Cleft Constructions in Japanese Syntax Mika Kizu
Cleft Constructions in Japanese Syntax
This page intentionally lef...
49 downloads
1384 Views
778KB 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
Cleft Constructions in Japanese Syntax Mika Kizu
Cleft Constructions in Japanese Syntax
This page intentionally left blank
Cleft Constructions in Japanese Syntax Mika Kizu Department of East Asian Studies University of Durham United Kingdom
© Mika Kizu 2005 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1T 4LP. Any person who does any unauthorized act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The author has asserted her right to be identified as the author of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2005 by PALGRAVE MACMILLAN Houndmills, Basingstoke, Hampshire RG21 6XS and 175 Fifth Avenue, New York, N.Y. 10010 Companies and representatives throughout the world PALGRAVE MACMILLAN is the global academic imprint of the Palgrave Macmillan division of St. Martin’s Press, LLC and of Palgrave Macmillan Ltd. Macmillan® is a registered trademark in the United States, United Kingdom and other countries. Palgrave is a registered trademark in the European Union and other countries. ISBN 1–4039–1235–1 This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Kizu, Mika, 1964– Cleft constructions in Japanese syntax / Mika Kizu. p. cm. Includes bibliographical references and index. ISBN 1–4039–1235–1 (cloth) 1. Japanese language—Grammar, Generative. 2. Japanese language— Syntax. I. Title. PL534.K58 2005 495.6′5′0182—dc22 10 14
9 13
8 12
7 11
6 10
2004054892 5 09
4 08
3 07
2 06
1 05
Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham and Eastbourne
This book is dedicated to my family; I can never thank them enough for their support, patience and fundamental trust in me
This page intentionally left blank
Contents
Acknowledgements
ix
List of Abbreviations
xi xiii
Foreword 1 Introduction Preliminaries Background: English cleft constructions The phenomenon of interest: Japanese cleft constructions Outline of the book
1 1 2 3 7
2 Topicalization and Cleft Constructions Introduction Topicalization in Japanese: a brief overview Our assumptions about topicalization Parallelisms Scrambling and clefting Summary
9 9 10 16 24 47 56
3 Nominalizations in Cleft Constructions Introduction Properties of no No in presuppositional clauses versus prenominal no Relative clauses and No-clauses of clefts The no-clause as a CP Summary
57 57 58 63 66 77 99
4 Resumptive A’-Dependencies Introduction Basic facts and proposal Reconstruction in long-distance clefts Properties of A’-movement Resumption Residual issues Some implications: highest clause sensitivity Summary vii
101 101 102 110 121 130 146 150 156
viii Contents
5 Ellipsis in Cleft Constructions Introduction Background and a proposal Basic facts about Japanese sluicing Sluicing as a cleft construction Other issues in Japanese sluicing Some consequences Summary
159 159 160 162 165 175 186 203
6 Conclusions
205
Notes
207
References
233
Index
242
Acknowledgements First and foremost, my deepest and most sincere gratitude goes to Mark Baker. I was extremely fortunate to be able to work with him at McGill. He set an extraordinary example of what a linguist should be. His enthusiasm for linguistics constantly amazed me, and his remarkably high standard of scholarship inspired me throughout my time at McGill. I benefited tremendously from his insightful comments and suggestions, which deepened my own work considerably. I would also like to thank Lisa Travis, Glyne Piggott, and Jonathan Bobaljik. I was indebted to Lisa Travis for her generous support. Her quick reactions and keen insight into language again set an excellent example to follow. I was grateful for her comments on part of an early draft of this work and on my research project done throughout my time at McGill. Glyne Piggott was a great mentor. I wish to thank him for his patience, and for always keeping his door open to me. I benefited from his warm encouragement in innumerable ways. It was such a privilege to have been able to work with Jonathan Bobaljik, who read an early version carefully, and made comments of great value and detail. The other McGill faculty members supported me in various ways. My gratitude goes especially to Nigel Duffield, Brendan Gillon, Heather Goad, and Lydia White. I would also like to thank my former fellow students and friends from the department: José Bonneau, Lotus Goldberg, Yahiro and Makiko Hirakawa, Hironobu Hosoi, Takako Kawasaki, Mikinari Matsuoka, Yuko Mochizuki, Junko Murai, Masanori Nakamura, Tomoyo Oda, Ileana Paul, Asya Pereltsvaig, Vivianne Phillips, Philippe Prévost, Roumyana Slabakova, O-T. Stewart, Mami Takahashi, Hidekazu Tanaka, Miwako Uesaka, Mikael Vinka, and Ching-huei Wu. I am particularly grateful to Hidekazu, who first inspired me with a fun trip to the ‘islands’ of Japanese. My intellectual debt to him has been huge, and his deep influence on my thinking will be obvious throughout the pages of this book. Lotus also deserves special mention. She kindly proofread the original version of this book, and, more importantly, her friendship and moral support helped me to overcome the difficulties in the course of doing graduate work. In addition to the people I met at McGill, I was fortunate to know other young linguists in North America. For their help and discussions, I would especially like to thank John Alderete, Kleanthes K. Grohmann, ix
x Acknowledgements
Chris Kennedy, Jeong-Seok Kim, Kiyomi Kusumoto, Ayumi Matsuo, Jason Merchant, Kunio Nishiyama, Masao Ochi, Philippe Spaelti, Ayako Yamagata, and particularly, Motoko Katayama. Stepping back more, I should mention the time that I spent in the incredibly rich educational environment of UC Santa Cruz. The courses, discussions, and conversations in which I participated there have been so influential to my development, both as a linguist and as a teacher. Special thanks to Judith Aissen, Sandy Chung (my MA thesis adviser), Donka Farkas, Junko Itô, Armin Mester, Jim McCloskey, and Jaye Padgett for their outstanding teaching and guidance. My gratitude goes beyond North America. It was my good fortune to have met Yasuaki Abe at Nanzan University. He helped and guided me towards the beginning of my studies in linguistics. For their help and encouragement since I was at Nanzan, many thanks also to Hiroshi Aoyagi, Noriko Ban, Shun-ichi Kato, Nanako Machida, Tadashi Sakamoto, and the late Akira Komai. I would also like to thank my colleagues in the Department of East Asian Studies, University of Durham, UK: especially, Don Starr, John Weste, Naomi Cross and Kazuki Morimoto, who provided me with their generous assistance in the final preparation of the script. The anonymous reviewer from Palgrave Macmillan was rigorous and helpful when reading the draft version of the book, and I am very grateful for the suggested revisions. Thank you to Jill Lake for her patience and consideration. MIKA KIZU
List of Abbreviations A(-movement) A’(-movement) ACC Adv AdvP Agr AgrO AgrOP AgrS AgrSP AP C(OMP) CL CLLD COP CP D(ET) DAT D-linked DP e ECP GEN HERC HIRC H-lessRC Hon Hum I(NFL) IP LCC LF Move F N NEG NM
(movement to) A-positions (movement to) A’-positions accusative adverb adverb phrase agreement object agreement object agreement phrase subject agreement subject agreement phrase adjective phrase complementizer classifier clitic left dislocation copula complementizer phrase determiner dative discourse linked determiner phrase empty category empty category principle genitive head-external relative clause head-internal relative clause headless relative clause honorific humble inflection inflection phrase linear crossing constraint logical form move a feature noun negative nominalizer xi
xii List of Abbreviations
NOM NP NPI NQ NumP Op P PAST PF pg Pol PP pro Pro PRO Q QP RC R-expression S SC SOV spec Subj t T(ns) TOP TopP TP V VOS VP WCO Wh X0 XP ʔ
nominative noun phrase negative polarity item numeral quantifier number phrase operator post- or preposition past tense phonetic form parasitic gap polite post- or prepositional phrase zero pronoun pronominal form pronominal anaphor question marker quantifier phrase relative clause referential expression sentence small clause subject-object-verb specifier subject trace tense topic topic phrase tense phrase verb verb-object-subject verb phrase weak crossover WH expressions head of category X maximal projection of category X glottal stop
Foreword This book is concerned with the syntactic properties of cleft constructions in Japanese within the Principles and Parameters framework with some consideration of the Minimalist Program. The constructions in question are divided syntactically into two sub-parts, a focus element and a presuppositional clause. This book claims that the focused element’s position is not derived via a process of movement such as scrambling, but is instead base-generated in a predicate position occupied by a single constituent. Presuppositional clauses of clefts are argued to involve movement of a null operator, which is analyzed on a par with topicalization. This is supported by various syntactic parallelisms between cleft and topic constructions in this language. While the presuppositional clause is marked by a nominalizer, it is shown that it does not project a nominal category; in this sense, cleft constructions are analogous to head-internal relative clauses. Furthermore, one of the most interesting properties of the cleft construction, the syntactic phenomenon of ‘connectivity’, is closely examined. I propose that long-distance cleft constructions in Japanese have peculiar structures: a null operator originates adjoined to the highest complement clause, and its thematic position is occupied by pro. This analysis is supported by empirical facts which involve binding relations, weak crossover effects, interactions with another A’-dependency, and clefting adjunct PPs. It is shown that these types of resumptive A’-dependencies are observed across languages as well as in different constructions within the Japanese language. Finally, the discussion is extended to so-called sluicing in Japanese. This book observes similar syntactic behavior in sluicing and clefts, which leads to the argument that sluicing sentences are derived from cleft constructions. MIKA KIZU
xiii
This page intentionally left blank
1 Introduction
Preliminaries This book examines the syntactic properties of cleft constructions in Japanese within the Principles and Parameters framework (Chomsky 1981, 1982, 1986a, 1986b and many others), with some consideration of the Minimalist Program (Chomsky 1995 and others). In a cleft construction, a focal element is displaced from the remainder of the clause, and then connected to this remainder with a copula. Although this construction has been broadly investigated in the literature on English, the analogous Japanese construction has still remained for further research, to my knowledge. This book deals with a number of facts concerning Japanese cleft constructions, and touches on larger issues in Japanese syntax, as well as in the grammar of human language in general. A major theoretical hypothesis of this book is that the cleft construction does not exist as a distinct theoretical entity. Following the insight of Chomsky (1977, 1981), syntactic constructions themselves are viewed as epiphenomenal, and are derived not by construction-specific rules but instead from the interaction of general principles. Cleft constructions in particular are argued to arise from a combination of the theories of A’-movement, predication relations and copulas. While this analytical view is not novel, we can note that this outcome lends support to the Principles and Parameters approach, and argues against approaches, such as Construction Grammar (Goldberg 1995 and references cited therein), which propose that each individual construction has its specific meaning, being accounted for by principles which apply only to that construction. Moreover, this book will present detailed discussion of the properties of A’-movement in clefts and other similar constructions in Japanese, which are crucially and systematically different from the 1
2 Cleft Constructions in Japanese Syntax
properties of A’-movement in English. This claim has strong empirical supports, some of which have not been observed in the literature on Japanese. The present chapter will review the fundamental facts of English and Japanese cleft constructions, and will outline empirical issues concerning Japanese cleft constructions which will be discussed in the following chapters. The larger organization of this book will be described at the end of the current chapter.
Background: English cleft constructions Cleft constructions in English have been explored extensively within the generative tradition (Akmajian 1970, Chomsky 1970, 1977, Higgins 1979, among many others). The sentences in (1) demonstrate two types of cleft constructions in English: (1)a shows an it-cleft, and (1)b a pseudocleft: (1) a It was [NP the movie] [CP that Mary saw] it-cleft b [CPWhat Mary saw] was [NPthe movie]
pseudocleft
Roughly speaking, the syntactic structures of it-cleft and pseudocleft constructions fulfill similar semantic functions.1 In terms of information structure, the CPs express a presupposition, and the NPs express the focus in (1). I will henceforth call the former a presuppositional clause and the latter a focus phrase. The analysis proposed in the literature assumes that in both it-clefts and pseudoclefts, Wh (operator)-movement of some kind takes place in the presuppositional clause. The structures in (2) are from Chomsky (1977) and Heggie (1988), which I take as standard analyses for English cleft constructions within Government and Binding theory: (2) a [IP it [VP be [CP XPi [CP Opi thati[IP . . . ei . . . ]]]]] (Chomsky 1977) b [CP[CPOpi . . . ei . . . ]j [IP be [SC XPi tj]]] (Heggie 1988) The sentences in (1) are represented as the schematized structures in (2), respectively. The null operator, which is coindexed with the focus phrase, moves to the spec of CP in the presuppositional clause, and the focus XP is licensed by a predication relation with the presuppositional clause which contains the coindexed null operator (Williams 1980, 1983).2 Thus, the structures in (2) account for the fact that English cleft constructions exhibit the standard constraints on A’-movement.3
Introduction 3
One of the most interesting issues concerning English cleft constructions is the syntactic phenomenon of ‘connectivity’ or reconstruction.4 Even though there is no c-command relation between the focus phrase and its antecedent in the presuppositional clause, the effects of binding dependencies are apparent: (3) a b
It was a picture of herselfi that Maryi bought What Maryi bought was a picture of herselfi
(4) a b
It was hisi own analysis that every studenti believed What every studenti believed was hisi own analysis
(5) a* b
It was any good books that he didn’t buy What he didn’t buy was any good books
The sentences in (3) show that the anaphor, herself, can be bound by the non-c-commanding Mary. The same is true for variable binding, as shown in (4): despite the absence of a c-command relation, the pronoun, his, can be coindexed with the quantified phrase, every student, yielding a bound variable reading. Although it-cleft constructions do not allow the clefting of a Negative Polarity Item (NPI), as shown in (5)a, the NPI is licensed in the pseudocleft sentence, as seen in (5)b.5 These facts are not expected under the standard Binding theory assumptions (Chomsky 1981, 1986a, 1986b), because binding basically requires c-commanding and coindexation. While numerous attempts have been made to account for the connectivity phenomena in the analyses of clefts, the implementation is still a matter of controversy.
The phenomenon of interest: Japanese cleft constructions With this background, let us now turn to Japanese cleft constructions. Unlike English, it appears that Japanese has only one kind of cleft construction, as shown in (6) below. The Japanese cleft construction resembles the English pseudocleft, with the presuppositional clause consisting of a ‘free relative’ clause followed by a focus element and a copular verb:6 (6) [Hanako-ga [e] atta no] -wa [sono hito-ni] da7 Hanako-NOM met NM TOP that person-DAT COP ‘It was that person that Hanako met’ Japanese has SOV word order. The presuppositional clause is marked by the nominalizer no and the entire clause is topicalized. This sequence is
4 Cleft Constructions in Japanese Syntax
followed by the focus phrase, sono hito-ni ‘that person-DAT’, and the copular verb, da.8 Given the facts we reviewed for English cleft constructions above, our examination of Japanese cleft constructions such as (6) will consider the following issues: (7) a Do Japanese cleft constructions involve operator movement like their English counterparts? b What is the syntactic structure of Japanese cleft constructions? c Are Japanese cleft constructions like English clefts in having connectivity effects? If so, how should we account for them? The questions formulated in (7) will be explored throughout the following chapters via the examination of the syntactic behavior of Japanese clefts. Why are the questions in (7) important? Concerning (7)a we can note that Japanese Wh-constituents appear in situ and are not considered to move to the designated spec position, at least overtly.9 Given this, it is not immediately obvious that the operator movement analysis shown in (2) is applicable to Japanese cleft constructions. Furthermore, although relative clauses are typical operator-movement constructions in English, the same constructions in Japanese do not obey the syntactic constraints on movement (Kuno 1973: 249): (8) a [[ ei sinda noni] daremo kanasimanakatta] hitoi died although anyone saddened-not-was person ‘the person who no one was saddened although (he/she) died’ b [[[ ei ej kiteiru] yoofukuj]-ga yogoreteiru] sinsii wearing suit-NOM dirty-be gentleman ‘the gentleman such that the suit that (he) is wearing is dirty’ These are well-known examples of relative clauses in Japanese. If these constructions are analyzed on a par with English counterparts and assumed to contain a null operator, then this operator would move outside an island: an adjunct clause in (8)a and a relative clause in (8)b. This would, of course, be a subjacency violation, and hence, the sentences in (8) should be ungrammatical. Both sentences in (8) sound perfect, however. Since the standard analysis for English (Chomsky
Introduction 5
1977) takes relative clauses and cleft constructions to both involve Wh-movement, these facts suggest that cleft constructions in Japanese might not involve A’-movement. Hoji (1990), in fact, argues that Japanese cleft constructions have two types of structures, one which involves movement and another which does not.10 Under this view, the sentence in (6) is analyzed either as in (9)a or in (9)b:11 (9) a [NP [IP . . . (proi) . . . ] [NP noi]] -wa NPi da NM-TOP COP b [CP Opi [IP . . . ti . . . ] no]-wa NPi-(CASE) da NM-TOP COP In (9)a, no movement has taken place. Instead, the structure is licensed by the so-called aboutness condition (Kuno 1973). This book, however, focuses on (9)b in which Japanese cleft constructions always involve movement. Concerning (7)b, cleft structures will be analyzed here as containing two parts, a presuppositional clause and a focus phrase. As we saw in (6), the presuppositional clause is marked with no, an element which has various uses in Japanese. For example, the following sentence contains a headless relative clause (or free relative clause) in subject position: (10)
[Taroo-ga e mita no]-wa omosirok-atta Taro-NOM saw NM-TOP be-interesting-PAST ‘The one Taro saw was interesting’
The headless relative clause in (10) is marked with no just like the presuppositional clause in (6). The question of how the two differ syntactically should be addressed in any theory of Japanese cleft constructions, and will thus be explored in this book. The focus phrase of the Japanese cleft construction presents an interesting puzzle as well. Contrasting with English cleft constructions, the Japanese cleft allows multiple foci:12 (11) [Taroo-ga e e ageta no]-wa [dare-ni nani (-o)]-na-no Taro-NOM gave NM-TOP who-DAT what-ACC-COP-Q ‘(lit.) To whom (and) what was it that Taro gave?’
6 Cleft Constructions in Japanese Syntax
The sentence in (11) shows that more than one Wh-constituent can occur in focus position. If we assume that this position is occupied by a single constituent, this fact arises unexpectedly. To begin our answer to (7)c, let us first consider the following examples:
(12) a
b
[Hanakoi-ga hon-o e ageta no]-wa Hanako-NOM book-ACC gave NM-TOP zibuni -no hahaoya-ni dat-ta self-GEN mother-DAT COP-PAST ‘(lit.) Who Hanakoi gave a book was herself’si mother’ ‘(lit.) It was herself’si mother that Hanakoi gave a book’ [Daremoi-ga e saisho-ni au no] -wa everyone-NOM first time meet NM-TOP zibuni-no hahaoya-ni da self-GEN mother-DAT COP ‘Who everyonei meets first is hisi own mother’ ‘It is hisi own mother that everyonei meets for the first time’
c* [Taroo-ga e aw-ana-katta no]-wa Taro-NOM meet-NEG-PAST NM-TOP da-ta dono sensei-ni-mo any teacher-DAT-also COP-PAST ‘Who Taro didn’t meet was any teacher’
(12)a contains the anaphoric pronoun, zibun ‘self’, in focus position. As in English connectivity facts, zibun is coindexed with its antecedent, Hanako, which is in the presuppositional clause. Zibun can also be interpreted as a bound variable, as shown in (12)b. Here again, the focused zibun is interpreted as bound by the quantifier, daremo-ga ‘everyoneNOM’, despite the fact that there is no c-command relation between the binder and the bindee. In contrast, the NPI, dono NP-mo ‘any NP-also’, in focus position in (12)c is not licensed. Thus, from what these sentences show, the behavior of Japanese cleft constructions parallels exactly that of English it-cleft constructions. Interestingly, however, a striking difference between Japanese and English cleft constructions emerges when more complex connectivity effects are examined. Consider the following long-distance cleft constructions:
Introduction 7
(13) a Tarooi-ga [Hanakoj-ga zibuni/j-no tomodachi-ni atta to] Taro-NOMHanako-NOMselfGEN friend-DAT met COMP sinziteita believed ‘(lit.) Taroi believed that Hanakoj met self’si/j friend’ b [Tarooi-ga [Hanakoj-ga e atta-to] sinziteita-no]-wa Taro-NOM Hanako-NOM met COMP believed-NM-TOP [zibuni/*j-no tomodachi-ni] dat-ta self-GEN friend-DAT COP-PAST ‘(lit.) It was self’s friendi/*j that Taroi believed that Hanakoj met’ (14) a Tarooi-ga [Hanakoj-ga zibun-nituitei/j hanasitato] omotta Taro-NOM Hanako-NOM self-about talked COMP thought ‘(lit.) Taroi thought that Hanakoj talked about selfi/j’ b [Tarooi-ga [Hanakoj-ga e hanasita to] omotta no]-wa Taro-NOMHanako-NOM talked COMP thought NM-TOP zibun-nituitei/*j dat-ta self-about COP-PAST ‘(lit.) It was about selfi/*j that Taroi thought that Hanakoj talked’ In the non-clefted (13)a and (14)a examples, zibun appears within the embedded CP complement. It is well known that zibun is a longdistance, subject-oriented anaphor. The zibun in (13)a and (14)a can therefore refer either to the matrix subject, Taroo, or to the embedded subject, Hanako. The standard analysis of English cleft constructions would predict successive-cyclic movement of a null operator, and thus zibun in focus position can be coindexed with either Taroo or Hanako. The fact is, however, the zibun in the focus position in (13)b and (14)b can be coindexed only with the matrix subject, and not with the embedded subject. This casts doubt on the idea that Japanese cleft constructions are to be analyzed on a par with English cleft constructions. We will consider this issue further in Chapter 4.
Outline of the book The organization of this book is as follows. Chapter 2 shows that Japanese cleft constructions are analogous syntactically to topic constructions, continuing a line of analysis begun by Hoji (1990). We will depart from Hoji’s analysis, however, and we will explore the parallelisms further. The ungrammaticality of NPI licensing in (12)c will be accounted
8 Cleft Constructions in Japanese Syntax
for by the analysis proposed for topicalization and clefting. Furthermore, I will argue that cleft constructions crucially involve movement within the presuppositional clause, a process which differs from scrambling but which is similar to topicalization. This will force us to assume that there is a null operator in presuppositional clauses. At the end of this chapter, the structure of focus phrases is investigated. The main concern in Chapter 3 is the nominalizer no and its syntactic category in presuppositional clause of clefts. I will argue that the no in cleft constructions is a complementizer, and that the presuppositional clause does not project a nominal category. This claim is based on a comparison of the presuppositional clauses of clefts and so-called headinternal relative clauses (HIRCs). These two clause types share similar syntactic property in that both are CPs rather than NPs. In this respect, presuppositional clauses and HIRCs are distinct from other types of relative clauses. Chapter 4 is concerned with the somewhat peculiar connectivity phenomena in long-distance cleft constructions. Empirical facts from reconstruction, weak crossover, and interactions with other A’-dependencies show that the null operator in the presuppositional clause of long-distance clefts is base-generated in some intermediate position rather than in its thematic position. Under the analysis I propose, the null operator in adjoined position is associated with a pro in thematic position, and this operator undergoes short movement to the spec of CP. This structure type is not specific to Japanese, but also occurs in A’-dependencies in other languages. Such ‘resumptive’ A’-dependencies appear in other Japanese constructions as well. Reconstruction effects in relative clauses and case conversion phenomena in Japanese provide support for our proposal that long-distance A’-dependencies may not be truly ‘long-distance’ movement after all, but rather appear only in the highest clause of the structure. As an extension of the discussion in Chapters 2–4, Chapter 5 deals with sluicing in Japanese. It has been pointed out by Kizu (1997), Kuwabara (1995, 1996) and Nishiyama, Whitman and Yi (1996), that Japanese sluicing involves some sort of copular construction, and more specifically, a cleft construction. The discussion in this chapter supports the cleft analysis of Japanese sluicing, and, more importantly, provides empirical evidence in favor of a PF deletion approach for ellipsis, addressing the central debate in the literature on ellipsis.
2 Topicalization and Cleft Constructions
Introduction The main purpose of this chapter is to show that cleft constructions in Japanese parallel topic constructions in syntactic structure. This basic line of analysis has already been claimed by Hoji (1990), but we will explore further into this issue here, and will argue for a slightly different analysis from Hoji’s. The points I would like to make are: 1) the cleft sentences must involve movement of a null operator, which is parallel to topicalization in Japanese, and 2) the focus position in a cleft construction is basegenerated as a predicate and is not created, for instance, by a movement operation of scrambling. In other words, the present chapter is concerned with properties of movement in presuppositional clauses and the structure of the focus position. The internal structures of presuppositional clauses will be closely examined in Chapters 3 and 4. The outline of Chapter 2 is as follows. First I briefly summarize the previous research on topicalization in Japanese, especially Kuno (1973), Saito (1985) and Hoji (1985). Then I present the assumptions that will be made in this chapter. Although it has been assumed that Japanese topic constructions are not necessarily derived by any movement operation, I will show that both topicalization and clefting involve movement. Chapter 2 also shows some parallelism between topicalization, relativization and clefting, and claims that cleft constructions as well as relative clauses are related to topicalization in Japanese; namely, they are all instances of A’-movement, as traditionally proposed in theories of generative grammar. The important consequence of the proposed analysis is to account for ungrammatical cases of connectivity with a Negative Polarity Item (NPI) in focus position. Chapter 2 goes on to 9
10 Cleft Constructions in Japanese Syntax
argue that the movement in clefts is distinct from scrambling and presents the structure for the focus position. The final section is a summary.
Topicalization in Japanese: a brief overview This section briefly reviews the previous analyses for topicalization in Japanese, especially base-generation analyses for topic phrases, and sketches out some important empirical facts concerning base-generation in the topic construction. It will be shown later that the syntactic properties of base-generated topic constructions we will see in this section can be explained in a different way. Kuno (1973) and Saito (1985) It has been acknowledged in the literature (Perlmutter 1972, Kuno 1973, Saito 1985, Hoji 1985, Kubo 1992 among many others) that topic phrases in Japanese may be base-generated in sentence-initial position.1,2 For instance, Saito (1985) argues, based on Kuno’s (1973) observation, that wa-marked phrases are licensed by a so-called ‘aboutness relation’ with the rest of the sentence, and do not have to bind a variable: (1)
a
Tarooi-wa [ proi sono hon-o yon-da] Taro-TOP that book-ACC read-PAST ‘As for Taro, (he) read that book’
b Sono honi-wa [ Taroo-ga proi that book-TOP Taro-NOM ‘As for that book, Taro read’
yon-da] read-PAST
In the sentences in (1), the subject, Taroo(-ga) ‘Taro(-NOM)’, and the object, sono hon(-o) ‘that book(-ACC)’, are topicalized with the topic marker, wa.3 Saito assumes that the topic phrases may not necessarily be derived by movement but can be base-generated sentence-initially (an adjoined position to IP), being licensed by an aboutness relation with the bracketed clause of proposition. As we can see from the meaning of the sentences above, the function of the topic phrases are to indicate an element which the proposition (bracketed clauses) is about. The evidence that topic sentences may not involve any movement operation is the following. First of all, it appears that topic constructions in Japanese do not obey island constraints:
Topicalization and Cleft Constructions 11
(2)
a
Sono hitoi- wa [[ ei sinda noni] daremo kanasimanakatta] that person-TOP died although anyone saddened-not-was ‘As for that person, no one was saddened although (he/she) died’
b Sono sinsii-wa [[[ ei ej kiteiru] yoofukuj]-ga yogoreteiru] that gentleman-TOP wearing suit-NOM dirty-be ‘As for that gentleman, the suit (he) is wearing is dirty’ (Kuno 1973: 249) The sentence in (2)a contains an adjunct clause in which the gap is coindexed with the sentence-initial topic phrase. If the gap is a trace of the topic phrase, we should expect ungrammaticality for the sentence; the topic phrase is moved from the adjunct island. The same is true for the sentence in (2)b. If the topic phrase is moved from the relative clause, the sentence would violate the complex NP constraint. Since both sentences are grammatical, the facts in (2) suggest that the topic phrases are not created by movement.4 Second, Kuno (1973) points out that Japanese allows a sentence in which a topic phrase appears not to bind any argument position in the clause: (3)
a
Sakana-wa [tai-ga oisii] fish-TOP red snapper-NOM be-delicious ‘As for fish, red snappers are delicious’
b Hana-wa [sakura-ga ii] flower-TOP cherry blossoms-NOM be-good ‘As for flowers, cherry blossoms are the best’ The topic phrases in (3) do not correspond to any argument of the predicates in the sentences, and therefore, we cannot assume that the topic phrases are derived by movement from the bracketed clause.5 Third, as noted by Kuno (1973) and discussed extensively by Saito (1985), topic constructions allow overt resumptive pronouns: (4)
a?
Sono hitoi-wa [John-ga karei-no imooto-o that person-TOP John-NOM he-GEN sister-ACC yoku sitteiru rasii] well know seem ‘As for that person, it seems that John knows his sister very well’
12 Cleft Constructions in Japanese Syntax
b?
Hirosimai-wa [Amerika-ni [[ ej sokoi-kara kita] Hiroshima-TOP America-in there-from came oozei iru] hitoj]-ga person-NOM many be ‘As for Hiroshima, there are many people in the United States who are from there’ (Saito 1985: 287)
While the sentences above do not sound perfect, they are not ungrammatical. The facts in (4) alone do not necessarily show that topicalization in Japanese involves base-generation; although English allows left dislocation constructions which contain an overt resumptive pronoun, topicalization involves movement in the same language (Chomsky 1977, Baltin 1987, Lasnik and Saito 1992 and others). However, given that Japanese allows null pronouns in general (Perlmutter 1972) unlike English, and together with the other pieces of evidence that we have seen so far, we can conclude that Japanese topic phrases can be base-generated. Hoji (1985) Notice that the sentences we have observed in the last subsection are a particular type of topic construction: they all have NP topics in sentenceinitial position. Hoji (1985) extends the data to other types of topic phrases and argues that Japanese topic phrases are divided into two types: thematic topics (or topic wa in Hoji’s term) and contrastive topics. Contrastive topics are interpreted as something like ‘at least’ or some kind of exhaustive listing whereas thematic topics do not have such an implication. Hoji claims that this semantic distinction between thematic and contrastive topics corresponds to a difference in syntax. According to Hoji, a contrastive topic in sentence-initial position is always derived by movement, whereas thematic topics can be base-generated there, being licensed by the aboutness relation with the rest of the sentence. The following sentences are basic examples of contrastive and thematic topic constructions: (5)
a
contrastive topic Montreal-nii-wa Taroo-ga ti nandomo itta Montreal-to-TOP Taro-NOM many -times went ‘To Montreal (but not anywhere else), Taro went many times’
Topicalization and Cleft Constructions 13
b thematic topic Montreali-wa, Taroo-ga proi nandomo itta Montreal-TOP Taro-NOM many-times went ‘As for Montreal, Taro went there many times’ The sentence in (5)a contains a PP topic phrase whereas (5)b has an NP topic. As the translation shows, the PP topic prefers to yield a contrastive reading, and the NP topic a thematic reading unless it has phonological stress or it is supported by a context that favors a contrastive reading.6 Hoji points out that it is unlikely that PP topics are interpreted thematically; their default reading is contrastive. However, it is intricate to draw a line between the two interpretations of topic phrases. I will come back to this issue; see pp. 16–18. The distinction between thematic and contrastive topics is not only based on their meaning but also on various syntactic behaviors. Hoji (1985) argues that contrastive topics in sentence-initial position must involve movement on the basis of the following facts:7 (6)
a contrastive topics obey subjacency b they have properties of A’-movement c they do not allow resumptive pronouns
First, let us look at the relevant examples to show (6)a: (7)
a
[Taroo-ga ti nandomo itta Montreal-nii-wa Hanako-ga Montreal-to-TOP Hanako-NOM Taro-NOM many-times went to] omotteiru COMP think ‘To Montreal, Hanako thinks that Taro went many times’
b?*
Montreal-nii-wa Hanako-ga [[ tj ti nandomo Montreal-to-TOP Hanako-NOM many-times itta] hitoj]-ni atta went person-DAT met ‘*To Montreal, Hanako met the man who went many times’
The example in (7)a shows that the contrastive PP topic phrase allows a long-distance dependency. However, when the associated gap of the topic PP is inside the relative clause, as shown in (7)b, the sentence is ruled out. Assuming that subjacency is a constraint on movement, the
14 Cleft Constructions in Japanese Syntax
PP topic in (7)b is not base-generated in sentence-initial position, but rather is moved from the relative clause. The evidence for (6)b comes from the fact that contrastive topics observe reconstruction effects; it is widely assumed that the effects can be observed when A’-movement is involved. The empirical facts of (6)b are shown in (8) (the capital WA indicates phonological stress on the topic marker, which forces a contrastive reading): (8)
a
Johni-ga [sono zibuni-nituite no hon]-o suteta John-NOM that self-about-GEN book-ACC threw-away ‘John threw away the book about himself’
b [Sono zibuni-nituite no hon]-WAj Johni-ga tj suteta that self about-GEN book-TOP John-NOM threw-away ‘That book about himself, John threw’ (Hoji 1985: 150) The sentence in (8)a shows that the subject, John, binds the anaphor, zibun. When the object which contains the anaphor is preposed by (contrastive-) topicalization, the zibun can still refer to John, as shown in (8)b.8 Hoji argues that this fact shows that the topic phrase exhibits reconstruction effects; in other words, the sentence in (8)b involves movement. The following weak crossover (WCO) test also tells us that contrastive topics are derived by A’-movement: (9)
a ?*
[[ e proi butta] hito-ga darei-o uttaeta no? hit person-NOM who-ACC sued Q ‘Who did the person who hit (him) sue?’
b
[[ e proi ] butta hito-WA darei-ga uttaeta no? hit person-TOP who-NOM sued Q ‘(lit.) The person who hit (him), who sued?’ (Hoji 1985: 150)
According to Hoji, the sentence in (9)a is a typical WCO configuration. The schematic LF representation for this sentence is as follows: (10)
[CP dare-oi [IP [NP [. . . proi . . .]] ti V] Q]
As shown in (10), the Wh-operator in object position moves at LF crossing the subject NP containing an empty pronoun. The pronoun is coindexed
Topicalization and Cleft Constructions 15
with the Wh-phrase but is not in a c-commanding relation with the Wh-trace.9 An exact analysis for WCO effects is not our main concern here.10 Although we will consider more about WCO effects in cleft constructions in Chapter 4, it is sufficient to assume the following generalization at the moment: (11)
A variable cannot be the antecedent of a pronoun that it does not c-command. (Reinhart 1976)
The sentence in (9)b, on the other hand, does not exhibit a WCO effect. This indicates that the contrastive topic phrase is moved from the object position of the sentence; if it were base-generated sentence-initially, (9)b would be as bad as (9)a. The grammaticality of (9)b can be explained in several ways. One possible account is that since the preposed topic phrase can ‘reconstruct’ to its original position, the sentence in (9)b does not exhibit WCO effects. Under this line of analysis, the LF representation for the sentence in (9)b after reconstruction is illustrated as in (12): (12)
[CPdare-gai [IP ti [NP [ . . . proi . . .]] V] Q]
As far as I understand, Hoji (1985) does not take any particular analysis for reconstruction effects. The point here is to show that the contrastive topics are derived by movement. Finally, Hoji’s (1985) analysis predicts that they do not allow an overt resumptive pronoun (see (6)c). This is because the thematic position for the contrastive topics must be occupied by a trace and not by a pro. The following sentence shows that this is correct: (13)
a
Johni-WA [ Mary-ga [ Bill-ga (*? karei-o) butta to] John-TOP Mary-NOM Bill-NOM he-ACC hit COMP omotteita] was-thinking ‘John, Mary thought that Bill hit (him)’ (Hoji 1985: 183)
The topic, John, which has phonological stress, is associated with the object position of the embedded clause. Hoji assumes that this position is ‘deep enough’ for an overt pronoun; however, the overt pronoun, kare ‘he’, is not allowed in (13)a. Ishii (1991), however, proposes that overt
16 Cleft Constructions in Japanese Syntax
resumptive pronouns are always a ‘last resort’, which means that they can be used only when movement is blocked for subjacency (or some other reasons). If this is correct, the ungrammaticality of (13)a may come from the fact that the resumptive pronoun appears in a complement clause, violating the last resort condition. However, even if a contrastive topic is associated with a resumptive pronoun within a relative clause, the sentence is ruled out, as shown below: (13)
b?*
Montreali-ni-WA Taroo-ga [[ e sokoi-ni itta] hito]-to Montreal-to-TOP Taro-NOM there-to went person-with hanasita talked ‘To Montreal, Taro talked to a person who went there’
Therefore, it seems that what Hoji (1985) calls contrastive topics do not allow resumptive pronouns in any case. Thus, these facts suggest that the contrastive topic phrase is not base-generated sentence-initially (a position adjoined to IP) but is moved from its thematic position. To summarize so far, the facts above clearly show that contrastive topic constructions are more like scrambling or English topicalization in the sense that the associated gap is a trace. On the contrary, thematic topic constructions may correspond to ‘as for’ constructions in English, as pointed out by Hoji (1985: 136); these do not involve any movement.
Our assumptions about topicalization In this section, we will reexamine the following two issues implied by the previous research. The first is whether PP topics are always interpreted contrastively, and the second is whether the examples in (2) are real violations of island constraints. It will be argued that the thematic/ contrastive distinction does not directly correspond to different syntactic positions for topic phrases, and the topic constructions which appear to violate island constraints are nevertheless derivable by movement. Thematic/contrastive distinctions Although Hoji’s (1985) argument for the dichotomy for topic constructions is succinct, it is not easy to see such a semantic distinction between thematic and contrastive topics in an independent sentence without any context. Saito (1985: 348, footnote 20) points out that the topic wa has only one meaning and the thematic/contrastive distinction comes
Topicalization and Cleft Constructions 17
from pragmatics.11 A topic phrase bearing old information is interpreted as thematic, and one bearing new information is contrastive (Kuroda 1965). Arguably, the issues of old/new information are related to pragmatics rather than syntax. For example, let us consider the following sentence: (14)
New York-ni-wa moo itta no? New York-to-TOP already went Q ‘Have you already been to New York?’
The category of the topic phrase in (14) is a PP, which, according to Hoji, must be interpreted contrastively. However, as far as I can see, this topic phrase is ambiguous. If the sentence in (14) is uttered when the speaker knows that the person he/she is asking has had a plan to go to New York, then the PP topic is not necessarily interpreted as a contrastive one; the speaker of (14) does not contrast any other places with New York but simply asks whether or not he/she has been to New York yet.12 The relevant example is shown in (15): (15)
Iku tte itteita kedo, New York-ni-wa moo itta no? go COMP said but New York-to-TOP already went Q ‘(You) said that (you) would go (there), but have you already been to New York?’
Contrary to the context in (15), when the speaker is talking about various places and asks (14), the topic phrase should be interpreted contrastively: (16)
Kimi-wa iroiro-na tokoro-ni itteiru ga, New York-ni-wa you-TOP various place-DAT went but New York-to-TOP moo itta no? already went Q ‘You’ve been to many places, but have you already been to New York?’
In the context of (16), the topic phrase, New York-ni-wa ‘New YorkDAT-TOP’, is rather understood contrastively. Therefore, if the observation here is correct, it seems that PP topics can be ambiguous in their function. The exact analysis of the semantic/pragmatic distinction between thematic and contrastive topics is out of the scope here; however, the point I would like to make here is that wa has only one meaning and the functional
18 Cleft Constructions in Japanese Syntax
difference in question comes from pragmatics. This assumption implies that the thematic/contrastive distinction does not necessarily reflect syntactic positions and derivation for wa-phrases. In fact, Hoji (1990) claims that the real dichotomy of movement and base-generation comes from whether topic phrases contain a case/prepositional marker. I discuss this issue in ‘Properties of movement’, p. 24. The next subsection will show that topic constructions basically involve movement regardless of the functional difference. Apparent island constraint violation If Japanese has only one type of topic in syntax, how do we account for the facts that topic constructions sometimes involve movement but sometimes do not? I would like to propose that topic constructions basically involve movement to the spec of CP or some functional category for feature checking (Kubo 1992).13,14 An obvious problem for this analysis would be sentences like (2), where a topic construction violates island constraints, and yet, the sentences are grammatical. However, as will be shown below, this problem is only apparent. Sakai (1994) claims that topic phrases such as the one in (2)b, repeated here as (17)a, are not base-generated in a topic position; rather, the derivational source for (17)a is (17)b: (17)
a
Sono sinsii-wa [[[ ei ej kiteiru] yoofukuj]-ga yogoreteiru] that gentleman-TOP wearing suit-NOM dirty-be ‘As for that gentleman, the suit (he) is wearing is dirty’ (Kuno 1973: 249)
b Sono sinsi-gai [[ proi ej kiteiru] yoofukuj]-ga yogoreteiru that gentleman-NOM wearing suit-NOM dirty-be ‘It is the gentleman whose suit which he is wearing is dirty’15 The sentence in (17)b is a so-called multiple nominative construction: more precisely, a major subject construction. According to Sakai, the topic phrase, sono sinsi ‘that gentleman’, in (17)a is not derived from the relative clause nor is it base-generated in a topic position, but comes from the major subject position, as shown in (17)b. Therefore, no movement is involved from the island. Before going into the supportive evidence presented by Sakai (1994), let us briefly talk about what a major subject construction is. A major subject construction is one of the multiple nominative constructions in Japanese. It consists of more than one nominative phrase, at least some
Topicalization and Cleft Constructions 19
of which are not theta-marked by the predicate of the sentence or the following NP (Kuroda 1987, Tateishi 1994). The following are examples cited from Tateishi (1994: 21): (18)
a
Nihon-no tabemono-ga sakana-ga umai16 Japan-GEN food-NOM fish-NOM be-delicious ‘It is Japanese food among which fish are good’
b
Tookyoo-ga jiko-ga at-ta Tokyo-NOM accident-NOM there-is-PAST ‘It was in Tokyo where there was an accident’
The first nominative phrases, Nihon-no tabemono-ga ‘Japan-GEN foodNOM’ in (18)a and Tookyoo-ga ‘Tokyo-NOM’ in (18)b are considered major subjects. These major subjects are distinct from nominative phrases in other types of multiple nominative constructions; the first nominative phrase and the second nominative phrase cannot have a genitive relation, as shown in (19): (19)
a*
Nihon-no tabemono-no sakana-ga umai Japan-GEN food-GEN fish-NOM be-delicious ‘Fish among the food in Japan are good’
b*
Tookyoo-no jiko-ga at-ta Tokyo-GEN accident-NOM there-is-PAST ‘There was an accident in Tokyo’
The facts in (19) suggest that the first nominative phrase in (18) cannot be the argument of the second noun phrase, nor an argument of the predicate. On the contrary, another type of multiple nominative construction allows such relations: (20)
a Taroo-ga titioya-ga otooto-ga Taro-NOM father-NOM younger-brother-NOM nyuuin-si-ta be-hospitalized-PAST ‘It was Taro whose father’s younger brother was hospitalized’ b Kyooto-ga tatemono-ga iro-ga utukusii Kyoto-NOM building-NOM color-NOM be-beautiful ‘It is Kyoto where the colors of buildings are beautiful’
20 Cleft Constructions in Japanese Syntax
(21)
a
Taroo-no titioya-no otooto-ga Taro-GEN father-GEN younger-brother-NOM nyuuin-si-ta. be-hospitalized-PAST ‘The younger brother of Taro’s father was hospitalized’
b
Kyooto-no tatemono-no iro-ga utukusii Kyoto-GEN building-GEN color-NOM be-beautiful ‘The colors of buildings in Kyoto are beautiful’
The first and second nominative phrases in (20) are not major subjects, because they have a genitive relationship with the third nominative phrase. In other words, these nominative phrases are arguments of the third nominative phrase, and hence, we need syntactically to differentiate these nominative phrases from those in (18). In previous research, Kuroda (1987) argues that major subjects are base-generated IP internally, but move to the position adjoined to the IP. Tateishi (1994) proposes that no movement is involved in major subject constructions (see also Mihara 1992); major subjects are generated either in spec of IP, spec of CP or spec of M(odal)P. These different assumptions, however, do not affect our main point in this chapter, and we do not have to go into details of analyses for major subject constructions. We thus simply assume that a major subject is generated in the spec of IP or a position adjoined to IP. With this much in mind, let us briefly review the claim in Sakai (1994). Sakai provides the following pieces of evidence for his claim. First, notice that the major subject, sono sinsi-ga ‘that gentleman-NOM’, in (17)b is adjacent to the coindexed pro in the relative clause (RC). Thus, one might wonder how we could know whether the major subject is outside the island. Sakai shows that a sentential modifier, fusigina-kotoni ‘strangely’, can appear between the major subject and the RC, modifying the matrix clause but not the RC: (22)
Sono sinsi-ga fusigina-kotoni [[ e e kiteiru] yoofuku]-ga that gentleman-NOM strangely wearing suit-NOM yogoreteiru be-dirty ‘Strangely, it is that gentleman whose suit which (he) is wearing is dirty’ ‘*It is that gentleman whose suit (he) is strangely wearing is dirty’
Topicalization and Cleft Constructions 21
If we assume that sono sinsi-ga ‘that gentleman-NOM’ is inside the RC, the adverb cannot modify the matrix clause. Second, if the major subject is a quantifier, it can take scope over the entire clause, as shown in (23): (23)
Daremo-ga [[ e e kiteiru] yoofuku]-ga yogoreteiru everyone-NOM wearing suit-NOM be-dirty ‘For each person, the suit which he/she is wearing is dirty’
The point here is that the sentence in (23) can be interpreted as if the quantifier took scope over the whole sentence. Crucially, this reading is obtained only when daremo-ga ‘everyone-NOM’ is outside the RC. This is schematically illustrated in an LF representation in (24)a, and (24)b is a structure at LF where daremo-ga occurs as the subject inside the RC: (24) a
IP everyonei
b IP
ti
IP
RC IP
IP I’
RC
proi wearing suit
I’
VP
everyonei I
VP
I
IP
ti wearing
Assuming that the quantifier is generated as a major subject, it adjoins to the IP by quantifier raising, as illustrated in (24)a. Under this configuration, the quantifier takes scope over the whole sentence. On the other hand, if the quantifier originates in the subject position inside the relative clause, as shown in (24)b, it only raises to the local IP, and hence, the interpretation given in (23) is impossible. Third, the major subject can bind the anaphor, zibun ‘self’, in the matrix clause, as shown in (25): (25)
Sono sinsii-ga [[ e e kiteiru] yoofuku]-ga zibuni-no that gentleman-NOM wearing suit-NOM self-GEN ki-ni-iranai be-fond-of-NEG ‘The gentleman is not fond of the suit which he is wearing’
22 Cleft Constructions in Japanese Syntax
Since the sentence in (25) is grammatical, it must be the case that sono sinsi ‘that gentleman’ c-commands the object anaphor, zibun ‘self’. This is not expected if the major subject is inside the relative clause, because it would not bind the anaphor. Sakai (1994) does not discuss the case of (2)a. It seems that two of the tests given in the text are applicable to (2)a as shown in (2)’a and (2)’b: (2)’
a
Sono hito-ga fusiginakotoni [ e sinda noni] daremo that person-NOM strangely died although anyone kanasimanakatta saddened-not-was ‘Strangely, for that person, nobody was saddened although he died’
b Daremo-ga [e sinda noni] sono kazoku-ga everyone-NOM died although that family-NOM kanasi-manakatta saddened-not-was ‘For each person, his family was not saddened although he died’ However, the sentence in (2c)’ does not go through the test of zibunbinding: (2)’ c*?
Sono hitoi-ga [e byooki-na noni] zibuni-no that person-NOM be-sick although self-GEN kazoku-ga kanasimanakatta family-NOM saddened- not-was ‘For that person, self’s family was not saddened although he was sick’
In contrast to (25), the anaphor in (2)’c cannot be coindexed with the major subject. Notice also that sono ‘it, that’ in (2)’b is bound by the quantified subject, yielding a bound variable reading. An anonymous reviewer, however, pointed out that the marginality of the sentence in (2)’c does not come from a syntactic reason but comes from a conflict in the meaning. The main clause is a statement that involves the speaker’s empathy with the family rather than with that of the referent of the subject. On the other hand, the fact that zibun ‘self’ is used to refer to the referent of the subject shows that the speaker is empathizing with
Topicalization and Cleft Constructions 23
that person. Therefore, these two contradictory empathy foci yield the marginality of the sentence. As the reviewer presented to me, when the main clause also expresses that person’s point of view, acceptability results, as shown in (2)’’: (2)’’
Sono hito-ga [e kurusimi-de nakisakendeiru- noni] that person-NOM in-agony crying although zibun-no kurusimi-ga dare-ni-mo tutawar-anakatta self-GEN agony-NOM to-anyone reached-not ‘In spite of the fact that the man was crying in agony, self’s suffering was not understood by anyone’
Therefore, these examples suggest that there is another way of analyzing the fact that topicalization does not appear to be subject to island constraints. I assume that topic phrases which seem to violate an island constraint can be derived by movement.17 As a consequence of this analysis, we can also account for the ungrammaticality of ‘contrastive’ PP topic sentences such as (7)b, repeated here as (26): (26)?*
Montreal-nii-wa Hanako-ga [[ tj ti nandomo itta] Montreal-to-TOP Hanako-NOM many-times went hitoj]-ni atta person-DAT met ‘*To Montreal, Hanako met the man who went many times’
The topicalized phrase, Montreal-ni ‘Montreal-to’, cannot be base-generated as a major subject. As shown in (27), such a derivational source would be ungrammatical: (27)*
Montreal-nii-ga Hanako-ga [[ tj proi nandomo Montreal-to-NOM Hanako-NOM many-times atta. itta] hitoj]-ni went person-DAT met ‘*It was to Montreal where Hanako met the man who went many times’
The ni-phrase cannot appear when the same phrase is marked by nominative ga. Thus, (26) has only one possible derivational source; that is, the topic phrase, Montreal-ni ‘Montreal-to’, is generated within the relative clause, and hence, the sentence is ruled out by subjacency.
24 Cleft Constructions in Japanese Syntax
Parallelisms Let us now consider cleft constructions. This section compares cleft constructions with topic constructions and head-external relative clauses (HERCs). First, I will argue that the apparent subjacency violation cases of cleft constructions can be accounted for in the same way as that of topicalization. The point here is that both topic and cleft constructions involve movement of the same kind. Clefts and HERCs behave on a par with topic constructions with respect to various syntactic contexts. This is because, as already pointed out by Kuno (1973), HERCs are related to topicalization and we argue that clefts are also related to topicalization. The goal of this section is, therefore, to show that cleft constructions involve movement just like topicalization. Properties of movement Hoji (1990) draws essentially the same dichotomy between PP and NP focus phrases in a cleft construction as Hoji (1985) does between PP and NP topics. The NP focus can correspond to a thematic topic, and the PP focus to a contrastive topic; namely, the NP focus can be base-generated in its position whereas the PP focus, which includes an NP with a case marker, is derived by movement. As discussed on pages 16–18, the thematic/contrastive distinction is not directly mapped onto the syntactic structure; however, Hoji’s dichotomy between the PPs and the NPs is still sound. These two types of sentences are shown below: (28)
a [John-ga proi atta-no]-wa Russelli-da John-NOM met-NM-TOP Russell-COP ‘It was Russell that John met’ b [John-ga ti atta-no]-wa Russell-nii -da John-NOM met-NM-TOP Russell-DAT-COP ‘It was with Russell that John met’
The difference between (28)a and (28)b is that the focus phrase, Russell, does not have any case marker in the former whereas it has a dative marker in the latter. Hoji assumes that a case-marked NP corresponds to a PP, and claims that sentences such as (28)a can contain a null pronoun in the presuppositional clause, but when the focus phrase is PP, as shown in (28)b pro, is not possible and the gap must be occupied by a trace.18
Topicalization and Cleft Constructions 25
As evidence for the dichotomy between a PP focus and an NP focus, Hoji (1990) presents the following examples:19 (29)
[John-ga [sono hito-ga e atta to] John-NOM that person-NOM met COMP omotteiru -no]-wa Russell(-ni)-da. think -NM-TOP Russell(-DAT)-COP ‘It was (with) Russell that John thought that that person met’
(30)
a(??)
b*?
[John-ga[[ ei ej atta-koto-ga-aru] nihonzini]-o oozei John-NOM have-met Japanese-ACC many sitteiru-no]-wa Russell da. knows-NM-TOP Russell-COP ‘It is Russell that John knows many Japanese that have met’ [John-ga [[ ei ej atta-koto-ga-aru] nihonzini]-o John-NOM have-met Japanese-ACC oozei sitteiru-no]-wa Russell-ni da. many knows-NM-TOP Russell-to-COP ‘It is with Russell that John knows many Japanese that have met’
The sentence in (29) shows that, whether or not the focus phrase is marked by a case marker, a long-distance dependency is possible in cleft constructions. As for cleft sentences involving a relative clause, Hoji claims that since the focus phrase without any case marker in (30)a is not necessarily created by movement, the sentence is grammatical. On the other hand, since the sentence in (30)b is ruled out, the PP focus obeys subjacency, and hence, it must move out of the relative clause. Based on these facts, Hoji proposes the following structures for sentences like (30)a and (30)b respectively: (31)
a [NP [IP . . .(proi). . .] [NP noi]] -wa NPi da NM-TOP COP b [CP Opi [IP . . .ti. . .] no]-wa NPi-(CASE) da NM-TOP COP
In (31)a aboutness licensing is possible between the focus NP and the pro, but it is not in (31)b and syntactic movement is obligatory.20
26 Cleft Constructions in Japanese Syntax
Note that the same contrast is obtained for the corresponding topic sentences (between (30)a and (32)a for instance):
(32)
a(??)
Russellj-wa John-ga [[ ei ej atta-koto-ga-aru] Russell-TOP John-NOM have-met nihonzini]-o oozei sitteiru Japanese-ACC many knows ‘As for Russell, John knows many Japanese that have met’
b *?
John-ga [[ei ej atta-koto-ga-aru] Russell-nij-wa Russell-DAT-TOP John-NOM have-met nihonzini]-o oozei sitteiru Japanese-ACC many knows ‘With Russell, John knows many Japanese that have met’
Therefore, Hoji argues that NP and PP focus phrases are analyzed on a par with NP and PP topic phrases. As summarized above, Hoji (1990) proposes that cleft constructions like (30)a and topic constructions like (32)a do not involve any movement because they do not appear to obey subjacency. However, I would like to argue that those constructions are derived by movement based on Sakai’s (1994) analysis for topicalization. Although it is not evident at this point why movement is crucial, it will be in Chapter 4. Now how do we account for (32)a under the assumption that the sentence involves movement? Notice that if Sakai’s analysis for topicalization is adopted, the derivational source for the presuppositional clause is represented as (33)a and not as (33)b:
(33)
a? Russell-gaj John-ga [[ ei proj atta-koto-ga-aru] Russell-NOM John-NOM have-met sitteiru nihonzini]-o oozei Japanese-ACC many knows ‘For Russell John knows many Japanese that have met’ b
John-ga [[ ei Russell-ni atta-koto-ga-aru] John-NOM Russell-DAT have-met nihonzini]-o oozei sitteiru Japanese-ACC many knows ‘John knows many Japanese that have met Russell’
Topicalization and Cleft Constructions 27
In (33)a, the focus phrase, Russell, is the major underlying subject and is not inside the relative clause; rather, the major subject is associated with the pro. If we assume that the derivational source for (30)a is (33)a and not (33)b, we can account for the grammaticality of the cleft sentence which appears to violate subjacency.21 If this is on the right track, we can assume that a structure in (31)b repeated here as (34), is the only possibility for cleft constructions in the grammar: (34)
[CP Opi [IP . . .ti. . .] no]-wa XPi- da NM-TOP COP
In the structure in (34), a null operator is involved in the presuppositional clause, which is coindexed with a focus phrase by a predication relation. We will discuss how a null operator in the presuppositional clause is linked to a focus XP later in this chapter. We will also discuss the status of no ‘nominalizer’ and the phrasal category (CP) of the presuppositional clause in cleft constructions in Chapter 3, showing that the structure in (31)a is not considered the one for a cleft sentence in a strict sense.22,23 I will not go into the issue of island constraints for clefts any further in this chapter; however, the important point that I would like to make is that we have a way to explain sentences with the apparent subjacency violation for topicalization, which can also explain the apparent subjacency violation in cleft constructions. This leads us to say that topic and cleft constructions involve movement of a major subject, except for true topics. We will see the instance of true topics in the following subsection. Topicalization and clefting The parallelism between topicalization and clefting extends beyond island effects. In addition, it should be pointed out that a HERC is also related to topicalization in Japanese and other languages (Kuno 1973, Schachter 1973 among many others). This subsection will illustrate the parallelism between these three constructions. First, let us consider the following examples:24 (35)
a
Taroo-ni tuma-ga aru Taro-DAT wife-NOM have ‘Taro has a wife’
28 Cleft Constructions in Japanese Syntax
b?*
Topicalization Tumai-wa Taroo-ni ei aru wife-NOM Taro-DAT have ‘Speaking of wife, Taro has one’
c?*
Cleft [Taroo-ni ei aru-no]-wa tumai -da Taro-DAT have-NM-TOP wife-COP ‘It is a wife that Taro has’
d?*
HERC [Taroo-ni ei aru] tumai Taro-DAT have wife ‘a wife that Taro has’
The sentence in (35)a has the possessor marked by dative case (Taroo-ni ‘Taro-DAT’) and the possessee (tuma ‘wife’) is marked by nominative case. Similar pattern (subject-ni object-ga . . .) can be found in sentences with stative predicates indicating competence and likes/dislikes. As shown in (35)b, topicalization of the nominative phrase makes the sentence ungrammatical. The same is true for the corresponding cleft construction and HERC; (35)c and (35)d demonstrate that clefting and relativizing the nominative phrase is ruled out.25 These constructions behave similarly in multiple nominative sentences, as shown in (36):
(36)
a
Gengogaku-ga hakaseronbun-ga taihen-da linguistics-NOM dissertation-NOM hard-COP ‘(Writing) a dissertation in linguistics is hard’
b Topicalization Gengogaku-wa hakaseronbun-ga taihen-da linguistics-TOP dissertation-NOM hard-COP ‘Speaking of linguistics, (writing) a dissertation is hard’ c
Cleft [hakaseronbun-ga taihen-na -no]-wa gengogaku-da dissertation-NOM hard-COP-NM-TOP linguistics-COP ‘It is linguistics where (writing) a dissertation is hard’
Topicalization and Cleft Constructions 29
d HERC [hakaseronbun-ga taihen-na] gengogaku dissertation-NOM hard-COP linguistics ‘linguistics where (writing) a dissertation is hard’ The first nominative phrase, gengogaku-ga ‘linguistics-NOM’, in (36)a can be topicalized as well as clefted without changing the meaning of the sentence, as illustrated in (36)b and (36)c respectively. The same is true for the HERC, as shown in (36)d. On the other hand, if the second nominative phrase is either topicalized, clefted, or relativized, the sentences are degraded, as in (37): (37)
a??
Topicalization Hakaseronbun-wa gengogaku-ga taihen-da dissertation-TOP linguistics-NOM hard-COP ‘Speaking of (writing) a dissertation, linguistics is hard’
b??
Cleft [Gengogaku-ga taihen-na -no]-wa hakaseronbun-da linguistics-NOM hard-COP-NM-TOP dissertation-COP ‘It is (writing) a dissertation for which linguistics is hard’
c??
