsky 1995, 270). This is conceptually desirable, as it allows for all movement operations to be analyzed as cyclic, roottargeting operations (see Fukui and Takano 1998 for arguments along the same lines).

Pied-Piping, Feature Movement, and Wh-Subjects

81

Though su‰cient for checking, the separation of F and its category in overt syntax gives rise to problems at PF. Properties of the phonological component therefore require F’s category to pied-pipe as well. Chomsky (1995) characterizes the situation as follows: For the most part—perhaps completely—it is properties of the phonological component that require such pied-piping. Isolated features and other scattered parts of words may not be subject to its rules, in which case the derivation is canceled; or the derivation might proceed to PF with elements that are ‘‘unpronounceable,’’ violating FI [Full Interpretation]. (1995, 262–263)

Under this approach, Chomsky suggests that overt movement is composed of a complex of chain-forming operations: Applied to the feature F, the operation Move thus creates at least one and perhaps two ‘‘derivative chains’’ alongside the chain CHF ¼ (F; tF ) constructed by the operation itself. One is CHFF ¼ (FF[F]; tFF[F] ), consisting of the set of formal features FF[F] and its trace; the other is CHCAT ¼ (a; ta ), a a category carried along by generalized pied-piping and including at least the lexical item LI containing F. CHFF is always constructed, CHCAT only when required for convergence. . . . CHCAT should be completely dispensable, were it not for the need to accommodate to the sensorimotor apparatus. (1995, 265)

Thus, if F is extracted in overt syntax, its category must pied-pipe; otherwise, the resulting structure will be defective for PF. Given the statement quoted immediately above, there appears to be an assumption that in overt syntax the formal features remain as part of the category (Cat) that pied-pipes. However, there is another possibility: that F and Cat move separately (Move F ﬁrst, followed by pied-piping of Cat) to the domain of H. I refer to the operation that raises Cat as Pied-Pipe (here I do not discuss the role—if any—of CHFF ). Pied-Pipe raises Cat to a position where it is in a local relation with its formal features in the domain of head H. I propose that the relation holding between F and Cat is PF adjacency, a local relation required at the PF interface. PF adjacency is deﬁned simply in (20). (20) X and Y are PF adjacent if no phonological features intervene between X and Y. ‘‘Intervention’’ in this case is deﬁned in terms of linear order: if X precedes both Y and Z, and Z precedes Y, then Z intervenes between X and Y. Pied-Pipe applied to Cat therefore forms a Spec in HP (it targets the root), making F and Cat adjacent. Both Move F and Pied-Pipe apply cyclically and in a local fashion: F is moved only if there is an uninterpretable formal feature to be checked at the root category; Cat raises to

82

Brian Agbayani

the root category only if its feature F has been extracted, and F and Cat are nonadjacent. Both operations, Move F and Pied-Pipe, target the root category, creating multiple Specs. This was shown in (19), repeated here. (19) a.

b.

This approach to movement potentially allows the featural content of a category to be in two places at once, which is what we need to resolve the VMH paradox. Let us redeﬁne the VMH as follows: (21) Vacuous Movement Hypothesis Move F applies; Pied-Pipe is blocked. Following the application of Move F prior to Spell-Out, we have the conﬁguration in (22). In (22), C is phonetically null and the arrow represents only the extraction of the wh-feature. (22)

In (22), Pied-Pipe is blocked by a simple economy condition: Avoid superﬂuous steps. Who and its wh-feature are adjacent in (22), since no phonological material intervenes between them. Thus, adjacency between the wh-feature and who holds without the application of Pied-Pipe, so the operation is blocked in this environment.8 Note that an EPP feature is unnecessary under this approach. PF considerations replace it as the driving force for overt category displacement.

Pied-Piping, Feature Movement, and Wh-Subjects

83

The following section will argue that the VMH paradox is resolved only under such an approach. 4.4.2

The Vacuous Movement Hypothesis Paradox Revisited

4.4.2.1 Clausal Typing The approach outlined in section 4.3.3 suggests that clausal typing is achieved principally by way of Move F, and only secondarily by Pied-Pipe, since the latter operation is dependent on the former. Let us then redeﬁne the CTH as in (23). (23) Clausal Typing Hypothesis (II) A wh-feature must be checked with C to type a wh-question. This seems plausible, since the interrogative property of C will be visible both at LF and at PF (by virtue of adjacency between an overt whcategory and the wh-feature that is in the domain of C). 4.4.2.2 Minimality Islands Recall that a wh-subject creates a wh-island, suggesting that it has moved to Spec,CP. Under the Move F approach, the feature to be checked is raised to the domain of C, and following Chomsky (1995) we assume that Move F is subject to locality restrictions on feature chains (i.e., the Minimal Link Condition). The locality restriction on feature chains restricts a more distant wh-feature from being moved past a feature closer to the target C. (24) a. *What i did Kim wonder [CP [Fwho ] [C 0 C [IP who bought t i ]]]? b. *Where i did Pat meet a student [CP [Fwho ] [C 0 C [IP who ate lunch t i ]]]? As (24) illustrates, the category of the wh-subject need not raise to Spec,CP; the island facts fall out from the behavior of Move F alone. 4.4.2.3 Condition on Extraction Domain Islands Pied-Pipe apparently is not constrained by the locality restriction on feature chains, and the question arises what restrictions, if any, it does obey. Recall that PiedPipe applies prior to Spell-Out only, so any island e¤ects induced by Pied-Pipe should hold of overt movement but not of covert movement. In fact, Pied-Pipe appears to respect subject and adjunct islands (both classic non-Subjacency or Condition on Extraction Domain islands), which show this very asymmetry. It is well known that these ‘‘domain barriers’’ fall outside of the locality restriction on feature chains (see Agbayani 1998a,b and Ochi 1999a,b for discussion of these facts under the two-chain approach).

84

Brian Agbayani

(25) a. Subject islands i. *Who(m) i did Kim say that [for Pat to marry t i ] would be a surprise? ii. Who said that [for Pat to marry who(m)] would be a surprise? b. Adjunct islands i. *Who(m) i did Kim get angry [because Robin insulted t i ]? ii. Who got angry [because Robin insulted who(m)]? As predicted, the island e¤ects are observed with overt movement only. 4.4.2.4 Sluicing Recall that the problem related to sluicing involved the presence of the wh-subject as a remnant. If the wh-subject remained in Spec,IP, it should be deleted along with the rest of the IP. Under the two-chain approach, the wh-feature of the subject raises to the domain of C. If Pied-Pipe does not apply prior to IP-deletion, then the wh-feature and the subject will not be adjacent at PF because the subject is gone. The stranded wh-feature is unpronounceable on its own, violating Full Interpretation at PF and therefore causing a PF crash. (26) a. I heard someone left early. Tell me [CP who i [C 0 [Fwho ] [C 0 C [ IP t i left early] ]]] b. I heard someone left early. *Tell me [CP [Fwho ] [C 0 C [ IP who left early] ]] (26b) fails because F is stranded at PF. 4.5

The View from LF

This section considers further evidence for the two-chain view of movement—speciﬁcally, facts bearing on LF interpretation that are contingent upon the application/nonapplication of the operation Pied-Pipe. 4.5.1 Evidence from Raising and ‘‘Expletive Replacement’’ Some intriguing facts that bear on this treatment of overt movement arise in the case of anaphor binding and negative polarity item (NPI) licensing. Consider the following examples from Lasnik 1995: (27) a.

[Some linguists] i seem to each other t i to have been given good job o¤ers b. *There i seem to each other t i to have been some linguists given good job o¤ers

Pied-Piping, Feature Movement, and Wh-Subjects

85

(28) a.

[No good linguistic theories] i seem to any philosophers t i to have been formulated b. *There i seem to any philosophers t i to have been no good linguistic theories formulated

These examples show a contrast between raising and insertion of expletives with respect to reciprocal binding and NPI licensing. (27a) and (28a) clearly involve overt raising of some linguists and no good linguistic theories, whereas it is assumed that (27b) and (28b) involve covert feature movement to the neighborhood of the expletive (Chomsky 1995; Lasnik 1995). Lasnik, assuming that anaphoric interpretation and NPI licensing are governed by output conditions at LF, argues that the deviant status of examples like (27b) and (28b) can be explained if features of the associate related to anaphora/NPI licensing remain in a lower position than the expletive, even at LF. On the other hand, in (27a) and (28a) the relevant licensing features raise to a higher position. If the formal features that undergo checking are not responsible for anaphora/NPI licensing, then the contrasts above are easily captured. In (27b) and (28b), the category of the associate never raises to the expletive, even though a feature of the associate has moved for checking purposes. The examples show that the (non)application of Pied-Pipe may have an e¤ect that is visible to LF, and they suggest that Pied-Pipe a¤ects an object that is distinct from the object a¤ected by Move F.9 4.5.2 Embedded Topics According to Lasnik and Saito (1992), local topicalization of subjects is not possible. This asymmetry likely falls under the VMH, assuming that topics move to Spec,CP. In light of the two-chain approach to overt movement, consider the following contrast, also noted by Lasnik and Saito (1992): (29) a. John thinks that himself, Mary likes t b. *John thinks that himself, t likes Mary The contrast is very clear for those who accept embedded topicalization. In (29a), the reﬂexive, topicalized from the object position in the embedded clause, takes the matrix subject as its antecedent (cf. the nontopicalized counterpart *John thinks that Mary likes himself ). However, if the reﬂexive is a topicalized subject as in (29b), it cannot take the matrix subject as its antecedent. I assume that a CP-recursion structure is present in the embedded clauses in (29), as suggested for embedded topicalization by Authier

86

Brian Agbayani

(1992). (29a) shows that pied-piping the reﬂexive to Spec,CP of the embedded clause allows the reﬂexive to ‘‘ﬁnd’’ its antecedent in the matrix clause. If local subject topicalization is blocked by the VMH, then (29b) is on a par with (30), where the reﬂexive is in the subject position of the embedded clause. (30) *John thinks that himself likes Mary Under the classical formulation of the VMH in (3), it is not clear why the subject should remain in situ. Under the two-chain analysis, economy forces the subject category in (29b) to stay in the subject position, even though the feature to be checked has moved to C in overt syntax. Features relevant for binding are pied-piped in (29a) and are therefore high enough to ﬁnd a local binder in the matrix clause. If the features that mediate binding are not part of the feature(s) checked, then the contrast in (29) again suggests that the location of the category, and not of the formal features, a¤ects binding (contra Chomsky 1995, 272). Note, ﬁnally, that the semantic e¤ects derived by the (non)application of Pied-Pipe conﬁrm that Pied-Pipe applies in narrow syntax and not in the post-Spell-Out derivation to PF. 4.6

Questions and Implications

A number of implications and new questions for further study arise if this analysis is on the right track. First, does the pattern predicted for wh-subjects in English occur in any language where phonologically null C and Spec,IP are linearly adjacent, or is there observable variation with respect to this pattern? For example, if a language allows for ‘‘relative pronoun deletion’’ of subject wh-elements in relative clauses and related constructions, then arguably the wh-subject has raised to Spec,CP in overt syntax. The same language should also allow subject topicalization. In any case, the analysis presented here provides for clear empirical tests of its validity.10 Second, note again that this analysis is silent about why there is no inversion in root subject wh-questions. One view consistent with the present approach would be that inversion in root wh-questions is triggered by the presence of a wh-phrase in Spec,CP. This view takes inversion to be an ancillary property of root wh-questions, perhaps applying in the post-Spell-Out derivation to PF (see Chomsky 2000, 149n68 and Boeckx and Stjepanovic´ 2001 for a suggestion along these lines for head movement).

Pied-Piping, Feature Movement, and Wh-Subjects

87

Third, the question also arises of how the adjacency requirement is satisﬁed in phrasal pied-piping cases, illustrated in (31), with [a picture of whose mother] in Spec,CP. (31) [CP [a picture of whose mother] [C 0 F [C 0 C . . . ]] One possibility, rooted in classical proposals for phrasal/clausal piedpiping, is that the wh-feature of whose percolates to the head of the upper NP prior to the application of Move F, so that the category associated with the wh-feature is the entire phrase a picture of whose mother. A related question arises concerning what rules out left branch extractions like *Whosei did you meet ti mother? Such facts suggest that the object affected by Pied-Pipe must be a maximal phrase and not a head, indicating a dichotomy between head movement and XP-movement that has also been noticed in recent work (in particular, Lasnik (2001) points out that the failure to apply head movement may be remedied by deletion, but such is not the case for XP-movement, suggesting that head movement is a PF phenomenon that interacts with other PF rules). Finally, ‘‘partial’’ wh-movement constructions such as those found in German may lend further support for the two-chain analysis. In these constructions, the matrix Spec,CP is occupied by an invariant wh-scope marker, with the actual wh-phrase occupying the Spec of the embedded clause. This is shown in (32) (from McDaniel 1989, 569). (32) Was glaubt Hans [CP [PP mit wem] i [IP Jakob jetzt t i spricht]]? wh thinks Hans with whom Jakob now talks ‘With whom does Hans think that Jakob is now talking?’ Cheng (2000) has analyzed such constructions as involving feature movement to the matrix CP and partial pied-piping of the wh-phrase to the embedded CP. The wh-feature in the matrix clause is spelled out as was (an invariant scope marker, homophonous with the wh-word was ‘what’), and the wh-phrase may actually take matrix scope (via the raised wh-feature), even though the category itself is not raised into the matrix clause. Under Cheng’s analysis, the wh-feature ﬁrst moves into the embedded CP, and this triggers pied-piping of the category. The whfeature then moves on to the matrix CP, where it is spelled out. This subsequent feature movement leaves the wh-phrase behind in the embedded CP. Given such an analysis, the existence of ‘‘partial’’ wh-movement, like the behavior of English wh-subjects, suggests that feature movement may indeed apply and that it may apply independently of category piedpiping in overt syntax.

88

4.7

Brian Agbayani

Conclusion

I have argued for a two-chain approach to overt movement that involves two operations: Move F and Pied-Pipe. The nature of each operation is summarized as follows: (33) Move F satisﬁes the feature-checking needs of the head H, which contains an uninterpretable feature. (34) Pied-Pipe raises an XP category to the domain of extracted F to satisfy the PF condition on scattered features. Move F and Pied-Pipe apply in the narrow syntax. Pied-Pipe creates a conﬁguration where a category and its extracted feature are PF adjacent, thus remedying the PF defectiveness of feature scattering. If the category is adjacent to the checked feature without the application of Pied-Pipe, then this operation is blocked by economy. This yields the pattern observed with wh-subjects in English under the Vacuous Movement Hypothesis, where some of the diagnostics of wh-movement are observed (e.g., it obeys Subjacency), but others are not (e.g., it apparently doesn’t leave a ‘‘gap’’/trace). Note that if the statement in (34) is correct, then the ‘‘Spec requirement’’ (i.e., ‘‘EPP’’) is not a requirement of an ‘‘attracting’’ functional head H, but a derivative property based on requirements for PF convergence. I have argued here that the peculiar properties of the subject in English provide evidence for such a two-chain analysis of overt movement. Notes I would like to thank the participants at the workshop ‘‘On Wh-Movement’’ for valuable discussion and comments. In particular, I’d like to thank Lisa Cheng, Norbert Corver, Jeroen van Craenenbroeck, Kleanthes Grohmann, Toru Ishii, Howard Lasnik, Ineke van der Meulen, Joe Trotta, Akira Watanabe, and an anonymous reviewer for valuable discussion on the ideas presented here. Of course, any errors and shortcomings are my own. This work was supported in part by a faculty development grant provided by California State University, Fresno. 1. This chapter extends and sharpens the analysis of wh-subjects pursued in previous studies, particularly Agbayani 1998a, 2000. 2. A reviewer points out that in dialects of English that allow subject contact relatives—such as Belfast English (Henry 1995; see also Doherty 1993)—subject clefts without an overt wh-element or complementizer are also allowed. The following examples from Belfast English are from Henry 1995, 124–125:

Pied-Piping, Feature Movement, and Wh-Subjects

89

(i) Subject contact relatives a. I have a sister lives in Dublin b. It’s always me pays the gas bill (ii) Subject clefts a. It was John told us about it b. It was one of my friends won the prize Doherty (1993) suggests that subject contact relatives are IP structures formed without movement. Rather, the ‘‘head’’ noun binds a resumptive pro in the subject position of the clause. Henry (1995) suggests that such constructions are ‘‘root-type’’ clauses that contain a null pro bound by a base-generated topic. Importantly, these authors seem to agree that subject contact constructions in Belfast English are not formed by (wh-)movement; if these constructions in fact do not involve movement (and I will assume they do not), then they do not stand as counterexamples to the claims made here. 3. This analysis might predict that in embedded nonsubject questions, where inversion does not apply generally, we should expect to ﬁnd VP-ellipsis. VP-ellipsis is blocked in these contexts, however. (i) a.

Pat will insult someone, but I don’t know who [Pat will insult t] (sluicing OK) b. *Pat will insult someone, but I don’t know who she will [insult t] (VP-ellipsis not OK)

Note that (ib) contains a nonsubject question; but if a subject question is embedded, VP-ellipsis is possible on a par with (9b). (ii) Someone will arrive early, but I don’t know who will [arrive early] This may suggest that although licensing under sisterhood with a nonnull functional head is necessary for the availability of ellipsis, it is not su‰cient in these contexts. In these embedded contexts, there appears to be an added requirement that the order of elements be parallel with the order in the left-hand conjunct clause under coordination. 4. Rizzi’s (1990, 1996) Wh-Criterion, stated in (i), also forces a wh-subject to raise locally to Spec,CP to license [þwh] C. (i) Wh-Criterion (Rizzi 1996, 64) a. A wh-operator must be in a speciﬁer-head conﬁguration with X 0 . [þwh] b. An X 0 must be in a speciﬁer-head conﬁguration with a wh-operator. [þwh] 5. Cheng points out that wh-in-situ languages that do not have an interrogative particle in wh-questions will nevertheless have one in yes/no questions. Generally, though, in languages with overt wh-movement, such a particle is lacking in both yes/no and wh-questions. However, there are apparently cases of optional wh-in-situ in a language that otherwise generally fronts its wh-phrases (and lacks an interrogative particle), as has been argued by Bosˇkovic´ (2000) for French. The cases discussed by Bosˇkovic´ involve phonetically null interrogative C lexically inserted covertly, which triggers

90

Brian Agbayani

covert wh-movement but allows wh-in-situ on the surface. French generally exhibits overt wh-movement, where it is assumed that interrogative C is lexically inserted prior to Spell-Out. However, what seems to be unattested is a language that exclusively uses the French-style wh-in-situ strategy (with covert insertion of interrogative C). 6. A reviewer suggests that if one accepts a less rigid approach to phrase structure, then clausal typing should address ‘‘relative’’ positions such as the clause edge (rather than ‘‘absolute’’ positions such as C or Spec,CP; see, for example, Epstein 1999). Under such an approach, subject questions in English need not project a CP, presumably unlike nonsubject questions. However, I will assume here that it is necessary to maintain, for both nonsubject and subject questions, that C is present as an attractor for the wh-feature. Thus, all wh-questions (both matrix and embedded) contain C, which mediates clausal typing (see section 4.2.3 for further discussion). 7. The ‘‘*’’ judgments given in (14), shared by me and other English speakers I have consulted, are at odds with a claim to the following e¤ect made in Chomsky 1986, where Chomsky adopts the VMH: If we adopt [the VMH], then there are a variety of consequences; for example, wh-island e¤ects will be removed for embedded wh-subjects, as in [(i)]: [(i)] what do you wonder [CP who saw t] Here what can move from its D-Structure position t to the speciﬁer of the embedded CP, which is not occupied by who, and then to the matrix speciﬁer position. (pp. 48–49)

Chomsky then argues that although the wh-subject remains in situ at S-Structure, it obligatorily raises to Spec,CP at LF. According to Chomsky, movement in (i) should not violate an island, as predicted by the VMH in (3) (see also Chung and McCloskey 1983). However, the judgment I have found after surveying a number of speakers is that this example and examples like it are quite degraded, completely on a par with violations of islands formed by nonsubject wh-elements. 8. This account predicts that a language with ‘‘doubly ﬁlled Comps’’—where an overt complementizer appears with a wh-phrase in its Spec—should force piedpiping of the wh-subject’s category. This is because the phonologically realized complementizer would otherwise intervene between the extracted wh-feature and the subject category. This is conﬁrmed by data from Dutch with embedded whquestions and relative clauses (from the Belgian Wambeek and Lier dialects, respectively; Jeroen van Craenenbroeck, pers. comm.). (i) a. Ik vruig ma af wou da mennen boek gepikt eit I ask myself o¤ who that my book stolen has ‘I wonder who stole my book.’ b. Ik weet nie mieje welk masken da ma da gezeid eit I know not anymore which girl that me that told has ‘I don’t recall which girl told me that.’ (ii) die man die dat da verhaal verteld ee that man who that that story told has ‘that man who told that story’

Pied-Piping, Feature Movement, and Wh-Subjects

91

In this case, the wh-subjects behave exactly like nonsubject wh-elements and do not exhibit the VMH properties. 9. The preceding runs contrary to Chomsky’s (1995, 272) claim that ‘‘FF(LI) [the formal features of a lexical item] includes the categorial feature of the nominal phrase and should have argument (A-position) properties, including the ability to serve as a controller or binder.’’ Chomsky points to the following contrast: (i) a. The DA [proved [the defendants to be guilty] during each other’s trials] b. *The DA [proved [that the defendants were guilty] during each other’s trials] According to Chomsky, covert raising of the defendants in (ia) allows the binding relation to be established with the anaphor each other. The contrast is unexpected given the facts in (27b) and (28b), where the moved feature does not license the anaphor or the NPI. It is important to note that a predicate-raising analysis of there-sentences may resolve the paradox (see Hoekstra and Mulder 1990; Moro 1991; Zwart 1992; Aoun and Li 1993; Den Dikken 1995; Authier and Reed 1999), but, unlike the two-chain approach adopted here, the predicate-raising analysis would not allow a uniﬁed account of Lasnik’s facts with the subject topicalization facts in section 4.5.2. 10. Vacuous movement of wh-subjects has been argued for in Icelandic and Kikuyu (Clements et al. 1983) on the basis of the distribution of stylistic fronting (Icelandic) and tonal downstep (Kikuyu). If the facts indeed suggest that vacuous subject movement is involved, then the question arises whether the facts are derived from movement of only the wh-feature (consistent with the analysis of English proposed here) or actual vacuous movement of the subject category as well. I leave the issue open for now. References Agbayani, Brian. 1998a. Feature attraction and category movement. Doctoral dissertation, University of California, Irvine. Agbayani, Brian. 1998b. Generalized pied-piping and island e¤ects. In Pius Tamanji and Kiyomi Kusumoto, eds., Proceedings of NELS 28, 1–14. Amherst: University of Massachusetts, GLSA. Agbayani, Brian. 2000. Wh-subjects in English and the Vacuous Movement Hypothesis. Linguistic Inquiry 31, 703–713. Aoun, Joseph, and Yen-hui Audrey Li. 1993. Wh-elements in situ: Syntax or LF? Linguistic Inquiry 24, 199–238. Authier, J.-Marc. 1992. Iterated CPs and embedded topicalization. Linguistic Inquiry 23, 329–336. Authier, J.-Marc, and Lisa Reed. 1999. On some syntactic conditions on presuppositions. In Elly van Gelderen and Vida Samiian, eds., Proceedings of the Western Conference on Linguistics 10, 27–40. Fresno: California State University, Department of Linguistics. Boeckx, Cedric, and Sandra Stjepanovic´. 2001. Head-ing toward PF. Linguistic Inquiry 32, 345–355.

92

Brian Agbayani

Bosˇkovic´, Zˇeljko. 2000. Sometimes in Spec,CP, sometimes in situ. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 53–87. Cambridge, Mass.: MIT Press. Cheng, Lisa L.-S. 1991. On the typology of wh-questions. Doctoral dissertation, MIT. Cheng, Lisa L.-S. 2000. Moving just the feature. In Uli Lutz, Gereon Mu¨ller, and Arnim von Stechow, eds., Wh-scope marking, 77–99. Amsterdam: John Benjamins. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1986. Barriers. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. Categories and transformations. In The Minimalist Program, 219–394. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1998. Some observations on economy in generative grammar. In Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky, eds., Is the best good enough?, 115–127. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chung, Sandra, and James McCloskey. 1983. On the interpretation of certain island facts in GPSG. Linguistic Inquiry 14, 704–713. Clements, George, James McCloskey, Joan Maling, and Annie Zaenen. 1983. String-vacuous rule application. Linguistic Inquiry 14, 1–17. Collins, Chris. 1997. Local economy. Cambridge, Mass.: MIT Press. Dikken, Marcel den. 1995. Binding, expletives, and levels. Linguistic Inquiry 26, 347–354. Doherty, Cathal. 1993. The syntax of subject contact relatives. Ms., University of California, Santa Cruz. Epstein, Samuel David. 1999. Un-principled syntax: The derivation of syntactic relations. In Samuel David Epstein and Norbert Hornstein, eds., Working minimalism, 317–345. Cambridge, Mass.: MIT Press. Fukui, Naoki, and Yuji Takano. 1998. Symmetry in syntax: Merge and Demerge. Journal of East Asian Linguistics 7, 27–86. George, Leland. 1980. Analogical generalization in natural language syntax. Doctoral dissertation, MIT. Halle, Morris, and Alec Marantz. 1993. Distributed Morphology and the pieces of inﬂection. In Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20, 111–176. Cambridge, Mass.: MIT Press.

Pied-Piping, Feature Movement, and Wh-Subjects

93

Henry, Alison. 1995. Belfast English and Standard English. Oxford: Oxford University Press. Hoekstra, Teun, and Rene´ Mulder. 1990. Unergatives and copular verbs: Locational and existential predication. The Linguistic Review 7, 1–79. Lasnik, Howard. 1995. Last resort. In Shosuke Haraguchi and Michio Funaki, eds., Minimalism and linguistic theory, 1–32. Tokyo: Hituzi Syobo. Lasnik, Howard. 2001. A note on the EPP. Linguistic Inquiry 32, 356–362. Lasnik, Howard, and Mamoru Saito. 1992. Move a: Conditions on its application and output. Cambridge, Mass.: MIT Press. Lobeck, Anne. 1990. Functional heads as proper governors. In Juli Carter, RoseMarie De´chaine, Bill Philip, and Tim Sherer, eds., Proceedings of NELS 20, 348– 362. Amherst: University of Massachusetts, GLSA. Lobeck, Anne. 1995. Ellipsis. Oxford: Oxford University Press. McDaniel, Dana. 1989. Partial and multiple wh-movement. Natural Language and Linguistic Theory 7, 565–604. Moro, Andrea. 1991. The raising of predicates: Copula, expletives and existence. In Lisa L.-S. Cheng and Hamida Demirdache, eds., More papers on whmovement, 119–181. MIT Working Papers in Linguistics 15. Cambridge, Mass.: MIT, Department of Linguistics and Philosophy, MITWPL. Ochi, Masao. 1999a. Constraints on feature checking. Doctoral dissertation, University of Connecticut, Storrs. Ochi, Masao. 1999b. Some consequences of Attract F. Lingua 109, 81–107. Radford, Andrew. 1997. Syntactic theory and the structure of English: A minimalist approach. Cambridge: Cambridge University Press. Rizzi, Luigi. 1990. Speculations on verb second. In Joan Mascaro´ and Marina Nespor, eds., Grammar in progress: GLOW essays for Henk van Riemsdijk, 375– 386. Dordrecht: Foris. Rizzi, Luigi. 1996. Residual verb second and the Wh-Criterion. In Adriana Belletti and Luigi Rizzi, eds., Parameters and functional heads: Essays in comparative syntax, 63–90. Oxford: Oxford University Press. Zwart, C. Jan-Wouter. 1992. Dutch expletives and small clause predicate raising. In Kimberley Broderick, ed., Proceedings of NELS 22, 477–491. Amherst: University of Massachusetts, GLSA.

PART II ¯ -Chains and Copy Theory A

Chapter 5 On the Form of Chains: Criterial Positions and ECP E¤ects

5.1

Luigi Rizzi

Introduction

It is widely recognized that natural language syntax makes extensive use of movement: elements are typically pronounced in positions di¤erent from those in which they receive some of their interpretive properties. ‘‘On Wh-Movement’’ (Chomsky 1977) made a major contribution to the ¯ -movement, ﬁrst and foremost study of the conﬁgurational properties of A its fundamentally local character and its mode of application in recursive (successive-cyclic) steps. The identiﬁcation of the basic conﬁgurational properties of movement was instrumental to abandoning a constructionbased conception of syntactic rules and to locating syntactic analysis at a more abstract level, a more appropriate level for identifying the basic computational ingredients of syntax. These conﬁgurational results also paved the way for a deeper understanding of the causes of movement, largely a contribution of more recent work, and of the Minimalist Program in particular. The present chapter follows this path backward: it starts with a discussion of the functional motivations of movement and then connects this level of analysis to the study of the form of chains, with special reference ¯ -system and the formal principles that constrain possible chain to the A conﬁgurations. The last part of the chapter addresses another traditional ¯ -syntax, the subject-object asymmetries arising in A ¯research topic of A extraction, showing that the system of principles proposed in the ﬁrst part provides an alternative to the classical analysis in terms of the Empty Category Principle. Section 5.2 addresses the issue of movement as a ‘‘last resort’’ and discusses the implementation of the operating mechanisms. Section 5.3 ¯ -chains connect an s-selection position (for arguments, a proposes that A y-position) to a criterial position, a position dedicated to the expression

98

Luigi Rizzi

of some scope-discourse property (Chomsky 2001, 2004) through a criterion, in the sense of Rizzi 1996 and related work. These two positions are ¯relevant for the interface with semantics and form the backbone of A chains. Sections 5.4–5.7 investigate whether and under what conditions other positions are allowed to occur in well-formed chains, in addition to the two interpretively relevant positions. In particular, they provide empirical evidence for a principle according to which criterial positions terminate chains: a phrase meeting a criterion is frozen in place, and its chain cannot extend further (Criterial Freezing). This principle makes sure that the chain will be assigned a unique scope-discourse property, basically in parallel with the assignment of a unique y-role, thus contributing to a parsimonious deﬁnition of chains as constituted by unique occurrences of the elementary ingredients. Sections 5.8–5.9 address the issue of subject chains. In classical Government-Binding (GB) analyses, the subject position is singled out by two separate principles, the Extended Projection Principle (EPP), which makes clausal subjects obligatory, and the Empty Category Principle (ECP), which bars subject extraction in certain environments. An analysis of the EPP is proposed in terms of the system of principles motivated in sections 5.2–5.7. A Subject Criterion is introduced, which requires a nominal expression in the higher part of the inﬂectional ﬁeld of which the VP is predicated, and the interpretive properties of this position are investigated. Criterial Freezing can thus provide a new explanation for subject-object asymmetries classically ascribed to the ECP. 5.2

Background: Movement as a Last Resort

Movement is pervasive in natural languages: it is quite typical that most or all of the pronounced elements in a given structure have moved from the positions in which they were ﬁrst merged. Consider for instance a French interrogative clause like the following, illustrating OVS order: (1) Qui rencontrera-t-il? whom will.meet-he ‘Whom will he meet?’ Under current assumptions, this surface conﬁguration is derived via three kinds of movement operations from the thematic nucleus expressed by the VP (or a more complex structure involving a vP layer, as in Chomsky 1995 and related works). (2) [VP il [rencontre- qui]]

On the Form of Chains

99

The lexical verb ﬁrst moves to the inﬂectional ﬁeld to pick up its inﬂectional a‰xes of tense, agreement, and so on, and then to the C ﬁeld, possibly, in this case to pick up -t-, which may be analyzed as an interrogative a‰x (alternatively, to check the relevant a‰xes; on the ‘‘building’’ versus ‘‘checking’’ approach to the syntax-morphology interface, see Baker 2002). The subject moves to the speciﬁer position (Spec) of some inﬂectional head to pick up Case and satisfy the EPP. The wh-object moves (possibly through intermediate steps) to its operator position in the C system. We thus end up with a representation like the following, with the thematic nucleus entirely vacated: (3) [CP Qui rencontreþraþt [ IP il tI [VP tS tV t O ]]]? A wealth of empirical evidence has accumulated over the years showing that movement is ubiquitous in natural language syntax. But why does movement exist? The Minimalist Program hypothesizes that movement is a device to satisfy certain interface requirements. Consider the movement of the verb to I and then to C in (3): following a line of analysis that goes back to Syntactic Structures (Chomsky 1957), through Emonds 1978, Pollock 1989, Belletti 1990, and much related work, this kind of movement appears to be motivated by the need to form morphologically well-formed words, by associating such a‰xes as -ra and -t-, which cannot survive as independent words, to an appropriate verbal root (but see Koster 2003 for an alternative approach). As for the movement of the object to the C system, a di¤erent requirement seems to be operative, involving the interface with semantic interpretation. Chomsky (2001, 2004) suggests that (certain) natural language expressions may need to be associated to two kinds of semantic properties: argumental properties (a y-role in the case of a direct object) and scope-discourse properties (clausal scope in the case of a wh-element). Linguistic expressions can receive both kinds of interpretive properties by ‘‘picking them up’’ from positions, as it were: they move from positions dedicated to properties of the ﬁrst kind to positions dedicated to properties of the second kind. So, qui receives the y-role ‘‘patient’’ from the verb in the position in which it is ﬁrst merged, and then the scopediscourse property ‘‘interrogative with main clause scope’’ in its ﬁnal position. Movement is a device to achieve this duality of interpretations. This kind of teleological motivation seems to be required of movement: movement is a last resort operation; there is no free or optional movement. I will assume the following informal version of the relevant economy principle (the concept goes back at least to Chomsky 1986;

100

Luigi Rizzi

important elaborations, essentially along the lines assumed here, were introduced in Fox 1995 and Reinhart 1997): (4) Movement as a last resort Movement takes place to satisfy some interface requirement. The interface level involved may be the syntax-morphology interface, internal to the narrow computational system (as in the head movement case), or the external interface with semantics, as in the case of left¯ -movement. As for the third kind of chain in (1), the one peripheral A involving the movement of the subject from its y-position to the Spec of an inﬂectional head, things are less straightforward. One standard assumption is that this movement is motivated by Case requirements on the subject, which would suggest again the relevance of the interface with morphology. On the other hand, if the subject position is endowed with special discourse properties (quasi-topicality, and the like; Chomsky 2002), and/or if external systems require events to be expressed in subjectpredicate format (Rothstein 1983), the interface with semantics may be relevant as well to motivate movement of the subject from the VPinternal position. Curiously, if the case of A-chains is less straightforward than that of other chain types in terms of the teleological motivation of movement, it is also the case that has basically inspired the formal implementation of movement in general in minimalism. Without entering into technical details here, I will assume that movement is formally triggered by matching of features: a head X, endowed with feature F, acts as the probe searching for a goal Y in its c-domain, endowed with the same feature. Y (or, more precisely, a phrase containing this element, selected in accordance with the principles responsible for ‘‘pied-piping’’) is then remerged in the immediate structural environment of X: basically, Y adjoins to X if Y is a head, and is merged as X’s speciﬁer if it is a phrase (see Chomsky 2000 and below for more detail and variants of these basic assumptions). (5) XF . . . YF . . . ! YF XF . . . t . . . ¯ -chains, referring to the In the rest of this chapter, I will focus mainly on A other chain types only occasionally. 5.3

Criterial Positions

In the following examples, the phrase [D book] must receive both the interpretive property ‘‘patient of the verb read ’’ and the interpretive properties ‘‘interrogative,’’ ‘‘topic,’’ ‘‘focus,’’ respectively:

On the Form of Chains

(6) a. Which book should you read b. This book, you should read c. THIS BOOK you should read

101

? (rather than something else)

Natural languages express this duality of interpretive properties by having the phrase occur in two positions, each assigning one kind of property. In standard terminology, an element is moved from a position of one kind to a position of the other kind. If one adopts the copy theory of traces (Chomsky 1993), according to which ‘‘traces of movement’’ are complete but silent copies of the moved element, it is more appropriate to say that an element can be merged several times in a structure. In other words, occurrences of the element can multiply, so that in (6) [D book] occurs both as complement of read and as left-peripheral focus, topic, or interrogative operator, the unpronounced lower occurrence being represented within angle brackets in (7). (7) a. Which book should you read hwhich booki? b. This book, you should read hthis booki c. THIS BOOK you should read hthis booki (rather than something else) Departing slightly from Chomsky’s (2001, 2004) terminology, I will call the two kinds of interpretive properties s-selectional and criterial, respectively. An element is ﬁrst merged in a position in which it is semantically selected, and then it may be merged again in a position dedicated to scope-discourse semantics. The two positions (and other positions hosting occurrences of the same element; see below) form a chain. Typical s-selectional properties are thematic properties, which I assume to be assigned by lexical items in their immediate structural neighborhood. But the concept should be generalized to whatever interpretive properties license the insertion of a nonargument in the structure: adverbials of various kinds, modiﬁers, and so on. Following a rather traditional terminology, I call these properties s-selection (semantic selection, as in Grimshaw 1979, Pesetsky 1982, etc.). The case of s-selection for arguments is y-assignment; the analogous case for, say, a temporal adverbial is s-selection by a T node, and similarly for modal and aspectual adverbs, according to much recent work on adverbial syntax, most notably in the systematic approach taken by Cinque (1999). As for Chomsky’s scope-discourse positions, I will assume that they are determined by a family of principles, the criteria, which require speciﬁerhead agreement with respect to features of the relevant class: Q, Top, Foc, R for questions, topic, focus, relatives, and so on (see Rizzi 1996

102

Luigi Rizzi

for an early formulation of this approach). As a ﬁrst approximation, I will take the format of criteria to be the following: (8) XPF and XF must be in a speciﬁer-head conﬁguration, for F ¼ Q, Top, Foc, R, . . . That the preposing processes create local speciﬁer-head conﬁgurations with dedicated heads is suggested by the fact that such heads are sometimes morphologically expressed. Some plausible cases of morphological expression of Q, Top, Foc, and R, respectively, are the following (the criterial heads are italicized):1 (9) a. Ik weet niet [wie of [Jan gezien heeft]] I know not who Q Jan seen has (Haegeman 1994) ‘I don’t know who Jan saw.’ `ı]]] b. Un se` [do [dan lo ya` [Koﬁ hu I heard that snake the top Koﬁ killed it (Aboh 1998, 2004) ‘I heard that the snake, Koﬁ killed.’ c. Un se` [do [dan lo we` [Koﬁ hu ]]] I heard that snake the foc Koﬁ killed (Aboh 1998, 2004) ‘I heard that THE SNAKE, Koﬁ killed.’ d. der Mantl [den wo [dea Hons gfundn hot]] the coat which rel the Hans found has (Bayer 1984) ‘the coat which Hans found’

(Dutch)

(Gungbe)

(Gungbe)

(Bavarian)

So, a head endowed with the relevant feature attracts a phrase bearing that feature, thereby designating a position dedicated to the relevant type of interpretation. The information is then passed on to the interpretive system (‘‘The phrase in this Spec is to be interpreted as an interrogative operator with scope over this clause, as a Topic, and so on’’). Such dedicated heads are sometimes overt, sometimes silent, but, I will assume, systematically present across languages (as should be expected under Chomsky’s (2001, 2) Uniformity Principle, essentially along the lines commonly assumed for Case theory). ¯ -chain therefore typically includes two positions, both A standard A expressing some interface property and both marked in the structure by the feature content of the local head. (10) . . . criterial . . . s-selection . . .

On the Form of Chains

103

¯ -chains in some generality, we should now deterTo explore the form of A mine what positions are required and what positions are allowed in a chain. Are there other positions in addition to s-selection and criterial positions? This question breaks up into three subcases: (11) a. Is there any position lower than the s-selection position? b. Is there any intermediate position between the criterial and sselection positions? c. Is there any position higher than the criterial position? The answer to (11a) is quite generally assumed to be no: arguments cannot be ﬁrst-merged in positions lower than their y-position, and then pick up their y-role through movement (but see Kayne 2002 for a di¤erent view, in the context of his attempt to derive binding conditions from movement). As for question (11b), the answer is clearly yes: there is rich and diverse evidence for intermediate traces between the s-selection and criterial positions. As for question (11c), it is rarely asked, and trying to address it in some generality is one of the main purposes of this chapter. Let us consider the three subcases in detail. 5.4

Chains Start at S-Selection Positions

¯ -chains uniquely. It can be asked of Question (11a) does not concern A phrasal chains in general, regardless of whether they stop inside IP or proceed to reach the left periphery of the clause. The question breaks up into two subcases (I will initially focus on arguments for simplicity): (12) a. Can an argument be merged in a non-y-position and then move to a y-position to pick up its y-role? b. Can an argument be merged in a y-position and then pick up another y-role through movement? The ﬁrst case is addressed by Chomsky (2000). Why can’t (13b) be derived alongside (13a), given the choice of lexical array appropriate for (13a)? (13) a. I expected [a proof to be discovered t] b. *I expected [t to be discovered a proof ] The critical point in the derivation is when the embedded T is merged, endowed with an EPP feature requiring its Spec to be ﬁlled. (14) TEPP [to be discovered a proof ]

104

Luigi Rizzi

If the pronoun I is part of the lexical subarray of the same phase, as under Chomsky’s assumptions, why can’t it be merged in Spec,T, thus satisfying T’s EPP requirement, and then be raised to the subject position of expect, thus receiving the y-role ‘‘experiencer of expect,’’ with representation (13b)? Why is movement of a proof the only viable option in (14), yielding (13a)? The question is particularly relevant given the idea that Merge, when possible, prevents Move, which Chomsky uses to account for the following pair: (15) a. *There was expected [a proof to be discovered t] b. There was expected [t to be discovered a proof ] If the expletive there is part of the relevant lexical subarray, Chomsky suggests, it must be merged at stage (14), to fulﬁll the EPP requirement on T (on the derivation of the more natural order . . . to be a proof discovered, see the discussion in Chomsky 2000). Then the expletive raises to the main subject position, yielding (15b). (15a) is excluded because the possibility of merging there in (14) precludes the more complex operation of moving a proof. But why is the pattern reversed in (13)? Chomsky (2000, 103) introduces the following principle: (16) Pure Merge in y-position is required of (and restricted to) arguments. So the argument I cannot be introduced into the structure by pure Merge in a non-y-position in (14), and the only available derivational option is movement of a proof to yield (13a). This approach must clearly be generalized to ﬁrst Merge of nonargumental material: an adverb cannot be inserted in a position lower than the position it is interpreted as a modiﬁer of. For instance, in a verb-second (V2) language a temporal adverbial cannot be ﬁrst-merged to satisfy embedded V2 and then be interpreted as construed with the main tense structure. Consider the following paradigm (provided by Gu¨nther Grewendorf, pers. comm.): (17) a.

Maria glaubte, Hans wu¨rde in zwei Jahren seinen 60. Maria believed Hans would in two years his 60th Geburtstag feiern birthday celebrate ‘Maria believed that Hans would celebrate his 60th birthday in two years’ time.’

On the Form of Chains

105

In zwei Jahren, glaubte Maria, t wu¨rde Hans t 0 seinen 60. in two years believed Maria would Hans his 60th Geburtstag feiern birthday celebrate ‘In two years’ time, Maria believed that Hans would celebrate his 60th birthday.’ c. *Vor zwei Jahren, glaubte Maria, t wu¨rde Hans seinen 60. before two years believed Maria would Hans his 60th Geburtstag feiern birthday celebrate ‘Two years ago, Maria believed that Hans would celebrate his 60th birthday.’ d. Vor zwei Jahren glaubte Maria, Hans wu¨rde bald seinen beofre two years believed Maria Hans would soon his 60. Geburtstag feiern 60th birthday celebrate ‘Two years ago, Mary believed that Hans would soon celebrate his 60th birthday.’

b.

(17b) shows that a temporal adjunct can be extracted from an embedded clause, satisfying ‘‘in passing’’ the V2 constraint in that clause, in which case the adjunct is interpreted as construed with the embedded clause. The deviance of (17c) shows that an adjunct satisfying embedded V2 must be construed with the embedded clause: vor zwei Jahren ‘two years ago’, semantically compatible with the main clause tense but not with the embedded tense, gives rise to ill-formedness in this case, while it is ﬁne with main clause construal in (17d), where embedded V2 is satisﬁed by another element. In conclusion, (12a) generally seems to require a negative answer. As for (12b), it is quite generally assumed to require a negative answer as well. For instance, languages do not express reﬂexive readings by moving an argument from object to subject position, to pick up both y-roles, and control structures cannot be analyzed through movement from the lower subject position to a main clause argument, as the lack of reconstruction e¤ects typical of raising shows (see (18)). (The analysis of control as feature movement, whether right or wrong, is a di¤erent matter (for discussion, see Hornstein 1999; Landau 2003); and the unaccusative-like analysis of Romance reﬂexives proposed by Burzio (1986) and Kayne (2001, chap. 8) does not involve the picking up of two y-roles via movement.)

106

Luigi Rizzi

(18) a. *John saw t (meaning ‘John saw himself ’) b. *John wants [t to leave] Consider, for instance, the familiar raising-control asymmetry with respect to the possibility of anaphor binding under reconstruction (Langendoen and Battistella 1982; Belletti and Rizzi 1988). The anaphor each other can be bound by the men under reconstruction in the raising structure (19a), a result immediately captured by the copy theory of traces, but not in the control structure (19b). (19) a. Friends of each other seemed [t to amuse (t 0 ) the men] b. *Friends of each other wanted [PRO to amuse (t 0 ) the men] (See also Burzio’s (1986) discussion of the construal of each under reconstruction, excluded in control structures.) We can now try to generalize Chomsky’s principle (16) to include these cases, and also to extend it to analogous patterns with nonarguments, adverbial material, and the like. Let us continue to refer to the traditional notion of s-selection to encompass y-marking of arguments and the licensing of adverbials (analyzed along the lines developed in Cinque 1999) and other nonargumental material. (20) S-selection is fulﬁlled by ﬁrst Merge. (We can understand ‘‘s-selection’’ here in a maximally general way, as involving both the selector and the selected element. Note that principle (20) basically expresses and generalizes the GB assumption that the yCriterion applies at D-Structure.) Therefore, the subject I cannot pick up its y-role ‘‘experiencer of expect’’ by second Merge in (13b), John cannot pick up the y-role ‘‘experiencer of see, want’’ in (18), the temporal adverb cannot pick up a selectional relation with the main clause T in (17c), and so on. On the other hand, the expletive there, which is not subject to s-selection, can be ﬁrst-merged in the subject position of (15) (and must be, according to Chomsky’s analysis). 5.5 Is There Any Intermediate Position between the Criterial and SSelection Positions? ‘‘On Wh-Movement’’ forcefully argued for a stepwise application of movement, hence, in present terms, for a positive answer to question (12b). Over the last 25 years or so, quite diverse and compelling evidence has accumulated in favor of this conclusion: in addition to the

On the Form of Chains

107

two positions relevant at the interface, chains must involve intermediate positions. Intermediate positions can trigger morphosyntactic phenomena, making them detectable in di¤erent languages. For instance, intermediate wh-movement to an embedded C system triggers embedded I-to-C movement in Belfast English (Henry 1995). (21) What did Mary claim [t did [they t steal t]]? French stylistic inversion is triggered by a wh-element in the local C system: intermediate movement triggers it, as in (22d) with representation (22f ), while a nonlocal wh-element per se is unable to trigger it, as in (22e) (Kayne and Pollock 1978). (22) a.

b. c.

d.

e. f.

Jean est parti Jean is left ‘Jean left.’ *Est parti Jean is left Jean Ou` est alle´ Jean? where is gone Jean ‘Where did Jean go?’ Ou` crois-tu qu’est alle´ Jean? where believe-you that is gone Jean ‘Where do you believe Jean went?’ *Qui croit qu’est parti Jean? who believes that is left Jean ou` crois-tu [hou`i que [est alle´ hou`i Jean]]

Other examples are past participle agreement in French (see Kayne 1989) and the C agreement of the Celtic (and many other) languages (Irish example from McCloskey 1996). (23) a. Jean a repeint les chaises Jean has repainted the chairs b. les chaises que Jean a t 0 repeintes t the chairs that Jean has repaintedþagr (24) Ce´n t-ursce´al aL [mheas me´ [aL duirt se´ [aL thuig se´]]]? which novel that thought I that said he that understood he ‘Which novel did I think that he said that he understood?’ Intermediate positions are not assigned special interpretive properties; nevertheless, they are visible at the interpretive interface, as shown by

108

Luigi Rizzi

reconstruction e¤ects (Lebeaux 1988). A sentence like (25d) is therefore ambiguous, admitting both Bill and John as antecedents of the anaphor himself. The latter interpretation must be made available by the intermediate trace in the embedded C system. Under the copy theory of traces, the relevant representation is (25e). (25) a. [Johnk thinks [that [Billi likes this picture of himselfi; k ]]] b. [Johnk wonders [which picture of himselfi; k [Billi likes ]]] c. [John wonders [which picture of himself [Bill likes hwhich picture of himself i]]] d. [Which picture of himselfi; k ] does [Johnk think [that [Billi likes ]]]? e. [which picture of himself ] does [John think [hwhich picture of himself i [that [Bill likes hwhich picture of himself i]]]] The evidence for intermediate positions in movement chains, between s-selection and criterial positions, is thus strong and diverse, ranging from purely syntactic, to morphosyntactic, to interpretive. See Chung 1998 for a detailed review. Why is it that chains involve such intermediate positions, positions that are not motivated by interface needs? The standard answer is that intermediate positions are enforced by the theory of locality: movement cannot land too far away; linguistic representations can be indeﬁnitely extended owing to the recursive nature of the system, but each step of the linguistic computation is local and can cover only a limited amount of structural ground. Chains can be indeﬁnitely long because the operations creating them can iterate indeﬁnitely, but each chain link itself is local. The fundamental intuition of ‘‘On Wh-Movement’’ (and previous work within the Extended Standard Theory tradition, starting from Chomsky 1973) has retained all its force here. Putting things in slightly more general terms, we can say that two competing forces determine the form of chains, both being construed as economy guidelines: Movement as a last resort: Movement must be motivated; it can only be performed to achieve an interface e¤ect; there is no ‘‘free movement.’’ Locality: Movement cannot land too far away.

The movement-as-a-last-resort guideline would favor chains with just two positions, the positions motivated by interface e¤ects; locality would favor a proliferation of positions very close to one another. Permissible chains therefore represent a point of equilibrium between these two desiderata.

On the Form of Chains

5.6

109

¯ -Movement On the Implementation of A

One salient property that intermediate positions have in common is that they normally cannot function as ﬁnal positions. Rather, they are gates, or escape hatches, through which elements must pass to reach higher positions. The paradox of these intermediate positions is that, on the one hand, they must autonomously cause a movement step (if we want to take seriously the idea that each step is locally determined, with no ‘‘look-ahead’’ to subsequent derivational steps), yet, on the other hand, we must ensure that movement will not stop there. Let us make more speciﬁc assumptions about the implementation of movement in general, and then return to intermediate movement. I will follow standard minimalist assumptions about the fact that movement is guided by a system of morphosyntactic features: basically, the attractor and the attractee are characterized by a certain shared feature speciﬁcation. In the system devised for A-chains in Chomsky 2000, uninterpretable features on the attractor and attractee play a crucial role: uninterpretable f-features on T enter into an Agree relation with a DP made active by uninterpretable Case features, thus establishing the conﬁguration that is a prerequisite for movement. (26) Tuf . . . DPif; uCase . . . ¯ -chains. In questions, for instance, the question feature Q, Now consider A expressed by special complementizers like if in English and si in French, and by question particles in various languages (sometimes limited to yes/ no questions, sometimes generalized to all questions), is presumably interpretable; it enters into clause typing (Cheng 1991) and selection from a higher predicate. The morphosyntactic feature characterizing a certain class of operators as question operators (wh- in English, qu- in French, etc.) is also visible to, and used by, interpretive systems. So, the minimal morphosyntactic feature system required by questions (and other criterial constructions) does not seem to involve morphologically overt uninterpretable features. We thus seem to have two theoretical options. Either we can assume abstract uninterpretable features associated to the interpretable features of attractor and attractee, making the system fully parallel to the Asystem in Chomsky 2000 (much as in Pesetsky and Torrego’s (2001) approach). Or we can take the lack of direct morphosyntactic evidence

110

Luigi Rizzi

¯ -uninterpretable features at face value and assume the criterial atfor A traction is achieved directly by the interpretable features; that is, in conﬁgurations like (27), C attracts XP bearing a matching criterial feature. (27) CF . . . XPF . . . Not much hinges on this point in the present context. The critical need for the computational system to work is to formally identify the attractor and the attractee, and this is achieved in both ways. For concreteness, in what follows I will take the second path, which minimizes the feature speciﬁcation to be postulated; I will therefore assume that interpretable criterial features can serve as attractors of movement. Whatever assumption one makes about attraction to a criterial posi¯ -chains. tion, the issue of uninterpretability arises in a di¤erent form in A Consider intermediate movement to a noncriterial position. How is movement triggered to, say, an intermediate C? (28) I wonder [what C [you think [t 0 that [I saw t]]]] One possibility is to allow non-feature-driven movement: movement is in fact free, but chains thus created must meet the requirement that the ﬁnal site is marked with a criterial feature. An alternative more in line with current ideas, and argued for in detail in McCloskey 2002, is to assume that interpretable criterial features have purely formal counterparts (such as the ‘‘pseudo’’ wh-feature discussed by McCloskey) and that intermediate movement is triggered by such purely formal features, which bring the moved phrase closer to the target, in accordance with locality principles. Capitalizing the labels for substantive criterial features (Q, Top, Foc, R, etc.) and lowercasing the labels for the corresponding purely formal features (q, top, foc, r, etc.), we have the following representation for a sentence like (28): (29) I wonder [what C Q [you think [t 0 thatq [I saw t]]]] We thus admit only one kind of movement, feature-driven movement (cutting across the distinction between substantive and purely formal features). Evidence favoring this view over the alternative admitting intermediate non-feature-driven movement is the fact that languages often manifest morphological reﬂexes on the heads hosting intermediate landing sites in their Specs, as shown in the work brieﬂy reviewed in section ¯5.5 (see McCloskey 2002 for a detailed argument based on Celtic A constructions). Agreement phenomena like past participle agreement in Romance, complementizer agreement in Celtic, and the triggering of I-

On the Form of Chains

111

to-C movement in cases like (21) are phenomena that would be expected if some feature is speciﬁed on the relevant head.2 Given this discussion, I will tentatively assume that intermediate movement is driven by features that are the purely formal counterpart of substantive criterial features. There is only one kind of movement, feature-triggered; the triggering features are sometimes substantive (criterial, interpretable), sometimes purely formal (uninterpretable). Natural language syntax includes the latter featural option in order to implement intermediate movement, so that a potentially distant criterial position can be reached in successive steps, each of which satisﬁes locality requirements. ¯ -constructions. Relatives also Now let us take a quick look at other A involve a (sometimes overt) relative head and morphologically marked operators (if one assumes the head-raising analysis of relativization, then it should work in a manner akin to topicalization). Topic and focus constructions involve special heads, sometimes overt (see (9b–c)); I will assume that the head of the topicalized or focalized phrase is freely assigned the corresponding feature as it enters syntax, much like a ffeature (so that no problem arises for the Inclusiveness Condition; see Chomsky 2000). This marking makes the phrase suitable for attraction and normally has PF e¤ects reﬂected by the special intonational contours of topics and foci. So, the same derivational mechanisms can be uni¯ -constructions, by and large the uniﬁcaformly assumed for all major A tion achieved by ‘‘On Wh-Movement’’ (but I continue to assume, in the spirit of the cartographic approach to the C system (Rizzi 1997, 2000, 2004a,b), that distinct landing sites are involved in the left periphery). 5.7

Is There a Position Higher Than the Criterial Position?

Can movement continue when the moved item reaches the ﬁrst criterial position, or does it have to stop there? In other words, can several criteria be met ‘‘in passing’’ by a single phrase? Analogy with the s-selection position would suggest a negative answer. By symmetry, one would expect that (1) chains begin at an s-selection position and end at a criterial position, and (2) such positions are unique, each chain having exactly (or at most) one position of each kind. Is the analogy correct? We will now look at some data supporting the analogy and suggesting that, when a criterion is met, the phrase is frozen in place. This issue is rarely addressed in the literature (if at all), so let us consider it systematically. There are four cases to consider:

112

Luigi Rizzi

(30) In principle, attraction of a phrase to two distinct criterial positions may involve a. a single feature on the same element of the phrase: [ . . . A þF . . . ]; b. two distinct features on the same element: [ . . . A þF1 ; þF2 . . . ]; c. the same feature on two elements in the same phrase: [ . . . A þF1 . . . BþF1 . . . ]; d. two distinct features on two elements in the same phrase: [ . . . A þF1 . . . BþF2 . . . ]. A straightforward instantiation of case (30a) is a structure involving both an indirect question and a main question: plainly, the same whelement cannot fulﬁll the requirements of the indirect question (satisfy the Question Criterion in the embedded C system) and then move on to the main C system. Given (31a) as an intermediate step, it is not possible to derive (31b), with representation (31c) in terms of the copy theory of traces. (31) a. Bill wonders [which book C Q [she read t]] b. *Which book C Q does Bill wonder [t 0 C Q [she read t]]? c. *Which book C Q does Bill wonder [hwhich booki C Q [she read hwhich booki]]? To the best of my knowledge, this is the only case of (30) discussed in the literature (e.g., Lasnik and Saito 1992). In terms of the approach and terminology used here, it suggests that the creation of a criterial conﬁguration freezes the structure, and the wh-phrase involved is not accessible to further movement. In other words, a constraint like this holds: (32) Criterial Freezing A phrase meeting a criterion is frozen in place. Principle (32) can be thought of as an economy condition that contributes to minimizing movement (hence a component of the formal implementation of the concept ‘‘movement as a last resort’’) and determines optimally simple chains with unique occurrences of the fundamental ingredients: unique y-role (or s-selection licensing), unique scope-discourse marking, and so on. Is (32) an independent formal principle, or does it follow from other properties of syntactic computations or of the interface systems? Case (31b–c) could be ruled out for purely interpretive reasons, as it does not seem to be amenable to any sound logical form. But if we turn to the other cases of (30), the interpretive approach does not seem to be generalizable.

On the Form of Chains

113

Consider (30b). A possible instantiation would be a phrase bearing both the Q-feature and the (contrastive) focus feature. In languages like Italian, contrastive focus may be associated to an element in situ, as in (33a), or may involve movement of the element to the left periphery, as in (33b). (Contrastive stress is expressed by capitalization in the following examples.) (33) a. Pensavo che avessero scelto la RAGAZZA, non il ragazzo I thought that they had chosen the GIRL not the boy b. La RAGAZZA pensavo che avessero scelto , the GIRL I thought that they had chosen non il ragazzo not the boy Now, a wh-phrase in an embedded question can be contrastively focused in its criterial position, in the embedded C system, but it cannot be moved to the left periphery of the main clause. (34) a.

Mi domandavo quale RAGAZZA avessero scelto, I wondered which GIRL they had chosen non quale ragazzo not which boy b. *Quale RAGAZZA mi domandavo avessero scelto, which GIRL I wondered they had chosen non quale ragazzo not which boy

As contrastive focus is clearly compatible with a wh-phrase (see (34a)), it does not seem plausible to assume that (34b) is ruled out for interpretive reasons.3 Case (30c) is illustrated by a complex phrase whose elements bear two ¯ A-features of the same kind—say, two Q-features. An example is [quanti libri di quale autore] ‘how many books by which author’. Suppose this complex phrase is inserted in a structure involving both a main and an embedded question. Consider an abstract conﬁguration like (35). (Here and in what follows, for the sake of simplicity I use old-style derivations with movement applying on completely formed structures; nothing hinges on this.) (35) C Q Ti domandi [C Q siano stati censurati you wonder have been censored [quanti libri di quale autore]] ! how many books by which author

114

Luigi Rizzi

Here, the only (marginally) possible grammatical output is the one in which the complex wh-phrase splits, and in a speciﬁc order: the whole wh-phrase moves to the embedded C system, yielding (36a), and then the most deeply embedded phrase, di quale autore ‘by which author’, is further extracted to the main C system, yielding the marginally acceptable (36b). (36) a. C Q Ti domandi [[quanti libri di quale autore] CQ you wonder how many books by which author [siano stati censurati t]] ! have been censored b. ?[Di quale autore] C Q ti domandi [[quanti libri t] CQ by which author do you wonder how many books [siano stati censurati t]]? have been censored The complex wh-phrase cannot split in the opposite order. That is, as (37) shows, it is impossible to ﬁrst extract the most deeply embedded subphrase di quale autore to the embedded C system and then move the remnant of the complex phrase to the main C system. (37) a.

C Q Ti domandi [[di quale autore] C Q [siano stati censurati you wonder by which author have been censored [quanti libri t]]] ! how many books b. *[Quanti libri t] C Q ti domandi [[di quale autore] C Q how many books do you wonder by which author [siano stati censurati t]]? have been censored

That the splitting can take place in only one way is an interesting observation that calls for an explanation. It presumably bears on the constraints on remnant movement; I assume it to be orthogonal to the present discussion.4 But the relevant point here is that the splitting must take place. It is not possible to move the large phrase to the embedded C system and then move it on to the main C system. (38) a.

C Q Ti domandi [[quanti libri di quale autore] C Q you wonder how many books by which author [siano stati censurati t]] ! have been censored

On the Form of Chains

115

b. *[Quanti libri di quale autore] C Q ti domandi [t C Q how many books by which author do you wonder [siano stati censurati t]]? have been censored Why is this derivational option not viable? Notice that the lower whelement is able to pied-pipe the whole phrase from an embedded declarative, at least marginally. (39) ?[I libri di quale autore] credi [t che the books by which author do you believe that [siano stati censurati t]]? have been censored Again, the impossibility of (38b) suggests that criteria cannot be satisﬁed in passing; hence, something like constraint (32) is operative. No interpretive problem seems to arise here: (38b) should receive the same logical form as (36b), obtainable under reconstruction (in fact, directly provided by the copy theory of traces). Notice that (38b) signiﬁcantly improves if the embedded C is ﬁlled by a yes/no complementizer (se ‘whether’). (38) b 0 . ??Quanti libri di quale autore ti domandi how many books by which author do you wonder se siano stati censurati? whether have been censored In fact, (38b 0 ) expresses a completely di¤erent structure from (38b): here, the Question Criterion is fulﬁlled by se (and/or a null yes/no operator) in the indirect question, and the complex wh-phrase is construed with the main C system as a multiple main question. The complex whphrase has passed through whatever escape hatch is available in the embedded C system in cases of extraction from a wh-island, where it is not involved in criterial satisfaction; the marginality of the example is due to the wh-Island and to whatever factor disfavors multiple questions in Italian (Rizzi 1982, chap. 2), but Criterial Freezing is not involved. The same argument provided by (38b) can be constructed with a complex phrase containing both an interrogative and a relative element [quanti libri del quale] ‘how many books by whom(rel)’. Starting from a conﬁguration like (40a), and after moving the complex phrase to the interrogative C system, as in (40b), it is necessary to split the complex

116

Luigi Rizzi

phrase, as in (41a); pied-piping the whole phrase to the relative C system, as in (41b), yields a sharply ungrammatical result. I assume that a Relative Criterion causes movement in relative clauses (see Bianchi 1999 for a more ﬁne-grained approach); for concreteness, in the following examples I call the feature attracting the relative phrase R. (40) a. Gianni [CR [mi domando [C Q [siano stati censurati Gianni I wonder have been censored [quanti libri [del quale]]]]] ! how many books by whom b. Gianni [CR [mi domando [[quanti libri [del quale]] CQ Gianni I wonder how many books by whom [siano stati censurati t]]] ! have been censored (41) a.

Gianni, [del quale CR [mi domando [[quanti libri t] CQ Gianni by whom I wonder how many books [siano stati censurati t]]]] have been censored b. *Gianni, [[quanti libri del quale] CR [mi domando [t C Q Gianni how many books by whom I wonder [siano stati censurati t]]]] have been censored

In the case of appositive relatives, the pied-piping of complex phrases by a deeply embedded relative pronoun is quite tolerable in other structures, as in (42b).5 (42) a. Gianni, del quale credo che siano stati censurati Gianni by whom I believe that have been censored [molti libri t] many books b. ?Gianni, [molti libri del quale] credo Gianni many books by whom I believe che siano stati censurati t that have been censored What goes wrong in (41b), then, is not heavy pied-piping per se but the attempt to satisfy a criterion in passing. In fact, in some cases an even heavier pied-piping is marginally tolerable, provided that no criterial conﬁguration is undone. Consider the following paradigm:

On the Form of Chains

117

Gianni, [del quale CR [non e` ancora stato chiarito Gianni by whom it has not been clariﬁed yet [[quanti libri tPP ] C Q [siano stati censurati tDP ]]]] how many books have been censored b. *Gianni, [[quanti libri del quale] CR Gianni how many books by whom [non e` ancora stato chiarito it has not been clariﬁed yet [tDP C Q [siano stati censurati tDP ]]]] have been censored c. ?(?)Gianni, [[[quanti libri del quale] C Q Gianni how many books by whom [siano stati censurati tDP ]] CR have been censored [non e` ancora stato chiarito t CP ]] it has not been clariﬁed yet

(43) a.

Here, given an abstract representation like (44a), after the complex phrase quanti libri del quale has moved to the Spec,C Q of the indirect question, yielding intermediate representation (44b), it is possible for the complex phrase to split, with movement of del quale alone to the relative C system, giving rise to (43a). (44) a. Gianni, [CR [non e` ancora stato chiarito [C Q Gianni it has not been clariﬁed yet [siano stati censurati [quanti libri del quale]]]]] ! have been censored how many books by whom b. Gianni [CR [non e` ancora stato chiarito [[quanti libri Gianni it has not been clariﬁed yet how many books del quale] C Q [siano stati censurati tDP ]]]] ! by whom have been censored It is not possible to continue the derivation from (44b) by pied-piping quanti libri del quale to the CR system; this would derive the sharply ungrammatical (43b), a violation of Criterial Freezing. But there is another (marginal) derivational option from (44b): the whole indirect question can be pied-piped to Spec,CR , yielding (43c). The steps involved in this derivation are repeated here: (45) a. Gianni [CR [non e` ancora stato chiarito [C Q Gianni it has not been clariﬁed yet [siano stati censurati [quanti libri del quale]]]]] ! have been censored how many books by whom

118

Luigi Rizzi

b. Gianni [CR [non e` ancora stato chiarito [[quanti libri Gianni it has not been clariﬁed yet how many books del quale] C Q [siano stati censurati tDP ]]]] ! by whom have been censored c. Gianni [[[quanti libri del quale] C Q Gianni how many books by whom [siano stati censurati tDP ]] CR [non e` ancora stato chiarito t CP ]] have been censored it has not been clariﬁed yet In this case, the complex phrase quanti libri del quale does not have to split because the Question Criterion and the Relative Criterion are satisﬁed by distinct phrases (the Question Criterion by the complex nominal phrase quanti libri del quale and the Relative Criterion by the whole indirect question quanti libri del quale siano stati censurati containing the complex nominal phrase); no satisfaction in passing is involved, and principle (32) is thus respected. Clausal pied-piping is a very marginal phenomenon in Italian (though quite common in other languages, such as Basque; Ortiz de Urbina 1989), whence the status of (43c). But the crucial point is that there is a clearly detectable di¤erence between the marginality of (43c) and the full ungrammaticality induced by a violation of Criterial Freezing in cases like (43b). Again, interpretive considerations seem to be irrelevant to rule out the crucial cases (41b) and (43b).6 In conclusion, the evidence presented in this section supports the view that there is a certain symmetry between s-selection and criterial positions. Consider the principles introduced here: (20) S-selection is fulﬁlled by ﬁrst Merge. (32) A phrase meeting a criterion is frozen in place. To stress the symmetry with (20), we could restate (32) as (32 0 ). (32 0 ) A criterion is fulﬁlled by last Merge. Both kinds of positions appear to be unique in chains and serve to delimit them: s-selection positions initiate chains (of all kinds); criterial positions ¯ -)chains.7 terminate (A We should ask why uniqueness is guaranteed in the way expressed by (20) and (32) (or (32 0 )): couldn’t natural languages work the other way around, with scope-discourse properties expressed in ﬁrst Merge positions and s-selection properties expressed by last Merge? Notice that s-selection is typically a property of lexical elements, while scope-discourse or criterial semantics is typically expressed by a system of functional heads. The

On the Form of Chains

119

particular way in which (20) and (32) ensure uniqueness thus seems to reﬂect the fact that lexical layers are generally embedded under functional layers, not the other way around.8 5.8

The EPP and the Subject Criterion

The classical EPP of the GB framework expresses the obligatoriness of the subject position of clauses. (46) Extended Projection Principle Clauses must have subjects. Direct evidence for this principle is provided by the very existence of expletives of various kinds. In structures in which there is no external argument, the subject position must be expressed by a nonreferential pronominal element (overt in non-null-subject languages). (47) a. There came a man b. It seems that John left (48) a. Il est entre´ une ﬁlle dans la salle it is entered a girl into the room ‘There entered a girl (into) the room.’ b. Il semble que Jean est parti it seems that Jean is left ‘It seems that Jean has left.’ The EPP remained an isolated principle within the GB framework, and minimalism has not signiﬁcantly changed things (apart from the terminological choice of generalizing the label EPP to the c-selection of a Spec potentially targeted by movement, a property more perspicuously renamed OCC(urrence) in Chomsky 2004; in what follows, I continue to use the term in its classical sense (46)). In this section, I would like to cap¯ -chains to propose a new way to italize on the results of the analysis of A integrate the EPP within an independent theoretical structure. I would also like to use this approach to reanalyze the subject-object asymmetries, classically ascribed to the ECP. Under the VP-internal subject hypothesis, the EPP position is typically ﬁlled by the external argument, if there is one, raised from its y-position. (49) John has [t seen Mary] The EPP position di¤ers from other A-positions, such as the position determining past participle agreement, in that the EPP position typically

120

Luigi Rizzi

is a ﬁnal target of movement, while the participial agreement position is not (see (50d)); the latter can only be the target of intermediate move¯ -position (as in the ment, with the relevant phrase further moved to an A relative construction in (50a)), or a clitic position (as in (50b)), or the EPP position itself (as in a passive construction like (50c)). la chaise que nous avons t 0 repeinte t the chair that we have repaintedþagr b. Nous l’avons t 0 repeinte t we it have repaintedþagr ‘We repainted it.’ c. La chaise a e´te´ t 0 repeinte t the chair has been repaintedþagr d. *Nous avons la chaise repeinte t we have the chair repaintedþagr

(50) a.

