University of Southern California Dissertations and Theses

On the nature of particles in Japanese and its theoretical implications

(USC Thesis Other)

On the nature of particles in Japanese and its theoretical implications

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely afiT ect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI A Bell & Howell Infoimation Com pany 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761^700 800/521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ON TH E NATURE O F PARTICLES IN JA P A N E S E AND ITS TH EO R ETIC A L IM PLICA TIO N S by Hiroshi Aoyagi A Dissertation Presented to the Faculty of the Graduate School University of Southern California In P artial Fulfillment of the Requirements for the Degree Doctor of Philosophy (Linguistics) August, 1998 © 1998 Hiroshi Aoyagi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 9919009 UMI Microform 9919009 Copyright 1999, by UMI Company. Ail rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 90007 This dissertation, written by Hiroshi Aoyagi under the direction of ..... Dissertation Committee, and approved by aü its members, has been presented to and accepted by The Graduate School, in partial fulfillm ent of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean cf Graduate Studies Date 10, 1998 DISSERTATION COMMITTEE ..... Chairperson C Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. u ACKNOW LEDGM ENTS It is hard to believe th at more time has already passed since I came back to Japan than those three years that I spent in Los Angeles. Upon finishing this thesis, I would like to th ank many people who 1 have met on either side of the Padfic. First of all, my deepest gratitude goes to Hajime Hoji, my thesis supervisor, for his encouragement and support over the years. Hajime has always been patient enough to listen to my “half-cooked” arguments, and his criticism has never failed to help put them into better shape. 1 would also like to thank the other members of my committee, Joseph Aoun, Nam-kU. Kim, Yen-hui Audrey Li, and Jean-Roger Vergnaud, for their time and support. Joseph’ s healthy skepticism has always made me think more deeply. Nam-kU. was my first Korean teacher, and 1 am grateful to him for his willingness to discuss Korean data with me; kansahapnita! 1 have not only learned much about Chinese syntax from Audrey, but she has also been a very good role model for us, graduate students firom Asia: xiexie nin! 1 would hke to thank Jean-Roger for his generosity in letting me share his genius. 1 have always enjoyed an appointment with him at Michel Richard. My thanks are extended to other linguists at USC and in the Great Los Angeles area. 1 am indebted to Bernard Comrie, Alicja Gorecka, Barry Schein, and M aria-Luisa Zubizarreta for whatever I know about typology, phonology, semantics, and Romance syntax, respectively. 1 would also Hke to thank Lisa Cheng, Naoki Fukui, Jim Huang, Kyle Johnson, Hilda Koopman, and Tim Stowell for their intellectual generosity. The departm ent of linguistics at USC was a much nicer place to be in (and really a multi-ethnic society!) owing to my fellow graduate students firom all over the world. Those are: Pablo Albizu, Joe Allen, Nancy Antrim, Jose Camacho, Dong-in Cho, Lina Choueiri, Dae-ho Chung, Abdesslam Elomari, Gorka Elordieta, Elena Herburger, Roland Hinterholzl, Kaoru Horie, Miao- lingHsieh, Susan Kalt, Charles Kim, Hae-won Kim, IbÜssam Kortobi, Hui-ju Grace Li, Yuki M atsuda, Karine Megerdoomian, Keiko Miyagawa, Hiro Oshita, Hong-keunPark, Liliana Sanchez, Patricia Schneider, Shu-ingIngrid Shyu, Linda Taylor, Ayumi Ueyama, Maki W atanabe, Shin Watanabe, and Xiu-zhi Zoe Wu. Above all, Maki-chan deserves specdal mention for everything she has done for me since 1 left LA. 1 would hke to thank them all for their friendship, love, and support. The time th at 1 spent with them in classrooms, on the beach, and in restaurants, 1 will never forget. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill A lot of thanks also go to the staff stationed at GFS #301: above all, Linda Culver, Kathy Stubaus, and one and only Laura Reiter. Without their assistance, my life at SC would have been much more difficult. My three years’ stay in the U.S. firom 1991 through 1994 was made possible by the Young Faculty Studies Abroad Program at Nanzan University. I would hke to express my sincere gratitude to N anzan University and the Graduate School of H um anities at USC for their financial support. Since I came back home, I have been intellectually indebted to a local but rapidly expanding community of linguists in Nagoya and its vicinity. My thanks go to, among others, Jun Abe, YasuaJd Abe, Kimiya Adachi, M asatake Arimoto, Setsuko ^Wta, Hiroko Azuma, Tadashi Eguchi, Hiroto Hoshi (now in London), Asako Miyachi, Yosuke Momiyama, Keiko Murasugi, Satoko Osawa, Mamoru Saito, M asaki Sano (in Kyoto), Ken’ ichiro Shirai, Tatsuya Suzuki, Masako Tsuzuki, and Mihoko Zushi. Among my teachers in Japan, I am especially grateful to Kazuko Inoue and M asatake M uraki for their incessant encouragement and support over the long, long years. All this started when they taught me at ICU. Among all people who have helped me over the years, there are two who deserve special mention: Hajime Lkawa and Torn Ishii. Hajime and Tom are not only my long-time pals but they also joined me for doctoral studies in Southern California. W ithout their fiiendship, support, and intellectual generosity, my life would have been almost unbearable. I would hke to extend my thanks to their wives, Miyoko Lkawa and Tomoe Ishii, for their fiiendship. Finally and most of all, I would hke to thank my parents, Hajime and Natsuyo Aoyagi, for their never-failing love and support. They have let me do whatever I have wanted to. This thesis is dedicated to them. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IV TABLE OF C O N TEN TS ACKNOWLEDGMENTS ü ABSTRACT OF THE DISSERTATION vi CHAPTER I Introduction 1 CHAPTER 2 Adjunct Clitic Hypothesis 7 2.1. A Paradox 8 2.1.1. An argum ent for headship: Su-support 10 2.1.2 An argum ent against headship: Categorial transparency 15 2.2. Q-particles as Adjunct Clitics 20 2.2.1. Clitics in the bare phrase structure theory 21 2.2.2. Clitics vs. inflectional affixes 29 2.3. Inflectional Morphology and Functional Categories in Japanese 39 2.3.1. Inflectional morphology 39 2.3.2. Functional categories 44 2.4. Deriving the Paradoxical Properties 52 2.4.1. Su-support as a morphological process 52 2.4.2. Selection 56 CHAPTER 3 Q-particles and Case Particles 60 3.1. Kakari-joshi (K-particles) vs. Fuku-joshi (F-particles) 61 3.2. Ordering Restrictions 64 3.2.1. Order between Q-particles and case particles 65 3.2.2. Order between K- and F-particles 69 3.2.3. The nominal nature of F-particles 70 3.3. Case Particles 74 3.3.1. Morphological case theory 74 3.3.2. Case marking in Japanese 90 3.3.3. Q-particles and D 109 3.3.4. Postpositions 120 3.3.5. Summ ary 128 3.4. “Case M arker Drop” 131 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 4 Association of Q-particles with Focus 140 4.1. Association with Focus 140 4.2. Licensing Functional Heads for K- and F-p articles 151 4.3. Conditions on Association with Wide Focus 168 4.3.1. [+Focus] assignment and propagation 168 4.3.2. An argument-adjunct asymmetry 175 4.3.3. Some further predictions 181 CHAPTER 5 An Implication to Word Order 188 5.1. Head param eter 189 5.2. Previous Analyses 197 5.2.1. Kayne(1994) 197 5.2.2. Takano (1996) and Fukui and Takano (1998) 201 5.3. An alternative 216 5.3.1. Deriving the strict head-finality of Japanese 216 5.3.2. “Mixed” orders in Chinese 221 APPENDIX I On Sells' (1995) Lexical Analysis of Inflection 228 APPENDIX E On Koizumi’ s (1995) Arguments for Overt V-raising 243 APPENDIX IE On the Irregularity of Adjectival Inflection 268 REFERENCES 274 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. V I ABSTRACT OF THE D ISSERTA TIO N On the N ature of Particles in Japanese and Its Theoretical Implications by Hiroshi Aoyagi Doctor of Philosophy (Linguistics) University of Southern California, Los Angeles, 1998 Professor Hajime Hoji, Thesis Supervisor This dissertation proposes new analyses of the two major classes of particles in Japanese: (quasi-)quantificational particles (Q-particles) and case particles. While Q-particles are morphosyntactically characterized as adjunct clitics, case particles serve as morphological spell-outs of CASE features in D. Q-particles are clitic-like elements, in the sense that they do not project (hence, being X° and simultaneously) and they need a phonological host to lean on. However, unlike pronominal clitics in Romance and other languages, they are adjuncts in their own right. The proposed characterization of Q- particles as adjunct clitics naturally accounts for the clustering morphosyntactic properties of them, including their wide freedom in distribution. On the assumption that case particles in Japanese are realization of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. vu morphological case th a t is assigned in reference to thematic positions, rather than th at of abstract Case, facts of the possible case arrays in the language are straightforwardly accounted for. Moreover, the proposed analysis of Q- particles, together with the morphological case theory adopted here, provides a principled account of the restricted orders among different types of particles. As focus inducers, Q-particles are associated with a target of focus, providing some additional implication to the sentence in which they appear. This association with focus is generally exercised under c-command. However, in certain contexts, Q-particles can be associated with a constituent that they do not c-command at Spell-Out. Assuming th at Q-particles are licensed by adjoining to a fimctional head at LF, this intriguing behavior of Q- particles in association with focus will be explained. Furthermore, this thesis presents a salient view of functional categories in Japanese, of which Q-particles constitute a part. All functional elements in Japanese are suffixal. This characterization of functional categories in the language gives a far-reaching consequence: i.e. the strict head-finality of languages like Japanese is the result of this param etric rendering of its repertoire of functional categories, rather than the conventional use of the head param eter. A careful comparison with the existing proposals reveals that the word order in a given language should better be captured by the morphological nature of functional categories in th at language. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C H A PTER 1 In tro d u c tio n As is weU known, Japanese is an agglutinative language which is replete with so-called particles (joshî) as well as inflectional affixes (katsuyoo-gobi). Due to its strictly head-final nature, while inflectional affixes are suffixal, particles always appear on the right of their hosts. Among different classes of particles, the one th at is most often discussed in the linguistic literature (e.g. Kuno, 1973; Kuroda, 1978, 1988; Saito, 1982; Takezawa, 1987; inter alias) is the class of case particles, as exemplified in the following. (1) a. John-ga ki-ta -NOM come-PAST ‘ John came.’ b. John-ga LGB-o yom-ta -NOM -ACC read-PAST John read LGB.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. John-ga Mary-ni hanataba-o age-ta -NOM -DAT bouquet-ACC give-PAST ‘ John gave a bouquet to Mary.’ Since the notion of case is inherent to nominality across languages, case particles in Japanese attach only to noun phrases or their equivalents. The three principal case particles found in the examples in (1) are the nominative ga, the accusative o and the dative ni. With some oft-mentioned exceptional instances, case particles basically represent a grammatical relation; ga marks subjects, o direct objects and ni indirect objects. ^ - To the extent th at ^ One exceptional use of the nominative ga that has often been discussed in the literature is its use as the m arker on the theme argument of a certain class of predicates that are seemingly transitive, as shown in the following. (i) a. John-ni eigo-ga /*o deki-ru -DAT English-NOM/ACC can.do-PRES "John has a good command of English.’ b. John-ga eigo-ga /(?)o deki-ru -NOM English-NOM/ACC can.do-PRES Kuno (1973) refers to the use of ga with the theme argument in (i) as object ga-marking. What is interesting in the examples in (i) is the possibility of case alternation. When the other noun phrase is marked with DAT as in (ia), ga does not alternate with o. On the other hand, when the other noun phrase is also m arked with ga as in (ib), it becomes easier to change ga on the theme into o. In chapter 3, we will discuss the nature of case particles in general from a particular perspective that takes them for morphological spell-out of a specific feature. 2 Ga is sometimes referred to as a focus marker, in opposition to wa as a topic marker (cf. Diesing, 1989). This rendition is based on the observation that nominative subjects with a class of stative predicates (or individual-level predicates, in Carlson’ s (1977) and Kratzer’ s (1989) terminology) are obligatorily focused in the root context. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 f-.his statem ent is correct, case particles only make a minimum contribution to the semantics of sentences. Japanese has another class of particles th at may cooccur with, or even replace, case particles. (2) a. John-mo ki-ta -also come-PAST ‘ JOHN also came.’ b. John-dake-ga LGB-o-s a e yom-ta -only-NOM -ACC-even read-PAST ‘ Only JOHN read even LGB.’ (i) Mary-ga kirei da -NOM beautiful COP ‘ MARY is beautiful.’ or ‘ It is Mary who is beautiful.' This is so-called ga as the m arker of exhaustive listing. However, as noted by Kuno (1973), the obligatoriness of focus reading for the nominative disappears if it is embedded in a subordinate clause. (ii) mosi Mary-ga kirei dat-tara ... if -NOM beautiful COP-subjunctive affix "if Mary is beautiful,...’ The sentence in (i) is embedded in the antecedent of a conditional in (ii), and the nominative phrase need not be interpreted as being focused. Furthermore, in multiple nominative constructions, only the first nominative phrase has to be interpreted as focus. (iii) John-ga imooto-ga kirei da -NOM sister-NOM beautiful COP ‘ It is John whose sister is beautiful.’ In (iii), only John, but not imooto ‘ sister’, m ust be construed as being focused. These facts and others indicate th at ga is not an inherent focus marker. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. John-ga M ary-ni-wa hanataba-o age-ta -NOM -DAT-TOP bouquet-ACC give-PAST ‘ John gave a bouquet to MARY (but not anybody else).’ Compare each of the sentences in (2) with its counterpart in (1). While dake only’ , sae even’ in (2b) and wa TOP* in (2c) cooccur with ga, o and ni, respectively, mo also’ in (2a) replaces ga in (la). Although each sentence in (2) shares its core meaning with its counterpart in (1), something is added to the semantics of the former, as indicated by the gloss. This sem antic addition is obviously the contribution of particles hke dake, mo, sae and wa. Borrowing from Kuroda’ s (1970) terminology, we will refer to these particles as (quasi-)quantificational particles, henceforth, Q-particles. Like focus adverbs in English, Q-particles are focus indicators, in the sense th at they are associated with a target of focus, such that each of them renders a particular impHcature th at is absent in the core proposition without it. For instance, (2a) produces an existential implicature that someone other than John came, as well as it expresses a core assertion th at John came. One major characteristic of Q-particles, in comparison to case particles, is th at they have greater freedom in distribution. While case particles are exclusively for noun phrases, Q-particles can attach not only to noun phrases, as seen in (2), but also to other categories, as exhibited in the following. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 (3) a. daihyoosya -wa [ppNew Y ork-kara]-dake/m o/sae ki-ta representatives-TOP -firom -only/also/even come-PAST ‘ Representatives came only/also/even firom New York.’ b. Mary-wa [vphon-o yom i]-dake/m o/sae si-ta -TOP book-ACC read -only/also/even do-PAST “ Mary only/also/even read books.’ c. Bill-wa [Apbuka-o isogasi-ku]-dake/mo/sae si-ta -TOP his men-ACC busy -only/also/even make-PAST ‘ Bill only/also/even made his men busy.’ d. John-wa [cpMary-ga sin-ta to ]-d ak e/m o /sae iw-ta -TOP -NOM die-PAST that-only/also/even say-PAST ‘ John only/also/even said that Mary was dead.’ The examples in (3a-d) show that Q-particles may attach to PP, VP, AP and CP as well as noun phrases. This is the reason why Q-particles in Japanese (and their equivalents in Korean) are often called “crosscategorial” particles (cf. Sells, 1995; Cho and Sells, 1995; and Yoon, 1995). The purpose of this thesis is to elucidate the grammatical nature of Q- particles. Chapters 2 and 3 discuss their morphosyntactic properties. Chapter 2 lays out the groundwork on which our later discussions will depend. In particular, we are presenting a hypothesis th a t Q-particles are adjunct clitics. From this, a lot of their properties, including the observed Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 “carosscategorial” nature, will follow. Chapter 3 will further look into the morphological nature of Q-particles and estabhsh a distinction between the two subclasses, suggested by some traditional Japanese grammarians. With a specific perspective of morphological case, we will also try to account for the observed restrictions on the order among the two subclasses of Q- particles and case particles. In chapter 4, we will consider the properties of Q-particles as focus indicators. Their intriguing behavior in association with focus will be derived firom their morphological nature, together with a minimal set of assumptions about focus. Finally, in chapter 5, we will discuss an imphcation about word order that our specific conception of functional categories renders. It will be suggested th at some particular morphological features on functional heads can alter the unmarked word order, which is established by the LCA that Kayne (1994) proposes. Throughout this thesis, we will presuppose certain familiarity on the part of the reader with recent tenets of generative grammar presented, for instance, in Chomsky (1991, 1993, 1994, 1995) and related works. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CH A PTER 2 A d ju n ct C litic H y p o th e sis This chapter and the next are mainly concerned with morphological aspects of Q-particles. The present chapter particularly focuses on a paradox concerning the morphosyntactic status of Q-particles. In section 2.1., we will first juxtapose argum ents for and against the view th a t they are heads of projections. While Q-particles behave hke heads in terms of inflectional morphology, they do not with respect to categorial selection. Ih section 2.2., drawing on Chomsky’ s (1994) analysis of Romance pronominal chtics, we will present a hypothesis that Q-particles are chtics, in the sense that they are heads th at do not project and they need a PF host to lean on. However, unlike pronominal chtics in Romance, Q-particles are obviously non arguments, and therefore, they are better termed adjunct chtics. Based on the plausible assumptions presented in section 2.3. about inflectional morphology and functional categories in Japanese, section 2.4. will show how the proposed hypothesis solves the noted paradox as well as other clustering properties of Q-particles. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 2.1. A Paradox As noted in the literature (Okutsu et al., 1986; Cho and Sells, 1995; Sells, 1995; Yoon, 1995; among others), Q-particles are “crosscategorial”, in the sense th at they can attach to a wide variety of categories, as exhibited in our earher examples in chapter 1, reproduced here as (la-e). (1) a. John-ga [opsusil-dakeAno/sae tabe-ta (DP) -NOM sushi -only/also/even eat-PAST ‘ John ate only/also/even sushi.’ b. daihyoosya-ga [ppNew York-kara]-dake/mo/sae ki-ta (PP) representatives-NOM -firom -only/also/even come-PAST ‘ Representatives came only/also/even firom New York.’ c. Mary-ga [vphon-o yom i]-dake/m o/sae si-ta (VP) -NOM book-ACC read -only/also/even do-PAST M ary only/also/even read books.’ d. BiU-ga [Apbuka-o isogasi-ku]-dakeAno/sae si-ta (AP) -NOM his men-ACC busy -only/also/even make-PAST ‘ Bill only/also/even made his m en busy.’ e. John-ga [cpMary-ga sin-ta to]-dake/m o/sae iw -ta (CP) -NOM -NOM die-PAST that-only/also/even say-PAST John only/also/even said th at M ary was dead.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 The examples in (la-e) show the “crosscategorial” nature of Q-particles, in the sense that they may attach to a variety of categories, including DP, PP, VP, AP and CP. A question naturally arises as to the syntactic status of Q- particles. There are a priori two alternative views as to merger of Q- particles. On the one hand, a Q-particle may be the target of merger and it projects, as exhibited in (2a); on the other, a Q-partide may be merged with a target and the target projects, as shown in (2b). (2) a. Q . b. Although the syntactic status of Q-particles has never been seriously considered in the literature, Aoyagi (1994, 1996) and M aki (1995) stand with (2a), assuming th at Q-particles are heads th at project; in this approach, the noted crosscategorial nature of Q-particles must be attributed to their ability to take any category as their complements. On the contrary. Sells (1995) argues for left-headedness of morphology in head-final languages hke Japanese and Korean, whereby a Q-particle only counts as a suffix attached to a root th at projects. For instance, the verbal complex in (Ic), hon-oyomi- mo si-ta %ook-ACC read-also do-PAST, will be represented in these approaches as (3a) and (3b), respectively. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 10 (3) a. y, IP si+ta mo N' S: hon-o 3 Suf2 9 Sufi Yi hon-o yomi yomi mo In (3a), mo ‘ also’ , designated as X, is a head th at projects a maximal projection XP, taking VP as its complement. In (3b), however, mo counts only as a suffix to the verb stem yom(i) ‘ read’ . Since, for Sells, morphology in Japanese is left-headed, the whole verbal complex V4 is headed by this verb stem V t. In w hat follows, we will take a look at arguments for these analyses in turn. 2 . 1. 1. An argument for headship; Su-support As noted by Kuroda (1965: chap. 3), when tense (or Aux in Kuroda’ s terms) is stranded from the verb, Japanese employs su-support, an analogue of do- support in English, in cases Hke (Ic), reproduced here as (4a). (4) a. Mary-wa hon-o yom i-dake/m o/sae si-ta -TOP book-ACC read-only/also/even do-PAST ‘ Mary only/also/even read books.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 11 b. *Mary-wa hon-o yom i-dake/m o/sae -ta -TOP book-ACC read-only/also/even -PAST(stranded) c. Mary-wa hon(-o)-dakeAno/sae yom-ta -TOP book-ACC-only/also/even read-PAST M ary read only/also/even books.’ In (4a), when a Q-particle appears VP-finally, the tense morpheme (i.e. the past tense ta) is isolated from the verb stem yom ‘ read’; in such a case, tense m ust be supported by su ‘ do’ (realized as si). W ithout support from su, the stranded tense would not survive, as indicated by the ungrammaticality of (4b). If a Q-particle attaches to other elements in the sentence, say, an argument DP as in (4c), adjacency between the verb stem and tense is maintained, and thus, tense is saved from being stranded. This state of affairs is quite similar to that which invokes do-support in English. One of the circumstances that force do-support in Enghsh is negation. Consider the following paradigm. (5) a. Mary read books b. *Mary not read books c. Mary didn't read books Negation in English and other languages has been a serious topic of discussion among syntactidans for the past decade. Despite several competing proposals for the exact locus of the negative morpheme not in. English, it Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 seems to be rath er established that negation creates its own phrasal projection, namely, Negation Phrase (NegP) (c£ Pollock (1989), Laka, (1990), Zanuttini (1990), Chomsky (1991), among others). * Let us assume with Pollock (1989) and Chomsky (1991) th at n o tin English occupies a head position ^eg°) th a t projects NegP, which intervenes between T and V. Let us also assume, in the spirit of these authors, th at in English, T lowers to V, rather than V raises to T, at S-structure, and subsequently at LF, V+T raises back to T to erase the potentially offensive trace created by T-lowering.- In this view, the inflected V in (5a) is formed as the result of lowering of T to V. 1 Although the text discussion assumes with Pollock (1989) and Chomsky (1991) that not is the head of NegP, there are several different proposals for the location of negation in English, Drawing on the “inner island” effect as in (i), Rizzi (1990) claims th at negation acts as an intervening A’ -specifier (i.e. Spec of NegP). (i) *how stronglyz do you not beheve [that the inflation will rebound tz]? (i) does not allow the embedded construal of the adjunct wh-phrase. According to Rizzi, this fact can be accounted for by the assumption th at not as Spec of NegP blocks antecedent government of the intermediate trace in the embedded Spec of CP by the fronted adjunct wh-phrase, a violation of his relativized minimality condition with respect to A -movement. On the other hand, Ouhalla (1990) proposes that while not is accommodated in the head of NegP, an empty Neg operator is base-generated in Spec of NegP. Roberts (1993) instead claims th at a lth o u ^ not is base-generated in Spec of NegP, it diticizes to Neg° and subsequently raises to T° Oience, its own trace in Spec of NegP is antecedent-governed), which is phonetically realized as n't. However, inasmuch as it is sound to assume th at the head of NegP, whether filled with not or not, intervenes between T and V, the choice among these alternatives is not our imm ediate concern here. 2 Pollock (1989) and Chomsky (1991) assum e th at AGR, which heads its own projection, intervenes between T and V. However, the text discussion is not contingent upon the existence of AGR, and thus, we discard it for a simplicity Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 13 If negation intervenes between T and V, however, LF raising of V+T is blocked by the head movement constraint (HMC) in (6). (6) The Head Movement Constraint (HMC)^ A head a can only move to another head p if and only if p properly governs the maximal projection th at a heads. Due to the HMC in (6), (5b) is ruled out.^ In such a case, the dummy verb do is inserted under T, as in (5c). Let us now compare the context of do-support in (5c) with th a t of su- support in (4a). These contexts are schematized in (7) and (8), respectively. reason. 3 Travis’ (1984: 131) original statem ent of the HMC is as follows. (i) The Head Movement Constraint An X° may only move into the Y° th at properly governs it. Formally speaking, the notion of proper government should be successively defined, as it is in Chomsky (1986a). However, for the present purposes, it seems good enough to understand th at a is properly governed by p iff a is the complement of, or the specifier of the complement of, p. For arguments for the possibility of incorporation from specifiers, see Baker (1988), Hale (1990) and Hale and Keyser (1993), among others. ^ The text discussion presupposes the last resort nature of syntactic operations and the principle of economy, as conceived of by Chomsky (1991), to the effect th at no operation is allowed unless it is motivated as a last resort. Since raising of V to T in English is prohibited for an independent (6- theoretic) reason, lowering of T to V is instead motivated, so th a t T will satisfy a morphological principle (cf. Baker’ s (1988) Stray Affix Filter). Furthermore, as noted by Pollock (1989), raising of V+T back to T is motivated as well, since T, as a sentential operator, m ust be interpreted under T at LF. However, neither lowering of T to Neg (followed by lowering of Neg+T to V) nor raising of V to Neg (again, followed by raising of Neg+V to T) is motivated; hence, prohibited. Note also that if lowering of T to V across Neg is allowed, raising of V+T back to T, skipping Neg, is ruled out by the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 (7) y. do eg y d; read books books Do-Support read (8) a. y IP yp 1 Î si+ta Su-Support EjP y mo hon-o yomi *--------------- hon-o yomi Because of the intervention of Neg between T and V, raising of V+T back to T at LF in (7a) wül be blocked; hence, do is inserted under T to save it 6om being stranded, as indicated in (7b). This explanation carries over to su- support in Japanese, if we assume that a Q-particle like mo ‘ also’ counts as an intervening head X which projects XP, in the same sense th at n o tin English does project NegP. As illustrated in (8a), raising of V to T is arrested by mo, and therefore, as a last resort, su is inserted to support T, as indicated in (8b). In return, this constitutes an argument that, hke not in Enghsh, Q-particles in Japanese are heads of projections that induce an intervention effect with respect to the HMC in (6). HMC in (6). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 2.1.2 An argum ent against headship: Categorial transparency Despite th e argument for headship of Q-particles above, Sells (1995) argues against such a view. The syntactic evidence th at he provides against the claim th at Q-particles are heads of projections is th at of selection. Some relevant examples are in the following. (9) a. Johng-wa [eg susi-o tabe]-te mi-ta -TOP sushi-ACC eat-ASP see-PAST ‘ John tried eating sushi.’ b. Johng-wa [eg susi-o tabe]-te-m o m i-ta -TOP sushi-ACC eat-ASP-also see-PAST ‘ John tried also eating sushi.’ c. *John-wa [eg susi-o tabe]-ni-mo m i-ta -TOP sushi-ACC eat-ASP-also see-PAST (10) a. Maryg-wa Tuldzi-e [eg susi-o kuruma-de tabe]-ni ik-ta -TOP Tsukiji-to sushi-ACC car-by eat-ASP go-PAST ‘ Mary went to Tsukiji by car to eat sushi.’ b. Maryg-wa Tukizi-e [eg susi-o kuruma-de tabel-ni-dake ik-ta -TOP Tsukiji-to sushi-ACC car-by eat-ASP-only go-PAST M ary went to Tsukiji by car only to eat sushi.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 c. *Mary2-wa Tukizi-e [eg susi-o kuruma-de tabe-te]-dake ik-ta -TOP Tsukiji-to sushi-ACC car-by eat-ASP-only go-PAST (11) a. John-wa [Mary-ga susi-o tabe-ru]-to omow-ta -TOP -NOM sushi-ACC eat-PRES-that think-PAST ‘ John thought th at Mary would eat sushi.’ h. John-wa [Mary-ga susi-o tahe-ru-to]-sae omow-ta -TOP -NOM sushi-ACC eat-PRES-that-even think-PAST ‘ (Lit.) John thought even th at Mary would eat sushi.’ c. * John-wa [Mary-ga susi-o tahe-rul -ka-s a e omow-ta -TOP -NOM sushi-ACC eat-PRES-Ques-even think-PAST (12) a. John-wa [Mary-ga susi-o tahe-ta]-ka siri-ta-gar-te-i-ru -TOP -NOM sushi-ACC eat-PAST-Ques know-want-PROG John wants to know if Mary ate sushi.’ h. John-wa [Mary-ga susi-o tahe-tal-ka-w a siri-ta-gar-te-i-ru -TOP -NOM sushi-ACC eat-PAST-Ques-TOP know-want-PROG ‘ John wants to know if Mary ate sushi (hut not anything else).’ d. *John-wa [Mary-ga susi-o tahe-tal-to-w a siri-ta-gar-te-i-ru -TOP -NOM sushi-ACC eat-PAST-that-TOP know-want-PROG It is generally considered th at a selectional relation holds under sisterhood; i.e., a head and a maximal category that it selects for m ust he canonical Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17 sisters.^ In each a-sentence of (9) - (12), categorial selection of the m atrix verb is straightforwardly satisfied. The verb m i ‘ see, try’, in its particular sense as in (9a), requires as its sister a category suffixed with te, a perfective aspectual marker.® As noted by Miyagawa (1987), verbs hke ik go’ and ku ‘ come,’ when functioning as “restructuring” verbs, take a purposive complement clause th at is sufhxed with n i an imperfective aspectual marker. ® “Selection” here means c(ategori.al)-selection, in the sense of Pesetsky (1982), who claims th at the canonical structural realization (CSR) of Grimshaw (1979, 1981) and Case theory will derive the c-selectional properties of a head firom its s(emantic)-selection, the former being redundant, hence, unnecessary. However, while P esetsk/s concern was restricted to lexical predicates th at assign 6-roles, we are concerned with a more general notion of the relation that a head, whether lexical or functional, bears with its complement, a combination of “selection” by a lexical head and “extended projections” formed by a functional head and its complement, in the sense of Grimshaw (1991). ® We will assume th at te and ni are aspectual morphemes that indicate perfectivity and imperfectivity, respectively. A case in point is given in the following. (i) a. Maryz-wa [ez gohan-o tabe]-ni uti-e kaet-ta -TOP meal-ACC eat-NI home-to retum-PAST "Mary went home to eat a meal.’ b. Maryz-wa [ez gohan-o tabe]-te uti-e kaet-ta -TOP meal-ACC eat-TE home-to retum-PAST ‘ Mary went home after eating a m eal.’ (Ü) ... ga, zissai gohan-wa tabe-na-katta but actually meal-TOP eat-NEG-PAST ... but actually, (she) did not eat a meal.’ (Ü) will m ake a felicitous discourse continuation for (ia), but not for (ib). This contrast indicates th at at the time of the matrix event (i.e. Mary’ s going home), while the event of Mary’ s eating a meal has not necessarily taken place in a ru-clause, it m ust have already taken place in a te-clause; hence, a logical inconsistency results, if (ib) is followed by (ü). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 18 as in (10a) J ® As shown in (11) and (12), verbs like omow ‘ think’ and ^ As noted by Miyagawa (1986), “restructuring” is blocked when the purposive clause is not a strict sister of the matrix verb. Compare (10a) with the following: (i) ?*Mary2-wa [eg sushi-o kuruma-de tabe]-m Tukizi-e ik-ta -TOP sushi-ACC car-by eat-ASP Tsukiji-to go-PAST In (10a), the purposive complement clause in ni appears in the sister position of the m atrix verb, and as the result of “restructuring,” kuruma-de ‘ by car,’ though appearing inside the complement clause, can modify the matrix verb. In (i), on the other hand, Tukizi-e to Tsukiji,’ a directional phrase, intervenes between the m atrix verb and the purposive complement; as a result, “restructuring” is prohibited, and the same adverbial, kuruma-de, in the complement clause cannot successfully modify the m atrix verb. Arguing against Miyagawa’ s drastic tree-changing analysis (i.e. structural reanalysis), Aoyagi (1991) claims th at “restructuring” should be taken as the result of head movement of the lower verb to the m atrix verb a t LF. If the purposive clause is in the matrix verb’ s complement position in (10a), but in an adjunct position in (i), the success/failure of the designated adverbial modification can be reducible to the HMC. For details of properties of “restructuring” predicates in Romance and Japanese, see Rizzi (1982) and Miyagawa (1986), respectively. ® The clausal category with te is not incompatible with the verbs ik ‘ go’ and ku ‘ come’ . In fact, the following two sentences are both fully acceptable. (ii) a. Maryz-wa [ez sushi-o tabe]-te ik-ta -TOP sushi-ACC eat-ASP go-PAST ‘ Mary left after eating sushi.’ b. Maryz-wa [ez sushi-o tabe]-te ki-ta -TOP sushi-ACC eat-ASP come-PAST M ary came after eating sushi.’ As indicated in the glosses, te renders perfective, rather than purposive, interpretation, and more importantly, it does not tr i^ e r “restructuring”. This indicates th a t the clausal category with te only counts as an adjunct in these contexts. Q-particles are, generally, either incompatible with adjuncts (e.g. yukkuh(to)- *mo/*sae/??dake ‘ slowly-also/even/only’ ,... node- *mo/*sae/*dake ... because-also/even/onlyO or, if compatible at all, not allowed association with wide focus when they are attached to adjuncts (see the relevant discussions in chapter 4). Te does not seem to be quite harmonious with dake ‘ only’ in (10c), and if the sentence is interpretable at all, the Q-particle will not be allowed wide scope association with focus (which is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19 (complex) verbs Kke siri-ta-gar ‘ want to know' require a complement clause headed by to th a t' and ka ‘ Ques(tion),’ respectively. If the complement clause in each a-example of (9) - (12) is headed by the designated morpheme Y, its selectional relation with the matrix verb can be straightforwardly captured, as represented in the following. (13) y ZP m i ‘ see’ Ak go'/omow ‘ th in k '/siri-ta-g ar wonder' te ‘ PERF'/ni ‘ IMPERF'/to ‘ th at'/k a ‘ Ques' In (13), each matrix verb selects for a maximal projection headed by the designated morpheme Y under sisterhood. However, as highlighted by Sells (1995), the complement in each a- example in (9) - (12) can be further suffixed with a Q-particle, as demonstrated in each b-example. Sells claims th a t if each Q-particle in cases like b-examples of (9) - (12) heads a projection XP, the selectional relation between the verb and its sentential complement, captured as a sister relation in (13), cannot be maintained, simply because XP intervenes and breaks the sisterhood between V and YP, as shown in (14). possible in (10b)), nor is the intended matrix construal of the adjunct kuruma- de by car’ perm itted. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 (14) V ik/om ow / s iri-ta -g a r X ‘ alsoV d a k e ‘ only’ / sa e ‘ even’ / w a T O P Y ZP te / n i/ to /k a One might be tem pted to argue th at in a- and b-examples in (9) - (12), what the matrix verb selects for is the projection of a Q-particle, i.e. XP in (14), with accidental “zero” occurrence of Q-particle in a-examples. However, also as noted by Sells, th is argument goes wrong with cases like c-examples in (9) - (12). Note th at b- and c-examples in each set of examples constitute a minimal pair, differing only in the morpheme th at attaches to the clausal complement. While all b-examples are perfectly acceptable, none of c- examples is. This clear-cut contrast between b- and c-examples indicates that the m atrix verb in each case selects for the maximal projection of the designated morpheme that immediately attaches to the complement clause, i.e. YP in (13) and (14). Thus, cases like b-examples in (9) - (12), as pointed out correctly by Sells (1995), pose a serious problem to the contention that Q- particles are heads th at project their own categories. 2.2. Q-particles as Adjunct Clitics In the previous section, we have seen arguments for and against the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 1 headship of Q-particles. While facts of su-support (section 2.1.1.) seem to support the view that Q-particles are heads of projections, selectional facts (section 2.1.2.) argue against th at view.s In an attem pt to solve this paradox, we will propose that Q-particles should be characterized as adjunct ditics, in the sense to be clarified shortly. 2.2.1. Clitics in the bare phrase structure theory It used to be widely assumed that, due to X’ -theory, the existence of a zero-level category in syntax entails that of a maximal, or double bar-level, projection. This is because zero-level categories, or heads, were assumed to project in the way th at they all satisfy the following schemata (firom Chomsky, 1986a:3). (15) a. X = X X”* b. X’ = X”* X * (where order param etrized, X ”* stands for zero or more occurrences of some maximal projection, and X = X ^ > 3 Arguing against the view that Q-particles are heads th at project. Sells (1995) proposes a lexical analysis of complex verbs in Japanese and Korean, whereby verb stems, together with intervening Q-particles and other affixes, form left-headed morphological complexes in the lexicon. Although we agree with Sells in th at syntactic movement, or V-raising above all, is not an appropriate mechanism to account for inflected verbs, his lexical analysis poses both empirical and conceptual problems. For a detailed discussion of such problems, see Appendix I. Also, for problems inherent to tem plate Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 Since a transitive verb like witness, for instance, selects for a DP (=D”) as its complement, the DP complement in (16b) must have as complex an internal structure as th at in (16a) in relevant respects, in conformity to the 5 C - schem ata in (15). (16) a. they witnessed [d’ the Germans’ destruction of the city] b. they witnessed [e t it] The structures of the verb combined with the object DP in (16a, b) are represented in (17a, b), respectively. (17) a. witnessed r the Germansa r b. P” 0 witnessed it r r destruction (of) the city According to the DP-hypothesis proposed by authors hke Abney (1987) and Fukui and Speas (1986), the object DP in (16a) is headed by the genitive ‘ s in D th at takes an NP as its complement and agrees with a DP in its Spec, as exhibited in (17a). Assuming with Postal (1969) that pronouns are Ds, we were forced to represent the pronominal object in (16b) as in (17b), where D° morphology th at SeUs assumes, see Yoon (1995). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 23 projects its maximal projection DP (=D”) w ithout taking a complement or a specifier. In other words, a representation hke the following was impermissible for (16b) under X’ -theory. (18) y o j ) o witnessed it Because in (18), the verb takes as its complement a zero-level category, without being projected, the schema in (15a) will not be satisfied. Thus, the projection of D° to D" in (17b), though seemingly superfluous, was inevitable, since the higher verb m ust take as its complement a m axim al projection, but not a zero-level or interm ediate, single bar-level, category, to satisfy the schema in (15a). However, the so-called bare phrase structure theory (BPST) proposed by Chomsky (1994, 1995) has brought about a drastic change in the view of phrase structure building in language, reducing the core properties a£X- theory to almost naught. Under the BPST, phrase structures are formed by a recursive operation called Merge, by which two syntactic objects are combined into a larger one. Now let a and p be syntactic objects and let Merge combine them to form another syntactic object y. In the formation of y, either a or p projects, y is defined set-theoretically as y = (6, {a, p}}, with Ô , the label (i.e. head) of y, being either a or p, and a and p being constituents of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 24 Y - Suppose th a t a projects, then Merge derives y = {a, {a, p}}. This can be represented rather informally as in the following. (19) Under this view of phrase structure building, projections are formed only as the result of Merge. Hence, there can be no non-branching projections, as in (17b). Furthermore, the “size” of a projection is determined only relatively. A syntactic object is minimal if it is a lexical item drawn from the lexicon, whereas it is m aximal if it does not project. For instance, p in (19), no m atter how simplex (or complex) it may be, is a maximal projection, simply because it does not project. Since phrase structure building in the BPST should anyway sta rt out with merger of lexical items, p as well as a in (19) can be a lexical item (i.e. minimal projection). This amounts to saying that a syntactic object can turn out to be both maximal and minimal at the same time. Consequently, (18) is a completely legitim ate representation for (16b), with a proviso th a t the pronominal object it is both minimal and maximal simultaneously (i.e. J> W n ia x ) lo Chomsky (1995, chap. 4: 245) restricts the term head to term inal elements, and makes a distinction between and X°; while the former is a term inal element (i.e. lexical item), the latter is either a head or a category formed by adjunction to the head X. To put it differently, X“ “ imphes X° b u t not vice versa. He further notes the possibility of a maximal zero-level projection, as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 25 Furthermore, as noted by Chomsky (1994: 15), th e BPST opens up a new array to account for the morphosyntacticaHy ambiguous status of clitics. In the following is an example of a pronominal clitic firom French. (20) a. M arie connaît le garçon knows the boy M ary knows the boy.’ b. Marie le connaît CL.3SM M ary knows him.’ c. *Marie connaît le What is obvious from the data in (20) is that unlike th e full DP object in (20a), a pronominal clitic in French m ust move to the left of the verb, as exhibited in the contrast between (20b) and (20c). We assume w ith Kayne (1991) that pronominal ditics in Romance are base generated in an argument position but must adjoin to T. It is obvious that pronominal clitics are ambiguous in terms of projection. On the one hand, they are X“« s because they receive a theta role in an argum ent position; on the other, they are X“ s because they adjoin to a head T°, and it is generally agreed that only heads can adjoin to heads. In addition, movement of a pronominal clitic exhibits a characteristic discussed in the text, referring to such an object as We will continue to use X P , rather ambiguously, to indicate either a term inal element or a complex Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 26 of movement of a maximal projection, since it does not obey the HMC; it necessarily crosses an intervening head V (and possibly, v, too). However, this ambiguous status of pronominal chtics does not bother us any more, since in the BPST, a pronominal chtic, as a D°, can also be a D™* if it is merged with a syntactic object (say, a verb) th at projects. Thus, the French pronominal clitic in (20b) is introduced to a derivation as in (2 la), and then adjoined to T° as in (2 Ib). b T )fmax •Ps connaît M arie y \ o po 02 connaît tz (In (21), the possibility of V-raising and other irrelevant details are suppressed.) In (2 la), a chtic le ‘ 3SM ’ , a D°, is merged with a verb connaît Tmow.3S’, and the latter projects. The chtic in (2 la) counts as both minimal and maximal simultaneously, since it is a lexical item (hence, minimal) and it does not project being embedded in a projection of V° (hence, maximal). As exhibited in (2 lb), after the derivation proceeds and T is introduced by Merge, head unless a distinction is necessary. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 27 the clitic adjoins to it, moving across intervenin g heads. This is possible, however, because, according to Chomsky, the chtic moves as a m aximal projection until the very last stage of adjunction to ^ Drawing on the BPST analysis of Romance pronominal chtics, we assume, in w hat follows, that those entities which (i) do not project and (ii) require a " One might argue, presumably correctly, th at V-raising to T via v, widely posited in French, will suppress the HMC effect in the case under discussion. However, this contention does not extend to cases of long chtic movement (or chtic climbing) noted, for example, by Rizzi (1982), as in the fohowing. (i) a. Piero verra a parlarti di parapsicologia wiU.come to speak.to you about parapsychology Piero will come to speak to you about parapsychology.’ b. Piero ti verra a parlare di parapsicologia In (ia), the second person pronominal chtic ti is adjoined to the embedded T (foUowed by adjunction of the infinitival parlar to T, according to Kayne, 1991). This is only an instance of short chtic movement like th at in (2 lb). In (ib), however, the same chtic moves to adjoin to the matrix T, necessarily crossing at least a few intervening heads, including the embedded verb and T. Hence, the argument for the suppression of the HMC effect in the text seems to be necessary anyway. 1 2 This adjunction operation ought to be characterized as Move, rather than Attract, in the sense of Chomsky (1995, chap. 4), simply because it is chtics, bu t not T, that have a drive for the operation. Pronominal chtics in Romance m ust adjoin to T, presumably due to their syntacticaUy defective character, defective to such a degree th at they cannot be interpreted on their own at LF. On the contrary, there seems to be no reason on the part of T th a t it has to attract chtics for its own sake. It is still unknown if aU cases of movement can be subsumed under Attract, as suggested by Chomsky (1995, chap. 4). Then, it m ight be the case th a t while [-interpretable] elements are Attracted, [+interpretable] elements, if not hcensed at LF, m ust undergo Move. Drawing on facts of focus movement, in contrast to wh-movement, in Serbo-Croatian, Boskovic (1997) claims that the computation C hl m ust allow Move, independently of Attract. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 28 phonological host to lean on can safely be called “clitics.” If this assumption is tenable, Q-particles in Japanese are qualified as clitics, since they are functional heads th at do not project and they require a phonological host to be attached to. However, unlike pronominal clitics in Romance, Q-particles are not arguments in their own right. Rather, they are more Hke adverbial elements, on a p ar with focus adverbs such as also, even and only in English. Thus, Q-particles in Japanese should most appropriately be called adjunct ditics. On the one hand, since Q-partides are introduced to the syntactic derivation only as adjuncts, the operation Merge introduces them as heads adjoining to a category of any type or any “size” (or any bar-level, in a more conventional sense) as exhibited in (22a). On the other hand, since they are ditics, they m ust be concatenated to the head of the category with which they are merged, as illustrated in (22b). ^ 3 This should not be taken as any attem pt to define “ditics” formally. In fact, “ditics” is one of those terms which are so firequently used in the Hterature but are not necessarily well-defined (but see section 2.2.2. below). I '* Larson and Siegel (1995, chap. 8) note th at onlyis semantically not a determiner, since it does not exhibit the property of conservativity, defined in the following (cf. Barwise and Cooper, 1981). (i) A determiner D is conservative if for any X, Y, D(X,Y)«»D(XnY,Y) Neither also nor even satisfies this definition. For semantics of focus adverbs in English, see Jackendoff (1972), Karttunen and Peters (1979), TagHcht (1984), Rooth (1985) and Konig (1991), among many others. 1 5 We will turn to the constituent order in Japanese in chapter 5. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 (22) a. at Spell-out b. in PF X“ QOAnax X°+Q° Since Q-particles are adjuncts, they neither select nor are selected for; they are simply merged with an arbitrary category (i.e. X“ in (22a)). This is the source of their “crosscategorial” nature noted above (cf. Okutsu et al., 1986; Cho and Sells, 1995; SeUs, 1995; Yoon, 1995). In addition, since Q-particles are clitics, they have to rely on a host (i.e. X° in (22b)) in PF. Not only does recognition of Q-particles as adjunct chtics in these term s bring about some significant consequences concerning facts of ordering among several distinct types of particles in Japanese, as we will extensively discuss in chapter 3, but it also accounts for the noted paradoxical nature of Q-particles, to which we will return in section 2.4. below. 2.2.2. Chtics vs. inflectional affixes We have recognized Q-particles in Japanese as chtics, in accordance with two independent criteria: i.e. their inabihty to project, and their requirem ent of a PF host. In fact, the term “chtic” is one of those which are so heavily used in the hterature but are not necessarily well-defined (cf. Crystal, 1991). What is widely agreed upon is only our second criterion, namely, their need for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 30 a phonological host, which necessarily leads to recognition of a wider range of entities as “clitics.” In this looser sense of “clitics,” however, there is still reason th at unlike Q-particles, other sim ilar categories hke inflectional afhxes should not be recognized as chtics. In contrast to independent words, both Q-particles and inflectional affixes are bound morphemes, so th at they both need a host or stem to lean on. (23) a. hon-wa (yom-ta) book-TOP (read-PAST) ‘ As for the book, (someone) (read it).’ a’ . * 0 -wa (yom-ta) -TOP (read-PAST) b. (pan-o) tabe-mo (si-ta) (bread-ACC) eat-also (do-PAST) ‘ (Someone) also ate (bread).’ b’ * (pan-o) 0-mo (si-ta) (bread-ACC) -also (do-PAST) c. Tokyo-kara-dake (ki-ta) -from-only (come-PAST) (Someone came) only from Tokyo.’ c’ . * 0-dake (ki-ta) -only (come-PAST) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 (24) a. (suteeki-o) tabe-ru (steak-ACC) eat-PRES ‘ (Someone) eats steak.’ a’. * (suteeki-o) e-ru (steak-ACC) -PRES b. (wuisukii-o) nom-ta (whiskey-ACC) drink-PAST ‘ (Someone) drank (whiskey).’ b’ * (wuisukii-o) o-ta (whiskey-ACC) -PAST c. tukare-ta-to (iw-ta) get tired-PAST-that (say-PAST) (Someone said) that (he) was tired.’ c’ . * 0 -to (iw-ta) -that (say-PAST) As shown in (23), Q-particles always need a host to attach to. Similarly, as shown in (24), tense morphemes hke ru ‘ PRES’ and ta ‘ PAST’ need to attach to a verb stem, and a so-called complementizer to th a t’ generally attaches to a tensed verb. However, a distinction still has to be made between these two types of bound morphemes. Zwicky and PuUum (1983) present several criteria th at help to distinguish chtics (in the looser sense of the term) from Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 32 inflectional affixes. In this section, we will argue th at Q-particles should indeed be characterized as clitics rather than as affixes. First of all, Zwicky and PuUum (1983: 503) m aintain that while inflectional affixes generally exhibit a b i ^ degree of selection with respect to their stems, clitics can exhibit a low degree of selection with respect to their hosts (also cf. Klavans, 1 9 8 5 ) . For instance, the tense affix -ed in English can only attach to a subclass of verbs, i.e. so-called regular verbs (hence, cook ed vs. *go-ed), the genitive marker ‘ s, as a ditic, can attach to a variety of hosts, as exhibited in the following.