HERC [Gengogaku-ga taihen-na ] hakaseronbun linguistics-NOM hard-COP dissertation ‘a dissertation of which linguistics is hard’
Although it is not clear why the sentences in (37) are marginal, at least it is found that topicalization and clefting from multiple nominative constructions pattern in the same way. In other words, we can say that cleft constructions as well as HERCs share the same syntactic properties with topicalization. One might ask whether the ungrammaticality in (35)b–d can be accounted for by the syntax or not.26 For instance, Kuno (1973) gives a functional account for ungrammatical sentences with topicalization and relativization; since a relativized or topicalized element must be the one whose property is described by the rest of the sentence, the topic or relativized positions are unlikely to be occupied by something that is not typically described in the proposition. The same explanation holds for cleft constructions; although the clefted element in focus position is not semantically similar to a theme of the topic/relative constructions, what is
30 Cleft Constructions in Japanese Syntax
at issue here is the derivation of the presuppositional clause and not that of the focus element. In other words, a null operator in presuppositional and relative clauses contains an identical property of a topic phrase. A ‘gapless’ topic construction also manifests the same distribution:27 (38)
a
Tarooi-wa [PP zibuni-ga inai aida-ni] ie-ga Taro-TOP self-NOM be-NEG during house-NOM yaketesimat-ta burnt-down ‘As for Taro, his house burnt down when he was out’
b [[PP zibuni-ga inai aida-ni] ie-ga self-NOM be-NEG during house- NOM yaketesimat-ta-no]-wa Tarooi-da burnt-down-NM-TOP Taro COP ‘It is Taro whose house burnt down when he was out’ c
aida-ni] ie-ga yaketesimat-ta] [[PP zibuni-ga inai self-NOM be-NEG during house-NOM burnt-down hitoi person ‘the person whose house burnt down when he was out’
Sentence (38)a has a topic phrase, Taroo-wa ‘Taro-TOP’, in the initial position, and it binds the anaphor in the adjunct clause. The sentence can be considered ‘gapless’, because the topic phrase does not appear to be associated with any argument within the rest of the sentence. Assuming that the sentence (38)b and the HERC in (38)c are derived from (38)a, clefting or relativizing of the topic phrase is allowed. Note, however, that it is possible that the derivational sources for (36) and (38) are something like (39)a and (39)b respectively: (39)
a
Gengogaku-no hakaseronbun-ga taihen-da linguistics-GEN dissertation-NOM hard-COP ‘(lit.) Linguistics’ dissertation is hard’
[zibuni-ga inai aida-ni] b Tarroi-ga ie-ga Taro-NOM house-NOM self-NOM be-NEG during yaketesimatta burnt-down ‘It is Taro whose house burnt down when he was out’
Topicalization and Cleft Constructions 31
It is reasonable to think that the topic phrase or a null operator in (36) comes from the possessive position of the subject as in (39)a, and for those in (38), from the first nominative phrase in the multiple nominative sentence, as in (39)b. If this is the case, the examples in (36) and (38) are not true instances of gapless constructions but cleft constructions involve movement. Our assumption that cleft constructions must involve movement predicts that true gapless constructions cannot be a derivational source for cleft sentences. What we call the true gapless topicalized sentences are those in (3), which are repeated here as (40): (40)
a
Sakana-wa [tai-ga oisii] fish-TOP red snapper-NOM be-delicious ‘As for fish, red snappers are delicious’
b
Hana-wa [sakura-ga ii] flower-TOP cherry blossoms-NOM be-good ‘As for flowers, cherry blossoms are the best’
I assume, following Kuno (1973) and Tateishi (1994), that the examples in (40) are true instances of gapless construction, because they cannot have any other possible derivational sources: (41)
(42)
a*
Sakana-ga tai-ga oisii fish-NOM red-snapper-NOM be-delicious ‘For fish, red snappers are delicious’
b*
Hana-ga sakura-ga ii flower-NOM cherry-blossoms-NOM be-good ‘For flowers, cherry blossoms are the best’
a*
Tai-ga sakana-wa oisii red-snapper-NOM fish-TOP be-delicious ‘For fish, red snappers are delicious’
b*
Sakura-ga hana-wa ii cherry-blossoms-NOM flower-TOP be-good ‘For flowers, cherry blossoms are the best’
In (41), the initial phrase is marked by ga ‘NOM’ rather than wa ‘TOP’, making the sentences multiple nominative, but they turn out to be
32 Cleft Constructions in Japanese Syntax
ungrammatical. The sentences in (42) show that the topic phrase cannot be base-generated in a sentence-internal position. Therefore, it is unlikely to assume that the sentences in (40) are derived from (41) or (42). In addition, the topic phrases in (40) cannot have a possessive relation with the following NP since such sentences are ungrammatical, as shown in (43): (43)
a*
Sakana-no tai-ga oisii fish-GEN red snapper-NOM be-delicious ‘(lit.) Red snappers of fish are delicious’
b*
Hana-no sakura-ga ii flower-GEN cherry blossoms-NOM be-good ‘(lit.) Cherry blossoms of flowers are the best’
These facts show that the topic constructions in (40) cannot have any derivational sources which are distinct from (40). Kuroda (1987), on the other hand, argues that the unacceptability of (41) should be explained by semantics rather than by syntax. According to his analysis for major subjects, they are interpreted as focus so that the meaning of (41)a, for instance, would be ‘it is fish among which red snapper is the best’. This does not make sense since red snapper does not cross-classify with fish. As evidence for his claim, he points out that the sentence in (44)b is acceptable in contrast to (41)a: (44)
a Oranda-no sakana-wa nisin-ga yoi Holland-GEN fish-TOP herring-NOM be-good ‘As for fish in Holland, herring is the best’ b Oranda-no sakana-ga nisin-ga yoi Holland-GEN fish-NOM herring-NOM be-good ‘Among fish in Holland, herring is the best’
In fact, the corresponding cleft sentence to (44) turns out to be more acceptable than (44)a: (45) [Nisin-ga yoi no]-wa oranda-no sakana-da herring-NOM be-good-NM-TOP Holland-GEN fish-COP ‘It is fish in Holland among which herring is the best’
Topicalization and Cleft Constructions 33
Therefore, these facts seem to suggest that all sentence-initial topic phrases are derived by movement, and the unacceptability of (41) comes from a semantic anomaly. However, we should notice that the sentence in (44)a can be analyzed in the way that it is derived from either of the following derivational sources: (44)’
a?
Oranda-no sakana-no nisin-ga yoi Holland-GEN fish-GEN herring-NOM be-good ‘Herring among fish in Holland is the best’
b
Nisin-ga oranda-no sakana-wa yoi herring-NOM Holland-GEN fish-TOP be-good ‘Herring, among fish in Holland, is the best’
As discussed above, such paraphrased sentences are ruled out for (40). While this issue depends on how much the meaning may differ between the derivational source and the derived form in linguistic theories, I assume that the facts here show that the topic phrases in (40) are true instances of gapless topics.28 Interestingly, the topic phrases in (40) cannot be clefted nor relativized, as shown in (46): (46)
(47)
a*
[Tai-ga oisii no]-wa sakana da red snapper-NOM be-delicious-NM-TOP fish-COP ‘It is fish that red snappers are delicious’
b*
[Sakura-ga ii no]-wa hana da cherry blossoms-NOM be-good-NM-TOP flowers-COP ‘It is a flower that cherry blossoms are the best’
a*
[Tai-ga oisii] sakana red snapper-NOM be-delicious fish ‘the fish that red snappers are delicious’
b*
[Sakura-ga ii] hana cherry blossoms-NOM be-good flowers ‘the flowers that cherry blossoms are the best’
If cleft sentences were represented as (31)a, repeated as (48)a below, it would be difficult to explain why sentences such as (46) are ruled out, in spite of the fact that the topicalization in (40) is possible. If true
34 Cleft Constructions in Japanese Syntax
gapless topic constructions are licensed by the aboutness relation, the structure in (48)a, which is also licensed by the aboutness relation, should make (46) grammatical.29 Assuming that cleft constructions are uniformly represented as in (48)b, we can account for the ungrammaticality of (46): (48)
a [NP [IP . . .(proi). . .] [NP noi]]-wa NPi-da NM-TOP COP b [CP Opi [IP . . .ti. . .] no]-wa NM-TOP
XPi-da COP
While topicalization and clefting, as well as HERCs, show a number of parallelisms, there is a crucial difference between the two; topic constructions can be gapless whereas cleft constructions cannot.30 In other words, cleft constructions are a proper subset of those for topicalization. I assume that this is because cleft constructions always involve movement in order to create a predication configuration, and they are not licensed by the aboutness condition.31 Except for such a difference between topicalization in (40) and clefting in (46), we can basically conclude that when a phrase can be clefted, it can also be topicalized. It seems reasonable to assume that these constructions involve A’-movement; more specifically, I would like to claim, holding Hoji’s (1990) basic insight, that a null operator in the presuppositional clause has a syntactic property shared with a topic phrase in Japanese. Licensing of negative polarity items Based on the assumption that a null operator in the presuppositional clause in Japanese cleft constructions is analogous to a topic phrase, what can we predict about cleft constructions? Importantly, our analysis readily accounts for the licensing of NPIs in cleft constructions, which is briefly introduced in Chapter 1; NPIs cannot occupy a focus position in cleft constructions, in contrast with pseudoclefts in English. The relevant example is repeated here as in (49)c and the corresponding simple and topicalized sentences are given in (49)a and (49)b respectively:32 (49)
a
Taroo-ga dono eiga-mo mi-nak-atta Taro-NOM any movie-also see-NEG-PAST ‘Taro didn’t see any movie’
Topicalization and Cleft Constructions 35
b*
[Dono eiga(-mo)]i-wa, Taroo-ga ei mi-nak-atta movie(-also)-TOP Taro-NOM see-NEG-PAST any ‘*Any movie, Taro didn’t see’
c*
[Taroo-ga ei mi-nak-atta -no] -wa [dono eiga-mo]i Taro-NOM see-NEG-PAST-NM-TOP any movie-also dat-ta COP-PAST ‘What Taro didn’t see was any movie’
(49)a is a simple sentence containing the NPI, dono eiga-mo ‘any moviealso’, which is licensed by the negative head, nak- ‘NEG’. Again, there is a parallelism between topicalization and clefting; as shown in (49)b and (49)c above, topicalizing or clefting the NPI is not allowed. Given that the null operator in the presuppositional clause in (49)c is a topic operator which is similar to an overt topic phrase, the ungrammaticality of cleft sentences containing an NPI follows naturally. Since NPIs are not qualified to be a topic, as shown in (49)b, sentences such as (49)c are ruled out. I assume that null operators of cleft constructions have a feature which does not coincide with NPIs, just like a topic feature does not coincide with them. One might point out that the fact of NPI licensing in cleft constructions might show anti-connectivity (Mark Baker, personal communication). This means that the focus phrase or the null operator in the presuppositional clause does not reconstruct at all. If this is the case, then we could say that the NEG head in the presuppositional clause does not c-command the NPI, and hence, the sentence is ruled out. However, as will be shown shortly, this is not plausible. Notice also that this kind of ‘anti-connectivity’ is also observed in English topic constructions. While the anaphor, himself, in (50)a can be bound by the antecedent, John, at some level of representation, the NPI, anyone, cannot be bound by the NEG head: (50)
a b*
Himselfi, Johni likes. Anyone, I don’t like.
There are two possible ways to analyze the ungrammaticality of (50)b. One is that, unlike an anaphoric pronoun, the NPI does not reconstruct at all, so that it cannot be bound by the NEG head. The other is that the
36 Cleft Constructions in Japanese Syntax
NPI cannot be topicalized in English either. In order to see which analysis is correct, let us consider the following sentences: (51)
a b*
John does not believe that Mary likes anyone John does not believe that anyone, Mary likes
The sentence in (51)a is grammatical, which suggests that the NPI, anyone, can be licensed from outside its clause. If we take the first analysis, namely, the one which assumes no reconstruction for NPIs, the topicalized NPI in the embedded clause in (51)b should be licensed and we would expect grammaticality. However, since the sentence in (51)b is ruled out, it seems plausible to think that NPIs are not topicalized in general. This still leaves a question why English allows pseudoclefting NPIs (see Chapter 1), but as far as Japanese NPIs are concerned, we can account for the ungrammaticality in (49).33 As briefly mentioned above, it might be possible that the cleft sentence in (49)c is ruled out simply because of the lack of a c-command relation between the NPI in focus position and its NEG head within the presuppositional clause; even though a null operator in the presuppositional clause is licensed by the NEG, the NPI in focus position is not, and hence, the sentence is ungrammatical. However, I would like to argue that the ungrammaticality does not come from the lack of c-commanding relation between the NPI in focus position and its licensing head in the presuppositional clause, but rather, an operator cannot share its feature with an NPI. I claim that an agreement chain is created between a null operator and a focus phrase in cleft constructions, as illustrated in (52):
(52)
XPi (focus)
CPi Opi
C’ IP
Ci
...ti...
A null operator in the presuppositional CP is coindexed with the C head by spec-head agreement. By hypothesis, the maximal projection of the C has the same index, and the CP forms a predication relation with the XP in focus position (Williams 1980, Browning 1987 and others). We should admit to say that the predication relation between the CP and the XP in
Topicalization and Cleft Constructions 37
(52) is not like the one in (pseudo)cleft constructions in English. We will discuss this on pages 38–40. Under this approach, the operator in the spec of CP has agreement or feature sharing with the focus XP. There are several empirical facts which lead us to assume agreement or feature sharing between a null operator and a focus phrase. Let us first look at the sentences in (53): (53)
a [Tarooi-ga e semeta no] -wa zibuni da Taro-NOM accused-NM-TOP self-COP ‘It was himself that Taro accused’ b [Taroo to Hanakoi-ga e hihansita no]-wa Taro and Hanako-NOM criticized-NM-TOP otagaii-no ronbun da each-other-GEN paper-COP ‘It was each other’s paper that Taro and Hanako criticized’
As we already saw in Chapter 1, the anaphoric pronoun in focus position can be coindexed with the subject NP in the presuppositional clause in (53). Assuming that these anaphors must be bound by their antecedents under the standard binding theory, we should consider that a null operator or its trace is bound by an antecedent. We can explain the grammaticality in (53) under the mechanism such that a null operator is linked to a focus element by agreement, as shown in (52); namely, when a null operator satisfies the relevant binding condition, a focus phrase does not need to satisfy the same condition independently. Another possible analysis to account for the connectivity in (53) is that the anaphor is directly bound by its antecedent in the course of derivation. This is, however, implausible, as we will discuss in the next section, and our analysis will become more evident later in this chapter; it will be argued that a focus phrase occupies the predicate position underlyingly and is not created by movement. More on connectivity facts in cleft constructions will be examined in Chapter 4. Furthermore, another piece of evidence to show that there is agreement or feature sharing between a null operator and a focus phrase is the following condition C effects: (54)
a
atta Kinoo Tarooi-ga [[ e karei-o sitteiru] hito]-ni Yesterday Taro-NOM he-ACC know person-DAT met ‘Taro met the person who knows him yesterday’
38 Cleft Constructions in Japanese Syntax
b
(55)
[Kinoo Tarooi-ga e atta no]-wa [ e karei-o yesterday Taro-NOM met NM-TOP he-ACC sitteiru] hito(-ni)-da know person(-DAT)-COP ‘It was the person who knows him that Taro met yesterday’
a* Kinoo karei-ga [[ e Tarooi-o sitteiru] hito]-ni atta yesterday he-NOM Taro-ACC know person-DAT met ‘*He met the person who knows Taro’ b
[Kinoo karei-ga e atta no]-wa yesterday he-NOM met NM-TOP [ e Tarooi-o sitteiru]hito(-ni)-da Taro-ACC know person(-DAT)-COP ‘It was the person who knows Taro that he met’
In (54)a the pronoun, kare ‘he, him’, is bound by Taroo in the subject position. When the object relative clause containing the pronoun is clefted, kare still can be coindexed with Taroo. On the other hand, the basic sentence in (55)a is ruled out presumably by the condition C of binding theory; Taroo in the relative clause is bound by its coindexed element, kare. Interestingly, the same is true for the corresponding cleft sentence, as shown in (55)b. This shows that reconstruction of the null operator is obligatory; otherwise, the ungrammaticality is difficult to explain, because it might be the case that an accidental coindexation is possible between the pronoun and Taroo. Note also that the fact in (55)b implies that the coindexation between the pronoun and Taroo in (55)b may not be accidental, either. In other words, a copy or a trace of the null operator in its thematic position must be bound by its antecedent in presuppositional clauses. A note on predication relations It should be noted that the predication relation in Japanese clefts, as illustrated in (52), is different from the predication relation in English cleft constructions proposed by Williams (1980) and Browning (1987). There are two peculiar facts about the predication relation between a presuppositional clause and a focus phrase. One is that a presuppositional CP is not like a standard predicate: in particular, a predicate of a small clause. The other is that the predicative CP is not in a local relation with a focus XP. While these questions will be still open for
Topicalization and Cleft Constructions 39
future research, it is worth pointing out some empirical issues concerning the predication relation in Japanese clefts. I will then present some speculation at the end of this subsection. Based on the analysis by Williams (1980), Browning (1987) discusses syntactic predication in null operator constructions in English. She assumes that a CP can function as a ‘complex predicate’ (Williams 1980) when it contains a predicate variable. Roughly speaking, if the CP has the following configuration, it can be interpreted as a predicate: (56)
[CP Opi/WHi [IP . . .ti. . .]]
Moreover, the predicate has to satisfy a locality condition; the predicate and its subject must mutually c-command and no phrase can intervene between the two. While the presuppositional clause in a Japanese cleft conforms to (56), it does not seem to have a mutual c-command relation with its subject (a focus phrase) at least at S-structure. However, as argued by Heggie (1988) for English pseudoclefts, it might be possible that a presuppositional clause and a focus element establish a predication relation within a small clause at D-structure. Heggie proposes that the specificational pseudocleft in (57)a is derived from the structure in (57)b: (57)
a What Johni is is important to himselfi b [VP is [CP[AP PRO important to himself] [CP what John is]]]
The underlying structure in (57)b has a CP small clause which consists of the subject AP and the predicate CP. In this sense, the pseudocleft sentence is regarded as a predicate inversion construction, where the predicate of the small clause moves to a higher position (in Heggie, spec of CP) crossing the subject AP.34 Notice that although the predicate (what John is) and the subject (important to himself) are not in a local relation at S-structure, the predication relation is formed locally at D-structure, as shown in (57)b. Let us consider whether the same line of analysis can be applied to Japanese cleft constructions. In fact, it is difficult to find independent motivation to assume a small clause configuration with the presuppositional clause and the underlying focus phrase. The underlying structure would be something like (58)b under the small clause analysis. (58)a is a surface structure:
40 Cleft Constructions in Japanese Syntax
(58)
a [Taroo-ga e atta-no] -wa zibun-no hahaoya-ni-da Taro-NOM met-NM-TOP self’s mother-DAT-COP ‘It was self’s mother that Taro met’ b [VP [CP[NP zibun-no hahaoya-ni] [CP Taroo-ga e atta-no]]-da] self’s mother-DAT Taro-NOM met-NM-COP
As shown in (58)b the presuppositional CP is base-generated as a predicate of the small clause and forms a predication relation with the subject NP, zibun-no hahaoya ‘self’s mother’ at D-structure, just as in (57)b. Under this analysis, we must assume that the predicate of the small clause moves to a higher position crossing the subject; however, a small clause predicate does not generally undergo any movement in Japanese (Kikuchi and Takahashi 1991). This is shown in (59): (59)
a
Taroo-ga [AP [NP zibun-no kodomo-o] [AP hokori-ni]] omotteiru Taro-NOM self-GEN child-ACC be-proud think ‘Taro is proud of his child’
b*
Taroo-ga [AP hokori-ni]i [AP [NP zibun-no kodomo-o] ti ] Taro-NOM be-proud self-GEN child-ACC omotteiru think
c*
[AP hokori-ni]i Taroo-ga [AP [NP zibun-no kodomo-o] ti] be-proud Taro-NOM self-GEN child-ACC omotteiru think
The sentence in (59)a contains a small clause, in which the subject NP has the predicate AP. In (59)b the predicate of the small clause is scrambled to an intermediate position, and it is scrambled to sentence-initial position in (59)c. Both sentences are ruled out. This fact casts doubt on the small clause analysis for Japanese clefts. At this point, I put this problem aside and yet stipulate that a presuppositional CP is base-generated either in spec of IP or a higher topic position, and that a focus phrase is generated as a sister to the copular verb. Clefting a Wh-constituent The next question which might be raised here is how we can account for cleft sentences whose focus position is occupied by a Wh-phrase.
Topicalization and Cleft Constructions 41
Although topicalized Wh-phrases generally appear to yield ungrammaticality, clefting Wh-phrases is allowed, as in (60): (60)
a
Taroo-ga nani-o mita no Taro-NOM what-ACC saw Q ‘What did Taro see?’
b*
Nanii-wa Taroo-ga ei mita no what-TOP Taro-NOM saw Q ‘What did Taro see?’
c
[Taroo-ga ei mita-no] -wa nan(i)i-na -no Taro-NOM saw-NM-TOP what-COP-Q ‘What is it that Taro saw?’
If the sentence in (60)c is analogous to (60)b, we might expect (60)c to be ruled out since topicalization of the Wh-constituent is ungrammatical. Contrary to our prediction, the sentence in (60)c is perfect, so that the analysis provided for clefting NPIs might not be on the right track. Here, we need to differentiate the syntactic aspect from the semantic/ functional properties of topic phrases and null operators. In other words, in order to be a felicitous topic, semantic/functional consideration is required. As noted by Hoji (1985), Wh-constituents are easily topicalized when they are PPs, as shown in (61)b: (61)
a*
Dare-wa tegami-o kaita no who-TOP letter-ACC wrote Q ‘Who wrote the letter?’
b?
Dare-kara-wa tegami-ga kita no who-from-TOP letter-NOM came Q ‘FROM WHOM did the letter come?’
The sentences in (61) show that the NP Wh-topic, dare-wa ‘who-TOP’, is ruled out; on the contrary, the PP Wh-topic, dare-kara-wa ‘who-from-TOP’, turns out to be grammatical. Moreover, if the NP Wh-constituent in (61)a has a phonological stress on it (capitalized wa-phrase), the sentence sounds much better:
42 Cleft Constructions in Japanese Syntax
(62)
?
DARE-WA tegami-o kaita no WHO-TOP letter-ACC wrote Q ‘Which of those wrote the letter (and which didn’t)?’
The Wh-interrogative sentence in (62) is grammatical and becomes felicitous when there is a presupposition something like ‘there is a person who did not write a letter’ and the speaker asks who wrote the letter. Based on the reading in (62), we consider that a Wh-phrase is possible in topic positions when it is D-linked, in the sense of Pesetsky (1987). Indeed, inherently D-linked Wh-phrases, such as dono NP ‘which NP’, can be topicalized without problems, even though this is not a PP: (63)
Dono-hon-wai Taroo-ga ei yonda no which-book-TOP Taro-NOM read Q ‘As for which book, did Taro read?’
On the contrary, ittai ‘the hell’, known as a phrase for ‘aggressively non-D-linked’ Wh-phrases (Pesetsky 1987), cannot make the following Wh-phrase topicalized, with or without phonological stress, as shown below: (64)
?*
Ittai nani-wa/WAi John-ga ei yonda no the-hell what-TOP John-NOM read Q ‘What the hell did John read?’
In addition, as pointed out by Hidekazu Tanaka (personal communication), a Wh-phrase which cannot be D-linked is impossible as a topic phrase. Even if we place a heavy stress on such phrases, the grammaticality does not change (contrast (62)): (65)
a*
Naze-WA sono hon-o katta no why-TOP that book-ACC bought Q ‘Why did you buy the book?’
b*
Doo-WA sono kuruma-o naosita no how-TOP that car-ACC fixed Q ‘How did you fix the car?’
Naze ‘why’ and doo ‘how’ cannot be D-linked. Such Wh-phrases are not possible in topic positions, as shown in (65). The ungrammaticality in
Topicalization and Cleft Constructions 43
(65) is not due simply to topicalizing an adjunct Wh-phrase. Adjunct Wh-phrases can be topicalized as long as they are D-linked: (66)
a
Dono heya-de-wa Taroo-ga hon-o yonda no which room-in-TOP Taro-NOM book-ACC read Q ‘As for which room, did Taro read a book in it?’
b
Doko-de-WA, Taroo-ga hon-o yonda no where-in-TOP Taro-NOM book-ACC read Q ‘WHERE did Taro read a book?’
c*
Doko-de-wa Taroo-ga hon-o yonda no where-in-TOP Taro-NOM book-ACC read Q ‘Where did Taro read a book?’
The topic phrases in (66) are all adjunct Wh-constituents. However, the judgments are not the same. The sentence in (66)a contains a doko NP Wh-phrase, which is considered an inherently D-linked Wh-element. This sentence, as well as (66)b with phonological stress on the Wh-phrase, is grammatical because the topic phrase is D-linked. The topic Wh-phrase in (66)c, however, is not felicitous; it is difficult to interpret it as a D-linked Wh-phrase without any phonological stress. Therefore, the facts here suggest that adjunct Wh-phrases can be topicalized but non-D-linked Wh-phrases cannot. Notice, on the other hand, that a focus position in cleft constructions does not require a D-linked element in it, so that a wider range of Wh-constituents can appear there. Going back to the ungrammatical cases of clefting NPIs, we find that even though a Wh-constituent can become a topic phrase in case it is D-linked, NPIs are not available for topicalization in general. As the following sentences show, NP topics as well as PP topics are not possible when the topic phrase is an instance of NPI: (67)
a*
Dono tomodati-wa Taroo-ga aw-anak-atta any friends-TOP Taro-NOM meet-NEG-PAST ‘Any friends, Taro didn’t meet’
b*
Dono tomodati-kara-wa Taroo-ga tegami-o any friends-from-TOP Taro-NOM letter-ACC uketor-anak-atta receive-NEG-PAST ‘From any friends, Taro didn’t receive a letter’
44 Cleft Constructions in Japanese Syntax
In addition, even if we place phonological stress on the NPIs in (67), or the NPIs are topicalized in-situ, the sentences are ruled out, as in (68) and (69): (68)
(69)
a*
DONO-TOMODATI-WA Taroo-ga aw-anak-atta any friends-TOP Taro-NOM meet-NEG-PAST ‘ANY FRIENDS, Taro didn’t meet’
b*
DONO-TOMODATI-KARA-wa Taroo-ga tegami-o any friends-from-TOP Taro-NOM letter-ACC uketor-anak-atta receive-NEG-PAST ‘FROM ANY FRIENDS, Taro didn’t receive a letter’
a*
Taroo-ga dono tomodati-wa aw-anak-atta Taro-NOM any friends-TOP meet-NEG-PAST ‘Taro didn’t meet ANY friend’
b*
Taroo-ga dono tomodati-kara-wa tegami-o Taro-NOM any friend-from-TOP letter-ACC uketor-anak-atta receive-NEG-PAST ‘Taro didn’t receive a letter from ANY friend’
I propose that the difference between NPIs and Wh-phrases is that while Wh-phrases may accord with an operator feature in clefts and topicalization, NPIs cannot whatever the context would be. Therefore, the ungrammaticality of clefting NPIs comes from some kind of mismatch between some feature of an operator and the one of its associated NPI. If the assumption proposed here is correct, the data discussed in this subsection still confirm that cleft constructions in Japanese are analyzed on a par with topicalization. Other focus particles and some puzzles Some more parallelisms between topicalization and cleft constructions are observed with another type of NPI, sika ‘only’, and with the focus marker dake ‘only’. The data presented in this subsection give us an interesting puzzle. First, like the NPIs we have already seen, sika phrases cannot be topicalized, and as a result, clefting a sika-phrase is not allowed either, as shown in (70):
Topicalization and Cleft Constructions 45
(70)
a
Taroo-ga tomodati-ni-sika aw-ana-katta Taro-NOM friend-DAT-only meet-NEG-PAST ‘Taro met only his friends’
b*
Tomodati(-ni)-sika-wa Taroo-ga aw-ana-katta friend-DAT-only-TOP Taro-TOP meet-NEG-PAST ‘Only his friends, Taro met’
c*
[Taroo-ga e aw-ana-katta no] -wa tomodati (-ni)-sika Taro-NOM meet-NEG-PAST-NM-TOP friend-DAT-only da COP ‘It is only his friends that Taro met’
The sentence in (70)a is a simple sentence in which the sika-phrase appears in object position. Although this sentence is grammatical, the sika-phrase cannot be topicalized, as shown in (70)b, nor clefted, as in (70)c. It is assumed that the sentences in (70)b and (70)c are accounted for in the same line of analysis in the previous section: some feature mismatch between an NPI and an operator. Dake also means ‘only’; it is almost synonymous with sika(-nai) ‘only(-not)’. Interestingly, a phrase marked with dake can be topicalized, as in (71)b, and clefted, as in (71)c: (71)
a Taroo-ga tomodati-ni-dake at-ta Taro-NOM friend-DAT-only meet-PAST ‘Taro met only his friends’ b Tomodati-ni-dake-wa Taroo-ga at-ta friend-DAT-only-TOP Taro-NOM meet-PAST ‘Only his friends, Taro met’ c [Taroo-ga e at-ta no] -wa tomodati(-ni)-dake da Taro-NOM meet-PAST- NM-TOP friend-DAT-only COP ‘It is only his friends that Taro met’
The examples in (70) and (71) clearly show that the parallelism between topicalization and clefting holds for other types of focus phrases. Thus, the facts given above confirm that our generalization is on the right track. However, as pointed out by Jonathan Bobaljik (personal communication), it is not clear what makes those phrases correspond to an operator feature. We could say that unlike sika ‘only’ and mo ‘also’, dake ‘only’ can match with the operator feature; however, this is a mere speculation.
46 Cleft Constructions in Japanese Syntax
I cannot give any independent motivation for the substantive properties of a null operator at this point, but rather present it as a descriptive puzzle which I hope will be solved in future research. Sika- and dake-phrases in cleft constructions do not necessarily behave in a different way. The following sentences demonstrate that there can be no binding relation between zibun ‘self’ in focus position and its antecedent. The sentences in (70)a and (71)a are basic structures, in (70)b and (71)b topicalized sentences, and in (70)c and (71)c cleft constructions. (72)
a
Tarooi-sika zibuni-o seme-nak-atta Taro-only self-ACC accuse-NEG-PAST ‘Only Taro accused himself’
b*
Zibuni-wa Tarooi-sika seme-nak-atta self-TOP Taro-only accuse-NEG-PAST ‘Himself, only Taro accused’
c*
(73)
[ Tarooi-sika seme -nak -atta -no] -wa zibuni dat-ta Taro-only accuse-NEG-PAST-NM-TOP self COP-PAST ‘It was himself that only Taro accused’
a
Tarooi-dake zibuni-o seme-ta Taro-only self-ACC accuse-PAST ‘Only Taro accused himself’
b*
Zibuni-wa Tarooi-dake seme-ta self-TOP Taro-only accuse-PAST ‘Himself, only Taro accused’
c*
[Tarooi-dake seme -ta -no] -wa zibuni dat-ta Taro-only accuse-PAST-NM-TOP self COP-PAST ‘It was himself that only Taro accused’
Our generalization still holds for the sentences in (72) and (73); when you cannot topicalize an element, you cannot cleft it either. However, the facts above may tell us that sika- and dake-phrases do not contrast with each other in terms of the referential binding properties. I am not sure whether this is related to a null operator feature or not; however, at least we should notice that (72)c and (73)c are not expected under our current proposal, since as we saw in Chapter 1, local cleft constructions display
Topicalization and Cleft Constructions 47
connectivity, as shown in (74)c whose basic structure is (74)a. The corresponding topic construction is also grammatical, as in (74)b: (74)
a
Tarooi-ga zibuni-o semeta Taro-NOM self-ACC accused ‘Taro accused himself’
b Zibuni-WA Tarooi-ga semeta35 self-TOP Taro-NOM accused ‘Himself, Taro accused’ c
zibuni dat-ta [Tarooi-ga e semeta-no]-wa Taro-NOM accused-NM-TOP self COP-PAST ‘It was himself that Taro accused’
The sentence in (74)c in which the focus position is occupied by the same anaphor, zibun ‘self’, is grammatical under the reading of coindexing the anaphor with Taro. Considering this is a standard phenomenon in cleft constructions (Higgins (1979) among many others), we should expect (72)c and (73)c to be grammatical. Therefore, there remains a question of why anaphors in focus position cannot refer to sika and dake phrases in the presuppositional clauses.36
Scrambling and clefting So far, we have argued that cleft constructions in Japanese involve movement; specifically, a null operator in the presuppositional clause moves to the spec of CP. A null operator and a focus phrase form an agreement chain, and they do not have to satisfy syntactic principles and conditions independently. However, we have not discussed an alternative analysis; that is, the movement in cleft constructions is more like scrambling rather than operator movement. This section will ask whether the following schematized structure would be a possible analysis for cleft constructions in Japanese, as proposed by Hasegawa (1997) and Matsuda (1997):37 (75)
a
[TopP[CP Taroo-ga ti atta no]j-wa [IP Hanako(-ni)i [VP tj da]]] Taro-NOM met NM-TOP Hanako(-DAT) COP ‘It was Hanako that Taro met’
48 Cleft Constructions in Japanese Syntax
b
TopP CPj-top
IP
I’
Taro ti met NPi VP
Hanako tj
I
COP
As shown in (75)b the focus phrase, Hanako(-ni), is extracted from the presuppositional CP within the VP.38 Then the presuppositional CP moves to the topic position. Under this analysis, properties of movement that we already saw are subsumed into properties of scrambling and not into those of operator movement. In other words, the analysis in (75) need not assume a null operator in the presuppositional clause with an agreement chain, but a focus phrase is directly created by scrambling. The analysis in (75) is one of the non-operator movement analyses proposed in the literature. A similar analysis is also proposed by Hiraiwa and Ishihara (2002); however, they adopt Rizzi’s (1997) articulated CP structures and so the landing site of the focus phrase (Hanako-ni in (75)) is the spec of Foc(us) P. As for another type of non-operator movement analysis, Kuwabara (1996) and Koizumi (2000) argue that Japanese cleft constructions involve string-vacuous verb movement; namely, within the presuppositional clause, a focus element undergoes remnant movement like German verb second phenomenon, and the focus element gets deleted at PF. This analysis is criticized by Takano (2002), in which he shows various empirical problems such as Proper Binding Condition violation and the nature of adverbial adjunct in focus position and remnant movement. This section focuses on the analysis represented as (75), however. We argue that (75) is not plausible, and that the focus phrase is generated within a VP, as a sister to a copula underlyingly. As will be shown, our analysis is supported by the facts that cleft sentences do not behave like sentences with scrambling. Scrambling of a subject One immediate question to raise concerning the analysis in (75) is whether a subject can be clefted at all, because it has been assumed in
Topicalization and Cleft Constructions 49
the literature (Saito 1985) that ga-marked phrases cannot be scrambled in Japanese. The following are the relevant examples: (76)
a*
Eigo-gai [Taroo-ga ti hanaseru] English-NOM Taro-NOM can-speak ‘English, Taro can speak’
b*
Hanako-gai [Taroo-ga [ ti sono hito-ni atta to] Hanako-NOM Taro-NOM that person-DAT met COMP omotteiru] think ‘Hanako, Taro thinks that (she) met that person’
In (76), the second ga-marked phrase is scrambled over the matrix subject. Unlike the other elements, scrambling of the ga-marked phrases is not allowed. In order to account for the facts such as (76), Saito (1985) argues that unlike the accusative case, the nominative case is inherent in that it is not assigned by any element (e.g. INFL) in Japanese. More specifically, the nominative case is only licensed in a certain syntactic context (in Saito’s term, [NP, S]), and the overt case marker must appear because a nominative phrase in question does not receive abstract case. Since scrambling creates an adjoined position, the landing sites for scrambled elements are not case-assigning positions. One piece of independent evidence for this claim is that Japanese allows multiple nominative constructions, as we saw in the previous sections. The other case markers do not allow multiple occurrences. This implies that the nominative case has a distinct case system from others. In addition, accusative o can be omitted in colloquial speech whereas nominative ga cannot. In order to make sure if a deleted case marker is not a topic marker, each case marker should be attached to a Wh-phrase; as we already saw, non-D-linked Wh-phrases cannot be topicalized: (77)
a
Nani(-o) katta no what(-ACC) bought Q ‘What did you buy?’
b
Dare*(-ga) kita no who(-NOM) came Q ‘Who came?’
50 Cleft Constructions in Japanese Syntax
As shown in (77), the accusative marker can be dropped but not the nominative marker. This fact shows that unlike the accusative phrase, the nominative phrase must have an overt case marker since it is not assigned abstract case. With this basic account in mind, let us look at the following cleft sentence: (78)
a
eigo(- ga)i-da [Taroo-ga ei hanaseru-no]-wa Taro–NOM can-speak-NM- TOP English (-NOM)-COP ‘It is English that Taro can speak’
b [Taroo-ga [ ei sono hito-ni atta to] omotteiru-no]-wa Taro–NOM that person- DAT met COMP think- NM-TOP Hanako(-ga)i-da Hanako(-NOM)-COP ‘It is Hanako that Taro thinks met that person’ In the sentences above, what is clefted is the nominative-marked phrases. Unlike the ungrammatical examples in (71), the sentences in (73) are perfect. If a cleft sentence is derived as in (70), nominative-marked phrases should not be focus phrases in clefts. It is intriguing that Hasegawa (1997), who proposes a similar scrambling analysis to (75) for some of the Japanese clefts, considers the focus with ga (but crucially, not o) ungrammatical. For those speakers who share the same judgments with Hasegawa’s, they may have a structure like (75); since nominative-marked phrases cannot be scrambled, ga-marked focus phrases are not allowed either. However, it is generally assumed that both ga and o are allowed to appear in focus position; even though some native speakers of Japanese who find focus NPs with ga and o degraded do not see any striking difference between the two case markers. Therefore, if Saito’s (1985) analysis for scrambling a nominative phrase is correct, it becomes implausible to assume the scrambling analysis for cleft sentences. Ordering Restrictions on Focus Position Second, a focus position in cleft constructions restricts the order between a numeral quantifier (NQ) and its associated noun phrase. These two occurrences are free in other positions, as shown in (79); however, this is not the case in cleft constructions, as in (80):39 (79)
a
Taroo-ga hon-o takusan katta Taro-NOM book-ACC many bought ‘Taro bought many books’
Topicalization and Cleft Constructions 51
(80)
b
Taroo-ga takusan hon-o katta Taro-NOM many book-ACC bought
a
[Taroo- ga e katta-no]-wa hon-o takusan dat-ta Taro-NOM bought-NM-TOP book-ACC many COP-PAST ‘It was many books that Taro bought’
b*
[Taroo-ga e katta-no]- wa takusan hon(-o) dat-ta Taro-NOM bought-NM- TOP many book(-ACC) COP-PAST
The examples in (79) are simple sentences in which the object, hon-o ‘book-ACC’, is associated with the NQ, takusan ‘many’. The order between the two elements can be either way; hon-o takusan and takusan hon-o are both possible, as shown in (79). Notice that the latter order is not allowed when it occurs in the focus position of the cleft sentence. This is shown in (80)b. When the NP and its NQ appear in focus position, the order between the two is fixed, as in (80)a. On the contrary, such restrictions are not observed when the NP and its NQ are scrambled sentence–initially, as shown below:
(81)
a
Hon-oi takusanj Taroo-ga ei ej katta book-ACC many Taro-NOM bought ‘Many books, Taro bought’
b Takusanj hon-oi Taroo-ga ei ej katta book-ACC many Taro-NOM bought
In (81), the NP and the NQ appear in either order in sentence-initial position. This is not expected if we assume the structure in (75), in which a scrambling operation is involved. One might say that this kind of ordering restriction is applied at a late stage of derivation to make the scrambled phrases a legitimate predicate by some kind of reanalysis. However, our analysis does not need such extra mechanisms. The fact that the cleft sentences in (80) observe the restriction on the order between an NP and its associated NQ is naturally accounted for if we assume that the focus position is base-generated as a predicate. I will continue the issue of ordering restriction on pages 53–4. The point here is that the analysis in (75) cannot straightforwardly explain the facts given in (80).
52 Cleft Constructions in Japanese Syntax
Resumptive Pronouns Ishii (1991) claims that Japanese allows a resumptive pronoun as a last resort strategy to save violations of island conditions. This analysis predicts that both topicalization and scrambling allow a resumptive pronoun strategy; however, it is argued in Saito (1985) and Ishii (1991) that topicalization allows an overt resumptive pronoun whereas scrambling does not. The following examples are cited from Saito (1985: 312): (82)
a?
Sono boosii-wa John-ga [[ ej sorei-o kabutteita] that hat-TOP John-NOM it-ACC was-wearing hitoj]-o yoku sitteiru rasii person-ACC well know seem ‘Speaking of that hat, it seems that John knows the person who was wearing it very well’
b*
Sono boosii-o John-ga [[ ej sorei-o kabutteita] hitoj]-o that hat-ACC John-NOM it-ACC was-wearing person-ACC yoku sitteiru rasii well know seem ‘That hat, it seems that John knows the person who was wearing it very well’
The sentences in (82) contain a relative clause. Each sentence-initial phrase (sono boosi-wa ‘that hat-TOP’ and sono boosi-o ‘that hat-ACC’) is linked to the pronoun, sore ‘it’, within the relative clause. Interestingly, the topic sentence in (82)a is grammatical with the overt pronoun in the relative clause, but the sentence involving scrambling in (82)b is not. The ungrammaticality of (82)b is accounted for in the following manner. According to Saito (1985), a scrambled phrase, which is in a position adjoined to IP, must bind a variable unless it is licensed in some other way. Since topic phrases can be licensed by aboutness conditions, they are licensed independent of the resumptive pronoun. Notice that under our analysis for topic constructions, the topic phrase in (82)a is derived from a major subject position. The major subject can be coindexed with a pro, and hence, the grammaticality of (82)a is expected. In order for the scrambled phrase to receive an interpretation, on the other hand, it has to inherit a theta role from an element which it binds. As only traces can transmit their theta role to the scrambled
Topicalization and Cleft Constructions 53
phrases, a resumptive pronoun is excluded from occurring in sentences such as (82)b. Now, let us turn to cleft constructions with resumptive pronouns. If we assume cleft constructions are derived by scrambling, specifically like (75), sentences such as (83) should have the same ungrammaticality as that of (82)b. This is not borne out, as shown in (83): (83)
?
[ John-ga [[ ej sore-oi kabutta] hitoj]-o John-NOM it-ACC was-wearing person-ACC yoku sitteiru-no]-wa sono boosii-da well know NM-TOP that hat-COP ‘It is that hat that John knows the person who was wearing it very well’
The sentence in (83) is a cleft construction in which the object in the relative clause, sono boosi ‘that hat’, is clefted and the resumptive pronoun, sore-o ‘it-ACC’, fills in that position. (83) sounds much better than the scrambled sentence in (82)b. Again, assuming the scrambling analysis for cleft constructions as in (75), we would expect the sentence in (83) to be ruled out; however, the fact is that (82) is grammatical. Therefore, it seems that the scrambling analysis has difficulties dealing with cleft constructions. Overall, it seems better to explain these facts in terms of null operator movement in presuppositional clauses. A single constituent restriction and multiple foci We observed that focus elements in cleft constructions must have a restriction on the order of an NP and its associated NQ in (80) on pages 52–3. This fact will lead us to think that the focus position is occupied only by a single constituent. It has been claimed that the NP-NQ order forms a single constituent but the other order does not (Kitahara 1993, Koizumi 1995). This is supported by the following coordination test: (84)
a
[ Hon-o takusan] to [ zassi-o sukosi] katta book-ACC many and magazine-ACC a few bought ‘(I) bought many books and a few magazines’
b*
[ Takusan hon(-o)] to [ sukosi zassi(-o)] katta many book(-ACC) and a few magazine(-ACC) bought
54 Cleft Constructions in Japanese Syntax
(84)a is grammatical, showing that the NP-NQ sequence can be coordinated with another NP-NQ. However, the NQ-NP phrases cannot be conjoined by to ‘and’, as shown in (84)b. This supports the claim that the NP-NQ forms a constituent but the NQ-NP does not. In addition, the judgments for multiple foci in Japanese clefts vary in the literature. For example, Koizumi (1995), Kuwabara (1995, 1996), and Hasegawa (1997) allow multiple foci in cleft constructions (with some restrictions; see Takano (2002: footnote 1)), but Inoue (1976: 101) and Noda (1996: 66–67) find these cases ungrammatical. I agree with the latter judgments and deal with the latter type of grammar in this book: (85)
*
[ Taroo-ga ei ej okutta-no]-wa [Hanako-nii hon(-o)j]-da Taro-NOM sent-NM-TOP Hanako-DAT book(ACC)-COP ‘(lit.) It was [to Hanako, a book] that Taro sent’
The cleft sentence in (85) contains more than one phrase in focus position. Unlike clefting, multiple scrambling is allowed, as shown in (86): (86)
kinoo Taroo-ga ti tj okutta Hanako-nii hon-oj Hanako-DAT book-ACC yesterday Taro-NOM sent ‘To Hanako, a book, Taro sent (it) (to her) yesterday’
The scrambled phrases in (86) are the same as the focus phrases in (85). Again, since multiple scrambling is allowed, as shown in (86), the analysis in (75) requires some more mechanisms to prohibit the multiple foci, such as (85). Based on these facts, we propose that focus phrases in clefts are not derived by scrambling but are base-generated in a predicate position which is occupied by a single constituent. Let us briefly discuss the structures for focus positions for the rest of this section. I assume that a copula da is a contracted form of de aru (Nishiyama 1998: 66, and references cited therein), where de is a postposition and aru is a main verb. For instance, da can be represented as de aru wherever the copula occurs, as shown below:40 (87)
a
Taroo-wa gakusei-da/-de aru Taro-TOP student-COP ‘Taro is a student’
b
Taroo-wa ganko-da/-de aru Taro-TOP stubborn-COP
Topicalization and Cleft Constructions 55
‘Taro is stubborn’ c
Taroo-ga hon-o yonda-no da/de aru Taro-NOM book-ACC read-NM-COP ‘It is that Taro read the book’
(87)a has a nominal predicate and (87)b a nominal adjective, followed by a copula. The sentence in (87)c is clefted as a whole, and the copula, da, is predicated of the sentence. In each of the examples in (87), the copula, da, can be replaced by de aru. The same is true for the standard cleft constructions, such as (88): (88)
[ Taroo-ga ei atta-no]-wa sono hito-nii-da/-de aru Taro–NOM met-NM-TOP that person-DAT-COP ‘It was that person that Taro met’
On the basis of the facts above, we assume the following schematized structure for the predicate in cleft constructions: (89)
VP
V’
XP
PP
V
P’
aru P de
The structure in (89) shows that the copula, da, is in fact a complex verb, which consists of the P, de, and the V, aru. De is considered a postposition, which is morphologically equivalent to the postposition of a location, a reason, or a purpose.41 I assume that the P incorporates into the V at some point of derivation, or it fuses phonologically under linear adjacency. According to this structure, a focus element appears as a complement of the P. This explains why the focus position must be occupied by a single constituent, because it is not likely that Ps can take more than one complement.
56 Cleft Constructions in Japanese Syntax
One might think that it is possible that the spec of P is filled in, so that more than one element can show up in a focus position in cleft constructions. Although it is controversial whether or not the spec position can be filled in the PP (see Stowell’s (1983) small clause analysis, and Hoekstra and Mulder (1990)), the facts are convincing that at least this kind of P behaves like an unaccusative predicate which projects only an internal argument (see Baker 1996). Assuming that this is correct, it is unlikely that the focus position in (89) is occupied by more than one element.
Summary We observed a number of parallel syntactic behaviors between topic and cleft constructions. Based on these facts, we claimed that Japanese cleft constructions involve some kind of covert topicalization; more precisely, a null topic operator plays a role in the presuppositional clauses of cleft sentences. Moreover, it was argued that this type of movement contrasts with an operation of scrambling, and that focus phrases in cleft constructions are generated in a predicate position underlyingly. We will further investigate properties of movement of a null operator in presuppositional clauses in Chapter 4.
3 Nominalizations in Cleft Constructions
Introduction In Chapter 2, we observed that cleft constructions are closely related to topicalization and do not involve a scrambling operation, and that the focus position in clefts must be occupied by a single constituent. While we briefly examined the internal structure of the predicate of clefts to explain this second property, we have not looked closely at the internal structure of the nominalized clauses (or presuppositional clauses) of the constructions. This will be the major concern of this chapter. The clause in question is the italicized portion of the following cleft sentence: (1) [Opi[Hanako-ga ei hana-o sensei-nii-dat-ta katta]-no]-wa Hanako-NOM flower-ACC bought-NM-TOP teacher-DAT-COP-PAST ‘It was the teacher that Hanako bought flowers for’ In Chapter 2, I assumed without argument that the presuppositional clause in (1) projects a CP, taking for granted the standard view for English cleft constructions and part of the analysis in Hoji (1990) for Japanese clefts. However, it is well known that the nominalizer no in (1) has multiple functions in Japanese, including a genitive marker, a pronominal form, and a complementizer. Therefore, it is worthwhile to investigate whether the no in the presuppositional clause of clefts is indeed a complementizer. Chapter 3 will provide various pieces of evidence that the structure of presuppositional clauses in clefts is indeed CP, which, in fact, parallels the structure of head-internal relative clauses (HIRCs) in Japanese. There is a brief review of previous research on no. I will claim that no in the presuppositional 57
58 Cleft Constructions in Japanese Syntax
clauses is not a prenominal marker and it will be pointed out that the no in cleft constructions is not a pro-form either, comparing several facts in relative clauses. Although the facts of case conversion might be used to argue that the presuppositional clauses can project a nominal category, I will argue that this is not the case, based on Watanabe’s (1996) analysis of case conversion in Japanese; the no-marked clauses in clefts are CPs, and not NPs or DPs. The claim here is also supported by extraction facts in HIRCs and cleft constructions. The final section is a summary of Chapter 3.
Properties of no This section briefly gives an overview of previous research on the marker, no, which appears as a nominalizer of the presuppositional clauses in cleft constructions. Following the proposals of Murasugi (1991) and others, the uses of no can be divided into three main types: a prenominal marker, a pronoun, and a complementizer (Harada (1971), Kamio (1983), Kitagawa and Ross (1982), Hoshi (1995), and references cited therein). The purpose of this section is to review the basic properties of no, which will help us decide which no is present in the presuppositional clauses of clefts. Prenominal marker no First, the phrases given in (2) contain a prenominal no, some instances of which correspond to a genitive marker in English. Although it is not clear whether the no in (2) is equivalent to English genitives, I translate a prenominal no into GEN(itive) in the gloss: (2) a Hanako-no hon Hanako-GEN book ‘Hanako’s book’ b Misima-no zisatu Misima-GEN suicide ‘Misima’s suicide’ c gengogaku-no kenkyuu linguistics-GEN research ‘research on linguistics’ d Canada-no ki Canada-GEN tree ‘trees in Canada’ e ki-no tukue wood-GEN desk ‘a wooden desk’
Nominalizations in Cleft Constructions 59
The NPs followed by a prenominal no are functionally interpreted as a possessor, subject, object, location, or type of the head noun, as shown in (2)a–e. Note also that the prenominal phrase in (2)a can be interpreted in a number of different ways, such as ‘an author of the book’ ‘the content of the book’, and so on as the English translation implies. There is another type of no in prenominal position, as shown in (2)’a: (2)’ a gakusee-no Taroo student-COP Taro ‘Taro as a student’ As the gloss shows, it is assumed that the no in this type of example can be interpreted as a copula. One of the reasons is that it is paraphrased by de aru (non-contracted form of a copula verb: see pages 54–5 in Chapter 2) without changing the meaning of the phrase, and it can also be inflected by the past tense form (dat-ta): (2)’ b gakusee de-aru Taroo student COP Taro ‘Taro as a student’ c gakusee dat-ta Taroo student COP-PAST Taro ‘Taro who was a student’ If these three phrases in (2)’ are analyzed in the same syntactic structure, the no in (2)’a is different from the no in presuppositional clauses; obviously, the no in presuppositional clauses of cleft constructions does not inflect like (2)’a and (2)’b. Therefore, it can be concluded that the no in presuppositional clauses does not belong to this type, and hence, I will exclude this possibility in this section. Prenominal no can also attach to PPs and CPs, as exemplified in (3): (3) a [PP gakkai -de]-no happyoo conference-at -GEN presentation ‘presentation at a conference’ b daigaku -kara-no [CP pro syurui-o uketotta -to] -no university-from-GEN document-ACC received-COMP-GEN hookoku report ‘a report from the university telling that (they) received the document’
60 Cleft Constructions in Japanese Syntax
In (3)a, no appears after the PP, gakkai de ‘at the conference’, and modifies the following NP, happyoo ‘presentation’. No also occurs after the CP in (3)b, modifying the NP, hookoku ‘report’. Pronominal no Second, no appears as a pronominal form, translated as ‘one’ in English. The examples are listed in (4): (4) a
akai-no red-Pro ‘the red one’
b
Canada-kara-no Canada-from-Pro ‘the one from Canada’
c
hattearu-no posted-Pro ‘the one which is posted’
d
Hanako-no Hanako-Pro ‘the one of Hanako’s’
e*
No-o mita Pro-ACC saw ‘(I) saw one’
(4)a–d show that a pronominal no can attach to APs, PPs, VPs (or IPs), and NPs respectively. In (4)e, unlike the English ‘one’ pro-form, Japanese no cannot stand alone; it is a bound morpheme and has to be preceded by another category. It should be noted that the nos in (4)b and (4)d are ambiguous between a pro-form and a genitive marker. When the nos in these examples are interpreted as a genitive marker in a certain linguistic context, we consider that the phrases involve NP- or N’-deletion. Adopting Saito and Murasugi’s (1990) NP-deletion analysis, such cases are roughly illustrated as follows: (5) a [DP Canada-kara-no [NP tegami]] Canada-from-GEN letter ‘a letter from Canada’ b [DP Canada-kara-no [NP e]] Canada-from-GEN
Nominalizations in Cleft Constructions 61
(6) a [DP Hanako-no [NP ronbun]] Hanako-GEN paper ‘Hanako’s paper’ b [DP Hanako-no [NP e]] Hanako-GEN The nos in (5)a and (6)a are genitive markers, and the NPs inside the phrases can be elided under identity with their linguistic antecedent, as shown in (5)b and (6)b. If the nos in (4)b and (4)d are interpreted as pronominal, the structures are the following: (7) a [DP/NP [PP Canada-kara] [N no]] Canada-from -Pro b [DP/NP [NP Hanako] [N no]] Hanako -Pro The no in (7) is the head of the entire phrase, which is modified by the PP or the NP. On the other hand, it seems that the nos in (4)a and (4)c are not genitive elements since APs and VPs do not need a genitive marker to modify the following noun phrase: (8) a * akai-no kuruma red-GEN car ‘a red car’ b*
hattearu-no syasin posted-GEN photo ‘the photo which is posted’
Therefore, the nos in (4)a and (4)c are unambiguously pro-form nos in any context.1 The following are other instances of the pronominal usage of no, where it occurs as ‘head’ of the headless relative clause (H-lessRC) in (9)a and of the complement clause of the perception verb in (9)b: (9) a [Taroo-ga ei katta] noi-o tabeta Taro-NOM bought Pro-ACC ate ‘(I) ate the one Taro bought’
62 Cleft Constructions in Japanese Syntax
b [ei tondeiru] -noi-o mita be-flying -Pro-ACC saw ‘(I) saw the one that is flying’ As argued in Kuroda (1976–77) and Hoshi (1995), this type of no can be replaced by a non-pronominal form, as shown below: (10) a [Taroo-ga ei katta] monoi-o tabeta Taro-NOM bought thing-ACC ate ‘(I) ate the thing that Taro bought’ b [ ei tondeiru] monoi-o mita be-flying thing-ACC saw ‘(I) saw the thing that is flying’ The difference between (9) and (10) is that the head is occupied by a pronominal no in the former but by a non-pro-form, mono ‘thing’ in the latter, and the meaning remains the same. It should be noted, however, that pronominal nos cannot refer to human beings or abstract entities (Kamio 1983), as shown in (11) and (12), although the example in (11)b is fine with a derogatory connotation: (11) a
[Taroo-ga ei suki-datta] kasyui Taro-NOM liked singer ‘the singer that Taro liked’
b # [Taroo-ga ei suki-datta]-noi Taro-NOM liked -pro ‘the one that Taro liked’ (12) a
b*
[kenkyuu-ni taisuru] jyoonetu research-DAT facing passion ‘a passion toward research’ [kenkyuu-ni taisuru]-no research-DAT facing -pro ‘the one toward research’
In (11)b, the head of the relative clause, kasyu ‘singer’ in (11)a, is replaced by no, but this becomes unacceptable. The replacement is not allowed in (12) either; the abstract noun, jyoonetu ‘passion’, cannot be pronominalized by no. We will come back to this issue to compare the pronominal nos and nos in clefts on pages 66–72.
Nominalizations in Cleft Constructions 63
Complementizer no The third type of no is the complementizer no. One of the relevant examples is illustrated in (13):2 (13) [pro teeburu-no ue-ni keekii-o oiteita]-noi table-GEN top-on cake-ACC left-NM ‘the one that (I) left [the cake] on the table’ (13) is a case of a head-internal relative clause (HIRC), where the head of the relative clause is not realized in a canonical position outside the relative clause but appears inside the clause. In (13), for example, although the entire phrase refers to keeki ‘cake’, the ‘head’ occurs in its theta position within the relative clause rather than the head position. While the exact structure of a HIRC has been controversial in the literature (Kuroda 1975–6, Itô 1986, Cole 1987, Williamson 1987, Ishii 1988, Culy 1990, Watanabe 1991, Bonneau 1992, Ohara 1994, Hoshi 1995, and many others), I will show that no in HIRCs, as well as no in cleft constructions, is not a prenominal marker or a proform on pages 63–72.