Under ‘‘movement as a last resort’’ guidelines, the EPP position must be associated to an interface e¤ect, morphological (Case-related), interpretive (some kind of criterial e¤ect), or both. Nominative Case assignment is often o¤ered as a morphological motivation for subject movement. But various considerations suggest an at least partial dissociation between the EPP position and the Case-agreement system (here I rely in part on Cardinaletti’s (2004) discussion). One is the existence of quirky (nonnominative) subjects illustrated here in Italian, with the preverbal subject position ﬁlled by an inherently Case-marked phrase and agreement construed with an internal argument, as in (51a), alternating with the canonical nominative-dative order, as in (51b). (51) a. A Gianni piacciono queste idee to Gianni please these ideas b. Queste idee piacciono a Gianni these ideas please to Gianni ‘These ideas please Gianni.’ That the dative experiencer actually is in a subject position is suggested by the fact that it does not induce any sort of island e¤ect, contrary to the e¤ect determined by a topic PP (the e¤ect is mild in Italian, further weakened if the subject is null; but even in this case there is a detectable di¤erence, for example, between (52a) and (52c)). (52) a.

Le idee che a Gianni piacciono di piu` sono queste the ideas that to Gianni please most are these ‘The ideas that please Gianni most are these.’

On the Form of Chains

121

b. ??Le idee che a Gianni Maria raccomanda sono queste the ideas that to Gianni Maria recommends are these ‘The ideas that Maria recommends to Gianni are these.’ c. ?Le idee che a Gianni raccomandiamo sono queste the ideas that to Gianni we recommend are these ‘The ideas that we recommend to Gianni are these.’ As the element satisfying the EPP and the agreeing element can be dissociated, I will assume two distinct heads as in (53), each related to a distinct DP position. (This follows from the natural assumption that a single head cannot carry a complex feature speciﬁcation, allowing it to enter into an Agree relation with a phrase and at the same time attract another phrase. That distinct subject positions should be postulated in di¤erent zones of the IP space is argued for by Belletti (2001, 2004), Cardinaletti (1997, 2004), and Shlonsky (2000) for reasons independent from the EPP; I will not discuss here how the positions they argue for relate to the EPP. I use the label Agr to designate the head carrying the agreement features, without committing myself to the existence of an independent Agr head; see Chomsky 1995, 2001, 2004 for discussion.) (53) . . . EPP . . . Agr . . . DP1 . . . DP2 At this point, we should inquire how the two heads can relate to distinct nominals without violating locality (in the sense of Rizzi 1990 and Chomsky 1995). Many options come to mind, but for concreteness I will assume that no minimality e¤ect arises in the derivation of (51a–b) because with the relevant verb class (the core of which consists of experiencing verbs), the small VP (containing the theme) can be scrambled past the (dative-marked) experiencer.9 (54) a. . . . EPP . . . Agr . . . [DPExp . . . [V DPTheme ]] ! b. . . . EPP . . . Agr . . . [V DPTheme ] . . . [DPExp . . . t] As the two DPs now belong to disjoint c-domains in (54b), Agree can relate Agr and DPTheme , and DPExp can be attracted to EPP, which gives rise to no Relativized Minimality violation.10 If the EPP position and the Case-agreement system are dissociated along these lines, we are naturally led to explore the possibility that the EPP position is a criterial position. In fact, it is often mentioned in the literature that some special interpretive property goes with the EPP position (a conclusion supported, for instance, by the intuition that an active and a passive sentence do not sound fully interchangeable). What special interpretive property could be associated with the EPP position? One frequently mentioned idea is that subjects are ‘‘topiclike’’

122

Luigi Rizzi

in some sense. A complete assimilation of subjects and topics is much too coarse for many reasons: even in a null subject language like Italian, preverbal subjects are possible in ‘‘out of the blue’’ contexts (e.g., as answers to ‘‘What happened?’’-type questions), in both active and passive structures ((55b) and (55b 0 )), while in such contexts bona ﬁde topics, expressed by clitic left-dislocation in Italian, are not felicitous (55b 00 ). Che cosa e` successo? ‘What happened?’ b. Un camion ha tamponato l’autobus per Roma a truck has bumped-into the bus for Rome ‘A truck bumped into the bus for Rome.’ b 0 . L’autobus per Roma e` stato tamponato da un camion the bus for Rome is been bumped-into by a truck ‘The bus for Rome was bumped into by a truck.’ b 00 . aL’autobus per Roma, un camion lo ha tamponato the bus for Rome a truck it has bumped-into ‘The bus for Rome, a truck bumped into it.’

(55) a.

Still, subject and topic have something in common: some kind of predication is involved in both cases, a process selecting an argument as the starting point of the event description and expressing the event as somehow involving that argument. Following a rather standard terminology, I will call this relation aboutness. So, the crashing event expressed in (55b) and (55b 0 ) is presented as being ‘‘about’’ a truck in (55b) and ‘‘about’’ the bus to Rome in (55b 0 ); in neither case does the nominal have the full topic interpretation of a left-dislocated phrase. Subject and topic share aboutness, but the two notions di¤er in that D(iscourse)-linking is a necessary component of topics (whence the illformedness of (55b 00 )) but not of subjects (on the notion of D-linking, see Pesetsky 1987, Cinque 1990, and much subsequent work). (56) Topic: þaboutness þD-linking (57) Subject: þaboutness D-linking I will then assume that the EPP is a manifestation of a Subject Criterion. To implement this hypothesis, I will follow Cardinaletti (2004) in assuming a nominal head in the higher functional ﬁeld, which attracts a nominal expression and determines the subject-predicate articulation, much as other dedicated heads determine topic-comment, focus-presupposition, and so on.

On the Form of Chains

123

(58) [XP [Top YP]] (59) [XP [Foc YP]] (60) [XP [Subj YP]] Subj occurs in the higher part of the inﬂectional ﬁeld, higher than Agr, or whatever head carries the agreement features if independent agreement heads do not exist (Chomsky 1995). (61) . . . Subj . . . Agr . . . One may speculate that a morphological realization of the Subj head is the subject clitic, an obligatory component of the clausal structure in some systems such as certain Northern Italian dialects, a position that appears to be distinct from, and higher than, the position occupied by the tensed verb, also carrying agreement features.11 (62) Le ragazze le son venute the girls subj.cl are come ‘The girls came.’ (adapted from Brandi and Cordin 1989)

(Northern Italian dialect)

The Subject Criterion may be satisﬁed in an interpretively vacuous way by the insertion (or movement) of an expletive as the speciﬁer of Subj. The possibility of such a purely formal satisfaction is sometimes considered a stumbling block for any approach to the EPP based on special interpretive properties of the subject position. I don’t think it is. It’s typical of natural languages to introduce interpretively motivated devices that, as it were, acquire a formal life of their own and operate blindly in special cases even in the absence of the interpretive conditions. Grammatical gender is a case in point—related to natural gender, but extending well beyond natural gender in determining formal nominal classes. So, expletive insertion corresponds to the case in which the speaker chooses not to present an event as being about a certain argument (an option limited by formal conditions that vary considerably across languages). See Chomsky 2002, 117, for discussion speciﬁcally related to the EPP. 5.9

The EPP and the ECP

The EPP and the classical ECP have the common property of singling out the subject position of clauses: the EPP says that the subject position is obligatory; the ECP says that the subject position is unmovable (in certain environments that permit object extraction, such as embedded clauses).

124

Luigi Rizzi

(63) a. *Who do you think [that [t Subj will come]]? b. Who do you think [that [Mary Subj will meet t]]? (64) a. *Qui who b. Qui who

crois-tu do you believe crois-tu do you believe

[que [t Subj viendra]]? that will come [que Marie rencontrera t]]? that Marie will meet

We can now connect the two properties: if the EPP requirement can be reduced to a criterion, we expect subjects to be unmovable, under Criterial Freezing (32). Who must move to Spec,Subj to fulﬁll the Subject Criterion, but then it is frozen there, unavailable for further movement. Subject-object asymmetries thus reduce to the fact that there is a Subject Criterion, but no Object Criterion. The possibility of moving the subject in certain environments now amounts to the ability of the thematic subject to avoid moving to the EPP position, thus escaping the freezing e¤ect. The most straightforward way to achieve this result is to ﬁll the criterial subject position with an expletive, thus making the thematic subject free to move from a lower position accessible to movement. The traditional analysis of the lack of asymmetries in null subject languages (Rizzi 1982, chap. 4) is straightforwardly interpretable in these terms: expletive pro fulﬁlls the Subject Criterion in (65c); hence, the thematic subject remains available for extraction from a noncriterial position. (65) a. Chi credi che verra`? who do you think that will come b. Credo che verra` Gianni I think that will come Gianni c. Chi credi [che [pro verra` t]] This option is also available in non–null subject languages—more marginally, though, because of the more stringent limitations on the occurrence of overt expletives like there. (66) a. I think that [something is [t in the box]] b. I think that [there is [something in the box]] (67) a. *What do you think that t is in the box? b. What do you think that there is t in the box? The critical point of the derivation is this: (68) Subj be [DP in the box]

On the Form of Chains

125

If DP is raised to Spec,Subj to satisfy the Subject Criterion, it is frozen there, under Criterial Freezing (32), and further movement as in (67a) is impossible. If the Subject Criterion is fulﬁlled by the expletive as in (66b), the DP subject of the small clause is accessible to further movement. Were one to adopt a theory of expletives as pro-predicates, as in Moro 1997, the result would be the same.12 The strategies for subject extraction that di¤erent languages use can now be seen as strategies for avoiding the freezing e¤ect connected to the Subject Criterion. See Rizzi and Shlonsky, in preparation, for an analysis of some major cases. 5.10

Conclusions

¯ -chains consist of two privileged positions that express two kinds of A interpretive properties relevant at the interface: the s-selection position and the criterial position. We have discussed evidence suggesting that these positions are unique and that they initiate and terminate chains, respectively. As a result, expressions cannot ‘‘pick up’’ y-roles, or other sselection properties, through movement, and criteria cannot be satisﬁed ‘‘in passing.’’ In particular, a phrase satisfying a criterion is frozen in place, unavailable for further movement. The principles determining this ¯ -chains optimally simple, with state of a¤airs have the e¤ect of keeping A unique occurrences of the basic essential ingredients. If s-selection and criterial positions delimit the chain, excluding additional positions lower than the former and higher than the latter, chains can include indeﬁnitely many positions intermediate between the two positions relevant at the interface. This is due to the operation of locality principles severely constraining the structural span of single chain links. If ‘‘movement as a last resort’’ and locality are to be reconciled, intermediate movement must be possible. We have assumed it can be triggered by purely formal equivalents of substantive criterial features, which serve to make chain connections possible between s-selection positions and (potentially remote) criterial positions in accordance with locality. Finally, we have considered why ‘‘subjects are special,’’ in that they are singled out by both the EPP and the ECP. If the EPP is to be understood as a particular criterion, the Subject Criterion, selecting an argument as the starting point in the description of the event, the immovability of subjects expressed by the ECP can be understood as a consequence of Criterial Freezing.

126

Luigi Rizzi

Notes For helpful comments, I would like to thank Adriana Belletti, Lisa Cheng, Noam Chomsky, Guglielmo Cinque, and Norbert Corver. ¯ -criteria involving the left periphery of 1. In this chapter, I will limit attention to A the clause, hence to constructions subsumed under wh-movement in Chomsky ¯ -positions that have been analyzed 1977; I will not try to integrate IP-internal A through criteria, such as negation (Haegeman 1995). ¯ -feature and 2. One could envision that there is only one, nonspeciﬁc formal A ¯ -chain is determined only when the moved that the nature of the particular A phrase reaches a criterial head. Evidence against this hypothesis is the fact that intermediate positions give rise to selective e¤ects: for example, embedded I-to-C movement in Belfast English is triggered only in questions like (21), not in topicalization or other constructions where the criterial head does not trigger I-to-C movement. (i) *My car, Mary claimed [t did [they steal t]] (Alison Henry, pers. comm.) Facts like this argue for separate formal features mirroring in part the properties of the respective criterial features. ¯ -feature is ‘‘valued’’ by the One could still envision that a nonspeciﬁc A attracted phrase as q, r, top, and so on, and then gives rise to distinct morphological reﬂexes accordingly. But this would lead to the expectation of minimality ¯ -extractions triggered by the nonspeciﬁc feature (e.g., e¤ects across types of A a Focus position should give rise to a minimality e¤ect on Topic extraction), ¯e¤ects that are not actually found: locality e¤ects are highly selective across A constructions (Rizzi 2004a). 3. It could be that the Q-feature and the contrastive focus feature are assigned to two distinct elements within the phrase, the wh-operator quale ‘which’ and the lexical restriction ragazza ‘girl’, respectively. If so, the conﬁguration would reduce to another subcase of case (30d), to be discussed directly, and the conclusions would remain unchanged. A reviewer points out that Criterial Freezing (32) has e¤ects analogous to those of Mu¨ller and Sternefeld’s (1994) Principle of Unambiguous Binding (at least for the cases of (30) in which distinct features are involved). 4. Starke (2001) proposes that wh-extraction of adnominal complements is only possible when they are speciﬁc, D(iscourse)-linked. In the reﬁned approach to relativized minimality proposed in Rizzi 2004a and Starke 2001, such phrases are attracted by a C head with a complex feature speciﬁcation—let us say [þQ, þTop], for concreteness. Because such a speciﬁcation is needed for extraction from a wh-island in this system, extraction of the remnant in (37b) gives rise to a relativized minimality e¤ect (attraction of [þQ, þTop] over a [þQ, þTop] element in the embedded C system). No such problem arises in the derivation of (36b), as wh-movement to the embedded C can involve attraction by a ‘‘pure’’ [þQ] feature. This approach may not be generalizable to the other cases discussed by Cecchetto (2001). Notice that the asymmetry in (37) also follows from the constraints on remnant movement proposed by Mu¨ller (1998) and Grewendorf

On the Form of Chains

127

(2003): according to the latter, the remnant may only be attracted by a feature higher than the feature attracting the extracted element in a feature hierarchy. 5. With some degradation induced by the quantiﬁer introducing the pied-piped phrase: compare (42b) with the more natural (i). (i) Gianni, i libri del quale credo che siano stati censurati Gianni the books by whom I believe that have been censored The crucial point is that the degradation of (42b) is mild compared to the total unacceptability of (41b); therefore, the status of the latter cannot be attributed to whatever factor causes degradation when the pied-piped phrase has an operator in its Spec. 6. The simplest case of the freezing e¤ect, (31b), would seem to follow immediately from the view that an uninterpretable feature on the goal is required in ¯ -chains to make the element ‘‘active’’ and available for attraction, much as in AA chains according to the analysis in Chomsky 2000, 2001 (see discussion in section 5.6). Such a feature would be deleted when the wh-element is moved to the embedded C system in (31a); hence, the element would cease to be active for further attraction. This approach does not seem to be immediately generalizable to the more complex cases in which distinct elements in the same phrase can be attracted to two distinct criterial positions. Consider for instance (40)–(41). Under the ‘‘inactivation’’ approach, both quanti and quale in the complex phrase would bear an uninterpretable feature, making them available for attraction. After quanti is attracted to the lower C (pied-piping the whole complex phrase), quale, still ‘‘active,’’ is attracted to the relative C. Why could it not, in turn, pied-pipe the whole complex phrase to yield (41b)—an operation it can perform in other structures, such as (42b)? Movement to the Relative Criterion position can involve pied-piping of a phrase smaller than the phrase satisfying the Question Criterion, as in (41a) and (43a), or larger than it, as in (43c), but not pied-piping of exactly the same phrase, as in (41b) and (43b), a fact that does not follow from the inactivation approach unless auxiliary assumptions are made. Possible reﬁnements may involve the sharpening of the notion ‘‘pied-piping.’’ I will not discuss the inactivation approach any further here. 7. One could perhaps understand the freezing e¤ect of (32) in terms of Chomsky’s cyclic Spell-Out. It could be that as soon as a phrase meets a criterion, it is sent to the interfaces, thus becoming unavailable for further movement, a rather natural economy principle. Developing this view would require some care, though: we have seen that part of the phrase involved in the satisfaction of a criterion remains available for further movement, as in the splitting illustrated by (41a) and (43a). So the freezing e¤ect is limited to the element actually carrying the feature involved in the satisfaction of the criterion (a more accurate wording of Criterial Freezing would then be something like ‘‘In a speciﬁer-head criterial conﬁguration, the element bearing the criterial feature in the speciﬁer position is frozen in place’’). I will not try here to integrate ‘‘partial movement’’ constructions (e.g., McDaniel 1989), in which a wh-operator is allowed to stop in an intermediate, noncriterial position.

128

Luigi Rizzi

8. In many languages and constructions, Foc and Q appear to be associated. In Gungbe, for instance, wh-phrases are moved to the overtly marked Foc position (Aboh 1998, 2004; the two positions can also be dissociated, on which see Rizzi 1997, 330n18). I will assume that this is made possible by head movement of Q to Foc, creating the complex head QþFoc, which then attracts the whphrase. This is only possible when Q and Foc belong to the same C system (note the ill-formedness of (34b), etc.), owing to the strongly local character of head movement. The cleft construction clearly permits further movement of the focused element (It is John that you saw t, Who is it t that you saw t?). To make this compatible with the analysis in the text, one could assume either that clefts are not to be assimilated to criterial constructions, or that the freezing e¤ect is limited to peripheral criterial positions (see also note 1; in view of the analysis to be developed in sections 5.7–5.9, the subject position should count as ‘‘peripheral’’ in this respect). 9. This derivational option is not available with agentive verbs if the small-VP scrambling takes place to a position lower than the one where the agent y-role is assigned. The alternation in (51) is not available with psych-verbs of the preoccupare ‘worry’ class, which allow only the order in (51b), if the inherent accusative Case assigned to the experiencer with this verb class is stuck in the position of assignment (Chomsky 1986; Belletti and Rizzi 1988); the dative of (51) is more mobile, as the preposition/assigner can be pied-piped in this case. This analysis is akin to (and inspired by) Cinque’s (2002) analysis of the right-peripheral position of adverbial PPs (in turn inspired by Kayne 2001, chaps. 14, 15), which also involves leftward movement of a verbal projection, except that the movement proposed by Cinque a¤ects a larger structure and targets a higher position, outside the thematic layers. 10. Note that EPP and Agr could not occur in the opposite order: in that case, the nominal closest to Agr would always be the one already attracted to the EPP position, and an Agree relation with another nominal could never be established. Much recent work suggests that person and number features of agreement have di¤erent roles and may well occupy di¤erent structural positions (Poletto 2000; SigurDsson 2000; Bianchi 2003; Di Domenico 2003). I omit these (potentially very relevant) reﬁnements here. 11. The text discussion should be reﬁned in various respects. In particular, an unqualiﬁed reference to ‘‘subject clitics in Northern Italian dialects’’ is an oversimpliﬁcation in that di¤erent subject clitics seem to occupy di¤erent positions in di¤erent varieties according to their feature speciﬁcation (Poletto 2000; Manzini and Savoia 2004). 12. In raising constructions, an embedded subject moves to the main subject position. How can it escape the freezing e¤ect in the embedded clause? Raising is clearly incompatible with the CP system (there is no raising from indirect questions, for instance), so it appears to take place from ‘‘truncated’’ clauses (undergoing ‘‘S 0 -deletion’’ in classical GB terms). We can now assume that the ‘‘truncation’’ also involves the higher part of the IP system, the Subj head, so that no freezing e¤ect arises (this is, in a sense, the analogue of Chomsky’s (2001, 2004)

On the Form of Chains

129

assumption about the ‘‘defective’’ character of the inﬂectional system in raising). Exceptional Case-marking clauses seem to be amenable to a similar analysis. Control clauses, on the other hand, are complete clausal structures with a C system and a full IP structure; hence, PRO is always frozen in Spec,Subj of the control clause under Criterial Freezing, and never raises further. Wh-extraction from ﬁnite complements in cases like Who do you think came? may in turn be amenable to a truncation of the CPþSubjP complex, in partial similarity to the cases of noncontrol inﬁnitives. On French que-qui, see the expletive analysis in Taraldsen 2001, which makes that case similar to subject extraction in null subject languages. References Aboh, Enoch. 1998. From the syntax of Gungbe to the grammar of Gbe. Doctoral dissertation, University of Geneva. Aboh, Enoch. 2004. The morphosyntax of complement-head sequences. New York: Oxford University Press. Baker, Mark. 2002. Building and merging, not checking: The nonexistence of (Aux)-S-V-O languages. Linguistic Inquiry 33, 321–328. Bayer, Josef. 1984. Comp in Bavarian. The Linguistic Review 3, 209–274. Belletti, Adriana. 1990. Generalized verb movement. Turin: Rosenberg and Sellier. Belletti, Adriana. 2001. ‘‘Inversion’’ as focalization. In Aafke Hulk and Jean-Yves Pollock, eds., Subject inversion in Romance and the theory of Universal Grammar, 60–90. New York: Oxford University Press. Belletti, Adriana. 2004. Aspects of the low IP area. In Luigi Rizzi, ed., The structure of CP and IP: The cartography of syntactic structures, vol. 2, 16–51. New York: Oxford University Press. Belletti, Adriana, and Luigi Rizzi. 1988. Psych-verbs and y-theory. Natural Language and Linguistic Theory 6, 291–352. Bianchi, Valentina. 1999. Consequences of antisymmetry: Headed relative clauses. Berlin: Mouton de Gruyter. Bianchi, Valentina. 2003. On ﬁniteness and logophoric anchoring. Ms., University of Siena. Brandi, Luciana, and Patrizia Cordin. 1989. Two Italian dialects and the null subject parameter. In Osvaldo Jaeggli and Kenneth Saﬁr, eds., The null subject parameter, 111–142. Dordrecht: Kluwer. Burzio, Luigi. 1986. Italian syntax: A government-binding approach. Dordrecht: Reidel. Cardinaletti, Anna. 1997. Subjects and clause structure. In Liliane Haegeman, ed., The new comparative syntax, 33–63. London: Longman. Cardinaletti, Anna. 2004. Toward a cartography of subject positions. In Luigi Rizzi, ed., The structure of CP and IP: The cartography of syntactic structures, vol. 2, 115–165. New York: Oxford University Press.

130

Luigi Rizzi

Cecchetto, Carlo. 2001. Proper Binding Condition e¤ects are Phase Impenetrability Condition e¤ects. In Minjoo Kim and Uri Strauss, eds., Proceedings of NELS 31, 99–115. Amherst: University of Massachusetts, GLSA. Cheng, Lisa L.-S. 1991. On the typology of wh-questions. Doctoral dissertation, MIT. Chomsky, Noam. 1957. Syntactic structures. Mouton: The Hague. Chomsky, Noam. 1973. Conditions on transformations. In Stephen Anderson and Paul Kiparsky, eds., A festschrift for Morris Halle, 232–286. New York: Holt, Rinehart and Winston. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1986. Knowledge of language. New York: Praeger. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays in minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2002. On nature and language. Cambridge: Cambridge University Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Chung, Sandra. 1998. The design of agreement: Evidence from Chamorro. Chicago: University of Chicago Press. Cinque, Guglielmo. 1990. Types of A¯-dependencies. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford: Oxford University Press. Cinque, Guglielmo. 2002. Issues in adverbial syntax. Ms., University of Venice. Di Domenico, Elisa. 2003. Placed, non-placed and anaphorically placed expressions: On the role of tense/person inﬂection in human language. Ms., University of Siena. Emonds, Joseph. 1978. The verbal complex V 0 -V in French. Linguistic Inquiry 9, 151–175. Fox, Danny. 1995. Economy and scope. Natural Language Semantics 3, 283– 341.

On the Form of Chains

131

Grewendorf, Gu¨nther. 2003. Improper remnant movement. Gengo Kenkyu— Journal of the Linguistic Society of Japan 123, 47–94. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10, 279–326. Haegeman, Liliane. 1994. An introduction to Government-Binding Theory. Oxford: Blackwell. Haegeman, Liliane. 1995. The syntax of negation. Cambridge: Cambridge University Press. Henry, Alison. 1995. Belfast English and Standard English: Dialect variation and parameter setting. Oxford: Oxford University Press. Hornstein, Norbert. 1999. Movement and control. Linguistic Inquiry 30, 69–96. Kayne, Richard. 1989. Facets of Romance past participle agreement. In Paola Beninca`, ed., Dialect variation and the theory of grammar, 85–103. Dordrecht: Foris. Kayne, Richard. 2001. Parameters and universals. Oxford: Oxford University Press. Kayne, Richard. 2002. Pronouns and their antecedents. In Samuel David Epstein and T. Daniel Seely, eds., Derivation and explanation in the Minimalist Program, 133–166. Oxford: Blackwell. Kayne, Richard, and Jean-Yves Pollock. 1978. Stylistic inversion, successive cyclicity, and Move NP in French. Linguistic Inquiry 9, 595–621. Koster, Jan. 2003. All languages are tense second. Ms., University of Groningen. Landau, Idan. 2003. Movement out of control. Linguistic Inquiry 34, 471–498. Langendoen, D. Terence, and Edwin Battistella. 1982. The interpretation of predicate reﬂexive and reciprocal expressions in English. In James Pustejovsky and Peter Sells, eds., Proceedings of NELS 12, 163–173. Amherst: University of Massachusetts, GLSA. Lasnik, Howard, and Mamoru Saito. 1992. Move a: Conditions on its application and output. Cambridge, Mass.: MIT Press. Lebeaux, David. 1988. Language acquisition and the form of grammar. Doctoral dissertation, University of Massachusetts, Amherst. Manzini, Maria Luisa, and Leonardo Savoia. 2004. Clitics: Cooccurrence and mutual exclusion patterns. In Luigi Rizzi, ed., The structure of CP and IP: The cartography of syntactic structures, vol. 2, 211–250. New York: Oxford University Press. McCloskey, James. 1996. On the scope of verb movement in Irish. Natural Language and Linguistic Theory 14, 47–104. McCloskey, James. 2002. Resumption, successive cyclicity, and the locality of operations. In Samuel David Epstein and T. Daniel Seely, eds., Derivation and explanation in the Minimalist Program, 184–226. Oxford: Blackwell. McDaniel, Dana. 1989. Partial and multiple wh-movement. Natural Language and Linguistic Theory 7, 565–604.

132

Luigi Rizzi

Moro, Andrea. 1997. The raising of predicates: Predicative noun phrases and the theory of clause structure. Cambridge: Cambridge University Press. Mu¨ller, Gereon. 1998. Incomplete category fronting. Dordrecht: Kluwer. Mu¨ller, Gereon, and Wolfgang Sternefeld. 1994. Scrambling as A-bar movement. In Norbert Corver and Henk van Riemsdijk, eds., Studies in scrambling, 331–385. Berlin: Mouton de Gruyter. Ortiz de Urbina, Jon. 1989. Parameters in the grammar of Basque. Dordrecht: Foris. Pesetsky, David. 1982. Paths and categories. Doctoral dissertation, MIT. Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Eric Reuland and Alice G. B. ter Meulen, eds., The representation of (in)deﬁniteness, 98–129. Cambridge, Mass.: MIT Press. Pesetsky, David, and Esther Torrego. 2001. T-to-C movement: Causes and consequences. In Michael Kenstowicz, ed., Ken Hale: A life in language, 355–426. Cambridge, Mass.: MIT Press. Poletto, Cecilia. 2000. The higher functional ﬁeld: Evidence from Northern Italian dialects. New York: Oxford University Press. Pollock, Jean-Yves. 1989. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20, 365–424. Reinhart, Tanya. 1997. Wh-in-situ in the framework of the Minimalist Program. Natural Language Semantics 6, 29–56. Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1996. Residual verb second and the Wh-Criterion. In Adriana Belletti and Luigi Rizzi, eds., Parameters and functional heads: Essays in comparative syntax, 63–90. Oxford: Oxford University Press. Reprinted in Rizzi 2000. Rizzi, Luigi. 1997. The ﬁne structure of the left periphery. In Liliane Haegeman, ed., Elements of grammar, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2000. Comparative syntax and language acquisition. London: Routledge. Rizzi, Luigi. 2004a. Locality and left periphery. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 223–251. New York: Oxford University Press. Rizzi, Luigi, ed. 2004b. The structure of CP and IP: The cartography of syntactic structures, vol. 2. New York: Oxford University Press. Rizzi, Luigi, and Ur Shlonsky. In preparation. Strategies of subject extraction. Ms., University of Siena and University of Geneva. Rothstein, Susan. 1983. The syntactic forms of predication. Doctoral dissertation, MIT. Shlonsky, Ur. 2000. Subject positions and copular constructions. In Hans Bennis, Martin Everaert, and Eric Reuland, eds., Interface strategies, 325–347. Amsterdam: Royal Netherlands Academy of Arts and Sciences.

On the Form of Chains

133

SigurDsson, Halldo´r. 2000. The locus of Case and agreement. Working Papers in Scandinavian Syntax 65, 65–108. Starke, Michal. 2001. Merge dissolves into Move. Doctoral dissertation, University of Geneva. Taraldsen, Knut Tarald. 2001. Subject extraction, the distribution of expletives, and stylistic inversion. In Aafke Hulk and Jean-Yves Pollock, eds., Subject inversion in Romance and the theory of Universal Grammar, 163–182. New York: Oxford University Press.

Chapter 6 On Overt and Covert Whand Relative Movement in Hindi and Punjabi

6.1

Henrietta Yang

Introduction

Since the publication of ‘‘On Wh-Movement’’ (Chomsky 1977), movement operations in syntax have been studied intensively (Huang 1982; Rudin 1988; Cheng 1991; Dayal 1996; Richards 1997; Bosˇkovic´ 2001, among many others). Movement operations appear to fall into two general types: covert movement and overt movement. If question or relative phrases stay in their base position in overt syntax, they are assumed to undergo covert movement. This kind of movement is illustrated in the following examples from Hindi:1 (1) siggi kis-ko pasand kartaa hai? Siggi who-acc like do is ‘Who does Siggi like?’ (question formation) (2) vo laRkii [rohit jis-ko pasand kartaa hai] hoshiyaar hai dem girl Rohit rel-acc like do is smart is ‘The girl (who) Rohit likes is smart.’ (relative construction) In contrast, when question or relative phrases are displaced from their base position in overt syntax, they are assumed to have undergone overt movement, as in English. (3) Whoi is she talking to t i ? (question formation; hereafter wh-movement) (4) The girl [whoi Rohit likes t i ] is smart (relative construction; hereafter rel-movement) Basing his argument on the homophony of relative and question phrases and the similar syntax of questions and relative clauses in English,

136

Henrietta Yang

Chomsky (1977) proposed a uniﬁed treatment of wh-movement and relmovement. However, the di¤erences between these constructions found in two closely related modern Indo-Aryan languages, Hindi and Punjabi, raise the question of whether a uniﬁed treatment is appropriate. The goal of this chapter is twofold: (1) to argue that apparently related constructions do not necessarily entail an identical movement operation; however, a uniﬁed analysis can be reached when some additional assumptions are taken into account, and (2) to o¤er empirical evidence that both the PF and LF interfaces constrain the overt realization of question and relative phrases in overt syntax; this ﬁnding will provide support for the copy theory (single-output syntax) as argued for by Bobaljik (1995, 2002). The rest of the chapter is organized as follows. In section 6.2, I provide background information regarding relative clause and question constructions in Hindi and Punjabi and brieﬂy discuss the basic ideas of Bobaljik’s copy theory. In section 6.3, I show that the movement strategy used in Punjabi rel-movement is di¤erent from that used in wh-movement; the ¯ -movement; the latter, scope marking. I further former involves cyclic A show that the ‘‘partial movement’’ phenomenon found in Punjabi relative clauses supports a cyclic Spell-Out approach and that the multiple realization phenomenon is constrained by a PF economy condition, the CPPhase Constraint. In section 6.4, I discuss an asymmetry regarding PF Relaxation between local and long-distance movement in Hindi and demonstrate that unifying these two movement operations requires additional assumptions about the realization of chains. I refer to the relevant assumptions as PF Relaxation (the pronunciation of either the higher or the lower copy of a relative or question phrase inside a ﬁnite clause) and Minimize Mismatch (to the extent possible, privilege the same copy at PF and LF) (Bobaljik 2002). I further suggest that PF Relaxation and Minimize Mismatch take place in a complementary fashion that accounts for the distinct realization pattern of relative and question phrases in Hindi and Punjabi. Section 6.5 concludes the chapter. 6.2

Background Information

6.2.1 Some Essential Properties of Relative Constructions and Questions in Hindi In both Hindi and Punjabi,2 unlike English, relative words and question words are not the same. The former usually begin with a j-sound, whereas the latter begin with a k-sound. There are three possible positions for ﬁnite relative clauses in both languages: (1) preceding the main clause (also

Overt and Covert Wh- and Relative Movement

137

known as the correlative construction; Dayal 1996), (2) immediately following the head noun (also called embedded relatives), and (3) following the main clause (also called extraposed relatives). In this chapter, I concentrate on embedded relative clauses in both languages (see (2)). Unlike in English, in Hindi the relative phrase (rel-phrase) is obligatory. Movement of the relative phrase is possible but not obligatory.3 If the relativization does not cross a ﬁnite CP, native speakers of Hindi allow the relative phrase to remain in situ, as shown in (5a), though they ﬁnd it more natural to front the relative phrase as in (5b). (5) a. vo laRkii [CP rohit *( jis-ko) pasand kartaa hai] hoshiyaar hai dem girl Rohit rel-acc like do is smart is ‘The girl (who) Rohit likes is smart.’ (no fronting; rel-in-situ) b. vo laRkii [CP jis koi rohit t i pasand kartaa hai] hoshiyaar hai dem girl rel-acc Rohit like do is smart is ‘The girl (who) Rohit likes is smart.’ (fronting; more natural) However, if the relativization crosses a ﬁnite CP, then the relative phrase must move to the highest CP in the relative clause. (6) a.

vo laRkii [CP jis-koi mona sochtii hai [CP t i ki rani t i dem girl rel-acc Mona think is that Rani pasand kartii hai]] like do is ‘the girl who Mona thinks that Rani likes’ b. *vo laRkii [CP mona sochtii hai [CP ki rani jis-ko pasand dem girl Mona think is that Rani rel-acc like kartii hai]] do is c. *vo laRkii [CP mona sochtii hai [CP jis-koi ki rani t i dem girl Mona think is rel-acc that Rani pasand kartii hai]] like do is

The examples in (6) indicate that full fronting is required. (6b) shows that the option of rel-in-situ is not available in embedded ﬁnite clauses. Neither is the option of partial movement (i.e., rel-phrases are not fully fronted), as shown in (6c). Like movement of the rel-phrase, movement of the wh-phrase in questions is possible but not obligatory. (7a) is an example of wh-in-situ,

138

Henrietta Yang

which is more natural to native speakers of Hindi, and (7b) shows a fronted wh-phrase. (7) a. siggi kis-ko pasand kartaa hai Siggi who-acc like do is ‘Who does Siggi like?’ b. kis-koi siggi t i pasand kartaa hai who-acc Siggi like do is ‘Who does Siggi like?’ (Mahajan 1990) Furthermore, in questions, as in relative clauses, wh-in-situ is not always permitted. If the question formation crosses a ﬁnite CP, then the whphrase must move out of its base-generated position. The wh-phrase in (8a) is in situ in the embedded ﬁnite clause and the matrix question reading is unavailable. However, the matrix question reading becomes available when the wh-phrase moves to the matrix clause, as in (8b). The matrix question reading is also possible when an invariant element, kyaa ‘what’, is inserted in the matrix clause while the wh-phrase remains inside the embedded clause (scope-marking strategy4), as in (8c). (8) a. jaun soctaa hai [CP ki meri kis-se baat karegii] John thinks is that Mary who-with will-talk ‘John wonders who Mary will talk to.’ not ‘Who does John think Mary will talk to?’ b. kis-sei jaun soctaa hai [CP ki meri t i baat karegii]? who-with John thinks is that Mary will-talk ‘Who does John think Mary will talk to?’ c. jaun kyaa soctaa hai [CP ki meri kis-se baat karegii]? John what thinks is that Mary who-with will-talk ‘Who does John think Mary will talk to?’ (what . . . wh-XP) 6.2.2 Some Essential Properties of Relative Constructions and Questions in Punjabi In Punjabi, as in Hindi, if the relativization does not cross a ﬁnite CP, it is optional to front the relative phrase in relative clauses. Unlike Hindi speakers, however, native speakers of Punjabi ﬁnd it more natural to use rel-in-situ. (9) a. o kuRii [CP rohit jeenu pasand karda hai] hoshiyaar hegi dem girl Rohit rel-acc like do is smart is ‘The girl who Rohit likes is smart.’

Overt and Covert Wh- and Relative Movement

139

b. o kuRii [CP jeenui rohit t i pasand karda hai] hoshiyaar hegi dem girl rel-acc Rohit like do is smart is ‘The girl who Rohit likes is smart.’ When a relative phrase originates in an embedded ﬁnite clause, it has the choice of either staying in situ as in (10a), moving partially to the Comp position in the embedded clause as in (10b), or moving to the matrix clause as in (10c).5 Notice that the counterparts of (10a–b) in Hindi, namely, (6b–c), are ungrammatical. (10) a. o kuRii [CP mona soch-di hagi [CP rani jeenu pasand dem girl Mona think is Rani rel-acc like kardi hai]] do is ‘the girl who Mona thinks Rani likes’ b. o kuRii [CP mona soch-di hagi [CP jeenui rani t i pasand dem girl Mona think is rel-acc Rani like kardi hai]] do is ‘the girl who Mona thinks Rani likes’ c. o kuRii [CP jeenui mona soch-di hagi [CP t i rani t i pasand dem girl rel-acc Mona think is Rani like kardi hai]] do is ‘the girl who Mona thinks Rani likes’ In question formation in Punjabi, if the wh-phrase does not cross a ﬁnite CP, then movement is possible but not obligatory, which is like relmovement in both Hindi and Punjabi, as shown in (11). (11) a. o kaun ai? she/he who is ‘Who is she/he?’ b. kauni o t i ai? who she/he is ‘Who is she/he?’ (Bhatia 1993) If question formation involves crossing a ﬁnite CP, wh-in-situ does not give a matrix question reading, as shown in (12a). In contrast to what we ﬁnd with wh-movement in Hindi, full fronting of a wh-phrase out of an embedded ﬁnite clause is not possible in Punjabi, as demonstrated in (12b).6 The only way to achieve a matrix question reading is through the

140

Henrietta Yang

scope-marking strategy, as shown in (12c). This strategy is also available in Hindi (see (8c)). (12) a.

mona sochi-di hagi [CP ki kal kaun jittegaa] Mona thinks is that tomorrow who win-fut.3msg ‘Mona wonders who will win tomorrow.’ not ‘Who does Mona think will win tomorrow?’ b. *kauni mona sochi-di hagi [CP kal t i jittegaa] who Mona thinks is tomorrow win-fut.3msg c. mona kii sochi-di hagi [CP ki kal kaun jittegaa]? Mona what thinks is that tomorrow who win-fut.3msg ‘Who does Mona think will win tomorrow?’ (what . . . wh-XP)

In sections 6.2.1 and 6.2.2, I have shown that Hindi and Punjabi are alike in cases where the relativization or question formation does not cross a ﬁnite CP. Movement operations in these two languages diverge once movement crosses a ﬁnite CP. 6.2.3 Bobaljik’s (2002) Copy Theory: Pronunciation of Chains Typically, in a nontrivial chain,7 the highest or the lowest copy is pronounced, as in the Hindi example in (13) and the Punjabi example in (14). (13) kis-sei jaun soctaa hai [ki meri t i baat karegii]? who-with John thinks is that Mary will-talk ‘Who does John think Mary will talk to?’ (The highest copy is pronounced.) (14) o kuRii [t i mona sochdi hagi [t i rani jeenui dem girl Mona think is Rani rel-acc pasand kardi hai]] like do is ‘the girl who Mona thinks Rani likes’ (The lowest copy is pronounced.)

(Hindi)

(Punjabi)

In Bobaljik’s (2002) approach, the syntax module of grammar generates a single representation that is interpreted by LF and PF. LF and PF are treated equally in the sense that they can both privilege a certain copy in a given chain. In other words, LF can choose to interpret the higher copy or the lower copy, and PF can choose to pronounce either copy as well. As a result, the distinction between overt and covert movement operations is just the matter of which copy is pronounced. Bobaljik (2002, 199) suggests that there is a four-way typology of movement operations,

Overt and Covert Wh- and Relative Movement

141

Table 6.1 Bobaljik’s (2002) four-way typology of movement Copy1. . . . . . . . . .Copy2 (high) . . . . . . . . . .(low) a. PF, LF b. LF c. PF d.

PF LF PF, LF

Example ‘‘overt’’ wh-movement in English ‘‘covert’’ wh-movement in Mandarin Chinese movement that undergoes reconstruction at LF object shift in the Scandinavian languages

demonstrated in table 6.1. Pattern (a), in which the higher copy is interpreted and pronounced, is the structure of ‘‘overt’’ movement in syntax such as the questions and relative constructions in (3)–(4). Pattern (b) is the structure of ‘‘covert’’ movement in which the higher copy is interpreted, but the lower copy is pronounced. This kind of movement is illustrated in (14). Pattern (c) is an instance of the reconstruction e¤ect (e.g., Barss 1986; Chomsky 1995), in which the higher copy is pronounced, while the lower copy renders the interpretation. This pattern is demonstrated in (15), an example given by Chomsky (1995, 74) regarding the binding condition. (15) a. [Which picture of himself ] did John say [t 0 that Bill liked t best]? b. [Which picture of each other] did they say [t 0 that we liked t best]? Here the antecedents for himself and each other must be Bill and we, respectively, even though both John and they are possible antecedent candidates. In order to capture the correct binding condition, both anaphors have to be interpreted at the lowest position, t. Finally, Bobaljik argues that object shift in Germanic languages conforms to the pattern in (d),8 in which both pronunciation and interpretation target the lower copy. A number of authors (see, e.g., Bobaljik 1995; Brody 1995; Pesetsky 1998) have questioned why PF always targets the highest copy, which Chomsky (1993) assumes without argument. They have convincingly shown that in certain cases PF can choose either the highest or the lowest copy of a nontrivial chain, a result that captures ‘‘overt’’ and ‘‘covert’’ movement in syntax. Two questions arise at this point: (1) Instead of the highest or the lowest copy being pronounced, is it possible for PF to choose to pronounce an intermediate copy of a nontrivial chain that consists of more than two copies? (2) Can PF allow more than one copy of a nontrivial chain? I address the ﬁrst question in section 6.3.1 and the second in section 6.3.2.

142

6.3

Henrietta Yang

Movement Operations in Punjabi Relative Clauses and Questions

As shown above, if we only examine wh-movement and rel-movement in a matrix ﬁnite CP domain, these two movement operations in Hindi and Punjabi seem to have the same properties. However, di¤erences arise when the constructions involve an embedded ﬁnite CP clause in which a rel-phrase or wh-phrase originates. In this section, I focus on movement in Punjabi and illustrate the di¤erences between these two operations. 6.3.1 Partial Movement Strategies As we saw in (10), repeated here as (16), the same rel-phrase appears in situ (16a), partially moved (16b), and fully fronted (16c). Partial movement in Punjabi relative clauses refers to movement of the rel-phrase out of its base position but not into a fully fronted position as in (16b). (16) a. o kuRii [mona sochdi hagi [rani jeenu pasand kardi hai]] dem girl Mona think is Rani rel-acc like do is ‘the girl who Mona thinks Rani likes’ (in situ) b. o kuRii [mona sochdi hagi [ jeenui rani t i pasand kardi dem girl Mona think is rel-acc Rani like do hai]] is ‘the girl who Mona thinks Rani likes’ (partial movement) c. o kuRii [ jeenui mona sochdi hagi [t i rani t i pasand kardi dem girl rel-acc Mona think is Rani like do hai]] is ‘the girl who Mona thinks Rani likes’ (full fronting) The term partial movement is often used to refer to a special strategy found in question formation, also known as the scope-marking strategy, illustrated in (17). (17) Partial movement in question formation a. mona kii sochidi hagi [CP ki kal kaun Mona what thinks is that tomorrow who jittegaa]? win-fut.3msg ‘Who does Mona think will win tomorrow?’

(Punjabi)

Overt and Covert Wh- and Relative Movement

143

b. Was glaubst du, mit wem Maria gesprochen (German) what think you with whom Maria spoken hat? has ‘Who do you think Maria has spoken to?’ (Van Riemsdijk 1982) c. jaun kyaa soctaa hai [CP meri kis-se (Hindi) John what thinks is Mary who-with baat karegii]? will-talk ‘Who does John think Mary will talk to?’ (Dayal 1994) d. kep itom [CP wen-ili (Passamaquoddy) what say who-Obv nemiy-ac-il t i ]? IC.see-3Conj-PartObv ‘Who did he say he saw?’ (Bruening 2001) The examples in (17) have a wh-pronoun in the embedded clause and an invariant ‘what’ in the matrix clause. The invariant ‘what’ is taken as a scope marker, and the form of the wh-phrase in the embedded clause depends on the syntactic context. Several questions concerning the scope-marking structure have drawn much investigation. The most general one is how the matrix question interpretation is achieved. There are two main competing analyses: the direct dependency approach (e.g., Van Riemsdijk 1983; McDaniel 1989) and the indirect dependency approach (e.g., Dayal 1994). The former suggests that the wh-phrase in the embedded position moves at LF to replace the scope marker, which is analyzed as a purely expletive element. The latter argues that the matrix clause is interpreted as a question over a proposition that is restricted by the embedded wh-question. No matter which analysis we adopt here, it is clear that the nature of the operation applied in (17) is di¤erent from the one applied in (16). Therefore, we can conclude that the strategy used in Punjabi rel-movement is distinct from the strategy used in Punjabi questions. The question left at this point is why we do not ﬁnd an analogue in relativization of the partial wh-movement strategy used for question formation. Such a relative clause would presumably look like (18a) or (18b).

144

(18) a. the girl [what Mona thinks [who Rani likes b. the girl [what Mona thinks [Rani likes who]]

Henrietta Yang

]]

Or, the other way around, why do we not ﬁnd an analogue in Punjabi questions of the partially or fully fronting rel-movement operation? Such a question would presumably look like either (19a) or (19b). (19) a. Who does Mona think [ will win tomorrow]? b. Mona thinks [who will win tomorrow]? We do ﬁnd the structure in (19b) in Punjabi; however, this structure will only give us an embedded question reading. The absence of a structure like (19a) or (19b) for a matrix question reading suggests that whmovement in Punjabi is subject to stricter locality constraints and is clause bound.9 I further suggest that movement of question phrases in Punjabi is the result of an A-scrambling e¤ect, which never crosses any CPs. The absence of a structure like (18) in Punjabi relative clauses is puzzling. A tentative proposal would be that the way semantic interpretations of relative clauses are derived di¤ers from the way those of questions are derived. If the indirect dependency approach is correct, a scope-marking structure is designed to deal with a question over a proposition that is restricted by the embedded wh-question. That is why whphrases do not have wide scope and are constrained inside embedded clauses. Since relative clauses are not designed to deal with propositions, a scope-marking structure is not available for relative constructions.10 6.3.2 Toward an Analysis of ‘‘Partial Movement’’ in Punjabi Relative Clauses Rel-movement in Punjabi is sensitive to island constraints, namely, the Complex NP Constraint, the Wh-Island Constraint, and the Adjunct Island Constraint. (20a) is ungrammatical because it violates the Complex NP Constraint. Note that (20b) is the in-situ variant and it is ungrammatical. This shows that even the derivation of ostensibly ‘‘in-situ’’ variants like (20b) involves movement. Without the NP island, the sentence is grammatical, as expected; see (20c–d). Hence, the derivation of Punjabi relative clauses involves movement. (20) Complex NP Constraint a. *o kuRii [ jeenui mona [eh gal ki [rani t i pasand dem girl rel-acc Mona this thing that Rani like kardi]] jandi hagi do know is ‘*the girl who Mona knows the thing that Rani likes’

Overt and Covert Wh- and Relative Movement

145

b. *o kuRii [mona [eh gal ki [rani jeenu pasand kardi]] dem girl Mona this thing that Rani rel-acc like do jandi hagi know is ‘*the girl who Mona knows the thing that Rani likes’ c. o kuRii [ jeenui mona jaandi hagi [rani t i pasand kardi dem girl rel-acc Mona know is Rani like do hai]] is ‘the girl who Mona knows that Rani likes’ d. o kuRii [mona jaandi hagi [ki rani jeenu pasand dem girl Mona know is that Rani rel-acc like kardi hai]] do is ‘the girl who Mona knows that Rani likes’ (21) and (22) demonstrate that it is not possible to move relative phrases out of an embedded question (the Wh-Island Constraint) or an adjunct clause (the Adjunct Island Constraint), respectively. (21) Wh-Island Constraint a. mein bill-nu puchiya [ki john-ne mary-nu kii ditta] I-erg Bill-acc ask-pfv that John-erg Mary-dat what give-pfv ‘I asked Bill what John gave to Mary.’ b. *o laRkii [ jis-nu mein bill-nu puchiya [ki john-ne that girl rel-dat I-erg Bill-acc ask-pfv that John-erg kii ditta]] what give-pfv (22) Adjunct Island Constraint11 a. john-ne mary-nu enne tofe ditte [ki sally John-erg Mary-dat so many presents give-pfv that Sally gusse ho gayi] angry be go-pfv ‘John gave Mary so many presents that Sally became angry.’ b. *o laRkii [ jo john-ne mary-nu enne tofe ditte that girl rel John-erg Mary-dat so many presents give-pfv [ki gusse ho gayi]] that angry be go-pfv ¯ -chain, we might expect If a movement is cyclic and forms an A the movement to go through the speciﬁer positions (Specs). However,

146

Henrietta Yang

whether all the copies of a movement chain are allowed to be overtly real¯ized would be another issue. We can schematize a successive-cyclic A movement chain as in (23). (23) DP/NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]] Copy3 Copy2 Copy1 ¯ -chain, which (23) shows that the movement of the rel-phrase forms an A 0 moves through Spec of the intermediate C and ﬁnally into the highest Spec,CP for interpretation. One piece of evidence that a rel-phrase moves into a Spec,CP comes from the examples in (24)–(25). In Punjabi, as in Hindi, the Doubly Filled Comp Filter does not seem to be operative. (24) and (25) show that rel-phrases can be overtly realized before complementizer ki, which can be taken as evidence that rel-phrases do move into Spec,CPs. (24) vo laRkaa [ jo ki siggi-ko pasand kartaa hai] dem boy rel that Siggi-acc like do is ‘the boy who likes Siggi’ (25) o kuRii [ jo ki khadi hagi] lumbi hai dem girl rel that standing is tall is ‘The girl who is standing is tall.’

(Hindi)

(Punjabi)

Following Bobaljik’s (2002) copy theory, when the lowest copy of the movement chain is pronounced, the result is the in-situ sentence shown in (26a), which is an example of so-called covert movement in syntax (pattern (b) of table 6.1). When the highest copy is pronounced, the result is a structure like (26c) that displays so-called overt movement in syntax (pattern (a)). It is well documented in the literature that PF and LF can privilege di¤erent copies in a movement chain (see patterns (b) and (c)); therefore, it is not surprising that an intermediate copy of a movement chain can be pronounced. (26b), illustrating ‘‘partial movement’’ in relative clauses, is an example of this sort.12 (26) a. o kuRii [mona soch-di hagi [rani jeenu pasand kardi dem girl Mona think-hab is Rani rel-acc like do hai]] is ‘the girl who Mona thinks Rani likes’ a 0 . [ IP . . .NP [CP Rel-XP i . . . [CP Rel-XP i [ C 0 . . .Rel-XPi . . . ]]]]

Overt and Covert Wh- and Relative Movement

147

b. o kuRii [mona soch-di hagi [ jeenu rani pasand kardi dem girl Mona think-hab is rel-acc Rani like do hai]] is ‘the girl who Mona thinks Rani likes’ b 0 . [ IP . . .NP [CP Rel-XP i . . . [CP Rel-XPi [ C 0 . . .Rel-XP i . . . ]]]] c. o kuRii [ jeenu mona soch-di hagi [rani pasand kardi dem girl rel-acc Mona think-hab is Rani like do hai]] is ‘the girl who Mona thinks Rani likes’ c 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XP i [ C 0 . . .Rel-XP i . . . ]]]] Under this view, there is no overt/covert distinction; it is PF that chooses which copy is to be pronounced. If we follow Chomsky’s (1993, 1995) view in which all overt movement precedes all covert movement or his (2001) proposal in which derivation is restricted by phases, we are able to explain all the possible PF outputs in (27), but this requires assuming in addition three distinct syntactic derivations, as in (27a 0 ,b 0 ,c 0 ). (27a 0 ) involves two covert movement operations; (27b 0 ), overt movement followed by covert movement; and (27c 0 ), two overt movement operations. (27) a. o kuRii [mona soch-di hagi [rani jeenu pasand kardi dem girl Mona think-hab is Rani rel-acc like do hai]] is ‘the girl who Mona thinks Rani likes’ a 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]]] " " covert covert b. o kuRii [mona soch-di hagi [ jeenu rani pasand kardi dem girl Mona think-hab is rel-acc Rani like do hai]] is ‘the girl who Mona thinks Rani likes’ b 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]]] " " covert overt

148

Henrietta Yang

c. o kuRii [ jeenu mona soch-di hagi [rani pasand kardi dem girl rel-acc Mona think-hab is Rani like do hai]] is ‘the girl who Mona thinks Rani likes’ c 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]]] " " overt overt However, Chomsky’s proposal runs into a problem in explaining the facts in (28).13 (28a/a 0 ) show the lowest copy and the highest one being pronounced, and (28b/b 0 ) the intermediate one and the highest one. (28) a. o kuRiii [ jeenui mona soch-di hagi [rani jeenui dem girl rel-acc Mona think-hab is Rani rel-acc pasand kardi hai]] like do is ‘the girl who Mona thinks Rani likes’ a 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XP i [ C 0 . . .Rel-XPi . . . ]]]] b. o kuRiii [ jeenui mona soch-di hagi [ jeenui rani dem girl rel-acc Mona think-hab is rel-acc Rani pasand kardi hai]] like do is ‘the girl who Mona thinks Rani likes’ b 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XP i . . . ]]]] This multiple realization phenomenon is not unique to Punjabi. There are crosslinguistic parallels in the literature such as participle agreement in Passamaquoddy (Algonquian).14 (29a–b) are examples of externally headed relative clauses, and (30) is an example of internally headed relative clauses (examples are from Bruening 2004). (29) a. Wo`t nı`t pahtolı´yasi [CP Ma´li this.prox that.in priest Mary ´ el-itahası-c-il [CP eli we`n IC.thus-think-3an-part:obv:sg thus someone kisi¼komutonom-a´-c-il t i ]] past¼steal.from-dir-3an-part:obv:sg ‘This is the priest that Mary thinks someone has robbed.’

Overt and Covert Wh- and Relative Movement

149

b. N-muhsa´l-a [NP wo`t skita`pi [CP 1-like-dir this.prox man Pı´yel ´ıt-o-k-il [CP el-itahası´-c-il Piyel say-th-3an-part:obv:sg IC.thus-think-3an-part:obv:sg [CP t i kisi¼komutonom-ı´hc-il ]]]] past¼steal.from-obv/prox-part:obv:sg ‘I like the man that Piyel said he thought robbed him.’ (30) No`t nı`t [el-itahası´-c-il that.prox that.in IC.thus-think-3an-part:obv:sg [kisi¼pson-a´-c-il otu´hk-ol ]] past¼catch-dir-3an-part:obv:sg deer-obv:sg ‘This is the deer that he thought he was going to get.’ The multiple realization phenomenon found in Passamaquoddy is not identical to the one in Punjabi. However, the goal of using this strategy, multiple realization, in both languages is the same: to mark the path of a movement chain. In Passamaquoddy, marking every participle form of the verb serves to clearly identify the path of a moving operator. As (29)–(30) show, more than one participle form can be marked by a single relative phrase. Further, even when the head is internally realized as in (30), it triggers marking in the higher clause. The same multiple Spell-Out phenomenon is found in German; see (31a–b). In addition to the scope-marking strategy, (31) seems to be the default way of forming long-distance questions in the Brandenburg dialect in German. (31) a. Wer denken Sie, wer Sie sind? who think you who you are ‘Who do you think you are?’ b. Wen denkst Du, wen sie meint, wen Harald liebt? who think you who she believes who Harald loves ‘Who do you think that she believes that Harald loves?’ (Fanselow and Mahajan 2000, 219) In this section, I have presented crosslinguistic facts from Punjabi, Passamaquoddy, and German to show that the multiple realization of a certain element of a movement chain is possible. In the following section, I show that multiple realization of a movement chain is subject to a PF economy restriction.

150

Henrietta Yang

6.3.3 Constraint on Multiple Realization So far, we have seen ﬁve options for realizing elements of a movement chain in Punjabi rel-movement. If the lowest copy is pronounced, (32a) results; if the intermediate one, (32b); and if the highest one, (32c). If the lowest copy and the highest copy are pronounced, (32d) results; if the highest copy and the intermediate copy are pronounced, (32e). (32) Five available options for realizing elements of a movement chain in Punjabi rel-movement a. o kuRiii [mona sochdi hagi [rani jeenui pasand kardi dem girl Mona think is Rani rel-acc like do hai]] is ‘the girl who Mona thinks Rani likes’ 0 a . [ IP . . .NP [CP Rel-XP i . . . [CP Rel-XP i [ C 0 . . .Rel-XPi . . . ]]]] b. o kuRiii [mona sochdi hagi [ jeenui rani pasand kardi dem girl Mona think is rel-acc Rani like do hai]] is b 0 . [ IP . . .NP [CP Rel-XP i . . . [CP Rel-XPi [ C 0 . . .Rel-XP i . . . ]]]] c. o kuRiii [ jeenui mona sochdi hagi [rani pasand kardi dem girl rel-acc Mona think is Rani like do hai]] is c 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XP i [ C 0 . . .Rel-XP i . . . ]]]] d. o kuRiii [ jeenui mona sochdi hagi [rani jeenui pasand dem girl rel-acc Mona think is Rani rel-acc like kardi hai]] do is d 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XP i [ C 0 . . .Rel-XPi . . . ]]]] e. o kuRiii [ jeenui mona sochdi hagi [ jeenui rani pasand dem girl rel-acc Mona think is rel-acc Rani like kardi hai]] do is e 0 . [ IP . . .NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XP i . . . ]]]] However, some options for multiple realization are unavailable. (33) Unavailable options a. *o kuRii [CP jeenui mona sochdi hagi [CP jeenui [ C 0 rani dem girl rel-acc Mona think is rel-acc Rani jeenui pasand kardi hai]]] rel-acc like do is

Overt and Covert Wh- and Relative Movement

151

a 0 . *[ IP . . .NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]]] b. *o kuRii [CP mona sochdi hagi [CP jeenui [ C 0 rani jeenui dem girl Mona think is rel-acc Rani rel-acc pasand kardi hai]]] like do is b 0 . *[ IP . . .NP [CP Rel-XP i . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]]] c. *o kuRii [CP mona sochdi hagi [CP [ C 0 rani pasand kardi dem girl Mona think is Rani like do hai]]] is 0 c . *[ IP . . .NP [CP Rel-XP i . . . [CP Rel-XP i [ C 0 . . .Rel-XP i . . . ]]]] Examples (33a–b) are not legitimate options. They show that multiplecopy pronunciation within a given CP domain is not permitted. In addition, without any overtly realized rel-phrase in rel-movement, the sentence is ruled out as in (33c/c 0 ). The ungrammaticality of (33c) suggests that a null operator is not available in Punjabi rel-movement; if it were, we would expect to ﬁnd a structure like (34). (34) a. *o kuRiii [CP Opi mona sochdi hagi [CP Opi [ C 0 rani Opi dem girl Mona think is Rani pasand kardi hai]]] like do is b. *[ IP . . .NP [CP Op i . . . [CP Op i [ C 0 . . .Op i . . . ]]]] In addition, in the following section, I show that a null operator is also not available in Punjabi wh-movement and that recoverability of deletion requires at least one copy of a movement chain to be overtly realized. Note that the copy theory used here for deriving (32a–e) also generates the ungrammatical sentences in (33a–b). Now the question is how we can rule out these ungrammatical constructions. I propose that multiple-copy pronunciation in rel-movement in Punjabi is constrained by a PF economy condition, the CP-Phase Constraint.15 (35) CP-Phase Constraint In a given CP phase, at most one copy of a movement chain can be overtly realized (or pronounced). In other words, two copies of the same element occurring in the same CP phase cannot be pronounced at the same time. According to Chomsky (2001), CP and vP are two phases in the derivation process. However, I have demonstrated above that rel-movement in Punjabi is an instance of

152

Henrietta Yang

successive-cyclic movement through Spec,CPs. Only CP phases seem to matter for the realization of chains. (35) is an economy constraint from the PF point of view; pronouncing one copy of a movement chain in a given CP phase is more economical than pronouncing two. Given this constraint, it is also true that pronouncing zero copies is even more economical than pronouncing one. However, that possibility is constrained by recoverability of deletion. If a movement chain does not cross a ﬁnite clause, pronouncing zero copies will violate recoverability of deletion. If it does cross a ﬁnite clause, then pronouncing zero copies in a given CP domain is a possible PF output as long as a copy of the movement chain is pronounced in another CP phase as in (32a–c). According to the CP-Phase Constraint, we can predict that in a movement chain schematized as in (23), repeated here as (36), pronouncing both Copy1 and Copy2 will violate the constraint since both copies appear in the same CP domain. (36) DP/NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]] Copy3 Copy2 Copy1 The prediction is borne out, as shown in (37). Examples (37a) and (37b) are (33b) and (33a), respectively, in which two copies of a movement chain are pronounced in the embedded CP domain, violating the PF economy constraint (35) (the CP-Phase Constraint). (37) a. *DP/NP [CP Rel-XP i . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]] Copy3 Copy2 Copy1 0 a . *o kuRiii [CP mona sochdi hagi [CP jeenui [ C 0 rani jeenui dem girl Mona think is Copy2 Rani Copy1 pasand kardi hai]]] like do is b. *DP/NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]] Copy3 Copy2 Copy1 b 0 . *o kuRiii [CP jeenui mona sochdi hagi [CP jeenui [ C 0 rani dem girl Copy3 Mona think is Copy2 Rani jeenui pasand kardi hai]]] Copy1 like do is c. *DP/NP [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]] Copy2 Copy1 0 c . *o kuRiii [CP jeenui [ C 0 rani jeenui pasand kardi hai]] dem girl Copy2 Rani Copy1 like do is

Overt and Covert Wh- and Relative Movement

153

The CP-Phase Constraint can further predict the ungrammaticality of (37c/c 0 ) in Punjabi, as well as (38) in German and (39) in English. (38) *Wen liebt er wen? who loves he who ‘Who does he love?’ (Fanselow and Mahajan 2000, 219) (39) *Who does he love who? In this section, I have shown that multiple realization of a relmovement chain is subject to both PF and locality constraints. In the following section, I focus on wh-movement in Punjabi. 6.3.4 Movement Operations in Punjabi Questions In section 6.2, I showed that wh-phrases in Punjabi questions do not undergo long-distance movement (i.e., movement crossing a CP clause). Examples are repeated in (40) (¼ (12)). The only way to achieve a matrix question reading is through the scope-marking strategy, as shown in (40c). Unlike wh-phrases in Hindi, wh-phrases in Punjabi are not allowed to undergo long-distance movement, as shown in (40b) (cf. (8b)). (40) a.

mona sochidi hagi [CP ki kal kaun jittegaa] Mona thinks is that tomorrow who win-fut.3msg ‘Mona wonders who will win tomorrow.’ not ‘Who does Mona think will win tomorrow?’ b. *kauni mona sochidi hagi [CP kal t i jittegaa] who Mona thinks is tomorrow win-fut.3msg c. mona kii sochidi hagi [CP kal kaun jittegaa]? Mona what thinks is tomorrow who win-fut.3msg ‘Who does Mona think will win tomorrow?’ (what . . . wh-XP)

(40a) is a grammatical sentence only under an embedded question reading. Now compare (40a) with (41), which is an ‘‘in-situ’’ version of rel-movement. (41) o kuRiii [ jeenu i mona sochdi hagi [ jeenu i rani jeenui dem girl rel-acc Mona think is rel-acc Rani rel-acc pasand kardi hai]] like do is ‘the girl who Mona thinks Rani likes’

154

Henrietta Yang

Even though the rel-phrase in (41) is overtly realized in the embedded clause, it does not lead to a deviant reading. The correct semantic interpretation can still be achieved by the highest unpronounced copy. On the other hand, in questions, the scope of a matrix question cannot be derived when the wh-phrase is overtly realized in an embedded clause. This fact suggests that the wh-phrase in (40a) does not move into the matrix clause; therefore, there is no unpronounced copy left in the matrix clause. Hence, a matrix question reading is not available. The ungrammaticality of (40b) further suggests that movement of a whphrase in Punjabi is clause bound and that a null wh-operator does not exist; otherwise, we would expect a structure like (42) to be available in Punjabi. (42) *mona Opi sochidi hagi [CP kal kauni jittegaa]? Mona think is tomorrow who win-fut.3msg ‘*Who does Mona think will win tomorrow?’ (Op i . . . wh-XPi ) What (40)–(42) show is that the scope of a question has to be marked overtly in Punjabi, as shown in (40c); however, this requirement does not hold in a relative clause. The facts presented here argue against the direct dependency approach as an explanation for the scope-marking strategy in Punjabi questions, in which the wh-phrase in the embedded clause moves at LF to replace the scope marker in the matrix clause. This leaves the other main analysis, the indirect dependency approach, which seems to be a more appropriate analysis for the facts presented here.16 Following Mahajan’s (2000) observation about Hindi questions, I ar¯ -movement through Spec,CPs is not available gue that a successive-cyclic A in Punjabi questions. The wh-movement examples in (43) and the relmovement examples in (44) demonstrate that the wh-phrase is not permitted to move into Spec,CP (see (43c)), while the rel-phrase is (see (44b)). (43) a.

tera kii khaayal hai [CP ki kal kaun jittegaa]? your what idea is that tomorrow who win-fut.3msg ‘Who do you think will win tomorrow?’ b. tera kii khaayal hai [CP ki kaun kal jittegaa]? your what idea is that who tomorrow win-fut.3msg ‘Who do you think will win tomorrow?’ c. *tera kii khaayal hagi [CP kaun ki kal jittegaa]? your what idea is who that tomorrow win-fut.3msg d. *tera kaun khaayal hagi [CP ki kal jittegaa]? your who idea is that tomorrow win-fut.3msg

Overt and Covert Wh- and Relative Movement

155

(44) a. o kataab [ jiit-ne akhiaa [ki kuldiip-ne jo paRii]] dem book Jiit-erg say-past that Kuldiip-erg rel read-past ethe naii hai here neg is ‘The book that Jiit said that Kuldiip read is not here.’ b. o kataab [ jiit-ne akhiaa [ jo ki kuldiip-ne paRii]] dem book Jiit-erg say-past rel that Kuldiip-erg read-past ethe naii hai here neg is ‘The book that Jiit said that Kuldiip read is not here.’ c. o kataab [ jo jiit-ne akhiaa [ki kuldiip-ne paRii]] dem book rel Jiit-erg say-past that Kuldiip-erg read-past ethe nahi hai here neg is ‘The book that Jiit said that Kuldiip read is not here.’ The facts shown here support the argument that (1) rel-movement and wh-movement in Punjabi utilize distinct movement operations and (2) assuming the same operation for these two seemingly related constructions is not appropriate. 6.4

Movement Operations in Hindi Relative Clauses and Questions

In this section, I focus on rel-movement and wh-movement in Hindi. I discuss the asymmetry between local and long-distance movement concerning the extent of ﬂexibility of pronunciation. I further present di¤erences between rel-movement and wh-movement in Hindi and show that these two constructions in Hindi, as in Punjabi, do not make use of the same movement operation. However, we can achieve a uniﬁed analysis for these two constructions if we take into account additional assumptions about the realization of chains such as PF Relaxation and Minimize Mismatch. 6.4.1 PF Relaxation: An Asymmetry between Local and Long-Distance Movement As shown in section 6.2, in Hindi, as in Punjabi, either the lower or the higher copy of a rel-phrase can be pronounced when rel-movement does not involve crossing a CP domain. The examples given in (5) are repeated here in (45). When the lower copy of a movement chain is pronounced, (45a) is derived. On the other hand, if PF chooses to pronounce the higher copy, the result is (45b).