*^ (25) a. the girl’ s h a t (noun) b. the girl th at I was talking to ’ s hat (preposition) c. the girl th at I saw ’ s hat (verb) d. the girl th at I saw yesterday’ s h at (adverb) The English genitive clitic ‘ s can take as its host either a noun, preposition, verb or adverb. In a sim ilar vein, the past tense affix -ta in Japanese can As noted by Zwicky (1985: 285), Zwicky and PuUum’ s criteria are mostly stated in terms of tendencies, and their imphcations work only one way. Thus, the first criterion in the text should be read as saying that if something is an affix, it wiU probably exhibit a high degree of selection. Hence, pronominal clitics in Romance do not necessarily devastate the m erit of this criterion, if they select for T, as argued by Kayne (1991). As long as ‘ s is identified as the head of DP, it does not qualify as a cUtic in our strict sense of the term (since it projects a phrasal category). However, it might be the case th a t ‘ s is only adjoined to either D° or Spec of DP in PF. If this is the correct approach, % can qualify as a clitic in our strict sense, since it Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 only attach to verbs, but, as noted above, Q-particles may attach to a wide variety of categories. (26) a. tabe-ta (verb) eat-PAST ‘ ate’ b. *gakusei-ta student-PAST c. *taka(-ku)-ta (noun) (adjective) bigb-PAST (cf. ta k a -k u -a r-ta ‘ bigb-INFL-be-PAS'F)^® d *LA-kara-ta (postposition) -from-PAST (27) a. oki-wa (si-ta) wake-TOP (do-PAST) at least woke up’ b. gakusei-dake student-only only students’ (verb) (noun) is a non-projecting category th a t unselectively seeks for a phonological host. For an apparent idiosyncrasy of adjectival inflection in M odem Japanese, see Appendix HL Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 34 c. taka-ku-sae (adjective) high-even ‘ even h i ^ ’ cL LA-kaia-bakari (postposition) -from-only ‘ only from LA’ As shown in (26), the Japanese tense suffîx -ta PA ST can attach to verbs, but not nouns, adjectives or postpositions. However, any of the four categories can serve as a host for Q-particles, as exemplified in (27). Secondly, Zwicky and PuUum (op. cit.: 504-5) note that accidental gaps in the paradigm are more characteristic of affixation than of diticization. They point out th at while it can happen th at a particular verb in E nglish, say, stride does not have its past participle form in the well-established register, it does not happen th at a particular host-clitic combination fails to occur. For instance, it is not Ukely that the English genitive ‘ s can diticize to one d ass of verbs b u t not to another. It seems true of Q-partides in Japanese, too. If they can take hosts of a certain category, aU the members of th a t category can be their hosts. It is h i ^ y unlikely th at a Q-partide mo also’ , for instance, can take certain verbs as its host hut it cannot take others. On the contrary, the potential morpheme -(rar)e can’ in Japanese can not attach Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 35 to su ‘ do’ .‘9 (28) a. tobi-xno su-ru (toft: consonant-stem) fly-also do-PRES tob -e -ru fly-POT-PRES (29) a. ki -mo su-ru (ü : vowel-stem) wear-also do-PRES b. ki -rare -ru wear-POT-PRES (30) a. ne -mo su-ru (ne: vowel-stem) sleep-also do-PRES b. ne -rare -ru sleep-POT-PRES A note on allomorphy of the potential morpheme is in order. Among the two allomorphs -e and -rare, the former is chosen for consonant-stem verbs (i.e. those whose stems end with a consonant) like tob Hy’ in (28a), and the latter is chosen for vowel-stem verbs (i.e. those whose stems end with a vowel) like ki ‘ wear’ in (28b) or ne ‘ sleep’ in (28c). In the tradition of Japanese grammar, consonant-stem verbs are often called five-low verbs (or godan doosbi), and vowel-stem verbs, one-low verbs (i.e. ichidan dooshi), the latter class being further classified into two groups: those whose stems end with i (upper one- low or kami-ichidan) and those whose stems end with e (lower one-low or shimo-icbidan). These two (or three) classes of regular verbs almost exhaust the fist of verbs in Modem Japanese, with only two exceptions: i.e. ku-ru ‘ come-PRES’ (k-irregular) and su-ru ‘ do-PRES’ (s-irregular), whose stem s consist of a single consonant, k and s, respectively. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 36 (31) a. ki -mo su-ru (ku: k-irregular) come-also do-PRES b. ko -rare -ru come-POT-PRES (32) a. si -mo su-ru (su: s-irregular) do-also do-PRES b. *s -e -ru /*si-rare-ru/*su-rare-ru do-POT-PRES On the one hand, a Q-particle mo ‘ also’ can attach to the two irregular verbs as well as all regular verbs, as shown in a-examples in (28) (32). On the other hand, as exhibited in b-examples, the potential morpheme -(rar)e can attach to the k-irregular verb, as well as the two classes of regular verbs, i.e. consonant-stem verbs and vowel-stem verbs, but it cannot attach to the s- irregular verb. In the case of the s-irregular verb su do’ , suppletion m ust take place to create deki-ru can-PRES’ instead of *s-e-ru. This is nothing but an accidental gap in afSxation. Furthermore, according to Zwicky and PuUum (ibid.), morphophonological idiosyncrasies are more of a sign of affixation than of cHticization. As they point out, nothing more than regular morphophonological rules like devoidng applies to diticization o f‘ s (is or has) and Ve (have) in En^ish, with their hosts unaffected, whereas irregularity often occurs to both stem s and affixes Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 37 in inflection, as exemplified in pairs like sleep-slept and ox-oxen. It is obvious th a t Q-particles never affect their hosts morphophonologically, except for the case where the host is a consonant-stem verb.^ In such a case, vowel epenthesis takes place (cf. (28a)). However, this is simply expected, because Japanese is a moraic language th at does not allow a non-nasal coda. On the contrary, morphophonological idiosyncrasies are very common in inflectional affixation in Japanese. For example, consonant-stem verbs which end with a velar sound lose th at stem-final velar, despite epenthesis w ith i, when suffixed with the past tense -ta: e.g. oyoida is derived firom oyog + ta ‘ swim- PAST, and kaita, from kak + ta ‘ write-PAST. This m ust be compared to the case of those consonant-stem verbs which end with s, where only epenthesis takes place: e.g. osita derived from os + ta ‘ push-PAST’. A lth o u ^ the verb ik ‘ go’ is among the class of verbs with a stem-final velar, exactly hke kak ‘ write’ , its past tense form is realized not as *iita, but unexpectedly as itta, with the stem-final consonant assimilated to the dental of the suffix. Furthermore, as noted by Kitagawa (1986), the present tense suffix -ru m ay shift accent, which Q-particles never do. Japanese is a pitch-accent language, and accent falls on the la st syllable of a high pitch peak (cf. Kindaichi, 1960; McCawley, 2 0 There seems to be one exception, though. The so-called topic m arker wa undergoes a phonological change when it chticizes to the accusative case particle o, rendering archaic o-ba instead of *o-wa, though the phonological sequence in the latter is permissible in Modem Japanese (e.g. o-wa nagai Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 1968). The verb stem ta be 'to eat', for instance, is inherently accented, and the accent falls on the penultim ate syllable. Although the past tense suffix -ta does not affect this contour, i.e. ta be-ta, -ru shifts it rightwards, hence, tabe’ -ru. All three criteria discussed above point to the conclusion that despite some similarities, clitics m ust be distinguished from inflectional affixes in Japanese as well as in English, and, more im portantly, Q-partides are more hke cUtics.2i ‘ tail-TOP long = The tail is long."). 2 1 Cardinaletti and Starke (1994) claim th at among personal pronouns across languages, three distinct classes (i.e. strong, weak and chtic pronouns) m ust be recognized, rather than the widely held dichotomy between fuU pronouns vs. chtics, and that distinct properties th at cut across these three dasses of pronouns should be accounted for by different degrees of deficiency, i.e. lack of features, th at results from reduction of projections. In their terms, strong, weak and chtic pronouns are nom in al CPs, ZPs th a t lack features in C and IPs that lack those in C and Z, respectively. Furtherm ore, they suggest th a t this three-ways distinction be extended to other categories as well. Although their recognition of three distinct dasses of pronouns is empirically weU motivated, there seems to be no evidence th a t Q -partides in Japanese have (or historically have had) strong or weak forms from which they are derived by structural reduction. For this reason, we would rather abstain from incorporating their theoretical implementation to our analysis of Q -partides at this point. I would like to thank Joseph Aoun for drawing my attention to their work. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 2.3. Inflectional Morphology and Functional Categories in Japanese 2.3.1. Inflectional morphology As discussed in section 2.1.1., facts of su-support, the Japanese analogue of do-support in English, would be straightforwardly accounted for if Q- particles were heads of projections that intervene between V and T, so th at they would cause movement of V to T to result in a HMC violation, as demonstrated in the following. (33) (John-wa) hon-o yomi-mo "(si)-ta -TOP book-ACC read -also (do)-PAST John also read books. hon-o -mo yonu hon-o -mo yomi Su-Support In our present approach, however, Q-particles are merely adjuncts, and generally, adjuncts are, if skipped, not expected to invoke HMC violations (e.g., VP-adjuncts do not block V-raising in French). This means th a t our analysis of Q-partides as adjunct ditics is not compatible with the account of su- support that rehes on the HMC, presented in section 2.1.1. above, and that if our analysis is to be maintained, we have to seek for an alternative account for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 40 su-support.22 2 2 J.-R. Vergnaud (p.c.) and M. Sano (p.c.) independently suggest that Q- particles, th o u ^ adjuncts, may still have an intervention effect on verb raising since they are heads. However, this possibility does not really arise. In our terms, nothing prevents a Q-particle &om adjoining to VP, to which T is mei^ed to form T in the following way (with the projection of v suppressed for ease of exposition). (i) [t [v p [v p ... V ] Q ] T ] Chomsky (1995: 299) defines the minimal domain of a chain and “closeness” as follows. (ii) (=Chomsky’ s (86)) If a is a feature or an 3?’ category, and CH is the chain (a, t) or (the trivial chain) a, then: a. Max(a) is the sm allest maximal projection including a. b. The domain ô(CH) of CH is the set of categories included in Max(a) that are distinct firom and do not contain a or t. c. The minimal domain Min(6(CH)) of CH is the sm allest subset K of ô(CH) such that for any ySÔ(CH), some pEK reflexively dominates y- (iii) (=Chomsky’ s (87) with modification) p is closer to a than y if P c-commands y and is not in the minimal domain of CH headed by a. According to the definition in (ii), Q is in the minimal domain of the (trivial) chain headed by T in (i). As a result, Q does count as being closer to T than V in(i). Vergnaud ^.c.) points out another possibility th at despite the contention in the text, Q-particles can project but they only allow projection by adjunction. In his terms, VP can be adjoined to Q, and the latter project in the following way. (iv) [t [q o /m ax [vP ... V ] Q O A n a x ] T ] Vergnaud suggests that T and VP c-command each other for selection in (iv) since VP is not dominated by Q o^nax^ being a two-segment category, but Q° still intervenes between V and T, bringing about an HMC effect. However, this proposal is immediately called into doubt. To put aside the question as to whether the head-complement relation between T and VP is appropriately captured in (iv), it is doubtful whether Q is closer to T than V in (iv) in relevant senses. According to the definition in (ii), the two-segment category <Q, Q> should be in the minimal domain of the trivial chain headed by T. Thus, Vergnaud’ s alternative, though interesting, is not adopted in the text. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 41 We agree with Sells (1995) that syntactic movement is not a proper device to account for inflectional morphology in languages like Japanese. Besides the syntactic argum ent firom selection, Sells draws on many morphophonological idiosyncrasies in inflectional morphology in Japanese and Korean, as an argum ent against the Baker-style syntactic analysis of inflection. It is clear th at his arguments are all directed against, what he calls, the strongest S3mtactic hypothesis, to the effect th at all inflectional processes are accounted for by head movement in the syntax. However, as is obvious from our discussion so far, we are taking the position in which inflectional morphology is taken care of by the morphological component on the PF branch of a derivation, along the lines of M arantz (1988, 1989, 1991), Halle and M arantz (1993), Lasnik (1995), and Takano (1996). If, as argued by Halle and M arantz, this component can deal with a certain class of morphological idiosyncrasies that do not exist in the syntax per se, we are immune to Sells’ criticism. It is only a m atter of the locus of the component that treats morphological idiosyncrasies; while it is only in the lexicon for Sells, it can exist in PF for us.^ Following Lasnik (1995) and Takano (1996), we claim that the surface 2 3 We are not arguing th at the lexicon can not deal with idiosyncrasies. If derivational morphology, in opposition to inflectional morphology, is taken care of by the lexicon, it m ust be able to deal with certain irregularities involved in the former, in the sense of Chomsky (1970). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 42 string of V+T (e.g. tabe-ta ‘ eat-PASTO in Japanese obtains as the result of a PF process of morphological concatenation th at appUes to two heads, independent heads, affixes and chtics alike, under string adjacency.^ This process can roughly be schematized in the following way. (34) Morphological Concatenation in PF: X - Z - Y ... —> ... X+Y (where X and Y are term inal strings, and Z = 0 ) Note that the rule in (34) is only sensitive to term inal strings. In principle, it applies not only to two heads of projections that are in a selectional relation, as in (35a), but also to the head of an adjunct and the head of the projection to which that adjunct is adjoined, as in (35b). (Order is irrelevant in (35a, b).) 2 4 Lasnik (1995) claims th at while auxiliaries including have emd be raise to T in overt syntax, as in (i), main verbs do not; tensed m ain verbs are formed by “affix hopping^, in the sense of Chomsky (1965), at PF, as in (ii). (i) ... V 2[+auxi+T ... t 2 ... (at Spell-out) (Ü) a. ... T ... V f-A U X ] ... (at Spell-out) b. ... V [ . a u x i+ T ... (in PF) For further details, see Lasnik (1995). Takano (1996) presents a far-reaching hypothesis, alternative to Kayne (1994), th at the base order is Spec-Complement-Head. (i.e. SOV) across languages, and he claims that the SVG order (as well as the VSO order) obtains only in the presence of V-raising to a higher functional category (i.e. movement of V to the light verb vin English and ofV-Hvto T in French). We will discuss Takano’ s (1996) and Fukui and Takano’ s (1998) proposals in detail in chapter 5. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43 (35) (=com plem ent) Y D XP Xo^-yo (=adjunct) Y D XP XOfYo Formation of a V+T sequence, e.g. tabe-ta ‘ eat-PAST, exemplifies the process in (35a), where VP (=XP) is the complement of T (=Y°), and a V+Q-particle sequence like tabe-sae ‘ eat-even’ , for example, is the result of the process in (35b), where Q-particle (=Y° and YP) is adjoined to VP (=XP). Furthermore, we will assume with Lieber (1983), Klavans (1985), M arantz (1988, 1989) and Speas (1991), among many others, that affixes, or functional heads above all, can be specified with morphological, as well as syntactic, subcategorization (or selectional) features.® Let us further assume th at a morphological subcategorization feature of a head is the specification of a PF property of that head th a t determines what type of category it is morphologically concatenated with, whereas a syntactic subcategorization feature is, as in its conventional sense, the specification of a ® Note th at only [+a£fixal] heads require this additional specification (cf. Lieber, 1983). In chapter 5, we will suggest th at this morphological specification is limited to non-lexical categories across languages, along the lines ofFukui’ s (1988, 1995) Functional Param etrization Hypothesis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 44 selectional property of that head relevant to the LF side of the derivation.^ 27 2.3.2. Functional categories It is still controversial which of the functional categories attested in other languages exist in Japanese. The success of an answer to this question depends on whether the known properties of the language can be explained in a principled way. As noted in the literature, English, for instance, has wh- movement and expletives, but Japanese does not have either of them. Fukui (1986) and Fukui and Speas (1986) claim th at the presence of wh-movement and expletives in English and their absence in Japanese should be reduced to 2 6 Sells (1995) argues against “double” specification of this sort, simply because it is redundant. However, it is an empirical issue whether the text approach will only bring about redundancy, resulting in an excessive power of the grammar. One might predict th a t if an affixal head H° syntactically selects for XP, it is morphologically concatenated with the head X° of XP, but this is not necessarily the case. As pointed out by Klavans (1985) and M arantz (1988, 1989), a head is not always morphologically concatenated with the head of the complement th a t it syntactically selects for. For instance, if the English genitive ‘ s is in D (but see footnote 17), it syntactically selects for an NP as its complement, but it will not be morphologically concatenated with the head of the complement; hence, John’ s book vs. *John book’ s. 2 7 Ikawa (1996) proposes that functional categories that induce overt movement for SPEC-head agreement in English and other languages are morphologically specified as suffixes, or (en)chtics in our term s; hence, they need something to attach to in their SPEC positions in PF. Thus, strength of features, Ikawa argues, can be reduced to this morphological properties of functional heads. However, Ikawa’ s rendition requires much comphcation due to the fact of successive cycHcity of movement in natural language (cf. CoUins, 1994). It is hard to imagine th at an interm ediate trace is able to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 45 the existence and non-existence of functional categories that induce agreement; while E n g lish has them, Japanese does not. In other words, these authors argue th a t the presence of functional categories should be parametrized across languages; the value for the param eter is set positively for E n glish but negatively for Japanese. However, Kuroda (1988) presents a different answer to the sam e question. Assuming, as a null hypothesis, th a t Japanese has the full set of functional categories th at are attested in languages like E n g lish , he m aintains th a t the noted crosslinguistic differences between Japanese and English must result from the choice between optionahty and obligatoriness of agreement. For Kuroda, Enghsh has wh- movement and expletives, because agreement is forced in this language, but Japanese does not have them because agreement is not forced. Thus, Kuroda proposes that languages m ust be param etrized as to whether agreement is forced or not.^ Full examination of these opposing proposals as to functional categories in Japanese is not our purpose in th is thesis. However, the m inim ahst view, which has been g a in in g ground in linguistic theory for the past several years. serve as a stem for a suffix to attach to. 2 8 Kuroda (1988) takes agreement (or “X-Agreement” in his terminology) as a wider notion than is widely understood; it is established under head- complement relation as well as Spec-head relation. However, we restrict our use of the term to its narrower sense (i.e. Spec-head agreement) in the text. See also Hoji (1997) for relevant discussions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 46 seems to shed l i ^ t on this issue. According to Chomsky (1995, chap. 4), legitim ate representations of the linguistic computational system are constrained at interface levels by bare output conditions required by extraneous performance systems; the principle of Full Interpretation (FI) requires linguistic representations to include all and only objects interpreted at the interface levels, i.e. PF and LF. If a linguistic object is interpreted at one or the other level, it m ust exist at that level. Conversely, we may assume that if a notion is necessary for instruction to a performance system at one of the interface levels, th at notion m ust be encoded in a linguistic object to be interpreted at th at level. If this is on the right track, we may further assume that a category th at is interpreted at one of the interface levels m ust exist there. This Une of reasoning forces us to opt for Kuroda’ s approach; Japanese m ust have functional categories as long as they are interpreted at one of the interface levels, whether or not they induce agreement. Assuming that the conclusion above is correct, we propose that Japanese has at least C (=complementizer), T (=tense) and D (=determiner) in its repertoire of functional categories. C is universally required, mostly because, as noted by Chomsky (1995, chap. 4), it determines the ülocutionary force of a clause, and also because, as claimed by Cheng (1991), it plays an important Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 47 role in clausal-typing.29 It is generally agreed th at T functions as a sentential operator at LF, in the sense of Higginbotham (1985), which spatio-temporally locates an action or process expressed by the predicate. Following Fukui (1995), we will stipulate that “noun phrases” in Japanese are headed by D, which is always phonetically nuU.^ We will consider, in chapter 3 below, th at case particles are morphological realizations of CASE features in Ds. All these considerations seem to suggest th at our proposal that Japanese has C, T and D is not unwarranted. Furthermore, we will assume th at the light verb, V, a half-functional, half-lexical, category th at assigns an external theta role, which has been originally proposed by Hale and Keyser (1993) and adopted by Chomsky (1995), is existent across languages, therefore, in Japanese. As is well known, Japanese is a strictly head-final a ^ u tin a tiv e language. This means th a t V, v, T and C are always linearized in this order at the end 2 9 However, Fukui (1995: 340ff.) argues th at since C (as well as AGR) does not seem to play any role at LF, Japanese need not have it, claiming that to is not like the English complementizer that, but it is a direct quotation introducer, which Fukui (1986) classified as a kind of P (as well as the question marker ka as a noun). However, it is uncertain whether his justification for the absence of C in Japanese (or any language, for th at matter) is grounded well enough to defy the text argument for the universal presence of C. 2 0 This is in fact an opposite position firom th at which Fukui took in his earher writings (e.g. Fukui, 1986, 1988; Fukui and Speas, 1986). Fukui (1995: 357) explicates that, given his restrictive theory of functional parametrization, he is forced to assume th at Japanese has D (though it does not induce agreement), if D is universally visible at LF. This is not unfounded, because D, lexically realized in many languages, seems to play a crucial role in scope and binding, which are universal phenomena. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48 of a clause, as exhibited in the following. (36) (John-wa) [Mary-ga pizza-o tabe-ta-to] (iw-ta) -TOP -NOM -ACC eat-PAST-that say-PAST ‘ (John said) th at Mary ate pizza.’ Notice th at V, v, T and C in the subordinate clause are concatenated into a verbal complex tabe-ta-to ‘ eat-PAST-thaf, with vphonetically nuU.3i As noted by Sells (1995), the V-v-T-C sequence always constitutes a phonological word, which may have at most one pitch peak, hence, tabe-ta-to. This is the result of morphological concatenation that takes place in PF, as assum ed above. As the source of the strict head-finality in Japanese, we will assum e 3 1 The light verb is not always phonetically nuU, however. When it selects for a verbal noun (VN), it m ust be phonetically realized as su do’. We consider th at VNs constitute a neutralized category, ambiguous between V and N, in the sense of Lefebvre and Muysken (1988). If they are embedded under v, they behave as Vs, and if embedded under D, they behave as Ns. Hence, the following case arrays obtain. (i) a. John-ga [vpeigo-o [v n benkyoo]]-siv-ta -NOM Enghsh-ACC study -do -PAST John studied English.’ b. John-ga [opeigo-no [vn benkyoo]-o] siv -ta -NOM English-GEN study-ACC do-PAST In (ia), the VN benkyoo study’ is embedded under the light verb su, and hence, its thematic object is marked by the accusative. In (ib), on the contrary, the VN is embedded under D, which is phonetically realized by the accusative case particle (see chapter 3), and the thematic object is marked by the genitive, which is only allowed DP-intemally. Note further th at the whole DP in (ib) is taken as the object of the heavy (i.e. main) verb su do’ , which is independently justified (cf. Grimshaw and Mester, 1988; Miyagawa, 1989a; Hasegawa, 1991; Kageyama, 1991, Saito and Hoshi, 1997; among many others). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 that the following param etric statem ent holds. (37) AU functional heads in Japanese are morphologicaUy sufSxal.^ We claim that (37) holds for Q-particles as weU as paradigmatic functional categories Kke T and C. Since Q-particles are clitics, (37) implies th at they need a host on their left, i.e. they are enclitics. Since Chomsky (1970), it has been a common exercise to classify the four major lexical categories with a combination of the values in two features, [±V] and [±N]. Fukui (1995) suggests in a footnote th at lexical and functional categories in natural language might better be classified with a set of two, rather than one, binary features, i.e. [±F(imctional)l and [±L(exical)], in the foUowing way. (38) Cross-classification of Lexical and Functional Categories (Fukui, 1995: 338, fii. 3) 4-L(exical) -L +F(imctional) functional elements with lexical nature ‘ pure’ functional elements -F lexical categories (substantive elements) ‘ minor’ categories (particles, etc.) This statem ent conforms to Fukui’ s (1995) functional param etrization hypothesis in (i), the inspiration behind which he credits to Borer (1984). (i) Functional Param etrization Hypothesis (Fukui, 1995: 342) Inside the lexicon, only functional elements are subject to param etric variation. In chapter 5, we wiU suggest that param etrization on morphological properties of functional categories decides order in a given language; hence, UG need not mention order at aU. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 Obviously, major lexical categories, i.e. N, V and A, are identified as [-F, +L]. As noted by Fukui, ‘ pirre’ functional ([+F, -Lj) elements are typical functional categories including T and C, well attested in English and other European languages. Heterogeneous [+F, +L] categories are, he states, those functional elements which “retain, to varying degrees, their characteristics as lexical categories,” firom which they have historically developed. The so-called light verbs, proposed in many languages, like su ‘ do’ in Japanese proposed by Grimshaw and Metser (1988) and the abstract “causative” verb first proposed by Hale and Keyser (1993) and then adopted by Chomsky (1995), belong to this class, as well as some types of prepositions (e.g. the prepositional complementizer ibr in English) and seemingly functional elements in East Asian languages (e.g. the causative and passive morphemes in Japanese and the passive and affective m arkers in Chinese).^ Given that inflectional heads have morphological subcategorization features and categories are crossclassified as in (38), the functional categories in Japanese proposed so far are summarized on the following table. 3 3 Fukui does not elaborate on the [-F, -L] type, except for an indication that it includes “particles”. However, we are claiming th at Q-particles are specified positively at least in their [F] feature. Furthermore, we will a i^ e , in chapter 3, th at case particles in languages like Japanese are phonetic realizations of CASE features in Ds. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 (39) functional categories examples categorial feature specification morphological subcategorization features complementizer (C) to ‘ th at’ , ka Ques’ [+F, -L] [To_] tense (T) ru “ present’ , ta ‘ past’ [+F, -LI [^_] light verb (v) su ‘ do’ , 0 [+F,+L] [Vo_l determiner (D) 0 [+F, ±L]34 [No_] Q-particles w a‘ TOF mo also’ dake ‘ only’ [+F, -L] or [+F,+L] not specified The crossclassification of functional categories on table (39) is still tentative, not final, and it requires further examination and justification. However, it still seems to serve for our present purposes. On the fourth column of the table above is indicated the morphological suhcategorization feature of each functional head. The notation that we employ here takes the general form PC °_ ] to indicate that the head in question m ust be concatenated with a head X P on its left. Recall that unlike inflectional affixes, Q-particles as clitics are not morphologically selective, and therefore, the fourth column for them does ^ The specification of the value of the [L] feature for D is not entirely clear. It seems to be generally considered as “pure” a functional category, i.e. [+F, -L], as C or T. However, once hybrid categories like vare introduced, and if v shares some lexical properties with V, then D can also be claimed to share certain lexical properties with N; hence, it may be [+F, +L] as well. See chapter 3 below for relevant discussions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 52 not have a specification. The process of morphological concatenation takes place in the morphological component, which we assume to exist on the PF branch of a derivation, following M arantz (1988, 1989, 1991), Halle and M arantz (1993) and Chomsky (1995).^ 2.4. Deriving the Paradoxical Properties 2.4.1. Su-support as a morphological process Assuming as above th a t verbal inflection results firom morphological concatenation rather than firom syntactic head movement and th at Japanese has a full repertoire of functional categories but they do not induce agreement at all, we will be able to account for su-support in a new light. Our earlier example in (33) is reproduced as (40). (40) (John-wa) hon-o yomi-mo *(si)-ta -TOP book-ACC read -also (do)-PAST ‘ John also read books.’ In (40), a Q-particle mo also’ is adjoined to a projection of either V or v If the Q-particle is adjoined to a projection of V, say VP, the representation in (41) 3 5 In one extreme of this line of approach, Halle and M arantz (1993) argue th at lexical items (i.e. XP-categories) introduced in the syntactic derivation are only bundles of features and actual vocabulary items are inserted in the morphological component on the PF branch, up to the compatibility with the set of features dominated by the topmost X°-level category after morphological rules (e.g. “merger” and “fusion” of X9s) applicable in th at component. For Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 53 obtains as the input to the morphological component on the PF branch of a derivation, where morphological concatenation takes place ts (41) DP ohn-ga VP y o hon-o -ta mo yomi In (41), mo, being a clitic, can attach to a host of any category; hence, V° can satisfy its morphological requirement. T can also be attached to v° under string adjacency. However, does not have a legitim ate stem to hold on to on its left, since it is morphologically subcategorized for [V°_ ] (cf. table (39)) but m ois not qualihed in itself and it breaks the linear adjacency between v° and V°. Hence, v° would not be able to satisfy its own morphological requirement, and it would end up as an illegitimate PF object. Hence, as a last resort, the epenthetic verb su ‘ do’ , which is V°, m ust be inserted for v° to attach to. details, see Halle and M arantz (1993) and the references cited therein. 3 6 If the Q-particle adjoins to V° instead, the text argum ent will still hold. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 54 If mo is attached to a projection of v, almost the sam e story holds. Consider the representation in (42), where mo is adjoined to (42) ohn-ga y, hon-o -ta In (42), a lth o u ^ mo and v ® can be attached to v° and V°, respectively, T ® cannot satisfy its morphological requirement by being attached to since mo intervenes. Again, su-support m ust be apphed to save T° firom being astray. In sum, su-support is employed, when adjacency between T° and v° or th at between v° and V° is broken, so th at the morphological subcategorization requirement of T ® or is not satisfied.^^®^ ^ Again, if mo is adjoined to v° or v * , it does not affect the text argument. The success of the account of su-support under our hypothesis th at Q- p articles are adjunct clitics is contingent upon the absence of (at least, overt) movement of V to T (possibly, via Notice that the Adjunct Clitic Hypothesis has stripped Q-particles of the intervention effect with respect to head movement, because they are now simply adjuncts. If movement of V to T should take place in the context of (41) or (42), nothing seems to prevent it. Although many Japanese linguists have postulated movement of V to T Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 55 during the course of syntactic derivation, few seem to have provided independent empirical evidence for it. Given th a t morphology is in PF, the mere fact th at verbs are suffixed with inflectional elements does not count as evidence for syntactic V-raising. For this m atter, Koizumi (1995) is an exception, in th at he draws on independent facts to argue for overt V-raising in Japanese. However, his arguments from “coordination” and “clefting” in the language are not exempted from skepticism (see Appendix H). In the meantime, following Huang (1988), W hitm an and Otani (1991) claim th at the “sloppy identity” reading of (ib) results from an elided VP after V-to-T raising, as indicated in (ü). (i) a. John-wa [zibun-no tegami]-o sute-ta -TOP self-GENletter-ACC discard-PAST ‘ Johna threw away selfa’ s letters.’ b. Mary-mo sute-ta -also M ary also threw away John’ s letters.’ (strict reading) OR Marys also threw away sel&’ s letters.’ (“sloppy” reading) (Ü) a. John-wa [vp [zibxm-no tegamij-o tv] sutev-ta? b. Mary-mo [vp e ] sutev-tax Roughly speaking. Whitman and Otani propose th a t the VP in (iia) is copied to the ehded VP position in (iib), thus deriving the “sloppy” reading of (ib). However, Hoji (1998) presents several empirical arguments against the view that cases like (ib) contain an ehded VP, and instead argues th at w hat those cases include is a null pronominal object. If Hoji is correct. W hitman and Otani’ s argum ent for V-raising will also become rather groundless. ^ K Miyagawa (1997) proposes an alternative in which raising of V to T takes place in the morphological component on the PF side of a derivation, and an intervening projection headed by a verb-final Q-particle brings about an HMC effect, hence prohibiting verb raising, in th at component. However, Sells’ (1995) criticism against the syntactic treatm ent of verbal inflection in term s of selection carries over to her analysis. Furthermore, it is rather questionable if conditions hke the HMC are apphcable in PF. Accordingto HaUe and M arantz (1993), much of whose tenet Miyagawa draws on, operations in the morphological component (e.g. “fusion”, “fission”, etc.) are allowed much more freedom than syntactic operations. In chapter 3 below, we will contend th a t one possible surface placem ent of Q-particles is the result of “morphological fusion” in the sense of H alle and M arantz. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 56 2.4.2. Selection As noted earlier, our analysis of Q-particles as adjunct clitics can nicely accommodate facts of selection, the m ain source of Sells’ (1995) criticism against syntactic inflection, as well. In the following, some relevant examples are reproduced from our earlier data base. (43) a. Johnz-wa [ez susi-o tabe]-te mi-ta -TOP sushi-ACC eat-ASP see-PAST ‘ John tried eating sushi.’ b. Johnz-wa [ez susi-o tabe]-te-m o m i-ta -TOP sushi-ACC eat-ASP-also see-PAST ‘ John tried also eating sushi.’ c. *John-wa [ez susi-o tabe]-ni-mo m i-ta -TOP sushi-ACC eat-ASP-also see-PAST (44) a. Maryz-wa Tuldzi-e [ez susi-o kuruma-de tabe]-ni ik-ta -TOP Tsukiji-to sushi-ACC car-by eat-ASP go-PAST ‘ Mary went to TsuMji by car to eat sushi.’ b. Maryz-wa Tukizi-e [ez susi-o kuruma-de tabe]-ni-dake ik-ta -TOP Tsukiji-to sushi-ACC car-by eat-ASP-only go-PAST M ary went to Tsukiji by car only to eat sushi.’ c. *Maryz-wa Tukizi-e [ez susi-o kuruma-de tabe-te]-dake ik-ta -TOP Tsukiji-to sushi-ACC car-by eat-ASP-only go-PAST Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 Recall th at Sells’ (1995) argument against the syntactic analysis of inflection comes from the fact th a t as exhibited in cases like the above, the matrix verb selects for a clausal category headed by te ‘ PERF’ or ni TMPERF”, with or without the intervention of a Q-particle. He m aintains that if, as suggested or simply assumed by proponents of syntactic inflection, the presence of a Q- particle entails an intervening projection between the verb and the category that it selects, the selectional relation will fail to be local, as shown in the following. (In (45), irrelevant details are neglected.) (45) X V pP Asp P te/ ni tabe miiik If the projection of a Q-particle indeed intervenes the verb and the selected category (i.e. AspP) as in (45), Sells’ criticism will not be without ground. Ih (45), strict locality of the intended selectional relation is broken. However, this locality problem does not arise in our approach. Recall that under our analysis, Q-particles are heads that do not project and they can be merged with any category. If this is the case, the representation of the relevant part of the examples in (43) and (44) wiU look, for example, hke (46), instead of (45). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 (46) P 0 AsdP îP P sp tabe Q O ' i n a * m i/ik In (46), the Q-particle is adjoined to the maximal projection headed by the designated aspectual morpheme. Since a Q-particle, by hypothesis, does not project, the category that it adjoins to (i.e. AspP) projects instead. Hence, it is always the case that strict sisterhood between the verb and the category that it selects for is maintained.^ Before concluding this subsection, it seems worth while making a comment on the noted “crosscategorial” status of Q-particles. Since they can attach to a category of any type, m any linguists (e.g. Cho and Sells (1995), Sells (1995), and Yoon (1995), among others) have characterized them as “crosscategorial” suffixes. It has been contended so far th a t Q-particles are ^ As noted by Sells himself, one way to escape firom his criticism against the representation in (45) is to state th a t XP headed by the Q-particle lacks categorial features; hence, all the necessary categorial information is supplied, instead, firom its complement, i.e. AspP, by a mechanism like “feature percolation” (cf. Lieber, 1983, 1992). However, such an approach necessarily raises a serious question. Note th at this feature percolation approach amounts to saying that XP is a category without categorial features at all. Even under the BPST of Chomsky's (1994), it is not entirely clear whether a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 adjunct clitics. They are clitics, since they are heads th at need a host to lean on. On the other hand, they are adjuncts; hence, they do not select, nor are they selected for, in syntax. As adjuncts, they can freely be introduced to a derivation, irrespective of selectional constraints, as long as they can be properly licensed and interpreted at the interface levels. These considerations will p u t their specification as “crosscategorial” into vacuity. To sum up, while propounders of Q-particles as heads of projections m ust be subjected to Sells’ (1995) challenge, we are not. Under our analysis, Q- particles are heads but they do not project. Hence, even in the presence of an intervening Q-particle, the selectional relation between the verb and the category that it selects for remains local. category without a categorial feature in its set of formal features can project. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 60 CHAPTER 3 Q -p a rtic le s a n d C ase P a rtic le s Based on the adjunct clitic hypothesis proposed in chapter 2, the present chapter further looks into the morphological nature of Q-particles. Since they are adjuncts, Q-particles exhibit relative freedom in distribution. However, as shown in section 3.2., there are very strict restrictions on the relative order, on the one hand, between Q-particles and case particles, and on the other, among Q-particles themselves. In section 3.1., it wiU be dem onstrated th at the distinction between two subclasses of Q-particles, recognized by traditional Japanese grammarians, is in fact real, and therefore, it m ust be accommodated in our analysis. While one class of Q- p articles are purely functional elements, the other class exhibits some nom m ai, hence, lexical, characteristics. After presenting a theory of morphological case that we adopt in this thesis, section 3.3. accounts for the observed ordering restrictions on different classes of particles. Finally, a new perspective to the so-called “case marker drop” will be presented in section 3.4. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 1 3.1. Kakari-joshi (K-particles) vs. Fuku-joshi (F-particles) In chapter 2, we have characterized all Q-particles as adjunct chtics in morphosyntactic and morphophonological terms. However, careful inspection reveals th a t the recognition of two distinct subclasses among them, i.e. kakari-joshi (K-particles) and fuku-joshi (F-particles) by traditional Japanese Gram m arians (e.g. M atsushita, 1930; Yamada, 1936; Hashimoto, 1969; inter aha) is empirically weU grounded, hence m ust be accommodated in our analysis. ^ ^ While particles hke wa TOP’ and mo ‘ also’ are K-particles, others hke dake only’, made even’ , and bakari ‘ only’ are F-particles.^ In fact. 1 The terms, kakari-joshi and fiiku-joshi, are due to Yamada (1936). Keeping kaku-joshi intact, Hashimoto (1969) further classifies Yamada’ s fuku-joshi into two subgroups, fuku-joshi and juntai-joshi ‘ quasi-nominal particle’. M atsushita’ s (1930) classification of Q-particles is almost identical to Hashimoto’ s, though he gives a different nam e to each class, i.e. teiji-joji ‘ presentational particle’ for kakari-joshi, fuku-joji adverbial particle’ for fuku- joshi, and mei-joji nominal particle’ for juntai-joshi. For a historical survey of studies on Q-particles, see Miyachi (1997). 2 For instance, Yamada (1936: 400) states th a t while K-particles are prepositional modifiers, F-particles are predicative modifiers. This statem ent is precursory of our contention in chapter 4 th at these two classes of Q-particles are hcensed by different functional heads at LF. 3 It is generally agreed among the traditional linguists th at while particles hke wa TOP’ , mo also’ , sae ‘ even’ , sika ‘ anything/ anybody but’ , hoka anything/ anybody but’ and koso (the) very’ are classified as K-particles, others hke dake ‘ only’, bakari ‘ only’, m ade‘ even’ , kurai ‘ at least, about’ , ka ‘ or’ , yara and’ , nado and (the like)’ are regarded as F-particles. However, the classification is often arbitrary, and it seems to be the case th at the class of K-particles in the present-day spoken Japanese is much smaller, leaving only wa and mo as core members. According to Shinmura’ s Kojien Dictionary (4th edition, 1991), Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 62 these two classes of particles exhibit several morphophonological and etymological differences. The two and only two core members of the class of K-particles, wa TO F and mo ‘ also’ , are both monosyllabic and inherently unaccented. More importantly, they are rem nants of the set of functional particles th at required a specific verb inflection in Old Japanese (OJ). This phenomenon is widely known as kakari-musubi ‘ particle-inflection agreement’ , which was first systematically described by Motoori (1796).4 s F-particles, on the other hand. sae has undergone the shift firom an Old Japanese F-particle to a Modem Japanese K-p article; hence, it is not surprising if it exhibits mixed characteristics. Koso, which used to be a member of the class of K-particles, has undergone the opposite shift, now being simply an emphatic F-particle. Furthermore, both sika and hoka are not only etymologically complex but also quite different firom the others in th at they are NPI-creating. Since detailed description of individual particles is beyond the scope of this thesis (but see Miyachi, 1997), we will only focus on those whose affiliation is out of question. ^ As is obvious, kakari-joshi (K-particle) is named after kakari-musubi ‘ particle-inflection agreement’, to indicate that K-p articles induce such agreement (cf. Yamada, 1936). 5 In OJ, spoken until about the 14th century, while wa TOP’ and mo ‘ also’ required the corresponding verb to take the conclusive form (or shuushi-kei), the two emphatic particles zo and nam and the two question particles ya and ka required the attributive form (or rentai-kei) of a verb, as exemplified in (i). (i) a. tuki -waAno idu -0 moon-TOP/also appear (conclusive) The moon appears.’ b. tuki -zo/nam idu-ru moon-EMPHATIC appear (attributive) THE MOON appears!’ c. tuki -k a ^ a idu-ru moon-Ques appear (attributive) ‘ Does the moon appear?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 63 are etymologically related to a variety of content words; e.g. dake is derived firom a noun take length, height’, and bakari is derived firom the attributive form of a verb, hakar *to m easure’ , possibly via its nominal bakari ‘ measure, scale’ . Many members of this class are inherently accented, e.g. dake’ , ba’ kari, m a’ de, although, after chtidzation to a stem, they naturally lose accent, hence, ho’ n-dake ‘ book-only’ , ho’ n-bakari ‘ book-only’, ho’ n-made ‘ book- even’ . It is rather obvious th at while K-particles constitute a class of purely functional elements, F-particles retain some lexical properties of their etymological ancestors. Then, it is desirable to be able to express the distinction between K-particles and F-particles in a formal way. Given the set of two binary features [±F] and [±L], as suggested by Fukui (1995: 338, fii. Furthermore, koso, another emphatic particle, required the perfective form (or izen-kei) of a verb, as in (id). (i) d. tuki -koso idu-re moon-EMPHATIC appear perfective) THE MOON has appeared (and therefore/but...)’ For the present tense in OJ, the conclusive form required no suffix as indicated in (ia), but the attributive and perfective forms required specific suffixes as shown in (ib-d). However, as the attributive form eradicated the conclusive form for the status as the simple present (or non-past) form after the 15th century, kakari- musubi faded away. Because the attributive ending now ceased to be specific, ksLkari-musubi, naturally, lost its rhetorical force. Koso also lost its kakari- musubi function, because the perfective ending became obsolete. According to Ohno (1993: 113£f.), the rationale and concessive use of the perfective form was substituted by the analytical (uninfiecting) conditional conjunctions -(re)ba and -(re)do(mo), respectively. For details of the history of kakari- musubi, see Ohno (1993). Incidentally, Ikawa (1995) argues th at kakari- musubi in OJ was an instance of syntactic SPEC-head agreement. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 64 3) (see section 2.3.2. in chapter 2, above), K-partides m ust be specified as [+F, -L] (a purely functional category), and F-partides as [+F, +L] (a functional category with lexical properties).® As we have seen, both K- and F-partides enjoy wide fireedom in distribution since they are adjuncts. However, there are some ordering restrictions. For instance, while F-partides can appear DP-intemally, K- partides cannot. This contrast will naturally follow firom the assumption made above th at while K -partides are purdy functional, F-partides are half- fimctional but half-lexical by retaining nominahty. Furthermore, as we will carefully see in chapter 4, different possibilities of assodation with focus result firom distinct functional heads th at K- and F-partides seek for as licensers at LF; while K -partides, as focus markers, adjoin to a [+F, -L] head (i.e. T) for hcensing, F-partides adjoin to a [+F, +L] head (i.e. 0 for the same purpose. 3.2. Ordering Restrictions As recognized by traditional Japanese grammarians (e.g. M atsushita, 1930; Yamada, 1936; Hashimoto, 1969; among others), there are some strict ordering restrictions gunong K-partides, F-partides and case partides. ® The much higher degree of “dosedness” of K-partides, whose core members are exhausted by only two p artid es wa and mo, seems to support this view. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 Yamada (1936; 447, 483) puts forth the following generalizations. (1) a. A case particle and a K-particle can be used simultaneously. In such a case, it is always the case th a t the K-particle is preceded by the case particle. b. A case particle and an F-particle can be used simultaneously. In such a case, either one can precede the other.^ c. An F-particle and a K-particle can be used simultaneously. In such a case, it is always the case th at the K-particle is preceded by the F- particle. As seen shortly, Yamada’ s generalizations are empirically correct, except for one case. However, our concern is whether our theory of particles can derive the facts stated in his generalizations in a principled way. In section 3.4., we will demonstrate th at our hypothesis of Q-particles as adjunct clitics, together with a very plausible theory of morphological case, nicely accounts for the ordering facts. 3.2.1. Order between Q-particles and case particles First, let us consider the order between Q-particles and case particles, including ga ‘ NOM’, o ‘ ACC’ , ni ‘ DAT and kara ‘ OBL(from)’, to see th at ^ Yamada (1936) repeatedly states (e.g. p. 439, p. 447) that the basic rule is that F-particles follow case particles Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 6 Yainada’ s generalizations in (la, b) are basically correct. (2) a. John-ga hon-wa/mo yom-ta -NOM book-TOP/also read-PAST ‘ (Ut) John read at least/also books’ b. John-ga h o n -d ak e/m ad e/b ak ari yom-ta -NOM book-only/even/only read-PAST John read only/even/only books’ cf- John-ga hon-o yom-ta -NOM book-ACC read-PAST ‘ John read books’ (3) a. *hon-ga-wa/mo (NP - NOM - K-particle) book-N OM-TOP/also b. *hon-ga-dake/m ade/bakari (NP - NOM - F-paurticle) book-N OM-only/even/only c. *hon-wa/mo-ga (NP - K-particle - NOM) book-TOP/also-NOM d. h o n -d ak e/m ad e/b ak ari-g a (NP - F-particle - NOM) book-only/even/only-NOM Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (4) a. hon-o-ba®/mo (NP - ACC - K-particle) 67 book-ACC-TOP/also b. h o n -o-dake/m ade/bakari (NP - ACC - F-particle) book-ACC-only/even/only c. *hon-wa/mo-o (NP - K-particle - ACC) book-TOP/also-ACC d. h on-dake/m ade/bakari-o (NP - F-particle - ACC) book-only/even/only-ACC (5) a. hon-ni-wa/mo (NP - DAT - K-particle) book-D AT-TOP/also b. h o n -n i-d ak e/m ad e/b ak ari (NP - DAT - F-particle) book-DAT-only/even/only c. *hon-wa/mo-ni (NP - K-particle - DAT) book-TOP/also-DAT d. h o n -d ak e/m ad e/b ak ari-n i (NP - F-particle - DAT) book-only/even/oiily-DAT * 0-wa ACC-TOP’ is phonetically realized as o-ba, which is rather archaic but still possible. It m ust be noted that the particle wa is still orthographically represented as ha, the consonant of which was pronounced as a bilabial firicative, [$] or [g], in OJ. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 (6) a. hon-kara-w a/m o (NP - OBL® - K-particle) book-from-TOP/also b. h o n -k ara-d ak e/m ad e /b ak a ri (NP - OBL - F-particle) book-from-only/even/only c. *hon-w a/m o-kara (NP - K-particle - OBL) book-TOP/also-firom d. h o n -d ak e/m ad e /b ak a ri-k ara (NP - F-particle - OBL) book-only/even/only-from As shown in (2), DPs may appear without a case particle as long as they receive a Q-particle. Yamada’ s generalization in (la) is correct, since, K- particles can only follow case particles but not precede them (cf. a- and c- examples in (4) - (6)), except for the nominative case particle ga, which K- particles cannot even follow (cf. (3a)). His generalization in (lb) also turns out to be true; as exhibited in b- and d-examples in (4) - (6), F-p articles can either precede or follow case particles, except, again, for ga, which they can only precede (cf. (3b, d)).^° ® Oblique particles hke kara from’ are often designated as postpositions which, taking a DP complement, project a phrasal category (i.e. PP). We wül turn to such a possibility in section 3.3.2. below. This exceptional status of ga seems to come from the widely known fact th at it used to share with no the function as a genitive marker in OJ, though the functions of these two particles diverged in later phases of the language. For the diachronic shifts th at ga has undergone, see Ishigaki (1955). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 69 The restriction on the order between Q-particles and case particles is summarized on the table in (7).ii (7) NOM-P P-NOM ACC-P P-ACC DAT-P P-DAT OBL-P P-OBL W 3 k * * V * V * V * m oK 4r * V * V * V * Sa0K or F * (?)V V V m v V dakep * V V V V V V V made? * V V V V V V V bakarip 4r V V V V V V V 3.2.2. Order between K- and F-p articles Yamada’ s generalization in (Ic) also holds true; the only order allowed between K- and F-p articles is th at in which a K-particle follows an F-particle, as indicated in the following. (8) a. John-dakeF-waK -only -TOP “ As noted above, sae ‘ even’ , often taken as a K-particle in Modem Japanese, retains its characteristics as an F-particle in OJ, and therefore, it is designated as ambiguous on the table. However, for some speakers, it seems to be a perfect K-particle; those speakers do not allow it to be followed by a case particle. This state of affairs is indicated by question marks put in parentheses in the relevant columns on table (7). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 b. *John-waK-dakeF -TOP -only (9) a. John-madeF-moK -even -also b. *John-moK-niadeF -also -even (10) a. John-bakariF-waK -only -TOP b. * John-w a K -b ak a r i F -TOP -only At the end of section 3.4. below, we will try to account for why this should be the case. 3.2.3. The nominal nature of F-p articles In both traditional and generative Hterature (e.g. M atsushita, 1930; Hashimoto, 1969; Fukui, 1986), the possibility of appearance inside (i.e. to the left oO a case particle has been taken as a firm testing stone for n o m i n a l i t y t2 As seen on table (7), F-partLcles, but not K-particles, may ^ 2 As noted above, F-p articles hke daJce only' can either precede or follow case particles. (i) a. mizu -o -dake nom-ta water-ACC-only drink-PAST Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 71 appear to the left of case particles like o ‘ ACC’, ni ‘ DAT’ , and kara “ OBL (from)’. However, we should not take this distributional fact for an argument th at only F-p articles can be nominal, because this fact itself calls for an account. In fact, there is some independent evidence that F-particles, but not K- particles, can be considered to be nominals. As noted by Miyagawa (1987), only nominal categories can be followed by the copulative da. (11) a. gakusei da (noun) student COP drank only w ater’ b. mizu -dake -o nom-ta water-only-ACC drink-PAST This fact motivated traditional linguists like M atsushita (1930) and Hashimoto (1969) to consider F-particles categoriaUy ambiguous; they recognized that while dake in (ia) is adverbial, th at in (ib) is nominal. Following these linguists, we take F-particles as categoriaUy ambiguous, being either Q-particles or (quasi-)nouns. As we wiU see, the latter status of F-particles aUow them to appear to the left of a case particle. InddentaUy, drawing on a similar fact, Fukui (1986: 22 Iff) argues th at the so-caUed question morpheme ka in Japanese is a noun. (ii) a. [[John-ga nani-o kaw-ta] ka] -ga mondai da -NOM what-ACC buy-PAST Ques-NOM problem COP The problem is w hat John bought.’ b. John-wa [[BiU-ga nani-o kaw-ru] ka] -o siritagarte i-ru -TOP -NOM what-ACC buy-PRES Ques-ACC want-to-know John wants to know what BUI is going to buy.’ Relying on the assum ption th at case particles can only attach to nouns, which requires an explanation, Fukui concludes th at ka is nominal. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 b. sizuka da (adjectival noun) quiet COP c. *tabe(-ru) da (verb) eat-PRES COP d. *taka(-i) da (adjective) high-PRES COP As exhibited in (11), nouns and adjectival nouns can safely be followed by the copula, but this is impossible for verbs or adjectives, with or without tense sufGx. However, if tensed verbs and adjectives receive an F-particle, they can appear immediately to the left of the copula, as shown in the following. (12) a. tabe-ru-bakarip da eat-PRES-only COP only eat' b. taka-i-dakep da high-PRES-only COP ‘ only h i ^ ’ On the contrary, K-p articles do not have this effect, as exhibited in the following. 1 3 Drawing on the evidence in (11) and others, Miyagawa (1987) claims th a t the Japanese counterparts of adjectives in English are adjectival nouns, in the sense th at they are categoriaUy specified as [+N, +V], but adjectives in Japanese lack specification for [±N]. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 (13) a. *tabe-ru-waK da eat-PRES-TOP COP b. *taka-i-moK da high-PRES-also COP Given th at da requires a nominal (i.e. [+N]) category immediately to its left, the contrast between (12) and (13) indicates that F-particles, but not K- particles, can behave as nominals. Furthermore, as demonstrated below, only F-particles can be immediately preceded by the attributive (or adnominal) form of an adjectival noun. (14) a. genki-na dakep-ga ziman da healthy only-NOM pride COP ‘ (He) is proud only of being healthy.’ b. *genki-na waK/moK ziman da healthy TOP/also pride COP (15) a. izyoo-na m ade? koohun-suru abnormal even excitement-do (He) is extraordinarily excited.’ b. *izyoo-na waK/moK koohun-suru abnormal TOP/also excitement-do Since it is established that adjectival nouns suffixed with -na can only be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 immediately followed by nouns, the contrast between a- and b-examples in (14) and (15) constitutes strong evidence for the nominality of F-p articles. 3.3. Case Particles In the preceding sections, we have established th at Q-particles m ust be classified into two subclasses: K-p articles and F-particles, and that only F- particles can be nominals. In this section, we will shed light on the nature of case particles firom a particular perspective of morphological case, and finally, we will account for the noted distributional facts of particles fiom the interaction of inherent properties of each class of particles and the morphological property of D. 3.3.1. Morphological case theory It has been generally assumed, since Chomsky (1981), that the distribution of argum ents is constrained, at least in part, by the theory of abstract Case, in the sense th at arguments m ust occupy positions in which they can receive abstract Case. However, drawing on facts of languages with rich morphological case and agreement, M arantz (1991) claims th at the proper treatm ent of morphological case, together with other independently motivated principles, will allow for the elimination of the theory of abstract Case. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 75 First, Marantz notes th a t in languages like Georgean, ergative case is found on subjects of transitive and unergative verbs in a certain series of tense, but not on subjects of unaccusatives.*'* The following examples from Georgian in (16) - (19) are firom M arantz (1991: 235). (16) a. vano [pikr-ob]-s marikaze (unergative) Vano(NOM) tbinka-INFLi Marika-on ‘ Vano is thinking about M arika.’ * '* According to Comrie (1989: llOff.), there are two basic patterns of case m ark in g in languages, schematically represented in the following (Linear order irrelevant). (i) Type One (nominative-accusative) a. S U B c a s e a v e r b O B J c a s e b (transitive) b. SUBcasea VERB (intransitive) (ii) Type Two (ergative-absolutive) a. S U B c a s e c v e r b O B J c a s e d (transitive) b. S U B c a s e d v e r b (intransitive) In Type One languages, subjects of transitive sentences and those of intransitive sentences share the same case m ark in g CASE A (=nominative), and objects of transitives bears a different case CASE B (=accusative). hi Type Two languages, on the other hand, objects of transitive sentences and subjects of intransitive sentences are marked by the same case CASE D, but subjects of transitives bear a distinct case CASE C; while CASE C is usually called ergative, CASE D is called absolutive (or simply, nominative). It is not the case th at all languages employ only one case marking system; many exhibit “spht ergativity”; for instance, while some languages exhibit Type One case marking in root contexts but Type Two case marking in embedded contexts, others shift between the two types according to tense (i.e. Georgean, discussed in the text, belongs to the latter type). For our purposes, however, it seems to suffice to define ergative languages as those which exhibit Type Two case marking in (ii) in certain contexts. See M arantz (1984a, chap. 6) for more detailed discussions of ergativity. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 76 b. vano-m [pikr-ob]-a marikaze (xmergative) Vano(ERG) thinka-INFLn Marika-on Vano thought about M arika.' (17) a. nino gia-s surateb-s [a-cven-eb]-s (transitive) NinoO^OM) GiaCDAT) pictures(DAT) showi-INFLi ‘ Nino is showing pictures to Gia.' b. nino-m gia-s surateb-i [a-cven-eb]-a (transitive) Nino(ERG) Gia(DAT) pictures(NOM) showi-INFLn ‘ Nino showed pictures to Gia.' As shown in (16a) and (17a), with unergative (class 3) and transitive (class 1) verbs in series I tenses (i.e. present, future, and others), Georgian exhibits the nominative-dative pattern (according to M arantz, dative and accusative are both realized as what is called dative in this language). On the contrary, we find the ergative-nominative (i.e. ergative-absolutive) pattern with the sam e classes of verbs suffixed with series II tenses (i.e. aorist or simple past), as exhibited in (16b) and (17b). However, this ergative-nominative pattern does not arise even in the aorist, if the verb belongs to the class of unaccusative verbs (class 2), which, hke passives, have syntactically derived subjects. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 77 (18) a. essaxl-i ivane-s a-[u-sendeb]-a (unaccusative) this house(NOM) Ivan(DAT) preV-builtrlNFLi This house will be built for Ivan.’ b. es saxl-i ivane-s a-[u-sendeb]-a (unaccusative) this house(NOM) Ivan(DAT) preV-builtz-INFLn This house was built for Ivan.’ For series I tenses, unaccusative verbs, like unergatives and transitives, take nominative subjects, as exhibited in (18a). What calls for an explanation is that even in the aorist, subjects of unaccusative verbs rem ain nominative, but they do not become ergative, as shown in (18b). The examples in (19) show that psych predicates (class 4), which take dative subjects and nominative objects, do not change the case marking pattern in the aorist. (19) a. sen pelamus-i g-[i-qyar]-s Opsych) you(DAT) pelamusi(NOM) AGR-hke4-INFL[ Tbu like pelamusi.’ b. sen pelamus-i g-[e-qyar]-e ^sych) you(DAT) pelamusi(NOM) AGR-like4-INFLn Tou liked pelamusi.’ According to Marantz, psych predicates (class 4) in Georgian are hke unaccusatives (class 2), in that their experiencer subjects are derived by Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 78 movement from a V P-intem al position. In Hindi and Basque as well, ergative case is not found on subjects of unaccusative verbs, as exhibited in the examples below. (20) a. süta(*-ne) aayii (unaccusative) Sita.F(*-ERG) arrived/came.F b. kuttoN-ne bhoNkaa (unergative) dogJPLM-ERG barked.SG.M c. raam-ne roTü khaayii thii (transitive) RamJVl-ERG breadJ* eat.F be.PAST.F (Hindi data from M ahajan (1990), cited by M arantz (1991: 236)) (2 Da. Ni etoni naiz (unaccusative) IJiBS come be.l.SG b. Nik Ian egin dut (unergative) I.ERG work do have.l.SG c. Nik libura ekarri. dut (transitive) I.ERG book.ABS bought have.l.SG (Basque data originally from Levin (1983), cited by M arantz (1984b: 65-66)) It has been widely accepted since BeUetti and Rizzi (1988) that psych verbs are uniformly base-generated with two internal arguments, i.e. experiencer and theme, in this hierarchical order, but they do not assign an external th e ta role. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 79 In either language, ergative case is not allowed on subjects of unaccusative verbs, although it is allowed or even obligatory on subjects of transitives and u n e r g a t i v e s . M arantz puts forth the following generalization concerning the distribution of ergative case. (22) Ergative case generalization M arantz, 1991:236): Even when ergative case may go on the subject of an intransitive clause, ergative case wül not appear on a derived subject. If (abstract) Case to the subject is always assigned by tense (or INFL), as proposed by Chomsky (1981, 1986a, b), the generalization in (22) rem ains a puzzle, M arantz notes th at whQe Basque shows ergativity with any tenses, Hindi shows it with perfect tense/aspect, and th a t ergative case on the subject of unergatives in Hindi is optional. That is, the subject of (20b) can appear without the ergative marker -ne, as in the following. (i) kutte bhoNke dog.PL.M barked.PL.M As noted by M arantz (1991), standard theories of abstract Case require a stipulation th a t unlike lexical DPs, PRO does not need abstract Case. However, Icelandic PRO, as shown by Sigurdsson (1991), needs morphological case in the sam e way that lexical DPs do. According to Sigurdsson, any morphological case of floating quantifiers m ust be licensed by an identical case on DP, as exhibited in the following (firom Sigurdsson, 1991: 331). a. strak am ir komust a l lir i skola the boys(NOM) got aU(NOM.PL.M) to school The boys aU managed to get to school.’ b. strak an a vantadi a l ia £ skolann the boys(ACC) lacked alI(ACC.PLJVO in the school The boys were all absent firom school.’ c. stràkunum leiddist o llu m £ skola the boys(DAT) bored all(DAT.PLJV[) in school Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 80 The boys were all bored in school.’ c L strakanna var a llr a getid i rædunni the boys(GEN) was all(GEN.PL.M) mentioned in the speech The boys were all mentioned in the speech.’ The floating quantifier a llir ‘ all’ agrees with the subject in case, person, number and gender in each sentence of (i), a lth o u ^ the subject is iu “quiriy” case in (ib-d). Now (i) should be compared with (ii) (firom Sigurdsson, 1991: 331-2), where allir appears in infinitival complement clauses. (ii) a. strakam ir vonasttil [ad PRO kom ast a llir i skola] the boys(NOM) hope for to (NOM) get all(NOM) to school b. strakam ir vonast til [ad PRO vanta ekki a lia i skolann] the boys(NOM) hope for to (ACC) lack not all(ACC) in the school c. strakam ir vonasttil [ad PRO leidast ekki o llu m £ skola] the boys(NOM) hope for to (DAT) bore not all(DAT) in school d. strakam ir vonast til [ad PRO verda a llr a getid £ rædunni] the boysCNOM) hope for to (GEN) be aU(GEN) mentioned in the speech The floating quantifier in each of (üa-d) must bear the same form as in the corresponding finite clause in (ia-d). If the floating quantifier indeed m ust agree in morphological case with the complement subject in (iia-d), PRO in Icelandic must bear morphological case in the same way th at lexical subjects do. Chomsky and Lasnik (1993) claim that PRO in fact needs abstract “null” Case. However, since this “null” Case is, obviously, not good enough to license lexical DPs, its postulation does not help to account for the Icelandic facts shown above. It is, incidentally, impossible to argue that the controller serves as a licenser for morphological case on the quantifier, because it would end up in case mismatch in (iib-d). Also as noted by M arantz (1991), Icelandic facts as in the following seem to defy any non-ad hoc explanation firom theories of abstract Case. According to Zaenen, et al (1990: 112), for ditransitive verbs th at exhibit the NOM-DAT- ACC case array, either of the intem al arguments can be passivized, as exemplified in the following. (i) a. am bàttin var gefin konunginum the slave(NOM.SG) was given.SGJF the king(DAT) The maidservant was given to the king.’ b. kon unginum voru gefiiar am bàttir the king(DAT) were given.PL.F slaves(NOM.PL.F) The king was given m aidservants.’ In (ia), the theme object, which receives ACC in the active, is passivized and it Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 M arantz (1991) also discusses a widely known generalization concerning object case. It h as been widely accepted th a t objects may not receive accusative in clauses th at do not have them atic subjects, a generalization known as Burzio’ s generalization (cf. Burzio, 1986). This generalization is empirically m otivated by examples hke the following. (23) a. *it broke the glass b. *it was killed a student (in the riot) Generally, objects of unaccusative and passive sentences may not rem ain in their base position in which they receive theta roles &om predicates, as shown in (23). In such a case, objects must move to subject position, as in (24). (24) a. the glass broke b. a student was killed (in the riot) On the contrary, in active transitive sentences, objects can remain in their base positions, as indicated in (25). (25) a. they broke the glass b. they killed a student (in the riot) If movement of objects in (24) is indeed motivated by the requirement of abstract Case, comparison between (23) and (25) will naturally lead one to gets NOM. On the contrary, the goal argument undergoes passive in (ib), and it retains its DAT (i.e. “quirky” case). This indicates th at movement for passive in (ib) is not motivated by the need for abstract Case. What is even more surprising is, however, that the rem aining object in (ib) gets NOM, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 something like B u rao ’ s generalization. In fact, the contrast between (23) and (25) can only be attributed to the presence and absence of a thematic role assigned to the subject position. However, M arantz (1991) points out another possible array to account for the ungrammaticality of cases like (23). (23a, b) are derived from (26a, b), respectively, by insertion of an expletive to the empty (i.e. non-tbematic) subject position. (26) a. ____broke the glass b. ____ was killed a student (in the riot) The use of expletives in languages like Engbsb is m otivated by the extended part of the projection principle (EPP), to the effect th at clauses m ust have a subject (cf. Chomsky, 1982). If, as suggested by M arantz, the insertion of an expletive is more costly than movement of object to subject, then the principle of economy of derivation will prefer (24a, b) over (23a, b). Thus, M arantz claims that (23a, b) can be excluded without reference to abstract Case.^^ instead of ACC, and the inflected participle agrees with it. The comparison th at M arantz makes between the cost of expletive insertion and th at of movement is not well grounded with respect to the principle of economy in Chomsky (1995, chap. 4). Since expletives are independent lexical item s (Us), the derivation for (23a), for instance, should have a different num eration, and therefore, a different reference set, from the derivation for (24a). This m eans th at these derivations cannot be compared in terms of economy. The same point is further illustrated by pairs of examples like the following. (i) a. there arrived a guest Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 Now m atter w hat im portant fact Burzdo’ s generalization captures and no m atter how ingeniously it does so, it is, after all, a mere observational generalization, which has no explanatory force. Rather, it would have to be derived from some general principles. Unfortunately, however, a theory of abstract Case like Chomsky’ s (1981, 1986a, b) has not been successful in reducing Burzio’ s generalization to any independently motivated principles in an insightful w a y . 20 b. a guest arrived (ii) a. (?)there was killed a student in the riot b. a student was killed in the riot If M arantz were correct in th at expletives are introduced in the course of a derivation, and their insertion is more costly than movement, (ia) and (iia) should be blocked by (ib) and (iib), respectively, which is an unfavorable result. The principle of economy should be made to allow pairs like (ia, b) and (iia, b). Although his economy account does not seem to be tenable, the existence of examples like (ia) and (iia) partially supports M arantz’ s main thesis th at the distribution of DPs cannot be accounted for by abstract Case. In examples hke (ia) and (iia), the object remains in situ , where it should not receive Case under the conventional Case theory (but see BeUetti (1988) for an argument th at postverbal DPs as in these cases are assigned partitive Case by INFL). Chomsky’ s (1995, chap. 4) minimaUst theory of Case has dissociated Case from overt A-movement, in reference to the existence of expletive constructions in Enghsh like (ia) and (iia) (as weU as transitive expletive constructions in Icelandic). Chomsky claims th at what triggers NP- movement is the strong D-feature in T, which has to be eliminated because it is uninterpretable; the strong D-feature in question can be checked off by there in cases like (ia) and (iia). Note th at this is the minimahst interpretation of the EPP. However, Chomsky stiU m aintains that Case features of NPs/DPs, which he considers to be also uninterpretable, m ust be eliminated by covert feature movement to T. 20 M arantz (1991) casts doubt on Burzio’ s generalization for its own validity. However, the counterevidence that he draws from Japanese indirect passives, Kichaga symmetric double object applied verbs and English expletive/raising Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 84 It has been made clear above th at while the conventional theory of abstract Case cannot give a principled account for Ergative generalization or Burzio’ s generalization, distributional facts of DPs as exhibited in (23) can be explained, at least in part, by some other independent principles Hke the EPP (i.e. subject requirement). M arantz (1991) argues th a t if Hcensing of DPs is teased apart from the theory of abstract Case, what rem ains to be necessary is a theory of morphological case.^* M arantz (1991) thus claims th at while the distribution of DPs should be accounted for by some formal principles of grammar including the EPP, realization (or assignment) of case is a m atter of morphology. As a theory of morphological case for such a purpose, he proposes that, in a given language, each noun phrase is generated with a CASE feature which m ust be spelled constructions with verbs like strike does not seem to be compelling. The point in the text discussion is, however, that the content of Burzio’ s generalization, to the extent th at it holds, must be accounted for. 2 1 Obviously, the text discussion remains silent concerning agreement. Noting the fact that in Georgian, even sentences in the aorist tense which exhibit the ERG case pattern show the NOM-ACC pattern in agreement, M arantz suggests that agreement is independent of morphological case (or abstract Case). Furthermore, it is not clear how languages Hke Chinese th at exhibit no agreement or case morphology at all are dealt with. It might be the case th at a theory of abstract Case m ust be maintained for Hcensing of argument DPs in those languages. This is a significant issue to be dealt with in future research. We will assume, in what foUows, that case is a m atter of morphology, at least, in languages Hke Japanese as well as Georgian, along the lines of M arantz (1991). Kuroda’ s (1978) linear case marking system can be taken as a proposal in this direction. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 85 out at the morphological component on the PF branch of a derivation (i.e. his Morphological Structure (MS)) .22 The realization of a CASE feature proceeds in accordance with the hierarchy in (27). (27) CASE realization disjunctive hierarchy (Marantz, 1991: 247) i. lexically governed case (inherent or “quirky” case) Ü. dependent case (accusative and ergative) ill unm arked case (environment-sensitive) iv. default case M arantz proposes the four distinct modes of case assignment in (27i-iv). Lexically governed case (271) is often called by different names (e.g. inherent case, “quirky” case and idiosyncratic case, among others), and it is assigned to an argument th at bears a specific them atic relation with a predicate (cf. Chomsky, 1986b). As noted by Zaenen and Maling (1990), lexically governed case is, typically, preserved under passivization. As will be made clear shortly, dependent case (27.ii) is assigned to a thematic position only in reference to another them atic position. Although M arantz does not elaborate on unm arked case assignment (27iü) or default case assignm ent (27.iv), these two modes of case assignm ent seem to be employed only when neither lexically governed case assignment nor dependent case assignm ent is 22 For the architecture of grammar th at M arantz proposes, see the series of his works, e.g. M arantz (1988, 1989, 1991) and Halle and M arantz (1993). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 available. The four modes of case assignment in (27i-iv) are hierachicaUy ordered, and the CASE feature in a given DP can be case m arked only once. If a CASE feature employs a certain mode of case assignm ent which morphologically spells it out, then the rest of the modes lower in the hierarchy become simply irrelevant. Noting th at the possibilities of ergative case on the subject and accusative case on the object depend on the thematic structure of the predicate, M arantz comes up with a generalization concerning the assignment of dependent case. His generalization can be stated for our purpose in the following way..^ (28) Dependent case is assigned to a thematic position in a given domain (e.g. a finite tense) when there is a distinct unm arked thematic position in the same domain. “Unmarked” in (28) means “not case marked yet”. Suppose th at a finite tense defines a domain of case assignment Oience, TP constitutes a domain to which the hierarchy in (27) apphes), then, if dependent case is to be assigned to one DP in a given TP, (28) requires the presence of a second DP within the same TP. Furthermore, there are two relevant cases in which, th o u ^ two DP positions are involved, dependent case assignment is prohibited by (28). 2 3 M arantz seems to assume V-to-I movement in all languages and states the condition on dependent case assignment accordingly. However, the condition Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 87 (29) a. [t p ... DP2* ... DPi ... ] (DP2 assigned inherent case) Î . b. [t p ... QPi ... [d p i £ ]* ... ] ( [ d p i e] trace of DP i) f In (29a), there are two DPs within TP. While DP2 is already assigned an inherent case (indicated by the asterisk), DPi is yet to be case marked, hi this case, DPi cannot be assigned dependent case, since DP2 is already case marked; hence, DPi is prohibited firom making reference to DP2 by (28). (This state of affairs is indicated by the arrow with an asterisk.) In (29b), DPi is related to its trace by movement within a single TP. In such a case, even if the trace is in a thematic position, DPi cannot refer to it. This is because, in terms of (28), the trace position does not count as distinct firom DPi (i.e., they are two b'nk.s in a single chain). Although two DP positions appear within one TP in (29a) and (29b), DPi cannot be assigned dependent case in either of the cases.24 Furthermore, according to Marantz, there are two param etric options concerning the directionality for dependent case to look for a reference can easily be rephrased without reference to V-to-1 movement, as in (28). ^ According to Chomsky (1981: 336£f.), the effects of the Case Filter follow firom the theta criterion on the assumption that Case m akes an argument chain “visible” for the purpose of theta marking (cf. Aoun, 1985: 76). However, note th a t the proposed morphological case theory takes this derivational relation in such an opposite way that theta role assignment Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 88 position. While ergative languages set the value of the param eter as (30a), (nominative-)accusative languages set it as (30b). (30) a. Dependent case assigned upward to subject is ERG. b. Dependent case assigned downward to object is ACC. Recall that while ERG goes with subjects of transitives (and unergatives), but not with subjects of unaccusatives or passives, as indicated in (31a, b), ACC goes with objects of transitives, but not with those of unaccusatives or passives, as schematized in (31c, d). (In (31), the arrow notation is intended to indicate the direction of “reference” that a DP makes to another DP in view of (28).) (31) a. Transitives (Unergatives) b. Unaccusatives/Passives ERG* T iRGok makes an argument chain “visible” to morphological case marking. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. Transitives d. Unaccusatives/Passives 89 TP (expletive) CC ACC ok In ergative languages, the subject of transitive sentences can be marked with ERG, as indicated in (3 la). This is because the subject can m ake reference downwards to a them atic object (i.e. DP3). In languages like Hindi and Basque where ERG also goes with subjects of unergatives, DPs in (31a) m ust still be visible. According to Hale and Keyser (1993), unergatives are underlyingly transitive, but they incorporate them atic objects. If this is the case, it is not surprising th at subjects of unergatives can still see DPs in (31a) for reference. On the other hand, the subject of unaccusative sentences is derived from a VP- intem al position, as indicated in (3 lb). Since this derivation constitutes a chain, the intem al position does not count as a distinct position in opposition to the derived subject, in term s of (28). Hence, the subject of unaccusatives cannot be assigned ERG. In languages in which Burzio’ s generalization holds (i.e. accusative Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 languages), object case is a dependent case. Hence, the object needs another unmarked thematic position upwards for reference. In transitive sentences schematically represented in (31c), such a position (i.e. DPg or the chain th a t it heads) is indeed available, the object can be assigned ACC. If, on the other hand, such a position is unavailable as in unaccusative or passive sentences represented in (3 Id), the object cannot get ACC. Insertion of an expletive in subject position does not help, since this position is not thematic. If insertion does not take place, the object has to move to subject position to satisfy the EPP. In this case, the trace of the object cannot make reference to subject position now occupied by an argument, because it belongs to the sam e chain and it does not count as distinct, in the sense of (28). Thus, the morphological case theory adopted here provides a possible accout for Burzio’ s generalization. 3.3.2. Case marking in Japanese M arantz’ s (1991) morphological case theory, which we adopt in this thesis, can also account for facts of the permissible and impermissible case arrays in Japanese. As noted by Kuroda (1978: 35), three basic case m ark in g patterns are found in Japanese. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 91 (32) a. (Di)traiisitive: NOM-(DAT)-ACC John-ga Mary-ni hon-o age-ta -NOM -DAT book-ACC give-PAST ‘ John gave a book to M ary/ b. Ergative: DAT-NOM John-ni tetugaku-ga wakar-u -DAT philosophy-NOM understand-PRES ‘ John understands philosophy/ c. Intransitve: NOM Mary-ga hasir-ta -NOM run-PAST M ary ran / Given th at Japanese is an accusative language, and that ni T)AT is lexically governed (or inherent) case, o ACC, dependent case, and ga ‘ NOM’ unm arked (or default) case, in the sense of the CASE realization disjunctive hierarchy in (27), it can be naturally explained why three paradigmatic case patterns— NOM-(DAT-)ACC ((di)transitive), DAT-NOM (ergative) and NOM (intransitive)—are allow ed.® (In (33), the arrow notation is employed for the ® InKuroda’ s (1978: 35) terms, those Canonical Sentence Patterns are m eant to function as filters in the sense of Chomsky and Lasnik (1977). If the case array on a given cycle th at is determined by the application of his Linear Case Marking does not match any of the patterns, th at case array will be filtered Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 92 same purpose as in (31).) (33) a. (Di)transitive P attern b. “Ergative” Pattern T T VP V * bhn y wakar ACC* age ACC ok Intransitive P attern VP -ta hasir M ary In ditransitive sentences like (32a), the indirect object is inherently case marked with m ‘ DAT’. Since the direct object can still make reference to another unmarked thematic position, i.e. subject, as indicated in (33a), it gets o ‘ ACC’. On the contrary, since the subject has no unm arked position out. See relevant discussions below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 upwards for reference, it gets unm arked (or default) ga ‘ NOM’. Suppose th a t so-called stative predicates (or “ergative” predicates) like w akar ‘ understand’ optionally takes a non-theme (presumably, experiencer) argument as well as a theme argument, and it assigns inherent case n i to the former. Then, the theme argument tetugaku philosophy’ in (32b) does not have any unm arked thematic position for reference, as indicated in (33b); the non-theme argument is invisible, because it is already m arked (cf. (28a)). Intransitive predicates as in (32c) take only one argument, as seen in (33c).^ Since this sole argument has no thematic position for reference, it can only get unmarked (or default) ga. The adopted morphological case theory makes several interesting predictions. First, it is predicted th a t if a mono-transitive predicate assigns inherent case (i.e. ni ‘ DAT) to one of its arguments, the other argum ent may not receive ACC; in other words, DAT-ACC array should be impossible. This is in fact home out, as demonstrated in the following. 2 ® The tacit assumption behind the text discussion here is th at surface intransitives in Japanese are all underlyingly unaccusatives, or all unergatives have alternative lexical representations as unaccusatives. Hale and Keyser (1993) argue that unergatives are underlyingly transitives whose internal argument is incorporated to the head of the lower verbal shell. However, no strong argument has been provided in the hterature for any class of verbs in Japanese th a t must solely be unergatives. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 (34) a. John-ni tetugaku-ga wakar-u (=32b) -DAT philosophy-NOM understand-PRES ‘ John understands philosophy.’ b. * John-ni tetugaku-o wakar-u -DAT philosophy-ACC understand-PRES Stative predicates like wakar *understand’ can be transitive, and they can assign the inherent ni to the non-theme argument. Hence, the theme argument, th o u ^ aligned lower than the non-theme (cf. Grimshaw, 1990; Takano, 1996), does not have an unmarked them atic position upwards for reference, and therefore, it m ust get NOM, as shown in (34a). The dependent ACC cannot be assigned to the theme object, as indicated by the ungrammaticahty of (34b), since the non-theme argument, which is already assigned an inherent case, does not count as a distinct position for the purpose of (28) .2^ 2 7 Stative predicates like wakar ‘ understand’ allow NOM-NOM, as well as DAT-NOM, but never NOM-ACC, as indicated in the following. (i) a. John-ga tetugaku-ga wakar-u -NOM philosophy-NOM understand-PRES b.. ?*John-ga tetugaku-o wakar-u -NOM philosophy-ACC understand-PRES For the sake of the text discussion, we assume th a t stative predicates Like wakar are only optionally transitive; however, once they are transitive, they assign inherent case to the non-theme argument, as in (34a). In (ia), the predicate is intransitive, and it only takes the them e argument tetugaku ‘ philosophy’; the first nominative phrase John-ga is a major subject (or focus, in the sense of Saito (1982)). Since the major subject does not occupy a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 95 Next, while the absence of ga NOM’ in certain classes of sentences poses a serious problem to theories of abstract Case that assume th a t the finite tense in Japanese m ust assign NOM (at least once), the adopted morphological case theory does not face such a problem. Takezawa (1987) argues th at DAT-ACC array as in (34b) is excluded because, in such a case, finite tense should fail to discharge its nominative case assigning feature. His contention is based on Shibatani’ s (1977, 1978a) observation th at NOM must be realized at least once for every finite clause in Japanese. In support of his own claim, Takezawa points out the fact th at examples like (35a) do not sound quite right to the ear of Japanese speakers. (35) a. John-ni hatarak-e -ru -DAT work-POT-PRES John can work.’ b. John-ga hatarak-e-u -NOM work-POT-PRES John can work.’ of. John-ga hatarak-u -NOM work-PRES John works.’ thematic position, the (sole) theme argument cannot refer to it for the purpose of dependent ACC assignment, as indicated by the unacceptability of (ib). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96 In (35a), the sole argument is assigned DAT by the potential auxiliary (rar)e. When it is assigned NOM instead, the sentence becomes perfect, as in (35b). Hence, Takezawa attributes th e oddity of (35a) to the absence of NOM in th at sentence. Although Takezawa is correct in th at sentences hke (35a) sound highly awkward, it is not entirely clear i f their status should be attributed to any grammatical principles. For instance, i f the DAT argument is topicahzed, the acceptabihty remarkably increases, as exhibited in the following.^s (36) a. John-ni -wa mada hatarak -e -ru -yo! -DAT-TOP stm work-POT-PRES-EXCL ‘ John can still work!’ b. watasi-ni-wa sonnnani hayaku hasir-e-masen 1 -DAT-TOP th at fast run-POT-NEG 1 cannnot run that fast.’ Since topicalization of the DAT phrase does not “m ask” potential assignment 2 8 (36b) is cited from McGloin (1980), who argues for pragmatic conditions th at make sentences like (36b) more acceptable than those like (35a). In her terms, addition of demonstratives hke sonnnani ‘ (to) th at (degree)’ increases the discourse anaphoridty (in the sense of Kuno (1973)) of the predicate, to the extent th at the predicate becomes the thematic topic of the sentence, and the sole DAT phrase, which can hardly be taken as a them atic topic even when topicahzed with wa, can rem ain contrastive; as a result, the acceptabihty of the sole DAT phrase increases. However, our text discussion does not rely on the vahdity of McGloin’ s argument. See also Inoue (1985) for relevant discussions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 of NOM, neither of th e sentences in (36) assigns NOM.^s Nevertheless, the sentences are both highly acceptable. Furthermore, it is not clear, either, if Shibatani’ s observation is empirically sound. As noted by Kuroda (1988: I IBflf), Japanese seems to have weather verbs, though small in number, such as hubuk ‘ to snowstorm’ and sigure to shower’ , neither of which does not seem to assign any them atic role, as exemplified in the following. (37) a. hubuk -te ki-ta snowstorm-PERF come-PAST ‘ It has begun to snowstorm.’ b. sigure-te ki-ta shower-PERF come-PAST ‘ It has begun to shower.’ These sentences do not require any discourse-anaphoric or pragmatic endeavor to supply m issing information. Although they lack thematic arguments, and therefore, they do not have NOM realized, they are perfectly 2 9 As noted by Shibatani (1977, 1978a), topicalization may “m ask” assignment of NOM. Compare (35a, b) with the following. (i) John-wa2 tg/ pro 2 hatarak-e-ru -TOP work-POT-PRES ‘ John can work.’ Shibatani would have to say that in (i), NOM is not realized but it is still assigned; it is assigned to pro if topicalization is base-generation, or to the chain headed by the topic, if topicalization involves movement. However, we will argue below th a t in cases like (i), Q-particles like wa T O F give phonetic content to a CASE feature in D, so that ga ‘ NOM’ , as an unmarked (or default) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 98 interpretable in their own right. Note that the morphological case theory adopted here does not exclude examples like (36a, b) or (37a, b). In (36a, b), the unique argum ent is inherently case marked, and in (37a, b), there is simply no argum ent DP, therefore, no CASE feature, to enter into the case assignment disjunctive hierarchy in (27). Furthermore, the morphological case theory assumed here does not require unnatural assumptions of the sort that Kuroda’ s (1978) linear case- marking system is forced to make. First, although there seems to be httle independent evidence for overt predicate raising in Japanese (as pointed out in chapter 2), Kuroda has to assume that it takes place in a rather convenient way. Kuroda (1978) proposes that in Japanese, ga ‘ NOM’ and o ‘ ACC’ are assigned in accordance with the case-marking system in (38a), and n i ‘ DAT’ on a subject is assigned by a specific rule in (38b). (38) a. Linear Case Marking (LCM) (Kuroda, 1978: 34) M ark the first unmarked noun phrase with ga, and m ark any other unm arked noun phrase or phrases with o. b. Subject M -M arking (based on Kuroda, 1978: 30-31) M ark the subject of an embedded clause with ni iff the predicate of th at clause raises to the predicate of the immediately higher clause. case, need not be employed. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 99 According to Kuroda, LCM applies to sentences cyclically in a bottom-up fashion, and if the predicate of one cycle raises to th at of the immediately higher cyde, the result of LCM on the former cyde is overridden by its application to the latter. To put it differently, if predicate raising takes place, the domain of application of LCM extends, and in such a case, ga “ NOM’ and o ‘ ACC’ assigned on the immediately lower cyde are erased when LCM apphes to the new cyde. First, reconsider the stative sentence with the permissible DAT-NOM case array in (34a) and th at with the impermissible NOM-ACC in (34b) . ((34a, b) are reproduced below as (39a, b), respectively.) (39) a. John-ni tetugaku-ga wakar -u (=34a) -DAT philosophy-NOM understand-PRES ‘ John understands philosophy.’ b. ?*John-ga tetugaku-o wakar-u (=34b) -NOM philosophy-ACC understand-PRES c. [cycie2 [cyciei John-ni tetugaku-ga tv] wakar -ov -u] -DAT philosophy-NOM understand -?? -PRES d. ([cyde2 )[cyciei John-ga tetugaku-0 wakar Q -ov) -u] -NOM philosophy-ACC understand -?? -PRES If his LCM is to work for cases like (39a), Kuroda has to stipulate th at those apparently simplex sentences like (39a) which exhibit the DAT-NOM case Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 100 array are indeed complex, and further th a t predates like w akar ^understand’ in the embedded clause raise to the empty predicate in the m atrix clause (or vice versa). This state of affairs is indicated in (39c), where w akar raises to 0 v, and as a result, the “embedded” subject John is assigned n i by Subject Ni- Marking; moreover, the apphcation of LCM on Cycle 2 assigns ga to the theme argum ent of wakar, since it is the first unmarked DP on th a t cyde. Thus, Kuroda’ s account for the DAT-NOM case array in sentences like (39a) rehes on two separate assumptions, nether of which seems to be well estabhshed; on the one hand, sentences like (39a) must indude sentential complementation, and on the other, predicate raising must take place in those sentences. Unless both assumptions turn out to be correct, LCM would generate the NOM-ACC case array on Cyde 1 as shown in (39d), which leads to the impermissible surface form in (39b).^ Now compare the stative sentence in (39a) with Kuno’ s celebrated example of the multiple nominative construction in (40a). ^ The two pairs of parentheses in (39d) are m eant to coindde. Note th at if the sentence in (39d) is two-storeyed and predicate raising does not take place, Cyde 1 will not be induded in the domain of LCM s apphcation on Cyde 2. In such a case, the apphcation of LCM on Cyde 2 wül be vacuous, and the case marking on Cyde 1 wiU be kept intact; hence, NOM-ACC will result. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 (40) a. [s 3 bunmeikoku-ga [s2 dansei-ga [si beikmzyumyoo-ga mizikai]]] civilized countries-NOM men-NOM average life span-NOM short I t is civilized countries in which the average life span of men is short/ b. *[s3 bunmeikoku-ga [s2 dansei-ga [si heikinzyumyoo-ni mizikai]]] civilized countries-NOM men-NOM average life span-DAT short c. *[s3 bunmeikoku-ga [sz dansei-ga [si heikinzyumyoo-o mizikai]]] civilized countries-NOM men-NOM average life span-ACC short d. *[s3 bunmeikoku-ga [sz dansei-o [si heikinzyumyoo-o mizikai]]] civilized countries-NOM men-ACC average life span-ACC short For Kuroda, each of the three subjects in (40a) belongs to a different cycle with respect to LCM. For this to be established, he has to assume that predicate raising does not take place in (40a) in the same way that it does in (39a); otherwise, the subject of the predicate m izik a i ‘ short’ would be assigned ni, as shown in (40b), which is an unfavorable result. However, under the theory adopted here, the fact of multiple ga m arking in cases like (40a) will follow ffom a simple assumption that all nominative phrases except for the most internal one are adjuncts; i.e. they do not occupy thematic positions.^i Then, since none of th e three subjects in (40a) has a distinct thematic position upwards for reference, they cannot be assigned dependent case (i.e. o The point here is th a t major subjects are not theta m arked in the usual sense; they are licensed either by “aboutness”, in the sense of Chomsky (1981), Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 ‘ ACC” ), as exhibited by the ungrammaticahty of (40c, d); instead, the only available case is the unmarked (or default) one (i.e. ga ‘ NOMO In addition, the adopted theory need not recourse to “Counter Equi” (cf. Harada, 1973), which Kuroda has to assiune to account for the case arrays found in ni-causatives, as opposed to o-causatives, as shown in the following. (41) a. John-ga Mary-ni (gakkoo-e) ik-ase-ta (ni-causative) -NOM -DAT school-to go- CAUS -PAST ‘ John made/let Mary go to school’ b. John-ga Mary-o (gakkoo-e) ik-ase-ta (o-causative) -NOM -ACC school-to go-CAUS-PAST ‘ John made Mary go to school.’ Kuroda assumes with many other grammarians th at Japanese causatives, whether ni- or o-causatives, involve sentence complementation. For him, while the causee in n i originates in subject position of the complement clause, the causee in o is generated in the m atrix clause. When the causee in o appears as in (41b), subject of the complement clause is deleted by Straight Equi (or controlled, in more recent terms). On the contrary, when the causee in ni surfaces as in (41a), the causee in the matrix clause (which is to be assigned o by LCM) m ust be deleted under identity with the causee in n i by “Counter Equi”, the theoretical status of which is not necessarily firm. or by predication, as suggested by Heycock (1993). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 103 However, assuming th at the causative verb (s)ase selects for VP, but not vP Oience, no external argument of the embedded verb is projected), and it optionally takes a causee, we can naturally derive the attested case arrays in the two types of causatives. Under our approach, the causative sentences in (41a, b) are represented as (42a, b), respectively. (42) a. [r [vP John-ga [vP 2 M ary 2-ni [vpi ecz (gakkoo-e) ikv] (s)asev] ta] -NOM -DAT school-to go CAUS PAST b. [r[vP John-ga [vP 2 [vpi Mary-o (gakkoo-e) ikv] (s)asev] vj ta] -NOM -ACC school-to go CAUS PAST Since the embedded verb ik ‘ go’ is an intransitive (or unaccusative) verb, its subject is projected within the lower VP. In (42a), this argument of the embedded verb is controlled by the causee in ni, which is projected by the causative verb. It is not surprising if a particular argum ent of a verb is inherently case marked; the causee, once projected, is always assigned DAT by the causative verb. In (42b), on the other hand, the causative verb does not take a causee argument, but the subject of the lower verb is instead realized. Because this argument of the lower verb can refer to the causer argument of (s)ase within the domain of the same tense, it gets dependent ACC. Needless to say, in either case, the causer argument can only get the unmarked (or default) NOM. As is well known, a transitive verb with an ACC object can be embedded Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 under ni-causatives as in (43a), but not o-causatives as in (43b). The la tte r is often excluded by a surface filter called the Double-O Constraint (c£ Miyagawa, 1989b). (43) a. John-ga Mary-ni hon-o yom-ase-ta -NOM -DAT book-ACC read-CAUS-PAST ‘ John made/let Mary read the book.’ b. * John-ga Mary-o hon-o yom-ase-ta -NOM -ACC book-ACC read-CAUS-PAST John made Mary read the book.’ However, as pointed out, correctly, by Kuroda (1988), Takano (1996) , Hoshi and Saito (1997), among others, (43b) is much worse than cases of “mild” violation of the constraint hke (44a, b). (44) a. ??Mary-ga John-o hoho-o but-ta -NOM -ACC face-ACC hit-PAST M ary slapped John on the face.’ b. ??Bill-ga ano ziken-o (kenmei-ni) soosa-o si-ta -NOM th at case-ACC earnestly investigation-ACC do-PAST ‘ Bill earnestly investigated th a t case.’ Although the sentences in (44a, b) each contain two occurrences of ACC, they exhibit much milder deviation than (43b) does. The crucial difference between (44a, b), on the one hand, and (43b), on the other, hes in th e fact th at Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 105 while the two arguments in ACC in the former seem to share one theta role assigned by the predicate (i.e. b u t t i t ' in (44a) and su ‘ do’ in (44b)), those in the latter are assigned different theta roles (i.e. agent for John and theme for bon ‘ book’ ). More importantly, as first noted by Kuroda (1988), the deviation found in (44a, b) can be ameliorated by separating one ACC firom the other as in cleft sentences in (45a, b), but the same operation will not improve the deviation found in (43b). (45) a. Mary-ga John-o but-ta no wa hoho-o da -NOM -ACC hit-PAST COMP TOP face-ACC COP ‘ (lit.)What Mary slapped John was (his) face.’ b. Bdl-ga (kenmei-ni) soosa-o si-ta no wa -NOM earnestly investigation-ACC do-PAST COMP TOP ano ziken-o da that case-ACC COP ‘ What John earnestly investigated was th at case.’ (46) a. * John-ga Mary-o yom-ase-ta no wa hon-o da -NOM -ACC read-CAUS-PAST COMP TOP book-ACC COP ‘ What John made Mary read was the book.’ b. * John-ga hon-o yom-ase-ta no wa Mary-o da -NOM book-ACC read-CAUS-PAST COMP TOP -ACC COP The one who John made read the book was Mary.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 106 It seems th a t while a surface filter hke the Double-0 Constraint is appropriate for marking the relatively weak deviation of examples hke (44a, b), the much more serious deviation found in (43b) should be accounted for by some core principle of the grammar. Kuroda (1978: 37) proposes to exclude cases hke (43b) by the Canonical Sentence Patterns; since o-causatives with transitive verb stems th a t aUow another occurrence of o on the theme argument do not match any of the canonical sentence patterns in (33), they should be filtered out. However, this “filtering” approach does not have any explanatory adequacy. A question naturally arises as to why this should be the case. In our approach, on the contrary, this question does not arise, since o-causatives like (43b) cannot be generated firom the beginning. The representation for (43a), after inherent case assignment but before dependent case marking, should look hke the foUowing. (47) [r [vP John [vp 2 Mary-ni [v p i hon yomv] (s)asev] v] ta] -DAT book read CAUS PAST ‘ John made Mary read the book.’ In (47), the causee argument is projected and it is inherently case m arked with DAT. We consider that its agentive reading results firom compositional theta role assignment by (s)ase and the VP th at it takes. The object of the embedded verb can refer to the causer argum ent within the domain of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 107 same tense, and so, it gets ACC. On the other hand, the causer argument has no unmarked thematic position upwards for reference, it can only get NOM. Recall that we have assum ed that the causative verb (s)ase takes VP, not vP, as its complement. This means that the external argument of the embedded verb can not be projected; hence, it simply has no chance to get marked with dependent ACC, as in (43b). Note, furthermore, that contrary to Kuroda (1978), we are arguing th a t while the subject of the embedded intransitive verb can be marked with o, the causee in n i always originates in the matrix clause; in other words, the ni- causee is selected for by (s)ase, but the o-subject is not. This explains the contrast of the sort noted by Shibatani (1978b: 313) . (48)a. karera-ga ame-o hur-ase-ta they-NOM rain-ACC faü-CAUS-PAST They made it rain.’ b. *karera-ga ame-ni hur-ase-ta they-NOM rain-DAT faU-CAUS-PAST Given that (s)ase selects for only a [+animate] (or [+volitional]) DP as a causee, (48b) is naturally excluded. On the other hand, since bur fall’ is an unaccusative intransitive verb, am e ‘ rain’ should be able to occur within the VP headed by this verb. Then, am e will get dependent ACC, as in (48a). hi this case, this inanim ate DP is projected within the embedded VP, which Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 108 (s)ase cannot look into for selection. However, it seems to be difficult to rule out cases like (48b) in Kuroda’ s approach. This is because as long as am e is the subject of th e embedded clause, it should automatically get ni by his Subject Ni-M arking in (38b).^ ^ Hiroto Hoshi O p-c.) notes that o-marked causées can be the antecedent of zibun ‘ self, hence, problematic to the analysis of o-causatives in the text. (i) a. Johnz-ga Marys-o zibunas-no heya-e ik-ase-ta -NOM -ACC self-GEN room-to go-CAUS-PAST (ht.) John made Mary go to selfs room.’ b. John 2-ga Marya-ni zibunas-no heya-e ik-ase-ta -NOM -DAT self-GEN room-to go-CAUS-PAST It is widely assum ed th at both sentences in (i) are ambiguous with respect to zibun; i.e., besides the causer in ga ‘ NOM, the causee in either o ACC’ or n i DAT qualifies as an antecedent of zibun. Hence, given “subject orientation” of zibun, the possible antecedence of zibun by the causee in o in (ia) challenges its proposed V P-intem al origin. However, a t least for some speakers, the causee in oin (ia) cannot antecede zibun as easily as that in ni in (ib). As noted by Inoue (1978: 123ff.), the assumed “subject orientation” of zibun is not as secure as it appears. Kuroda (1979) persuasively argues th at n i passives, in contrast to niyotte passives, allow derivations with sentential complementation but without movement, as widely assumed for indirect passives (cf. also Kitagawa and Kuroda, 1992). Given this, Inoue’ s point is illustrated by the following example, which she cites firom H arada (1977). (Ü) Taro2-ga Hanakoa-ni zibim2/? * 3-no heya-kara proa oidas-are-ta -NOM -DAT self-GEN room-firom drive out-P ASS-PAST ‘ (ht.) Taro was driven out of selfs room by Hanako.’ In (ii), H anako-ni should be able to be taken as the complement subject. However, for m any speakers, zibun in (ii) is not ambiguous; it can be construed with the m atrix subject Taro, b u t not Hanako. More significantly, nonsubjects seem to be able to serve as antecedents of zibun at least marginally. (hi) Taroa-ga Hanakoa-o zibim*a3-no otto-no moto-e oikaesi-ta -NOM -ACC self-GEN husband-to send back-P AST ‘ (ht.) Taro sent Hanako back to selfs husband.’ The context of (hi) is pragmatically controlled in such a way th at the direct Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 3.3.3 Q particles and D Along with M arantz’ s (1991) morphological case theory, we will assume th at Japanese has D, which, as claimed by Fukui (1995), is universally required for interpretive purposes at LF (i.e. quantification, definiteness, binding, etc.), and th at the locus of CASE features is also D .33 Recall that, in chapter 2, we have proposed a morphological condition parametrized for Japanese th at functional categories in this language m ust satisfy in PF, hke the following. (49) AH functional (i.e. [+F]) heads in Japanese are morphologically suffixal. object Hanako will be the only plausible candidate for the antecedent of zibun, and in fact, it can serve as such for some speakers. Thus, to the extent th a t “subject orientation” of zibun is insecure, the noted antecedence of zibun by the causee in o, if possible at aU, does not necessarily argues against the text analysis of o-causatives. 