No in presuppositional clauses versus prenominal no Since, as we saw in (3)b no can be attached to some sentential element in prenominal position, one might wonder whether this type of no is the same as the no in presuppositional clauses in cleft constructions. This section, however, argues that the no in presuppositional clauses is not a genitive marker. If we suppose that no in presuppositional clauses is a prenominal marker, we must assume that the nominal head is always phonologically unrealized in the presuppositional clause because of some process like NP (N’)– deletion. The structure under this analysis is illustrated in (14): (14) [[Hanako-ga ei atta-no] Ø ]-wa Taroo-nii dat-ta Hanako-NOM met-NM -TOP Taro-DAT COP-PAST ‘It was Taro who Hanako met’ However, as discussed in Hoshi (1995) among others for relative clauses, it is implausible to assume an NP (N’)-deletion construction for the presuppositional clause, because the elided noun is never phonologically realized in Japanese:
64 Cleft Constructions in Japanese Syntax
(15) * [[Hanako-ga ei atta-no] hito ]-wa Taroo(-ni)i dat-ta Hanako-NOM met-GEN person-TOP Taro(-DAT) COP-PAST ‘The person who Hanako met was Taro’
The sentence in (15), where the head is occupied by hito ‘person’, is semantically felicitous but syntactically impossible. As we already observed on pages 60–62 (see (8)), categories other than NPs, PPs and CPs cannot be marked by no to modify the following noun phrase. Just like these examples, relative clauses such as (15) should not be marked by no to modify the head noun in the language. In other words, we may say that no and hito occupy the same syntactic position in such a configuration, and hence, (15) is ruled out.3 Thus, since the genitive no must appear before its modifying noun phrase (e.g., Hanako-no mono ‘Hanako’s thing’), the facts above show that the no in the presuppositional clause does not behave like the no as a prenominal marker. In fact, NP (N’)-deletion is not allowed when the prenominal no is attached to a CP in general:
(16) a* [DP/NP [CP syorui-o uketotta to]-no hookoku]-wa document-ACC received COMP-GEN report-TOP atta ga, [DP/NP [CP gookaku-sita to] -no ø]-wa mada nai existed but passed COMP-GEN-TOP be-not-yet ‘(I) had the report that (they) received the documents, but have not had (the one) that (I) passed yet’ b* [DP/NP[CPitu harau-ka]-no sirase]-wa atta ga, when pay-Q -GEN notice-TOP arrived but harau-ka]-no ø]-wa mada nai [DP/NP[CPikura how-much pay -Q-GEN -TOP be-not-yet ‘(I) received a notice about when (I) should pay, but have not had (the one) about how much (I) should pay’
The second sentences in (16) contain the deleted head noun preceded by the CP, and they are ruled out. One might think that the nos in (16)a and (16)b are proforms rather than genitive markers and the sentences do not involve any NP (N’)-deletion. If this would be the case, we can explain the ungrammaticality due to the fact that a pro-form no cannot be replaced by an abstract entity, such as hookoku ‘report’ and sirase
Nominalizations in Cleft Constructions 65
‘report, notice’. However, this line of analysis does not account for general facts of NP(N’)-deletion. The following sentence becomes grammatical even though the noun phrase in question is an abstract entity, as argued in Saito and Murasugi (1990):
(16) c [DP Gakubusei-no [NP sensei-e-no izon]]-wa yuruseru undergrad-GEN teacher-on-GEN reliance-TOP can-tolerate ga, [DP insei -no [NPø]]-wa yurusenai but grad-student-GEN -TOP cannot-tolerate ‘(I) can tolerate the undergraduates’ reliance on the faculty, but not the graduate students’’
In (16)c, we cannot assume that the no in the second DP is a pro-form, because it clearly refers to the noun, izon ‘reliance’, which is an abstract entity. Saito and Murasugi (1990) point out that this is a true instance of NP-deletion in Japanese, and if this is correct, the ungrammaticality in (16)a and (16)b comes from the impossibility of NP-deletion and not the impossibility of proform no.4 Furthermore, there is another indirect argument against analyzing presuppositional no as a genitive marker; some dialects of Japanese morphologically distinguish a prenominal (or genitive) marker from other types of no. Murasugi (1991) points out that in the Toyama dialect, the genitive case marker is realized as no; however, the pro-forms and complementizers are realized as ga. Crucially, presuppositional clauses in cleft constructions are marked by ga in this dialect, and not by no. The following examples in the Toyama dialect are cited from Murasugi (1991: 72, 95):
(17) a John-no-hon John-GEN-book ‘John’s book’ b siroi-ga white-Pro ‘the white one’ c hasitteiru-ga running-Pro ‘the one who is running’
66 Cleft Constructions in Japanese Syntax
d [Yamada-ga e atta ga]-wa Russell-da Yamada-NOM met NM-TOP Russell-COP ‘It was Russell that Yamada met’ As shown in (17)a, the prenominal marker is no as in other dialects; however, the proforms are realized as ga, as shown in (17)b and (17)c. Since the nominalizer in the cleft sentence is also ga in this dialect, it is suggested that no in presuppositional clauses is not the same element as the genitive marker. On the basis of the facts observed in this section, we conclude that the no in presuppositional clauses is not a genitive marker. This still leaves a possibility that it is a pro-form; however, we will argue that the no in presuppositional clauses is not a pro-form either.
Relative clauses and No-clauses of clefts In the last section, we saw that no in cleft constructions is not a genitive marker followed by an elided NP. It will be claimed in this section that the no in presuppositional clauses in clefts is distinct from a pro-form. We further investigate the syntactic status of no in presuppositional clauses by comparing them with sentential modifiers, such as Headless Relative Clauses (H-less RCs), Head-External Relative Clauses (HERCs) and Head-Internal Relative Clauses (HIRCs) in Japanese. The following discussion shows that a distinction should be drawn between H-less/ HERCs on the one hand and HIRCs and presuppositional no-clauses on the other. This distinction is revealed by certain inherent semantic properties, adjective- and numeral-quantifier (NQ)-modification, and case conversion. Assuming that the no in H-less RCs is a pro-form, we conclude that nos in presuppositional clauses, as well as those in HIRCs, are not pronominal. The facts in this section, therefore, will lead us to the idea that the HIRCs and no-clauses in clefts are not projections of a nominal category. It should be noticed that there are some cases where the no in ‘cleft’ sentences is interpreted ambiguously, either as a complementizer, as we will discuss shortly, or as a pro-form. The sentence in (18)a is one of such cases: (18) a [Hanako-ga e katta no] -wa sono hon-da Hanako-NOM bought-NM/Pro-TOP that book-COP ‘What Hanako bought was the book’ ‘The one that Hanako bought was the book’
Nominalizations in Cleft Constructions 67
b [Hanako-ga e katta mono]-wa sono hon-da Hanako-NOM bought thing-TOP that book-COP ‘The one that Hanako bought was the book’ The no in (18)a can be interpreted as a pro-form, because what it indicates is an inanimate, concrete entity, which is qualified as pro-form no. As shown in (18)b the no in this sentence can be paraphrased as mono ‘thing’. Thus, under this interpretation, the sentence in (18)a is not really a cleft sentence with which we are concerned in this book, but it is an identificational copular construction whose subject is a H-lessRC. Since the pronominal usage of no is not our main interest here, I will use examples which do not give us such ambiguity, namely those with a (case-marked) human focus phrase. Inherent semantic properties It has been known in the literature that pronominal no has some derogatory connotation when it refers to a human being (Kuroda 1976–77), as pointed out on page 62. For example: (19) a
Wakai sensei-ga oozei orareru young teachers-NOM many there-are(Honorific) ‘There are a lot of young teachers’
b # Wakai-no-ga oozei orareru young-pro-NOM many there-are(Honorific) The sentence in (19)a has an honorific form on the verb. This is perfectly natural, but the sentence in (19)b is not acceptable; since sensei ‘teachers’ is replaced by a no pro-form, which contains a derogatory meaning, it does not go well with the honorific form of the verb. Interestingly, as pointed out by Itô (1986), Murasugi (1994) and others, this type of anomaly is not observed in HIRCs. The following example is cited from Itô (1986: 125): (20) [sensei-ga kenkyuusitu-kara dete-irassyatta-no]-ni teacher-NOM office-from came-out(Hon)-NM-DAT guuzen oaisuru-koto-ga dekita accidentally meet(Hon) was-able ‘I happened to be able to meet the teacher who was coming out of his office’
68 Cleft Constructions in Japanese Syntax
Although the internal head of the HIRC in (20) is sensei ‘teacher’ which can introduce the honorific form of the verb, the no in the HIRC does not conflict with the honorific morphology, in contrast with (19). Assuming that pronominal nos have a derogatory connotation, we conclude that the no in the HIRC does not belong to the same kind. How about nos in presuppositional clauses of clefts? Let us look at the following sentences: (21) [watasi-ga e oaisita-no]-wa Tanaka-sensei(-ni) da I-NOM meet(Hon)-NM-TOP Tanaka-teacher(-DAT) COP ‘It was Professor Tanaka that I met’ The cleft sentence in (21) has an honorific form of the verb inside the presuppositional clause. The point here is that the no in (21) is perfectly natural with its referent, Tanaka sensei ‘Professor Tanaka’ and the honorific form of the verb. This fact suggests that nos in clefts are not pronominal, just like HIRCs. The second piece of semantic evidence which distinguishes pronominal nos from those in HIRCs and cleft constructions comes from a difference in what they can refer to (Kamio 1983), as discussed in the last section. It is noted that a pronominal no cannot refer to some kind of abstract entity, as shown in (22): (22) a
taihen-na jyookyoo tough situation ‘tough situation’
b?* taihen-na no tough Pro ‘tough one’ The head of the noun phrase in (22)a is jyookyoo, ‘situation’, which is considered an abstract entity. Therefore, it cannot be replaced by a pronominal no, as shown in (22)b. The same is true for H-less RCs, as we saw in (12) on page 62, repeated here as (23): (23) a
b*
[kenkyuu-ni taisuru] jyoonetu research-DAT facing passion ‘a passion toward research’ [kenkyuu-ni taisuru] no research-DAT facing -pro ‘the one toward research’
Nominalizations in Cleft Constructions 69
In contrast, nos in HIRCs are not restricted by this semantic constraint: (24) Taroo-wa [jyookyoo-ga taihen-na no]-o norikoeta Taro-Top situation-NOM tough-COP NM-DAT overcame ‘Taro overcame the situation that was tough’ The sentence in (24) contains a HIRC, in which the abstract noun, jyookyoo ‘situation’, serves as its internal head. Unlike (22)b, the relative clause in (24) can refer to the abstract noun; we conclude that nos in HIRCs do not behave like pronominal nos. Here again, we can find a parallelism between HIRCs and presuppositional clauses. If we assume that nos in presuppositional clauses are pronominal, we should expect that the corresponding focus phrase could not be an abstract entity. This is not the case, as shown in (25): (25) [e taihen-na no]-wa kono jyookyoo da tough-COP-NM-TOP this situation COP ‘What is tough is this situation’ Since both HIRCs and the presuppositional clauses of clefts do not have any semantic restriction on the no, we conclude that the nos in these constructions are different from pronominal nos. Adjective-/numeral-quantifier modification Ohara (1994) and Hoshi (1995) discuss a difference between HIRCs and other types of relative clauses in terms of the location of modifying adjectives. For example, the adjective takasoona ‘expensive-looking’, which modifies the nominal head of relative clause, can be located either before or after a HERC in (26) and a H-less RC in (27): (26) a Taroo-wa [Hanako-ga ei katta] takasoona honi-o Taro-TOP Hanako-NOM bought expensive-looking book-ACC karita borrowed ‘Taro borrowed the expensive-looking book that Hanako bought’ b Taroo-wa takasoona [Hanako-ga ei katta] honi-o Taro-TOP expensive-lookingHanako-NOM bought book-ACC karita borrowed
70 Cleft Constructions in Japanese Syntax
(27) a Taroo-wa [Hanako-ga ei katta] takasoona noi-o Taro-TOP Hanako-NOM bought expensive-looking one-ACC karita borrowed ‘Taro borrowed the expensive-looking one that Hanako bought’ b Taroo-wa takasoona [Hanako-ga ei katta] noi-o Taro-TOP expensive-looking Hanako-NOM bought one-ACC karita borrowed However, Tsubomoto (1981) (cited in Hoshi (1995)) observes that neither position is possible to modify adjectives in HIRCs, as shown in (28): (28) a?*
Taroo-wa [Hanako-ga keekii-o katteoita] Taro-TOP Hanako-NOM cake-ACC bought tabeta noi–o takasoona expensive-looking NM-ACC ate ‘Taro ate the expensive–looking cake that Hanako bought [cakes]’
b ?* Taroo-wa takasoona [Hanako-ga keekii-o Taro-TOP expensive-looking Hanako-NOM cake-ACC tabeta katteoita] noi-o bought NM-ACC ate Let us consider the following presuppositional clauses modified by an adjective: (29) a
[Hanako-ga ei atta-no]-wa sono gakusei(-ni)i-da Hanako-NOM met-NM-TOP that student(-DAT)-COP ‘It was the student that Hanako met’
b ?* [Hanako-ga ei atta] sinsetusoona-no-wa sono Hanako-NOM met gentle-looking-NM-TOP that gakusei(-ni)i-da student(-DAT)-COP ‘It was the gentle-looking student that Hanako met’
Nominalizations in Cleft Constructions 71
c?*
Sinsetusoona [Hanako-ga ei atta-no]-wa sono gentle-looking Hanako-NOM met-NM-TOP that gakusei(-ni)i-da student(-DAT)-COP
(29)a is a basic cleft sentence. The presuppositional phrase is ungrammatical when it is either preceded or followed by the adjectives, sinsetusoona ‘gentle-looking’ in (29)b and (29)c. Thus, these sentences contrast with the H-lessRCs and HERCs in (26) and (27), but do not contrast with the HIRC in (28). In other words, presuppositional clauses behave more like HIRCs in Japanese. The same is ture for the cases where a Numeral Quantifer (NQ) modifies a relative clause. Ohara (1994) argues that, unlike HERCs, HIRCs cannot be modefied by an NQ, as shown in (30):
(30) a [Kare-ga sasidasita hon] go-satu-o uketotta he-NOM handed-out book five-CL-ACC received ‘(I) received the five books that he handed out’ b * [[Kare-ga hon-o sasidasita] no] go-satu-o uketotta he-NOM book-ACC handed-out NM five-CL-ACC received
The sentence in (30)a contains a HERC, in which the NQ, go-satu ‘fiveCL’, modifies the head, hon ‘book’. While this sentence is grammatical, the one in (30)b is not; the NQ cannot modify the HIRC.5 A parallel ungrammaticality can be obtained in the following cleft sentence:
(31) a
b*
[Kare-ga e deatta no] -wa gakusei-ni dat-ta he-NOM came-across-NM-TOP students-DAT COP-PAST ‘It was students that he came across’ [Kare-ga e deatta no] go-nin-wa gakusei-ni he-NOM came-across NM five-CL-TOP students-DAT dat-ta COP-PAST ‘It was students that he came across five of’
72 Cleft Constructions in Japanese Syntax
In (31)b, the NQ, go-nin ‘five-CL’ appears after the presuppositional noclause, and the sentence is not grammatical.6 Again, we find that HIRCs and presuppositional clauses pattern in the same way. The facts shown above suggest that the no in presuppositional clauses is not a pro-form but rather a complementizer under the assumption that the no in HIRCs is C0 (Kuroda 1976–77, Itô 1986, Watanabe 1992, Hoshi 1995). The next question is whether it is possible to think that presuppositional clauses in cleft constructions contain a nominal category with a null head, as proposed by Hoshi (1995). A proposal and an alternative I argue for the following schematic structure for no-clauses of cleft constructions:7
(32)
CP
C’
Opi IP
C no
...ti...
The structure in (32) is distinct from the one for H-lessRCs, in which no is not a complementizer but a noun (see (33) below). (32) is like a wh-interrogative clause in that an operator-variable relation creates an open sentence. This is a similar structure for presuppositional clauses in English it-cleft constructions, as proposed by Chomsky (1977), Barss (1986) and Heggie (1988). Notice that (32) can explain the facts we have seen so far in this section. First, the analysis in (32) can account for the fact that no in presuppositional clauses does not have any inherent semantic restriction on its referent, because the no is not a noun but a C head which does not refer to any argument. Second, we can also give an account for the fact that presuppositional clauses cannot be modified by an adjective or an NQ if we assume (32). Under standard assumptions, prenominal adjectives are predicated of an NP inside a DP projection, according to Abney (1987), and an NQ has its own projection (QP in Terada (1990), and NumP in
Nominalizations in Cleft Constructions 73
Kitahara (1993) and Kawashima (1994)), being predicated of an NP. Therefore, it follows that the nos in presuppositional clauses cannot be modified by adjectives and NQs since presuppositional clauses do not project an NP, as illustrated in (32). I will come back to this issue shortly. However, the following subsection will show some apparent counter evidence to our analysis; the presuppositional clause in cleft constructions behaves in the same way as HERCs and H-lessRCs with respect to case conversion. Therefore, one might still consider that the structure of the no-clause of clefts is not the one in (32) but should be something like (33):8 (33)
NP NP/N [e]
CP
Opi
C’
IP
C no
...ti...
Unlike (32), the structure in (33) projects an NP on top of the clause, which contains an empty N or NP. This structure would be supported by the facts of case conversion; however, I will argue against (33) on pages 77–90, showing that the case conversion does not necessarily imply the existence of nominal category. Apparent counter-evidence It is well known that Japanese allows a genitive case for a phrase which is generally marked by a nominative case within a clause modifying a noun phrase. This is called ga-no conversion in the literature (Harada 1971, 1976, Shibatani 1978, Inoue 1976, among others). As shown below, we observe ga-no conversion phenomena in HERCs, H-lessRCs and no-complement clauses of perception verbs: (34) a [Hanako-ga ei mita] syasini Hanako-NOM saw picture ‘the picture that Hanako saw’
74 Cleft Constructions in Japanese Syntax
b [Hanako-no ei mita] syasini Hanako-GEN saw picture (35) a [Hanako-ga e totta]-no Hanako-NOM took -pro ‘the one that Hanako took’ b [Hanako-no e totta]-no Hanako-GEN took -pro The clauses in (34) are instances of HERCs. The nominative subject can be marked by a genitive marker within the relative clause, as shown in (34)b. The same is true for the H-lessRC in (35). The nos in the head of the clauses in (35) are considered pro-forms, as we already discussed on pages 60–2. The standard generalization for ga-no conversion is that a nominative case marker can be replaced by a genitive case marker when it is dominated by a projection of nominal category (Saito 1983, Miyagawa 1993 and others). Although there are some complications for the generalization, which will be discussed in the next section, the examples above conform to the generalization of case conversion. According to Miyagawa’s (1993) analysis within the framework of Chomsky (1993), a genitive phrase in case conversion phenomena moves to have its case feature checked in the spec of DP at LF, as illustrated in (36):
(36) a [[kinoo Hanako-no katta] hon] yesterday Hanako-GEN bought book ‘the book that Hanako bought yesterday’ b [DP Hanako-noi [kinoo ti katta] hon D] Hanako-GEN yesterday bought book (Miyagawa 1993:217)
The relative clause in (36)a has an LF representation like (36)b. The genitive phrase within the lower clause moves to the spec of DP to check its Case feature. Miyagawa crucially assumes that the no in caseconversion phenomena is the realization of a formal feature, which has to be checked by a functional category. A piece of evidence that a genitive subject is not in spec of DP at S-structure is that, as shown in (36), an adverb can occur on the left of the genitive phrase within the same clause. This means that the genitive phrase is not located in the spec of DP prior to LF in (36)a.
Nominalizations in Cleft Constructions 75
Another fact which strongly supports LF movement of a genitive subject in Miyagawa’s (1993) analysis is that the case alternation affects scope interpretation: (37) a [DP [(kinoo) [John-ka Mary]-ga kita] riyuu]-o osiete yesterday John-or Mary-NOM came reason-ACC tell-me ‘Tell me the reason why John or Mary came (yesterday)’ reason > [John or Mary], *[John or Mary] >reason b [DP [ (kinoo) [John-ka Mary]-no kita] riyuu]-o osiete yesterday John-or Mary-GEN came reason-ACC tell-me reason > [John or Mary], [John or Mary] >reason The difference between the sentences in (37) is that the subject is marked by nominative ga in (37)a whereas it is marked by genitive no in (37)b. Intriguingly, although the nominative subject in (37)a does not allow a wide-scope reading over the head, riyuu ‘reason’, the wide-scope interpretation is possible in (37)b when the subject is genitive.9 Therefore, this fact supports Miyagawa’s analysis; the genitive phrase moves higher than the head noun at the level of interpretation. Assuming the structure in (32) for presuppositional clauses of clefts, we would not expect ga-no conversion based on the standard analysis such as Miyagawa’s (1993). (32) does not project a nominal category, such as N or D, and therefore, there is no position to license a genitive phrase of the case conversion. However, as shown in (38), presuppositional clauses allow ga-no conversion, contrary to our prediction:10 (38) a
[Hanako-ga ei atta-no]-wa sono hito(-ni)i-da Hanako-NOM saw-NM-TOP that person-(DAT)-COP ‘It was that person that Hanako saw’
b? [Hanako-no ei atta-no]-wa sono hito(-ni)i-da Hanako-GEN saw-NM-TOP that person(-DAT)-COP In (38), the subject, Hanako-ga ‘Hanako-NOM’, in the presuppositional clause can also be realized as Hanako-no ‘Hanako-GEN’ without changing the meaning. It appears that this fact should not be expected under our analysis, because Miyagawa’s analysis tells us that the genitive marker, no, in ga-no conversion is licensed in the spec of DP. Therefore, this fact suggests that the no-clause in (38) can be embedded in a DP. Interestingly, HIRCs, which pattern with presuppositional clauses of clefts in other respects, also allow ga-no conversion:
76 Cleft Constructions in Japanese Syntax
(39) a Taroo-wa [ringo-ga sara-no ue-ni aru-no]-o totte tabeta Taro-TOP apple-NOM plate-GEN top-on be-NM-ACC take ate ‘Taro took the apple which was on the plate and ate it’ b Taroo-wa [ringo-no sara-no ue-ni aru-no]-o totte tabeta Taro-TOP apple-GEN plate-GEN top-on be-NM-ACC take ate (Itô 1986: 130) The object HIRC in (39)a has an internal head, ringo ‘apple’, which is marked by a nominative marker. This nominative case can be replaced by the genitive marker, as shown in (39)b. Again, presuppositional clauses pattern together with HIRCs, but crucially, they also behave like H-lessRCs and HERCs. Thus, these facts might show that presuppositional clauses can be analyzed on a par with all the relative clauses. It should be noted that, Itô (1986), following Kuroda (1976–77), argues that (39)b is not a true instance of case conversion, and that the internally headed genitive phrase is in an appositive relation to the rest of the clause. In other words, the no in the internal head is not a genitive marker, such as the one in John-no hon ‘John’s book’, but is the one in bengosi-no John ‘John, the lawyer’. The proposed structure, under the NP hypothesis for noun phrases, is roughly the following: (39)’ a [NP [NP1 ringo]-no [NP2 [CP sara no ue-ni aru no] [NP e]]] apple-COP plate-GEN top-DAT be-NM ‘the apple, which is on the plate’ It is assumed that the no within NP1 is inserted under the appositive relation between NP1 and NP2, as shown in (39)’a. However, notice that the PP, sara no ue-ni ‘on the plate’, can appear before the genitive phrase, as shown in (39)’b, and the case conversion is still allowed: (39)’ b [sara no ue-nii [NP1ringo]-no [NP2 [CP ti aru no] [NP e ]]] plate-GEN top-DAT apple-GEN be NM Under Itô’s analysis, the PP must be considered to be scrambled out of the NP2. This configuration is not generally allowed, as shown below: (39)’ c [[NP1 gakusei-no] [NP2 tosyokan-de yoku benkyoo-suru Taroo]] student-COP library-in often study Taro ‘Taro, who is a student and often studies in the library’
Nominalizations in Cleft Constructions 77
d * [tosyokan-dei [NP1 gakusei-no] [NP2 ti yoku benkyoo-suru library-in student-COP often study Taroo]] Taro As illustrated in (39)’d, extraction out of one of the appositive noun phrases is not allowed. This suggests that it is unlikely that the genitive phrase in (39)’a holds the appositive relation with the NP2. Therefore, I assume that (39)b is a true instance of case conversion. It initially appears that what we have seen in (38)b and (39)b may contradict the facts of adjective and NQ modification in the previous subsections. Assuming that ga-no conversion is expected only within nominal clauses, presuppositional clauses and HIRCs cannot be represented as in (32). Thus, the facts here might support the analysis in (33); however, the next section will argue against this interpretation of the facts.
The no-clause as a CP In the last section, we started looking at the empirical evidence for the structure (32) for presuppositional clauses of cleft constructions. However, the case conversion phenomena we saw on pages 73–7 suggest that this may not be on the right track. The reason why the case-conversion facts may support the structure in (33) rather than the one in (32) for presuppositional clauses of clefts is that under the standard assumption, the case conversion should be licensed only in nominal clauses. This section, however, reexamines this prevailing assumption, and proposes following Watanabe (1996), that ga-no conversion does not appear only within nominal clauses. In the following subsections, I will first take a look at Watanabe’s (1996) analysis of case conversion phenomena, and then, assuming that HIRCs share the same structure with presuppositional clauses in cleft constructions, I will closely examine extraction from HIRCs and presuppositional clauses. Based on the empirical evidence, it is argued that the no-clauses of clefts and HIRCs do not constitute complex NP structures. This will vindicate the proposed structure in (32). Properties of ga-no conversion Watanabe (1996) argues that ga-no conversion displays some of the properties of Wh-agreement and does not necessarily appear within a
78 Cleft Constructions in Japanese Syntax
nominal clause in Japanese. He points out several important properties of the case conversion to start with: 1) ga-no conversion is optional, 2) it is subject to a transitivity restriction, and 3) it is limited to the nominative case (no o-no ‘ACC-GEN’ conversion takes place in the language).11 Since these properties are also found in French stylistic inversion, Watanabe accounts for both facts with a unified analysis. First, let us look at the descriptive facts to show the properties of ga-no conversion listed above. As for the optionality in 1), we have already seen the examples on pages 73–4. The following sentences illustrate the second property: a transitivity restriction, in which no accusativemarked phrase may intervene between a genitive subject and a predicate in ga-no conversion configurations. The examples in (40) are instances of HERCs, and those in (41) are, according to Watanabe (1996: 375–6), complex noun phrases whose nominal complements are subjunctive clauses: (40) a
b*
(41) a
b*
[John-ga ei LGB-o kasita] hitoi John-NOM LGB-ACC lent person ‘the person to whom John lent LGB’ [John-no ei LGB-o kasita] hitoi John-GEN LGB-ACC lent person ‘the person to whom John lent LGB’ [John-ga uso-o tuita] koto John-NOM lie-ACC told fact ‘the fact that John told a lie’ [John-no uso-o tuita] koto John-GEN lie-ACC told fact ‘the fact that John told a lie’
(40)a and (41)a are nominal clauses, which we can expect to license ga-no conversion. However, as shown in (40)b and (41)b, the nominative marker cannot be replaced by a genitive marker when the direct object intervenes between the genitive phrase and the verb in the clause. We might want to consider that the ungrammaticality is caused by mere phonological intervention of the object between genitive phrases and the predicate in nominal clauses; however, even when the object is scrambled to clause-initial position, the ungrammaticality does not change:
Nominalizations in Cleft Constructions 79
(42)*
[LGB-oj John-no ei tj kasita] hitoi LGB-ACC John-GEN lent person ‘the person to whom John lent LGB’
(43)*
[uso-oi John-no ti tuita] koto lie-ACC John-GEN told fact ‘the fact that John told a lie’
(42) and (43) show that the object is preposed to clause-initial position, so that nothing appears phonologically between the genitive subject and the verb. However, the clauses are still ungrammatical, contrary to our expectation.12 Remember that, as we already saw on pages 113–8, when the object is relativized, case conversion is licensed. Thus, we can say that Wh-movement licenses case conversion but scrambling does not. The third property is that only the nominative case changes into genitive; there is no o-no ‘ACC-GEN’ conversion phenomenon in Japanese. The following examples illustrate this point: (44) a [eiLGB-o katta] hitoi LGB-ACC bought person ‘the person who bought LGB’ b*
[ei LGB-no katta] hitoi LGB-GEN bought person
In (44)a, the subject is relativized leaving the object inside the clause. As shown in (44)b, the accusative marker cannot change to a genitive case even if no element intervenes between the genitive phrase and the predicate in the structure. Now, let us go back to the second property of ga-no conversion, and see if this is also observed in presuppositional clauses and HIRCs: (45) a [Taroo-ga e ringo-o ageta-no]-wa Hanako-ni-da Taro-NOM apple-ACC gave-NM-TOP Hanako-DAT-COP ‘It was Hanako that Taro gave an apple to’ b*
[Taroo-no e ringo-o ageta-no]-wa Hanako-ni-da Taro-GEN apple-ACC gave-NM-TOP Hanako-DAT-COP
80 Cleft Constructions in Japanese Syntax
(46) a
Taroo-wa [Hanako-ga katteoita-no]-o ringo-o Taro-TOP Hanako-NOM apple-ACC bought-NM-ACC tabetesimatta ate-up ‘Taro ate up the apples that Hanako had bought’
b* Taroo-wa [Hanako-no ringo-o katteoita-no]-o Taro-TOP Hanako-GEN apple-ACC bought-NM-ACC tabetesimatta ate-up (Itô 1986: 130) As we can observe in (45) and (46), when the direct object, ringo-o ‘apple-ACC’, appears between the genitive subject and the verb, the case conversion is not licensed in the presuppositional clause and the HIRC. In addition, scrambling the object does not make the sentences in (45)b and (46)b grammatical, as shown in (47) and (48): (47)* [ringo-oi Taroo-no e ti ageta-no]-wa Hanako(-ni)-da apple-ACC Taro-GEN gave-NM-TOP Hanako-DAT-COP ‘It was Hanako that Taro gave an apple’ (48)*
Hanako-no ti katteoita-no]-o Taroo-wa [ringo-o i Taro-TOP apple-ACC Hanako-GEN bought-NM-ACC tabetesimatta ate-up ‘Taro ate up the apples that Hanako had bought’
Needless to say, o-no conversion is not allowed in presuppositional clauses and HIRCs, as shown in (49) and (50): (49) a
[e ringo-o tabeta-no]-wa Hanako-da apple-ACC ate-NM-TOP Hanako-COP ‘It was Hanako who ate an apple’
b* [e ringo-no tabeta-no]-wa Hanako-da apple-GEN ate-NM-TOP Hanako-COP (50) a Taroo-wa [pro ringo-o katteoita-no]-o tabetesimatta Taro-TOP apple-ACC bought-NM-ACC ate-up ‘Taro ate up the apples that (he) had bought’ b* Taroo-wa [pro ringo-no katteoita-no]-o tabetesimatta Taro-TOP apple-GEN bought-NM-ACC ate-up
Nominalizations in Cleft Constructions 81
Therefore, it is reasonable to assume that the case conversion in presuppositional clauses and HIRCs can be accounted for on a par with the case conversion in relative clauses and nominal complements.13 Case checking with Agr Before going into Watanabe’s (1996) analysis of ga-no conversion, let us briefly look at his basic theoretical assumptions, especially the background of the Case theory that he proposes (Watanabe 1993, 1995). First of all, Watanabe (1996) assumes the split INFL hypothesis proposed by Pollock (1989) and its extension in Chomsky (1991). A schematic phrase structure under this theory is illustrated as follows:14 (51)
CP C’ C
AgrSP AgrS’ AgrS
TP T’ Tns
AgrOP AgrO’ AgrO VP
The Case theory proposed by Chomsky (1993) relies on the structure given in (51). Within this framework, there is no role of a notion of government; instead, a simple X’-theoretic notion of a spec–head relation is important for the Case theory. As shown in (52), structural Case checking takes place in the following manner:15 (52)
AgrP Agr’
DP Agr Xi
Agr
XP ti
82 Cleft Constructions in Japanese Syntax
A head X, which bears its Case feature, is adjoined to Agr and the Case feature is checked with that of DP in the spec of AgrP. Under this approach, Case licensing relations are restricted to only spec-head relations. We should note, however, that in the correct derivation, the object which is in the complement of V must move over the subject in the spec of VP. This movement would cause violation of the Relativized Minimality (Rizzi 1990).16 In order to avoid this problem, Chomsky (1993) proposes the mechanism of equidistance, as shown in (53):17 (53) If α and β are in the same minimal domain, they are equidistant from γ. (54) a MIN(S), S a set of categories, is the smallest subset K of S such that for any γ ε S, some β ε K reflexively dominates γ b The domain of a head chain CH=(α1 . . . , αn) is the set of nodes contained in MAX (α1) and not containing any αi c MAX(α), where a is a head, is the least full-category maximal projection dominating α Based on the definitions in (53) and (54), the following derivation is not ruled out:
(55)
AgrOP DPi
AgrO’
AgrO Vj AgrO
VP Subject tj
V’ ti
The spec of AgrOP and the spec of VP are in the same minimal domain of the chain (AgrO, tj), and hence, the two positions are equidistant from the complement of V by definition. The object DP can thus move to the spec of AgrO over the subject in the spec of VP without violating any constraint on movement under the framework of Chomsky (1993). Furthermore, the subject in the spec of VP can move to the spec of TP (and AgrSP), crossing the object DP under the same mechanism.
Nominalizations in Cleft Constructions 83
With this much as background, let us take a look at the mechanism of Case checking in Watanabe (1993, 1995), which is slightly different from the standard assumption. Watanabe assumes that Case features have to be eliminated from structure by the end of derivation. This is made possible by the mechanism that Agr can absorb a Case feature either from a DP or from a Case-bearing head and that the head further moves to another functional category. In other words, the Case feature of the Case-bearing head is not transferred to Agr when Case checking takes place, because it is assumed that Agr can take only one Case feature in. In order for the derivation to converge, Watanabe claims that the Case-bearing head must be adjoined to the immediately higher head to allow the follow-up checking to take place. Considering nominative Case checking, which is directly relevant to ga-no conversion, the process of the Case checking takes place in the following way:
(56)
a
AgrSP DP [Nom]
AgrS’
AgrS
TP
Tnsi AgrS [Nom]
ti
b
AgrSP AgrS’
DP [φ] AgrS
AgrS Tnsi [Nom] [Nom]
TP ti
As shown in (56)a, the head of TP is raised to adjoin to AgrS, and the nominative DP moves to the spec of AgrSP. In this configuration, the AgrS absorbs the nominative Case of the DP and the Case feature which the DP has is eliminated, as shown in (56)b. Notice that assuming that Agr can only take in one Case feature, the Case feature of the Tns still
84 Cleft Constructions in Japanese Syntax
remains. What is needed in Watanabe’s (1993, 1995) theory is that the complex Tns/Agr head undergoes further movement and adjoins to C, as shown below: (57) a
CP C AgrSj
AgrSP C
DP [φ]
Tnsi AgrS [Nom] [Nom]
b
AgrS’ AgrS
TP
tj
ti
CP C
AgrSj Tnsi [φ]
AgrS [Nom]
AgrSP C
DP [φ]
AgrS’ AgrS
TP
tj
ti
In (57)a, the Tns/Agr complex head adjoins to the C. Then, as shown in (57)b, the C deletes the nominative case feature of the Tns as a followup checker. Although it appears that the nominative Case feature of AgrS would make the derivation crash, the derivation converges because Agr becomes invisible at LF due to its lack of semantic content. Further discussion on Case checking is not our main concern here. See, for details of this line of analysis, Watanabe (1993, 1995). What is more relevant here is that Watanabe’s (1993, 1995) theory of Case checking predicts a tight relation between Tns, AgrS and Complementizer. The structure involving Tns, AgrS and C forms one unit which underlies nominative Case checking.18 The case conversion in Japanese is one of such instances, as will be discussed below.
Nominalizations in Cleft Constructions 85
French stylistic inversion and ga-no conversion One of the appealing claims made by Watanabe (1996) is that his analysis captures cross-linguistic parallelisms that put Japanese case conversion in a broader context. For instance, he points out striking similarities between French stylistic inversion and case-conversion in Japanese. French stylistic inversion is optional, as shown in (58) (Kayne and Pollock 1978: 595), and it is subject to a transitivity restriction, as in (59): (58) a Je me demande quand ton ami partira I wonder when your friend will leave ‘I wonder when your friend will leave’ b Je me demande quand partira ton ami I wonder when will leave your friend? (59) a*
Je me demande quand mangera sa pomme Marie I wonder when will eat her apple Mary ‘I wonder when Mary will eat her apple
b*
Je me demande quand mangera Marie sa pomme I wonder when will eat Mary her apple
However, when the direct object is extracted through Wh-movement, as shown in (60), cited in Valois and Dupuis (1992: 327), stylistic inversion becomes possible: (60) Je me demande quel livre préfère Pierre I wonder which book prefers Pierre ‘I wonder which book Pierre prefers’ The facts in (58)–(60) are reminiscent of Japanese case-conversion phenomena. Based on the analyses by Déprez (1988, 1990) and Valois and Dupuis (1992), Watanabe (1996) claims that the inverted subjects above occupy the spec of VP. In other words, these subjects can remain in situ in the context of Wh-extraction or in some kind of subjunctive clause. This analysis can explain the ungrammaticality of (59). Within the framework of Chomsky (1993), Watanabe (1996) proposes that the
86 Cleft Constructions in Japanese Syntax
intervening object causes a violation of Minimality (Collins and Branigan 1997), as illustrated in (61): (61) [AgrSP Subji AGR [TP T [AgrOP Objj AGR [VP ti tj ]]]] X Let us assume that the spec of TP is not available at LF, as proposed by Bobaljik and Jonas (1996), and movement proceeds cyclically. Then, LF movement of the subject to the spec of AgrSP for Case checking violates Minimality, because it crosses the object in the spec of AgrOP; the spec of AgrOP and the spec of AgrSP are not in the same minimal domain (see (54)). On the contrary, if the object is a Wh-constituent, as we have seen in (60), it can adjoin to the AgrOP on its way to the spec of CP (Kayne 1989 and Ura 1993), which does not create a Minimality violation.19 Watanabe (1996) extends this analysis for French stylistic inversion to case-conversion in Japanese as follows. He assumes that spec of AgrS does not need to be occupied in overt syntax in the context of Wh-agreement (Drijkoningen 1988). A genitive subject occupies the spec of VP in overt syntax just like a subject in French stylistic inversion configurations, and it checks its Case feature in the spec of AgrSP at LF. This analysis actually implies that there is another type of no in Japanese: a ‘nominative’ Case of no. Watanabe also points out that the no in ga–no conversion is a disguised form of the nominative Case. One of the independent reasons for the claim that a genitive subject does not move to spec of AgrSP in overt syntax is that a manner adverb, which is generally assumed to occur in a position adjoined to VP, cannot appear between a genitive subject and its predicate, as shown below: (62) a?* [kinoo Taroo-no isoide ei kaita] ronbuni yesterday Taro-GEN quickly wrote paper ‘the paper which Taro wrote quickly yesterday’ b
[kinoo isoide Taroo-no ei kaita] ronbuni yesterday quickly Taro-GEN wrote paper
The manner adverb, isoide ‘quickly’, is between the genitive subject and the verb, as shown in (62)a, and the sentence is not grammatical. On the other hand, if the adverb occurs before the genitive subject, the sentence is perfect, as shown in (62)b. Assuming that the manner adverb indicates a VP boundary, the examples in (62) show that the genitive subject stays inside the VP at S-structure.
Nominalizations in Cleft Constructions 87
The same explanation as the one for French stylistic inversion applies to the transitivity restriction in Japanese case conversion; when the object appears in the same clause of the genitive subject, then movement of the genitive subject at LF causes a violation of Minimality. Under this analysis, it is assumed that nominative subjects are raised out of VP in overt syntax whereas accusative objects remain in-situ in Japanese. This is illustrated in the LF representation in (63)b:20 (63) a*
[John-no hon-o kasita] hito John-GEN book-ACC lent person ‘the person to whom John let a book’
b* [AgrSP John-noi [TP [AgrOP hon-oj [VP ti tj Kasita ] AgrO] t ] AgrS/Tns X Notice that Watanabe’s (1996) analysis shows that the case conversion is observed only in the AgrS-Tns-C system; a Wh-agreement configuration licenses ga-no conversion as French stylistic inversion does. This means that Watanabe’s approach does not generate case conversion of different types, such as o-no ‘ACC-GEN’ conversion in Japanese; Agr-o does not form a unit with the C head in his Case theory. More importantly, the consequence of this line of analysis predicts that case conversion takes place not only within relative and subjunctive clauses, but also within some other clauses involving Wh-movement. The following examples cited from Watanabe (1996: 394) show that the prediction is in fact borne out: (64) a John-wa [Mary-no yonda yori] takusan-no hon-o yonda John-TOP Mary-GEN read than many-GEN book-ACC read ‘John read more books than Mary read’ b John-wa [Mary-no hataraita yori] isshookenmei hataraita John-TOP Mary-GEN worked than hard worked ‘John worked harder than Mary read’ The sentences in (64) are known as comparative deletion constructions. Intuitively, the bracketed clauses do not look like nominal clauses; for instance, apparently the yori ‘than’ clause can never be marked by a case marker in Japanese. However, the fact is that the subject can be marked by a genitive marker within the yori-clause. Kikuchi (1987) argues that these constructions involve null operator movement, which is an
88 Cleft Constructions in Japanese Syntax
instance of wh-movement (Chomsky 1977). Thus, the grammaticality of the examples in (64) supports Watanabe’s analysis.21,22 To sum up so far, we have seen that Watanabe’s analysis can explain various facts concerning ga-no conversion. It has been shown that case conversion is not a typical phenomenon within a nominal clause; rather, it generally occurs in Wh-agreement configurations. However, it seems that one piece of evidence which supports Miyagawa’s (1993) analysis may not be accounted for under Watanabe’s (1996) theory. That is the scope facts in (37), repeated here as (65): (65) a [DP [(kinoo) [John-ka Mary]-ga kita] riyuu]-o osiete yesterday John-or Mary-NOM came reason-ACC tell–me ‘Tell me the reason why John or Mary came (yesterday)’ reason > [John or Mary], *[John or Mary] >reason b [DP [(kinoo) [John-ka Mary]-no kita] riyuu]-o osiete yesterday John-or Mary-GEN came reason-ACC tell-me reason > [John or Mary], [John or Mary] > reason The ambiguous reading given in (65) becomes possible under Miyagawa’s analysis; namely, the genitive phrase moves to the spec of DP at LF, so that the subject can take scope over the relative head. However, when the adverb, kinoo ‘yesterday’, appears left to the genitive subject in (65)b, it becomes difficult to get the wide-scope reading of the genitive subject for at least some native speakers of Japanese: (66) a [DP [ kinoo [John-ka Mary]-no kita] riyuu]-o osiete yesterday John-or Mary-GEN came reason-ACC tell-me reason > [John or Mary], *[John or Mary] >reason b [DP [[John-ka Mary]-no kita] riyuu]-o osiete John-or Mary-GEN came reason-ACC tell-me reason > [John or Mary], [John or Mary] > reason If the observation in (66) is correct, contrary to that of Miyagawa (1993), it is possible to think that when the genitive subject takes scope over riyuu ‘reason’ in (66)b, it is not within the relative clause, but rather, it is outside the relative clause, modifying the entire relative clause with its head. The structure under such reading is suggested as follows:
Nominalizations in Cleft Constructions 89
(67) [DP [NP [John-ka Mary]-no[NP [CP pro kita] riyuu]]] John-or Mary-GEN came reason ‘John’s or Mary’s reason that he or she came’ The genitive subject is in a position adjoined to the NP containing the relative clause and its head. As it is widely observed that relative clauses are modified by a genitive phrase in Japanese, as shown in (67)’, it is not surprising that (66)b has a structure something like (67): (67)’ a [kyonen]-no [Taroo-ga e katta] hon last-year-GEN Taro-NOM bought book ‘last year’s book that Taro bought’ b
[yooroppa]-no [Taroo-ga e itta] kuni Europe-GEN Taro-NOM went country ‘the country which Taro went in Europe’
Thus, strictly speaking, the no in (67) is not the one for case conversion but rather a prenominal no (see pages 58–60). Therefore, the facts of scope interpretation in case conversion do not undermine Watanabe’s analysis. If the analysis we have seen in this subsection is correct, the case conversion within presuppositional clauses as well as HIRCs does not require the structure of (33), which is repeated as in (68): (68) NP CP Opi IP
NP/N C’
[e]
C no
...ti...
Assuming that a genitive phrase in ga-no conversion is not licensed in the spec of DP but in the spec of VP because of a Wh-agreement configuration, we do not have any reason to assume a nominal projection that dominates the CP. Although the structure in (68) involves a null
90 Cleft Constructions in Japanese Syntax
operator movement which potentially licenses ga-no conversion, the caseconversion phenomenon itself cannot be taken as evidence for nominal clauses. Therefore, considering this together with other facts discussed on pages 66–77. I continue to assume that the structure for a presuppositional clause of clefts is something like (32), repeated as (69) below: (69) CP Opi IP
C’
C no
...ti...
Extraction from HIRCs In the last subsection, we saw that the facts of case conversion do not require the structure in (68). The discussion in this subsection will provide further evidence for (69). Here, we will look at some potential evidence concerning extraction from HIRCs and no-clauses of clefts, and will show that these clauses do not form a complex NP like (68). If the structure in (68) is correct, we may expect HIRCs to induce complex NP effects. Hoshi (1995: Appendix, 273–80), however, points out that unlike other kinds of complex NPs, HIRCs do not count as islands for extraction, as shown in (70): (70) a
John-ga [[[Mary-ga teeburu-no ue-ni ringo-o John-NOM Mary-NOM table-GEN top-DAT apple-ACC oitekureta]-no]-o [e]] tabeta put -NM-ACC ate ‘John ate the apple which Mary put on the table’
b? Teeburu-no ue-nii table-GEN top-DAT oitekureta]-no]-o [e]] put -NM-ACC
John-ga [[[ Mary-ga ti ringo-o John-NOM Mary-NOM apple-ACC tabeta ate
The sentence in (70)a contains a HIRC in its object position. Even though the PP, teeburu-no ue-ni ‘on the table’, is scrambled out of the
Nominalizations in Cleft Constructions 91
HIRC to sentence-initial position, the sentence is grammatical, as shown in (70)b.23 Note here that the PP still modifies the predicate within the HIRC, and it does not provide any interpretation such that the action expressed by the matrix verb takes place on the table. This fact is not expected if we assume the structure in (68) for the HIRC, because in Japanese, scrambling from out of a complex NP yields ungrammaticality, as shown in (71) and (72): (71) a
Taroo-ga [[Hanako-ga heya-de benkyoositeiru] soba]-o Taro-NOM Hanako-NOM room-in be-studying side-ACC toorikakatta passed-by ‘Taro passed by the place where Hanako was studying in the room’
b?* Heya-dei Taroo-ga [[Hanako-ga ti benkyoositeiru] soba]-o room-in Taro-NOM Hanako-NOM be-studying side-ACC toorikakatta passed-by (72) a
Taroo-ga [[Hanako-ga teeburu-no ue-ni e oita] ringo]-o Taro-NOM Hanako-NOM table-GEN top-on put apple-ACC tabeta ate ‘Taro borrowed the book that Hanako gave to Masao’
b?* Teeburu-no ue-nii Taroo-ga [[Hanako-ga ti e oita] table-GEN top-on Taro-NOM Hanako-NOM put ringo]-o tabeta apple-ACC ate The sentence in (71)a involves a so-called pseudo-relative clause, and the PP within the clause cannot be scrambled to sentence-initial position, as shown in (72)b. The same is true for the relative clause in (72); the PP, teeburu-no ue-no ‘on the table’, cannot scramble out of the HERC.24 It is crucial that the sentence in (70)b clearly contrasts with those of (71)b and (72)b; the HIRC does not behave like standard complex NPs. Now, let us look at presuppositional no-clauses in clefts. As shown in (73) and (74), the presuppositional clauses exhibit the same behaviour:
92 Cleft Constructions in Japanese Syntax
(73) a
Taroo-ga [[Hanako-ga kodomo-o e syookaisita-no] -wa Taro-NOM Hanako-NOM child-ACC introduced-NM-TOP sono hito-ni-da to] itta that person-DAT-COP COMP said ‘Taro said that it was that person who Hanako introduced to the child’
b? Kodomo-nii Taroo-ga [[Hanako-ga ti e syookaisita-no]-wa child-DAT Taro-NOM Hanako-NOM introduced-NM-TOP sono hito-da to] itta that person-COP COMP said (74) a
Taroo-ga [[Hanako-ga heya-de e atta -no] -wa Taro-NOM Hanako-NOM room-in met -NM-TOP sono hito-ni-da to] itta that person-DAT-COP COMP said ‘Taro said that it was that person that Hanako met in the room’
b? Heya-dei Taroo-ga [[Hanako-ga ti e atta-no]-wa room-in Taro-NOM Hanako-NOM met-NM-TOP sono hito-ni-da to] itta that person-DAT-COP COMP said First, scrambling the dative phrase from out of the no-clause does not make the sentence ungrammatical, as illustrated in (73)b. Scrambling the PP does not make the sentence ungrammatical either, as shown in (74)b. These facts are unexplained if we assume the structures in (68) for no-presupposition clauses. On the other hand, the structure in (69) can account for the facts; the proposed structure in (69) is a bare CP and does not constitute a complex NP structure.25 Here, let us consider how Hoshi (1995) accounts for the grammaticality in (70)b. Hoshi proposes that HIRCs are dominated by a null N or NP. Under his analysis, it is suggested that the CP within NP can move overtly to a position adjoined to IP or VP, just like a case of relative clause extraposition. Hoshi claims that this analysis is supported by the following examples concerning adverbial placement: (75) a
John-ga teeburu-no musaboruyooni [[[ Mary-ga John-NOM greedily Mary-NOM table-GEN ue-ni oitekureta]-no]-o [e]] tabeta ringo-o top-DAT apple-ACC put-NM-ACC ate ‘John greedily ate the apple which Mary put on the table’
Nominalizations in Cleft Constructions 93
b
(76) a*
John-ga [[[ Mary-ga teeburu-no ue-ni ringo-o John-NOM Mary-NOM table-GEN top-DAT apple-ACC oitekureta] -no]-o [e]] musaboruyooni tabeta put -NM-ACC greedily ate Teeburu-no ue-nii John-ga musaboruyooni table-GEN top-DAT John-NOM greedily [[[Mary-ga ti ringo-o oitekureta]-no]-o [e]] tabeta Mary-NOM apple-ACC put -NM-ACC ate
b? Teeburu -no ue-nii John-ga [[[Mary-ga ti ringo-o table -GEN top-DAT John-NOM Mary-NOM apple-ACC oitekureta]-no]-o [e]] musaboruyooni tabeta putNM-ACC greedily ate In the examples in (75), we find that the manner adverb, musaboruyooni ‘greedily’, can be situated either before or after the HIRC. Interestingly, if the PP, teeburu-no ue-ni ‘on the table’, is scrambled to the front of the sentence, then the adverb cannot occur right before the HIRC, as shown in (76)a. In contrast, the sentence is still grammatical when the adverb appears after the HIRC, as in (76)b. Hoshi (1995) proposes that the reason for the contrast provided in (75) and (76) is that the head-internal CP is moved to adjoin to the VP; this means that the CP is no longer part of a complex NP. The following are the schematic structures for (76)a and (76)b, respectively: (77) a∗
b
IP
IP VP
VP
∧ Adv
VP NP V
∧
CPi t
VP Adv NP
CP N t
VP
ti
V N
(77)a represents the sentence in (76)a, and (77)b the sentence in (76)b. Thus, if this assumption is correct, scrambling is not ruled out as long as a HIRC appears before the manner adverbs.
94 Cleft Constructions in Japanese Syntax
Hoshi’s (1995) account might explain the following sentences which involve a cleft construction. As we have seen in (75) and (76), the cleft sentences behave exactly the same way as the HIRCs: (78) a
b
Taroo-ga hayakuti-de [[Hanako-ga heya-de e atta-no] Taro-NOM fast Hanako-NOM room-in met-NM -wa sono hito-ni-da-to] itta -TOP that person-DAT-COP-COMP said ’Taro spoke fast and said that it was that person who Hanako introduced to the child’ Taroo-ga [[Hanako-ga heya-de e atta -no] -wa sono Taro-NOM Hanako-NOM room-in met -NM-TOP that hito-ni da-to] hayakuti-de itta person-DAT COP-COMP fast said
As we have seen in the HIRC constructions, the manner adverb (hayakuti-de ‘speaking-fast’) can occur either before or after the no-clause in clefts, as shown in (78). In addition, PP preposing from out of the no-clause is prohibited when the adverb is located before the clause. This is shown in (79)a. In contrast, the sentence (79)b is grammatical, in which we find the adverb after the no-clause: (79) a*
Heya-dei Taroo-ga hayakuti -de [[Hanako-ga ti e] room-in Taro-NOM fast Hanako-NOM atta-nowa sono hito-ni-da to] itta met-NM- TOP that person-DAT-COP-COMP said
b? Heya-dei Taroo-ga [[Hanako-ga ti e atta-no]-wa sono room-in Taro -NOM Hanako-NOM met-NM-TOP that hito-ni da to] hayakuti-de itta person-DAT COP-COMP fast said However, as we will see in the following discussion, the assumption of extraposition of the entire head-internal CP in (77) is not plausible. We should notice that the ungrammatical cases involving a manner adverb with a scrambled phrase are more generally observed. Even scrambling from a complement clause does not give us grammatical sentences when an adverb intervenes between the preposed phrase and the complement clause. Let us look at the following sentence:
Nominalizations in Cleft Constructions 95
(80) a
b
Taroo-ga hayakuti-de [Hanako-ga teeburu-no ue-ni Taro-NOM fast Hanako-NOM table-GEN top-DAT ringo-o oita-to] itta apple-ACC put-COMP said ‘Taro spoke fast and said that Hanako put the apple on the table’ Taroo-ga [Hanako-ga teeburu-no ue-ni ringo-o Taro-NOM Hanako-NOM table-GEN top-DAT apple-ACC oita-to] hayakuti-de itta. put-COMP fast said
(81) a ?* Teeburu-no ue-nii Taroo-ga hayakuti-de [Hanako-ga ti table-GEN top-DAT Taro-NOM fast Hanako-NOM ringo-o oita-to] itta apple-ACC put-COMP said b?
Taroo-ga [Hanako-ga ti ringo-o Teeburu-no ue-nii table-GEN top-DAT Taro-NOM Hanako-NOM apple-ACC oita-to] hayakuti-de itta put-COMP fast said
The sentences in (80) show that the distribution of adverbs is the same as the one for HIRCs and no-clauses in clefts; namely, the manner adverb modifies the verb either before or after the complement clause. Also, the same fact we saw above is observed when the PP, teeburu-no ue-ni ‘on the table’, is scrambled to sentence-initial position, as demonstrated in (81). The standard analysis for to ‘that’ takes it to be a complementizer and assumes a CP projection without any NP projection above the CP. Therefore, it seems that the ungrammatical sentences we have seen in this subsection do not come from violation of complex NP constraint but from some other reason. These facts suggest that Hoshi’s (1995) assumption of CP extraposition is not on the right track. The licensing of adverbs in Japanese has not been investigated very much in the literature (but see Ueda (1990, 1993)). Although the interaction between adverbs and other phrases must be closely examined, let me just introduce a descriptive observation on this issue to show that there must be another way to account for the sentences in (76) and (79). Looking at the sentences in (76) and (79), we notice that the scrambling should not cross a manner adverb. However, there is a case where scrambling does cross a manner adverb, as shown in (82)a:
96 Cleft Constructions in Japanese Syntax
(82) a Teeburu-no ue-nii [Hanako-ga ti ringo-o oitekureta-no]-oj table-GEN top-on Hanako-NOM apple-ACC put-NM-ACC Taroo-ga musaboruyooni tj tabeta Taro-NOM greedily ate ‘On the table, the apple that Hanako put [apples], Taro greedily ate’ b Taroo-ga musaboruyooni teeburu-no ue-nii [Hanako-ga ti Taro-NOM greedily table-GEN top-on Hanaka-NOM oitekureta-no]-o tabeta ringo-o apple-ACC put-NM-ACC ate ‘Taro greedily ate, on the table, the apple that Hanako put [apples]’ In the sentence in (82)a, it appears that the adverb, musaboruyooni ‘greedily’, intervenes between the scrambled phrase and its trace. However, scrambling of the HIRC is supposed to be local, and hence, can be considered A-movement; for some reason, A-movement is not prohibited by an intervening adverb. In addition, the chain of the scrambled PP, teeburu-no ue-ni ‘on the table’, and its trace is not disrupted by the adverb, so that we can get the grammatical sentence. Considering (82)b, since the adverb is not in between the PP and its trace, we expect a grammatical sentence. The same pattern of judgments is observed in corresponding cleft constructions and sentences with a complement clause, as exemplified in (83) and (84):
(83) a Kodomo-nii [[Hanako-ga ti e syookaisita-no]-wa sono hito child-DAT Hanako-NOM introduced-NM-TOP that person Taroo-ga hayakutidetj itta. -da to]j -COP-COMP Taro-NOM fast said ‘To the child, it was that person who Hanako introduced, Taro fast spoke’ b
Taroo-ga hayakutide kodomo-nii [Hanako-ga ti e Taro-NOM greedily child-DAT Hanako-NOM syookaisita-no-wa sono hito-da-to] itta introduced-NM-TOP that person-COP-COMP said ‘Taro fast spoke and said that to the child, it was that person who Hanako introduced’
Nominalizations in Cleft Constructions 97
(84) a Kodomo-nii [Hanako-ga ti sono hito-o child-DAT Hanako-NOM that person-ACC syookaisita-to]j Taroo-ga hayakutidetj itta introduced-COMP Taro-NOM fast said ‘To the child, Hanako introduced that person, Taro spoke fast’ b
Taroo-ga hayakutide kodomo-nii [Hanako-ga ti sono Taro-NOM fast child-DAT Hanako-NOM that hito-o syookaisita to] itta person-ACC introduced COMP said ‘Taro fast spoke and said that to the child, Hanako introduced that person’
The facts here show that the ungrammatical sentences in (76) and (79) are not to be accounted for under the complex NP analysis. It seems rather that the data involving an adverb should be accounted for by an analysis in terms of Relativized Minimality effects; A’-movement cannot cross another element in an A’-position. Taking (81) as an example, the PP, teeburu-no ue-ni ‘on the table’, is preposed by long-distance scrambling, which is considered A’-movement (Saito 1992). This movement takes place skipping another A’-position, namely, the one for the adverb, hayakuti-de ‘(speaking) fast’, and hence, violation of shortest move under the principles of economy. If this idea is correct to explain the extraction facts, we still do not have to conclude that no-clauses in clefts is dominated by a nominal head. Rather, the pure CP analysis fits the facts better. We have seen so far the cases of scrambling of an element which is not an internal head of HIRCs. For completeness, let us now consider how our analysis accounts for a scrambled phrase which is the internal head. Note that this test cannot be applicable to no-clauses of clefts because the operator in question is null. However, the following discussion relies on the assumption that an internal-head of HIRCs is analogous to the null operator of clefts. If an internal head undergoes overt movement outside the HIRC, the sentence becomes ungrammatical, as shown below:26 (85) a Taroo-ga [okaasan-ga Hanako-ni keeki-o child-NOM mother-NOM Hanako-DAT cake-ACC age-yootosita no]-o totte tabeta give-tried NM-ACC took ate ‘Taro took and ate the cake which his mother tried to give to Hanako’
98 Cleft Constructions in Japanese Syntax
b?
Hanako-nii Taroo-ga [okaasan-ga ti keeki-o Hanako-DAT Taro-NOM mother-NOM cake-ACC age-yootosita no]-o totte tabeta give-tried NM-ACC took ate
c ?* Keeki-oi Taroo-ga [okaasan-ga Hanako-ni ti cake-ACC Taro-NOM mother-NOM Hanako-DAT age-yootosita no]-o totte tabeta give-tried NM-ACC took ate The sentence in (85)a has a HIRC in which the internal head is keeki ‘cake’. The non-head dative phrase, Hanako-ni ‘to Hanako’, can be scrambled, as shown in (85)b. This is expected under our analysis because the HIRC is a pure CP. Strikingly, however, the internal head, keeki, cannot be extracted out of the HIRC, as in (85)c. It is possible to think that the internal head must check its operator feature either overtly (as an invisible operator, as proposed by Watanabe (1992)) or covertly. In either case, the internal head in (85)c has to be lowered to the spec of CP, violating the Proper Binding Condition (Fiengo 1977 and others). Thus, the ungrammaticality in (85)c does not come from impossibility of scrambling out of the HIRC, but it can be explained in this general way. Moreover, if this line of analysis is correct, then extraction of the internal head within embedded clauses should be possible:27 (86) a
Taroo-ga [okaasan-ga [kinoo pro Hanako-ni keeki-o Taro-NOM mother-NOM yesterday Hanako-DAT cake-ACC ageta to] itteita no]-o totte tabeta gave COMP said NM-ACC took ate ‘Taro took and ate the cake which his mother said that she gave (it) to Hanako’
b
Taroo-ga [keeki-o okaasan-ga [kinoo pro Hanako-ni ti Taro-NOM cake-ACC mother-NOM yesterday Hanako-DAT ageta to] itteita no]-o totte tabeta gave COMP said NM-ACC took ate
The sentence in (86)a contains a HIRC where the internal head is embedded in the complement CP. The internal head, keeki-o ‘cake-ACC’, can be moved by scrambling as long as it stays within the HIRC. This fact can be explained by our analysis.
Nominalizations in Cleft Constructions 99
Summary This chapter has explored properties of no in presuppositional clauses of clefts, and proposed that it is neither a genitive Case nor a pro-form, but a complementizer. We further argued that presuppositional clauses are not dominated by a nominal head, based on the facts that the no does not have any semantic restrictions and adjective/NQ cannot modify presuppositional clauses. Although the presence of case conversion in presuppositional clauses might seem to indicate the presence of nominal clause, we observed that this is not in fact the case; ga-no conversion occurs in Wh-agreement configurations, following Watanabe (1996). Furthermore, the facts of extraction from a presuppositional clause indicate that there is no null N or NP dominating the presuppositional CP in the structure.
This page intentionally left blank
4 Resumptive A’-Dependencies
Introduction In Chapter 3, we examined the basic properties of presuppositional clauses in Japanese clefts. Let us now extend the domain of inquiry from simple cleft constructions to more complex cleft sentences. This chapter is particularly concerned with long-distance cleft constructions in Japanese. The main purpose here is to present a striking fact, namely, that, unlike English clefts, long-distance clefts in Japanese involve some kind of ‘resumptive A’-dependencies’; a fact which, as far as I know, has not been discussed in previous literature. I will argue that, in long-distance clefting, a null operator undergoes movement from the position adjoined to the highest embedded clause. The organization of this chapter is as follows. On pages 102–10, I briefly present the basic facts of English cleft constructions with respect to so-called ‘connectivity’ phenomena, and compare the data with corresponding Japanese cleft sentences. I propose that, unlike local cleft constructions, long-distance clefts have a peculiar structure that decomposes into two dependencies, where the higher dependency exhibits properties of movement but the lower one does not involve any movement. I go on to investigate reconstruction effects in long-distance clefts on pages 110–21. Further empirical evidence for the proposed analysis, such as weak crossover effects and interactions with another A’-dependency, will be discussed on pages 121–30. Following this, more evidence in favor of our proposed analysis will be presented on pages 130–45, including three-layered cleft constructions and clefts with true adjunct PPs. In addition, it will be noted that the proposed structure is also found in other languages, such as Modern Greek and 101
102 Cleft Constructions in Japanese Syntax
Selayarese. Pages 146–50 will deal with some other residual issues: island effects in long-distance cleft constructions, and structures for local clefts. Similar phenomena to long-distance cleft constructions, namely, phenomena of the highest clause sensitivity, will be presented in Japanese and other languages on pages 150–6. The discussion will be summarized at the end of Chapter 4.
Basic facts and proposal Connectivity in English cleft constructions As we saw in Chapter 1, one of the interesting discoveries in previous research on English (specificational) cleft constructions is that they exhibit the syntactic phenomenon of ‘connectivity’ (Higgins 1979, among many others, for pseudoclefts, and Pinkham and Hankamer 1975 and Percus 1997 for it-clefts).1 In other words, despite the apparent lack of a c-command relation between a binder in the presuppositional clause and its bindee in the focus position, a binding dependency is possible, as shown in (1)–(3): Referential binding (1) a It was a picture of herselfi that Maryi bought b What Maryi bought was a picture of herselfi Variable binding (2) a It was hisi own analysis that every studenti believed b What every studenti believed was hisi own analysis Licensing of negative polarity item (3) a* It was any good books that he didn’t buy b What he didn’t buy was any good books The sentences in (1)a and (1)b show that the anaphor, herself, can be bound by its antecedent, Mary, even though Mary does not c-command the anaphor in surface structure. Within the basic binding theory of Government and Binding (Chomsky 1981, 1986a, 1986b), we would not expect the grammaticality since the binding relation is not established under the theory. The same phenomenon is found in the case of variable binding in (2). Apparently, the quantified phrase, every student, does not
Resumptive A’-Dependencies 103
c-command the pronoun, and yet the bound variable reading is available in (2)a and (2)b. (3) demonstrates a slightly different distribution; the negative polarity item (NPI), any, is not licensed by the negative (NEG) head in the it-cleft sentence in (3)a but is licensed in the pseudocleft sentence in (3)b. Although the contrast between (3)a and (3)b should be explained, it is assumed that (3)a has an independent reason to be ruled out.2 Putting this problem aside, we can make the generalization that English cleft constructions observe connectivity (Peters and Bach 1968 and Akmajian 1970) or reconstruction effects (Chomsky 1976). A number of studies have been made of the English connectivity phenomena within the framework of generative grammar. Tracing back to the transformational approach, Peters and Bach (1968), Akmajian (1970), Chomsky (1970), Hankamer (1974), Pinkham and Hankamer (1975), Higgins (1979) among others have closely examined (pseudo)cleft constructions in terms of connectivity. Exact analyses differ among these studies, but the technical details of how to account for the connectivity under the transformational grammar will not be discussed here. It should be noted, however, that most of the analyses above assume that the underlying representations of (pseudo)clefts are similar to the corresponding simple sentences in (4) and derive the surface structures:3
(4)
a Maryi bought a picture of herselfi b Everyi boy believed hisi own analysis. c He didn’t buy any good books.
The sentences in (4) are paraphrased from the cleft constructions in (1)–(3), and in these underlying representations, the anaphor is bound by its antecedent satisfying the basic binding theory. In fact, within current linguistic theories, Heycock and Kroch (1996) and Boškovic (1997) also propose that the LF representations of the pseudocleft constructions are something like the ones in (4), assuming that LF is the relevant level of binding relations. Percus (1997) proposes a similar analysis to Akmajian (1970) although the structure proposed by Percus contains a null operator within a presuppositional clause. I will not go into details of these analyses here, and will focus on Japanese cleft constructions within the approach similar to Heggie (1988, 1993) shown below.
104 Cleft Constructions in Japanese Syntax
Within the Government and Binding theory, Heggie (1989, 1993), based on Chomsky (1977) and Barss (1986), proposes the following structures for it-clefts and pseudoclefts: (5)
a
[IP it [VP be [CP XPi [CP Opi thati [IP . . . ei . . .]]]]] (Chomsky 1977)
b [CP [CP Opi . . . ei . . .]j [IP be [SC XPi tj]]] (Heggie, 1988) The schematic structures in (5)a and (5)b represent the one for it-clefts and pseudoclefts, respectively. This analysis can account for the connectivity if we assume that a focus element is linked to its antecedent by coindexation through a null operator. Thus, previous analyses of English cleft constructions assume either that a direct binding relation between a focus element and its antecedent is possible at some relevant level of derivation, or that it is licensed by coindexation with predication theories. I will examine connectivity phenomena under the latter line of approach.4 Let us now turn to long-distance cleft constructions in terms of connectivity. Like a simple cleft construction, reconstruction to a thematic position is possible but not necessarily in long-distance clefting in English, as shown in the following sentences involving referential binding: (6)
a
It was [a picture of himselfi/j] that Maxi thought that Johnj bought [e]
b What Maxi thought that Johnj bought [e] was [a picture of himselfi/j] (6)a is an it-cleft construction and (6)b a pseudocleft sentence with the same meaning. In these examples, the anaphor in focus position can be coindexed with either the higher subject, Max, or the lower one, John. Assuming that the cleft sentences involve null operator movement within the presuppositional clause (Chomsky 1977, Barss 1986, Heggie 1988) and the operator moves successive-cyclically, there is a trace of the operator not only in its theta position (the object position of bought) but also in the intermediate position (spec of the embedded CP: that Max bought).5 Therefore, the binding relation with the lower trace yields the interpretation of Max as the antecedent, and the one with the higher trace yields the interpretation of John as the antecedent.6 Connectivity in Japanese cleft constructions Let us now consider the corresponding Japanese cleft constructions with respect to connectivity phenomena.7 As the following examples show, we
Resumptive A’-Dependencies 105
initially find similar behavior in Japanese cleft constructions as in English: (7) a [Hanakoi-ga e atta no]-wa zibuni–no hahaoya-ni dat-ta Hanako-NOM met NM-TOP self-GEN mother-DAT COP-PAST ‘Who Hanakoi met was herself’si mother’ ‘It was herself’si mother that Hanakoi met’ b [Hanakoi-ga e hanasita no]-wa zibun-nituitei dat-ta Hanako-NOM talked NM-TOP self-about COP-PAST ‘Who Hanakoi talked about was herselfi’ ‘It was about herselfi that Hanakoi talked’ (8) a [Daremoi-ga e saisho-ni au no]-wa zibuni-no everyone-NOM first time meet NM-TOP self-GEN hahaoya-ni da mother-DAT COP ‘Who everyonei meets first is hisi own mother’ ‘It is hisi own mother that everyonei meets for the first time’ b [Daremoi-ga e hanasita no]-wa zibuni-nituite da-ta everyone-NOM talked NM-TOP self-about COP-PAST ‘What everyonei talked about was himselfi’ ‘It was about himselfi that everyonei talked’ (9)* [Taroo-ga e aw-ana-katta no]-wa dono sensei-ni-mo Taro-NOM meet-NEG-PAST NM-TOP any teacher-DAT-also da-ta COP-PAST ‘Who Taro didn’t meet was any teacher’ The sentences in (7) are cases of referential binding, (8) variable binding, and (9) licensing of a NPI. As we saw in English it-clefts, while referential/variable binding remains in the cleft constructions, as shown in (7) and (8), the NPI is not licensed, as in (9). The reason why sentences such as (9) are ruled out was already discussed in Chapter 2; the NPI does not coincide with the properties of null operators. The difference between the examples in (a) and those in (b) is that (a) contains a dative NP in the focus position whereas (b) has a PP focus phrase. The sentences in (7) and (8) suggest that in the local environment, both NPs and PPs exhibit connectivity just like English cleft constructions.
106 Cleft Constructions in Japanese Syntax
Although the same connectivity is observed in Japanese cleft constructions as in English in (7) and (8), a difference between Japanese and English clefts emerges clearly when one looks at the following longdistance cleft constructions:8,9 (10) a Tarooi-ga [Hanakoj-ga zibuni/j-no hahaoya-ni atta to] Taro-NOM Hanako-NOM self’s mother-DAT met COMP sinziteita believed ‘Taroi believed that Hanakoj met self’si/j mother’ b [Tarooi-ga [Hanakoj-ga e atta to] sinziteita-no]-wa Taro-NOM Hanako-NOM met COMP believed-NM-TOP [zibuni/*j-no hahoya-ni dat-ta self’s mother-DAT COP-PAST ‘It was self’si/*j mother that Taroi believed that Hanakoj met’ (11) a Tarooi-ga [Hanakoj-ga zibun-nituitei/j hanasita to] Taro-NOM Hanako-NOM self-about talked COMP omotta thought ‘Taroi thought that Hanakoj talked about selfi/j’ b [Tarooi-ga [Hanakoj-ga e hanasita to] omotta no]-wa Taro-NOM Hanako-NOM talked COMP thought NM-TOP zibun-nituitei/*j dat-ta self-about COP-PAST ‘What Taroi thought that Hanakoj talked about was selfi/*j’ ‘It was about selfi/*j that Taroi thought that Hanakoj talked’ (10)a has a complement clause containing the anaphor, zibun ‘self’. Zibun is known as a subject-oriented, long-distance anaphor in the literature (Kuroda 1965, Kuno 1973, Katada 1991 and others), and thus, in this sentence, zibun can refer to either the matrix subject (Taroo) or the embedded subject (Hanako).10 Interestingly, if the anaphor in the complement clause is clefted, as illustrated in (10)b, the sentence loses one of the possible interpretations: ‘*Taro believed that Hanako accused herself’, and zibun can be coindexed only with Taroo in the matrix subject position. The same contrast can be observed in (11), in which the focus phrases are PPs rather than NPs.11 This fact is striking if we assume that Japanese cleft constructions are analyzed on a par with English cleft constructions. Unlike English connectivity facts, it seems that the focus phrase is not reconstructed to its theta position in Japanese long-distance clefts.
Resumptive A’-Dependencies 107
The same is true for variable binding in long-distance clefts. Let us look at the examples below:
(12)
a
Sono gakushai-ga [daremoj-ga zibuni/j-no bunseki-ni that scholar-NOM everyone-NOM self’s analysis-DAT tayotteiru to] itta count-on COMP said ‘The scholari said that everyonej counted on hisi/j own analysis’
b
[Sono gakushai-ga [daremoj-gae tayotteiru to] that scholar-NOM everyone-NOM count-on COMP dat-ta itta-no]-wa zibuni/*j-no bunseki-ni said-NM-TOP self’s analysis-DAT COP-PAST ‘What the scholari said that everyonej counted on was hisi/*j own analysis’ ‘It was hisi/*j own analysis that the scholari said that everyonej counted on’
The sentence in (12)a shows that zibun can be bound either by the referential expression, sono gakusha ‘the scholar’, or by the quantifier, daremo ‘everyone’. However, if you cleft the embedded object containing zibun, the anaphor cannot refer to the embedded subject, as shown in (12)b. Note that, as discussed in Chapter 2, we argued that a cleft sentence is closely related to a topic construction. If this is correct, we should find the same connectivity phenomena involving topicalization. This is, in fact, borne out, as shown below:
(13) a Zibuni/?*j-no hahaoya-ni-wa [Tarooi-ga [Hanakoj-ga e self’s mother-DAT-TOP Taro-NOM Hanako-NOM atta to] sinziteiru] met COM believe ‘As for self’s mother, Taro believes that Hanako met (her)’ b Zibuni/?*j-nituite-wa [Tarooi-ga [Hanakoj-ga e hanasita self-about-TOP Taro-NOM Hanako-NOM talked to] omotteiru] COMP think ‘About himself, Taro thinks that Hanako talked’
108 Cleft Constructions in Japanese Syntax
(14)
Zibuni/?*j-no bunseki-ni-wa [sono gakushai-ga self’s analysis-DAT-TOP that scholar-NOM [daremoj-ga e tayotteiru to] itta] everyone-NOM count-on COMP said ‘As for his own analysis, that scholar said that everyone counted on’
While the judgements in (13) and (14) may be subtle, it seems difficult for zibun ‘self’ to be bound by the embedded subject. We will focus on the data of cleft constructions in this chapter, but the facts here suggest that long-distance topic constructions are analysed on a par with longdistance cleft constructions. Considering more complex cleft constructions, namely, those which have two embedded clauses, we find that reconstruction takes place somewhere between the matrix clause and the higher embedded clause: (15)
a
b
Tarooi-ga [Hanakoj-ga [Masaok-ga zibuni/j/k-no Taro-NOM Hanako-NOM Masao-NOM self’s hahaoya-ni atta to] itta to] omotteiru mother-DAT met COMP said COMP think ‘Taroi thinks that Hanakoj said that Masaok met self’si/j/k mother’ [Tarooi-ga [Hanakoj-ga [Masaok-ga e atta to] itta Taro-NOM Hanako-NOM Masao-NOM met COMP said to] omotteiru-no]-wa zibun-noi/*j/*k hahaoya-ni COMP think-NM-TOP self’s mother-DAT dat-ta COP-PAST ‘It was self’si mother that Taroi thinks that Hanako said that Masao met’
The sentence in (15)a contains the anaphor, zibun ‘self’, in the most embedded object position. Due to the properties of the anaphor, it refers to either Taro, Hanako or Masao. However, when zibun is clefted from this three-layered structure, it can only be coindexed with the matrix subject, Taro, as shown in (15)b. This fact narrows down our generalization. The intermediate position is not somewhere between the two embedded clauses but somewhere below the matrix subject and higher than the second subject. This will be discussed more on pages 130–4. Reconstruction effects have been analyzed by two distinct ways in the literature. One is that reconstruction effects are observed even in non-movement dependencies (Cinque 1990), and the other is that they are observed only when movement is involved (Chomsky 1976, Aoun
Resumptive A’-Dependencies 109
1986, Barss 1986 among many others). Let us simply assume the latter view at the moment, but we will see empirical evidence later in this chapter which clearly shows that reconstruction effects are related only to movement. Assuming that reconstruction effects indicate movement, we find that the theta position within the embedded clause in long-distance clefts should not be occupied by a trace created by movement. In other words, movement of a null operator does not start from its theta position, but from some intermediate position bound by the matrix subject; the generalization we have seen in the data above is that an anaphoric pronoun in focus position does not reconstruct to its thematic position but reconstructs to a (higher) intermediate position. Proposal Here, I propose that Japanese cleft constructions are divided into two distinct structures: one is for local clefting and the other is for longdistance clefting. Based on the standard analysis for English cleft constructions, these structures are illustrated in (16) followed by schematic structures in (17): (16)
a
[CP Opi [IP Hanako-ga ti atta]-no]-wa [NP sono hito-ni]i Hanako-NOM met-NM-TOP that person-DAT dat-ta COP-PAST ‘It was that person who Hanako met’
b
[CP Opi [IP Taroo-ga [CP ti [CP Hanako-ga proi atta to]] Taro-NOM Hanako-NOM met COMP omotteita]-no] -wa [NP sono hito-ni]i dat-ta thought -NM-TOP that person-DAT COP-PAST ‘It was that person that Taro thought that Hanako met’
(17) a
TopP CP-top Opi
IP I’
IP
...ti...
VP XPi
I
110 Cleft Constructions in Japanese Syntax
b
TopP CP-top Opi
IP
IP ...CP... ti
CP
I’ VP
I
XPi
proi
(16)a is the case of a local cleft and (16)b a long-distance cleft. Op stands for a null operator which is closely related to a topic phrase in Japanese (see Chapter 2). Both local and long-distance cleft constructions are schematized as (17)a and (17)b, respectively. The XPs within the matrix VP are focus phrases coindexed with the operator by agreement or feature sharing (see Chapter 2). I argue that both local and long-distance clefts always involve movement of a null operator; however, unlike local clefts, long-distance clefts contain an operator base-generated in a position adjoined to the highest embedded CP.12 Then, it undergoes movement to the spec of CP to check its feature. Thus, the proposal here contrasts with Hoji (1990) in the sense that his dichotomy of distinct derivations for clefts basically comes from whether the focus phrase is a bare NP or not. The current proposal, on the other hand, implies that the dichotomy comes from whether clefting is local or long-distance. In the following sections, I will argue that syntactic facts of cleft constructions in Japanese can be explained naturally under our proposal in (17).
Reconstruction in long-distance clefts The present section turns to the connectivity phenomena of referential and variable binding in Japanese cleft constructions. The facts given here will lead to the proposal of a striking structure for long-distance clefts in Japanese, which, as mentioned in the previous section, exhibits some kind of resumptive A’-dependencies. It will also be examined whether anaphoric pronouns in Japanese can be interpreted as logophoric pronouns. Binding relations The first piece of evidence for resumptive A’-dependencies in Japanese long-distance clefts is that a focus phrase seems to be reconstructed to an intermediate position for purposes of binding relations. First, let us
Resumptive A’-Dependencies 111
look at an instance of local clefting again. The sentence in (10)a is reproduced as in (18)b: (18)
a
Hanakoi-ga zibuni-no hahaoya-ni atta Hanako-NOM self’s mother-DAT met ‘Hanako met self’s mother’
b
[Hanakoi-ga e atta -no] -wa zibuni-no hahoya-ni mother-DAT Hanako-NOM met-NM-TOP self’s dat-ta COP-PAST ‘It was self’s mother that Hanako met’
Sentence (18)a is a simple sentence where the object anaphor is bound by its antecedent in the subject position. Even if the object is clefted as represented in (18)b, still the anaphor in focus position becomes coindexed with the subject in the presuppositional clause. Thus, the focus phrase, or more specifically, the corresponding operator is reconstructed into its theta position in the presuppositional clause. Interestingly, cleft sentences with a complement clause exhibit distinct behavior, as we observed before. The following sentences show that the null operator cannot reconstruct to its thematic position inside the complement clause: (19)
a
b
Tarooi-ga [Hanakoj-ga Zibuni/j-no hahaoya-ni atta to] Taro-NOM Hanako-NOM self’s mother-DAT met COMP sinziteiru believe ‘Taro believes that Hanako met (him/her)self’s mother’ [Tarooi-ga [Hanakoj-ga e atta to] sinziteiru-no]-wa Taro-NOM Hanako-NOM met COMP believe-NM-TOP [zibuni/?*j-no hahaoya-ni] dat-ta self’s mother-DAT COP-PAST ‘It was (him/*her)self’s mother that Taro believes that Hanako met’
In (19)a, the anaphor in the embedded object position can be bound by either the embedded subject (Hanako) or the matrix subject (Taroo), since zibun ‘self’ is subject-oriented and bound by its long-distance antecedent, as we already saw in the previous section. If we adopt the analysis for connectivity observed in English cleft constructions, we
112 Cleft Constructions in Japanese Syntax
would predict that clefting the embedded object should not change the interpretation that we get in the simple sentence. However, as shown in (19)b, the clefted anaphor can only refer to the matrix subject and not the embedded subject. Hence, it is reasonable to assume that the null operator associated with its focus phrase is reconstructed to a position that the matrix subject c-commands but the embedded subject does not. Notice that Japanese has other types of anaphors: zibun-zisin ‘self-self’ and kare-zisin/kanojo-zisin ‘himself/herself’. The former is known as a subject–oriented local anaphor, and the latter is a local anaphor with no orientation (Katada 1991, and others). First, let us consider the following sentences containing zibun-zisin ‘self-self’. Surprisingly, the same connectivity effect as we saw in (19) holds, even though the binding relation in simple sentences is different: (20)
a
Tarooi-ga [Hanakoj-ga zibun-zisin*i/j-ni yasasi-sugiru to] Taro-NOM Hanako-NOM self-self-DAT be-too-kind COMP omotteiru think ‘Taro thinks that Hanako is too kind to her-/*himself’
b [Tarooi-ga [Hanakoj-ga e yasasi-sugiru to] Taro-NOM Hanako-NOM be-too-kind COMP omotteiru-no]-wa zibun-zisini/*j-ni dat-ta think-NM-TOP self-self-DAT COP-PAST ‘It was himself that Taro thought that Hanako was too kind to’ The anaphor in the sentence in (20)a can refer to Hanako but not Taro. Zibun-zisin ‘self-self’ is a local anaphor, so that Taro cannot be coindexed with the anaphor in the embedded clause in the simple sentence. However, as shown in (20)b, the binding relation in the sentence is reversed when the anaphor is clefted; zibun-zisin ‘self-self’ can refer to Taro and not Hanako. Therefore, it seems that the null operator associated with the focus phrase is reconstructed to somewhere between the matrix subject and the embedded subject at the relevant level of interpretation. Next, let us look at the examples containing kare-zisin ‘himself’. The following sentences make the same point: (21)
a
Tarooi-ga Zirooj-ni [Masaok-ga kare-zisin*i/*j/k-ni Taro-NOM Ziro-DAT Masao-NOM himself-DAT yasasi-sugiru to] osie-ta be-to-kind COMP told ‘Taroi told Ziroj that Masaok was too kind to himself*i/*j/k’
Resumptive A’-Dependencies 113
b
[Tarooi-ga Zirooj-ni [Masaok-ga e yasasi-sugiru to] Taro-NOM Ziro-DAT Masao-NOM be-to-kind COMP osie-ta-no]-wa kare-zisini/?j/*k -ni da-ta told-NM-TOP himself -DAT COP-PAST ‘It was himself that Taro told Ziro that Masao was too kind to’
The sentence in (21)a shows that the anaphor, kare-zisin ‘himself’, can be bound only by the embedded subject, Masao, because kare-zisin ‘himself’ is a local anaphor. (21)b, where the anaphor is clefted, changes the reading of the binding relation; kare-zisin ‘himself’ in the focus position does not refer to the embedded subject, Masao, but can be coindexed either with the indirect object, Ziroo, or with the matrix subject, Taroo.13 Again, this fact would not be expected if we assume that the null operator is reconstructed to its theta position like English cleft constructions. Thus, the generalization we saw in zibun ‘self’ and zibun-zisin ‘self–self’ binding with respect to connectivity is also obtained in examples with kare-zisin ‘himself’. One might wonder, by looking at (18), (19) and (20), whether it is posssible to think that an anaphoric pronoun in focus position is directly bound by the highest subject in cleft constructions. Note, however, that the fact that the kare-zisin ‘himself’ can be coindexed with Ziroo in (21)b suggests that the focus position is not directly bound by the subject. We will come back to this issue in the next subsection. Not surprisingly, the same facts can be observed in cleft sentences with a bound variable, which confirm our generalization about long–distance cleft constructions in Japanese: (22)
a
Daremoi-ga [sono gakusyaj-ga zibuni/j-no bunseki-ni everyone-NOM that scholar-NOM self-GEN analysis-DAT tayori-sugiteiru to] it-ta count-on-too-much COMP said ‘Everyonei said that the scholarj counted on hisi/j analysis too much’
b? [Daremoi-ga [sono gakusyaj-ga e tayori-sugiteiru everyone-NOM that scholar-NOM count-on-too-much to] itta-no]-wa [zibuni/*j-no bunseki-ni] da-ta COMP said-NM-TOP self-GEN analysis-DAT COP-PAST ‘What everyonei said that the scholarj counted on too much was hisi/*j analysis’
114 Cleft Constructions in Japanese Syntax
(23) a
Sono gakusyai-ga [daremoj-ga zibuni/j-no bunseki-ni that scholar-NOM everyoneNOM self-GEN analysis-DAT tayori-sugiteiru to] it-ta count-on-too-much COMP said ‘That scholari said that everyonej counted on hisi/j analysis too much’
b? [Sono gakusyai-ga [daremoj-ga e tayori-sugiteiru to that scholar-NOM everyone-NOM count-on-too-much COMP itta-no]-wa [zibuni/*j-no bunseki-ni da-ta said-NM-TOP self-GEN analysis-DAT COP-PAST ‘What that scholari said that everyonej counted on too much was ‘hisi/*j analysis’ Since zibun ‘self’ can be interpreted as an anaphor or a bound variable (and can have a long-distance antecedent), zibun-no bunseki ‘self’s analysis’ in (22)a and (23)a is construed as either ‘the scholar’s analysis’ (referential binding) or ‘everyone’s analysis’ (variable binding). However, if you look at the corresponding cleft sentence in (22)b and (23)b, the same phrase in the focus position in cleft constructions loses one of the possible interpretations. It cannot mean ‘*the scholar’s analysis’ in (22)b and ‘*everyone’s analysis’ in (23)b. The distribution given here is exactly the same as the one we have seen so far; the null operator is not reconstructed in its theta position in the embedded clause but seems to be situated in the intermediate position. According to the facts shown above, we conclude that the LF representation for the presuppositional clause in long-distance cleft constructions is schematized as (24)b but not (24)a: (24) a*
Long-distance cleft [IP XPi [CP YPj . . . Opi/j . . .]]
b
[IP XPi [CP Opi/*j [CP YPj . . . (pro) . . .]]]
(24)a is a logical possibility and actually can be the LF representation for English long-distance clefts. However, the peculiar reconstruction facts cannot be accounted for by (24)a. Assuming that the structure at the level of binding interpretation is represented as something like (24)b, we can explain why the anaphoric focus pronoun can be only coindexed with the matrix element; the operator is base-generated in the position adjoined to the embedded clause, and since its thematic position is occupied by a pro, it cannot be reconstructed in the embedded clause.
Resumptive A’-Dependencies 115
A direct binding relation Another possibility to account for the data above is that the highest subject directly binds the focus phrase. This subsection will argue that this is not on the right track.14 Under the analysis that the highest subject binds a focus element directly, Taroo in (19)b is assumed to be adjoined to the no-clause, and c-commands zibun in focus position, as shown in (25):
(25)
a
b
[CP1 Tarooi-ga [CP2 ti [Hanakoj-ga e yasasi-sugiru to] Taro-NOM Hanako-NOM be-too-kind COMP omotteiru-no-wa zibuni/*j -ni datta think-NM-TOP self-DAT COP-PAST
TopP CP1
IP
Taroi CP2 ti....
I’ VP
I
zibuni
Taro in (25) c-commands the zibun if we assume that the highest subject is not dominated by the node that dominates all other nodes in the no-clause, adopting the category/segment distinction (May 1985).15 In other words, the CP1 does not dominate Taro in the adjoined position, and hence, Taro can c-command out of the CP. However, the following sentences show that this is not in fact the case:
(26)
a
Taroo to Hanakoi-ga [sensei-gataj-ga otagai*i/j-ni Taro and Hanako-NOM teachers-NOM each-other-DAT tayotteiru to] omotteiru rely-on COP think ‘Taro and Hanakoi think that the teachersj relied on each other*i/j’
116 Cleft Constructions in Japanese Syntax
b
Taroo to Hanakoi-ga [[sensei-gataj-ga e tayotteiru Taro and Hanako-NOM teachers-NOM rely-on no]-wa otagaii/j -ni dat-ta to] omotteiru COMP-TOP each-other-DAT COP-PAST COMP think ‘Taro and Hanakoi think that it was each otheri/j that the teachersj relied on’
The sentence in (26)a contains a complement clause, in which the reciprocal anaphor, otagai ‘each other’, can be coindexed with the local antecedent, sensei-gata ‘teachers’, but not with Taroo to Hanako ‘Taro and Hanako’. In contrast, if the anaphor is clefted within the complement clause, the sentence becomes ambiguous, as shown in (26)b. Assuming that otagai ‘each other’ can be coindexed with its closest antecedent which binds it, the fact in (26)b is not fully expected under the direct binding analysis. While sensei-gata ‘teachers’ is the closest antecedent for the anaphor, the matrix subject can also be coindexed with it. Our analysis can predict this fact; since the null operator associated with otagai moves to the spec of no-clause, the closest antecedent is the matrix subject. As for the other interpretation, the null operator reconstructs to its thematic position because this is a local cleft, and the embedded subject becomes the closest antecedent. Therefore, the direct binding analysis faces this difficulty whereas our analysis does not.16 Logophoricity As pointed out by Hiroshi Aoyagi and Mark Baker (personal communication) independently, there is a possibility that zibun ‘self’ in the focus position of long-distance cleft constructions is actually a logophoric pronoun. In other words, the linking between zibun in focus position and the matrix subject might not be licensed by syntactic binding but by pragmatically determined coreference.17 An anonymous reviewer also points out that we may be able to argue that when a reflexive is not c–commanded by a potential antecedent in surface sentences, it is interpreted as a discourse reflexive. While the issue has been controversial in the literature, I adopt the claim proposed by Abe (1997) that zibun is ambiguous between a logophoric pronoun and an anaphor. It will be argued that zibun in focus position in long-distance clefts is not an instance of a logophoric pronoun but is licensed by binding principles at the relevant level of representation. This is also supported by instances of cleft sentences with another anaphoric pronoun, such as soko ‘that (place)’. Unlike zibun, soko is argued to be a pure anaphor when it
Resumptive A’-Dependencies 117
is bound by a certain type of antecedent in Japanese (Hoji 1995). This will be discussed at the end of this subsection. According to Sells (1987), the fact that zibun ‘self’ can have a longdistance antecedent is not explained by parametrized binding conditions but should be accounted for by the discourse structure. The following sentences are cited from Sells (1987), and cannot be explained by the binding conditions:18 (27)
a
Takasii-wa Taroo-ni [Yosiko-ga zibuni-o nikundeiru Takasi-TOP Taro-DAT Yosiko-NOM self-ACC hate koto]-o hanasita fact-ACC told ‘Takasii told Taro that Yosiko hated selfi’
b
[Yosiko-ga zibuni-o nikundeiru koto]-ga Mitikoi-o Yosiko-NOM self-ACC hate fact-NOM Mitiko-ACC zetuboo e to oiyatta desperation drove ‘That Yosiko hated selfi drove Mitikoi to desperation’
c(?*)
Taroo-wa Takasii-kara [Yosiko-ga zibuni-o Taro-TOP Takasi-from Yosiko-NOM self-ACC nikundeiru koto]-o kiita hate fact-ACC heard ‘Taro heard from Takasii that Yosiko hated selfi’
The sentence in (27)a is a case of long-distance binding such as we have already seen in the previous sections. Since Condition A restricts the occurrence of an anaphor and its antecedent in a local domain, the fact here does not show Condition A effects. (27)b is an instance of backward binding, in which an anaphor precedes its antecedent and yet the sentence is grammatical. Because there is no c-command relation between the anaphor and its antecedent in this example, it could also be considered as violation of the binding conditions. Sells assumes that (27)c is grammatical under the coreference reading between Takasi and zibun. Because the antecedent is embedded inside the PP, it cannot c-command zibun, and hence, is a violation of the binding principle. These coreferential facts in (27) can be explained under the notion of discourse structure. Sells (1987) argues that there are three notions of logophoricity: ‘source’, ‘self’ and ‘pivot’. Looking at (27)a, we find that the ‘source’ of information is Takasi. In (27)b, ‘self’, which is the person with respect to whose consciousness the report is made, is Mitiko. The sentence in (27)c contains ‘pivot’, which is the person from whose point of view the report is made; Takasi is regarded as a pivot in the
118 Cleft Constructions in Japanese Syntax
sentence. Therefore, all the sentences in (27) are readily accounted for by the notions of logophoricity. Notice that in order to pursue the idea that zibun in the focus position is a logophoric pronoun, we need not assume any kind of reconstruction operation for the null operator. Since zibun in long-distance clefts, such as (19)b repeated as in (28), is not ambiguous, it must be interpreted in the focus position under this approach: (28)
[Tarooi-ga [Hanakoj-ga e atta to] sinziteiru-no]-wa Taro-NOM Hanako-NOM met COMP believe-NM-TOP [zibuni/?*j-no hahaoya-ni] dat-ta self’s mother-DAT COP-PAST ‘It was (him/*her)self’s mother that Taro believes that Hanako met’
The hypothesis that zibun in a certain position may not reconstruct is not new. Hasegawa (1988) points out that zibun does not exhibit reconstruction effects in non–restrictive relative clauses, as shown in (29) (the judgment is Hasegawa’s): (29)*
[Johni-ga e taipu-sita] zibuni-no ronbun John-NOM typed self-GEN paper ‘self’s paper that John typed’
In (29), zibun in the head of the relative clause cannot refer to John, which is the subject of the relative clause. Assuming that reconstruction effects are observed depending on types of anaphoric pronouns, it might be reasonable to say that zibun is the kind of pronoun which does not reconstruct. However, the logophoric approach encounters the following potential problem. Recall the facts, discussed in the last subsection, that when a cleft sentence is embedded in a complement clause, an anaphoric pronoun in focus position can be coindexed with the matrix subject. The following examples show that zibun in focus position still has an ambiguous interpretation when the cleft sentence appears in the complement clause: (30)
a
Tarooi-ga [[Hanakoj-ga e atta -no]-wa zibuni/j -no Taro-NOM Hanako-NOM met NM-TOP self’s hahaoya-ni dat-ta to] omotteiru mother-DAT COP-PAST COMP think ‘Taroi thinks that it was selfi/j’s mother that Hanakoj met’
Resumptive A’-Dependencies 119
b Masaoi-ga [[Tarooj-ga [Hanakok-ga e atta to] Masao-NOM Taro-NOM Hanako-NOM met COMP omotteiru-no]-wa [zibuni/?j/*k-no hahaoya-ni] dat-ta think-NM-TOP self’s mother-DAT COP-PAST to] itta COMP said ‘Masaoi said that it was selfi/j/*k’s mother that Taroj thought that Hanakok met’ (30)a contains a local cleft sentence and (30)b a long-distance cleft sentence. While the judgments are not very clear in the sentences in (30), it appears that zibun in (30)a and (30)b are ambiguous; zibun can refer to either the matrix subject outside the cleft sentence or the highest subject within the cleft sentence. The logophoric analysis may explain why zibun refers to the matrix subject, because the matrix subject is considered a ‘source’ of information. However, it is difficult for the analysis to account for why it can also be associated with the highest subject in the cleft. Although there might be a way to explain the cases of long-distance zibun with ambiguous reading under the logophoric analysis, it is not clear at this point.19 Moreover, it should be noted that the reconstruction effects similar to these in cleft sentences containing zibun can be observed in soko ‘it, that (place)’ binding, which is considered a true instance of bound variable anaphora with an appropriate antecedent (Hoji 1995). According to Hoji (1995: 256–7), referential dependency between soko and a conjoined NP cannot be one of coreference but must be that of binding.20 In other words, when soko is referentially linked to a quantified and/or conjoined NP, the anaphoric interpretation requires formal licensing rather than some other discourse factors. For example: (31)
a*
dokoi-o uttaeta no? Sokoi-no kumiai-ga it-GEN labor-union-NOM where-ACC sued Q ‘Which institution did its labor union sue?’
b
Dokoi-ga sokoi-no kumiai-o uttaeta no? where-NOM it-GEN labor-union-ACC sued Q ‘Which institution sued its labor union?’
The sentence in (31)a is ruled out under the coindexed reading between soko ‘it’ and doko ‘where’. This ungrammaticality is explained as a weak
120 Cleft Constructions in Japanese Syntax
crossover effect, which we will discuss in the context of cleft constructions in the next section. On the contrary, when soko is c-commanded by doko, as shown in (31)b, the coindexed interpretation is allowed. Notice that soko also allows an ‘accidental’ coreference reading if its antecedent is not a quantifier, as shown in (32): (32)
Sokoi-no kumiai-ga sono kaisyai-o uttaeta it-GEN labor-union-NOM that company-ACC sued ‘Its labor union sued that company’
(32) is grammatical under the interpretation that soko ‘it’ refers to sono kaisha ‘that company’. Comparing the sentence in (32) with that in (31)a, we find that soko must be bound by its antecedent if it is a quantifier. Hoji (1995) also argues that a conjoined NP is interpreted distributively when it binds soko ‘it, that place’, because in general, soko ‘it, that place’ cannot take a plural antecedent, and thus, when it is coindexed with a conjoined NP, it forces distributed reading on the conjunction. In this sense, a conjoined NP behaves like a quantifier. Let us look at the following example: (33) [Toyota to Nissan]i-ga Mazda-ni [CP CIA-ga Toyota and Nissan-NOM Mazda-DAT CIA-NOM sokoi-o sirabeteiru to] tugeta it-ACC is-investigating COMP told ‘Each of [Toyota and Nissan]i told Mazda that CIA was investigating iti’ In (33), soko in the embedded clause is bound by the conjoined subject, Toyota to Nissan ‘Toyota and Nissan’ in the matrix clause, yielding the bound variable interpretation. Soko has a singular denoting nature, which cannot allow split antecedents, and observes weak crossover effects. Therefore, the association between soko and Toyota to Nissan in (33) must require a formal dependency, namely, binding. Let us consider soko binding in long-distance cleft constructions. (34)a is a non-cleft sentence, in which soko appears inside the embedded object position. Soko in (34)a can be associated with either the matrix subject, subete-no sinbunsha ‘all newspaper offices’, or the embedded subject, Toyota to Nissan ‘(each of) Toyota and Nissan’. Interestingly, if the object phrase containing soko is clefted, the ambiguity disappears, as shown in (34)b:21
Resumptive A’-Dependencies 121
(34)
a
Subete-no sinbunshai-ga [[Toyota-to Nissan]j-ga all-GEN newspaper-office-NOM Toyota and Nissan-NOM sokoi/j-no keiretu-gaisya-ni tayori-sugiteiru to] it-GEN subsidiary-DAT rely-on-too-much COMP hoodoo-sita reported ‘All newspaper officesi reported that each of [Toyota and Nissan]j relied on itsi/j subsidiary company too much’
[[Toyota-to Nissan]j-ga e b [Subete-no sinbunshai-ga all-GEN newspaper-office-NOM Toyota and Nissan-NOM tayori-sugiteiru to] hoodoo-sita no]-wa sokoi/*j-no rely-on-too-much COMP reported NM-TOP it-GEN keiretu-gaisya-ni da subsidiary-DAT COP ‘It is itsi/*j subsidiary company that all newspaper officesi reported that each of [Toyota and Nissan]j relied on too much’ In (34)b, soko in focus position cannot be linked to the conjoined NP, whereas the coindexed reading between the soko and the matrix subject is possible. The facts in (34) can be readily accounted for if we assume that the operator associated with the focus phrase undergoes movement from the position adjoined to the embedded CP. Since the focus phrase reconstructs to this intermediate position, soko can be bound by the matrix subject but not by the embedded conjoined NP.22 To summarize so far, we have proposed that null operator movement in long-distance clefts starts from a position adjoined to the embedded clause, based on the ‘partial’ reconstruction effects. We have also observed that these effects cannot in general be explained by having the anaphoric focus phrase interpreted as a logophoric pronoun; a soko-phrase for example must be licensed by binding.
Properties of A’-movement In addition to the binding facts we observed in the last section, there are further pieces of evidence for our proposal concerning long-distance cleft constructions. This section focuses on weak crossover (WCO) effects and some interaction with other A’-dependencies. It will be shown that higher dependencies in long-distance clefts exhibit properties of A’-movement. As for local cleft constructions concerning these phenomena, we will come back to them on page 148.
122 Cleft Constructions in Japanese Syntax
Weak crossover effects We already observed that Japanese exhibits weak crossover (WCO) effects in Chapter 2 and in the last section in this chapter. As has been discussed since Saito and Hoji (1983), Japanese WCO phenomena are exemplified in (35): (35)
a
b??
Hanako-wa [Taroo-ga proi yomu maeni] sudeni sono Hanako-TOP Taro-NOM read before already that honi-o yondeita book-ACC read ‘Hanako already read the book before Taro read (it)’ Hanako-wa [Taroo-ga proi yomu maeni] sudeni nanii-o Hanako-TOP Taro-NOM read before already what-ACC yodeita no? read Q ‘What did Hanako already read before Taro read (it)?’
The sentence in (35)a contains a null pronoun within the adjunct clause, which is coindexed with the object in the matrix clause. If the object is replaced by a Wh-constituent, nani-o ‘what-ACC’, as shown in (35)b, then the grammaticality of the sentence becomes degraded. A widely assumed descriptive generalization for WCO effects is roughly the following, according to Lasnik and Stowell (1991: 690):23 (36)
In a configuration where a pronoun P and a trace T are both A’-bound by a category C, T must c-command P.
Lasnik and Stowell also claim that whatever the principles responsible for WCO are, it is crucial that they apply only to variables that are locally bound by true quantifiers. The marginality of the sentence in (35)b is explained by the generalization in (36) along with this claim. Since the true quantifier, nani ‘what’, moves to the spec of CP at LF, or, if Watanabe (1992) is correct, an invisible operator undergoes movement before spell-out, it binds the coindexed pro as well as its trace, but the trace does not c-command the pro. Therefore, (35)b becomes marginal. Let us now consider a long-distance cleft sentence with respect to WCO effects. Provided that a null operator is base-generated in a position adjoined to the embedded clause, the sentence would not exhibit WCO effects when an adjunct clause containing a pro is generated within the
Resumptive A’-Dependencies 123
embedded clause. This is in fact borne out, as shown in the following long-distance cleft sentence: (37)
a
[Hanako-ga [kinoo pro [AdvP Taroo-ga proi atta toki] Hanako-NOM yesterday Taro-NOM met when sudeni proi attako-koto-ga-aru to] omotta-no] -wa already had-met COMP though-NM TOP darei-ni-na-no? who-DAT-COP-Q ‘Who is it that Hanako thought that (Taro) had already met when Taro met (him/her) yesterday?’
b
CP1
Opi
C’
IP
C NM
Hanako
I’
VP
CP2
I
V thought
ti
CP2 C
IP yesterday
IP
Taro
that
VP AdvP
..proi..
VP proi
V had-met
124 Cleft Constructions in Japanese Syntax
In the sentence in (37)a, the adverbial clause contains a null pronoun which is coindexed with the null operator (and thus linked to the focus Wh-constituent). Looking at the schematized structure for the presuppositional clause in (37)b, the null operator is generated in the position adjoined to the embedded CP2 under our analysis for longdistance clefts. In other words, a movement dependency is only observed between the two elements in boldface type in (37)b. The pro in question occurs inside the AdvP within the CP2, and therefore, the sentence does not fail to conform to the generalization in (36). This suggests that the topic operator which is base-generated in the position adjoined to the embedded clause c-commands the coindexed pro in the adjunct clause in (37)a. Interestingly, the corresponding non-cleft sentence does exhibit a WCO effect, as shown in (38): (38)??
Hanako-ga [CP kinoo pro [Taroo-ga proi atta toki sudeni Hanako-NOM yesterday Taro-NOM met when already atta-koto-ga-aru to] omotteiru-no? dare-nii who-DAT had-met COMP think-Q ‘Who does Hanako think that (Taro) already knew when Taro met (him/her) yesterday?’
Since at LF, the Wh-phrase moves to the spec of CP (or an invisible operator moves to the position before spell-out, as argued by Watanabe (1992) and Tanaka (1997)), its trace does not c-command the pro; hence, marginality is observed. Furthermore, we predict that if the adjunct clause containing a pro appears in the matrix clause in long-distance clefts, then a WCO effect should be obtained under our analysis. The sentence in (39) demonstrates exactly this: (39)
a??
[[Taroo-ga proi atta toki] Hanako-ga [kinoo pro Taro-NOM met when Hanako-NOM yesterday omotteiru-no]-wa sudeni proi atta-koto-ga-aru to] already had-met COMP think-NM-TOP darei-ni-na-no? who-DAT-COP-Q ‘Who is it that Hanako thought that (Taro) already knew when Taro met (him/her)?’
Resumptive A’-Dependencies 125
b
CP1 C’
Opi IP AdvP proi
C NM IP VP
Hanako
CP2 ti
V
thought CP2 IP
C
that yesterday IP pro (Hanako)
VP proi
V knew
The sentence in (39)a has the presuppositional clause schematized in (37)b. The structure shows that the adjunct clause in question which contains a pro appears in the matrix clause. The judgment for the sentence in (39)a might not be so clear; however, it does not contrast with the marginality in (38) and is clearly worse than (37)a. This is explained by our analysis; since, as shown in (39)b, the null operator moves from the position adjoined to CP2, it crosses the AdvP containing the pro coindexed with it. Although (39)a by itself can also be explained even if we assume that the theta position in the complement clause is occupied by a trace, the contrast between (37) and (39) suggests that the movement of the null operator takes place above the complement clause. Here, one might wonder why Japanese cleft constructions observe WCO effects at all under the assumption that Japanese clefts are closely related to topicalization (see Chapter 2). Since English topicalization is a case of ‘weakest crossover’ (Lasnik and Stowell (1991: 697)), no WCO effects are observed in the following instance of topic constructions: (40)
Johni [Opi [I believe hisi mother loves ti]]
Recall that Lasnik and Stowell claim that WCO effects emerge when variables are locally A’-bound by true quantifiers. In (40), because the
126 Cleft Constructions in Japanese Syntax
null operator associated with the referential NP, John, is not considered a true quantifier, we do not observe WCO effects.24 However, Lasnik and Stowell (1991) also discuss cleft constructions in English with respect to WCO effects; they argue that both it-clefts and pseudoclefts do not usually exhibit WCO effects; however, when the focus phrases are quantificational, they do observe the WCO. The following examples are instances of pseudoclefts, cited from Lasnik and Stowell (1991: 715): (41)
a?
Whoi hisi mother was talking about ei was Johni
b*
Whati itsi author read ti to me was nothingi
The difference between these sentences is that the sentence in (41)a contains a referential focus phrase whereas the one in (41)b is a quantifier: the former does not observe WCO effects and the latter does. We can propose that the quantificational status of the null operator is determined by that of the focus phrase which shares its feature with the associated null operator. Therefore, in order to observe WCO effects, it is crucial that a null operator is associated with a true quantifier in focus position.25 In fact, if the Wh-phrase in focus position in (39) is replaced by an R-expression, the sentence becomes grammatical: (42)
[CP1 Opi [IP1 [Taroo-ga proi atta toki] Hanako-ga [CP2 ti [IP2 kinoo pro Taro-NOM met when Hanako-NOM yesterday sudeni proi atta-koto-ga-aru to]] omotteiru-no]]-wa sono already had-met COMP think-NM TOP that hitoi-ni-da person-DAT-COP ‘It was that person that Hanako thought that (Taro) had already met when Taro met (him/her)’
The long-distance cleft sentence in (42) contains sono hito ‘that person’, instead of dare ‘who’, in its focus position. Although the sentence in (39) exhibits a WCO effect, this sentence does not, because, as we already discussed in Chapter 2, the operator is associated with a non-quantificational phrase by agreement or feature sharing. This fact corresponds to the case in English pseudoclefts, in which a quantified focus but not an R-expression induces the WCO. The data observed in this subsection tell us that the theta position of the focus phrase is not occupied by a trace. In other words, the A’-dependency
Resumptive A’-Dependencies 127
in long-distance clefts is not a uniform chain but involves some kind of resumptive pronoun.26 Interaction with another A’-dependency Another piece of evidence for the structure of long-distance cleft constructions which involve resumptive A’-dependencies is pointed out by Hidekazu Tanaka (personal communication). According to Tanaka (1997, 1998), the A’-dependency lines between two operators and their licensing heads must not cross at S-structure.27 His argument is based on NPI constructions in Japanese. For example: (43)
a?*
Taroo-sika nani-o yoma-nai no Taro-only what-ACC read-NEG Q ‘What does only Taro read?’
b?*
Taroo-sika nani-o yoma-nai no
The sentence in (43) contains two operators: Taroo-sika ‘Taro-only’ and nani-o ‘what-ACC’, which must be licensed by their designated heads. Based on Tanaka’s analysis, (43)a is represented as having two A’-association lines. This sentence is ruled out because, as illustrated in (43)b, two A’-dependency lines (sika-nai ‘only-NEG’ and nani-no ‘what-Q’) are crossed at S-structure. Tanaka suggests that the ungrammaticality of sentence (43)a is explained under the Linear Crossing Constraint (LCC), which accounts for sentences containing two A’-dependencies.28 On the other hand, if the two A’-dependency lines are not crossed but nested, we predict that a sentence turns out to be grammatical. This is demonstrated in (44): (44)
a
Nani-o Taroo-sika yoma-nai no what-ACC Taro-only read-NEG Q ‘What does only Taro read?’
b
Nani-o Taroo-sika yoma-nai no
The difference between (43)a and (44)a is that in the sentence in (44)a, the Wh-phrase is scrambled to sentence-initial position, which can be
128 Cleft Constructions in Japanese Syntax
shown to be an A-position (Saito 1992). The scrambling creates the representation in (44)b in which the two A’-dependencies do not violate the LCC. Therefore, unlike (43), the sentence in (44) is grammatical. Now, let us consider the following long-distance cleft sentence with another A’-dependency internal to the presuppositional clause. I assume that a null operator in cleft constructions is associated with the nominalizer (no): (45)
a?*
[Taroo-sika [Hanako-ga proi atta to] omottei-nai-no]-wa Taro-only Hanako-NOM met COMP think-NEG-NM-TOP sono hitoi-ni-da that person-DAT-COP ‘It is that person that only Taro thinks that Hanako met’
b?*
[CP1[IP Taroo-sika [CP2 Op [CP . . . . .] NEG]]-NM]
The ungrammaticality of the sentence in (45)a suggests that not only the A’-dependency line between Taroo-sika ‘Taro-only’ and the NEG head but also another A’-dependency exists in the presuppositional clause. Assuming that a null operator undergoes movement from a position adjoined to the embedded CP, we find that the two A’-dependency lines are crossed, as shown in (45)b: hence, there is a violation of the LCC. If the embedded clause in (45) is scrambled to the sentence-initial position, the sentence satisfies the LCC and becomes grammatical, as shown in (46): (46)
a
Taroo-sika tj omottei-nai-no]-wa [[Hanako-ga proi atta to]j Hanako-NOM met COMP Taro-only think-NEG-NM-TOP sono hitoi-ni-da that person-DAT-COP ‘It is that person that [Hanako met], only Taro thinks’
b
[CP1 [CP2 Op [CP2 . . .]]j [IP Taroo-sika . . tj . . . NEG] NM]
The null operator moves from the position adjoined to the scrambled embedded clause, and this A’-dependency line does not cross another association line connecting by Taroo-sika ‘Taro-only’ and its licensing NEG head. Therefore, (46) does not violate the LCC. If we assume that
Resumptive A’-Dependencies 129
the LCC is correct, the fact in (46) also tells us that the operator crucially undergoes movement from a position adjoined to the embedded clause. In other words, the grammaticality of (46)a suggests that the operator is not base-generated in a position adjoined to the highest VP within the presuppositional clause. Notice, however, that the sentences in (45) and (46) do not necessarily show that the topic operator is generated in the adjoined position; the same grammaticality can be explained under the assumption that the operator moves from its thematic position in the embedded clause, because the two A’-dependency lines still cross each other in (45) and do not cross in (46). However, the following example crucially demonstrates that the null operator is not generated in its thematic position but in the adjoined position: (47)
a
[Taroo-ga [Hanako-sika proi awa-nak-atta to] Taro-NOM Hanako-only meet-NEG-PAST COMP omotteiru-no]-wa sono hitoi-ni-da think-NM-TOP that person-DAT-COP ‘It is that person that Taro thinks that only Hanako met’
b
[CP1 [IP . . . [CP2 Op [CP2 Hanako-sika . . . NEG]]] NM]
Unlike the sentence in (45), (47)a is perfect. This fact indicates that there are no crossing A’-dependencies in the representation, as schematized in (47)b. If we analyze the sentence in (47)a as the structure in (48) shown below, where movement of the null operator starts out from the thematic position in the embedded clause, then the grammaticality of (47)a would not be expected: (48)
*
[CP1 [IP . . . [CP2 Hanako-sika . . Op . . . NEG]] NM]
The representation in (48) violates the LCC because the two A’-dependency lines are crossed; however, since the sentence is grammatical, this must not be the structure for the sentence in (47)a. Thus, we can conclude that the null operator is generated outside the embedded clause. To summarize this section, we have seen that the binding facts concerning ‘partial’ reconstruction are not the only ones to support the
130 Cleft Constructions in Japanese Syntax
structure of long-distance cleft constructions. Considering WCO effects and interaction with NPIs, we have concluded that long-distance clefts involve operator movement from a position adjoined to the embedded CP.
Resumption Three layered clefts: resumptive pro or movement? This subsection discusses the status of the theta position of null operators in long-distance clefts. As pointed out by Mark Baker and Jonathan Bobaljik (personal communication) independently, the analysis in the last subsection might suggest that the scrambling analysis for simple clefts can also deal with long-distance cleft constructions. In other words, while we have assumed that a theta position for a focus phrase is occupied by pro in long-distance clefts, it may be possible to think that the theta position in question is a trace of an operator which is first scrambled locally (hence, by A-movement), and then undergoes A’-movement to the spec of CP. This subsection, however, shows that the resumptive pro analysis, which was originally proposed in this chapter, correctly explains the facts concerning Japanese three-layered clefts. The following are the schematic structures that each analysis can predict in terms of anaphor binding (SELF stands for an operator associated with an anaphoric focus phrase): (49)
a
[CP SELFi [IP1NPi [CP1 t [IP2NP [CP2NP pro]]]]]
b [CP SELFi/j [IP1NPi [CP1 t [IP2NPj [CP2 t [IP3 NP pro]]]]]] (50)
[CP SELFi/j [IP1NPi [CP1 t [IP2NPj [CP2 t [IP3 NP t ]]]]]] A’
A’
A
The structures in (49) and (50) are possible representations for the presuppositional portion of the cleft sentences containing two embedded clauses. The structures under the resumptive pro analysis are shown in (49). In (49)a, the null operator associated with a focus anaphor (represented as SELF) is base-generated in the position adjoined to CP1. Thus, this structure predicts that the anaphor can be coindexed only with the subject of IP1. On the other hand, the structure (49)b shows the null operator (SELF) as generated in the position adjoined
Resumptive A’-Dependencies 131
to CP2. This predicts that the anaphor can be coindexed either with the subject of IP1 or the subject of IP2. (50) is the representation under the scrambling analysis. Although the null operator is base-generated in its theta position, the first step counts as A-movement and the rest as A’-movement. Therefore, if we assume that the A-movement is not subject to reconstruction (Lasnik 1999), then the anaphor is reconstructed to the base position for the A’-dependency. If this is on the right track, we could predict that the anaphor can be coindexed with either the subject of IP1 or the subject of IP2, but not just the topmost subject.29 Let us now consider the following sentences with three layers:30 (51) a Tarooi-ga [Hanakoj-ga [Masaok-ga zibuni/j/k-no Taro-NOM Hanako-NOM Masao-NOM self’s hahaoya-ni atta to] itta to] omotteiru mother-DAT met COMP said COMP think ‘Taroi thinks that Hanakoj said that Masaok met self’si/j/k mother’ b [Tarooi-ga [Hanakoj-ga [Masaok-ga e atta to] Taro-NOM Hanako-NOM Masao-NOM met COMP itta to] omotteiru -no] -wa zibuni/*j/*k-no hahaoya-ni said COMP think -NM-TOP self’s mother-DAT dat-ta COP-PAST It was himselfi that Taroi thinks that Hanako said that Masao met’ The sentence in (51)a has two embedded clauses. Since zibun ‘self’ is a long-distance anaphor, it can refer to any of the subjects in the sentence. Sentence (51)b is similar, except zibun has been clefted out of all these clauses. Because of the property of center-embedding, the grammaticality judgments are hard to obtain. However, most of the native speakers of Japanese that I consulted agree that the focus anaphor in (51)b can be coindexed with the matrix subject, Taro, but not with the other more deeply embedded subjects. The same is true for three-layered structures containing soko ‘it, that (place)’. (52)a is a non-clefted sentence in which the anaphoric pronoun, soko, appears inside the most embedded clause. The sentence in (52)b demonstrates a structure where the soko phrase is clefted:
132 Cleft Constructions in Japanese Syntax
(52)
a
Nihon-juu-no kaisyai-ga [subete-no Japan-all-over-GEN company-NOM all-GEN sinbunshaj-ga [[Toyota to Nissan]k-ga e newspaper-office-NOM Toyota and Nissan-NOM hoodoo-sita to] sokoi/j/k-no keiretu-gaisya-ni tayotteiru to] it-GEN subsidiary-DAT rely-on COMP reported COMP omotteiru think ‘Companiesi all over in Japan think that all newspaper officesj reported that each of [Toyota and Nissan]k relied on itsi/j/k subsidiary company’
[subete-no b [Nihon-juu-no kaisyai-ga Japan-all-over-GEN company-NOM all-GEN sinbunshaj-ga [[Toyota-to Nissan]k-ga e tayotteiru newspaper-office-NOM Toyota and Nissan-NOM rely-on to] hoodoo-sita to] omotteiru-no]-wa sokoi/*j/*k-no COMP reported COMP think-NM-TOP it-GEN keiretu-gaisya-ni-da subsidiary-DAT-COP ‘It is itsi subsidiary company that companiesi all over in Japan think that all newspaper offices reported that each of [Toyota and Nissan] relied on’ The sentence in (52)a shows that soko in the most embedded clause can refer to any subject in the sentence. However, when the anaphoric pronoun is clefted, as shown in (52)b, the possible antecedent is only the highest subject, nihon-juu-no kaisya ‘Japan all-over-GEN companies’. Again, this fact of three-layered clefts parallels that of zibun binding in (51)b. Therefore, the reconstruction facts support (49)a rather than (49)b or (50). We can also ask whether the structure in (49)a is correct with respect to the interaction with sika-nai ‘only-NEG’ constructions. What is expected under our current analysis is that when sika ‘only’ attaches to the highest subject with a NEG head in a relevant position, the two A’-lines are crossed, violating the LCC. On the other hand, if sika appears with the second highest subject or the lowest subject, the sentence does not become ungrammatical. As we see in (53), this is in fact borne out:
Resumptive A’-Dependencies 133
(53)
a
[Taroo-ga [Hanako-ga [Masao-sika e Taro-NOM Hanako-NOM Masao-only aw-ana-katta to] itta to] omotteiru-no]-wa meet-NEG-PAST COMP said COMP think-NM-TOP sono hito-ni da that person-DAT COP ‘It was that person that Taro thinks that Hanako said that only Masao met’
b?