156

Henrietta Yang

(45) a. vo laRkii [CP rohit jis-ko pasand dem girl Rohit rel-acc like ‘The girl who Rohit likes is smart.’ b. vo laRkii [CP jis-ko rohit pasand dem girl rel-acc Rohit like ‘The girl who Rohit likes is smart.’

kartaa hai] hoshiyaar hai do is smart is kartaa hai] hoshiyaar hai do is smart is

However, if a rel-phrase crosses a ﬁnite CP, then the highest copy of the movement chain must be the only copy that is pronounced. (46) a.

vo laRkii [CP jis-ko mona sochtii hai [CP ki rani dem girl rel-acc Mona think is that Rani kartii hai]] do is ‘the girl who Mona thinks that Rani likes’ b. *vo laRkii [CP mona sochtii hai [CP ki rani jis-ko dem girl Mona think is that Rani rel-acc kartii hai]] do is c. *vo laRkii [CP mona sochtii hai [CP jis-ko ki rani dem girl Mona think is rel-acc that Rani kartii hai]] do is

pasand like

pasand like

pasand like

The grammaticality of (46a) indicates that pronunciation of the highest copy is required. (46b) shows that the option of pronouncing the lower copy is not available in embedded ﬁnite clauses; neither is the option of ‘‘partial movement,’’ as shown in (46c). Note that (46b–c) are legitimate in Punjabi. What we have observed here is that pronunciation of a relphrase inside a ﬁnite clause—local movement—is free; either the higher or the lower copy can be pronounced. I will call this kind of PF ﬂexibility PF Relaxation. However, PF Relaxation is not operative when relmovement involves crossing a CP domain, in that only the highest copy is pronounced. As for the requirement to pronounce the highest copy in a longdistance movement chain, I adopt both Bobaljik’s (2002) and Diesing’s (1997) proposals that it is the result of an economy condition that prefers isomorphism between PF and LF. This condition is called Minimize Mismatch in Bobaljik’s terms, and the Scope Principle in Diesing’s. (47) Minimize Mismatch (To the extent possible) privilege the same copy at PF and LF. (Bobaljik 2002, 251)

Overt and Covert Wh- and Relative Movement

157

Under this view, it is more economical when PF and LF choose to pronounce and interpret the same copy, as we have seen with Hindi long-distance rel-movement. An asymmetry between local and long-distance movement is not unusual. Richards (1997) provides a detailed discussion of asymmetries between local and long-distance wh-movement; for example, he shows that in Serbo-Croatian, superiority e¤ects occur only with long-distance movement and not with local movement. (48) a. Ko je koga vidjeo? who aux whom seen ‘Who saw whom?’ b. Koga je ko vidjeo? whom aux who seen ‘Who saw whom?’ (Richards 1997, 32) (49) a.

Ko si koga tvrdio da je who aux whom claimed that aux ‘Who did you claim beat whom?’ b. *Koga si ko tvrdio da je whom aux who claimed that aux (Richards 1997, 32)

istukao? beaten istukao? beaten

A similar asymmetry between local and long-distance movement is found in Japanese,17 where superiority e¤ects constrain only the latter. An asymmetry is also found in weak crossover (Richards 1997), where longdistance movement appears to be subject to stricter constraints than local movement. In Serbo-Croatian and Japanese, it is superiority e¤ects that create the asymmetry between local and long-distance movement, and in Hindi, it is the phonological condition PF Relaxation. 6.4.2 Questions in Hindi Question formation in Hindi has drawn considerable attention in the literature, mainly because of work by Dayal (1994, 1996, 2000) and Mahajan (1990, 2000). Local wh-movement in Hindi seems to behave like local rel-movement in having the option of pronouncing either the lower copy or the higher copy; the phonological condition PF Relaxation is also operative. The examples in (50) are repeated from (7). (50) a. siggi kis-ko pasand kartaa hai? Siggi who-acc like do is ‘Who does Siggi like?’

158

Henrietta Yang

b. kis-koi siggi t i pasand kartaa hai? who-acc Siggi like do is ‘Who does Siggi like?’ Furthermore, if question formation involves crossing a ﬁnite CP, like long-distance rel-movement in Hindi, then the highest copy must be pronounced in order to derive a matrix question reading. In (51a), the only possible reading is the embedded question reading. In order to derive a matrix question reading, the wh-phrase in (51b) must be moved into the matrix clause. As demonstrated in (51c), there is an alternative way to achieve the matrix question reading—namely, to use the scope-marking strategy. (51) a. jaun soctaa hai [CP ki meri kis-se baat karegii] John thinks is that Mary who-with will-talk ‘John wonders who Mary will talk to.’ not ‘Who does John think Mary will talk to?’ b. kis-se jaun soctaa hai [CP ki meri baat karegii]? who-with John thinks is that Mary will-talk ‘Who does John think Mary will talk to?’ c. jaun kyaa soctaa hai [CP ki meri kis-se baat karegii]? John what think is that Mary who-with will-talk ‘Who does John think Mary will talk to?’ Following Mahajan’s (2000) proposal, I assume that (51b) and (51c) involve two distinct constructions, which coexist in languages like Hindi and German. Mahajan suggests that (51b) involves long-distance scrambling, which is distinct from successive-cyclic wh-movement.18 As shown ¯ -movement moves through Spec,CPs to reach above, successive-cyclic A its scope position; therefore, it leaves copies behind. However, longdistance scrambling does not work the same way; it does not proceed through Spec,CPs and does not leave copies in those positions. A question arises at this point: even if we assume that long-distance scrambling does not leave any copies along the way, why do we not have the choice of pronouncing the original copy as in (52)? (52) kis-se jaun soctaa hai [CP ki meri kis-se baat karegii] who-with John thinks is that Mary who-with will-talk ‘John wonders who Mary will talk to.’ not ‘Who does John think Mary will talk to?’ The answer to this question is that, like rel-movement in Hindi, longdistance wh-movement in Hindi is constrained by Minimize Mismatch,

Overt and Covert Wh- and Relative Movement

159

Table 6.2 The distribution of PF Relaxation and Minimize Mismatch Hindi

Punjabi

Local

Longdistance

Local

Longdistance

Movement

Rel-

Wh-

Rel-

Wh-

Rel-

Wh-

Rel-

Wh-

PF Relaxation Minimize Mismatch

B x

B x

x B

x B

B x

B x

B x

N/A N/A

according to which PF and LF prefer to pronounce the same copy. As a result of this economy condition, PF and LF will choose the higher copy inside the matrix clause, yielding (51b). The grammaticality of (51b) suggests that wh-phrases in Hindi, unlike those in Punjabi (see (40b)), are not clause bound. A question arises at this point: why is long-distance relmovement in Punjabi not constrained by Minimize Mismatch? I suggest that, in Punjabi, PF Relaxation constrains both local and long-distance rel-movement; therefore, the economy condition Minimize Mismatch does not come into play. In sum, PF Relaxation and Minimize Mismatch operate in complementary distribution, as illustrated in table 6.2.19 This complementary distribution suggests that it may be possible to unify PF Relaxation and Minimize Mismatch or at least derive them from the same initial premises. I leave this uniﬁcation for further work. 6.5

Conclusion

In this chapter, I have shown that wh-movement and rel-movement do not necessarily make use of the same movement operation. Rel-movement in ¯ -movement, while both Hindi and Punjabi undergoes successive-cyclic A wh-movement in both languages is the result of a scrambling e¤ect. A uniﬁed analysis for these two movement operations is feasible but requires additional assumptions about the realization of chains, namely, PF Relaxation and Minimize Mismatch. I have suggested that the di¤erence found in long-distance rel-movement in the two languages can be taken as the consequence of the interaction between PF Relaxation and Minimize Mismatch, which operate in complementary distribution. These two phonological conditions also account for the asymmetry between local and long-distance movement operations in Hindi. I have provided empirical evidence demonstrating that the PF and LF interfaces can constrain the overt realization of question and relative

160

Henrietta Yang

phrases in overt syntax. Some facts from the multiple realization of relphrases in Punjabi support the analysis of Bobaljik’s (2002) copy theory in which PF can choose to pronounce a certain copy or copies. However, the ﬂexibility of pronunciation permitted by PF in Punjabi is constrained by a PF economy condition, the CP-Phase Constraint. This constraint can also account for the fact that all the copies in a nontrivial chain are not phonetically realized in a language that permits multiple Spell-Out of a certain element. Finally, the complementary distribution between PF Relaxation and Minimize Mismatch raises the question whether the two constraints are derived from the same initial premises. I leave this question for future research. Notes Unless otherwise noted, all Hindi data are from Rajesh Bhatt, and Punjabi data are from Amit Garg and Madhu Garg. I am very grateful to all of them; they are in no way responsible for any misinterpretation on my part of the data they provided. All remaining errors, of course, are my own. 1. Later, I will show that the in-situ examples presented in (1) and (2) indeed undergo an ‘‘overt’’ movement operation. 2. Punjabi, a modern Indo-Aryan language, is mainly spoken in the Punjab states of both India and Pakistan. Traditionally, there are at least eight recognized dialects of Punjabi, which vary from one another: Majhi, Bhattiani, Rathi, Ludhianwi, Doabi, Patialwi, Powadhi, and Malwi. The informants I consulted are from Ludhiana in Punjab, India, and consider their dialect to be close to Powadhi. 3. I will discuss the mechanism of movement later in the text: speciﬁcally, whether ¯ -movement operation or an A-movement operation. The terms moveit is an A ment and fronting used in section 6.2 refer only to an element that is not realized in its base position. More discussion is provided in sections 6.3 and 6.4. 4. See section 6.3.1 for more discussion about the scope-marking strategy. 5. Bhatia (1993) also reports cases similar to (10a) and (10c) in the dialect he studies. The following examples (and glosses) are from Bhatia 1993, 61: (i) o kuRii [ jiit-ne aakhiaa [ki jo j paRaandii ai aaii ai]] dem girl Jiit-erg say-past.ms that rel j teach-pres.fs aux come is ‘The girl that Jiit said teaches has come.’ [ki kuldiip ne (ii) o kataabi [ jo ji [ jiit-ne aakhiaa dem book rel j Jiit-erg say-past.ms that Kuldiip erg ethe naii ai i paRii]]] read-past.fs here neg is ‘The book that Jiit said that Kuldiip read is not here.’ 6. Both of my Punjabi informants also speak Hindi natively. In their Hindi, they do not allow long-distance extraction of a wh-phrase out of an embedded clause. (8b) is ungrammatical to them.

Overt and Covert Wh- and Relative Movement

161

7. Following Chomsky’s (1993) proposal, I assume that traces are copies of the moved element. A nontrivial chain consists of sequences of copies of a moved element. 8. One might question how we know that movement takes place in pattern (d) of table 6.1 if PF and LF both choose the lowest copy. See Bobaljik 2002 for detailed discussion of this pattern. 9. A similar phenomenon is reported by Dayal (2000); she argues that ﬁnite clauses in Hindi form scope islands for wh-in-situ. 10. I thank Rajesh Bhatt (pers. comm.) for bringing this tentative proposal to my attention. 11. In Punjabi, as in Hindi, adjunct clauses are often realized in nonﬁnite clauses from which extraction is possible. Therefore, I choose a ﬁnite result clause to demonstrate the island constraint. 12. Overstriking indicates unpronounced copies. 13. A reviewer points out that if we assume Chomsky’s (2001) Phase Impenetrability Condition (PIC), we can also account for the data in (28). It is true that Chomsky’s proposal can account for (28a); however, it cannot account for (28b). Spell-Out is subject to the PIC, deﬁned as in (ii) (Chomsky 2001, 13). (i) [ ZP Z . . . [ HP a [H YP]]] (HP and ZP are phases) (ii) The domain of H is not accessible to operations at ZP; only H and its edge are accessible to such operations. To account for (28a) under Chomsky’s (2001) proposal, the reviewer suggests the following. First, the rel-phrase in the embedded CP moves to Spec,CP as in (iii). (iii) [CP Rel-XP [ . . . Rel-XP . . . ]] Since the embedded CP does not undergo Spell-Out until the next phase, we do not have to decide at this point which copy is pronounced. After the matrix CP phase is constructed, Rel-XP moves from the Spec of the embedded CP to the Spec of the matrix CP as in (iv). (iv) [CP . . .Rel-XP. . . [CP Rel-XP [ . . .Rel-XP. . . ]]] Copy3 Copy2 Copy1 At this point, the complement of the embedded C undergoes Spell-Out. Since the ﬁrst part is covert movement, Rel-XP is pronounced. Then, at the next phase, the complement of the matrix C undergoes Spell-Out. Since this is an overt movement, Rel-XP in the Spec of the embedded C is unpronounced. (28a) now can also be derived under Chomsky’s analysis. The reviewer’s proposal can plausibly be extended to handle the other cases, (28b–c). It should be noted, however, that the reviewer takes the overt/covert distinction as a primitive, while Chomsky (2001) does not do so. In Chomsky’s system, it is the timing of pronunciation that results in the di¤erences between overt and covert movement. ‘‘Covert movement’’ takes place when a wh-phrase moves to the edge of its phase after the phase undergoes Spell-Out, and ‘‘overt movement’’ takes place when a wh-phrase moves to the edge of its phase before its phase undergoes Spell-Out.

162

Henrietta Yang

14. What Bruening calls participle agreement is an instance in which an operator moves successive-cyclically and marks every participle form of the verb along the path of extraction. 15. From a syntactic point of view, we need to delete one copy of a movement chain if there are two possible copies in a given CP domain. If deleting a copy counts as an operation, then it is more economical not to apply any copy-deletion operation at all. Therefore, a reviewer suggests that the CP-Phase Constraint should count as an absolute constraint rather than an economy condition. The reviewer further suggests that if we apply Kayne’s (1994) Linear Correspondence Axiom (LCA) and assume that it applies at every CP phase, then we can dispense with the CP-Phase Constraint since copies are subject to the LCA. The reviewer’s proposal, applying the LCA at every CP phase, turns out to be too strong and rules out some optimal outputs in Punjabi. If we take rel-phrases as nondistinct copies in a very general sense (without considering formal features), overtly realizing two nondistinct copies violates the LCA. Hence, we will derive either (i) or (ii) but not (iii). This accounts well for the ungrammaticality in (37). (i) DP/NP [CP Rel-XP i [ C 0 . . .Rel-XPi . . . ]] Copy2 Copy1 (ii) DP/NP [CP Rel-XPi [ C 0 . . .Rel-XP i . . . ]] Copy2 Copy1 (iii) *DP/NP [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]] Copy1 Copy2 Now, let’s consider long-distance movement as in (iv). (iv) DP/NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XPi . . . ]]] Copy3 Copy2 Copy1 Again, we can derive the optimal outputs as in (32d–e), repeated in schematic form in (v)–(vi). (v) DP/NP [CP Rel-XPi . . . [CP Rel-XP i [ C 0 . . .Rel-XPi . . . ]]] Copy3 Copy2 Copy1 (vi) DP/NP [CP Rel-XPi . . . [CP Rel-XPi [ C 0 . . .Rel-XP i . . . ]]] Copy3 Copy2 Copy1 Under the LCA, deletion of a copy is triggered by considerations of linearization. However, (vii) (cf. (32c)) shows that deleting Copy1 and Copy2 is not triggered by such considerations. (vii) DP/NP [CP Rel-XPi . . . [CP Rel-XP i [ C 0 . . .Rel-XP i . . . ]]] Copy3 Copy2 Copy1 From (i)–(vii), we can conclude that while certain deletions can be forced by the LCA, others still seem to be optional. The dilemma we encounter here is that while the LCA seems promising in explaining the CP-Phase Constraint, additional optional deletion is still required. A syntactic approach would need to allow for free deletions (up to recoverability) as well as obligatory deletions (forced by the LCA).

Overt and Covert Wh- and Relative Movement

163

16. I would like to thank Junko Shimoyama for calling my attention to this interesting potential correlation. 17. For Japanese examples, see Richards 1997, 33. 18. See Mahajan 1990 for an exposition of long-distance scrambling, 19. As a reviewer points out to me, the facts presented in this chapter seem to suggest that PF Relaxation is parameterized. Such a parameter should not be allowed under a widely accepted theory of parameters, which claims that parametric variation should be restricted to the lexicon (see Borer 1984; Chomsky 1995). References Barss, Andrew. 1986. Chains and anaphoric dependence: On reconstruction and its implications. Doctoral dissertation, MIT. Bhatia, Tej K. 1993. Punjabi: A cognitive-descriptive grammar. London: Routledge. Bobaljik, Jonathan David. 1995. Morphosyntax: The syntax of verbal inﬂection. Doctoral dissertation, MIT. Bobaljik, Jonathan David. 2002. A-Chains at the PF-interface: Copies and ‘‘covert’’ movement. Natural Language and Linguistic Theory 20, 197–267. Borer, Hagit. 1984. Parametric syntax. Dordrecht: Foris. Bosˇkovic´, Zˇeljko. 2001. On the nature of the syntax-phonology interface: Cliticization and related phenomena. Amsterdam: Elsevier. Brody, Michael. 1995. Lexico-Logical Form: A radically minimalist theory. Cambridge, Mass.: MIT Press. Bruening, Benjamin. 2001. Syntax at the edge: Cross-clausal phenomena and the syntax of Passamaquoddy. Doctoral dissertation, MIT. Bruening, Benjamin. 2004. Two types of wh-scope marking in Passamaquoddy. Natural Language and Linguistic Theory 22, 229–305. Cheng, Lisa L.-S. 1991. On the typology of wh-questions. Doctoral dissertation, MIT. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Dayal, Veneeta. 1994. Scope marking as indirect wh-dependency. Natural Language Semantics 2, 137–170.

164

Henrietta Yang

Dayal, Veneeta. 1996. Locality in WH quantiﬁcation. Dordrecht: Kluwer. Dayal, Veneeta. 2000. Scope marking: Cross-linguistic variation in indirect dependency. In Uli Lutz, Gereon Mu¨ller, and Arnim von Stechow, eds., Wh-scope marking, 157–194. Amsterdam: John Benjamins. Diesing, Molly. 1997. Yiddish VP order and the typology of object shift in Germanic. Natural Language and Linguistic Theory 15, 369–427. Fanselow, Gisbert, and Anoop Mahajan. 2000. Towards a minimalist theory of wh-expletives, wh-copying, and successive cyclicity. In Uli Lutz, Gereon Mu¨ller, and Arnim von Stechow, eds., Wh-scope marking, 195–230. Amsterdam: John Benjamins. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT. Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press. Mahajan, Anoop. 1990. The A/A-bar distinction and movement theory. Doctoral dissertation, MIT. Mahajan, Anoop. 2000. Towards a uniﬁed treatment of wh-expletives in Hindi and German. In Uli Lutz, Gereon Mu¨ller, and Arnim von Stechow, eds., Whscope marking, 317–332. Amsterdam: John Benjamins. McDaniel, Dana. 1989. Partial and multiple wh-movement. Natural Language and Linguistic Theory 7, 565–604. Pesetsky, David. 1998. Some optimality principles of sentence pronunciation. In Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky, eds., Is the best good engough?, 337–383. Cambridge, Mass.: MIT Press and MITWPL. Richards, Norvin W. 1997. What moves where when in which language? Doctoral dissertation, MIT. Riemsdijk, Henk van. 1983. Correspondence e¤ects and the Empty Category Principle. Tilburg Papers in Language and Literature 12. Tilburg: University of Tilburg. Rudin, Catherine. 1988. On multiple questions and multiple wh-fronting. Natural Language and Linguistic Theory 6, 445–501.

Chapter 7 Movement, Wh-Agreement, and Apparent Wh-in-Situ

7.1

Chris H. Reintges, Philip LeSourd, and Sandra Chung

Introduction

In current minimalist syntax (Chomsky 2000, 2001, 2004), movement is an operation that does not introduce a trace; rather, it leaves behind a complete copy of the moved element, with the result that structures formed by movement exhibit multiple copies of the moved element. In the phonology, it is determined which of these copies is phonetically realized, the typical case being that only the highest copy is pronounced (see Nissenbaum 2000). (1)

This copy theory of movement allows for an option not available in trace theory, namely, that language-speciﬁc phonological constraints might dictate that the copy privileged for pronunciation is the lowest, as opposed to the highest. This possibility has been productively explored by Brody (1995), Groat and O’Neil (1996), Pesetsky (1997, 1998, 2000), Bosˇkovic´ (2001), Bobaljik (2002), and others (but see Nunes 2004 for a di¤erent view). Following this line of thought, we argue here that certain types of wh-in-situ instantiate ‘‘hidden’’ movement conﬁgurations that are created by wh-movement in overt syntax, followed by pronunciation of only the lowest copy. (2)

166

Chris H. Reintges, Philip LeSourd, and Sandra Chung

Our evidence for these hidden movement conﬁgurations is supplied by the phenomenon of wh-agreement, which is widely believed to track whmovement in overt syntax, before Spell-Out. In brief, we show that there are languages in which the morphological patterns of wh-agreement occur in certain wh-in-situ contexts. Continuing to assume that this special morphology provides a record of the history of wh-movement before the derivation is shipped to PF, its occurrence in certain wh-in-situ constructions reveals that here too, movement has occurred in overt syntax. The only di¤erence from overt wh-movement is that the lower rather than the higher copy is pronounced. Our proposal thus invokes the copy-plusdelete theory of movement to integrate one strand of research on wh-insitu with research on the morphology of extraction. Section 7.2 o¤ers some background on the morphological proﬁle of wh-agreement and its interaction with overt wh-movement, using Chamorro as an illustration. Section 7.3 makes explicit our claim that whagreement is a morphological reﬂex of the Agree relation that must hold before wh-movement can occur. Sections 7.4 and 7.5 o¤er two case studies of languages in which wh-agreement is visible in wh-in-situ contexts: in internally headed relative clauses in Passamaquoddy, an Eastern Algonquian language, and in wh-in-situ interrogatives in Coptic Egyptian, an Afroasiatic language from the fourth through fourteenth centuries ad. Section 7.6 places our ﬁndings in the context of recent discussions of the diversity of wh-in-situ. Section 7.7 brieﬂy discusses the interaction of whagreement and wh-in-situ in Palauan, the Austronesian language of the Palau Islands. Our proposal is that in Palauan, wh-in-situ does not undergo movement in narrow syntax; rather, the scope of the interrogative phrase results from unselective binding. Section 7.8 provides a brief summary of our conclusions. 7.2

The Morphosyntactic Proﬁle of Wh-Agreement

In languages such as Chamorro, Hausa, Irish, Kikuyu, Moore, and Palauan, wh-constructions exhibit a special type of morphosyntactic ﬂagging, which has come to be known as wh-agreement (see Haı¨k 1990; Watanabe 1996; Chung 1998; and the references cited there). This agreement phenomenon has the following typological proﬁle: A. It occurs only in the classic wh-constructions, such as constituent questions, relative clauses, and focus constructions, which we assume, following ‘‘On Wh-Movement’’ (Chomsky 1977), are derived by whmovement. In some of these constructions, Chomsky’s diagnostics reveal

Movement, Wh-Agreement, and Apparent Wh-in-Situ

167

that wh-movement has occurred even though there seems to be no overt element at all that instantiates the wh-phrase. We take no ﬁrm position on whether such constructions involve a deleted wh-phrase (as Chomsky originally proposed) or a null operator (Browning 1987). For convenience, we will sometimes refer to the moved element in such constructions as a null operator. B. The morphological e¤ects of wh-agreement surface on V or on T, C (see Zaenen 1983), or on some other designated head in the functional layer of the clause, such as Focus (see Green 1997). C. In general, wh-agreement is not sensitive to person, number, or gender (f-features).1 D. The agreement explicitly registers the presence of a moved wh-phrase. In some languages, such as Hausa (Tuller 1986) and Irish (McCloskey 1979, 1990, 2001), this is all that wh-agreement registers; in other languages, such as Palauan (Georgopoulos 1985, 1991), it registers additional features of the moved wh-phrase, such as category type or Case. E. In long-distance constructions, in which wh-movement crosses a clause boundary, wh-agreement comes in two ﬂavors. Either (i) the agreement is recursive, in the sense that it surfaces on every designated head on the path of wh-movement; or else (ii) the agreement is nonrecursive, meaning that it surfaces only on the highest designated head on the path of whmovement. As far as we know, no other patterns are attested. We illustrate this typological proﬁle via the two types of wh-agreement found in Chamorro, an Austronesian language of the Mariana Islands. Chamorro has a standard inventory of wh-constructions, including constituent questions, relative clauses, focus constructions, and comparatives, which display all the earmarks of constructions derived via wh-movement. The language also has not just one but two types of whagreement. The ﬁrst type of agreement, called ‘‘wh-agreement’’ by Chung (1998), surfaces on V (or A) and registers the Case of the moved whphrase. Compare (3), which is a simple declarative clause, with the constituent questions in (4a) and (4b). In (4a), the nominative Case of the wh-subject is ﬂagged by the wh-agreement morpheme -um-; in (4b), the objective Case of the wh-object is ﬂagged by the morpheme -in- plus possessor agreement. These morphemes are given in italics below.2 (3) Chamorro declarative VSO clauses Ha-bendi si Maria i kareta agr-sell Maria the car ‘Maria sold the car.’

168

Chris H. Reintges, Philip LeSourd, and Sandra Chung

(4) Chamorro ‘‘wh-agreement’’ in constituent questions a. Hayi bumendi i kareta? who? wh[nom].sell the car ‘Who sold the car?’ b. Hafa bininde-nn˜a si Maria? what? wh[obj].sell-agr Maria ‘What did Maria sell?’ The same agreement pattern reappears in the context of relative clauses, which are externally headed and can either precede or follow the head NP. We consider here only postnominal relative clauses. Notice that in these and all other relative clauses, there is no overt relative pronoun distinct from the head. For simplicity’s sake, we will refer to the moved element in such constructions as a null operator, although other analyses are possible. (5) Chamorro ‘‘wh-agreement’’ in relative clauses a. Kao un-li’i’ i palao’an [ni bumendi i kareta]? Q agr-see the woman C rel wh[nom].sell the car ‘Did you see the woman who sold the car?’ b. Hu-fahan i kareta [ni bininde-nn˜a si Maria] agr-buy the car C rel wh[obj].sell-agr Maria ‘I bought the car that Maria sold.’ The second type of wh-agreement is called ‘‘operator-C agreement’’ by Chung (1994, 1998). This agreement surfaces on C and registers the category type of the moved wh-phrase, as well as its operator feature (i.e., the feature that identiﬁes it as an interrogative phrase, a relativized DP, or a focus). The realizations of ‘‘operator-C agreement’’ are most transparent when the wh-phrase is a DP denoting a location in time or space, as in the examples below. In the question (6a), which is formed on the locative phrase manu ‘where?’, the agreement is realized as the complementizer na (in the Guam dialect), ni, or nai (in the Saipan dialect). In the focus construction (6b), in which the focus is the locative gi gima’mami ‘in our house’, the agreement has the same realization. (6) Chamorro ‘‘operator-C agreement’’ in questions and focus constructions a. Manu na suma´saga hao? where? C Q agr.live.prog you ‘Where are you living?’

Movement, Wh-Agreement, and Apparent Wh-in-Situ

169

b. Gi gima’-mami nai suma´saga si Jess loc house-agr C emph agr.live.prog Jess ‘Jess is living in our house.’ But in the relative clause (7), the null operator corresponding to the relativized DP is a locative, and the agreement is realized as the complementizer a¨nai. (7) Chamorro ‘‘operator-C agreement’’ in relative clauses Hu-li’i’ i gima’ [a¨nai suma´saga si Maria] agr-see the house C rel agr.live.prog Maria ‘I saw the house that Maria is living in.’ We turn now to the patterning of these types of Chamorro whagreement in long-distance constructions. Recall from point E above that wh-agreement comes in two ﬂavors: either it surfaces (i) on every relevant head on the path of wh-movement, or it surfaces (ii) only on the highest relevant head. The agreement that Chung called ‘‘wh-agreement’’ comes in ﬂavor (i): it appears on every V on the path of wh-movement, as can be seen from the italicized verbal morphology in the long-distance constructions in (8) and (9). (8) Chamorro ‘‘wh-agreement’’ in long-distance constituent questions Hafa sinangane-nn˜a si Joaquin nu hagu [bininde-nn˜a]? what? wh[obj].tell-agr Joaquin obl you wh[obj].sell-agr ‘What did Joaquin tell you that he sold?’ (9) Chamorro ‘‘wh-agreement’’ in long-distance relative clauses Hu-tungu’ ha’ i palao’an [ni malagu’-n˜iha agr-know emph the woman C rel wh[obl].want-agr [pa¨ra ali’e’-mu]] fut wh[obl].meet-agr ‘I know the woman who they want that you meet with.’ But ‘‘operator-C agreement’’ comes in ﬂavor (ii): it appears only on the highest C on the path of wh-movement, as can be seen from the italicized complementizer alternations in (10) and (11). (10) Chamorro ‘‘operator-C agreement’’ in long-distance constituent questions Amanu na ha-tagu’ si Dolores i lahi-n˜a where? C Q wh[obj2 ].agr-order Dolores the son-agr [pa¨ra u-tohgi]? fut agr-stand ‘Where did Dolores tell her son that he should stand?’

170

Chris H. Reintges, Philip LeSourd, and Sandra Chung

(11) Chamorro ‘‘operator-C agreement’’ in long-distance relative clauses Manggi i gima’ [a¨nai ma-sangani i pa¨li’ where?.is the house C rel wh[obj2 ].agr-tell the priest [na pa¨ra u-saga]]? C fut agr-live ‘Where is the house where they told the priest that he should live?’ The agreement patterns just illustrated are not limited to constituent questions and relative clauses, but instead occur in the full range of Chamorro wh-constructions, exactly as expected if both types of agreement track the history of wh-movement. At the same time, their di¤erent distribution in long-distance constructions serves to reinforce our point that there can be parametric variation in the way the wh-agreement phenomenon is realized. This variation can occur not just across languages, but— strikingly—within a single language, as Chamorro shows quite clearly. We will have occasion to return to this point later. 7.3

Wh-Agreement as the Morphological Reﬂex of Agree

In pre-minimalist syntax, it was natural to analyze wh-agreement as the morphological reﬂex of speciﬁer-head agreement between the moved wh-phrase and a functional head, such as C (see, e.g., Watanabe 1996). Such an approach dealt well with some language-speciﬁc patterns of whagreement, but not others. (The complexity of the Chamorro pattern is problematic; see Chung 1998, 259.) Minimalist syntax o¤ers a promising alternative. Instead of appealing to the speciﬁer-head relation, we can analyze wh-agreement as a morphological reﬂex of Agree, the relation of feature valuation that must hold for movement to occur (see also Bruening 2001, 209–210). In the minimalist view of movement, in order for one category (the goal ) to move to the vicinity of another (the probe), the two must stand in the Agree relation: the probe must c-command the goal, the relation between the two must be local, and each category must have active (i.e., uninterpretable or unvalued) features whose values can be set by the corresponding features of the other (see, e.g., Chomsky 2001, 4–5). Movement is then implemented by merging the goal in the vicinity of the probe, an operation that leaves the original copy of the goal untouched. If we apply this view to the case at hand, the probe for whmovement, and therefore for wh-agreement purposes, is C, T, or one of

Movement, Wh-Agreement, and Apparent Wh-in-Situ

171

the other designated heads mentioned in point B of section 7.2; the goal is a wh-phrase. In minimalist syntax, access to previously constructed constituent structure is regulated by the Phase Impenetrability Condition (PIC), which permits operations outside a phase Ph to have access only to the head of Ph and to its edge (Chomsky 2000, 108). If we assume the PIC and take CP to be a strong phase, then it becomes clear how to analyze the two ﬂavors of wh-agreement that are manifested in long-distance whconstructions. Flavor (i) (illustrated earlier by ‘‘wh-agreement’’ in Chamorro) is the morphological reﬂex of Agree holding between a probe that bears the active feature that drives wh-movement (henceforth [X]) and a goal that bears the active feature [wh].3 Morphological agreement of this type signals the Agree relations indicated in the schematic tree in (12) (where u signiﬁes ‘‘uninterpretable’’ or ‘‘unvalued’’). (12) The probe-goal relations underlying wh-agreement, ﬂavor (i)

Flavor (ii) (illustrated earlier by ‘‘operator-C agreement’’ in Chamorro) is also a morphological reﬂex of Agree holding between a probe that bears the feature [X] and a goal that bears the feature [wh], but here the morphology also reﬂects the fact that both probe and goal bear an operator feature—[Q] in questions, [pred] in relative clauses, [emphasis] or [contrast] in focus constructions (see Rizzi 1990, 1996, 1997)—that is active (i.e., unvalued) on the probe but inactive on the goal. Morphological agreement of ﬂavor (ii), then, signals the higher Agree relation indicated in the schematic diagram in (13).

172

Chris H. Reintges, Philip LeSourd, and Sandra Chung

(13) The probe-goal relations underlying wh-agreement, ﬂavor (ii)

Notice that there is a second Agree relation in (13), one that holds between the lower F and the lowest instance of the wh-phrase. The point is that this Agree relation, though syntactically present, is not signaled morphologically by wh-agreement of ﬂavor (ii). More generally, we propose that the two types of wh-agreement spell out the abstract Agree relation holding between a wh-goal and two types of probes. While the feature composition of the probe determines the ﬂavor of wh-agreement, it is the Agree relation itself that ultimately lies behind the occurrence of this special morphology. Exactly when in the derivation wh-agreement is spelled out is a separate issue, to which we return in section 7.5. All that matters here is that the setup of the computational system does indeed permit the Agree relation to have phonetic e¤ects; on this point, see Chomsky 2001, 5. When this view of wh-agreement as the spell-out of Agree is combined with the copy theory of movement, a clear prediction emerges: it should be possible for a language to exhibit wh-agreement as well as whmovement in overt syntax, followed by pronunciation of only the lowest copy. The result should be ‘‘apparent’’ wh-in-situ whose participation in syntactic wh-movement leaves visible morphological e¤ects. This prediction is borne out by empirical evidence from Passamaquoddy and Coptic Egyptian, as we now show.

Movement, Wh-Agreement, and Apparent Wh-in-Situ

7.4

173

Case Study I: Internally Headed Relative Clauses in Passamaquoddy

Relative clauses in Passamaquoddy, an Eastern Algonquian language, may be either externally headed, as in (14), or internally headed, as in (15). The understood head of the relative clause is italicized in each example. In general, external heads precede the relative clause; internal heads occur nested within the relative clause or at its right edge. (14) Passamaquoddy relative clauses with external heads a. [No`t skita`p [nemı´y-uk wolaku`]] mace´he sepa´wonu that.prox man see-1/3 yesterday leave-(3) tomorrow ‘That man that I saw yesterday is leaving tomorrow.’ b. Nı`l n-koti¼pse´hl-a pem-kı´skah-k [no`t otu´hk I 1-going.to¼skin-dir along-be.day-3in that.prox deer [n-ta´tat nehpah-a´-c-il wolaku`]] 1-dad kill-dir-3an-part:obv:sg yesterday ‘I am going to skin today that deer that my dad killed yesterday.’ c. [Nı´ktok otu´hk-ok [n-ta´tat nehpah-a-c-ihı`]] those.prox deer-prox:pl 1-dad kill-dir-3an-part:obv:pl psı´¼te kinkil-uhtu´-w-ok all¼emph be.big-pl-3-prox:pl ‘Those deer that my dad killed are all big.’ (15) Passamaquoddy relative clauses with internal heads a. [Wolaku` skita`p nemı´y-uk] mace´he sepa´wonu yesterday man see-1/3 leave-(3) tomorrow ‘The man that I saw yesterday is leaving tomorrow.’ b. Nı`l n-koti¼pse´hl-a pem-kı´skah-k [n-ta´tat I 1-going.to¼skin-dir along-be.day-3in 1-dad nehpah-a´-c-il otu´hk-ol wolaku`] kill-dir-3an-part:obv:sg deer-obv:sg yesterday ‘I am going to skin today the deer that my dad killed yesterday.’ c. We`n kis-anku´m-ot [Pı´yel tuwihpu`t kis-ı´ht-a-q]? who? past-sell.to-2/3 Peter table past-make-th-3an ‘To whom did you sell the table that Peter made?’ (adapted from Bruening 2001, 161, (386b)) It is generally not possible for a relative clause to include both an external and an internal head. Thus, example (16) with a resumptive demonstrative pronoun nı´htol ‘that (obviative)’ is unacceptable. (The personal pronoun ne´kom ‘he, she, it (animate)’ has no obviative forms, so the only possible pronoun here is a demonstrative.)

174

Chris H. Reintges, Philip LeSourd, and Sandra Chung

(16) Ungrammaticality of relative clauses with external and internal head (¼ resumptive demonstrative) *Koma`c kı´nkil [no`t otu´hk [n-ta´tat nı´htol very.much be.big-(3) that.prox deer 1-dad that.obv nehpah-a´-c-il]] kill-dir-3an-part:obv:sg ‘That deer that my dad killed is very big.’ Evidence that external heads surface outside the relative clause, whereas internal heads surface inside, is provided by the language’s system of obviation. Roughly speaking, any expression that refers to the most topical or otherwise most prominent third person within a particular span of discourse is proximate. Expressions referring to any other third person in the same context are obviative. (However, proximate or obviative status is overtly marked only for nouns and pronouns that belong to the animate grammatical gender.) Within an obviation span, only one DP can be proximate; all other third person expressions that di¤er in reference from the proximate DP must be obviative. The minimal such domain in Passamaquoddy is ordinarily the clause. As Bruening (2001, 206) has observed, however, the external head of a relative clause can be proximate even when some other NP within the relative clause is proximate. Thus, no`t otu´hk ‘that deer’ is proximate in (17), but the subject NP nta´tat ‘my dad’ within the relative clause itself is also proximate. This argues that the external head lies outside the minimal obviation domain—here, the relative clause. (17) Proximate external head co-occurring with another proximate in the relative clause Koma`c kı´nkil [no`t otu´hk [n-ta´tat very.much be.big-(3) that.prox deer 1-dad nehpah-a´-c-il ]] kill-dir-3an-part:obv:sg ‘That deer that my dad killed is very big.’ The obviative status of the relativized correspondent of the NP no`t otu´hk ‘that deer’ is registered by two verbal markers, one of which is the socalled direct theme sign -a that is used in forms with a third person subject (as here) only if the object is third person and obviative. The other is the participle agreement su‰x -il, which indexes the relativized object NP as obviative and singular. Since the head of the relative clause in this example is proximate, but is nonetheless coreferential with the obviative object of nehpah-a´-c-il ‘he (proximate) kills him (obviative)’, this head can only be external to the relative clause itself.

Movement, Wh-Agreement, and Apparent Wh-in-Situ

175

The distribution of proximate and obviative forms also provides evidence that when the relativized constituent appears on the right periphery of the clause, the clause can only be internally headed. In (18), for example, the obviative status of the relativized constituent otu´hkol ‘deer (obviative)’ is expected if the relative clause here is head internal, since the subject NP within the clause, nta´tat ‘my dad’, is proximate. (18) Right-peripheral heads can only be internal Koma`c kı´nkil [n-ta´tat nehpah-a´-c-il otu´hk-ol ] very.much be.big-(3) 1-dad kill-dir-3an-part:obv:sg deer-obv:sg ‘The deer that my dad killed is very big.’ If ‘deer’ could instead be an external head here, however, then we would expect (19) with ‘deer (proximate)’ to be acceptable, since the head of a relative clause need not be construed as occurring within the same obviation domain as nominals contained within the clause. But (19) is, in fact, unacceptable. This argues that a right-peripheral head lies within the relative clause. (19) Proximate right-peripheral head incompatible with another proximate in the relative clause *Koma`c kı´nkil [[n-ta´tat nehpah-a´-c-il ] otu´hk] very.much be.big-(3) 1-dad kill-dir-3an-part:obv:sg deer ‘The deer that my dad killed is very big.’ Bruening (2001, 117–124) analyzes the obviation system of Passamaquoddy as analogous to abstract Case. Though we do not adopt the details of his analysis, our approach is similar in that it takes proximate and obviative morphology to encode formal features (minimally, the feature [proximate]), which are uninterpretable in the sense that they have no truth-conditional e¤ects on semantic interpretation. This assumption will come into play shortly. As Bruening (2001, 161–162) observes, relative clauses are derived by wh-movement whether their heads are internal or external. This is shown by the fact that both types of relatives exhibit island e¤ects. (20) Adjunct island e¤ects in a relative clause with external head *Wo`t nı`t [skita`p [palitaha´m-uk Pı´yel this.prox that.in man think.highly.of-1/3 Peter [eli¼kis-akonim-a´-c-il]]] thus¼past-tell.about-dir-3an-part:obv:sg ‘This is the man that I’m proud of Peter because he told about.’ (adapted from Bruening 2001, 162, (387b))

176

Chris H. Reintges, Philip LeSourd, and Sandra Chung

(21) Negative island e¤ects in a relative clause with internal head *Nı´htol nı`t [ikonewa´t-o-k Peta`k that.obv that.in deny-th-3an Petak [eli¼kisi¼komutonom-a´-c-il pahtoliya´s-ol ]] thus¼past¼steal.from-dir-3an-part:obv:sg priest-obv:sg ‘That is the priest that Petak denied that he had stolen from.’ (adapted from Bruening 2001, 162, (390a)) In Bruening’s analysis, internally headed relatives undergo covert whmovement in Logical Form (LF). We, too, assume that Passamaquoddy has both externally and internally headed relative clauses: the relativized NP in relative clauses of the ﬁrst type is a null operator, whereas the relativized NP in relative clauses of the second type is the internal head. But unlike Bruening, we propose that in both types of relative clauses, the relativized NP undergoes wh-movement in overt syntax, landing ultimately in the speciﬁer of the C bearing the active feature [pred] at the top of the relative clause. In the phonology, the lowest copy of the moved wh-phrase is pronounced (if it can be); higher copies are not pronounced. These details of pronunciation emerge clearly in internally headed relatives; in externally headed relatives, they are concealed by the fact that the moved wh-phrase is the null operator, which has no pronunciation. Evidence for our approach is supplied by participle agreement in relative clauses, which we analyze as wh-agreement. This agreement, which surfaces optionally on V, registers features of the relativized NP, including number and obviation. Consider, for instance, the externally headed relatives in (22), where the participle agreement su‰x -il indicates that the relativized NP is obviative and singular. (22) Externally headed relative clauses with participle agreement a. Koma`c kı´nkil [no`t otu´hk very.much be.big-(3) that.prox deer [n-ta´tat nehpah-a´-c-il ]] 1-dad kill-dir-3an-part:obv:sg ‘That deer that my dad killed is very big.’ b. ’Pecı´m-a-n [ putepe-hsı´s-ol (3)-call.hither-dir-sub whale-dim-obv:sg [nuta´-hc-il ]] hear-obv/prox-part:obv:sg ‘He calls (to him) a little whale that hears him.’ (adapted from Bruening 2001, 207, (512))

Movement, Wh-Agreement, and Apparent Wh-in-Situ

177

c. Piyemkil-u` [ putepi-yı` be.bigger-(3)-obv.pl whale-obv:pl [pethul-a-hti-c-ihı` ]] bring.by.boat-dir-prox:pl-3an-part:obv:pl ‘The whales that they bring in (by boat) are bigger.’ Participle agreement also occurs in internally headed relatives like (23). (23) Internally headed relative clauses with participle agreement a. Nı`l n-koti¼pse´hl-a pem-kı´skah-k I 1-going.to¼skin-dir along-be.day-3in [n-ta´tat nehpah-a´-c-il otu´hk-ol wolaku`] 1-dad kill-dir-3an-part:obv:sg deer-obv:sg yesterday ‘I am going to skin today the deer that my dad killed yesterday.’ b. Psı´¼te kinkı´l-uk [n-ta´tat nehpah-a-c-ihı` all¼emph be.big-(3)-prox:pl 1-dad kill-dir-3an-part:obv:pl otu`hk] deer-(obv:pl) ‘All of the deer that my dad killed are big.’ In long-distance relative clauses, participle agreement shows itself to be wh-agreement of ﬂavor (i): the agreement surfaces (optionally) on every V along the path of wh-movement. Crucially, this happens whether the relative clause is externally headed, as in (24), or internally headed, as in (25). (24) Participle agreement in a long-distance relative clause with external head Wo`t nı`t pahtolı´yas [Ma´li el-itahası´-c-il this.prox that.in priest Mary thus-think-3an-part:obv:sg [eli we`n kisi¼komutonom-a´-c-il ]] thus someone past¼steal.from-dir-3an-part:obv:sg ‘This is the priest that Mary thinks someone stole from.’ (adapted from Bruening 2001, 208, (516b)) (25) Participle agreement in a long-distance relative clause with internal head N-koti¼pse´hl-a [kisi¼ih-ı´-c-il n-mı´htaqs 1-going.to¼skin-dir past¼tell-3/1-3an-part:obv:sg 1-father [nehpah-a´-c-il otu´hk-ol wolaku`]] kill-dir-3an-part:obv:sg deer-obv:sg yesterday ‘I am going to skin the deer that my father told me that he killed yesterday.’

178

Chris H. Reintges, Philip LeSourd, and Sandra Chung

Bruening (2001, 229) takes this pattern to indicate that in relative clauses of type (25), the internal relative head undergoes covert wh-movement in LF. However, if that were so, the participial agreement on the higher V could not reﬂect any Agree relation holding in overt syntax between it and the internal head. The problem is that in overt syntax, the two would be lodged in separate phases: because the internal head would not yet have moved, it would remain embedded within the lower CP, which forms its own strong phase, and so the Phase Impenetrability Condition would prevent the higher V from having access to it. It is worth asking whether the higher V’s participial agreement could reﬂect any LF relation of Agree holding between it and the internal head, where by ‘‘LF relation’’ we mean one established in the covert part of the derivation, after Spell-Out. (This is evidently Bruening’s proposal; see also Bruening 2001, 209.) The issues here are controversial. So far, we have taken the Agree relation reﬂected by participial agreement to involve the uninterpretable feature [proximate]. We have further been assuming, following Chomsky (2001, 5), that ‘‘Spell-Out removes LFuninterpretable material from the syntactic object K . . . shortly after the uninterpretable features have been assigned values (if they have not been assigned values at this point, the derivation will crash, with uninterpretable features at the interface).’’ This set of assumptions would make it impossible for [proximate] to persist past Spell-Out and therefore for this uninterpretable feature to be involved in any LF relation of Agree. Alternatively, one could take the position that uninterpretable features can indeed persist past Spell-Out to the covert part of the derivation. If so, the challenge would be to ensure that an Agree relation not present in overt syntax but established only in covert syntax could have phonetic e¤ects. While an analysis along these lines could be constructed, it would be deeply subversive to the spirit of the grammatical architecture assumed by Chomsky in much work (e.g., 1981, 1995, 2001), since that architecture claims that covert syntax (LF) and PF are independent interpretive components that do not directly interact. No such di‰culties arise in our analysis. Under this analysis, participle agreement is analogous to ‘‘wh-agreement’’ in Chamorro. This agreement transparently reﬂects the relation of Agree in overt syntax that holds between V and the relativized NP, which bears the active feature [wh], as sketched in diagram (12). The main di¤erence between Passamaquoddy participle agreement and Chamorro ‘‘wh-agreement’’ lies in which copy of a wh-moved NP is pronounced. In Chamorro, only the highest copy is pronounced, whereas in Passamaquoddy, the lowest copy is pronounced,

Movement, Wh-Agreement, and Apparent Wh-in-Situ

179

leading to an internal head (when the relativized NP is not a null operator to begin with). 7.5

Case Study II: Interrogative Wh-in-Situ in Coptic

In Coptic Egyptian, the ﬁnal developmental stage of the Ancient Egyptian language, most constituent questions involve wh-in-situ. Both wharguments and wh-adjuncts can occur in situ, as shown in (26). (The canonical word order is SVO. For details of the textual sources of the examples, see Reintges 2004, 597–600.) (26) No argument-adjunct asymmetry in Coptic wh-in-situ constructions a. nt-a nim tsˇpo na-f n-tei-hypomonE . . . ? rel-perf who? acquire for-3sg:m DO-dem:sg:f-endurance ‘Who has acquired such an endurance . . . ?’ (Hilaria 12, 29; ed. Drescher) b. e-i-na-tsˇe u na-k? rel(-fut)-1sg-aux-say what? to-2sg:m ‘What shall I say to you?’ (Apophth. Patrum no. 28, 5, 25; ed. Chaıˆne) c. e-tetPn-sˇine nsa nim? rel(-pres)-2pl-search for who? ‘Whom are you looking for?’ (John 18, 4) d. awO nt-a-u-ei ebol t n? and rel-perf-3pl-come pcl where? ‘From where did they come?’ (Apocalypse 7, 14; ed. Budge) e. Pnt-a-k-ei e-pei-ma n-asˇ n-he? rel-perf-2sg:m-come to-dem:sg:m-place in-which? of-manner ‘How did you get here?’ (Coptic Martyrd. 206, 29; ed. Budge) e

c

e

Wh-in-situ has a broad syntactic distribution and occurs in matrix and embedded clauses. Main and embedded wh-in-situ di¤er, however, with respect to the scope of the wh-phrase. In main clauses, the wh-phrase takes matrix scope and the clause is interpreted as a direct question, as seen in (27a). If the wh-phrase occurs in situ in a ﬁnite embedded clause, it must take embedded scope and only the indirect question reading is available, as seen in (27b).

180

Chris H. Reintges, Philip LeSourd, and Sandra Chung

(27) Matrix versus embedded wh-in-situ a. Pnt-a u sˇOpe Pmmo-k pe.n-tsˇoeis rel-perf what? happen to-2sg:m def:sg:m.1pl-lord p-Prro? def:sg:m-king ‘What happened to you, our lord and king?’ (Eudoxia 36, 24; ed. Orlandi) b. awO n-ti-sown an [tsˇe nt-a u sˇOpe and neg(-pres)-1sg-know not C rel-perf what? happen mmo-s] to-3sg:m ‘And I don’t know what happened to her.’ (Hilaria 7, 30–31) It is also possible, though not common, for the interrogative phrase to surface at the left edge of the clause, in a position that, following Green and Reintges (2004) and Reintges and Green (2004), we analyze as the speciﬁer of a Focus head located below C and above T. This displacement, we assume, is the result of wh-movement. Constituent questions with overt wh-movement are exempliﬁed in (28a–c). (28) Coptic constituent questions with overt wh-movement a. nim a-f-ent-k e-pei-ma? who? perf-3sg:m-bring-2sg:m to-dem:sg:m-place ‘Who has brought you here?’ (KHML I 3, 7–8) b. ebol t n a-tetPn-ei e-pei-ma? pcl where? perf-2pl-come to-dem:sg:m-place ‘From where did you come here?’ (Coptic Martyrd. 220, 8) c. awO n-asˇ n-he anon tPn-sOtPm p-wa and in-which? of-manner we (pres)-1pl-hear def:sg:m-one p-wa hraı¨ hPn te.f-ape . . . ? def:sg:m-one pcl in def:sg:m.3sg:f-language ‘And how do we hear each one (of us) in his (mother-) tongue . . . ?’ (Acts 2, 8; ed. Thompson) c

e

Because Coptic is no longer spoken, it is impossible to establish with certainty that constituent questions of types (26)–(28) are sensitive to islands, and so involve wh-movement. The best that can be said is that

Movement, Wh-Agreement, and Apparent Wh-in-Situ

181

there are no attested examples of questions of either type in which the interrogative phrase originates inside an island and takes scope outside the island. What interests us here is that wh-in-situ questions are ﬂagged by relative tenses, so called because they are derived from ‘‘basic’’ (absolute) tenses by the addition of the relative markers e- and nt- to the left edge of the verbal cluster. The relative preﬁx nt- is selected in the context of the perfect marker a, while the allomorph e- is chosen in all other contexts. Since the morphological alternation of the relative marker is dependent on the following tense inﬂection (a null morpheme in the present tense), it provides prima facie evidence for the correlation between this special inﬂection and the tense/ﬁniteness system. In general, relative tenses occur in the classic constructions formed by wh-movement, such as relative clauses and declarative focus sentences.4 In relative constructions (which in Coptic are externally headed), there is no subordinating complementizer besides the relative tense marker, as shown in (29a–b). e

e

(29) Coptic relative clauses a. u-hOb [ere p-nute moste mmo-f ] indef:sg-thing rel(-pres) def:sg:m-god hate DO-3sg:m ‘a thing that God hates’ (Acts Andrew & Paul 202, 126–127; ed. Jacques) b. e-p-ma [ nt-a-k-ky nt-f nhEt-f ] to-def:sg:m-place rel-perf-2sg:m-ﬁnd-3sg:m inside-3sg:m ‘the place where you have found it (the boat)’ (Acts Andrew & Paul 204, 145–146) e

In the question-answer pair in (30), the focus-in-situ e-p-topos n-Apa MEna ‘to the shrine of Apa Mena’ in the answer corresponds to the whin-situ locative argument t n ‘where’ in the question. c

(30) Coptic focus-in-situ constructions Q: e-r-bEk e-t n? rel(-pres)-2sg:f-go to-where A: e-i-bEk e-p-topos n-Apa MEna rel(-pres)-1sg-go to-def:sg:m-shrine of Apa Mena nta-sˇlEl conj:1sg-pray ‘Where are you (woman) going to? – I am on my way to the shrine of Apa Mena to pray.’ (Mena, Mirc. 27 b , 22–25; ed. Drescher) c

182

Chris H. Reintges, Philip LeSourd, and Sandra Chung

As might be expected by this point, we claim that relative tense is a form of wh-agreement: it spells out the Agree relation holding between a probe of some type and a goal that bears the active feature [wh]. Wh-agreement takes the form of a relativizing complementizer that is merged in the leftperipheral Focus head. To determine whether this wh-agreement is of ﬂavor (i) or ﬂavor (ii), we must ask how relative tense patterns in longdistance constituent questions. Reintges’s (2002, 2003, 2004) work has uncovered no examples in which the wh-in-situ occurs within an embedded clause containing a ﬁnite verb, but takes scope over the matrix clause (see above, (27b)). However, there are attested examples in which a wh-in-situ occurs as the nonverbal predicate of an embedded clause, but nonetheless takes matrix scope. In these examples, relative tense marking surfaces in the left periphery of the matrix clause. (31) Coptic relative tense in a long-distance question with a wh-in-situ predicate nominal eye PntOtPn e-tetPn-tsˇO Pmmo-s ero-i [tsˇe ang nim]? Q you(-pl) rel(-pres)-2pl-say DO-3sg:f about-1sg C I who? ‘Who are you saying of me that I (am)?’ (Mark 8, 29; ed. Horner) There are also examples in which the wh-in-situ occurs within an embedded inﬁnitive, but nonetheless takes matrix scope.5 Once again, relative tense marking occurs in the highest clause over which the wh-in-situ takes scope. (32) Coptic relative tense in a long-distance question with wh-in-situ in an inﬁnitival a. e-n-Pr khria k y e Pm-mPntre [e-Pr u tsˇin rel(-pres)-make need pcl as-witness to-make(-inf) what? since tenu]? now ‘What further witness do we need to bear now?’ (Mark 14, 63) b. e-k-wOsˇ [e-tre-n-sobte na-k t n rel(-pres)-3sg:m-want to-caus:inf-1pl-prepare for-2sg:m where? Pm-p-paskha [e-wom-Pf ]]? DO-def:sg:m-paskha to-eat-3sg:m ‘Where do you want to let us prepare the Pasha(-meal) for you to eat?’ (Matthew 26, 17) c

Movement, Wh-Agreement, and Apparent Wh-in-Situ

183

This pattern argues that relative tense is wh-agreement of ﬂavor (ii): it surfaces in the highest clause along the path of wh-movement, ﬂagging only the highest copy—which, of course, is unpronounced. In other words, relative tense is entirely analogous to Chamorro ‘‘operator-C agreement,’’ diagrammed in (13) above. Crucially, the parallel holds even though constituent questions in Chamorro require overt whmovement, whereas in the relevant constituent questions in Coptic the wh-phrase surfaces in situ. The alternation in the form of the relative marker reveals that Coptic operator-C agreement interacts with the realization of tense and aspect in ways that could indicate movement of T to C, or some equivalent (see Pesetsky and Torrego 2001). We pass over this here (but see Reintges 2003). What is important is that we take relative tense in Coptic to transparently reﬂect the Agree relation in overt syntax, where this relation holds between a probe that bears the active features [X] and [Q] (presumably because it has acquired these features from the closest C above it) and a goal that is a wh-phrase. The main di¤erence between Coptic relative tense and Chamorro ‘‘operator-C agreement’’ lies in which copy of wh-movement is pronounced: in Chamorro, only the highest copy is pronounced, whereas in Coptic wh-in-situ constructions such as (26), (27), (31), and (32a–b), only the lowest copy is pronounced. We can take the analysis further by recognizing that when interrogative phrases in Coptic undergo overt wh-movement, which raises them to the speciﬁer of FocusP, relative tense marking does not occur. Consider (28) above as well as the long-distance constituent question in (33), which exhibit overt wh-movement but no relative tense marking. (33) Long-distance constituent question with overt wh-movement and no relative tense n-asˇ n-he Pntok k-tsˇO Pmmo-s in-which? of-manner you(-sg:m) (pres-)2sg:m-speak DO-3sg:f [tsˇe tet(n)-na-Pr rPmhe]? C (fut-)2pl-aux-make free.man ‘How do you say that you will become a free man?’ (John 8, 33) e

e

How is the absence of relative tense in these constructions to be explained? An answer emerges within our approach, which maintains that constituent questions in Coptic always involve wh-movement in overt syntax. Suppose further that relative tense is spelled out only in the phonology, at roughly the point when it is determined which copy of

184

Chris H. Reintges, Philip LeSourd, and Sandra Chung

wh-movement is pronounced. Then the two pronunciation issues would be interrelated. If only the lowest copy of the interrogative phrase is pronounced, then relative tense is spelled out overtly (see (34a)). But if only the highest copy of the interrogative phrase is pronounced, then the spellout of relative tense, which would occur within the same minimal projection, is blocked (see (34b)).6 (34) The syntactic distribution of relative tenses in Coptic wh-questions a. C [FocP wh rel [FinP T [IP . . .wh. . . ]]] b. C [FocP wh rel [FinP T [IP . . . wh . . . ]]] The situation is reminiscent of the Doubly Filled Comp Filter in English (Chomsky and Lasnik 1977), which allows the pronunciation of either a complementizer or the wh-phrase in its speciﬁer, but not both. 7.6

The Diversity of Wh-in-Situ

In short, participle agreement in Passamaquoddy bears a striking resemblance to Chamorro ‘‘wh-agreement’’; relative tense in Coptic bears a striking resemblance to Chamorro ‘‘operator-C agreement.’’ These resemblances can be explained if all four patterns are instances of the wh-agreement phenomenon, and wh-agreement is the morphological reﬂex of Agree holding between a probe that bears the feature [X] and a goal that bears the feature [wh]. When both probe and goal also bear an operator feature, what emerges is the Coptic/‘‘operator-C agreement’’ pattern; when the probe bears no other relevant features, what emerges is the Passamaquoddy/‘‘wh-agreement’’ pattern. Crucial to this account is the assumption that some instances of wh-in-situ are only apparent, in that they arise as a consequence of wh-movement in overt syntax, followed by pronunciation of the lowest as opposed to the highest copy. (For discussion of another type of case in which wh-in-situ is only apparent, see Simpson and Bhattacharya 2003.) In this section, we situate our conclusions in the context of recent theoretical discussions of the diversity of wh-in-situ. Our arguments that instances of wh-in-situ in Passamaquoddy and Coptic are derived via syntactic wh-movement before the derivation is shipped to the phonology do not entail that every instance of wh-in-situ in natural language involves a hidden movement conﬁguration. In fact, our proposal is fully compatible with the idea that there is no unitary wh-in-situ phenomenon: some types of wh-in-situ may be derived by movement in overt syntax, while other types may involve covert movement after Spell-Out (Nissenbaum 2000; Richards 2001), and still others

Movement, Wh-Agreement, and Apparent Wh-in-Situ

185

may participate in no movement whatsoever (Reinhart 1997). The possibility that there are distinct types of wh-in-situ was ﬁrst raised by Cole and Hermon (1994). The claim has been investigated in depth by Cheng and Rooryck (2002), who demonstrate the diversity of wh-in-situ not only across languages but also within individual languages (such as European Portuguese and Mandarin). Our analysis is compatible with the spirit of Cheng and Rooryck’s proposal, though not with the details, since we document a type of wh-in-situ that they do not recognize. The claim that there is no uniform analysis of wh-in-situ makes a prediction when combined with our assumption that wh-agreement tracks the history of wh-movement before Spell-Out. The prediction is this. One might expect to ﬁnd languages that di¤er from Coptic and Passamaquoddy in that wh-agreement ﬂags overt wh-movement but, crucially, not wh-in-situ. But such languages should provide other evidence besides the absence of this special inﬂection that wh-in-situ does not involve movement in overt syntax. Palauan seems to be just such a language. 7.7

Case Study III: ‘‘True’’ Wh-in-Situ in Palauan

Wh-movement, wh-agreement, and wh-in-situ in Palauan are investigated by Georgopoulos (1985, 1991), whose work we draw on here. In this Austronesian language, overt wh-movement is ﬂagged by a type of whagreement that registers the Case of the moved wh-phrase (similar to Chamorro ‘‘wh-agreement’’) but is signaled by mood morphology on the verb.7 Compare the simple declarative clause in (35) with the constituent questions in (36a) and (36b). In (36a), the nominative Case of the whsubject is ﬂagged by realis morphology and the absence of subject-verb agreement; in (36b), the nonnominative Case of the wh-object is ﬂagged by irrealis morphology (even though the semantic mood is realis). (35) Palauan declarative VOS clauses ng-’illebed-ii a bilis a buik real-3sg-hit-3sg dog boy ‘The boy hit the dog.’ (Georgopoulos 1991, 26, (7d)) (36) Palauan wh-agreement in constituent questions a. ng-te’a [a kileld-ii a sub]? cl-who? real-perf-heat-3sg soup ‘Who heated up the soup?’ (Georgopoulos 1991, 70, (19a))

186

Chris H. Reintges, Philip LeSourd, and Sandra Chung

b. ng-te’a [a l-ulekod-ir a rubak]? cl-who? irreal-3-perf-kill-3sg old.man ‘Who did the old man kill?’ (Georgopoulos 1991, 70, (19c)) The long-distance questions in (37a–b) reveal that Palauan wh-agreement is of ﬂavor (i): it surfaces on every verb on the path of wh-movement. (37) Palauan wh-agreement in long-distance questions with overt wh-movement a. ng-ngera [a ’om-dilu [el lo-ngiil er ngak cl-what? irreal-2-said C irreal-3-wait prep me [el bo kuruul er ngii]]]? C irreal-fut irreal-do prep it ‘What did you say that they are waiting for me to do?’ (Georgopoulos 1991, 93, (55d)) b. ng-mele’ede’ [el kmo ng-ngera a lo-ngiil er ngii real-3s-wonder C cl-what? irreal-3-wait prep him [el bo loruul ]] C irreal-fut irreal-do ‘He is wondering what she is waiting for him to do.’ (‘*What is he wondering that she is waiting for him to do?’) (Georgopoulos 1991, 153, (25a)) Alongside questions formed by overt wh-movement, Palauan productively allows wh-in-situ, as (38a–b) show. In (38a), the wh-in-situ is a subject; in (38b), a direct object. Crucially, wh-agreement is not triggered by whin-situ, as can be seen by comparing these examples with (37a–b). Notice further that wh-in-situ is possible with matrix scope (as in (38a)) or embedded scope (as in (38b)); for discussion, see Georgopoulos 1991, 152–155. (38) Palauan wh-in-situ constructions a. ng-osiik er ngak a te’ang? real-3sg-look.for prep me who? ‘Who is looking for me?’ (Georgopoulos 1991, 69, (18d)) b. ng-mele’ede’ [el kmo ng-mengiil er ngii [el mo real-3sg-wonder C real-3sg-wait prep him C real-aux meruul a ngerang]] real-do what? ‘He is wondering what she is waiting for him to do.’ (Georgopoulos 1991, 153, (25b))

Movement, Wh-Agreement, and Apparent Wh-in-Situ

187

We therefore expect there to be other Palauan evidence that argues that wh-in-situ does not involve movement in overt syntax, followed by pronunciation of the lowest copy. There is such evidence, though of a rather surprising sort. As mentioned in note 7, overt wh-movement in Palauan can evidently cross the boundaries of certain islands, including complex DP islands and embedded question islands (see Georgopoulos 1991, 80–82). But whin-situ observes embedded question islands in the following sense: it is impossible for a wh-in-situ to take scope outside an embedded question. Consider (39), in which the matrix verb obligatorily selects an embedded question. Even though this embedded question contains two instances of interrogative wh-in-situ, neither interrogative phrase can take matrix scope. (39) Multiple embedded wh-in-situ with no matrix scope a Tmerukl a uleker [el kmo ng-mengiil er ngii [el mo Tmerukl real-ask C real-3sg-wait prep him C real-aux meruul a ngera a te’ang]] real-do what? who? ‘Tmerukl asked who is waiting for him to do what.’ (‘*Who did Tmerukl ask is waiting for him to do what?’) (‘*What did Tmerukl ask who is waiting for him to do?’)8 (Georgopoulos 1991, 172, (62a)) Georgopoulos (1991, 171–173) also shows that the judgments do not change if one of the interrogative phrases in (39) undergoes overt whmovement within the embedded question. The broader generalization, then, is that wh-phrases that surface within an embedded question, including instances of wh-in-situ, cannot take scope outside it. Further evidence for the correctness of this generalization is provided by question-answer pairs like (40). (40) Multiple embedded wh-phrases must have identical scopes Q: ng-te’a a medengei [el kmo ng-te’a a mil’erar a agr-who? real-remember C agr-who? real-bought ngerang]? what? ‘Who remembers who bought what?’ A: a. ngak I ‘I (do).’