3 3 This latter idea is not entirely new. Tonoike (1991), for instance, claims th at particles, including wa TOP’ as well as case particles, occupy D positions. He further argues th at while wa has universal quantificational force, case particles hke ga are inherently existential. Although it might be the case with wa, the interpretation of a DP with a case particle, despite Tonoike, depends on either the content of NP inside the DP (e.g. [d p [n p dareka]-ga] ‘ someone-NOM’ is existential, but [ d p [n p daremo]-ga] everyone-NOM’ is universal) or external factors (e.g. a bare noun with a case particle gakusei-ga ‘ student-NOM is generic in generic sentences, but the same DP is construed as existential in episodic sentences). Hence, it is concluded th at alth o u ^ case particles are morphological speU-out of CASE features in D, they lack quantificational force. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 0 We continue to assume that functional heads, C, T and v, each select for T, v and V, respectively. From this assumption follows the strict head-finality of the language, and it also induces su-support since T must attach to v, and vto V, under string adjacency at PF (cf. section 2.3.2. in chapter 2). If Japanese noun phrases are headed by D, D, being a functional category, m ust satisfy (49) by attaching to N in PF. We claim below that this accounts not only for the fact th at case particles always appear to the right of an NP but also for the possible orders among Q-particles and case-particles noted in section 3.2. The permissible/impermissible particle arrays in Japanese are schematically reproduced in (50) - (56), with relevant examples on the right. (50) a. N P -K hon-wa/mo book-TOP/also b. N P -F hon-dake/m a de/b a k a ri (51) a. *N P -N O M -K book-only/even/only *hon-ga-wa/mo book-NOM-TOP/also b. *N P -N O M -F (52)a. *N P-K -N O M *hon-ga-dake/m ade/b a k a ri book-NOM-only/even/only *hon-wa/mo -ga book-TOP/also-NOM Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill b. NP - F - NOM h o n -d ak e/m ad e/b ak ari-g a book-only/even/only-NOM (53) a. NP - ACC - K hon-o-ba/mo book-ACC-TOP/also b. NP - ACC - F hon-o-dake/m a de/b a k a ri book-ACC-only/even/only (54)a. *N P-K -A C C *hon-wa/mo-o book-TOP/also-ACC b. NP - F - ACC h o n -d ake/m ade/bakari-o book-only/even/only-ACC (55) a. NP - DAT - K hon-ni-wa/mo book-DAT-TOP/also b. NP - DAT - F h o n -n i-d ak e/m ad e/b ak ari book-DAT-only/even/only (56) a. *NP - K - DAT *hon-wa/mo-ni book-TOP/also-DAT b. NP - F - DAT h o n -d ak e/m ad e/b ak ari-n i book-only/even/only-DAT If, as assumed here, the location of case particles is the direct reflex of the position of D, the contrast between a- and b-examples in (52), (54) and (56) indicates th at only F-particles can appear inside DP (i.e. to the left of D). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 Except for the cases with NOM in (51a, b), both K- and F-p articles may appear to the right of D (cf. (53) and (55)). We assume with M arantz (1988, 1989, 1991) and Halle and M arantz (1993) th at structural information is preserved in the morphological component of PF. We have argued in chapter 2 th at Q-particles are adjunct ditics; hence, they can be adjoined to any category, as long as other requirements are met. In the cases under discussion, Q-particles can, in principle, be adjoined to any projection of N or D. In the following, however, we will lim it our attention to cases where Q-particles are adjoined to a head , jsjo or D°, and a maximal projection, N™^ (=NP) or D ™ ® * (=DP) for ease of exposition.3^ Q ip P b. P T Q N O CASE CASE P Q CASE P NO N 0 N O Q 0 4 Note th at we will not lose generahty if we ignore the possibility th a t Q- p articles are adjoined to an intermediate category of N or D. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 113 In (57a, b), a Q-particle is adjoined to DP and D°, respectively, and in (57c, d), it is adjoined to NP and N°. Among these four possibilities, the first two cases are immediately known to be permissible. As stated in (49), e lU functional categories including D° as well as Q-particles m ust attach to a head in the morphological component of PF. In (57a, b), this morphological requirements on both D ® and the Q-particle are satisfied, since, under string adjacency, D° can attach to N°, for which it is subcategorized (see section 2.3.2., chapter 2), and the Q-particle can attach to D°. Notice that in these two cases, the relevant p art of the representation after morphological concatenation wiU look Hke the following. (58) ... N O + Ijo + Q CASE Along the lines of M arantz (1991), we assume th at the theory of morphological case requires all CASE features m ust be spelled out in one way or another (presumably, for PF convergence), and the presence of a case particle is to provide a morphological spellout to the CASE feature in D. That is, aside firom inherent (or lexically governed) case markers Hke ni ‘ DAT, dependent and unmarked (or default) case particles, o ‘ ACC and ga ‘ NOM’, are only employed to fulfill this morphological requirem ent on CASE features. Given this assumption, the spellout requirement on the CASE feature in (58) can be considered to be satisfied by suffixation of the Q-particle to D°, the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 114 former inherently having phonetic content. In cases like (50a, b), the morphological requirement of those DPs w ith a Q-particle but without a case particles is satisfied, because the CASE features in those DPs are spelled out by the Q-particle.® At first sight, it looks as if the pre-Spell-Out representations in (57c, d) will not lead to legitimate morphological strings. In these cases, the Q- particle can attach to N° under linear adjacency, so th at its morphological requirement is met. However, since D^ is subcategorized for N° as its stem, it will be astray, as shown below, which would result in an illegitim ate PF representation. (59) ... r ^ + Q p ... * However, as noted by Klavans (1985) and M arantz (1988, 1989), it need not always be the case that cUtics choose particular syntactic positions as their phonological hosts (cf. section 2.2.2., chapter 2). For instance, according to M arantz, in Papago, AUX takes VP as its complement on its right, as shown in the following. ® A tacit assumption here is th at assignm ent of non-inherent cases in accordance with the hierarchy in (27) takes place after morphological concatenation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 115 (60) [s Uux ‘ o] [vp pi iam-hu cdkpan] g Huan] NEG there work ART John If some constituent is preposed to the sentence-initial position for topicahzation, the AUX b, characterized as an enclitic, suffixes to it. However, if nothing is preposed, the AUX diticizes to the first word in the VP instead, as shown in (61). (61) [s [vp [pi+’ o] iam-hu dkpan] g Huan] NEGr+AUX there work ART John M arantz calls this phenomenon “morphological fusion”, and he states that it is typical of clitics. Since, as noted in chapter 2, clitics are often unselective with respect to categories of their hosts, they can trade off their syntactic positions for dependency on a local phonological host. As noted by M arantz, morphological fusion is a local phenomenon, and it takes place only when a strai^tforw ard application of morphological concatenation does not satisfy the morphophonological requirements on the p art of the clitic. If morphological fusion is not a possible option for clitics, it can be considered to apply to cases like (57c, d), which would, otherwise, end up as an illegitimate PF object in (59). In (59), the requirem ent on D° th at it must attach to N° is not m et due to the presence of a Q-p article. However, if morphological fusion applies to (57c, d), and it transposes the Q-particle around D° for the purpose of satisfying the la tte r’ s morphological requirement. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 116 the following morphological representations in (62b, d) will resu lt. (62) a. (=46c) DJ Q P r N: NO c. (=46d) P 0 N O CASE NO Q In either case, morphological concatenation can reapply to the derived morphological structure, and it will eventually bring about a legitimate terminal string in the following. (63) ... No + Do + Q It can be considered th at morphological fusion m ust take place in (57c, d), because this is the only way (i.e. the “last resort”) th at can make the morphological requirem ents on Do and the Q-particle (as a clitic) simultaneously satisfied.^^ 3 6 We m ust adm it th a t the proposed application of morphological fusion in the case under discussion is “altruistic”, because the Q-p article, as a chtic, can satisfy its morphophonological requirement by attaching to N°. 3 7 J.-R. Vergnaud O p-c.) has raised an im portant question as to why morphological fusion is a possible option in the nominal context, but not in the verbal context th a t requires su-support. This is presum ably because the use Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 117 We propose th at the three modes of case assignment and morphological fusion are ordered in the following way. (64) a. inherent case assignment (i.e. ni T)AT) Spell-Out b. dependent case assignment (i.e. o ‘ ACC*) c. morphological fusion d. unmarked (or default) case assignment (i.e. ga ‘ NOM^ We assume th at morphological concatenation appHes whenever morphological requirements on heads are satisfied. This does not seem to be implausible if morphological fusion takes place after inherent and dependent case assignment. This is because, on the one hand, inherent case is standardly assumed to be contingent upon theta role assignment (cf. Chomsky, 1986b), hence, presumably assigned before SpeU-Out, and dependent case assignment, in the approach adopted here, obviously refers to them atic structures of syntactic trees, but, on the other hand, morphological fusion exhibits a rather idiosyncratic character th at is typical of morphophonemics.^ The pre-Spell- of the dummy verb su ‘ do’ is less costly than morphological fusion in terms of economy, and since there exists no dummy noun in Japanese, morphological fusion is really a last resort in the nominal context discussed above. ^ We m ust assume th a t them atic properties, together with configurations, are represented at the morphological component of PF. As long as morphological case is concerned, this view forces us to take the oppoate direction firom Aoun (1985) with respect to visibility; i.e., theta roles make arguments visible for the purpose of morphological case assignement, but not the other way around. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 118 Out representations in (62a, c) (=(57c, d)), after morphological fusion, will yield the derived morphological structures in (62b, d), which will surface as N- DAT-Q and N-ACC-Q, as exhibited in (55) and (53), respectively. However, NOM, as unm arked (or default) case, is employed only as a “la st resort”, if and only if CASE features would otherwise not be spelled out. Since the CASE feature in D° can be spelled out by suffixation with a Q-particle after morphological fusion, NOM is not inserted in (62b, d). Hence, the surface string in (51), N-NOM-Q, is not found. Now recall th at F-particles, but not K-particles, can alternatively be used as (quasi-)nouns (cf. section 3.2.3.). This can be taken th a t F-particles, as nouns, are able to head NPs, as exhibited in the following. (65) [ n o F ] c a s e We take this to be the source for the surface string, NP - F - case particle, found in b-examples in (52), (54) and (56). Since F-particles as nouns are defective, in the sense that they are semantically incomplete, they always need a complement or a modifier.^ Thus, in hon-dake-ga ‘ book-only-NOM’ in (52b), for instance, bon book' modifies dake ‘ only’ , the latter being the head of ^ Or alternatively, F-particles as nouns might preserve their clitic nature, and so, they need hosts to lean on. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 119 the N P . However, th is alternative nom in al use is not available to K- particles, they never appear in the position of the head N° in (54); hence, the strings in a-examples in (52), (54) and (56) are not found. Before concluding this subsection, we m ust explain the fact th at the permissible order between K - and F-particles is an F -particle followed by a K- particle, but not the other way around. This restriction is dem onstrated as follows ((8) - (10) are reproduced here as (66) - (68).) (66) a. John-dakep-waK -only -TOP b. *John-waK-dakeF -TOP -only (67) a. John-madep-moK -even -also b. *John-moK-madep -also -even (68) a. John-bakarip-waK -only -TOP b. *John-waK-bakarip -TOP -only For some (presumably, interpretive) reasons, a constituent, say, a DP, can not accommodate more than one K-particle; i.e., neither John-mo-wa ‘ John-also- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 0 TOF nor Jobn-wa-mo ‘ John-TOP-also’ is an acceptable string. If one category can receive at most one Q -partide, the noted restriction follows. Since only F-particles can appear to the left of D for the reason noted above, the only permissible order should be [ n o F] - D - K, with D phonetically empty 3.3.4. Postpositions In this subsection, we will see how postpositions can satisfy the morphological requirement on the CASE feature in D. In the literature, obhque case m arkers such as kara ‘ from’, de at, in, with’ , e ‘ to’, etc. are often recognized as postpositions th at take DP complements and project PPs. One strong piece of evidence for this view comes from so-called foating numeral quantifiers (FNQs). It has widely been accepted that if FNQs are to successfully modify DPs, Miyagawa’ s (1989b: 30) m utual c-command requirement m ust be met, to the effect th at an FNQ (or its trace) and the host DP th at it modifies (or its trace) m ust c-command each other. In our terms, case particles Kke ga ‘ NOM’ and o ‘ ACC’ are only morphological spellouts of CASE features in D. Since those particles are properly included in DPs, they should not block the required m utual c-command between DPs and FNQs, ^ At first s i ^ t , sae-mo looks hke an exception. However, it seems to be a variation of sae in its use as an F-particle, because the sequence can further be sufhxed with a case particle, e.g. John-sae-mo-ga, as Jobn-sae-ga is allowed for m any speakers, but mo alone does not allow it, cf. *Jobn-mo-ga. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 which is hom e out hy examples hke the following. (69) a. gakusei-ga 3-nin hon-o kaw-ta students-NOM 3-CL books-ACC buy-PAST Three students bought books.’ b. Yamada-sensei-ga gakusei-o 3-nin sikar-ta -teacher-NOM students- ACC 3-CL scold-PAST M r. Yamada scolded three students.’ In (69a, b), each FNQ can properly be construed as modifying the intended host DP (i.e. gakusei ‘ student’ ) with ga or o. On the contrary, when intended hosts are suffixed with kara ‘ firom’ , de ‘ at, in, with’ , e ‘ to’, etc., modification by FNQs is not successful, as shown by a-examples in (70) - (72) (taken firom Miyagawa, 1989b: 31) (70) a. *hito-ga tiisai mura -k ara 2-tu ki-ta people-NOM small villages -firom 2-CL come-PAST ‘ People came firom two small villages ’ b. hito-ga 2-tu-no tiisai m ura -k a ra ki-ta people-NOM 2-CL-GEN small villages -from come-PAST (71) a. *gakusei-wa kuruma-de 2-dai ki-ta students-TOP cars -in 2-CL come-PAST S tudents came in two cars.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 b. gakusei-wa kurum a-de 2-dai ki-ta students-TOP cars-in 2-CL come-PAST (72) a. *Hanako-wa kooen-e 2-tu ik-ta -TOP parks-to 2-CL go-PAST ‘ Hanako went to two parks.’ b. Hanako-wa 2-tu-no kooen-e ik-ta -TOP 2-CL-GEN parks-to go-PAST The (“non-floating”) num eral quantifiers in b-examples in (70) - (72) are in a prenominal position, therefore receiving the genitive no, and they can successfully modify the noun suffixed with an oblique particle. On the contrary, as shown in the corresponding a-examples above, DPs suffixed with any of those oblique particles cannot be modified by an FNQ. The contrast between (69a, b), on the one hand, and a-examples in (70) - (72), on the other, can be accounted for, if we assume with Miyagawa th at obhque particles hke kara, but not case particles hke ga and o, are postpositions th at project, as repreesnted below.^i In (73), following Miyagawa (1989b), FNQs are assumed to be inserted in a sister position of an argument (i.e. subject, object or PP-complement). However, if FNQs are adjoined to DP or PP, the text discussion will not be affected. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 123 (73) a. FNQ ... IP ^ m a x FNQ ... P P D max gakiisei ga/o gakusei In either case, the FNQ c-commands the intended host DP. Since case particles like ga. and o occupy D as exhibited in (73a), they do not prohibit DP from c-commanding the FNQ. On the contrary, since oblique particles like kara, taking a DP complement, project PP as shown in (73b), DP should fail to c-command the FNQ. Hence, the mutual c-command requirement is not m et in any of a-examples in (70) - (72).^ ^ As noted by Miyagawa (1989b: 34), if a particle assigns a thematic role of its own, the particle is a postposition that projects, but if the NP-particle constituent receives its thematic role from an external source, the particle is a case m arker th at does not project (also cf. Inoue (1978) for relevant discusions). The correctness of this criterion is supported by the ambiguity of the particle ni, as exhibited in examples like the following (cited from Miyagawa, 1989b: 35-36). (i) a. boku-wa yuumei-na gakusya-ni 3-nin aw-ta I -TOP famous scholars-DAT 3-CL meet-PAST 1 m et three famous scholars.’ b. *kodomotati-wa kooen-ni 2-tu ik-ta children -TOP parks-to 2-CL go-PAST ‘ Children went to two parks.’ c. kodomotati-wa 2-tu-no kooen-ni ik-ta children -TOP 2-CL-GEN parks-to go-PAST The success and failture of construal with an FNQ can straightforwardly be accounted for, if ni in (ia) is a case particle, but ni in (ib) is a postposition th a t projects PP. In fact, it seems to be the case th at while predicates hke aw Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 If morphological concatenation applies to the heads dominated by PP in (73b), the following term inal string obtains. (74) ... No + po + po _ CASE Given that P is morphologically subcategorized for D, the morphological requirement on P is satisfied in (74). Furthermore, the CASE feature in D can also satisfy its morphological requirement by being suffixed with P which has phonetic content. In other words, P provides morphological speUout to the CASE feature in D, in the same way th a t the Q-particle does in (58) above. Finally, an account for the relative order between Q-p articles and oblique particles is in order. Since it has been established that oblique particles ‘ meet’ select for a DP and assign it inherent case, those hke ik ‘ go’ select for a directional PP. Compare the following examples with those in (i). (ii) a. *boku-wa yuumei-na gakusya-e aw-ta I -TOP famous scholars-to meet-PAST 1 met famous scholars.’ b. *boku-wa yuumei-na gakusya-o aw-ta I -TOP famous scholars-ACC meet-PAST c. kodomotati-wa kooen-e ik-ta children -TOP parks-to go-PAST Children went to parks.’ Although ni in (ic) can be replaced with e ‘ to’ , which is semantically minimally different firom ni, this replacement is impossible for ni on the them e argument in (ia). Furthermore, ni in (ia) cannot be replaced with the accusative o, either. These facts stron^y suggest th at while ni in (ia) represents the inherent case assigned by aw, what ik selects for is a directional PP, the head of which can be either ni or e, which assigns its own theta role to the DP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 125 like kara * 6 :0111’ are postpositions that project, the possible and impossible arrays in (6 ) can be schematically reproduced in th e following way. (75) a. NP - P- K bon-kara-w a/m o book&om-TOP/also b. NP - P - F b o n -k ara-d ak e/m ad e/b ak ari book-from-only/even/only c. *NP -K-P *bon-wa/m o-kara book-TOP/also-6 rom d. NP-F-P b o n -d ak e/m ad e/b ak ari-k ara book-only/even/only-6 rom (75) shows that both K- and F-particles can follow Ps hke kara * 6 rom’ , but only F-particles can precede them. Since Q-particles can, by hypothesis, be adjoined to a projection of an arbitrary category, relevant cases include adjunction of a Q-particle to a projection of N, D or P. First, let us take a look at cases where a Q-particle is adjoined to a projection of N. Wbüe in (76a) a Q-particle is adjoined to N®“ (=NP), in (76b) it is adjoined to N°, (Again, the possibility of adjunction to an intermediate category of N is ignored for the sake of discussion.) complement. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 (76) a. po d : D po P M O N O Q In either case, the application of morphological concatenation yields the following term inal string. (77) ... *No + Q + Ij)o + Po ... CASE This result does not satisfy the morphological requirement on the part of D°, since it fails to suffix to N° (a lth o u ^ its CASE feature can be spelled out by P°). As suggested earlier, morphological fusion should be a potential savor in such a case. However, as noted by M arantz (1988, 1989) and Halle and M arantz (1993), morphological fusion is a strictly local phenomena, the domain of application of which seems to be restricted to the im m ediately adjacent morpheme (or constituent). If morphological fusion applies to (77) and flips Q around D ® , then the following terminal string will result, in which the morphological requirem ent of D ® is met. (78) ... *No + Do + Q + Po ... This is still illegitimate, because has now become unable to satisfy its morphological requirement, with its string adjacency to D° having been broken Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 127 by Q. Hence, neither of the representations in (76) yields a legitimate term inal string, and th at is why the string in (75c) is excluded. On the other hand, the legitimate string in (75d) results firom the noted possibility of an F- particle occurring as N°, as shown below (cf. (65)). (79) ... [noF] + Do + Po ... Secondly, consider the possibility of adjunction of a Q-partdce to a projection of D. b. P n Q I? p d; p A N O Q CASE N O CASE From either of the representations in (80) will the following terminal string result. (81) ... * N o + | ) o + Q + p o ... CASE Although the case requirement of I)o is met, this string is not felicitous, because Po fails to attach to Do. However, if morphological fusion places Q around P, the resultant string will be an impeccable one in (82). (82) ... No + Do + po + Q... Since the CASE feature in Do can now be spelled out by Po (hence, assignm ent Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 of the unm arked (or default) NOM unnecessary), the expected surface string is N-P-Q, which is attested by the examples in (75a, b). Furthermore, the terminal string in (82) can result firom the third possibility, i.e. adjunction of a Q-particle to a projection of P, in a more straightforward way. Since nothing seems to prohibit this derivational ambiguity, we assume that the surface strings in (75a, b) are structurally ambiguous on the LF branch of derivations. Despite the surface enclisis of Q to P, Q can be either internal to P, as in (80a, b), or external to it, as in (83a, b) below, on the LF side of a derivation. (83 Q p D p o n ; N O d : n ; 3.3.5. Sum m ary In the previous subsection, all the possible particle arrays in Japanese have been accounted for. The relevant mappings firom syntactic structures to surface strings in the morphological component of PF are summarized on the following table. (In the third column of (84), C and X stand for a case particle and an arbitrary category, respectively.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (84) 129 syntactic structure morph. surface examples (pre-Spell-Out) fusion string (57a) (50a) [dp [dp [n p —N°] D°] Q] not hon-moK applicable N-Q book-also’ (57b) (50b) [dp [NP... N°] [do D° Q]] hon-dakep book-only’ (57c) (53a) [dp [np [n p ...N °] Q] D°] hon-o-moK applicable N-C-Q book-ACC-also’ (57(1) (53b) [dp [n p ... [no N° Q]] D°] hon-o-dakep book-ACC-only’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 130 (65) [dp [n p X [n o F]] D°] not applicable X-F-C (54b) hon-dakep-o book-only-ACC’ (83a) [pp [pp [dp [n p —N ° ] D°] P°] Q] (83b) [pp [dp [n p ... N ° ] D °] [po p o Q ]] not applicable N-P-Q (75a) hon-kara-moK book-firom-also’ (75b) hon-kara-dakep book-firom-only’ (80a) [pp [dp [dp [np...N O ] D ® ] Q] P O ] (80b) [pp [dp [n p ... N°] [do D ® Q]] P ® ] applicable N-P-Q (75a) hon-kara-moK book-£rom-also’ (75b) hon-kara-dakep book-firom-only’ [pp [dp [np X [n o F]] D ® ] P ® ] not applicable X-F-P (75d) hon-dakep-kara book-only-firom’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 131 [d p [d p [n p X [no F]] D°] KJ not (66a) apphcable X-F-K John-dakep-waK [d p [n p X [no F]] [do D° K]] ‘ John-only-TOP’ It is important to note th at the surface strings in (53a, b) are dissociated firom the corresponding syntactic structures, due to morphological fusion. The same is true of one of the two possible derivations for (75a, b). As we will see in chapter 4, this dissociation of surface strings from syntactic structures will bring about a significant consequence with respect to the possibility of association with focus. 3.4. “Case M arker Drop” We have seen in the preceding sections that DPs can appear without a case particle if a Q-particle is present and the Q-particle serves the purpose of a case particle, i.e. spellout of the CASE feature in D. However, there are cases where DPs appear without a Q-particle or a case particle, as exhibited in examples like (85b) and (86b). (85) a. John-ga nani-o kaw-ta (ka osie-te kudasai) -NOM what-ACC buy-PAST Ques tell.me please ‘ (Please tell me) what John bought.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 132 b. John-ga nani-o kaw-ta (ka osie-te kudasai) -NOM what buy-PAST Ques tellm e please (86) a. John-ni nani-ga deki-ru (ka osie-te kudasai) -DAT what-NOM can.do-PRES Ques tell.me please ‘ (Please tell me) what John can do.’ b. ?John-ni nani-o deki-ru (ka osie-te kudasai) -DAT what can.do-PRES Ques tellm e please While the object DP nani ‘ what’ is assigned ACC in (85a), it appears ‘ Tbare”, without the case particle, in (85b). Similarly, nani in (86a) is assigned NOM, but it appears “bare” in (86b).^ The examples in (85b) and (86b), as well as the corresponding a-examples, are considered to be acceptable in spoken Japanese. The absence of case particles as in (85b) and (86b) is generally referred to as “case marker drop”. As noted by Saito (1983, 1985), Takezawa (1987) and Kuroda (1988), it is generally the case that “case m arker drop” is only possible with a DP th at is immediately followed by a verb. Hence, a ^ Note th at in our terms, if the experiencer argument of verbs like deki can.do’ is projected, it m ust be assigned inherent DAT, and in such a case, the theme object ran not, be assigned ACC. (For the impossibility of the *DAT-ACC case array, see section 3.3.2. above) The following is indeed impossible as a source for (86b). (i) * John-ni nani-o deki-ru (ka osiete kudasai) -DAT what-ACC can.do-PRES Ques tell.me please (Please tell me) what John can do.’ Hence, it is certain th at the case particle that is dropped in (86b) is NOM, but not ACC. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 contrast in acceptability arises in pairs like the following. (87) a. nani-o John-ga kaw-ta (ka osie-te kudasai) what-ACC -NOM buy-PAST Ques tell.me please ‘ (Please tell me) w hat John b o u ^ t.’ b. ?*nani- 0 John-ga kaw-ta (ka osie-te kudasai) what -NOM buy-PAST Ques tell .me please (88) a. dare-ga kuruma-o kaw-ta (ka osie-te kudasai) who-NOM car-ACC buy-PAST Ques tell.me please (Please teU me) who bought the car.’ b. *dare- 0 kuruma-o kaw-ta (ka osie-te kudasai) who car-ACC buy-PAST Ques tellm e please In (87a, b), the object DP is scrambled to the sentence-initial position, and the presence of ACC is obhgatory. The examples in (88a, b) show that NOM th a t is assigned to the subject cannot be dropped. However, actual deletion of a case particle, especially th a t of ga ‘ NOM, which is only employed as a l a s t resort” for the purpose of spellout of a CASE feature, is not quite compatible with our view of morphological case. Rather, taking “case marker drop” as epiphenomenal, we would like to suggest that a CASE feature in D can optionally be incorporated to a verb before SpeU-Out in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 the following way.^ (89) b. V •P NP f P V NP CASE CASE t(CASE) We assume th a t since D lacks a phonetic matrix, the CASE feature alone can incorporate to V (as an instance of Move F, in the sense of Chomsky (1995, chap. 4)). As suggested by Baker (1988), incorporation of the head of an argum entai category to V seems to serve as one mode of fu lfilling case requirement on th at argumentai category. If this is on the right track, the morphological case requirement imposed on the CASE feature in (89) can be taken to be satisfied by incorporation to V .^ ^ It is not clear, at this point, whether the CASE feature alone or the whole set of formal features of D to which the CASE feature belongs moves. Furthermore, the text discussion assum es th at the optionality of “case m arker drop” is attributed to the optionality of movement of the CASE feature, and this in turn leads to an unsafe assumption that this feature movement is no more costly than morphological case assignment at PF with respect to the principle of economy of derivation. However, we m ust leave these open questions for future research. ^ Hoji ^.c.) notes th a t certain adjuncts seem to be invisible to the adjacency requirem ent on “case marker drop”. Thus, (ib) is only little degraded in comparison to the perfect (ia), at least, for some speakers. (i) a. John-ga nani-o ano mise-de kaw-ta (ka osie-te kudasai) -NOM what-ACC that store-at buy-PAST Ques tell.me please ‘ (Please tell me) what John bought at th at store.’ b. ? John-ga nani-o ano mise-de kaw-ta (ka osie-te kudasai) -NOM what th at store-at buy-PAST Ques tell.me please Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 Note th at the adjacency requirem ent follows from the present approach to “case m arker drop”. If, as suggested above, “case m arker drop” is only possible when the CASE feature is incorporated to V, it m ust be in a position firom where movement to V is possible. That is, it m ust be in the head of DP th at is c-commanded by V. If object DPs are dislocated by scrambling, the CASE feature in D cannot move to V, and as a result, “case m arker drop” is impossible. Our approach to “case m arker drop” makes certain predictions. First, since it is made possible as the result of syntactic incorporation, surface A proper account for this observation requires one of the two assumptions to be made; either [i] certain adjuncts can be adjoined to V, and the CASE feature in D still can move to the complex V thus formed, or [ii] the CASE feature incorporates to a higher head hke v, rather than V. Inasmuch as examples like (iib), as weU as those like (ib), are only mildly degraded, the latter option seems to be preferred. (Ü) a. John-ga nani-0 2 Mary-ni ts age-ta (ka osie-te kudasai) -NOM what-ACC -to give-PAST Ques tell.me please ‘ (Please tell me) what John gave to Mary.’ b. ?(?)John-ga nani-02 Mary-ni t 2 age-ta (ka osie-te kudasai) -NOM what -to give-PAST Ques tell.me please The seemingly mild degration of (iib) would remain a puzzle, if “case marker drop” is attributed to incorporation of the CASE feature to V. Because the DP firom which the CASE feature is to move is scrambled to a VP-adjoined position in (iib), the expected movement would end up in an illicit one. If, on the other hand, the target of movement is v, and if “short” scrambling by adjunction to VP is A-movement (cf. Saito, 1992), movement of the CASE feature firom the scrambled DP in (iib) should be expected to be possible. (Note th at incorporation firom subjects to vis still impossible. See the discussions below.) However, the judgment of sentences like (ib) and (iib) varies among speakers, we will put aside the question of the exact target of movement for future research. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 136 adjacency does not suffice. This prediction is in fact home out. Compare the following with (88). (90) a. kuruma-0 2 dare-ga ta kaw-ta (ka osie-te kudasai) car-ACC who-NOM buy-PAST Ques tellm e please ‘ (Please tell me) who bought the car.’ b. *kuruma-0 2 dare-o ta kaw-ta (ka osie-te kudasai) car-ACC who buy-PAST Ques teU.me please In (90a), the object DP kuruma-o car-ACC is scrambled sentence-initially, and as a result, the subject dare ‘ who’ is adjacent to the verb on the surface string. However, deletion of ga ‘ NOM’ is still impermissible, as indicated in (90b). This is because incorporation to V is impossible from the subject. Furthermore, if inherent case is, as proposed above, assigned at an early stage of the derivation, then an inherent case particle should not be “dropped”, because the morphological case requirement on the CASE feature is already satisfied. Again, this prediction seems to be borne out. (91) a. John-ga dare-ni/*-o aw-ta (ka osie-te kudasai) -NOM who-DAT/ACC meet-PAST Ques teU.me please (Please tell me) who John met.’ b. *John-ga dare-o aw-ta (ka osie-te kudasai) -NOM who meet-PAST Ques tellm e please Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 (92) a. John-ga dare-m/*-o soodansi-ta (ka osie-te kudasai) -NOM who-DAT/ACC consult-PAST Ques tellm e please ‘ (Please tell me) who John consulted.’ b. * John-ga dare-o soodansi-ta (ka osie-te kudasai) -NOM who consult-PAST Ques teU.me please Although the verbs in (91) and (92) take a theme object, as weU as an agentive subject, they assign in h eren t DAT to their objects; dependent ACC is impermissible, as indicated. As indicated in (91b) and (92b), the inherent case particle ni cannot be absent, even if it is adjacent to the verb. Since inherent case is tied to theta role assignment, it is plausible to assume th at inherent case assignment takes place at the merger of a DP with a predicate th at assigns it a particular th eta role. If the case requirem ent on the CASE feature in D is satisfied as soon as it is assigned inherent case, it need not move to V; hence, movement is prohibited by the general principle of economy. RecaU that DPs with a Q-particle can appear without a case particle in cases hke the foUowing. (93) a. John-ga LGB-moK yom-ta (koto) -NOM -also read-PAST the fact th a t ‘ (ht.) John read also LOB.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 138 b. John-ga LG B -daker yom-ta (koto) -NOM -only read-PAST the fact th a t ‘ John read only LGB.’ One might be tem pted to claim that “case m arker drop” operates on the objects in the examples in (93). However, the adjacency requirement on “case marker drop” need not hold for cases similar to (93). Compare the following with (87b), (88b) and (90). (94) a. LGB-moK2 John-ga ecg yom-ta (koto) -also -NOM read-PAST the fact th at ‘ (lit.)Also LGB, John read.’ b. LGB-dakera John-ga ecz yom-ta (koto) -only -NOM read-PAST the fact th at (ht.)Only LGB, John read.’ (95) a. John-moK LGB-o yom-ta (koto) -also -ACC read-PAST the fact th a t ‘ (Lit.)Also John read LGB.’ b. John-dakep LGB-o yom-ta (koto) -only -ACC read-PAST the fact th a t ‘ Only John read LGB.’ In (94a, b), the sentence-initial DP with a Q-particle appears without a case particle. No m atter how the sentence-initial DP is related to the gap in the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 object position in these cases, it is obvious th at the DP is not adjacent to the verb. In (95a, b), the subject, which is not adjacent to the verb, appears with a Q-particle but without a case particle. These facts indicate that, as we have suggested, Q-particles in fact can play the role of case particles in term s of morphological Hcensing of CASE features in D; Q-particles serve the purpose of morphologically spelling out CASE features by clitidging to D. If this is correct, we m ust conclude that since the morphological requirement on the CASE feature is satisfied by enclisis of a Q-pzurticle to D in cases like (93a, b), incorporation of the CASE feature to the verb is unnecessary, therefore, prohibited by the economy piindple. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 C H A PTE R 4 A sso ciatio n of Q -p a rtic le s w ith F ocus In the preceding two chapters, we have elaborated on the morphological aspects of Q-particles. In tbis chapter, we are concerned with the function of Q-particles as focus inducers. As shown in section 4.1., Q-particles are, like focus adverbs in Englim b, associated w ith a constituent th a t is focused, and this association of Q-particles with focus exhibits several intriguing characteristics. In sections 4.2. and 4.3., we will try to derive those characteristics &om a minimal set of existing grammatical devices like head movement and propagation of focus. Some consequences of our proposal will also be discussed at the end of section 4.3. 4.1. Association with Focus As extensively discussed by Kuroda (1965, 1969, 1970), Jackendoff (1972), Rooth (1985, 1992, 1996), Taglight (1984) and others, so-called focus adverbs in English like also, even and only are associated with focus. Roughly speaking, a focus adverb can be associated with any constituent with focal Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 141 stress that it c-commands. For instance, the following sentences in (1), cited &om Jackendoff (1972: 248), can be construed differently, depending on which constituent is focused and thereby, associated w ith the adverb. (1) John even gave his daughter a new bicycle i. John even GAVE his daughter a new bicycle Ü. John even gave HIS daughter a new bicycle iii. John even gave his DAUGHTER a new bicycle iv. John even gave his daughter a NEW bicycle V. John even gave his daughter a new BICYCLE In each of (li-v), the capitalized constituent has focal stress, and even is associated with it, adding a specific implication th a t the sentence without even does not have.^ ^ 1 According to Rooth’ s (1985) theory of focus in situ, the semantics of even has existential and scalar implicatures. Hence, the sentence in (i), for instance, can semantically be understood as the combination of the assertion in (iia), the existential im phcature in (iib), and the scalar im plicature in (üc). (i) John even bought a car. (Ü) a. a = [[John bought a car]] b. 3p [C(p) & "p & [p 9^ a]] c. Vp [[C(p) & p # a] -* exceed’ (likelihood’ 0?), hkelihoodCa))] The assertion a in (iia) is the proposition in (i) m inus even. The existential implicature in (iib) indicates th at there is an alternative proposition p, which is determined by the context C, p is true, and p is not equal to a. The scalar implicature in (iic) states th at for every p, if p is determined by C and not equal to a, then the likelihood of p exceeds the likelihood of a. For fiirther details of semantics of focus adverbs, see Rooth (1985, 1992, 1996) and references cited therein. 2 As we will see below, a [+focus] feature may “propagate” up the syntactic Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 142 Like focus adverbs in Finnish, Q-particles in Japanese participate in association with focus. As noted by Kuroda (1970), a Q-p article can be associated with any constituent th at it c-commands, as illustrated in the following. (2 ) Mary-wa [vp [d p pan-o] [v yaki]]-mo si-ta -TOP bread-ACC bake -also do-PAST M ary also baked bread.’ i. In addition to doing something else (e.g. waxing the kitchen floor), Mary baked bread. (VP focused) Ü. In addition to baking something else (e.g. an apple pie), Mary baked bread. (DP focused) iii. In addition to doing something else with the bread (e.g. eating it), Mary baked it. (V focused) In (2), mo ‘ also’ is attached to VP, and it may take as its focus associate either the whole VP, the object DP or the verb alone.^ As a result, (2) is at least three-ways ambiguous, as represented in (2 i-üi). tree &om the node th at is assigned focal stress in a certain way. Hence, the focus adverb in examples like (l.v) can be associated with a larger constituent th at properly includes the node th a t is phonetically stressed. 3 Since, as we have argued, Q-particles can be adjoined to any category, there are several other possibilities for the adjunction site of mo in (2). For instance, it might have been adjoined to V°, V’ or a projection of v. However, we ignore these possibilities for ease of exposition. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 Furthermore, as also noted by Kuroda (1965), a Q-partide may be associated even with a category that dominates it, in cases like the following. (3) (kinoo-no paatdi-de Johng-wa sake-o nom-ta yesterday-GEN party-at -TOP sake-ACC drink-PAST dake-de n a k u ...) not-only b u t ... (At yesterday’ s party, John not only drank sake, b u t...’ ) a. (kareg-wa) [vpsusi-o tabe]-sae si-ta he-TOP sushi-ACC eat-even do-PAST ‘ (he) even ate sushi.’ b. (kareg-wa) [v p [d p susi]-sae tabe]-ta he-TOP sushi-even eat-PAST (he) ate even sushi.’ (4) (kinoo Johng-wa hoka-ni nanimo se-zu...) yesterday -TOP other-as anything do-NEG... (‘ Yesterday, John did not do anything else, b u t...") a. (kareg-wa) [vpmanga-o yom ]-te-bakari i-ta he-TOP comics-ACC read-PROG-only be-PAST (he) was only reading comics.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 b. (kareg-wa) [vp [op m anga]-bakari yom]-te i-ta he-TOP comics-only read-PROG be-PAST ‘ (he) was reading only comics.’ Each of the given precontexts in (3) and (4) can naturally be followed by the corresponding a-sentence, in which a Q-p article is attached to VP, so th at the VP as a whole can be focused, in opposition to the VP in the sentence of the precontext. Since sae even’ in (3a) adjoins to, therefore c-commands, the VP susi-o tabe ‘ sushi-ACC eat’ , it can take the whole VP as the associate of focus, in opposition to the VP sake-o nom ‘ sake-ACC drink’ in the precontext sentence, as expected from the discussion about (2) above. It is, however, rather surprising that although sae is attached to the object DP included in the VP in (3b), (3b), as well as (3a), can be construed as: “although eating sushi (in opposition to drinking sake) was the least expected for John to do, he did it”, and above aU, (3b) need not be taken as: “although sushi was the le ast expected among the things th at John was to eat, he ate it”, where only the object DP sushi is the target of focus for sae* This indicates th at in (3b), sae can be associated with the VP th at contains it. In the same way, bakari ‘ only’ in (4b), which adjoins to the object, can be associated with the VP th a t * The latter interpretation is of course among the possible readings th at (3b) renders. The point here is that Q-particles need not be associated with constituents th at they c-command a t Spell-Out, b u t rather, higher categories th at include them may qualify as targets of focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 145 includes it, in th e sam e way that bakari in (4a), which is adjoined to the VP, can. K-particles like mo ‘ also’ can be associated with a still more inclusive constituent, as exhibited in the following. (5) (kinoo-no paatü-de-wa Mary-ga odor-ta dake-de n a k u ...) yesterday’ s party-at-TOP -NOM dance-PAST not-only b u t... (‘ At yesterday’ s party, not only M ary danced, b u t...’ ) a. [vpJohn-ga [vppiano-o hiki]]-mo si-ta -NOM -ACC play-also do-PAST I t also happened th at John played the piano.’ b. John-ga [vp [oppianoj-mo hik]]-ta -NOM -also play-PAST (ht.) John played also the piano.’ The precontext in (5) can naturally be followed by (5a), which can be taken as: “(at yesterday’ s party, in addition to Mary’ s dancing,) it also happened th a t John played the piano”. Note th at the precontext given in (5) is shghtly different firom th a t in (3). While the precontext in (3) includes John’ s action that is to be followed by his further action in the continuation, that in (5) includes an event of M ary’ s dancing, which is to be followed by another event. Consequently, it is likely th at a natural continuation to (5) will allow the core prepositional category, inclusive of a subject, th a t represents an event to be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 146 focused. It is not surprising if mo to the right of the verb in (5a) is capable of this, since it can be taken as adjoining to v f, which is, as we have assumed in the preceding chapters, the minimal prepositional category th at includes an agentive subject. On the other hand, it is rather surprising th at the sentence in (5b), in which mo is adjoined to the object, can be synonymous with the sentence in (5a). This indicates th at mo in (5b), as well as that in (5a), can be associated with the vP. Most importantly, again, (5b) does not have to be interpreted as: “in addition to some other musical instrum ents, John played the piano”, in which case only the piano is focused. In (6), the two cases of association of Q-particles with focus that we have seen so far are schematically represented. (6) a. b. Q a (6a) is the straightforward case, exemplified by (2), where a Q-particle (Q) is adjoined to a consituent a, which dominates p, and Q is associated with p by c-command. (6b) represents cases like b-examples in (3) - (5), where Q, which is adjoined to a, is associated with p, which Q does not c-command. In what follows, we will rrfer to what happens in the latter cases as association with wide focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 147 Cases of association with wide focus represented in (6b) led Kuroda (1965) to propose an attachm ent transformation analysis, whereby a Q- particle is base-generated at the end of a sentence at Deep Structure and then it is attached to an arbitrary category within the same sentence at Surface Structure. Kuroda (1965: 80) formulates the attachm ent transformation, for example, for mo ‘ also’ in the following way. (7) a. Mb-Attachment [s X-Y-Z] - mo —^ [s X - Y"t"m o - Z ] - mo b. Mo-Deletion [s X - Y'J-mo - Z ] - mo —> [g X - Y+mo - Z ] - 0 In (7), Y is the target of attachment, and X and Z are arbitrary categories (i.e. variables). This rule consists of two parts, i.e. “copy" and “deletion”. In (7a), mois copied and right-adjoined to the target Y, and then, in (7b), the base generated m ois deleted accordingly. It is im portant to note that Kuroda (1965) was w ritten at the times of Chomsky (1965), when Deep Structure was considered to be the sole input to semantic in terp retatio n C o n seq u en tly , 5 Particles like mo ‘ also, V‘ and ka ‘ Ques., 3’ can be considered to be bundles of features. In Kuroda’ s terms, words like daremo ‘ everyone’ and dareka someone’ are derived by attachment of mo and ka, respectively, to dare ‘ who’ , which is further derived by attachement of an indeterminate feature (END) to a pronominal feature (PRO) in an argument position. This latter analysis of dare is obviously descendant of Khma’ s (1964) analysis of wh-words in English (e.g. who = someone + WH). Hoji (1997) notes in passing that in this respect, Kuroda’ s (1965) attachment transformation, and hence, Khma’ s Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 148 for Kuroda, the base-generated position of a Q-partide was m eant to capture the fact of association with wide focus. By being base-generated at the end of a sentence as in (7a), m ois guaranteed to be able to focus as wide a domain as the whole sentence. Starting out in the base-generated position at Deep Structure, as indicated in (8a), if Mb-Attachment in (7a) targets VP (or V), (8b) will result, and if it targets the object, (8c) will be derived. (8) a. [s John-ga piano-o hik ta] - mo (Deep Structure) -NOM -ACC play PAST also b. [s John-ga piano-o hik+mo ta] - mo c. (s John-ga piano+mo hik ta] - mo In either case, the subsequent apphcation of Mo-Deletion, followed by su- support in the case of (8b), will derive the correct surface form, (5a) or (5b). One prediction to be naturally made &om Kuroda’ s attachm ent transformation analysis is th a t mo in (7a) should be able to attach to the subject as well, while m aintaining the possibility of focusing the whole sentence.® This prediction is indeed borne out. Mb -Attachment targeting (1964), too, are precursory of feature movement of Chomsky (1995, chap. 4). ® Kuroda (1965: 80-81) provides the following examples to illustrate the point, (i) a. musuko-ga daigaku-ni hairi-mo si-ta son-NOM coUege-to enter-also do-PAST I t also happened th a t (his) son entered college.’ b. musuko-mo daigaku-ni hair-ta -also college-to enter-PAST ‘ (His) son also entered college.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 the subject in (7a), followed by Mo-Deletion, will derive the following surface form. (9) [s John-mo piano-o hik-ta] -also -ACC play-PAST ‘ (ht.) Also John played the piano.’ One of the possible readings of (9) is indeed the one in which the event of John’ s playing the piano is focused by mo; i.e., (9) can be construed in the same way that (5a) is. To update Kuroda’ s attachm ent transform ation analysis, Aoyagi (1996), assuming the principles-and-p aram eters approach, proposes th a t a Q- particle raises to a higher (functional) head at LF, from where it can be associated with a target of focus th at it c-commands, as indicated in the following. (10) a. [ t p [vP John-ga [vp [oppianoj-mo hik]]-ta] (=5b) -NOM -also play -PAST ‘ (ht.) John played also the piano.’ As noted by Kuroda, mo in (ib), as well as th at in (ia), can be associated with the whole prepositional content. However, since predicates hke hair ‘ enter’ require the ga-ni case array, which can be taken as that of unaccusatives (cf. section 3.3.2. of chapter 3 above), sentences hke (i) are avoided and genuine transitive sentences are used in the text. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 150 b. [T P [% f John-ga [ v p [ d p piano]-t(mo) hik]]-ta+mo] -NOM play -PAST+also Under this LF head movement analysis, the surface form in (10a) (=5b) is the underlying form, &om which the representation in (10b) is derived by movement of mo ‘ also’ at LF. Since (10b) is, in relevant respects, identical to Kuroda’ s Deep Structure representation in (8a), this analysis looks as if it is merely an “upside down” recapitulation of Kuroda’ s, to the extent that either approach can account for the fact of association with wide focus. As will become obvious in what follows, however, our LF head movement approach presents more than a simple recapitulation, with some significant consequences. We will argue th at association with focus m ust be explained in three separate terms: (i) copying of a [+focus] feature onto a category that a Q-particle c-commands at the time of merger, (ii) propagation of the copied [+focus] feature, and (iii) raising of a Q-particle to a (relevant) functional head at LF. In respect of (iii), we will m aintain th at LF movement of Q-particles is selective, due to feature compatibility; while K-p articles (i.e. kakari-joshi), specified as [+F, -L], raises to T, which is also specified as [+F, -L], F-particles (i.e. fuku-joshi), specified as [+F, +L], raises to v, which shares the same categorial specification (see the discussion in section 3.1. of chapter 3). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 151 4.2. Licensing Functional Heads for K- and F-p articles As seen in th e previous section, K-particles th a t are adjoined to a sentence-internal constituent can be associated with the whole sentence, as exemphhed in our earher examples in (5a, b) and (9), reproduced below as (lla-c). (11) (kinoo-no paatü-de-wa Mary-ga odor-ta dake-de n a k u ...) yesterday’ s party-at-TOP -NOM dance-PAST not-only b u t... (‘ At yesterday’ s party, it not only happened that Mary danced, b u t...’ ) a. [w p John-ga [vp piano-o hiki]]-mo si-ta (=5a) -NOM -ACC play-also do-PAST I t also happened th at John played the piano.’ b. John-ga [vp [d p piano]-mo hik]]-ta (=5b) -NOM -also play-PAST ‘ (lit.) John played also the piano.’ c. [v P John-mo [vp [d p piano-o] hik]]-ta] (=9) -also -ACC play-PAST ‘ (ht.) Also John played the piano.’ The sentence with mo also’ in (1 lb) and (1 Ic) as well as th at in (1 la) can be interpreted as “it also happened that John played the piano”, in which mo takes the whole clause as its focus associate. A sim ilar observation seems Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 152 to hold for the other two K-particles, sae and wa, as weU.^ (12) (kinoo-wa iroiro mezurasii koto-ga ar-ta. 4-gatu nanoni yesterday-TOP many unusual things-NOM be-PAST April though ooyuki -ga fur-ta sosite ...) heavy snow-NOM fall-PAST and ... (Yesterday, many unusual things happened. Although it is April by now, it snowed heavily, and ...’ ) a. [v p Mary-ga [vp uta-o yomi]]-sae si-ta -NOM song-ACC read-even do-PAST I t even happened th at Mary made a song.’ b. [v p Mary-ga [ v p [ d p u ta]-sae yom]]-ta -NOM song-even read-PAST ‘ (Lit.) Mary made even a song.’ c. [vpMary-sae [vp [op uta-o] yom]]-ta -even song-ACC read-PAST ‘ (lit.) Even Mary made a song.’ For some speakers, association with wide focus by sae th at is adjoined to the subject is only marginal. This is presumably due to the ambiguous status of sae between a K-p article and an F-particle. See section 3.1. in chapter 3. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 153 (13) (konoyooni genkan-ni asiato-ga nokor-teiru tokoro-o m iru-to...) this way entrance-at footprints-NOM remain fact-ACC looking (‘ Judging from the fact that some footprints remain at the entrance,...’ ) a. [v P dareka-ga [vp ie-ni hairi]]-wa si-ta (hazu da) someone-NOM house-to enter-TOP do-PAST it m ust be the case ‘ (It m ust be the case that) someone has entered the house.’ b. [rf> dareka-ga [vp ie-ni-wa hair]]-ta (hazu da) someone-NOM house-to-TOP enter-PAST it must be the case c. [v p dareka-w a [vp ie-ni hair]]-ta (hazu da) someone-TOP house-to enter-PAST it m ust be the case In (12) and (13), b- and c-sentences can be taken as synonymous with a- sentence, in which the content of the whole clause is focused by the K-particIe. Given th at association with focus requires c-command and th at the minimal propositional category th a t includes the external argument is vP, K-particles must be able to raise to T or higher. If a topic is accommodated in Spec of TP,® and if Q-particles can raise as high as C, they can take a topic within ® This assum ption is not uncontroversial. Many authors who worked on topicalization in Japanese (e.g. Hoji, 1985; Saito, 1985; Kuroda, 1988; inter alia) have shared a view that a thematic topic in wa is accommodated in Spec of CP (or COMP, in a more traditional sense). It is widely assum ed in the literature th at the constituent accommodated in Spec of CP, either by movement or by base-generation, is an operator (or an A’ -element). However, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 154 the thematic topic does not seem to show any typical operatorhood. Consider the following examples in the context where representatives are arriving in Beijin for a convention of world cities from all over the world. (i) a. Tokyo-kara2-wa John-ga ima [dare-ga ec2 kita ka] -from-TOP -NOM now who-NOM came Ques sirabete iru investigate be ‘ (ht.) It is from Tokyo that John is now investigating who came.’ b. Tolqro2-wa John-ga ima [dare-ga ecg kita ka] -TOP -NOM now who-NOM came Ques sirabete iru investigate be (ht.) As for Tokyo, John is now investigating who came.’ In (ia), a PP Tokyo-kara “ Tokyo-from’ is topicalized sentence-initially, and it is construed with the gap in the embedded question clause. If, as suggested by Saito (1985), topicalization of PP (i.e. necessarily contrastive, in his terms), in contrast to that of DP, m ust involve A’ -movement, a wh-island effect is expected to emerge, which is not borne out. On the other hand, Kuroda, who derives both them atic and contrastive topics by movement (cf. Kuroda, 1986), and Hoji (1985), who claims th at while thematic topicalization is base generation, contrastive topicalization is movement, might claim that the PP topic in (ia) m ust have been scrambled (long-distance) from within the embedded question clause as in (ü). (ii) John-ga im a [dare-ga Tokyo-kara kita ka] -NOM now who-NOM -from came Ques sirabete iru (koto) investigate be (the fact that) John is now investigating who came from Tokyo.’ (See Saito (1989) for the possibhty of scrambling out of an indirect question clause. Saito ^.c.) presently agrees to derive (ia) from (ii) by scrambling.) However, (ib), which is as acceptable as (ia), poses a problem to all these authors. Although they aH agree that the DP topic in (ib), as a thematic topic, is accommodated in Spec, CP (by movement for Kuroda, by base generation for Hoji, and by either of the two ways for Saito), the construal of the topic with the gap in the indirect question clause does not exhibit a wh-island effect or a violation of the m in im al binding requirement (MBR), in the sense of Aoun and Li (1989, 1993a, b). (ib) should be contrasted with the following. (iii) a. ?*New Yorkg John wonders [who came from ecs]? b. ?*John-ga im a [dare^ga dokoa-kara kita kaa] -NOM now who-NOM where-from came Ques Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 155 their focus domain. However, this does not seem to be the case, as indicated in examples like the following. (14) (kinoo-no paatü-de-wa Mary-ga sake-o nom-ta yesterday-GEN party-at-TOP -NOM -ACC drink-PAST dake-de n a k u ...) not-only but a. #[TP John-wa [vP [vp susi-o tabe]]-mo si-ta] (C) -TOP sushi-ACC eat -also do-PAST ‘ John also ate sushi.’ b. # [ tp John-wa [vp [d p susi]-mo tabe]]-ta] (C) -TOP sushi-also eat-PAST Either (14a) or (14b), if taken as a natural continuation to the given context. sirabete iru kaa (osiete) investigate be Ques (teU.me) ‘ (ht.) (Tell me) where John is investigating who came from.’ (ib) does not show either the typical wh-island effect in Enghsh in (nia) or a sim ilar efiEect (or an MBR violation) in Japanese in (inb). This observed absence of a wh-island (or MBR) effect indicates th a t the thematic topic in wa, as well as the contrastive one, is a non-operator, hence, not an element of the sort that appears in Spec of CP. Given the morphological case theory adopted in chapter 3, subjects in Japanese can be assigned morphological case in situ (i.e. within vP). Since nothing seems to prohibit topic from being accommodated in Spec of TP, we assume that this is indeed the case. This assumption is indeed compatible with the historical fact th a t topicalization with wa is a reminiscence of kakari-musubi ‘ particle-inflection agreement’ . It is not surprising if topic appears in Spec of TP for agreement with T, insofar as T is the locus of inflection. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 156 should be interpreted as: "in addition to drinking sake (as M ary did), John also ate sushi.” This indicates th a t mo in (14a) or (14b) only focuses on the action depicted by the VP, not the event in which John is the agentive participant. This in turn m eans th a t the topic in (14a, b) is located outside mo’ s focus domain. Thus, we conclude th at the target of LF movement for K- particles is T, firom where they can be associated with the m inim al propositional category, under c-command. Unlike K-particles, F-p articles do not seem to be able to extend their domain of focus beyond VP. F-particles that are adjoined to a VP-intemal constituent can be associated with the inclusive VP, as suggested by the possible synonymy between a- and b-examples in (15) and (16). (15) a. John-wa [vpmanga-o yom] -u-b ak ari-de zenzen benkyoo-si-na -TOP comics-ACC read-PRES-only at all study-do-NEG -katta -PAST ‘ John only read comics and did not study at all.’ b. John-wa [ v p [ d p m an g a]-b ak ari yom]-te zenzen benkyoo-si-na -TOP comics-only read at all study-do-NEG -katta -PAST (ht.) John read only comics and did not study at all.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 157 (16)a. John-wa [vpmanga-o yom ]-ta-dake-de zenzen benkyoo-si-na -TOP comics-ACC read-PAST-only at all study-do-NEG -katta -PAST Jo h n only read comics and did not study at all.’ b. John-wa [vp [d p manga]-dake yom]-te zenzen benkyoo-si-na -TOP comics-only read at all study-do-NEG -katta -PAST ‘ (lit.) John read only comics and did not study at all.’ However, F-particles do not seem to be able to take the whole sentence as a possible target of focus, as indicated by the non-synonymy between a- and b- examples in (17) and (18). (17) (koko-kara mi-e-ru no w a ...) here-from see-POT-PRES COMP TOP (W hat we can see from here is ...’ ) a. [v P John-ga [vpsara-o araw ]]-u-bakari -no yoosu desu -NOM dish-ACC wash -PRES-only -GEN scene COP-PRES only the scene of John’ s washing dishes’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 158 b . [vP John-ga [vp [ d p sara]-b ak ari araw]]-u yoosu desu -NOM dish-only wash -PRES scene COP-PRES ‘ the scene of John’ s washing only dishes’ (18) (watasi-no Idoku-de-wa ...) I-GEN memory-by-TOP (To the best of my m em ory,...’ ) a. [,f John-ga [vpsara-o araw]]-ta-dake desita -NOM dish-ACC wash -PAST-only COP-PAST i t only happened th at John washed dishes’ b. [v P John-ga [vp [d p sara]-dake arawi]] m asita -NOM dish-only wash POL-PAST ‘ John washed only dishes’ Since the F-particles appear to the right of tense in (17a) and (18a), they can naturally take the whole clause within their focus domain, as indicated by the translations.^ On the other hand, the F-particles in the corresponding b- sentences are attached to the object DP, and these sentences cannot be interpreted in the same way as a-sentences. This indicates that the F- ® An F-particle appearing immediately to the right of tense is taken to be N° th a t takes TP as its complement (or modifier). See relevant discussions in section 3.1. of chapter 3. 1 0 We are not stating that b-sentences are not possible continuations to the given precontexts; they indeed are. The point here is, instead, that each b- sentence cannot be taken as a paraphrase of the corresponding a-sentence. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 159 particles in (17b) and (18b) cannot take the whole clause, especially, the subject, within their domain of focus. This state of affairs is contrasted with the possibility of the K-p article mo ‘ also’ to take the whole clause within its focus domain. The following (19a, b) are only minimally different, in relevant aspects, from (17a, b) or (18a, b); th at is, the F-particle in the latter is replaced with mo. (19) (Paatii-no ato-wa, Mary-ga yuka-o hui-ta dake-de n ak u ,...) (party-GEN after-TOP -NOM floor-ACC wipe-PAST not-only-but (‘ After the party, not only did Mary wipe the floor, b u t ...’ ) a. [v P John-ga [vp sara-o arawi]]-mo si-ta -NOM dish-ACC wash -also do-PAST It also happened th at John washed dishes’ b. [vp John-ga [ v p [d p sara]-m o araw j]-ta .-NOM dish-also wash-PAST John washed also dishes’ Not only the postverbal mo in (19a) but also mo attached to the object DP in (19b) can be associated with the whole propositional content inclusive of the subject. Hence, (19b), as well as (19a), can be taken to mean th a t the event of John’ s washing dishes took place in addition to that of M ary’ s wiping the floor. These two contrasts—the one between (15) - (16) and (17) - (18), on the due to the restriction on the former for the possibility of wide focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 160 one hand, and the other between (17) - (18) and (19), on the other—suggest th a t VP-intemal F-particles, as well as K-particles, may move h i ^ e r than VP, but unlike K-particles, they do not move high enough to be able to c- command the subject in Spec of Thus, it can be concluded th at the only possible target of LF movement for F-partides is v. An expectation th at immediately arises is th at unlike K-particles, F- p articles may not extend their domain of focus &om subject. This is indeed the case, as indicated in examples like the following. (20) (koko-kara mi-e-ru no w a ...) here-firom see-POT-PRES COMP TOP (W hat we can see from here is ...’ ) a. [v p John-ga [vp sara-o araw H -u-bakari-no yoosu desu (=17a) -NOM dish-ACC wash -PRES-only-GEN scene COP-PRES ‘ only the scene of John’ s washing dishes’ b. [v p Jo h n -b ak ari [vp sara-o araw R-u yoosu desu -only dish-ACC wash -PRES scene COP-PRES the scene of only John’ s washing dishes’ (21) (watasi-no kioku-de-wa ...) I-GEN memory-by-TOP (To the best of my memory,...’ ) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 161 a. [v P John-ga [vp sara-o araw ]]-ta-dake desita (=18a) -NOM dish-ACC wash-PAST-only COP-PAST I t only happened that John washed dishes’ b. [vP [d p John]-dake [vp sara-o arawi]] m asita -only dish-ACC wash POL-PAST ‘ only John washed dishes’ Neither (20b) nor (2 lb) can be interpreted in the same way th at the corresponding a-sentence is. Also in this respect, the behavior of the K- particle mo ‘ also’ is different from th at of F-particles like bakari only’ and dake only’. (22) (Paatii-no ato-wa, Mary-ga yuka-o hui-ta dake-de naku,...) (party-GEN after-TOP -NOM floor-ACC wipe-PAST not-only-but (‘ After the party, not only did M ary wipe the floor, b u t...") a. [v P John-ga [v? sara-o arawi]]-mo si-ta -NOM dish-ACC wash -also do-PAST ï t also happened that John washed dishes’ b . [vP John-mo [vp [ d p sara]-o araw ]]-ta -also dish-ACC wash-PAST ‘ Also John washed dishes’ The possible synonymy between (22b) and (22c) indicates that unlike F- p articles, K-particles may extend their domain of focus from subjects. It has Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 162 been made clear by now th at the possibility for association with wide focus is more restricted with F-particles than with K-particles. ■ In sum, while K-particles raise to T, F-particles to v, which results in the capability for the former, but the incap abdity for th e latter, of including subjects for association with wide focus, as schem atized below. (23) Association with Wide Focus K-particles a. b. F-particles object V T subject F [+focus| object V Notice that the choice of the Hcensing functional category seems to be determined by feature compatibihty. As noted, while K-particles are purely functional (i.e. [+F, -L]), F-particles are half-functional but half-lexical (i.e. [+F, +L]). Then, it is not surprising that K-particles and F-particles each seek for a hcensing category of the same type. In fact, it is generally agreed that while T is [+F, -L] vis [+F, +L].ii Recall that F-partides, but not K-partides, m ay head NPs, so that they Sano (1997, 1998), on independent grounds, argues th at LF-movement of dake ‘ only’ must be strictly raising. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 163 may appear to the left of D° (see section 3.3. in chapter 3). In such a case, as expected, F-p articles are not ehgihle for association with wide focus. (24)a. [vpmanga-o yom ]-u-bakari zya-nakute benkyoo-mo si-nasai comics-ACC read-PRES-only do-not(IMP) study-also do-IMP T)on’ t only read comics but also do studies.’ b. [vp [op m anga]-bakari yom]-te i-naide benkyoo-mo si-nasai comics-only read be-not(lMP) study-also do-IMP ‘ Oit.) Don’ t read only comics but also do studies.’ c. ? ? [v p [d p m anga-bakari-o] yom]-te i-naide benkyoo-mo si-nasai comics-only -ACC read be-not(IMP) study-also do-IMP ‘ (ht.) Don’ t read only comics but also do studies ’ (25) a. [vp manga-o yom]-u-dake zya-nakute benkyoo-mo si-nasai comics-ACC read-PRES-only do-not(IMP) study-also do-IMP ‘ Don’ t only read comics but also do studies ’ b. [v p [d p m anga]-dake yom]-te i-naide benkyoo-mo si-nasai comics-only read be-not(EMP) study-also do-IMP ‘ (ht.) Don’ t read only comics but also do studies.’ c. ??[vp [d p manga-dake-o] yom]-te i-naide benkyoo-mo si-nasai comics-only -ACC read be-not(EMP) study-also do-IMP ‘ (ht.) Don’ t read only comics but also do studies.’ The possibüity of VP focus in a-examples in (24) and (25) is straightforward. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 164 As noted above, F-particles th a t are adjoined to the object DP, as in (24b) and (25b), may take VP focus; hence, each b-sentence can be interpreted in the same way th at the corresponding a-sentence is. However, this possibility of VP focus seems to be rath er marginal with F-particles th at are followed by a case particle, as in (24c) and (25c). It seems that in c-examples, the domain of association with focus is restricted to be internal to DP. 12 We have argued above th a t the possibility of wide focus association for F- particles is contingent upon the success of LF movement to v. It is thus plausible to suspect th a t the impossibility of association with wide focus results from the failure of such movement. Suppose th a t F-particles as well as K-particles share some formal feature with D. This does not seem to be an unnatural assumption, because, as argued in section 3.3. of chapter 3 above, both K- and F-particles may play the role of case particles which morphologically spell out the CASE feature in D. If this assumption is in the right direction, the impossibility of LF movement of F-particles across D can 12 This point can be illustrated by the fact th at acceptability (or discourse felicitousness) increases if the first conjunct of c-examples is followed by a conjunct as in the following. (i) a. [v p [d p m anga-bakari-o] yom]-te i-naide kyookasyo-mo yomi-nasai comics-only -ACC read be-not(IMP) textbook-also read-IMP Don’ t read only comics but also read textbooks.’ b. [v p [dp manga-dake-o] yom]-te i-naide kyookasyo-mo yomi-nasai comics-only -ACC read be-not(IMP) textbook-also read-IMP Since the two conjuncts share the same verb in (i), only the object of the first conjunct has to be focused. As a result, (ia, b) are more felicitous than (24c) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 165 be attributed to a minimality violation, since D, which, by hypothesis, shares some formal feature with F-particles, is closer to them than vin the case under consideration, schematically represented in the f o l l o w i n g . (26) a. b . DP vH"F[+focusl DP NP D N O J(F) NP D ^■F [+focus| NO t ( F ) As indicated in (26a), if an F-particle that heads NP were to move to v at LF, it would be arrested by the intervening D; otherwise, the m inimality condition would be violated. The DP-internal F-particle must, instead, adjoin to D for hcensing, and from there, it can be associated with a target of focus, as shown in (26b). In this case, however, the focusable domain is lim ited to the NP that the F-particle heads. Hence, only the narrowest focus reading is possible in cases like (24c) and (25c). It is interesting to note that for many speakers, Q-p articles on the right of a case particle are also prohibited from undertaking association with wide focus. Compare c-examples with a- and b-examples in the following. and (25c), respectively. Recall th a t in PF, the CASE feature in D is morphologically spelled out by a case particle in (25). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (27) a. [vpJohn-ga [vppiano-o hiki]]-ino si-ta -NOM -ACC play-also do-PAST I t also happened th at John played the piano.’ b. [»p John-ga [vp [d p piano]-mo hik]]-ta -NOM -also play-PAST ‘ (ht.) John played also the piano.’ c. [v P John-ga [vp [d p piano-o]-mo hik]]-ta -NOM -ACC-also play-PAST (ht.) John played also the piano.’ (28) a. John-wa [vp manga-o yom ]-te-bakari i-ta -TOP comics-ACC read-PROG-only be-PAST John was only reading comics.’ b. John-wa [ v p [d p m anga]-bakari yom]-te i-ta -TOP comics-only read-PROG be-PAST John was reading only comics.’ c. John-wa [vp [d p m anga-oj-bakari yom]-te i-ta 166 (=5a) (=5b) (=4a) (=4b) -TOP comics-ACC-only read-PROG be-PAST ‘ John was reading only comics.’ Although the Q-particles in b-examples in (27) and (28) are attached to the object DP, they can take VP focus, in the same way th at the Q-particles in a- examples do. Each c-example is minimally different from the corresponding Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 b-example. While the Q-particle in each b-example is directly preceded by the head noun, a case particle intervenes in between the head noun and the Q- p article in each c-example. To repeat the same point (cf. section 3.3. in chapter 3), a Q-particle in cases like (27b) and (28b) is adjoined to a projection of D, and the morphological case requirement on D is satisfied by th at Q-particle which chticizes to it. On the contrary, a Q-particle in cases like (27c) and (28c) is in fact adjoined to a projection of N. However, due to the morphological subcategorization requirement on D (i.e., it m ust suffix to N), the Q-particle m ust be transposed around D by morphological fusion. If this is indeed the case, the pre-Spell-Out representation of the object DP in (27c) or (28c) should be dissociated 6rom its morphological representation, as indicated below. (29) Pre-Spell-Out representation Morphological representation P P Q CASE n ; NO A 0 CASE n ; NO Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 168 Since morphological fusion is only relevant in th e morphological component in PF, the pre-SpeU-Out representation in (29a or c), as it is, goes through the LF side of the derivation, and the level of representation that is relevant to association with focus is LF. The noted m arginality of wide focus for Q- particles in cases like (27c) and (28c) can be attributed to the impossibility of their movement across D. If the Q-particle in (27c) or (28c) is to move across D to a higher functional head, it will violate the m inimality condition, in the way that is indicated in (26). 4.3. Conditions on Association with Wide Focus 4.3.1. [+Focus] assignment and propagation As noted at the beginning of this chapter, a Q-particle, as a focus marker, can be associated with any constituent th at it c-commands at Spell-Out. Furthermore, it has been argued in the preceding sections that, due to movement at LF, a Q-particle can take as its target of focus a more inclusive category that it is not able to c-command at Spell-Out. A relevant example with a K-particle mo ‘ also’ in (11) is reproduced below as (30). (30) (kinoo-no paatii-de-wa Mary-ga odor-ta dake-de n a k u ...) yesterday-GEN party-at-TOP -NOM dance-PAST not-only b u t ... (‘ At yesterday’ s party, it not only happened th at Mary danced, b u t...’ ) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 169 a. [vP John-ga [vppiano-o hild]]-mo si-ta -NOM -ACC play-also do-PAST Tt also happened that John played the piano/ b. [if John-ga [vp [DPpiano]-mo hik]]-ta -NOM -also play-PAST ‘ (ht.) John played also the piano/ c. [vpJohn-mo [vp piano-o hik]]-ta -also -ACC play-PAST ‘ (lit.) Also John played the piano.' Due to movement to T at LF, m ois able to take vP as its target of focus under c-command at th at level. Hence, m oin (30b, c), as well as th a t in (30a), can subsume the whole prepositional content under its domain of focus. However, it does not seem to be the case that a Q-particle can focus on any constituent th at it c-commands after movement at LF. (31) (kinoo Maryz-wa Johna-o karakaw-ta s i ...) yesterday -TOP -ACC make fun of-PAST and ... ("Yesterday, Mary made fun of John, a n d ...") a. (kanozyoz-wa) karea-o buti-mo si-ta she-TOP him-ACC slap-also do-PAST ‘ (She) also SLAPPED him.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 b. #(kanozyo2-wa) kares-mo but ta she-TOP him-also slap-PAST (ht.) (She) slapped also HIM.’ The context given in (31) will naturally continue, if the verb alone in the continuation is focused. This is achieved in (31a), but not in (3 lb); while (31a) can he taken as: “M ary also SLAPPED John as well as making fun of him”, but (3 lb) can not be interpreted in this way. Although mo in (3 lb) should be able to raise to T at LF, so th at it can c-command vP and any category included in it at th at level, it may not be associated with the verb alone. This suggests th at a domain of association with focus expands, but not shifts. As is widely noted in the hterature (e.g. Chomsky, 1972; Jackendoff, 1972; Selkirk, 1984; Cinque, 1993; Zubizarreta, 1993; among many others), expansion of the domain of focus is a common phenomenon observed in many languages. In English, for example, focus is often indicated by emphatic stress. If stress falls, for instance, on the object, as in (32), the sentence can be construed in several different ways with respect to focus. (32) John fixed THE CAR. (33) a. What did John fix? b. What did John do? c. What happened? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 Because of different presuppositions, each of the three questions in (33a-c) requires a different p art of the sentence to be focused. In fact, (32) serves as a felicitous answer to any of the questions in (33a-c). Zubizarreta (1993) captures this fact by assuming that a [+focus] feature propagates (or percolates up the tree). In her terms, emphatic stress in languages like English is an indication of assignment of a [+focus] feature, and this feature may propagate, as dem onstrated in the following. (34) S[+focusj V P [+ fo c u s j John fixed D P [+ fo c u s I As indicated in (34), the [+focus] feature originates in the object DP; hence, this DP receives stress, and this feature can propagate up the tree to VP or even to the whole S. This possibility of wide focus enables (32) to be an appropriate answer to the questions in (33b, c) as well as th a t in (33a). Although focus propagates, it does not shift sideways. Although (32) can be a fehcitous answ er to any question in (33a-c), it cannot be a good answer to either of the following. (35) a. W hat did John do with the car? b. Who fixed the car? The question in (35a) requires the verb alone to be focused in its answer, and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 172 th at in (35b) requires the subject to be so. The unfebcitousness of (32) as an answer to (35a) or (35b) is accounted for by the impossibibty of shift of [+focus], as shown in the following. (36) John fixed THE CAR We will assume th at a Q-particle with a [-f-focus] feature makes a copy of the feature onto any constituent that it c-commands as soon as it is introduced to a derivation, and the copied feature optionally propagates up the tree as the derivation proceeds, along the lines of Zubizarreta (1993). In LF, the Q-particle moves to a functional category, and firom there it is associated with the constituent up to which the copied [+focus] feature propagates. Taking (30b) for example, this procedure in association with Zubizarreta (1993: 7) proposes a constraint on focus propagation as follows, (i) Focus may propagate upward firom the constituent that bears the un m arked accent along a continuous path th at includes the nodes on the recursive side of the tree and the nodes th at are projections of the head. According to (i), in languages hke English, focus can propagate from the right branch. It is not entirely clear what kind of role stress (or accent, in Zubizarreta’ s terms) plays in Japanese, since it is a pitch-accent language. However, taking [+focusl copying in Japanese as virtually equivalent to stress assignment in English, we will assume, in what follows, that focus can propagate from the left branch and along projections in Japanese in accordance with (i). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 173 wide focus is illustrated in the following. (30) b . [v P John-ga [vp [ d p piano]-mo hik]]-ta -NOM -also play-PAST ‘ (lit.) John played also the piano.’ (37) a. [+focus] copying n b. propagation of [+focus] D P [+ fo cU s| K {+focusI piano DP K hik piano mo movement of a Q-particle to a functional head A sso cia tio n w ith Focus DP ta t(mo) hik piano In (37a), mo is adjoined to DP, and it copies its [+focus] feature onto that DP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 174 (or one of its constituents). In (37b), the copied [+focus] feature propagates along the tree up to the vP node.^® In (37c), mo, being a K-particIe, adjoins to T at LF, whereby the [+focus] feature of mo is associated with the sam e feature with another constituent, vP, in this case, under c-command. This is, we claim, how association with (wide) focus is estabhshed. Note th at since a copied [+focus] feature propagates but not shifts sideways, the V node alone has no chance to receive th at feature. This is why the verb alone cannot be the target of association with focus in (30b). 1 5 In principle, the copied [+focus] feature can propagate further up. However, if it does, it will not be c-commanded, hence not licensed, by (the same feature on) the Q-particle at LF. Thus, we ignore this possibility in the text. 1 5 The same point holds for subjects, too. In the following, the Q-particle on the object DP cannot focus the subject alone. (i) (kinoo-wa Maryz-ga Johna-o karakaw -ta s i ...) yesterday-TOP -NOM -ACC m ake fun of-PAST a n d ... (Testerday, M ary made fun of John, and ...” ) a. Lucy-ga karea-o karakawi-mo si-ta -NOM him-ACC make fun of-also do-PAST TiUCY also made fun of him.' b. Lucy-ga karea-mo karakaw-ta -TOP him-also make fun of-PAST ‘ (lit.) Lucy made fun also of HEM.’ The subject in (ia) can be taken as being focused by mo ‘ also’ , since this postverbal mo can c-command the subject by adjoining to vP. On the other hand, the subject in (ib) cannot be solely focused. This is because mo in (ib), which is adjoined to the object, cannot c-command, and therefore, cannot assign a [-Hocus] feature to, the subject. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 175 4.3.2. An argument-adjunct asymmetry As noted by Aoyagi (1996), there is an asymmetry in the possibility of wide focus association between argum ents and adjuncts. The point is illustrated in the following series of examples, (38) (John-wa mai-asa tyuusya-o ut-ta dake-de naku...) -TOP every morning shot-ACC take-PAST not only but ‘ (John not only took a shot every morning, b u t...)’ a. iti-niti san-kai kusuri-o nomi-mo si-ta one-day three-times medicine-ACC take-also do-PAST (he) also took medicine three times a day.' b. iti-niti san-kai kusuri-mo nom-ta one-day three-times medicine-also take-PAST (ht.) (he) took also medicine three tim es a day.’ c. iti-niti. san-kai-mo kusuii-o nom-ta one-day three-tûnes-also medicine-also take-PAST (he) took medicine as often as three times a day.’ (39) (Bill-wa kata-asi-de booru-o ker-ta dake-de n a k u ...) -TOP one leg-with ball-ACC kick-PAST not-only-but (Bill not only kicked a ball with one leg, b u t...)’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 a. kata-te-de batto-o huri-sae si-ta one-hand-with bat-ACC swing-even do-PAST ‘ (he) even swung a bat with one hand.’ b. kata-te-de batto-sae hur-ta one-hand-with bat-even swing-PAST ‘ (lit.) (he) swung even a bat with one hand.’ c. kata-te-de-sae batto-o hur-ta one-hand-with-even bat-ACC swing-PAST (he) swung a bat even with one hand.’ (40) (uti-ni kaer-te k a ra ...) home-to came since... (‘ Since she came home, ...’ ) a. Mary-wa ip-pai mizu-o nom-ta -dake da -TOP one-CL water-ACC drink-PAST-only COP (ht.) it is only th at Mary has drunk one glass of water.’ b. Mary-wa ip-pai mizu-dake nom-ta -TOP one-CL water-only drink-PAST M ary h as drunk only [one glass of water].’ c. Mary-wa ip-pai-dake mizu-o nom-ta -TOP one-CL-only water-ACC drink-PAST M ary has drunk only [one glass] of water.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 177 In each of (38) - (40), the most natural interpretation of a-sentence in the given context is that in which th e Q-particle takes the content of the whole VP as focus. Since the Q-particle in each a-sentence appears postverbally, its association with the VP focus is straightforwardly successful. In each b- sentence, the Q-particle attaches to an argument (i.e. the theme object), but it can still be associated with VP as focus; hence, each b-sentence can be taken as synonymous with the corresponding a-sentence. These are cases of association with wide focus. However, in each c-sentence, the Q-particle is adjoined to an adverbial constituent: a frequency adverbial in (38c), an instrumental in (39c) and a so-called numeral quantifrer in (40c), and none of c-sentences seems to allow the construal with VP focus, Hoji (p.c.) notes that examples Hke (ib) allow asscociation with wide focus, (i) (Kinoo-wa gozen-tyuu kaze-ga tuyo-katta si, ...) (yesterday-TOP moming-in wind-NOM strong-PAST a n d ...) ‘ (Yesterday, the wind brew strong in the morning, and ...)’ a. gogo-ni ame-ga huri-mo si-ta aftemoon-in rain-NOM fall-also do-PAST I t also rained in the afternoon.’ b. ?gogo-ni-mo ame-ga hur-ta aftemoon-in rain-NOM fall-PAST ‘ (Ht.) It rained also in th e afternoon.’ Furthermore, (Hb), as weU as (iia), seems to allow wide focus reading, at least marginaUy. (n) (John-wa Kyoto-de kabuki-o mi-ta dake-de n ak u ,...) -TOP -in -ACC see-PAST not-only-but (‘ John not only saw kabuki in Kobe, b u t...’ ) a. Kobe-de sukiyaki-o tabe-m o si-ta -in -ACC eat-also do-PAST ‘ (he) also ate sukiyaki in Kobe.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 This contrast indicates th a t Q-particles can exercise association with wide focus from an argument, but not from an adjunct. Aoyagi (1996) suggests th at this argument-adjunct asymmetry be reduced to B aker’ s (1988) version of the HMC; i.e., heads can move from complements, b u t not from adjuncts. If this is on the right track, the possibility of association with wide focus in (38a) and (38b) and its impossibility in (38c) can be shown as follows. T i+focus] subject +focus] ta"t"m o [+focus| Adv kusuri-o nom subject ta^Tn o [+ fo c u s| :us Adv nom kusuri b. ?Kobe-de-mo sukiyaki-o tabe-ta -in-also -ACC eat-PAST (ht.) (he) ate sukiyaki also in Kobe.’ Note, however, th a t the adverbials in (ib) and (üb) are tem poral and locative, respectively. Since locative and temporal adverbs are known to behave hke arguments in certain respects (cf. Huang, 1982; Aoun, et al, 1987; Cinque, 1990; Rizzi, 1990; inter aha), it is not totally unexpected th a t Q-particles attached to those adverbials can exercise association with wide focus.. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 179 [+focns| subject V ? iti-niti san-do t(mo) kusuri-o nom In (38a), a K-p article mo ‘ also’ appears to the right of VP, which is an indication th at it is adjoined to a projection of V or v. Suppose th at it is adjoined to VP, as indicated in (41a).^® At the time of merger of mo, it copies its [+focus] feature onto a category th at it c-commands, for instance, the lower segment of VP. From there, the [+focus] feature can percolate up to the upper segment of VP by projection. Since VP is on the left of v, the [+focus] feature can go further up to v * , and again, by projection, it can reach the vP node. At LF, mo, being a [+F, -L] K-particle, adjoins to T, and from there, it can associate its own [+focus] feature with the same feature on the vP node as one of the possibilities. Thus, association with wide focus is established in (38a). (38b) is different from (38a) only in th at mo is adjoined to the If it is adjoined to other projections of V or v, it does not affect our text discussion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 180 object DP- In this example, too, the copied [+focusJ feature can propagate up the projections to yP, and mo, which is adjoined to T at LF, can be associated with the feature under c-command at th at level, as demonstrated in (41b). On the other hand, m oin (38c) is adjoined to an adverbial adjunct, as indicated in (41c). Mo can copy its [-+focus] feature to the adverbial, and the copied feature may propagate beyond the upper segment of the adverbial because this adverbial is on the recursive side of VP. However, movement of mo to T is prohibited by the HMC, because it is adjoined to an adverbial adjunct. As a result, association with mo with vP for wide focus is 1 9 In our approach, Q-particles are always adjoined to some category. However, w hat counts in the text explanation is the status of the category to which a Q-particle is adjoined. If it is adjoined to a complement, it may move; however, if it is adjoined to an adjunct, it may not. A plausible alternative that suggests itself is that movement of Q- particles &om adjuncts is indeed prohibited by the CED, insofar as movement in question counts as extraction in the sense of Huang (1982). Note further that in the m inim alist terms, only features move at LF. Given this, only the [+focus] feature of a Q-particle needs to move to a relevant functional head. It is not entirely clear if feature movement is constrained by conditions like the CED or, for th at m atter, the HMC (see Chomsky (1995: chap. 4) for relevant discussions). Yet another imaginable alternative is th at movement of the [+focus] feature out of adjuncts is not prohibited by the HMC or the CED, but propagation of [+focus] across adjuncts is barred. This amounts to saying that despite Zibizarreta’ s formulation propagation of focus in term s of tree geometry (cf. footnote 14), languages like Japanese must see the distinction between argum ents and adjuncts. The fact th at such a distinction is invisible to English is exemplified by a question-answer pair like the following. (i) Q: W hat did John do? A: He fixed the clock CAREFULLY. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 181 impossible. The contrast in the possibility of wide focus between a- and b- examples in (39) and (40), on the one hand, and c-examples, on the other, can be accounted for in the same way.^o 4.3.3. Some further predictions Our account of association with wide focus makes some further predictions. If a DP with a Q-particle is dislocated from its base position by either scrambling or topicalization, th at Q-particle should not be eligible for association with wide focus. This is so because LF movement of such a Q- particle which targets T or v would be illicit. Suppose, first, that the DP in question is dislocated by scrambling, which is generally considred to be an adjunction operation.-^ Since the scram bled phrase should behave hke an adjunct in the sense th at it is in a non-theta position, movement of the Q-particle attached to the scrambled phrase is expected to be banned, as shown in the previous section. Suppose, on the The [+focus] feature assigned to the m anner adverb carefully seems to be able to propagate to make the discourse in (i) fehcitous. Although much future work is required to make a choice from among these alternatives, we will keep on assuming that movement of Q-particles from adjuncts is barred by the HMC, as in the text. 2 0 It is not entirely clear how Q-particles th at cannot move to vor T are hcensed. It is probable th at they are hcensed in situ in a m arked way. For instance, the interpretation of mo ‘ also’ in (38c) (i.e. “as often as”) is very idiosyncratic. 2 1 However, see Kuroda (1988) and Fukui and Saito (1996) for alternative Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 other hand, th a t th e DP in question is dislocated by topicalization, and th a t topic is, as assum ed above, accommodated in Spec of T P .2 2 Then, the Q- particle should not be able to move to T or v, because it is too high up in the tree. In either case, wide focus association is expected to be impossible. This prediction indeed seems to be borne out by examples Hke the following. (42) (kinoo-wa iroiro mezurasH. koto-ga ar-ta. 4-gatu nanoni yesterday-TOP many unusual things-NOM be-PAST ^ r i l though ooyuki-ga hur-ta. sosite ...) heavy snow-NOM fall-PAST and ("Yesterday, m any unusual things happened. Although it is ^ r i l by now, it snowed heavily, and ...* ) a. [vf Mary-ga [vpmeshi-o taki]]-m o si-ta -NOM rice-ACC cook-also do-PAST ‘ (Ht.) M ary also cooked rice.’ ‘ It also happened that Mary cooked rice.’ b. [v P Mary-ga [vp meshi-mo ta k ]]-ta -NOM rice-also cook -PAST ‘ (Ht.) M ary cooked also rice.’ views. 22 See footnote 8 above. If topic is located higher than Spec of TP, the text argument will not be affected. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 c meshi-mo 2 [.f Mary-ga [vp eca tak]] -ta rice-also -NOM cook -PAST ‘ (lit.) Also rice, M ary cooked.’ (43) (kinoo uti-ni kaer-te kara...) yesterday home-to come since (Since he came home...") a. John-wa [vp[vpinu-ni esa-o yari]-dake] si-ta -TOP dog-to food-ACC give -only do-PAST ‘ It only happened th at John gave food to his dog.’ h. John-wa [v P [vp inu-ni esa-dake yar]] -ta -TOP dog-to food-only give -PAST John gave only food to his dog.’ c. John-wa [v P [vp esa-dakea [vp inu-ni eca yar]]] -ta -TOP food-only dog-to give -PAST ‘ (lit.) John, only food, gave to his dog.’ The Q-particles mo ‘ also’ and dake onl/ are used in (42) and (43), respectively. The most salient interpretation of each a-sentence in the given context is th at which takes the widest focus possible; i.e. vP for mo in (42a) and VP for dake in (43a), since they are a K- and F-particle, respectively. While the most natural interpretation of (42a) is that in which the event of Mary’ s cooking rice is focused hy mo and th at of (43a) is the interpretation in which the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 184 action of feeding his (=John’ s) dog is focused by dake. The success of these readings in (42a) and (43a) is straightforward, since the Q-particle in each case is in a postverbal position. These wide focus readings are also possible with (42b) and (43b). This is as expected, because mo in (42b) and dake in (43b) can raise to T and v at LF, and the associate [+focus] feature may propagate to vP and VP, respectively. On the contrary, neither of c-examples seems to permit the wide focus reading hke the corresponding a- and b- examples. The sentence-initial DP with mo in (42c) is either topicalized or scrambled, and the DP with dake in (43c) m ust be scrambled. If these DPs are scrambled, they are adjoined to some categories. As seen above, LF movement of a Q-particle tirom an adjunct is prohibited by the HMC. Hence, irrespective of the possibüity of focus propagation, wide focus association wül not be allowed. If, on the other hand, the sentence-initial DP in (42c) is topicalized (either by movement or by base-generation) and topic is in Spec of TP (or higher), then association with wide focus should, again, be prohibited. In this latter case, the Q-particle mois located too high to move to T. Consequently, (42c) and (43c) are both expected to allow only narrow focus readings, which seems to be the case. Another prediction th at our approach to association with focus m akes is that it is upperbounded. This is so because association with wide focus, as assumed here, necessarily incorporates movement of a Q-particle, which m ust Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 be subject to general constraints on movement, including the minimality condition. Again, the prediction seems to be borne out. (44) (kinoo Johng-wa uti-de yakuza-eiga-o mi-ta yesterday -TOP home-at gangster movie-ACC see-PAST dake de n a k u ...) not only but (Yesterday, John not only saw a gangster movie at home, b u t...’ ) a. TuJdzi-e [ecg susi-o tabe-ni] iki-mo si-ta TsuJdji-to sushi-ACC eat-ASP go-also do-PAST ‘ (he) also went to eat sushi in Tsukiji.’ b. Tiddzi-e [ecg susi-m o tabe-ni] ik-ta Tsukiji-to sushi-also eat-ASP go-PAST (ht.) (he) went to eat also sushi in Tsukiji.’ (45) (John-wa Mary-ga sake-o nom-e-ru to omotte iru -TOP -NOM -ACC drink-POT-PRES that think dake de n a k u ...) not only but a. [Nancy-ga sasimi-o tabe-rare-ru to] sinzi-te-mo iru -NOM raw fish-ACC eat-POT-PRES th at beheve-also be (he) also beheves th a t Nancy can eat raw fish.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 b. [Nancy-ga sasimi-mo tabe-rare-ru to] sinzi-te iru -NOM raw fish-also eat-POT-PRES that believe be ‘ (he) believes th at Nancy can eat also raw fish.’ While the complement clause in (44a, b) (presumably, AspP or vP headed by an imperfective aspectual morpheme ni) lacks T, that in (45a, b) is tensed. The Q-particle mo also’ in (44b), being a K-particle to be hcensed by T at LF, can take the m atrix predicate within its scope of association with focus, in the same way th a t mo in (44a) naturally does. On the contrary, mo in (45b), unlike th at in (45a), cannot take the m atrix scope. This follows firom the natural assumption th at LF movement of Q-particles is susceptible to a constraint sim ilar to th at which prohibits chtics in Romance firom undergoing long movement. As noted by Rizzi (1982), long chtic movement in Itahan is arrested by finite tense, as indicated by examples like the following (taken firom Rizzi, 1982: 1-3). (46) a. Piero verra a parlarti di parapsicologia wiU.come to speak.to you about parapsychology Piero will come to speak to you about parapsychology.’ b. Piero ti verrà a parlare di parapsicologia (47) a. Credo che Gianni In présentera a Francesco I.beheve th at her wdldntroduce to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 187 b. *la credo che Gianni présentera a Francesco î believe that Gianni will introduce her to Francesco/ In (46), the pronominal chtic ti *2S’ can be adjoined to either the embedded [-finite] tense (followed by adjunction of the infinitival parlar to T , according to Kayne, 1991) or the [+finite] tense in the matrix clause. In (47), however, the chtic la “ SSF’ cannot move to be adjoined to the matrix tense. This is because, according to Rizzi, the tense of the embedded clause is finite in (47). If this explanation can be extended to LF (long) movement of Q-particles in Japanese, the matrix scope of m oin (45b) is impossible, because its movement is arrested by the closest finite tense. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 CHAPTER 5 An Im p licatio n to W ord O rder It has been commonly held, among the piindples-and-param eters camp, that the word order in a given language results from setting the value of the head parameter. In this view, children decide the value of the param eter as either [+head-initial] or [-head-initial] by being exposed to the data of the language th at they are learning. If the theory of grammar is to attain explanatory adequacy, it should be made clear w hat children need to know from the raw data, so th at they can fix the param eter. In other words, the choice between head-initiality and head-finality m ust be reduced to a core property of a given language. In this section, we will suggest a view th a t while the default order is S(pedfier)-H(ead)-C(cmplement) across languages, as proposed by Kayne (1994), morphological properties of fimctional categories in a given language can bring about an alternative order. In particular, we will argue th at the strict head-finality of Japanese results from specific morphological properties of functional categories in this language, as proposed in chapter 2. Furthermore, it is suggested that a perspective from Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 189 morphology like ours will naturally capture the fact of “mixed” orders exhibited by languages like Chinese. 5.1. Head param eter It is well known th at Japanese exhibits head finality across categories, as illustrated in the following. ^ (1) a. hon-o yom(-ta) (DP-V) book-ACC read(-PAST) ‘ read a book’ b. *yom(-ta) hon-o (V-DP) read(-PAST) book-ACC (2) a. (hon-o) yom ta (VP-T) (book-ACC) read PAST ‘ read a book’ ^ As assumed in section 3.3. in chapter 3, Japanese has D and D is the locus of case particles hke ga ‘ NOM’, o ACC’ and ni ‘ DAT’. (Fukui (1986: 206, fii. 11) suggests that if Japanese has D at all, it may accommodate case particles.) We further assume that, unlike case particles and Q-particles, particles Hke kara firom’ , de In, a t’ and to ‘ with’ are postpositions th a t project maximal projections, i.e. PPs. Postpositions in Japanese are arguably a half functional and half-lexical ([+F, +L]) category, in the sam e sense that the English prepositional complementizer for is (cf. section 2.3.2. in chapter 2). We assume that like Q-particles, postpositions can be employed to satisfy the morphological requirement on the CASE feature in D by suffixing to it in PF. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. *ta (hon-o) yom 190 (T-VP) PAST (book-ACC) read (3) a. (John-ga hon-o yom) ta to (TP-C) (John-NOM book-ACC read) PAST th a t th a t John read a book' b. *to (John-ga hon-o yom) ta th at (John-NOM book-ACC read) PAST (4) a. (John-ga hon-o yom ta) to iw(-ta) (John-NOM book-ACC read) PAST th at say(-PAST) (C-TP) (CP-V) ‘ said th a t John read a book’ b. *iw(-ta) (John-ga hon-o yom ta) to (V-CP) say(-PAST) (John-NOM book-ACC read) PAST th at (5) a. hon o (NP-D) book -ACC b. *o hon (D-NP) ACC book (6) a. hon kara (DP-P) book &om ‘ from a book’ b. *kara hon (P-DP) £rom book Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 191 (7) a. hon-ga hosi(-i) (DP-A) book-NOM desirous(-PRES) ‘ w ant a book’ b. *hosi(-i) hon-ga (DP-A) desirous(-PRES) book-NOM From the d ata above, it can safely be concluded that Japanese is a strictly head-final language. In the principles-and-parameters approach, Japanese, unlike English, has been said to set the head param eter uniformly as [ head- initial] across categories. Among the observed characteristics of Japanese, in opposition to languages like English, is its productivity in complex predicate formation. Unlike English, Japanese abounds with complex predicates, as exhibited in examples like the following. (8) a. (John-wa Mary-ni hon-o) yom -ase -ra re (-ta) -TOP -DAT book-ACC read-CAUS-PASS(-PAST) (John was) made to read (a book by Mary.)’ b. (John-wa Mary-ni hon-o) y o m i-ta g a r-a re ( ta) -TOP -DAT book-ACC read-DESID-PASS(-PAST) (John was) affected by (Mary’ s) wanting to read (a book.)’ Fukui (1988) attem pts to derive the presence of productive complex predicate formation in Japanese and its absence in English firom his theory of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 192 parametric syntax. According to him (op. cit., 266fif.), acquisition of a particular language requires determination of the core lexicon of th at language as well as fixing parameters. With a view to restricting the possible param eters and the range of possible core lexicons, Fukui presents a hypothesis about param etrization and proposes restrictions on possible parameters, as follows. (9) Lexical categories are essentially invariant across languages; only functional categories are subject to crossHnguistic variation.^ (10) Possible param eters are restricted to those having to do with i. linear order, and/or ii. the existence or absence of fimctional categories (as well as features that they contain) .3 Fukui m aintains th a t the strict head-finality and the absence of functional categories in Japanese contribute to satisfy the general adjacency requirement on morphological processes, as schematized in (11a), but the - This statem ent is recapitulated by Fukui (1995: 337) as the Functional Parametrization Hypothesis, as follows. (i) Functional Param etrization Hypothesis: Only [+F] elements in the lexicon are subject to parametric variation. 3 As noted in chapter 2, despite his earlier contention as in Fukui (1986), Fukui (1995) claims th at those fimctional categories which are interpreted a t LF are existent across languages, although they can still be param etrized as to whether they induce agreement or not. Thus, as long as functional categories in languages like Japanese are concerned, Fukui’ s (1995) position is almost indistinguishable firom Kuroda’ s (1988). See the discussion in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 presence of functional categories like I (and C) and Spec-Head agreement in English would lead to the same requirem ent, as exhibited in (1 lb). (11) a. [... [... [... Vi] V I V2]v 2 Vajva (Japanese) I ÎI I b. V 2 ([cp... [c C) [if XP [r I [vi V i ... (English) t______ . IÎ__ I Fukui states th at while each step in (11a) is legitimate, the second step in (1 lb) violates either the HMC/ECP (because of the intervening head C) or the general morphological requirement of string adjacency (because of the existence of XP in Spec of IP for Case and/or agreement). However, Shibatani (1990) correctly points out, first, th a t there is a priori nothing th a t prohibits an inflected verb (i.e. V+I) firom raising to C in English; such a process is in fact attested in subject-auxiliary inversion,^ and furthermore, th at complex predicate formation is a productive process in some SVO languages such as Chichewa and Southern Tiwa, citing the following examples firom Baker (1988). section 2.3.2. in chapter 2 above. '‘ Even in V2 languages in the Germanic family, where V-to-I-to-C movement is attested, complex predicate formation does not seem to be productive (cf. Booij, 1988). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 194 (12) a. Chichewa M phunzitsi a-na-lemb-ets-a ana teacher AP-PAST-write-CAUS-ASP child The teacher made the children write.’ b. Southern Tiwa I-’ u’ u-kur-’ am-han lsS:2sO-bahy-hold-CAUS-PAST I made you hold the baby.’ Shibatani concludes th at the presence or absence of productive complex predicate formation in a given language is correlated, not with word order, but with the bound vs. free morphological distinction of auxiliary verbs in th a t language.5 6 5 Shibatani (1990) notes that Lahu is an SOV language, but unlike Japanese, it does not allow complex predicates. Since relevant data from Lahu has not been available to us at this point, it is not clear whether our contention in what follows is seriously jeopardized. ® Fukui (1995) withdraws his earlier proposal in response to Shibatani’ s criticism, and contends that the contrast in question between Japanese and English is reducible to the fact th at while the former is head-final and its verbs are bound morphemes, verbs in the latter are free morphemes. However, this characterization of the distinction between the two languages would, in a strict sense, not agree with his own functional param etrization hypothesis, to the effect th at only properties of functional categories can be param etrized across languages. This is so because verbs in any language are considered to constitute a major lexical category. Furthermore, examples Uke the following indicate th at Fukui’ s contention is not empirically sound. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 195 Note th a t Shibatani’ s (1990) contention in the previous paragraph is compatible with our hypothesis concerning the morphological nature of functional categories in Japanese, which is reproduced below firom chapter 2. (13) AH functional heads in Japanese are morphologically suffixal. It has been assum ed in chapter 2 th at verbal inflection in Japanese obtains as the result of morphological concatenation in PF. If auxiliary verbs like the causative and passive verbs are [+F] (as well as [+L], on a par with the hght verb then the proposed morphological condition in (13) applies to them. From this, the productivity of complex predicates simply follows. Take yom- ase-rare ‘ read-CAUS-PASS’ in (8a) as an example. Given th at the causative verb (s)ase and the passive verb rare are both morphologically subcategoried for V ® or v°, morphological concatenation applies to the syntactic representation in (14a) to produce the linear string in (14b). (i) John-wa susi-o tabe, sake-o nom -ta -TOP sushi-ACC eat drink-PAST ‘ John ate sushi and drank sake.’ Examples like (i) are often taken as cases of coordination. Tomioka (1991), for example, claims that (i) wiU be derived from two coordinated TPs by movement of T to C across the board. No m atter how cases Like (i) are derived, the im portant fact is th at the verb of the first conjunct appears without any overt s uffix Positing a zero suffix (or the trace of T, in Tomioka’ s terms) would not give a good loophole, since bound morphemes are, generally, those which m ust be supported by a phonologically overt morpheme. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 196 (14) a. (John-ga) (Mary-ni) rare (hon-o) Î Y (s)ase ‘ CAUS’ yom ‘ read’ ‘ PASS’ Morphological Concatenation b. ... yom + (s)ase + rare ‘ read+CAUS+PASS’ The exact positions of the arguments are not our immediate concern. In (14a), the auxiliary verbs yom read’ , (s)ase ‘ CAUS’ and rare PASS’ each head a maximal projection th a t is the complement of the next. Since the causative (s)ase and the passive rare are both specified as [V°/v°_ ] by hypothesis, each m ust attach to a host verb immediately to its left. As the result of morphological concatenation, the Linear string in (14b) obtains. To the extent th a t the proposed morphological condition for Japanese in (13) refers only to functional categories in this language, it is compatible with Fukui’ s (1988, 1995) Functional Param etrization Hypothesis, which is intended to restrict the range of possible param etrization across languages. Note that the condition in (13) contains a param etric specification of linearity. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 197 This is because a sufGx must come on the n ^ t of the stem or host th at it is attached to. Along the lines of Fukui, we consider that children fix the head parameter, [±head-initial], in one way or the other by learning morphological properties of functional categories in a given language th at encode the linearity of th a t language, in the way th at the condition in (13) does, hi terms of explanatory adequacy, this seems to be an appropriate way to fix a parameter. After reviewing some existing proposals, we will suggest in section 5.2. th a t the marked word order in a given language should be attributed to the specified morphological properties of functional categories in that language. 