[Taroo-ga [Hanako-sika [Masao-ga e atta to] Taro-NOM Hanako-only Masao-NOM met COMP iw-ana-katta to] omotteiru-no]-wa sono hito-ni da say-NEG-PAST COMP think-NM-TOP that person-DAT COP ‘It was that person that Taro thinks that only Hanako didn’t say that Masao met’
c?*
[Taroo-sika [Hanako-ga [Masao-ga e atta to] Taro-only Hanako-NOM Masao-NOM met COMP itta to] omottei-nai -no]-wa sono hito-ni da say-PAST COMP think-NEG-NM-TOP that person-DAT COP ‘It was that person that only Taro thinks that Hanako said that Masao met’
In (53), we find that when the sika phrase with its NEG head occurs in the most embedded clause ((53)a) or in the second embedded clause ((53)b), the sentences are grammatical.31 However, when the sika phrase is generated as the subject of the matrix clause (53)c, then the sentence is ruled out. The schematic structures for the sentences in (53) are represented as in (54), respectively:
(54)
a
[CP [IP1 Taro [CP1 Op [IP2 Hanako [IP3 Masao-sika pro . . . NEG] . . .]] . . .] NM]
b
[CP[IP1 Taro [CP1 Op [IP2 Hanako-sika [IP3 Masao pro . . .] NEG]] . . .] NM]
c*
[CP[IP1 Taro-sika [CP1 Op [IP2 Hanako [IP3 Masao pro . . .] . . .]] NEG] NM]
134 Cleft Constructions in Japanese Syntax
As shown in the structures in (54), the A’-dependency lines are crossed only in the (54)c example under our analysis. The facts in (53) can be accounted for if the null operator undergoes movement from the position adjoined to the highest embedded clause. Notice that if the movement starts from other lower positions, the facts in (54)a and (54)b cannot be explained. Therefore, it is concluded that the operator in long-distance cleft constructions moves from the highest embedded clause. In contrast, the scrambling analysis, in which the first step is A-movement, does not predict the facts given above. Clefting true adjunct PPs We have argued that the thematic position for a null operator in longdistance presuppositional clauses is occupied by a pro rather than a trace. If this is on the right track, we can predict that ‘true’ adjuncts which do not have corresponding pros cannot modify the embedded clause in long-distance presuppositional clauses. In previous research (Huang 1982, Lasnik and Saito 1984, Saito 1985, Murasugi 1991, and references cited therein), it is assumed that PPs are divided into two types: true adjunct PPs and quasi adjunct PPs. Roughly speaking, the former include reason and manner PPs and the latter time and place PPs, as exemplified in (55): (55)
a
Taroo-ga sono riyuu-de sigoto-o yameta Taro-NOM that reason-for job-ACC quit ‘Taro quit the job for that reason’
b
Taroo-ga soko-de sigoto-o hajimeta Taro-NOM there-at job-ACC started ‘Taro started working there’
(55)a contains the reason PP, sono riyuu-de ‘for that reason’, and (55)b has the place PP, soko-de ‘there’. These two types of PPs allow long-distance scrambling, as shown in (56) and (57): (56)
a
Taroo-ga [kyonen Hanako-ga sono riyuu-de sigoto-o Taro-NOM last-year Hanako-NOM that reason-for job-ACC yameta to] omotteiru quit COMP think ‘Taro thinks that Hanako quit the job for that reason last year’
Resumptive A’-Dependencies 135
(57)
b
Taroo-ga sono riyuu-dei [kyonen Hanako-ga ti sigoto-o Taro-NOM that reason-for last-year Hanako-NOM job-ACC yameta to] omotteiru quit COMP think
c
Sono riyuu-dei Taroo-ga [kyonen Hanako-ga ti sigoto-o that reason-for Taro-NOM last-year Hanako-NOM job-ACC yameta to] omotteiru quit COMP think
a
Taroo-ga [kyonen Hanako-ga soko-de sigoto-o Taro-NOM last-year Hanako-NOM there-at job-ACC hajimeta to] omotteiru started COMP think ‘Taro thinks that Hanako started working there last year’
b
Taroo-ga soko-dei [kyonen Hanako-ga ti sigoto-o Taro-NOM there-at last-year Hanako-NOM job-ACC hajimeta to] omotteiru started COMP think
c
Soko-dei Taroo-ga [kyonen Hanako-ga ti sigoto-o there-at Taro-NOM last-year Hanako-NOM job-ACC hajimeta to] omotteiru started COMP think
The reason PP in (56) as well as the place PP in (57) behave in the same way in that they can be preposed by long-distance scrambling. The b and c sentences in (56) and (57) can be interpreted ambiguously; namely, the PPs modify either the embedded clause or the matrix clause. However, what is important here is that both true and quasi adjunct PPs allow the lower reading even when they are scrambled out of the embedded clause. The contrast between the two types emerges clearly in the following HERCs: (58)
a
[Taroo-ga [Hanako-ga ei sigoto-o hajimeta to] Taro-NOM Hanako-NOM job-ACC started COMP omotteiru] tokoroi think place ‘the place that Taro thinks that Hanako started working’
136 Cleft Constructions in Japanese Syntax
omotteiru] b* [Taroo-ga [Hanako-ga ei sigoto-o yameta to] Taro-NOM Hanako-NOM job-ACC quit COMP think riyuui reason ‘the reason that Taro thinks that Hanako quit the job’ Both relative clauses contain a complement clause, in which the gap is associated with the relative head. The head of the relative clause in (58)a indicates a place whereas the one in (58)b a reason. Interestingly, unlike (58)a, riyuu ‘reason’ cannot be relativized from the complement clause in (58)b; a possible interpretation is the reason why Taro thinks that way, and not the reason why Hanako quit the job. Murasugi (1991) proposes that Japanese allows a pro which corresponds to a quasi adjunct PP but not to a true adjunct PP. Based on this assumption, she argues that within the Government and Binding theory, the ungrammaticality of (58)b comes from the fact that the intermediate trace of an operator is not properly governed. Her analysis crucially assumes only X0 can be a governor (Stowell 1981, Rizzi 1986, Lasnik and Saito 1984), and Japanese relative clauses are IPs to which a null operator adjoins.32,33 Under Murasugi’s analysis, the relative clause in (58)b is schematically illustrated as follows: (59)*
NP IP Opi
NP IP reason I’ VP
CP ungoverned--> ti
I V
C’ IP
Ci
ti
The trace in the base position for the null operator is properly governed, assuming that the spec-head agreement gives the same index to the C0, which can be qualified as a governor for the trace.34 The intermediate trace in the spec of CP is not licensed, however. The potential antecedent
Resumptive A’-Dependencies 137
governor is the operator in the position adjoined to the highest IP, but this is not an X0 category and hence, the trace is not governed.35 With this much as a background, let us consider the following cleft sentences: (60)
a
[Taroo-ga [Hanako-ga ei sigoto-o hajimeta to] Taro-NOM Hanako-NOM job-ACC started COMP omotteiru no] -wa sono basyo-dei da think-NM-TOP that place-in COP ‘It was in that place that Taro thinks that Hanako started working’
b*
[Taroo-ga [Hanako-ga ei sigoto-o hajimeta to] Taro-NOM Hanako-NOM job-ACC quit COMP omotteiru-no]-wa sono riyuu -dei da think-NM-TOP that reason -for COP ‘It was for that reason that Taro thinks that Hanako quit the job’
As we have seen in (58), the same contrast between the place and reason PPs can be observed in cleft sentences, as shown in (60). Our proposed analysis can readily account for this fact, even without assuming the mechanism proposed by Murasugi (1991). The gap in the most embedded clause in long-distance clefts must be occupied by a pro; however, since true adjunct PPs do not have corresponding pros, (60)b is not possible. Under our analysis, the presuppositional clause in (60)b is roughly represented as follows: (61)*
CP Opi (reason)
C’ C NM
IP Taro
I’
VP CP
V think
Cp
ti IP
...ti/proi...
C
I
138 Cleft Constructions in Japanese Syntax
Under our analysis, the null operator in long-distance cleft constructions is generated in the position adjoined to the (highest) embedded CP. Therefore, the lowest empty position in (61) should not be occupied by a trace. In addition, since the null operator agrees with the focus phrase, which is a true adjunct PP, a pro is not possible in the lowest empty position. Thus, the structure in (61) is not possible.36 Let us further consider the following three-layered cleft constructions: (62)
a* [Masao-ga [Taroo-ga [Hanako-ga ei sigoto-o yameta Masao-NOM Taro-NOM Hanako-NOM job-ACC quit to] omotteiru to] itta-no]-wa sono riyuu-dei da COMP think COMP said-NM-TOP that reason COP ‘It was for that reason that Masao said that Taro thought that Hanako quit the job’ b*
[Masao-ga [Taroo-ga ei [Hanako-ga sigoto-o yameta Masao-NOM Taro-NOM Hanako-NOM job-ACC quit to] omotteiru] to itta-no]-wa sono riyuu-dei da COMP think COMP said-NM-TOP that reason COP
c
[Masao-ga ei [Taroo-ga Masao-NOM Taro-NOM to] omotteiru to] COMP think COMP
[Hanako-ga sigoto-o yameta Hanako-NOM job-ACC quit itta-no]-wa sono riyuu-dei da said-NM-TOP that reason COP
The cleft sentences in (62) suggest that the true adjunct PP cannot be construed as having originated in either of the embedded clauses, as shown in (62)a and (62)b. (62)c shows that sono riyuu-de ‘for that reason’ can only modify the highest clause of presupposition. The facts here are also compatible with our analysis; a null operator can undergo movement from the highest clause of the presuppositional clause but not from more deeply embedded clauses. Since a true adjunct PP does not allow the occurrence of the corresponding pro, (62)a and (62)b are not possible structures.37 Other languages In the previous sections, we discussed the fact that Japanese long-distance clefts exhibit resumptive A’-dependencies; namely, when a phrase within the complement clause is clefted, its thematic position is occupied by pro and the associated null operator is base-generated in the position adjoined to the embedded clause, from whence it moves to the spec of
Resumptive A’-Dependencies 139
CP. Thus, while there is no movement operation taking place within the embedded clause, we see some syntactic effects of A’-movement outside the complement clause. This type of derivation for Japanese long-distance clefts appears to be a highly language-specific operation; however, as we will see in the following subsections, it can be shown that our analysis presented here is also motivated in other languages. The following sections will discuss similar resumptive A’-dependency analyses proposed in the literature; however, some crucial differences between these languages and Japanese will be also presented.38 Clitic left dislocation in Modern Greek Clitic Left Dislocation (CLLD) in languages like Greek involves a left dislocated phrase in an adjoined position which is associated with a resumptive clitic on the predicate, as exemplified in (63):39 (63)
[ton Kosta], i Maria ton idhe DET Kosta-ACC DET Mary-NOM him saw ‘Kosta, Mary saw him’
One of the major issues concerning CLLD is whether the left dislocated phrase is base-generated in its surface position or is moved from its theta position. Cinque (1990), for instance, takes the former view; based on Italian data, he argues that CLLD constructions do not involve movement and the dislocated element is base-generated in its left peripheral position.40 Iatridou (1991) also claims that CLLD in Modern Greek shows some properties of base-generation. For example, she argues that a left-dislocated object in the CLLD constructions is different from an object moved from its theta position. As shown in (64), the CLLD construction in (64)a does not exhibit WCO effects whereas the moved object construction in (64)b does observe WCO effects:41 (64)
a
[ton Kosta] i mitera tu ton agapa DET Kosta DET mother-his him loves ‘Kosta, his mother loves’
b*
[ton Kosta] i mitera tu agapa DET Kosta DET mother-his loves
The only difference between (64)a and (64)b is that the former has a clitic on the verb and the latter does not have a clitic. In spite of this slight
140 Cleft Constructions in Japanese Syntax
difference, the CLLD construction and the moved object construction behave differently in terms of WCO effects. In addition, the moved object can license parasitic gaps; however, the left-dislocated object in CLLD cannot, as shown in (65): (65) a* [Afto to arthoro] i Maria to arxiothetise xoris na dhiavasi this DET article DET Mary it filed without reading ‘This article, Mary filed without reading’ b
[Afto to arthoro] i Maria arxiothetise xoris na dhiavasi this DET article DET Mary filed without reading
(65)a is a CLLD construction, and (65)b has also the left-dislocated object but involves movement from its theta position. Unlike the moved object in (65)b, the left-dislocated object in (65)a cannot license the parasitic gap. Assuming that a WCO effect and licensing a parasitic gap are related to A’-movement, Iatridou concludes that the facts in (64) and (65) show that the left-dislocated element is base-generated in a surface position in CLLD constructions.42 While the examples in (64) and (65) involve local CLLD, Iatridou (1991) further examines long-distance CLLD. She attempts to solve the so-called ‘Cinque’s Paradox’, where the relationship between the left dislocated phrase and the clitic is not created by movement and yet is still constrained by strong islands. This is shown in (66): (66)*
[ton Kosta], sinandisa tin kopela pu ton idhe DET Kosta (I) met DET girl who him saw ‘Kosta, I met the girl who saw him’
If the left-dislocated phrase, ton Kosta ‘DET Kosta’, is base-generated in the surface position, coindexed with the clitic pronoun inside the relative clause, the sentence should not observe island effects under standard assumptions. Contrary to Cinque’s analysis of this paradox, which is that the sensitivity to strong islands can be explained by binding chains, Iatridou (1991) argues that island effects are sensitive to a movement operation as the previous analyses proposed in the literature (Chomsky 1975 and many others). According to Iatridou (1991), the left-dislocated phrase is base-generated adjoined to the minimal CP containing the clitic, because the left-dislocated element appears before a Wh-phrase in Modern Greek, which does not allow multiple spec positions, as shown in (67).43,44
Resumptive A’-Dependencies 141
(67)
a
tin Maria [CP pios tin idhe] DET Mary who her saw ‘Mary, who saw her?’
b
anarotieme [ton Kosta [CP pios ton idhe]] wonder DET Kosta who him saw ‘I wonder Kosta, who saw him?’
Thus, based on the facts above, the sentence with a complement clause is represented as (68): (68)
[ton Kosta]i nomiza [CP ti [CP oti i Maria ton idhe]] DET Kosta (I) thought that DET Mary him saw ‘Kosta, I thought that Mary saw him’
In (68), the left-dislocated phrase, ton Kosta ‘DET Kosta’, is generated in the position adjoined to the embedded CP, which contains the associated clitic pronoun in it.45 The CLLD construction, represented in (68), is licensed by a predication relation. The left-dislocated phrase, ton Kosta, is considered a subject, and the rest of the clause is the predicate. The clitic is the predicate variable, which makes the clause a legitimate predicate (Williams 1980). The ungrammaticality of (66) follows assuming that the left-dislocated NP is adjoined to the smallest CP with the theta position. The basegenerated position for the left-dislocated element is in the position adjoined to the CP relative clause. The movement of the left-dislocated phrase takes place from an adjoined position, and extraction from such a position across an island is predicted to have a strong island violation. The same is true for extraction from adjunct clauses, sentential subjects, and complex NP islands in Modern Greek. Therefore, Iatridou’s analysis correctly explains why CLLD constructions show effects of both base-generation and movement of the left-dislocated element. What is more relevant to see whether CLLD in Modern Greek exhibits resumptive A’-dependencies are the following parasitic gap examples: (69)
a(?)
ton Yanisi i Maria ipe [xoris na agapa pg] ti oti DET John DET Mary said [without loves] that (she) tha ton pandrefti will him marry ‘John, Mary said that she will marry him without loving (him)’
142 Cleft Constructions in Japanese Syntax
b*
ton Yanisi i Maria ipe ti oti tha ton pandrefti DET John DET Mary said that (she) will him marry [xoris na agapa pg] [without loves]
The sentences in (69) are parasitic gap constructions, in which a parasitic gap (pg) in the adjunct clause is licensed by A’-movement (Chomsky 1982, 1986a, among others). (69)a has the adjunct clause containing the pg, and the sentence is grammatical under the relevant reading. This fact suggests that A’-movement takes place to license the parasitic gap; on the other hand, Cinque’s (1990) base-generation analysis cannot explain it. On the contrary, in (69)b, the adjunct clause containing the pg seems to appear inside the embedded clause under the same reading. The trace of the left-dislocated element c-commands the pg, which is considered a violation of the ‘anti-c-command’ requirement for parasitic gap licensing; hence, it is ungrammatical. Therefore, these parasitic gap constructions in Modern Greek show that there is indeed A’-movement from a position adjoined to an embedded clause in CLLD constructions. The CLLD in Modern Greek is similar to Japanese cleft constructions in that both involve resumptive A’-dependencies in long-distance configurations. However, there are at least two properties which make the two constructions differ. First, according to Iatridou (1991), the CLLD in a local environment does not involve any movement in Modern Greek (see (64) and (65)).46 However, as will be shown on pages 148–501, local clefting in Japanese needs to be derived by movement. One of the examples which support this analysis is a reconstruction effect, as shown in (70): (70)
a
Tarooi-ga zibuni-no hahaoya-ni atta Taro-NOM self’s mother-DAT met ‘Taro met self’s mother’
b
[Tarooi-ga e atta no]-wa zibuni-no hahaoya-ni da Taro-NOM met NM-TOP self’s mother-DAT COP ‘It was self’s mother that Taro met’
The anaphor, zibun ‘self’, is bound by Taroo in the simple sentence in (70)a. When the anaphor is clefted, it still gets coindexed with Taroo, as shown in (70)b. Under the assumption that reconstruction effects are observed only when movement takes place, we need to assume that the local cleft sentence involves movement. More about local clefting will be discussed later.
Resumptive A’-Dependencies 143
Second, whereas Modern Greek and Japanese have similar chains in two-clause structures in CLLD and cleft constructions, they apparently differ in how resumptive A’-chains are formed in three-clause structures. Iatridou (1991) proposes that a left-dislocated element in CLLD is generated adjoined to a minimal clause containing its clitic. Whether this is also true in three-layered sentences is not attested in her work; however, the following sentences with three-layered structures suggest that this is indeed the case:47 (71)
a
ton Kosta nomiza oti i Maria ipe oti tha DET Kosta thoughtthat DET Mary said that (she) will ton pandrefti him marry ‘Kosta, I thought that Mary said that she will marry him’
b?
ton Kosta nomiza oti i Maria ipe [xoris na agapa pg] DET Kosta thought that DET Mary said without loves oti the ton pandrefti that (she) will him marry ‘Kosta, I thought that Mary said that she will marry him without loving (him)’
The sentence in (71)a shows that CLLD is possible from the most embedded clause in the three-layered structure. The grammaticality of (71)b further suggests that the analysis in Iatridou (1991) is correct. Because the left-dislocated phrase is base-generated in the position adjoined to the most embedded CP (a minimal CP containing a clitic), the parasitic gap is not c-commanded by the trace. On the contrary, if the left-dislocated phrase is generated in the adjoined position to the second highest CP, the grammaticality is not expected. Recall that a null operator is not generated adjoined to the minimal CP containing its theta position in Japanese clefts. This is supported by various facts concerning three-layered cleft constructions (see pages 130–4). Although both languages allow resumptive A’-dependencies, they are crucially different in terms of the base position of the operator. I do not have any satisfactory solution for these language differences, and will leave these questions open for future research. Wh-movement in Selayarese Another type of language which exhibits resumptive A’-dependencies is Selayarese, discussed by Finer (1997). This language manifests morphology that indicates extraction. For instance, according to Finer, Wh-movement
144 Cleft Constructions in Japanese Syntax
is allowed only when overt complementizers and agreement suffixes on verbs do not appear:48 (72)
a
apa mu-isseʔ la-ʔalle i Basoʔ what 2fam-know 3-took DET Basoʔ ‘What do you know Basoʔ took?’
b*
apa mu-isseʔ wha 2fam-know
c*
apa mu-isseʔ la-ʔalle-i i Basoʔ what 2fam-know 3-took-3 DET Basoʔ
muko la-ʔalle i Basoʔ COMP 3-took DET Basoʔ
The sentence in (72)a is grammatical in the language, because no overt complementizer nor agreement suffixes occur between the Wh-phrase and its original position; if the overt complementizer appears, the sentence is ruled out, as shown in (72)b. In addition, when the verb has an agreement suffix, as shown in (72)c, extraction of the Wh-phrase is also blocked.49 Relevant here is what Finer (1997) calls the second Wh-constructions, which contain a base-generated Wh-operator binding a phonologically null category that is not generated by movement. An example is shown below: (73)
[CP apai [IP mu-isseʔ [CP ti muko [IP la-isseʔ-i i Ali [CP lako what 2fam-know COMP 3-know-3 DET Ali COMP [IP la-/alle-i proi i Basoʔ]]]]]] 3-take-3 DET Basoʔ ‘What do you know that Ali knows that Basoʔ took?’
Based on the analysis proposed for (72) and others, Finer claims that the presence of overt complementizers and verbal affixes indicates that cyclic movement is not involved in (73). Hence, the Wh-phrase is base-generated in the COMP of the first embedded CP and moves to the spec of the top CP. One piece of evidence to confirm that sentences like (73) contain a null pronoun in the theta position of the Wh-phrase is that when complementizers are overt, no crossover effects are observed: (74)
inai mu-kua muko la-isseʔ-i andoʔ-na lako la-sumbele-i who 2fam-say COMP 3-know-3 mom-3 COMP 3-slaughter-3 tedong injo buffalo DET ‘Whoi did you say hisi/j mom knows slaughtered the buffalo?’
Resumptive A’-Dependencies 145
If the Wh-phrase had been in some lower position (its theta position or COMP in the most embedded CP), we should observe a WCO effect, because the trace of the Wh-phrase would not c-command the pronoun. However, the sentence in (74) is grammatical, which suggests that the Wh-movement does not start from its theta position, but presumably, it starts from the COMP of the highest embedded clause.50 The Wh-interrogatives in Selayarese are also reminiscent of cleft constructions in Japanese. They both allow resumptive A’-dependencies in long-distance configurations; an operator is generated in some intermediate position, binding a coindexed null element in its theta position. The difference between the two languages is that in Selayarese, types of chains are determined by the realization of agreement on a verb and a complementizer. Thus, unlike Japanese long-distance clefts, A’-movement does not necessarily take place from the highest embedded clause, but as shown in (75), the agreement morphology shows that the operator movement starts from the most embedded clause: (75)
[CP apai [IP mu-isseʔ [CP ti [IP to-isse [CP ti toko what 2fam-know 1exc-know COMP [IP la-taro-i proi ri lamari]]]]]] 3-put-3 in cupboard ‘What do you know that we know that he put in a cupboard?’
The structure in (75) shows that the Wh-phrase is generated in COMP in the most embedded clause. Since the verbs in the higher clauses do not have agreement suffixes and no overt complementizer appears above the most embedded clause, the sentence is grammatical. In this section, we have looked at A’-movement in Modern Greek and Selayarese, and have found that the structure for resumptive A’-dependencies proposed for Japanese long-distance cleft constructions is observed in other languages. What is common in these languages is that an operator is generated in some intermediate position, binding a coindexed null element in its theta position. This dependency does not show any diagnosis of movement; however, since the operator undergoes movement to the spec of the highest CP, properties of movement are observed in the higher dependency. A full analysis of the crosslinguistic difference that I have pointed out above is beyond the scope of this book. However, the brief survey of resumptive A’-dependencies across languages indirectly supports our claim for the structure of a long-distance cleft construction in Japanese.
146 Cleft Constructions in Japanese Syntax
Residual issues This section will discuss some residual issues that we have not considered under our analysis so far. One is about island effects in long-distance cleft constructions, and the other is the structure for local clefts. Strong islands in long-distance clefts So far, we have examined the structures of long-distance cleft constructions which involve a complement clause. According to the analysis proposed here, it might be expected that long-distance clefts containing a syntactic island would not be sensitive to subjacency (or some equivalent principle of movement). For instance, since we assume that a null operator associated with a focus phrase is base-generated in the position adjoined to the embedded clause, subjacency should not be observed when an island appears inside the embedded clause. Subjacency is obeyed, however, as shown in (76) and (77):51 (76)
(77)
a
Hanako-ga [CP Tarooi-ga [PP proi sono hito-ni au-maeni] Hanako-NOM Taro-NOM that person-DAT meet-before dekaketa to] itta left COMP said ‘Hanako said that Taro left before (he) met the person’
b?*
[Hanako-ga [CP Tarooi-ga [PP proi [e]j au-maeni] dekaketa Hanako-NOM Taro-NOM meet-before left to] itta no]-wa sono hito-nij-da COMP said NM-TOP that person-DAT-COP ‘?*It is that person that Hanako said that Taro left before meeting’
a.
Hanako-ga [CP Taroo-ga [CP/IP ei tomodati-ni atta] hitoi-o Hanako-NOM Taro-NOM friend-DAT met person-ACC sitteiru to] itta know COMP said ‘Hanako said that Taro knew the person who met the friend’
b.?*
[Hanako-ga [CP Taroo-ga [CP/IP ei [e]j atta] hitoi-o Hanako-NOM Taro-NOM met person-ACC sitteiru to] itta-no]-wa tomodati-nij-da know COMP said-NM-TOP friend-DAT-COP ‘?*It is the friend that Hanako said that Taro knew the person who met’
Resumptive A’-Dependencies 147
(76)a and (77)a are non-clefted sentences. If the object phrase, sono hito-ni ‘that person-DAT’, inside the adjunct clause in (76)a or tomodati-ni’ friendDAT’ inside the relative clause in (77)a is clefted, the sentences become degraded, as shown in (72)b and (73)b, respectively. This fact should not be expected under our proposal, because each theta position in the islands is assumed to be occupied by a pro. Therefore, the sentences in (76)b and (77)b should be free from violation of subjacency under our analysis. Here, we assume, following Cinque (1990), that sensitivity to strong islands is not only a property of movement but also a property of binding chains created by a base-generated element and its associated pro.52 In other words, subjacency is a condition on representation (Cinque 1990: 57). Cinque’s analysis is based on CLLD constructions in Italian, where, for instance, the clitic cannot license a parasitic gap.53 However, there is a difference between Italian CLLD constructions and Japanese long-distance cleft constructions. According to Cinque, the CLLD constructions show total connectivity even in long-distance CLLD constructions, as illustrated in (78): (78)
a
Per séi immagino che Mariai si sia comprata for herself (I) imagine that Maria her (self) has got un nuovo borsellno a new purse ‘For herself, I imagine that Maria got a new purse for herself’
b* Per séi Mariai immagina che io si sia comprato for herself Maria imagines that I her (self) has got un nuovo borsellino. a new purse ‘For herself, Maria imagines that I got a new purse for her’ (Cinque 1997:102–3) The sentence in (78)a shows that the left dislocated anaphor, per sé ‘herself’, reconstructs to its base position, and the sentence does not violate the binding condition A. On the contrary, the sentence in (78)b is ruled out, because the same anaphor in the reconstructed position cannot be bound within its governing category. Therefore, this fact suggests that the left dislocated element must have reconstructed to its thematic position even in long-distance CLLD constructions. As we already saw in the previous discussion, this is not the case for the cleft constructions in Japanese. However, this difference between
148 Cleft Constructions in Japanese Syntax
Italian CLLD constructions and Japanese clefts might be due to the occurrence of the clitic pronoun in Italian CLLD; in Italian, it could be that overt clitic pronouns somehow allow total reconstruction whereas in Japanese, since there is no overt element left, reconstruction is not obtained. This is a mere stipulation and further investigation will be required. Local clefting While we have closely examined long-distance cleft constructions, we have not carefully considered the structure for the presuppositional clause in local cleft constructions. This subsection will address some problematic issues concerning local clefts under our analysis. On pages 109–10, we proposed that structures for local clefts should be schematized as (79)a and not (79)b: (79)
a
[CP Opi [IP . . . ti . . .]]
b*
[CP Opi [IP . . . proi . . .]]
The difference between (79)a and (79)b is that the null operator is generated in its thematic position in the former, whereas the null operator is base-generated in the spec of CP, binding a thematic position occupied by pro in the latter. We would like to propose the structure in (79)a, rather than (79)b, to account for the connectivity phenomenon, such as (7)a, repeated here as (80): (80)
[Hanakoi-ga e atta no]-wa zibuni-no hahaoya-ni dat-ta Hanako-NOM met NM-TOP self’s mother-DAT COP-PAST ‘It was self’si mother that Hanakoi met’
Since the anaphor, zibun ‘self’, is bound by its antecedent, Hanako, in (80), it is plausible to assume that the theta position of the focus phrase has a trace created by movement. This fact is difficult to explain under the structure in (79)b. The structure in (79)a correctly accounts for the facts of reconstruction. However, we need more detailed structures for local clefting in Japanese, because the interaction with another A’-dependency in local clefts does not appear to support the structure in (79)a. The relevant example and structures are shown below: (81)
a
[Taroo-sika ti aw-ana-katta-no]-wa sono hitoi-ni-da Taro-only meet-NEG-PAST-NM-TOP that person-DAT-COP ‘It is that book that only Taro reads’
Resumptive A’-Dependencies 149
b*
[Taroo-sika Op aw-ana- . . . no]-wa [sono hitoi-ni]-da
In spite of our prediction, the sentence in (81)a is perfect. If we assume the structure in (79)a, the representation for the relevant A’-dependency lines is as illustrated in (81)b, which violates the LCC. However, since the sentence is grammatical, we should reexamine the structure we have for local clefts. As pointed out by Hidekazu Tanaka (personal communication), one of the possible analyses for the local cleft cases such as (81)a is to assume that the operator does not move from the object position directly to the spec of CP but is scrambled first to sentence-initial position. This scrambling counts as A-movement rather than A’-movement under the standard analysis. Then, the operator further moves to the spec of CP. Thus, if the derivational source for (81)a is represented as a scrambled sentence like (82), we can account for (81)a, while maintaining our proposed analysis: (82)
Sono hito-nii Taroo-sika ti aw-anak-atta that person-DAT Taro-only read-NEG-PAST ‘That person, only Taro met’
Since the clause-internal scrambling is considered A-movement (Saito 1992), the LCC is satisfied in the sentence in (82). Assuming that this is the derivational source for (81)a, the structure, with relevant dependency lines, can be represented as the following: (83)
[Op
Taroo-sika ti aw-anai-no]-wa sono hitoi-ni-da
If we assume that (83) is the correct representation for (81)a, then the grammaticality of the sentence follows. In addition, assuming that scrambling can be undone at LF following Saito (1992), we can also account for why local clefts observe reconstruction to the thematic position. Thus, the schematized presuppositional clauses in local cleft constructions are represented as below: (84)
[CP Opi [IP ti [IP . . . ti . . .]]]
150 Cleft Constructions in Japanese Syntax
Let us now look at local cleft sentences with respect to WCO effects, and see if the following examples can be explained by the analysis in (84). Under the assumption that a null operator in the presuppositional clause of local cleft constructions undergoes A-movement to some clause-initial position, we would not expect WCO effects in such a sentence even if a focus phrase is a true quantifier. This is in fact borne out: (85)
[CPTOPi [IP ti
Taroo-ga [Hanako-ga proi au maeni] Taro-NOM Hanako-NOM meet before sudeni ti atta-no]]-wa darei-ni-na-no? already met-NM-TOP who-DAT-COP-Q ‘Who is it that Taro already met before Hanako met (him/her)?’
The sentence in (85) has an adjunct clause containing a pro coindexed with the operator. Although the focus position is occupied by a Whphrase, the sentence does not exhibit a WCO effect. Our analysis given in (84) can correctly account for this grammaticality. Since the operator is locally scrambled (A-movement) to the position adjoined to the IP before moving to the spec of CP, we expect that no WCO effects arise; a null pronoun in the adjunct clause is bound by the intermediate trace. This fact as well as the one in sika constructions suggests that we need a structure something like (84) rather than the simpler structure in (79)a.
Some implications: highest clause sensitivity Let us return to some more discussion about long-distance cleft constructions in Japanese. We have seen in this chapter that presuppositional clauses in long-distance clefts involve short A’-movement of a null operator in the highest clause. While the analysis at first appears to be special to this particular construction, we will find that this is not true. This section will look at some other constructions which also exhibit the highest clause sensitivity within Japanese as well as in another language. Head external relative clauses (HERCs) In Chapter 3, we argued that the structure for HERCs is distinct from that of presuppositional clauses in clefts in that the latter does not project a nominal category. However, as is well known since Chomsky
Resumptive A’-Dependencies 151
(1977), both constructions are types of operator movement constructions. In fact, peculiar structures for long-distance clefts can also be observed in HERCs in Japanese. Let us first look at reconstruction facts:54 (86)
a
[Hanako to Tarooi-ga e atta] otagaii-no yuujin Hanako and Taro-NOM met each other-GEN friend ‘each other’s friend that Hanako and Taro met’
b [Hanako to Tarooi-ga [sensei-gataj-ga e atta to] Hanako and Taro-NOM teachers-NOM met COMP omotteiru] otagaii/*j-no yuujin think each other-GEN friend ‘each other’s friend that Hanako and Taro think that the teachers met’ (87)
a
[Toyota to Mazdai-ga e izonsiteiru] sokoi-no keiretugaisha Toyota and Mazda-NOM count-on it-GEN subsidiary ‘its subsidiary that each of Toyota and Mazda counts on’
b [subete no sinbunshai-ga [Toyota to Mazdaj-ga e all-GEN newspaper office-NOM Toyota and Mazda-NOM izonsiteiru to] hoodoosita] sokoi/*j-no keiretugaisha count-on COMP reported it-GEN subsidiary ‘its subsidiary that all newspaper offices reported that each of Toyota and Mazda counted on’ The HERCs in (86)a and (87)b are instances of local relativization, and those in (86)b and (87)b involve long-distance relativization. Interestingly, we can find behavior parallel to what we saw in long-distance clefts; the anaphoric pronoun in the head position can be coindexed with the highest subject but not with the lower subject. Therefore, under the assumption that reconstruction effects are one of the manifestations of movement, we can conclude that movement of a null operator in HERCs takes place only in the highest clause. This analysis can be confirmed by the other tests to show highest clause sensitivity. For example, HERCs containing a sika phrase in the embedded clause do not violate the LCC, as shown in (88): (88)
a
[Taroo-ga [Hanako-sika e aw-anak-atta to] Taro-NOM Hanako-only meet-NEG-PAST COMP omotteiru] think ‘the person that Taro thinks that only Hanako met’
152 Cleft Constructions in Japanese Syntax
b [CP . . . [CP Op [CP Hanako-sika proi NEG] . . .] hito
If the HERC in (88)a undergoes A’-movement from the thematic position, we would expect crossing lines, and hence, violation of the LCC. Since the HERC is grammatical, this must not be the right derivation. Rather, as we proposed, the A’-dependency lines should be illustrated as in (88)b. In contrast, when the sika phrase appears in matrix subject position, the sentence is ruled out by the LCC: (89)
a?*
[Taroo-sika [Hanako-ga e atta to] omow-anai] Taro-only Hanako-NOM met COMP think-NEG hito person ‘the person that only Taro thinks that Hanako met’
b?*
[CP Taroo-sika [CPOp [CP . . . proi . . .] NEG]] ronbun
Although the ungrammaticality in (89)a is also explained even if the null operator undergoes movement from the theta position, it is crucial that we can observe a contrast between (88) and (89).55 Thus, the facts in this subsection suggest that HERCs behave like long-distance clefts in that they are sensitive to highest clauses in terms of operator movement. Case conversion Recall that Watanabe (1996) argues that case conversion is allowed when some kind of Wh-agreement takes place, as discussed in Chapter 3. In Japanese, typical constructions which allow case conversion are relative clauses and other nominalized clauses: (90)
(91)
a
[Taroo-ga e katta] hon Taro-NOM bought book ‘the book that Taro bought’
b
[Taroo-no e katta] hon Taro–GEN bought book
a
[Hanako-ga kita] koto Hanako-NOM came fact ‘the fact that Hanako came’
Resumptive A’-Dependencies 153
b
[Hanako-no kita] koto Hanako-GEN came fact
Putting aside the complications that we already saw in the last chapter, we generally find nominative markers can be replaced by genitive markers, as shown in (90) and (91). What is interesting in ga-no conversion is that, as noted by Watanabe (1996: 390), it is only allowed in the highest clause, as shown in (92): 56 (92)
a
[[John-ga e katta to]i Mary-no ti omotteiru] hon John-NOM bought COMP Mary-GEN think book ‘the book that Mary thinks that John bought’
b*
[[John-no e katta to]i Mary-ga/no ti omotteiru] hon John-GEN bought COMP Mary-NOM/GEN think book
c*
[Mary-ga/no [John-no e katta to] omotteiru] hon Mary-NOM/GEN John-GEN bought COMP think book
The HERCs in (14) contain a complement clause. (14)a shows that the highest subject Mary can be marked by a genitive marker; however, the HERCs in (92)b and (92)c show that the subject inside the complement clause cannot be marked by a genitive marker. To be more precise, we can observe the highest clause sensitivity in the following three-layered HERCs: (93)
a?
[[John-ga e katta to]i Mary-no [Bill-ga ti itteita John-NOM bought COMP Mary-GEN Bill-NOM said to] omotteiru] hon COMP think book ‘the book that Mary thinks that Bill said that John bought’
b
[[Bill-ga Bill-NOM omotteiru] think
c*
[[John-ga e katta to]i Mary-ga/no [Bill-no ti John-NOM bought COMP Mary-NOM/GEN Bill-GEN itteita to] omotteiru] hon said COMP think book
[John-ga e katta to] itta to]i Mary-no ti John-NOM bought COMP said COMP Mary-GEN hon book
154 Cleft Constructions in Japanese Syntax
d*
[[Bill-no [John-ga e katta to] itta to]i Bill-GEN John-NOM bought COMP said COMP Mary-ga/no ti omotteiru] hon Mary-NOM/GEN think book
e*
[Mary-ga/no [Bill-no [John-ga e katta to] Mary-NOM/GEN Bill-GEN John-NOM bought COMP itta to] omotteiru] hon said COMP think book
The HERCs in (93) clearly show that only the highest nominative phrases can be replaced by no. The following are schematic structures for each example: (94)
a?
[CP1[CP3 . . .]i Mary-no [CP2 . . . ti . . .] . . .] hon
b
[CP1 [CP2 . . .]i Mary-no . . . ti . . .] hon
c*
[CP1 [CP3 . . .]i . . . [CP2 Bill-no ti . . .] . . .] hon
d*
[CP1 [CP2 Bill-no . . .]i . . . ti . . .] hon
e*
[CP1 . . . [CP2 Bill-no . . .] . . .] hon
The structures in (94) show that only the highest genitive subject (the one in CP1 and not those in CP2 and CP3) is allowed in the caseconversion constructions containing two embedded clauses. If Watanabe (1996) is correct in that ga-no conversion takes place when Wh-movement is involved, we can say that only the highest clause involves Wh-movement in these HERCs. Therefore, not only can our analysis account for long-distance cleft constructions but also it can extend to explain other types of constructions in Japanese. Although further research is required to verify this line of analysis, it is possible to assume that a null operator, in general, undergoes movement from the position adjoined to the highest CP in Japanese. Similar phenomena in Mooré There are other languages in which the verbal morphology changes in Wh-extraction domains. Those languages include, for example, Chamorro (Chung 1982, 1994), Edo (Stewart 1998), Hausa (Haïk 1990), Irish (McCloskey 1979) and Kikuyu (Haïk 1990). One such language is Mooré (Haïk 1990, Tellier 1989), where it is assumed that irrealis forms of verbs (or INFL) are instances of Wh-agreement:
Resumptive A’-Dependencies 155
(95)
a
a Pok yã-a /*yã ãnda zaame? Poko see-R/*see-IR who yesterday ‘Who did Poko see yesterday?’
b
ãnda (n) ri-a /*ria-lame? who eat-it-IR /*eat-it-R ‘Who ate it?’ (Haïk 1990: 354, 349)
The sentences in (95) are Wh-interrogative constructions. (95)a shows that when the Wh-phrase is in situ, the verb is realized as the realis form. On the other hand, when the Wh-phrase is preposed, the verbal morphology is affected and changes to the irrealis form. The following are instances of long-distance Wh-interrogative sentences in Mooré: (96)
a
bwε la fo yeel /*yeel-a t’ a Bil ri-i /*ri zaame? what FOC 2s say-IR /*say-R that Bila eat-R /*eatIR yesterday ‘What did you say that Bila ate yesterday?’
b bwεla fo yeel /* yeel-a t’ a Bil tagsame what FOC 2s say-IR/* say-R that Bilathought-R /*tags t’ a Pok ri-i/*ri zaame? /*thought-IR that Poko ate-R/*ate-IR yesterday ‘What did you say that Bila thought that Poko ate yesterday?’ (Haïk 1990: 354) The sentence in (96)a contains a complement clause, and (96)b has a complement clause in which another clause is embedded (threelayered). As we observe in both sentences, only the highest verb is realized as the irrealis form even though the Wh-phrase is associated to the most embedded object position. The same is true for head-internal relative clauses (HIRCs). In Mooré, the relative marker, sen ‘REL (Infl)’, appears inside the relative clause in HIRCs, agreeing with the internal head which is italicized below: (97)
m mii [rawã sen seg wã] biig ninga 1s see man REL meet child NINGA DET ‘I saw the child that the man met’ (Tellier 1989: 2)
156 Cleft Constructions in Japanese Syntax
According to Tellier (1989:15), when a HIRC has a complement clause, only the highest INFL shows agreement with the internal head, as shown in (98): (98)
a
mam sen mi ti fo sibga kamb ninsb wã waame 1s REL know that 2s punish children NINGA DET came ‘the children that I know that you punished came’
b
mam sen mi ti fo toola neb nins sebrã waame 1s REL know that 2s send people NINGA book-DET came ‘the people that I know that you sent the book to came’
c
mam sen mi ti fo karma sebra zilg ninga wã yaa 1s REL know that 2s read book place NINGA DET is noodo interesting ‘the people where I know that you read a book is interesting’
The HIRCs have their internal heads in the most embedded object positions. If we assume some agreement between the relative marker and the internal head, it seems that such agreement does not take place locally in the examples in (98). Here again, we can see the highest clause sensitivity in constructions which are generally assumed to have operator movement. It must be noted, however, that we cannot simply conclude from these facts that irrealis morphology on verbs indicates the domain of true operator movement. For example, Haïk (1990) argues that the intermediate trace is deleted in Mooré, and Watanabe (1996:187) suggests that morphological realization is different between intermediate traces and operators. In other words, the data here do not necessarily show the same syntactic configuration as the one of long-distance clefts in Japanese. In addition, the task remains to show why some languages, such as Japanese, show highest clause effects, and others, such as English, do not. However, I believe that this section at least gives us some indirect support for our analysis; the peculiar structures for long-distance clefts can be observed in other null operator constructions.
Summary Chapter 4 claimed that Japanese cleft constructions always involve A’-movement, especially, movement of a null operator in the presuppositional clauses. A dichotomy within the cleft constructions can be observed between local and long-distance clefts. Unlike local clefts,
Resumptive A’-Dependencies 157
long-distance clefts demonstrate resumptive A’-dependencies; an operator is base-generated in the position adjoined to the highest complement CP, binding pro in its thematic position. In other words, the operator movement starts from the position adjoined to the embedded clause rather than from its thematic position. This analysis is supported by reconstruction effects, WCO effects, and interactions with other A’-dependencies. Furthermore, the facts of true adjunct PPs in focus position in Japanese and some facts from other languages support our analysis. Chapter 4 shows that the structures for local and long-distance clefts are empirically well motivated. As for theoretical perspectives, short movement in the highest clause can be supported by the theory of economy (Chomsky 1995). In presuppositional clauses of cleft constructions, a structure with local null operator movement becomes the optimal one under the Minimal Link Condition. However, notice that it is not clear whether other structures involving a trace in the theta position are really comparable with our proposed structure in terms of numeration, because the structure in long-distance clefts contains a pro. The economy account, therefore, would become plausible only if we can assume that the two structures have the same numeration. I will leave this question open for future research.
This page intentionally left blank
5 Ellipsis in Cleft Constructions
Introduction Chapter 5 is an extension of our investigation of cleft constructions in Japanese; in particular, we will examine so-called ‘sluicing’ in the language. It has been proposed in previous research that, unlike English elided structures, Japanese sluicing is derived from cleft constructions (Kizu 1997, Kuwabara 1995, 1996, Nishiyama 1995, Nishiyama, Whitman and Yi 1996). In this analysis, presuppositional clauses are phonologically empty or deleted given an appropriate linguistic antecedent, and only focus phrases occur in elided sentences. Chapter 5 is based on the proposal in Kizu (1997), and provides further evidence for the analysis of sluicing as a cleft construction. On the basis of the proposed analysis, I will confirm the idea presented in Chapter 2 that focus phrases in clefts are not directly generated by scrambling, but are base-generated in predicative nominal position, and are associated with a null operator in the presuppositional clause by agreement. The organization of this chapter is as follows. First there is a presentation of the analysis introducing the argument for Japanese sluicing. This is followed by a look at basic facts about Japanese sluicing with a brief review of the arguments presented in Takahashi (1994). Chapter 5 goes on to present a significant parallelism between the sluicing and cleft constructions in Japanese, and then, on pages 175–86 discusses other issues which arise in the literature, namely, island (in)sensitivity and multiple Wh-foci. The penultimate section discusses some consequences of the proposed analysis, and argues that PF deletion operations can account for the facts correctly. The final section is a summary. 159
160
Cleft Constructions in Japanese Syntax
Background and a proposal The properties of ellipsis in English have been extensively discussed since the days of the standard theory (Ross 1967, Sag 1977, Williams 1977, Chao 1987, Fiengo and May 1994 among many others, and also see Johnson 1997 and references cited therein). For example, Lobeck (1991), assuming the Government and Binding theory, points out that a maximal projection (NP, VP and IP) governed by a certain functional head can be empty if and only if the functional head agrees with its specifier.1 Under Lobeck’s proposal, it is assumed that ellipsis is a unified syntactic phenomenon across categories. The next question to raise is whether this hypothesis is tenable in other languages, especially since there has been little agreement about the actual properties of functional categories in languages such as Japanese (Fukui 1986). One of the controversies in Japanese which is relevant to this claim is whether or not the language, which has the Wh-in situ property, permits overt Wh-movement when constructing elliptical sentences on a par with English IP ellipsis, or ‘sluicing’ in the terminology of Ross (1969). Takahashi (1994) argues that Japanese sluicing provides a piece of evidence that a remnant Wh-phrase in sluiced clauses occupies the spec of CP position, with the COMP triggering spec-head agreement at S-structure and thus licensing an empty IP complement.2 Therefore, Japanese sluiced clauses can be analyzed as having a structure parallel to English sluicing. Before going into Takahashi’s (1994) analysis, let us first look at English sluicing, as shown in (1) and (2): (1)
a b
Somebody just left. Guess who? . . . Guess [CP whoi [IP (ti just left)]]
(2)
a
Ralph is going to invite somebody from Kankakee to the party, but they don’t know who . . . they don’t know [CP whoi [IP (Ralph is going to invite ti . . .)]] (Ross 1969: 252)
b
The sluiced structures in (1)a and (2)a are represented as (1)b and (2)b, respectively. The clauses in brackets are either deleted at PF or reconstructed by copying the antecedents, depending on analyses.3 Following the analogy with the analysis of Wh-interrogative constructions generally assumed, we find that the Wh-remnants, who and what, in (1)a and (2)a
Ellipsis in Cleft Constructions
161
are situated in the spec of CP at S-structure or before spell-out, as shown in (1)b and (2)b. Takahashi (1994) claims that the example of Japanese sluicing in (3)a and (4)a can be represented analogously to English sluicing, as shown in (3)b and (4)b. Here, contrary to my proposal in this chapter, Wh-remnants move to the spec of CP of the complement clause and the IP selected by the C-head gets deleted:4 (3)
(4)
a
Dareka-ga ki-ta rasii ga, somebody-NOM come-PAST I-heard but watasi-wa dare(*-ga) (-da)-ka siranai I-TOP who(NOM) (COP)Q don’t know ‘I heard that somebody came, but I don’t know who’
b
. . . watasi-wa [CP dare [IP (ti kita)]-ka] . . . I-TOP who came -Q
a
Taroo-ga nanika-o yon-da rashii ga, Taro-NOM something-ACC read-PAST I-heard but watasi-wa nani(*-o)(-da)-ka wakaranai I-TOP what(ACC)(COP)-Q don’t-know ‘I heard that Taro read something, but I don’t know what’
b
. . . watasi-wa [CPnani(-o) [IP (Taroo-ga ti yonda)]-ka] I-TOP what-ACC Taro-NOM read-Q wakaranai don’t-know
The sentences in (3) and (4) show that the Wh-remnants, dare ‘who’ and nani ‘what’, occupy the spec of CP. Spec-head agreement between the interrogative C0 and the Wh-phrase in its spec position makes the ensuing empty IP licit. It appears that Japanese sluicing can be accounted for by the same analysis proposed for English sluicing, assuming that Japanese allows overt Wh-movement. In Chapter 5, contrary to Takahashi’s analysis, I would like to claim that Japanese sluicing does not involve overt Wh-movement, but instead involves a cleft construction. Given that Japanese sluicing is explained under a PF deletion approach, which I will come back to discuss on pages 189–97, the derivational source for (4)a is illustrated in the following:
162
Cleft Constructions in Japanese Syntax
(5)
a
. . . watasi-wa [CP2[CP1Opi[IP Taroo-ga ti yon-da]-noi]-wa I-TOP Taro-NOM read-PAST-NM-TOP [nanii-(da)]-ka] wakaranai what-COP-Q don’t-know ‘I don’t know what it is that Taro read’
b
CP2 C’ TopP Delete
CP1
C ka(Q) IP I’
C’
OPi IP
C no i(NM)
...ei...
VP
I V’
nanii(what) da(COP)
I propose that the elided sentence in (4)a has a derivational source something like (5)a. The structure of the bold-faced type of (5)a is illustrated in (5)b. The structure in (5)b shows that the topicalized CP1 (Taroo-ga yon-da no ‘what Taro read’), which serves as a presupposition in the cleft construction, becomes empty when it has a linguistic antecedent. Only the Wh-phrase in focus position with the copula appears phonologically in this construction. My proposal is, therefore, distinct from Takahashi (1994), and supports insights presented by Kuwabara (1995, 1996), Nishiyama (1995), and Nishiyama, Whitman and Yi (1996), in which various correlations between sluicing and cleft constructions are discussed.5 In addition to the discoveries presented in Kizu (1996), I will show some empirical evidence that Japanese sluicing should be accounted for by a PF deletion operation. While the present discussion is based primarily on Japanese sluicing, it is my hypothesis that the analysis can be extended to other Wh-in situ languages.
Basic facts about Japanese sluicing This subsection reviews some of the basic facts about Japanese sluicing, which are also observed in Takahashi (1994). The fact that Japanese
Ellipsis in Cleft Constructions
163
sluicing shares some properties of English sluicing suggests that the constructions in Japanese are instances of ellipsis; however, it is crucial that our proposal does not assume that Japanese sluicing is analyzed on a par with English sluicing. As pointed out by Takahashi, Japanese sluicing requires a linguistic antecedent just as in English (Hankamer and Sag 1976); this is shown in (6) and (7): (6)
(7)
a
Taroo-wa dareka-ni syukudai-o tetsudatte-moratta Taro-TOP someone-DAT assignment-ACC was-helped yoo-da seems ‘It seems that Taro had somebody to help his assignment’
b
Dare(-ni)-ka wakaranai who(-DAT)-Q don’t-know-Part ‘I don’t know who’
[When the speaker found out Taro did a very good job on his assignment] #Dare(-ni)-ka wakaranai ‘I don’t know who’
The sentence in (6)b is uttered following the linguistic antecedent (6)a, and (7) is uttered without such an antecedent. Interestingly, the former sounds appropriate whereas the latter does not.6 The same is pointed out for English sluicing by Hankamer and Sag (1976), who argue that sluicing cannot be controlled by pragmatics but must be licensed by syntax.7 Moreover, as in English sluicing, Japanese sluicing permits so-called sloppy identity. Notice that the following sentence in (8)a is ambiguously interpreted, as translated in (8)b and (8)c: (8)
a
Tarooi-wa [[zibuni-ga naze sikarare-ta] -ka] wakatteiru ga, Taro-TOP self-NOM why be-scold-PAST-Q know but Hanako-wa [naze(-da)-ka] wakaranai Hanako-TOP why-COP-Q doesn’t-know ‘Taro knows why he was scolded, but Hanako doesn’t know why’
b
Hanakoi doesn’t know why shei was scolded
c
Hanako doesn’t know why Taro was scolded
The sentence in (8)a, where zibun ‘self’ can be interpreted as a bound variable, shows that the sluiced clause permits both a sloppy reading, as
164
Cleft Constructions in Japanese Syntax
shown in the English translation in (8)b, and a strict reading, as in (8)c. It is well known that the same results can be obtained in English sluicing (Ross 1967 and others), and hence, Japanese sluicing appears to exhibit properties also found in English ellipsis. Furthermore, Takahashi (1994: 269) provides the same kind of syntactic behavior using another type of pronoun, soko ‘it, that place’, which can be interpreted as a bound variable:8 (9)
(10)
a
UConni-ga [sokoi-no basukettobooru tiimu-ga UConn-NOM it-GEN basketball team-NOM dare-o sukautosita ka] happyoosita who-ACC scouted Q announced ‘UConn announced who its basketball team scouted’
b
Duke-mo [dare-ka] happyoosita Duke-also who-Q announced ‘Duke also announced who’
a
Duke also announced who Duke’s basketball team had scouted
b
Duke also announced who UConn’s basketball team had scouted
When the sentence in (9)b is uttered after (9)a, it is interpreted either as (10)a or (10)b: the former is a sloppy reading and the latter a strict reading. Roughly speaking, under the analysis of PF deletion operation, there are two distinct underlying structures possible for the sluiced sentence in (9)a: one is similar to (10)a and the other is like (10)b.9 In addition to the facts of sloppy identity above, Japanese sluiced clauses exhibit another syntactic characteristic in common with English sluicing: sluicing is only allowed if the empty clause is followed or preceded by certain types of complementizers. In other words, according to Lobeck (1991, 1995) and Takahashi (1994), Wh-elements in the spec of CP can license the elided clause but other complementizers cannot, as in the following: (11)
a*
Taroo-ga hon-o yon-da to it-ta ga, Taro-NOM book-ACC read-PAST COMP say-PAST but watasi-wa [[e] kadooka] wakaranai I-TOP whether don’t-know ‘*Taro said that he read a book, but I don’t know whether/if’
Ellipsis in Cleft Constructions
b*
165
Taroo-ga hon-o yon-da to iw-anak-atta ga, Taro-NOM book-ACC read-PAST COMP say-NEG-PAST but watasi-wa [[e] to] omou I-TOP that think ‘*Taro didn’t say that he read a book, but I think that’
The English translations in (11) show that neither whether/if nor that can license an empty complement IP, and the same turns out to be true for Japanese sluiced clauses containing the corresponding complementizers, kadooka ‘whether/if’ and to ‘that’. This fact can be accounted for by the analysis proposed by Lobeck (1991, 1995); since whether/if and that are not proper governors for licensing the empty IP, the ungrammaticality of (11) is expected.10 Therefore, it appears that these examples suggest that Japanese sluicing may be further analogous to English sluicing; Japanese sluicing exhibits many of the characteristic properties of ellipsis. Based on the observations above, Takahashi (1994) argues that the Japanese IP can be deleted when selected by a [+WH] complementizer which agrees with a Wh-remnant in the spec position. Moreover, he points out that Japanese sluicing exhibits island effects and reconstruction effects, which crucially show that the remnant Wh-phrase undergoes syntactic movement to the spec of CP. We will come back to these issues on pages 175–203.
Sluicing as a cleft construction While the argument by Takahashi (1994) poses an interesting issue concerning overt Wh-movement in languages with the Wh-in situ property, it is not clear why sluicing should trigger overt Wh-movement in Japanese, given that such movement does not occur (or, at least, it is not obligatory) in ordinary Wh-questions in this language.11 In other words, the Wh-movement analysis is not the only way to deal with sluicing in Japanese and other Wh-in situ languages. This section presents striking distributional similarities between sluiced sentences and cleft constructions. I will claim that Japanese sluicing is derived from a cleft construction, and will argue against the Wh-movement analysis. The copula in sluiced clauses First, as pointed out by Kuwabara (1995), Nishiyama (1995), and Nishiyama et al. (1996), the Japanese sluiced clause optionally contains the
166
Cleft Constructions in Japanese Syntax
copula da, as shown in the examples in the previous section. Another instance is illustrated in (12)a below. Notice that the copula appears in cleft constructions, as in (12)b, and the embedded cleft sentence also allows an optional copula, as shown in (12)c:
(12)
a
Dareka-ga sono hon-o yonda rasii ga, someone-NOM that book-ACC read I-heard but watasi-wa [dare(-da)-ka] wakaranai I-TOP who COP-Q don’t-know ‘I heard that someone read that book, but I don’t know who’
b
[CPOpi[IP ei sono hon-o yonda]-no]-wa Tarooi-da that book-ACC read -NM-TOP Taro-COP ‘It is Taro who read that book’
c
Watasi-wa [CP[CPOpi[IP ei sono hon-o yonda]-no]-wa I-TOP that book-ACC read-NM-TOP Tarooi(-da)-to] omou Taro-COP-COMP think ‘(I) think that it is Taro that read that book’
If the sentence in (12)a involves overt Wh-movement, the derivational source of the sluiced clause with a copula after Wh-movement should be represented either as (13)a or (13)b. However, under Takahashi’s (1994) Wh-movement analysis, there is no convenient place to generate this copular element:
(13)
a
*. . . [CP dare-gai[IP ti sono hon-o yonda-da] ka] . . . who-NOM that book-ACC read-COP -Q ‘. . . who read that book . . .’
b
*. . . [CP dare-dai[IP ti sono hon-o yonda] ka]. . . who-COP that book-ACC read -Q
Neither sentence in (13) is grammatical, because the copula da cannot be attached to the verb, as shown in (13)a, or the Wh-phrase with a copula cannot be preposed in Japanese, as in (13)b. Therefore, the sentences in (13) cannot be the underlying structures for (12)a.12
Ellipsis in Cleft Constructions
167
Notice also that the copula can be inflected with a past tense morpheme in both the sluiced clause and the cleft sentence, as shown in (14): (14)
a
Watasi-wa kyonen gakkai-de dareka-ni atta ga, I-TOP last-year conference-at who-DAT met but [dare(-ni) dat-ta -ka] oboeteinai who-DAT COP-PAST-Q don’t-remember ‘I met someone at the conference last year, but (I) don’t remember who (it was)’
b
[CPOpi[IP Kyonen gakkai-de ei atta]-noi]-wa Hanakoi(-ni) last-year conference-at met-NM-TOP Hanako-DAT dat-ta COP-PAST ‘It was Hanako who (I) met at the conference last year’
The sentence in (14)a involves a sluiced clause, and the one in (14)b a cleft construction. Since Wh-phrases in interrogative sentences do not occur with a copula, as we observed in (13), the Wh-remnant in sluicing examples in (12)a and (14)a differs syntactically from the Wh-phrase in standard interrogative sentences, but is similar to the cleft sentences. This is predicted readily from the assumption that Japanese sluicing is derived from a cleft construction; as we saw in (12)c, the copula in the embedded clause also occurs optionally. Interestingly, other Wh-in situ languages, such as Korean and Chinese, show that sluiced clauses must contain a copula as well: (15)
Motwu-nun John-i nwukunka-lul salanghan-ta-ko everyone-TOP John-NOM someone-ACC love-ind-COMP malha-ciman na-nun [nwukwu *(i-n)-ci] molu-n-ta say but I-TOP who- COP Q don’t-know ‘Everyone said John loves someone, but I don’t know who’
(16)
Meige ren dou shuo Zhangsan ai-shang le shenme ren le, each person all say Zhangsan love-up PFT what man PFT keshi meiren zhidao [*(shi) shei] but no one know be who ‘Everyone said that Zhangsan fell in love with someone, but no one knew who’ (Nishiyama 1995)
168
Cleft Constructions in Japanese Syntax
The sentence in (15) is a Korean sluicing example, and the one in (16) is an example of Chinese sluicing. The copula in sluicing constructions in these languages is obligatory, unlike Japanese sluiced sentences. In addition, I ‘be’ in Korean and shi ‘be’ in Chinese are found in the corresponding cleft constructions, as shown in (17) and (18):13, 14
(17)
John-i mek-un kes-un sakwa i-ta John NOM eat NM-TOP apple be-Dec ‘What John ate is an apple’
(18)
wo kanjian diao dao wuding le Shi shuzhi be tree branch I see fall onto roof ASP ‘It was a tree branch that I saw fall on the roof’ (Hoh and Chiang 1990: 48)
Another Wh-in situ language, Turkish, exhibits facts similar to those seen above for Japanese sluicing and cleft constructions. While the Turkish copula does not appear in present tense sluiced clauses because it is generally phonetically null, the past tense copula is found both in sluiced and cleft constructions:15
(19)
a
Mehmet birsey söyle-di ama, ne-y-di hatirla Mehmet one thing told but what-bePAST remember -mi -yor -um -NEG -PROG -1st.sg. ‘Mehmet said something, but I don’t remember what (it) was’
b
Mehmet’un kir -dig -i bir vazo-y-du Mehmet-GEN break-NM-POSS one vase -bePAST ‘What Mehmet broke was a vase’
In both the sluiced sentence in (19)a and the cleft construction in (19)b, the same copular verb is observed. Therefore, putting aside the issue of optionality of the Japanese copula, it seems that the existence of copula is observed widely in sluicing constructions in Wh-in situ languages. This fact supports the idea that sluicing involves a cleft construction since a cleft sentence is a certain type of copular construction.
Ellipsis in Cleft Constructions
169
Case marker drop in sluicing The second issue concerns a case marker occurring in the focus position of cleft constructions as well as in the remnant position of sluicing. It should be pointed out that most native speakers of Japanese that I consulted prefer Wh-remnants without a major case marker (nominative ga or accusative o) to those with such a case marker. This contrast is illustrated in (20): (20)
a
Dareka-ga Taroo-o nagut-ta ga, watasi-wa someone-NOM Taro-ACC hit-PAST but I-TOP [dare(*-ga)]-ka siranai who-NOM -Q don’t-know ‘Someone hit Taro, but I don’t know who’
b
Taroo-ga dareka-o nagut-ta ga, watasi-wa Taro-NOM someone-ACC hit-PAST but I-TOP [dare(*-o)]-ka siranai who-ACC-Q don’t-know ‘Taro hit someone, but I don’t know who’
c
Taroo-ga hon-o dareka-ni age-ta ga, Taro-NOM book-ACC someone-DAT give-PAST but watasi-wa [dare(-ni)]-ka siranai I-TOP who-DAT-Q don’t-kn*-ow ‘Taro gave a book to someone, but I don’t know who’
In the sentences in (20)a and (20)b, when the case marker, ga or o, is attached to the Wh-remnant, the grammaticality of the sentence becomes much degraded. The dative marker in (20)c is truly optional; either with or without the case marker, the sentence is fully grammatical. This fact follows naturally under our cleft construction analysis. The speakers who have the judgments in (20) also have similar judgments for the following corresponding cleft sentences: (21)
a
[CP[CP Opi [IP ei Taroo-o nagut-ta]-no]-wa Taro-ACC hit-NM-TOP siranai don’t-know ‘(I) don’t know who it is that hit Taro’
[dare(*-ga)]ika] who-NOM -Q
170
Cleft Constructions in Japanese Syntax
b
[CP[CP Opi [IP Taroo-ga ei nagut-ta]-no]-wa [dare (*-o)]i-ka] Taro-NOM hit-PAST-NM-TOP who-ACC-Q siranai don’t-know ‘(I) don’t know who it is that Taro hit’
c
[CP[CP Opi[IP Taroo-ga hon-o ei age-ta]-no]-wa Taro-NOM book-ACC give-PAST-NM-TOP [dare(-ni)]i ka] siranai who-DAT-Q don’t-know ‘(I) don’t know who it is that Taro gave a book’
The focus phrase, dare ‘who’, cannot be marked by a nominative or accusative case particle, as shown in (21)a and (21)b. On the other hand, the focus phrase without such a case marker makes the sentence grammatical. (21)c shows that the dative marker is optional on the focus phrase, just like the corresponding sluicing sentence. This is perfectly parallel to (20).16 In contrast, if a Wh-phrase is preposed by scrambling (or by Wh-movement) to sentence-initial position, it must be marked by a case marker: (22)
a
[CP
dare*(-ga)i [IP ti Taroo-o nagut-ta]-ka] siranai who-NOM Taro-ACC hit-PAST-Q don’t-know ‘(lit.) I don’t know who hit Taro’
b
[CP dare*(-o)i [IP Taroo-ga ti nagut-ta]-ka] siranai who-ACC Taro-NOM hit-PAST-Q don’t-know ‘(lit.) I don’t know who Taro hit’
c
[CP dare*(-ni) [IP Taroo-ga hon-o ti age-ta]-ka] who-DAT Taro-NOM book-ACC give-PAST-Q siranai don’t-know ‘(I) don’t know to whom Taro gave a book’
If we assume that the sentences in (22) are derivational sources for (20), we need to explain why these case-markers obligatorily or optionally disappear in sluicing. What we can find in (21) and (22) is, therefore, that the sluiced clauses behave more like focus constituents in cleft constructions rather than Wh-constituents created by movement. These facts present another piece of empirical evidence for the cleft analysis.17, 18
Ellipsis in Cleft Constructions
171
Ordering restriction on Wh–remnants We observed in Chapter 2 that focus phrases of clefts have a restriction on the order of a numeral quantifier (NQ) and its associate NP; the NQ must follow the NP. The relevant examples are reproduced here in (23): (23)
a
[CPOpi [IP Taroo-ga ei katta]-no]-wa [hon-o takusan]i Taro-NOM bought-NM-TOP bookACC many da-ta COP-PAST ‘It was many books that Taro bought’
b*
[CPOpi [IP Taroo-ga ei katta]-no]-wa [takusan hon(-o)]i Taro-NOM bought-NM-TOP many bookACC da-ta COP-PAST
It has been argued in Kitahara (1993), Koizumi (1995) and references therein that the ‘NPi-CASE + NQi’ sequence in (23)a forms a single constituent, whereas the ‘NQi + NPi(-CASE)’ in the (23)b sentence consists of two separate constituents. This is supported by the fact that the former string can be coordinated with to ‘and’, but the latter sequence cannot. For example, ‘hon-o san-satsu to zasshi-o ni-satsu (bookACC threeCL and magazineACC twoCL)’ is grammatical; however, the converse order ‘*san-satsu hon-o to ni-satsu zasshi-o (threeCL bookACC and twoCL magazineACC)’ is ruled out. Since only a single constituent can occupy focus position, (23)b is ruled out. What is interesting is that the same ordering restriction holds for a Wh-remnant of a sluiced sentence: (24)
a
Taroo-ga takusan nanika-o katta rasii ga, Taro-NOM many something-ACC bought I-heard but watasi-wa [nani-o takusan-ka] wakaranai I-TOP what-ACC many-Q don’t know ‘I heard that Taro bought a lot of something, but I don’t know many of what.
b*
Taroo-ga takusan nanika-o Taro-NOMmany something-ACC watasi-wa [takusan nani(-o) -ka] I-TOP many what-ACC -Q
katta rasii ga, bought I-heard but wakaranai don’t know
172
Cleft Constructions in Japanese Syntax
In the first half of (24)a and (24)b, the object, nanika-o ‘something-ACC’, and its associate quantifier, takusan ‘a lot’, can occur in either order, so that nanika-o takusan ‘something-ACC many’ and takusan nanika-o ‘many something-ACC’ in the antecedent sentences are both grammatical under the identical interpretation. However, the Wh-remnant in a sluiced clause must precede the numeral quantifier, with the converse order being ungrammatical, as shown in the second half of (24)a and (24)b. Therefore, the remnants in the sluiced clauses in (24), as well as the focus phrases of the cleft sentences in (23), must be single constituents. However, as we already discussed in Chapter 2 for cleft constructions, this is not the case when an NQ and its associated NP are fronted by scrambling in an embedded complement clause: (25)
a
b
[nani-o takusan[IP Taroo-ga t t katta]-ka] wakaranai what-ACC many Taro-NOM bought-Q don’t-know ‘(roughly=) (I) don’t know what Taro bought many of’ [takusan nani-o [IP Taroo-ga t t katta]-ka] wakaranai many what-ACC Taro-NOM bought-Q don’t-know
The grammaticality of the sentences in (25) contrasts with that of (24); although the NQ-NP sequence is not allowed in the sluiced sentence, the NQ-NP sequence is perfect when the following elements are fully overt in the clause, as shown in (25)b. If we assume that sluicing involves overt Wh-movement, we should explain why the preposed NQ-Wh-phrase sequence in (25)b cannot be the underlying structure for the sluiced clause.19 The data in (24) are explained naturally under the cleft analysis. On such a view, the underlying sluiced sentence in (24)a is represented as follows: (26)
[nani-o [CP1[CP2 Opi [IP Taroo-ga ei katta]-no]-wa Taro-Nom bought-NM-TOP what-ACC takusan]i-ka wakaranai many-Q I-don’t-know ‘(I) don’t know what it was that Taro bought many of’
This analysis does not need further mechanisms to account for the ordering restriction. The ungrammatical sequence in (24)b, namely, takusan nani-o ‘many whatACC’, is not generated because the Wh-remnant must be underlyingly a single constituent of the cleft construction’s focus phrase.
Ellipsis in Cleft Constructions
173
Concessive Wh-phrases and non-Wh-remnants Although English sluicing is licensed solely by a strong [+WH] head which agrees with a Wh-phrase in its spec position (Lobeck 1991, 1995), the Japanese sluiced construction does not necessarily contain the same kind of [+WH] head in it. As shown in the following examples, the concessive Wh-phrases, dare-demo ‘whoever’ or nan-demo ‘whatever’, can also license an empty clause in Japanese unlike English sluicing: (27)
a
Dareka-ga conpuutaa-ga kai-tai to itta ga, someone-NOM computer-NOM buy-want COMP said but watasi-wa [dare-demo] kamawanai I-TOP who-no-matter don’t-care ‘Someone said that (he/she) wants to buy a computer, but I don’t care *whoever/who’
b
Taroo-ga nanika-ga yomi-tai to itta ga, Taro-NOM something-NOM read-want COMP said but watasi-wa [nan(i)-demo] kamawanai I-NOM what-no-matter don’t-care ‘*Taro said that (he) wants to read something, but I don’t care *whatever/what’
Notice that concessive Wh-phrases cannot license sluicing in English, as the translations of (27) show. Putting aside the reason why these English sentences are ungrammatical, we can at least say that if Japanese sluicing were to be analyzed in the same way as English sluicing with respect to the licensing conditions on ellipsis, this kind of non-interrogative Wh-phrase should not be able to license the empty IP.20 In contrast, if we assume that the sluiced sentences in Japanese are derived from a cleft construction, the grammaticality of (27) can be readily predicted, because dare-demo ‘whoever’ and nan(i)-demo ‘whatever’ can occur grammatically in focus position of a cleft sentence: (28)
a
[CP [IP ei ei conpuutaa-ga kai-tai]-no]-wa computer-NOM buy-want-NM-TOP [dare-demo]i-da who-no-matter-COP ‘(*)Who wants to buy a computer is anybody’
b
[CP [IP Taroo-ga ei yomi-tai]-no]-wa [nan(i)-demo]i-da Taro-NOM read-want-NM-TOP what-no-matter-COP ‘What Taro wants to read is anything’
174
Cleft Constructions in Japanese Syntax
Thus, by analyzing the sentences in (28) as the underlying structures for (27), and assuming that the sluiced sentences have an empty presuppositional clause, we can successfully account for the licensing ability of non-interrogative Wh-phrases in a way that appears not to be possible under the overt Wh-movement analysis. Extending this line of reasoning, the cleft construction analysis of sluicing can account for the following elliptical sentences where the remnant is not a Wh-phrase of any kind: (29)
a
Taroo-ga Hanako-no ronbun-o hometa rasii ga, Taro-NOM Hanako-GEN paper-ACC praised I-heard but watasi-wa [zettai-ni Taroo(-da)-ka] utagatteiru I-TOP definitely Taro-COP-Q doubt ‘I heard that Taro praised Hanako’s paper, but I doubt that (it is) definitely Taro’
b
Taroo-ga Hanako-ni sono kuruma-o ageta rasii ga, Taro-NOM Hanako-DAT that car-ACC gave I-heard but watasi-wa [tasikani kuruma(-da) ka] siritai I-TOP surely car-COP-Q want-to-know ‘I heard that Taro gave the car to Hanako, but I want to know if (it is) surely the car’
c
Watasi-wa Hanako-ni niteiru hito-o mita ga, I-TOP Hanako-DAT resemble person-ACC saw but [hontooni Hanako(-da)-ka] wakaranai truly Hanako-COP-Q don’t-know ‘I saw a person who resembled Hanako, but I don’t know if (it is) truly Hanako’
In (29), what is required to produce a sluiced clause containing a nonWh-phrase is that an adverb, which has some kind of emphatic or focal connotation, supports the occurrence of the remnant (non-Wh-)NP. Here, if zettaini ‘definitely’, tasikani ‘surely’ and hontooni ‘truly’ are not realized in the elided clause, the sentences in (29) become less acceptable, although even then they are not totally ungrammatical. These examples suggest that the remnant phrase has to be focused with some contrastive relationship to the first sentence. This fact is naturally explained under the cleft analysis on sluicing since the remnant of the sluiced clause is equivalent to the focus portion of a cleft construction.
Ellipsis in Cleft Constructions
(30)
175
[CP[CP Opi [IP watasi-ga ei mita]-no]-wa hontooni [Hanako]i da-ka] I-NOM saw-NM-TOP truly Hanako-COP-Q ‘. . . what I saw is truly Hanako . . .’
In (30), adverb hontooni ‘truly’ adds a strong focal connotation to the meaning of the focus clause; however, this is not required when the presuppositional clause is phonologically realized. While I do not have an exact account for why a focal adverb should appear in the sluiced clauses, it seems that the overt occurrence of the presuppositional clause also contributes to the meaning of the focus position of clefts; the meaning of focus is understood by the presence of a presuppositional clause, but it is not only by the overt occurrence of a phrase in focus position. Thus, an element in focus position of clefts must be inherently focused (like Wh-constituents and quantified phrases), or it should be supported by a focal adverb. The facts above have shown that our analysis can account not only for typical sluicing involving a Wh-remnant but also for other kinds of ellipsis in Japanese. The Wh-movement analysis has some difficulty in dealing with the latter case, because they are not [+WH] elements. To sum up so far, we have observed various similarities between sluicing and cleft constructions in Japanese, and have proposed that a sluiced clause is derived not by overt Wh-movement or scrambling, but rather, from a cleft sentence. In this section, we have seen in addition that our analysis can extend to sluiced clauses containing a non-Wh-remnant, which cannot be accounted for by a Wh-movement analysis of sluicing.
Other issues in Japanese sluicing Island effects This subsection is concerned with island effects in sluicing. In the literature, there seem to be different views for the facts of Japanese sluicing with respect to island effects. Here, I would like to show that island effects are in fact observed in sluicing in Japanese, which are also accounted for by the cleft analysis. Takahashi (1994) points out that Japanese sluicing exhibits island effects, as shown in (31)–(33):21, 22 (31)
a
Mary-ga [NP[CP John-ga nanika-o katta to]-no Mary-NOM John-NOM something-ACC bought that-GEN uwasa]-o kiita sooda rumor-ACC heard I-heard ‘I heard that Mary heard the rumor that John bought something’
176
(32)
(33)
Cleft Constructions in Japanese Syntax
b?*
Watasi-wa [nani-ka] siritai I-TOP what-Q want-to-know ‘I want to know what’
a
Mary-ga [NP [CP John-ni nanika-o ageta] Mary-NOM John-DAT something-ACC gave onna]-ni atta woman-DAT met ‘I heard that Mary met the woman who gave something to John’
b?*
Watasi-wa [nani-ka] siritai I-TOP what-Q want-to-know ‘I want to know what’
a
b?*
Mary-ga [PP dareka-ga kubi-ni natta kara] Mary-NOM someone-NOM was-fired because okotteiru sooda is-angry I-heard ‘I heard that Mary was angry because somebody was fired’ Watasi-wa [dare-ka] siritai I-TOP who-Q want-to-know
The sentences in (31) contain a noun complement, (32) a relative clause, and (33) an adjunct clause. Since the remnant Wh-phrases are all associated with positions inside syntactic islands, the sentences in (31)–(33) suggest that Japanese sluicing involves some kind of movement. We should note that, unlike the judgments in (31)–(33), some native speakers of Japanese turn out to accept the sentences in which a sluiced IP appears within a syntactic island (see, for instance, Nishiyama 1995). I would like to argue, assuming a deletion operation under identity, that this is due to a possible derivational source which does not in fact contain an island. Let us look at the following long-distance sluicing sentence, where the antecedent sentence contains a complement clause:
Ellipsis in Cleft Constructions
(34)
177
a
nanika-o kat-ta] Taroo-ga [CP[IP Hanako-ga Taro-NOM Hanako-NOM something-ACC buy-PAST to] it-ta ga, watasi-wa [nani-ka] wakaranakat-ta COMP say-PAST but I-TOP what -Q don’t-know-PAST ‘Taro said that Hanako bought something, but I didn’t understand what’
b
[CPOpi[IP Taroo-ga [CPti [IP Hanako-ga proi kat-ta]-to] Taro-NOM Hanako-NOM buy-PAST-COMP it-ta] no]-wa [nani]i -ka said NM-TOP what -Q ‘. . . what it is that Taro said that Hanako bought . . .’
(34)a is grammatical as is the corresponding cleft sentence in (34)b, which is supposed to be one of the possible derivational sources of the sluiced clause in (34)a under our current analysis. The remnant Wh-phrase can get its antecedent within the embedded complement clause, as shown in (34)b, and the intended meaning of the sluiced clause is ‘I didn’t understand what (it is that) Taro said that Hanako bought’. The grammaticality of the sentence under this reading suggests that a longdistance dependency is possible in Japanese sluicing and clefts; as for the latter, we already saw this specifically in Chapter 4. Notice also that there is another possible derivational source for the sluiced sentence in (34)a. The following representation, in addition to (34)b, could also be possible:
(35)
. . . [CPOpi[IP
Hanako-ga ti kat-ta]-no]-wa [nani]i -ka . . . Hanako-NOM buy-PAST-NM-TOP what-Q ‘. . . what it is that Hanako bought . . .’
The representation in (35) shows that the gap associated with the focus phrase is not deeply embedded; rather, the hidden presuppositional clause contains only the embedded CP of the antecedent sentence in (34)a. The structure in (35) gives the interpretation something like ‘I didn’t understand what (it is that) Hanako bought’. This should be licit as the underlying structure, because the presuppositional clause is identical to one of the clauses in the preceding sentence, which makes the deletion possible. Thus, it has two possible antecedents in the linguistic context.23
178
Cleft Constructions in Japanese Syntax
Keeping this in mind, let us look at the sluicing sentences containing a complex NP. If the representation in (35) is also available as a derivational source for sluicing, the sluiced clause in (36)a can have two distinct underlying representations, as shown in (36)b and (36)c. We assume that a sluiced sentence is derived from a cleft counterpart, in which the presuppositional clause is deleted at PF. (36)b is the case in which the higher IP, including the complex NP, occupies the presuppositional slot. On the other hand, (36)c has only the lower IP, which does not involve any island: (36)
a
#[IP2 Taroo-ga[NP[CP[IP1 Hanako-ga nanika-o Taro-NOM Hanako-NOM something-ACC kat-ta] to-yuu] uwasa]-o sinjiteiru] ga, watasi-wa bought COMP rumor-ACC believe but I-TOP [nani-ka] wakaranai what-Q don’t-know ‘Taro believes the rumor that Hanako bought something, but I don’t know what’
b
*. . . [CPOpi[IP2 Taroo-ga [NP[IP1 Hanako-ga proi kat-ta to-yuu] Taro-NOM Hanako-NOM bought-COMP uwasa]-o sinjiteiru]-no]-wa [nani]i-ka . . . rumor-ACC believe -NM-TOP what-Q ‘. . . what it is that Taro believes the rumor that Hanako bought . . .’
c
. . . [CPOpi[IP1 Hanako-ga ti kat-ta]-no]-wa [nani]i-ka . . . Hanako-NOM bought-NM-TOP what-Q ‘. . . what it is that Hanako bought . . .’
The two possible underlying structures for the sluiced sentence in (36)a are illustrated in (36)b and (36)c. If the interpretation of the second half of the sentence in (36)a is the one in (36)b in which the IP2 is generated in the presuppositional clause, the sluiced sentence should be ruled out. This is because, as discussed on pages 146–81 in Chapter 4, the null operator base-generated in the intermediate position cannot be associated with the pro in the island to form a binding chain (Cinque 1990). However, the structure in (36)c, in which only the lower IP1 appears underlyingly, is also possible as a derivational source for the sluiced clause in the context of (36)a.24 The sentence is acceptable under this analysis. While it is said that unlike cleft constructions, sluiced sentences are free from
Ellipsis in Cleft Constructions
179
island constraints (Nishiyama 1995), we should be aware of the fact that this is due to the fact that the sentence is acceptable with the reading in (36)c. Essentially, the same situation obtains for sentences which contain an adjunct island, such as (37): (37)
a
#[IP2 Taroo-ga [PP[IP1 Hanako-ga nanika-o katta] Taro-NOM Hanako-NOM something-ACC bought kara] okotteiru] rasii ga, watasi-wa [nani-ka] because is-angry I-heard but I-TOP what-Q siranai don’t-know ‘I heard that Taro is angry because Hanako bought something, but I don’t know what’
b
*[CPOpi[IP2 Taroo-ga[PP Hanako-ga proi kat-ta kara] Taro-NOM Hanako-NOM bought because okotteiru]-no]-wa [nani]i–ka is-angry-NM-TOP what Q ‘. . . what it is that Taro is angry because Hanako bought . . .’
c
. . . [CPOpi[IP1 Hanako-ga ei kat-ta]-no]-wa [nani]i-ka . . . Hanako-NOM bought-NM-TOP what-Q ‘. . . what it is that Hanako bought . . .’
Again, (37)a can be grammatical only under the interpretation in (37)c where the lower clause, IP1, appears as a derivational source. (37)b, on the other hand, is ruled out since the null operator cannot form a binding chain with the pro inside the adjunct clause. Interestingly, a sluiced sentence with a relative clause is not as good as one with a complex NP or an adjunct clause, even for those speakers who accept (36)a and (37)a. Consider the following examples: (38)
a
?*[IP2 Hanako-ga [NP[IP1 ei Taroo-ni nanika-o age-ta] Hanako-NOM Taro-DAT something-ACC give-PAST at-ta] sooda ga, watasi-wa hitoi]-ni person-DAT meet-PAST I-heard but I-TOP [nani-ka] wakaranai what-Q don’t-know ‘I heard that Hanako met a person who gave something to Taro, but I don’t know what’
180
Cleft Constructions in Japanese Syntax
In fact, the ungrammaticality of (38)a can be directly explained under the proposed cleft analysis for sluicing. As we have seen above, the derivational source in (38)b below is ruled out because the null operator cannot be associated with the pro in the complex NP. However, unlike (36)c and (37)c, the structure in (38)c, where only IP1 is included in the presuppositional clause, is not available either: (38)
b
*. . . [CPOpj[IP2 Hanako-ga [NP[IP ei Taroo-ni ej age-ta] Hanako-NOM Taro-DAT give-PAST hitoi]-ni at-ta]-noj]-wa [nani]j-ka . . . person-DAT met-NM-TOP what Q ‘. . . what it is that Hanako met the person who gave to Taro . . .’
c
*. . . [CPOpj[IP1 ei Taroo-ni ej age-ta]-noj]-wa [nani]j ka.... Taro-DAT give-PAST- NM-TOP what Q ‘. . . (lit.) what it is that gave to Taro . . .’
In (38)a and (38)b, the subject inside the relative clause can be identified by the head (hito ‘person’). On the contrary, structure (38)c, which only the relative clause appears, is not interpretable because the subject variable in the IP1 remains unbound. Thus, the fact that the sluiced sentence with a relative clause such as (38)a is degraded can be consistently accounted for under the cleft analysis. Coming back to the facts of complex NP and adjunct constraints on sluicing, we actually have a way to distinguish a derivational source with an island from the one without an island. Let us look at the contrast between (39)a and (39)b. Notice that the latter sentence contains an honorific predicate, which is pragmatically appropriate in the context of (39): (39)
a
[Yamada-sensei-ga [Taroo-ga ei at-ta to] Yamada-teacher-NOM Taro-NOM meet-PAST COMP omot-ta-no]-wa Hanako(-ni)i-da think-PAST-NM-TOP Hanako-DAT-COP
b
[Yamada-sensei-ga [Taroo-ga ei at-ta to] Yamada-Prof.-NOM Taro-NOM meet-PAST COMP oomoininat-ta-no]-wa Hanako(-ni)i-desu think(Hon)-PAST-NM-TOP Hanako-DAT-COP(Pol) ‘It was Hanako that Prof. Yamada thought that Taro met’
Ellipsis in Cleft Constructions
181
Both (39)a and (39)b are grammatical; however, the latter is more appropriate in a certain discourse context; when a speaker expresses his/her respect, which is socially determined, toward a listener or a participant in a sentence, honorific/humble forms are used. In (39), due to the subject of the presuppositional clause, Yamada-sensei ‘Prof. Yamada’, its predicate has the honorific form. Now, let us further consider the following sentences involving honorific forms: (40)
a [CP1Yamada-sensei-ga [CP2 Taroo-ga dareka-ni Yamada-Prof.-NOM Taro-NOM someone-DAT at-ta to] oomoininat-ta] soodesu meet-PAST COMP think(Hon)-PAST I-heard ‘I heard that Prof. Yamada thought that Taro met someone’ b
Demo, watasi-wa dare(-ni)-ka zonziage-masen but I-TOP who(-DAT)-Q know(Hum)-NEG ‘But I don’t know who’
The sentence in (40)a is the antecedent sentence for the sluiced clause in (40)b, in which the predicate has a humble form. Assuming that the sluiced clause contains a hidden presuppositional clause, we have two possible presuppositional clauses as derivational sources: CP1 and CP2. However, as the following cleft sentences show, the humble form is appropriate only when Yamada-sensei is involved in the presuppositional clause: (41)
a
[[Yamada-sensei-ga [Taroo-ga ei atta to] Yamada-Prof.-NOM Taro-NOM met COMP oomoininat-ta-no]-wa darei(-ni)-ka] zonziage-masen think(Hon)-PAST-NM-TOP who(-DAT)-Q know(Hum)-NEG ‘(I) don’t know who Prof. Yamada said that Taro met’
b.#
[[Taroo-ga ei atta-no]-wa darei(-ni)-ka] zonziage-masen Taro-NOM met-NM-TOP who(-DAT)-Q know(Hum)-NEG. ‘(I) don’t know who Taro met’
The sentences in (41) indicate that, whereas non-honorific/humble examples can be ambiguous, the sluiced sentence in (40)b should have the underlying structure in (41)a but not the one in (41)b. Thus, this honorific/humble test forces the underlying structure which involves the long-distance sluicing to be present.
182
Cleft Constructions in Japanese Syntax
On the basis of the honorific/humble tests, let us reexamine sluicing sentences containing a complex NP and an adjunct island. As shown below, when the presence of an honorific predicate forces the elided presuppositional clause to contain an island, the sluiced clause is more sharply ruled out when the antecedent of a Wh-remnant is inside an island: (42)
(43)
a
Yamada sensei-ga [[Taroo-ga dareka-ni Prof.Yamada-NOM Taro-NOM someone-DAT syukudai-o tetudattemoratta toyuu] uwasa]-o assignment-ACC was-helped COMP rumor-ACC okikininatta soodesu heard(Hon) I-heard ‘I heard that Prof. Yamada heard that Taro had somebody help his assignment’
b*
Demo, watasi-wa [dare(-ni)(-da)-ka] zonziagemasen but I-TOP who(-DAT)(-COP)-Q don’t-know(Hum) ‘But I don’t know who’
a
Yamada sensei-ga [Taroo-ga dareka-ni syukudai-o Prof.Yamada-NOM Taro-NOM someone-DAT assignment-ACC tetudattemoratta kara] ookorininatt soodesu was-helped because was-angry(Hon) I-heard ‘I heard that Prof. Yamada was angry because Taro had somebody help his assignement’
b*
Demo, watasi-wa [dare(-ni)(-da)-ka] zonziagemasen but I-TOP who(-DAT)(-COP)-Q don’t-know(Hum) ‘But I don’t know who’
The antecedent sentences in (42)a and (43)a contain a complex NP and an adjunct clause, respectively. The remnant Wh-constituents in (42)b and (43)b are associated with the indefinite phrase, dareka-ni ‘somebody-DAT’, which is inside the island. These two sentences are comparable to those in (31)b and (33)b: however, (42)b and (43)b are more clearly ruled out because of the presence of honorific/humble forms. The ungrammaticality of (42)b and (43)b comes from the fact that the derivational sources violate subjacency; under our analysis, the predicates of these sluiced sentences have a humble form, which forces the higher CP containing an honored participant and containing the island to be present in the underlying presuppositional clause.
Ellipsis in Cleft Constructions
183
The sentences in (42)b and (43)b clearly contrast with the following simple sluicing sentence: (44)
a
Yamada sensei-ga dareka-ni oaininatta soodesu Prof. Yamada-NOM someone-DAT met(Hon) I-heard ‘I heard that Prof. Yamada met someone’
b
Demo, watasi-wa [dare(-ni)(-da)-ka] zonjiagemasen but I-TOP who(-DAT)(-COP)-Q don’t-know(Hum) ‘But I don’t know who’
If the observation here is correct, then we can conclude that sluicing in Japanese does obey island constraints once all factors are suitably controlled. Multiple Wh-phrases in focus position Another interesting fact observed in previous research is that the Japanese sluiced clause permits a multiple Wh-remnant, as pointed out by Takahashi (1994) and Kuwabara (1995, 1996). Let us now discuss how to account for the multiple Wh-remnant under the present analysis. Consider the following example: (45)
Dareka-ga nanika-o yon-da rasii ga, someone-NOM something-ACC read-PAST I-heard but [dare-ga nani(-o) -ka] wakaranai who-NOM what-ACC -Q don’t-know ‘Someone read something, but (I) don’t know who what’
Unlike English sluicing, the remnant in a sluiced clause can consist of more than one Wh-phrase in Japanese. This fact initially appears to contradict our analysis, because if the remnant Wh-phrase is the focus portion of the cleft construction, it should not be able to contain more than one focused element, as we discussed in Chapter 2. The following sentence illustrates this point: (46) *
[CP Opi Opj [IP ti tj yon-da]-noi/j]-wa [Taroo-gai hon-oj] -da read-PAST-NM-TOP Taro-NOM book-ACC COP ‘(lit.) It is Taro, a book that read’
Since, as shown in (46), multiple foci are not allowed in Japanese cleft constructions, Taroo-ga hon-o ‘TaroNOM bookACC’ is not qualified to be in focus position. Hence, the present analysis needs to explain why (45) is grammatical and why (46) is not.
184
Cleft Constructions in Japanese Syntax
This fact, however, can be attributed to a special property of Wh-phrases in Japanese, and thus does not pose a difficulty for the cleft analysis. Let us first look at the sentence in (47):
(47)
Watasi-wa [[ti tj yon-da]-no]-wa [dare-gai nani-oj]-ka] I-TOP read-PAST-NM-TOP who-NOM what-ACC-Q wakaranai don’t-know ‘I don’t know who read what’
The sentence in (47) is an overt cleft construction in which the focus position is occupied by more than one Wh-remnant. Unlike (46), examples such as (45) and (47) are grammatical when the focus phrases contain Wh-constituents; ungrammaticality results, as seen in (45), when more than one non-Wh-phrase is focused. These examples can be accounted for under the analysis proposed in previous research that Japanese Wh-phrases can be amalgamated into a single constituent at some level of representation. According to Saito (1992), Takahashi (1994) and Sohn (1994), the level is LF, whereas Tanaka (1998) proposes that the amalgamation takes place either at S-structure or at LF.25 Under our analysis for cleft constructions, the amalgamation takes place between two null operators within the presuppositional clause at some point of derivation prior to licensing the agreement/ feature sharing between the null operator(s) and focus element(s). The multiple Wh-phrases in focus position are base-generated as one constituent, agreeing with the null operator in the presuppositional clause. If this assumption is correct, the grammaticality of (45) and (47) is not surprising. Since the multiple Wh-remnant and the multiple Wh-foci form a single constituent, these multiple Wh-phrases become grammatical. Let us now briefly review the analysis presented by Takahashi (1994) following Saito (1992), and see how we can incorporate his analysis into our present analysis for sluicing. According to Saito (1992) and Takahashi (1994), the lower Wh-phrase adjoins to the higher one and forms a single constituent.26 However, there are some restrictions on this syntactic process. One of the restrictions is that the lower Wh-phrase cannot adjoin to the higher one across a clause boundary (Takahashi 1994, Tanaka 1998). Thus, if the two Wh-expressions are not clausemates, we can predict that the sentence with such multiple Wh-foci will be ungrammatical. This is borne out, as we see in (48):
Ellipsis in Cleft Constructions
(48)
*
185
[CP [Opi Opj] [IP ti [CP Taroo-ga tj yon-da-ka] kii-ta]-no] Taro-NOM read-PAST-Q hear-PAST-NM -wa [darei(-ga) nanij(-o)]-ka . . . -TOP who-NOM what-ACC-Q ‘* . . . it is who, what that heard that Taro read’
In sentence (48), one of the focus phrases, dare-ga ‘who-NOM’, is associated with a position in the matrix clause, and the other, nani-o ‘what-ACC’, is associated with the one in the embedded clause. Such a sentence is thus ruled out due to the clausemate condition on Wh-amalgamation. Takahashi (1994) claims that the clausemate condition on multiple Wh-phrases can be explained by the ECP along with constraints on improper movement. Assuming that Wh-phrases in question must form one constituent, the adjunction operation of nani-o ‘what-ACC’ in the embedded clause to dare-ga ‘who-NOM’ in the matrix clause either violates the ECP, as shown in (49)a, or the ban on improper movement, as shown in (49)b: (49)
a*
[CP [whoi-whatj]k [IP [ti-tj]k . . . [CP [IP . . . tj . . .]]]]
b*
[CP [whoi-whatj]k [IP [ti-tj]k . . . [CP tj [IP . . . tj . . .]]]]
In the structure in (49)a, the lower operator does not move successive cyclically. Hence, the representation does not satisfy the ECP, because the Opj crosses the CP boundary. We could consider another type of derivation, as shown in (49)b; however, assuming that the adjunction counts as A-movement, this movement does not form a uniform chain since it stops at A’-position (spec of CP), creating an instance of improper movement. Therefore, (49)b is not plausible either.27 Under our analysis, the amalgamation of Wh-elements is taken as that of two null operators in a presuppositional portion. Recall also that in a long-distance configuration, an operator is generated in a position adjoined to the highest embedded clause. If this is correct, the derivation for amalgamation of multiple Wh-phrases in (48) is illustrated as follows: (50)*
[CP [Opi-Opj] [IP ti-tj [CP tj [CP proj]]]]
(50) is a schematized structure of the overt presuppositional clause in (48). In this structure, the lower operator is base-generated in adjoined position, being associated with the pro in the embedded clause. If we
186
Cleft Constructions in Japanese Syntax
assume, following Takahashi (1994), that the adjunction of the lower operator to the higher one counts as A-movement, (50) is also ruled out by an instance of improper movement; the original trace of Opj is bound by the higher trace in A-position, which causes a violation of Condition C in binding theory (May 1981). What is important to note here is that the same generalization holds for the corresponding sluicing sentence: (51)
*
Dareka-gai [CP[IP Taroo-ga nanika-oj yon-da] ka] someone-NOM Taro-NOM something-ACC read-PAST Q kii-ta ga, [dare-gai nani(-o)j -ka] wakaranai ask-PAST but who-NOM what-ACC-Q don’t-know ‘*Someone said that Taro read something, but (I) don’t know who what’
The facts here suggest that multiple Wh-elements in clefts and sluicing can be accounted for by the same analysis, which is required to explain multiple Wh-constructions in Japanese on every account.
Some consequences So far, we have seen some parallelism between sluicing and cleft constructions in Japanese. Based on the empirical evidence, we have argued that Japanese sluicing contains a hidden presuppositional clause, leaving a focus portion phonologically realized. If this analysis is correct, we should expect the same kind of syntactic behavior as we already saw in Chapter 4. This prediction is borne out, as we will see in this section; Japanese sluicing is sensitive to weak crossover (WCO) effects and interaction with another A’-dependency although we will see some mixed results for reconstruction effects. Based on the discussion in this section, I will construct an argument that Japanese sluicing should be accounted for by a PF deletion operation rather than by LF reconstruction. Weak crossover effects In Chapter 4, we saw that cleft constructions exhibit weak crossover (WCO) effects in certain cases. The question raised here is whether the sluiced sentences observe the same WCO effects as cleft constructions. One of the difficulties in applying syntactic tests to long-distance sluicing is that, as we saw on pages 175–83, there is more than one possible structure generated underlyingly. It is important to know whether the
Ellipsis in Cleft Constructions
187
sluiced sentences in question actually involve long-distance dependencies, so that we need to control the data to obtain relevant results. Thus, we will henceforth use honorific/humble forms of predicates to produce desirable syntactic configurations. Let us now consider WCO effects in sluicing. The sentence in (52)a is an antecedent sentence for the following elided sentence. (52)a contains an adjunct clause with a null pronoun, which is associated with the object of the verb. The associate of the remnant phrase in (52)b (=Hanako) can potentially cause a WCO effect in the sluiced clause:28, 29 (52)
a
Yamada-sensei-ga [Taroo-ga proi au maeni] Yamada-Professor-NOM Taro-NOM meet before Hanakoi-ni oaininatta soodesu Hanako-DAT met(Hon) I-heard ‘I heard that Professor Yamada met Hanakoi before Taro met (heri)’
b(?)
Demo, watasi-wa hontoowa dare(-ni-dat-ta)-ka but I-TOP actually who(-DAT-COP-PAST)-Q zonziagemasen know(Hum)-NEG ‘But I don’t know who it actually was’
It seems that the sluiced sentence in (52)b does not observe WCO effects when uttered after the sentence in (52)a. Because of the occurrence of the humble form of the predicate in (52)b, the sluiced clause allows only the presuppositional clause which contains the matrix subject as its derivational source. In this way, we can avoid another possible derivational source for (52)b (*Taroo-ga [e] atta no-wa dare-ka . . . ‘who it was that Taro met’). The relevant part of the LF representation is illustrated as follows: (53) . . . [CPOpi[IP ti Yamada-sensei-ga [PP Taroo-ga proi au maeni]ti Yamada-Professor-NOM Taro-NOM meet before zonziage-masen oaininatta]-no]-wa dare(-ni-datta)i-ka met(Hon)-NM-TOP who(-DAT-COP-PAST)-Q know(Hum)-NEG ‘. . . I don’t know actually whoi it was that Professor Yamada met before Taro met (him/heri)’ In (53), the humble form of the predicate requires having Yamada sensei ‘Professor Yamada’ in the presuppositional clause. As argued in our
188
Cleft Constructions in Japanese Syntax
previous discussion, the operator undergoes A-movement to the sentence-initial position by scrambling, and thus, the representation of the presuppositional clause does not manifest the WCO configuration. This is why the sluiced sentence in (52)b is grammatical. Let us turn to long-distance sluicing which contains the same type of adjunct clause. Recall that long-distance cleft constructions have a peculiar structure in which a null operator is base-generated in the position adjoined to the highest embedded clause. If this analysis as well as the analysis for sluicing here is correct, we can predict that sentences such as (54)b are not ruled out. While the judgment for the sentence is subtle, it seems that this is in fact the case:
(54)
a
Yamada-sensei-ga [Taroo-ga [kyoo pro proi au Yamada-Professor-NOM Taro-NOM today meet maeni] Hanakoi-ni atta to] osyatteimasita before Hanako-DAT met COMP said(Hon) ‘Professor Yamada said that Taro met Hanakoi before (he = Taro) met (heri) today’
b
Demo, watasi-wa hontoowa dare(-ni-datta)-ka but I-TOP actually who(-DAT-COP-PAST)-Q zonziagemasen know(Hum)-NEG ‘But I don’t know actually who it was’
The antecedent sentence in (54a) contains the embedded clause with the adjunct clause. Based on the analysis proposed in Chapter 4, the LF representation in (54b) is illustrated as follows:
(55)
[CPOpi[IP Yamada-sensei-ga [CP ti [CP Taroo-ga [PP Yamada-Professor-NOM Taro-NOM maeni] proi atta to]] ossyatteita]-no]-wa kyoo pro proi au today meet before met COMP said(Hon)-NM-TOP hontoowa dare(-nii)-ka zonziage-masen actually who(-DAT)–Q know(Hum)NEG ‘. . . I don’t know actually whoi it was that Professor Yamada said that Taro met before he met (him/heri) today’
Ellipsis in Cleft Constructions
189
Since the content structure for the sluiced sentence in (54)b does not exhibit a WCO configuration, as shown in (55), we can predict the grammaticality of this sentence. If we assume that a WCO effect is sensitive to LF representations, the facts given in this subsection suggest that the empty portion in sluicing is occupied by some content clause at LF. This is, as far as I know, a standard assumption regardless of whether one adopts a PF deletion or LF copying analysis for ellipsis; however, the following subsection will provide significant evidence for the PF deletion approach to Japanese sluicing.30 Interaction with sika phrases Another piece of evidence for the structure of long-distance clefts is interaction with a second A’-dependency, as discussed in the last chapter. Here again, we will start examining whether long-distance sluicing behaves in the same way as long-distance cleft constructions. In order to control the possible underlying structures, honorific/humble forms of predicates are used in the following examples, as we did in the previous sections. By doing so, the deleted presuppositional clauses always contain an embedded CP clause. The example in (56)a is the antecedent sentence of (56)b. (56a) contains a sika phrase with its licensing NEG head. Following the sentence in (56)a, the sluiced sentence in (56)b turns out to be ungrammatical:31 (56)
a
Yamada-sensei-sika [Hanako-ga hoka-no gakusei-ni Yamada-Prof.-only Hanako-NOM another-GEN student-DAT syukudai-o tetudattemoratta-to] omotte-irassyar-anai assignment-ACC was-helped-COMP think(Hon)-NEG yoodesu it-seems ‘It seems that only Prof. Yamada thought that Hanako had another student to help her assignment’
b?*
Demo, watasi-wa dono gakusei(-ni)(-datta)-ka but I-TOP which student(-DAT)(-COP-PAST)-Q zonziagemasen -NEG know(Hum) ‘But I don’t know which student it was’
190
(57)
Cleft Constructions in Japanese Syntax
a?*
. . . [CP [IP Yamada-sensei-sika [CP Opi [CP Hanako-ga proi Yamada-Prof.-only Hanako-NOM syukudai-o tetudatte moratta to]] assignment-ACC was-helped COMP omotte-irassyar-anai]-no]-wa dono gakusei(-ni) think(Hon)-NEG-NM-TOP which student(-DAT) (-datta)i-ka zonziage-masen. (-COP-PAST)-Q know(Hum)-NEG ‘I don’t know which student only Prof. Yamada thought that Hanako had to help her assignment’
b?*
[CP[IP Yamada-sensei-sika [CPOp [CP . . .]] . . NEG] NM]
Because of the existence of the humble form of the predicate, the only possible derivational source is (57)a for the sluiced sentence in (56)b. In this long-distance structure, the null operator is generated in the position adjoined to the embedded CP. (57)b is the schematized structure for (57)a, where we find that the two A’-dependency lines are crossed; hence, the sentence violates the Linear Crossing Constraint (LCC). Another instance of sika phrases interacting with sluicing is shown in (58): (58)
a
Yamada-sensei-ga [Hanako-sika hoka-no gakusei-ni Yamada-Prof.-NOM Hanako-only another-GEN student-DAT syukudai-o tetudattemoraw-nak-atta to] assignment-ACC was-help-NEG-PAST COMP omotte-irassyaru yoodesu think(Hon) it-seems ‘I heard that Prof. Yamada thought that only Hanako had another student to help her assignment’
b?
Demo, watasi-wa dono gakusei(-ni)(-datta)-ka but I-TOP which student(-DAT)(-COP-PAST)-Q zonziagemasen know(Hum)NEG ‘But I don’t know who it was’
In the sentence in (58)a, the sika phrase and its licensing head appear within the embedded clause. The fact here slightly contrasts with that of (57); when the sluiced sentence in (58)b is uttered after the one in (58)a, the sentence sounds better than (57)b. This is because the presuppositional clause deleted in (58)b does not violate
Ellipsis in Cleft Constructions
191
the LCC; the sika phrase is lower, and therefore does not cross the cleft dependency: (59)
a
. . . [CP [IP Yamada-sensei-ga [CP Opi [CP Hanako-sika proi Yamada-Prof.-NOM Hanako-only aw-anak-atta to]] omotte-irassyaru]-no]-wa meet-NEG-PAST COMP think(Hon)-NM-TOP darei(-ni)-ka zonziage-masen what(-DAT)-Q know(Hum)-NEG ‘I don’t know what Prof. Yamada thought that only Hanako read’
b
[[[Op [Hanako-sika . . .]] . . . NEG] NM]
Looking at the schematized structure in (59)b, we find that the A’dependency paths are nested in the derivational source in (59)a. Since the deleted presuppositional clause satisfies the LCC, the grammaticality of the sentence in (58)b is expected under our analysis. It should also be noted that examples such as (59)a are crucial for our analysis proposed in Chapter 4; if we assume that the null operator is generated in its theta position, the sentence would violate the LCC, yielding the ungrammatical sentence.32 Therefore, what we have seen in this subsection reconfirms that Japanese sluicing can be accounted for on a par with cleft constructions. Furthermore, the facts concerning the interaction with sika phrases in sluicing point to an important theoretical conclusion; that is, whether sluicing involves an LF copying operation or a PF deletion. The facts we have seen in this section already suggest that the empty slot is not merely occupied by a pro (Lobeck 1995) throughout the derivation, but it has to be filled by some syntactic content. For the rest of this subsection, we will discuss which approach is most plausible for Japanese sluicing, and it will be shown that the PF deletion approach correctly captures the empirical facts in a way that LF reconstruction does not. Before going into the empirical evidence of interactions with other A’-dependencies for the PF deletion approach, let us look at what we can predict under each approach within the Principles and Parameters framework. First, consider the following simple example of Japanese sluicing:
192
(60)
Cleft Constructions in Japanese Syntax
a
Taroo-ga dareka-ni atta rasii ga, Taro-NOM someone-DAT met I-heard but watasi-wa dare(-ni)(-da)-ka wakaranai I-TOP who-DAT-COP-Q don’t-know ‘I heard that Taro met someone, but I don’t know who’
b
[[Opi [Taroo-ga ti atta] no]-wa dare(-ni)(-da)-ka] Taro-NOM met-NM-TOP who(-DAT)(-COP)-Q wakaranai don’t-know ‘(I) don’t know who it was that Taro met’
The second half of (60)a contains a sluiced clause. The derivational source for this sluiced clause is illustrated in (60)b, in which the presuppositional clause (bold-faced) is phonologically null. If we assume an LF copying approach, the sluiced clause in (60) is represented roughly in the following way, at each level of representations: (61)
a
S-structure [Ø] dare(-ni)(-da) -ka who(-DAT)(-COP)-Q
b
LF [Opi [Taroo-ga ti atta]-no]-wa dare(-ni)(-da)-ka Taro-NOM met-NM-TOP who(-DAT)(-COP)-Q
c
PF dare(-ni)(-da)-ka who(-DAT)(-COP)-Q
As illustrated in (61), the LF copying analysis predicts that only the representation at LF has a contentful presuppositional clause. On the other hand, a PF deletion approach predicts that each level of representation is the following: (62)
a
S-S [Opi [Taroo-ga ti atta]-no]-wa dare(-ni)(-da)-ka Taro-NOM met-NM-TOP who(-DAT)(-COP)-Q
b
LF [Opi [Taroo-ga ti atta]-no]-wa dare(-ni)(-da)-ka Taro-NOM met-NM-TOP who(-DAT)(-COP)-Q
Ellipsis in Cleft Constructions
c
193
PF dare(-ni)(-da)-ka who(-DAT)(-COP)-Q
Notice that the LF and PF representations under each approach are identical, as shown in (61)b, c and (62)b, c. The only difference between these two approaches is whether or not the presuppositional portion has its content at S-structure. As illustrated in (62)a, a contentful presuppositional clause must occur at S-structure under the PF deletion approach. If this hypothesis is on the right track, then we should observe some syntactic facts sensitive to S-structure in Japanese sluicing. Assuming here that there is a level called S-structure in which some syntactic principle is applied, we are able to examine whether Japanese sluicing is accounted for by the PF deletion approach or LF reconstruction. Now, let us return to the interaction with another A’-dependency. Tanaka (1997, 1998) argues that the LCC is a constraint at S-structure. Let us look at the basic sentence given on page 127, Chapter 4, repeated here as (63):33 (63)
a?*
Taroo-sika nani-o yom-anai-no Taro-only what-ACC read-NEG-Q ‘What did only Taro read?’
b?*
[CP[IPTaroo-sika nani-o yoma-nai] no]
(Tanaka 1997: 159, 161) The sentence in (63)a contains the sika phrase in the subject position and the Wh-constituent as the object. As shown in (63)b, this structure violates the LCC because the two A’-dependencies are crossed. This does not necessarily show that the LCC applies at S-structure; since Tanaka (1997) assumes that the spec position is on the right of its head, we can say that the LCC is also violated at LF under his analysis: (64)
?*
[CP[[TP[NegP [ti tj yoma-nai] Taroo-sikai]] no] nani-oj]
However, there is a crucial piece of evidence for the claim that the LCC applies at S-structure. Let us first look at the following sentence:
194
(65)
Cleft Constructions in Japanese Syntax
a?*
Taroo-wa [[Hanako-sika nani-o yom]-anai]-to Taro-wa Hanako-only what-ACC read-NEG-COMP Tomoko-ni itta-no Tomoko-DAT said-Q ‘What did Taro tell Tomoko that only Hanako read?’
b?*
[[. . . [Hanako-sika nani-o . . .]-NEG]-COMP . . .]-Q
The sentence in (65)a contains the sika-nai ‘only-NEG’ dependency within the embedded clause. Since the embedded Wh-phrase, nani-o ‘what-ACC’, has its licensing head in the matrix clause, the sentence is ruled out by the LCC, as illustrated in (65)b. However, when the sentence in (65) has an additional Wh-phrase, it becomes grammatical: (66)
a
Taroo-wa [[Hanako-sika nani-o yom]-anai]-to Taro-Top Hanako-only what-ACC read-NEG-COMP dare-ni itta-no Who-DAT said-Q ‘Who did Taro tell that only Hanako read what?’
b
[[. . . [Hanako-sika nani-o . . .]-NEG]-COMP dare-ni . . .]-Q
The difference between (65) and (66) is that the dative object in the matrix clause is not a Wh-phrase in the former (Tomoko-ni ‘Tomoko-DAT’) and is a Wh-phrase in the latter (dare-ni ‘who-DAT’). This contrast is presumably of the same nature of the following English examples (Watanabe 1992): (67)
a?? b
What did John ask Tom whether Mary bought? Who did John ask whether Mary bought what?
Based on these contrasts, Tanaka (1997) assumes that there is one and only one operator [ + Wh] COMP at S-structure, and only dare-ni ‘whoDAT’ but not nani-o ‘what-ACC’ is directly associated with the COMP.34 If the LCC applies at LF, then the LF representation for (66)a would be as follows:
Ellipsis in Cleft Constructions
(68)
195
[[[. . . [[ti . . . tk . . .]-NEG-Hanako-sikai]-COMP tj told]]-Q-darej nanik]
The standard analysis for Wh-in situ is that Wh-constituents move to the spec of CP at LF (Huang 1982), so that both dare-ni ‘who-DAT’ and nani-o ‘what-ACC’ move to the spec of CP. The association paths in the (66) sentence would be crossed at LF, as shown in (68). This representation makes a wrong prediction; the sentence should be ruled out by the LCC. However, since the sentence in (66)a is grammatical, as represented in (66)b, we conclude that the LCC is a constraint at S-structure. Now let us return to the examples in (56), which are repeated here as (69): (69)
a
Yamada-sensei-sika [Hanako-ga hoka-no Yamada-Professor-only Hanako-NOM another-GEN gakusei-ni syukudai-o tetudattemoratta to] student-DAT assignment-ACC was-helped COMP omotte-irassyar-anai yoodesu think(Hon)-NEG it-seems ‘It seems that only Professor Yamada thought that Hanako had another student to help her assignment’
b?*
Demo, watasi-wa dono gakusei(-ni)(-datta)-ka but I-TOP which student(-DAT)(-COP-PAST)-Q zonziage-masen know(Hum)-NEG ‘But I don’t know which student it was’
We have discussed that the sentence in (69)b is ruled out by the LCC under our analysis. We have argued that the sluiced sentence in (69)b contains a presuppositional clause underlyingly, but it is deleted at PF. The derivational source for (69)b is reproduced in (70)a: (70) a?* . . . [CP [IP Yamada-sensei-sika [CP Opi [CP Hanako-ga proi Yamada-Professor-only Hanako-NOM syukudai-o tetudattemoratta to]] assignement-ACC was-helped COMP omotte-irassyar-anai]-no]-wa dono gakusei(-ni)(-datta)i-ka think(Hon)-NEG-NM-TOP which student(-DAT)(-COP-PAST)-Q zonziage-masen know(Hum)-NEG ‘I don’t know which student only Professor Yamada thought that Hanako had to help her assignment’
196
Cleft Constructions in Japanese Syntax
b?*
[CP[IP Yamada-sensei-sika [CPOp [CP . . .]] . . NEG] NM]
Under the PF deletion approach, the sentence in (69)b has the structure (70)a at S-structure. The LF copying approach, however, cannot assume that (70)a is present at that level; as we have seen in (61), there is no contentful structure in the presuppositional clause at S-structure. Therefore, our analysis correctly explains the ungrammaticality of (69)b whereas the LF copying approach cannot. Our analysis also accounts for grammatical sluicing sentences which contain a sika-phrase. The examples in (58) and (59), repeated here as (71) and (72), are the grammatical instances: (71)
(72)
a
Yamada-sensei-ga [Hanako-sika hoka-no Yamada-Professor-NOM Hanako-only another-GEN gakusei-ni syukudai-o tetudattemoraw-anak-atta student(-DAT) assignment-ACC was-help-NEG-PAST to] omotte-irassyaru yoodesu COMP think(Hon) it-seems ‘I heard that Professor Yamada thought that only Hanako had another student to help her assignment’
b?