188

Chris H. Reintges, Philip LeSourd, and Sandra Chung

b.

ak-medengei [el kmo a Sabeth a mil’ar a real-1sg-remember C Elisabeth real-bought sasimi me a Miriam a mil’ar a rode’] sashimi and Miriam real-bought fruit ‘I remember that Elisabeth bought sashimi and Miriam bought fruit.’ c. ??ak-medengei [el kmo ng-te’a a mil’ar a real-1sg-remember C cl-who? real-bought sasimi e a Joe a medengei [el kmo ng-te’a a sashimi and Joe real-remember C cl-who? mil’ar a rode’] real-bought fruit ‘I remember who bought sashimi and Joe remembers who bought fruit.’ (Georgopoulos 1991, 169, (54))

In the question in (40), the matrix verb can select either an embedded question or a declarative clause as its complement. One argument of the matrix verb is a displaced wh-phrase; both arguments of the embedded verb are wh-phrases that surface within the embedded clause, one displaced and the other in situ. Structures of this type are clearly questions in which the matrix wh-phrase has wide scope. What are the possible interpretations of the embedded wh-phrases? The answers in (40) reveal that the embedded wh-phrases can both take matrix scope (in which case the matrix verb selects a declarative clause; see answer (b)) or they can both take embedded scope (in which case the matrix verb selects an embedded question; see answer (a)). But it is impossible for one embedded wh-phrase to have wide scope while the other has embedded scope (see answer (c) and the discussion in Georgopoulos 1991, 168– 171). In other words, it is (once again) highly deviant for wh-phrases that surface within an embedded question, including instances of wh-in-situ, to take scope outside that question. To recapitulate, overt wh-movement in Palauan can evidently cross the boundaries of embedded question islands, whereas wh-in-situ appears to be sensitive to embedded question islands in the sense just outlined. This surprising di¤erence conﬁrms that wh-in-situ in Palauan is not, after all, derived via syntactic wh-movement. Were it (counterfactually) the case that all instances of wh-in-situ involved movement in overt syntax, the asymmetry between wh-in-situ and overt wh-movement with respect to island violations would be hard to explain. But the Palauan pattern

Movement, Wh-Agreement, and Apparent Wh-in-Situ

189

receives a straightforward account if we adopt an analysis in the spirit of Pesetsky 1987 and much subsequent work (see, e.g., Cheng 1991; Cole and Hermon 1994, 1998; Reinhart 1997; Cheng and Rooryck 2002), according to which some types of wh-in-situ are indeﬁnites, interpreted as free variables eligible for unselective binding. Suppose we assume that in Palauan, in-situ wh-phrases are interpreted as free variables that must be bound by the nearest c-commanding Q operator (i.e., the nearest ccommanding C that bears the feature [Q]). Then the fact that wh-in-situ cannot take scope outside embedded questions, and in this sense appears to respect embedded question islands, will follow immediately.9 The analysis that we are suggesting is not unusual from the standpoint of the typology of indeﬁnites; see Matthewson 1999 on St’at’imcets ku, another morphologically marked indeﬁnite that must be bound by a ccommanding operator and might conceivably force narrowest scope. At the same time, the analysis makes further predictions whose status remains unclear. For instance, if wh-in-situ in Palauan never undergoes movement of any sort, it should be able to take scope outside adjunct islands. Evidence bearing on this prediction is, as far as we know, not available. Setting aside these uncertainties, we provisionally conclude that wh-in-situ in Palauan does not undergo movement; rather, it is licensed by unselective binding. 7.8

Conclusions

According to the copy theory of movement, moving a category does not introduce a trace; instead, it leaves behind a complete copy of that category. Which copy is pronounced is then determined in the phonology. This theory allows for an option not available in trace theory, namely, that the copy of the moved element designated for pronunciation in a particular language may be the lowest one rather than the highest. The result is movement in overt syntax that is not revealed by the pronounced form of a sentence. We have argued that exactly this predicted type of overt but hidden movement is instantiated in internally headed relative clauses in Passamaquoddy and in Coptic Egyptian questions involving apparent wh-in-situ. Both in Passamaquoddy and in Coptic, the application of whmovement is overtly signaled by wh-agreement. In Passamaquoddy relative clauses, every verb along the path of wh-movement may bear a participle su‰x, which registers properties of the moved category. In constituent questions in Coptic, relative tense occurs in the highest clause

190

Chris H. Reintges, Philip LeSourd, and Sandra Chung

along this path. That movement plays a role in the Passamaquoddy construction is conﬁrmed by the fact that relativization obeys island constraints. Since Coptic is no longer spoken, direct conﬁrmation of this kind is not available, but all of the attested examples are consistent with the claim that island constraints hold here as well. Despite superﬁcial similarities, the wh-in-situ constructions of Palauan contrast sharply with Passamaquoddy relatives and Coptic constituent questions. While wh-agreement is found in Palauan questions that are formed with overt wh-movement, this agreement is not triggered by whin-situ. Moreover, while wh-in-situ constructions in Palauan appear to respect embedded question islands, this e¤ect results not from movement but from the interpretation of wh-in-situ as indeﬁnites that function as free variables eligible for unselective binding. In conclusion, we have used wh-agreement to argue that wh-in-situ in some languages is the product of wh-movement in overt syntax, followed by pronunciation of the lowest copy. At the same time, our use of whagreement as a diagnostic for wh-movement o¤ers further evidence that the inventory of wh-in-situ is richer than one might suppose. Internally headed relative clauses in Passamaquoddy and wh-in-situ in Coptic involve wh-movement in overt syntax followed by pronunciation of the lowest copy, while wh-in-situ in Palauan evidently does not. Notes Our thanks to Benjamin Bruening, Marcel den Dikken, James McCloskey, Jairo Nunes, Jean-Yves Pollock, and a reviewer for their thoughtful, constructive comments. None of them necessarily agrees with our conclusion. 1. Hausa might conceivably provide a counterexample to the claim that whagreement is not sensitive to f-features. In Hausa, verbal inﬂection is realized on the so-called person-aspect complex (PAC), which occurs immediately before the verb and is morphologically independent of it. The PAC consists of the subject agreement pronoun and a tense-aspect-mood marker, which are sometimes fused together. In this language, wh-agreement is instantiated in the PAC as well. Depending on how one chooses to decompose the PAC in such cases, one might infer that wh-agreement either is fused with a morpheme that encodes f-features or else sometimes encodes f-features itself. Further, some wh-constructions in Hausa display a morpheme known as the ‘‘focus copula,’’ which registers the number and gender of the moved wh-phrase. If one chooses to view the focus copula as a realization of wh-agreement, then in such cases, wh-agreement would encode f-features. See Newman 2000 and Jaggar 2001 for further discussion. 2. The following abbreviations are used in the glosses: 1 ‘ﬁrst person’; 2 ‘second person’; 3 ‘third person’; 1/3 etc. ﬁrst person subject with third person object, etc.; agr ‘agreement’; an ‘animate’; aux ‘auxiliary verb’; C ‘subordinating complementizer’; C Q ‘interrogative complementizer’; C rel ‘relative complementizer’;

Movement, Wh-Agreement, and Apparent Wh-in-Situ

191

caus:inf ‘causative inﬁnitive’; dim ‘diminutive’; dir ‘direct’; DO ‘direct object marker (oblique Case)’; emph ‘emphatic’; fut ‘future’; in ‘inanimate’; inf ‘inﬁnitive’; irreal ‘irrealis’; loc ‘local morphological case’; nom ‘nominative Case’; obj ‘objective Case’; obl ‘oblique morphological case’; obv ‘obviative’; part ‘participle’; pcl ‘particle’; perf ‘perfect’; pl ‘plural’; prep ‘preposition’; pres ‘present tense’; prog ‘progressive’; prox ‘proximate’; Q ‘yes/no question particle’; real ‘realis’; rel ‘relative marker’; sg ‘singular’; sub ‘subordinative’; th ‘theme-forming su‰x’; wh ‘wh-agreement’. Glosses are given in parentheses for morphemes that have no surface-segmental shape. 3. The precise identity of the feature that drives wh-movement in successive-cyclic cases is not known (for some commentary, see Chomsky 2000, 128n91; 2001, 6). Furthermore, it is not clear what feature of the probe in wh-movement sets the value for the feature [wh] of the goal. We are not able to deal with these issues here. 4. Relative tenses can also occur in operator-variable constructions that are not included in Chomsky’s (1977) original taxonomy, for instance, yes/no questions, broad focus constructions (with VP focus), and subject and object depictives; for a more detailed discussion, see Reintges 2003, 2004. 5. Ouhalla (1996) shows that in Iraqi Arabic, wh-in-situ can take matrix scope if embedded in an inﬁnitival clause, but not if embedded in a ﬁnite clause. This is reminiscent of the Coptic facts reported here. Note, however, that when a wh-insitu predicate in Coptic is embedded in a ﬁnite clause, it can take matrix scope, and this is ﬂagged overtly by relative tense. We do not know what the corresponding facts are in Iraqi Arabic. 6. This generalization has an apparent exception: interrogative adverbials expressing cause or reason can co-occur with matrix relative tense marking even when they surface in the speciﬁer of Focus, as in (i). (i) awO etbe u e-k-kEk ahEu? and for what? rel(-pres)-strip.o¤stat naked ‘And why are you naked?’ (Coptic Martyrd. 207, 1) Reintges (2002, 375n8) analyzes why-questions of the kind in (i) as truncated cleft sentences, meaning something like ‘For what (reason) (is) it that you are naked?’. Truncated cleft sentences di¤er from their nontruncated counterparts in the absence of an agreement-marking deictic copula. Under this analysis, the relative tense marker e- would be used in its prototypical function as a relative complementizer. 7. Wh-constructions in Palauan involve either a gap or a resumptive pronoun. Georgopoulos (1991) argues, citing the distribution of gaps and resumptive pronouns, that all wh-constructions involve resumption as opposed to movement. But she also shows that wh-dependencies respect adjunct islands (1991, 116), though they are insensitive to embedded question islands and complex DP islands (1991, 80–82). How to handle this complexity is not entirely obvious. Here we follow Watanabe (1996) in assuming that at least some wh-constructions involving gaps are produced by overt wh-movement. Georgopoulos achieves some of the same results by proposing that all resumptive pronouns, both null and overt, are syntactic variables.

192

Chris H. Reintges, Philip LeSourd, and Sandra Chung

8. The intended meaning is roughly ‘Tmerukl asked who is waiting for Tmerukl to do something, and what is that something?’ 9. The discussion in the text assumes, following Georgopoulos (1991), that overt wh-movement of interrogative phrases can cross the boundaries of an embedded question island. Should that prove not to be so, the Palauan wh-in-situ discussed in this section would be amenable to an analysis involving LF wh-movement, along the lines suggested by Richards (2001) for Japanese. References Bobaljik, Jonathan David. 2002. A-chains at the PF-interface: Copies and ‘‘covert’’ movement. Natural Language and Linguistic Theory 20, 197–267. Bosˇkovic´, Zˇeljko. 2001. On the nature of the syntax-phonology interface: Cliticization and related phenomena. Amsterdam: Elsevier. Brody, Michael. 1995. Lexico-Logical Form: A radically minimalist theory. Cambridge, Mass.: MIT Press. Browning, M. A. 1987. Null operator constructions. Doctoral dissertation, MIT. Bruening, Benjamin. 2001. Syntax at the edge: Cross-clausal phenomena and the syntax of Passamaquoddy. Doctoral dissertation, MIT. Cheng, Lisa L.-S. 1991. On the typology of wh-questions. Doctoral dissertation, MIT. Published, New York: Garland, 1997. Cheng, Lisa L.-S., and Johan Rooryck. 2000. Licensing wh-in-situ. Syntax 3, 1–19. Cheng, Lisa L.-S., and Johan Rooryck. 2002. Types of wh-in-situ. Ms., Leiden University. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Chomsky, Noam, and Howard Lasnik. 1977. Filters and control. Linguistic Inquiry 8, 425–504. Chung, Sandra. 1994. Wh-agreement and ‘‘referentiality’’ in Chamorro. Linguistic Inquiry 25, 1–44. Chung, Sandra. 1998. The design of agreement: Evidence from Chamorro. Chicago: University of Chicago Press.

Movement, Wh-Agreement, and Apparent Wh-in-Situ

193

Cole, Peter, and Gabriella Hermon. 1994. Is there LF wh-movement? Linguistic Inquiry 25, 239–262. Cole, Peter, and Gabriella Hermon. 1998. The typology of wh-movement: Whquestions in Malay. Syntax 1, 221–238. Georgopoulos, Carol. 1985. Variables in Palauan syntax. Natural Language and Linguistic Theory 3, 59–94. Georgopoulos, Carol. 1991. Syntactic variables: Resumptive pronouns and A 0 binding in Palauan. Dordrecht: Kluwer. Green, Melanie. 1997. Focus and copular constructions in Hausa. Doctoral dissertation, School of Oriental and African Studies, London. Green, Melanie, and Chris H. Reintges. 2004. Syntactic anchoring in Hausa and Coptic wh-constructions. In Andrew Simpson, ed., Proceedings of the Twentyseventh Annual Meeting of the Berkeley Linguistics Society, 61–73. Berkeley: University of California, Berkeley Linguistics Society. Groat, Erich, and John O’Neil. 1996. Spell-Out at the LF interface: Achieving a uniﬁed syntactic computational system in the minimalist framework. In Werner Abraham, Samuel David Epstein, Ho¨skuldur Thra´insson, and C. Jan-Wouter Zwart, eds., Minimal ideas: Syntactic studies in the minimalist framework, 113– 139. Amsterdam: John Benjamins. Haı¨k, Isabelle. 1990. Anaphoric, pronominal and referential INFL. Natural Language and Linguistic Theory 8, 347–374. Jaggar, Philip J. 2001. Hausa. Amsterdam: John Benjamins. Matthewson, Lisa. 1999. On the interpretation of wide-scope indeﬁnites. Natural Language Semantics 7, 79–134. McCloskey, James. 1979. Transformational syntax and model theoretic semantics: A case study in Modern Irish. Dordrecht: Reidel. McCloskey, James. 1990. Resumptive pronouns, A 0 -binding, and levels of representation in Irish. In Randall Hendrick, ed., The syntax of the modern Celtic languages, 199–248. New York: Academic Press. McCloskey, James. 2001. The morphosyntax of WH-extraction in Irish. Journal of Linguistics 37, 67–100. Newman, Paul. 2000. The Hausa language: An encyclopedic reference grammar. New Haven, Conn.: Yale University Press. Nissenbaum, Jonathan W. 2000. Investigations of covert phrase movement. Doctoral dissertation, MIT. Nunes, Jairo. 2004. Linearization of chains and sideward movement. Cambridge, Mass.: MIT Press. Ouhalla, Jamal. 1996. Remarks on the binding properties of wh-pronouns. Linguistic Inquiry 27, 676–707. Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Eric J. Reuland and Alice G. B. ter Meulen, eds., The representation of (in)deﬁniteness, 98–129. Cambridge, Mass.: MIT Press.

194

Chris H. Reintges, Philip LeSourd, and Sandra Chung

Pesetsky, David. 1997. Optimality Theory and syntax: Movement and pronunciation. In Diana Archangeli and D. Terence Langendoen, eds., Optimality Theory: An overview, 134–170. Oxford: Blackwell. Pesetsky, David. 1998. Some optimality principles of sentence pronunciation. In Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky, eds., Is the best good enough?, 337–383. Cambridge, Mass.: MIT Press and MITWPL. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: MIT Press. Pesetsky, David, and Esther Torrego. 2001. T-to-C movement: Causes and consequences. In Michael Kenstowicz, ed., Ken Hale: A life in language, 355–426. Cambridge, Mass.: MIT Press. Reinhart, Tanya. 1997. Wh-in-situ in the framework of the Minimalist Program. Natural Language Semantics 6, 29–56. Reintges, Chris H. 2002. A conﬁgurational approach to Coptic second tenses. Lingua Aegyptia—Journal of Egyptian Language Studies 10, 343–388. Reintges, Chris H. 2003. Syntactic conditions on special inﬂection in Coptic interrogatives. In Jacqueline Lecarme, ed., Research in Afroasiatic grammar II: Papers from the Fifth Conference on Afroasiatic Languages (Paris, June 2000), 363–408. Amsterdam: John Benjamins. Reintges, Chris H. 2004. Coptic Egyptian (Sahidic dialect): A learner’s grammar. Cologne: Ru¨diger Ko¨ppe Verlag. Reintges, Chris H., and Melanie Green. 2004. Coptic second tenses and Hausa relative aspects: A comparative view. Lingua Aegyptia – Journal of Egyptian Language Studies 12, 157–177. Richards, Norvin. 2001. Movement in language: Interactions and architectures. Oxford: Oxford University Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1996. Residual verb second and the Wh-Criterion. In Adriana Belletti and Luigi Rizzi, eds., Parameters and functional heads: Essays in comparative syntax, 63–90. Oxford: Oxford University Press. Rizzi, Luigi. 1997. The ﬁne structure of the left periphery. In Liliane Haegeman, ed., Elements of grammar: Handbook in generative syntax, 281–337. Dordrecht: Kluwer. Simpson, Andrew, and Tanmoy Bhattacharya. 2003. Obligatory overt whmovement in a wh-in-situ language. Linguistic Inquiry 34, 127–142. Tuller, Laurie. 1986. Bijective relations in Universal Grammar and the syntax of Hausa. Doctoral dissertation, UCLA. Watanabe, Akira. 1996. Case absorption and wh-agreement. Dordrecht: Kluwer. Zaenen, Annie. 1983. On syntactic binding. Linguistic Inquiry 14, 469–504.

PART III Cyclicity and Locality

Chapter 8 Conceptions of the Cycle

8.1

Howard Lasnik

The Development of the Cyclic Principle

Since the mid-1960s, the cyclic principle, in one form or another, has played an important role in syntactic theorizing. The core concept has remained relatively stable, but there have been a variety of technical implementations, and one substantial augmentation, ‘‘successive cyclicity,’’ introduced in the early 1970s and ﬁguring prominently in ‘‘On WhMovement’’ (Chomsky 1977). It will be my modest purpose here to survey some of these developments, exploring their motivations and consequences.1 The cyclic principle, along with analysis into distinctive features, was one of two major syntactic innovations of the mid-1960s borrowed from phonology. The principle was ﬁrst formulated in Chomsky, Halle, and Luko¤ 1956, and applied in the phonological analysis of a variety of languages from the early 1960s on. Chomsky and Halle 1968, the monumental examination of the phonology of English, made crucial use of the cycle in virtually every analysis. The fundamental idea is that complex words have nested internal structure and that (certain) phonological rules apply ﬁrst to the most deeply embedded constituent, then to the next most deeply embedded, and so on. In (1), for example, ﬁrst the Word Stress Rule applies to the individual components black, board, and eraser, assigning each a primary stress indicated by ‘‘1’’; then the Compound Stress Rule, which reassigns primary stress to the leftmost primary stress in its domain, applies to black board; and ﬁnally it applies to black board eraser.2 Each reassignment of primary stress weakens all nonprimary stresses in the domain by one degree. (1) [ N [ N [ A black] [ N board]] [ N eraser]] 1 1 1

198

Howard Lasnik

(2) a. [ N [ A black] [ N board]] 1 2 b. [ N [ N [ A black] [ N board]] [ N eraser]] 1 3 2 In this case, and in a vast number of others, such cyclic application correctly accounts for the facts. Chomsky and Halle sum up the situation as follows: ‘‘it is natural to suppose that in general the phonetic shape of a complex unit . . . will be determined by the inherent properties of its parts and the manner in which these parts are combined, and that similar rules will apply to units of di¤erent levels of complexity. These observations suggest a general principle for the application of rules of the phonological component, namely, what we shall call the principle of the ‘transformational cycle’ ’’ (1968, 15). In Aspects of the Theory of Syntax (1965), Chomsky introduced the cyclic principle into syntactic theory as part of a revised organization of the grammar. In the original Chomskyan framework, outlined in ‘‘The Logical Structure of Linguistic Theory’’ (LSLT; Chomsky 1955), the phrase structure rules produced only simple monoclausal structures, which could then be merged by generalized transformations conjoining two (or more) separate structures, or embedding one into another. In that framework, the generalized transformations constituted the recursive component of the grammar. The Aspects alternative was that the phrase structure rule component itself had a recursive character. Chomsky’s major arguments for this new organization were that it resulted in a simpler overall theory, and at the same time it explained the absence of certain kinds of derivations that seemed not to be needed. The second of these points is our concern here. As just discussed, the LSLT model had generalized transformations to combine a ‘‘forest of trees’’ into a single tree. It also had singulary transformations (Passive, Subject-Auxiliary Inversion, etc.), which operated on single trees, whether basic ones or ones produced by the operation of generalized transformations. Following Fillmore (1963), Chomsky argued that while there is extensive ordering among singulary transformations (situations where a derivation produces an unacceptable sentence if two transformations are applied in reverse order), ‘‘there are no known cases of ordering among generalized embedding transformations although such ordering is permitted by the theory of Transformation-markers’’ (1965, 133). One classic example of strict ordering between singulary transformations centers on passive sentences, which, in the classical model, had active underlying forms. Consider, in this light, the following pair, which arise from the same underlying form:

Conceptions of the Cycle

199

(3) a. The professor praised the students b. The students were praised by the professor The Number Agreement transformation is crucially ordered after the Passive transformation. The reverse order of application would give (4) instead of the desired (3). (4) *The students was praised by the professor Further, Chomsky claimed that while there are many cases of singulary transformations that must apply to a constituent sentence before it is embedded, or that must apply to a ‘‘matrix’’ sentence after another sentence is embedded in it, ‘‘there are no really convincing cases of singulary transformations that must apply to a matrix sentence before a sentence transform is embedded in it’’ (1965, 133). In the Aspects theory, as in the LSLT theory, there was extensive ordering among singulary transformations. In both frameworks, the set of singulary transformations was seen as a linear sequence: an ordered list. Given the Aspects modiﬁcation, this list of rules applied cyclically, operating ﬁrst on the most deeply embedded clause, then on the next most deeply embedded, and so on, working up the tree until the rules applied on the highest clause, the entire generalized phrase marker. Thus, singulary transformations applied to constituent sentences ‘‘before’’ they were embedded, and to matrix sentences ‘‘after’’ embedding had taken place. ‘‘The ordering possibilities that [were] permitted by the theory of Transformation-markers but apparently never put to use [were] now excluded in principle’’ (Chomsky 1965, 135). Curiously, it was almost three decades before it became clear that this was not, as Chomsky presented it, a real argument against generalized transformations. For it was not recursion in the base that excluded the unwanted derivations: it was an implicit version of the cyclic principle. And three decades later, a version of the cyclic principle was to be grafted onto a theory with generalized transformations. The part of the cyclic principle that was actually articulated in Aspects was the Insertion Prohibition. (5) Insertion Prohibition ‘‘[N]o morphological material . . . can be introduced into a conﬁguration dominated by S once the cycle of transformational rules has already completed its application to this conﬁguration (though items can still be extracted from this constituent of a larger ‘matrix structure,’ in the next cycle of transformational rules).’’ (Chomsky 1965, 146)

200

Howard Lasnik

As noted by Chomsky (1973, 243), ‘‘The Insertion Prohibition . . . is a step toward a stricter interpretation of the cycle: it asserts that once a stage of the cycle has been passed, we cannot introduce material into it from the outside.’’ Observe, though, that (5) did not exclude all of the unwanted derivations. Even with this constraint, it was still possible to manipulate the contents of an embedded structure even after its cycle had been completed. There was a further implicit requirement, ﬁnally made explicit in Chomsky 1973, where Chomsky observed that (6) sharpened the notion of the cycle. (6) Strict Cycle Condition ‘‘No rule can apply to a domain dominated by a cyclic node A in such a way as to a¤ect solely a proper subdomain of A dominated by a node B which is also a cyclic node.’’ (Chomsky 1973, 146) Later, Chomsky (1977, 73) asserted that ‘‘transformational rules, e.g., [Move NP and Move wh-phrase], meet the condition of the (strict) cycle.’’ 8.2

Cyclicity and Minimalism

In the course of his minimalist critique of all of syntactic theory, Chomsky (1993) argues that deep structure—the level of representation resulting from recursion in the base—should be eliminated and generalized transformations reinstated. He notes that in Aspects, ‘‘Elimination of generalized transformations in favor of cyclic base generation [sic] is . . . justiﬁed in terms of explanatory adequacy. But the questions under discussion then do not arise in the far more restrictive current theories’’ (p. 46). But it seems that the questions do still arise, and the answer is the same as it was, as noted above. Chomsky continues to assume that we must eliminate the kinds of derivations that concerned him in Aspects. Recursion in the base (and the concomitant level of deep structure) was not the antidote. Rather, the (strict) cycle was, as pointed out by Freidin (1978). The minimalist approach to structure building and cyclicity that Chomsky proposes is as follows. We have ‘‘bottom-up’’ creation of a phrase marker by generalized transformations. There is no deep structure. The derivation is constrained by an ‘‘extension condition.’’ (7) Extension Condition ‘‘GT [Generalized Transformation] and Move a extend K to K*, which includes K as a proper part.’’ (Chomsky 1993, 22) Chomsky observes that ‘‘[(7)] yields a version of the strict cycle’’ (p. 23).

Conceptions of the Cycle

201

Actually, I regard the necessity of the principle of strict cyclicity (in any of its incarnations) as still an open question. Certainly, it is not unreasonable to regard as best the theory that allows the smallest set of derivations (the classic motivation for strict cyclicity). But it is also not unreasonable to regard as best the theory that makes the fewest stipulations. In this instance, the two methodological desiderata appear to be at odds. Subsequent to the formulation of the classic theoretical argument for strict cyclicity, empirical arguments were developed for cyclicity. The ﬁrst type of argument centers on derivations where one transformation must both precede and follow another in order to produce a particular sentence. One standard example from the late 1960s involved Passive in both an embedded and a matrix sentence, as in (8). (8) Mary is believed by John to have been unfairly criticized At the time, it was taken for granted that Passive can only move an ‘‘object’’ into subject position. Thus, in (8), following Passive in the embedded clause that moves Mary into embedded subject position, Subject Raising makes Mary the object of the matrix clause. At this point, it is available for Passive in the matrix. Thus, Passive must be able to both precede and follow Subject Raising. Strict ordering of Subject Raising before Passive combined with cyclic application of rules makes this possible. However, Freidin (1978) shows that this entire class of arguments—potential undergeneration—carries no weight. Assuming free ordering (i.e., eliminating the stipulation of strict linear ordering), all of the grammatical sentences that could have been generated by cyclic derivations can be generated without it. As Freidin discusses, of greater potential import is overgeneration. The standard case involves illicit extraction out of an embedded question, as in (9). (9) *Where j did [you wonder [what i [John put t i t j ]]]? (9) is, of course, immediately excluded by Subjacency, but only if where is prevented from moving ‘‘from Comp to Comp.’’ If what undergoes whmovement ﬁrst, the embedded Comp (Spec,CP in later developments of the theory) is unavailable for the ﬁrst step of the movement of where. But if where moves ﬁrst, the embedded Comp is available. Then, once where moves to the higher Comp, what can move into the now vacated lower one. It is this kind of derivation that strict cyclicity forbids. However, again as noted by Freidin, under plausible assumptions this kind of derivation is independently excluded. Given that movement leaves a trace, the position into which what must move is not empty, being occupied by the trace of where. This su‰ces to prevent the derivation.

202

Howard Lasnik

Chomsky (2000, 136) proposes a new theoretical argument for (the minimalist version of ) strict cyclicity. The idea is that derivations conform to a condition demanding as little tampering as possible with existing structure. As Chomsky puts it, ‘‘operations preserve existing structure.’’ I will use a schematic example to illustrate this point. Consider the phrase structure tree in (10). (10)

Now suppose b is adjoined to X, in conformity with (7), the Extension Condition. The resulting tree is (11), which clearly includes (10) as a subtree, thus preserving the existing tree structure. (11)

Suppose, on the other hand, we were to adjoin b to C, the most ﬂagrant possible violation of the Extension Condition. The result is shown in (12). (12)

This time the original tree is not a subtree of the derived tree. Thus, if it really is a priori desirable to preserve existing structure, we have a strong argument for strict cyclicity as instantiated by the Extension Condition. Interestingly, this strong argument might actually depend on a somewhat arbitrary notational choice. Given the graph-theoretic representation

Conceptions of the Cycle

203

of phrase structure, the argument trivially went through. But Chomsky’s original conception of phrase structure, as in Chomsky 1955 and numerous subsequent works, was set-theoretic. Consider, then, the corresponding derivations on phrase markers construed as sets. The initial phrase marker corresponding to (10) is (13). (13) {X, ZA, ZBC} With b adjoined ‘‘high’’ (i.e., to the node that dominates all others), the derived phrase marker is (14), corresponding to (11). (14) {X, bX, bZA, bZBC} With b adjoined ‘‘low,’’ the derived phrase marker is (15), corresponding to (12). (15) {X, ZA, ZBC, ZBCb} Surprisingly, with the ‘‘low’’ attachment, the original structure is a subset of the derived one, while with the ‘‘high’’ attachment, it is not. More generally, the higher the attachment site, the more radically the set is altered; the lower, the less radically. Thus, surprisingly, the relationship between the Extension Condition and preservation of structure appears to be notation dependent. Again, an argument for strict cyclicity is less conclusive than it originally seemed.3 Chomsky (1995) gives what can be construed as an argument against the Extension Condition, though he doesn’t present it quite that way. Developing his notion of ‘‘feature strength,’’ which forces overt movement, he proposes that strong features are elements that are not tolerated by the derivation and therefore must be eliminated (almost) immediately upon insertion into the structure. More precisely, he deﬁnes strength as follows: (16) Suppose that the derivation D has formed S containing a with a strong feature F. Then, D is canceled if a is in a category not headed by a. Chomsky observes that ‘‘cyclicity follows at once’’ (p. 233). This is because this conception of strength disallows acyclic checking of heads with strong features. Bosˇkovic´ and Lasnik (1999) explore this deduction in detail, particularly emphasizing what Chomsky implies: the fact that the Extension Condition and strength radically overlap in their e¤ects. Typically, if overt insertion of a head a with a strong feature takes place acyclically, then the derivation will immediately cancel, since, by hypothesis, a is contained in a category not headed by a. Similarly, if overt

204

Howard Lasnik

movement takes place acyclically, then the strong feature F of a driving the movement must not have been checked until later structure has been created headed by something other than a. Consider, for example, an acyclic derivation of (9), repeated here as (17). (17) *Where j did [you wonder [what i [John put t i t j ]]]? Suppose Subjacency, at least in wh-island conﬁgurations, reduces to Relativized Minimality. Then there is a potential derivation in which where moves directly to the matrix Spec,CP before what moves into the embedded Spec,CP. Relativized Minimality would not prevent this derivation, given the acceptability of (18). (18) Where did you put what? That is, (18) shows that prior to wh-movement, where can be higher than what. The Extension Condition rules out (17) on this derivation. However, (16) redundantly blocks the derivation, since the strong feature of the embedded interrogative C is not checked until too late, or, more precisely, the derivation is canceled as soon as the complement CP is embedded. As Bosˇkovic´ and Lasnik (1999) note, this redundancy is conceptually problematic. Bosˇkovic´ and Lasnik propose removing the redundancy by eliminating the Extension Condition, since they argue that the condition is problematic in several respects. In particular, head adjunction and LF movement appear to freely violate it. First, consider head adjunction. Suppose X has merged with YP, creating X 0 , as in (19). (19)

Now suppose Y is to raise to adjoin to X, the typical case. The di‰culty is that this adjunction does not extend the tree upward, as (20) shows. (20)

Similarly, if there is covert wh-movement in a module following all of the overt operations, such movement in an embedded question like (21) will fail to extend the tree upward (with what moving to join who in the embedded Spec,CP).

Conceptions of the Cycle

205

(21) Mary wonders who bought what If cyclicity is reduced to feature strength as in (16), then head adjunction and LF movement are no longer exceptional. For example, no additional stipulations are now needed to accommodate overt (strong-feature-driven) V-to-I raising in French, as long as the relevant IP projection is not embedded within another category before V-to-I adjunction takes place. Also, since LF movement is, by deﬁnition, not driven by strong feature checking, it is automatically consistent with the strength conception of the cycle. An interesting property of this conception of the cycle is that, unlike the Extension Condition, it does not in principle ban elements that do not have any strong features from entering the structure acyclically. Bosˇkovic´ and Lasnik (1999) show that permitting such derivations has several positive e¤ects. One instance involves Empty Category Principle/ Subjacency asymmetries. Quite generally, traditional ECP violations with extraction of adjuncts go hand in hand with Subjacency violations with extraction of arguments. More precisely, in the contexts in which extraction of adjuncts leads to an ECP violation, extraction of arguments leads to a Subjacency violation. This is illustrated in (22) with an adjunct island. (22) a. ??Whoi did Mary leave for London [after Peter had visited t i ]? b. *Whyi did Mary leave for London [after Peter had visited her t i ]? Chomsky and Lasnik (1993) develop a system in which traditional ECP violations with extraction of adjuncts and Subjacency violations with extraction of arguments reduce to the same economy condition, the only distinction between the two being that with argument Subjacency violations the o¤ending trace is deleted in LF, whereas with adjunct ECP violations it remains present in the ﬁnal LF representation. In this system, we would not expect to ﬁnd a conﬁguration in which extraction of adjuncts would lead to an ECP-type violation, but extraction of arguments would not lead to a Subjacency-type violation. It is well known, however, that such conﬁgurations exist. For example, as discussed by Rizzi (1990), pseudo-opacity e¤ects and inner island e¤ects obtain with adjunct extraction, but not with argument extraction. (23) illustrates this for pseudo-opacity. (23) a.

[Combien de livres]i a-t-il beaucoup consulte´s t i ? how-many of books has-he a-lot consulted ‘How many books did he consult a lot of?’

206

Howard Lasnik

b. *Combieni a-t-il beaucoup consulte´s [t i de livres]? how-many has-he a-lot consulted of books ‘How many did he a lot consult of books?’ c.

Compare: Combieni a-t-il consulte´s [t i de livres]? how-many has-he consulted of books ‘How many books did he consult?’

The strong feature approach to cyclicity can explain this surprising asymmetry between the ECP and Subjacency, provided that we adopt Lasnik and Saito’s (1984, 1992) theory that adjunct traces are checked with respect to locality restrictions only in LF, whereas argument traces can be checked in overt syntax.4 Assuming that beaucoup does not have any strong features to check and is not required to be present in the structure to check strong features of another element, it could enter the structure acyclically. However, since beaucoup is phonologically realized, we know that it must have entered the structure in overt syntax. Given Lasnik and Saito’s proposal, beaucoup then does not have to be present in the structure when argument chains are checked with respect to locality restrictions, but it does have to be present when adjunct traces are checked. The surprising asymmetry between the ECP and Subjacency exhibited by pseudo-opacity is thus captured in a relatively straightforward way. 8.3

Successive Cyclicity and A-Movement

Chomsky (1973) introduced yet another, and independent, aspect of cyclicity: successive cyclicity. Successive cyclicity, following from Subjacency (and perhaps other locality constraints as well), demands that apparent long movement is actually the result of a successive series of short ¯ -movement (as in movements. There is a massive literature on this for A interrogatives or relative clauses), and a wealth of evidence has been uncovered conﬁrming the theoretical prediction of Subjacency. Torrego (1983, 1984) and McCloskey (1990), among many others, have shown ¯ -movement how the successive short steps composing a long instance of A are overtly marked, for example in Spanish and Irish. In both instances, the marking shows up in or near the complementizer position (via obligatory fronting of a ﬁnite verb or by a change in the form of the complementizer), signiﬁcant in light of Chomsky’s Subjacency proposal that long ¯ -movement proceeds via that position. A-movement has been much less A

Conceptions of the Cycle

207

discussed with respect to successive cyclicity. In the remainder of this chapter, I will consider potential arguments for and against successivecyclic A-movement, ultimately concluding that A-movement does indeed proceed in this way. Chomsky (1977, 74) presents the following example, noting that ‘‘if the rule of NP-movement that yields [(24)] applies successive cyclically . . . then the rule will observe subjacency.’’ (24) John seems [to be certain [to win]] In Lectures on Government and Binding (LGB; 1981, 44), he discusses a similar example, (25), with the representation in (26). (25) John was believed to have been killed (26) John INFL be believed [S t 0 INFL have been killed t] This time, though, he o¤ers a di¤erent theoretical motivation for the assumed successive cyclicity of the movement, saying, ‘‘In the case of sentence [(26)], we were led by the projection principle to assume that the rule Move-a applies twice, leaving the two traces t and t 0 , successively.’’ This is evidently incorrect. While the movement might be successive (in fact, I will argue that it is), it cannot be the Projection Principle that is forcing this, since that principle demands only that y-positions be present at all syntactic levels of representation; but crucially in the LGB framework, positions to which, or through which, movement takes place are never y-positions. Instead, Subjacency might be the relevant principle, as proposed in Chomsky 1977. Or perhaps not. Certainly in the 1970s theory of Subjacency, with S (¼ IP) as a bounding node, successive movement is forced. But in a version of the Barriers framework (Chomsky 1986a), IP is never an ‘‘inherent’’ barrier, and in the typical structure of A-movement such as (24), there might not be anything for the embedded IPs to ‘‘inherit’’ barrierhood from. With that in mind, it is worth considering whether A-movement is in fact successive cyclic, and, if it is, what forces this. Chomsky (1981, 44– 45) presents an empirical argument for intermediate A-traces, hence, for successive-cyclic A-movement. He observes that (27) is acceptable, indicating that Condition A is satisﬁed. (27) They are likely [t 0 to appear to each other [t to be happy]] On this basis, he argues for intermediate traces (hence, for successivecyclic movement). His line of reasoning is as follows: ‘‘The GF-y [nonthematic grammatical function] ﬁlled by medial traces such as t 0 in [(26)]

208

Howard Lasnik

may . . . be relevant to LF; for example in the sentence [(27)], . . . where the medial trace serves as the antecedent of each other, which requires an antecedent in the same clause in such cases in accordance with binding theory’’ (1981, 44–45). Interestingly, though, it actually isn’t completely clear that the antecedent must be in the same clause, given the binding theory developed in Chomsky 1981 (or those developed in Chomsky 1973 and Chomsky 1986b, for that matter). Consider the formulation in Chomsky 1981. (28) ‘‘b is a governing category for a if and only if b is the minimal category containing a, a governor of a, and a SUBJECT accessible to a.’’ (p. 211) (29) SUBJECT ¼ AGR in a ﬁnite clause; NP of S in an inﬁnitival; . . . (30) ‘‘g is accessible to a if and only if a is in the c-command domain of g and . . . ’’ (p. 212) If there is a trace in the intermediate clause, then that clause is the governing category (GC) of each other. But that is no argument that there is such a trace. Suppose there were none. Then the intermediate clause, lacking a SUBJECT, would not be the GC, and Condition A would correctly be satisﬁed with they as the binder of each other. The result is the same under the formulation in Chomsky 1986b. Here, I will give only the portion of the theory relevant to Condition A: (31) ‘‘I is BT-compatible with (a; b) if: (A) a is an anaphor and is bound in b under I.’’ (p. 171) (32) Licensing condition for a category a governed by a lexical category g in the expression E with indexing I: ‘‘For some b such that [(i)], I is BT-compatible with (a; b): [(i)] a is an anaphor . . . and b is the least CFC [complete functional complex] containing g for which there is an indexing J BT-compatible with (a; b).’’ (pp. 170–171) As Chomsky puts it, ‘‘For . . . an anaphor . . . the licensing condition amounts to saying that the relevant governing category for a is the minimal one in which the binding theory could have been satisﬁed by some indexing’’ (p. 172). Once again, if there is an intermediate trace, each other must be (and is) bound in that clause. But in the absence of that trace, the licensing condition for the anaphor could not, hence need not, be satisﬁed in the intermediate clause. Therefore, matrix binding is allowed. Once again, then, there is no argument for successive cyclicity.

Conceptions of the Cycle

209

I will return to this issue, but ﬁrst, I will consider an argument that Amovement is not successive cyclic. Epstein and Seely (1999) note that successive-cyclic A-movement should create an A-chain, by a chain formation mechanism. According to Chomsky (1995), a chain is a set of ‘‘occurrences,’’ where each occurrence is deﬁned in terms of sisterhood. The occurrences at issue here are then intermediate projections of I, I 0 s, since we are considering movement through Spec,IPs. But it is widely assumed that syntactic operations can’t target intermediate projections; thus, it is reasonable to assume that the chain formation mechanism can’t either. Therefore, the needed chain links can’t exist, so movement must take place in one fell swoop.5 On the other hand, even if intermediate projections are not available for any purposes, thus possibly rendering Chomsky’s deﬁnition of ‘‘occurrence’’ unstatable, an alternative deﬁnition in terms of ‘‘motherhood’’ instead of ‘‘sisterhood’’ seems equally capable of characterizing what Chomsky set out to characterize. And ‘‘motherhood’’ raises no issues at all about intermediate projections. Further, even if chains as representational objects are ultimately eliminated (for Epstein and Seely’s reasons or others), the successive movement itself is still perfectly statable in derivational terms. That is, at the derivational point where a head—say, I—is merged into the structure, the resulting conﬁguration is, by deﬁnition, a maximal projection. It should also be recalled that there is overwhelming ¯ -movement is successive cyclic. But it is evidence, alluded to above, that A very hard to see how we could accept Epstein and Seely’s argument for ¯ -movement, since type of movement is irA-movement but reject it for A relevant to their line of reasoning. With this argument against successive-cyclic A-movement called into question, I return to Chomsky’s Condition A argument for successive cyclicity. Above, I showed that Chomsky’s argument does not go through on standard formulations of the binding domain for anaphors. Chomsky’s argument relied on a clause-mate condition, but those formulations do not give that result in the crucial cases. However, in Lasnik 2002, 2003, I argue that clause-mate is, in fact, the correct characterization, as argued much earlier by Postal (1966, 1974). My argument is initially based on a very interesting verb-particle construction ﬁrst discussed by Kayne (1985) and later analyzed by Johnson (1991) in terms relevant to the present discussion. Johnson provides an insightful account of examples like (33) involving overt raising of the ECM (exceptional casemarking) subject John.

210

Howard Lasnik

(33) Mary made John out to be a fool Both Kayne and Johnson convincingly treat (33) as an inﬁnitival counterpart of (34). (34) Mary made out that John is a fool Very plausibly, John in (33) has overtly raised into the matrix clause. Now consider that for many speakers, the raising seen in (33) is optional. For those speakers, (35) is an acceptable alternative to (33). (35) Mary made out John to be a fool This time, plausibly John has remained in the embedded clause, and in Lasnik 2002, 2003, I provide evidence that this is so. Now note that an anaphor on the upstream side of out is much better than one on the downstream side, with antecedent in the matrix clause. (36) a. Jack made himself out to be honest b. ?*Jack made out himself to be honest (37) a. They made each other out to be dishonest b. ?*They made out each other to be dishonest None of the standard formulations of the notion of governing category predict the disparity between the (a) and (b) versions of these examples, but, under the stated assumptions, a clause-mate formulation could. In each of the (b) examples, the anaphor is separated from its antecedent by the boundary of the inﬁnitival clause, while in the (a) examples, no clause boundary of any sort intervenes. It must be acknowledged that there is an interfering factor: verb-particle-NP order with reﬂexives and reciprocals is somewhat degraded even with only one clause, as in (38)–(39). (38) ?Jack called up himself (39) ?They called up each other However, the deviance is less severe than that seen in ECM instances like (36)–(37). Thus, the argument for the clause-mate condition stands. Given the clause-mate requirement on anaphors, examples like the following, pointed out to me in this connection by Adolfo Ausı´n (also attributed to Danny Fox, via David Pesetsky, in Castillo, Drury, and Grohmann 1999), argue for successive-cyclic A-movement: (40) John appears to Mary [to seem to himself/*herself [to be the best candidate]]

Conceptions of the Cycle

211

In the absence of successive-cyclic movement, himself would incorrectly be predicted to violate Condition A, and, on fairly standard assumptions, herself to be in conformity. A similar though somewhat less direct argument can be constructed on the basis of Condition B. Examples (41)–(42) display standard Condition B e¤ects. (41) *Johni injured himi (42) *Johni believes himi to be a genius These contrast with (43)–(44), where the antecedent is more distant in relevant respects. (43) Johni believes Mary to admire himi (44) Johni believes that hei is a genius As would be expected, (41) remains unacceptable under VP-ellipsis. (45) *Mary injured himi and Johni did too What is not expected is that (42) is substantially improved under VPellipsis. (46) ?Mary believes himi to be a genius and Johni does too How is it that deletion, which I take to be a PF process, remediates a presumably semantic violation? Suppose that in addition to their properties with respect to anaphora, the pronouns in question also have to satisfy a morphosyntactic requirement: as weak pronouns, they must cliticize onto the verb, as suggested by Oehrle (1976), who cites data like these from Chomsky 1955: (47) The detective brought John in (48) The detective brought in John (49) The detective brought him in (50) *The detective brought in him If the relevant structural conﬁguration for Condition B is based on the notion of clause-mate, an account of the ellipsis paradigm presents itself. In particular, one might hypothesize that it is not actually the Condition B violation (presumably an LF e¤ect) in (42) that is repaired by ellipsis in (46). Rather, it is failure to cliticize (a PF violation) that is repaired (by PF deletion). And without cliticization, the pronoun can remain in the lower clause (if, as I argue in Lasnik 1999, 2001, ‘‘subject raising’’ in these

212

Howard Lasnik

constructions is generally optional). On the other hand, in (45), the pronoun and its antecedent are clause-mates independent of cliticization, so there is no possibility of ‘‘repair’’ of the Condition B violation by ellipsis.6 Given the clause-mate character of Condition B, (51) constitutes evidence for successive-cyclic movement, as John must have moved through the intermediate Spec,IP to yield the observed obviation. John in its surface position is not a clause-mate of him. (51) *Johni is believed [to seem to himi [to be a genius]] This raises the question of what forces the movement to be successive cyclic. It is possible that it is a locality constraint on movement, as argued most recently by Bosˇkovic´ (2002). But it is not so clear that the most ‘‘minimal’’ theories of locality guarantee this. At the moment, it seems to me that the likeliest culprit is the Extended Projection Principle, as I argue in Lasnik 2003. 8.4

Cyclic Output

The ﬁnal issue I will consider, if only brieﬂy, is the ‘‘timing’’ of the relation between syntax on the one hand and semantics and phonology on the other. In the Government-Binding ‘‘T-model,’’ there is one speciﬁc syntactic level, S-Structure, which connects to the LF and PF interface modules. In the LSLT model, on the other hand, the interface is the Tmarker, which includes all of the syntactic structures created in the course of the derivation. Given that the minimalist approach to structure building is much more similar to that in the LSLT model than to that in any of the intervening models, the possibility presents itself that interpretation in the minimalist model also could be more like that in the LSLT model, distributed over many structures. Interestingly, already in the late 1960s and early 1970s, arguments were occasionally made for such a model of interpretation, for phonological interpretation as well as semantic interpretation. For example, Bresnan (1971) argued that the phonological rule responsible for assigning English sentences their intonation contour applies cyclically, following each cycle of transformations, rather than at the end of the entire syntactic derivation. Similar proposals for semantic phenomena involving anaphora were put forward by Jackendo¤ (1969). Jackendo¤ showed that given standard assumptions about derived constituent structure, the rule assigning a reﬂexive its antecedent must apply before certain transformations. He showed that the right result is obtained if the reﬂexive rule operates at the end of each syntactic cycle. In

Conceptions of the Cycle

213

Lasnik 1972, I suggested, on largely conceptual grounds, that the scope of negation is also determined cyclically (i.e., at the end of each syntactic cycle). And in Lasnik 1976, I argued that strong crossover can be accounted for if the scope rule determining bound variable anaphora is cyclic in the same sense. In recent work, Chomsky (2000, 2001) suggests a more general instantiation of this distributed approach to phonological and semantic interpretation, based on proposals by Epstein (1999) and Uriagereka (1999). At the end of each cycle (or ‘‘phase’’ in Chomsky 2000, 2001), the syntactic structure thus far created is encapsulated and sent o¤ to the interface components for all phonological and semantic interpretation. Thus, while there are still what might be called PF and LF components, there are no levels of PF and LF. There is just one cyclic derivation, which, in e¤ect, encodes what the much earlier T-markers did. 8.5

Conclusion

Needless to say, this discussion only scratches the surface of a huge mountain of a topic. If there is any conclusion to be drawn, it is one of general applicability, though perhaps especially so for the present topic: What goes around comes around. Notes 1. I recommend Freidin 1999 for an illuminating perspective on the history and current status of cyclicity. 2. Here I oversimplify, as eraser also has internal cyclic structure. 3. See Freidin 1999 for an intriguing alternative derivation of cyclicity, based on core properties of the elementary operations available to transformations. 4. See Lasnik and Saito 1984, 1992 for motivation for this proposal, and for a potential way to deduce this di¤erence between adjuncts and arguments from independent mechanisms of the grammar. 5. At this point, the question arises how even one-fell-swoop movement will be allowed. Epstein and Seely (1999) argue that in this instance, unlike the successive one, no chain formation algorithm is necessary. I am indebted to Tomohiro Fujii for very helpful discussion of this point. 6. Notice that coordinated pronouns, which are known not to behave like clitics, also behave di¤erently with respect to Condition B in ECM conﬁgurations. As Lisa Cheng and Norbert Corver (pers. comm.) observe, (i) is much better than Chomsky’s original example. (i) The detective brought in [him and Peter] Correspondingly, (ii), with its coordinated him, seems considerably better than (42).

214

Howard Lasnik

(ii) Johni believes himi and Mary to be geniuses Similarly, as Cheng and Corver also observe, a focus marker on the pronoun allows it to remain uncliticized. (iii) The detective brought in only him Thus, as expected, the focus marker on the ECM subject also considerably remediates the Condition B e¤ect. (iv) Johni believes only himi to be a genius References Bosˇkovic´, Zˇeljko. 2002. A-movement and the EPP. Syntax 5, 167–218. Bosˇkovic´, Zˇeljko, and Howard Lasnik. 1999. How strict is the cycle? Linguistic Inquiry 30, 691–703. Bresnan, Joan W. 1971. Sentence stress and syntactic transformations. Language 47, 257–281. Castillo, Juan Carlos, John Drury, and Kleanthes K. Grohmann. 1999. Merge over Move and the Extended Projection Principle. In Sachiko Aoshima, John Drury, and Tuomo Neuvonen, eds., Papers in linguistics, 66–104. University of Maryland Working Papers in Linguistics 8. College Park: University of Maryland, Department of Linguistics. Chomsky, Noam. 1955. The logical structure of linguistic theory. Ms., Harvard University and MIT. Revised 1956 version published in part by Plenum Press, New York, 1975; University of Chicago Press, Chicago, 1985. Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1973. Conditions on transformations. In Stephen R. Anderson and Paul Kiparsky, eds., A festschrift for Morris Halle, 232–286. New York: Holt, Rinehart and Winston. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. Chomsky, Noam. 1986a. Barriers. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1986b. Knowledge of language. New York: Praeger. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In Kenneth Hale and Samuel Jay Keyser, eds., The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, 1–52. Cambridge, Mass.: MIT Press. Reprinted in Noam Chomsky, The Minimalist Program, 167–217. Cambridge, Mass.: MIT Press, 1995. Chomsky, Noam. 1995. Categories and transformations. In The Minimalist Program, 219–394. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press.

Conceptions of the Cycle

215

Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam, and Morris Halle. 1968. The sound pattern of English. New York: Harper and Row. Chomsky, Noam, Morris Halle, and Fred Luko¤. 1956. On accent and juncture in English. In Morris Halle, Horace Lunt, and Hugh MacLean, eds., For Roman Jakobson, 65–80. The Hague: Mouton. Chomsky, Noam, and Howard Lasnik. 1993. The theory of principles and parameters. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, and Theo Vennemann, eds., Syntax: An international handbook of contemporary research, vol. 1, 506–569. Berlin: Walter de Gruyter. Epstein, Samuel David. 1999. Un-principled syntax: The derivation of syntactic relations. In Samuel David Epstein and Norbert Hornstein, eds., Working minimalism, 317–345. Cambridge, Mass.: MIT Press. Epstein, Samuel David, and T. Daniel Seely. 1999. SPEC-ifying the GF ‘‘subject’’: Eliminating A-chains and the EPP within a derivational model. Ms., University of Michigan and Eastern Michigan State University. Fillmore, Charles J. 1963. The position of embedding transformations in a grammar. Word 19, 208–231. Freidin, Robert. 1978. Cyclicity and the theory of grammar. Linguistic Inquiry 9, 519–549. Freidin, Robert. 1999. Cyclicity and minimalism. In Samuel David Epstein and Norbert Hornstein, eds., Working minimalism, 95–126. Cambridge, Mass.: MIT Press. Jackendo¤, Ray. 1969. Some rules of semantic interpretation for English. Doctoral dissertation, MIT. Johnson, Kyle. 1991. Object positions. Natural Language and Linguistic Theory 9, 577–636. Kayne, Richard. 1985. Principles of particle constructions. In Jacqueline Gue´ron, Hans-Georg Obenauer, and Jean-Yves Pollock, eds., Grammatical representation, 101–140. Dordrecht: Foris. Lasnik, Howard. 1972. Analyses of negation in English. Doctoral dissertation, MIT. Lasnik, Howard. 1976. Remarks on coreference. Linguistic Analysis 2, 1–22. Reprinted in Howard Lasnik, Essays on anaphora, 90–109. Dordrecht: Kluwer, 1989. Lasnik, Howard. 1999. Chains of arguments. In Samuel David Epstein and Norbert Hornstein, eds., Working minimalism, 189–215. Cambridge, Mass.: MIT Press. Lasnik, Howard. 2001. Subjects, objects, and the EPP. In William D. Davies and Stanley Dubinsky, eds., Objects and other subjects: Grammatical functions, functional categories, and conﬁgurationality, 103–121. Dordrecht: Kluwer.

216

Howard Lasnik

Lasnik, Howard. 2002. Clause-mate conditions revisited. Glot International 6, 94–96. Lasnik, Howard. 2003. On the Extended Projection Principle. Studies in Modern Grammar 31, 1–23. Lasnik, Howard, and Mamoru Saito. 1984. On the nature of proper government. Linguistic Inquiry 15, 235–289. Reprinted in Howard Lasnik, Essays on restrictiveness and learnability, 198–255. Dordrecht: Kluwer, 1990. Lasnik, Howard, and Mamoru Saito. 1992. Move a. Cambridge, Mass.: MIT Press. McCloskey, James. 1990. Resumptive pronouns, A 0 -binding, and levels of representation in Irish. In Randall Hendrick, ed., Syntax and semantics 23: The syntax of the modern Celtic languages, 199–248. New York: Academic Press. Oehrle, Richard. 1976. The grammatical status of the English dative alternation. Doctoral dissertation, MIT. Postal, Paul M. 1966. A note on ‘understood transitively’. International Journal of American Linguistics 32, 90–93. Postal, Paul M. 1974. On raising: One rule of English grammar and its theoretical implications. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Torrego, Esther. 1983. More e¤ects of successive cyclic movement. Linguistic Inquiry 14, 561–565. Torrego, Esther. 1984. On inversion in Spanish and some of its e¤ects. Linguistic Inquiry 15, 103–129. Uriagereka, Juan. 1999. Multiple Spell-Out. In Samuel David Epstein and Norbert Hornstein, eds., Working minimalism, 251–282. Cambridge, Mass.: MIT Press.

Chapter 9 On the Relaxation of Intervention E¤ects

9.1

Toru Ishii

Introduction

On the basis of insights captured in Rizzi’s (1990) Relativized Minimality, Chomsky (2000, 2001, 2004) proposes the Defective Intervention Constraint in (1) (adapted from Chomsky 2000, 123). (1) *a > b > g, where (a) ‘‘>’’ indicates c-command, and (b) b and g match the probe a, but b is inactive so that the e¤ects of matching are blocked. This constraint is appropriate as a ﬁrst approximation, but certain anomalies remain unaccounted for (see, e.g., Rizzi 2004). In this chapter, I will focus on the Wh-Island Constraint imposed on wh-movement. The Wh-Island Constraint, exempliﬁed by (2a–b), has been widely assumed to be a defective intervention e¤ect of type (1). (2) a. *Which of the new booksi do you wonder [when j John bought ti t j ]? b. *Which pastai do you wonder [howj the famous Italian chef cooked ti t j ]? Let us consider (2a). During its derivation, we come to the stage where the Q-feature of the matrix C, being a probe, seeks a goal. (3) [C [Q] [you wonder [when [Q; wh ]j [C [ Q ] [John bought which [Q; wh] of the new books t j ]]]]] In (3), the wh-phrase when raises to the speciﬁer position (Spec) of the embedded C for checking purposes. The Q-feature of the embedded C and the wh-feature of when, both of which are uninterpretable, are deleted (valued). Given that uninterpretable features render the goal active, when, whose uninterpretable wh-feature has been deleted (valued), is not active

218

Toru Ishii

as a goal. When the Q-feature of the matrix C seeks a goal, when blocks movement of which of the new books to the Spec of the matrix C as a result of the Defective Intervention Constraint (1). This is because when, which is not active but visible, c-commands which and thus functions as an intervener.1 I will show, however, that the wh-island e¤ects are relaxed when the following two conditions are met: (1) indirect questions are inﬁnitive or subjunctive, and (2) extracted wh-phrases are D(iscourse)-linked. I will argue that this hitherto unexplained anomaly concerning the Wh-Island Constraint is given a principled minimalist account by an operation of covert restructuring and the Phase Impenetrability Condition (PIC) proposed by Chomsky (2000, 2001, 2004). I will show that contrary to the widespread view that the Wh-Island Constraint can be subsumed under the Defective Intervention Constraint (1), we still need a traditional Subjacency-type account like that advocated by Chomsky (1973, 1977, 1986) to fully accommodate the wh-island e¤ects. The chapter is organized as follows. In section 9.2, I argue that inﬁnitive/subjunctive complements undergo covert restructuring. Speciﬁcally, I argue that an Aspect Phrase (AspP) within an inﬁnitive/subjunctive complement undergoes covert phrasal movement to the Spec of the embedding v for reasons of temporal interpretation. I then show that the covert restructuring analysis accounts not only for the relaxation of the defective intervention e¤ects with inﬁnitive/subjunctive indirect questions but also for transparency (clause-mate) e¤ects with inﬁnitive/subjunctive complements. In section 9.3, I ﬁrst show that the relaxation of the defective intervention e¤ects can only be observed with D-linked wh-phrases. I argue that the selective extractability of a wh-phrase out of an inﬁnitive/subjunctive indirect question should be given a traditional Subjacency-type account based on the notion of the phase. In section 9.4, I argue that this analysis also accounts for the ‘‘double wh-island’’ e¤ects found with inﬁnitive/subjunctive indirect questions. I also show that relativization can be dealt with in the same way as D-linked wh-phrases. In section 9.5, I o¤er concluding remarks. 9.2

Covert Restructuring

If the Defective Intervention Constraint (1) is operative in grammar, we must account for why the intervention e¤ects are relaxed when extraction takes place out of an inﬁnitive/subjunctive indirect question (see, e.g., Chomsky 1986; Cinque 1990; Boeckx 2001).

On the Relaxation of Intervention E¤ects

219

(4) a. Which of the new booksi do you wonder [when j to buy ti t j ]? b. Which pastai do you wonder [how to cook ti t j ]? (5) a. Which of the new booksi do you wonder [when j you should buy ti t j ]? b. Which pastai do you wonder [how you should cook ti t j ]? In (4) and (5), even though the wh-phrase is extracted out of the indirect question, the result is relatively acceptable. In this section, I will propose that inﬁnitive/subjunctive complements undergo covert restructuring, arguing that it accounts for the relaxation of the intervention e¤ects. First, though, I will explicate temporal interpretations of inﬁnitive/subjunctive complements, as these provide the basis for the covert restructuring account. 9.2.1 Temporal Interpretations of Inﬁnitive/Subjunctive Complements It is well known that inﬁnitive/subjunctive complements always depend on higher clauses for tense interpretations (see, e.g., Picallo 1984; Kempchinsky 1986; Manzini and Wexler 1987; Zagona 1988; Hornstein 1990; Thompson 1996). I claim with, among others, Zagona (1988), Hornstein (1990), Stowell (1995, 1996), Thompson (1996), and Giorgi and Pianesi (1997), that tenses are represented syntactically. Among these syntactic approaches to tenses, I adopt the neo-Reichenbachian approach advocated by Hornstein (1990) and Thompson (1996). I will argue that temporal interpretations of inﬁnitive/subjunctive complements can be captured by covert restructuring under the neo-Reichenbachian theory of tense. 9.2.1.1 A Neo-Reichenbachian Theory of Tense The neo-Reichenbachian approach has extended Reichenbach’s (1947) system, arguing that tenses are composed by linearly ordering three times: the Event time (E), the Reference time (R), and the Speech time (S). As an illustration, let us consider (6). (6) John had left the restaurant In (6), the Event time is the time of John’s leaving the restaurant, the Reference time is the time by which John leaves, and the Speech time is the time at which this sentence is uttered. The basic tense structure of the past perfect (6) is represented by (7). (7) E_R_S When two points are separated by an underscore, the one to the left is interpreted as temporally preceding the one to the right. In (7), E is

220

Toru Ishii

located before R, which is itself located before S. This is in accord with the natural interpretation of (6). The neo-Reichenbachian approach also enables us to account for the properties of complex tense conﬁgurations. Let us consider adverbial modiﬁcation, taking (8a–c) as examples. (8) a. John is leaving at this very moment/right now b. John is leaving tomorrow c. *John is leaving yesterday (Hornstein 1990, 16) Hornstein (1990) argues that an adverb can combine with a basic tense structure (BTS) to form a derived tense structure (DTS). The temporal structures of (8a–c) are given in (9). (9) a. BTS now S,R,E !

DTS S,R,E

b. BTS tomorrow S,R,E !

DTS S_R,E

now

tomorrow c. BTS yesterday S,R,E !

DTS E,R_S

yesterday (Hornstein 1990, 19) In (9), the points separated by a comma, which are called ‘‘associated’’ points, are interpreted as contemporaneous. Hornstein assumes the following constraint on the derivation of a derived tense structure: (10) Constraint on DTS (CDTS) DTS must preserve BTS. (Hornstein 1990, 15) The notion of preservation is deﬁned as follows: (11) BTSs are preserved i¤ a. no points are associated in DTS that are not associated in BTS; b. the linear order of points in DTS is the same as that in BTS. (Hornstein 1990, 15)

On the Relaxation of Intervention E¤ects

221

In (9a), the adverb now modiﬁes E or R without altering the basic tense structure; there is no violation of the CDTS (10). In (9b), although the adverb tomorrow, which modiﬁes E or R, alters the basic tense structure, dissociating R and E from S, there is no violation of the CDTS (10). In (9b), there are no SER points that are associated in the derived tense structure but not in the basic tense structure. Furthermore, the order of the SER points in the derived tense structure is the same as their order in the basic tense structure. Hence, the basic tense structure is preserved; (9b) satisﬁes the CDTS (10). In (9c), on the other hand, the order of the SER points in the derived tense structure is not the same as their order in the basic tense structure. Modiﬁcation by yesterday alters the order of the SER points; this violates the second part of the notion of preservation (11) relevant to the CDTS (10). Hence, the neo-Reichenbachian approach can account for the contrast in acceptability between (8a–b) and (8c). 9.2.1.2 Sequence-of-Tense Phenomena Hornstein (1990) argues that sequence-of-tense phenomena, which are illustrated in (12), can also be accommodated under the neo-Reichenbachian approach. (12) John heard that Mary was pregnant (Hornstein 1990, 120) (12) is ambiguous. On one reading, (12) is true if at some past time John heard that Mary was pregnant at a prior time. On the other reading, the time of being pregnant is the same as the time of hearing. The latter reading, where the temporal interpretation of the embedded Event time is dependent on the matrix Event time, is taken as an instance of the sequence-of-tense phenomenon. Hornstein claims that when a ﬁnite clause is embedded under a past time clause as in (12), the temporal dependency on the matrix Event time is signaled by a change in the surface form of the embedded verb. In (12) with the sequence-of-tense interpretation, the underlying tense of the embedded verb is actually the simple present, and the surface past tense form is just the morphological manifestation of the temporal dependency on the matrix Event time. Under Hornstein’s analysis, the temporal structure of (12) with the sequence-of-tense interpretation is represented by (13), where the sequence-of-tense (SOT) rule applies to the basic tense structure, yielding the derived tense structure. (13)

BTS E1 ,R1 _S1 S2 ,R2 ,E2

SOT

!

DTS E1 ,R1 _S1 S2 ,R2 ,E2

222

Toru Ishii

In (13), the matrix S point (S1 ) is interpreted as denoting the utterance time. The sequence-of-tense rule links the embedded S point (S2 ) with the matrix E point (E1 ); S2 is interpreted as contemporaneous with E1 . Since the embedded E point (E2 ) is associated with S2 , E2 is also interpreted as contemporaneous with E1 . This yields the correct interpretation; the time of being pregnant is the same as the time of hearing. Hornstein argues that the sequence-of-tense rule is optional with ﬁnite clauses. The non-sequence-of-tense analogue of (12) is given in (14). (14) John heard that Mary is pregnant Here, the surface present tense form is signals that the sequence-of-tense rule is not applied. Its tense structure is represented by the left side of the arrow in (13). Since the embedded S point (S2 ) is not associated with the matrix E point (E1 ) or anything else, the default rule requires S2 to be interpreted as denoting the utterance time. This yields the correct interpretation for (14), namely, Mary’s pregnancy is contemporaneous with the moment of utterance. Hornstein (1990) argues that while the sequence-of-tense rule is optional with ﬁnite clauses, it is obligatory with inﬁnitive clauses. This captures the fact that the temporal interpretation of an embedded inﬁnitive clause always depends on that of the embedding clause. Let us ﬁrst consider (15a–c) (cf. Baker 1989, 442). (15) a. John seems to be in the correct room b. John appears to be a rich man c. John believes Bill to be the smartest Under Hornstein’s analysis, the temporal structure of (15a–c) is represented by (16). (16)

BTS S,R1 ,E1 R2 ,E2

SOT

!

DTS S,R1 ,E1 R2 ,E2

In (16), the sequence-of-tense rule applies to the basic tense structure, yielding the derived tense structure. Note that in (16), the S point is lacking in the temporal structure of the embedded inﬁnitive clause. Hornstein assumes that the S point is tied to a tense morpheme. In an inﬁnitive clause, where the tense morpheme is lacking, the S point is absent as well. When the S point is absent in the embedded clause, the sequenceof-tense rule associates the embedded R point (R2 ) with the matrix E point (E1 ). This is the case in (16), and R2 is interpreted as contempora-

On the Relaxation of Intervention E¤ects

223

neous with E1 . Since the embedded E point (E2 ) is associated with R2 , E2 is also interpreted as contemporaneous with E1 . In (15a), for example, this yields the correct interpretation that the time of being in the correct room is contemporaneous with the matrix Event time. As pointed out by Baker (1989) and Enc¸ (1996), among others, there are cases where a di¤erent temporal interpretation is assigned to an inﬁnitive complement. (17) a. John hopes to be in the correct room (Baker 1989, 442) b. I expect John to win the race (Stowell 1982, 566) c. I promise to do it d. John wants to buy a new computer I argue that the temporal structure of (17) is represented by (18). (18)

BTS S,R1 ,E1 R2 ,E2

SOT

!