5.2. Previous Analyses 5.2.1. Ka)me(1994) Kayne (1994) presents a drastic proposal that completely eliminates the head param eter firom the theory of grammar. The Linear Correspondence Axiom (LCA) establishes that precedence is the direct reflection of c-command, such th at if a asymmetrically c-commands p, a must also precede p. Since specifiers (as adjuncts in Kayne’ s terms) asymmetrically c-command heads and complements, and heads asymmetrically c-command aU the m aterial inside complements, the basic constituent order within a phrase m ust be S^edfi.er)-H(ead)-C(omplement) across languages, with SVO being the direct Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 198 consequence of this constituent order. However, in languages like Japanese th at exhibit the strict OV order, the LCA necessarily predicts th at objects should have been moved leftward hrom a postverbal position, as indicated in the following. (15) OBJ VP t(OBJ) If the object moves around the verb to Spec of X, a designated functional category, then the OV order results. Furthermore, if the subject also moves to Spec of a functional category that is located higher than XP in (15), the resultant order will be SOV. However, Japanese seems to lack independent evidence for such leftward movement. As noted by Takano (1996), if Kasme is correct, complements of verbs, whatever category they belong to, m ust move leftwards, as shown below. (16) a. John-ga [opsakana-o] tabe-ta -NOM fish-ACC eat-PAST ‘ John ate fish.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 199 b. John-ga [pp Amerika-kara] ki-ta -NOM America-firom come-PAST ‘ John came from America.’ c. John-ga [a p kasiko-ku] nar-ta -NOM sm art become-PAST ‘ John became sm art.’ cL John-ga [a spP sakana-o tabe-te] mi ta -NOM fish-ACC eat-ASP try-PAST John tried eating fish.’ e. Mary-ga [cp John-ga sakana-o tabe-ta to] iw-ta -NOM -NOM fish-ACC eat-PAST th at say-PAST M ary said that John ate fish.’ Since Japanese is a strictly head-final language, complements, no m atter what category they belong to, always appear to the left of verbs, as exhibited in (16). As assumed widely, functional categories ti i ^ e r movement for the purpose of checking a particular feature. For instance, a [+wh] complementizer always attracts a constituent that has the same feature, and as assumed by Chomsky (1995, chap. 4), T with a strong D-feature induces movement of a DP to its Spec in the view of satisfying th e EPP. However, in the case observed above, Kayne would have to assume th a t the designated functional category, Xin (15), attracts whatever category appears in the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 complement position of a verb. In short, Kayne would be forced to trade off a restrictive theory of functional categories for the validity of the LCA. Furthermore, Kayne (1994) attributes the lack of overt wh-movement in languages like Japanese to the presence of movement of TP to Spec of C. If this is correct, then a somewhat strange picture concerning complementizers in Japanese will emerge. Compare the following with (16e). (17) [cp [tp Mary-ga [cp [ tp John-ga nani-o tabe-ta] to] -NOM -NOM what-ACC eat-PAST th a t iw-ta] ka] (osiete kudasai) say-PAST Ques (please tell me) ‘ (Please tell me) what Mary said th a t John ateT (17) is only minimally different firom (16e), in th a t the object of the embedded verb is replaced with a wh-phrase and, accordingly, the matrix clause ends with a question particle. As is well known, Japanese lacks overt wh- movement, and the wh-phrase is found in situ in (17). The comparison of (17) with (16e) would force us to say, in accordance with Kayne, that although the [+wh] complementizer ka does not induce wh-movement, even the [ wh] complementizer to th a t' triggers movement of TP. This would be an unexpected situation, in view of the general characteristics of [+wh] and [-wh] complementizers. Generally speaking, a one-way imphcation seems to hold that if [ wh] complementizers in a given language trigger movement, then Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 [+wh] complementizers in that language induce wh-movement, as is the case in V2 languages in the Grermanic family. 5.2.2. Takano (1996) and Fukui and Takano (1998) Assuming with Chomsky (1994; 1995, chap. 4) th at word order is decided only at PF, Takano (1996) makes a very interesting proposal that the basic constituent order is S-C-H, instead of S-H-C, across languages. In his terms, the S-H-C order is derived by movement of H to a h i ^ e r head. Takano assum es th a t linearization of constituents is a top-down process and it consists of two components. Demerge and Concatenate, exphcated as follows: (18) Linearization (Takano, 1996: 40) Applied to 2, Demerge yields {a, {2-a)}, a a constituent of 2, and Concatenate turns {a, ^-a}} into a+(2-a). (where (2-a) indicates the object resulting from detachm ent of a jfrom the syntactic object 2, and + indicates concatenation) Let us see how Takano’ s linearization works. In the following, first, 2 = XP (=X““ ). (19) a. 2=XP(=X®«) b. Y + 2= X’(=X”“ ) X P Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 202 c. Y + Z + 2=XP(=X««) d. Y +Z + X Takano postulates th a t only maximal projections are visible to Demerge as well as Merge, m aximality bmng defined relatively as in the BPST of Chomsky (1994) (cf. section 2.2.1. in chapter 2 above). In (19a), when Demerge targets XP as 2, Y is first detached since Y is maximal (since it does not project) but X is not. After Y is detached firom XP, Demerge can take X * as 2, as indicated in (19b), because it is now a maximal projection (=X““ ). Then, Z, a maximal projection, is detached firom X * . Finally, as shown in (19c), Demerge targets X P, which has become a maximal projection. Since it does not break down any further, it enters into concatenation as it is. Since Takano assumes th at the order of detachment by Demerge directly corresponds to the order for Concatenate, the resultant order is Y+Z+X, as exhibited in (19d). For Takano, param etrization with respect to word order is encoded in the presence or absence of V-raising and how far it raises, such th at SVO, but not SOV, should be derived by movement of V to a h i ^ e r head. This can be demonstrated in the following way. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 203 (20) a. b. A Takano assumes th a t the basic constituent order is universally S-C-H, as indicated in (20a). If nothing happens, linearization directly apphes to (20a). Demerge first takes T (=T*»«) as its target and it, as well as Concatenate, recursively applies from top to bottom to derive: vP+T = (SUBJ-h v O+T = (SUBJ+(VP+v^)+T = (SUBJ+((OBJ+V)-t- v))+T = SUBJ+OBJ+V-i-v+T, which yields the SOV surface order found in Japanese. If, as assumed by Chomsky (1995: chap. 4), V raises to vand subject raises to Spec of TP in English, the representation in (20b) obtains. In the same vein, linearization apphes to the structure in (20b) from top down. When it gets down to v', however. Demerge m ust detach V before VP. The reason is as follows. If VP were detached prior to V, w hat remains would be an adjunction structure <v, v> ; however, <v, v> does not dominate V, in the sense of May (1985). Since Takano defines constituents of 2 in term s of domination, V would not be a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 204 constituent of <v; v>, hence, invisible to Demerge J On this assum ption, linearization applies to (20b) to derive: SUBJ+T = SUBJ+(vP+T) = SUBJ+((tsuBJ+vO+T) = SUB J+((tsuBj+(V+ VP+\^+T) = SUBJ+((tsuBJ+((V+(OBJ+tv)+\^+T) = SUB J+tsuBj+V+OB J+tv+v+T, which yields the surface SVO order in English.® Although Takano’ s view of linearization in PF is compatible with the view of morphological concatenation adopted in this thesis, it seems to face some empirical and conceptual problems. F irst of all, since he assum es S-C-H as the universal base order across categories, he would predict th a t in head- initial languages like English, N, A and P, as well as V, should raise.® Indeed, he stipulates th a t there are “hg^t heads”, namely, «, a andp, th a t induce overt movement of these categories (Takano, op .cit.: 50). However, the existence of such functional categories does not seem to be independently motivated. A criticism similar to th a t against Kayne (1994) m ight apply. " The intuition behind this assumption is, Takano maintains, th a t V th at is adjoined to vbecomes an immediate constituent of v'. For details, see Takano (1996: p.42£f). ® IfV +vfurther raises to T, as assumed in French, V followed by v m u st be detached prior to vP when Demerge targets T’ for the noted reason. Given that French is different 6om English only in this respect, Takano’ s linearization derives SUBJ+V+v+tsuBJ+OBJ+tv+tv+T, which, correctly, corresponds to SVO. Moreover, if V further raises to C, for instance, and subject stays lower than TP, the surface VSO order will result. ® Fukui and Takano (1998) argue th at so-called complementizers in head- initial languages, e.g. th a t in English, are not in C, but rather they are in Spec of C as “C-checkers”. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 205 Fukui and Takano (1998), on the other hand, present a different view concerning the consistent head-finality in Japanese. Building upon Takano’ s (1996) idea of linearization, they propose that head movement is not adjunction to head, as is conventionally assumed, b u t it is a subcase of “substitution” to Spec. In their view, V and N in English, for example, raise to Spec of vand Spec of D, respectively, instead of adjoining to a head. Compare the following with (20b). b. (2 D a. John kissed tv ary the N NP D about John In (2 la, b), V is substituted to (the inner) Spec of vand N to (the inner) Spec of D, rather than being adjoined to v and n, respectively. Although it is generally agreed that only XPs (or X ™ » * s) can undergo substitution, Fukui and Takano, who, as we do in this thesis, assume the BPST of Chomsky (1994, > 0 Note th a t in Fukui and Takano’ s (1998) terms, determ iners (such as a, the, this, that, etc.) are not D heads, but they are merged as specifiers of Ds and as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 206 1995: chap. 4), state that substitution of a head to Spec is legitim ate since the raised head, which does not project, counts as as well as X ® (see the discussion in section 2.2.1. of chapter 2).^^ ^^linearization applies to (21a) and (2 lb) above, the resultant term inal strings wiU be DPsubj+tsubj+V+DPobi+tv+v+T for t h e former and th e + N + P P + tw + D for t h e la tte r , both of which yield the correct surface orders, John kissed M ary and the book about John, respectively. Furthermore, they maintain that the SVO order in French and the VSO order in Irish are derived by movement of V to Spec of T, as foUows. a result, formal features of Ds (i.e. FF(D)) are checked off. Fukui and Takano claim that their analysis of head movement as substitution to Spec is not necessarily incompatible with the chain uniformity requirement, suggested by Chomsky (1994, 1995: chap. 4), because the head and the tail of a chain created by head movement are “non-distinct”, in the sense that they are both X P (though the former is also X™ ^). Furthermore, they m aintain th at head movement ceases to be an exception to the extension requirement proposed by Chomsky (1993), to the effect th at merger always applies to the root node. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 207 (22) a. French b. Irish V T V DPsubj T As argued by Pollock (1989) and Chomsky (1991), French is different from English, in that V raises to T. In Fukui and Takano’ s terms, V raises to the inner Spec of T in French, as exhibited in (22a). In contrast, they m aintain, V raises to the outer Spec of T in Irish, as shown in (22b). ^ In each case, linearization will produce the correct surface order. Fukui and Takano indicate that the difference between French and Irish is so param etrized th a t T in the former attracts V prior to the subject DP, and T in the latter, vice versa. However, they do not give any explicit reason as to why this should be the case. In the minim alist terms, overt movement is motivated only for the purpose of checking a strong formal feature on a functional category. In the French and Irish cases under discussion, the relevant features are D- and V- 1 2 Following Bobalijk and Camie (1992), Fukui and Takano (1998) assum e that in Irish, V does not raise as high as C nor does subject stay in situ. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 208 features on T (i.e. [-D] and [-V], respectively). These features should be equally strong; hence, overt movements of the subject DP and V are both motivated. It is not clear why more than one strong features on a single functional head m ust be eliminated in a particular order in one language and in a different order in another. Unless they provide a principled reason, their analysis m ight lose its explanatory force. Furtherm ore, the proper treatm ent of adjuncts seems to be difhcult for the analysis proposed by Takano (1996). Adjuncts are categories adjoined to a maximal projection, as exhibited in the following. (23) Z YP X adjunct In (23), an adjunct (=X) is adjoined to YP (=Y ™ “ ). For the noted reason, when Demerge targets T (=Z ™ “ ), it must detach X prior to YP. Then, Concatenate will yield X+YP+Z. This amounts to saying th at an adjunct always precedes the category th a t it is adjoined to in the surface order. However, this does not always seem to be the case. It is widely assumed th at by-phrases in English passives are VP-adjuncts. Their adjunct status is dem onstrated as their optionality in examples like the following. (24) a. John was arrested by the pohce. b. John was arrested. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 209 A ssu m in g th a t the Enghsh passive construction involves the use of a specific l i ^ t verb th at absorbs the external theta role of the main verb and be raises to T, as argued by Emonds (1978) and Pollock (1989), we can represent the derivation of (24a) in Takano’ s terms, as follows. (25) 'Pobj John 'ASS by the police topbbj tv arrest In (25), the by-phrase is adjoined to VP, and T attracts D Pobj in its Spec to check its D-feature. As always, linearization applies in a top-down fashion. When Demerge targets T, be m ust be detached before for the reason noted above. Suppose th a t when it comes to targeting v f, it detaches V prior to the VP-adjunct. Then, Concatenate will yield the string: DPobj+be+V+PP+ toPbiy+tv+vfASS +T, as desired. However, it is not entirely clear why V should be detached prior to the by-phrase when Demerge applies to vP. The opposite order would result in a wrong surface string, i.e., *John was by the police arrested. In this respect, Fukui and Takano’ s (1998) conception of 1 3 Be might be either raised firom a lower head or directly introduced to the structure by merger to T. However, the choice between the two options is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 head movement would yield a better result. Under their analysis, be and the main verb are accommodated in Spec of T and Spec of v, respectively. (26) John vP arrest Vt>ASS by the pohce toP bbj The ambiguity in the order of detachment between V and the by-phrase does not arise in (26). V is detached by Demerge when it targets vP, but the by phrase is detached only when it applies to v * , one cycle down. Thus, the surface order th at is produced by Concatenate will be unambiguously: DPobj+be+V+PP+ tD Pbbi+tv+vt>A SS +T, which is the desired result. Here, a tacit assumption is th at the auxiliary be is merged to T, so that it will occupy the inner Spec of T, before the object DP is raised to check the D-feature in T. Then, the sam e question as we have raised concerning the SVO order in French and the VSO order in Irish might apply; it is unknown why T in English m ust elim inate the feature th at can be checked off by an auxiliary like immaterial to the text discussion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 211 be, prior to its D-feature for the EPP. Putting this potential problem aside, passives of ditransitive verbs like the following seem to cast doubt on the straightforward application of Takano’ s or Fukui and Takano’ s view of phrase structure and linearization. (27) a. The villagers provided the soldiers with food and water. b. The soldiers were provided with food and w ater by the villagers. c. ??The soldiers were provided by the villagers with food and water. (28) a. Mary put the vase on the coffee table. b. The vase was p u t on the coffee table by Mary. c. ??The vase was p u t by Mary on the coffee table. Verbs like provide and p u t in Enghsh are ditransitive, in the sense th at they take a PP complement as well as a DP object. When these verbs undergo passivization, the object DP is moved to subject position, leaving the PP complement internal to VP. The unmarked string order between the PP complement and the by-phrase is PP-by, as in b-sentences. The reversed order seems to be marked, as indicated by c-sentences. However, the opposite would be predicted in Takano’ s and Fukui and Takano’ s approach. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 212 (29) vP b y ... P P com plem ent When linearization applies to v', the hy-phrase must imm ediately be detached, since it is only adjoined to VP. On the other hand, the PP complement can be detached only when Demerge targets VP, which dominates it. Hence, the straightforward application of linearization proposed by Takano and Fukui and Takano would yield by-PPcompiement as the unmarked order, contrary to fact. The French .feire-par construction and its Italian equivalent fare-da construction point to the same problem. According to Burzio (1986), these constructions involve VP-complementation and par-phrases in French and da- phrases in Italian are adjuncts ju st like by-phrases in English passives. The following examples are firom Italian, adopted firom Burzio (1986: 228, 232). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 213 (30) a. Maria ha fatto riparare la macchina da Giovanni. has made repair the car by M aria had the car repaired hy Giovanni.’ b. Maria ha fatto riparare la macchina. has made repair the car M aria had the car repaired.’ c. *Maria ha fatto (da) Giovanni riparare la macchina. has made by repair the car The adjunct status of the da-phrase in the fare-da construction is indicated by its optionality; compare (30a) and (30b). However, it cannot precede the m aterial inside VP, as shown by the ungrammaticality of (30c). If, like the by-phrase in Enghsh, the da-phrase is adjoined to VP, then its VP-finality is not expected by Takano’ s or Fukui and Takano’ s linearization convention. This is because, in their view, the da phrase m ust be detached prior to the m aterial internal to VP. Recall that for Takano (1996) and Fukui and Takano (1998), the difference between the VO and OV orders is param etrized in term s of the presence or absence of V-raising. As is weU known, Chinese exhibits the surface VO order. Following Huang (1996), Fukui and Takano state th a t Li (1990) claims th at Chinese is head-final across categories. This is not ungrounded, since despite the VO order, categories like CP and DP are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 214 the VO order in Chinese is the resu lt of movement of V out of VP. However, this is not as obvious as they might assume it to be. Transitive verbs in Chinese often take a resultative complement (RC) th at modifies the object, as illustrated in examples like the following. (31) a. Ta xie zi [Rczai zhuozi-shang] he write character on table-top H e is writing characters (and as a result, they are) on the surface of the table top.’ b. Ta [zai zhuozi-shang] xie zi he on table-top write character ‘ Sitting on top of the table, he is writing characters (say, on paper).’ or H e is writing characters (and as a result, they are) on the surface of the table top.’ The postverbal PP in (31a) modifies the object, and it renders a resultative interpretation. In this case, one could imagine that V has been raised out of VP. What is rather surprising is the fact that in (3 lb), where the same PP obviously head-final in Chinese, and “prepositions” in this language seem to have been historically derived firom verbs (hence, they are often called “coverbs”; cf. Li and Thompson, 1981). In Li’ s terms, the object, which is base-generated in a preverbal position, is moved postverbally for a case- theoretic reason. However, such a postverbal case position is not well motivated under our approach. Hence, we will assume th a t Chinese indeed exhibits “mixed” orders, depending on types of categories. The data presented in (31) is provided by Zoe Wu. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 215 appears preverbaUy, the resultative reading is still maintained. It can modify the object as well as the subject; hence, the sentence is ambiguous, as indicated. Since it seems fair to assume th a t only VP-intemal PPs can modify objects as RCs, and there is no compelling reason that the PP in (3 lb) should be raised higher than V, the retained resultative interpretation in (3 lb) indicates th a t V remains internal to VP. If V in Chinese does not raise at all, its VO order cannot be captured by Takano’ s (1996) or Fukui and Takano’ s (1998) system. Although Takano (1996) and Fukui and Takano (1998) present a very interesting alternative to Kayne's (1994) LCA, and we share many of the basic assumptions conceminginflectional morphology with those authors, their particular analysis of crosslinguistic word orders leaves some im portant conceptual questions yet to be answered, and empirically speaking, the straightforward apphcation of their linearization convention does not always bring about correct results. It should also be noted th at their linearization convention predicts th a t sentences in SVO and VSO languages always end with an empty functional head. It is unclear whether this is conceptually and empirically favorable. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 216 5.3. An alternative 5.3.1. Deriving the strict head-finality of Japanese In this section, we propose th at the order of constituents within a phrase is S-H-C across languages, as proposed by Kayne (1994), unless required otherwise. Along the hnes of Kayne (1994), Takano (1 9 9 6 ) and Fukui and Tak an o (1 9 9 8 ), we assume that VSO languages necessarily involve movement of V to a higher functional head. However, contrary to Takano and Fukui, SVO languages do exhibit the unm arked order in our view, On the assumption th a t V does not overtly raise in Chinese, the VO order in th a t language should be taken not as the result of V-raising, but as the simple reflection of the unmarked order. Following the spirit of Fukui (1995), we consider th a t an alternative surface order in a given language is possible only if, otherwise, morphological requirements on functional categories in th at language would not be satisfied. Suppose th a t a functional head F° takes as its complement a projection of a lexical head L°. Assuming with Chomsky (1994, 1995: chap. 4) th at linear order is irrelevant to the core syntactic computation, the only relevant relation We do not commit ourselves to the choice between head movement as substitution to Spec, as Fukui and Takano (1998) claim, and head movement as adjunction to head. Furthermore, we are not claiming th a t every major sentential constituent must stay in situ in SVO languages. If some constituent is raised for an independent reason in, say, English, our proposal will not be jeopardized. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 217 is the hierarchical relation in which F° c-commands the maximal projection of IP. Following H alle and M arantz (1993), we also assume th at this (limited) configurational information is preserved in the morphological component of PF. In the u n m ark ed case, the hierarchical relation where F° c-commands LP (= L ™ ax) is linearized in PF, as follows. (32) X Y L° In the un m arked case, constituents are ordered in such a way that if a asymmetrically c-commands p, a precedes p, as established by Kayne’ s LCA.‘ ® However, if F° is morphologically specified as a suffix and is subcategorized for L°, i.e., if it has morphological features (MFs) Kke [-f-sufGx] and [L°_ ], the linearization in (32) would not satisfy the requirem ent on F°. If it is to be met, linearization m ust produce the following alternative order. Kayne (1994) assum es th at the LCA is at work throughout a derivation; hence, the constituent order is universally S-H-C firom the base. However, we consider that, following Chomsky (1994, 1995, chap. 4), the LCA is only relevant to the PF side of a derivation, and furthermore th at it simply defines the unm arked order. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33) F X Y 218 MF(FO)={[+suffix], [L° _ ],...} If F° is to suffix to L° under string adjacency, F° must appear immediately to the right of L°. Thus, only in the order in (33) will the morphological requirement on F° be satisfied. Recall th at as we have claimed above, functional categories in Japanese have the following morphological property. To repeat: (34) AU functional (i.e. [+F]) heads in Japanese are [+suffix]. We have also claimed in chapter 2 th at except for Q-particles, aU functional categories in Japanese are subcategorized for a specific head, as foUows. (35) functional categories examples categorial feature specification morphological subcategorization features complementizer (C) to ‘ th a t’ , ka ‘ Ques’ [+F, -LI [To_i tense (T) ru ‘ present’ , ta ‘ p ast’ (+F, -LI ['^ -l Hght verb (\^ su do,’ 0 [+F,-fL 1 [V°_l determiner (D) 0 [+f ,± li [No_l In Japanese, C is a suffix and subcategorized for T. These morphological Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 219 properties of C require linearization to apply in such a way th at it places C immediately to the rig^t o£T in linear string. In a sim ilar vein, T is placed immediately to the right of v, and v appears im m ediately to the right of V. Thus, the Japanese CP, exemplified by (36a), will have the morphological structure after linearization in (36b). (36) a. (Mary-ga) [cpJohn-ga hon-o yom ta to] -NOM -NOM book-ACC read PAST th at (iw-ta) (say-PAST) ‘ (Mary said) th a t John read a book’ b. C0MF(C0)={[+su£fix], [To_],...} MF(To)={[+suffix], [v O _ ] ,...} MF(v°)={[+s\iffix], [V°_ ],...} D 0 Noun phrases in Japanese also exhibit the strict head-finality, as demonstrated in the following. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 220 (37) a. sono hon(-o) th a t book(-ACC) th a t book’ b. *hon sono(-o) book that(-ACC) (38) a. akai hon(-o) red book(-ACC) ‘ read book’ b. *hon akai(-o) book red(-ACC) (39) a. [John-ga ec2 yom-ta] hon2(-o) -NOM read-PAST book(-ACC) ‘ the book that John read’ b. *hon2 [John-ga ec2 yom-ta] ( o) book -NOM read-PAST(-ACC) (40) a. [John-ga sono hon-o yom-ta] (to yuu) zizitu(-o) -NOM th at book-ACC read-PAST (that) fact(-ACC) ‘ the fact that John read the book, b. *zizitu [John-ga sono hon-o yom-ta] (to yuu) ( o) fact -NOM th at book-ACC read-PAST (that) ( ACC) As illustrated in (37) - (40), demonstratives, adjectives, relative clauses and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 221 noun complement clauses all precede a head noun. This is required, since, as we have assumed in chapter 2, Japanese noun phrases are DPs and D has MFs including [+suffix] and [N°_ ]. / V D O MF(DO)={[+su£fix], [No_],.„l X (where X and Y are modifiers or Y N O complements) Since D is a suffix and subcategorized for N, linearization must place it immediately to the right of N in linear string. The morphological structure in (41) is the only possibility th at satisfies the MFs of D. This is, we claim, the source of the head-finality of DPs as well as the other categories in Japanese and other languages th a t share the same property. 5.3.2. “Mixed” orders in Chinese As noted above, M andarin Chinese exhibits the VO order in the unmarked case, although objects may be preposed due to topicalization and localization (cf. Huang, 1982; Cheng, 1991; Shyu, 1995). 1 9 However, Li (1990) presents a different view. See footnote 14. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 222 (42) a. Zhangsan meitian nian na-ben shu every day read that-CL book ‘ Zhangsan reads that book every day.’ b. Zhangsan meitian na-ben shu nian every day that-CL book read TEvery day, Zhangsan reads th a t book (but not, say, this book).’ c. na-ben shu Zhangsan m eitian nian that-CL book every day read ‘ As for th a t book, Zhangsan reads it every day.’ (42a) exhibits the unmarked VO order in Chinese. If the object is preposed around the verb as in (42b), it gets contrastive focus. If the object is preposed to the sentence-initial position, it usually receives a topic reading. This head-initiality is more strictly observed in PPs. (43) a. Zhangsan [ppzai Meiguo] zhu in the U.S. live ‘ Zhangsan lives in the U.S.’ b. *Zhangsan [ppMeiguo zai] zhu the U.S. in live The contrast between (43a) and (43b) indicates that PP, as well as VP, is head-initial in Chinese. However, as noted in the literature, Chinese noun phrases exhibit strict Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 223 head-finality, as dem onstrated in the prenominal demonstrative in the examples in (42) and the following. (44) a. wo-de shu I-PNM book (PNM: prenominal marker) ‘ my book’ b *shu wo(-de) book I-PNM (45) a. hen chang-de shu very long-PNM book a very long book’ b. *shu hen chang(-de) book very long-PNM (46) a. [Zhangsan du-guo ec2]-de shuz read-PERF -PNM book the book th a t Zhangsan has read’ b. *shu2 [Zhangsan du-guo ec2l(-de) book read-PERF -PNM (47) a. [Zhangsan du-guo na-ben shu]-de shishi read-PERF that-CL book-PNM fact the fact th a t Zhangsan has read th a t book’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 224 b. *shishi [Zhangsan du-guo na-ben shu](-de) fact read-PERF that-CL book-PNM As dem onstrated above, as in Japanese, possessives, adjectives, relative clauses and noun complement clauses m ust all precede a head noun in Chinese. Furthermore, the so-called question particles in Chinese, m a for yes-no questions and ne for wh-questions, appear at the end of clauses. (48) a. Zhangsan zuotian chi le shenme ne yesterday eat-ASP what Q u e s [ + w H i W hat did Zhangsan eat yesterday?* b. Zhangsan meitian zuo fan m a every day make meal Ques[.vrai Does Zhangsan cook every day?’ If these question particles are complementizers, CPs m ust also be head-final in Chinese. In sum, Chinese exhibits “mixed” word orders. While VPs and PPs are head-initial, noun phrases and CPs are head-final. As we have seen above, it is hard to believe that V overtly raises in Chinese, and hence, its VO order cannot be the result of V-raising, as Fukui and Takano (1998) argue. It seems to be the case that the VO order in Chinese is what the LCA brings about in the unmarked case. We assume th at the head-initiality of PPs is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 225 also a case of the unm arked linearization forced by the LCA. However, this unmarked linearization does not extend to noun phrases and CPs in Chinese. Our proposal about the head-finahty of noun phrases and CPs in Chinese is that these categories have some spedfic MFs. Suppose that noun phrases in Chinese are DPs as well as those in Japanese, and also suppose th at like the Japanese counterpart, D in Chinese is a sufGx th at is subcategorized for N. Then, the head-finality of Chinese noun phrases follows in exactly the same way that the Japanese equivalents are head-final, as demonstrated below. (49) a. yi-ben hen chang-de shu one-CL very long-PNM book ‘ one very long book’ D O MF(Do)={[+su£fix],[No_],„.l yi-ben hen chang-de shu Since D in Chinese is, by hypothesis, [+suffix] and subcategorized for N, it must appear immediately to the right of N, as indicted above. On the other hand, this analysis does not directly extend to Cs, since they do not seem to be subcategorized for a particular head. They appear to the right of object DPs in (48a, b), and they may immediately follow verbs as well. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 226 as exemplified in the following examples. (50) a. shenme ren zuotian lai ne what person yesterday come Ques[+wHi Who came yesterday?* b. Zhangsan meitian qu m a every day go Ques[.wH] T)oes Zhangsan go every day?’ This indicates th at C in Chinese is a phrasal suffix th at is morphologically subcategorized for TP. However, since C is syntactically subcategorized for TP, its repeated specification in morphology is only redundant. Hence, it suffices to spedJfy C as [+suffix] in morphology. If this is on the right track, the morphological representation of the sentence in (48b) will look like the following. 2 0 2 0 In the morphological representation below, we simply assum e th at the temporal adverb is adjoined to T and the subject is moved to Spec of T (possibly, for topicalization). There are of course some other possihilities about the location of subjects and adjuncts in Chinese. However, this is not an immediate concern in the text discussion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 227 (51) ma MF(C)={[+suffix], Zhangsan m eitian Although Chinese is head-initial with respect to V, y and T, C appears at the end of a sentence. Since C in Chinese, by hypothesis, has MFs including [4-sufGx], linearization m ust place it immediately on the right of TP. Since ordering is an absolute necessity only for articulatory-perceptual purposes, it is natural to assume that linearization is a PF process. Linearization ahgns phrasal constituents in the S-H-C order unless otherwise required. However, we have claimed th at a particular set of morphological features (MFs) on functional categories can give rise to alternative orders. Since it is possible th at only some functional categories but not others have such MFs within a single language, “mixed” orders can arise in such a language. Only along these hnes can the fact of “mixed” orders of constituents in languages like Chinese be naturally captured. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 228 APPENDIX I On S e lls ’ (1995) L exical A n aly sis of In fle ctio n Although we agree with Sells (1995) in that syntactic head movement is not an appropriate mechanism to account for inflectional morphology in languages like Japanese, we m ust depart firom him because of his particular analysis of inflected words. He claims that inflected words in Japanese (and Korean), either verbal or nominal, are not ri^t-headed as standardly assumed, but left-headed. In his terms, morphological combination between a root and a suffix takes place in the following way (firom Sells, 1995: 308). (1) categorial - -»XP^-------- combinatoric information n. information (what heads (what can be Xs select for) Root ... Suffix right sister) According to Sells, it is always the root (or ultimately, the initial stem of a lexical category in the case of a complex including more than one such stems) that decides the category of the whole complex, whereas a suffix contributes to the complex with a combinatoric feature, or TYPE feature, th at specifies what category the complex with th a t suffix can be the left sister of. The class of TYPE features that Sells propounds includes [V-SIS], [N-SIS] and [ROOT]. The root attached to by a suffix with the [V-SlS] feature can only appear as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 229 the left sister of a verb, and the one suffixed with an affix with the [N-SIS] feature appears to th e left of a noun. [ROOT] demarcates the right edge of a complex; the complex suffixed with an element having this feature can stand alone, not having to form a further complex word with anything on its right. In this view of morphology, the verbal complex in (2), for instance, will be represented as in (3). Oh (3), the indices on verbal segments are only for expository purposes.) (2 ) John-wa susi-o tabe-te-mo m i-ta -TOP sushi-ACC eat-ASP-also see-PAST ‘ John also tried eating sushi.’ (3) U-TYPE] tabe -te -ta mi -mo By the convention in (1), categorial features are always inherited from the lefthand side of the morphological tree. The verbal complex in the left half of the tree in (3), tabe-te-mo ‘ eat-ASP-also’ , is ultim ately headed by the verb stem Vi tabe eat’. Although V i has no TYPE indication (hence, [-TYPE]), meaning that it cannot be a sister of anything on its own, the suffix -te has a combinatoric feature [V-SIS], and it is inherited by Vg. This is consistent Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 230 with the fact th at while the verb stem alone cannot appear to the left of the verb m i ‘ see’ , suffixation with-te (without any further suffixation) makes it possible, as indicated in the following. (4) a. *John-wa susi-o tabe mi-ta -TOP sushi-ACC eat see-PAST ‘ John also tried eating sushi.’ b. John-wa susi-o tabe-te m i-ta -TOP sushi-ACC eat-ASP see-PAST John also tried eating sushi.’ Sells assumes th at Q-particles like -mo ‘ also’ have a [V-SIS] indication. According to the convention in (1), the [V-SIS] feature of -mo percolates to V 3 in (3). Note th at since V 2 already has [V-SIS], inherited from -te, -mo virtually makes no contribution to morphological combination. This explains the noted “transparency” of Q-particles in term s of selection (see the related discussion in chapter 2). In the meantime, on the righthand half of the tree in (3), the verb stem m i see’ is combined with the suffix - ta TA ST. Since the latter is spedfred with the [ROOT] feature, the combination can stand alone, as indicated in (5a), with the possibility of further suffixation with something hke a complementizer as in (5b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 231 (5) a. Mary-ga nizi-o m i-ta -NOM rainbow-ACC see-PAST M ary saw a rainbow.’ b. John-ga Mary-ga nizi-o mi-ta-to it-ta -NOM -NOM rainbow-ACC see-PAST-COMP say-PAST (koto) (the fact that) ‘ John said th at Mary saw a rainbow.’ In (5b), the combinatoric feature [ROOT] associated with the verbal complex mi-ta see-PAST, supplied by -ta, is changed into [V-SlS] by further suffixation with -to th a t’ ; hence, the complex mi-ta-to see-PAST-COMP’ may appear to the left of another verb stem iw say’.^ On the root level of V° in (3), the two complex verbals are compounded together with relevant features all inherited from both roots V3 and V5. Although Sells (1995) is successful in capturing certain facts of verbal complexes in languages like Japanese, his view of morphology with the convention in (1) seems to pose both empirical and conceptual problems.- To 1 Indeed, this TYPE feature-changing seems to pose a conceptual problem, to which we will return shortly. 2 Besides the selectional “transparency” of Q-particles, his analysis straightforwardly accounts for the fact th a t a sentence-final complex verbal in Japanese always counts as a phonological word. As indicated in (i), the verb stems tabe ‘ eat’ and m i ‘ see’ are both inherently accented. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 232 begin with, since Sells’ lexical analysis of verbal inflection in Japanese amounts to saying th a t a sentence-final verbal sequence, no m atter how complex it is, is formed as a single verb in the lexicon, it is expected that all sentences in this languages virtually turn out to be monoclausal. However, this resultant state of affairs wül make syntactic locality hard to express. For instance, the following examples including negative polarity items (NPIs) with -sika ‘ anything b u t’ . The bracketings in (6 ) are given in Sells’ terms.^ (i) a. TA be (-te) H L L b. MI (-te) H L It is generally accepted th at phonological words in pitch-accent languages hke Japanese may have a t most one pitch peak, and accent falls on the last syllable of a stretch of high tones (cf. Kindaichi, 1960; McCawley, 1968). Hence, each of the verb stems in (i) can potentially constitute the nucleus of an independent phonological word. However, if the two verb stems are combined as in (2 ), the resultant complex will have only one pitch peak, as exhibited in (ii) (ii) TA be -te -mo -mi -ta H L L L L L This simply follows firom the lexical analysis of morphological complexes that Sells proposes, since, in his view, they are formed as words in the lexicon and then introduced to a syntactic derivation. Note, however, th at it is not necessary th at those morphological complexes be words in syntax. In our approach, while constituents of those complexes can be separate heads in syntax, they are combined into phonological words by morphological concatenation at PF. 3 Sells assumes with Fukui (1986) that Japanese sentences are projections of V, with recursion fireely allowed at the V’ -level. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 233 (6 ) a. [\r Mary-ga [v- nizi-sika [vo mi-na-katta]]] (koto) -NOM rainbow-SIKA see-NEG-PAST (the fact that) M ary did not see anything but a rainbow.’ b. ?* [ v John-ga [\^ Mary-ga [ v nizi-sika [vo mi-ta-to-iw -NOM -NOM rainbow-SIKA see-PAST-COMP-say -ana-katta]]]] (koto) -NEG-PAST (the fact that) ‘ John did not say that Mary saw anything but a rainbow.’ c. [\r John-ga [\r Mary-ga [ v nizi-sika [vO m i-na-katta-to -NOM -NOM rainbow-SIKA see-NEG-PAST-COMP -iw-ta]]]] (koto) -say -PAST (the fact that) As first noted by MurakL (1978), the NPI-creating suffix -sika anything but’ is only hcensed by a “clausemate” negation, “clausemate”, in the conventional sense.4 In Sells’ terms, all sentence-final verbal complexes are formed as words (i.e. in (6 )) in the lexicon. Then, the three clauses in (6 ) should all count as monoclausal and the complex verb of each sentence includes NEG, but the NPI in (6 b) is not legitimate. Note, especially, th at the contrast between (6 b) and (6 c) seems minimal; while the first verb stem is suffixed For recent proposals on NPI hcensing in Japanese, see, for example, Aoyagi and Ishii (1994a, b) and Nishioka (1994). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 234 with NEG in (6 c), the second one is, in (6 b). Given the feature inheritance procedure noted above, the topmost node of the complex predicate in either case should inherit the [+NEG] feature th at can license an NPI. However, this is not home out. Thus, Sell’ s lexical analysis of complex verbals is problematic, in th at it may obscure syntactic locality. Another objection to the lexical analysis of inflectional morphology th a t Sells (1995) proposes is a conceptual one. The principle th at constrains feature inheritance in Sells’ theory, if existent at all, seems to contradict the standardly accepted conventions of feature percolation in important respects. One widely assum ed set of feature percolation conventions is provided by Lieber (1992). The Feature Percolation Conventions (Lieber, 1992: 92) (7) Head Percolation Morphosyntactdc features are passed from a head morpheme to the node dominating the head. Head Percolation propagates the categorial structure. (8 ) Backup Percolation If the node dominating the head rem ains unm arked for a given feature after Head Percolation, then a value for th at feature is percolated from an immediately dominated non-head branch m arked for th at feature. Backup Percolation propagates only values for unmarked features and is strictly local. The point of Lieber’ s conventions in (7) and (8 ) is th at whenever a head and a non-head compete for a value of the same feature, the head always wins; conversely, percolation of a feature 6 om the non-head is possible only if the head is not specified for th a t feature. This is the standardly accepted view of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 235 how feature percolation proceeds. Despite the first impression th at the principle of feature inheritance in ( 1) gives, a suffîx, though a non-head in Sells’ terms, m ust be able to propagate some features other than TYPE features. One such feature is that which determines the complementation type, i.e. a FORM feature for SeUs. This is necessary in order to make a distinction between, for instance, the following pair of sentences firom our earlier data base. (9) a. John-wa susi-o tabe-te-mo-mi-ta -TOP sushi-ACC eat-PERF-also-see-PAST ‘ John tried also eating sushi.’ b. * John-wa susi-o tabe-ni-mo-mi-ta -TOP sushi-ACC eat-IMPERF-also-see-PAST cf. John-wa susi-o tabe-te-mi-ta -TOP sushi-ACC eat-PERF-see-PAST ‘ John tried eating sushi.’ The difference between (9a) and (9b) is quite minimal; whereas the first verb stem tabe ‘ eat’ in the former is suffbced with the perfective aspectual morpheme -te, the same verb stem in the latter is suffixed with -ni, the imperfective aspectual morpheme. Since the rest is aU the same in the two examples, the contrast in grammaticahty m ust be attributed to the difference in aspectual mode between the two suffixes th at each attach to the first verb Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 236 stem; the second verb stem m i ‘ see, try', in this particular use, requires th a t the verb that it selects for be in the perfective aspect. In order to account for this contrast. Sells assigns each of the suffixes in question (i.e. COMPs in his notation) a FORM feature that decides a particular type of complementation. Suppose that -te is assigned a FORM feature, say, [COMPa], distinct from [COMPp] assigned to -ni. The representation of the verbal complex of (9a), more elaborately annotated than (3), will look like the following.^ ' The types of features th at SeUs introduces are summarized as follows. PRED: predicative (semantic) feature FORM: complementation type feature TNS: tense feature -TYPE: no type V-SIS: sister to a projection of V N-SIS: sister to a projection of N X-LJM: inner particles (e.g. -mo, -sae, -dake) Z-UM: outer particles (Nom, Acc, Gen, Top) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 237 ( 1 0 ) VO Va PRED = read' 1 FORM = COMPa X-LIM = also' TYPE: V-SIS V2 PRED = ‘read’ FORM = COMPa TYPE: V-SIS Suf Suf tabe -te -mo PRED = ‘read’ FORM = COMPa X-LIM = also -TYPE TYPE; V-SIS TYPE: V-SIS Vs PRED = ‘try- TNS = PAST TYPE; ROOT Suf -mi PRED = 'try' -TYPE -ta TNS = PAST TYPE: ROOT For the sake of discussion, we focus on how the selectional property of m i is met by feature percolation in (1 0 ). Because V i is not specified for a complementation type, the FORM feature of -te percolates to V? (by (8 )). Since Vzis the head of Va, the FORM feature [COMPa], now associated with V 2, percolates to Va (by (7)). Now th at [COMPa] has entered a local domain, the selectional property of V4 (or the suffixed verb V5) is satisfied. As discussed above, inheritance of FORM features follows firom standard principles of feature percolation like those in (7) and (8 ). However, percolation of TYPE features does not conform to the general rule. Sells Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 238 (1995: 309) notes, “various sufSxes change the TYPE specification of the word that they are part of, b u t in the syntax TYPE remains invariant, always inherited firom the right.” A case in point is given in the following. ( 1 1 ) a. (tegami-ga) Canada-kara Id-ta letter-NOM -firom come-PAST ‘ (A letter) came firom Canada.’ b. Canada-kara-no tegami -firom-GEN letter a letter firom Canada’ c. *Canada-kara tegami In (11a), the postpositional phrase Canada-kara firom Canada’ appears as the left sister of a verb ki(-ta) ‘ come(-PAST)’ . This means th at the postpositional phrase m ust be associated with a [V-SIS] feature. Since the noun stem Canada is, by nature, unspecified for a combinatoric type, this TYPE feature should be inherited firom -kara. On the other hand, the same postpositional phrase is further sufhxed with the genitive case marker -no in (1 lb). Because, as indicated in ( 1 Ic), the postpositional phrase without -no may not appear as the left sister of a noun, it is -no that changes the TYPE specification firom [V- SIS] to [N-SIS]. The steps which determine the TYPE specification are represented in the foUowing. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (12) a. ,V ’ [ R O O T ] N ’ lv^is] V’ [ R O O T ] ‘ Z)i :) Narv-sis] V [R O O T ] N I [-TYPE] Suf[ V-SIS] Canada -kara ki-ta 239 N’ [-ty pe] N [N-SIS] N3[N-SIS] N [-TYPE] Z ) N[-type] N 2[V-SIS] Suf [N-SIS] NI [ - T Y P E ] Suf[V -SIS] Canada -kara -no tegarni Feature calculation proceeds bottom up. As shown in (12a), the noun root Ni, th o u ^ being the head of Ng, is not specified for a combinatoric type; hence, the TYPE feature [V-SIS] is supplied firom the suffix -kara. Now th at Ng has a TYPE specification, [V-SIS], it may appear as the lefthand sister of V. This process conforms to the feature percolation conventions in (7) and (8). In (12b), the first step of feature inheritance is identical to th at in (12a), Ng being provided with [V-SIS] firom the non-head suffix. At the next cycle, the TYPE specification is changed to [N-SIS] by the second suffix -no. However, this should not be permitted by the feature percolation conventions in (7) and (8), because the head of Na is Ng by definition, and Ng has a [V-SIS] specification by inheritance. The fact is, as noted by SeUs himself, th at unUke categorial features, TYPE features are always suppUed firom the right. It is now obvious that TYPE features have a very special status, very Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 240 different from th a t of categorial features or FORM features, in Sells’ theory. Recall that TYPE features determine w hat can appear as the right sister in syntax and other features like categorial features constrain the possibility of morphological concatenation. This m eans th at TYPE features are syntactic in nature, whereas other features are morphological. Although Sells does not provide a particular theory of feature inheritance in general, his postulation of TYPE features with their distinct inheritance properties would amount to saying th at while complex words in languages hke Japanese are left-headed in morphology, they are right-headed in syntax, as indicated in the following. (13) a. Morphology b. Syntax X[toot Y suf X r o oI Y su f [categorial features] [TYPE features] The state of affairs in (13) is reminiscent of the theory of autolexical syntax, proposed by Sadock (1991), where derivations in morphology and syntax proceed in a parallel fashion. A full evaluation of the tenet of autolexical syntax is well beyond the scope of our im m ediate concern. However, it has now become clear th a t the postulation of TYPE features as a specific theoretical apparatus seriously weakens Sells’ (1995) central claim th at complex words are uniformly left-headed in Japanese (and Korean) throughout the derivation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 241 While we agree with Sells (1995) th at inflectional morphology is not the direct resu lt of head movement in syntax, we depart from him because we cannot accept his lexical analysis of inflectional morphology for the noted reasons. Notice th at under our analysis of Q-p articles, the facts th at Sells wants to account for with his lexical theory can straightforwardly be captured. (14) (John-wa) susi-o tabe-te-mo m i-ta -TOP sushi-ACC eat-ASP-also see-PAST ‘ (John) also tried eating sushi.' A sn™ ™ susi-o tabe In syntax, as exhibited in (14), the verb m i ‘ see’ , in this particular sense, selects for Asp™ »* headed by the perfective aspectual -te, which, in turn, selects for V ™ »*. The Q-particle -mo also' is a head but it does not project. Since it is only adjoined to a projection of Asp by Merge, it does not intervene in the selectional relation between V° and Asp™ »* (or ultim ately Asp°). When the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 242 structure in (14) enters the morphological component in PF, the heads are morphologically concatenated, according to the morphological properties of each head (cf. section 2.3. of chapter 2 and chapter 5); hence, a phonological word tabe-te-mo-mi-ta ‘ eat-ASP-also-see-PAST will be yielded in PF. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 243 A PPENDIX n O n K oizum i’s (1995) A rg u m en ts fo r O v ert V -raisin g The success of our account of su-support in section 2.4.1. iu chapter 2 is contingent upon the absence of (at least, overt) movement of V to T. Recall that our analysis of Q-particles has stripped them of the intervention effect with respect to head movement, because they are now simply adjuncts. If movement of V to T should take place, nothing seems to prevent it. Many Japanese linguists have postulated syntactic movement of V to T with a view to accounting for facts of verbal morphology. Under our current assumption that inflectional morphology is in PF, the mere existence of inflected verbs and complex verbs does not constitute evidence for syntactic verb movement. However, few have provided independent empirical evidence for it. ^ A lthou^ I Following Huang (1988), Otani and Whitman (1991) claim that the “sloppy identity” reading of (ib) results &om an elided VP after V-to-T raising, as indicated in (ii). (i) a. John-wa [zibun-no temagi]-o sute-ta -TOP self-GENletter-ACC discard-PAST ‘ John2 threw away selfz's letters.’ b. Mary-mo sute-ta -also M ary also threw away John’ s letters.’ (strict reading) OR M arys also threw away selfc’ s letters.’ (“sloppy” reading) (ii) a. John-wa [vp [zibun-no temagi]-o tv] sutev-tar b. Mary-mo [vp e ] sutev-tar Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 244 Koizumi (1995) is an exception in that he draws on some independent facts to argue for overt V-raising in Japanese, those facts do not seem to be as reliable as they first appear. Koizumi (1995: chap. 7) argues for string vacuous overt V-raising in Japanese, drawing facts firom clefting and coordination. To begin with, Koizumi (1995: 166ff ) provides examples of clefting like the following. (1) a. Mary-ga John-ni ringo-o 3-tu age-ta (koto) -NOM -DAT apple-ACC -CL gLve-PAST (the fact that) M ary gave three apples to John.’ b. Mary-ga age-ta no-wa [John-ni ringo-o 3-tu] da COMP-TOP COP ‘ (Lit.) It was [three apples to John] th a t Mary gave.’ While (la) is a regular ditransitive sentence, (lb) is, Koizumi notes, one of its possible clefted variations.- The focus of clefting in (lb) is seemingly an indirect object followed by a direct object, which, in standard assumptions, is not considered to be a constituent. Keeping to the general assumption th at Roughly speaking, Otani and Whitman propose th at the VP in (iia) is copied to the elided VP position in (üb), thus deriving the “sloppy” reading of (ib). However, Hoji (1998) presents several empirical arguments against the view that cases Hke (ib) contain an elided VP, and instead argues th a t what those cases include is a null pronominal object. If Hoji is correct. W hitman and Otani’ s argum ent for V-raising will become rather groundless. - Concerning examples like (lb), judgment indeed varies among speakers; its range seems to spread firom perfection to mere marginahly. In the text Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 245 the focus of a cleft sentence is a syntactic constituent, however, Koizumi argues th at what is clefted in (lb) is in fact a VP with its head V extracted. The underlying form of d eft sentences hke (lb) is schematically displayed in the following. (2) Subject [v p Indirect Object Direct Object tjv ] y +T Koizumi m a in ta in s th at after V moves out of VP and adjoins to T, as indicated in (2), the VP dom inating both direct and indirect objects can be defted as a constituent in cases like (lb) (no m atter w hat process is responsible for deft formation) Next, Koizumi (1995: 170fO turns to “coordinated” structures such as the following. discussion, we simply foUow Koizumi’ s judgment. 