Demo, watasi-wa dono gakusei(-ni)(-datta)-ka but I-TOP which student(-DAT)(-COP-PAST)-Q zonziage-masen know(Hum)-NEG ‘But I don’t know who it was’
a
. . . [CP [IP Yamada-sensei-ga [CPOpi [CP Hanako-sika proi Yamada-Professor-NOM Hanako-only yom-anakatta to]] omotte-irassyaru]-no]-wa read-NEG COMP think(Hon)-NM-TOP nanii-ka zonziage-masen what-Q know(Hum)-NEG ‘I don’t know what Professor Yamada thought that only Hanako read’
b
[[ [Op [Hanako-sika . . . .]] . . NEG] NM]
Ellipsis in Cleft Constructions
197
The sluiced sentence in (71)b is grammatical, as the underlying structure shows in (72). The LF copying approach can predict this grammaticality, because at the level in which the LCC applies, the contentful structure of the presuppositional clause does not exist, and hence, there is no LCC violation. However, the contrast between (70)b and (71)b cannot be accounted for under this approach. If Tanaka’s (1997, 1998) claim is correct, the fact that sluicing is sensitive to the LCC suggests that the hidden presuppositional clause is not empty at S-structure but must have a contentful syntactic structure in it. Thus, the discussion here supports a PF deletion operation to generate Japanese sluicing. Under the Minimalist Program (Chomsky 1995), S-structure is discarded as a significant level of syntax. If this view is pursued, the facts concerning the LCC should be reexamined. There might be a way to assume that the LCC is not an S-structure condition, however. In other words, this is not a condition on representations but a condition on derivation before spell-out. While it is not necessary for the purpose of this book to enter into a detailed discussion of this issue, we can point out that such reinterpretation of the LCC can conform to the spirit of Minimalism, and still the idea of the PF deletion analysis is maintained for Japanese sluicing. Reconstruction in sluicing Assuming that Japanese sluicing is generated by a PF deletion operation, it is expected that a representation at S-structure can be identical to the LF representation in the construction. The examples we saw in previous subsections in fact point this out; given that WCO effects are sensitive to LF representations, and the LCC should be satisfied at S-structure, there must be a contentful structure in the elliptical cleft constructions at both these levels. The reconstruction facts that we will see in this subsection basically accord with this line of analysis. In Chapter 4, we looked at reconstruction effects in cleft constructions. The relevant examples are as follows: (73)
a
[Yamada senseii-ga ti izon-nasatta-no]-wa go-zibuni-no Professor Yamada-NOM rely-on(-Hon)-NM-TOP (Hon)self-GEN bunseki(-ni) dat-ta analysis(-DAT) COP-PAST ‘It was self’s analysis that Professor Yamada relied on’
198
Cleft Constructions in Japanese Syntax
b
[Yamada senseii-ga [Hanakoj-ga proi izon-sita to] Professor Yamada-NOM Hanako-NOM relied-on COMP ossyatta-no]-wa go-zibuni/*j-no bunseki(-ni) desu said(Hon)-NM-TOP (Hon)self-GEN analysis(-DAT) COP(Pol) ‘It was self’s analysis that Professor Yamada said that Hanako relied on’
(73)a is an instance of a local cleft, and (73)b is a long-distance cleft. Note that both anaphoric pronouns in focus position are coindexed with the matrix subject. As for (73)b, if the anaphor is not marked by a honorific (polite) prefix, the sentence becomes stylistically unacceptable, because, as we discussed in Chapter 4, the zibun ‘self’ in focus position of long-distance clefts can only be coindexed with the highest subject of the presuppositional clause, Yamada sensei ‘Professor Yamada’, which triggers the honorific/polite forms. In addition, it should be noted that when Yamada sensei ‘Professor Yamada’ and Hanako are switched with stylistically appropriate forms of the predicates, go-zibun ‘Hon-self’ turns out to be inappropriate. These two cases are shown in (74): (74)
a#
[Yamada senseii-ga [Hanako-ga pro izon-sita to] Professor Yamada-NOM Hanako-NOM relied-on COMP desu ossyatta-no]-wa zibuni-no bunseki(-ni) said(Hon)-NM-TOP self-GEN analysis(-DAT) COP(Pol) ‘It was self’s analysis that Prof. Yamada said that Hanako relied on’
b#
[Yamada senseii-ga [Hanako-ga pro Professor Yamada-NOM Hanako-NOM izon-nasatta to] itta-no]-wa go-zibuni-no relied-on(Hon) COMP said-NM-TOP (Hon)self-GEN bunseki(-ni) desu analysis(-DAT) COP(Pol) ‘It was self’s analysis that Hanako said that Prof. Yamada relied on’
In (74), if the anaphor in focus position can be coindexed with the embedded subject, these sentences should be perfect stylistically; however, the examples show that this is not the case.
Ellipsis in Cleft Constructions
199
If our proposal is on the right track and the binding conditions apply at LF, we expect exactly the same reconstruction effects as in (73) in the corresponding sluiced sentences. This is borne out, as shown in (75): (75)
a
Yamada senseii-ga go-zibuni-no bunseki-ni Professor Yamada-NOM (Hon)self-GEN analysis-DAT izon-nasatta soodesu ga, watasi-wa go-zibuni-no dono relied-on(Hon) I-heard but I-TOP (Hon)self-GEN which bunseki(-ni)-datta-ka yoku zonziagemasen analysis(-DAT)-COP-PAST-Q well don’t-know ‘I heard that Prof. Yamada relied on self’s analysis, but I don’t know very well which of self’s analysis it was’
b
Yamada senseii-ga [Hanakoj-ga zibun#i/j-no Professor Yamada-NOM Hanako-NOM self-GEN bunseki-ni izonsita to] ossyatta soodesu ga, watasi-wa analysis-DAT criticized COMP said(Hon) I-heard but I-TOP zibun#i/*jno dono bunseki(-ni)-datta-ka yoku self-GEN which analysis(-DAT)-COP-PAST-Q well zonziagemasen don’t-know(Hon) ‘I heard that Professor Yamada said that Hanako relied on self’s analysis, but I don’t know very well which analysis it was’
The sentence in (75)a shows that go-zibun ‘Hon-self’ in the remnant position can be construed as Yamada sensei ‘Professor Yamada’. Although this is straightforward, we need some explanation for the unacceptability in (75)b. In the long-distance sluicing in (75)b, zibun in the antecedent clause can be appropriate only if it is coindexed with Hanako and not Professor Yamada due to the lack of honorific marker. This already provides the context such that the bunseki ‘analysis’ is Hanako’s and not Professor Yamada’s. However, because of the presence of the honorific predicate, zonziagemasen ‘don’t-know(Hon)’, the hidden presuppositional clause has only one possible structure; that is, ‘what Professor Yamada said that Hanako criticized’ and not merely ‘what Hanako criticized’. Since under our analysis, syntax forces zibun in the focus position to be coindexed with the highest subject, Professor Yamada, this contradicts the presupposition set up in the antecedent sentence. Therefore, the sluiced sentence in (75)b is unacceptable.
200
Cleft Constructions in Japanese Syntax
On the contrary, the following long-distance sluiced sentences are acceptable: (76)
a
Yamada senseii-ga [Hanakoj-ga go-zibuni/#j-no Professor Yamada-NOM Hanako-NOM self-GEN bunseki-ni izonsita to] ossyatta soodesu ga, watasi-wa analysis-DAT relied-on COMP said(Hon) I-heard but I-TOP go-zibuni/*jno dono bunseki(-ni)-datta-ka yoku self-GEN which Analysis(-DAT)-COP-PAST-Q well zonziagemasen don’t-know(Hon) ‘I heard that Professor Yamada said that Hanako relied on his own analysis, but I don’t know very well which analysis it was’
b
Hanakoi-ga [Yamada senseij-ga zibuni/#j-no bunseki-ni Hanako-NOM Professor Yamada-NOM self-GEN analysis-DAT izonsita to] itta soodesu ga, watasi-wa zibun?i/*jno relied-on COMP said I-heard but I-TOP self-GEN dono bun-seki(-ni)-datta-ka yoku sirimasen which analysis (-DAT)-COP-PAST-Q well don’t-know ‘I heard that Hanako said that Professor Yamada relied on her own analysis, but I don’t know very well which analysis it was’
In contrast with (75)b, (76)a shows that the anaphor in the remnant phrase with an honorific prefix can refer to Professor Yamada. This is expected under our analysis; since the hidden presuppositional clause includes the highest subject, the go-zibun can be coindexed with it. (76)b can also be explained in the same way; the zibun is coindexed with the highest subject Hanako.35 Not only zibun binding but also otagai ‘each other’ makes the same point. Consider the following sentences:36 (77)
a
Taroo to Hanakoi-ga otagaii-no bunseki-ni Taro and Hanako-NOM each-other-GEN analysis-DAT izon-sita rasii ga, watasi-wa otagaii-no dono relied-on I-heard but I-TOP each-other-GEN which bunseki(-ni)-datta-ka yoku siranai analysis (-DAT)-COP-PAST well don’t-know ‘I heard that Taro and Hanako relied on each other’s analysis, but I don’t know which of each other’s analysis they were’
Ellipsis in Cleft Constructions
b
201
Yamada sensei to Tanaka senseii-ga [gakusei futarij-ga Professor Y and Professor T-NOM students two-NOM otagai*i/j-no bunseki-ni izonsita to] ossyatta each-other-GEN analysis-DAT relied-on COMP said(Hon) dono bunseki(-ni) soodesu ga, watasi–wa otagai#i/*j-no I-heard but I-TOP each-other-GEN which analysis (-DAT) datta-ka zonziagemasen COP-PAST-Q don’t-know (Hon) ‘I heard Professor Yamada and Professor Tanaka said that two of the students relied on each other’s analysis, but I don’t know which of each other’s analysis they were’
The sentence in (77)a contains a case of local sluicing. As our analysis predicts, the reciprocal anaphor in the remnant phrase can be coindexed with Taro and Hanako in the antecedent clause. On the other hand, (77)b is unacceptable. Recall that otagai ‘each other’ is a local anaphor, so that the otagai in the antecedent sentence can be coindexed only with gakusei futari ‘two students’. This pragmatic context produces some mismatch with the syntactic structure; the sluiced clause contains a long-distance presuppositional clause due to the honorific form of the predicate, and the reciprocal anaphor in the sluiced clause is allowed to be coindexed only with the highest subject, Professor Yamada and Professor Tanaka. In other words, a mechanism of deletion under identity forces each clause to have a different indexing, but the two must be parallel in interpretation. Therefore, this coindexed reading does not accord with the presupposed context, and hence, the sentence is unacceptable. There is a piece of evidence that is not in favor of our analysis. Fukaya (1998: 14) argues that soko ‘it, that place’ binding does not show reconstruction effects in sluicing:
(78)
a
[[Toyota-saei-ga [sokoi-no aru syuu-no Toyota-even-NOM it-GEN a-certain state-GEN koozyoo]-o heisasita] rasii ga,*[sokoi-no dono syuu-no factory-ACC closed seem but it-GEN which state-GEN koozyoo](-o) ka sitteimasu ka factory(-ACC) Q know Q ‘(lit.) I hear that even Toyota closed its factory in a certain state, but do you know its factory in which state?’
202
Cleft Constructions in Japanese Syntax
b
[sokoi-no [[kanarino kazu no zidoosya gaisha]i-ga a large number-GEN automobile companies-NOM it-GEN aru syuu-no koozyoo]-o heisasita] rasii ga, *[sokoi-no a-certain state-GEN factory-ACC closed seem but it-GEN dono syuu-no koozyoo](-o) ka sitteimasu ka which state-GEN factory(-ACC) Q know ‘(lit.) I hear that a large number of automobile companies closed its factory in a certain state, but do you know its factory in which state?
The sentence in (78)a is, in fact, reminiscent of the cleft constructions on pages 34–8, Chapter 2; the subject is marked by a focus particle and the focus position is occupied by a bound pronoun: (79)
a*
[Tarooi-sika e aw-anak-atta no]-wa zibuni-no Taro-only meet-NEG-PAST NM-TOP self’s hahaoya-ni da mother-DAT COP ‘It was his own mother that only Taro met’
b*
[Tarooi-dake e atta no]-wa zibuni-no hahaoya-ni da Taro-only met NM-TOP self’s mother-DAT COP ‘It was his own mother that only Taro met’
The subject in the presuppositional clause in (79)a is a negative polarity item, and the one in (79)b is another type of focus particle. The zibun in the focus position is coindexed with these subjects, the sentences are ruled out. The same is true for sae ‘even’: (80)
*
[Tarooi-sae e atta no]-wa zibuni-no hahaoya-ni da Taro-even met NM-TOP self’s mother-DAT COP ‘It was his own mother that even Taro met’
If we assume that Japanese sluicing involves a cleft construction, the ungrammaticality in (78)a is not surprising. Although I do not have any principled account for these facts at this point, it can be accounted for by the same analysis for (79) and (80). As for (78)b, our analysis for sluicing cannot explain the ungrammaticality, as the corresponding cleft sentence is grammatical. See also the discussion on pages 110–14, Chapter 4. It has been controversial whether binding conditions apply at LF or at S-structure, the PF deletion analysis can conform to either view.37 The
Ellipsis in Cleft Constructions
203
observation in this section shows that Japanese sluicing involves a contentful structure in the presuppositional clause at both LF and S-structure. The facts concerning WCO effects force us to argue that a sluiced sentence has a structure with content at LF, and the facts of interaction with another A’-dependency show that it also has a full structure in an elided portion at S-structure. This crucially suggests that a PF deletion operation is responsible for the phenomenon of Japanese sluicing.
Summary This chapter presented the following three points. First, given the striking similarities between Japanese sluicing and cleft constructions, it is argued that the remnant Wh-phrase is analogous to the focus element of a cleft sentence in Wh-in situ languages. The implication of the proposed analysis is that Japanese Wh-phrases are not situated in the same position as English Wh-phrases in sluicing; more specifically, it may be the case that Japanese does not have overt Wh-movement triggered by ellipsis. Second, based on our approach for sluicing, the analysis proposed in Chapter 4 is further confirmed by examining sluicing data in Japanese. The empirical facts, such as WCO effects, interaction with another A’-dependency, and reconstruction effects exhibit some more parallelisms between sluicing and cleft constructions. Third, we concluded from the syntactic behaviors with respect to the interaction with another A’-dependency, that the Japanese sluicing is explained by a PF deletion approach rather than an LF copying approach.
This page intentionally left blank
6 Conclusions
This book has explored the syntactic nature of cleft constructions in Japanese. The discussions in the preceding chapters have led to the conclusion that the cleft construction is not a distinct theoretical entity, but rather, that the construction results from a combination of other basic structures which are required in the grammar of the Japanese language on all accounts. Let us now reconsider the questions listed at the beginning of this book, and review the solutions which have been given: (1) a Do Japanese cleft constructions involve operator movement like their English counterparts? b What is the syntactic structure of Japanese cleft constructions? c Are Japanese cleft constructions like English clefts in having connectivity effects? If so, how should we account for them? Question (1)a was answered in Chapters 2 and 4, where it was proposed that both local and long-distance cleft sentences involve a null operator in the presuppositional clause which agrees with a focus phrase. This analysis was also supported by the fact that the focus element is not directly generated by scrambling. While operator movement is involved in Japanese as well as English clefts, these two languages differ crucially in terms of how movement takes place. In English, it is generally assumed that the null operator moves successive-cyclically; however, this is not what we have found to hold in Japanese. Instead, the null operator of Japanese long-distance clefts is base-generated adjoined to the highest embedded clause in the presuppositional clause, and undergoes short movement to the spec of CP. Short A’-movement is also 205
206 Cleft Constructions in Japanese Syntax
observed in local clefts; the operator at first undergoes A-movement to sentence-initial position, and then further A’-movement takes place. In reply to (1)b, I claimed in Chapter 2 that the focus element is underlyingly a single constituent in predicate position. Concerning the presuppositional clause, we saw in Chapter 3 that the head of the no-clause is a complementizer whose phrasal category is CP rather than NP or DP. The same no can be found in HIRCs, which also demonstrate similar syntactic behavior such as case conversion and extraction from the clause. Furthermore, this book has presented interesting data which have not to my knowledge been noticed in previous studies. These new syntactic facts involving reconstruction, weak crossover, and interaction with another type of A’-dependency, have shown that A’-movement of the null operator in Japanese long-distance configurations is sensitive only to the highest clause. This was examined closely in Chapter 4. In addition to answering the major questions in (1), we examined Japanese sluicing constructions in Chapter 5. It was argued there that Japanese sluiced sentences are derived from a cleft construction in which the presuppositional clause is phonologically empty, and the focus element is overt. Sluicing was seen to share several syntactic properties which were shown in other chapters to argue in cleft constructions. The analysis which I proposed has an important theoretical implication, providing the empirical evidence for a PF deletion approach rather than an LF copying approach for ellipsis. The data support the view that the empty clause in Japanese sluicing has syntactic content which corresponds to the overt clause before spell-out, but which is deleted at PF. A number of tasks which are outside the scope of this book remain for future research. One such question is what causes Japanese to exhibit a sensitivity to the highest clause, whereas other languages, including English, do not. While it remains to be seen what deeper principles might explain such a typological difference, this book has provided a first step in bringing this difference to light, within the Principles and Parameters approach and the Minimalist Program. It is hoped that the conclusions of this book will be able to serve as a starting point for further work outside of its domain.
Notes
1
Introduction
1. Although these two types of cleft sentences are similar in their meaning when uttered out of context, it is pointed out by Prince (1978) and Declerck (1984) that the two construction types differ in their usages at the level of discourse. It is outside the scope of this book to explore this area; for details, see Prince (1978), Declerck (1984) and references cited therein. 2. It-cleft and pseudocleft constructions may be distinct in the kind of element that can occupy the focus position. From the viewpoint of syntactic category, it appears that it-cleft constructions are more restricted than pseudoclefts; for instance, a predicate cannot be clefted in it-cleft sentences (Heggie 1988, 1993). Although pseudoclefts do not have such a restriction, they are restricted in the sense that the focus element should be a constituent that can serve as the answer to a ‘what’-interrogative sentence (McCawley 1998: 65). 3. Alternative analyses proposed are found in Heycock and Kroch (1996) and Boškovic (1997) for pseudoclefts, and in Percus (1997) and Kiss (1998) for it-clefts. 4. Strictly speaking, the connectivity phenomenon is observed only in specificational (pseudo)cleft constructions, and not in other types of cleft constructions, including predicational cleft constructions. The difference between the two types of pseudoclefts is noticed by Akmajian (1970): (i)
a What John bought was junk predicational b What John bought was a house specificational
The post-copular element, when it is interpreted as an adjective, as in (i)a describes a property or state of the pre-copular clause, whereas in (i)b the post-copular element specifies semantic information that the pre-copular clause is lacking. Thus, it is said that (i)a has a ‘predicational’ reading while (i)b has a ‘specificational’ reading. This distinction does not arise solely from the lexical category of the element which appears in the post-copular position. Consider the following example: (ii)
What John is is silly
Like (i)a in which only the predicational reading is possible, the sentence in (ii) also contains an AP in the post-copular position. As has been discussed in the literature, however, sentence (ii) is in fact ambiguous between the predicational reading and the specificational reading. The former reading is that the relevant property of John is silly; for example, John is a grad student and being a grad student is silly. The latter reading is equivalent to ‘John is 207
208 Notes silly’. Therefore, the type of category which appears in post-copular position does not determine whether the construction is predicational or specificational in general. Williams (1983), Heggie (1988), and Heycock (1991) argue that the basic structure of predicational pseudoclefts differs from that of specificational pseudoclefts. In the specificational pseudocleft, the underlying focus phrase is a subject, and the underlying Wh-clause is a predicate of a small clause. Therefore, the specificational pseudocleft is considered an inverted predicate sentence, with the predicative Wh-clause becoming a surface subject. In contrast, this is not assumed for the underlying structure of the predicational pseudocleft here; the Wh-clause is regarded as the subject throughout the derivation. I will deal here only with specificational cleft constructions, the predicational constructions being outside the present scope. 5. Percus (1997) shows a parallelism in NPI licensing between it-cleft constructions and copular sentences whose subjects contain a definite description. He assumes, roughly speaking, that the it–clefts in the following sentences (i)a and (ii)a are represented as the corresponding sentences in (i)b and (ii)b: (i)
a b
It isn’t [anyone I KNOW] that John saw The one that John saw isn’t anyone I know
(ii)
a* It is [anyone I KNOW] that John didn’t see b* The one that John didn’t see is anyone I know
Based on the examples above and others, he proposes that at the point at which negative polarity licensing takes place, the clefts have structures identical to the sentences in (i)b and (ii)b. 6. It has been generally agreed in the literature that Japanese ‘cleft’ constructions correspond to English it-clefts (cf. Matsuda 1997), because in Japanese, it is said that only NPs and PPs can be clefted whereas not only NPs and PPs, but also APs and VPs can be the focus elements in English pseudocleft constructions. Since English it-clefts do not usually take AP/VP focus elements, it seems that Japanese ‘clefts’ are more like it-clefts rather than pseudoclefts in this respect. However, Heggie (1993) points out that English it-clefts can take APs as a focus element in a certain type of clefting: (i)
It is [AP quite happy] that Bill is
Following Horn (1985), she proposes that this kind of cleft construction is ‘metalinguistic’ in the sense that it is not clear whether the sentence obeys syntactic restrictions. I will not discuss this issue in this book; however, it should be noticed that the types of lexical categories in the focus position may not determine whether Japanese ‘clefts’ are similar to English it-clefts, and it is still worthwhile to compare Japanese ‘clefts’ with English pseudoclefts. 7. I put NM (=nominalizer) on no in (6); however, this is more likely to be a pronominal form of no. For various usages of no in Japanese, see Chapter 3.
Notes 209 8. It is pointed out by an anonymous reviewer that a true case of cleft construction requires a case marker in focus position; otherwise, it is a construction with a relative clause complex NP in subject position (also see Hiraiwa and Ishihara 2002). I will come back to this issue in Chapter 2, and will use in this book a focus NP with a dative case particle as a representative of the cleft constructions in Japanese. 9. Note, however, that Takahashi (1993, 1994) claims that some instances of scrambling and sluicing can be counted as ‘overt Wh-movement’ or within the framework of Minimalist Program, ‘phrasal movement’. 10. See also Murasugi (1991), Hasegawa (1997) and Matsuda (1997) for analyses similar to that of Hoji (1990). Alternatives are Koizumi (1995, 2000) and Kuwabara (1996) where string-vacuous overt verb raising is involved in clefts. Hiraiwa and Ishihara (2002) argue that cleft constructions are derived by focus movement and topicalization based on Rizzi’s (1997) articulated CP structures. Here I will take Hoji’s analysis as representative of the previous research. 11. No is glossed as NM, a nominalizer, which does not indicate any specific analysis. The status of no will be argued in Chapter 3. 12. An anonymous reviewer pointed out that it is not the case that multiplefocus cleft sentences are always banned in English. The examples provided by the reviewer are: (i)
a It was in Paris in 1945 that John met Mary for the first time b It was in 1945 in Paris that John met Mary for the first time c (?)What I bought was ‘Hamlet’ at bookstore X and ‘Syntactic Structure’ at bookstore Y
In addition, it is not the case either that Japanese allows any kind of multiple foci in cleft constructions; there are some restrictions on the order and case markers of each phrase in focus position. I will discuss this issue in Chapter 4.
2
Topicalization and Cleft Constructions
1. What I call a topic phrase is the NP or PP marked by wa ‘TOP’. 2. Kuroda (1987) claims that all sentence-initial topics are derived by movement. On the contrary, Mihara (1992) suggests that the sentence-initial topics are base-generated either in the position adjoined to IP or in the spec of CP. As we will see later, we basically follow Saito (1985) in that Japanese topic constructions are derived by movement or base-generated; however, our analysis is slightly distinct from Saito’s. It will crucially be argued that only a special type of topic is base-generated and most other topic phrases are derived by movement. 3. Note that when a phrase marked by a major case marker (ga ‘NOM’ and o ‘ACC’) is topicalized, the case marker disappears. While postpositions usually remain when they are marked by wa, the occurrence of ni ‘DAT’, which in some constructions is interpreted as a postposition, is optional. 4. However, Hasegawa (1984/85), Imai (1987) and Sakai (1994) argue against this generalization. I review the main point made by Sakai, pp. 20–2 and
210 Notes
5. 6.
7.
8.
argue that topic constructions such as (2) may be derived by movement without violating subjacency. See, however, Kuroda’s (1987) insight on this fact, which will be discussed, pp. 32–3. NP topics can be interpreted contrastively if they occur in a relative clause and/or if they have phonological stress. Hoji (1985) notes that thematic topics can appear in a complement clause of bridge verbs and koto-clauses (‘the fact that. . .’) although they sound marginal. Hoji (1985) also discusses the deletability of wa and the syntactic positions where each wa is able to occur. For further details of these two arguments, see Hoji (1985). The basic notion of the theory of binding is defined as follows (Chomsky 1981): α is bound by β iff α and β are coindexed, and β c-commands α. a An anaphor is bound in its governing category. b A pronominal is free in its governing category. c An R-expression is free. (iii) β is a governing category for α iff β is the minimal category containing α, a governor of α, and a SUBJECT accessible to α. (iv) a α is accessible to β iff β is in the c-command domain of α and assignment to β of the index of α would not violate (b). b *[γ. . .σ. . .], where γ and σ bear the same index. (i) (ii)
9. The definition of c-command is as follows: (i)
X c-commands Y if neither X nor Y dominates the other and the first branching node that dominates X also dominates Y. (Reinhart 1976)
10. A number of analyses have been proposed for the reconstruction effects. For example, Chomsky (1977) proposes that preposed items literally move back to the original position, and Barss (1986) claims that reconstruction is an effect of chain binding. It is also argued that the copy theory (Chomsky 1993) accounts for the reconstruction effects. The choice between these alternatives is largely irrelevant here. 11. See also Mihara (1992) and Tateishi (1994) who argue the same point. 12. Needless to say, if the speaker put a heavy stress on the topic phrase, it gives us a contrastive interpretation. I put aside the issue of phonological effects here. 13. For the purpose of the present chapter, the landing site for a sentence-initial topic phrase can be either a position adjoined to a sentence (Saito 1985, Hoji 1985), spec of CP (Kubo 1992), spec of M(odal)P (Ueyama 1989), or spec of TopP (Lasnik and Saito 1992). For cleft constructions, I will argue that a null operator moves to the spec of CP in the presuppositional clause. This will be discussed in Chapter 3. 14. Kuroda (1987) also proposes the same line of analysis although he assumes that a sentence-initial topic is in a position adjoined to a sentence (=IP). He does not present an exact analysis for sentence-internal topic phrases, which I will also put aside in this book. For other analyses for topicalization in Japanese, see Mihara (1992), Tateishi (1994) and references cited therein.
Notes 211 15. A multiple nominative construction is translated as a cleft sentence or a sentence with ‘For NP (=major subject)’, depending on the sentence. No particular analysis is implied in the translation. 16. The sentence in (18)a can be paraphrased as follows: (i)
Nihon-ga tabemono-ga sakana-ga umai Japan-NOM food-NOM fish-NOM be-delicious ‘It is in Japan where fish among the food is delicious’
It should be noted that the first nominative phrase, Nihon-ga ‘Japan-NOM’, is not a major subject since it is considered to be derived from the genitive phrase like (18)a; Nihon-ga can be an argument of tabemono ‘food’ in (i). 17. I am not claiming that all topic constructions must involve movement. The examples in (3) are true instances of base-generated topicalization (but see Kuroda 1987). In addition, Japanese has another type of wa: that is, a so-called conditional topic: (i)
18.
19.
20. 21.
yomi-tai] hitoi]-wa, koko-ni arimasu [[ ei sinbun-o newspaper-ACC read-want person-TOP here-DAT be-there ‘If you want to read a newspaper, here is one’ (Tateishi 1994: 131)
Except for these special types of wa-phrases, I assume that topic constructions are derived by movement, and a major subject position is licensed by the aboutness condition. A similar discussion is found in Italian Clitic Left Dislocation (Cinque 1990). It will be shown, however, that some types of PPs have corresponding pros in Japanese. These are called quasi adjunct PPs which include time/ place PPs. True adjunct PPs, such as reason/manner PPs, on the contrary, do not have corresponding pros. These are the issues in Chapter 4. While Hoji (1990) finds the sentence in (30)a grammatical, the judgment on this sentence is not evident. As far as I can tell from my own judgments and those from my informants, it is difficult to find such a striking distinction between the two sentences. In fact, Saito (1985) puts two question marks on the NP topic construction in (32)a (and *? for the PP topic construction in (32)b) although he draws a crucial line between NP topics and PP topics; the former are grammatical and the latter are not. This book assumes judgments in Hoji and Saito, however. The pro in (31)a cannot be inside a PP. Therefore, the structure only takes care of clefting of an NP. As pointed out by an anonymous reviewer, we can consider that another possible derivational source would be the following: (i)
Russell-waj John-ga [[ ei proj atta-koto-ga-aru] nihonzini]-o Russell-TOP John-NOM have-met Japanese-ACC oozei sitteiru many knows ‘Speaking of Russell, John knows many Japanese that have met’
212 Notes
22. 23.
24.
25.
Whether or not the sentence above is derived from a major subject construction (= (32)a), this clearly shows the parallelism between topicalization and clefting. Chapter 4, however, will present structures for long-distance cleft constructions, which are different from (34). I am not claiming that a structure like (31)a does not exist at all in the grammar of Japanese; the same structure can be found in a copular sentence with a free relative clause in subject position. In the following examples of cleft constructions in this chapter, (un)grammaticality does not change with or without any case marker although some native speakers of Japanese including myself find clefts with case-marked focus NP (NP-ga and NP-o) degraded. As pointed out by Hidekazu Tanaka (personal communication), if tuma ‘wife’ in (35)a is topicalized in-situ, the sentence is grammatical: (i)
Taroo-ni tuma-wa aru Taro-DAT wife-TOP have ‘A wife, Taro has’
The contrast between (i) and (35)b indicates that the derivational source for (35)c and (35)d is not like (i) but more like (35)b. In other words, in order to derive the examples in (35)c and (35)d, it is crucial that the phrase in question is preposed (if the ungrammaticality comes from syntax). Thanks also to an anonymous reviewer for pointing out that the same argument can be made for (33)a. In Chapter 4, I will argue that local clefting and local relativization involve local scrambling before A’-movement to the spec of CP takes place. The fact here supports the proposed analysis. 26. Thanks to Mark Baker for making me aware of this issue and to an anonymous reviewer for correcting my original argumentation. 27. Thanks to Hidekazu Tanaka for pointing this out to me. This kind of sentence is discussed in Kitagawa (1986) among others. 28. Remember that the topic phrase in (36)b, which is repeated here as (i)a, is not a true topic, since we can assume that (i)b or (i)c is a derivational source: (i)
a Gengogaku-wa hakaseronbun-ga taihen-da linguistics-TOP dissertation-NOM be-heard-COP ‘Speaking of linguistics, (writing) a dissertation is hard’ b Gengogaku-ga hakaseronbun-ga taihen-da linguistics-NOM dissertation-NOM be-heard-COP ‘It is linguistics where (writing) a dissertation is hard’ c Gengogaku-no hakaseronbun-ga taihen-da linguistics-GEN dissertation-NOM be-heard-COP ‘(Writing) a dissertation in linguistics is hard’
On the contrary, the sentences in (40) do not have such possible derivational sources. 29. Another possible account is based on semantics. For example, (46)a intends to mean ‘in the set of all the things under discussion (of which fish is
Notes 213 a member), it is fish that red snappers are the most delicious (in).’ However, this does not make sense because the proposition ‘the red snappers are the most delicious’ represents a property that is applicable only to fish, and would not be relevant to any other members of the sentence under discussion. Thanks to an anonymous reviewer for pointing this out to me. 30. More about comparison between clefts and HERCs will be discussed in Chapters 3 and 4. 31. Note that the trace in (48)b can be in a theta-position or in a major subject position, but not in a non-theta-marked topic position. In addition, although cleft constructions are not directly licensed by the aboutness relation, a major subject is licensed by the aboutness relation. 32. According to an anonymous reviewer, it is said that the following cleft sentences involving an NPI improve considerably: (i)
33. 34. 35. 36.
a
[Taroo-ga ei mi-nakatta-no]-wa [tada hitotu-no Kurosawa-no Taro-NOM didn’t-see-NM-TOP only one-GEN Kurosawa-GEN eiga mo] i da movie-also COP ‘What Taro didn’t see was even a single movie by Kurosawa’
b
[ Watasi-ga ei yom-itakunai-no]-wa [aitu-ga kaita I-NOM read-don’t-want-NM-TOP that-guy wrote ittuu-no tegami-mo]i da one-GEN letter-also COP ‘What I don’t want to read is even a single letter written by that guy’
However, I do not agree with the judgment given, and thus, I would like to leave this issue aside here. These facts need to be investigated further in future research. Thanks to Mark Baker and Jonathan Bobaljik for pointing out this issue. For details of this analysis, see Heggie (1988). Hoji (1985) points out that a reconstruction effect is observed when the topic phrase is interpreted contrastively. One might wonder whether the coindexation in (74)c is an ‘accidental’ coreference, and since the dake-phrase is quantificational in (73), it needs to c-command the anaphor in focus position rather than the null operator in the presuppositional clause. However, as will be discussed in Chapter 4, a quantifier can bind a null operator which is coindexed with a focus phrase. The relevant examples are shown below: (i)
a
[Daremoi-ga e hihansita -no]-wa zibuni-no ronbun da everyone-NOM criticized-NM-TOP self-GEN paper-COP ‘It was his paper that everyone criticized’
b
[Subete-no kaisyai-ga e uttaeta-no]-wa all-GEN company-NOM sued-NM-TOP sokoi-no kanrengaisya-da it-GEN subsidiary-COP ‘It was its subsidiary company that all companies sued’
214 Notes
37.
38.
39. 40.
41.
3
In (i)a, daremo ‘everyone’ does not c-command zibun ‘self’, which can be interpreted as a bound variable. However, the binding relation is allowed in the sentence. The same is true for (i)b; the quantifier, subete-no kaisya ‘all companies’, can be coindexed with soko ‘it’ under the bound variable reading. Unlike NPIs in focus position being displaced from its licensing head, these examples are grammatical. Therefore, the facts in (i) suggest that a quantifier does not have to bind a pronoun in focus position, but can bind a trace or the null operator within the presuppositional clause. The difference between Matsuda (1997) and Hasegawa (1997) is that Matsuda assumes that a focus phrase is base-generated in the spec of IP whereas Hasegawa assumes that it is scrambled to the spec of IP or an adjoined position to IP. Kuroda (1988) argues that a scrambled position can be the spec of IP under the subject internal hypothesis. Whether the landing site of the focus phrase, Hanako, is the spec of IP or a position adjoined to IP (Saito 1985) does not affect our discussion here. This fact is first noted by Hidekazu Tanaka (personal communication). Needless to say, this is not exactly correct if we consider stylistic matters for da and de aru in a certain context. The de aru form sounds more formal than the da form, which is presumably common across languages. This idea follows Nishiyama (1998) among others. In fact, the type of category for de as part of copula is not our main concern in this book. However, it is important that a focus phrase is a complement of the de for the purpose of theta role and/or case.
Nominalizations in Cleft Constructions
1. However, the conclusion that we have arrived at here crucially depends on the assumption that the language does not generate ungrammatical derivational sources. If we have a no-deletion rule, as proposed by Kitagawa and Ross (1982), the nos in (4)a–d may be all considered genitive markers. Kitagawa and Ross assume that (4)a and (4)d are derived in the following way: (i)
a
akai-no red-GEN Pro ‘the red one’
b
Hanako-no Hanako-GEN Pro ‘the one of Hanako’s’
This analysis, however, faces several problems concerning semantic restrictions and some dialectal facts for nos, as will be shown shortly. See also the discussion against this analysis in Murasugi (1991: 68–73). 2. The internal head of HIRCs is indicated by italics whereas the ‘syntactic (external) head’ is (marked by) no. I will gloss this type of no with NM (nominalizer) here.
Notes 215 3. Note that when the dative case marker appears on the focus position, the sentence would be ungrammatical even without the genitive marker in the presuppositional clause: (i)*
[[Hanako-ga e atta] hito]-wa Taroo-ni dat-ta Hanako-NOM met person-TOP Taro-DAT COP-PAST ‘*The person who Hanako met was to Taro’
The sentence in (i) is considered a topicalized copular construction, in which the two arguments have both to be NPs. 4. Under the analysis of Saito and Murasugi (1990), the ungrammaticality of (16)a and (16)b is due to the fact that the genitive CPs are considered adjuncts. They argue that only genitive phrases which have the subject, object or possessor theta role can license NP-deletion on a par with English NP-deletion. 5. One type of analysis for HIRCs in Japanese (Murasugi 1994, Mihara 1994 and others) assumes that the no-clause in these constructions is base-generated as a circumstantial adverbial clause which is associated with an empty argument. I do not specifically commit myself to this view; however, if it is the case, then NQs may modify the empty argument. The following example shows that this is presumably correct although the sentence gives us a partitive reading, which is different from the one in (30): (i)
[Kare-ga hon-o sasidasita no]-o pro go-satu uketotta he-NOM book-ACC handed-out NM-ACC five-CL received ‘(I) received five books out of those he handed out’
In order to avoid this possibility, I placed the NQ before the case-marker, o in (30)a. 6. In order to be consistent with the sentence in (30)b, the NQ is placed right after the no-clause in (31). If the no in (31b) is interpreted as a pro-form, the dative marker on the focus phrase is not allowed but the sentence becomes grammatical. 7. This analysis supports part of Hoji’s (1990) analysis for cleft constructions, where he proposes two types of structures for cleft constructions: (i)
a b
[NP [IP . . .(proi). . .] [NP noi]] -wa NPi da NM-TOP COP [CP Opi [IP . . .ti. . .] no]-wa NPi-(CASE) da NM-TOP COP
The structure in (32) is the same as the one in (i)b, and we argue that this is the only possible structure for Japanese clefts. For the discussion about this issue, see pages 24–7 in Chapter 2. 8. Similar structures like (33) have been proposed for HIRCs in previous research. Kuroda (1976–77), Tsubomoto (1981), Itô (1986), and Hoshi (1995) assume that there is an empty N or NP in HIRCs. 9. As Miyagawa (1993) notes, some native speakers of Japanese think that the wide-scope reading of the genitive phrase becomes difficult when an adverb
216 Notes appears before the genitive phrase. Overall, I also agree with the judgment that an adverb left to a genitive phrase prevents the wide-scope reading of the genitive phrase. We will come back to this issue later in this chapter. 10. Murasugi (1991) gives ?? to sentences such as (38)b; however, as Watanabe (1996) notes, the sentence is not considered ungrammatical. The discussion here is based on the latter type of judgment. At this point, I do not have any account of this marginality. 11. Watanabe (1996) also lists another property: ga-no conversion is limited to relative clauses and nominal complements. However, this is not always the case, as pointed out in the discussion by Watanabe. 12. Watanabe (1996) notes that while the transitivity effect is syntactic, some kind of intervening effect still matters for the following relative clauses: (i)
a??
b
[ John-no Mary-ni ei kasita] honi John-GEN Mary-DAT lent book ‘the book that John lent to Mary’ [ Mary-nij John-no tjei kasita] honi Mary-DAT John-GEN lent book
The relative clause in (i)a contains a genitive subject followed by a dative phrase. This is not considered ungrammatical but degraded compared with the one which does not have any intervening phrase between a genitive phrase and its predicate. Thus, as shown in (i)b, scrambling the dative phrase makes the sentence more acceptable. As far as I can see, Watanabe does not give an account for the nature of the distinct acceptability. I will also put this aside in this chapter. 13. In addition, a dative phrase between a genitive subject and its predicate makes cleft sentences degraded, as shown in (i)a; however, when the dative phrase is scrambled to clause-initial position, it becomes more acceptable, as in (i)b: (i) a?? [ Taroo-no Mary-ni e syookaisita no]-wa sono hito dat-ta Taro-GEN Mary-DAT introduced-NM-TOP that person COP-PAST ‘It was that person that Taro introduced to Mary’ (i)
b?
Taroo-no ti e syookaisita no]-wa [ Mary-nii Mary-DAT Taro-GEN introduced-NM-Top dat-ta COP-PAST
sono hito that person
The same behavior is observed for case conversion in relative clauses, as we have seen in the previous footnote. 14. It is also assumed that all arguments are base-generated within the VP (internal subject hypothesis). 15. However, Watanabe (1993, 1996) modifies the mechanism of (52), which will be shown later. 16. The definition of the Relativized Minimality is as follows (Rizzi 1990:7)
Notes 217 (i)
X α-governs Y only if there is no Z such that: a Z is a typical potential α-governor for Y b Z c-commands Y and does not c-command X
In terms of the Minimalist Program, this is considered the Minimal Link Condition, which says that α must make the ‘shortest move’. 17. The notions of dominance and containment are as follows (Chomsky 1993: 12): (i)
a The category α dominates β if every segment of α dominates β b The category α contains β if some segment of α dominates β
18. For how this system is reworked under the Agr-less theory of Chomsky (1995), see Watanabe (1995: 17–21). 19. As pointed out by Watanabe (1996), we need to assume that adjunction to AgrOP is only available for Wh-movement; otherwise, A-movement of the direct object would not induce a transitivity effect. Moreover, note that scrambling of an object cannot be movement to the spec of AgrS under Watanabe’s theory. 20. Miyagawa (1993) also succeeds in explaining the transitivity effects in case conversion. He also assumes that unlike nominative subjects, genitive subjects can remain within VP and move to the spec of DP at LF. If an accusative object occurs in the same clause with a genitive subject, movement of the object over the subject violates Minimality, because Miyagawa proposes that movement over elements with unchecked feature constitutes a Minimality violation. However, it is difficult for this analysis to account for the facts of comparative deletion constructions, which will be shown in (64). See also Watanabe (1996: 395–6) for arguments against Miyagawa (1993). 21. Here, we might ask a question of why standard Wh-interrogatives do not observe ga-no conversion: (i)
a*
Taroo-no dare-o semeta no Taro-GEN who-ACC accused Q Who did Taro accuse?’
b*
Taroo-no ti semeta no Dare-oi who-ACC Taro-GEN accused Q
As shown in (i), whether or not the Wh-phrase occurs between the genitive subject and the predicate, the sentences do not allow case conversion. As pointed out by Mark Baker (personal communication), the ungrammaticality in (i) is expected because the sentences are not instances of Wh-movement; Wh-in-situ never triggers inversion or morphological Wh-agreement across languages (e.g., French, Chamorro, and some African languages). If this is true, the facts in (i) naturally follow under Watanabe’s analysis. 22. As for optionality of ga-no conversion, Watanabe (1996) assumes that the NP-feature of Tns is optionally weak and the NP-feature of AgrS is always weak.
218 Notes 23. As presented in Hoshi (1995), the sentence in (70)b is somewhat degraded; however, it is crucially better than standard cases of complex NP violation, as we will see shortly. We should note that the fact in (70)b is not expected under the assumption that HIRCs are circumstantial adverbial phrases. I will put this problem aside here. 24. The same ungrammaticality is observed when a scrambled element is an argument: (i) a ?* Hon-oi Taroo-ga [[ Hanako-ga ti yondeiru] soba]o toorikakatta book-ACC Taro-NOM Hanako-NOM is-reading side-ACC passed-by ‘Taro passed by the place where Hanako was reading a book’. b ?*
Masao-nii Taroo-ga [[ Hanako-ga ti e ageta] hon]-o karita Masao-DAT Taro-NOM Hanako-NOM gave book-ACC borrowed ‘Taro borrowed the book that Hanako gave to Masao’
25. If a HIRC is regarded as a circumstantial adverbial clause, as proposed by Murasugi (1994) and Mihara (1994), extraction from the HIRC should not be allowed due to an adjunct island. The extraction facts show that this might not be the case; however, I will leave this question open at the moment. 26. As pointed out by Mark Baker (personal communication), we must assume either no reconstruction takes place at LF, or an invisible operator moves to the spec of CP at S-Structure (Watanabe 1992) to account for the fact in (85)c. 27. Thanks to Mark Baker for pointing this out to me.
4
Resumptive A’-Dependencies
1. The connectivity phenomenon is observed only in specificational (pseudo) cleft constructions. See footnote 4 in Chapter 1 to know what kind of cleft sentence is considered specificational. 2. For a specific analysis to deal with licensing of NPIs in it-clefts, see Percus (1997). 3. Taking a pseudocleft construction as an example, Peters and Bach (1968) propose the underlying structure for (1)b is something like (i)a, and the relevant portion of the structure is deleted. Chomsky (1970) and Hankamer (1974) assume that the underlying structure is (i)b which derives a surface sentence by extraposition: (i)
a
The thing that Mary bought something was Mary bought a picture of herself. b [NP it [S Mary bought a picture of herself]] was
In either way, the structures in (i) contain a paraphrased simple sentence, and the anaphor is bound by the underlying antecedent. Higgins (1979), on the other hand, claims that an underlying structure of pseudoclefts is the same as the surface structure, and the syntactic connectivity is accounted for by some semantic properties of specificational copular constructions. See Higgins (1979) for details of this line of analysis.
Notes 219 4. In Chapter 2, we already discussed some empirical reason why a direct binding relation between a focus phrase and its antecedent in a presuppositional clause is impossible. 5. There is a case where the spec of embedded CP is occupied by another Wh-constituent and we still have an ambiguous interpretation: (i)??
It is [a picture of himselfi/j] that Maxi wonders where Johnj bought [e]
The same result can be obtained in the following Wh-interrogative sentence: (ii)?? [Which pictures of himselfi/j] did Maxi wonder where Johnj bought [e] ?
6.
7.
8.
9.
10.
The sentences in (i) and (ii) are degraded due to subjacency; otherwise, the anaphor in focus position and in the fronted Wh-phrase can refer to either Max or John. This is not expected if the Wh-phrase needs a trace in the intermediate spec of CP to be coindexed with the higher subject. However, Barss (1986) and Chomsky (1986b) suggest that Wh-movement involves intermediate adjunction to the matrix VP, and the trace adjoined to the VP is bound by the moved Wh-phrase. See also Fox (1998: 158) which argues for adjuction to VP (or somewhere between subject and object positions) based on reconstructions facts. Thanks to Jonathan Bobaljik for pointing this out to me. At this point, this chapter need not take any particular theory of reconstruction. Although we henceforth use the terminology of ‘reconstruction’ and ‘to reconstruct’, I do not necessarily imply the adoption of literal reconstruction (Chomsky 1976) but also by the chain binding analysis (Barss 1986) or the copying/deletion approach (Chomsky 1993, among others). This chapter deals with the cleft constructions in which the presuppositional clause is marked by a topic marker, wa. Interestingly, Hasegawa (1997) points out that the clause marked by a nominative marker, ga, does not exhibit connectivity at all. I will put this aside here, but see Hasegawa (1997) for details. Ueda (1984) and Fukui (1984), cited in Kitagawa (1986), observe that zibun ‘self’ in a sentence such as (10)a is likely to be disjoint from the local antecedent, so that one might think that the grammaticality in (10)b resides in the marginality of local-binding of zibun in (10)a. However, as will be seen later, the same kind of contrast between non-clefted and clefted sentences as (10) can be observed even if the anaphor is replaced by a local anaphor (zibun-zisin ‘self-self’, kare-zisin ‘himself’, and otagai ‘each other’). This book, therefore, assumes that the contrast concerning binding facts does exist between non-clefted and clefted sentences, and the judgment by Ueda and Fukui can be explained by some pragmatic factors. Thanks to Mikinari Matsuoka for pointing this out to me. Most of the native speakers of Japanese that I know of agree with the judgments given in (10)b and (11)b; however, one of the informants allows ambiguous reading of these sentences. This book deals with the grammar of the former speakers and does not take into consideration the grammar which produces ambiguity for sentences such as (10)b and (11)b. As is well known, zibun ‘self’ may be construed as a logophoric pronoun (Sells 1987 and others). The possibility that zibun in focus position may not
220 Notes be licensed by syntactic binding but by discourse will be discussed on pages 116–21. 11. Since we do not find any difference between NP with a case marker and PP foci in terms of reconstruction effects, we will look at examples with an NP focus phrase only. 12. This goes against the adjunction prohibition proposed by McCloskey (1992, 1996): (i)
Adjunction to a phrase which is s-selected by a lexical head is ungrammatical. (McCloskey 1992: 11, 1996: 58)
Chomsky (1986a) argues that adjunction to IPs and VPs is allowed since they are non-argument categories whereas adjunction to an argument category, such as a CP complement, is prohibited. As far as I know, this targets only movement-derived adjunction; however, McCloskey claims that this is also the case for base-generated adjunction. His analysis is supported by the fact that an adverbial element cannot occur on the left of the complementizer to modify an IP inside the complement clause in English. The same test cannot be applied to Japanese because Japanese is strictly head-final, so that we do not know whether the position of the operator in long-distance clefts is the one adjoined to a CP or an IP. It is possible for us to assume that, following McCloskey, (i) is universal and the base-generated position for operators in Japanese clefts is a position adjoined to an IP. This does not affect our discussion in this book. However, the facts concerning Clitic Left Dislocation in Modern Greek suggest that a null operator is generated in a position adjoined to a CP rather than an IP (see pages 139–43). 13. The marginality of the coindexed reading between Ziroo and kare-zisin ‘himself’ may come from some pragmatic or processing reasons. The point here is that the coreference interpretation is obtained much easier than that of Masao and the anaphor. 14. This is first noted by Hidekazu Tanaka (personal communication). 15. The relevant definitions for c-command and dominance are shown below: (i)
a
α c-commands β iff: i) neither α nor β dominates the other, and ii) β is dominated by the first branching node that dominates α
b
α dominates β iff all segments of α dominate β
16. There is even another possible adjunction site for a null operator in cleft constructions; a position adjoined to the highest VP in a presuppositional clause. This might be plausible because in some languages, such as Hungarian (Kiss 1986 and others), the position adjoined to a VP is reserved for a focus phrase. Japanese cleft constructions, however, do not share such syntactic properties, as we will see on pages 127–30. 17. In most of the examples above, I tried to present examples in which zibun ‘self’ appears in argument position of a verb. This is because English anaphors can be interpreted as a logophoric pronoun typically in non-argument position
Notes 221 of a verb, as discussed in Reinhart and Reuland (1991, 1993). Therefore, the sentences we have seen so far are at least free from the logophoric possibility in the sense of Reinhart and Reuland. However, as we will see shortly, zibun can be used a logophoric pronoun even when it appears in argument position (Sells 1987). 18. The gloss is my own in order to be consistent with other examples in this book. The judgment in (27)c is grammatical in Sells (1987); however, I do not agree with his judgment. 19. Zibun is not the only type of anaphor which exhibits intermediate reconstruction effects in cleft constructions. As is shown in the previous sections, local anaphors, zibun-zisin ‘self–self’ and kare-zisin ‘himself’, give us the same distribution as that of zibun. Otagai ‘each other’ also shows the same intermediate reconstruction effects:
(i)
a
[sensei-gataj-ga otagai*i/j-no Taroo to Hanakoi-ga Taro and Hanako-NOM teachers-NOM each-other’s bunseki-ni tanotteiru to] omotteiru analysis-DAT count-on COMP think ‘Taro and Hanako think that the teachers counted on each other’s analyses’
b
[sensei-gataj-ga e tayotteiru to] [Taroo to Hanakoi-ga Taro and Hanako-NOM teachers-NOM count-on COMP omotteiru no]-wa otagaii/?*j-no bunseki-ni dat-ta think NM-TOP each-other’s analyses-DAT COP-PAST ‘It was each other’s analyses that Taro and Hanako thought that the teachers counted on’
However, Aikawa (1994) proposes that zibun-zisin can be interpreted as a logophoric pronoun. In addition, Hoji (1997) claims that otagai is not a reciprocal anaphor corresponding to English ‘each other’; he argues that what we have considered as a binding relation between otagai and its antecedent should be understood as a coreferential relation. These reciprocal pronouns have always a possibility of logophoric usage. 20. Thanks to Hiroshi Aoyagi for pointing this out to me. 21. The judgment for the sentence in (34)b is subtle. In fact, one of the native speakers of Japanese that I consulted (from approximately five informants) does not seem to change the interpretation between (34)a and (34)b; they are both ambiguous. However, the other informants admit that there is a contrast, as shown in (34). 22. In fact, the ungrammatical linking between the soko and the conjoined NP in (34)b should be examined carefully. According to Hoji (1995: 259), the following sentence is ruled out by weak crossover: (i)*
[Toyota to Nissan]i-o uttaeta Sokoi-no kumiai-ga it-GEN labor-union-NOM Toyota and Nissan-ACC sued ‘Itsi labor unions sued [Toyota and Nissan]i’
222 Notes If conjoined NPs in Japanese are considered quantifiers which raise at the level of interpretation, we can say that the conjoined NP in (i) moves crossing over the soko-phrase at LF, and hence, it violates a condition of weak crossover. If this is the case, then the conjoined NP in (34)b may move over the reconstructed focus phrase in the position adjoined to the embedded clause, which gives us an instance of weak crossover. However, since the sentence is grammatical, we should crucially assume that the conjoined NP in (34)b adjoins to the closest clause, namely IP, and does not move over the reconstructed focus operator in the position adjoined to CP. 23. Lasnik and Stowell (1991) also point out that the generalization can be restated as follows: (i)
In a configuration where a category C A’-binds a pronoun P and a trace T, P may not be contained in an argument phrase XP that c-commands T.
Either generalization would be adopted to see the point we are making in this subsection. 24. There are cases where no WCO effects arise even when a null operator is bound by a Wh–constituent or a quantifier, as shown in (i): (i)
a
Which boyi [ti will be easy [Opi [PRO to persuade hisi wife to talk about ei]]]
b
No childi is old enough for us [Opi [PRO to ask [hisi mother] to give up ei]] (Lasnik and Stowell 1991:696)
The grammaticality of these sentences can be explained by the fact that the pronoun is not locally A’-bound by the true (overt) quantifier, even though it is A’-bound by the null operator which is externally bound by the quantifier. In order to account for the WCO effects in Japanese clefts, we should assume that unlike English tough-constructions, a null operator in Japanese clefts has the same logical status as its antecedent (= focus element) by agreement. See the discussion concerning agreement relations between a null operator and a focus phrase in cleft constructions in Chapter 2. 25. Lasnik and Stowell (1991) propose more a complicated analysis for the sentences in (41). Their analysis predicts that (41)a should be grammatical, because an operator in (pseudo)clefts has a logical function of the Wh-operators (Chomsky 1977) with some covert quantifier (meaning only). The contrast between (41)a and (41)b is explained by reconstruction and quantifier raising; even though the focus phrase, nothing, in (41)b is reconstructed to its base position, it undergoes quantifier raising, which triggers a WCO effect. In contrast, since the focus phrase in (41)a is not a quantifier, it does not trigger a WCO effect. 26. In this chapter, I assume that the theta position in long-distance clefts is occupied by pro. An alternative analysis is proposed in Tanaka and Kizu
Notes 223 (1999), where we argue that the theta position is empty by a deletion operation on a par with comparative deletion in English (Bresnan 1976). Both possibilities should be investigated for future research. 27. Here, I follow Tanaka’s (1997) proposal that A’-dependencies are licensed at S-structure. See Tanaka (1997, 1998) for details. 28. Tanaka (1997, 1998), as well as Watanabe (1992), assumes that the spec of CP is not located in the left of the head but in the right of the head. Thus, Tanaka’s definition of the LCC is presented as follows: (i)
29.
30.
31. 32.
Linear Crossing Constraint (LCC): Suppose that Opi precedes Opi + 1 and that ti, the trace of Opi, precedes ti + 1, the trace of Opi + 1, then ti + 1 cannot precede Opi at S-Structure.
Therefore, A’-dependency lines are created by an operator and its licensing specifier. Since the aim of this chapter is to show striking facts of longdistance clefts in Japanese, I do not follow technical details proposed by Tanaka (1997, 1998). This book simply assumes that a specifier is on the left of its head, and association paths of A’-dependencies are represented as linking an operator base-generated in an adjoined position and its licensing head. Saito (1992) assumes that scrambling is freely undone at LF. Therefore, it is predicted that the anaphor, zibun ‘self’, can refer to the most embedded subject NP under the strict sense of the scrambling analysis. The scrambling analysis, therefore, needs to assume that an A-moved phrase is not put back at LF; however, we lose force to account for the reconstruction facts in local clefts, as will be argued later in this chapter. The facts of local clefts in terms of reconstruction effects imply that the scrambling analysis for reconstruction in clefts predicts total reconstruction in any case. One of the eight informants that I consulted allows coreferent reading between the second highest subject and the anaphor. This book does not take this type of grammar into consideration but examines the grammar which presents the judgments in (51)b. I have no explanation of the marginality of (53)b; however, it is crucial that the sentence in (53)b is better than (53)c. The definition of Empty Category Principle is as follows (Murasugi 1991): (i)
ECP: α properly governs β iff α c-commands β, and a) α theta-marks or Case-marks β, or b) α is coindexed with β and β is subjacent to α.
33. Under the framework of the Minimalist Program, Takahashi (1997) gives an account for the absence of true adjunct relativization in Japanese. Assuming also that no resumptive pro can take the place of a true adjunct PP, he claims that only the relevant operator feature of the null operator (and not the entire operator) undergoes movement to the highest spec of CP. This move-F hypothesis can explain why relative clauses such as (58)b are ruled out; moving a feature from an adjunct operator violates the adjunct condition.
224 Notes 34. Murasugi (1991) does not show exactly where the operator originates. However, the point here is that the trace of the base position is governed by the C0 under her analysis. 35. This analysis poses a problem for local relativization of a true adjunct PP: (i)
[Hanako-ga ei sigoto-o yameta] riyuui Hanako-NOM job-ACC quit reason ‘the reason that Hanako quit the job’
The relative clause in (i) is perfect; however, the analysis given in (59) predicts that the trace is not properly governed for the same reason that the intermediate is not governed in (58)b. Murasugi (1991) argues that the relative clause in (i) does not involve a trace, but is a pure instance of complex NP, similar to the following: (ii)
a
[sakana-ga yakeru] nioi fish-NOM burn smell ‘(lit.) the smell that the fish burns’
b
[doa-ga simaru] oto door-NOM shut sound ‘(lit.) the sound that the door shut’
Murasugi’s structure for these complex NPs is illustrated as follows: [NP [IP . . .] [N’ reason]]
(iii)
36. One might wonder whether a similar contrast we have seen in (58) is also observed in the corresponding topic constructions. For some reason, a topicalized reason PP does not sound perfect even in a simple topic sentence: (i)
??
Sono riyuu-dei-wa [Hanako-ga ei sigoto-o yameta] that reason-for-TOP Hanako-NOM job-ACC quit ‘For that reason, Hanako quit the job’
The topicalized reason PP sounds more natural when it appears in some negative clause; however, it seems that the interpretation relies on the position of the NEG head, as follows: (ii)
a?*
[Taroo-ga [Hanako-ga ei sigoto-o yameta Sono riyuu-dei-wa that reason-for-TOP Taro-NOM Hanako-NOM job-ACC quit to] omow-anai COMP think-NEG ‘For that reason, Taro doesn’t think that Hanako quit the job’
b
Sono riyuu-dei-wa [Taroo-ga [Hanako-ga ei sigoto-o that reason-for-TOP Taro-NOM Hanako-NOM job-ACC yamenakatta to] omotteiru quit-NEG COMP think
Notes 225 I do not have any analysis to account for at the moment, and will leave them for future research. 37. The same paradigm is observed for HERCs, as shown below: (i)
a*
[Masao-ga [Taroo-ga [Hanako-ga ei sigoto-o yameta Masao-NOM Taro-NOM Hanako-NOM job-ACC quit to] omotteiru to] itta] riyuui COMP think COMP said reason ‘the reason that Masao said that Taro thought that Hanako quit the job’
b*
[Masao-ga [Taroo-ga ei [Hanako-ga sigoto-o yameta Masao-NOM Taro-NOM Hanako-NOM job-ACC quit to] omotteiru to] itta] riyuui COMP think COMP said reason
c
sigoto-o yameta [Masao-ga ei [Taroo-ga [Hanako-ga Masao-NOM Taro-NOM Hanako-NOM job-ACC quit to] omotteiru to] itta] riyuui COMP think COMP said reason
38. Thanks to Mark Baker for pointing this out to me. 39. All the examples cited in this subsection are from Iatridou (1991). 40. Cinque (1990) argues that Italian CLLD constructions exhibit only properties of unbounded dependencies. They do not demonstrate typical diagnoses for Wh-movement, such as licensing of parasitic gaps and ne-cliticization. 41. As we discussed on pages 125–7, a similar construction to (64)b, such as topicalization, is a case of weakest crossover in English (Lasnik and Stowell 1991). I do not go into details of why such a difference is observed here. 42. Iatridou (1991) also argues that CLLD is not derived from a clitic doubling construction, because this construction has some semantic restrictions on the left-dislocated element in Modern Greek and Rio Platese Spanish. Note also that Italian does not have a clitic doubling construction which can derive a corresponding CLLD sentence. 43. Iatridou (1991) does not provide relevant examples to show that the language does not allow more than one Wh-phrase in the spec of CP. 44. Iatridou (1991) suggests that the same range of data can be explained even if the base-generated position for a left-dislocated phrase is a position adjoined to IP. Mark Baker (personal communication) also pointed out to me that a left-dislocated element comes after a complementizer in some Romance languages and Mohawk. 45. This structure violates the adjunction prohibition discussed in McCloskey (1992, 1996). See note 12 in this chapter. 46. As we will see on pages 148–50, Japanese local clefts do not exhibit WCO effects either. However, my analysis is crucially dependent on reconstruction effects in local clefts, assuming that reconstruction is observed only when movement is involved. 47. Thanks to Ekaterina Klepousniotou for the judgments. 48. The examples cited in this subsection are from Finer (1997). Here, ʔ stands for a glottal stop. In Selayarese, which demonstrates VOS word order,
226 Notes Wh-interrogatives must have a Wh-phrase in a preverbal position, as shown in the following:
49. 50.
51. 52.
53. 54.
(i)*
ku-isse?-i kuko la-kco-i i Ali inai 1-know-3 COMP 3-call-3 DET Ali who ‘I know who called Ali’
(ii)
ku-isse?-i kuko inai la-keo-i i Ali 1s-know-3 COMP who 3-call-3 DET Ali
The sentence in (i) is ruled out since the Wh-phrase appears in post-verbal position. The sentences in (i) and (ii) suggest that the Wh-phrase has to move obligatorily. Finer argues that this is because the Wh-phrase must check its strong feature before spell-out. For the technical details of the analysis to account for this fact, see Finer (1997). According to Finer (1997), the theoretical reason why the Wh-phrase is not base-generated in its surface position is that the Wh-feature has to be checked by movement and not by merge (Chomsky 1995). The thematic position for an operator is indicated by [e]. Another possibility is to assume that the gap in question is created by deletion (Bresnan 1976). This analysis is presented in Tanaka and Kizu (1999); however, it has to distinguish a case of deletion under the copy theory (Chomsky 1993) from a case of deletion for the gap in cleft constructions. Since this analysis needs more work, I will not pursue this possibility in this book. This fact contrasts with parasitic gap constructions in Modern Greek. I will put this difference aside here. As discussed in the literature, when the head of HERCs contains zibun ‘self’, the grammaticality is degraded: (i)?? [Hanako-ga e hihansita] zibun-no ronbun Hanako-NOM criticized self-GEN paper ‘self’s paper that Hanako criticized’ I do not have an account to explain the unacceptability of (i), but it should be pointed out that when the head contains another type of anaphoric pronoun, such as zibun-zisin ‘self-self’, the HERCs sound perfect: (ii)
[Hanako-ga e hihansita] zibun-zisin-no ronbun Hanako-NOM criticized self-self-GEN paper ‘self’s paper that Hanako criticized’
In what follows, anaphoric pronouns other than zibun ‘self’ are used for the examples with HERCs. 55. See also the discussion on pages 127–30. 56. In (92)a, the complement clause is scrambled to the clause-initial position. This is to avoid intervention effects between the no-phrase and its predicate.
Notes 227
5
Ellipsis in Cleft Constructions
1. This is intuitively clear for the cases of NP (or N’) ellipsis and sluicing. As for VP ellipsis, Lobeck (1995) assumes that empty VPs must be properly governed by AGR through verb raising (Chomsky 1993). The AGR properly governs empty VPs under the Generalized Government Transparency Corollary proposed by Baker (1988). 2. As far as I know, a ‘remnant’ originally indicates the Wh-constituent which overtly appears after deletion of the following IP. However, I will use this term for any item which remains in the elided clause, and do not intend to commit myself to a particular analysis. 3. The exact analysis for ellipsis, namely, whether it involves PF deletion or LF copying, has been controversial in the literature. A deletion operation requiring identity with its antecedent clause is proposed by Ross (1967), Hankamer and Sag (1976), Sag (1977) and others, and an interpretive hypothesis is adopted by Wasow (1972), Williams (1977), Chao (1987), and Chung, Ladusaw and McCloskey (1995). As far as I can see, both analyses present problems. For instance, the deletion analysis has difficulty in determining how strict the identity is at PF, and while the interpretive approach may have a way to get rid of the problem, it still needs an explanation for subjacency sensitivity in some English and Japanese elliptical structures (Johnson 1997). The Japanese facts which will be shown later in this chapter, in fact, tell us that Japanese sluicing is accounted for by the PF deletion approach. 4. Judgments on Wh-remnants with a major case marker, such as ga ‘nominative’ and o ‘accusative’, vary among native speakers of Japanese. Although Takahashi (1994) considers the sluiced clause containing these case markers to be perfect, Wh-remnants without a nominative or accusative case marker are much more preferable for the native speakers of Japanese that I consulted (see also Nishiyama, Whitman, and Yi 1996 for the same judgments). I will provide an account for cases without such major case markers, and will put aside the examples with major case markers here. For sluicing with a major case marker, see also Kuwabara (1995, 1996) and Fukaya (1998, 2003). 5. Nishiyama (1995), in fact, does not claim that Japanese sluicing is a cleft construction on the grounds of contrasting facts of subjacency which he observes. Rather, he proposes that sluicing involves a covert copular construction. The same analysis as Nishiyama is proposed by Fukaya (1998) for ‘sluicing without case markers’; however, he argues that ‘sluicing with case markers’ is explained in a different way. See also Fukaya (2003). Kuwabara (1995, 1996) clearly states that Japanese sluicing is derived from an equivalent cleft structure where the underlying presupposition portion has a VP operator. This operator moves to the spec of CP, and thus gets coindexed with the focused portion, which is also a VP. I, however, assume, following Hoji (1990), that a cleft construction involves movement of a null operator to the spec of CP, as already discussed in the previous chapters. 6. Fukaya (1998) has a different interpretation; when the remnant is not marked by a case marker, it does not require any linguistic antecedent. I do not find such a contrast here, and have no account for the difference in judgments at this point. 7. For details, see Hankamer and Sag (1976).
228 Notes 8. Takahashi’s (1994) original sentences contain an accusative marker on the Wh-remnant. This is omitted in the examples in (9) in order to make the sentence sound perfect for the native speaker’s grammar studied in this book (see footnote 4). However, the absence of the case marker does not affect the interpretation with respect to sloppy/strict reading even for the speakers who also allow a major case marker on remnants. 9. Under the interpretive hypothesis, the phenomenon of sloppy identity is analyzed using lambda notation; a semantic rule will affix a lambda and a variable to a VP, and it will place a variable bound by the lambda in the position of the logical subject of the verb of the VP (Williams 1977:115). For instance, the sentence in (i)a is represented as (i)b when it is interpreted under the sloppy reading: (i)
a John shot himself and Bill did too b John [VPλx (x shot x)] and Bill [VPλx (x shot x)] too (Williams 1977: 117)
However, Fiengo and May (1994:130) point out several problems with this analysis. For instance: (ii)
Max said he saw his mother, and Oscar did, too
This sentence expects four logical possibilities in the interpretation of the second sentence; however, one of them is impossible: (iii)
a . . . and Oscar said that Max saw Max’s mother b . . . and Oscar said that Oscar saw Oscar’s mother c . . . and Oscar said that Oscar saw Max’s mother d* . . . and Oscar said that Max saw Oscar’s mother
This kind of eliminative effect of ellipsis can be observed in other types of elided constructions. According to Fiengo and May, these can be accounted for by the ‘dependency theory’ (see, for the details, Fiengo and May 1994). This issue is out of the scope of this chapter, and we will not discuss it further here. 10. Our analysis here does not care whether an element in COMP is considered a proper governor or not. On our account, the reason why the sentences in (11) are ungrammatical is that they lack focus phrases which must remain overtly in elided constructions. 11. Takahashi (1993) takes a certain case of scrambling involving a Wh-phrase to be an instance of syntactic Wh-movement in Japanese. If this is the case, then sluicing would be a second construction in which overt Wh-movement takes place. Recall, however, that a standard Wh-interrogative sentence does not exhibit properties of typical Wh-agreement with or without scrambling. See the discussion of case conversion in Chapter 3. 12. However, the following sentence might be a possible derivational source for (12)a under the analysis involving Wh-movement: (i)
yonda-no-da] ka] . . . . . . [CP dare-gai[IP ti sono hon-o who-NOM that book-ACC read-NM-COP-Q ‘. . . who it is read that book . . .’
Notes 229
13. 14.
15.
16.
17.
18.
19.
20.
21.
The structure in (i) shows that the sentence as a whole is nominalized by the no-marker. From within this nominalized clause, the Wh-constituent, dare-ga ‘who-NOM’, moves to the spec of CP. Putting aside the exact structure of the nominalized sentences for now, we find that this might be plausible since this underlying sentence is grammatical unlike (13). However, as will be discussed later in this chapter, this type of derivation requires an overt case marker with a Wh-remnant. Therefore, for those speakers who accept sluicing with a major case marker, the structure in (i) would be a possible derivational source, although some of the facts in the following subsection still need to be explained under the Wh-movement analysis. Thanks to Jeong-Seok Kim for the Korean sentence. Hoh and Chiang (1990) state that there are two types of shi in Chinese: one is the copula, and the other a focus marker. Although they argue that the shi in cleft constructions is in fact a focus marker, what is crucial for our analysis is that the same morpheme appears both in sluicing and cleft constructions, just like Japanese. Thanks to Ays Gurel who provided Turkish examples for me. In these sentences, the -y before the copula is an epenthetic segment, and the vowel in the past tense form of the copula (di, du) undergoes vowel harmony with the preceding vowel. Korean does not allow nominative and accusative markers on Wh-remnants of sluicing or on focus phrases in clefts. However, as for a dative marker, it is obligatory with the Wh-remnant but is degraded when it appears in the focus position. I do not have any account for this discrepancy at this point. Thanks to Myunghyun Yoo for helping me out with these Korean facts. Furthermore, recall Saito’s (1985) analysis of scrambling of a subject (see the discussion in Chapter 2). The Wh-movement analysis for sluicing cannot assume this line of analysis, because clefting a subject is analogous to scrambling a subject under the Wh-movement analysis. While the Wh-movement analysis can stipulate a rule of case marker drop obligatorily or optionally, the facts here at least show that the Wh-movement/ scrambling analysis for Japanese sluicing requires further mechanisms. As pointed out by Mark Baker (personal communication), a possible solution under this analysis is to assume that the focus position is reanalyzed as a predicate position, which does not need a case, following Safir (1985, 1987). Thus, the true case markers, ga and o, do not appear in the remnant Wh-phrases. It might be possible that the Wh-movement analysis accounts for the ungrammaticality of (24)b. Assuming that the spec of CP is occupied by a single constituent, (24)b is ruled out because the NQ-NP in (25)b adjoins to the IP by scrambling independently, and hence, this operation is not considered Wh-movement. This line of analysis is plausible; however, there is another fact which can be accounted for by our present analysis but not by the Wh-movement analysis. This will be shown in the next subsection. It should be noted here that we ignore the question whether the Japanese concessive Wh-phrases are comparable to those in English. This should be investigated further in further research. The examples in (31)–(33) are cited from Takahashi (1994: 17–18) with some modification to be consistent with the examples throughout in this
230 Notes
22.
23. 24.
25.
26.
27. 28.
29.
30.
31.
chapter. Note also that the judgment on (31)b is ?? according to Takahashi (1994); however, I cannot see a contrast between (31)b and (32)b/(33)b. Fukaya (1998) divides sluicing constructions into two types: sluicing with a case marker and sluicing without a case marker. He observes that the former exhibits a clause-bounded nature whereas the latter has no subjacency effects. There is a way to generate only one possible derivational source for sluicing constructions. This will be discussed shortly. Here, we should note that it must be assumed that the ‘deletion under identity’ relation does not obey subjacency; a clause inside an island can be the antecedent of the empty presuppositional clause in Japanese sluicing. Under our analysis for cleft constructions, the amalgamation takes place between two null operators within the presuppositional clause, and the two Wh-elements in the focus position are base-generated as one constituent. Therefore, it is not clear whether the two operators are amalgamated before spell-out or at LF. See Tanaka (1998), which gives a different analysis from Saito (1992) and Takahashi (1994) based on extensive data in Japanese multiple Wh-constructions. As for other possible derivations, see Takahashi (1994). In the antecedent sentence in (52)a, we do not use an indefinite NP, such as dareka ‘someone’, as an antecedent of the Wh-remnant, because the sentence itself would exhibit a WCO effect. Instead, Hanako is used as the antecedent so that the sentence alone does not show any WCO effects. However, since the hidden presuppositional clause in (52)b has a null operator, which is associated with dare(-ni) ‘who(-DAT)’, the sluiced sentence may be considered a WCO configuration, as will be discussed shortly. Concerning (52)b, the dative marker on the Wh-remnant, dare ‘who’, must appear to avoid an ambiguous reading. When the antecedent of the Wh-remnant is indefinite, this is not the case. The exception for this standard view is the analysis proposed by Fukaya (1998). He assumes, following Ueyama’s (1997) framework, that the sluiced IP is empty at LF but has a copy from its antecedent in semantics. Here again, we cannot use dareka ‘someone’ or nanika ‘something’ in the antecedent sentence to generate the following sluiced clause, because cooccurrence of these indefinite NPs and the sika-nai ‘only-NEG’ makes the sentence degraded to a simple sentence, regardless of linear order: (i)
a??
Taroo-sika nanika-o kaw-ana-katta Taro-only something-ACC buy-NEG-PAST ‘Only Taro bought something’
Taroo-sika ti kaw-ana-katta b?? Nanika-oi something-ACC Taro-only buy-NEG-PAST ‘For something, only Taro bought’ Putting aside the reason why the sentences in (i) do not sound perfect, I will not use indefinite NPs in the antecedent sentences to avoid this effect.
Notes 231 32. Local sluicing constructions exhibit the same syntactic behaviors as cleft constructions in terms of sika phrase interaction: (i)
Hanako-sika hoka-no gakusei-ni syukudai-o Hanako-only another-GEN student-DAT assignment-ACC tetudattemoraw-anak-atta rasii ga, watasi-wa [dono was-helped-NEG-PAST I-heard but I-TOP which gakusei(-ni)(-datta)]-ka wakaranai student(-DAT)(-COP-PAST)-Q don’t-know ‘I heard that only Hanako had another student to help her assignment, but I don’t know who.’
For local cleft constructions, we proposed that the operator can scramble to sentence-initial position, so that the derivational source relevant to the LCC in (i) is represented as in (ii): (ii) [Opi Hanako-sika ti syukudai-o tetudattemoraw-ana-katta-no]-wa Hanako-only assignment-ACC was-helped-NEG-PAST-NM-TOP dare(-ni)(-datta)-ka . . . who(-DAT)(-COP)-Q ‘. . . who it is that only Hanako had to help her assignment’ Assuming that the italicized operator and the underlined sika phrase are associated with their corresponding head in the structure in (ii), we can account for the grammaticality of the sluiced sentence in (i) as well. 33. Strictly speaking, Tanaka (1997, 1998) assumes that invisible operators from a sika phrase and a Wh-element move to the designated spec position at S-structure, which is crucially on the left of its head. Since our claim in this book is not concerned with this particular mechanism, the schematized structures are illustrated simply with dependency paths between sika/ Wh-phrases and their licensing head. 34. Remember that Tanaka (1997, 1998) assumes that the association line links the overt Wh-phrase and its invisible operator moved to the spec of CP at S-structure. 35. For some unknown reason, zibun in the remnant phrase does not make a sentence perfect irrespective of local and long-distance sluicing: (i)??
zibuni-ni izonsiteiru rasii ga, watasi-wa Hanakoi-ga Hanako-NOM self-DAT rely-on I-heard but I-TOP yoku siranai hontooni zibuni(-ni-datta)-ka truly self(-DAT-COP-PAST)-Q well don’t-know ‘I heard that Hanako relied on herself, but I don’t know (if it was) truly herself.’
I speculate that the marginality of sentences such as (i) comes from the occurrence of the matrix subject watasi ‘I’. For some kind of processing reason, an overt element which is available as a possible antecedent can be taken over the other covert possible antecedents. Note also that using
232 Notes go-zibun ‘Hon-self’ eliminates this possibility because the honorific prefix on the anaphoric pronoun cannot be used when it indicates the speaker. Therefore, the sentences in the text are not marginal. 36. Otagai ‘each other’ does not have a corresponding honorific form. 37. While it is assumed in the Minimalist Program that binding conditions apply at LF, Lasnik (1999) suggests that S-structure is crucial for matters for binding.
References
Abe, Jun, 1997, ‘The locality of zibun and logophoricity’, Report (1): Researching and Verifying an Advanced Theory of Human Language (COE Research): 595–626, Kanda University of International Studies. Abney, Steven, 1987, ‘The English noun phrase and its sentential aspect’, doctoral dissertation, MIT, Cambridge, Mass. Aikawa, Takako, 1994, ‘Logophoric use of the Japanese reflexive zibun-zisin ‘self-self’, Formal Approaches to Japanese Linguistics 1: MIT Working Papers in Linguistics 24: 1–22, eds Masatoshi Koizumi and Hiroyuki Ura, MIT, Cambridge, Mass. Akmajian, Adrian, 1970, ‘Aspects of the grammar of focus in English’, doctoral dissertation, MIT, Cambridge, Mass. Aoun, Joseph, 1986, Generalized Binding, Dordrecht: Foris. Baker, Mark, 1988, Incorporation: A Theory of Grammatical Function Changing, Chicago: University of Chicago Press. Baker, Mark, 1996, The Polysynthesis Parameter, New York: Oxford University Press. Baltin, Mark, 1987, ‘A landing site theory of movement rules’, Linguistic Inquiry 13: 1–38. Barss, Andrew, 1986, ‘Chains and anaphoric dependence: on reconstruction and its implications’, doctoral dissertation, MIT, Cambridge, Mass. Bobaljik, Jonathan David and Dianne Jonas, 1996, ‘Subject positions and the roles of TP,’ Linguistic Inquiry 27(2): 195–236. Bonneau, José, 1992, ‘The Structure of internally headed relative clauses: implications for compositionality’, doctoral dissertation, McGill University. Bošcovic, Zeljko, 1997, ‘Pseudoclefts’, Studia Linguistica 51 (3): 235–77. Bresnan, Joan, 1976, ‘Evidence for a theory of unbounded transformation’, Linguistic Analysis 2: 353–93. Browning, Marguerite, 1987, ‘Null operator constructions’, doctoral dissertation, MIT, Cambridge, Mass. Chao, Wyn, 1987, ‘On ellipsis’, doctoral dissertation, University of Massachusetts, Amherst, Mass. Chomsky, Noam, 1970, ‘Remarks on nominalization’, in Readings in English Transformational Grammar, ed. Roderick A. Jacobs and Peter S. Rosenbaum, 184–221, Waltham: Ginn and Co. Chomsky, Noam, 1973, ‘Conditions on transformations’, A Festschrift for Morris Halle: 232–86, eds Stephen R. Anderson and Paul Kiparsky, New York: Holt, Rinehart and Winston. Chomsky, Noam, 1976, ‘Conditions on rules of grammar’, Linguistic Analysis 2: 303-51. Chomsky, Noam, 1977, ‘On Wh-movement’, Formal Syntax, eds Peter W. Culicover, Thomas Wasow and Adrian Akmajian: 71–132, New York: Academic Press. Chomsky, Noam, 1981, Lectures on Government and Binding, Dordrecht: Foris. 233
234 References Chomsky, Noam, 1982, Some Concepts and Consequences of the Theory of Government and Binding, Cambridge, Mass: MIT Press. Chomsky, Noam, 1986a, Knowledge of Language: Its Nature, Origin and Use, New York: Praeger. Chomsky, Noam, 1986b, Barriers, Cambridge, Mass: MIT Press. Chomsky, Noam, 1991, ‘Some notes on economy of derivation and representation’, in Principles and Parameters in Comparative Grammar: 417–54, ed. Robert Freidin, Cambridge, Mass.: MIT Press. Chomsky, Noam, 1993, ‘A minimalist program for linguistic theory’, The View from Building 20: 1–52, eds Kenneth Hale and Samuel Jay Keyser, Cambridge, Mass.: MIT Press. Chomsky, Noam, 1995, The Minimalist Program, Cambridge, Mass.: MIT Press. Chung, Sandra, 1982, ‘Unbounded dependencies in Chamorro grammar’, Linguistic Inquiry 13: 39–77. Chung, Sandra, 1994, ‘Wh-agreement and “referentiality” in Chamorro’, Linguistic Inquiry 25: 1–44. Chung, Sandra, William Ladusaw, and James McCloskey, 1995, ‘Sluicing and Logical Form’, Natural Language Semantics 3: 239–82. Cinque, Guglielmo, 1990, Types of A’-Dependencies, Cambridge, Mass.: MIT Press. Cinque, Guglielmo, 1997, ‘“Topic” constructions in some European languages and “connectedness” ’, in Materials on Left Dislocation: 93–118, eds Elena Anagnostopoulou, Henk Van Riemsdijk and Frans Zwarts, John Benjamins Publishing Company. Cole, Peter, 1987, ‘The structure of internally headed relative clauses’, Natural Language and Linguistic Theory 5: 277–302. Collins, Chris and Phil Branigan, 1997, ‘Quotative inversion’, Natural Language and Linguistics Theory 15: 1–41. Culy, Christopher D., 1990, ‘The syntax and semantics of internally headed relative clauses’, doctoral dissertation, Stanford University. Declerck, Renaat, 1984, ‘The pragmatics of it-clefts and Wh-clefts’, Lingua 64: 251–89. Déprez, Vivianne, 1988, ‘Stylistic inversion and verb movement’, Proceedings of the Eastern States Conference on Linguistics. Déprez, Vivianne, 1990, ‘Two ways of moving the verb in French’, MIT Working Papers in Linguistics 13: 27–46, eds Lisa L.S. Cheng and Hamida Demirdash, Cambridge, Mass.: MIT Press. Drijkoningen, Frank, 1988, ‘Stylistic inversion in French and the extended projection principle’, Linguistics in the Netherlands: 39–47, eds Peter Coopmans and Aafke Huld, Dordrecht: Foris. Fiengo, Robert, 1977, ‘On trace theory’, Linguistic Inquiry 8: 35–61. Fiengo, Robert and Robert May, 1994, Indices and Identity, Cambridge, Mass.: MIT Press. Finer, Daniel L., 1997, ‘Contrasting A’-dependencies in Selayarese’, Natural Language and Linguistics Theory 15: 677–728. Fox, Danny, 1998, ‘Economy and semantic interpretation’, doctoral dissertation, MIT, Cambridge, Mass. Fukaya, Teruhiko, 1998, ‘On the so-called “sluicing” in Japanese’, MS, University of Southern California. Fukaya, Teruhiko, 2003, ‘Island (in)sensitivity in Japanese sluicing and stripping and some implications’, Proceedings of the West Coast Conference on Formal Linguistics 22: 179–92.
References 235 Fukui, Naoki, 1984, ‘Studies on Japanese anaphora I: the adjunct subject hypothesis and zibun,’ MS, MIT, Cambridge, Mass. Fukui, Naoki, 1986, ‘A theory of category projection and its applications’, doctoral dissertation, MIT, Cambridge, Mass. Goldberg, Adele, 1995, Constructions: A Construction Grammar Approach to Argument Structure, Chicago: The University of Chicago Press. Haïk, Isabelle, 1990, ‘Anaphoric, pronominal and referential infl’, Natural Language and Linguistic Theory 8: 347–74. Hankamer, Jorge, 1974, ‘On the non-cyclic nature of WH-clefting’ Proceedings of the Tenth Regional Meeting of the Chicago Linguistic Society, 221–33. Hankamer, Jorge and Ivan Sag, 1976, ‘Deep and surface anaphora’, Linguistic Inquiry 7: 391–428. Harada, Shin–ichi, 1971, ‘Ga-no conversion and idiolectal variations in Japanese’, Gengo-Kenkyu 60: 25–38. Harada, Shin-ichi, 1976, ‘Ga-no conversion revisited’, Gengo-Kenkyu 70: 23–38. Hasegawa, Nobuko, 1984/85, ‘On the so-called “zero pronouns” in Japanese’, The Linguistic Review 4: 289–341. Hasegawa, Nobuko, 1988, ‘Remarks on “zero pronominals”: in defense of Hasegawa (1984/85)’, Proceedings of Japanese Syntax Workshop: Issues of Empty Categories: 50–76, eds Mineharu Nakayama and Wako Tawa. Hasegawa, Nobuko, 1997, ‘A copula-based analysis of Japanese clefts: Wa-cleft and ga-cleft’, in Report (1): Researching and Verifying an Advanced Theory of Human Language (COE Research): 15–38, Kanda University of International Studies. Heggie, Lorie, 1988, ‘The syntax of copular sentences’, doctoral dissertation, University of Southern California. Heggie, Lorie, 1993, ‘The range of null operators: evidence from clefting’, Natural Language and Linguistic Theory 11:45–84. Heycock, Caroline, 1991, ‘Layers of predication: the non-lexical syntax of clauses’, doctoral dissertation, University of Pennsylvania. Heycock, Caroline and Anthony Kroch, 1996, ‘Pseudocleft connectivity: implications for the LF interface level’, Edinburgh Occasional Papers in Linguistics: 96(1). Higgins, Francis Roger, 1979, The Pseudo-cleft Construction in English, New York: Garland Publishing. Hiraiwa, Ken and Shinichiro Ishihara, 2002, ‘Missing links: clefts, sluicing, and “no-da” construction in Japanese’, MIT Working Papers in Linguistics 43: 35–54. Hoekstra, Teun and René Mulder, 1990, ‘Unergatives as copular verbs: locational and existential predication’, The Linguistic Review 7: 1–79. Hoh, Pau-San and Wen-yu Chiang, 1990, ‘A focus account of moved Wh-phrases at S-structure in Chinese’, Lingua 81: 47–73. Hoji, Hajime, 1985, ‘Logical form constraints and configurational structures in Japanese’, doctoral dissertation, University of Washington. Hoji, Hajime, 1990, ‘Theories of anaphora and aspects of Japanese Syntax’, MS, University of Southern California. Hoji, Hajime, 1995, ‘Demonstrative binding and principle B’, Proceedings of the North East Linguistic Society 25: 255–71. Hoji, Hajime, 1997, ‘Otagai (Each other),’ MS, University of Southern California. Horn, Laurence, 1985, ‘Metalinguistic negation and pragmatic ambiguity’, Language 61: 121–74.
236 References Hoshi, Koji, 1995, ‘Structural and interpretive aspects of head-internal and headexternal relative clauses’, doctoral dissertation, University of Rochester. Huang, C.-T. James, 1982, ‘Logical relations in Chinese and the theory of grammar’, doctoral dissertation, MIT, Cambridge, Mass. Iatridou, Sabine, 1991, ‘Clitics and island effects’, MS, MIT, Cambridge, Mass. Imai, Takashi, 1987, ‘Some consequences of move-α and Japanese grammar’, Issues in Japanese Linguistics: 203–27, eds Takashi Imai and Mamoru Saito, Dordrecht: Foris. Inoue, Kazuko, 1976, Henkeibonpoo to Nihongo (Transformational Grammar and Japanese), Japan: Taishukan. Ishii, Yasuo, 1988, ‘Head-internal relative clauses in Japanese’, Proceedings of the Eastern States Conference on Linguistics 8: 234–45. Ishii, Yasuo, 1991, ‘Operators and empty categories in Japanese’, doctoral dissertation, University of Connecticut. Itô, Junko, 1986, ‘Head-movement at LF and PF: the syntax of head-internal relatives in Japanese’, University of Massachusetts Occasional Papers in Linguistics 11: 109–38, eds Nobuko Hasegawa and Yoshihisa Kitagawa. Johnson, Kyle, 1997, ‘When verb phrases go missing’, MS, University of Massachusetts, Amherst. Kamio, Akio, 1983, ‘Meisiku no koozoo (structures of noun phrases)’, Nihongo no Koozoo (Structure of Japanese), ed. Kazuko Inoue, Japan: Sanseido. Katada, Fusa, 1991, ‘The LF representation of anaphors’, Linguistic Inquiry 22: 287–313. Kawashima, Ruriko, 1994, ‘The structure of noun phrases and the interpretation of quantificational NPs in Japanese’, doctoral dissertation, Cornell University. Kayne, Richard S., 1989, ‘Facets of romance past participle agreement’, Dialect Variation and the Theory of Grammar: 85–103, ed. P Benincà, Dordrecht: Foris. Kayne, Richard S., and Jean-Yves Pollock, 1978, ‘Stylistic inversion, successive cyclicity, and move NP in French’, Linguistic Inquiry 9: 595–621. Kikuchi, Akira, 1987, ‘Comparative deletion in Japanese’, MS, Yamagata University. Kikuchi, Akira and Daiko Takahashi, 1991, ‘Agreement and small clauses’, in Topics in Small Clauses: 75–105, eds Heizo Nakajima and Shigeo Tonoike, Japan: Kuroshio Publishers. Kiss, Katalin É, 1986, Configurationality in Hungarian, Dordrecht Kluwer. Kiss, Katalin É, 1998, ‘Identificational focus versus informational focus’, Language 74(2): 245–73. Kitagawa, Chisato and Claudia N.G. Ross, 1982, ‘Prenominal modification in Chinese and Japanese’, Linguistic Analysis 9(1): 18–53. Kitagawa, Yoshihisa, 1986, ‘Subjects in Japanese and English’, doctoral dissertation, University of Massachusetts, Amherst. Kitahara, Hisatsugu, 1993, ‘Numeral classifier phrases inside DP and the specificity effect’, Japanese/Korean Linguistics 3: 171–86, ed. Soonja Choi, CSLI, Stanford University, California. Kizu, Mika, 1997, ‘Sluicing in Wh-in-situ languages’, Proceedings of the Thirty-third Regional Meeting of the Chicago Linguistic Society: 231–44. Koizumi, Masatoshi, 1995, ‘Phrase structure in minimalist syntax’, doctoral dissertation, MIT, Cambridge, Mass. Koizumi, Masatoshi, 2000, ‘String vacuous overt verb raising’, Journal of East Asian Linguistics 9: 227–85.
References 237 Kubo, Miori, 1992, ‘Japanese syntactic structures and their constructional meaning’, doctoral dissertation, MIT, Cambridge, Mass. Kuno, Susumu, 1973, The Structure of the Japanese Language, Cambridge, Mass: MIT Press. Kuroda, Shige-yuki, 1965, ‘Generative grammatical studies in Japanese’, doctoral dissertation, MIT, Cambridge, Mass. Kuroda, Shige-yuki, 1975–76, ‘Pivot-independent relativization in Japanese II’, Papers in Japanese Linguistics 4: 85–96. Kuroda, Shige-yuki, 1976–77, ‘Pivot-independent relativization in Japanese III’, Papers in Japanese Linguistics 5: 157–79. Kuroda, Shige-yuki, 1987, ‘Movement of noun phrase in Japanese’, Issues in Japanese Linguistics, eds Takashi Imai and Mamoru Saito, Foris Publications. Kuroda, Shige-yuki, 1988, ‘Whether we agree or not: a comparative study of English and Japanese’, ed. W. Poser, Papers from the Second International Workshop on Japanese Syntax, CSLI Publications. Kuwabara, Kazuki, 1995, ‘Some problems of sluiced clauses in Japanese’, MS, Kanda University of International Studies, Japan. Kuwabara, Kazuki, 1996, ‘Multiple Wh-phrases in elliptical clauses and some aspects of clefts with multiple foci’, Formal Approaches to Japanese Linguistics 2: The MIT Working Papers in Linguistics 29: 97–116, eds Masatoshi Koizumi, Masayuki Oishi and Uli Sauerland, MIT, Cambridge Mass. Lasnik, Howard, 1995, ‘A note on pseudogapping’, eds R. Pensalfani and H. Ura, Papers on Minimalist Syntax, The MIT Working Papers in Linguistics 27: 143–63, MIT, Cambridge, Mass. Lasnik, Howard, 1999, Minimalist Analysis, Blackwell Publishers. Lasnik, Howard, 1999, ‘Chains of arguments’, Working Minimalism: 189–215, eds Samuel Epstein and Nobert Hornstein, Cambridge, Mass.: MIT Press. Lasnik, Howard and Mamoru Saito, 1984, ‘On the nature of proper government’, Linguistic Inquiry 14: 235–89. Lasnik, Howard and Mamoru Saito, 1992, Move a: Conditions on its Application and Output, Cambridge, Mass.: MIT Press. Lasnik, Howard and Tim Stowell, 1991, ‘Weakest crossover’, Linguistic Inquiry 22: 687–720. Lobeck, Ann, 1991, ‘The phrase structure of ellipsis’, ed. S. Rothstein, Syntax and Semantics 25, Perspectives on Phrase Structure: Heads and Licensing: 81–103, San Diego: Academic Press. Lobeck, Ann, 1995, Ellipsis, Oxford: Oxford University Press. Matsuda, Yuki, 1997, ‘A syntactic analysis of focus sentences in Japanese’, MIT Working Papers in Linguistics 31: Papers from the Eighth Student Conference in Linguistics: 291–310, ed. Benjamin Bruening, MIT, Cambridge, Mass. May, Robert, 1981, ‘Movement and binding’, Linguistic Inquiry 12: 215–43. May, Robert, 1985, Logical Form: Its Structure and Derivation, Cambridge, Mass.: MIT Press. McCawley, James, 1998, The syntactic phenomena of English, (second edition), Chicago: The University of Chicago Press. McCloskey, James, 1979, Transformational Syntax and Model Theoretic Semantics: A Case Study in Modern Irish, Dordrecht: D. Reidel. McCloskey, James, 1992, ‘Adjunction, selection and embedded verb second’, Linguistic Research Center, University of California, Santa Cruz.
238 References McCloskey, James, 1996, ‘On the scope of verb movement’, Natural Language and Linguistic Theory 14: 47–104. Mihara, Kenichi, 1992, Jiseikaisyaku to Toogogensyo (The Interpretation of Tense and Syntactic Phenomena), Japan: Kuroshio. Mihara, Kenichi, 1994, ‘Iwayuru Syuyoobunaizaigata-kankeisetu ni tuite (on the so-called head-internal relative clauses)’, Nihongogaku (Japanese Linguistics) 13, Japan: Meijishoin. Miyagawa, Shigeru, 1993, ‘Case-checking and minimal link condition’, Papers on Case and Agreement II: MIT Working Papers in Linguistics 19: 213–54, eds Jonathan D. Bobaljik and Colin Phillips, MIT, Cambridge, Mass. Murasugi, Keiko, 1991, ‘Noun phrases in Japanese and English: a study in syntax, learnability, and acquisition’, doctoral dissertation, University of Connecticut. Murasugi, Keiko, 1994, ‘Head-internal relative clauses as adjunct pure complex NPs’, Synchronic and Diachronic Approaches to Language: A Festschrift for Toshio Nakao on the Occasion of His Sixtieth Birthday: 425–37, eds S. Chiba et al., Japan: Liber Press. Nishiyama, Kunio, 1995, ‘Sluicing without Wh-movement’, eds. N. Díaz-Insensé and L. López, Cornell Working Papers in Linguistics 13: 85–96, New York: Cornell University. Nishiyama, Kunio, 1998, ‘The morphosyntax and morphophonology of Japanese predicates’, doctoral dissertation, Cornell University. Nishiyama, Kunio, John Whitman and Eun-Young Yi, 1996, ‘Syntactic movement of overt Wh-phrases in Japanese and Korean’, Japanese/Korean Linguistics 5, eds. Noriko Akatsuka, Shoichi Iwasaki and Susan Strauss, CSLI Publications, Stanford, California. Noda, Hisashi, 1996, Wa to Ga (Wa and Ga), Kuroshio, Japan. Ohara, Kyoko Hirose, 1994, ‘A correlative analysis of so-called internally-headed relativization in Japanese’, paper presented at the Annual Meeting of Linguistic Society of America, Boston. Otani, Kazuyo and John Whitman, 1991, ‘V-raising and VP ellipsis’, Linguistic Inquiry 22. Percus, Orin, 1997, ‘Prying open the cleft’, Proceedings of the North East Linguistic Society 27: 337–51, ed. Kiyomi Kusumoto, GLSA, University of Massachusetts, Amherst. Perlmutter, David, 1972, ‘Evidence for shadow pronouns in French relativization’, The Chicago Which Hunt, eds Paul M. Peranteau, Judith N.Levi and Gloria C. Phares, Chicago Linguistic Society, Chicago, Illinois. Pesetsky, David, 1987, ‘Wh-in situ: movement and unselective binding’, The Representation of (In)definiteness: 98–129, eds Eric J. Reuland and Alice ter Merlen, Cambridge, Mass: MIT Press. Peters, Stanley and Emon Bach, 1968, ‘Pseudo-cleft sentences’, MS, MIT, Cambridge, Mass. Pinkham, Jessie and Jorge Hankamer, 1975, ‘Deep and shallow clefts’, Proceedings of the Eleventh Regional Meeting of the Chicago Linguistic Society: 429–50. Pollock, Jean-Yves, 1989, ‘Verb movement, universal grammar, and the structure of IP’, Linguistic Inquiry 20: 265–424. Prince, Ellen, 1978, ‘A comparison of Wh-clefts and it-clefts in discourse’, Language 54(4): 883–906.
References 239 Reinhart, Tanya, 1976, ‘The syntactic domain of anaphora’, doctoral dissertation, MIT, Cambridge, Mass. Reinhart, Tanya and Eric Reuland, 1991, ‘Anaphors and logophors: an argument structure perspective’, Long-distance Anaphora: 283–321, eds Jan Koster and Eric J. Reuland, Cambridge: Cambridge University Press. Reinhart, Tanya and Eric Reuland, 1993, ‘Reflexivity’, Linguistic Inquiry 24: 657–720. Rizzi, Luigi, 1986, ‘Null objects in Italian and the theory of pro’, Linguistic Inquiry 17: 501–57. Rizzi, Luigi, 1990, Relativized Minimality, Cambridge, Mass.: MIT Press. Rizzi, Luigi, 1997, ‘The fine structure of the left periphery’, Elements of Grammar: Handbook in Generative Syntax: 281–337, ed Liliane Haegeman, Dordrecht: Kluwer. Ross, John, 1967, ‘Constraints on variables in syntax’, doctoral dissertation, MIT, Cambridge, Mass. Ross, John, 1969, ‘Guess Who?’, Proceedings of the Chicago Linguistic Society: 252–86. Safir, Ken, 1985, Syntactic Chains, Cambridge: Cambridge University Press. Safir, Ken, 1987, ‘The syntactic projection of lexical thematic structure’, Natural Language and Linguistic Theory 5: 561–601. Sag, Ivan, 1977, Deletion and Logical Form, Bloomington, Indiana: Indiana University Linguistics Club. Saito, Mamoru, 1983, ‘Case and government in Japanese’, Proceedings of the West Coast Conference on Formal Linguistics 2: 247–59. Saito, Mamoru, 1985, ‘Some asymmetries in Japanese and their theoretical implications’, doctoral dissertation, MIT, Cambridge, Mass. Saito, Mamoru, 1992, ‘Long distance scrambling in Japanese’, Journal of East Asian Linguistics 1: 69–118. Saito, Mamoru and Hajime Hoji, 1983, ‘Weak crossover and move-α in Japanese’, Natural Language and Linguistic Theory 1: 245–59. Saito, Mamoru, and Keiko Murasugi, 1990, ‘ N’-deletion in Japanese: a preliminary study’, ed. H. Hoji, Japanese/Korean Linguistics: 285–301, CSLI Publications, Stanford, California. Sakai, Hiromu, 1994, ‘Complex NP constraint and case-conversion in Japanese’, Current Topics in English and Japanese: 179–203, ed. Masaru Nakamura, Tokyo: Hituji Syobo. Schachter, Paul, 1973, ‘Focus and Relativization’, Langage 49: 19–46. Sells, Peter, 1987, ‘Aspects of logophoricity’, Linguistic Inquiry 18: 445–79. Shibatani, Masayoshi, 1978, Nihongo no Bunseki (Analyses of the Japanese Language), Japan: Taishukan. Sohn, Keun-Won, 1994, ‘Adjunction to argument, free ride and a minimalist program’, Formal Approaches to Japanese Linguistics 1: MIT Working Papers in Linguistics 24: 315–34, eds Masatoshi Koizumi and Hiroyuki Ura, MIT, Cambridge, Mass. Stewart, O.T., 1998, ‘The serial verb construction parameter’, doctoral dissertation, McGill University. Stowell, Tim, 1981, ‘Origins of phrase structure’, doctoral dissertation, MIT, Cambridge, Mass. Stowell, Tim, 1983, ‘Subjects across categories’, The Linguistic Review 2: 285–312. Takahashi, Daiko, 1993, ‘Movement of Wh-phrases in Japanese’, Natural Language and Linguistic Theory 11: 655–78. Takahashi, Daiko, 1994, ‘Sluicing in Japanese’, Journal of East Asian Linguistics 3: 265–300.
240 References Takahashi, Daiko, 1997, ‘Move-F and null operator movement’, The Linguistic Review 14: 181–96. Takano, Yuji, 2002, ‘Surprising constituents’, Journal of East Asian Linguistics 11: 243–301. Tanaka, Hidekazu, 1997, ‘Invisible movement in sika-nai and the linear crossing constraint’, Journal of East Asian Linguistics 6: 143–88. Tanaka, Hidekazu, 1998, ‘Conditions on logical form derivations and representations’, doctoral dissertation, McGill University. Tanaka, Hidekazu and Mika Kizu, 1999, ‘Chain binding in nominal constructions’, paper presented at Move and Interpret α: Workshop on Movement Related Issues in Syntax and/or Semantics, Kanda University of International Studies, Japan. Tateishi, Koichi, 1994, The Syntax of ‘Subjects’, CSLI Publications and Kuroshio Publishers. Tellier, Christine, 1989, ‘More on Mooré relatives: licensing internal topics,’ MS, Université de Québec à Montreal. Terada, Michiko, 1990, ‘Incorporation and argument structure in Japanese’, doctoral dissertation, University of Massachusetts, Amherst. Tsubomoto, Atsuro, 1981, ‘It’s all no: unifying function of no in Japanese’, Proceedings of the Seventeenth Regional Meeting of the Chicago Linguistic Society: 393–403. Ueda, Masanobu, 1984, ‘On a Japanese reflexive zibun: a non-parametrization approach’, MS, University of Massachusetts, Amherst. Ueda, Masanobu, 1990, ‘Japanese phrase structure and parameter setting’, doctoral dissertation, University of Massachusetts, Amherst. Ueda, Masanobu, 1993, ‘On the phrase structure of Japanese and English clauses’, Japanese Syntax in Comparative Grammar: 9–44, ed. Nobuko Hasegawa, Japan: Kuroshio Publishers. Ueyama, Ayumi, 1989, ‘Focus no “ga” to Nihongo no Kukoozoo (focus ga and phrase structures of Japanese)’, Kansai Linguistic Society 9: 1–16. Ueyama, Ayumi, 1997, ‘Two types of dependency’, MS, University of Southern California. Ura, Hiroyuki, 1993, ‘On feature checking for Wh-traces’, Papers on Case and Agreement II: MIT Working Papers in Linguistics 19: 377–99, eds Jonathan D. Bobaljik and Colin Phillips, MIT, Cambridge, Mass. Valois, Daniel and Fernande Dupuis, 1992, ‘On the status of (verbal) traces in French: the case of stylistic inversion’, Romance Languages and Modern Linguistic Theory: 325–38, eds Paul Hirschbüler and K. Koerner, Amsterdam: John Benjamins. Wasow, Thomas, 1972, ‘Anaphoric relations in English’, doctoral dissertation, MIT, Cambridge, Mass. Watanabe, Akira, 1991, ‘Wh-in-situ, subjacency, and chain formation’, MS, MIT, Cambridge, Mass. Watanabe, Akira, 1992, ‘Subjacency and s-structure movement of Wh-in-situ’, Journal of East Asian Linguistics 1: 255–91. Watanabe, Akira, 1993, ‘AGR-based case theory and its interaction with the A-bar system’, doctoral dissertation, MIT, Cambridge, Mass. Watanabe, Akira, 1995, Case Absorption and Wh-Agreement, Dordrecht: Kluwer. Watanabe, Akira, 1996, ‘Nominative-genitive conversion and agreement in Japanese: a cross-linguistic perspective’, Journal of East Asian Linguistics 5: 373–410.
References 241 Williams, Edwin, 1977, ‘Discourse and logical form’, Linguistic Inquiry 8, 101–39. Williams, Edwin, 1980, ‘Predication’, Linguistic Inquiry 11: 203–38. Williams, Edwin, 1983, ‘Against small clauses’, Linguistic Inquiry 14: 287–308. Williamson, Janis, 1987, ‘An indefiniteness restriction for relative clauses in Lakhota’, The Representation of (In)definiteness: 168–90, eds Eric J. Reuland and Alice ter Merlen, Cambridge, Mass.: MIT Press.
Index
Abe, J., 116 Abney, S., 72 aboutness, 5, 10, 12, 25, 34, 52, 211n, 213n A’-dependency, 8, 101, 110, 121, 127–9, 128, 129, 138, 141–3, 145, 157, 191, 193, 223n adjunct clause, 4, 11, 30, 122, 124–5, 141–2, 147, 150, 157, 176, 179, 187, 188 constraint, 180, 223n island, 11, 179, 182, 218n PP, 101, 134–8, 157, 211n, 212n, 224n, 225n adjunction, prohibition of, 220n adverbial circumstantial, 218n placement, 92 Agr-less theory, 217n agreement, 37, 109, 126, 156, 159, 184, 222n (see also feature sharing) chain, 36, 47–8 spec-head, 136, 160, 161 suffix, 144–5 wh-, 77, 86–9, 99, 152, 154, 217n, 228n Aikawa, T., 221 Akmajian, A., 2, 103, 207n amalgamation, of wh-phrases, 184–5, 230n A-movement, 96, 130–1, 134, 149, 150, 185–6, 188, 206, 217 A’-movement, 1–2, 5, 9, 13–14, 34, 97, 121, 130–1, 139–40, 142, 145, 149–50, 152, 156, 205, 206, 212n anaphor, 37, 110, 116, 117, 119, 121, 130–2, 213n, 218n, 219n, 220n reciprocal, 116, 151, 201 reflexive, 3, 6–7, 14, 21–2, 30, 35, 47, 102–4, 106–9, 111–14, 117–18, 131, 142, 147–8, 198, 200, 213n, 220n, 221n, 223n, 226n, 232n A’-position, 97, 185 appositive, 76, 77 articulated CP, 48
Barss, A., 72, 104, 109, 210, 219 binding, 22, 105, 114, 116, 119, 120, 121, 130, 132, 144–5, 147, 148, 157, 200, 201, 220n backward, 117 chain, 140, 147, 178–9, 210n condition, see binding theory principle, see binding theory referential, 114, 130, 132 theory, 37, 38, 103, 116, 117, 186, 199, 202, 210n, 232n Bobaljik, J. and Jonas, D., 86 Boškovic, Z., 103, 207 bound variable, 3, 6, 22, 103, 113, 114, 120, 164, 214 Browning, M., 36, 38, 39 case checking, 81, 83, 84, 86 conversion, 8, 58, 66, 73–81, 83, 84, 85, 87, 89–90, 99, 152–4, 206, 216n, 217n theory, 81, 87 category/segment distinction, 115 c-command, definition of, 210n, 220n Chamorro, 154, 217 Chao, W., 227n Chinese, 167, 168, 229n Chomsky, N., 1, 2, 3, 4, 12, 72, 74, 81, 82, 85, 88, 102, 103, 104, 108, 140, 142, 150, 157, 197, 210n, 217n, 218n, 219n, 220n Chung, S., Ladusaw, W., and McCloskey, J., 227n Cinque, G., 108, 139, 140, 142, 147, 178, 211n, 225n Cinque’s Paradox, 140 clausemate condition, 185 clitic CLLD (Clitic Left Dislocation), 139, 140, 141, 142, 143, 147, 148, 211, 220n, 225n doubling, 225n complementizer, 8, 57, 58, 63, 66, 72, 95, 99, 144, 145, 165, 206 242
Index 243 complex NP, 77, 90, 92, 93, 97, 178, 179, 180, 209n, 224n constraint, 11, 141, 182, 218n extraction, 90–1 concessive, wh-phrase, 173 connectivity, 3, 4, 6, 8, 9, 35, 37, 47, 101–7, 110–13, 147, 148, 205, 207n, 218n, 219n anti-, 35 see also reconstruction construction grammar, 1 containment, 217n contrastive, 12, 13, 14, 15, 16, 17, 18, 23, 24, 174, 210 coordinate structure, 53–4, 171 copula, 1, 48, 54, 55, 59, 162, 165, 166–8, 214n copy LF, 189, 191–2, 196–7, 203, 206, 227n theory, 226n CP extraposition, 95 Declerck, R., 207 deletion comparative, 87, 217n, 223n PF, 8, 159, 161, 162, 164, 186, 189, 191–7, 202, 203, 206, 227n see also ellipsis dependency theory, 228n Déprez, V., 85 discourse structure, 117 D-link, 42, 43, 49 dominance, 217n economy, 97, 157 ECP (Empty Category Principle), 185, 223n Edo, 154 ellipsis, 8, 160, 163, 164, 165, 173, 175, 189, 203, 206 see also deletion empathy, 22–3 equidistance, 82 exhaustive listing, 12 feature checking, 18 feature sharing, 37, 126, 184 see also agreement Finer, D., 143, 144 Fox, D., 219n French, 78, 85, 86, 87, 217n Fukaya, T., 201 Fukui, N., 160, 219
ga-no conversion, see case conversion Generalized Government Transparency Corollary, 227n German, verb second phenomenon, 48 government, 81 government and binding theory, 2, 102, 104, 136, 160 Greek, see Modern Greek Haïk, I., 154, 155, 156 Hankamer, J., 102, 103, 163, 218n Hasegawa, N., 47, 50, 54, 118, 209n, 214n, 219n Hausa, 154 Heggie, L., 2, 39, 72, 103, 104, 207, 208n, 213n Heycock, C. and Kroch, A., 103, 207n Higgins, F., 2, 47, 102, 103, 218n Hiraiwa, K. and Ishihara, S., 48, 209n Hoh, P. and Chiang, W., 229n Hoji, H., 5, 7, 9, 10, 12, 13, 14, 15, 16, 17, 18, 24, 25, 26, 34, 41, 57, 110, 117, 119, 120, 122, 209n, 210n, 211n, 213n, 215n Horn, L., 208n Hoshi, K., 58, 62, 63, 69, 70, 72, 90, 92, 93, 94, 95, 215n, 218n Hungarian, 220n Iatridou, S., 139, 140, 141, 142, 143, 225n Imai, T., 209n improper movement, 185, 186 Inoue, K., 54, 73 internal subject hypothesis, 216n interpretive hypothesis, 228n Irish, 154 Ishii, Y., 15, 52, 63 island, 4, 10, 11, 16, 18, 20, 23, 27, 52, 102, 140, 141, 146, 159, 165, 175–83, 218n Italian, 139, 147, 148, 211 it-cleft, 2, 3, 6, 72, 103, 104, 207, 208n Itô, J., 63, 67, 72, 76, 80, 215n Kayne, R. and Pollock, J., 85 Kikuchi, A., 40, 87 Kikuyu, 154 Kiss, K., 207n Kitagawa, C. and Ross, C., 58, 214n Kitagawa, Y., 212n, 219n Kitahara, H., 53, 73, 171 Kizu, M., 8, 159, 162 Koizumi, M., 48, 53, 54, 171, 209n
244 Index Korean, 167, 168 Kuno, S., 4, 5, 9, 10, 11, 18, 24, 27, 29, 31, 106 Kuroda, S., 17, 19, 20, 32, 62, 63, 67, 72, 76, 106, 209n, 210n, 211n, 214n, 215n Kuwabara, K., 8, 48, 54, 159, 162, 165, 183, 209n, 227n Lasnik, H. and Stowell, T., 122, 125, 126, 222n LCC (Linear Crossing Constraint), 127–9, 132, 149, 151, 152, 190, 191, 193–5, 197, 223n, 231n Lobeck, A., 160, 164, 165, 173, 191 logophoric pronoun, 110, 116, 118, 121, 219n, 220n major subject construction, 18, 212n see also multiple nominative construction manner adverb, 86, 93, 94, 95 Matsuda, Y., 47, 208n, 209n, 214n McCloskey, J., 220n, 225n merge, 226n Mihara, K., 20, 209n, 210n, 215n, 218n minimal domain, 82, 86 Minimal Link Condition, 157, 217n Minimalist Program, 1, 197, 206, 209n, 217n, 223n minimality, 82, 86, 87, 217n Miyagawa, S., 74, 75, 88, 215n, 217n Modern Greek, 101, 139–43, 145, 220n, 225n, 226n Mohawk, 225n Mooré, 154, 155, 156 move-F, 223n multiple foci, 5, 53, 54, 183 multiple nominative construction, 18, 19, 211n see also major subject construction Murasugi, K., 58, 60, 65, 67, 134, 136, 137, 209n, 214n, 215n, 216n, 218n NPI (Negative Polarity Item), 3, 6, 7, 9, 35, 36, 43, 44, 45, 102, 103, 105, 127, 208n, 213n Nishiyama, K., 8, 54, 159, 162, 165, 167, 176, 179, 214n, 227n Nishiyama, K., Whitman, J. and Yi, E., 8, 162 Noda, M., 54 NQ (numeral quantifier), 50, 51, 53, 54, 66, 71, 72, 77, 99, 171, 172, 215n
null pronoun, 24, 122, 124, 144, 150, 157, 187 see also pro NumP (number phrase), 72 Ohara, K., 63, 69, 71 operator invisible, 98, 122, 218n, 231n movement, 4 null, 2, 4, 7–9, 27, 30, 31, 34, 35–9, 46–8, 53, 56, 87, 97, 101, 103, 104, 109, 111–14, 116, 118, 121, 122, 124–6, 128–30, 131, 134, 136, 138, 143, 146, 148, 150, 151, 156, 157, 159, 179, 180, 184, 188, 190, 191, 205, 206, 210n, 213n, 214n VP, 227n wh-, 14, 144 parasitic gap, 140, 141–2, 143, 147, 225n Percus, O., 102, 103, 207, 208, 218n Pesetsky, D., 42 Peters, S. and Bach, E., 103, 218n Pinkham, J. and Hankamer, J., 103 Pollock, J., 81, 85 predication, 1, 2, 27, 34, 36, 38, 39, 40, 104, 141 predicational cleft construction, 207n Prince, E., 207n Principles and Parameters theory, 1, 191 pro, 8, 14–15, 20, 24, 25, 27, 52, 114, 122, 124, 125, 130, 134, 136, 137, 138, 147, 148, 150, 157, 178, 179, 180, 185, 188, 191, 198, 211n, 215n, 222n pro-form (pronominal form), 57–68, 72, 74 Proper Binding Condition, 48, 98 pseudocleft, 2, 3, 39, 103, 104, 207, 208n, 218n quantifier, 6, 21, 50, 66, 69, 107, 120, 122, 126, 150, 171, 172, 213n, 214n reconstruction, 3, 8, 14, 15, 36, 38, 101, 103–9, 114, 118, 119, 121, 129, 131, 132, 142, 148, 149, 151, 157, 165, 186, 191, 193, 197, 199, 201, 203, 206, 210n, 213n, 218n, 219n see also connectivity
Index 245 relative clause, 4, 13, 14, 16, 18, 20–3, 25, 38, 52, 53, 61, 62, 88, 118, 136, 141, 147, 179, 180, 209n, 210n, 216n free, 3, 5, 212n HERC (head-external relative clause), 24, 27–30, 34, 66, 69, 71, 73, 74, 76, 78, 91, 135, 150, 151, 152, 153, 154, 213n, 225n, 226n HIRC (head-internal relative clause), 8, 57, 58, 63, 66, 67, 68, 69, 70–2, 75–7, 79, 80–1, 89–90, 92, 94–8, 155, 156, 206, 214n, 215n, 218n H-less RC (headless relative clause), 61, 67, 71, 72, 73, 74, 76 non-restrictive, 118 Relativized Minimality, 82, 97, 216n Reo Platese Spanish, 225n Rizzi, L., 48, 82, 136, 209n, 216n Ross, J., 58, 160, 164, 214n Saito, M., 9, 10, 11, 12, 16, 49, 50, 52, 60, 65, 74, 97, 122, 128, 134, 136, 149, 184, 209n, 210n, 211n, 214n, 215n Sakai, H., 18, 20, 22, 26, 209n scope, 21, 75, 88, 89, 215n, 216n scrambling, 8, 9, 10, 16, 47, 48, 49, 50, 51, 52, 53, 54, 56, 57, 79, 80, 91–8, 128, 130, 131, 134, 135, 149, 159, 170, 172, 175, 188, 205, 209n, 212n, 216n, 217n, 223n Selayarese, 102, 143, 145, 225n Sells, P., 117, 219n, 221n sloppy identity, 163, 164, 228n small clause, 38, 39, 40, 56, 208n Sohn, K., 184 specificational cleft, 39, 102, 207n, 208n, 218n split INFL hypothesis, 81 standard theory, 160 stylistic inversion, 78, 85–7 subjacency, 4, 13, 16, 23, 24, 25, 26, 27, 146, 147, 182, 210n, 219n, 227n, 230n subjunctive, 78, 85, 87 Takahashi, D., 40, 159, 160, 161, 162, 163, 164, 165, 166, 175, 183, 184, 185, 186, 209n, 223n, 227n, 228n, 230n
Takano, Y., 48, 54 Tanaka, H., 42, 68, 124, 127, 149, 184, 193, 194, 197, 201, 212n, 214n, 223n, 230n, 231n Tanaka, H. and Kizu, M., 222n Tateishi, K., 19, 20, 31, 210n, 211n Tellier, C., 154, 155, 156 topic contrastive, 12, 13, 14, 15, 16–18, 23–4, 174, 210n gapless construction, 30–4 thematic, 12, 13, 16–18, 24, 210n tough-construction, 222n Toyama dialect, 65 transitivity restriction, 78, 85, 87 Tsubomoto, A., 70, 215n Turkish, 168 Ueda, M., 95, 219n Ueyama, A., 230n Valois, D. and Dupuis, F., 85 variable, bound, 3, 6, 22, 103, 113, 114, 120, 164, 214n verb movement raising, 209n, 227n string-vacuous, 48 Wasow, T., 227n Watanabe, A., 58, 63, 72, 77, 78, 81, 83, 84, 85, 86, 87, 88, 89, 98, 99, 122, 124, 152, 153, 154, 156, 194, 216n, 217n, 218n, 223n WCO (Weak Crossover), 8, 14, 15, 101, 120–2, 124–6, 130, 139, 140, 145, 150, 157, 186–9, 197, 203, 206 weakest crossover, 125 Wh agreement, 228n extraction, 85, 154 movement, 5, 79, 85, 87, 143, 145, 154, 160, 161, 165, 166, 170, 172, 174, 175, 203, 209n, 217n, 219n, 228n, 229n multiple, 183–6, 230n operator, see operator second wh-construction, 144 Williams, E., 2, 36, 38, 39, 63, 141, 160, 208n, 227n