DTS S,R1 ,E1 R2 _E2

In (18), the sequence-of-tense rule applies to the basic tense structure, yielding the derived tense structure, where the matrix E point (E1 ) is associated with the embedded R point (R2 ). As shown in (9), adverbial modiﬁcation alters the basic tense structure in accordance with the meaning of an adverb. I argue that in the same way, the lexical meanings of verbs like hope, promise, want, and expect may alter the basic tense structure of their inﬁnitive complement. These verbs assign a later time to their inﬁnitive complement. They dissociate the embedded E point (E2 ) from the embedded R point (R2 ), shifting the former to the right of the latter. In (17a), for example, this yields the correct interpretation that the time of being in the correct room is in the future relative to the matrix Event time, that is, the time of hoping. Note that this dissociation of the embedded E and R points in the derived tense structure does not violate the CDTS (10). Baker (1989) also observes that adjectives like certain assign either the same or a later time to their inﬁnitive complement. (19) John is certain to be at home (Baker 1989, 443) The temporal structure of (19) is either (16) or (18). This analysis correctly captures the fact that the tense interpretation of an inﬁnitive complement is determined by the lexical meaning of the matrix predicate.2

224

Toru Ishii

I also argue that exactly like inﬁnitive complements, subjunctive complements also obligatorily undergo the sequence-of-tense rule. As examples, consider (20a–b). (20) a. I demand that he (should) go there b. His teacher requires that he (should) go to bed early I argue that exactly like the matrix verbs in (17), verbs selecting subjunctive complements like demand, request, and insist assign a later time to their complement, dissociating the embedded E point (E2 ) from the embedded R point (R2 ) and shifting the former to the right of the latter. The temporal structure of (20a–b) is also represented by (18). In (20a), for example, this correctly yields the interpretation that the time of going there is in the future relative to the matrix Event time, that is, the time of demanding. 9.2.1.3 A Syntactic Representation of the Sequence-of-Tense Phenomena Having shown how the sequence-of-tense phenomena can be accommodated under the neo-Reichenbachian approach, I now consider how they should be represented syntactically. Essentially following Hornstein (1990) and Thompson (1996), I assume that the three times (i.e., the Event, Reference, and Speech times) are syntactically represented as follows: the Event time is represented by V; the Reference time by Aspect (Asp), which is the head of the Aspect Phrase (AspP) that resides between TP and vP; and the Speech time by a [þtense] T, as shown in (21).3 (21)

On the Relaxation of Intervention E¤ects

225

Hornstein claims that tense points can interact only under government, deﬁned as follows: (22) A governs B i¤def all maximal projections that dominate B dominate A, and if A governs B then A governs the head of B. (Hornstein 1990, 168) On the assumption that T (INFL in Hornstein’s analysis) is the head of CP (S 0 in his analysis), the embedding V governs the embedded CP and its head, namely, T. Hence, in (12) (repeated here as (23)), the embedding E point (E1 ), which is provided by the embedding verb heard, can interact with the embedded S point (S2 ), which is provided by the embedded T, as represented by (24). (23) John heard that Mary was pregnant (24) [TP John [T[Past] [ AspP Asp [vP v [VP hear [CP that [TP Mary [T[Present] S1 R1 E1 S2 [be pregnant]]]]]]]]] This allows the sequence-of-tense rule to link the embedded S point (S2 ) with the embedding E point (E1 ), which yields the derived tense structure in (13). Hornstein’s analysis, however, cannot account for the sequence-oftense phenomena with inﬁnitive/subjunctive clauses. Let us again consider (17a) (repeated here as (25)). (25) John hopes to be in the correct room Its structure is represented in (26). (26) [TP John [T[Present] [ AspP Asp [vP v [VP hope [CP C [TP PRO to S1 R1 E1 [ AspP Asp [be in the correct room]]]]]]]]] R2 Recall that the S point is lacking in the temporal structure of an inﬁnitive complement clause, and thus the sequence-of-tense rule links the embedding E point (E1 ), which is provided by the embedding verb hope, with the embedded R point (R2 ), which is provided by the embedded Asp. In (26), however, the embedding V does not govern the embedded Asp, since CP, which is assumed to be the maximal projection of T under Hornstein’s analysis, dominates the embedded Asp but not the embedding V. Hence, Hornstein’s analysis would wrongly predict that the sequence-oftense rule cannot apply to an inﬁnitive/subjunctive complement clause.

226

Toru Ishii

Note also that Hornstein’s analysis crucially relies on the notion of government, which is no longer available in the Minimalist Program (MP). I depart from Hornstein’s analysis, arguing that tense points can interact only under a local structural relation, that is, a speciﬁer-head or headcomplement relation. I assume the widespread view that the head of CP is C. I also claim that T, which hosts an S point, covertly raises to C. Let us ﬁrst look at how this analysis accounts for the sequence-of-tense phenomena with ﬁnite clauses. (27) . . . [vP v-V [VP tv [CP C-T [TP Subj [tT . . . As shown schematically in (27), the embedded [þtensed] T, which provides the embedded S point, covertly raises to the embedded C.4 The matrix V (more precisely, the trace/copy left by V-raising to v), which provides the matrix E point, enters into a head-complement relation with the embedded CP, to whose head the embedded T is raised. Hence, the two tense points, the matrix E and embedded S points, can interact; the sequence-of-tense rule can link these two points. Let us now turn to the sequence-of-tense phenomena found with inﬁnitive/subjunctive complement clauses. On the basis of insights by Burzio (1986) and Boeckx (2001), I propose covert restructuring, whereby inﬁnitive/subjunctive complements undergo covert clause-union, becoming part of the higher clauses covertly. More speciﬁcally, I argue that the embedded AspP, whose head Asp provides the embedded R point, covertly raises to the Spec of the matrix v to which the matrix V, which provides the matrix E point, raises, as shown schematically in (28). (28) . . . [vP [AspP . . . ] [ v 0 Subj1 [ v 0 v-V [VP tv [CP C [TP Subj2 [T [AspP . . . ] ... In (28), the matrix subject (Subj1 ) originates in the Spec of v and then AspP raises to the outer Spec of v. The embedded AspP enters into a speciﬁer-head relation with the v-V complex. The matrix E and embedded R points can interact under the speciﬁer-head relation; the sequence-oftense rule can link these two points. Hence, the covert restructuring analysis can account for the sequence-of-tense phenomena found with inﬁnitive/subjunctive clauses. 9.2.2 The Relaxation of the Intervention E¤ects Having shown that the covert restructuring analysis accounts for the sequence-of-tense phenomena found with inﬁnitive/subjunctive com-

On the Relaxation of Intervention E¤ects

227

plements, I will now argue that this analysis, which is independently motivated by the temporal interpretations of inﬁnitive/subjunctive complements, also accounts for the fact that the Defective Intervention Constraint (1) is relaxed when extraction takes place out of an inﬁnitive/ subjunctive indirect question, as shown in (4) and (5) (repeated here as (29) and (30), respectively). (29) a. Which of the new booksi do you wonder [when j to buy ti t j ]? b. Which pastai do you wonder [how to cook ti t j ]? (30) a. Which of the new booksi do you wonder [when j you should buy ti t j ]? b. Which pastai do you wonder [how you should cook ti t j ]? Let us consider (29a) as an example. Given the covert restructuring analysis, we construct the matrix vP (31) during its derivation.5 (31) [vP [AspP tPRO buy which of the new books] [ v 0 you " [VP wonder [CP when [TP PRO to [AspP tPRO buy which of the new books]]]]]] In (31), the embedded AspP tPRO buy which of the new books is required to raise covertly to the Spec of the matrix v for its temporal interpretation, since inﬁnitive/subjunctive complements obligatorily undergo the sequence-of-tense rule. The present analysis assumes the single-output syntax advocated by, among others, Bobaljik (1995), Brody (1995), Groat and O’Neil (1996), and Pesetsky (2000), which claims that Spell-Out takes place after covert operations. In the single-output model, there is just one movement component; the phonological principles of chain pronunciation determine whether the head of a chain is the only position pronounced (overt movement) or whether a trace position is pronounced (covert movement). I argue that this AspP-movement is a covert phrasal movement in the sense that the moved phrase is pronounced in the original position of its chain and unpronounced in its moved position in the PF component. Under this analysis, the wh-phrase which of the new books, which is buried in the moved AspP, can bypass the intervening element when without inducing any intervention e¤ects.6 The wh-phrase which of the new books raises from within the inner Spec of the matrix v to the outer Spec of the matrix v and then to the Spec of the matrix C, yielding (32).7

228

Toru Ishii

(32) [CP Which of the new books do [TP you [vP which of " the new books [ AspP tPRO buy which of the new books] " [ v 0 tyou [VP wonder [CP when PRO to [ AspP tPRO buy which of the new books]]]]]]] Following Pesetsky (2000), let us assume that the ﬁrst instance of phrasal movement of a wh-phrase to the Spec of an interrogative C is overt in that the wh-phrase is pronounced in the head position of its chain and unpronounced in the other positions in the PF component. In (32), the wh-phrase which of the new books is pronounced in the Spec of the matrix C and unpronounced in the other positions. This yields (29a).8 (29b) and (30a–b) can be accounted for in the same say. Hence, the covert restructuring analysis can account for the relaxation of the intervention e¤ects with inﬁnitive/subjunctive indirect questions. 9.2.3 Transparency E¤ects with Inﬁnitive/Subjunctive Complements It has been observed that inﬁnitive/subjunctive clauses pattern alike regarding binding phenomena. In particular, inﬁnitive/subjunctive complements are typically more transparent, showing clause-mate e¤ects like anaphor binding and obviation. Here, we will look at another piece of evidence for the covert restructuring analysis: it can also account for transparency e¤ects with inﬁnitive/subjunctive complements. Let us ﬁrst look at anaphor-binding facts. There are languages in which an anaphor can refer to an antecedent across an inﬁnitive or subjunctive clause, but not across an indicative clause. Icelandic is the best-studied language of this type (see, e.g., Anderson 1986; Manzini and Wexler 1987; Koster 1987; Manzini 2000). (33) a. *Jo´ni segir aD Maria elskar sigi John says that Mary loves.ind himself ‘John i says that Mary loves himself i .’ b. Jo´ni segir aD Maria elski sigi John says that Mary loves.subj himself ‘John i says that Mary loves him i .’ c. Mariai skipaDi Harald aD PRO raka sigi Mary ordered Harold to shave herself ‘Mary i ordered Harold to shave her i .’ (Manzini and Wexler 1987, 417)

On the Relaxation of Intervention E¤ects

229

As shown in (33b–c), when the reﬂexive pronoun sig is embedded in an inﬁnitive or subjunctive clause, it can be bound by the matrix subject. However, as shown in (33a), when embedded in an indicative clause, it cannot be. Let us next look at obviation facts. In languages like Catalan, Icelandic, Italian, and Spanish, the subject of a subjunctive clause is disjoint in reference from the matrix subject (see, e.g., Picallo 1985; Kempchinsky 1986; Manzini 2000). The following examples are from Catalan: (34) a. Ili dis qu’ili ne montrera cette photo he say that he neg show.ind this picture ‘Hei says that hei will not show this picture.’ b. *Ili veut qu’ili vienne he want that he come.subj ‘Hei wants that hei comes.’ (Picallo 1985, 1–2) As shown in (34a), the coreference reading between the embedded and matrix subjects is possible when the embedded clause is indicative. When the embedded clause is subjunctive, however, the coreference reading is not possible, as shown in (34b). The anaphor-binding and obviation facts show that inﬁnitive/subjunctive clauses are more transparent than indicative clauses, though there are parametric variations among languages with respect to binding domains and the existence of long-distance reﬂexives that determine whether or not the transparency e¤ects are observed in a given language (see, e.g., Manzini and Wexler 1987 for detailed discussion). The covert restructuring analysis accounts for these transparency e¤ects. Let us assume that TP constitutes a binding domain for reﬂexives in languages like Icelandic and pronouns in languages like Catalan. Let us also assume with Chomsky (1995), among others, that binding conditions apply at LF. Then, in the Icelandic example (33a), the antecedent Jo´n ‘John’ is outside the minimal TP that contains the reﬂexive pronoun sig ‘himself ’; anaphor binding is not possible. In the Catalan example (34a), the antecedent il ‘he’, the matrix subject, is outside the minimal TP that contains the embedded subject pronoun il ‘he’; there is no obviation e¤ect. Let us next consider the Icelandic examples (33b–c) and the Catalan example (34b). Under the covert restructuring analysis, since the complement clause is inﬁnitive or subjunctive in (33b–c) and (34b), the embedded AspP undergoes covert phrasal movement to the Spec of the

230

Toru Ishii

matrix v. The relevant structures of (33b) and (34b) are represented in (35) and (36), respectively. (35) [TP Jo´ni [vP [AspP Mariaj elski sigi ] [ v 0 tJon ´ [VP segir [CP aD . . . (36) [TP Ili [vP [AspP ili vienne] [ v 0 tIl [VP veut [CP que . . . In (35), the antecedent Jo´n ‘John’ is within the minimal TP that contains the reﬂexive pronoun sig ‘himself ’ in the covertly moved AspP in the Spec of the matrix v. Anaphor binding is possible, and hence (33b) is acceptable. (33c) can be accounted for in the same way. In (36), since the antecedent il ‘he’, the matrix subject, is within the minimal TP that contains the embedded subject pronoun il ‘he’ in the covertly moved AspP, they cannot be coreferential with each other. Hence, (34b) is deviant. The covert restructuring analysis receives further support from a subject/nonsubject asymmetry with respect to anaphor-binding and obviation facts. Let us ﬁrst look at anaphor binding. As shown in (37), in Icelandic, a reﬂexive in the subjunctive complement clause may not be interpreted as coreferential with a matrix nonsubject. (37) *E´g sagDi Jo´ni aD Marı´a hefDi boDiD se´ri I told John that Mary had (subj) invited himself ‘I told Johni that Mary had invited himi .’ (Anderson 1986, 67) Note that the Icelandic reﬂexive is not subject oriented, as exempliﬁed by (38). (38) E´g sendi Jo´ni fo¨t a sigi I sent John clothes for himself ‘I sent John i clothes for him i .’ (Manzini and Wexler 1987, 437) Hence, the unacceptability of (37) is not due to any non-subject-oriented property of the Icelandic reﬂexive. Similar facts are found with obviation. In Catalan, an embedded subject and a matrix nonsubject can corefer, as shown in (39). (39) La Merce` deia al Perei que [ proi busque´a una feina the Merce` told to.the Peter that (he) looked.for a better millor] job ‘Merce` told Peteri that (hei ) looked for a better job.’ (Picallo 1985, 83)

On the Relaxation of Intervention E¤ects

231

Under the covert restructuring analysis, during the derivation of (37), the embedded AspP, which contains the reﬂexive se´r ‘himself ’, moves covertly to the Spec of the matrix v. ´ ni [CP (40) [TP E´g [vP [AspP Marı´a hefDi boDiD se´ri ] [ v 0 t Eg ´ [VP sagDi Jo aD . . . Since the matrix indirect object Jo´n ‘John’ is within the matrix VP, it does not c-command the reﬂexive sig ‘himself ’ within the covertly moved AspP, which resides in the Spec of the matrix v. Hence, anaphor binding is not possible. Similarly, in (39), the indirect object Pere ‘Peter’, which is within the matrix VP, does not c-command the empty pronoun pro within the covertly moved AspP, which is in the Spec of the matrix v. Hence, the obviation e¤ect is not observed. 9.3

A Subjacency-Type Account of Selective Extractability

9.3.1 Selective Extractability In the previous section, I pointed out that examples like (29) and (30), where the wh-phrases are extracted out of inﬁnitive/subjunctive indirect questions, are acceptable, and I argued that such relaxation of the intervention e¤ects can be accounted for by the covert restructuring analysis. In this section, I will ﬁrst point out that the intervention e¤ects are not always relaxed with inﬁnitive/subjunctive indirect questions; only Dlinked wh-phrases in the sense of Pesetsky (1987) can be extracted out of inﬁnitive/subjunctive indirect questions. I will then argue that the selective extractability of a wh-phrase out of an inﬁnitive/subjunctive indirect question should be given a Subjacency-type account based on the Phase Impenetrability Condition (PIC). In (29) and (30), where the extracted wh-phrases which of the new books and which pasta are D-linked, inﬁnitive/subjunctive indirect questions do not function as islands. Inﬁnitive/subjunctive indirect questions, however, are not always devoid of islandhood. First, non-D-linked wh-phrases like who and what cannot be extracted out of inﬁnitive/subjunctive indirect questions, as shown in (41) and (42) (cf. Cinque 1990; Rizzi 1990; Boeckx 2001).9 (41) a. ??Whati do you wonder [when j to buy ti t j ]? b. ??Whati do you wonder [howj to cook ti t j ]? (42) a. ??Whati do you wonder [howj you should buy ti t j ]? b. ??Whati do you wonder [howj you should cook ti t j ]?

232

Toru Ishii

Although judgments regarding extraction out of a subjunctive indirect question vary among speakers, there are speakers who see a clear contrast between extraction of a D-linked wh-phrase as in (29) and (30) and that of a non-D-linked wh-phrase as in (41) and (42). Second, it has been observed that wh-adjuncts like how, when, and why cannot be extracted out of inﬁnitive/subjunctive indirect questions, as shown in (43)–(45) (see, e.g., Huang 1982). (43) a. *Howj did you wonder [which problemi to solve ti tj ]? b. *Howj did you wonder [which problemi you should solve ti tj ]? (44) a. *Whenj did you wonder [wherei to go ti tj ]? b. *Whenj did you wonder [wherei you should go ti tj ]? (45) a. *Whyj did you wonder [wherei to go ti tj ]? b. *Whyj did you wonder [wherei you should go ti tj ]? Note that this fact has traditionally been attributed to the Empty Category Principle (ECP). Although the ECP accounts for the wh-island e¤ects with wh-adjunct extraction in (43)–(45), it cannot be invoked in a minimalist account, since ECP approaches employ notions such as head government, antecedent government, proper government, and indexing that are not allowed in the Minimalist Program. We therefore need to give an alternative account of (43)–(45) that is compatible with the minimalist spirit. 9.3.2 The Phase Impenetrability Condition and Successive Cyclicity Before we turn to an account of the selective extractability of wh-phrases out of inﬁnitive/subjunctive indirect questions, a few remarks are in order concerning the Phase Impenetrability Condition (PIC) and successivecyclic movement. Chomsky (2000, 2001, 2004) assumes the PIC (46) (adapted from Chomsky 2001, 14), which ensures that derivations proceed phase by phase, thereby reducing computational burden.10 (46) In [ZP Z . . . [HP a [H 0 H YP]]], where HP is a phase and ZP is the next phase (where phases are vP and CP), the domain of H is not accessible to operations at ZP. Chomsky claims that the PIC follows from the fact that Spell-Out is subject to the general condition on operations (47) (adapted from Chomsky 2001, 14). (47) A phase Ph1 is interpreted/evaluated at the next phase Ph2 .

On the Relaxation of Intervention E¤ects

233

In order to ensure successive-cyclic movement under the PIC, Chomsky makes the following assumption (adapted from Chomsky 2001, 12): (48) The head of a phase, C and v, may be assigned an EPP feature. This provides an ‘‘escape hatch’’ for successive-cyclic movement through the edge of a phase. Chomsky assumes the probe-goal theory of movement, where three kinds of uninterpretable formal features are involved in overt movement. In overt wh-movement to the Spec of an interrogative C, the following uninterpretable formal features are involved: the Q-feature of C, the whfeature of a wh-phrase, and the EPP feature of C, as shown in (49a). (49) a. [C [Q; EPP] [ . . . wh-phrase [Q; wh] . . . ]]] b. [C [ Q ; EPP] [ . . . wh-phrase [Q; wh ] . . . ]]] c. [wh-phrase [Q] [C [ EPP ] [ . . . twh-phrase . . . ]]] The uninterpretable Q-feature of C, which counts as a probe, seeks a goal, namely, a matching feature. The Q-feature of C enters into a matching relation with that of the wh-phrase, which is interpretable. As shown in (49b), the Q-feature of the probe C and the wh-feature of the goal whphrase, being uninterpretable, are both deleted (valued). The Q-feature of the wh-phrase, being interpretable, remains. As shown in (49c), the EPP feature of C, being a selectional feature, merges the wh-phrase in Spec,C. Since the EPP feature is uninterpretable, it is deleted (valued). This is essentially the mechanism responsible for overt wh-movement. Chomsky (2000) applies the probe-goal theory of movement, which assumes three kinds of uninterpretable formal features to be involved in overt movement, to successive-cyclic movement without change. He assumes that a noninterrogative C or v without undeleted (unvalued) ffeatures may also be assigned a nonspeciﬁc periphery feature (P-feature), which is contingent on the assignment of the EPP feature to the head of a phase in accordance with (48). While essentially following Chomsky’s analysis of successive-cyclic movement, I depart from it in claiming that only C may be assigned an EPP feature (and a P-feature); v may not. Let us look at how successivecyclic movement proceeds under this analysis, taking (50) as an example. (50) Where did John buy it t? During its derivation, we construct the vP phase (51). (51) [vP John [v[f ] [VP buy it[f; Case ] where [Q; wh] ]]]

234

Toru Ishii

In (51), Agree of v with the object it establishes a Case-agreement relation, deleting (valuing) the f-features of v and the Case feature of it. Although v is not assigned an EPP feature or a P-feature, where may raise to the vP-edge position, as shown in (52). (52) [vP where[Q; wh] [ v 0 John [v [VP buy it[f] twhere ]]]] Although this movement is not triggered by any formal feature at this vP-phase level, it does not violate the economy condition that bans superﬂuous steps in a derivation. This is because according to (47), evaluation/ interpretation for the vP phase takes place at the next phase, namely, the CP phase. That evaluation includes whether or not movement of where to the vP-edge position satisﬁes the economy condition. At the CP-phase level, C is assigned an EPP feature and a Q-feature. These features trigger movement of where to Spec,C, which licenses movement of where to the vP-edge position. This ensures successive-cyclic movement without assuming that v may be assigned an EPP feature (and a P-feature).11 The analysis of successive-cyclic movement proposed here should be preferred over Chomsky’s analysis on theoretical grounds, because it does not have to assume v with a P-feature. Chomsky (2000) assumes that P-features are defective, arguing that they are analogous to the [person] feature of a defective T. There is a case where C is assigned a Q-feature, a full complement of peripheral features. By analogy with the T-system, it is reasonable to claim that defective C with a P-feature also exists. In contrast to C, v is never assigned a Q-feature; this is clear from the fact that a wh-phrase can never be stranded in the vP-edge position. This casts serious doubt on Chomsky’s analysis, which assumes v with a P-feature. 9.3.3 D-Linking/Speciﬁcity and a Subjacency-Type Account Returning to the main subject, I now propose a minimalist account of the selective extractability of wh-phrases out of inﬁnitive/subjunctive indirect questions based on the claim that speciﬁc DPs in the sense of Enc¸ (1991) move out of VP to the vP-edge position while nonspeciﬁc DPs do not, as advocated by, among others, Mahajan (1990), Diesing (1992), Runner (1994), and Torrego (1998). Essentially following Diesing (1992), I argue that speciﬁc (what Diesing calls ‘‘presuppositional’’) DPs are required to move out of VP to the vP-edge position before a derivation reaches LF, in accordance with the Mapping Hypothesis. One of Diesing’s arguments supporting this contention is based on the antecedent-contained deletion (ACD) construction (see, e.g., Bouton 1970; Sag 1976; May 1985; Larson and May 1990; Fiengo and May 1994; Pesetsky 2000).

On the Relaxation of Intervention E¤ects

235

(53) Mary read every book that John will The meaning of the elided VP complement of will cannot be understood as the matrix VP read every book that John will, since this would lead to an inﬁnite regress. If the DP every book that John will moves out of the VP covertly, then it leaves behind a VP of the form read t, which provides the appropriate antecedent for the elided VP, as shown in (54). (54) Mary [vP [DP every book that John will [VP read t]] [VP read t]] Hence, as argued by Pesetsky (2000), the ACD construction can be used as a diagnostic for phrasal movement: unless phrasal movement takes place, there is no way of licensing the ACD construction. Diesing observes that while speciﬁc DPs allow the ACD construction, as shown in (55), nonspeciﬁc DPs do not, as shown in (56). (55) a. b. c. d.

I read every book that you did I read each book that you did Robert played many of the pianos that Clara did Dulles suspected the spy that Angleton did ((55a–c) from Diesing 1992, 71–72; (55d) from Fiengo and May 1994, 242)

(56) a. *?I read many books that you did b. *I read books that you did c. *Max put some things he could in his pockets (Diesing 1992) The contrast between (55) and (56) supports the Mapping Hypothesis, which requires speciﬁc DPs, but not nonspeciﬁc DPs, to move out of VP covertly. As argued by Cinque (1990), among others, D-linked wh-phrases also count as speciﬁc, since they are understood as presupposing the existence of a known set of speciﬁc elements. In (29a) and (30a), for example, the D-linked wh-phrase which of the new books presupposes the existence of a known set of speciﬁc new books, and the question is interpreted as asking which of such new books a person has bought. Further evidence for this contention comes from ACD facts. As observed by Pesetsky (2000), exactly like speciﬁc DPs, D-linked wh-phrases in situ also license the ACD construction, which suggests that D-linked wh-phrases are themselves speciﬁc and thus required to move out of VP covertly in accordance with the Mapping Hypothesis.

236

Toru Ishii

(57) Which girl invited [which student that John did]? (Pesetsky 2000, 6) Let us consider (29a) (repeated here as (58)) as an example. (58) Which of the new booksi did you wonder [when j to buy ti t j ]? During its derivation, we construct the embedded vP phase (59). (59) [vP PRO [v [VP buy which of the new books when]]] According to Diesing’s analysis, the D-linked wh-phrase which of the new books, being speciﬁc, is required to move out of the VP covertly in accordance with the Mapping Hypothesis. Recall that we are assuming that a phase Ph1 is interpreted/evaluated at the next phase Ph2 (47). Hence, this semantic/interpretive condition on speciﬁc DPs requires the D-linked/ speciﬁc wh-phrase which of the new books to move out of the VP into the vP-edge position covertly until the derivation reaches the next phase level, namely, the embedded CP phase (60). (60) [CP when PRO to buy [ AspP [vP which of the new books " 0 0 0 [ v t when [ v tPRO buy which of the new books t when ]]]]] Note in passing that in (60), the other wh-phrase when moves ﬁrst to the vP-edge position and then to the Spec of the embedded C.12 As the derivation proceeds, we construct the matrix vP phase; the relevant structure is represented in (61). (61) [ vP 1 [AspP [vP 2 which of the new books [v 0 2 t 0 when [v 0 2 tPRO buy which of the new books twhen ]]]] [ v 01 you [VP wonder when . . . In (61), the embedded AspP undergoes covert restructuring. Recall that the D-linked/speciﬁc wh-phrase which of the new books resides in the embedded vP-edge (vP 2 -edge) position. From there, it raises to the matrix vP-edge (vP 1 -edge) position, as shown in (62). (62) [ vP 1 which of the new books [ v 01 [ AspP [ vP 2 which of " the new books [ v 0 2 t 0 when [ v 0 2 tPRO buy which of the new books t when ]]]] [ v 01 you [VP wonder when . . . Although this movement is not triggered by any formal feature at the matrix vP-phase level, its evaluation with respect to the economy condition takes places at the next phase, namely, the matrix CP phase, in accordance with (47). At the matrix CP-phase level, the Q-feature and

On the Relaxation of Intervention E¤ects

237

EPP feature of the matrix C trigger movement of which of the new books to its Spec, as shown in (63). (63) [CP which of the new books [C [you [ vP 1 which of the " new books [ v 01 [ AspP [ vP 2 which of the new books [ v 0 2 t 0 when [ v 0 2 tPRO buy which of the new books t when ]]]] [ v 01 you [VP wonder when j . . . Recall that the covertly moved embedded AspP is not pronounced in the head position of its chain (in the matrix clause); instead, it is pronounced in its trace position (in the embedded clause). Recall also that according to Pesetsky’s (2000) principle of pronunciation regarding wh-movement, the ﬁrst instance of phrasal movement of a wh-phrase to the Spec of an interrogative C is overt in that the wh-phrase is pronounced in the head position of its chain and unpronounced in the other positions. In this derivation, the wh-phrases which of the new books and when are pronounced in Spec,C and unpronounced in the other positions. This derivation converges; (58) is acceptable. (29b) and (30a–b) can be explained in the same way. Let us next consider (41a) (repeated here as (64)). (64) ??Whati do you wonder [when j to buy ti t j ]? During its derivation, we construct the matrix vP (vP 1 ) phase; the relevant structure is shown in (65). (65) [ vP 1 [AspP [vP 2 tPRO to buy what thow ]] [ v 01 you [VP wonder how . . . In (65), the embedded AspP undergoes covert restructuring. Note that there is no way of raising the non-D-linked/nonspeciﬁc wh-phrase what to the embedded vP-edge (vP 2 -edge) position. Recall that under the proposed analysis of successive-cyclic movement, v may not be assigned an EPP feature or a P-feature; there are no formal features that trigger movement of what to the embedded vP-edge (vP 2 -edge) position. Furthermore, since what is non-D-linked/nonspeciﬁc, the Mapping Hypothesis does not require what to move out of VP to the vP-edge position. It should also be noted that what cannot undergo non-feature-driven movement to the embedded vP-edge (vP 2 -edge) position because of the economy condition. This is because what does not undergo feature-driven movement at the next phase (the matrix vP (vP 1 ) phase), where movement in the embedded vP (vP 2 ) phase is evaluated, since under the proposed analysis of successive-cyclic movement, v may not be assigned an EPP feature or a P-feature. The wh-phrase what stays within the embedded vP (vP 2 ) phase and thus is not accessible to operations at the

238

Toru Ishii

matrix CP because of the PIC. There is no way of satisfying the uninterpretable EPP feature and Q-feature of the matrix C. This derivation crashes; (64) is therefore unacceptable. (41b) and (42a–b) can be ruled out in the same way.13 As shown in (43)–(45), adjunct wh-phrases cannot be extracted out of inﬁnitive/subjunctive indirect questions, either. This fact can be explained on a par with the inability of non-D-linked whphrases like who and what to be extracted out of inﬁnitive/subjunctive indirect questions. This is because the notion of speciﬁcity is only relevant to DP denotations; thus, adjunct wh-phrases like why, how, and when, which are not DPs, do not count as speciﬁc.14 9.4

Double Wh-Islands and Relativization

The analysis proposed here receives further support from the fact that Dlinked/speciﬁc wh-phrases, though extractable from inﬁnitive/subjunctive indirect questions, exhibit ‘‘double wh-island’’ e¤ects with inﬁnitive/subjunctive indirect questions, as shown in (66). (66) ??Which assignmenti did you ask [to whomj to ﬁnd out [whenk to give ti t j tk ]]? As the derivation proceeds, we construct the matrix vP (vP 1 ) phase (67). (67) [ vP 1 [AspP 2 [ vP 2 [AspP 3 [ vP 3 which assignment [ v 03 tPRO give " which assignment tk ]]] [ v 0 2 tPRO ﬁnd out whenk to . . . ]] [ v 01 you ask to whomj . . . In (67), covert restructuring (covert AspP-movement) applies twice. The D-linked/speciﬁc wh-phrase which assignment is required to move out of VP to the most deeply embedded vP-edge (vP 3 -edge) position in accordance with the Mapping Hypothesis. It cannot raise any further because of the economy condition, however, since its movement is not triggered by any formal feature at the intermediate vP-phase (vP 2 -phase) level or at the matrix vP-phase (vP 1 -phase) level. Which assignment is not accessible to operations at the matrix CP because of the PIC. There is no way of satisfying the uninterpretable EPP feature and Q-feature of the matrix C. This derivation crashes; the ‘‘double wh-island’’ e¤ects follow. The proposed analysis can also account for the fact that relativization behaves exactly like movement of a D-linked/speciﬁc wh-phrase with respect to the Wh-Island Constraint, as shown in (68) and (69).

On the Relaxation of Intervention E¤ects

239

(68) a. My old Dodge, which my mechanic knew [how to ﬁx t], . . . b. ??My old Dodge, which I know [how my mechanic ﬁxed t], . . . (69) ?These are the only vegetables which I don’t know [how to ﬁnd out [where to plant t]] (Frampton 1990, 62–63) As shown in (68a) and (68b), respectively, relativization may take place out of an inﬁnitive indirect question but not out of a ﬁnite indirect question. Relativization out of an inﬁnitive indirect question exhibits the ‘‘double wh-island’’ e¤ects, as shown in (69). These facts also follow from the proposed analysis if we assume with Cinque (1990) that relative operators are speciﬁc, just like D-linked wh-phrases. 9.5

Conclusion

In this chapter, I have investigated the relaxation of the intervention e¤ects. I have shown that indirect questions do not always constitute islands; D-linked/speciﬁc wh-phrases can be extracted out of inﬁnitive/ subjunctive indirect questions. I have argued that this hitherto unexplained anomaly concerning the Wh-Island Constraint is given a principled minimalist account by covert restructuring and the PIC. If the present analysis is on the right track, it supports the view that we still need a traditional Subjacency-type account based on the PIC to fully accommodate wh-island e¤ects, contrary to the widespread view that the Wh-Island Constraint can be subsumed under the Defective Intervention Constraint. Before closing this chapter, I will brieﬂy point out an interesting theoretical issue raised by the proposed analysis. Recall that this analysis crucially assumes that derivations are interpreted/evaluated locally. More speciﬁcally, following Chomsky (2001, 2004), I have assumed here that a phase is interpreted/evaluated at the next phase. Within the theory of computational complexity, it is generally agreed that local considerations induce less computational burden than global ones (see, e.g., Chomsky 1995; Fukui 1996; Ishii 1997). As Chomsky (2000) argues, however, it is not clear whether computational complexity matters for a cognitive system like language, which does not involve any processing but only stores information; in other words, there is no a priori reason to claim that the computation should be local to avoid the problem of computational complexity. Nevertheless, we need to seek a resolution of this local versus global issue on empirical grounds. Even if language is local in nature, we

240

Toru Ishii

still need to ﬁnd out, empirically, how local the computation should be. The analysis proposed here gives a local account based on the notion of phase to the relaxation of the intervention e¤ects and the selective extractability of wh-phrases with inﬁnitive/subjunctive indirect questions, providing evidence that language is phasally local in nature. Notes I have greatly beneﬁted from discussions with Brian Agbayani. My special thanks are also due to an anonymous reviewer for carefully reading an earlier version of this chapter and providing critical comments. I would also like to thank the participants at the workshop ‘‘On Wh-Movement’’ for their comments. Remaining errors and omissions are, of course, solely my responsibility. This work was supported in part by the Japan Society for the Promotion of Science under grant Scientiﬁc Research C2 14510543 and by a grant from the Institute of Humanities at Meiji University. 1. See section 9.3 for an explication of the more precise mechanism responsible for overt wh-movement in the probe-goal theory of movement. 2. Hornstein’s (1990) analysis of inﬁnitive clauses is di¤erent from Stowell’s (1982). Citing (i) as an example, Stowell claims that a control inﬁnitive clause has an independent temporal interpretation in the sense that its tense is always understood as being unrealized. (i) John remembered to bring the wine (Stowell 1982, 563) According to Stowell, in (i), John has not yet brought the wine at the time when he remembers to do so. Hornstein, however, adduces arguments against Stowell’s unrealized-tense analysis, two of which I outline here. First, when the control inﬁnitive clause is modiﬁed by an adverb, its tense is not understood as being unrealized; instead, the matrix and embedded events are interpreted as contemporaneous with each other, as shown in (ii). (ii) At 6 o’clock, John remembered to bring the wine (Hornstein 1990, 226) In (ii), John’s bringing the wine is temporally located at 6 o’clock and contemporaneous with John’s remembering. Second, contrary to what Stowell claims, (i) does not involve an ordered set of events. It does not imply that John has a memory event. Rather, it implies that John did what he was supposed to do. This is clearer with the verb forget. (iii) Yesterday, John forgot to bring the wine (Hornstein 1990, 226) (iii) does not involve an ordered set of events (i.e., a forgetting that precedes a wine-bringing); rather, it implies that John did not do what he was supposed to do. Karttunen (1971) adduces similar arguments for Hornstein’s view. Karttunen classiﬁes verbs like remember, forget, manage, happen, and bother as implicative verbs. He claims that an implicative verb does not represent a separate event, but

On the Relaxation of Intervention E¤ects

241

expresses a necessary and su‰cient condition that alone determines whether the event of the complement clause took place. The verb remember, for example, states that the subject is obligated to carry out the act described in the complement clause and also willing to carry out the obligation. According to Karttunen, in (i), for example, all that took place when John remembered to bring wine is that he did it. Remembering to bring wine is inseparable in space and time from bringing wine; it is the same event. What (i) means is that John was obligated to bring wine and intended to do so, and whether or not he did depended only on whether he remembered his commitment. 3. Here, I assume Thompson’s (1996) analysis, which claims that the R point is provided by the head of an Aspect Phrase. Hornstein (1990) assumes, however, that the R point is provided by the perfective morpheme, which is adjoined to VP. Note that crucial to the present analysis is the assumption that the R point is provided by a functional head that resides between T and v, whatever the functional head may be. 4. Hornstein (1990) argues against the covert T-to-C movement analysis advocated here. His argument is based on the fact that the application of the sequence-of-tense rule has phonological e¤ects. In (12) (repeated here as (i)) with the sequence-of-tense interpretation, for example, although the underlying tense of the embedded verb is the simple present, the embedded verb has the surface form of the past tense element was. (i) John heard that Mary was pregnant Hornstein claims that since LF operations do not feed PF operations under the standard model of grammar, the covert T-to-C movement analysis is untenable. As we will see later, however, the present analysis assumes single-output syntax, which claims that Spell-Out takes place after LF operations; LF operations feed PF operations. The present analysis therefore does not face the problem raised by Hornstein. 5. More precisely, the embedded AspP ﬁrst moves to the Spec of the embedded C and then to the Spec of the matrix v because of the PIC. Following Chomsky (2001, 2004), I assume here that a phase is interpreted/evaluated at the next phase (for details, see section 9.3.2). Then, although nothing requires the embedded AspP to raise to the Spec of the embedded C, evaluation/interpretation for the embedded CP takes place at the next phase, that is, the matrix vP phase. At the matrix vP-phase level, the embedded AspP is required to raise to the Spec of the matrix v for its temporal interpretation, which licenses its movement to the Spec of the embedded C at the embedded CP-phase level. I thank a reviewer for bringing this issue to my attention. 6. Boeckx (2001) also claims that an inﬁnitive/subjunctive complement becomes part of the higher clause, which voids the intervention e¤ects. Under Boeckx’s analysis, ﬁrst the embedded TP (IP in his analysis) raises into the matrix clause overtly (Boeckx 2001, 213). (i) Vmatrix [IP Iinﬁnitive=subjunctive [VP . . . ]]i [CP C ti ] Then the embedded CP undergoes remnant movement, reordering the CP-IP sequence (Boeckx 2001, 214).

242

Toru Ishii

(ii) Vmatrix [[CP [C [ti ]]]j [[IP Iinﬁnitive=subjunctive [VP . . . ]]i tj ]] Boeckx’s analysis is problematic in that it is unclear about the landing sites of these operations and their driving forces. 7. See section 9.3.2 for a more precise explication of how overt wh-movement proceeds in a successive-cyclic manner under the notion of phase. 8. In the derivation of (29a), the chain of the wh-phrase which of the new books consists of the following three copies: the copy in the Spec of the matrix C, the one in the Spec of the matrix v, and the one inside the moved AspP. Now, how can we ensure that the copy of which of the new books in the original position, which is not part of the chain, is unpronounced? Bobaljik (1995) points out that it is not su‰cient to claim that one member of every chain is pronounced. Rather, we should assume that maximally one copy of a single element is pronounced regardless of the number of chains in which it occurs. In (i), for example, although there are two chains—an A-chain created by movement to the subject position ¯ -chain created by movement to the Spec of the matrix C— and a subsequent A only one copy of who is pronounced (Bobaljik 1995, 350). (i) Who did Sam think [t 0 had been arrested t]? Recall that in the derivation of (29a), Pesetsky’s (2000) principle of chain pronunciation requires the wh-phrase which of the new books to be pronounced in its head position, namely, the Spec of the matrix C. It then follows that all the other copies of the wh-phrase must be unpronounced regardless of whether they are part of the chain. Hence, the copy of which of the new books in its original position is unpronounced. I thank a reviewer for bringing this issue to my attention. 9. Cinque (1990) classiﬁes the wh-island as a weak island. However, the inability of a non-D-linked wh-phrase to be extracted out of an inﬁnitive/subjunctive indirect question as in (41) and (42) suggests that wh-islands should not be treated on a par with weak islands like inner islands, factive islands, and extraposition islands. Unlike wh-islands, these weak islands allow extraction of a wh-argument even though the wh-phrase is not D-linked (see, e.g., Cinque 1990; Manzini 1992). (i) a. Inner (negative) island To whom didn’t you speak t? b. Factive island To whom do you regret that you could not speak t? c. Extraposition island To whom is it time to speak t? (Cinque 1990, 1–2) 10. Unlike the discussion in Chomsky 2001, 2004, the discussion to follow does not assume a distinction between strong and weak phases. 11. See Ishii 2003 for the view that this non-feature-driven movement of a whphrase to the vP-edge position is scrambling in the sense of Fukui (1993) and Saito and Fukui (1998). 12. Under the assumption that a phase is interpreted/evaluated at the next phase, although when in (60) is not required to raise to the Spec of the embedded v by any formal features, evaluation/interpretation of the embedded vP takes place at

On the Relaxation of Intervention E¤ects

243

the next phase, namely, the embedded CP phase. At that level, when is required to raise to the Spec of the embedded C by the EPP and P-features of the embedded C, which licenses its movement to the Spec of the embedded v at the embedded vP-phase level. 13. It is crucial in the proposed analysis that contrary to what is advocated by, among others, Johnson (1991) and Koizumi (1995), there is no overt/covert phrasal movement to a VP-external position for reasons of Case; otherwise, nonD-linked wh-objects like what in (41) and (42) may also move out of the VP and reside in the vP-edge position. Evidence for this view comes from the ACD facts. If there were any overt/covert phrasal movement to the vP-edge position for Case reasons, we would lose an explanation of the contrast in acceptability between (55) and (56) regarding the ACD facts. See also Pesetsky 2000 for more arguments supporting the view that there is no overt/covert phrasal movement to a VP-external position for Case reasons. 14. A reviewer points out that (i) is deviant, apparently contrary to what the present analysis predicts. (i) *Which datej do you wonder [howi Mary should cook the pasta ti tj ]? Huang (1982) argues that wh-adjuncts like where and when are DPs (NPs in his analysis) inserted in the environment [PP P [DP ]], where P may or may not be phonetically realized. This contention is supported by the fact that where and when can be complements of prepositions, as shown in (ii). (ii) a. From where did he come t? b. Since when has he been here t? (Huang 1982, 536) Huang furthermore argues that overt extraction of where and when cannot a¤ect only a DP within a PP regardless of whether P is phonetically realized or not, because of the Condition on Extraction Domain (CED). The CED states that an element may be extracted from a domain only if the domain is properly governed. Since the domain of an adjunct is by deﬁnition not properly governed, the CED blocks extraction of a DP out of a PP. It follows that overt extraction of where and when must a¤ect the whole PP. Like where and when, the wh-phrase which date may also appear in the complement position of P. (iii) On which date do you draw your salary t? It is then reasonable to claim that which date is also inserted in the environment [PP P [DP ]] and its overt extraction must a¤ect the whole PP. It then follows that in (i), the extracted wh-phrase which date is a PP and thus does not count as speciﬁc, since the notion of speciﬁcity is relevant only to DP denotations. Hence, the proposed analysis can account for the fact that examples like (i) are deviant. References Anderson, Stephen R. 1986. The typology of anaphoric dependencies: Icelandic (and other) reﬂexives. In Lars Hellan and Kirsti Koch Christensen, eds., Topics in Scandinavian syntax, 65–88. Dordrecht: Reidel.

244

Toru Ishii

Baker, C. L. 1989. English syntax. Cambridge, Mass.: MIT Press. Bobaljik, Jonathan David. 1995. Morphosyntax: The syntax of verbal inﬂection. Doctoral dissertation, MIT. Boeckx, Cedric. 2001. Mechanisms of chain formation. Doctoral dissertation, University of Connecticut. Bouton, Lawrence. 1970. Antecedent-contained pro-form. In Mary Ann Campbell, et al., Papers from the Sixth Regional Meeting, Chicago Linguistic Society, 154–167. Chicago: University of Chicago, Chicago Linguistic Society. Brody, Michael. 1995. Lexico-Logical Form: A radically minimalist theory. Cambridge, Mass.: MIT Press. Burzio, Luigi. 1986. Italian syntax. Dordrecht: Reidel. Chomsky, Noam. 1973. Conditions on transformations. In Stephen R. Anderson and Paul Kiparsky, eds., A festschrift for Morris Halle, 232–286. New York: Holt, Rinehart and Winston. Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1986. Barriers. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Cinque, Guglielmo. 1990. Types of A-dependencies. Cambridge, Mass.: MIT Press. Diesing, Molly. 1992. Indeﬁnites. Cambridge, Mass.: MIT Press. Enc¸, Mu¨rvet. 1991. The semantics of speciﬁcity. Linguistic Inquiry 22, 1–25. Enc¸, Mu¨rvet. 1996. Tense and modality. In Shalom Lappin, ed., The handbook of contemporary semantic theory, 345–358. Oxford: Blackwell. Fiengo, Robert, and Robert May. 1994. Indices and identity. Cambridge, Mass.: MIT Press. Frampton, John. 1990. Parasitic gaps and the theory of wh-chains. Linguistic Inquiry 21, 49–77. Fukui, Naoki. 1993. Parameters and optionality. Linguistic Inquiry 24, 399–420. Fukui, Naoki. 1996. On the nature of economy in language. Cognitive Studies 3, 51–71.

On the Relaxation of Intervention E¤ects

245

Giorgi, Alessandra, and Fabio Pianesi. 1997. Tense and aspect: From semantics to morphosyntax. Oxford: Oxford University Press. Groat, Erich, and John O’Neil. 1996. Spell-Out at the LF interface. In Werner Abraham, Samuel David Epstein, Ho¨skuldur Thra´insson, and C. Jan-Wouter Zwart, eds., Minimal ideas, 113–139. Amsterdam: John Benjamins. Hornstein, Norbert. 1990. As time goes by. Cambridge, Mass.: MIT Press. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT. Ishii, Toru. 1997. An asymmetry in the composition of phrase structure and its consequences. Doctoral dissertation, University of California, Irvine. Ishii, Toru. 2003. Weak crossover and scrambling in English. English Linguistics 20.2, 562–572. The English Linguistic Society of Japan. Johnson, Kyle. 1991. Object positions. Natural Language and Linguistic Theory 9, 577–636. Karttunen, Lauri. 1971. Implicative verbs. Language 47, 340–358. Kempchinsky, Paula. 1986. Romance subjunctive clauses and Logical Form. Doctoral dissertation, UCLA. Koizumi, Masatoshi. 1995. Phrase structure in minimalist syntax. Doctoral dissertation, MIT. Koster, Jan. 1987. Domains and dynasties: The radical autonomy of syntax. Dordrecht: Foris. Larson, Richard, and Robert May. 1990. Antecedent containment or vacuous movement: Reply to Baltin. Linguistic Inquiry 21, 103–122. Mahajan, Anoop. 1990. The A/A-bar distinction and movement theory. Doctoral dissertation, MIT. Manzini, Maria Rita. 1992. Locality: A theory and some of its empirical consequences. Cambridge, Mass.: MIT Press. Manzini, Maria Rita. 2000. Sentential complementation: The subjunctive. In Peter Coopmans, Martin Everaert, and Jane Grimshaw, eds., Lexical speciﬁcation and insertion, 241–267. Amsterdam: John Benjamins. Manzini, Maria Rita, and Kenneth Wexler. 1987. Parameters, binding theory, and learnability. Linguistic Inquiry 18, 413–444. May, Robert. 1985. Logical Form. Cambridge, Mass.: MIT Press. Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Eric Reuland and Alice G. B. ter Meulen, eds., The representation of (in)deﬁniteness, 98–129. Cambridge, Mass.: MIT Press. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: MIT Press. Picallo, Carme. 1984. The Inﬂ node and the null subject parameter. Linguistic Inquiry 15, 75–102. Picallo, Carme. 1985. Opaque domains. Doctoral dissertation, City University of New York.

246

Toru Ishii

Reichenbach, Hans. 1947. Elements of symbolic logic. New York: Macmillan. Reprint, New York: Free Press, 1966. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 2004. Locality and left periphery. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 223–251. Oxford: Oxford University Press. Runner, Je¤rey. 1994. A speciﬁc role for AGR. In Elena Benedicto and Je¤rey Runner, eds., Functional projection, 153–177. University of Massachusetts Occasional Papers in Linguistics 17. Amherst: University of Massachusetts, GLSA. Sag, Ivan. 1976. Deletion and Logical Form. Doctoral dissertation, MIT. Saito, Mamoru, and Naoki Fukui. 1998. Order in phrase structure and movement. Linguistic Inquiry 29, 439–474. Stowell, Tim. 1982. The tense of inﬁnitives. Linguistic Inquiry 13, 561–570. Stowell, Tim. 1995. What do the present and past tenses mean? In Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham, and Mario Squartini, eds., Temporal reference, aspect, and actionality. Vol. 1, Semantic and syntactic perspectives, 381–396. Turin: Rosenberg and Sellier. Stowell, Tim. 1996. The phrase structure of tense. In Johan Rooryck and Laurie Zaring, eds., Phrase structure and the lexicon, 277–291. Dordrecht: Kluwer. Thompson, Ellen. 1996. The syntax of tense. Doctoral dissertation, University of Massachusetts, Amherst. Torrego, Esther. 1998. The dependencies of objects. Cambridge, Mass.: MIT Press. Zagona, Karen. 1988. Verb phrase syntax: A parametric study of English and Spanish. Dordrecht: Kluwer.

PART IV Wh-Constructions

Chapter 10 Top Issues in Questions: Topics–Topicalization– Topicalizability

10.1

Kleanthes K. Grohmann

On Wh-Movement and ‘‘On Wh-Movement’’

Noam Chomsky, in his seminal paper ‘‘On Wh-Movement,’’ mentions in passing an unacceptable structure involving an embedded wh-phrase moved beyond a fronted topic. The datum in question is (1a), taken from Chomsky 1977, 105, (132b); (1b) highlights the (updated) analytical details relevant here. (1) a. *I wonder [who2 , this book, [(which1 ) t2 really likes t1 ]] b. [CP whoi C 0 [TopP this bookk Top 0 [ IP t i I 0 (tk ) [vP really t i likes tk ]]]] Taking it out of its historical context and disregarding notational di¤erences, I want to investigate structures that superﬁcially look like grammatical counterparts of (1b): a fronted wh-phrase moved over a preposed topic. The structures this chapter revolves around are questions that involve two wh-phrases in between which a topicalized constituent may or may not be sandwiched in German, the primary language of concern; comparative data will be presented in support of the unfolding proposal. The upshot of this study is that looks are deceiving: what may look like a multiple wh-question involving a fronted topic is in fact multiple topicalization and does not serve as a proper information question. Empirically, the discussion starts o¤ with the main fact that might be drawn from the paradigm in (2) and related ones, discussed in detail by Beck (1996a): in certain positions, certain elements seem to ‘‘intervene’’ in a certain way and thus destroy an otherwise well-formed dependency. (2) a. *Warum haben wenige Linguisten wem geglaubt? why have few linguists whom believed ‘Why did few linguists believe whom?’ b. Warum haben wem wenige Linguisten geglaubt?

250

Kleanthes K. Grohmann

In essence, I will revisit and extend a partial set of data that served as motivation for the so-called intervention e¤ect, sometimes even dubbed ‘‘Beck e¤ect’’ (Hagstrom 1998), a term popularized by Pesetsky (2000), who investigates a similarly small empirical subset in his study on phrasal movement. The source of apparent ‘‘intervention’’ in the structures discussed here is Beck’s intriguing article, the base of my discussion.1 I will argue against the existence of intervention as understood there and account for the facts observed on the basis of topicalizability. The discussion combines results from and follows up on previous work of mine on multiple wh-questions in German (Grohmann 1998; 2000a, chap. 5; 2003a). Once the notion of topicalizability is on the table and its relevance to the present subject matter clariﬁed, I will turn to a number of consequences: the role of topicalizability in the formation of multiple wh-questions in German; conclusions to be drawn from the notion of topicalizability for the structure of and derivational processes involved in the German clause; crosslinguistic implications of topicalizability for the syntax and semantics (as well as discourse properties/pragmatics) of (multiple) wh-questions in grammar.

In this sense, the chapter addresses one of the larger issues that Chomsky (1977) presented, namely, how (the rule of ) wh-movement ties in with other operations in the grammar, possibly of the same or a related type ¯ -movement), an issue that has subse(what later became known as A quently been the focus of many studies on the interaction of such operations as well as intervention in syntax (Cinque 1990; Rizzi 19902). It will be mainly concerned with the formation of multiple wh-questions (in German) and the connection with topicalization at large (across languages). 10.2

Topicalizability in Questions: Against an Intervention E¤ect

This section introduces the notion, illustrates the idea, and discards the existence of an intervention e¤ect in German multiple wh-questions as currently widely understood. While something like it may be at work in other languages—or for that matter, even in other structures in German3—there is reasonable doubt that it should be applied to mono-

Top Issues in Questions

251

clausal questions involving two wh-phrases. Instead, I will suggest an alternative approach that capitalizes on the notion of topicalizability. 10.2.1 Enter and Exit: Intervention Investigating the position and role of quantiﬁers in German multiple whquestions, among other structures, Beck (1996a) proposes the informal ban in (3),4 illustrating abstractly her deﬁnitions reproduced in (4) and (5). (3) Restriction on LF (ROLF; Beck 1996a, 1) *[ . . . X i . . . [ Q . . . [ . . . t iLF . . . ]]] (4) Quantiﬁer-Induced Barrier (QUIB; Beck 1996a, 39) The ﬁrst node that dominates a quantiﬁer, its restriction, and its nuclear scope is a Quantiﬁer-Induced Barrier. (5) Minimal Quantiﬁed Structure Constraint (MQSC; Beck 1996a, 39) If an LF trace b is dominated by a QUIB a, then the binder of b must also be dominated by a. The paradigmatic instance of a ROLF violation of interest here (ignoring other phenomena discussed by Beck) is arguably a structure like (6), in which the negative quantiﬁer niemand ‘nobody’ sits in between two whquestion words, apparently inducing the ungrammaticality by the combination of (3)–(5). (6) *Wen hat niemand wo gesehen? whom has nobody where seen ‘Where did nobody see whom?’ (Beck 1996a, 1) The relevant contrasting fact from German is that the variant of (6) in which the negative quantiﬁer follows both wh-expressions is well formed.5 (7) Wen hat wo niemand gesehen? whom has where nobody seen ‘Where did nobody see whom?’ Intervention qua MQSC apparently accounts for the contrast. Note, however, that ‘‘[t]he distinction between S-Structural and LF movement is absolutely crucial’’ (Beck 1996a, 18) for such an explanation, within the limits of Beck’s syntax (see note 9 below). In other words, LF must involve phrasal movement of all wh-material into a position like Spec,CP

252

Kleanthes K. Grohmann

in order for ROLF to apply. (As a matter of fact, Beck goes to some lengths to account for additional LF movements of items that it is not at all clear should move, including the assumption that jeder ‘every(one)’ ‘‘is raised to a CP-adjoined position at LF’’ (p. 22).) Thus, (8), slightly adjusted from Beck’s work, must be the relevant simpliﬁed LF structure of (6), where IP forms a QUIB (induced by niemand ‘nobody’) and ti cannot be bound by wo ‘where’ (because of the MQSC). (8) [CP woi wen hat [ IP niemand t i gesehen]] However, the syntactic assumptions Beck makes are not at all shared by everyone. In particular, within the Minimalist Program (Chomsky 1995), LF movement of the type ﬁrst suggested by Huang (1982) is rejected: on the basis of well-motivated economy conditions, covert phrasal movement is replaced by feature movement. This is not the right place to indulge in a more elaborate discussion.6 As it turns out, one might cite works like Pesetsky 2000 and argue that covert phrasal movement is compatible with minimalism after all. And indeed Pesetsky endorses Beck-type intervention; he even couches a deﬁnition of it in a minimalist framework. (9) Intervention e¤ect (universal characterization; Pesetsky 2000, 67) A semantic restriction on a quantiﬁer (including wh) may not be separated from that quantiﬁer by a scope-bearing element. Again, however, this does not make the approach any more valid. As argued at length by Mathieu (2002), among others, there are empirical reasons to doubt the existence of covert phrasal movement in the relevant cases, that is, wh-questions.7 One cannot help but agree with Mathieu’s appraisal that none of the explanations o¤ered for this type of intervention (neither Beck’s nor Pesetsky’s, for example8) ‘‘follow from anything’’ and that they ‘‘only amount to a restatement of the facts’’ (p. 174). Maybe the reasoning behind it, possibly more explicit in (9) than (5), can be semantically motivated or even ﬂeshed out; but with regard to the syntax involved here, I see very little hope. If anything, it should fall under some version of locality restrictions of the Relativized Minimality type (Rizzi 1990)—indeed, one could envision a greater relevance for negative islands, mentioned only in passing by Beck; see Boeckx 2003 and references cited for recent attempts in this direction.9 Another problem with explicitly (and ‘‘crucially’’) deﬁning intervention as an LF condition concerns the Uniformity Condition (see also Mathieu

Top Issues in Questions

253

1999). With the rise of the Minimalist Program and the denial of the existence of levels or components of representation other than those needed for the interfaces (LF, PF), uniformity as understood in (10) is important in the coherence of the grammar. (10) Uniformity Condition (Hornstein, Nunes, and Grohmann 2006, 74) The operations available in the covert component must be the same ones available in overt syntax. Finally, I want to mention in passing a further problem with intervention as understood in Beck and Kim 1997 (which is the same as Beck 1996a), noted by Kim (2003): as Beck and Kim’s discussion of whscrambling in Korean shows, scrambling must be semantically e¤ective (otherwise, intervention could not be called upon). This stands in stark contrast to the received wisdom that scrambling is semantically vacuous and involves radical reconstruction (Saito 1989). Here the facts for Korean seem to be duplicated in Japanese, the language Saito examines (see also Lee and Tomioka 2000). Beyond theoretical concerns, the analysis encounters some empirical problems as well, which will play a major role in formulating an alternative. For example, Beck (1996a, 17) addresses some predictions of QUIB/ MQSC; one is that because of the barrier postulated by these conditions, elements should not be able to occur in the restriction of a (negative) quantiﬁer. Among her supportive data is (11). (11) *Wen hat kein Pferd, das wer gefu¨ttert hat, gebissen? whom has no horse that who fed has bitten ‘Whom did no horse that who fed bite?’ What is puzzling here is that none of the imaginable positive counterparts, without a negative quantiﬁer inducing a barrier (or any of the other quantiﬁers that work for the QUIB/MQSC; see below for more), is any good either (see (12)).10 Beck’s formulation of intervention might have something to say about the ill-formedness of (11), but it does not relate at all to (12). (12) a. *Wen hat ein Pferd, das wer gefu¨ttert hat, gebissen? whom has a horse that who fed has bitten ‘Whom did a horse that who fed bite?’ b. *Wen hat das Pferd, das wer gefu¨ttert hat, gebissen? whom has the horse that who fed has bitten ‘Whom did the horse that who fed bite?’

254

Kleanthes K. Grohmann

c. *Wen haben viele Pferde, die wer gefu¨ttert hat, gebissen? whom have many horses that who fed has bitten ‘Whom did many horses that who fed bite?’ This said, since more narrowly deﬁned multiple wh-questions constitute the focus of this chapter, I will simply disregard this puzzle (and the potential mismatch between what intervention is supposed to be ruling out and how that ties in with grammatical structures). Rather, my empirical concerns about intervention relate to those structures that look very much like (2a) versus (2b) and therefore might be expected to receive an identical explanation, but fail to fall under (3)–(5) or any reformulation thereof. In other words, the intervention e¤ect investigated by Beck is more general than she claims it is, even for a subset of her structures. Moreover, the alternative account presented below does follow from other, independent properties. 10.2.2 Introducing Topicalizability Let me present the full paradigm relevant for the subsequent discussion. As Beck (1996a) notes, many elements may occur in between two whphrases or follow them. (The examples presented in this section are variations of constructions ﬁrst discussed in Grohmann 2000a, 2003a.) (13) a. Wer hat Maria/seine Freundin/einen Porsche wo geku¨sst? who has Maria/his girlfriend/a Porsche where kissed ‘Who kissed Maria/his girlfriend/a Porsche where?’ b. Wer hat wo Maria/seine Freundin/einen Porsche geku¨sst? This is not so for all quantiﬁcational phrases, though. Considering both ordering options for quantiﬁcational elements, we can observe a stark contrast (see (2)): (some) universal quantiﬁers are able to appear in between the two wh-elements, while negative ones are not. If all wh-elements really must end up in CP at one point (through wh-movement at LF, for example), Beck’s barrier approach would provide one possible account for these facts. Note that nothing rules out the co-occurrence of two whelements and a negative quantiﬁer: if the apparently intervening quantiﬁer follows all wh-elements, the question is well formed, as the examples in (14) and (15) illustrate. (14) a. Wer hat alle Bu¨cher wo gekauft? who has all books where bought ‘Who bought all (the) books where?’ b. Wer hat wo alle Bu¨cher gekauft?

Top Issues in Questions

255

(15) a. *Wer hat kein Buch wo gekauft? who has no book where bought ‘Who bought no book where?’ b. Wer hat wo kein Buch gekauft? However, not only negative quantiﬁers are prohibited from intervening, nor is the class of possible interveners restricted to universal quantiﬁers.11 (16) a. Wer hat viele/die meisten/mehr als drei Bu¨cher wo gekauft? who has many/the most/more than three books where bought ‘Who bought many/the most/more than three books where?’ b. Wer hat wo viele/die meisten/mehr als drei Bu¨cher gekauft? And again, if the quantiﬁers follow the wh-items, everything is all right. (17) a. *Wer hat wenige/ho¨chstens drei/weniger als drei Bu¨cher wo who has few/at-most three/fewer than three books where gekauft? bought ‘Who bought few/at most three/fewer than three books where?’ b. Wer hat wo wenige/ho¨chstens drei/weniger als drei Bu¨cher gekauft? These contrasts seem to allow for the following generalization: (18) Quantiﬁer-Intervention Generalization (Grohmann 2003a, 116) Monotone-increasing quantiﬁers may appear in between two whphrases, while monotone-decreasing quantiﬁers may not appear in between two wh-phrases. Now, this generalization seems to ﬁt the characterizations of intervention presented earlier rather well—but does that make it any more plausible? I do not think so (see also note 11). Rather, I believe, this state of a¤airs begs to be explained by something deeper, preferably by being tied to other properties of the grammar. Let me introduce and investigate further one such property: topicalizability, the property of being able to serve as a topic. As it turns out, there is evidence for the assumption that (monotone-) decreasing quantiﬁers are nontopicalizable. Consider the contrast in (19)–(20), where capitals indicate focus.12 (19) a. VIELE Bu¨cher hat Peter gestern gelesen (, nicht ALLE) many books has Peter yesterday read not all ‘MANY books Peter read yesterday (, not ALL).’

256

Kleanthes K. Grohmann

b. Viele Bu¨cher hat Peter gestern gelesen. (Er ist ﬂeissig.) many books has Peter yesterday read he is industrious ‘Many books, Peter read yesterday. (He is a hard worker.)’ (20) a.

WENIGE Bu¨cher hat Peter gestern gelesen (, nicht VIELE) few books has Peter yesterday read not many ‘FEW books Peter read yesterday (, not MANY).’ b. *Wenige Bu¨cher hat Peter gestern gelesen. (Er ist faul.) few books has Peter yesterday read he is lazy *‘Few books, Peter read yesterday. (He is lazy.)’

Next, consider the following questions (see also note 12): (21) a. b. c. d.

What happened? What did Peter do yesterday? Did Peter read few/many books yesterday? How many books did Peter read yesterday?

For a felicitous answer to (21a–b) in German, the new information appears in what we may call a default focus position, where the type of focus is information focus (as summarized in E´. Kiss 1998) and the default position is rather low in the structure (see Meinunger 2000 for useful discussion and references). In replying to (21b), neither construction in (19) or (20) can be used; the only word order that results in a felicitous answer is that in (22). (22) a. Peter hat gestern viele Bu¨cher gelesen Peter has yesterday many books read ‘Peter read many books yesterday.’ b. Peter hat gestern wenige Bu¨cher gelesen Peter has yesterday few books read ‘Peter read few books yesterday.’ However, (19a) and (20a) can be used as felicitous answers to questions like (21c–d), distinguishing them from the context in note 12, where the answer contains a contrastive element as part of the new information. (19b) and (20b) are simple topic structures: there is no contrastive stress, and the fronted elements can only pick up a referent already established in the discourse; both are infelicitous.13 Let me propose a ﬁnal argument against any intervention-based account of the ‘‘Beck e¤ect’’ (which we can now understand as the phenomenon of speciﬁc word order incompatibilities between nontopicalizable elements and the two wh-phrases in German multiple wh-questions). We

Top Issues in Questions

257

can construct paradigms that are identical to the ones discussed above, except that they do not contain any quantiﬁcational element. If it can be shown that here, too, we ﬁnd a topicalizability e¤ect, (any type of ) intervention is ruled out for good. 10.2.3 Adverbial Incompatibilities and Topicalizability As noted in Citko and Grohmann 2001, certain adverbs may not occur in between two wh-elements either, and it can be shown that the property they all share is nontopicalizability—the impossibility of be(com)ing topicalized. Obviously, this could hardly be captured under a barrierinspired, quantiﬁer-induced intervention approach, even leaving aside LF issues. I will present the relevant facts in this section and address an interesting twist as well. For the sake of presentation, I conﬁne myself to manner adverbs, the class of adverbial modiﬁers that arguably sit very low in the clause structure. In German declarative clauses, these must follow the direct object, unless it is focused. (23) a.

Peter hat das Buch gerne/komplett/sorgfa¨ltig gelesen Peter has the book with-pleasure/completely/carefully read ‘Peter read the book with pleasure/completely/carefully.’ b. *Peter hat gerne/komplett/sorgfa¨ltig das Buch gelesen c. Peter hat gerne/komplett/sorgfa¨ltig das BUCH gelesen

In combination with multiple wh-phrases, these adverbs may not appear between them, regardless of their order. (Recall that German does not show any obvious superiority e¤ects in monoclausal interrogative structures.) (24) a.

Wer hat was gerne/komplett/sorgfa¨ltig gelesen? who has what with-pleasure/completely/carefully read ‘Who read what with pleasure/completely/carefully?’ b. *Wer hat gerne/komplett/sorgfa¨ltig was gelesen? c. Was hat wer gerne/komplett/sorgfa¨ltig gelesen? d. *Was hat gerne/komplett/sorgfa¨ltig wer gelesen?

Moreover, such adverbs may not be fronted, in contrast to sentence adverbs like gestern ‘yesterday’, adverbs implying a propositional argument such as vielleicht ‘perhaps’, or epistemic modal adverbs like wahrscheinlich ‘probably’. Since the latter cases exhibit neutral intonation and stress, I understand this kind of fronting as topicalization (taken throughout to be a syntactic operation).

258

Kleanthes K. Grohmann

(25) a. *Gerne/Komplett/Sorgfa¨ltig hat der Peter das Buch with-pleasure/completely/carefully has the Peter the book gelesen read *‘With pleasure/Completely/Carefully, Peter read the book.’ b. Wahrscheinlich/Vielleicht/Gestern hat der Peter das Buch probably/perhaps/yesterday has the Peter the book gelesen read ‘Probably/Perhaps/Yesterday, Peter read the book.’ As it turns out, however, such prima facie nontopicalizable adverbs may be fronted if they are contrastively stressed. In these cases, we would expect them also to be legitimate in between two wh-elements. And indeed this expectation is borne out. When any of these adverbs is fronted through focalization, it may also appear in between the two wh-phrases (see also the data in note 13). ¨ LTIG hat der Peter das Buch (26) a. GERNE/KOMPLETT/SORGFA with-pleasure/completely/carefully has the Peter the book gelesen read ‘WITH PLEASURE/COMPLETELY/CAREFULLY, Peter read the book.’ ¨ LTIG was gelesen? b. Wer hat GERNE/KOMPLETT/SORGFA who has with-pleasure/completely/carefully what read ‘Who read what WITH PLEASURE/COMPLETELY/ CAREFULLY?’ The same contrasts can also be found in embedded contexts, thus ruling out that this is a matrix property possibly related to the verb-second constraint (a traditional indication of topicalization). (27) Maria hat gesagt, . . . Maria has said a. *dass gerne/komplett/sorgfa¨ltig der Peter das Buch that with-pleasure/completely/carefully the Peter the book gelesen hat read has ¨ LTIG der Peter das b. dass GERNE/KOMPLETT/SORGFA Buch gelesen hat

Top Issues in Questions

259

*‘Maria said that with pleasure/completely/carefully, Peter read the book.’ (28) Maria hat gefragt, . . . Maria has asked a. *wer gerne/komplett/sorgfa¨ltig was gelesen hat who with-pleasure/completely/carefully what read has ¨ LTIG was gelesen hat b. wer GERNE/KOMPLETT/SORGFA *‘Maria asked who with pleasure/completely/carefully read what.’ In this section, I have argued that the phenomenon of (non)topicalizability is not only more general than a QUIB approach in terms of intervention, it also follows from more general (and independently justiﬁed) principles of grammar, and—‘‘crucially’’—may involve deﬁnitely nonquantiﬁcational material (such as adverbs). A Beck-inspired approach would be more than hard pressed to capture these data—under any version of intervention presented in section 10.2.1. It now looks like interveners are ﬁnally a nonissue, at least as far as German multiple wh-questions are concerned. Previous, interventiondriven approaches (starting with Beck 1996a) rely on their being quantiﬁcational, however that is to be deﬁned (universally or on the basis of monotonicity). This simply does not cut the pie right. The data in this section and the last suggest that the class of interveners—which block binding of a trace created by LF movement or one’s favorite interpretation of ROLF—has one more property, which does not depend on interaction with any other material (such as those traces purportedly created at LF): they cannot be topicalized. For this, their quantiﬁcational properties do not matter at all.14 Let me sum up the anti-intervention discussion and put any such notion to rest (including the revision provided by (18)) with the Topicalizability Generalization. (29) Topicalizability Generalization Only topicalizable elements may appear in between two wh-phrases; nontopicalizable elements may only follow them (in German multiple wh-questions). Admittedly, generalization (29) is only that, a generalization, with the proviso that it is possibly restricted to German monoclausal questions with two wh-elements—but it is one that works, unlike intervention. As a generalization, (29) has no explanatory power. This is what section

260

Kleanthes K. Grohmann

10.3 is concerned with: ﬂeshing out some more interesting consequences of a topicalizability approach to the present subject matter and removing a stipulated axiom, be it intervention or topicalizability. 10.3

Topicalization: On the Complexity of ‘‘Move Wh’’

This section expands on an important consequence of the topicalizability account of apparent intervention phenomena in German multiple wh-questions. I ﬁrst sketch an analysis of (the left periphery of ) German multiple wh-questions in terms of multiple topicalization. I then integrate further phenomena that might not at ﬁrst glance be related, but seem somehow also to be predicted by the multitopicalization approach. And throughout, I introduce the notion of a wh-topic, to be solidiﬁed in section 10.4. 10.3.1 An Alternative Derivation of Multiple Wh-Questions in German Is there anything interesting for the analysis of multiple wh-questions in German that would relate to the conclusion that apparent interveners are nontopicalizable? I want to argue for a resounding yes. After all, if real, such nontopicalizability would be a striking property. To begin with, the purported role of topicalizability looks rather obscure under a traditional analysis of multiple wh-questions in German, whose derivation would be analogous to that of multiple wh-questions in English: one wh-phrase fronts into the left periphery overtly, the other stays in situ (with ‘‘?P’’ ¼ VP; see von Stechow and Sternefeld 1988). (30) [CP WH1 C 0 . . . [?P . . . WH2 . . . ]] Working from a rough structure like (30), I want to propose a radical alternative. After all, how could we tell from (30) that the intervener, an illegitimately topicalized nontopicalizable element, is indeed in a topic position? What is a topic position, anyway? Let me brieﬂy lay out some background assumptions and sketch my analysis (drawing from the more detailed expositions in Grohmann 1998, 2000a, 2003a). First of all, topicalization as understood here is an operation in the overt component; it is the result of movement (some form of the rule Move wh from Chomsky 1977), targeting a position like Spec,CP—or rather, within an articulated Comp, the position explicitly identiﬁed by Rizzi (1997) and others as Spec,TopP. That is, minimally (building on (30)), we are dealing with something like (31).

Top Issues in Questions

261

(31) [CP WH1 . . . [TopP XP . . . [?P . . . WH2 . . . ]]] (31) is then our starting point for the structures at hand. The potential intervener—or rather, (non)topicalizable constituent—is signaled by XP: as a topic, it must move to Spec,TopP, and if it is nontopicalizable, it cannot move there. Our ﬁrst stab at formulating a ban thus yields (32). (32) *[CP WH1 . . . [TopP XPTOP . . . [?P . . . WH2 . . . ]]] Now the 64,000-dollar question is how we can ensure that the structures discussed above actually do involve some XP in a purported topic projection. In other words, can we show that what appears in between two wh-phrases in German is by necessity in a syntactic topic position? Working with standard assumptions as in (30), the answer is clearly no—and the whole endeavor is jeopardized. But with a simple suggestion ﬁrst advanced in Grohmann 1998, we get a positive answer instead. Assume that WH2 does not sit so low in the clause structure (‘‘in situ’’). Assume rather that, for some reason, WH2 has moved quite high, into the left periphery (i.e., our C domain), in the overt syntax. Assume further that this movement of WH2 also targets Spec,TopP. And ﬁnally, assume that in addition to WH2 and XP, WH1 occupies Spec,TopP and that fronted topics like to stick together. We now have a potentially more interesting picture. (33) C 0 [TopP WH1 XP WH2 Top 0 [?P . . . ]] The line of the argument is clear: if both wh-phrases are topics and if they can sandwich only topics, then only topics may appear in between them. In other words, when we see any structure involving two wh-phrases, we know one thing for sure: whatever sits in between the two sits in Spec,TopP. Problem solved. As nice as this argument may look, it carries a big if—and the big if consists at least of the number of assume thats, rephrased as follows: (34) a. In German multiple wh-questions, WH1 overtly moves to Spec,TopP. b. In German multiple wh-questions, WH2 overtly moves to Spec,TopP. c. Any XP occurring in between WH1 and WH2 must occupy Spec,TopP. What I have done in previous work is motivate (34a–c) with some variation.15 In the interest of space, I will conﬁne myself to a brief exposition.

262

Kleanthes K. Grohmann

The speciﬁc details arguably play only a minor role, so that we can basically consider (35) to contain the most salient ingredients of the alternative analysis. (35) Deus ex machina (Grohmann 1998, 2000a, 2003a) In German monoclausal multiple wh-questions, all wh-phrases move overtly to CP, driven by properties such that the landing sites sandwich only topics. Obviously, the argumentation I have developed elsewhere does not involve some deus ex machina, but derives the desiderata on mostly empirical grounds (see the works cited). We can then just adopt (35) as make-belief in the absence of a more elaborate discussion in the present context. The single most important aspect of the derivation involved, however, is the topichood of both wh-phrases in a German multiple whquestion; I return to this in more detail below (section 10.4.1), and I also address the notion of a wh-topic from a crosslinguistic perspective (section 10.4.3). Given the explosion of CP in (35) qua articulated C layer as in (36b) or ¯any version thereof, and the purported landing sites for the overtly A moved wh-phrases of (35) indicated in (37) (Grohmann 2003a, 114), any material appearing between the two wh-phrases must target TopP. This particular analysis (as developed in Grohmann 2000a, chap. 5) is a modiﬁed Q-based analysis of wh-question formation, what I call elsewhere the ‘‘Hagstrom-Bosˇkovic´ approach’’ (see also note 15). More in line with what I have said above, and will say below, about (so far, purported) wh-topics, one might actually want to adopt an even further reﬁned structure of the C domain, such as (36c).16 Again, I believe that the details of the proverbial devil, important though they certainly are, are not critical for present purposes. (36) a. ForceP > TopP* > FocP > TopP* > FinP > IP b. CP > TopP* > FocP > TopP* > FP > IP c. CP > TopP* > FocP > FP > IP

(Rizzi 1997, 297) (Grohmann 2000a, 76) (additional reﬁnement)

Top Issues in Questions

263

(37)

Under this analysis, questions are licensed by a question morpheme [Q] that has to end up in C, in a position high enough to take scope over the entire structure. As Hagstrom (1998) lays out in detail, [Q] may move on its own, from a high base-generated position (here, F 0 ), or it may tag along with the lowest wh-phrase (here, WH2 ), its alternative locus of generation. Bosˇkovic´ (2001) exploits this system to derive pair-list and single-pair readings, slightly revising Hagstrom’s original mechanism, which is further reﬁned in Grohmann 2003a. As argued there, a derivation as sketched in (37) accounts not only for the topicalizability issues discussed so far, but also for the peculiar behavior of German multiple wh-questions with respect to the readings they allow (or require). For the latter account to work, nothing hinges on the exact (labels of the) positions, as long as the relational properties of (37) remain. In other words,

264

Kleanthes K. Grohmann

we would not lose anything from the explanatory power of that analysis if the landing sites of WH1 and WH2 were in fact the same (type of ) projection, Spec,TopP (see note 15). Note further that pursuing an alternative to the observed word order restrictions as hastily sketched above does not depend at all on the speciﬁc generalization, that is, on either universal-based intervention as in (3) or monotonicity-inspired topicalizability concerns as in (18)— empirical shortcomings of either aside. It does not matter how we deﬁne inherent properties of illegitimate in-betweeners; what matters is that any constituent XP that wants to follow one wh-phrase and precede another (WH1 , WH2 ) must be a topic. If it cannot be a topic, for reasons that go beyond this investigation, it cannot appear in-between, but must strictly follow. ¯ -Movement 10.3.2 Further Evidence for Multiple A Next, I want to consider additional evidence supporting the assumption that even the lower wh-phrase WH2 moves overtly quite high up in the structure. Leaving aside the details of how topicality and interrogativity are to be married (section 10.4), the proposed step must be an in¯ -movement; the landing site, whether stance of what used to be called A ¯Spec,TopP or any of the other candidates in (36)–(37), is arguably an A position (terminology I continue to use for purely expository reasons). ¯ -move overtly into an articulated C doIf all wh-elements do indeed A main in German, we would expect them to behave syntactically in ways ¯that go beyond word order restrictions or intervention e¤ects, such as A binding. As it turns out, such e¤ects are notoriously di‰cult to show for German. But interestingly, we can ﬁnd some—and while these do not ¯ -movement proper, like wh-movement in English, necessarily suggest A they do suggest some sort of movement, even (and especially) of the lower wh-element. The following discussion is taken from Grohmann 2003a, but the data are considerably enriched: whereas the arguments provided there suggested that even the lower wh-element, WH2 , moves somewhere, there is now nearly conclusive evidence that it moves into the C domain, ¯ -element. as an A ¯Consider parasitic gaps, for example, which are licensed by overtly A moved elements only—at least in English (pace Nissenbaum 2000; see, e.g., Hornstein 2001 and Nunes 2004 for recent discussion and ample references about this core property of parasitic gaps and others). As it turns out, the lower wh-element in German may license a parasitic gap ¯ -movement. ( pg), thus at ﬁrst glance suggesting overt A

Top Issues in Questions

265

(38) a. Wer hat weni ohne pgi zu kennen eingeladen? who has who without to know invited ‘Who invited whom without knowing *(him)?’ b. Wer hat wasi ohne pgi gelesen zu haben rezensiert? who has what without read to have reviewed ‘Who reviewed what without reading *(it)?’ Before we can use parasitic gaps as a bona ﬁde argument in favor of overt displacement of the lower wh-element into the C domain, it has to be shown that German indeed has parasitic gaps analogous to those of English.17 But since wh-phrases do not scramble in German (unlike, say, in Japanese; see Grewendorf and Sabel 1999), the landing site of the lower wh-phrase cannot be the typical scrambling position. In e¤ect, the licensing of parasitic gaps by the lower wh-phrase can indeed be integrated into the analysis that the landing site is within the C domain, and even as things stand now concerning the status of parasitic gaps in German, data such as (38a–b) do not provide a counterargument to my proposal—quite the opposite, in fact, as the ensuing discussion shows. ¯ -licenser of the parasitic gap has to c-command it Note ﬁrst that the A in the overt syntax (Webelhuth 1992); thus, the adjunct clause containing the parasitic gap cannot scramble over the licenser (or whatever analysis one wants to apply to German word order, clause structure, and phenomena like scrambling). (39) *Wer hat [ohne pgi gelesen zu haben]k wasi tk rezensiert? who has without read to have what reviewed *‘Who reviewed without reading (it) what?’ With this in mind, consider the following contrasts: (40) a.

Wann hat Peter wasi [ohne pgi gelesen zu haben] sorgfa¨ltig when has Peter what without read to have carefully rezensiert? reviewed ‘When did Peter carefully review what without reading *(it)?’ b. *Wann hat Peter [ohne pgi gelesen zu haben] was sorgfa¨ltig rezensiert? c. *Wann hat Peter sorgfa¨ltig was [ohne pgi gelesen zu haben] rezensiert?

¯ -character noted by Webelhuth, Mahajan Regarding the mixed A/A (1990) o¤ers an interesting solution for German: prior to the step of

266

Kleanthes K. Grohmann

¯ -movement licensing the parasitic gap, the licenser undergoes AA movement (from the landing site of which it can license anaphors, one of the issues of contention). I adopt this basic line of analysis, which, interestingly, should also work for anyone’s take on headedness of the German VP (see, e.g., Sabel 1996, 37n26). In other words, regardless of the particular derivational operations and structural positions involved, ¯ -position and contrasts the parasitic gap is eventually licensed from an A with lower positions (scrambled noninterrogatives or multiple wh-phrases as traditionally conceived). This approach makes a number of testable predictions. One concerns the contrasts in (40). The grammaticality of (40a) would be expected since we have seen in section 10.2.3 that an adverb like sorgfa¨ltig ‘carefully’ must follow all wh-phrases. This said, the ungrammaticality of (40b) is equally expected since the adjunct containing the parasitic gap is scrambled over its licenser, the lower wh-phrase was ‘what’. Interestingly, now, (40c) is bad: as the parasitic gap seems to be correctly positioned with respect to its licenser, one might hypothesize that what rules this structure out is the relative ordering of the adverb and the lower wh-phrase, for the by now well-known reasons. A similar line of reasoning can be constructed for a di¤erent type of adverb. Frey (2000) identiﬁes the adverb vermutlich ‘presumably’, among others, as a marker for sentence topics: what stands to its left must be a topic. Consider some relevant data: (41) a. Wer hat vermutlich was rezensiert? who has presumably what reviewed ‘Who presumably reviewed what?’ b. Wer hat was vermutlich rezensiert? Both ordering relations between adverb and lower wh-phrase are possible—at ﬁrst glance unexpectedly under current assumptions, since, as a topic marker, vermutlich ‘presumably’ should follow was ‘what’. (One property of topic markers is that they do not allow the topic to follow.) Admittedly, Frey and I disagree on the implication that topic markers (and his discussion of sentence topics in general) carry for the syntax. He identiﬁes the notion ‘‘sentence topic’’ with a Mittelfeld position, that is, within the German middle ﬁeld (here, roughly IP-internal). But it seems to me that unless the analysis hastily sketched above is ﬂeshed out in full detail, we can remain agnostic about exactly which projection in the structure is identiﬁed here. Rather, I would like to focus in on the contrast between (41) and (42).

Top Issues in Questions

267

(42) a. *Wer hat vermutlich wasi rezensiert [ohne pgi gelesen zu who has presumably what reviewed without read to haben]? have ‘Who presumably reviewed what without reading *(it)?’ b. Wer hat wasi vermutlich rezensiert [ohne pgi gelesen zu haben]? While the lower wh-phrase may follow the topic-indicating adverb, it may not license a parasitic gap; this it can only do from the preadverbial position. (If the sentence receives any interpretation, the adverb in (42a) can only be understood as modifying was ‘what’.) This suggests that (41) contains two quite di¤erent structures, one in which the wh-phrase has moved to a left-peripheral topic position, and one in which it has not. So far, I have not considered the latter possibility at all. In fact, in previous work I have always argued that multiple whfronting is obligatory in German. With data such as the above at hand, however, this conclusion might have to be revised. At this point, I do not see any di‰culties with such a revision; on the contrary, it might come in handy. Consider, for example, the following examples: (43) a.

Wer hat was letztes Jahr jedenfalls zum Glu¨ck rezensiert? who has what last year at-least for-the luck reviewed ‘Last year at least, who fortunately reviewed what?’ b. *Wer hat letztes Jahr jedenfalls was zum Glu¨ck rezensiert? c. Wer hat letztes Jahr jedenfalls zum Glu¨ck was rezensiert? d. *Wer hat was zum Glu¨ck letztes Jahr jedenfalls rezensiert?

(44) a.

Wer hat was i letztes Jahr jedenfalls zum Glu¨ck rezensiert who has what last year at-least for-the luck reviewed [ohne pgi gelesen zu haben]? without read to have ‘Last year at least, who fortunately reviewed what without reading *(it)?’ b. *Wer hat letztes Jahr jedenfalls zum Glu¨ck wasi rezensiert [ohne pgi gelesen zu haben]?

Drawing on Frey’s work again, we see that the structures in (43)–(44) contain two relevant elements: the expression zum Glu¨ck ‘fortunately’ serves a topic-marking purpose similar to vermutlich ‘presumably’, while the expression [NP jedenfalls ‘at least’] must refer to some aforementioned NP; that is, it must be a topic (and as such cannot follow the topic marker, as in (43d)). The clear contrast in the complex examples here tells

268

Kleanthes K. Grohmann

us one thing for certain: the lower wh-phrase is in a higher position in the (a) examples than in the others. This conclusion su‰ces for the purpose of this section: suggestive evidence that the lower wh-phrase moves overtly ¯ -character exhibiting properties into a high left-peripheral position with A reminiscent of topics. If this is a fair and safe conclusion, the deus ex machina from (35) is ﬁnally demystiﬁed, and so are the big assumptions leading to (34). The parasitic gap data thus lend support to the hypothesis of multiple whtopicalization endorsed here from a syntactic point of view.18 Regarding the purported optionality, I leave for further research additional properties that a nontopicalized WH2 may carry. What matters is that we now have the tools to deﬁne a topicalized WH2 : for example, only a moved WH2 licenses parasitic gaps, thus lending plausible support to the topic analyses proposed here and in previous work.19 This is the type of lower wh-phrase that I concentrated on in the previous sections and will return to below. 10.4

Topics: Interrogative Topicality across Languages

Let me ﬁnally address the dimension that topics give the structures in question. If, so the proposal goes, monoclausal multiple wh-questions in German are in fact multiple topicalizations of wh-phrases (as well as noninterrogative constituents, potentially), we are basically dealing with whtopics. This section explores this notion, reﬁnes its meaning, and presents crosslinguistic support for it. 10.4.1 Discourse-Restricted Quantiﬁcation Side by side with the evidence suggested above, a further type of argument for multiple wh-displacement into the C domain in German comes from discourse restrictions. In this section, I present the relevant data and introduce a means to account for them, Discourse-Restricted Quantiﬁcation (Grohmann 1998). Compare, for example, the following two discourse contexts and the (in)felicity (marked with ‘‘a’’) of a multiple wh-question in English and German: (45) Context I: A man comes to a newsstand and sees three people leaving in di¤erent directions, all ﬁddling with their purchases, obviously excited. The man asks the newsagent whether he sold anything interesting to those three, and the newsagent replies, ‘‘I

Top Issues in Questions

269

can’t believe it! Within two minutes, I just sold the Anarchist newsletter, a Fascist magazine, and the Christian church news.’’ a. Who bought what? b. Wer hat was gekauft? who has what bought ‘Who bought what?’ (46) Context II: A jeweler comes home to his wife for lunch and exclaims excitedly, ‘‘I had a great morning, honey! I sold a platinum watch, a gold necklace, and a titanium wedding band.’’ a. Who bought what? b. aWer hat was gekauft? who has what bought ‘Who bought what?’ As these cases show, German allows a felicitous binary wh-question only if the set of individuals is part of the common ground between speaker and hearer, introduced in the discourse. The English example in (46a) is a perfectly reasonable question, and a potential answer would be something like (47).20 (47) A businessman bought the platinum watch, an old lady bought the gold necklace, and a young groom bought the titanium wedding band. This felicity requirement is captured by a condition I called, for better or worse, ‘‘Discourse-Restricted Quantiﬁcation’’ in earlier work (Grohmann 1998). (48) Discourse-Restricted Quantiﬁcation (DRQ; Grohmann 2000a, 269) Questions involving two wh-expressions are well formed if the value of both wh-expressions is determined by the context; determination of values is satisﬁed by providing a set of at least two possible referents in the discourse. In other words, DRQ requires the referents of all wh-phrases in a multiple wh-question to be salient to a certain degree. As the above contrasts show, it applies strictly to German, but not to English; this is the notion of topicality understood throughout (see note 12). What DRQ essentially expresses is the intuition that all wh-phrases in a German multiple wh-question must be D(iscourse)-linked (possibly, but not necessarily in the sense of Pesetsky 1987). While a wh-phrase in an information question represents a novel set of individuals (new

270

Kleanthes K. Grohmann

information), the set of individuals represented by D-linked wh-phrases is to some degree known (old information). The term DRQ can be used to distinguish a condition of D-linking on all wh-phrases (in German), while D-linking simply applies to those elements commonly identiﬁed as such.21 I present one type of interpretation of DRQ next, and another in section 10.4.3. 10.4.2 Absence of Multiple Information Questions Rizzi (1978) notes a curious property of Italian (see also McCloskey 1979 for a similar claim about Irish): this language does not allow multiple whquestions at all. (49) a. *Chi ha scritto che cosa? who has written what ‘Who wrote what?’ b. *Chi e` partito quando? who is left when ‘Who left when?’ An intriguing explanation for this ban has been o¤ered by Calabrese (1984, 1992): Italian does not have the second focus position required to license the additional wh-element. Calabrese’s reasoning goes as follows. A multiple wh-question, if not used as an echo or quiz question, asks for new information regarding all wh-elements; new information in a declarative sentence is construed with nondefault sentential stress—or focus. The type of focus relevant for a felicitous reply to a question is information (or presentational) focus, not identiﬁcational (or contrastive) focus (see the discussion in E´. Kiss 1998). In Italian, the position for information focus is unique, and so it follows that multiple wh-questions are ill formed: the multiple information requested cannot be realized in the answer. (50) a. *MARIO ha scritto una LETTERA Mario has written a letter ‘MARIO wrote a LETTER.’ b. *FRANCO e` partito alle CINQUE Franco is left at ﬁve ‘FRANCO left at FIVE O’CLOCK.’ This state of a¤airs is independent of interpretation in terms of singlepair and pair-list readings, as Calabrese shows (see Grohmann 2003a for discussion).

Top Issues in Questions

271

(51) aMARIO ha scritto una LETTERA, FRANCO un ARTICOLO, Mario has written a letter Franco an article CARLO un LIBRO . . . Carlo a book ‘MARIO has written a LETTER, FRANCO an ARTICLE, CARLO a BOOK . . . ’ (Calabrese 1984, 67) The intended interpretation of (51) is not contrastive, but information focus (cf. ‘What did people write, and who did it?’), hence the a. (Presumably, an utterance like (51) is felicitous and well formed in a context that focuses contrast.) As it turns out, we can detect a similar e¤ect in German: the analogue of (51) can only be used contrastively as well. It seems that German and Italian pattern alike in the relevant properties. In a sense, then, German multiple wh-questions are not really questions. However, German has the option of marking contrastiveness in two positions: by fronting the relevant element or by assigning it heavy stress in low position. Thus, if a question is formed that asks for contrastive information (in a syntactically well-formed way), a possible answer is also well formed syntactically (see, e.g., section 10.2.3).22 If German shares with Italian the absence of ‘‘real’’ (information) multiple wh-questions, the notion of wh-topic—or the analysis of movement to Spec,TopP, for that matter—loses some of its initial awkwardness. More evidence against such awkwardness follows directly. Let me just brieﬂy remark on and relativize earlier conclusions that the Italian connection might o¤er a clue about why some structures do not seem to involve movement to TopP of the lower wh-phrase (see especially the data in the discussion of parasitic gaps in section 10.3.2): maybe German does have real multiple wh-questions, but they are (possibly) rare and show di¤erent properties than those I have considered throughout. 10.4.3 The Wh/Clitic Connection There is a further link one can make between wh-elements and some notion of topicality. In this ﬁnal section, I will discuss how the phenomenon of clitic doubling in Bulgarian, as well as other languages, correlates with fronted wh-phrases, or rather vice versa. All in all, we can observe a close connection between topical wh-phrases and clitics—or, alluding to the work of Boeckx and Stjepanovic´ (2005), there is more to the ‘‘wh/clitic connection’’ than meets the eye.

272

Kleanthes K. Grohmann

Citko and Grohmann (2001) advance the bold thesis (expanded in Grohmann 2000a, 2003a) that German is like Bulgarian. The main parallel, they argue, is that both languages have the obligatory fronting requirement on all wh-phrases in the overt component, presented above. In an intriguing recent paper, Jaeger (2004) presents a curious set of data from Bulgarian that is much more in line with the view on German multiple wh-questions espoused here. Moreover, the analysis Jaeger suggests bears striking similarities to the one presented here. Let me summarize the data and suggest an interpretation that ﬁts nicely into the present set of assumptions, drawing also from joint work with Cedric Boeckx. The parallelism between German and Bulgarian multiple wh-question formation hypothesized elsewhere faces the problem that Bulgarian interrogative syntax has a well-known property that German lacks: all whitems are obligatorily fronted en bloc or adjacent to one another (Rudin 1988); that is to say, fronted wh-phrases may not be split up.23 Rudin already o¤ered some data pertaining to parentheticals (see Bosˇkovic´ 2002 for recent discussion). More relevant to the present issues, another type of constituent that may occur in the left periphery and is presumably the result of movement may not intrude either: topics. Richards (1997, 2001) presents data involving the interaction between topics and fronted wh-elements which suggest ﬁrst, that the two are not incompatible, and second, that topics must precede, rather than follow, the wh-constituent(s).24 (52) a. Ivan vcˇera kakvo kupi? Ivan yesterday what bought ‘What did Ivan buy yesterday?’ b. Vcˇera Ivan kakvo kupi? (Richards 1997, 111) (53) a. *Koj vcˇera kakvo kupi? who yesterday what bought ‘Who bought what yesterday?’ b. Vcˇera koj kakvo kupi? The same e¤ects can be found in Romanian (Ileana Comorovski, pers. comm.). What makes Romanian interesting is a fact ﬁrst observed by Dobrovie-Sorin (1990). This language allows clitic doubling, and as it turns out wh-phrases may be doubled by a clitic just in case they are Dlinked—in fact, D-linked wh-phrases must be doubled in this way, as the

Top Issues in Questions

273

contrast in (54) shows.25 (The particle pe occurs only in D-linked questions, for reasons that do not bear on the present issue.) (54) a. *Pe care (baiat) ai vazut? pe which boy have.2sg seen ‘Which one (Which boy) did you see?’ b. Pe care (baiat) l-ai vazut? pe which boy cl-have.2sg seen Non-D-linked wh-words, on the other hand, do not allow doubling at all. (55) a.

Cine ai vazut? who have.2sg seen ‘Who did you see?’ b. *Cine l-ai vazut? who cl-have.2sg seen

Picking up on the discussion in Billings and Rudin 1998, Jaeger (2004) now examines a curious property of (colloquial) Bulgarian: clitic doubling is allowed not only with overtly marked D-linked wh-elements, but with bare wh-words as well.26 Consider these contrasting structures: (56) a.

Kogo kakvo go iznenada? whom what cl.acc surprised ‘What surprised whom?’ b. ?*Kakvo kogo go iznenada?

An important initial observation is that the doubled wh-phrase precedes all others—in particular, a subject wh-phrase (see Jaeger 2004 for discussion). This ﬁts in well with the proposal developed above for German: clitic-doubled wh-phrases seem to underlie di¤erent syntactic operations than nondoubled ones. The particular operation one could envision here, with the proposal from German in mind, is topic movement. One result of the many Rizzi-inspired works on the left periphery, I take it, is that they have identiﬁed a crosslinguistic tendency that topics precede whitems if the two co-occur. The data in (52)–(53) suggest that this holds at least for Bulgarian. And indeed, Jaeger explores an account along the same lines. He introduces a deﬁnition of topicality that (in the spirit of note 12) I will not recreate. Su‰ce it to say, the two approaches are compatible in many respects. One discourse function of doubled wh-phrases in (colloquial) Bulgarian is illustrated by the following context and the question relating to it:

274

Kleanthes K. Grohmann

(57) Some of the most popular painters in town recently made portraits of a couple of John’s friends. John knows that each of his friends wanted to be painted by a particular artist, but he doesn’t know by which one. So he asks: Kogo koj go e narisuval? whom who cl.acc aux painted ‘Who painted who?’ To the best of my knowledge, the interpretation of the question would be analogous to something one could reconstruct in English as ‘For each of our friends, who painted him?’ or simply ‘Which friend was painted by which artist?’. (Jaeger o¤ers the latter as the sole translation, reﬂecting native speakers’ intuitions about the interpretation.) Another property of clitic-doubled multiple wh-questions in (colloquial) Bulgarian is that the doubled wh-phrase may also precede a Dlinked subject. (58) Kogo/Koj koja zˇena *(go) obicˇa? whom/who which woman cl.acc love.3sg ‘Who does which woman love?’ By now, it should come as no surprise that D-linked wh-phrases may also be doubled, as shown in (59). Here, one function of the doubling clitic seems to be disambiguation, since, as the pair in (60) shows, Dlinked wh-phrases may front in either order (in which case the interpretation is, of course, unambiguous, as suggested already in (59); see Jaeger 2003, 2004). (59) Koj ma¯zˇ koja zˇena (go) obicˇa? which man which woman cl.acc love.3sg Without clitic doubling: ‘Which man loves which woman?’ With clitic doubling: ‘Which woman loves which man?’ (60) a. Koj ma¯zˇ koja zˇena obicˇa? which man which woman love.3sg b. Koja zˇena koja ma¯zˇ obicˇa? which woman which man love.3sg ‘Which man loves which woman?’ or ‘Which woman loves which man?’ I leave the discussion of Bulgarian at this point. Clearly, much more can be—and needs to be—said, but that will have to wait for future research. The goal of this section was to draw a connection between wh-

Top Issues in Questions

275

elements that are in one way or another linked to the discourse (whtopics) and another discourse-heavy element of grammar, clitics (viz. clitic doubling; see also note 16). This wh/clitic connection has been noted in di¤erent contexts and might be worth exploring further. 10.5

Finding Closure and Looking Outward

To sum up the main proposal of this chapter: I have argued in favor of an analysis of German monoclausal multiple wh-questions that involves topic movement of the two wh-elements into the left periphery of the clause in the overt syntax. The starting point was the so-called intervention e¤ect (Beck 1996a), which I argued against on theoretical and empirical grounds. Especially the latter demonstrated that any attempt to capture the ‘‘Beck e¤ect’’ using some notion of intervention undergenerates, hence fails. The alternative developed here capitalizes on a property other than quantiﬁcation that illegitimate in-betweeners share: they are all nontopicalizable. Fleshing out this notion led to further empirical justiﬁcation of the validity not only of this generalization, but also of the theoretical proposal whereby both wh-phrases (as well as a legitimate in-betweener, for example) move to a left-peripheral topic position. One possible conclusion to draw from the analysis is to deny that multiple wh-questions exist in German in the ﬁrst place, by analogy with Italian, on the basis of ﬁner discourse properties distinguishing German from English, for example (Pesetsky 1987). But this is not the only necessary conclusion, and it need not be the ﬁnal word either. Finally, I connected the structures under discussion to clitic-doubling wh-phrases in (colloquial) Bulgarian. It has been argued that here too topicality in a wider sense plays a role (Jaeger 2004), something Cedric Boeckx and I have independently explored on the basis of languages such as Romanian (Boeckx and Grohmann 2004). How deep this wh/clitic connection goes remains to be seen. Let me ﬁnally comment on some possibilities. As noted above, the phenomenon of doubling wh-phrases with a clitic is not unique to Bulgarian, and Jaeger’s study does not really break new ground on an empirical level. Nor does the attempt to tie in doubled whphrases with some notion of topicality. What is new and highly intriguing, however, is the double explicitness of Jaeger’s study. First, in that it explicitly spells out such intuitions rather than mentioning them as a side remark. Bear in mind that the language under consideration is Bulgarian; while there has been a tremendous amount of

276

Kleanthes K. Grohmann

research on clitic strategies in Romance languages, including the distinction between clitic doubling and clitic left-dislocation, the body of research on related phenomena in Slavic is considerably smaller. The consensus seems to be that Slavic has one strategy only, which is commonly referred to as ‘‘clitic doubling.’’ The relation between these clitic strategies and information structure is as obvious as it is contested. While the two obviously go hand in hand, it is not necessarily clear which (type of ) strategy goes with which type of discourse function. In Romance, for example, there is a long-standing tradition of associating clitic leftdislocation with topicality (in the present sense), but not necessarily and/ or exclusively clitic doubling (see Villalba 2000 for a summary). Whichever way the dice fall, the phenomenon at hand is clearly one that shows topicality e¤ects, if not even restrictions akin to D-linking or DRQ. For what it is worth (see also, e.g., Leafgren 1997), I assume that Bulgarian clitic doubling is of the ‘‘right’’ variety to correlate with this kind of function. Future research will surely shed more light on the issue. It is in this sense that Jaeger’s results are doubly interesting. To the extent that the tentative yet explicit connection to my main proposal holds up to scrutiny, there would be independent support for some of the wilder stabs at German left-peripheral syntax taken above. This is the topic of ongoing research—and if nothing else, this hypothesis and some correlating predictions are empirically testable. Note, incidentally, that the parallelism between clitic doubling and whphrases, the type of wh/clitic connection I am interested in (for further interpretations, see Boeckx and Stjepanovic´ 2005; Poletto and Pollock 2002), is not restricted to the Balkan Sprachbund, and neither is the notion of a wh-topic, even in the explicitness presented above (for both Bulgarian and German). Another language that has been argued to contain wh-topics is (Modern) Greek. On the basis of a wide range of properties, Androulakis (1998) suggests that some wh-phrases are in fact wh-topics—namely, (at least) those that are doubled by a clitic. Her analysis accounts for the ambiguity she observes between a fronted wh-phrase and a lower quantiﬁer in regular questions, which disappears when the wh-phrase is doubled, but it also extends to data pertaining to parasitic gaps, weak crossover e¤ects, and others. Importantly, she also explores a notion of topicality, presuppositionality, and discourse properties of such wh-topics basically driven by the need to say something in addition to more or less well-established concepts like D-linking (see my remarks above).

Top Issues in Questions

277

And just to stress that the connection between wh-phrases and topicdiscourse properties is not the ﬁgment of limited imagination, let me mention Reglero’s (2003) and Lambova’s (2003) work on multiple whquestions in Basque and Bulgarian, respectively, which bears heavily on these issues, as well as Poletto and Pollock’s (2002) explorations in Romance (Northern Italian dialects, predominantly). The latter work in particular picks up on the wh/clitic connection made explicit in Boeckx and Stjepanovic´ 2005, though it also relates to present concerns. In the larger picture, then, we might add ‘‘topicality’’ to the connection and explore even more intricate patterns. While I do not wish to appear to take too functional a perspective on grammar, one wonders how narrow syntax is and how narrow syntax should be—or, more rhetorically, following up on the discussion in Chomsky 2000 and much subsequent work: what exactly is narrow syntax? Notes The material o¤ered here expands considerably on my presentation at the workshop ‘‘On Wh-Movement,’’ as well as talks held at the workshop ‘‘SyntaxSemantics Interface in the CP-Domain’’ (ZAS Berlin, March 6–8, 2002) and at the Linguistischer Arbeitskreis (University of Cologne, December 4, 2002). I am grateful to the audiences for feedback and suggestions, especially Howard Lasnik, Luigi Rizzi, and Peter Svenonius; to an anonymous reviewer; and to the editors of this volume, Lisa Cheng and Norbert Corver. For the preparation of this chapter and the work leading to it, Cedric Boeckx, Zˇeljko Bosˇkovic´, Norbert Hornstein, Florian Jaeger, Horst Lohnstein, Anna Roussou, Joachim Sabel, and Juan Uriagereka deserve special mention for invaluable help and discussion. 1. In Beck 1996b, she addresses the issue as well, but as far as I can see, the relevant chapter is only a slightly revised reproduction of Beck 1996a. An additional chapter on Korean, which later appeared as Beck and Kim 1997, serves as crosslinguistic support for her proposal. 2. Or even de Swart 1992, de Swart being one of the contenders for the title of ‘‘inventor of intervention’’ (though from a strictly semantic perspective). 3. Note that Beck also applies her analysis to intervention e¤ects observed in separation constructions of the form was . . . alles ‘what . . . all’, in the was fu¨r ‘what for’ split, and in wh-scope-marking structures, for example, which are not considered here. 4. ROLF is my acronym for a term that Beck does not employ; she introduces (3) as ‘‘a restriction on LF movement . . . informally expressed as in [(3)]’’ (Beck 1996a, 1). As we will see, however, the LF part of ROLF is ‘‘absolutely crucial’’ to Beck’s proposals (p. 18), so the acronym ROLF zeroes in on its importance.

278

Kleanthes K. Grohmann

5. Note that, while starring (6) on page 1, Beck (1996a, 3n3) then qualiﬁes the degraded status, subsequently marking such structures with two question marks. Since I am not interested in ﬁner factors of interpretation, I adopt the former notation throughout. The analysis proposed here can easily be adjusted to capture either state of a¤airs. 6. Nevertheless, in what follows I do criticize the ‘‘crucial’’ aspect of a separation between overt and covert components. Note that I do not address the reintroduction of barriers into syntax here. 7. Mathieu’s (1999, 2002) main concern is wh-in-situ constructions in French (see his work for further references). Doubts about Huang-inspired covert phrasal whmovement are, however, far more general and widespread (for explicit arguments and alternatives, see, e.g., Hornstein 1995; Hagstrom 1998; Sabel 1998; Simpson 2000; and see section 10.3.1 below for some discussion). 8. To complete the brief survey of the treatment of Beck-inspired intervention in the literature: Hagstrom (1998) also gives an informal deﬁnition of intervention qua ‘‘Beck e¤ect’’: (i) Beck e¤ect (Hagstrom 1998, 80) Q cannot be attracted over a certain set of elements (which includes at least negative items like niemand ‘nobody’). For a revised implementation of Hagstrom’s Q(uestion morpheme) that ﬁts into the current set of assumptions, see the exposition of the ‘‘Hagstrom-Bosˇkovic´ approach’’ in Grohmann 2000a, 2003a, drawing from Bosˇkovic´ 2001, and section 10.3.1 below. 9. Since Beck acknowledges that there is a huge body of literature on negative islands, with which her data set has little in common, it is surprising that so many linguists jump on the ‘‘Beck e¤ect’’ ship (beyond those already cited, starting with Beck and Kim (1997), a number of researchers have looked for crosslinguistic evidence for intervention: see, e.g., Lee and Tomioka 2000; Lipta´k 2001; Lotﬁ 2003). In fact, according to Beck herself, there is no such ship: In contrast to this range of papers [syntactic and semantic works dealing with negative islands], my proposal does not deal with wh-phrases that occur structurally above a negation at S-Structure. I am only concerned with LF movement. Accordingly, I am looking at a somewhat di¤erent set of data. . . . Thus, none of the existing proposals are prepared to deal with my data. The distinction between S-Structural and LF movement is absolutely crucial. (Beck 1996a, 18)

I stress my point again: regardless of whether some form of intervention exists in grammar (see note 3), it does not play a role in German multiple wh-questions. After the conceptual arguments against it laid out here, I will turn to empirical reasons. 10. This is not to say that such a structure may not be found (see, e.g., Lipta´k 2004 on Hungarian); however, Beck mentions that such multiple wh-structures within nominals are really what she has in mind. 11. Since Beck (1996a) goes into detail on the issue of negative quantiﬁers, which forms the basis for her more general constraints, I stress these particularly. This becomes relevant in (18) (see also Grohmann 1998, 2000a; Citko and Grohmann

Top Issues in Questions

279

2001; as well as Boeckx 1999 on French), where I suggest an alternative characterization of the class of impossible interveners in terms of monotonicity—one that Beck in fact considers and rejects in favor of universal quantiﬁers. This cannot be the whole story either, however. Space does not permit a more detailed discussion, but I think it su‰ces to say that the generalization arrived at below in the main text is just as untenable as intervention accounts (pace a possible between-thelines reading of the works just cited). Something else has to be said to tie together all observations, and the proposal in terms of topicalizability below arguably does. In her early rejection of (18), Beck (1996a, 33–34) mentions some misﬁts, such as the quantiﬁer jeder ‘every(one)’ (as well as fast jeder ‘almost every(one)’), which seems to exhibit the same intervening powers as monotone-decreasing (or downward-monotonic) quantiﬁers without being one. Some quantiﬁcational adverbials behave similarly (Beck exempliﬁes with zweimal ‘twice’, meistens ‘mostly’, oft ‘often’, and mindestens zweimal ‘at least twice’). I address the issue of adverbs in section 10.2.3, and it should become clear that once the correct approach to apparent intervention is spelled out, these misﬁts ﬁt. 12. Admittedly, the notion of topic—and its relevance to the syntactic component—is one of the hotbeds of research and contention. See the discussion in Grohmann 1998 and references cited as far as an integration for current purposes is concerned; see also Jaeger 2004 for a more recent treatment basically along those lines. Important at this point is that some notion of topic in terms of old or given information and possibly aboutness does play a role in the syntax, which I take to be some property driving displacement to a topic-devoted projection, for simplicity, Top(ic)P. With respect to topicalizability as mentioned in the text, it is interesting to note that the same observation seems to hold for English, possibly even for other languages as well. There seems to be disagreement among speakers, however, as to whether universal (monotone-increasing) quantiﬁers may indeed be topicalized, or whether fronting into the ﬁrst position of a matrix clause is necessarily focus driven. Beck’s (1996a) suggestion to test with left-dislocation structures is useful, but caution is advised; see Grohmann 2000a, 2003b for detailed discussion. The observation that monotone-increasing quantiﬁers may topicalize, but monotonedecreasing ones may not, was to my knowledge ﬁrst made explicit in this context in Citko and Grohmann 2001. Note that Grewendorf (2002, 77–79) comes to virtually the same conclusion. Much the same can be said for focus as for topic. At this point, it is only relevant that focalized elements displaced into the left periphery do not target TopP, and that all topics in this domain are adjacent to one another (viz. the notation TopP* in Rizzi 1997). As spelled out below, I also assume that a Foc(us)P may exist, which would then also host an overtly moved wh-phrase. For the following data, I employ contrastive focus rather than information focus as the response to a question (see E´. Kiss 1998 for discussion and literature, and also Brunetti 2003 for more recent work). The latter would work too, but it sounds odd for decreasing quantiﬁers (as it does in English, indicated in the translations in (i)). I have nothing to say about these facts and am not even sure that they are relevant.

280

Kleanthes K. Grohmann

(i) What did Peter read yesterday? ¨ CHER hat Peter gestern gelesen a. Viele BU many books has Peter yesterday read ‘Peter read many BOOKS yesterday.’ ¨ CHER hat Peter gestern gelesen b. ??Wenige BU few books has Peter yesterday read ??‘Peter read few BOOKS yesterday.’ 13. In this connection, it may be interesting to note that once contrastively stressed, the elements identiﬁed by Beck as interveners, monotone-decreasing quantiﬁers, suddenly become good (even clearer with the focus particles sogar/ nur ‘even/only’). (i) a. Wer hat (sogar) VIELE Bu¨cher wo gekauft? who has even many books where bought ‘Who bought (even) MANY books where?’ b. Wer hat (nur) WENIGE Bu¨cher wo gekauft? who has only few books where bought ‘Who bought (only) FEW books where?’ These questions carry quite a di¤erent meaning, however, and I will thus leave them aside. For a recent discussion of focus, see for example Brunetti 2003 and references cited. 14. At least not in a direct way: while inherent quantiﬁcational power might make them bad topics, what is of interest here is the fact that they make bad topics. This property, nontopicalizability, is what gives rise to apparent intervention. 15. The technical details of the previous works cited are as irrelevant here as their di¤erences (for clariﬁcation, see Grohmann 1998; 2000a, chap. 5; 2003a). In fact, one can envision a number of ways to capture the essentials of the present analysis, including the use of multiple speciﬁers or multiple projections (see also Richards 1997; Sabel 1998; Grewendorf 2001; Bosˇkovic´ 2002; and further literature cited in these works). 16. The present reﬁnement (36c) beneﬁts from Rizzi’s own reconsiderations (Rizzi 2004), inspired by much productive work on the left periphery over the past decade. My ‘‘FP’’ is in fact an adaptation of Uriagereka’s (1988) ‘‘Further Projection’’; see also Uriagereka 1995a,b and Boeckx and Stjepanovic´ 2005, among others, for a ﬁner understanding of a discourse-driven, ‘‘point-of-view’’ FP as endorsed here (relevant especially in section 10.4.3). In Grohmann 2000b, I present further amalgamation of the discourse properties of F and the topicality of clitics, just as I do here for the discourse properties of F and the topicality of wh-expressions. 17. This is not at all clear. See Sabel 1996 for discussion and references; but note that Grewendorf (2002, 234) also employs the idea that German wh-movement licenses parasitic gaps. See Webelhuth 1992 for early discussion of basic properties of the German parasitic gap construction, such as restrictions on the adjunct clause that contains it (e.g., that it must be a zu-inﬁnitive like those shown in (38a–b)).

Top Issues in Questions

281

¯18. In Grohmann 2003a, I present another potential indicator of (high) A movement of the lower wh-phrase based on (the absence of ) weak crossover e¤ects. 19. Sura´nyi (2004) also argues on the basis of parasitic gap licensing (among other things) that apparent wh-in-situ in Hungarian must involve movement of WH2 , at least to some position licensing the parasitic gap. As a reviewer notes, Sura´nyi shows that the lower wh-phrase occupies a focus position in Hungarian and therefore would not be expected to interact with topic-marking adverbials the way it does in German. 20. Peter Svenonius (pers. comm.) points out that this state of a¤airs carries over to English. While in a situation that conforms to Discourse-Restricted Quantiﬁcation (see (48)), such as (45), even the structure violating Superiority is well formed, it is not so in (46). In other words, English speakers may employ What did who buy? in a context that identiﬁes potential referents of both wh-phrases (see Baker 1970; Pesetsky 1987). This further conﬁrms Haider’s (2000) intuition that Superiority is not a uniform condition on the grammar, evident in much current research, but a complex epiphenomenon of the interplay of various grammatical properties. More needs to be said about which properties it actually concerns. Haider concentrates on syntax-semantics relation and parsing, and Bosˇkovic´ (2001) focuses on syntactic word order and interpretive constraints, while I investigate the role of discourse proper on the syntax (viz., D-linking/DRQ); others do so too, maybe with a little less syntax (Comorovski 1996), no formal syntax at all (e.g., the Sorting Key Hypothesis of Kuno 1982), or something much more in line with the present proposal (Jaeger 2004). 21. The latter refers, of course, to morphologically marked D-linking—for example, through the prototypical which-phrase. What prevents me from fully admitting that DRQ (or whatever it is that makes German multiple wh-questions special) is equivalent to D-linking is that as far as I can see there does not exist an explicit theory of D-linking. Pesetsky’s instrumental paper notwithstanding, very little has been done to formalize the intuitions expressed there (which are, partially, already found in Baker 1970 and other earlier work; see also Grohmann 1998); the work reported in Hirose 2003 is, despite the title ‘‘The Syntax of DLinking,’’ rather narrow. Perhaps the syntax sketched in Boeckx and Grohmann 2004 will give new impetus. In the absence of such a theory, I develop the analysis independently of D-linking, but with it, and its potential usefulness, in mind. (A note to this note. There does exist a theory of the opposite of D-linking, also ﬁrst made precise and labeled in Pesetsky 1987: the syntax and semantics of aggressively non-D-linked wh-phrases pursued in Den Dikken and Giannakidou 2002.) 22. However, German and Italian apparently di¤er with respect to D-linking. (i) *Quale ragazza ha dato un bacio a quale ragazzo? which girl has given a kiss to which boy ‘Which girl kissed which boy?’ (Calabrese 1984, 67) I leave this discrepancy between Italian and German, as well as D-linking and DRQ, for future research. But note that if a multiple question with two bare

282

Kleanthes K. Grohmann

wh-elements is D-linked, it might provide a clue about why the Superiority Condition (Chomsky 1973, though admittedly not topic of the present chapter) may be violated in German: the two wh-elements are actually D-linked, which—for reasons beyond the scope of this discussion—may circumvent Superiority (see Boeckx and Grohmann 2004 for suggestions). 23. A great number of theoretically di¤erent analyses have been proposed for Rudin’s ﬁndings, including a modernized version of her original clustering hypothesis (e.g., Grewendorf 2001; Sabel 2002), the implementation of multiple speciﬁers of C (e.g., Richards 1997; Pesetsky 2000), and the interaction of di¤erent C heads or at least di¤erent C-related features, such as topic, focus, and interrogative C proper as endorsed here (as in Citko 1998 and Lambova 2001, or the work of Zˇeljko Bosˇkovic´; see Bosˇkovic´ 2002 for a summary and ample references). 24. (53a) is also from Richards 1997; (53b), home-made and conﬁrmed. Lambova (2001) also discusses the interaction of topics and wh-phrases in Bulgarian and provides a similar example from a multiple wh-question (cited in Jaeger 2004). (i) Decata na cirk koj koga sˇte vodi? children.def to circus who when will take ‘Who will take the children to the circus when?’ 25. Boeckx and Grohmann (2004) capitalize on the relation between clitic doubling and D-linking (observed across languages, such as Hebrew and possibly Greek) by syntactically implementing Rullmann and Beck’s (1998) proposal regarding the structure of D-linked wh-phrases motivated on semantic grounds, which involves an additional D head carrying the required presuppositionality (another notion I leave open here). 26. Apparently, Billings and Rudin (1998) restrict this phenomenon to unaccusative verbs, as in (i). (i) Koj na kogo mu xaresva? who to whom cl.dat pleases ‘Who likes whom?’ Dimitrova-Vulchanova and Hellan (1999, 502¤.) also brieﬂy look at clitics doubling a fronted wh-phrase. They o¤er some relevant data involving a ditransitive verb. (ii) a. Knigata na kogo (*mu) ja dadoxa? book.def to whom cl.dat cl.acc gave.3pl ‘To whom did they give the book?’ b. Na kogo kakvo mu dadoxa? to whom what cl.dat gave.3pl ‘What did they give to whom?’ To my knowledge, however, Jaeger’s study is the ﬁrst to systematically investigate the relation between (fronted) wh-items and (doubled) clitics; see also Jaeger 2002. I will therefore concentrate on his data and exposition, and then tie it in with related data in other languages. See his work for further speciﬁcs regarding Bulgarian syntax.

Top Issues in Questions

283

References Androulakis, Anna. 1998. Wh- and direct object clitics revisited. In Brian D. Joseph, Geo¤rey C. Horrocks, and Irene Philippaki-Warburton, eds., Themes in Greek linguistics II, 131–167. Amsterdam: John Benjamins. Baker, C. L. 1970. Notes on the description of English questions: The role of an abstract question morpheme. Foundations of Language 6, 197–217. Beck, Sigrid. 1996a. Quantiﬁed structures as barriers for LF-movement. Natural Language Semantics 4, 1–56. Beck, Sigrid. 1996b. Wh-constructions and transparent Logical Form. Doctoral dissertation, Eberhard-Karls-Universita¨t Tu¨bingen. Beck, Sigrid, and Shin-Sook Kim. 1997. On wh- and operator-scope in Korean. Journal of East Asian Linguistics 6, 339–384. Billings, Loren, and Catherine Rudin. 1998. Animacy and focus in Bulgarian. Paper presented at the Workshop of Comparative Slavic Morphosyntax, Indiana University and the U.S. Department of Education, McCormick’s Creek State Park, Spenser, Ind., June 5–7, 1998. Boeckx, Cedric. 1999. Decomposing French questions. In Jim Alexander, Na-Rae Han, and Michelle Minnick Fox, eds., Proceedings of the 23rd Annual Penn Linguistics Colloquium, 69–80. University of Pennsylvania Working Papers in Linguistics 6.1. Philadelphia: University of Pennsylvania, Penn Linguistics Club. Boeckx, Cedric. 2003. Islands and chains. Amsterdam: John Benjamins. Boeckx, Cedric, and Kleanthes K. Grohmann. 2004. SubMove: Towards a uniﬁed account of scrambling and D-linking. In David Adger, Ce´cile de Cat, and George Tsoulas, eds., Peripheries: Syntactic edges and their e¤ects, 241–257. Dordrecht: Kluwer. Boeckx, Cedric, and Sandra Stjepanovic´. 2005. The wh/clitic connection. In Lorie Heggie and Francisco Ordo´n˜ez, eds., Clitics and a‰xation, 301–314. Amsterdam: John Benjamins. Bosˇkovic´, Zˇeljko. 2001. On the interpretation of multiple questions. Linguistic Variation Yearbook 1, 1–15. Bosˇkovic´, Zˇeljko. 2002. On multiple wh-fronting. Linguistic Inquiry 34, 351–383. Brunetti, Lisa. 2003. A uniﬁcation of focus. Doctoral dissertation, Universita` di Firenze. Calabrese, Andrea. 1984. Multiple questions and focus in Italian. In Wim de Geest and Yvan Putseys, eds., Sentential complementation, 67–74. Dordrecht: Foris. Calabrese, Andrea. 1992. Some remarks on focus and logical structure in Italian. In Susumu Kuno and Ho¨skuldur Thra´insson, eds., Harvard working papers in linguistics 1, 91–127. Cambridge, Mass.: Harvard University, Department of Linguistics. Chomsky, Noam. 1973. Conditions on transformations. In Stephen R. Anderson and Paul Kiparsky, eds., A festschrift for Morris Halle, 232–286. New York: Holt, Rinehart and Winston.

284

Kleanthes K. Grohmann

Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal syntax, 71–132. New York: Academic Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1990. Types of A¯-dependencies. Cambridge, Mass.: MIT Press. Citko, Barbara. 1998. On multiple wh-movement in Slavic. In Zˇeljko Bosˇkovic´, Steven Franks, and William Snyder, eds., Formal Approaches to Slavic Linguistics: The Connecticut Meeting, 97–114. Ann Arbor, Mich.: Michigan Slavic Publications. Citko, Barbara, and Kleanthes K. Grohmann. 2001. The non-uniqueness of multiple wh-fronting: German ¼ Bulgarian. In Steven Franks, Tracy Holloway King, and Michael Yadro¤, eds., Formal Approaches to Slavic Linguistics: The Indiana Meeting, 117–136. Ann Arbor, Mich.: Michigan Slavic Publications. Comorovski, Ileana. 1996. Interrogative phrases and the syntax-semantics interface. Dordrecht: Kluwer. Dikken, Marcel den, and Anastasia Giannakidou. 2002. From hell to polarity: ‘‘Aggressively non-D-linked’’ wh-phrases as polarity items. Linguistic Inquiry 33, 31–61. Dimitrova-Vulchanova, Mila, and Lars Hellan. 1999. Clitics and Bulgarian clause structure. In Henk van Riemsdijk, ed., Clitics in the languages of Europe, 469– 513. Berlin: Mouton de Gruyter. Dobrovie-Sorin, Carmen. 1990. Clitic doubling, wh-movement, and quantiﬁcation in Rumanian. Linguistic Inquiry 21, 351–397. ¨ ber die syntaktische Position der Satztopiks im Deutschen. Frey, Werner. 2000. U In Kerstin Schwabe, Andre´ Meinunger, and Dieter Gasde, eds., ZAS papers in linguistics 20, 137–172. Berlin: ZAS. Grewendorf, Gu¨nther. 2001. Multiple wh-fronting. Linguistic Inquiry 32, 87– 122. Grewendorf, Gu¨nther. 2002. Minimalistische Syntax. Tu¨bingen: Francke. Grewendorf, Gu¨nther, and Joachim Sabel. 1999. Scrambling in German and Japanese: Adjunction versus multiple speciﬁers. Natural Language and Linguistic Theory 17, 1–65. Grohmann, Kleanthes K. 1998. Syntactic inquiries into discourse restrictions on multiple interrogatives. Groninger Arbeiten zur germanistischen Linguistik 42, 1– 60. Grohmann, Kleanthes K. 2000a. Proliﬁc peripheries: A radical view from the left. Doctoral dissertation, University of Maryland, College Park.

Top Issues in Questions

285

Grohmann, Kleanthes K. 2000b. Towards a syntactic understanding of prosodically reduced pronouns. Theoretical Linguistics 25, 149–184. Grohmann, Kleanthes K. 2003a. German is a multiple wh-fronting language! In Cedric Boeckx and Kleanthes K. Grohmann, eds., Multiple wh-fronting, 99–130. Amsterdam: John Benjamins. Grohmann, Kleanthes K. 2003b. Proliﬁc domains: On the anti-locality of movement dependencies. Amsterdam: John Benjamins. Hagstrom, Paul. 1998. Decomposing questions. Doctoral dissertation, MIT. Haider, Hubert. 2000. Towards a superior account of superiority. In Uli Lutz, Gereon Mu¨ller, and Arnim von Stechow, eds., Wh-scope marking, 231–248. Amsterdam: John Benjamins. Hirose, Tomio. 2003. The syntax of D-linking. Linguistic Inquiry 34, 499–506. Hornstein, Norbert. 1995. Logical Form: From GB to minimalism. Oxford: Blackwell. Hornstein, Norbert. 2001. Move! A minimalist theory of construal. Oxford: Blackwell. Hornstein, Norbert, Jairo Nunes, and Kleanthes K. Grohmann. 2006. Understanding minimalism. Cambridge: Cambridge University Press. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT. Jaeger, T. Florian. 2002. On the optional and obligatory realization of the Bulgarian direct object clitic. Ms., Stanford University. Jaeger, T. Florian. 2003. Topics ﬁrst! In- and outside of Bulgarian wh-interrogatives. In Stefan Mu¨ller, ed., The Proceedings of the 10th International Conference on Head-Driven Phrase Structure Grammar, Michigan State University, July 18– 20, 2003, 181–202. Stanford, Calif.: CSLI Publications. Jaeger, T. Florian. 2004. Topicality and superiority in Bulgarian wh-questions. In Olga Arnaudova, Wayles Browne, Maria Luisa Rivero, and Danijela Stojanovic´, eds., Formal Approaches to Slavic Linguistics: The Ottawa Meeting, 2003, 207– 228. Ann Arbor, Mich.: Michigan Slavic Publications. Kim, Rhang-Hye-Yun. 2003. Scrambling and Minimalist Program. Studies in Generative Grammar 13, 3–19. E´. Kiss, Katalin. 1998. Identiﬁcational focus versus information focus. Language 74, 245–273. Kuno, Susumu. 1982. The focus of the question and the focus of the answer. In Robinson Schneider, Kevin Tuite, and Robert Chametzky, eds., Papers from the Parasession on Nondeclaratives, 134–157. Chicago: University of Chicago, Chicago Linguistic Society. Lambova, Mariana. 2001. On A 0 -movements in Bulgarian and their interaction. The Linguistic Review 18, 327–374.

286

Kleanthes K. Grohmann

Lambova, Mariana. 2003. Patterns of wh-fronting in Bulgarian. Paper presented at the 4th GLOW in Asia, Seoul National University, Korea, August 20–23. Leafgren, John R. 1997. Bulgarian clitic doubling: Overt topicality. Journal of Slavic Linguistics 5, 117–143. Lee, Kisuk, and Satoshi Tomioka. 2000. Intervention e¤ects are topic e¤ects: Whquestions in Japanese and Korean. Paper presented at the Conference on Peripheral Positions, University of York, September 9–12. Lipta´k, Aniko´. 2001. On the syntax of wh-items in Hungarian. Doctoral dissertation, Leiden University. Lipta´k, Aniko´. 2004. Scope marking constructions in Dayal-type indirect dependency. In Anne Breitbarth and Henk van Riemsdijk, eds., Triggers, 259–292. Berlin: Mouton de Gruyter. Lotﬁ, Ahmad R. 2003. Persian wh-riddles. In Cedric Boeckx and Kleanthes K. Grohmann, eds., Multiple wh-fronting, 161–186. Amsterdam: John Benjamins. Mahajan, Anoop. 1990. The A/A-bar distinction and movement theory. Doctoral dissertation, MIT. Mathieu, Eric. 1999. French wh in situ and the intervention e¤ect. In Corinne Iten and Ad Neeleman, eds., UCL working papers in linguistics 11, 441–472. London: University College London, Department of Phonetics and Linguistics. Mathieu, Eric. 2002. The syntax of non-canonical complementation: A comparative study. Doctoral dissertation, University College London. McCloskey, James. 1979. Transformational syntax and model theoretic semantics. Dordrecht: Reidel. Meinunger, Andre´. 2000. Syntactic aspects of topic and comment. Amsterdam: John Benjamins. Nissenbaum, Jon. 2000. Investigations of covert phrase movement. Doctoral dissertation, MIT. Nunes, Jairo. 2004. Linearization of chains and sideward movement. Cambridge, Mass.: MIT Press. Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Alice G. B. ter Meulen and Eric Reuland, eds., The representation of (in)deﬁniteness, 98–129. Cambridge, Mass.: MIT Press. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: MIT Press. Poletto, Cecilia, and Jean-Yves Pollock. 2002. On the left periphery of Romance interrogatives. GLOW Newsletter 48, 47–48. Reglero, Lara. 2003. Non-wh-fronting in Basque. In Cedric Boeckx and Kleanthes K. Grohmann, eds., Multiple wh-fronting, 187–227. Amsterdam: John Benjamins. Richards, Norvin. 1997. What moves where when in which language? Doctoral dissertation, MIT.

Top Issues in Questions

287

Richards, Norvin. 2001. Movement in language: Interactions and architectures. Oxford: Oxford University Press. Rizzi, Luigi. 1978. Violations of the Wh-Island Constraint in Italian and the Subjacency Condition. In Colette Dubuisson, David Lightfoot, and Yves Charles Morin, eds., Montreal working papers in linguistics 11, 155–190. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1997. The ﬁne structure of the left periphery. In Liliane Haegeman, ed., Elements of grammar: Handbook of generative syntax, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2004. Locality and left periphery. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 223–251. Oxford: Oxford University Press. Rudin, Catherine. 1988. On multiple questions and multiple wh-fronting. Natural Language and Linguistic Theory 6, 445–501. Rullman, Hotze, and Sigrid Beck. 1998. Presupposition projection and the interpretation of ‘which’-questions. In Devon Strolovitch and Aaron Lawson, eds., Proceedings from Semantics and Linguistic Theory VIII (SALT VIII), 215–232. Ithaca, N.Y.: Cornell University, CLC Publications. Sabel, Joachim. 1996. Restrukturierung und Lokalita¨t: Universelle Beschra¨nkungen fu¨r Wortstellungsvarianten. Berlin: Akademie Verlag. Sabel, Joachim. 1998. Principles and parameters of wh-movement. Habilitationsschrift, Johann Wolfgang Goethe-Universita¨t Frankfurt am Main. Sabel, Joachim. 2002. Intermediate traces, reconstruction and locality e¤ects. In Artemis Alexiadou, ed., Theoretical approaches to universals, 259–313. Amsterdam: John Benjamins. Saito, Mamoru. 1989. Scrambling as semantically vacuous A-bar movement. In Mark Baltin and Anthony Kroch, eds., Alternative conceptions of phrase structure, 182–200. Chicago: University of Chicago Press. Simpson, Andrew. 2000. Wh-movement and the theory of feature checking. Amsterdam: John Benjamins. Stechow, Arnim von, and Wolfgang Sternefeld. 1988. Bausteine syntaktischen Wissens. Wuppertal: Opladen Verlag. Sura´nyi, Bala´zs. 2004. The left periphery and cyclic spell-out: The case of Hungarian. In David Adger, Ce´cile de Cat, and George Tsoulas, eds., Peripheries: Syntactic edges and their e¤ects, 49–74. Dordrecht: Kluwer. Swart, Henrie¨tte de. 1992. Intervention e¤ects, monotonicity and scope. In Chris Barker and David Dowty, eds., SALT II: Proceedings of the Second Conference on Semantics and Linguistic Theory, 387–406. OSU Working Papers in Linguistics 40. Columbus: Ohio State University, Department of Linguistics. Uriagereka, Juan. 1988. On government. Doctoral dissertation, University of Connecticut, Storrs. Uriagereka, Juan. 1995a. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26, 79–123.

288

Kleanthes K. Grohmann

Uriagereka, Juan. 1995b. An F position in Western Romance. In Katalin E´. Kiss, ed., Discourse conﬁgurational languages, 153–175. Oxford: Oxford University Press. Villalba Nicola´s, F. Xavier. 2000. The syntax of sentence periphery. Doctoral dissertation, Universitat Auto`noma de Barcelona. Webelhuth, Gert. 1992. Principles and parameters of syntactic saturation. Oxford: Oxford University Press.

Chapter 11 Mechanisms of WhSaturation and Interpretation in Multiple Wh-Movement

11.1

Bala´zs Sura´nyi

Introduction

Multiple wh-constructions, mostly of the Slavic type, have received considerable attention in recent minimalist literature. This chapter demonstrates that Hungarian exhibits (at least) three syntactic patterns of multiple wh-questions, and it relates these three syntactic patterns in a principled way to di¤erent focus structures. Modulo distinctions in focus structure, I relate the three multiple wh-constructions to three distinct sets of answerhood conditions they are shown to be matched with, arguing that multiple (overt or covert) movement to the same projection results in a pair-list interpretation, while otherwise only a single-pair reading is available. Providing a typological outlook, I show that this account of answerhood conditions can be maintained more generally. Brieﬂy, the chapter is structured as follows. In section 11.2, I establish that Hungarian multiple wh-fronting (henceforth MF) cannot be reduced either to multiple focus fronting (cf. Bosˇkovic´ 1997, 1998, 1999, 2000, 2002), or to universal quantiﬁer fronting (cf. E´. Kiss 1994, 2002), or to multiple topicalization. In section 11.3, I examine the topic/focus structure of both the Slavic-type MF and the apparently English-type multiple wh-questions to be found in Hungarian, and I propose a syntactic analysis that holds that in the MF pattern in Hungarian exactly one wh-phrase bears both [foc] and [wh], whereas the rest of the fronted wh-elements carry only [wh], all occupying multiple speciﬁers of a single FocP projection. The seemingly uniform English-type multiple wh-pattern is in fact to be factored into two covert structures corresponding to two focus structures. Then in section 11.4, I derive the way the three di¤erent syntactic structures I have found match up with three distinct sets of answerhood conditions. In section 11.5, I probe various MF and non-MF languages, showing that the account of answerhood conditions I have put forward

Bala´zs Sura´nyi

290

can be maintained in the context of a more general typology of interpretations available in multiple wh-constructions. 11.2

Movement in Multiple Wh-Fronting

11.2.1 Multiple Focusing It is a standard view that wh-operators in single wh-questions are a special case of focus operators (e.g., Rochemont 1978, 1986; Culicover and Rochemont 1983; Horva´th 1986), and indeed in Hungarian (or Italian, Greek, Catalan, Bengali, etc.) single wh-movement and focus movement apparently target the same left-peripheral position, identiﬁed as FocP in Brody 1990. Bosˇkovic´ (1997, 2000, 2002) argues that MF in Serbo-Croatian matrix short wh-movement contexts (as well as in analogous MF contexts in other languages) derives from multiple syntactic focalizations of wh-items (following Stjepanovic´ 1995, 1998; see also Citko 1998), where the o¤ending [foc] feature is located on the moved wh-elements and not on the attractor—hence the order of checking the wh-elements is free, hence it can violate Superiority. Bosˇkovic´ (1998, 1999, 2002) extends this to Bulgarian, and Stepanov (1998) to Russian. Now Hungarian multiple whfronting can freely violate Superiority, in matrix, long movement, and embedded contexts. I argue, however, that the analysis of Superiorityviolating MF involving multiple focus fronting proposed by Bosˇkovic´ cannot be extended to Hungarian. More precisely, treating MF in Hungarian as multiple syntactic focalizations is on the one hand unjustiﬁed, and on the other, not a necessity. It is not a necessity because Bosˇkovic´’s account of the lack of a superiority e¤ect depends not on the nature of the feature (being [foc]), but on its location (being on the moved elements). Consequently, the very same account can be transposed to [wh] on the moved wh-items. Independently of this option, the apparent violations of Superiority can be easily put down in Hungarian to the attested freedom in the basic order of constituents of the (extended) VP (for the basic order within VP, see E´. Kiss 1994, 2002). On the other hand, a Bosˇkovic´-style analysis cannot characterize Hungarian, given that multiple overt focusing—which is available in SerboCroatian—is strictly prohibited in Hungarian. In Hungarian, only a unique preverbal focus is allowed, whereas MF involves multiple instances of preverbal wh-elements. Bosˇkovic´ argues that in echo questions, wh-elements are focused but do not carry [wh]. Now in this regard what is interesting is that although

Wh-Saturation and Interpretation in Multiple Wh-Movement

291

Hungarian single echo-questions must front the (only) wh-element, a multiple echo-question cannot front all wh-items (in contrast to SerboCroatian and Bulgarian; see Bosˇkovic´ 1997, 1999, 2000): it has to front exactly one of them. This should mean once again that Hungarian MF involves only a unique focused wh-item, the rest of the wh-expressions being nonfocused. 11.2.2 Covert Wh-Movement and Multiple Wh-Fronting via Quantiﬁer Movement? Even though syntactic focusing cannot be held responsible for MF in Hungarian, the standard view that the fronted wh-element in single questions is focused and occupies a syntactic focus position may be adopted nonetheless. The question then is still open whether the single fronted wh-element has also undergone wh-movement in overt syntax, or whether wh-movement takes place covertly (in the spirit of Bosˇkovic´’s proposal) to a higher functional projection of the CP system. In recent work (though without explicit reference to the connection with Bosˇkovic´’s analysis), Lipta´k (2001) sides with the latter analysis, shown schematically in (1). (1) [CP C 0 [wh]

[FocP WH [foc]

"

" covert whmovement

Foc 0 . . . [foc] [wh]

overt focus movement

In order to treat the nonﬁnal (and nonfocused) fronted wh-elements in MF constructions (i.e., WH1 and WH2 in (3) below), Lipta´k adopts a common assumption from the relevant literature on Hungarian (e.g., Horva´th 1998; Puska´s 2000; E´. Kiss 2002), namely, that these wh-items have the status of universal quantiﬁers of the every-QP type, a view that originated with E´. Kiss (1992, 1993) and Comorovski (1989) (see also Garrett 1996). If every-QPs (or other increasing distributive quantiﬁers) co-occur with and outscope a focus, they are typically fronted to a position above fronted focus (their scope position) (2). E´. Kiss (1992, 1993) claims that nonﬁnal fronted wh-elements are analogous both syntactically and semantically to every-QPs; hence, (3a) can be paraphrased as (3c). ´ NOST mutatta (2) a. Mindenki minden la´nynak JA be everybody every girl-dat Ja´nos-acc introduced pref ‘Everybody introduced JOHN to every girl.’ b. [every1 [every2 [focus . . .