3 In fact, how the derivation proceeds in deft sentences in Japanese still seems to be controversial. While Kuroda (1983) suggests (rightward) direct movement of the focused constituent to the coda position, Hoji (1990) posits null operator movement in the presuppositional p art of the sentence with the focused constituent base-generated in the coda position. However, the choice between the two alternatives is not our immediate concern in the text. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 246 (3) Mary-ga [[John-ni ringo-o 2-tu] to [Bob-ni banana-o -NOM -DAT apple-ACC -CL and ba n a n a 3-bon]] age-ta (koto) -CL give-PAST (the fact that) M ary gave two apples to John and three bananas to Bob.’ In (3), what is coordinated by the conjunctive to ‘ and’ appears to be the sequence lO+DO, which is, again, standardly considered to be a non constituent. Koizumi argues that overt V-to-T movement accounts for the possibility of coordination as in (3). The schematic representation th a t he posits for (3) is in (4). (4) Subject [v p [v p 10 DO tv ] and [vp 10 DO 'V 1] y + T T He explicates that, as indicated in (4), in fact two VPs are conjoined in (3) and the common verb is extracted out of the coordinated structure in an across-the- board fashion (cf. Williams, 1978). Thus, Koizumi concludes th at both clefting and coordination support his hypothesis that V overtly raises to T in Japanese. If Koizumi is correct in th at V overtly raises to T in Japanese, our account of su-support above will not be tenable. However, his arguments firom clefting and those firom coordination both seem to be called into question. Let us begin with clefting. Although Koizumi assumes, as is standard in the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 247 literature, that the focus of a cleft sentence is a single syntactic constituent, this assumption itself seems to be rather shaky. Consider examples like the following. (5) a. sono seiyaku-gaisya-ga kousei-daizin-ni th at pharmaceutical company-NOM Welfare M inister-to tagaku-no wairo-o okutta (koto) a lot of bribe-ACC sent (the fact that) T h at pharmaceutical company sent a lot of bribes to Welfare M inister.’ b. tagaku-no wairo-o okutta no-wa sono seiyaku-gaisya-ga COMP-TOP kousei-daizin-ni da COP ‘ (Lit.) It was that pharmaceutical company, to Welfare M inister that sent a lot of bribes.’ (5a) is a regular ditransitive sentence, and (5b) is a cleft sentence derived &om (5a), with the sequence of a subject and an indirect object put in the coda position. Although (5b) might sound a bit unnatural to some speakers, it seems to be no worse than Koizumi’ s examples like (lb). Note, more importantly, that in (5b), the focus of clefting includes the subject and the indirect object but it excludes the direct object and the verb. At first sight, it Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 248 seems very difficult to conceive of this (multiple) focus as a constituent. Koizumi m ight argue that the focus of clefting in (5b) is still a constituent. This is because clefting may apply after scrambling, as shown in (6), and because, as he suggests, V might even raise to C, as required in cases hke (7) in bis term s ((7) firom Koizumi, 1995: 167). (6) a. tagaku-no wairo-0 2 sono seiyaku-gaisya-ga a lot of bribe-ACC th at pharm aceutical company-NOM kousei-daizin-ni tz okutta (koto) Welfare M inister-to sent (the fact that) ‘ (Lit.) A lot of bribesz, th at pharm aceutical company sent tz to Welfare M inister.’ b. tagaku-no wairo-oz sono seiyaku-gaisya-ga es tz okutta a lot of bribe-ACC that pharm aceutical company-NOM sent no-wa kousei-daizin-nis da COMP-TOP Welfare M inister-to COP ‘ (Lit.) It was to Welfare M inister th a t a lot of bribesz, th at pharm aceutical company sent tz.’ (7) a. Mary-ga John-ni ringo-o 3-tu ageta (koto) -NOM -DAT apple-ACC 3-CL gave (the fact that) M ary gave three apples to John.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 249 b. ageta no-wa Mary-ga John-ni ringo-o 3-tu da gave COMP-TOP -NOM -DAT apple-ACC 3-CL COP (Lit.) It was [Mary, three apples, to John] th at gave.’ (6a) is derived firom (5a) by scrambling the direct object to the sentence-initial position presum ably, adjoining it to TP). If the indirect object is clefted firom (6a), (6b), which is perfectly gra m m atical, results. This indicates that clefting may apply to scrambled sentences. (7b) is derived firom a ditransitive sentence in (7a), with all the three argum ents defted.^ This is possible, Koizumi claims, because V may raise to C via T and the remnant TP may be clefted. If clefting can fireely apply to a scrambled sentence and if V may raise to C and the residual TP may be clefted, one wiU be able to derive examples like (5b) in the following way.® 4 For many speakers, examples like (7b), where three argum ents are clefted, are only marginal, much worse than those hke (lb) with two arguments in the focus position. There seem to be two alternative explanations for this downgrading acceptabüity. On the one hand, it might be the case that the more “complex” the focus of clefting is, the lower the acceptability may be. On the other, the lower acceptabüity of (7b) might result firom the “hghtness” of the presuppositional part of the cleft sentence. Whüe the first choice seems to be more of syntactic, the second, of semantic. However, we are not immediately concerned with the choice between the two, and we simply follow Koizumi’ s judgment for cases like (7b) for the sake of the text discussion. ® Strictly speaking, the derivation in (8) is not the only possibility. If, as Kayne (1994) suggests, multiple scrambling is adjunction of a constituent to another already scrambled, and if nominative subjects can be scrambled, a second possiblity wül arise, as in the following. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 250 (8) a. Scrambling of DO and V-to-T-to-C movement [cp [t p DO2 [t p SUBJ [v p 10 t 2 tv ] tx ]] V+T+C] b. Clefting of the rem nant TP [cp [tp DO2 e r p ] V+T+no]-wa [t p SUBJ [v p 1 0 t 2 t v ] tx] da In (8a), the direct object is adjoined to TP by scrambling and V raises to T followed by movement of V+T to C. If the lower segment of TP in (8a) can be clefted, (8b), a schematic representation of (5b), obtains. If this is how Koizumi derives cases hke (5b), almost any sequence of sentential arguments could be made a constituent by scrambling. Furthermore, his theory of clefting would predict th at any scrambled element could be left in the presuppositional p art of a cleft sentence. However, as we will see shortly, this prediction is not home out. One relevant case to which attention m ust be drawn involves control, as in the following. (8) a. Scrambling of 1 0 and movement of V to T [cp [t p IO2 [xp SUBJ [vp t2 DO t v ] V+T] C] b. Scrambling of SUB J [cp [xp [SUBJ3 10]2 [xp ta [vp t2 DO tv ] V+T] C] c. Clefting of lO with SUBJ adjoined to it [c p [xp 02 [xp ta [vp t2 DO tv ] V+T] no]-wa [SUBJa IO]a da However, Saito (1985) convincingly argues th at nominative subjects cannot undergo scrambling for various empirical reasons. (For more recent proposals as to why subjects do not scramble, see Abe (1993), Takano (1996) and references cited therein.) Following Saito, we will not pursue this possibihty any further. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 251 (9) a. John-ga Mary-niz [ecz BiU-ni sonohon-o kaes-u -NOM -to -to that book-ACC retum-PRES yooni] iw-ta (koto) COMP tell-PAST (the fact that) ‘ John told Mary to return the book to Bill.’ b. John-ga Mary-niz es iw-ta no-wa [ecz Bill-ni -NOM -to teU-PAST COMP-TOP -to sono hon-o kaes-u yoonijs da th at book-ACC retum-PRES COMP COP "What John told Mary to do was to return the book to Bill.’ c. Mary-niz John-ga tz [ecz Bül-ni sono hon-o kaes-u -to -NOM -to that book-ACC retum-PRES yooni] iw-ta (koto) COMP tell-PAST (the fact that) ‘ (Lit.) Maryz, John told tz to retum the book to Bill.’ d. ?*Mary-niz es iw-ta no-wa [John-ga tz [ecz BiU-ni -to teU-PAST COMP-TOP -NOM -to sono hon-o kaes-u yooni]]s da th at book-ACC retum-PRES COMP COP ‘ (Lit.) It was John, to retum the book to Bill that told Mary.’ (9a) is a complex sentence that contains a control complement clause, the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 252 subject empty category of which is controlled by the indirect object of the matrix clause.® (9b) is a cleft sentence derived from (9a) by putting the whole complement clause in the focus position. Notice that a lth o u ^ the controlled empty category is separated from the controller in (9b), control is still successfully exercised. (9c) is derived from (9a) by scrambling the m atrix 10 to the sentence initial position, presumably adjunction to TP. If, as indicated in (8b), the “rem n a n t TP” can be clefted, (9d) should result from (9c). This is not borne out, however. The resultant cleft sentence in (9d) is hardly acceptable. Note th a t the low acceptability of (9d) should not be attributed to failure in control. This is because a (clause-intemally) scram bled argument can still serve as a controller, as exhibited in (9c), and because separation of the controlled empty category from the controller need not end in deviation, as indicated by the fuU grammaticality of (9b); note, furthermore, that the “residual TP” clefted in (9d) should carry the trace of the controller along with it. Thus, the existence of cases like (9d) makes us skeptical about the derivation of (5b) with recourse to scrambling and clefting of the “residual ® It is generally assum ed in the literature (cf. Nemoto, 1993) th a t the subject of such a control clause is PRO, even if the clause is apparently tensed as in (9a). It m ight be the case that all “controlled” empty categories, whether in object position or in subject position, are pro rather than PRO in Japanese. However, we will not commit ourselves to the choice between the two alternatives; hence, simply indicating “controlled” empty categories as ecin the text. For relevant discussions, see Hasegawa (1984-85), Hoji (1985: ^ p e n d ix A) and N akam ura (1991). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 253 TP”, schematically represented in (8). Once the derivation in (8) is questionable, it is suspected th at there is something more than constituency involved in the formation of cleft sentences hke (5b), with SUBJ +10 dislocated together. Then, to the extent th at our counterargument to the derivation in (8) is tenable, Koizumi’ s claim that V overtly raises to T (or even to C) is weakened. Another piece of counterevidence to the proposed constituency of multiply clefted elements comes from constructions th at employ a verb su do’ . Consider the foUowing."^ (10) a. ??John-ga M ary-ni tagaku-no saiken-o zyooto-o-sae -NOM -to a lot of bonds-ACC handover-ACC-even si-ta (koto) did-PAST (the fact that) ‘ John even handed over a lot of bonds to Mary.’ b. John-ga eg es zyooto-o-sae si-ta no-wa Mary-niz -NOM handover-ACC-even do-PAST COMP-TOP -to , tagaku-no saiken-os da a lot of bonds-ACC COP ‘ (Lit.) It was a lot of bonds to Mary th at John even handed over.’ ^ The addition of sae even’ at the end of the verbal noun in (10) is simply for a styhstic purpose. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 254 ]h (10a) is a sentence th at contains a so-called verbal noun (VN) zyooto handover’ and the verb su ‘ do’ inflected as si-ta ‘ did’ .* The goal and theme arguments in (10a) are thematically related to the VN, which appears as a nom in al with a case marker. In (10b), these two arguments are clefted while the VN rem ains in the presuppositional part of the sentence. Since Grimshaw and M aster’ s (1988) sem inal work, a great many linguists have proposed alternative analyses for sentences with a VN and su (cf. Miyagawa, 1990; Hasegawa, 1991; Eageyama, 1991; Aoyagi, 1991; Poser, 1991; Hoshi and Saito, 1993; Sells, 1993, among many others). However, one common assumption shared by almost all proposals is that the VN, e.g. zyooto in (10a), instead of su, assigns 0-roles to (at least internal) arguments w ithin its maximal projection (i.e. VN“» ® * ) in cases like (10a). Depending on how the two argum ents of the VN are case-marked (see section 3.3. in chapter 3 for the theory morphological case marking adopted in this thesis), the sentence in (10a) can be represented in either of the following two ways (with many irrelevant details suppressed). * The slight anomaly of (10a) m ust be attributed to a surface constraint th at prohibits a simple clause ftom containing more than one accusative phrases (i.e. the Double-O Constraint: cf. Kuroda (1988), Sells (1990), Poser (1991), Takano (1996), and Saito and Hoshi (1997), among others). As noted by Kuroda (1988), such a superficial deviance can be ameliorated if th e two accusative phrases are separated by, for instance, clefting, as in (10b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 255 (11) a. [ t p John-ga [vp [vN m ax Mary-ni [v N tagaku-no saiken-o zyooto]]-o-sae (tv)] siv-ta] b . [ t p John-ga [ x p Mary-niz [y p tagaku-no saiken-03 [vp [vNma% tz [vN t a zyooto]]-o-sae (tv)]]] siv-ta] If the goal argument Mary-ni ‘ to Mary' and the theme argument tagaku-no saiken-o a lot of bonds-ACC’ are case-marked in situ, (1 la) counts as the surface representation of (10b). On the other hand, if these two argum ents of the VN m ust be licensed in certain functional projections, say, XP and YP, respectively, or if they each undergo scrambling to adjoin to XP and YP, (1 lb) obtains. Notice th a t even if, as Koizumi claims, su raises to T (or, alternatively, is inserted under T by su-support), there seems to be no single constituent th at includes the goal and theme arguments but sim ultaneously excludes the VN. Thus, cases like (10b) cast doubt on the assumption th a t only a single syntactic constituent per clause can be focused in clefting in Japanese. Then, again, one of the basic premises on which Koizumi’ s arguments for overt V-raising are built disappears. In the preceding paragraphs, we have presented some counterevidence against the rather common view that only a single constituent can be clefted, one of the two crucial assumptions on which Koizumi (1995) bases his argument for V-raising. In what follows, we will argue against the other assumption th at he makes, concerning those cases which, he claims, involve Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 256 “coordination” with to ‘ and’ . In th e following are some relevant examples, drawn firom Koizumi (1995: ITOffl). (The bracketings in the examples in (12) that indicate constituency are Koizumi’ s.) (12) a. Mary-ga [[John-ni ringo-o 2-tu] to [Bob-ni banana-o -NOM -DAT apple-ACC -CL and banana 3-bon]] age-ta (koto) -CL give-PAST (the fact that) (=3) M ary gave two apples to John and three bananas to Bob ’ b. [[Mary-ga John-ni ringo-o 2-tu] to [Nancy-ga Bob-ni -NOM -DAT apple-ACC -CL and banana-o 3-bon]] age-ta (koto) banana -CL give-PAST (the fact that) ‘ (Lit.) [Mary two apples to John] and [Nancy three bananas to Bob] gave. = Mary gave two apples to John and Nancy, three bananas to Bob.’ It is generally agreed that the conjunctive particle to and’ coordinates only nominal projections. In Koizumi’ s term s, however, it can conjoin two projections of an arbitrary type It coordinates two VPs after V-to-T movement in (12a), and two TPs after V-to-T-to-C raising in (12b), as schematically represented in (13a) and (13b), respectively. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 257 (13) a. SUBJ [vp[vpIODOty] and [vpIODOty]] V+T bj] and [vpIODOt|r]] b. [tp[tpSUBJ [lODOtv] tr] and [ tp SUB J [10 DO ty] tx]] Y ^ + C Obviously, his analysis of cases like (12) is crucially contingent upon the assumption th a t only constituents can be “coordinated” by to. However, for some empirical reasons, such an assum ption does not seem to be firm. To begin with, consider the following example. (14) John-ni 2-hon to Bob-ni 3-bon Mary-ga banana-o -to -CL and -to -CL -NOM banana-ACC age-ta (koto) give-PAST (the fact that) ‘ (Lit.) [two to Johnl and [three to Bobl Mary gave bananas. = Mary gave two bananas to John and three bananas to Bob.’ What is seemingly conjoined by to in (14) is 10 + Numeral Floating Quantifier (NFQ). Apparently, this sequence is not a constituent, because each NFQ modifies DO banana-o ‘ bananas-ACC’ , b u t not 10, John nor Bob. This casts doubt on the assumption that to is a constituent coordinator. However, propounders of such an assumption m ight claim th a t to indeed conjoins two constituents in (14). The derivation th a t they m ight claim to be possible will proceed in the following way. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 258 (15) a. “B ase” structure Mary-ga [[John-ni banana-o 2-hon age] to [Bob-ni -NOM -to banana-ACC -CL give and -to banana - 0 3-bon age]] ta (koto) banana-ACC -CL give PAST (the fact that) b. Across-the-board scrambling of DO and across-the-board V-raising Mary-ga banana -0 2 [[John-ni tg 2-hon tv] to [Bob-ni -NOM banana-ACC -to -CL and -to t 2 3-bon tv]] age-ta (koto) -CL give-PAST (the fact that) c. Scrambling of the “coordinated” VPs [[John-ni t 2 2-hon tv] to [Bob-ni t 2 3-bon tv]]3 Mary-ga -to -CL and -to -CL -NOM banana -0 2 ta age-ta (koto) banana-ACC give-PAST (the fact that) (15b) is derived firom the “base” structure in (15a) by scrambling DO and raising V to T, both in an across-the-board manner. If the residual conjoined VPs are scrambled together to the firont of the subject, (15c) (=(14)) will be successfiiUy derived. Thus, if each step of the derivation in (15) were legitimate, (14) would be obtained without discarding the assumption that to is a crosscategorial constituent coordinator. However, the final step th at Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 259 maps (15c) firom (15b) is rather questionable. Note that in (15c), the scrambled DO does not bind its trace in either of the conjuncts. This is a case that violates th e Proper Binding Condition (PEG). As noted by Saito (1989), a typical violation of the PEG is found in cases of multiple scrambling like the following. (16) a. Mary-ga [John-ga sono ringo-o tabetesim atta to] itta -NOM -NOM that apple-AGG has eaten GOMP said (koto) (the fact that) M ary said th at John had eaten th a t apple.’ b. sonoringo-0 2 Mary-ga [John-ga tz tabetesim atta to] itta that apple-AGG -NOM -NOM has eaten GOMP said (koto) (the fact that) ‘ (Lit.) T hat apple, Mary said th at John had eaten.’ c. [John-ga sono ringo-o tabetesim atta tola Mary-ga ta itta -NOM th at apple-AGG has eaten GOMP -NOM said (koto) (the fact that) (Lit.) T hat John had eaten th at apple, Mary said.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 260 c L *[John-ga tz tabetesim atta to]a sonoiingo -0 2 Mary-ga ta -NOM has eaten COMP th a t apple-ACC -NOM itta Qcoto) said (the fact that) ‘ (Lit.) T hat John had eaten, that apple, M ary said.’ While (16b) is derived from (16a) by scrambling the complement DO to the front of the m atrix sentence, (16c) is obtained by scrambling the whole complement CP. On the contrary, as exhibited by the ungrammaticality of (16d), multiple application of scrambling in the noted order will yield an illegitimate representation, where the trace of DO (=ts) is not bound by its antecedent. If both (15c) and (16d) are results of m ultiple application of scrambling, then the PBC or whatever prohibits the latter should also prohibit the former.^ However, this is not home out. This strongly suggests ® Strictly speaking, (15c) and (16d) are not quite parallel, since while in the former, scrambling of the DO, which applies across the board, places it to the right of the subject, th e complement DO in the latter is scrambled to the left of the matrix subject. If the DO and then the containing conjoined VPs in (15a) were both scram bled to the left of the subject, in the same way th at (17d) is derived, the following representation would result: (i) [[John-ni t 2 2-hon tv] to [Bob-ni t 2 3-bon tvlja banana -02 -to -CL and -to -CL banana-ACC Mary-ga ta agev-ta (koto) -NOM give-PAST (the fact that) This representation would violate whatever condition th at excludes (17d). However, if NFQs can be base-generated to the left of their hosts, or if they can somehow be scram bled around their hosts, then the string in (i) will allow an alternative representation such as the following. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 261 that (15c) is not the correct representation of (14). Since no alternative derivation with coordination seems to yield a legitimate representation for (14), the identification of to as a crosscategorial coordinator becomes dubious. Another argument against the contention that to can be a large-scale constituent coordinator is drawn fi-om scope facts. Consider examples like the following. (17) a. Mary-dake-ga keeki-o zenbu tabe-oe-ta (koto) -only-NOM cake-ACC whole eat-finish-PAST (the fact that) ‘ (Lit.) Only Mary finished eating the whole cake.’ b. Mary-ga keeki-dake-o zenbu tabe-oe-ta (koto) -NOM cake-only-ACC whole eat-finish-PAST (the fact that) ‘ (Lit.) Mary finished eating only the whole cake.’ Among other things, attention m ust be drawn to scope interaction between the DP with an Q-particle dake only’ and the predicate oe finish’ in each example in (17).i° While the DP with dake is in subject position in (17a), it is in object (ii) [[John-ni 2-hon proz tv] to [Bob-ni 3-bon banana -0 2 tvjja -to -CL and -to -CL banana-ACC Mary-ga ta agev-ta (koto) -NOM give-PAST (the fact that) The representation in (ii) does not contain any offensive trace. Note, incidentally, th at (15c) does not allow an alternative representation sim ilar to (Ü), due to the intervention of the subject between the NFQ 3-bon ‘ 3-CL’ and the host banana-o ‘ banana-ACC’ . It is for this reason that we have considered (14) instead of the string given in (i) in the text. At first sight, the NFQ zenbu, hterally, all’ or whole’ , attached to the object Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 262 position in (17b). If dake and oe freely interact with respect to scope, then (at least) two distinct interpretations will be potentially possible for each example, as indicated in the following. (18) (for (17a)) i. [only > finish] reading (Among all members present) it was only M ary that finished up the whole cake. Ü. [finish>only] reading What came to an end was the event th a t (among aU members present) only Mary ate the whole cake. (19) (for (17b)) i. [only>finish] reading (Among all things present) it was only the cake that Mary finished up. Ü. [finish>only] reading What came to an end was the event th a t M ary ate up only the cake (among all things present). in each of the examples in (17), only seems to complicate the situation, but some NFQ hke it is necessary, as noted by Koizumi (1995: 204-207), for to to attach to in later examples. Since its intended holistic meaning ‘ wholly’ or ‘ as a whole’ seems more compatible with, hence less affective on, the semantics of oe ‘ finish’ than cardinal NFQs like 3-tu three pieces’ , zunbu is used in the text. However, even if it interacts with either of the scope bearers in (17) or the examples to follow, it does not affect our argument which is based on the facts of relative scope between dake and oe only. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 263 K dake takes wide scope with respect to oe, (17a) and (17b) will be interpreted as (18i) and (19i), respectively. If, on the other hand, oe takes wide scope, each example in (17) wül have the second interpretation given in (18) and (19). Despite the potential ambiguity, it is very obvious to native speakers th at each example in (17) only allows the first reading in (18) and (19), with dake taking wide scope. “ In this connection, let us make an experiment by putting the sentences in (17) in “to-coordinated” structures, as in the following. (20) a. Mary-dake-ga keeki-o zenbu to John-ga piza-o -only-NOM cake-ACC whole and -NOM pizza-ACC yooyaku hanbun tabe-oe-ta (koto) as little as half eat-finish-PAST (the fact that) O nly Mary finished (eating) the whole cake and John finished (eating) as Little as half of the pizza.’ iiPor a possible explanation for the impossibüity of narrow scope reading on dake only’ even in object position, see Koizumi (1995: chap. 4). However, we are not concerned why this is the case. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 264 b. Mary -ga keeld-dake -o zenbu to John-ga piza-o -NOM cake-only -ACC whole and -NOM pizza-AC C yooyaku hanbun tabe-oe-ta O^oto) as little as half eat-finish-PAST (the fact that) M ary finished (eating) only the whole cake and John finished (eating) as little as h alf of the pizza.’ In (20a) and (20b), (17a) and (17b), respectively, appear as the first “conjxmct” of a “to-coordinated” sentences. Crucially, in each case, wide scope reading of dake is preserved. While the semantics of (20a) includes (18i), th at of (20b) includes (19i). Note th at this is very hard to expect in Koizumi’ s analysis of “to-coordination”. In his terms, (20a) and (20b) will be schematically represented as (21a) and (21b), respectively. (21) a. [t p [tp SUBJ-dake [DO t v ] tT] and [t p SU B J [DO t v ] tij] V+oe+T+C b. [t p [tp SUBJ rnO-dflke t v ] tT] and [t p SU B J [DO t v ] tij] V-t-oe+T-i-C In Koizumi’ s analysis, when “conjuncts” contain a subject, two TPs are “coordinated” and V m ust raise to C via T. U nder standard assumption of scope and c-command, if oe raises to C together with the main verb and T, as indicated in (21), it will c-command, therefore, have scope over, everything th a t C c-commands, i.e. the whole content of the “coordinated” TPs. One way for a scope-bearing argum ent to escape firom the scope domain of a c-commanding element is to undergo QR to a higher position. However, this is impossible Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 265 for either of the arguments with dake in (21), since they are both contained in a conjunct of coordination. If they move higher than C, necessarily out of the conjuncts th a t contain them, the coordinate structure constraint (CSC) wiU be violated. In fact, as indicated in examples hke the following, a quantified element contained in a coordinated structure cannot take scope outside it. (22) a. every boyg hates hisg mother b. *[every boyz and Mary] hate hisz mother (23) a. subeteno otokonokog-ga soitug-no hahaoya-o Idratteiru all boy-NOM his-GEN mother-ACC hate (koto) (the fact that) ‘ Every boyg hates hisg mother.’ b. *[subeteno otokonokog to Mary]-ga soitug-no hahaoya-o all boy and -NOM his-GEN mother-ACC Idratteiru (koto) hate (the fact that) ‘ [Every boyg and Mary] hate hisg mother.’ It is generally assumed th a t if a pronoun is to be bound by a quantifier, it must be in the scope domain of th at quantifier. In each a-example in (22) and (23), the pronoun contained in the object is c-commanded by, hence, in the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 266 scope domain of, the quantifier in subject position, and bound variable construal of the pronoun is felicitous. On the contrary, in each b-example, the quantifier intended to bind a pronoun is embedded in a coordinated structure, and bound variable anaphora does not hold. This can be accounted for as follows. On the one hand, as long as it is embedded in a coordinated structure, the quantifier in each b-example cannot c-command th e pronoun; hence, bound anaphora fails. If, on the other hand, the quantifier undergoes QR for the purpose of c-commanding the pronoun at LF, it will violate the CSC; hence, bound anaphora, again, fails. If scope of a quantified expression is thus constrained by conditions Hke the CSC, the facts of wide scope of dake in (20) suggest that the representations in (21) are not correct. Then, Koizumi's analysis of to as a crosscategorial coordinator is further questioned. In chapter 2, we have presented a morphological account of su-support that is compatible with our analysis of Q-particles; Q-particles are adjunct clitics that do not project. Our account of su-support is contingent upon the assumption that V stays in situ at Spell-Out in Japanese. Although Koizumi (1995) argues to the contrary, we have found that neither of the arguments that he draws firom clefting and coordination is as empirically sound as they first look. Especially, we cannot help but question his basic assumptions th at only a single constituent per clause can be clefted and to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 267 ‘ and’ is a crosscategorial coordinator. Thus, to the extent th at the existence of overt V-raising is dubious, our analysis of su-support remains tenable. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 268 A PPENDIX m O n th e Irre g u la rity of A d jectiv al In flectio n Tense inflection with adjectives in Japanese appears to be irregular, and it seems to defy any systematic analysis. First of all, the present and past forms do not constitute what is generally called a minimal pair. (1) a. taka -i high -PRES b. taka -ku -ar -ta (phonetically realized as takakatta) high -INFIX?) -be -PAST The present tense form is simply created by suffixation of -i to the adjective stem, as in (la). On the contrary, the past tense form is more complex; the stem is suffixed with a series of morphemes, as indicated in (lb). Above all, the status of -ku is yet to be made clear. We tentatively consider it to be an inflectional suffix particular to adjectives. One piece of evidence th at -ku is inflectional is th a t it is employed in inflection, but not in derivation. (2) a. taka *(-ku) si-ta hig^ make-PAST hiade (something) h i ^ ’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 269 b. taka *(-ku) nar-ta h i ^ become-PAST ‘ became h i ^ ’ c. taka *(-ku) tob-ta h i ^ jump-PAST jumped (3) a. taka (*-ku) -sa high nominalizing suffix ‘ b e i ^ f b. tak a (*-ku) -mi high n om in a liz in g suffix “ height, high place' c. taka (*-ku) -nari h i ^ -soimding ‘ pounding’ When embedded under the causative verb su make' or the inchoative verb nar become' or used with an action verb like tob jump', the adjective stem m ust be suffixed with -ku, as exhibited in (2a-c). Conversely, in derivational processes like nominalization as in (3a, b) or synthetic compounding as in (3c), -ku can not be involved. This indicates that -ku is an inflectional, rather than derivational, suffix. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 270 The disparity in tense inflection with adjectives, illustrated in (1), has a historical origin. It is a well-known fact th at in Old Japanese, spoken until the 15th century, verbal and adjectival conjugation for the conclusive form (or shuusbi-keî) was different from that for the attributive form (or rentai-kei), a distinction lost in the subsequent period (i.e. Muromachi Period, in the late 15th century). (4) a. toki(-wa) sugu (conclusive) time(-TOP) pass Time passes.’ b. sugu-ru toki (attributive) pass ‘ time th a t passes’ (5) a. yama(-wa) taka-si (conclusive) mountain(-TOP) high The mountain is h i^ .’ b. ta ka -ki yama (attributive) h i ^ mountain ‘ a m ountain th at is h i^ , or a h i ^ m ountain’ In Old Japanese, verbs in the conclusive form as in (4a) were untensed, indicating non-past, and the past tense was indicated by an auxiliary verb hke -tari, whose modem descendant is the past tense suffix -ta. A s is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 271 obvious, the present tense suffîx for verbs -ru in Modem Japanese originates &om the attributive ending in Old Japanese as in (4b). The present tense adjectival ending -i in Modem Japanese has sprung from the attributive ending -in in Old Japanese, with the loss of the velar sound (i.e. k —> 0 ). Since the Old Japanese past tense auxiliary -tari selected for a verb stem, it did not directly attach to an adjective stem. Hence, a copular verb a r he', together with -ku, was employed when past tense had to be expressed with an adjective, as in taka-ku-ari-tari hi^-IN FL-be-PA ST’, the direct origin of the past tense form in Modem Japanese in (lb). If the gap between the present and past tense forms of adjectives is to be bridged, one m ay consider th at morphophonological suppletion takes place to derive the simple ending -2 from the complex iorm -ku-ar-(r)u TNFL-be-PRES’, the latter being expected on the analogy of the p ast tense ending. In fact, this view seems to be supported by examples like the following. (6) a. sono yama-wa taka-i th at mountain-TOP hi^-P R E S T h at m ountain is hig^.’ b. sono yama-wa taka -ku -sae ar-u (cf. *taka(-ku)-sae-i) th at mountain-TOP high-INFL-even be-PRES T h at m ountain is even h i ^ .’ As noted, in the present tense, the adjective stem is simply suffixed with 2, as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 272 shown in (6a). However, when a Q-particle is involved in adjectival inflection as in (6b), ku TNFL’ , a r h e ’ and (r)u ‘ PRES’ m ust be employed. This suggests that the underlying form for adjectival inflection in the present tense is ku-ar- (r)u, as indicated above, and i substitutes this complex form only if the three inflectional elem ents are under string adjacency. Furthermore, there are cases where the underlying complex form interchanges with the suppletive form, or the former is even more appropriate than the latter. (7) a. taka -ku -ar -ru hazu-no nedan-ga... high-INFL-be-PRES expectancy-GEN price-NOM the price th a t should be h i ^ ...’ b. taka -i hazu-no nedan-ga... h i^-P R E S expectancy-GEN price-NOM (8) a. kokorozasi-wa taka -ku -ar -ru held da ambition-TOP high-INFL-be-PRES necessary COP (lit.) (Your) ambition must be h i ^ . = You must aim h i ^ . ’ b. ?*kokorozasi-wa taka -i beki da ambition-TOP high-PRES necessary COP Hazu expectancy’ in (7) is an abstract noun and beki necessary’ in (8) is one of the few Old Japanese adjectives that have survived up to date; notice that the latter takes the attributive form ending with -M, and its conclusive form is be- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 273 si (see the historical discussion above). Both take a complement clause. H azu allows either the complex or the simple ending, but beki allows only the complex form. Although the reason that the simple form is disallowed in (8b) is unknown, examples like (7a) and (8a) suggest th a t the suppletion process noted above is in fact necessary. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 274 R E FE R E N C E S Abe, J. 1993. Binding Conditions and Scrambling without A/A’ Distinctions. PhD Thesis. U niverâty of Connecticut. Abney, S. 1987. The English Noun Phrase and Its Sentential Aspect. PhD Thesis. MIT. Aoun, J. 1985. A Grammar o f Anaphora. Cambridge, MA: MET Press. Aoun, J., N. Hom stein, D. lightfoot, and A. Weinberg. 1987. Two types of locality. Linguistic Inquiry, 18, 537-577. Aoun, J. and A Y.-H. Li. 1989. Constituency and scope. L inguistic Inquiry, 20, 142-172. Aoun, J. and A . Y.-H. Li. 1993a. Syntax of Scope. Cambridge, M A MIT Press. Aoun, J. and A. Y.-H. Li. 1993b. Wh-elements in situ: syntax or LF. Linguistic Inquiry, 24, 199-238. Aoyagi, H. 1991. Light verb constructions and restructuring at LF. Ms. University of Southern California and Nanzan University. Aoyagi, H. 1994. On association with focus and scope of focus particles in Japanese. Formal Approaches to Japanese Linguistics, 1 (M IT W orking Papers in Linguistics, 24), 23-44. MET. Aoyagi, H. 1996. On Kuroda’ s scope mystery. Academia, 61, 97-133. Nanzan University. Aoyagi, H. and T. Ishii. 1994. Agreement-inducing vs. non-agreement- inducing NPIs. NELS, 24, 1-15. Baker, M. 1988. Incorporation. Chicago: University of Chicago Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 275 Barwise, J. and R. Cooper. 1981. Generalized quantifiers and natural language. U nguistics and Philosophy, 4, 159-219. BeUetti, A. The Case of unaccusatives. L inguistic Inquiry, 19, 1-34. Belletti, A. and L. Rizzi. 1988. Psych verbs and theta theory. N atural Language and Linguistic Theory, 6, 291-352. Borer, H. 1984. P aram etric Syntax. Dordrecht: Foris. Boskovic, Z. 1997. Superiority and economy of derivation: multiple wh- firontmg. Paper presented at the Workshop on Theoretical Linguistics, Nanzan University, January 17, 1997. Burzio, L. 1986. Italian Syntax. Dordrecht: Kluwer. Cardinaletti, A. and M. Starke. 1994. The typology of structural deficiency: on the three gram m atical classes. Ms. University of Venice and University of Geneva. Carlson, G. 1977. Reference to Kinds in E n g lish , PhD thesis. University of M assachusetts, Amherst. Cheng, L. L.-S. 1991. On the Typology of Wh-Questions. PhD Thesis. MIT. Cho, Y.-M. and P. Sells. 1995. A lexical account of inflectional sufhxes in Korean. Journal o f E ast Asian Linguistics, 4, 119-174. Chomsky, N. 1965. A spects o f the Theory o f Syntax. Cambridge, MA: MIT Press. Chomsky, N. 1970. Remarks on nominalizations. In Readings in English Transform ational Grammar, ed. R.A. Jacobs and P S. Rosenbaum. Waltham: Ginn. (Reprinted in Chomsky, 1972) Chomsky, N. 1972. S tu d ies on Sem antics in G enerative Grammar. The Hague: Mouton. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 276 Chomsky, N. 1986a. Barriers. Cambridge, MA: MET Press. Chomsky, N. 1986b. Knowledge o f Language. New York: Praeger. Chomsky, N. 1991. Some notes on economy of derivation and representation. In Principles and Parameters in Com parative Grammar, ed. R. Freidin. Cambridge, MA: MET Press. (Reprinted as chapter 2 of Chomsky, 1995) Chomsky, N. 1993. A m inim alist program for linguistics theory. In The View Grom Building 20, ed. K. Hale and S.J. Keyser. Cambridge, MA: MIT Press. (Reprinted as chapter 3 of Chomsky, 1995) Chomsky, N. 1994. Bare phrase structure. MZT Occasional Papers in Linguistics, 5. MIT. Chomsky, N. 1995. The M inim a list Program. Cambridge, MA: MIT Press. Chomsky, N. and H. Lasnik. 1977. Filters and control. L inguistic Inquiry, 8, 425-504. Chomsky, N. and H. Lasnik. 1993. Principles and param eters theory. In Syntax: An International H andbook o f Contemporary Research, ed. J. Jacobs, A. von Stechow, W. Stem feld and T. Vennemann. Berlin and New York: Walter de Gruyter. (Reprinted as chapter 1 of Chomsky, 1995) Cinque, G. 1990. Tirpes of A'-Dependencies. Cambridge, MA: MIT Press. Cinque, G. 1993. A null theory of phrase and compound stress. Linguistic Inquiry, 24, 239-297. Collins, Chris. 1994. Economy of derivation and the generalized proper binding condition. L inguistic Inquiry, 25, 45-61. Comrie, B. 1989. Language U niversels and Linguistic Typology (2nd Edition). Chicago: University of Chicago Press. Crystal, D. 1991. A D ictionary o f Linguistics and Phonetics (3rd edition). Oxford: Basil Blackwell. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 277 Diesing, M. 1988. Bare plural subjects, inflection and mapping to LF. In Papers on Quantification, ed. E. Bach, A. K ratzer and B. Partee. University of M assachusetts, Amherst. Diesing, M. 1992. IndeSnites. Cambridge, MA: MIT Press. Emonds, J. 1978. The verbal complex V’ -V in French. Linguistic Inquiry, 9, 151-175. Fukui, N. 1986. A Theory of Category Projection and Its implications. PhD Thesis. MIT. Fukui, N. 1988. Deriving differences between Enghsh and Japanese. English Linguistics, 5, 249-270. Fukui, N. 1995. The principles and param eters approach: a comparative syntax of Enghsh and Japanese. In Approaches to Language Topology, ed. M. Shibatani and T. Bynon. Oxford: Clarendon Press. Fukui, N. and M. Saito. 1996. Order in phrase structure and movement. Ms. University of California, Irvine and N anzan University. Fukui, N. and M. Speas. 1986. Specifiers and projections. M IT W orking Papers in Linguistics, 8, 128-172. Fukui, N. and Y. Takano. 1998. Symmetry in syntax: Merge and Demerge. Journal o f E ast A sian Linguistics, 7, 27-86. Grimshaw, J. 1979. Complement selection an d the lexicon. Linguistic Inquiry, 10, 279-326. Grimshaw, J. 1981. Form, function, and the language acquisition device. In The Logical Problem o f Language A cquisition, ed. C.L. Baker and J.J. McCarthy. Cambridge, MA MIT Press. Grimshaw, J. 1990. A rgum ent Structure, Cambridge, MA MIT Press. Grimshaw, J. 1991. Extended projection. Ms. Brandeis University. Grimshaw, J. and R.A Mester. 1988. Light verbs and theta marking. Linguistic Inquiry, 19, 205-232. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 278 Hale, K and S.J. Keyser. 1993. On argum ent structure and the lexical expression of syntactic relations. In The View Grom B uilding 20, ed. K Hale and S.J. Keyser. Cambridge, MA: MET Press. Halle, M. and A. M arantz. 1993. D istributed morphology and the pieces of inflection. In The View Grom B uilding 20, ed. K. Hale and S.J. Keyser. Cambridge, MA: MET Press. Harada, S.-I. 1973. Counter equi NP deletion. A nnual B ulletin, 7, 113- 147. The Research Institute of Logopedics and Phoniatrics, University of Tokyo. Harada, S.-I. 1977. Nihongo-ni “henkei”-wa hitsuyoo da. Gengo, 6: 10, 88- 95, 6: 11, 96-103. Hasegawa, N. 1984-85. On the so-called “zero-pronouns” in Japanese. The L inguistic Review, 4, 289-341. Hasegawa, N. 1991. On head movement in Japanese: the case of verbal nouns. The Proceedings o f Sophia L inguistic Society, 6, 8-32. Hashimoto, S. 1969. Joshi-Jodooshi no Kenkyuu. Tokyo: Iwanami. Heycock, C. 1993. Syntactic predication in Japanese. Journal o f E ast A sian Linguistics, 2, 167-211. H i^inbotham , J. 1985. On semantics. Linguistic Inquiry, 16, 547-593. Hoji, H. 1985. Logical Form Constraints and Configurational Structures in Japanese. PhD thesis. University of Washington. Hoji, H. 1990. Theories of anaphora and aspects of Japanese syntax. Ms. University of Southern California. Hoji, H. 1997. A review of Japanese S yn ta x and Sem antics by S.-Y. Kuroda. Ms. University of Southern California. (To appear in Language.) Hoji, H. 1998. Null object and sloppy identity in Japanese. Linguistic Inquiry, 29, 127-152. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 279 Huang, J. C.-T. 1982. Logical Relations in Chinese and the Theory of Grammar. PhD Thesis. MIT. Huang, J. C.-T. 1988. Comments on Hasegawa’ s Paper. Jn Proceedings o f Japanese S yn ta x Workshop, ed. W. Tawa and M. Nakayama. Connecticut College, New London. Ikawa, H. 1995. On kakaribusubi in Old Japanese: a possibility under a generative gram m atical perspective. Ms. University of California, hvine. Ikawa, H. 1996. Overt Movement as a Reflex of Morphology. PhD Thesis. University of California, Irvine. Inoue, K. 1978. Nihongo no Bunpoo Kisoku. Tokyo: Taishukan. Inoue, K. 1985. Case marking and property reading. In The Final Report to the M in istry o f Education, ed. K Inoue. International Christian University. Ishigaki, K. 1955. Joshi no R ekishiteki Kenkyuu. Tokyo: Iwanami. Jackendoff, R. 1972. Sem antic Interpretation in G enerative Grammar. Cambridge, MA: MET Press. Kageyama, T. 1991. Light verb constructions and the syntax-morphology interface. In Current Enghsh Linguistics in Japan, ed. H. Nakajima. Berlin and New York: Mouton de Gruyter. Karttunen, L. and S. Peters. 1979. Conventional imphcature. In Presupposition (Syntax and Semantics, 11), ed. C.-K. Oh and DA. Dinneen. New York: Academic Press. Kayne, R. 1991. Romance cHtics, verb movement and PRO. Linguistic Inquiry, 22, 647-686. Kayne, R. 1994. The A ntisym m etry o f Syntax. Cambridge, MA MET Press. Kindaichi, H. 1960. M eikai Akusento Jiten. Tokyo: Sanseidoo Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 280 Kitagawa, Y. 1986. Subjects in Japanese and English. PhD Thesis. University of M assachusetts, Amherst. Kitagawa, Y. and S.-Y. Kuroda. 1992. Passive in Japanese. Ms. University of Rochester and University of California, San Diego. Klavans, J. 1985. The independence of syntax and phonology in cliticization. Language, 61, 95-120. Klima, E.S. 1964. Relatedness between grammatical systems. Language, 40, 1-20. Koizumi, M. 1995. Phrase Structure in M inimalist Syntax. PhD Thesis. MIT. Konig, E. 1991. The M eaning o f Focus Particles. London: Routledge. Kratzer, A. 1989. Stage-level and individual-level predicates. In Papers on Q uantification, ed. E. Bach, A. Kratzer and B. Partee. University of M assachusetts, Amherst. (Reprinted in The Generic Book, ed. G. Carlson and F.J. Pelletier, Chicago: University of Chicago Press, 1995.) Kuno, S. 1973. The Structure o f the Japanese Language. Cambridge, MA: MIT Press. Kuroda, S.-Y. 1965. A Generative-Grammatical Studies of the Japanese Language. PhD Thesis. MIT. Kuroda, S.-Y. 1969. Remarks on the notion of subject with reference to words hke also, even and only. P art I. A nnual B ulletin, 3, 111-130. The Research Institute of Logopedics and Phoniatrics, University of Tokyo (Reprinted as part 1, chapter 2 of Kuroda, 1992) Kuroda, S.-Y. 1970. Remarks on the notion of subject with reference to words hke also, even and only. P art H. A nnual B ulletin, 4, 127-152. The Research Institute of Logopedics and Phoniatrics, University of Tokyo (Reprinted as part 2, chapter 2 of Kuroda, 1992) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 281 Kuroda, S.-Y. 1978. Case-marMng, canonical sentence patterns, and Counter Equi in Japanese. In Problems in Japanese S yn ta x and Sem antics, ed. J. Hinds and I. Howard. Tokyo: K aitakusha. (Reprinted as chapter 6 of Kuroda, 1992) Kuroda, S.-Y. 1979. On Japanese passives. In Explorations in Linguistics: Papers in Honor o f K azuko Inoue, ed. G. Bedell, E. Kobayashi and M. Muraki, Tokyo: Kenkyusha. (Reprinted as chapter 5 of Kuroda, 1992) Kuroda, S.-Y. 1983. W hat can Japanese say about government and binding. WCCFL, 2, 153-164. (Reprinted as chapter 7 of Kuroda, 1992) Kuroda, S.-Y. 1986. Movement of noun phrases in Japanese. In Issues in Japanese Linguistics, ed. T. Imai and M. Saito. Dordrecht: Foris. (Reprinted as chapter 8 of Kuroda, 1992) Kuroda, S.-Y. 1988. W hether we agree or not. In Papers from the Second International W orkshop on Japanese Syntax, ed. W. Poser. CSLI, Stanford. (Reprinted as chapter 10 of Kuroda, 1992) Kuroda, S.-Y. 1992. Japanese Syntax and Sem antics. Dordrecht: Kluwer. Laka, I. M. 1990. Negation in Syntax: on the N ature of Functional Categories and Projections. PhD thesis. MIT. Larson, R. and G. Siegel. 1995. Knowledge o f M eaning. Cambridge, MA: MIT Press. Lasnik, H. 1995. Verbal morphology: Syntactic Structures m eets the minimalist program. In Evolution and Revolution in L inguistic Theory, ed. H. Campos and P. Kempchinsky. Washington, D C.: Georgetown University Press. Lasnik, H. and M. Saito. 1992. Move a. Cambridge, MA: MIT Press. Lefebvre, C. and P. Muysken. 1988. M ixed Categories. Dordrecht: Kluwer. Levin, B. 1983. On the N ature of Ergativity. PhD Thesis. MIT. Li, A. Y.-H. 1990. Order and Constituency in M andarin Chinese. Dordrecht: Kluwer. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 282 li, A. Y.-H. 1992. Indefinite wh in M andarin Chinese. Journal o f East A sian U nguistics, 1, 125-155. Lieber, R. 1983. Argument linking and compounds in English. U nguistic Inquiry, 14, 251-286. Lieber, R. 1992. Deconstructing Morphology. Chicago: University of Chicago Press. Mahajan, A. 1990. A/A-bar distinction and movement theory. PhD Thesis. MIT. Malin g, J. and A. Zaenen, ed. M odem Icelandic Syntax (S yn ta x and Sem antics, 24). New York: Academic Press. Marantz, A. 1984a. On the N a tu re o f G ram matical R elations. Cambridge, MA: MET Press. Marantz, A. 1984b. Predicting ergative agreement with transitive auxiliaries. ESCOL, ‘ 84, 58-68. Marantz, A. 1988. Clitics, morphological merger, and the m apping to phonological structure. In Theoretical Morphology, ed. M. Hammond and M. Noonan. New York: Academic Press. Marantz, A. 1989. Clitics and phrase structure. Jn A ltern a tive Conceptions o f Phrase Structure, ed. M. Baltin and A. Kroch. Chicago: University of Chicago Press. Marantz, A. 1991. Case and licensing. ESCOL, *91, 234-253. M atsushita, D. 1930. Hyoojun Nihonkoogohoo. Tokyo: Benseisha. May, R. 1985. Logical Form. Cambridge, MA: MET Press. McCawley, J. 1968. The Phonological Component o f a G ram m ar o f Japanese. The Hague: Mouton. McGloin, N. H. 1980. Ga-ni conversion re-examined. P apers in Japanese Linguistics, 7, 65-77. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 283 Miyachi, A. 1997. Hiteihyoogen ni kakawaru ko’ oogenshoo. MA Thesis. Nagoya University. Miyagawa, K. 1997. The Japanese dummy verbs and the organization of grammar. To appear in Japanese/Korean Linguistics, 7. Miyagawa, S. 1986. Restructuring in Japanese. In Issues in Japanese Linguistics, ed. T. Imai and M. Saito. Dordrecht: Foris. Miyagawa, S. 1987. Categories in Japanese. Lingua, 73, 29-51. Miyagawa, S. 1989a. Light verbs and the ergative hypothesis. Linguistic Inquiry, 20, 659-668. Miyagawa, S. 1989b. Structure and Case M arking in Japanese (Syntax and Sem antics, 22). San Diego: Academic Press. Motoori, N. 1796. Genji Monogatari Tama no Ogushi. (Reprinted as volume 7 of M otoori Norinaga Zenshuu, ed. K. Motoori, Tokyo: Yoshikawa Koobunkan, 1927) Muraki, M. 1978. The sika nai construction and predicate restructuring. In Problems in Japanese S yn ta x and Sem antics, ed. J. Hinds and I. Howard. Tokyo: Kaitakusha. Nakamura, M. 1991. Japanese as a pro language. The Linguistic Review, 6, 281-296. Nemoto, N. 1993. Chains and Case Positions: A Study from Scrambling in Japanese. PhD Thesis. University of Connecticut. Ohno,S. 1993. K akari-m usubi no Kenkyuu. Tokyo: Iwanami. Okutu K., Y. N um ata and T. Sugimoto. 1986. Iwayuru Nihongo Joshi-no Kenkyuu. Tokyo: Bonjinsha. Otani, K. and J. Whitman. 1991. V-raising and VP-ellipsis. Linguistic Inquiry, 22, 345-358. OuhaHa, J. 1990. Sentential negation, relativized minimality and the aspectual status of auxiliaries. The Linguistic Review, 7, 183-231. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 284 Pesetsky, D. 1982. Paths and Categories. PhD Thesis. MET. Pollock, J.-Y. 1989. Verb movement, universal grammar and the structure of IP. U nguistic Inquiry, 20, 365-424. Poser, W. 1991. Lexical and phrasal periphrastic verbs in Japanese. Ms. Stanford University. Rizzi, L. 1982. Issues in Italian Syntax. Dordrecht: Foris. Rizzi, L. 1990. Relativized M inim ality. Cambridge, MA: MET Press. Roberts, I. 1993. Verbs suid Diachronic Syntax. Dordrecht: Kluwer. Rooth, M. 1985. Association with Focus. PhD thesis. University of M assachusetts, Amherst. Rooth, M. 1992. A theory of focus interpretation. N atural Language Sem antics, 1, 75-116. Rooth, M. 1996. Focus. In The Handbook o f Contemporary Sem antic Theory, ed. S. Lappin. Oxford: Blackwell. Rothstein, S. 1991. Heads, projections and category determination. In View s on Phrase Structure, ed. K. Leffel and D. Bouchard, Dordrecht: Kluwer. Sadock, J. 1991. Autolexical Syntax. Chicago: University of Chicago Press. Saito, M. 1982. Case marking in Japanese. Ms. MET. Saito, M. 1983. Case and government in Japanese. WCCFL, 2,247-259. Saito, M. 1985. Some Asymmetries in Japanese and Their Theoretical Implications. PhD thesis. MET. Saito, M. 1989. Scrambling as semantically vacuous A'-movement. In A lternative Conceptions o f Phrase Structure, ed. M. Baltin and A. Kroch. Chicago: University of Chicago Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 285 Saito, M. 1992. Long-distance scrambling in Japanese. Journal o f E ast A sian Linguistics, I, 69-118. Saito, M. and H. Hoshi. 1997. Japanese light verb constructions and the m inim alist program. Ms. N anzan University and SOAS, University of London. Sano, M. 1997. Fuku-joshi dake no LF idoo ni okeru target ni tsuite. Ms. Ritsum eikan University. Sano, M. 1998. On the direction of movement: a case study from LF- movement of dake ‘ only*. JE LS, 15. Selkirk, E. 1984. Phonology and Syntax. Cambridge, MA: MIT Press. Sells, P. 1990. More on light verbs and 0-marking. Ms. Stanford University. Sells, P. 1995. Korean and Japanese morphology from a lexical perspective. Linguistic Inquiry, 26, 277-325. Shibatani, M. 1977. Grammatical relations and surface cases. Language, 53, 789-809. Shibatani, M. 1978a. Mikami A kira and the notion of‘ subject’ in Japanese gram m ar In Problems in Japanese S yn ta x and Sem antics, ed. J. Hinds and I. Howard. Tokyo: K aitakusha. Shibatani, M. 1978b. Nihongo no BunseM . Tokyo: Taishukan. Shibatani, M. 1990. On parametric syntax. Studies in Generative Grammar, 1, 243-270. Shyu, S.-I. 1995. The Syntax of Focus and Topic in Mandarin Chinese. PhD Thesis. University of Southern California. Sigurdsson, H A . 1990. Icelandic Case-marked PRO and the licensing of lexical arguments. N atural Language and Linguistic Theory, 9, 327- 363. Speas, M. 1991. Functional heads and inflectional morphemes. The Linguistic Review , 8, 389-417. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 286 Taglicht, J. 1984. M essage and Em phasis. London: Longman. Takano, Y. 1996. Movement and Param etric Variation in Syntax. PhD Thesis. University of California, Irvine. Takezawa, K. 1987. A Configurational Approach to Case-Marking in Japanese. PhD Thesis. University of Washington. Tomioka, S. 1993. Verb movement and tense specification in Japanese. WCCFL, 11, 482-494. Tonoike, S. 1991. The comparative syntax of English and Japanese. In C urrent E nglish Linguistics in Japan, ed. H. Nakajima. Berlin and New York: Mouton de Gruyter. Watanahe, A. 1992. Wh-in situ, subjacency, and chain formation. M IT Occasional Papers in Linguistics, 2. Williams, E. 1978. Across-the-board rule application. Linguistic Inquiry, 9,31-43. Yamada, Y. 1936. Nihonbunpoogaku Gairon. Tokyo: Hoobunkan. Yoon, J. H.-S. 1995. Nominal, verbal, and cross categorial affixation in Korean. Journal o f East Asian Linguistics, 4, 325-356. Zaenen, A. and J. M aling. 1990. Unaccusative, passive and quirky case, hi M odem Icelandic Syntax (Syntax and Sem antics, 24), ed. Maling, J. and A. Zaenen. New York: Academic Press. Zaenen, A., J. M aling and H. Thrainsson. 1990. Case and grammatical functions: the Icelandic passive. In M odem Icelandic S yn ta x (Syntax and Sem antics, 24), ed. Maling, J. and A. Zaenen. New York: Academic Press. Zanuttini, R. 1991. Syntactic Properties of Sentential Negation: A Comparative Studies of Romance Languages. PhD Thesis. University of Pennsylvania. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 287 Zubizarreta, M.L. 1993. The grammatical representation of topic and focus: implications for the structure of the clause. Ms. University of Southern California. Zwicky.A. 1985. Clitics and particles. Language, 61, 283-305. Zwicky, A. and G. Pullum. 1983. Cliticization vs. inflection: English n ’ t. Language, 59, 502-513. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A y : , . IMAGE EVALUATION TEST TARGET (Q A -3 ) % 1. 0 l.l 1 . 2 5 L: L # 12.2 K 6 liO 1.4 1 . 8 1 . 6 150mm V V W o / ’/J yflP P L IE D A IIVMGE . In c 1653 East Main street ■ Rochester. NY 14609 USA . a = r".3= Phone: 716/482-0300 — Fax: 716/288-5989 0 1993. AppUM Image. Inc.. A ll Righls Reserved Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Representation of focus and presupposition in Japanese