Bala´zs Sura´nyi

292

(3) a. Ki melyik la´nynak kit mutatott be? who which girl-dat who-acc introduced pref ‘Who introduced who to which girl?’ b. [WH1 [WH2 [WH3 . . . c. ‘For every person, tell me for every girl who that person introduced to her.’ I now show brieﬂy that on the one hand, wh-movement per se is overt in Hungarian, and on the other, the universal quantiﬁer treatment of nonﬁnal fronted wh-items is unsupported. Elsewhere, I have presented considerations showing that Lipta´k’s arguments in favor of a covert wh-movement step in Hungarian are inconclusive (Sura´nyi 2005). Below, I will concentrate on two arguments that directly vitiate the covert wh-movement analysis. The ﬁrst point to be made is that if in a MF question like (3), WH3 by assumption covertly raises to a C position higher than WH1 and WH2 and takes widest scope, then the sorting key (cf. Kuno 1982; Kuno and Takami 1993) should be provided by WH3 . This is contrary to fact, since in MF questions the sorting key is invariably supplied by the leftmost wh-element, namely, WH1 . Ki melyik ta´rgyat tanı´tja? who which subject-acc teaches ‘Who teaches which subject?’ b. Pa´l a szintaxist tanı´tja, Ma´rk a szintaxist e´s a morfolo´gia´t, . . . ‘Paul teaches syntax, Mark teaches syntax and morphology, . . . ’ c. a(5b)

(4) a.

Melyik ta´rgyat ki tanı´tja? A szintaxist Pa´l e´s Ma´rk tanı´tja, a morfolo´gia´t Ma´rk, . . . ‘Syntax is taught by Paul and Mark, morphology by Mark, . . . ’ c. a(4b)

(5) a. b.

An interesting contrast between syntactic focusing and single wh-fronting, which are by assumption identical in overt syntax, comes from Lipta´k 2001. Focus in a headless relative clause may or may not overtly piedpipe the containing headless relative to a matrix focus position. However, the same is obligatory for a wh-element. (6) a. Az jo¨het be [aki PE´TERT ismeri] that-nom come-pot-3sg in who-rel Peter-acc knows b. ?(Csak) [aki PE´TERT ismeri] jo¨het be (only) who-rel Peter-acc knows come-pot-3sg in ‘(Only) those who know PETER can come in.’

Wh-Saturation and Interpretation in Multiple Wh-Movement

293

(7) a. *Az jo¨het be [aki kit ismer]? that-nom come-pot-3sg in who-rel who-acc knows b. ? (Csak) [aki kit ismer] jo¨het be? (only) who-rel who-acc knows come-pot-3sg in ‘Who is such that (only) those who know him can come in?’ If in overt syntax (single) wh-movement were merely focus movement, then the obligatory overt movement in (7) would be completely unexpected. It is all the more unsurprising if bona ﬁde wh-movement is in fact overt, that is, if [wh] is checked overtly. I turn now to the other component of the analysis, namely, identifying nonﬁnal fronted wh-items (henceforth ‘‘high’’ wh-phrases) as being fronted qua universal quantiﬁers. I demonstrate in Sura´nyi 2002 that even though such a general account is able to derive a number of properties of MF constructions, inasmuch as the same properties can be derived in alternative ways, this constitutes no argument in favor of such an analysis. Let me restrict attention here to two types of arguments that in fact appear to go against this treatment. The ﬁrst type of argument comes from the interpretation of MF constructions that involve more than two fronted wh-elements, as in (3) above. If the interpretation of (3a) did correspond to the paraphrase in (3c), appropriate answers would be expected to be ones that are exhaustive not only with respect to the set quantiﬁed over by WH1 (appropriate answers are such), but also with respect to the set quantiﬁed over by WH2 —contrary to fact. (3a) is not synonymous with Hungarian (8a); nor is its interpretation or its appropriate answer analogous to those of English (8b), both of which require exhausting the set of girls. (8) a. Mondd el, mindenki minden la´nynak kit mutatott be tell-imp pref everybody every girl-dat who-acc introduced pref ‘Tell (me) who is it that everybody introduced to every girl?’ b. What did every boy give every girl? This questions treating the high wh-phrases as universal quantiﬁers. Further doubt is cast on such a treatment by distributional di¤erences between high wh-phrases and every-QP-type quantiﬁers. Let me point out a few of these syntactic discrepancies. First, while wide scope every-QPs typically appear above the preverbal focus site, they may also stay postverbal. The same is impossible for high wh-items (9). Wide scope every-QPs may undergo long overt movement to a superordinate clause, but the same is impossible for high wh-phrases

Bala´zs Sura´nyi

294

(10). Wide scope universals can be separated by certain high adverbs and parentheticals, whereas high wh-expressions cannot (11). ´ NYT hı´vott fel minden ﬁu´ KE´T LA two girl-acc called up every boy ‘Every boy phoned TWO GIRLS.’ (every > two) b. *Kit hı´vott fel ki? who-acc called up who On the interpretation who > whom, with who ¼ sorting key

(9) a.

Minden la´nyt PE´TER szeretne´, hogy felhı´vjunk every girl-acc Peter like-cond-3sg that up-call-subj-1pl ‘For every girl, it’s Peter that would like us to phone.’ b. *Melyik la´nyt PE´TER ke´rdezte, hogy mikor hı´vjunk fel? which girl-acc Peter asked that when call-subj-1pl up ‘Which girl is such that it’s Peter that asked when we should phone her?’

(10) a.

Minden ﬁu´ szerintem minden la´nyt felhı´vott every boy in-my-opinion every girl-acc up-called ‘In my opinion, every boy phoned every girl.’ b. *Ki szerinted kit mikor hı´vott fel? who in-your-opinion who-acc when called up ‘In your opinion, who phoned who when?’

(11) a.

It appears justiﬁed to draw the following two conclusions in view of these observations. First, neither the fronted wh-element in single questions nor the linearly last fronted wh-element in MF undergoes further covert whmovement. Second, MF does not result from quantiﬁer fronting; that is, high wh-phrases are fronted by some other movement operation. I brieﬂy examine next whether this movement could be identiﬁed as syntactic topicalization. 11.2.3 Multiple Topicalization I argued above that high wh-phrases, in contrast to the linearly last, immediately preverbal fronted wh-element in MF, are not focused. Two questions arise. The ﬁrst is related to another syntactic pattern of multiple wh-interrogatives also found in Hungarian (as well as in several Slavic languages), which appears to be analogous to English-type multiple whinterrogatives, with one wh-element fronted, further wh-elements apparently in situ (henceforth English-type questions). The question is whether in-situ wh-phrases in English-type questions are also focused or not. I re-

Wh-Saturation and Interpretation in Multiple Wh-Movement

295

turn to this in section 11.3. Topic/focus structure is central to the second issue as well, to which I turn ﬁrst. Given the heavy use of syntactic topicalization in Hungarian (see E´. Kiss 1994, 2002), it could be claimed that high wh-items in MF are in fact topics, and undergo topic movement. This may appear prima facie justiﬁed, since topics also occupy a (recursive) position above the focus site, as (12) illustrates. ´ NOS mutatta (12) A la´nyt a ﬁu´nak JA be the girl-acc the boy-dat John-nom introduced pref ‘As for the girl, as for the boy, it’s John who introduced her to him.’ However, high wh-phrases are syntactically distinct from topicalized constituents, as I demonstrate presently. First, topics can intersperse with certain adverbs (e.g., temporal, epistemic) or parentheticals, while high wh-expressions cannot (13). ´ NOS A la´nyt minden valo´szı´nu˝se´g szerint a ﬁu´nak JA the girl-acc in all probability the boy-dat John-nom mutatta be introduced pref ‘As for the girl, in all probability, as for the boy, it’s John who introduced her to him.’ b. *Ki minden valo´szı´nu˝se´g szerint kinek kit mutatott who in all probability who-dat who-acc introduced be? pref ‘Who introduced who to who in all probability?’

(13) a.

Naturally, two topics can be separated by a topicalized constituent, whereas two high wh-elements cannot (14). A la´nyt a ﬁu´nak Pe´ter TEGNAP mutatta be the girl-acc the boy-dat Peter-nom yesterday introduced pref ‘Peter introduced the girl to the boy YESTERDAY.’ b. *Ki a ﬁu´nak kit mikor mutatott be? who the boy-dat who-acc when introduced pref ‘Who introduced who to the boy when?’

(14) a.

Long movement of a topic to a matrix clause is possible, but long movement of high wh-phrases to a similar position is severely degraded (15).

Bala´zs Sura´nyi

296

Pe´tert Gabi azt mondta, hogy elu¨to¨tte egy Peter-acc Gaby-nom that-acc said that over-ran a auto´ car-nom ‘Gaby said that Peter was run over by a car.’ b. *Kit Gabi azt ke´rdezte, hogy mi u¨to¨tt who-acc Gaby-nom that-acc asked that what-nom ran el over ‘Gaby asked who was run over by what.’

(15) a.

¯ -position reconstruct, unlike Topics with pronouns bound from a lower A analogous high wh-phrases. In (16), binding of the pro possessor within the initial topicalized DP by the every-quantiﬁer is possible in (a), but the pro in the ﬁrst fronted wh-phrase cannot be bound by the second whoperator in (b). A ﬁa minden apa´nak bu¨szkese´get jelent [the son-his [every father-dat pride-acc means]] ‘Every father takes pride in his son.’ b. *Melyik gyereke melyik fe´rﬁnek jelent bu¨szkese´get? [which child-his [which man-dat means pride-acc]] ‘Which man takes pride in which of his children?’

(16) a.

I conclude that high wh-phrases cannot be collapsed syntactically with topics: their fronting cannot be put down to topicalization. In the foregoing, I have demonstrated, rejecting relevant alternatives— especially treatments by E´. Kiss (1993, 1994, 2002) and Lipta´k (2001)— that the Hungarian MF construction is not brought about either by multiple focusing, or by quantiﬁer fronting, or by topicalization. In section 11.3, I propose a syntactic analysis of the MF construction as well as the English-type pattern in Hungarian, taking their focus structure into account, to which I presently turn. 11.3

Topic-Focus Structure and the Structure of Multiple Wh-Questions

11.3.1 Topic-Focus Structure I have adopted the view from the pertinent literature that the linearly last fronted wh-element in MF is focused and occupies Spec,FocP. This is conﬁrmed not only by its complementary distribution with preverbal focus, but also by the fact that it corresponds to (immediately) preverbal fo-

Wh-Saturation and Interpretation in Multiple Wh-Movement

297

cus in the appropriate answers, and by the fact that it bears pitch accent. I have contended that high wh-phrases in MF cannot have been moved by topicalization, or by (universal) quantiﬁer fronting, or by focusing. The latter assumption was justiﬁed in section 11.2.1 simply by reference to the fact that Hungarian does not allow multiple preverbal foci. I now add that high wh-elements are interpreted at the interfaces as topics (and not as foci). This appears to be the case given that high wh-expressions invariably quantify over presupposed sets (whereas the immediately preverbal whphrase may not). Furthermore, constituents that correspond in appropriate answers to a high wh-element in the question must appear in a topic position. Ki mit la´tott? who what-acc saw ‘Who saw what?’ b. Pe´ter a Termina´tor 3-at la´tta, e´s Ja´nos a Peter-nom the Terminator 3-acc saw and John-nom the Ma´trix 2-t (la´tta) Matrix 2-acc (saw) ‘Peter saw Terminator 3, and John (saw) Matrix 2.’ c. aA Termina´tor 3-at la´tta Pe´ter, . . . the Terminator 3-acc saw Peter ‘Peter saw Terminator 3, . . . ’ (unacceptable as an answer to (a))

(17) a.

In short, high wh-elements are not syntactically topicalized; nevertheless, they have the discourse semantic status of a topic.1 I examine next the discourse-functional status of wh-in-situ in Englishtype multiple questions in Hungarian—that is, whether or not they are focused, similarly to the immediately preverbal wh-element. As a consequence of the picture I have o¤ered of the focus status of whphrases in MF, wh-phrases cannot bear [foc] as a lexical feature; rather, this property must be an optional feature assigned to them in the numeration (Num, see Chomsky 1995; or in the relevant lexical array, LA, see Chomsky 2000)—just like in the case of ordinary foci. In these terms, the common assumption of the parallelism of wh-operators and focus, which I have also adopted here, holds that the wh-pronoun in single-constituent questions must be assigned [foc] in the Num/LA in order to function and be interpreted as a question operator. Transposing this to multiple

298

Bala´zs Sura´nyi

questions, it should follow that at least one wh-item must be focused there too.2 I demonstrated above that in Hungarian MF, exactly one whpronoun must bear [foc]: namely, the linearly last one in the fronted cluster. Given that [foc] is not an inherent feature of wh-pronouns, if it is su‰cient to have one wh-phrase functioning as focus in a (multiple) question,3 the second wh-item in English-type multiple interrogatives a priori may or may not be assigned [foc]. Indeed, there is empirical evidence that both options are realized. Focused constituents undergo obligatory syntactic fronting and are immediately followed by the verb in Hungarian, in both root and embedded clauses. Of interest now is to construct English-type biclausal multiple interrogatives that have one wh-phrase in (the focus position of ) the matrix clause and a second wh-phrase in the complement clause. What we ﬁnd in such cases is that the second wh-element only optionally fronts within the embedded clause to the immediate left of the embedded verb.4 (18) a. Melyik ﬁu´ a´llı´tja, hogy felhı´vta melyik la´nyt? which boy claims that up-called which girl-acc b. Melyik ﬁu´ a´llı´tja, hogy melyik la´nyt hı´vta fel? which boy claims that which girl-acc called up ‘Which boy claims that he phoned which girl?’ This contrasts with the patterning of an analogous sentence with foci, where the focused constituent in the embedded clause must raise to focus position. (19) a. *CSAK E´N a´llı´tom, hogy felhı´vta Pe´ter CSAK only I-nom claim that up-called-3sg Peter-nom only MARIT Mary-acc b. CSAK E´N a´llı´tom, hogy CSAK MARIT hı´vta fel only I-nom claim that only Mary-acc called-3sg up Pe´ter Peter-nom ‘It’s only me who claims that Peter phoned only Mary.’ The apparent optionality in (18), then, derives from the optional assignment of [foc] to the in-situ wh-element in the Num/LA. I have shown here that in MF, high wh-phrases have a topic status instead of focus (though they are not syntactically topicalized, as argued in

Wh-Saturation and Interpretation in Multiple Wh-Movement

299

the previous section), and exactly one wh-phrase (the linearly last one) bears [foc]. On the other hand, in English-type multiple questions the in-situ wh-phrase bears [foc] as a free option, as determined in the Num/ LA. 11.3.2 Focusing, Wh-Saturation, and the Structure of Multiple Questions Since potential alternatives advocated in the literature are unsuitable, as shown in section 11.2, and since fronted wh-elements form an undisruptable cluster, the obvious remaining candidate producing MF is bona ﬁde overt multiple wh-movement to a multiple speciﬁer conﬁguration, triggered by strong [wh]. Given that in English-type double questions, the second wh-element does not overtly raise, one might be tempted to conclude that [wh] is weak on the wh-items themselves, and strong only on the attracting functional head Foc. Then, Slavic-type MF and English-type multiple questions in Hungarian could be distinguished by assuming that Foc could have either an Attract-1 [wh] or an Attract-All [wh] (the latter overtly attracting an indeﬁnite number of wh-elements), in the manner entertained in Bosˇkovic´ 1999 (see also Chomsky 1995; Pesetsky 2000). However, such an analysis is not only conceptually unattractive in that it postulates lexical ambiguity in the Foc element (which serves merely to capture the problematic patterns), but also empirically unworkable. This is because Hungarian allows multiple questions with both MF and an insitu wh-phrase, as in (20).5 Such sentences cannot be generated in such a bifurcated system of [wh]s on Foc. (20) a. (Mondd el) melyik ﬁlmben ki o¨l meg kit? (tell-imp pref ) which ﬁlm-in who kills pref who-acc ‘(Tell me) who kills who in which ﬁlm?’ b. (Mondd el) ki mit te´vesztett o¨ssze mivel? (tell-imp pref) who what-acc confused pref what-with ‘(Tell me) who confused what with what?’ This supports the alternative of the lexically ambiguous Foc head analysis: in MF, [wh] is strong on the moved wh-pronouns themselves (see also Lasnik 1999 and Chomsky 2000 for relevant discussion of ‘‘Attract’’ vs. ‘‘Move’’—that is, the location of o¤ending features). The structure and feature composition I am assuming for the MF construction is given schematically in (21).

300

Bala´zs Sura´nyi

(21)

Multiple wh-speciﬁer structures (vs. single wh-speciﬁer structures) are generally implicationally related to an obviation of wh-island e¤ects, as well as to the availability of long multiple wh-movement constructions, as for instance in Bulgarian (see Rudin 1988; Richards 1997; Bosˇkovic´ 2000). Indeed, Hungarian parallels Bulgarian in this regard, as expected.6 Now English-type multiple questions in Hungarian stand out as special in that even though the in-situ wh-pronouns also bear strong [wh], they do not overtly raise. I adopt the view that Hungarian wh-pronouns are bare indeﬁnites (see Nishigauchi 1990; Cheng 1991; also Lipta´k 2001 for Hungarian), and I assume they carry [wh], which is checked in MF in Spec,FocP against Foc. I propose to treat English-type Hungarian multiple interrogatives, following Reinhart (1997), as involving binding of choice function variables. In motivating this treatment, it is important to observe that Englishtype multiple questions in Hungarian in general cannot involve adverbial in-situ wh-phrases, as exempliﬁed in (22). (22) a. *Ki jo¨tt el mie´rt? who came along why ‘Who came along why?’ b. *Ki viselkedett hogyan? who behaved how ‘Who behaved how?’ Exceptions to this generalization are speciﬁc/referential (D(iscourse)linked) adverbials, as in (23).

Wh-Saturation and Interpretation in Multiple Wh-Movement

301

(23) Melyik e´rkezett melyik mo´don? which-nom arrived which way-on ‘Which one arrived in which way?’ Adapting Reinhart’s account of a similar dichotomy between nominal and speciﬁc/referential adverbial wh-phrases, on the one hand, and nonreferential adverbials, on the other (cf. also Tsai 1994), this di¤erence can be related to the presence or absence of a D head (the syntactic locus of reference/speciﬁcity). In particular, nonreferential adverbials arguably lack such a D projection. This is signiﬁcant given that Reinhart suggests that choice function (variable)s in in-situ wh-elements are a type of determiner. Thus, nonreferential adverbials cannot combine with a choice function (variable), lacking an appropriate syntactic position.7 I propose that the syntactic reﬂex of combining a bare wh-pronoun with a D head ﬁlled by a choice function (variable) in Hungarian is that the [wh] of the wh-pronoun will be checked. For concreteness, assume that choice function variables combining with wh-pronouns carry [wh]. This means that the [wh] property of the bare wh-pronoun will be able to be checked locally within the full wh-phrase, plausibly, via head movement to D.8 This derives why nonreferential adverbials cannot appear as in-situ wh-elements in English-type multiple questions at all, lending initial plausibility to a choice function analysis. Choice function variables associated with wh-elements are bound longdistance, are quantiﬁed existentially, and by deﬁnition ‘‘choose’’ (i.e., output) an individual from a set (Reinhart 1997). The mechanism entails that island boundaries or quantiﬁcational interveners (e.g., Beck 1996; Chang 1997; Cheng and Rooryck 2000) will not disrupt the dependency, which is indeed the case; see (24a) and (24b), respectively. (24) a. Melyik ﬁu´ lett ideges miuta´n felhı´vta melyik la´nyt? which boy became angry after up-called which girl-acc ‘Which boy got angry after he phoned which girl?’ b. Ki mutatott meg mindenkinek kit? who showed pref everybody-dat who-acc ‘Who showed everybody who?’ Recall from section 11.3.1 that I established that the in-situ whelements in English-type questions may optionally be associated with [foc]. In that case, an English-type question that involves a [foc] in-situ wh-phrase (English-typefocused for short) is essentially analogous to a sentence with multiple foci. I will follow Brody (1990) and Szabolcsi (1997)

Bala´zs Sura´nyi

302

here in taking in-situ foci in Hungarian to raise to the preverbal focus covertly. Notice then, however, that two kinds of English-type questions are being entertained in Hungarian, sharing the same surface pattern: one involves covert movement, the other does not. (25) English-typenonfocused questions [FocP WH Foc [ . . . f – WH . . . ]] [wh] [wh] [wh] [wh] [foc] [foc] (26) English-typefocused questions [FocP WH Foc [ . . . f – WH . . . ]] [wh] [wh] [wh] [wh] [foc] [foc] [foc] " In brief, if the above reasoning concerning the optionally focused status of in-situ wh-phrases is on the right track, then the empirical consequence is that the in-situ wh-expressions may or may not be associated syntactically (by a movement dependency) with the preverbal wh-operator. Having established what patterns of focus structure materialize in the Slavic- and English-type multiple wh-interrogatives, and having proposed that they are respectively characterized as involving true multiple whmovement, on the one hand, and as involving the option of covert focus movement or simple binding of choice function variables, on the other,9 in the next section I consider what predictions this account makes for how these syntactic setups match with answerhood conditions. 11.4

Syntactic Structure and Answerhood Conditions

In general, a multiple interrogative can be answered either with a single pair (SP) or with a list (typically a pair list, PL). More speciﬁcally, certain multiple interrogative constructions may allow both of these answer types as appropriate, others may be expected to be answered only by an SP, or only by a PL. I demonstrate now that the distribution of the SP and PL interpretations across the constructions identiﬁed above is straightforwardly derived on the basis of the structural descriptions I have posited. Here I will adopt Higginbotham and May’s (1981) proposal, also maintained by Barss (2000), that Absorption is a necessary and su‰cient condition for a PL reading to obtain. Absorption is a rule of the syntax-

Wh-Saturation and Interpretation in Multiple Wh-Movement

303

¯ -elements to be structursemantics interface that requires the relevant A ally adjacent. Practically, this entails that if Absorption is to take place, the wh-phrases must raise to the same FocP, the checking position of whoperators. If correct, this assumption directly explains a prominent fact about Hungarian MF questions: they can only receive PL answers, not SP answers. For instance, a question like (4a), repeated here as (27a), can be answered with a list, as in (27b) (¼ (4b)), but not with a single pair of a person and a subject taught (27c). Ki melyik ta´rgyat tanı´tja? who which subject-acc teaches ‘Who teaches which subject?’ b. Pa´l a szintaxist tanı´tja, Ma´rk a szintaxist e´s a morfolo´gia´t, . . . ‘Paul teaches syntax, Mark teaches syntax and morphology, . . . ’ c. aPa´l a szintaxist tanı´tja ‘Paul teaches syntax.’

(27) a.

English-typenonfocused questions do not have the surface in-situ and the preverbal wh-phrase in structurally adjacent positions; hence, only an SP reading is expected. On the other hand, English-typefocused questions do have the surface in-situ wh-element structurally adjacent to the preverbal wh-phrase in covert syntax;10 hence, a PL reading is predicted. That is, for the English-type surface pattern it is predicted that either an SP or a PL reading should be available. All sources on Hungarian describe English-type questions as allowing an SP reading. However, it has gone unnoticed that they also license a PL answer.11 (28) a. Ki ne´zett ra´ kire? who looked on who-on ‘Who looked at who?’ b. Ja´nos ne´zett ra´ Marira ‘John looked at Mary.’ c. Ja´nos ne´zett ra´ Marira, Pali Gabira, . . . ‘John looked at Mary, Paul looked at Gaby, . . . ’ Question (28a) can be asked in the context of a single looking event (28b), but also in the context of attempting to reconstruct what looking events occurred during a group game that is based on silent eye-gestures (28c). What is remarkable is that although both MF questions and (28a)-type (i.e., English-typefocused ) questions can be answered with a list, this list

304

Bala´zs Sura´nyi

strictly must involve individuals as second members of the pairs (nth members of the n-tuples) in the case of (28a)-type (i.e., English-typefocused ) interrogatives (‘‘matching questions’’), whereas an answer to MF questions can also have sets of individuals as second (nth) members (see, e.g., (27b)). Thus, (29) is not felicitous as an answer to (28a). (29) Ja´nos ne´zett ra´ Marira e´s Sa´ra´ra, Pali Gabira, . . . ‘John looked at Mary and Sarah, Paul looked at Gaby, . . . ’ In fact, the analysis I have proposed already explains this restriction with respect to PL answers to English-type questions. English-type questions involve the application of a choice function variable to a wh-pronoun. In English-typefocused questions (like (28a)), the (determiner of the) in-situ wh-item undergoes Absorption with the (determiner of the) preverbal wh-operator, hence the list reading. Given that choice functions by deﬁnition output individuals only, the list will contain individuals as second members. This is suggestive evidence of the role played by choice functions in such ‘‘matching’’ questions. A further correct consequence of the account developed here is the following. Given that Absorption requires movement of the in-situ whphrase in English-type questions, in contexts where this movement is not conceivable—for instance, because of locality restrictions—it is predicted that such English-type questions can expect only an SP answer, not a PL answer. Thus, (30a), repeated from (24a), which involves an island, should allow only an SP answer. Similarly, (30b), involving a quantiﬁcational intervener between the preverbal and the in-situ wh-phrases, is expected to be incompatible with a PL reading. These predictions are indeed veriﬁed. (30) a. Melyik ﬁu´ lett ideges miuta´n felhı´vta melyik la´nyt? which boy became angry after up-called which girl-acc ‘Which boy got angry after he phoned which girl?’ b. Melyik szervezo˝ mutatta be minden jelo¨ltnek melyik which organizer introduced pref every candidate-dat which opponense´t? opponent-his-acc ‘Which organizer introduced which of his opponents to every candidate?’ Recall now the discussion of biclausal English-type double questions from section 11.3.1, where I demonstrated that either the second whexpression undergoes local focus movement within the complement clause

Wh-Saturation and Interpretation in Multiple Wh-Movement

305

(when it bears [foc]) or it does not (when it does not bear [foc]); see (31) (¼ (18)). (31) a. Melyik ﬁu´ a´llı´tja, hogy felhı´vta melyik la´nyt? which boy claims that up-called which girl-acc b. Melyik ﬁu´ a´llı´tja, hogy melyik la´nyt hı´vta fel? which boy claims that which girl-acc called up ‘Which boy claims that he phoned which girl?’ I have established that English-typenonfocused questions must receive an SP answer. This correctly holds of (31a) as well. On the other hand, I have demonstrated that English-typefocused questions may also receive a PL answer (of a restricted kind). On the present account, (31b) is an Englishtypefocused question. However, given that the second wh-element moves to FocP of the lower clause, no structural adjacency obtains; hence, in turn, Absorption cannot apply. Thus, for (31b) only an SP reading is generated—once again, the right prediction: (31b) is indeed incompatible with a PL answer. On the other hand, if the second wh-element in (31) can both be assigned [foc] and associated with the preverbal wh-phrase in the matrix clause, then a list reading is expected to reappear, exceptionally. The relevant case is (32). (32) Annyi riportot ja´tszottak be volt sportolo´kro´l meg re´gi versenyekro˝l, hogy a ve´ge´n ma´r nem tudtam ko¨vetni . . . ‘They broadcast reports about so many sportsmen and so many old sports events that in the end I was unable to follow . . . ’ . . . melyik versenyzo˝ dicsekedett, hogy hol ve´gzett melyik which sportsman boasted that where ﬁnished which versenyen competition-on ‘ . . . which sportsman boasted about where he ﬁnished in which competition.’ Here the matrix verb embeds an interrogative itself. This interrogative clause has its own wh-operator hol ‘where’ in Spec,FocP. This allows the in-situ wh-phrase to be construed as [foc], which then can be syntactically associated with the wh-operator of the matrix clause (as is characteristic of such ‘‘wh-triangle’’ situations), via covert movement. In this case, a PL interpretation is predicted to be induced by the matrix and the in-situ wh-phrases. Indeed, this is correct: (32) can be answered as in (33).

306

Bala´zs Sura´nyi

´ NOS dicsekedett, hogy hol ve´gzett a VB-n, (33) JA John boasted that where ﬁnished the World Cup-at ´ PAL dicsekedett, hogy hol ve´gzett az Olimpia´n, . . . Paul boasted that where ﬁnished the Olympics-at ‘It’s John who boasted of where he ﬁnished in the World Cup, it’s Paul who boasted of where he ﬁnished in the Olympics, . . . ’ Let us take stock of the results of this section. Relying on H&M’s approach to pair-list readings, I exploited the ﬁndings regarding the proper structural description of multiple wh-questions, their focus structure, and the application of choice functions to systematically derive answerhood conditions for the three discovered syntactic patterns of this type of question. I showed that the account of the answerhood conditions naturally and correctly extends to some more complex types of multiple wh-interrogatives as well. In the remainder of the chapter, I demonstrate that the treatment of the PL/SP distinction defended here can be supported in a typological context. 11.5

A Typology of Interpretations in Multiple Wh-Questions

11.5.1 Single-Pair and Pair-List Readings in Eastern European Multiple Wh-Fronting Let us consider the typological implications of the proposals made here, situating Hungarian ﬁrst within Eastern European MF languages.12 Following Higginbotham and May (1981), I have maintained that the list reading in Hungarian MF is a consequence of Absorption, a rule of the syntax-semantics interface, conditioned by structural adjacency of the absorbed operators.13 Then the expectation is that other MF constructions in Eastern European languages should also invariably receive a PL reading. That expectation, however, is not borne out. As I will show directly, this is unsurprising if, as is standardly assumed, Wh-Absorption applies not just to any two adjacent wh-phrases, but to two wh-operators in their wh-operator position. Bosˇkovic´ (1998, 2000) demonstrates that Bulgarian moves its whphrases overtly to CP (CP is taken to be the locus of [wh] checking; not all wh-elements are moved overtly in order to check [wh]). Indeed, Bulgarian MF receives a PL reading (see Bosˇkovic´ 2001). Romanian is analogous to Bulgarian and Hungarian in the relevant regard. Serbo-Croatian, on the other hand, is unlike Hungarian/Bulgarian/ Romanian in that in simple matrix clauses MF licenses both PL and SP

Wh-Saturation and Interpretation in Multiple Wh-Movement

307

readings (Bosˇkovic´ 2001). Bosˇkovic´ (1997, 1998) argues that in such syntactic contexts, Serbo-Croatian wh-items undergo focus-driven movement to a position lower than CP, from where wh-movement to CP takes place in covert syntax (and in this sense, fronted wh-elements count as wh-in-situ). Now, according to the picture developed in the foregoing sections, wh-in-situ can be related to a C 0 housing [wh] via either of two mechanisms: movement or binding of a choice function variable. Conceptually, to the extent that movement and variable binding are both well-motivated and independently existing mechanisms in the syntax of wh-in-situ, both should be available for a given wh-element, unless further restrictions apply.14 This means that in a double question in Serbo-Croatian, the two fronted wh-elements are each expected to be able to relate to C 0 either by movement or by binding (of a choice function variable). This in turn predicts that when both wh-elements move to C 0 , Absorption results in a PL reading; in other scenarios, the question should receive an SP interpretation. This prediction is borne out, since such questions in Serbo-Croatian may receive either a PL or an SP reading. On the other hand, as Bosˇkovic´ (2001) points out, Serbo-Croatian MF in long movement and embedded contexts receives only a PL answer. That in such syntactic setups Serbo-Croatian MF involves wh-movement to CP is shown independently of the PL/SP interpretation issue in Bosˇkovic´ 1997, 1998. In present terms, inasmuch as such long and embedded MF involves true wh-movement to CP, the PL interpretation is predicted: Absorption takes place in (multiple) Spec,CP. Russian is demonstrated by Stepanov (1998), in the footsteps of Bosˇkovic´, to be a ‘‘low-fronting’’ language (according to Stepanov, whelements undergo focus fronting); and the same extends to Polish along the lines of this genre of analysis. That means that Russian and Polish display what Serbo-Croatian exhibits in simplex clauses: no overt [wh]checking movement. Wh-movement, as in Serbo-Croatian simplex-clause short MF too, may or may not happen covertly to all wh-elements, resulting in the same PL/SP ambiguity (see Stepanov 1998).15 11.5.2 Consequences for a Wider Typology Let us look beyond Eastern European MF languages. Pesetsky (2000) argues that English is what he terms a ‘‘multispeciﬁer complementizer’’ (Cm-spec ) language. Grossly simplifying here, this means that in multiple questions, English C 0 requires multiple wh-movement to itself, where multiple stands for ‘minimally two’. Bulgarian is demonstrated to also

308

Bala´zs Sura´nyi

strictly require (at least) two wh-phrases (overtly) in its speciﬁer (Pesetsky 2000), and English is taken to be like Bulgarian in its covert structure. I will adopt this analysis of English C 0 here. If English C 0 attracts both whelements in a double question, then on the account developed here only a PL reading is generated—correctly, since English (ordinary) multiple whquestions are known to yield only a PL interpretation (Bosˇkovic´ 1998). When the in-situ wh-phrase is located in a strong island, unexpectedly, only an SP reading is obtainable (as ﬁrst pointed out in Sura´nyi 2002 for Hungarian and in Dayal 2002 for English). In present terms, this is accounted for, since the island-internal wh-element can be related to C 0 only by binding of a choice function variable, not by movement (a prerequisite for a PL reading). A similar observation, noted by Pesetsky (2000), is that certain quantiﬁcational interveners (cf. Beck 1996) have an identical e¤ect on the interpretation of English multiple questions: only an SP reading is produced. This is analogous to the previous case: again, the in-situ wh-phrase cannot use the movement strategy; hence, instead of a PL reading, an SP reading is produced. Further, it is well known that one context where English can (apparently) violate Superiority arises when a lower wh-phrase crosses over a structurally superior D-linked wh-in-situ, as in (34). (34) Which girl did which boy invite? Barss (1990, 2000) notes that in such sentences only an SP reading is available. Assuming, as I did above for Hungarian wh-in-situ, that the choice function variable (a determiner) checks [wh] o¤ the wh-in-situ in English, it follows that no Superiority violation per se occurs here: given that [wh] on the wh-subject is not visible externally (being checked internal to the wh-expression), it is the structurally highest wh-phrase bearing an unsaturated [wh] that is attracted. This would explain why an apparent Superiority violation can occur here. At the same time, it accounts for why only an SP reading is generated: this is because the in-situ wh-phrase in Superiority-violating cases does not undergo movement to CP, but is saturated via binding of a choice function variable—an approach essentially analogous to Barss’s (2000). Evidence for this analysis of wh-in-situ in Superiority-violating cases comes from an observation made by Pesetsky (2000): namely, that such in-situ wh-phrases—in contrast to ordinary English wh-in-situ (see (35a)) —do not license antecedent-contained deletion (see (35b)). Since resolution of antecedent-contained deletion requires (category) movement (May 1985; Larson and May 1990) (here of the containing wh-expression

Wh-Saturation and Interpretation in Multiple Wh-Movement

309

marked by italics in (35)), antecedent-contained deletion is not licensed in the wh-in-situ headed by which boy in (35b), given that such an in-situ whphrase does not undergo covert movement at all: it must be interpreted via binding of a choice function variable for the apparent Superiority violation to have taken place.16 (35) a.

Which girl [VP invited [which student that John (also) did [VP e]]]? b. *I need to know which girl i Sue [VP ordered [which boy that Mary (also) did [VP e]] to congratulate t i ]

German appears to be reducible to the English case, along the same lines, including the SP e¤ect of quantiﬁcational intervention (Pesetsky 2000, pers. comm. to Sigrid Beck). In contrast to their English counterparts, however, apparently Superiority-violating German multiple questions still receive a PL reading. This can be explained if apparently Superiority-violating questions in German can be reduced to the case of regular multiple wh-interrogatives. Precisely this is what obtains if the Superiority violations are not real—which then is the reason why they are routinely possible in German in the relevant contexts. Assuming that the patterns at hand involve scrambling or some other reordering prior to wh-movement (cf. Fanselow 1991, 2001) su‰ces to make this reduction step: then Superiority is not in fact violated at all.17 French is similar to English and German when it overtly exhibits the English pattern of wh-phrases—hence, it is liable to the same analysis. On the other hand, French binary questions, besides exhibiting the English pattern, are known to optionally have both wh-phrases in situ. French in-situ wh-questions have been analyzed by Cheng and Rooryck (2000) as involving an intonational(ly realized) particle in the CP projection. If French C 0 has the [multiple] value in multiple questions, then having one particle in CP does not su‰ce: one of the in-situ wh-phrases needs to move to CP too. As for a second in-situ wh-phrase, similarly to the case of Serbo-Croatian simple clauses, it may be related to C 0 either via movement (resulting in a PL reading) or via binding of a choice function variable (resulting in an SP reading). Hence, if French C 0 is like English C 0 , then French binary in-situ questions are predicted to optionally receive either a PL or an SP reading. This appears once again to be a correct prediction. I will conclude this brief typological outline with a pure in-situ language: Japanese. Recall that Serbo-Croatian simplex clauses and Russian/Polish have been taken to be masked in-situ languages. Mutatis

Bala´zs Sura´nyi

310

mutandis, the analysis I proposed for them should hold for a par excellence wh-in-situ language like Japanese as well: either multiple movement or binding of choice function variables may apply. Indeed, analogously to what is found in the masked in-situ languages, in Japanese, multiple questions may receive both a PL and an SP answer (Hagstrom 1998). Interestingly, as Kitagawa, Roehrs, and Tomioka (2004) show, if two wh-phrases are located inside a wh-island, or in the focus of a cleft (which may be taken to be an island itself, too; Mamoru Saito, pers. comm.), then the PL reading is lost. Brody (1995), building on an important insight by Longobardi (1991), clearly demonstrates that the wh-phrase of single questions and the ‘‘primary’’ wh-operator of multiple questions obey Subjacency (e.g., a wh-phrase embedded in two Condition on Extraction Domain islands is ungrammatical). Then Kitagawa, Roehrs, and Tomioka’s ﬁnding is not unexpected: it can be assimilated to the facts cited from English above: namely, the wh-island and the focus of clefts allow only the binding strategy, hence only an SP interpretation. Signiﬁcantly, Kitagawa, Roehrs, and Tomioka note that an interpretive additional wh-e¤ect arises when one wh-element is outside the whisland and the other is inside: in this scenario, the PL reading (besides the SP reading) becomes available again. Richards (1998) has analyzed such constructions (and ones discussed in Longobardi 1991 and Brody 1995) as a reﬂex of the Principle of Minimal Compliance: if the ﬁrst (primary) wh-item obeys Subjacency (pays the ‘‘Subjacency tax’’), then a further wh-element moving to the same projection (‘‘secondary’’ wh-element, in Brody’s terms) is exempted from obeying it. In the present context, this means that the secondary wh-item in such constructions is free to be related to the matrix C 0 via movement; hence, the PL reading is correctly expected to be available again. I have argued in this section that the approach to the PL/SP distinction advocated on the basis of Hungarian in terms of presence versus absence of multiple movements is indeed justiﬁable from a typological perspective. 11.6

Conclusion

In this chapter, I hope to have substantiated four general points: (1) true multiple wh-movement exists, and is realized in Hungarian; (2) focus structure plays a crucial role in determining answerhood conditions of multiple questions in Hungarian; (3) multiple movement correlates with

Wh-Saturation and Interpretation in Multiple Wh-Movement

311

list interpretations in multiple wh-interrogatives; and (4) binding choice function variables and covert movement are both necessary in the grammar of wh-in-situ. Notes I gratefully acknowledge the support of OTKA (Hungarian Scientiﬁc Research Fund) grant D 048454. 1. Bolinger (1978) calls the wh-phrase in a multiple question that contributes the sorting key the ‘‘topic’’ of the multiple question. Note that in English-type languages, the sorting key can be provided either by the fronted or by the surface in-situ wh-elements. See also Grohmann 2000, this volume, Boeckx and Grohmann 2004, and Wu 1996, 1999, regarding topicality of wh-phrases and whtopicalization in Chinese. Jaeger (2004) argues that topicality is a crucial factor responsible for the ordering of fronted wh-elements in Bulgarian. 2. Or one wh-cluster if in a given language wh-elements adjoin to each other to form a complex phrase, as in Rudin 1988, or more recently in Grewendorf 2001 and Sabel 2001. 3. As suggested in Krifka 2001, even a multiple interrogative can only question one entity (though this one object can be internally complex). 4. Though (18a) may be also construed as an echo question, it need not appear in an echo context. When it is used as an echo question, the wh-in-situ must have a rise-fall intonation with the fall on the last syllable. The in-situ wh-phrase in (18a), however, can receive the normal intonation associated with English-type double questions in Hungarian: one with an initial rise on the ﬁrst syllable of the whphrase followed by a continuous fall in the remainder of the wh-expression. Some speakers prefer (18a) with pitch compression from after the ﬁrst wh-element up to the in-situ wh-phrase. Japanese also requires pitch compression up to an embedded wh-in-situ element when that takes matrix scope; see Deguchi and Kitagawa 2002 and Ishihara 2002. 5. In fact, Hungarian also allows one wh-phrase to be fronted, and two (or more) wh-phrases to be left in situ (Hungarian contrasts in this respect with Bulgarian, where such patterns are ungrammatical; see Pesetsky 2000). 6. A Wh-Island violation is illustrated in (i), and multiple long wh-movement is exempliﬁed in (ii). (i) Kit ke´rdezte´l, hogy mit mondott? who-acc ask-past-2sg that what-acc say-past-3sg ‘Who did you ask said what?’ (ii) Ki kinek gondolod, hogy bajt okozhat? who who-dat think-2sg that trouble-acc cause-pot-3sg ‘Who do you think can cause problems for who?’ Moreover, variants of (i) with a cluster of wh-elements either in the matrix or in the embedded clause (or in both) are also grammatical. On the present analysis, Hungarian forms a minimal pair with Bulgarian if Bosˇkovic´’s (1998, 2000) analysis of Bulgarian is correct. Bosˇkovic´ claims that in

312

Bala´zs Sura´nyi

Bulgarian exactly one wh-phrase is fronted to check [wh] (and [foc]), and further wh-phrases check only [foc], while in Hungarian exactly the reverse is true. 7. (22) involves truly adverbial wh-in-situ, in contrast to quasi-argumental ‘when’/ ‘where’, which are admissible as wh-in-situ in appropriate contexts, similarly to the situation in English. Reinhart (1997) builds on work by Higginbotham (1985) in taking true adverbials to lack an index argument, for Reinhart, an N-set. By assuming that true adverbials lack D, as I do here, we can derive several of their properties other than inability to stay in situ in languages like English and Hungarian, once they are related syntactically to the absence of a D head and D projection: nonpresuppositionality/untopicalizability (D being responsible for referentiality), lack of Superiority violations (choice function variables are D heads), inability to escape weak islands (only DPs/topicalizable elements can escape weak islands), lack of adverbial resumptives (resumptive pronouns are determiners). See Rullmann and Beck 1998, where D-linked wh-phrases are decomposed into a deﬁnite determiner, a descriptive restriction, and a wh-component; and also Boeckx and Grohmann 2004, which capitalizes on Rullmann and Beck 1998 in deriving radical reconstruction e¤ects regularly arising with D-linked wh-phrases—as well as a number of other properties shared by D-linked wh-phrases as scrambled elements—by assuming (sub)movement to extract material from below a (stranded) D head. See also Rizzi 2001 for a proposal that nonlocal, weak-island¯ -chains must involve DPs, given that only DPs can form binding escaping A chains, entering a genuine referential dependency. 8. For example, in [DP f[wh] [NP who[wh] ]], the N who checks against D, occupied by f. 9. Note that I am assuming that the in-situ wh-phrase containing a choice function determiner can (covertly) raise to FocP. This is similar to what Hagstrom (1998) assumes for pure wh-in-situ languages like Japanese, with the di¤erence that according to Hagstrom, in those languages an existential quantiﬁer over choice functions (dubbed ‘‘Q’’) raises alone, stranding wh-in-situ phrases. 10. If such covert movements are instances of pure Agree (Chomsky 2000), then Agree with the relevant functional head (here, Foc) brings the second wh-element to a position structurally adjacent (in the sense relevant for Absorption) to the ﬁrst one by establishing a syntactic relation with the same position. Nevertheless, there is evidence (e.g., from antecedent-contained deletion contexts) that such secondary movements may involve not only Agree, but also category movement (e.g., Pesetsky 2000). See section 11.5.2 for related discussion. 11. As a consequence, clauses with an English-type pattern can be arguments of verbs like ‘list’ or ‘enumerate’; see (i). This property is used to control for the PL reading of multiple interrogatives in Sura´nyi 2002 and in Kitagawa, Roehrs, and Tomioka 2003. (i) Sorold fel, ki ne´zett ra´ kire list-imp pref who looked on who-on ‘Please list who looked at who.’

Wh-Saturation and Interpretation in Multiple Wh-Movement

313

12. The crosslinguistic distribution of the availability of PL and SP readings is investigated in much recent work. Hagstrom (1998) suggests that a Q morpheme, interpreted as an existential quantiﬁer over choice functions, is moved in wh-insitu languages to the C domain (see also note 9). Depending on whether Q is generated low on a low wh-element, or high (i.e., above all wh-elements), a PL or an SP reading results, respectively. Bosˇkovic´ (2001) extends this analysis to languages with overt wh-movement. He assumes that wh-elements and the Q morpheme are members of the same class for Relativized Minimality. As a consequence, in singular wh-fronting languages like English, the independently required movement of a wh-phrase across a high Q morpheme is ruled out, which in turn derives the unavailability of an SP reading in the basic case. Grohmann (2000) adapts the Hagstrom-Bosˇkovic´ account to his domain-driven framework. Barss (2000), assuming as I do here that Absorption is responsible for PL readings, derives the loss of the PL interpretation in Superiority-violating cases in English (and several other interpretive e¤ects of Superiority violations) by proposing that the superior wh-phrase that is crossed over does not move at all, but is interpreted via choice function variable binding. See Pesetsky 2000 for an alternative analysis of Superiority violations, and their interpretation. Dayal (2002) shows for English what I elsewhere show independently for Hungarian (Sura´nyi 2002): the PL reading is lost if the wh-in-situ in the English-type multiple wh-interrogative is within an island. For reasons of space, I will not attempt to evaluate these accounts or compare them with the present one; I only note that my approach is consonant with Dayal’s and Barss’s treatment of English in assuming that multiple movement is necessary for a PL reading to obtain. 13. Structural adjacency in present terms means speciﬁers of the same feature (i.e., [wh]). 14. Reinhart (1997) takes binding of choice function variables to be the sole mechanism for in-situ wh-interpretation. Pesetsky (1987) also assumes (unselective) binding to be the only mechanism available for D-linked wh-in-situ. It is not clear, however, why such an apparently stipulative restriction should hold. The descriptive coverage in Pesetsky’s and Reinhart’s papers is in fact undiminished if this stipulation is dropped. 15. Czech, as reported by Meyer (2003), is the real odd one out among the Eastern European MF languages reviewed so far: MF in Czech receives only an SP reading. Meyer argues that Czech MF is di¤erent from Polish/Russian MF in one crucial regard: the cluster of fronted wh-elements forms a constituent (see Rudin 1988), and this cluster formation can arguably happen prior to the last movement step, which takes the cluster to CP (cf. Grewendorf 2001; Sabel 2001). This may well be the reason for the unavailability of a PL interpretation if the absorbed wh-interpretation requires distinct wh-elements in the speciﬁer of a head bearing [wh]. This is because the cluster creates an island already below C 0 for movements of the individual wh-elements that would take them to Spec,C (which would be excorporation). 16. In a minimalist approach that incorporates both covert category movement and covert feature movement (or pure Agree) (like Pesetsky’s), the obvious alternative to the account of the contrast in (35) that I have provided is to claim that

314

Bala´zs Sura´nyi

the in-situ wh-element undergoes pure feature movement, which does not su‰ce to license antecedent-contained deletion. This is the explanation given in Pesetsky 2000. 17. Sabel (1998) credits Grohmann (1997) with a pre-wh-A-movement solution for German Superiority violations. Grohmann’s (2000) account of German multiple wh-interrogatives di¤ers sharply from the present one: he proposes that German fronts all its wh-elements to the CP domain, thereby assimilating German to the Bulgarian case. References Barss, Andrew. 1990. Optional movement, Absorption and the interpretation of wh-in-situ. Paper presented at NELS 21, Universite´ du Que´bec a` Montre´al; ms., University of Arizona, Tucson. Barss, Andrew. 2000. Minimalism and asymmetric wh-interpretation. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 31–52. Cambridge, Mass.: MIT Press. Beck, Sigrid. 1996. Quantiﬁed structures as barriers for LF-movement. Natural Language Semantics 4, 1–56. Boeckx, Cedric, and Kleanthes Grohmann. 2004. SubMove: Towards a uniﬁed account of scrambling and D-linking. In David Adger, Ce´cile de Cat, and George Tsoulas, eds., Peripheries: Syntactic edges and their e¤ects, 241–257. Dordrecht: Kluwer. Bolinger, Dwight. 1978. Asking more than one thing at a time. In Henry Hiz, ed., Questions, 87–105. Dordrecht: Reidel. Bosˇkovic´, Zˇeljko. 1997. Superiority e¤ects with multiple wh-fronting in SerboCroatian. Lingua 102, 1–20. Bosˇkovic´, Zˇeljko. 1998. Multiple wh-fronting and economy of derivation. In Emily Curtis, James Lyle, and Gabriel Webster, eds., Proceedings of the 16th West Coast Conference on Formal Linguistics, 49–63. Stanford, Calif.: CSLI Publications. Bosˇkovic´, Zˇeljko. 1999. On multiple feature checking. In Samuel David Epstein and Norbert Hornstein, eds., Working minimalism, 159–187. Cambridge, Mass.: MIT Press. Bosˇkovic´, Zˇeljko. 2000. What is special about multiple wh-fronting? In Masako Hirotani, Andries Coetzee, Nancy Hall, and Ji-Yung Kim, eds., Proceedings of NELS 30, 83–107. Amherst: University of Massachusetts, GLSA. Bosˇkovic´, Zˇeljko. 2001. On the interpretation of multiple questions. Linguistic Variation Yearbook 1, 1–15. Bosˇkovic´, Zˇeljko. 2002. On multiple wh-fronting. Linguistic Inquiry 33, 351–383. Brody, Michael. 1990. Remarks on the order of elements in the Hungarian focus ﬁeld. In Istva´n Kenesei, ed., Approaches to Hungarian 3, 95–121. Szeged: JATE. Brody, Michael. 1995. Lexico-Logical Form: A radically minimalist theory. Cambridge, Mass.: MIT Press.

Wh-Saturation and Interpretation in Multiple Wh-Movement

315

Chang, Lisa. 1997. Wh-in-situ phenomena in French. Master’s thesis, University of British Columbia. Cheng, Lisa L.-S. 1991. On the typology of wh-questions. Doctoral dissertation, MIT. Published, New York: Garland, 1997. Cheng, Lisa L.-S., and Johan Rooryck. 2000. Licensing wh-in-situ. Syntax 3, 1– 19. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Citko, Barbara. 1998. On multiple wh-movement in Slavic. In Zˇeljko Bosˇkovic´, Steven Franks, and William Snyder, eds., Formal Approaches to Slavic Linguistics: The Connecticut Meeting, 97–114. Ann Arbor, Mich.: Michigan Slavic Publications. Comorovski, Ileana. 1989. Discourse and the syntax of multiple constituent questions. Doctoral dissertation, Cornell University. Culicover, Peter W., and Michael S. Rochemont. 1983. Stress and focus in English. Language 59, 123–149. Dayal, Veneeta. 2002. Single-pair versus multiple-pair answers: Wh-in-situ and scope. Linguistic Inquiry 33, 512–520. Deguchi, Masanori, and Yoshihisa Kitagawa. 2002. Prosody and wh-questions. In Masako Hirotani, ed., Proceedings of NELS 32, 73–97. Amherst: University of Massachusetts, GLSA. Fanselow, Gisbert. 1991. Minimale Syntax. Habilitationsschrift, University of Passau. Fanselow, Gisbert. 2001. Features, y-roles, and free constituent order. Linguistic Inquiry 32, 405–437. Garrett, Edward. 1996. Wh-in-situ and the syntax of distributivity. In Edward Garrett and Felicia Lee, eds., Syntax at sunset, 129–145. UCLA Working Papers in Linguistics and Semantics 1. Los Angeles: UCLA, Department of Linguistics. Grewendorf, Gu¨nther. 2001. Multiple wh-fronting. Linguistic Inquiry 32, 87–122. Grohmann, Kleanthes. 1997. German superiority. Groninger Arbeiten zur germanistischen Linguistik 40, 97–107. Grohmann, Kleanthes. 2000. Proliﬁc peripheries: A radical view from the left. Doctoral dissertation, University of Maryland, College Park. Hagstrom, Paul. 1998. Decomposing questions. Doctoral dissertation, MIT. Higginbotham, James. 1985. On semantics. Linguistic Inquiry 16, 547–593. Higginbotham, James, and Robert May. 1981. Questions, quantiﬁers, and crossing. The Linguistic Review 1, 41–80. Horva´th, Julia. 1986. Focus in the theory of grammar and the structure of Hungarian. Dordrecht: Foris.

316

Bala´zs Sura´nyi

Horva´th, Julia. 1998. Multiple wh-phrases and the wh-scope-marker strategy in Hungarian interrogatives. Acta Linguistica Hungarica 45, 31–60. Ishihara, Shinichiro. 2002. Invisible but audible wh-scope marking: Whconstruction and deaccenting in Japanese. In Line Mikkelsen and Christopher Potts, eds., Proceedings of the 21st West Coast Conference on Formal Linguistics, 180–193. Somerville, Mass.: Cascadilla Press. Jaeger, T. Florian. 2004. Topicality and superiority in Bulgarian wh-questions. In Olga Arnaudova, Wayles Browne, Maria Luisa Rivero, and Danijela Stojanovic, eds., Formal Approaches to Slavic Linguistics 12: The Ottawa Meeting 2003, 207– 228. Ann Arbor, Mich.: Michigan Slavic Publications. E´. Kiss, Katalin. 1992. A to¨bbszo¨ro¨s ke´rde´sekro˝l (On multiple questions). In La´szlo´ Hunyadi, Zsolt Lengyel, Kinga Klaudy, and Ga´bor Sze´kely, eds., Ko¨nyv Papp Ferencnek (Festschrift for Ferenc Papp), 79–90. Debrecen: KLTE. E´. Kiss, Katalin. 1993. Wh-movement and speciﬁcity. Natural Language and Linguistic Theory 11, 83–120. E´. Kiss, Katalin. 1994. Sentence structure and word order. In Ferenc Kiefer and Katalin E´. Kiss, eds., The syntactic structure of Hungarian, 1–90. New York: Academic Press. E´. Kiss, Katalin. 2002. Hungarian syntax. Cambridge: Cambridge University Press. Kitagawa, Yoshihisa, Dorian Roehrs, and Satoshi Tomioka. 2004. Multiple whinterpretation. In Hang-Jin Yoon, ed., Generative grammar in a broader perspective: Proceedings of the Fourth GLOW in Asia, 209–233. Seoul: Seoul National University. Krifka, Manfred. 2001. Quantifying into question acts. In Caroline Fe´ry and Wolfgang Sternefeld, eds., Audiatur vox sapientia: A festschrift for Arnim von Stechow, 287–319. Berlin: Akademie der Wissenschaften. Kuno, Susumu. 1982. The focus of the question and the focus of the answer. In Robinson Schneider, Kevin Tuite, and Robert Chametzky, eds., Papers from the Parasession on Nondeclaratives, 134–157. Chicago: University of Chicago, Chicago Linguistic Society. Kuno, Susumu, and Ken-ichi Takami. 1993. Grammar and discourse principles: Functional syntax and GB Theory. Chicago: University of Chicago Press. Larson, Richard, and Robert May. 1990. Antecedent containment or vacuous movement: Reply to Baltin. Linguistic Inquiry 21, 103–122. Lasnik, Howard. 1999. On feature strength: Three minimalist approaches to overt movement. Linguistic Inquiry 30, 197–217. Lipta´k, Aniko´. 2001. On the syntax of wh-items in Hungarian. Leiden: LOT. Longobardi, Giuseppe. 1991. In defence of the Correspondence Hypothesis: Island e¤ects and parasitic constructions in Logical Form. In C.-T. James Huang and Robert May, eds., Logical structure and linguistic structure, 149–198. Dordrecht: Reidel.

Wh-Saturation and Interpretation in Multiple Wh-Movement

317

May, Robert. 1985. Logical Form: Its structure and derivation. Cambridge, Mass.: MIT Press. Meyer, Roland. 2003. Superiority e¤ects and structural distinctions in Polish and Czech wh-questions. In Adam Przepio´rkowski and Piotr Banski, eds., Morphosyntactic investigations: Proceedings of the GLiP-5 Conference held in Warszawa, Poland. Warsaw: Polish Academy of Sciences, Institute of Computer Science. Nishigauchi, Taisuke. 1990. Quantiﬁcation in the theory of grammar. Dordrecht: Kluwer. Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Eric Reuland and Alice G. B. ter Meulen, eds., The representation of (in)deﬁniteness, 98–129. Cambridge, Mass.: MIT Press. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: MIT Press. Puska´s, Genoveva. 2000. Word order in Hungarian: The syntax of A-positions. Amsterdam: John Benjamins. Reinhart, Tanya. 1997. Wh-in-situ in the framework of the Minimalist Program. Natural Language Semantics 6, 29–56. Richards, Norvin. 1997. What moves where when in which language? Doctoral dissertation, MIT. Richards, Norvin. 1998. The Principle of Minimal Compliance. Linguistic Inquiry 29, 599–629. Rizzi, Luigi. 2001. Reconstruction, weak island sensitivity and agreement. In Carlo Cecchetto, Gennaro Chierchia, and Maria-Teresa Guasti, eds., Semantic interfaces: Reference, anaphora and aspect, 155–176. Stanford, Calif.: CSLI Publications. Rochemont, Michael S. 1978. A theory of stylistic rules in English. Doctoral dissertation, University of Massachusetts, Amherst. Rochemont, Michael S. 1986. Focus in generative grammar. Amsterdam: John Benjamins. Rudin, Catherine. 1988. On multiple questions and multiple wh-fronting. Natural Language and Linguistic Theory 6, 445–501. Rullmann, Hotze, and Sigrid Beck. 1998. Presupposition projection and the interpretation of which-questions. In Devon Strolovitch and Aaron Lawson, eds., Proceedings of Semantics and Linguistic Theory VIII (SALT VIII), 215–232. Ithaca, N.Y.: Cornell University, CLC Publications. Sabel, Joachim. 1998. Principles and parameters of wh-movement. Habilitationsschrift, Johann Wolfgang Goethe-Universita¨t Frankfurt am Main. Sabel, Joachim. 2001. Deriving multiple head and phrasal movement: The Cluster Hypothesis. Linguistic Inquiry 32, 532–547. Stepanov, Arthur. 1998. On wh-fronting in Russian. In Pius N. Tamanji and Kiyomi Kusumoto, eds., Proceedings of NELS 28, 453–467. Amherst: University of Massachusetts, GLSA.

318

Bala´zs Sura´nyi

Stjepanovic´, Sandra. 1995. Short distance movement of wh-phrases in SerboCroatian matrix clauses. Ms., University of Connecticut, Storrs. Stjepanovic´, Sandra. 1998. Movement of wh-phrases in Serbo-Croatian matrix clauses. Paper presented at the Comparative Slavic Morphosyntax Workshop, Spencer, Ind. Sura´nyi, Bala´zs. 2002. Multiple operator movements in Hungarian. Utrecht: LOT. Sura´nyi, Bala´zs. 2005. Triggering wh-fronting. In Christopher J. Pino´n and Pe´ter Sipta´r, eds., Approaches to Hungarian 9, 231–259. Budapest: Akade´miai. Szabolcsi, Anna. 1997. Strategies for scope taking. In Anna Szabolcsi, ed., Ways of scope taking, 109–154. Dordrecht: Kluwer. Tsai, Wei-Tien Dylan. 1994. On economizing the theory of A-bar dependencies. Doctoral dissertation, MIT. Wu, Jianxin. 1996. Wh-topic, wh-focus and wh-in-situ. In Juan Carlos Castillo, Viola Miglio, and Julien Musolino, eds., University of Maryland working papers in linguistics 4, 173–192. College Park: University of Maryland, Department of Linguistics. Wu, Jianxin. 1999. Syntax and semantics of quantiﬁcation in Chinese. Doctoral dissertation, University of Maryland, College Park.

Chapter 12 On the Structural Height of Reason Wh-Adverbials: Acquisition and Consequences

12.1

Heejeong Ko

Introduction

In this chapter, I discuss the structural position of the reason wh-adverbial ‘why’ in three typical wh-in-situ languages: Korean (K), Japanese (J), and Mandarin Chinese (C). I argue that ‘why’ in these languages is externally merged in Spec,CP as a CP modiﬁer, and that this high base position of ‘why’ plays a central role in determining its unique syntax in in-situ languages. I argue for this proposal on the basis of a peculiar interaction between ‘why’ and scope-bearing elements in Korean and Japanese (K/J) (section 12.2; Ko 2005b). I show that the proposal is further supported by child Korean acquisition data (section 12.3). Speciﬁcally, I demonstrate that child Korean makes systematic distinctions between way ‘why’ and other wh-adjuncts such as eti ‘where’ with respect to their placement, distinctions that my proposal predicts. Furthermore, I explore the implications of the ﬁne structure of CP for the syntax of ‘why’ in in-situ languages (section 12.4). Investigating the antisuperiority e¤ect (Saito 1982; Watanabe 1992; Chung 1996), I argue that there are two distinct C heads in in-situ languages for licensing wh-interrogatives. In particular, ‘why’ is licensed by the lower C (C Int ), whereas the other whphrases are licensed by the higher C (C Foc ) (cf. Rizzi 1999). I show that this approach, in conjunction with other independent properties of in¯ -scrambling, in particusitu languages (the presence or absence of A lar), accounts for crosslinguistic variations between K/J and Chinese in multiple wh-questions. This chapter is in line with a series of studies arguing that a reason wh-adverbial can be licensed in its base position without undergoing movement (Rizzi 1990, 1999; Collins 1991; Bromberger 1992; Hegarty 1992; Lin 1992; Bosˇkovic´ 2000; McCloskey 2002; Ko 2005b).

320

12.2

Heejeong Ko

External Merge Position of ‘Why’ in Korean and Japanese

In this section, I discuss a puzzle concerning the interactions between ‘why’ and a scope-bearing element (SBE) in Korean and Japanese and introduce the main proposal of this chapter.1 12.2.1 Puzzles: Nonuniform Behavior of ‘Why’ in Korean and Japanese Korean and Japanese are SOV scrambling languages. In most cases, elements in K/J may freely undergo scrambling, yielding various types of noncanonical word order.2 Interestingly, however, there are some environments in which possible word orders are restricted. In particular, it is well known that K/J impose restrictions on relative word order between a wh-phrase and an SBE (e.g., Hoji 1985; Kim 1991; Beck and Kim 1997; Hagstrom 1998). As illustrated in (1) and (2), when a wh-phrase and an SBE co-occur in a sentence, the SBE cannot precede (and c-command) the wh-phrase in K/J. The SBEs inducing this e¤ect include negative polarity items (NPIs) such as amwuto/dareka ‘anyone’, -pakkey/-sika ‘-only’; negation anh/nai ‘not’; and some nonpolarity items such as man ‘only’ (K), to/mo ‘also’, nwukwunka/dareka ‘(nonspeciﬁc) someone’, and nwukwuna/daremo ‘everyone’. (For the sake of space, I present only the examples with NPIs in this section, but see Ko 2005b for other K/J examples.) (1) a. *Amwuto mwues-ul ilk-ci-anh-ass-ni? anyone what-acc read-ci-not-past-Q b. Mwues-uli amwuto t i ilk-ci-anh-ass-ni? what-acc anyone read-ci-not-past-Q ‘What did no one read?’

(Korean)

(2) a. *Hanako-sika nani-o yoma-nai no? Hanako-only what-acc read-not Q b. Nani-oi Hanako-sika t i yoma-nai no? what-acc Hanako-only read-not Q ‘What will only Hanako read?’

(Japanese)

A variety of approaches have been proposed to capture the paradigm in (1)–(2) (Hoji 1985; Takahashi 1990; Kim 1991; Sohn 1995; Beck and Kim 1997; Tanaka 1997; Hagstrom 1998; Lee and Tomioka 2000; Pesetsky 2000; Kim 2002; Kratzer and Shimoyama 2002). Although this chapter can be made compatible with other approaches, I present an analysis slightly modiﬁed from Beck and Kim 1997. This permits a

Structural Height of Reason Wh-Adverbials

321

reasonably simple presentation of my arguments in a manner consistent with other current work in syntax. Following Chomsky (1995), I assume that a wh-phrase contains an uninterpretable wh-feature [uWH] to be checked o¤ by [þQ], and that [þQ] is hosted by a question morpheme Q in a head C. Adapting Beck and Kim’s proposal, I take the Intervention E¤ect Constraint stated in (3) as a constraint on wh-movement at LF.3 (3) Intervention E¤ect Constraint (modiﬁed from Beck and Kim 1997) At LF, a wh-phrase cannot move across an SBE to its checking (scope) position. a. *[ . . . Q SBE WH . . . ]: (1a), (2a) b. t [ . . . Q WHi SBE t i . . . ]: (1b), (2b) a a

As described in (3a), ‘what’ in (1a) and (2a) cannot move across the NPI to its checking position at LF, and the unlicensed ‘what’ thereby renders the sentence ungrammatical. On the other hand, a scrambled ‘what’ as in (1b) and (2b) can be licensed because the NPI does not block the LF movement of the overtly scrambled ‘what’, as illustrated in (3b). What is interesting for the present discussion is the distribution of way ‘why’ in Korean and naze ‘why’ in Japanese. As shown in (4) and (5), in some contexts ‘why’ may precede or follow an SBE, unlike the paradigm seen in (1) and (2) (Miyagawa 1997b; Cho 1998; Kuwabara 1998; Watanabe 2000).4 (4) a. Amwuto way ku chayk-ul ilk-ci-anh-ass-ni? anyone why that book-acc read-ci-not-past-Q ‘Why did no one read that book?’ b. Way amwuto ku chayk-ul ilk-ci-anh-ass-ni?

(Korean)

(5) a. Taroo-sika naze sono hon-o yoma-nakat-ta Taroo-only why that book-acc read-not-past no? Q ‘Why did only Taroo read that book?’ b. Naze Taroo-sika sono hon-o yoma-nakat-ta no?

(Japanese)

Given the contrast between way/naze and other wh-phrases shown above, one might think that way/naze is just exempt from the intervention e¤ect (for this approach, see Cho 1998; Kuwabara 1998; Watanabe 2000). This conjecture, however, is incorrect. Way/Naze does show the

322

Heejeong Ko

intervention e¤ect in other contexts. This point is illustrated in (6)–(9) (see also Miyagawa 1999 for Japanese). As illustrated in (6a) and (7a), when way/naze is merged in an embedded declarative clause, it cannot be preceded by an NPI in the higher clause. The grammaticality of (6b) and (7b), on the other hand, shows that way/naze may obtain a long-distance reading if it is not preceded by an NPI.5 If way/naze were simply an exception to the intervention e¤ect, we would expect (6a) and (7a) to be as grammatical as (6b) and (7b), contrary to fact. Hence, the claim that way/naze is just an exception to the intervention e¤ect is untenable. (6) a. *Amwuto [John-i way saimha-yess-ta-ko] (Korean) anyone [John-nom why resign-past-dec-C] malha-ci-anh-ass-ni? say-ci-not-past-Q ‘What is the reason x such that no one said that John resigned for x?’ b. Mary-nun [John-i way saimha-yess-ta-ko] malha-yess-ni? Mary-top [John-nom why resign-past-dec-C] say-past-Q ‘What is the reason x such that Mary said that John resigned for x?’ (7) a. *Hanako-sika [Taroo-ga naze kuru to] (Japanese) Hanako-only [Taroo-nom why come C] iwa-nakat-ta no? say-not-past Q ‘What is the reason x such that only Hanako said that Taroo will come for x?’ b. Hanako-ga [Taroo-ga naze kuru to] itta no? Hanako-nom [Taroo-nom why come C] said Q ‘What is the reason x such that Hanako said that Taroo will come for x?’ The ungrammaticality of (8) and (9) provides further support for the observation that way/naze is subject to the intervention e¤ect in certain contexts. As demonstrated in (8) and (9), way/naze cannot be preceded by an NPI even when they are clause-mates.6 (8) *John-un [amwuto way ku chayk-ul John-top [anyone why that book-acc ilk-ci-anh-ass-ta-ko] malha-yess-ni? read-ci-not-past-dec-C] say-past-Q

(Korean)

Structural Height of Reason Wh-Adverbials

‘What is the reason x such that John said that book for x?’ ‘What is the reason x such that John said that that book?’ (9) *John-wa [Mary-sika naze sono hon-o John-top [Mary-only why that book-acc yoma-na-katta-to] itta no? read-not-past-C] said Q ‘What is the reason x such that John said that book for x?’ ‘What is the reason x such that John said that read that book?’

323

no one read that for x, no one read (Japanese)

only Mary read that for x, only Mary

The paradigm discussed in this section indicates that what is crucial for the distribution of way/naze in K/J is the type of the clause that way/naze is merged into in the overt syntax. If way/naze is merged within an interrogative clause, as in (4) and (5) (local construal), it can be preceded by an NPI. In this respect, way/naze behaves di¤erently from the other whphrases. On the other hand, when way/naze is merged within a declarative clause, as in (6)–(9) (nonlocal construal), it cannot be preceded by an NPI. In this respect, way/naze behaves just like other wh-phrases. In what follows, I show that this nonuniform behavior of way/naze in K/J can be naturally accounted for by assuming a particular external Merge position of way/naze. 12.2.2 Proposal and Analysis I argue that way/naze is always subject to the intervention e¤ect and that the nonuniform behavior of way/naze discussed in the previous section stems from its external Merge position. More speciﬁcally, I propose that ‘why’ in in-situ languages (including way/naze) is an adverb that is externally merged (i.e., base-generated) in Spec,CP of the clause it modiﬁes (i.e., it is a CP modiﬁer). Crucially, the CP that ‘why’ modiﬁes may be interrogative or declarative. I call this proposal the CP-Modiﬁer Hypothesis (CMH) (see Ko 2005b for crosslinguistic variation in the external Merge position of reason wh-adverbials).7 Let us ﬁrst consider the case where way/naze is merged into an interrogative clause, as in (4)–(5). For presentation purposes, the structure of the Korean example (4a) is given here in (10). (The account for (4a) extends to Japanese (5a).)

324

Heejeong Ko

(10) Tree structure of (4a) under the CMH

As illustrated in (10), under the CMH, way is merged into Spec,CP of an interrogative clause. Notice that the uninterpretable wh-feature [uWH] of way in (10) is licensed in the overt syntax, since the interrogative clause has the licenser Q morpheme for way. This implies that way in (10) does not undergo any movement at LF to be licensed. The NPI amwuto ‘anyone’ may undergo scrambling over way in (10) (see Lee 1993 and Sohn ¯ -scramble in K/J; and 1995 for independent evidence that NPIs may A see note 11 for discussion of subject scrambling).8 Crucially, however, this scrambling of the NPI does not induce the intervention e¤ect, because way has already been licensed in the overt syntax. Hence, the grammaticality of (4a) is correctly expected. Now let us turn to the case where way/naze is merged into an (embedded) declarative clause, as in (6)–(9). The Korean examples (6a) and (8) are repeated here with their structures in (11a) and (11b), respectively. (The same account extends to Japanese (7a) and (9).) Under the CMH, way in (11a) is merged into Spec,CP of a declarative clause. Note, however, that way in (11a) cannot be licensed in its external Merge position, in contrast to way in (10). Since the declarative clause does not contain a Q morpheme, way in (11a) must undergo LF movement to a higher clause to be licensed. The NPI preceding way, however, induces the intervention e¤ect, blocking the LF movement of way. Hence, (11a) is ungrammatical. The same account extends to (11b). (In (11b), the NPI precedes way via scrambling, unlike the NPI in (11a). Both (11a) and (11b), however, are ruled out for the same reason: violation of the Intervention E¤ect Constraint.)9

Structural Height of Reason Wh-Adverbials

(11) a. Tree structure of (6a) under the CMH

b. Tree structure of (8) under the CMH

325

326

Heejeong Ko

To sum up, way/naze merged in an interrogative clause does not undergo any movement at LF, so it does not show the intervention e¤ect, unlike other wh-phrases. Way/Naze merged in a declarative clause, on the other hand, must move at LF, so it does show the intervention e¤ect, just like other wh-phrases. As discussed in Ko 2005b, this asymmetry between K/J ‘why’ with local construal and K/J ‘why’ with nonlocal construal is comparable to the nonuniform behavior of perche` ‘why’ in Italian (Rizzi 1999) and ce´n fa´th ‘what reason’ and cad chuige ‘why’ in Irish (McCloskey 2002). In the next section, I discuss predictions of the CMH and provide further evidence in favor of the CMH from child Korean data. 12.3

The Position of ‘Why’ in Child Korean

Under the CMH, ‘why’ in Korean and Japanese occupies Spec,CP in the overt syntax. This implies that if an XP may precede ‘why’, either it ¯ -movement over Spec,CP, or it must be basemust be able to undergo A generated above Spec,CP. This is described in (12). In this section, I investigate the implications of (12) with respect to child Korean.10 (12) Word order in ‘why’-questions under the CMH view a. [CP ‘why’ [ IP XP YP V]] b. [CP XPi [CP ‘why’ [ IP (t i ) YP V]]] 12.3.1 Prediction Consider the Korean sentences in (13). (13) a. Way John-i ku chayk-ul ilk-ess-ni? why John-nom that book-acc read-past-Q ‘Why did John read that book?’ b. John-i way ku chayk-ul ilk-ess-ni? Both (13a) and (13b) are grammatical and are freely uttered in adult Korean. Under the CMH, however, there is a crucial di¤erence between the two in terms of their derivational history. Speciﬁcally, the CMH predicts that (13a) reﬂects the base order between way ‘why’ and the subject John-i, whereas (13b) reﬂects the derived order via scrambling of John-i over way, as illustrated in (14).11 (14) [CP John-ii [CP way t i ku chayk-ul ilk-ess-ni]]? " scrambling

Structural Height of Reason Wh-Adverbials

327

Given that adult Korean quite freely allows scrambling, the grammaticality of (13b) is expected. Suppose, however, that some Korean speakers do not allow scrambling. The CMH then predicts that (13b) will not be uttered by this group of speakers. Of course, it is di‰cult to test this prediction with adult Korean owing to the possibility of scrambling. Child Korean data, however, do provide such an opportunity. It has been attested crosslinguistically that children rarely employ operations such as scrambling in the early stages of language acquisition (2;00–3;00) (see Slobin 1966 for Russian, Barbier 2000 and Schae¤er 2000 for Dutch; but see Otsu 1994 for perception tests with Japanese). This generalization extends to child Korean, as reported by Cho (1981) and Kim (1997).12 Given that Korean children utter only a few sentences with scrambling, the CMH makes the following prediction about child way-questions: (15) Way ‘why’ will generally precede the subject in child Korean, since the base order way < subject is preserved. (‘‘A < B’’ indicates that A precedes B.) To evaluate this prediction, I examined the naturalistic production data of a Korean child, JK, from age (2;00;06) to age (2;11;29).13 The data come from ﬁles recorded on a weekly basis by his mother—53 in all. To compare way-questions with other wh-adjunct questions, I also examined eti ‘where’ questions in the same database. Since it was not clear from the literature whether child Korean allows topicalization in this early stage, I focus here on analyzing the sentences clearly involving scrambling rather than topicalization. (In this study, I deﬁne scrambling as an operation that moves an overtly Case-marked item from its canonical position. When the Case marker for the subject is dropped, it is not clear whether the subject has undergone topicalization, scrambling, or some other processes. Thus, I counted only the subjects clearly marked with the nominative Case marker).14 12.3.2 Results Overall, JK’s ﬁles contained 111 tokens of way-questions. Unfortunately, however, most of them did not represent the subject overtly (72.97%, 81/ 111) or did not mark the subject with nominative Case (12.6%, 14/111). Given that Korean is a null subject language and that Case dropping is common even in adult Korean, this is not a surprising result. Crucially, however, when the subject was overtly marked with a nominative Case marker, I did ﬁnd a remarkably consistent pattern. This is summarized

328

Heejeong Ko

Table 12.1 Way-questions with a nominative Case-marked subject in JK’s speech Word order

Tokens

%

way SNOM (O) (V) SNOM way (O) (V) Total

12 4 16

75 25 100

in table 12.1. As the table illustrates, way precedes the nominativemarked subject at a very high rate in JK’s speech, 75% (12/16). Thus, the data in the table support the prediction of the CMH in (15). (I also note that there were extremely rare instances of object scrambling (O < way) in JK’s way-questions (2/111, 1.8% of the time)).15 Some examples of JK’s way-questions are given in (16) and (17). (16) Way Joonkyu-ka balp-ass-nunde? why Joonkyu-nom step-past-Q? ‘Why did Joonkyu step (on something)?’ (JK 2;06;07) (17) Appa-ka way an-o-ci? Daddy-nom why not-come-Q ‘Why does Daddy not come?’ (JK 2;05;16) In addition, I suggest that the remaining instances of subject < way order in table 12.1 can be attributed to JK’s limited ability to scramble. There were some sentences with noncanonical word order (non-SOV) in JK’s speech, amounting to 16% of all the observed transitive sentences (73/ 465). This indicates that JK can utter some sentences with scrambling, even though the instances are rather rare. Since the number of subject < way utterances is comparable to the overall scrambling rate in JK’s speech, I conclude that the prediction made by the CMH holds consistently in JK’s data.16 12.3.3 Further Comparison: Eti ‘Where’ versus Way ‘Why’ in Child Korean One might think that the pattern found with way in table 12.1 could be a general pattern found with any wh-adjunct in child Korean. To test this alternative hypothesis, I examined eti ‘where’ questions in the same database. Strikingly, the data show a sharp contrast between way ‘why’ and eti ‘where’ questions.

Structural Height of Reason Wh-Adverbials

329

Table 12.2 Eti-questions with a nominative Case-marked subject in JK’s speech Word order

Tokens

%

eti SNOM (O) (V) SNOM eti (O) (V) Total

0 14 14

0 100 100

Table 12.3 Comparison between way and eti in JK’s speech Child wh-word

wh S NOM

S NOM wh

way ‘why’ eti ‘where’

12 (75%) 0 (0%)

4 (25%) 14 (100%)

w 2 ¼ 17:5 , p < :001; Fisher exact, p < :0001

Overall, the eti-questions exhibited a large number of subject omissions (45.2%, 66/146) and nominative Case marker omissions (45.2%, 66/146). In clear contrast to way, however, eti never preceded the nominativemarked subject. This is illustrated in table 12.2. (There were extremely rare instances of object scrambling in JK’s eti-questions (3.42%, 5/146).) An example of JK’s eti-questions is given in (18). (18) Kunde, appa-ka eti ka-ass-e?17 but Daddy-nom where go-past-Q ‘By the way, where is Daddy?’ (JK 2;02;29) Statistically, the distributional di¤erences between way and eti are highly signiﬁcant, as shown in table 12.3.18 This contrast indicates that the word order pattern with way in table 12.1 is not accidental, but systematic. As the CMH predicts, way is merged higher than the subject, and so precedes the subject in nonscrambled child speech. The absence of eti < subject order, on the other hand, suggests that eti is merged lower than the subject, in contrast to way. 12.3.4 Adult versus Child Korean Way-Questions Finally, one might wonder whether the pattern found with way in table 12.1 results from mere imitation of adult input. However, as table 12.4 indicates, the child data show the opposite pattern from the adult input with respect to way-questions. In contrast to the order found in JK’s

330

Heejeong Ko

Table 12.4 Adult way-questions in JK’s ﬁles Word order

Tokens

%

way SNOM (O) (V) SNOM way (O) (V) Total

32 63 95

34 66 100

Table 12.5 Child versus adult way-questions in JK’s ﬁles

Child Adult

way SNOM

SNOM way

12 32

4 63

w 2 ¼ 9:769 , p < :01; Fisher exact, p < :002

own speech, shown in table 12.1, the subject < way order was the dominant order (66%) for the adults communicating with JK. This contrast argues strongly that the pattern reported in table 12.1 did not originate from the inﬂuence of the adult input. In fact, the child reverses the predominant subject < way order in the adult input into the way < subject order in his own speech. This point is illustrated clearly in the following dialogue between JK and his mother. JK hears his mother’s way-question, which contains the daddy-NOM < way order. In repeating the question, however, JK switches the word order to way < daddy-NOM. (19) Mother: Appa-ka way dwul-i-ni? daddy-nom why two-be-Q ‘Why are daddy two?’ (‘Why do you have two daddies?’) JK: Way appa-ka dwul-i-ni? why daddy-nom two-be-Q ‘Why are daddy two?’ (‘Why do you have two daddies?’) (JK 2;06;07) The di¤erences between the adult and child data in the placement of way reached statistical signiﬁcance as well, as shown in table 12.5.19 As for eti-questions, there was no signiﬁcant di¤erence between child and adult. As illustrated in table 12.6, eti in adult speech follows the nominative-marked subject in most cases (95%), as in child speech (table

Structural Height of Reason Wh-Adverbials

331

Table 12.6 Adult eti-questions in JK’s ﬁles Word order

Tokens

%

eti SNOM (O) (V) SNOM eti (O) (V) Total

5 91 96

5 95 100

12.2). The di¤erence between the child and the adult in placement of eti was not statistically signiﬁcant (w 2 ¼ 0:76, p < :5; Fisher exact, p < :499). The comparison between the child and the adult eti-questions suggests two possibilities. Either the child simply reproduces the adult pattern in uttering eti-questions, or eti is base-generated lower than the subject and remains below the subject in adult speech. I suggest that the latter is the plausible interpretation given the contrast between the child and adult way-questions.20 In further research, I hope to investigate a large corpus to determine whether wh-adjuncts generally resist scrambling in adult speech.21 12.3.5 Implications of the Acquisition Data In this section, I have demonstrated the distributional di¤erences between way ‘why’ and eti ‘where’ in child Korean: namely, that way ‘why’ precedes the subject most of the time, whereas eti ‘where’ never precedes the subject. I argued that the distributional di¤erences between the two whadjuncts are not accidental, but predicted by the CMH. I argued further that this word order pattern cannot be traced to adult input. Before closing the section, I would like to address the theoretical import of the child Korean data. First, the discussion in this section provides another piece of empirical evidence for the thesis that child data are needed to evaluate predictions that cannot be tested in adult data. In adult Korean, the base order of the sentence may always be obscured by scrambling. Investigation of child data, in contrast, provided a clear test for verifying the base word order in way-questions. Second, the discussion in this section contributes to the observation that acquisition of ‘why’ shows special patterns in many languages. A series of works on child English, in particular, has shown that why-questions do not require subject-aux inversion, in contrast to other wh-questions (Stromswold 1990; Berk 2003; see also Thornton 2004 regarding the asymmetry between why in local versus nonlocal construal in child English). This

332

Heejeong Ko

observation has been attributed to merger of ‘why’ at the top of the clause (cf. Rizzi 1990, 1999). The data presented here show that the same type of argument extends to child Korean. Way ‘why’ in Korean is directly merged in Spec,CP, and this results in the ‘‘noninversion’’ of the base order way < subject in child Korean. 12.4

The ‘‘Fine Structure of CP’’ in In-Situ Languages

So far, I have been agnostic about the ‘‘ﬁne structure of CP’’ (Rizzi 1997, 1999) in in-situ languages. In this section, I investigate a possible consequence of the CMH in a split-CP system. Issues regarding multiple whquestions in K/J/C are of special concern. 12.4.1 Split-CP System in Wh-Interrogatives Rizzi (1997, 1999) has argued that the complementizer system is actually a structural zone containing several distinct functional heads and their projections, much like the IP system and the DP system under many proposals. On the basis of examining the position of focus (Foc), topic (Top), se ‘if ’ (Int), che ‘that’ (Force), wh-phrases (Foc), and perche` ‘why’ (Int) in Italian, Rizzi proposes the following CP structure: (20) Force (Top) Int (Top) Foc (Top) Fin IP . . . (Rizzi 1999, 5) In this section, I do not aim to provide a full picture of the split-CP system in in-situ languages. Instead, I focus on the interactions between the two positions that may be occupied by wh-phrases in (20): Int and Foc. Rizzi (1999) argues that Int is a base-generation position for perche` ‘why’ in Italian, whereas Foc is a landing site for wh-movement (including perche` in nonlocal construal). I argue that this idea of positing an interrogative head for ‘why’ distinct from the landing site of wh-movement has the potential to resolve certain puzzles that arise in in-situ multiple wh-questions. Speciﬁcally, I suggest that in-situ languages have two types of C heads for licensing wh-phrases, and that Int (the merger site for ‘why’) is structurally lower than Foc (the landing site for wh-movement) (cf. (20)). This is illustrated in (21). To avoid unnecessary confusion with Rizzi’s (1999) proposal, I refer to the C head hosting ‘why’ as CInt and the C head hosting wh-movement as CFoc .22 In the following subsection, I present evidence for this structure. (21) [C Foc . . . C Int [ IP . . . ]]

Structural Height of Reason Wh-Adverbials

333

12.4.2 Evidence for the Split-CP System in In-Situ Languages The ﬁrst argument in support of the structure (21) comes from the interaction between multiple wh-questions and the intervention e¤ect ((3); see section 12.2). Consider the Korean examples in (22). (According to my informants, the same contrast holds in Japanese.)23 (22) a. *?Amwuto mwues-ul way mek-ci-anh-ass-ni? anyone what-acc why eat-ci-not-past-Q ‘Why did no one eat what?’ b. Mwues-ul amwuto way mek-ci-anh-ass-ni? what-acc anyone why eat-ci-not-past-Q ‘Why did no one eat what?’ c. Mwues-ul way amwuto mek-ci-anh-ass-ni? what-acc why anyone eat-ci-not-past-Q ‘Why did no one eat what?’

(Korean)

Under the single-C system, we might predict that (22a) would be grammatical because mwues ‘what’ may scramble to the Spec,CP domain (crossing way) in the overt syntax and thus can obviate the intervention e¤ect via speciﬁer-head agreement with C, as schematized in (23). (23) [CP amwutoi mwues-ulj way C [ IP t i t j . . . ]] Without further assumptions about the internal structure of CP or the path of multiple scrambling to Spec,CP, the ungrammaticality of (22a) is di‰cult to explain.24 However, if we assume that (21) represents the internal structure of CP in K/J, (22a) is straightforwardly ruled out. This is illustrated in (24). amwutoi mwues-ulj way C Int [ IP t i t j . . . ]]

a

(24) *[CP CFoc

Under the structure in (21) (shown in (24)), mwues ‘what’ in (22a) cannot be licensed, because amwuto ‘anyone’ blocks LF movement of mwues to the licensing head C Foc . Hence, the ungrammaticality of (22a) follows. Another argument for the structure in (21) comes from the phenomenon called the antisuperiority e¤ect (ASE; e.g., Saito 1982, 1994; Watanabe 1992; Watanabe 1995; Chung 1996). Descriptively speaking, the ASE is a word order restriction between way/naze ‘why’ and other wh-phrases in K/J such that way/naze ‘why’ cannot precede the other wh-phrases, retaining a multiple wh-question reading. This is illustrated in (25) and (26).25 (25) a. *John-i way mwues-ul sa-ss-ni?26 John-nom why what-acc buy-past-Q ‘Why did John buy what?’

(Korean)

334

Heejeong Ko

b.

John-i mwues-ul way sa-ss-ni? John-nom what-acc why buy-past-Q ‘Why did John buy what?’

(26) a. *John-wa naze nani-o katta John-top why what-acc bought ‘Why did John buy what?’ b. John-wa nani-o naze katta John-top what-acc why bought ‘Why did John buy what?’

no? Q

(Japanese)

no? Q

I argue that the ASE can be naturally captured under the structure in (21) if we adopt a particular hypothesis about the semantic property of ‘why’. I suggest that way/naze ‘why’ in K/J is itself an SBE that induces the intervention e¤ect. In other words, the ASE is just a variant of the intervention e¤ect (cf. Cho 1998 and Richards 2001, 287–299, for an approach unifying the intervention e¤ect and the ASE), and the ungrammaticality of (25a) and (26a) is ascribed to the intervention e¤ect, as shown in (27). (27) a. *[C Foc ‘why’ C Int ‘what’]: (25a), (26a) b. [C Foc ‘what’i ‘why’ C Int t i ]: (25b), (26b) a

The analysis in (27) assimilates the ASE to an already existing phenomenon, the intervention e¤ect. Hence, we do not have to resort to additional constraints invented only for the ASE (cf. Watanabe 1992). Furthermore, there is evidence that ‘why’ does share some semantic properties with SBEs triggering the intervention e¤ect. As Kim (2002) argues, it is focus-related items that induce the intervention e¤ect. Interestingly, why-questions typically induce focus association. As Bromberger (1992) points out, the answer to a why-question in English crucially di¤ers depending on the association between why and a focus with emphatic stress. This is shown in the question-answer pairs in (28). (28) a. Why did Adam eat the apple? —Because God intended that to happen b. Why did ADAM eat the apple? —Because he was the one that Eve worked on c. Why did Adam EAT the apple? —Because he couldn’t think of anything else to do with it d. Why did Adam eat the APPLE? —Because it (the apple) was the only food around (Bromberger 1992, 160–161)

Structural Height of Reason Wh-Adverbials

335

Note that this focus association disappears in other wh-questions. For instance, compare the why-questions in (28) with the when-questions in (29). (29) a. When did Adam eat the apple? —At 4 PM on July 7, 24,000 BC b. When did ADAM eat the apple? —At 4 PM on July 7, 24,000 BC c. When did Adam EAT the apple? —At 4 PM on July 7, 24,000 BC d. When did Adam eat the APPLE? —At 4 PM on July 7, 24,000 BC Unlike what we ﬁnd with why-questions, the placement of emphatic stress on di¤erent words in (29) does not impose di¤erent conditions on what counts as an answer in when-questions.27 The same fact holds in Korean, as shown in (30) and (31). (The same paradigm is attested in Japanese.) The answer for the way-questions in (30) di¤ers depending on the focus, as in English (28), whereas the answer for encey ‘when’ questions in (31) (and all other types of wh-questions) remains the same regardless of the focus, as in English (29). (30) a. Way Adam-i sakwa-lul mek-ess-ni? (Korean) why Adam-nom apple-acc eat-past-Q ‘Why did Adam eat the apple?’ —Hanunim-i kukes-ul uytohasiessu-nikka God-nom that-acc intended-because ‘Because God intended it (to happen).’ b. Way ADAM-i sakwa-lul mek-ess-ni? —Ivu-ka Adam-eykey kwonhayssu-nikka Eve-nom Adam-dat recommended-because ‘Because he was the one that Eve recommended (to eat the apple).’ c. Way Adam-i SAKWA-lul mek-ess-ni? —Kukes-i cwupyeney iss-ten yuilhan umsik iessu-nikka that-nom around be-Rel only food was-because ‘Because it (the apple) was the only food around.’ d. Way Adam-i sakwa-lul MEK-ess-ni? —Kukes-ulo mwues-ul halci mollassu-nikka that-with what-acc to.do was.ignorant.of-because ‘Because he couldn’t think of anything else to do with it.’

336

Heejeong Ko

(31) a. Encey Adam-i sakwa-lul mek-ess-ni? when Adam-nom apple-acc eat-past-Q ‘When did Adam eat the apple?’ —Chil wol chil il ohwu ney si 7 month 7 day afternoon 4 hour ‘July 7, 4 pm.’ b. Encey ADAM-i sakwa-lul mek-ess-ni? —Chil wol chil il ohwu ney si ‘July 7, 4 pm.’ c. Encey Adam-i SAKWA-lul mek-ess-ni? —Chil wol chil il ohwu ney si ‘July 7, 4 pm.’ d. Encey Adam-i sakwa-lul MEK-ess-ni? —Chil wol chil il ohwu ney si ‘July 7, 4 pm.’

(Korean)

Given that ‘why’ in Korean (and Japanese) invokes focus associations, like many other SBEs inducing the intervention e¤ect, it seems plausible to assume that the ASE is a variant of the intervention e¤ect.28 The contrast between Chinese and K/J with respect to the ASE provides further support for the analysis in (27). Compare the Chinese sentences in (32) with the K/J sentences in (25) and (26).29 (32) a.

Zhangsan weishenme bu chi shenme (dongxi)? Zhangsan why not eat what thing ‘Why did Zhangsan not eat what?’ b. *?Zhangsan shenme dongxi weishenme bu chi? Zhangsan what thing why not eat ‘Why did Zhangsan not eat what?’ c. Zhangsan na wan fan weishenme bu chi? Zhangsan the bowl rice why not eat ‘Why did Zhangsan not eat the bowl of rice?’ (Hooi Ling Soh and Feng-Fan Hsieh, pers. comm.)

Interestingly, (32) illustrates that the ASE does not hold in Chinese. In fact, the judgments about Chinese weishenme ‘why’ questions go in the opposite direction from judgments about their K/J counterparts: weishenme may precede shenme ‘what’ in (32a), whereas shenme cannot precede weishenme in (32b). The same point can be made with embedded weishenme-questions in (33).

Structural Height of Reason Wh-Adverbials

337

(33) Ni xiang-zhidao [Lisi weishenme mai-le shenme]? you wonder [Lisi why buy-asp what] ‘What is the thing x such that you wonder why Lisi bought?’ (Huang 1982, 526; see also Lasnik and Saito 1984, 243) I argue that the CMH and the analysis in (27) provide an account of this asymmetry between K/J and Chinese. Suppose that the CMH holds in Chinese as in K/J (see Ko 2005b for various arguments in favor of this assumption; see also Lin 1992 for Chinese), and that the ASE is a variant of the intervention e¤ect. Under these assumptions, the contrast between K/J (25a)/(26a) and Chinese (32a) is reduced to the independent fact that Chinese wh-nominals (wh-phrases except weishenme and zenme ‘how’ with a manner reading) do not show the intervention e¤ect (Cheng and Rooryck 2002; Gue´rin and Soh 2003; Soh 2005). As illustrated in (34), Chinese wh-nominals may be preceded by an SBE, which may induce the intervention e¤ect in K/J (for other examples in Chinese, see Cheng and Rooryck 2002; Gue´rin and Soh 2003; Soh 2005; see also Ko 2005b for the interactions between weishenme and the intervention e¤ect). (34) {Meiyouren/Henshaoren/Zuiduo liang-ge ren} gan gen shei dajia? nobody/few person/at most two-cl person dare with who ﬁght ‘Who is the person x such that {nobody/few people/at most two people} dare(s) to ﬁght with x?’ (Soh 2005) Given that Chinese wh-nominals are insensitive to the intervention e¤ect, it is expected from (27) that Chinese lacks the ASE (a variant of the intervention e¤ect) in (32). Furthermore, it is also correctly predicted that K/J wh-nominals may show the ASE in (25) and (26), as a variant of the intervention e¤ect, since they are sensitive to the intervention e¤ect (section 12.2). One remaining question is why Chinese (32b) is degraded, in contrast to K/J (25b) and (26b). I argue that this contrast is, in fact, naturally captured by the CMH. If ‘why’ in K/J/C is merged in Spec,CP in the overt syntax, wh-arguments may precede ‘why’ in K/J/C only when it ¯ -movement (e.g., A ¯ -scrambling or A ¯ -topicalization) over may undergo A a C boundary (see (12)). It is well known that K/J wh-phrases may ¯ -movement (via A ¯ -scrambling), as illustrated in (35) (see Takaundergo A hashi 1993 for Japanese examples). Thus, it is not surprising that mwues/ nani ‘what’ may scramble over way/naze ‘why’ in (25b) and (26b).30

338

Heejeong Ko

(35) a. John-un [Mary-ka mwues-ul mek-ess-ta-ko] (Korean) John-top Mary-nom what-acc eat-past-dec-C malhay-ss-ni? say-past-Q ‘What did John say that Mary ate?’ b. Mwues-uli John-un [Mary-ka t i mek-ess-ta-ko] malhay-ss-ni? what-acc John-top Mary-nom eat-past-dec-C say-past-Q ‘What did John say that Mary ate?’ ¯Now let us turn to Chinese. In contrast to K/J, Chinese does not allow A scrambling (cf. Soh 1998 regarding short scrambling in Chinese). Thus, in order for shenme ‘what’ to precede weishenme ‘why’ in (32b), it must be ¯ -topicalization. However, as illustrated in (36), whable to undergo A ¯ -topicalization in Chinese. Hence, the CMH predicts that phrases resist A shenme cannot precede weishenme in (32b).31 (36) a.

Zhangsan suo Mali chi le shenme (dongxi)? Zhangsan said Mary eat asp what thing ‘What did Zhangsan say that Mary ate ?’ b. *?Shenme (dongxi) i [Zhangsan suo [Mali chi le t i ]]? what thing Zhangsan said Mary eat asp ‘What did Zhangsan say that Mary ate ?’ c. Pingguoi [Zhangsan suo [Mali chi le t i ]] apples Zhangsan said Mary eat asp ‘Apples, Zhangsan said that Mary ate .’

To wrap up, in this section I explored the implications of the CMH concerning the ﬁne structure of CP in in-situ languages. I suggested that there may be two distinct C heads for licensing wh-interrogatives in insitu languages (cf. Rizzi 1999). By assuming that ‘why’ is merged in the lower head C Int and that the other wh-phrases covertly move to the higher head C Foc , we can account for the emergence of the intervention e¤ect within the CP ﬁeld. I also argued that the ASE is a variant of the intervention e¤ect and that this proposal accounts for otherwise unexpected contrasts between K/J and Chinese with respect to the ASE. 12.5

Conclusion

In this chapter, I argued that the base position of ‘why’ in in-situ languages sheds new light on its unique properties in syntax. I proposed that ‘why’ in Korean, Japanese, and Chinese is directly merged into

Structural Height of Reason Wh-Adverbials

339

Spec,CP as a CP modiﬁer (CMH) and that this proposal accounts for various puzzles concerning ‘why’ in in-situ languages. In particular, the local versus nonlocal asymmetry of K/J ‘why’ with respect to the intervention e¤ect is explained under this proposal. Predictions concerning the left periphery of ‘why’ were conﬁrmed with data from K/J/C. Korean child acquisition data provided further support for the CMH, showing that ‘why’ is externally merged higher than the subject. The acquisition data are theoretically important, in that predictions concerning the base order in way-questions cannot be easily tested with adult Korean. I also pointed out that child Korean data contribute to the crosslinguistic observation that acquisition of ‘why’ shows special patterns because of its high base-position. Investigating the ﬁne structure of CP in in-situ languages, I proposed that the ASE is a variant of the intervention e¤ect. This proposal not only accounts for the ASE in K/J, but also explains the asymmetry between K/J and Chinese concerning the ASE. Notes This chapter is a developed version of the paper presented at the workshop ‘‘On Wh-Movement.’’ I am grateful to the participants at the workshop and two reviewers for helpful comments. I am deeply thankful to David Pesetsky and Danny Fox for insightful discussions on the earlier versions of the chapter. I am also indebted to Rajesh Bhatt, Sylvain Bromberger, Stephen Crain, Paul Hagstrom, James Huang, Sabine Iatridou, Shigeru Miyagawa, Norvin Richards, Luigi Rizzi, Mamoru Saito, William Snyder, Akira Watanabe, and Ken Wexler for comments and suggestions. I thank Ressy Ai, Feng-Fan Hsieh, Sachiko Kato, Youngjoo Lee, Zhiqiang Li, Hideki Maki, Hooi Ling Soh, and Shoichi Takahashi for grammaticality judgments and discussion. Of course, all errors in the chapter are my own. 1. This section has been developed from Ko 2005b. It focuses on the facts in Korean and Japanese, but see Ko 2005b for detailed discussion concerning Chinese weishenme ‘why’ (see also section 12.4). 2. The motivation for scrambling in K/J has been controversial. Some researchers argue that discourse factors like topic/focus (most likely optionally) derive scrambling (Choi 1996), whereas others argue that morphological features like Case or EPP (obligatorily) trigger scrambling (Lee 1993; Miyagawa 1997a). This issue is immaterial here, however, since this chapter is concerned with the distribution of the scrambled elements, regardless of the trigger for scrambling. As for the landing site of a scrambled phrase, see note 8. 3. Two remarks are in order regarding (3). First, Beck and Kim (1997) implemented (3) within the semantic frameworks of Hamblin (1973) and Karttunen (1977): a wh-phrase must move to Spec,CP for semantic reasons, and an SBE blocks the binding relationship between the LF trace of the wh-phrase and its binder. As far as this chapter is concerned, however, the

340

Heejeong Ko

empirical consequences of Beck and Kim’s proposal and (3) are the same. The term intervention e¤ect was ﬁrst introduced by Hagstrom (1998), who implemented (3) with di¤erent assumptions about wh-movement from the one adopted here. For alternative approaches to (3), see the references cited in the text. Second, in Chomsky 1995 a bundle of formal features is attracted to its licensing head at LF. Thus, (3) needs to be understood as an LF constraint on wh-feature movement rather than wh-phrasal movement (see Pesetsky 2000 for implications). For presentation purposes, however, I abstract away from the distinction between feature and phrasal movement. For concreteness, I assume Chomsky’s (1995, chap. 4) framework (copy theory of movement, a single Ymodel, etc.), but the chapter can be made compatible with Chomsky’s other recent approaches (2000, 2001, 2004) (see Ko 2005b for discussion). For convenience, a copy of a moved element is represented with a trace bearing an index, but this has no theoretical import. The notation for features (e.g., [uWH]) is borrowed from Pesetsky and Torrego 2001. 4. It is only way/naze ‘why’ that shows the pattern in (4)–(5). Other wh-adjuncts in K/J such as eti/doko ‘where’, encey/itsu ‘when’, and ettehkey/donoyounisite ‘how’ cannot be preceded by an NPI, just like wh-arguments (see also Beck and Kim 1997 for relevant Korean examples). 5. Negation alone (without an accompanying NPI) may induce the intervention e¤ect when way/naze is embedded in a declarative clause. That is, a question like (i) is ungrammatical in Korean. (The same is true of Japanese.) I thank Daeho Chung for clarifying this. (i) *Mary-nun [John-i way saimha-yess-ta-ko] (Korean) Mary-top [John-nom why resign-past-dec-C] malha-ci-anh-ass-ni? say-ci-not-past-Q ‘What is not the reason x such that Mary said that John resigned for x?’ 6. Miyagawa (1999) argues that naze is exempt from the intervention e¤ect only when the NPI and naze are clause-mates. Example (9) argues against this proposal. For some speakers, (8) marginally allows a matrix reading of way, with the interpretation ‘Whyi did John say t i that no one read the book?’. This reading is irrelevant for the present discussion because the intervention e¤ect is at stake in (8) when way modiﬁes the embedded clause and is interpreted across the NPI. 7. As I will show in section 12.4, this proposal extends to weishenme ‘why’ in Chinese (see also Lin 1992 and Ko 2005b for other evidence in Chinese). Similar approaches have been proposed for reason wh-adverbials in some wh-fronting languages: pourquoi ‘why’ in French (Rizzi 1990; Bosˇkovic´ 2000), perche` ‘why’ and come mai ‘how come’ in Italian (Rizzi 1999), cad chuige ‘why’ and ce´n fa´th ‘what reason’ in Irish (McCloskey 2002), and why (Bromberger 1992; Hegarty 1992) and how come (Collins 1991) in English. See also Tsai 1999 and Tsai and Chang, to appear, for arguments that in Tsou, causal mainci ‘why’ and causal mainenu ‘how’ are merged higher than IP as sentential operators. 8. Following the common assumption in the scrambling literature (e.g., Mahajan 1990; Saito 1992), I assume that a scrambled phrase adjoins to the closest maxi-

Structural Height of Reason Wh-Adverbials

341

mal projection that it may land in. For instance, in (10) the subject amwuto adjoins to the CP where way is merged, and projects a segment of CP. The possibility that there might be a speciﬁc position designated for a scrambled phrase in the Spec,CP domain leaves my arguments intact, however. Following Mahajan (1990), when scrambling occurs across a C boundary, I will call it A¯-scrambling (whether it is clause-internal or clause-external). Since amwuto in (10) scrambles ¯ -scrambling (see Ko 2005b and over a C boundary, this scrambling counts as A ¯ -scrambling). I thank Lisa Cheng and Chris references therein for clause-internal A Reintges for clarifying this. 9. This analysis assumes that LF reconstruction of an NPI does not bleed the application of the Intervention E¤ect Constraint at LF. For relevant discussion, see Beck and Kim 1997 and Ko 2005b. ¯ -movement refers to the movement crossing a C boundary (Mahajan 10. A 1990; note 8). Since movement over ‘why’ in (12) crosses Spec,CP, it counts as ¯ -movement. This movement may be instantiated by A ¯ -scrambling (in K/J) or A ¯ by A-topicalization (in K/J/C; see section 12.4 for further discussion). See Ko 2005b for other evidence supporting the prediction in (12). In Ko 2005b, I argue that this prediction accounts for the crosslinguistic variation between K/J and Chinese ‘why’-questions. K/J SBEs (e.g., -man/-sika NP ‘only NP’, -to/-mo NP ‘also NP’, -kkaci/-sae NP ‘even NP’, nukwunka/dareka ¯ -movement (via A ¯‘someone’, nwukwuna/daremo ‘everyone’) may undergo A scrambling) and so can precede ‘why’ in K/J. In contrast, Chinese SBEs (e.g., zhiyou NP ‘only NP’, meiyouren ‘nobody’, henshaoren ‘few people’) may not ¯ -movement (topicalization) and so cannot precede weishenme ‘why’ in undergo A Chinese, unlike its K/J counterparts. Furthermore, in Korean, epistemic adver¯ -movement like Chinese SBEs, bials (e.g., amato ‘probably’) cannot undergo A and they cannot precede way ‘why’ in Korean (see Ko 2005b for the relevant examples). ¯ -scrambling in K/J has been contro11. Whether a subject may undergo A- or A versial. Some researchers argue that subject scrambling exists (Kurata 1991; Lee 1993; Sohn 1995) while others argue the opposite (Hoji 1985; Saito 1985). The CMH has an important implication for this debate. Since a subject may precede way/naze, as in (14), the CMH crucially entails that a subject is able to undergo ¯ -scrambling in (14)). Thus, to the extent that the scrambling (more precisely, A CMH is successful, we have direct support for the existence of subject scrambling. (See also Ko 2005a for extensive disscussion of syntactic restrictions on subject scrambling.) 12. Kim (1997) shows that the rate of the noncanonical word order (non-SOV) in the speech of ﬁve children varies from 6% to 27%. Cho (1981), investigating the word order patterns produced by three Korean children, reports that in nonelliptical transitive sentences two children employed non-SOV order 5.7% and 17.4% of the time, respectively, and that one child used the noncanonical order 56.8% of the time (cited in Kim 1997, 351). This exceptional child seems to be an outlier or a bilingual English speaker, given that her rate of SVO order was extremely high at 24.3%, in contrast to the other children whose speech is examined in Cho 1981 and Kim 1997, whose SVO rate ranged from 4.5% to 0%.

342

Heejeong Ko

13. I am grateful to Seungbok Lee for allowing me to study JK’s data for this chapter. 14. Tokens were counted as one utterance. Each utterance of self-repetition was counted as one token. Total imitation was not counted as a token. Partial or transformed imitation was counted as one token of voluntary child utterance, given that partial imitation often changes the word order of the preceding adult utterances (see (19); see also Stromswold 1996 for methods of analyzing children’s spontaneous speech). In counting way- and eti-questions, I counted only sentences containing more than one word. 15. There was one token of way-doubling in a sentence containing a nominativemarked subject. As shown in (i), way ‘why’ appeared twice in the sentence: once before the subject, and again after the subject. This was not counted in table 12.1 because it was not clear which of the two instances of way should be considered. (i) Way kangaci-ka way an-mek-ni? why puppy-nom why not-eat-Q ‘Why does not the puppy eat (it)?’ (JK 2;09;09) 16. It is worth noting that the overall rate of scrambling in transitive sentences in the adult input (i.e., non-SOV sentences) was 11.4% (95/832). This indicates that when adults spoke to JK, they preferred to use the canonical word order in transitive sentences (but see also note 19). 17. As Chris Reintges (pers. comm.) notes, eti in (18) is a complement selected by a directional verb, ka ‘go’ (and may be viewed as a direct object; see Levin and Rappaport Hovav 1985). It will be interesting to see whether a true VP-adjunct like ettehkey ‘how’ (with manner reading) shows a distribution di¤erent from that of eti. I leave this for further research. 18. I thank Tania Ionin, Ken Wexler, and William Snyder for helping me to conduct the statistical analysis. 19. This leaves the question of why the scrambling rate in way-questions (subject < way) in the adult input amounts to 66% while the noncanonical order (i.e., non-SOV) in transitive sentences is limited to 11.4% (note 16). I leave this issue for further research. 20. In Hindi, when scrambling of a wh-phrase places the wh-phrase ﬁrst in the sentence, the scrambling is generally blocked for stylistic reasons (Rajesh Bhatt, pers. comm.). I leave it open whether this generalization may explain the tendency that subject < eti order is preferred in Korean. 21. As Chris Reintges (pers. comm.) points out, it will be interesting to see whether scrambled order in child Korean is acquired gradually or appears suddenly. In addition, comparison of JK’s data with data from other corpora and comparison between the linear ordering of way and that of wh-arguments would be helpful in evaluating the results reported here. I thank William Snyder and Bonnie Schwartz for raising this point. 22. Nothing in sections 12.2 and 12.3 is crucially a¤ected by this modiﬁcation. In other words, under the split-CP system proposed in (21), ‘why’ is externally

Structural Height of Reason Wh-Adverbials

343

merged in Spec,C Int , and an element may scramble to the left of ‘why’ and adjoin to Spec,C Int . For present purposes, I leave it open whether the CP ﬁeld between C Int and C Foc can be occupied by a topic phrase in in-situ languages, as Rizzi (1997, 1999) argues for Romance languages. 23. I thank Norvin Richards for bringing this example to my attention. 24. In Ko 2005b, I suggest that if we adopt Yatsushiro’s (1996) analysis of multiple scrambling, the ungrammaticality of (22a) can be explained under the single-C system. Noting the scope rigidity between a scrambled indirect object and a scrambled direct object, Yatsushiro argues that when multiple scrambling occurs, the maximal projection dominating the scrambled items moves to the target (but see also Sauerland and Elbourne 2002 for a critique of Yatsushiro 1996). If we extend this analysis to multiple scrambling over ‘why’, (22a) is ruled out, as shown in (i). Given that ‘what’ is c-commanded by the NPI in (i), ‘what’ cannot be licensed at LF because of the intervention e¤ect. (i)

25. In contrast to way/naze ‘why’, all the other wh-phrases in K/J including other wh-adjuncts (e.g., eti/doko ‘where’, encey/itsu ‘when’, ettehkey/donoyounisite ‘how’) may precede or follow other wh-phrases irrespective of their structural height. The ASE is limited only to the interactions between way/naze ‘why’ and the other wh-phrases. 26. This sentence is ﬁne with the indeﬁnite reading of ‘what’: ‘Why did John buy something?’. 27. As David Pesetsky (pers. comm.) notes, the di¤erence between why and when shown in (28) and (29) can be weakened in some contexts. For instance, suppose that there are several dishes on the breakfast menu (apples, bagels, eggs, etc.) that Adam or Eve might have eaten. If the speaker is interested in the overall time when Adam or Eve started to eat the breakfast, the answer for the questions in (29) would be the same. But if the speaker is interested in the exact time when Adam or Eve started to eat a speciﬁc food, the answers for the questions in (29) could di¤er. It might be the case that the focus association for why is more contextually salient and easily invoked than the one for when, just for pragmatic reasons. If this suggestion is correct, the argument that ‘why’ is the only wh-operator inducing focus associations would also be challenged. 28. It is known that the ASE is ameliorated by an additional wh-phrase ccommanding ‘why’, as in (i) (the ‘‘additional-wh e¤ect’’ discussed in Saito 1994; see also Chung 1996).

344

Heejeong Ko

(i) Dare-ga naze nani-o katta-no? who-nom why what-acc buy-Q ‘Who bought what why?’

(Japanese)

It is not clear whether this additional-wh e¤ect holds for the intervention e¤ect generally. While Tanaka (1998) and Richards (2001) argue that an additional wh-phrase improves the ungrammaticality of the intervention e¤ect in Japanese, Beck and Kim (1997) report the opposite judgments in Korean. I leave it for further research to determine the nature of this di¤erence. 29. Negation was added in (32) in order to elicit a pragmatically more plausible answer for the weishenme-question. Negation, however, is not the factor that contributes to the di¤erences between K/J and Chinese. Judgments about the sentences in (32) remain the same even though negation is omitted there. See also Huang’s example in (33) for the same point. Discussion about weishenme in this section is limited to its reason/causal reading. It remains to be seen whether the arguments about causal weishenme in this section extend to wei(-le)shenme with a purpose reading. (See also Tsai 1994, 1999 for some di¤erences between causal and purpose weishenme.) 30. Crosslinguistically, a wh-phrase generally resists topicalization. As shown in (i), K/J wh-phrases cannot host a topic marker (implying that they cannot undergo topicalization according to Lee and Tomioka 2000). It is important to note that a wh-phrase may precede way/naze in K/J (25b) and (26b) because K/J allow scrambling of wh-phrases, not because K/J allow topicalization of a whphrase. (i) *Nwukwu-nun Mary-lul manna-ss-ni? who-top Mary-acc meet-past-Q ‘Who met Mary?’

(Korean)

31. Lisa Cheng (pers. comm.) suggests that (32b) might be degraded because the indeﬁnite shenme cannot undergo object shift (Soh 1998), unlike the deﬁnite NP na wan fan in (32c). Since both topics and elements undergoing object shift in Chinese require deﬁniteness, it is di‰cult to tease the two analyses apart as far as (32b) is concerned. Note, however, that this alternative analysis does not explain the crosslinguistic variation in the ASE between K/J (25a)/(26a) and Chi¯nese (32a). In contrast, the current analysis explains the correlation between A movement and the ASE in K/J/C. A cautionary note is in order, however. Huang (1982) observes that shei ‘who’ may precede weishenme ‘why’ in example (i). This might constitute a counterexample to the current generalization that wh-phrases cannot undergo topicalization in Chinese. (i) [Ni xiang-zhidao [shei weishenme mai-le shu]]? you wonder who why buy-asp book ‘Who is the person x such that you wonder why x bought books?’ (Huang 1982, 525)

(Chinese)

I speculate that there are two reasons why (i) can be grammatical. First, according to my Chinese informant (Feng-Fan Hsieh, pers. comm.), weishenme in (i) does not have the causal reading; (i) has only the purpose reading. It might be the

Structural Height of Reason Wh-Adverbials

345

case that weishenme with a purpose reading may be merged lower than Spec,CP (see Tsai 1999 for arguments that purpose weishenme is merged in VP-adjoined position). Second, there might be some elements that may precede weishenme but cannot undergo clause-external topicalization. Example (i) is compatible with the CMH if we assume that some speakers allow shei ‘who’ to undergo clause-internal ¯ -topicalization, but not clause-external A ¯ -topicalization. I leave this issue open. A References Barbier, Isabella. 2000. An experimental study of scrambling and object shift in the acquisition of Dutch. In M. Susan Powers and Cornelia Hamann, eds., The acquisition of scrambling and cliticization, 41–69. Dordrecht: Kluwer. Beck, Sigrid, and Shin-Sook Kim. 1997. On wh- and operator scope in Korean. Journal of East Asian Linguistics 6, 339–384. Berk, Stephanie. 2003. Why ‘‘why’’ is di¤erent. In Barbara Beachley, Amanda Brown, and Frances Conlin, eds., BUCLD 27: Proceedings of the 27th annual Boston University Conference on Language Development, 127–137. Somerville, Mass.: Cascadilla Press. Bosˇkovic´, Zˇeljko. 2000. Sometimes in [Spec,CP], sometimes in situ. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 53–87. Cambridge, Mass.: MIT Press. Bromberger, Sylvain. 1992. On what we know we don’t know: Explanation, theory, linguistics, and how questions shape them. Chicago: University of Chicago Press. Cheng, Lisa L.-S., and Johan Rooryck. 2002. Types of wh-in-situ. Ms., Leiden University. Cho, Eun. 1998. Why, contrastive topic, and LF movement. In David Silva, ed., Japanese/Korean linguistics 8, 403–415. Stanford, Calif.: CSLI Publications. Cho, Sook Whan. 1981. The acquisition of word order in Korean. Master’s thesis, University of Calgary. Choi, Hye-Won. 1996. Optimizing structure in context: Scrambling and information structure. Doctoral dissertation, Stanford University. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A life in language, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Adriana Belletti, ed., Structures and beyond: The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Chung, Daeho. 1996. On the representation and licensing of Q and Qdependencies. Doctoral dissertation, University of Southern California.

346

Heejeong Ko

Collins, Chris. 1991. Why and how come. In Lisa L.-S. Cheng and Hamida Demirdache, eds., More papers on wh-movement, 31–45. MIT Working Papers in Linguistics 15. Cambridge, Mass.: MIT, Department of Linguistics and Philosophy, MITWPL. Gue´rin, Valerie, and Hooi Ling Soh. 2003. Intervention e¤ect and Case checking in Mandarin Chinese and French wh-in-situ. Paper presented at the annual meeting of the Linguistic Society of America, Atlanta, Ga. Hagstrom, Paul. 1998. Decomposing questions. Doctoral dissertation, MIT. Hamblin, Charles Leonard. 1973. Questions in Montague English. Foundations of Language 10, 41–53. Hegarty, Michael. 1992. Adjunct extraction without traces. In Dawn Bates, ed., Proceedings of the 10th West Coast Conference on Formal Linguistics, 209–223. Stanford, Calif.: CSLI Publications. Hoji, Hajime. 1985. Logical Form constraints and conﬁgurational structures in Japanese. Doctoral dissertation, University of Washington. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT. Karttunen, Lauri. 1977. Syntax and semantics of questions. Linguistics and Philosophy 1, 3–44. Kim, Shin-Sook. 2002. Focus matters: Two types of intervention e¤ect. Paper presented at the 21st West Coast Conference on Formal Linguistics. Kim, Soo Won. 1991. Chain scope and quantiﬁcation structure. Doctoral dissertation, Brandeis University. Kim, Young-joo. 1997. The acquisition of Korean. In Dan I. Slobin, ed., The crosslinguistic study of language acquisition, vol. 4, 336–443. Mahwah, N.J.: Lawrence Erlbaum. Ko, Heejeong. 2005a. Syntactic edges and linearization. Doctoral dissertation, MIT. Ko, Heejeong. 2005b. Syntax of why-in-situ: Merge into [Spec,CP] in the overt syntax. Natural Language and Linguistic Theory 23, 867–916. Kratzer, Angelika, and Junko Shimoyama. 2002. Indeterminate pronouns: The view from Japanese. In Yukio Otsu, ed., The Proceedings of the Third Tokyo Conference on Psycholinguistics, 1–25. Tokyo: Hituzi Syobo. Kurata, Kiyoshi. 1991. The syntax of dependent elements. Doctoral dissertation, University of Massachusetts, Amherst. Kuwabara, Kazuki. 1998. Overt wh-movement and scope-ﬁxing scrambling: A preliminary study. In Kazuko Inoue, ed., Report (2) researching and verifying an advanced theory of human language, 115–127. Chiba, Japan: Kanda University of International Studies. Lasnik, Howard, and Mamoru Saito. 1984. On the nature of proper government. Linguistic Inquiry 15, 235–289.

Structural Height of Reason Wh-Adverbials

347

Lee, Ki-Suk, and Satoshi Tomioka. 2000. Intervention e¤ects are topic e¤ects: Wh-questions in Japanese and Korean. Paper presented at Japanese/Korean Linguistics 10. Lee, Young-Sook. 1993. Scrambling as Case-driven obligatory movement. Doctoral dissertation, University of Pennsylvania. Levin, Beth, and Malka Rappaport Hovav. 1985. Unaccusativity: At the syntax– lexical semantics interface. Cambridge, Mass.: MIT Press. Lin, Jo-Wang. 1992. The syntax of zenmeyang ‘how’ and weishenme ‘why’ in Mandarin Chinese. Journal of East Asian Linguistics 1, 293–331. Mahajan, Anoop. 1990. The A/A-bar distinction and movement theory. Doctoral dissertation, MIT. McCloskey, James. 2002. Resumption, successive cyclicity, and the locality of operations. In Samuel David Epstein and T. Daniel Seely, eds., Derivation and explanation, 184–226. Malden, Mass.: Blackwell. Miyagawa, Shigeru. 1997a. Against optional scrambling. Linguistic Inquiry 28, 1–25. Miyagawa, Shigeru. 1997b. On the nature of wh-scope. Ms., MIT. Miyagawa, Shigeru. 1999. Reconstruction and ‘‘ECP’’ e¤ects. Ms., MIT. Otsu, Yukio. 1994. Early acquisition of scrambling in Japanese. In Teun Hoekstra and Bonnie D. Schwartz, eds., Language acquisition studies in generative grammar, 253–264. Amsterdam: John Benjamins. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: MIT Press. Pesetsky, David, and Esther Torrego. 2001. T-to-C movement: Causes and consequences. In Michael Kenstowicz, ed., Ken Hale: A life in language, 355–426. Cambridge, Mass.: MIT Press. Richards, Norvin. 2001. Movement in language. New York: Oxford University Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1997. The ﬁne structure of the left periphery. In Liliane Haegeman, ed., Elements of grammar, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 1999. On the position of ‘‘Int(errogative)’’ in the left periphery of the clause. Ms., Universita` di Siena. Saito, Mamoru. 1982. Scrambling, topicalization, and strong crossover. Ms., MIT. Saito, Mamoru. 1985. Some asymmetries in Japanese and their theoretical implications. Doctoral dissertation, MIT. Saito, Mamoru. 1992. Long distance scrambling in Japanese. Journal of East Asian Linguistics 1, 69–118. Saito, Mamoru. 1994. Additional-wh e¤ects and the adjunction site theory. Journal of East Asian Linguistics 3, 195–240.

348

Heejeong Ko

Sauerland, Uli, and Paul Elbourne. 2002. Total reconstruction, PF movement, and derivational order. Linguistic Inquiry 33, 283–319. Schae¤er, Jeannette. 2000. Object scrambling and speciﬁcity in Dutch child language. In M. Susan Powers and Cornelia Hamann, eds., The acquisition of scrambling and cliticization, 71–93. Dordrecht: Kluwer. Slobin, Dan I. 1966. The acquisition of Russian as a native language. In Frank Smith and George A. Miller, eds., The genesis of language: A psycholinguistic approach, 129–152. Cambridge, Mass.: MIT Press. Soh, Hooi Ling. 1998. Object scrambling in Chinese. Doctoral dissertation, MIT. Soh, Hooi Ling. 2005. Wh-in-situ in Mandarin Chinese. Linguistic Inquiry 36, 143–155. Sohn, Keun-Won. 1995. Negative polarity items, scope and economy. Doctoral dissertation, University of Connecticut, Storrs. Stromswold, Karin. 1990. Learnability and acquisition of auxiliaries. Doctoral dissertation, MIT. Stromswold, Karin. 1996. Analyzing children’s spontaneous speech. In Dana McDaniel, Cecile McKee, and Helen Smith Cairns, eds., Methods for assessing children’s syntax, 23–54. Cambridge, Mass.: MIT Press. Takahashi, Daiko. 1990. Negative polarity, phrase structure, and the ECP. English Linguistics 7, 129–146. Takahashi, Daiko. 1993. Movement of wh-phrases in Japanese. Natural Language and Linguistic Theory 11, 655–678. Tanaka, Hidekazu. 1997. Invisible movement in Sika-Nai and the Linear Crossing Constraint. Journal of East Asian Linguistics 6, 143–188. Tanaka, Hidekazu. 1998. Conditions on Logical Form derivations and representations. Doctoral dissertation, McGill University. Thornton, Rosalind. 2004. Why continuity. In Alejna Brugos, Linnea Micciulla, and Christine E. Smith, eds., BUCLD 28: Proceedings of the 28th annual Boston University Conference on Language Development, 620–632. Somerville, Mass.: Cascadilla Press. Tsai, Wei-tien Dylan. 1994. On economizing the theory of A-bar dependencies. Doctoral dissertation, MIT. Tsai, Wei-tien Dylan. 1999. The hows of why and the whys of how. In Francesca Del Gobbo and Hidehito Hoshi, eds., UCI working papers in linguistics 5, 155– 184. Irvine: University of California, Department of Linguistics. Tsai, Wei-tien Dylan, and Melody Yayin Chang. To appear. Two types of whadverbials: A typological study of how and why in Tsou. Linguistic Variation Yearbook. Watanabe, Akira. 1992. Subjacency and S-Structure movement of wh-in-situ. Journal of East Asian Linguistics 1, 255–291. Watanabe, Akira. 2000. Absorption: Interpretability and feature strength. In Grant-in-aid for COE research report (4): Researching and verifying an advanced

Structural Height of Reason Wh-Adverbials

349

theory of human language: Explanation of the human faculty for constructing and computing sentences on the basis of lexical conceptual features. Chiba, Japan: Kanda University of International Studies. Watanabe, Shin. 1995. Aspects of questions in Japanese and their theoretical implications. Doctoral dissertation, University of Southern California. Yatsushiro, Kazuko. 1996. On the unaccusative construction and nominative Case licensing. Ms., University of Connecticut, Storrs.

Index

Aboutness, 122 Absorption, 302–306 Agree, 10, 22, 47, 71 agree at LF, 178 A-Movement, 206–212 Antecedent-contained deletion, 22, 234–235, 308 Anti-D-linking marker, 53 Bavarian, 102 Belfast English, 107 Breton, 28 Bridge verbs, 4 Bulgarian, 29, 35 Cartographic approach, 111 Case-agreement system, 120 Catalan, 229–230 Catalyst, 64 Chains, 101–125 A-chains, 101 ¯ -chains, 102, 109 A Chamorro, 167–170 Chinese, 50, 56, 336–338 Choice function, 300–301 Clausal Typing Hypothesis, 75 Clause-mate condition, 209 Cleft, 73 Clitic, 25 Clitic left dislocation, 276 Comparatives, 6, 33 comparative deletion, 6 comparative subdeletion, 6 Condition on Extraction Domains, 83 Condition on Uniformity of Chains, 29 Control, 106 Coptic Egyptian, 179–183 Copy, 9, 140, 183–185 multiple copies, 148–153 Covert restructuring, 219, 236–238 Covert wh-movement, 291–294

CP-Phase Constraint, 151–153 CP phases, 152 CP-Modiﬁer Hypothesis (CMH), 323 Criteria, 101, 106, 110 Criterial Freezing, 112, 125 Subject Criterion, 119–125 Croatian, 29 Cyclic principle, 197 Cyclic Spell-Out, 48 Cyclicity, 7, 10 Danish, 64 Deep structure, 200 Defective Intervention Constraint, 10, 227 Diagnostics of wh-movement, 4 Distributed Morphology, 80 D-Linking, 122, 234–238, 269, 272–275 Do-support, 72 Double wh-island e¤ect, 238–239 Doubly Filled Comp Filter, 146 Dutch, 102 Echo question, 290–291 Ellipsis, 73 VP-ellipsis, 73, 211 Empty Category Principle (ECP), 123–125, 232 English, 71 EPP-feature, 47 Expletive replacement, 84 Extended projection, 25 Extended Projection Principle (EPP), 119– 125 Extension Condition, 10, 24, 200 Feature movement, 22, 77, 79, 252 Feature strength, 203 Features focus feature, 50 pied-piper feature, 47, 63 Q-feature, 49

352 Features (cont.) uninterpretable feature, 109 wh-feature, 49, 82 Focus, 101, 113, 128, 255–258, 334–336 identiﬁcational focus, 270 information focus, 270 Focus movement, 290–293 Focus particle, 60 Freezing, 36 French, 107, 120, 205–206 Generalized transformation, 198 German, 87, 104–105, 249–269 Governing category, 208 Head versus phrase, 30 Head conﬂation, 24 Gungbe, 102 Head movement, 23, 63 Head Movement Constraint, 27 Hindi, 135–160 Hungarian, 292–306 Icelandic, 228, 230 Inclusiveness Condition, 9, 60 Insertion Prohibition, 199 Intervention E¤ect, 252, 301, 304, 308, 321 Irish, 107 Italian, 32, 36–38, 113–124 Japanese, 51, 55–57, 320–326 old Japanese, 51–53, 58–59 Kikuyu, 60 Korean, 320–326, 333–338 Child Korean, 326–332 Lakhota, 50, 54–55, 56–57 Last Resort, 9, 98, 108 Merge, 22, 47, 104 External Merge, 10 Internal Merge, 10, 22 Second Merge, 77–78 Minimal Link Condition, 27, 83 Minimal Quantiﬁed Structure Constraint, 251 Minimize Mismatch, 156–159 Modal auxiliary, 53 Move, 47, 104 Multiple Specs, 82, 300 Multiple wh-questions, 270, 298, 332 Northern Italian Dialects, 123 Null operator, 73, 167 Obviation, 174, 229 Okinawan, 61

Index Palauan, 185–188 Parasitic gap, 264 Partial movement, 142–149 Participle agreement, 120 Passamaquoddy, 143, 148–149, 173–177 Past Participle agreement, 120 PF-adjacency, 81 PF Relaxation, 156–159 Phase, 10, 27, 213 Phase Impenetrability Condition (PIC), 10, 38, 171, 231–232 Phrasal movement, 27, 251 Phrase Structure Graph-theoretic representation of- , 202– 203 Pied-piping, 22, 47, 77, 81, 116 anti-pied-piping, 32 Preposition stranding, 38 Probe goal relation, 31, 47, 50, 170, 233 Propositional Island Condition, 2 Punjabi, 135–160 Quantiﬁer-Induced Barrier, 251 Quantiﬁer-Intervention Generalization, 255 Quirky subjects, 120 Raising, 84, 106, 128 Reconstruction, 106, 108 Relative tenses, 181 Relativization, 238–239 Relatives, 73, 173, 116–118 appositive relatives, 116 embedded relatives, 137 extraposed relatives, 137 free relatives, 32 full relatives, 33 head-internal relatives, 54, 173 Relativized Minimality, 10, 28, 121, 204, 252 Restriction on LF, 251 Rhetorical questions, 52 Romance, 25 Romanian, 35 Scope-Bearing Element, 321 Scope-discourse properties, 99, 101 Scope-marking strategy, 138, 142, 154–155 Scrambling, 320 ¯ -scrambling, 324, 337 A A-scrambling e¤ect, 144 Serbo-Croatian, 157 Single Output Syntax, 227 Single pair versus Pair list, 302–310 Sinhala, 62 Sluicing, 76 Spanish old Spanish, 29 Speciﬁed Subject Condition, 2

Index Split CP-system, 332–338 S-selectional properties, 101, 103, 106 Strict Cycle Condition, 200 Subjacency, 2, 75, 201 Subject-AUX-Inversion, 72 Successive-cyclic movement, 4, 10, 154, 158, 197, 232–234 Superiority, 257, 290, 308–309 Anti-superiority e¤ect, 333 Tense Neo-Reichenbachian approach, 219–221 sequence of tense, 221–226 Topicalization, 74, 273 ¯ -Topicalization, 338 A Embedded Topicalization, 85 Multiple Topicalization, 260, 294–296 Topicalizability Generalization, 259 Transformational cycle, 198 Transparency e¤ects, 228–229 Unbounded dependency, 6 Uniformity Condition, 253 Unselective binding, 189 Vacuous Movement Hypothesis (VMH), 72, 79–84 Verb-second, 62, 258, 104 Wh-agreement, 167–170 Wh-agreement as Spell-Out of Agree, 172 Wh-indeterminate, 56–59 Wh-in-situ, 50, 184–189, 297, 300 Wh-movement multiple Wh-movement, 293–294, 299–310 optional Wh-movement, 298 partial Wh-movement, 87 wh-head movement, 33 Wh-subjects, 71 Wh-topic, 268

353

Current Studies in Linguistics Samuel Jay Keyser, general editor 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31.

A Reader on the Sanskrit Grammarians, J. F. Staal, editor Semantic Interpretation in Generative Grammar, Ray Jackendo¤ The Structure of the Japanese Language, Susumu Kuno Speech Sounds and Features, Gunnar Fant On Raising: One Rule of English Grammar and Its Theoretical Implications, Paul M. Postal French Syntax: The Transformational Cycle, Richard S. Kayne Pa¯nini as a Variationist, Paul Kiparsky, S. D. Joshi, editor ˙ Semantics and Cognition, Ray Jackendo¤ Modularity in Syntax: A Study of Japanese and English, Ann Kathleen Farmer Phonology and Syntax: The Relation between Sound and Structure, Elisabeth O. Selkirk The Grammatical Basis of Linguistic Performance: Language Use and Acquisition, Robert C. Berwick and Amy S. Weinberg Introduction to the Theory of Grammar, Henk van Riemsdijk and Edwin Williams Word and Sentence Prosody in Serbocroatian, Ilse Lehiste and Pavle Ivic´ The Representation of (In)deﬁniteness, Eric J. Reuland and Alice G. B. ter Meulen, editors An Essay on Stress, Morris Halle and Jean-Roger Vergnaud Language and Problems of Knowledge: The Managua Lectures, Noam Chomsky A Course in GB Syntax: Lectures on Binding and Empty Categories, Howard Lasnik and Juan Uriagereka Semantic Structures, Ray Jackendo¤ Events in the Semantics of English: A Study in Subatomic Semantics, Terence Parsons Principles and Parameters in Comparative Grammar, Robert Freidin, editor Foundations of Generative Syntax, Robert Freidin Move a: Conditions on Its Application and Output, Howard Lasnik and Mamoru Saito Plurals and Events, Barry Schein The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Kenneth Hale and Samuel Jay Keyser, editors Grounded Phonology, Diana Archangeli and Douglas Pulleyblank The Magic of a Common Language: Jakobson, Mathesius, Trubetzkoy, and the Prague Linguistic Circle, Jindrˇich Toman Zero Syntax: Experiencers and Cascades, David Pesetsky The Minimalist Program, Noam Chomsky Three Investigations of Extraction, Paul M. Postal Acoustic Phonetics, Kenneth N. Stevens Principle B, VP Ellipsis, and Interpretation in Child Grammar, Rosalind Thornton and Kenneth Wexler

32. Working Minimalism, Samuel David Epstein and Norbert Hornstein, editors 33. Syntactic Structures Revisited: Contemporary Lectures on Classic Transformational Theory, Howard Lasnik with Marcela Depiante and Arthur Stepanov 34. Verbal Complexes, Hilda Koopman and Anna Szabolcsi 35. Parasitic Gaps, Peter W. Culicover and Paul M. Postal 36. Ken Hale: A Life in Language, Michael Kenstowicz, editor 37. Flexibility Principles in Boolean Semantics: The Interpretation of Coordination, Plurality, and Scope in Natural Language, Yoad Winter 38. Phrase Structure Composition and Syntactic Dependencies, Robert Frank 39. Representation Theory, Edwin Williams 40. The Syntax of Time, Jacqueline Gue´ron and Jacqueline Lecarme, editors 41. Situations and Individuals, Paul D. Elbourne 42. Wh-Movement: Moving On, Lisa Lai-Shen Cheng and Norbert Corver, editors