PDF

Definiteness types in Spanish: A study of natural discourse

PDF

On the status of possessives

PDF

The accidence and syntax in john Wallis' 1653 "grammatica linguae anglicanae": a translation and a commentary on its alleged relationship to the 1660 port-royal "grammaire generale et raisonnee"

PDF

The syntax of focus and topic in Mandarin Chinese

PDF

The syntax of clitic doubling in modern Greek

PDF

The syntax of the Chinese BA-constructions and verb compounds: A morphosyntactic analysis

PDF

Emotions: Linguistic representation and cultural conceptualization

PDF

Language services planning in the banking industry: An example of unplanned language policy

PDF

Aspects of questions in Japanese and their theoretical implications.

PDF

On the representation and licensing of Q and Q-dependents

PDF

Morphology and the lexicon: Exploring the semantics-phonology interface

PDF

Morpheme recognition in prosodic morphology

PDF

In the event of focus

PDF

Rhetorical abstraction as a facet of expected response: A structural equation modeling analysis

PDF

The phonemics and morphology of Hokkaido Ainu

PDF

Syntactic structures in nominals: A comparative study of Spanish and Southern Quechua

PDF

Evaluating heaviness: Relative weight in the spoken production of heavy-NP shift

PDF

Intra-lexical noun-verb dissociations: Evidence from Chinese aphasia

PDF

Restructuring infinitives and the theory of complementation.

Asset Metadata

Creator Aoyagi, Hiroshi (author)

Core Title On the nature of particles in Japanese and its theoretical implications

Degree Doctor of Philosophy

Degree Program Linguistics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag language, linguistics,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-373600

Unique identifier UC11350832

Identifier 9919009.pdf (filename),usctheses-c17-373600 (legacy record id)

Legacy Identifier 9919009.pdf

Dmrecord 373600

Document Type Dissertation

Rights Aoyagi, Hiroshi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA