University of Southern California Dissertations and Theses

Aspects of questions in Japanese and their theoretical implications.

(USC Thesis Other)

Aspects of questions in Japanese and their theoretical implications.

PDF

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 affect 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 Information Company 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761-4700 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. A SPEC TS O F Q U E ST IO N S IN JA PA N ESE A N D THEIR TH EO RETICAL IM PLICATIONS by Shin Watanabe A Dissertation Presented to the FA C U L T Y OF TH E G R A D U A T E SCH O O L U N IV E R SIT Y OF SO U T H E R N C A LIFO R N IA In Partial Fulfillm ent o f the Requirements for the D egree D O CTO R OF PH ILO SO PH Y (Linguistics) A ugust 1995 C opyright 1995, Shin W atanabe Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 9621643 UMI Microform 9621643 Copyright 1996, by UMI Company. All 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 90089 This dissertation, written by W a ta n a b e , S h in under the direction of h...i.s.... Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean Date D IS S E R T A T IO N CO M M ITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A cknow ledgm ents It w as a long w ay. A lm ost eight years ago, I started m y graduate study in a little city in Japan called Tsukuba. At that tim e I did not even dream o f finishing up m y thesis in this huge city o f Los A ngeles. S in ce the very beginning o f m y life as a graduate student, I have been looking forward to writing this portion o f m y dissertation . I w ould like to thank all the m em bers on m y dissertation com m ittee; Hajime H oji, N am -kil K im , Barry Schein and Jean-R oger Vergnaud. It w as such a privilege that I could stand on their shoulders for the past five years. H ajim e w as m y academ ic advisor throughout those years, and I am very glad that I can finally release him from the heavy duty o f being chair o f my dissertation com m ittee. He spent so m uch tim e w ith me in discussing ideas that were oftentim es undeveloped and som etim es even hopeless. I am very proud that people outside U SC som etim es call m e a “cop y” o f Hajim e. I am indebted to N am -kil for his extrem e patience and invaluable com m ents from a view point o f typology. I ow e to Barry Schein much o f m y little k now ledge o f formal sem antics. It w as the series o f his lectures that taught m e how sem antic aspects o f languages can be incorporated into a scientific inquiry o f linguistics. A bove all, his persistent intellectual pursuit in a D avidsonian approach dem onstrated to me (and to all the students at U SC , I think) the im portance o f establishing "your ow n world." I am grateful to Jean-R oger for his constant help and encouragem ent. W henever I w as stuck, he w as there to offer me a helping hand. Jean-R oger is so smart that I cannot believe the fact that he and I Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. are at the sam e biological stage o f evolution. I also thank him for his treats at "Larchmont Village." I also benefited from many teachers and friends in the U SC linguistics com m unity. T hey include N ancy Antrim, H iroshi A oyagi, Joseph A oun, Jose C am acho, Bernard Com rie, D aeho Chung, A licia Grecka, Audrey Li, Yuki M atsuda, K eiko M iyagaw a, Hiro Oshita, Liliana Sanchez. M aki W atanabe and M aria-luisa Zubizarreta. Keiko M iyagaw a deserves special m ention for her assistance at the various phases o f the writing o f this dissertation. M any thanks go to m y teachers w ho first taught me linguistics at U niversity o f Tsukuba: Y ukio H irose, N obuhiro Kaga, M inoru Nakau, M am oru Saito Takashi Shim aoka, Haraguchi Shosuke, H idekazu Suzuki, K oichi T akezaw a R yuuichi W ashio. It is Minoru Nakau w ho introduced m e to this academ ic field and sh ow ed m e what it takes to be a real linguist. I also want to thank those w ho used to be m y fellow students there. A m ong them are Jun A be, Shinsuke Honm a, H iroto H oshi, D aisuke Inagaki, Masaharu Shim ada, Yuji Takano, K azue Takeda and Shin-ichi Tanaka Finally and m ost o f all, I thank m y parents Akira W atanabe and H iroko W atanabe for their love, patience, and constant support, and m y aunt H atsue G lass for her lim itless encouragem ent. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. iv Table o f Contents C h ap ter O ne. Introd u ction .................................................................................................1 C hapter T w o . (A n ti-)S u p eriority as W eak C ro sso v er............................................. 7 2 .1 . I n tr o d u c tio n ..................................................................................................................7 2 .2 . A nti-superiority, ECP and R igid ity C onstraint 1......................................11 2 .2 .1 . A S u m m a r y ...................................................................................................12 2 .2 .2 . A S u sp e n sio n E ffe c t.................................................................................. 16 2 .2 .3 . C om parative D e le tio n ........................................................................18 2 .2 .4 . T he ECP and the M in im alist T heory o f C h om sk y...........................20 2.3. The A nti-superiority E ffect as an Instance o f W eak C rossover...............22 2 .3 .1 . C la u se-In tern a l S c r a m b lin g .....................................................................23 2 .3 .2 . L o n g -d ista n c e S c r a m b lin g ....................................................................... 25 2 .3 .3 . A S u sp e n sio n E ffe c t.................................................................................. 27 2 .4 . E x ten d in g W eak C r o sso v e r ............................................................................29 2 .4 .1 . C hierchia (1 9 9 1 , 1 9 9 2 -1 9 9 3 ) and S om e rev isio n s.........................30 2 .4 .2 . H ornstein (1 9 9 4 , Ch. 7) and S om e R ev isio n s.............................. 37 2.5. Subsum ing the A nti-superiority E ffect under the W C O -phenom ena.......4 4 2 .5 .1 . C a n o n ic a l C o n tra st...................................................................................... 4 4 2.5 .2 . On Low er W h-e ffe c ts and R elated P h en om en a ................................ 55 2.5 .3 . On H igher W h-e ffe c ts and R elated E x a m p les...................................64 2.5 .4 . A N ote on C lause-boundedness o f the Anti-superiority E ffects......68 2 .5 .5 . Suspension o f A nti-superiority E ffects by Scram bling.......................69 2 .5 .6 . T he A n ti-su p eriority E ffect in E n g lish .............................................. 71 2 .5 .7 . A S u p eriority E ffect in J a p a n ese........................................................73 2 .6 . C o n c lu s io n ......................................................................................................... 79 Chapter Three. T w o M odes o f Scope A ssignm ent to W h-p h r a s e s.......................... 81 3 .1 . In tr o d u c tio n .............................................................................................................81 3.2. An A ssum ption C oncerning The Nature o f LF W h-m o v e m e n t.................87 3.3. T he “D om ain” o f an Interrogative C om plem entizers and....................... 97 Suspension o f W h-island Effects 3 .3 .1 . F oci o f Y es-N o Q u e s t io n s .......................................................................... 9 9 3 .3 .2 . E xtension to W h-questions: On "U nselective Binding" and 104 Scope A ssignm ent to W h-phrases Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.4. A Formal D efinition o f the “Dom ain" o f an Interrogative C om plem entizer 110 3 .4 .1 . A D efin itio n and V erb -R aisin g............................................................I l l 3 .4 .2 . On C o p u la S e n te n c e s ..............................................................................123 3 .4 .3 . N e g a tio n and F o c u s ............................................................................. 131 3.5 . C on cep tu al and E m pirical Im p lication s.......................................................136 3.5 .1 . T he C om p lex N P Constraint and “P ied -p ip in g” ............................136 3.5.2. Extraction o f C onjoined W h-phrases Out o f a W h -isla n d ...............143 3 .5 .3 . S o m e C o p u la S e n te n c e s......................................................................146 3.5.4. A nother T ype o f W h-isla n d V io la tio n ............................................... 148 3 .6 . C o n c lu s io n ..................................................................................................................152 Chapter Four “S tylistic Restriction on Q uestions in Japanese: D escrip tion s 154 and a Possible A venue Toward Explanation 4 .1 . In tro d u ctio n ........................................................................................................... 154 4 .2 . B a sic F a cts ............................................................................................................160 4 .3 . A “G o v ern m en t” R eq u irem en t...................................................................163 4.4 . Y es-N o Q u e s t io n s ................................................................................................ 168 4 .5 . E m p irica l J u stific a tio n .......................................................................................175 4 .5 .1 . A Paradigm in the N o-da / desu C o n str u c tio n ......................................... 176 4 .5 .2 . A n o th er P a ra d ig m ..................................................................................... 179 4.6. A Puzzle in Y es-N o Q u e s t io n .....................................................................................182 4 .7 . A S p e c u la tio n ......................................................................................................... 184 4.8. More on Y es-N o Q u e s t io n s .............................................................................. 194 4 .9 . Sum m ary and C o n c lu sio n ........................................................................... 197 R e fe r e n c e s ...............................................................................................................................201 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 Chapter One Introduction The goal o f this thesis is to present proposals in generative grammar where I intend to account for various aspects o f interrogative sentences primarily in Japanese.1 In this brief introductory chapter, I shall outline what em pirical issues w ill be considered and what conclusions w ill be reached based on those em pirical argum ents. Chapter 2, focusing on m ultiple w h-questions. w ill discuss a phenom enon called "anti-superiority effect" by A. W atanabe (1991). In particular, I w ill be concerned w ith contrasts such as that in (1) w hich have been w idely discussed in the literature (e.g. Saito (1 9 8 2 , 1992b), Kuno andT akam i (1993) and A. W atanabe (ibid.) am ong others): (1) ( = ( 1 ) , Saito (1992b ) ) a. *John-w a [ y p naze [ y p nani -o katta ]] no ? -T om w hy w hat-A cc bought Q 'John bought w h v w hat? 1 b. John-w a [ y p nanit-o [ y p naze [y p tj katta]]] no ? -T op w hat-A cc w hy bought Q 'John bought w hat w h y . ' 1 This dissertation presupposes the reader's familiarity with generative grammar, esp ecially its current developm ent within Chom sky's (1992, 1994, 1995) "minimalist program." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The rudimentary descriptive generalization is that in a multiple w h-question, naze ’w h y’ must be c-com m anded by another w h-phrase.: Thus, the exam ple in (la ) is ungrammatical, for nani 'what' fails to c-com m and naze. In ( lb ), on the other hand, this deviant exam ple is rem edied by scram bling nani to the VP-adjoined position, where it can c-com m and naze. W hy is it that naze must be c-com m anded by another wh-phrase in a multiple question? I shall first assum e, follow in g Chierchia (1991, 1992-1993) and H om stein (1 9 9 4 ), that in a m ultiple w h-question. naze ought to sem antically depend upon another w h-phrase. and that this dependency is form ally represented at LF by a "bound variable anaphora" o f the form in (2): whi naze (x It w ill be argued that naze is construed as a reason-valued "Skolem Function" (cf. G roenendijik and Stok h of (1989)) w hose argument is bound to another w h-phrase w h i in this schem atic multiple w h-question. By virtue o f this formal relation at LF, the value o f naze "covaries with" w h [, yielding a pair-list interpretation / answer. 2 Let us sim ply assum e a custom ary "first-branch definition" o f c-com m and (cf. Reinhart (1 9 8 3 )). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 Under this assum ption, (la ) w ill have an LF-representation w hich has a feature as in (3): (3) (sim plified) naze(x) Here x is a variable bound to nani. I shall argue that this bound variable anaphora is illicit, since it violates the w ell-know n weak crossover constraint ju st like (4): (4) H isi mother loves every b o v i . T his sentence is excluded, since his^ is not c-com m anded by every b o y i . (3) (hence ( la)) is analogously ruled out. Chapter 3 w ill consider how w h-phrases in Japanese are assigned scop e in different linguistic contexts. Putting forth rather traditional ideas, I shall propose that the scope o f a w h-phrase is represented either (i) by LF w h - m ovem ent (cf. C hom sky (1973)), or (ii) by a “binding relation” betw een a w h-phrase and an interrogative com plem entizer such as ka (cf. C. L. B aker (1 9 7 0 ) and P esetsky (1987)). (5a) and (5b) schem atically represent the options in (i) and (ii), respectively: (5) (The linear orders are irrelevant.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 a. b. wh] I shall propose that the LF w h-m ovem ent in (5a) is an operation that is analogous to M ay’s (1977, 1985) Q(uantifier) R (aising) in that they are both triggered solely to assign scope to scope-baring expressions. I shall further maintain that w henever the option in (5b) is available, this LF w h-m ovem ent does not have to take place (and hence must not take place in the spirit o f the m inim alist theory), sim ply because an application o f this operation is not necessary to assign scope to a given w h-phrase. In this chapter, I shall pay close attention to the fact that Japanese allow s w h - questions (at least) in tw o distinct constructions; a "bare-form" as in (6a) and no - desu as in (6b): a. John-w a nani -o kaimashita ka? -T op w hat-A cc boughtp0|ite Q 'John bought w hat? ' b. [John-w a nani -o katta no] desu ka? -T op w hat-A cc bought C om p isp0iite Q ’It is that John bought w hat? ' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (6) (6a) is a sim ple w h-question. (6b), on the other hand, basically em beds (6a) in a clause headed by the so-called "nominal complementizer" no, and this no-clause is im m ediately follow ed by a copula desu ‘is.’ I shall suggest that the binding option in (5b) is available only in w h-questions o f this latter form. This option not being available, LF w h-m ovem ent must apply in (6a) in order to assign scope to nani ‘w h at.’ Chapter 4 w ill be a preliminary descriptive study regarding what conditions need to be satisfied by questions in the "bare" form and in no-desu. A n interesting fact, pointed out initially by S. M iyagaw a (1987) and U eyam a (1991, 1992), is that the polite verb forms used in (6) cannot be replaced by their neutral counterparts: (7) a. (based on S. M iyagaw a's (4)) * John-w a nani -o katta ka? -Top w hat-A cc b oughtN eutral Q 'John bought w hat? ' b. * [John-w a nani -o katta no] da ka? -Top w hat-A cc bought C om p isjveutrai Q 'It is that John bought w hat?' T hese lexical changes equally render the tw o types o f questions rather marginal. Parallelism , how ever, collap ses, w hen w e turn our attention to ves-n o questions: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 (8) a. John-w a hon -o katta ka? -Top book-A cc boughtneutrai Q 'John bought a book?' b. *[John -w a hon -o katta no] da ka? -Top book -A cc bought C om p isneutral Q 'It is that John bought a book?' T hese sentences m inim ally differ from those in (7) only by nani 'what' is replaced by hon 'book' so as to obtain ves-n o questions. The result is that only (8a) turns out to be ill-form ed. A ssum ing that these properties o f Japanese interrogative sentences, am ong many others presented in this chapter, are to be explained in syntax, the purpose o f this chapter is three-fold. First, I shall suggest that S. M iyagaw a's (1897) "government restriction" on ka explains parts o f intricate restrictions on Japanese w h-questions. I shall particularly argue that such w h-questions as those in (7) are ill-form ed, since ka is left "ungovem ed" in these exam ples. S econ d , unlike S. M iyagaw a, I w ill su ggest that ves-n o questions are free from this constraint. Third, I shall su ggest that a v es-n o question is, how ever, subject to a rather different constraint from the "government" restriction on w h -q u estion s. In discussing this matter, I w ill inquire about w hy da, a non-polite copula, may not occur in a v es-n o question, as (8b) illustrates. The discu ssion o f this chapter w ill be descriptive rather than conceptual. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter T w o (Anti-)Superiority as W eak Crossover' 2. 1. Introduction Since A . W atanabe (1991), som e phenom ena found in Japanese have attracted m uch attention in the theory o f syntax within the generative tradition. A. W atanabe generalizes several related observations under the nam e o f anti-superioritv effects. The rough descriptive generalization in (1) "extracts" from his original what I shall be concerned with in this chapter:1 * The idea in this chapter cam e out o f a discussion that I had with K eiko M iyagaw a in 1993. 1 A. W atanabe’s generalization is given in (i): (i) ( = (22), A . W atanabe (1 9 9 1 )) Anti - superiority effects The wh-phrase that is m oved first cannot c-com m and the other w h- phrase at S-structure w hich takes the sam e scope. G iven the standard ECP argument in C hom sky ( 1986b), a pure adjunct w h like naze 'w h y * m ust m ove prior to the other w h-phrases in a multiple w h-question: O therw ise, the trace o f naze cannot be antecedent-governed. Then the generalization in (i) requires o f naze not to c-com m and w h-phrases if they take the sam e scope. (2b), not (2a), is thus excluded, sin ce naze c-com m ands nani 'what' and they take the sam e scope. with permission of the copyright owner. Further reproduction prohibited without permission. ( 1) the anti-superioritv effect In a Japanese multiple w h-question. naze 'why' must not c-com m and another wh-phrase when they take the sam e scope. Consider the exam ples in (2) and (3): (2) ( = (1). Saito (1992b ) ) a. John-w a nani -o naze katta no ? -N om w hat-A cc w hy bought Q 'John bought what w h y . ' b. *John-w a naze nani -o katta no ? -T om w h y w hat-A cc bought Q 'John bought w h y w h a t.' Nani 'what' and naze 'why' take the sam e scope in these sentences. (2a), conform ing to the generalization in (1), is a fully acceptable question.2 On the other hand, (2b) sounds rather unnatural, if not im possible, for naze c-com m ands nani in this question. This latter exam ple sharply contrasts in acceptability with (3): 2 Hajim e Hoji (p.c., 1995) pointed out to me that there may be an "adjacency condition" on m ultiple w h-questions o f this sort: (i) (based on exam ples suggested by H oji)\ a. Dane -ga naze kinoo ano kooen-de Mary -ni atta no? w h o-N om w h y yesterday that park -at -to m et Q 'W hy w h o m et Mary at that park yesterday.' b. (?)Dare-ga kinoo naze ano kooen-de M ary-ni atta no? c.(??)D are-ga kinoo ano kooen-de naze Mary-ni atta no? d.(??‘ ?) Dare-ga kinoo ano kooen-de M ary-ni naze atta no? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 (3) M ary-wa naze [ John-ga nani -o katta ka ] shiri-tagatteiru no ? -T op w hy -N om w hat-A cc bought Q know -w ant Q 'W hy d o es M ary want to know w hat John bought ?' Despite that naze c-com m ands nani. (3) is fully acceptable. T his is because the scope dom ains o f the wh-phrases are distinct; naze takes scope over the matrix clause, w hereas nani takes scope over the em bedded clause. The goal o f this chapter is to suggest that the anti-superiority effect as illustrated by (2b) falls under the general w eak crossover phenom ena. R oughly, I shall argue that in a multiple w h-question. naze obligatorily involves a variable bound to another w h. Thus postulated, (2b) w ill have an LF-representation w hich has a feature indicated in (4): (4) (sim plified) naze(x) nani Here x is a variable bound to nani. This bound variable anaphora is illicit, since it violates the w ell-know n w eak crossover constraint just like (5):3 T hese exam ples are all consonant w ith the generalization in (1), for dare 'who' c- com m ands naze 'why.' H ow ever, (ib) - (id) appear to be degraded com pared with (ia), as naze "displaced" further from dare. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 10 (5) H is i mother loves every b o v i . T his sentence is excluded, sin ce hisj ^ fails to be c-com m anded by every b o v i . (4) is an alogously ruled out. The nature o f this phenom enon is not clear to me at this point. N or is it clear how solid a generalization w e can make based on this observation. T hough the questions in (ib) - (id) are indeed deviant if presented in isolation, they can be rem edied by putting them in discourse contexts: (ii) A: Yuki to K eiko to Maki-ga kinoo ano kooen-de sorezore and and -N om yesterday that park -at respectively betsu-no -riyuu -de M ary-ni aim ashita. different-reason-for -to met 'Yuki, K eiko and M aki m et M ary at that park yesterday for different reasons.' B: (= (ic)) (G utaiteki-ni) dare -ga kinoo ano kooen-de naze M ary-ni atta no (to be sp ecific) w ho-N om yesterday that park -in w hy -to m et Q (ka osh iete kudasai.) (please tell me) Here I put the question in (ic) in a discourse context where a particular set o f people com prising Yuki. K eiko and Maki is introduced by speaker A . W ith this specific set g iv en , speaker B's utterance, i.e. (ic), seem s felicitous. Incidentally the presence o f a phrase w ith a stress affects the acceptability o f a w h-q u estion w ith naze: (iii) * Dare-ga kinoo ano kooen-de naze M A R Y -N I atta no? T his exam ple is exactly the sam e as (ic) except that M ARY is assigned an emphatic stress. T his exam ple seem s h op elessly bad. I leave this issue open for future d iscu ssio n . 1 S ee footnote 19 for a form ulation o f the "weak crossover constraint." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This chapter is organized as follow s. In section 2.2, I w ill briefly review A. W atanabe's w ell-know n account for the anti-superiority effect, and then point out som e potential empirical and conceptual problems. Section 2.3 w ill provide initial em pirical justification o f the W CO analysis o f the anti-superiority effect. I w ill show three sim ilarities betw een the anti-superiority effect and the W C O phenom ena w hich lead m e to reduce the former to the latter. In section 2.4, a conceptual fram ework w ill be introduced as a basis o f the later discussion o f the W C O analysis. T o be more concrete, I w ill sum m arize, with som e m odifications, Chierchia's (1991, 1992-1993) treatment o f w h-questions with quantifiers and H ornstein's (1994) reanalysis o f superiority effects in English. In section 2.5, the m ain section o f this chapter, I w ill show that the anti-superiority effect is an instance o f w eak crossover. Various em pirical considerations w ill be presented including the claim that superiority effects in English and the anti-superiority effect in Japanese should be both considered as cases o f w eak crossover in a unified m anner. Section 2.6 w ill be a brief concluding remark. 2. 2. A nti-superiority, ECP and R igidity Constraint First this section w ill briefly review A. Watanabe's (1991) analysis o f the anti superiority effect in question, and then show what appear to m e problem atic with his approach. Som e conceptual issues regarding his fram ework w ill also be d iscu ssed in relation to C hom sky's (1 9 9 2 , 1994, 1995) m inim alist theory. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 2.2 .1 . A Sum m ary B asically extending Saito (1982), A . W atanabe characterizes the anti-superiority effect in terms o f what he calls the R elation Preservation condition (= (6)), the ECP and other auxiliary assum ptions.4 A certain fact concerning relative scope interpretations in Japanese leads A. W atanabe (1991) to formulate the Relation Preservation condition in (6), where relation is defined in terms o f the notion o f seg(m entVcom m and in (7); (6) ( = (2 1 5 ), A. W atanabe (1 9 9 1 )) Relation Preservation : A relation established at a certain point o f derivation m ust be maintained throughout. (7) ( = (208), A. W atanabe (1 9 9 1 )) X seg(m enti-com m ands Y iff X d oes not dominate Y and every segm ent that dom inates X dom inates Y , w here X , Y are categories. In particular, as Kuroda (1 971, 1992 Ch. 2) and Hoji (1 9 8 5 , Ch. 4) exten siv ely discuss, w hen tw o quantified N P s occur in the positions w here they originate at the point o f Spell-O ut, the one higher than the other obligatorily takes w ider scope. C onsider the sentence in (6): 4 A lthough A. W atanabe (1991) adopts the framework o f C hom sky (1986b), I w ill sum m arize his proposal within the m inim alist theory, except the use o f the ECP, to make the argument coherent with other sections o f this chapter. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 13 (8) Dareka-ga daremo -o aishiteiru. som eon e-N om everyone-A cc love a. There is a person w ho loves everyone. b. *For every o n e i, there is a person w ho lov es him i. N o m ovem ent being involved,5 dareka 'som eone' and daremo 'every' appear in their original p ositio n s.6 Thus, as indicated, the former necessarily takes w ide scope with respect to the latter. G iven Q(uantifier) R(aising) o f M ay (1 977, 1985), the readings in (8a) and (8b) are represented by the LF-representations in (9a) and (9b), respectively: (9) a. [ip dareka-ga| [ip darem o- 0 2 [ip tj t2 aishiteiru ]]] (= (6a)) b.* [ip d arem o - 0 2 [ipdareka-gai [ip ti t2 a ish iteiru ]]] (=(6b )) W e need to rule in (9a), but rule out (9b). 5 For discussion concerning how m ovem ent operations affect scope interpretations, I w ould like the reader to refer to Kuroda (1 971, 1992 Ch. 2), H o j i( l9 8 5 , Ch. 4) and A . W atanabe (1 9 9 1 ). 6 The "original" position o f a N P is w here it is inserted by G (eneralized) T(ransform ation) in the sen se o f C hom sky (1 9 9 2 , 1994, 1995). On this v iew , a question arises w hy it is that in (8) the “them e” N P is inserted prior to the “agent” N P. A n answ er to this question obtains with recourse to the follow in g custom ary assum ptions: (i) Them atic roles are “hierarchically” ordered in the lexical entry o f a lexical item (cf. G rim shaw (1990)), and (ii) in a given lexical entry, an assignm ent o f “low er” thematic role m ust “precede” those o f the higher roles. Supposing further that the them e role is low er than the agent role, it follow s that the form er must be assigned First. For m uch relevant discussion, see H ale and K eyser (1993). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 This is what the Relation Preservation condition purports to characterize. The licit LF-representation in (9a) conform s to the condition; dareka seg-com m ands daremo at the point o f Spell-O ut, and this relation is preserved at LF. On the other hand, the illicit one in (9b) is in violation o f this condition, since the seg-com m and relation at Spell-O ut betw een dareka and daremo is not retained at LF. G iven this Relation Preservation condition, A. W atanabe associates the grammatical exam ple in (10) with the LF-representation in (11): ( 10) Dare-ga naze soko-ni itta no? w h o-N om w hy there-to w ent Q 'W ho w hy w ent there?' ( 1 1 ) [cp lNPdare-gaifNPnaze? [NP Op h h h tlP 11 tlP 12 soko-ni itta ]] no ] Supposing the standard D P -hypothesis (A bney (1987)), A. W atanabe assum es that a null operator raises to the C P-spec position before Spell-O ut from the D P -spec position o f a w h-phrase7. 8 M oreover, he posits that a wh-phrase that has launched this operator m ust adjoin to the operator at LF to form a constituent. 7 H e assum es that in a m ultiple w h-question. a null-operator is discharged from one o f the w h-phrases in volved in the sentence. x A . W atanabe proposes that adjunct wh-phrases including naze projects to D P 's. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 The licit LF in (11) indicates that a null operator originates in the spec o f naze and m oves before Spell-O ut to the CP-spec position. In order to form a constituent, naze adjoins to this operator at LF, and subsequently dare 'who' adjoins to this adjoined structure. (11) is legitim ate with respect to the ECP; naze, m oved prior to dare, can antecedent-govem its trace, and the trace o f dare is lexically governed by the Infl.9 The Relation Preservation condition is also observed; dare seg-com m ands naze at the point o f Spell-O ut and this relation is retained at LF. Finally let us consider the typical anti-superiority violation in (12): ( 12) *N aze dare -ga soko-ni itta no? w hy w ho-N ofn there-to w ent Q 'W hy w ho w ent there?' Plainly, this sentence is forced to map to the LF in (13) in order to satisfy the ECP: (1 3 ) * [cp [Npdaffi=gai[Npnaze2 [N P Op h h k [lP l2 U soko-ni itta ] no ] 9 S ee H uang (1 9 8 2 ) and Lasnik and Saito (1984, 1992). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 A s in (11), a null operator originates in the spec o f naze and m oves before Spell- Out to the C P -spec position. In order to form a constituent, naze 'w hy' adjoins to this operator at LF. and subsequently dare-ga 'w h y-N om ' adjoins to this am algam . T he ECP is satisfied on a par with (11). (13) is h ow ever ruled out by the Relation Preservation condition; naze seg-com m ands dare-ga at the point o f Spell-O ut, w hereas the latter seg-com m ands the former at LF. Consequently (12) is exciuded. 2.2.2. A S u sp en sion Effect A. W atanabe's account appears to be challenged by an observation in Saito ( 1992b) concerning how the anti-superiority effect is circum vented. C onsider the contrast in (1 4 ):1 0 (14) a. (Saito's (2b )) * Naze dare -ga soko -ni itta no? w hy w ho-N om the place-to went Q 'Why w ho went there?' b. (Saito's (3b )) ?N aze dare-ga doko -ni itta no? w hy w ho-N om what place-to w ent Q 'W hy w ho went where?' For the sake o f fairness, A. W atanabe (1 9 9 1 ) considers exam ples such as (14) and (15) all ungrammatical. So his factual claim s and his theory are quite consistent. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17 (15) a. *N aze dare -ga John-ni hon -o ageta no ? why w ho-N om -to book-A cc gave Q 'Why w ho gave to John a book?' b. ?Naze dare -ga dare-ni hon -o ageta no ? w hy w h o-N om w ho-to book-A cc gave Q 'Why w ho gave to w h o a book?' c. ?Naze dare -ga John-ni nani -o ageta no ? why w ho-N om -to what -A cc gave Q 'W hy w ho gave to John what?' W e observe in these exam ples that an effect o f anti-superiority is suspended or at least weakened, if a "third" w h occurs low er than an "offending" w h v -w h sequence. This fact is unpredictable on A. W atanabe’s account. Take ( 14b) for instance, where a potential anti-superiority violation is remedied (or, at least w eakened) by doko 'which place.' T he LF o f ( 14b), given in (16), in volves null operator m ovem ent from the (D P -)spec o f naze to satisfy the ECP: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 18 (16) CP [N3 d a re -g a ] NPl N Pi IP [c no ] [PP2 d o k o -n i] N P i t, IP [Advi nazej O pi t3 VP [vitta] T his LF is how ever in clear violation o f the Relation Preservation C ondition; naze seg-com m ands dare and doko at the point o f S p ell-O u t, and this relation is not preserved at LF." Thus, ( 13b) w ould be falsely excluded. 2.2.3. Com parative D eletion Another em pirical problem, less relevant for the present discussion o f the anti- superiority effect, is regarding his assum ption that a Japanese w h-question involves pre-Spe 1 1 -Out null-operator m ovem ent from the D P-spec o f a w h -p h rase.1 2 1 1 The exact working o f the Relation Preservation condition is som ew hat more intricate, albeit in an irrelevant sense for the present discussion. For details, see A. W atanabe (1991). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19 In so far as I can judge, his strongest argument for such a pre-Spell Out operation is based on an exam ple such as (17): (17) (cited from Ishii (1991. p. 189), Ishii o w es this exam ple to Kikuchi (1 9 8 9 )) B ill-w a [cp O pi [ip John-ga [ ei kaita ato ] Pari-de t| happyooshita ] -T op -N om wrote after -in published yori-m o ] ookuno ronbun-o Am erica-de kaiteita. than many article-A cc -in has written 'Bill wrote more articles in Am erica than John had published t in Paris after he wrote e in London.' A ccording to K ikuchi (1989), w hich A. W atanabe esp ou ses. e | is a parasitic gap. G iven that a parasitic gap needs to be licensed by a variable created by pre-Spell- Out m ovem ent (C hom sky, 1986b), this licensing, if true, is a strong indication o f null-operator m ovem ent before Spell-O ut. H ow ever, Ishii (1991, 191) con vin cin gly argues that the gap in question is not necessarily a parasitic gap. F ollow in g C hom sky ( 1986b, Ch. 1 0 ), suppose that the parasitic gap itself is a trace left behind by null operator m ovem ent. W e thus predict a Subjacency violation when a gap such as ej in (17) is em bedded in an island. T his prediction is not borne out how ever, as Ishii points out: 1 2 Supposing that only pre-spell-out m ovem ent obeys the Subjacency condition (H uang (1 9 8 2 , Lasnik and Saito (1 9 8 4 , 1992), A. W atanabe needs to assum e this null-operator m ovem ent to take place before Spell-O ut in order to explain effects o f the w h-island constraint. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 20 (18) (= (214), Ishii (1991) ) B ill -\va [cp Opi [ip John-ga [ ei kaita ] hito -ni atta ato ] Pari-de t| -Top -N om wrote person-to met after -in katta ] yori-m o ] ookuno ronbun-o Am erica-de katta. bought than many article-A cc -in bought 'Bill bought more books in Am erica than John bought t in Paris after he m et the person w ho wrote e in Paris.' The grammaticality o f this exam ple appears to ascertain that e | in (1 8 ) (and in (17)) does not need to be analyzed as a parasitic gap.1 3 This fact then significantly undermines Kikuchi's analysis o f the com parative deletion construction and, in turn, A. W atanabe loses his strongest empirical justification o f his pre-Spell-O ut null operator m ovem ent in w h-questions. 2.2.4. The ECP and the M inim alist Theory Recall that A, W atanabe's account has a crucial reference to the ECP. The notion, as w e custom arily assum ed in the EST framework, is a principle o f the language faculty, and thus it is underivable from more elem entary considerations or conceptions. 1 3 To be precise, Ishii's claim is much stronger in that the gap in question cannot be regarded as a parasitic gap; it is a pro. This is an em pirical result o f his theory o f "comparative deletion," w hich argues that an operator m oved in the construction is a floating quantifier incapable o f licensing a parasitic gap. For details, see Ishii (1 9 9 1 , C hw 3). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In the recent years, how ever, this assum ption has been questioned, and C hom sky (1992. 1994. 1995) totally elim inates this notion from the considerations o f the grammar. A s is w ell-know n, his M inim alist theory regards a particular language as an instance o f the language faculty with the options in the functional part o f the lexicon fixed. Phenom ena o f sound and m eaning o f the language are determined by representations at PF and LF, formed by m axim ally econom ical derivations, w hich conform to the Principle o f Full Interpretations and conditions on the representational levels. G iven this language design, C h om sk y! 1992, p.21 ), Kitahara (1994) and Takano (1994) am ong many others convincingly argue that large portions o f the data that have been accounted for with recourse to the ECP are properly characterized by econ om y conditions such as the M (inim al) L(ink) C (ondition) and the Principle o f G reed.1 4 (C hom sky, 1994, p. 14) If the ECP cannot indeed "sneak in" the form ulation o f the grammar. A . W atanabe's account o f the anti superiority effect, w hich crucially relies on it, is to be reconsidered.1 5 1 4 The MLC states that an elem ent makes the "shortest move" in the sense determ ined in term s o f X'-schem atic relations such as "equidistance." D ue to Greed, m ovem ent is licit only when som e m orphological property o f the m oved entity is satisfied by the operation. For discussion, see C hom sky (1 994, p. 14). 1 5 H ow ever A. W atanabe's important work should not be criticized for this conceptual matter, since the paper was written in 1991, w hen the entire picture o f the M inim alist theory w as beyond our reach. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2. 3. The A nti-superiority E ffect as an Instance o f W eak C rossover Having seen that A. W atanabe's (1991) account is not without em pirical and conceptual problem s. I w ould like to seek an alternative analysis where the anti superiority effect is subsum ed under the general weak crossover (W CO) phenom ena. Before going into theoretical details, this section w ill present initial em pirical justification o f this approach C onsider the exam ple in (19): (19) : t : John-w a naze nani -o katta no ? -Top w hy w hat-A cc bought Q '*John bought w h y what ?' On the W CO account o f the anti-superiority effect to be defended below . (19) is excluded alongside o f (20): (20) (based on exam ples given in Hoji (1 9 8 5 )) ♦?r T ei H itom e soitsu->-o mita ] onna])-ga dare->-o sukini natta no? a-glance he -A cc saw w om an -Norn w ho-A cc fell-in-love Q 'WhoT did [ the w om an that took a glance at hiirn ] fell in love with t2 ?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 23 A w ell-know n constraint is that a bound pronoun must be c-com m anded by its quantified antecedent. (20) is ruled out by this constraint, for dare 'who' fails to c- com m and soitsu 'he.' The W CO account o f the anti-superiority effect w ill maintain that naze in (19) contains an im plicit variable that purports to be bound, but fails to be c-com m anded, by nani 'what.' It follo w s from this assum ption that (1 9 ) is excluded as a W C O violation. 2.3.1. C lause-Internal Scram bling It has been noted in the literature (e.g. A be (1993), Hoji (1985), Saito (1 9 8 5 , 1992a). and Y oshim ura (1991) am ong many others), clause-internal scram bling cures a W C O violation, as evidenced by (21):"’ (21) (based on exam ples given in Hoji (1985)) a. (= (2 0 )) *?[ [ ej H itom e so itsin -o mita ] onna]( -ga dare->-o sukini natta no ? a-glance he -A cc saw w om an -N om w ho-A cc fell-in-love Q '[(The w om an that took a glance at him 2 ] fell in love with who->?' b. Dare-)-o [ [ e i hitom e soitsu?-o mita ] onna] i -ga t2 sukini natta no? ' W hoo [i the w om an that took a glance at hirm ] fell in love with t2 ? ' lft T his chapter w ill not discuss how to characterize this fact. For d iscu ssion , see the cited works and the references therein. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (21a) is ruled out, since soitsu fails to be c-com m anded by dare. In (21b), the clause-internal scram bling o f dare rem edies this deviance. If the anti-superiority effect is indeed an instance o f weak crossover, it should be attenuated by clau se internal scrambling. T his prediction appears to be borne out: ( 22 ) a. *?A nata-w a naze dare -o naguttano ? you-T op w h y w h o-A cc hit Q 'You hit w h y who?' b. A nata-w a dare-o naze nagutta no ? 'You hit w h o why?' Sim ply fo llo w in g Saito ( 1992b), I assum e that naze is base-generated in a V P- adjoined p o sitio n .1 7 Then it is indicated by (22b) that an anti-superiority violation is alleviated by clause-internal scrambling ( V P-intem al scrambling in this case). This parallelism betw een the W CO phenom ena and the anti-superiority effect is naturally exp ected , if the latter reduces to the former. 1 7 For an alternative view on the position o f naze, see Kuno and Takam i (1 9 9 3 ), where they argue that it is base-generated under S". Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.3.2. Long-distance Scrambling L ong-distance scram bling, viz. scram bling out o f a finite clause, does not suspend a W CO violation unlike clause-internal o n e .ls (23) exem plifies this point: ls A ccording to Saito ( 1992a. I l l , FN. 8), som e speakers accept exam ples like (23b). (H e ow es this suggestion to Y oung- Suk Lee and M asaru Nakamura (p.c.)). A s for this "apparent dialectical variation," Saito h im self g ives no explanation. Incidentally, to exam ine the nature o f "long distance " scram bling, the follow ing factor must be taken into account. C onsider the exam ples in (i) and (ii): (i) a. John-ga [M ary-ga baka da to ] om otta (koto) - N om - N om fool is C om p thought 'John thought that Mary is a fool.' b. John-ga Marv (no koto)- o [p ro ib a k a d a to] om otta (koto) -N om (-G en fact )-A cc fool is C om p thought 'John thought o f M ary that she is a fool.' (ii) a. H anako-ga [Y uki-ga K eik o -n i atta to] itta(koto) -N om -N om -Dat m et Com p said 'Hanako said that Yuki m et K eiko. ’ b. 7 H anako-ga K eiko ( no koto I- o [Yuki- ga proi atta to ] itta (koto) -N om -A cc -N om m et C om p said 'Hanako said o f K eiko that Yuki met h e r .' A s Hoji (1991, R ochester Japanese W orkshop) and Oka (1988) convincingly argue, verbs such as om ow 'think' in (i) and jw 'say' in (ii) allow an "extra accusative phrase " possibly as their argument. In (iib), for exam ple, Keiko ( no koto ) - o is licensed in the matrix clause and coindexed w ith grot • (For more details, see Hoji (1991) and O ka (1988).). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 26 (23) ( based on exam ples given in Saito (1992b)) a.*?[ [ e | H itom e so itsm -o mita ] onna] i -ga [M ary-ga dare-»-ni a-glance he -A cc saw w om an -N om -N om w ho-D at horeteiru to ] om otta no ? loved C om p thought Q '[ The w om an that took a glance at hirm ] thought that Mary was attracted to w h o ? ?' b. *?Dare-nb [ [ e i hitom e so itsm -o mita ] onna] i -ga [M ary-ga t2 horeteiru to ] om otta no ? 'W hoo [the w om an that took a glance at hirm ] thought that Mary w as attracted to t2 ?' It should be the case on the W CO account that long-distant scram bling cannot circum vent the anti-superiority effect. This appears to be a correct prediction: (24) a. *?T aro-w a naze M ary-ni [ anata-ga dare-ni atta to ] itta n o ? -T op w h y -to you -N om w ho-to m et Com p said Q 'Taro w h y said to Mary [ that you met to w hom ] ?' b. *?Dare-nii Taro -w a naze Marv-ni T anata-ga t| atta to ] itta n o ? w ho-to -T op w hy -to you -N om met Com p said Q ' T o w hom | Taro w h y said to Mary f that you met ti ] ?' T o avoid com plications related to this "extra accusative" phrase, the dative phrase dare-ni ‘w h o -to ’ is m oved in (23b). Since there is possibly no "extra dative phrases," this guarantee that long-distance scram bling is actually involved in the sentence. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (25) a .“John -vva naze Tom -ni f Marv-ni nani-go -ga vvakara -nai to ] -T op w hy -to -to what lan.-N om understand-not C om p itta no? said Q 'John w h y said to T om [ that Mary does not understand what languagel ?' b. * N an i-go -g a i John -w a naze T om -ni \ M arv-ni t| wakara -nai to ] what lan.-N om -Top w hy -to -to understand-not Com p itta no? said Q 'W hat language i John w hy said to T om [ that Mary does not understand t| ] ?' Long-distance scram bling in the (b)-exam ples fails to suspend the effects o f anti- superiority. 2 .3 .3 . A Suspension Effect The anti-superiority effect and the W C O phenom ena show another interesting parallelism as to how they are im proved. H om stein (1994, Ch. 6) points out that a W CO effect is slightly but decidedly w eakened by a "licit" bound pronoun in terms o f the W CO constraint. Let me reconstruct his observation in Japanese as in (26): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2 6 ). a *?fSoitsu i -no hahaova-ga dare i -ni seeseebunpoo -o he -Gen m other-Nom w ho-D at generative grammar-Acc hihanshita no ka] wakaranai. criticized Q know -not '*1 don't know [his i m other to w h o i G enerative Grammar criticized].” b. ??rS o itsu i-n o hahaova-ga dare 1 -ni soitsu i -no kenkvuu -o his m other-N om w h o-D at his research-Acc hihanshita no ka] wakaranai. criticized Q 'His i mother to w ho i his i research criticized.' Interestingly, the acceptability o f (26b) slightly im proves in spite o f the fact that the higher pronoun soitsu is outside the c-com m and domain o f its quantified antecedent. This "incom plete” rem edy o f a weak crossover violation is caused by the existen ce o f the other bound pronoun (the lower soitsu ) within the c-com m and dom ain o f its antecedent. Lacking this rescuing bound pronoun, (26a) is strictly ruled out as a w eak crossover v io la tio n .1 9 G iven this fact, the W CO analysis im plies that an anti-superiority violation should be made to be less deviant in a sim ilar fashion. T o exam ine whether this is the case, consider the exam ples below : 1 9 T his suspension fact along w ith som e others lead H ornstein (1994, Ch. 6) to form ulate the w eak crossover constraint in (i), which I w ill adopt as an LF- interface condition: (i) If a pronoun P is linked to a variable V. then V must c-com m and P. (= (6 ), H ornstein (1 9 9 4 , Ch. 6)) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 (2 7 ). a.*'?Naze darei-g a M ary-o [ T om -no uchi ]-ni tsureteitta no? w hy w ho-N om -A cc -Gen hom e-to took Q '*W hy w ho took Mary to Tom 's place?' b. ??Naze dare i - ga M ary-o [ soitsu i-n o uchi ] -ni tsureteitta no ? why w h o-N om -A cc he -Gen hom e -to took Q 'W hy w h o] took Mary to [ h isi ow n place ]?' Although the judgm ents are subtle, (27b) appears to me decidedly better than (27a). O bviously the im provem ent o f the latter case is induced by supplying a bound pronoun soitsu just like in (26b). This fact is readily understood if the anti superiority effect is subsum ed under the general weak crossover phenom ena. 2.4. Extending W eak C rossover T his section w ill introduce technical assumptions to characterize the anti- superiority effect as an instance o f w eak crossover. I shall base m y analysis on Hornstein's (1994, Ch. 7 ) treatment o f superiority effects in E nglish, w hich itself is an extension o f Chierchia's (1991, 1992-1993) account o f w h / OP interactions. I w ill first review Chierchia (1991, 1992-1993), and then turn to Hornstein (1994). Since I w ill make som e revisions necessary for m y argument o f the anti-superiority effect, w hat w ill follow should not be considered as exact sum m ations o f their analyses. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 .4 .1 . C hierchia (1991, 1992-1993) and Som e R evisions^ ’ 30 C hierchia (1 9 9 1 , 1992-1993) proposes to subsum e under the general W C O phenom ena w ell known interpretive asym m etries exten sively discussed in the literature (e.g. M ay (1985)). C onsider (28), w hich sh ow s an interesting contrast regarding what answers the w h-questions are able to support: (28) a. W h o i does every bov like t[? b. W h o i t| likes every b o v ? O bviously, this definition entails that a grammatical relation o f "antecedence" is to be represented by H igginbotham 's (1983, 1985) "linking" rather than by a custom ary indexing m echanism . G iven (i), consider the abstract LF- representations in (iia) and (iib), corresponding to (26a) and (26b), respectively: (ii) a.?*[ ... pronoun ... [ ... variable... ]] pronoun ... [ ... van^ti b.??[ ... pronoun ... [ ... variable ... [ ... pronoun ... ]]] R oughly, "variable" stands for dare in (26). (iia) clearly violates (i), since the pronoun is linked to the variable, that does not c-com m and it. (ii), on the other hand, conform s to (i); T he low er pronoun is linked to the c-com m anding variable. N ote that the higher pronoun is linked to the low er pronoun, not the variable. This relation is an instance o f "backward pronominalization," rather than bound variable anaphora, for w hich (i) is irrelevant. Though it is indifferent to (i), this “backward pronom inalization” itself renders the sentence far from perfect as is usually the case for such anaphora. In what follow s, the use o f indices to represent bound variable anaphora shall be just for the sake o f expediency. :u This subsection w ill rely on Hornstein (1994, Ch. 6), w hich presents a lucid sum m ary o f C hierchia (1 9 9 1 , 1992-1993). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 T hese exam ples crucially differ in that (28a) admits all the three types o f answers in (29), w hereas (28b) can only support an answ er o f the type in (29c): (29) ( ( 2 ) + (1 7 ), Chierchia (1 9 9 3 )) a. H is | m other. ( Functional A nsw er ) b. B ill lik es Sm ith, Sue Jones, ... ( Pair List A nsw er ) c. P rofessor Sm ith. ( Individual A nsw er ) (29a) is a f(unctional)-answ er,2 1 w hich associates to each person (belonging to the dom ain o f individuals associated w ith every b o v ) his i mother. T he answ er in (29b) and that in (29c) are called a pair-list ansv/er and an i(ndividual)-answer. Chierchia's account o f this contrast begins by asking w hy (28a), not (28b ), can support functional-interpretations. F ollow in g Ham blin (1958, 1973) and esp ecially Karttunen (1977), he assum es that a question denotes a set o f true propositions that jointly constitute a com plete answer to the question.2 2 On this view (28a), under its functional reading, is mapped to the rudimentary logical form in (30). obtained by custom ary com positional interpretation from a certain appropriate LF-representation: 2 1 T o m y know ledge, the f-answ er w as first discussed in Engdahl (1 9 8 6 ). 22 For recent alternatives, see G roenendijk and Stokhof (1984) and H igginbotham (1991) am ong others. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (30) A.pEf [p = AV x [B oy' (x) — > Aloves'(x, p e r so n /(x ))] In this logical form, w h o is decom posed into tw o parts; a function variable f w hose range is over persons and a singular variable x that serves as an argument o f the function denoted by f. Since x is bound to every b ov. this function associates each boy | in the dom ain o f individuals presupposed for every bov w ith a person that he | loves. To com p osition ally yield this 'functional interpretation,' w h o should have a due com plex syntactic structure consisting o f parts corresponding to the person-valued function and its argument. Slightly m odifying Chierchia’s theoretical fram ework along the lines o f Hornstein (1994, Ch. b ),2 2 but keeping his basic idea in tact, let us speculate that the LF-representation in (31) underlies (30): (31) (Sim plified) [cp[lP every_boy.2 [iP t2 love [dpPRC>2 fNP whoillin Hornstein assum es that an "implicit" argument o f a functional w h is a pro rather than a PR O . Here I take it to be a PRO sim ply to avoid a com plication in relation to a custom ary assum ption that English is not a pro-drop language. N othing essen tial is affected by this change. Incidentally, C hinque (1 9 9 0 , Ch. 3) suggests a possibility that English might contain a pro- In the m ain text, I suppress problem s that m ay w ell arise in conjunction with the so-called PR O -gate phenom ena Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 I assum e with C hom sky (1992, 1994) that m ovem ent leaves behind a full copy o f a m oved elem ent24. An in-situ copy o f w h o i is a full-fledged DP w hose spec is occupied by PR C b.2 3 Subsequently w h o i is translated as a person-valued function, and PR O i its argument Provided that this argument is correct, w e can naturally explain w hy (28b) lacks a functional-interpretation. The LF-representation in (32) is associated with the exam ple on the intended interpretation: (32) [ c p w h o i fip evervbov-> Iip rPRO? w h o i] love t^]]] 2 4 T his LF-representation is incom plete in that the CP-spec position is left em pty. H ow ever, whatever elem ent is in this position is irrelevant for the present d iscu ssion . 2 3 W hat is necessary here is to assum e an open position in the structure assigned to w h o. If H igginbotham (1985) is correct, it is possible that this position corresponds to the external argument in the 9-grid o f w h o . I leave this possibility open for the future. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This representation is a clear W C O configuration and hence excluded, as desired. Chierchia further suggests that this account is extended to the asym m etry betw een (28a) and (28b) with regard to the availability o f pair-list interpretations. F ollow ing Engdahl (1986), he assum es that such an interpretation is a special case o f a functional interpretation. The intuition behind this idea is that functions, if taken extentionally, are just sets o f ordered pairs; nam ely a pair-list.26- 2' If so, the ill- form ed representation in (32) must underlie the list interpretation o f (28b). A s a result, such an interpretation is unavailable for this sentence. T his account o f pair-list interpretations, Hornstein (1994, Ch. 6) points out, naturally explains the contrast given in (33), where only (33b) can m anage to yield a pair-list interpretation; (33) (=H orn stein’s (44)) a. W ho told everyone that B ill must shut up? b. W ho told everyone i P R O i to shut up? The LF-representations o f these exam ples are as in (34) on the given account: 2 6 S c h e in (1 9 9 3 , Ch. 11, fn. 10) points out several potential problem s o f this approach on both sem antic and syntactic grounds. I leave open for the future research how Schein's problem s are to be reconciled with the present v iew o f the pair-list interpretations. Such discussion, not directly related to the later discussion o f the anti-superiority effect, w ould be o f substantial size. 2 7 For a precise sem antic form ulation o f pair-list interpretations, see Chierchia (1 9 9 2 , pp. 2 0 9 - 213). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (34) (sim plified) a. *rP R O i w h o i told everyon ei that B ill must shut up. b. fP R O i w h o ! told everyon ei P R O i to shut up. (34a) is ruled out by the W CO constraint, since P R O i fails to be c-com m anded by everyon ei. T his W C O violation is w eakened in (34b) since the lower P R O i m eets the W CO constraint.2 8 Finally, for a wh-phrase to have a functional interpretation is necessary but not sufficient to elicit a pair-list interpretation; the argument o f such a function m ust be bound by the quantifiers that Hornstein (1 9 9 4 , Ch. 6)) dubs as generators, and uncontroversial generators are universals.2 9 28 R ecall H ornstein's (1994, Ch. 6) observation, d iscu ssed in 3.3. with Japanese exam ples, that a W CO violation is slightly but decidedly im proved by a bound pronoun c-com m anded by its quantified antecedent. Thus, the contrast in (I): (i) (based on H ornstein's (1 9 9 4 , Ch. 6) exam ple) a. *?H isi m other told every b o v i that B ill must shut up b. ?‘ ?H is i mother told every b o v ; PR O i to shut up. 29 There are divergent discussions in the literature regarding whether definite and indefinite descriptions m ay yield pair-list interpretations. The most com prehensive argument, in m y opinion, is given in S zab olcsi (1 9 9 3 ), w here she su ggests follow ing Krifka (1992) in part that they fail to elicit pair-list interpretations (at least in matrix w h-questions). This claim is evidenced by the exam ples in (i); (i) (The judgm ents are on intended pair-list readings.) a. * W hich boy did the d ogs bite? (= Szabolcsi's (14)) b. *W hich boy did m ore than tw o dogs bite? (= S zab olcsi's (14)) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 36 A function m ediates a relation from an elem ent in a domain to an entity in a range. In order to extentionally spell-out such a m apping as a pair-list, it is necessary for a dom ain o f the relevant function to be specified in a given discourse context. Therefore, only a quantifier ranging over a specific and fixed dom ain o f entities may support a pair-list interpretation. A universal quantifier, w hich presents us with such a fixed domain o f entities , can easily generate a pair-list interpretation. On the other hand, a quantifier w hich fails to specify an explicit dom ain o f a function elicits no such construal: (35) W hat did m ost / no / few men say? (V f-answ er, *pair-list answer) Take m ost men for exam ple. Being presuppositional (M ilsark (1974)), this quantifier is actually associated with a set o f individuals in a given discourse context. T his N P yet fails to function as a generator, since it is associated w ith not a single presuppositional set but rather any set o f individuals consisting o f more than half o f the m em bers o f such a presuppositional set. Due to this sem antic property, m ost m en cannot present a fixed dom ain o f individuals to be paired up with a thing that each o f them said. c. ??W hich boy did tw o dogs bite? (=Szabolcsi's (14)) For Szabolcsi, a custom ary "choice" reading associated with an indefinite description is a cum ulative interpretation (Scha (1981)) obtained betw een the indefinite and a plural-denoting wh-phrase. For details, see Szab olcsi (1 9 9 3 ). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 37 In sum . on C hierchia's account, QP / w h sentences produce pair-list interpretations only w hen the tw o conditions are met: (i) a w h is rendered as a function, and (ii) the argument o f this function is bound by a generator. . 2. 4. 2. H ornstein (1994; Ch. 7) and Som e R evisions Extending Chierchia's account o f pair-list interpretations to multiple w h - questions in E nglish. Hornstein (1994. Ch. 7) proposes that the so-called superiority effect (C hom sky (1973)) is subsum ed under the general w eak crossover phenom ena. The relevant (rudimentary) generalization is that a multiple w h-question must m ove into the C P -spec position the highest wh-phrase. Thus the follow in g contrast obtain: (36) a. W h o i ti bought w hat? b. * What i did w h o buy t j ? (36a) is a w ell-form ed sentence, since w h o , higher than w hat, is raised into the C P -spec position. T he deviance o f (36b) is yielded by raising what into the C P- spec position, low er than w h o prior to m ovem ent. First, before g oin g into Hornstein's analysis o f this contrast, let us very briefly consider som e previous studies o f the superiority effect for the later com parison with Hornstein's alternative. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 B ased on the notion o f ECP, a w ell-know n account o f this phenom enon is given by C hom sky (1986b ) and others. Supposing that the wh-phrase in a CP- spec position at Spell-O ut is adjoined by an in-situ w h-phrase at LF. (36a) and (36b) are m apped to (37a) and (37b), respectively, at LF: (37) a. [cp [what2 [who]]] [jp tj bought to ]] b. *[cp [whO|[what2]] [ip tj bought t2]] N ot being lexically governed, the subject trace ti must be antecedent governed to satisfy the ECP. A ssum ing that a wh-phrase antecedent-governs its trace only if it m oves into a vacant C P -sp ec,'0 (37a) but not (37b ), can be ruled in as desired. A s w id ely noted, this ECP-based account, with a crucial reference to the notion o f "subject," runs afoul o f handling a "pure" superiority case (H endrick and R ochem ont (1 9 8 2 )) such as (38): (38) a. W h o i did you persuade ti to buy what ? b. * W hat | did you persuade w ho to buy ti? O bviously these exam ples do not involve a subject trace, and therefore the deviance o f (38b) cannot be characterized as an ECP violation .'1 '° A s is w ell-k n ow n , it has been argued that the C om p-indexing procedure (c, H ornstein and Sportiche (1981)) yields this result. A m ong the authors w hich adopt this rule are Lasnik and Saito (1984) and A oun, H ornstein, Lightfoot and W einberg (1987). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 Chom sky (1 9 9 2 , 21), very roughly though, suggests a possible alternative explanation in terms o f his econom y considerations. Suppose that the com putational system associates to a sentence the most econom ical convergent derivation, and that a longer m ovem ent is more costly than a shorter alternative.'2 On this view , (38b), for instance, is ruled out since the exam ple involves the m ovem ent o f w hat, w hich is longer than the possible alternative, that is, the m ovem ent o f w h o . H ow ever this explanation, too, faces an em pirical problem . In sp ecifics, it fails to capture an important aspect o f superiority effect as observed in the contrast in (39): (39) ( cited from Saito (1992b)) a.?*W hat b o o k si do you expect w h o to give ti to Bill ? b. What b o o k s| do you expect w ho to give t| to w h om ? This contrast illustrates that a superiority violation is rem edied (or at least, w eakened) by another occurrence o f a wh-phrase in a position low er than the original trace o f a m oved w h-phrase.” Since his econom y condition requires o f the uppermost wh-phrase to m ove in a multiple w h-question. this additional w h - effect is totally unexpected on Chom sky's account. ■ " For m uch relevant discussion, see P esetsky (1987). ” I like the reader to refer to C ollins (1994) for a hierarchical definition o f the length o f m ovem ent. ” See K ayne (1 9 8 3 ) for a detailed discussion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 40 Hornstein (1994, Ch. 7) naturally explains the empirical problem s that these previous studies face. A s the reader m ay have surm ised from Chierchia's proposal in the last subsection, he suggests that the superiority violation in (36b ), for exam ple, is rendered as an instance o f W C O by associating to the sentence the LF- representation in (40): (4 0 )* [c p w h ati [ip [n p PR O , w h o i buy w h ati ]] The fundamental assum ption here is that exactly as in the cases o f w h-questions w ith quantifiers, functional w h-phrases w h ose arguments are bound by generators yield pair-list interpretations associated with multiple w h-questions. In particular, (40) renders w h o as a person-valued function w hose argument is bound to the generator w hat. Thus, the pair-list interpretation, inform ally given in (4 1 ), is what this representation yields: (41) Xxe things A./person [ / ( * ) bought x ] O bviously the LF-representation in (40) is ruled out by the W C O constraint as desired, since P R O i fails to be c-com m anded by the copy o f w h at), a variable. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 41 N ow the question is w hy w hat, not w h o , must be a generator in (40). He speculates that an inherently non-d(iscourse)-linked wh-phrase (P esetsky (1987)) such as w h o and what qualifies as a generator, only if it is in a C P -specifier p o sition .3 4 Recall from the last subsection that a distinctive property o f a universal quantifier, the best supporter o f pair-list interpretations, is that in a given discourse context, it is associated with a specific set o f individuals in terms o f w h ich w e generate a pair-list. W ith this in mind, consider the follow ing sentence: (42) W ho bought what ? Hornstein points out that there is a striking semantic sim ilarities betw een w h o , in [Spec, C P ], and universal quantifiers,35and that these very properties m ake w h o , or an inherently non-d-linked w h-phrase in [Spec. CP] in general, a generator. 34 In contrast, an inherently d-linked wh-phrase o f the form w hich N' does not need to m ove into [Spec, CP] in order to be rendered as a generator. This assum ption is crucial for Hornstein in explaining the grammaticality o f a w ell- known exam ple like ( i ) : (i) W hich book did w hich person buy? T his sentence illustrates that inherently d-linked wh-phrases do not sh o w the custom ary superiority effect. On Hornstein's account, this sentence is acceptable, since the follow in g w ell-form ed LF-representation is associated with it: (ii) [cp w hich book [|p w hich person i buy [PROi which book ]]] B eing inherently d-linked, w hich person can becom e a generator w ithout m oving into the C P -spec. Taking w hich book as a function o f w hich person, this sentence does not violate the w eak crossover constraint. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. First, being strong, a universal quantifier ranges over a presuppositional set o f individuals (M ilsark (1974) and D iesing (1992) am ong others). Sim ilarly the felicity o f (42) seem s to require for w h o a presuppositional set o f people o f w hich the speaker asks what each o f the m em bers bought. In other w ords, w h o is d- linked. Lack o f such pragmatic know ledge makes (42) very awkward. On the other hand, there is no need to presuppose such a set for what; even total ignorance regarding the references o f what is still felicitous with (42). S econ d , in addition to this presuppositionality (or, d-linkedness), w ho exhibits "maximality," again a characteristic property o f universals. Thus, the felicity o f (42) requires o f all the m em bers in the set associated with w h o to be mapped to a thing that he (= w h o ) bought. A s for what, although nothing seem s to keep it from being d-linked, such a property o f m axim ality is not attributed to it. H ornstein concludes that being d-linked and m axim al, an inherently non-d-linked w h-phrase in [Spec, CP] qualifies as a generator, whereas those in-situ cannot.3 6 3 5 Scrutinizing syntactic and sem antic properties o f m ultiple w h -q u estion s. K iss (1 9 9 3 ) independently proposes to analyze w ide-scope w h-phrases o f such questions as universal quantifiers. 3 6 T his argument, as it stands, is not consonant with the M inim alist theory. G iven that w h-m ovem ent is solely m otivated by checking o f a m orphological feature o f som e sort, this sem antic disparity betw een a dislocated M l and an in-situ w h is a total mystery. Take (i) as an LF-representation o f (42): (i) [Cp w h o [ip w h o bought what ]] K eep to that a displaced elem ent is interpreted in its original position at LF and that w h o leaves behind its exact cop y . H ence there should not be any difference betw een w h o , a copy, and what at LF. This may w ell indicate that a "copy" is not just a cop y and that m ovem ent itself does have som e sem antic import. Unfortunately I have nothing interesting to say about this issue. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43 This W C O analysis, since it has no reference to the notion o f "subject," can readily be extended to Hendrick and Rochem ont's "pure” superiority cases. On the account, (38a) and (38b) are m apped to (43a) and (43b) at LF, respectively : (43) a. [cp w h o i did [jp you persuade w h o i to buy [PRO] what 1 1 1 b. *[cp w h a t; did [ip you persuade [PRO ) w h o ] to buy w h at11 1 In (43a) w h o , a generator, binds the argument (i.e. P R Q i) o f the thing-valued function introduced by w hat. O bviously this bound variable anaphora obeys the W C O constraint. On the other hand (43b) renders what as a generator, w hile w h o as a function w h ose argument is bound to w h at|. The low er w hatj, a variable, not c-com m anding PR Q i. this representation is ruled out by the W C O constraint. T his W C O -based analysis o f the superiority effect explains the low er w h-effect by associating to (39b) the follow ing LF-representation: (44) [cp what b o o k si do [ip you exp ect [PRO] w h o ] to g ive w hat b o o k ) to [PRO] w hom ]]] T his LF-representation renders the low er wh-phrase w hom as a function o f what b o o k s alongside o f w h o . Even though the upper P R O ) fails to be c-com m anded by the cop y o f what b ook i (=a variable), a W C O violation is attenuated here since there is a licit bound pronoun, viz. the low er P R Q i within the c-com m and dom ain o f the variable. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 44 2.5. Subsum ing the Anti-superiority Effect under the W C O -phenom ena C om bining Chierchia (1991, 1992-1993) and H ornstein (1 9 9 4 ), a unified treatment is possible for tw o apparently distinct phenom ena; (i) pair-list interpretations o f w h-questions with quantifiers, and (ii) the superiority effect in English. O ne o f the key assum ptions in their analyses is that pair-list interpretations are elicited by rendering a wh-phrase as a function and letting its argument be bound to a quantifier ranging over a definite set o f individuals. G iven this and others, (i) and (ii) are successfully explained in terms o f the W C O constraint, that is independently m otivated on solid empirical grounds. T his section, as I hope to have em pirically justified in section 2.3, w ill com e back to the proposal that the anti-superiority effect is subsum ed under the general W C O phenom ena. The idea w ill be developed along the lines o f Chierchia (1991, 1992-1993) and H ornstein (1994). This extension, if correct, lends strong em pirical support to their analyses. 2 .5 .1 . C anonical Contrast Let us begin our discussion by considering a by-now fam iliar exam ple: (45) *?N aze dare-ga Amerika-ni itta no? w hy w h o-N om -to w ent Q '?*W ho w hy w ent to Am erica.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 45 A multiple w h-question is m arginal, when naze 'w hy' c-com m ands another w h- phrase in the sentence. R ecall that section 2. 3 suggested a possibility where this anti-superiority effect is explained in term o f the W CO constraint, assum ing that naze contains a variable bound to another wh-phrase. T his view , com bined with Chierchia (1991, 1992-1993) and Hornstein (1 9 9 4 ), naturally leads us to assum e that a multiple w h-question renders naze as a function o f another wh-phrase. G iven this is indeed the case, I propose that (45b) is mapped to (46) at LF: (46) [cp tip [a d v p PR O , nazel [n> darei-ga Amerika-ni itta]] no] This LF-representation is excluded as a W CO violation, as desired;17- 1 8 PR O , the argument o f the reason-valued function denoted by naze, fails to be c- com m anded by dare.1 9 1 7 I tentatively assum e a custom ary "first-branch definition" o f c-com m and (Reinhart (19 8 3 )) w ithout the segm ent / category distinction (C hom sky (1986b)). 1 8 If w e assum e, as noted earlier, that naze is "base-generated" in a VP-adjoined position, (45) raises a question regarding "reconstruction effects." That is to say, it ought to be plausible under this particular assum ption that (45) is mapped to the LF-representation in (i) along with that in (46): (>) fcptip darei-ga [vp [a d v p PRO) nazelIvp Am erika-ni itta]]] no] In (i). naze, scram bled before Spell-O ut, is “reconstructed” to the V P-adjoined position, and it is rendered as a function o f dare in this position. For the PRO is c- com m anded by dare, this LF-representation does not violate the W CO constraint. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 46 N ow the question is w hy it is that in (46) naze is construed as a function, w hereas dare as a generator. N o longer could w e maintain our explanation o f the Reconstruction effects, how ever, obtain only marginally for “reason" expressions. Thus, in (ii), it is rather difficult for soitsu ‘h e,’ contained in the sentence initial clause that denotes a reason, to be bound to subete-no otoko ‘every man:’ (ii) ?*(Soitsu i -no kodom o -ga bvooki na node] subete -no otoko l-ga he -G en child-N om sick is because every-G en m an-N om yakkyoku -e itta (koto) drug store-to went 'B ecause hds i child is sick, every m am went to a drug store.' T his observation appears to have m uch to do with the fact that adjuncts in general are hard to scramble: (iii) a. :|:?N aze | John-wa [q M ary-ga yakkyoku -e itta to] itta no? w hy -Top -N om drug store-to went Com p said Q ‘W h v i John said [that Mary w ent to a drug store ti]’ b.*?[| kodom o-ga byooki na node] John-wa [t) M ary-ga child -N om sick is because -Top -N om yakkyoku-e itta to] itta. drug store-to went C om p said ‘[i B ecause her child w as sick ] John said [that Mary w ent to a drug store t,]’ If w e can generalize that adjuncts cannot be scrambled, it im m ediately follow s that they do not exhibit reconstruction effects. O f course, a question rem ains what keeps adjuncts from being scram bled. I leave this question for future discussion. 1 9 (46) assum es that in-situ w h-phrases are interpreted in their original positions. S ince this question ends in no(-desu), to interpret dare in-situ is not problem atic. Should it be true that naze m oves to a C P-spec position even in n o(-d esu ). naze in (46) can be regarded as a "copy" left behind by naze. I w ill not g o into details on this matter due to its irrelevance to the present discussion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 47 ungram m atically o f (45), if the reverse were possible. To present a possible answ er, let us ask the question from a different point o f view ; W hy should it be that naze cannot be interpreted as a generator? O ne possibility that im m ediately com es to my mind is related to a certain sem antic / pragmatic property o f naze. R ecall that on Hornstein's account, a generator wh-phrase must be d-linked in that it ranges over a set o f individuals know n both to the addresser and the respondent(s) o f the relevant question. A pair- list answ er is assum ed to be constructed based on this set. W ith this in mind, consider the follow ing sentence: (47) * A to B to C to - yuu riyuu no uchi-de anata-wa naze and and and say reason-G en am ong you-T op why M ary-o kiratteiru no ? -A cc dislike Q '*A m ong the reasons A , B and C, w hy do you hate Mary ?' The unacceptability o f this exam ple seem s to illustrate that naze cannot be d-linked, w hich, if true, keeps it from being a generator. M y judgm ent is very clear, w hen the sentence is presented in a context such as the follow ing for instance; There are tw o persons, say Hanako and Y osh ik o. Three statem ents A, B and C are given on a blackboard as possible reasons for Hanako's hatred against M ary. M oreover Y osh ik o know s that Hanako hates M ary for one o f those reasons, but she d oes not know w hich. Then Y oshiko utters (47) to H anako. A d-linked interpretation is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48 clearly intended here and this situation renders the use o f naze quite awkward. Failing to be d-linked, naze cannot be a generator o f a pair-list interpretation. This sem antic / pragmatic behavior o f naze, in turn, forces dare to be construed as a generator in (45). T his assum ption seem s unproblem atic in terms o f'd - linkability;' (48) A to B to C no naka-de anata-wa dare-ga ichiban kiraidesu and and -G en am ong you -T op w ho-N om m ost hate k a ? Q ' A m on g A, B and C, w h o do you hate m ost ? 1 A d-linked interpretation is intended and actually forced on dare in this context. U nlike (47). this sentence is fully acceptable. H ence dare is d-linkable. Capable o f being d-linked, dare can serve as a generator o f a pair-list interpretation in (4 6 ).4 1 ’ 4,1 A problem seem s to arise regarding a sentence like ( i ) : (i) ittai dare -ga naze Am erika -ni itta n o ? the-hell w ho-N om w hy A m erica-to w ent Q 'W ho the hell w hy w ent to A m erica ?' T o m y know ledge, it has been argued since Takubo (1985) that ittai. taken as a Japanese equivalent o f the hell, forces a w h-phrase to be non-d-linked. If this is correct, the w ell-form edness o f (i) illustrates that a non-d-linked wh-phrase ittai dare ’the hell w h o ’ can serve as a generator o f a pair-list interpretation, contra the present argument. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 T he discussion thus far has tacitly assum ed that a multiple w h-question with naze is to be responded with a pair-list answer. It is o f course possible that a question o f the sort solicits just a "single-pair" answer instead o f a pair-list answer, depending upon what discourse context it is a part o f .41 A com m ent is in order regarding a w h-question with naze in such a context in relation to the above argum ent o f d-linking. C onsider (49): (49) Speaker A: Aru otoko-ga tandoku-de aru shiteki-na riyuu-kara H irom i-o a m an-N om alone a personal reason-from -A cc satsugaishim ashita. murdered 'A certain man killed H irom i for a certain personal reason.’ E ven though the semantic / pragmatic properties o f ittai is far from clear, it is dubious at best that the expression actually has a function o f ridding a wh-phrase o f a d-linked interpretation. Indeed ittai can readily be attached to a typically d-linked w h -phrase o f the form dono N' (N ishigauchi (1990, 38)): (ii) K ono guruupu kara -w a ittai dono gakusee -ga kaigi -ni this group from -Top the hell w hich student-Nom m eeting-to shusseki-suru no ? attend Q ' *From this group, w hich student the hell w ill attend the m eeting ? ' A related discussion w ill appear in 5.7. 41 N eed less to say, it is plausible that a single-pair constitutes a pair-list that is "singleton." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 Speaker B: G utaiteki-ni, |(i) dare- g a naze / (ii) *naze dare-ga } H irom i-o concretely { (i) w ho-N om w hy / (ii) w hy w h o-N om ) -A cc koroshita no desu ka? O shie-te kudasai. killed Q tell please 'In particular, ( (i) W ho w hy / (ii) * w hv w h o 1 killed Hiromi? Please tell m e.’ Speaker A: K obayashi-to-yuu-otoko-ga renai kankee-no motsure kara -by-the-nam e-of love relation-Gen trouble from kanojyo-o satsugaishita no desu. she -A cc lolled 'A man by the name o f Kobayashi killed her due to a trouble in his love relationship w ith her.' Im m ediately after the first utterance o f speaker A , information is shared betw een the tw o speakers in w hich there is a unique man w ho murdered Hirom i for som e personal reason. Speaker B then utters a w h-question asking for the identity o f the killer and the m otivation for the killing. N ow these m ultiple w h-questions. as indicated, exhibit the anti-superiority effect. Exhausting the logical possibilities, (i) and (ii) are associated w ith the LF- representations in (50) and (51), respectively: (50) (=(i), Irrelevant details are om itted.) a. [c p [ c [IP darezgai [Vp [PRO i naze] ... ]] ka]] b .?*[cP [ c [IP [P R O i dare-gal [yp ... nazei ... ]] ka]] (51) (= (ii)) a .* ? [cp [ c ElP [PR O i naze] [ip dare-ga | [Vp... ]] ka]] b .* ? [cp t c ElP nazei [IP [P R O i dare -ga] [Vp... ]] ka]] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 (50b) and (51a) are excluded by the W CO constraint.42 Setting aside (51b) for a m om ent, let us focus on the representation in (50a). Despite that dare is construed as a generator,43 this wh-phrase does not appear to be d-linked in the sense custom arily assum ed for the notion; A wh-phrase is d-linked only if its value(s) is(are) to be assigned (or, chosen) from a set such that the questioner and the respondent both know what individuals this set com prises. Regarding dare in (5 0 ), w e may claim that the discourse context presents the wh-phrase with a unit set w hose m em ber is the unique killer. H ow ever this wh-phrase is still non-d-linked, since his identity should be unknown to the questioner, or asking this question w ould possibly be just pointless. If the questioner knew that John, say, is the killer, he w ou ld rather ask "John-wa naze H irom i-o koroshita no desu ka?” (=W hy did John murder Hirom i?) instead o f the question in (49): A question im m ediately arises as to what excludes in this context the representation in (5 lb ), including the bound variable anaphora in (52): nazej 42 B esid es, (50b ) renders naze as a generator. W e w ish to exclude this option. 43 Since the question under discussion solicits just a single pair answer, one m ight feel it inadequate to call dare as a generator (o f a pair-list interpretation). A t this point, I w ould like to avoid this problem sim ply by assum ing that a single pair is a special case o f pair-list on w hich there happens to be only one pair. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 52 Since the multiple w h-question under discussion solicits a single-pair answer, w e may not sim ply say that naze, not being d-linked, cannot be a generator. T his is because dare in (50a) can be a generator in the very context under discussion even if it is not d-linked . I suggest, basically follow in g Kuno and Takami (1 9 9 3 , Ch. 3), that the oddity o f a relation o f the sort in (52) is due to a certain "functional" factor outside the realms o f form al sem antics and syntax. For the sake o f expediency, let me first speculate that a bound variable anaphora in general "expresses," som ew hat loosely, a cognitive act (or Kurodaian (K uroda, 1974, 1992) "judgment") o f, say, association in o n e :s mind. C onsider the exam ple in (53): (53) Every b o v i loves [hisi mother] For every bov binds h]s, w e may think o f his mother as a mother-valued function from the set that every bov ranges over. This linguistic relation o f bound variable anaphora, as it seem s natural to assum e, expresses a cognitive act in the speaker's mind o f the follow in g form: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (54) boyi boy2 boy 3 boy 4 boy 5 mother) m oth en mother mother4 53 Here in this cognitive act o f association, each boy in the given universe o f discourse is m apped to his ow n mother. In this light, consider the exam ple in (55) with its associated LF-representations in (56): (55) (based on Kuno and Takam i's (1 2 2 )) *?Naze dare -ga paatii-ni kita no? w hy w h o-N om party-to cam e Q ' W hy w h o cam e to the party?' (56) (sim plified) a. * ? [ c p t c [IP [PR O i nazel [vp d arei-ga] ... ]]]] b . . * ? [ c P t c [lP f l i ^ l [v p [P R O i dare-ga] ... ]]]] (56a) is trivially ruled out by the W C O constraint. (56b) involves a m apping that w e w ish to exclu d e independent o f the notion o f d-linking so as to accom m odate the anti-superiority effect in a multiple question with naze that purports to be replied with a sin gle pair-answer. The bound variable anaphora betw een naze and PRO inside dare expresses the follow ing schem atic association: (57) A B reason i _ _ reason 2 - reason 3 - reason 4 person i p erso n 2 p erso n 3 (A and B may be unit sets.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 54 Since w e are dealing with a multiple wh-question. this cogn itive act is not fully carried out in the questioner's mind; If he had this full know ledge about the given universe o f discourse, his utterance o f (55) w ould be just pointless. On the other hand, (56b) "induces" this cognitive act in the respondent's mind, and he spells it out as a pair-list answ er as in (58): (58) (K uno and Takam i's (125b)) *?G otisoo -o tabetakute Taroo to Ziroo-ga, Marv-ni aitakute Hanako-ga, feast-A cc want-to-eat and -N om -Dat w ant-to-m eet -N om watashi-to hanashi-o shitakute Taroo-to K yooko-ga paatii-ni kita. -w ith want-to-talk -with -N om party-to cam e ' B ecause Taro and Ziro wanted to eat the feast, they cam e; because Hanako wanted see M ary, she cam e; and because Taro and K voko wanted to talk with m e. they cam e.' In this pair-list answer, the relevant pieces o f inform ation, so to speak, are organized with the reasons w hich are the values assigned to naze. T his answ er sounds very unnatural, if not im possible, since w hile in our universe o f discourse w e are accustom ed to mapping party-comers to the reasons they cam e, it is not custom ary to do the opposite association. The LF-representation in (56), I assum e, is deviant, since it expresses a cognitive act like (57) w hich does not fit in w ell with our pragmatic know ledge o f the world. B efore leaving this subsection, I note that the foregoing discussion based on our pragmatic know ledge o f the world Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with 55 is not to deny m y first suggestion regarding w hy naze fails to he a generator; Naze cannot be a generator, for a d-linked interpretation is not possible for this w h- phrase. This argument, I b elieve, is still valid when a multiple w h-question with naze is answ ered w ith a pair-list, not single-pair, answer. Due to its apparently w ider em pirical coverage, how ever, the discussion henceforward w ill refer to the pragmatic based explanation. T he remainder o f this section w ill turn to em pirical consequence o f this analysis. O ne appeal o f the present account o f the anti-superiority effect is that assum ing naze cannot be a generator, it enables us to uniform ly explain ihe anti-superiority effect in Japanese and the superiority effect in English; they are both considered as instances o f w eak crossover violations. T o show that this is not an unreascnable direction, the subsequent subsections w ill indeed present several striking sim ilarities am ong the tw o sets o f phenom ena and the general w eak crossover phenom ena. 2 .5 .2 . On L ow er W h-effects and Related Phenom ena T he w eak crossover analysis sheds new light on the low er w h -effects d iscu ssed in Saito (1992b ). Consider the exam ples in (59); (59) a. ( Saito's (2b )) * N aze dare -ga soko-ni itta no ? w hy w ho-N om there-to went Q 'Why w ho w ent there?' permission of the copyright owner. Further reproduction prohibited without permission. 56 b. ( Saito's (3b )) ?Naze dare-ga doko -ni itta n o ? w hy w ho-N om w here-to went Q 'W hy w h o w ent where?' The acceptability o f (59b) illustrates that an anti-superiority violation is suspended by a w h-phrase (doko 'where' in this case) added in a position low er than an 'offending' naze-w h sequence. The present W C O analysis ties this fact with a similar phenom enon concerning superiority effects in English. Consider the exam ples in (60), repeated from Subsection 4.2: (60) (cited from Saito ( 1992b)) a.?*W hat b o o k si do you expect w h o to give tj to Bill? b. W hat books i do you expect w h o to give ti to w hom ? The superiority effect in (60a) is circum vented by replacing to Bill w ith to w h om in (60b). H ornstein accounts for this contrast by associating (60a,b) w ith (61a,b), respectively, at LF: (61) (sim plified) a.?*what b o o k si do you expect [ PR O i w h o ] to give t| to Bill b. w hat b o o k s i do you expect [ PRO] w h o ] to give ti to [PRO i w h om ] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 In Hornstein (1 994). (the copy of) an inherently non-d-linked wh-phrase must be functionally rendered unless it is m oved into the specifier o f C P .44 (61a) is thus ruled out as an instance o f w eak crossover, as desired. On the other hand, in (61b). the occurrence o f w hom (precisely. P R O i in the function) rem edies this w eak crossover violation. Hornstein's account totally assim ilates the contrast in ( 6 1 ) to that in (6 2 ): (62) ( based on (6 1 ), Hornstein (1994)) a. * His mother i gave every boy i a book. b. H is m otheri gave every b o v i his b o o k i. Even though the first occurrence o f his i is outside o f the c-com m and dom ain o f everv b o v i. (62b ), unlike (62a), is saved by a "rescuing bound pronoun" ( the second his i.). A n alogou sly, the low er P R O i cancels a potential W C O violation in ( 6 lb ). This argument is readily extended to the contrast in (59). (59a) and (59b) are m apped to the (rough) LF-representations in (63a) and (63b), respectively: (63) ( sim p lifie d ) a. * LPRO| nazel darei -ga soko-ni itta no 44 Recall that this generalization is based on his observation that an inherently d- linked wh-phrase can be d-linked only w hen it m oves into a C P-spec position. Although I w ill not go into details, this generalization needs a qualification, w hen a question like (61b) solicits a single-pair answ er, not a pair-list answer. In such a context, what books in (61b) is not d-linked in the custom ary sense. T he evasion from discussing this matter w ill not be crucial in w hat follow s. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 b. [P R O | naze] darei -ga \ PROi doko 1 -ni itta no F ollow in g the 'guideline' from the last subsection, (63a) renders dare as a generator binding the argument o f the function denoted by naze. C onsequently, the representation violates the weak crossover constraint as desired. (63b), on the other hand, is saved due to another occurrence o f functional wh-phrase (v iz. [ P R O | doko ] ) with a "licit" bound pronoun in it. The W C O approach to the anti-superiority effect also explains a contrast from Section 2.3. repeated in (64): (64) a. *N aze dare i-g a M ary-o [Tom -no uchi ]-ni tsureteitta no? w hy w h o-N om -A cc -Gen hom e-to took Q ' w ho took Mary to T om ’s place?' b. N aze dare i -ga M ary-o [ soitsu i -no uchi ] -ni tsureteitta no? w hy w h o-N om -A cc se lf -G en hom e -to took Q 'W hy w h o i took Mary to [ h isl ow n place ]?' The anti-superiority effect is suspended by a bound pronoun soitsu in (64b ). This contrast is rather expected on our analysis, since the LF-representations o f the exam ples are as in (65): (65) (Sim plified) a. *[ P R O i naze ] darel-g a M ary-o [ T om -no uchi ]-ni tsurete itta no ? b. [ P R O i naze 1 darei- ga M ary-o [ so itsu i-n o uchi ] -ni tsureteitta no ? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 (65b) is w ell-form ed , since a weak crossover effect is suspended by soitsu bound to dare i . The W CO analysis is further supported by another similarity betw een the w eak crossover phenom ena and the (anti-)superiority phenom ena regarding how they are attenuated. C onsider the sentences in (6 6 ) - (69): (66) * H is | mother said [ that every student i hated his l advisor ] (H ornstein (1994)) (67) *W h o i did h is) mother say [ ti hated h isi b oss ] (68) * [soitsu l -n o hahaoya]-ga [ subeteno gakusee ) -ga soitsu | -no he-G en m other -N om every student-N om he -Gen adobaizaa -o kiratteiru to ] itta. advisor -A cc hate Com p said ’*H is) m other said that every student) hated h isl advisor.’ (69) *[ Soitsu i no hahaoyal -ga fdono gakusee i-g a \soitsu i no adobaizaa]- his -G en m other-N om w hich student-N om he -G en advisor -o kiratteiru to ] itta no -A cc dislike C om p said Q '*W hich student) did h is) mother say [ t) disliked h isl advisor]' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 60 W e w itness that the suspension o f a W C O violation under discussion is "clause- bound." Thus (6 8 ) is ill-form ed despite the occurrence o f the rescuing bound pronoun (the low er soitsu ‘h e’). Setting aside the w h-questions in (67) and (69) for a m om ent, this appears to be an expected lim itation since a finite clause is a scope island; a bound pronoun must be in the scope dom ain o f a quantifier to w hich it is bound (H eim (1982)). (6 8 ) ( ( 6 6 ) as w e l l ) is then excluded, sim ply because the higher soitsu . an intended bound pronoun o f subeteno gakusee. is not w ithin the quantifier's scope dom ain, viz. the em bedded clause. From this perspective, the ill-form edness o f the w h-questions in (67) and (69) needs an additional explanation. Take (69) for exam ple. This sentence is ungram m atical, even though the higher soitsu ‘he’ apparently falls under the "scope" o f dono gakusee ‘w hich student.’ N otice that what is meant by "scope" here is actually "scope o f question." T his should be at least intuitively clear, since this sentence is a matrix (as opposed to em bedded) constituent question. G iven the account o f ( 6 6 ) and (6 8 ) above, I take the ill-form edness o f this exam ple to suggest that a w h -phrase is a m ulti-scopal elem ent. B eing m ulti-scopal, it is associated w ith tw o distinct scope relations; a scope o f question and a scope o f a quantifier.45 T o be more concrete, I assum e that the sentence looks like (70) at LF: (70) (sim p lifie d ) * [ C P Q l [IP [ N P - h e i •••] [C P [IP w h l [V P [N P - bSl •••]]]]]] 45 Independently, som e linguists entertain this possibility including A be (1993, C h. 4 ) C u licover (1 9 9 2 ), K eiko M iyagaw a (1994a) and N ishigauchi (1 9 9 0 , 58 - 6 3 ). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission 61 T his logical form has tw o essential features; (i) the question operator in the matrix [Spec, CP] represents the scope o f question and (ii) (a copy o f ) a w h-phrase is interpreted as an ordinary quantifier in situ ( or at an adjoined position after Q R ). A s for the question operator, this may be either a remnant o f Chom sky's (1992, 58) deletion operation to m inim ize a restrictor in [Spec. CP] or just a question morphem e. The ch oice does not really affect the present argument. Regarding (ii), on the other hand, even though I w ill not go into the issue here, interesting proposals have been indeed developed along these lines and convincingly defended, (e.g. K im , 1990). N o w the origin o f the ill-form edness o f (70) (hence (6 9 )) can readily be understood as follow s; the higher h e js a variable bound to w h i, w hich d oes not contain this variable in its scope dom ain, viz. the em bedded clause. T his analysis can be straightforwardly extended to (6 7 ), which the reader w ill readily surm ise. The W C O analysis im plicates that the sam e sort o f clause boundedness holds in the low er w h -effects concerning (anti-)superiority phenom ena. Indeed, this appears to be the case. Consider (71) w ith its LF in (72): (71) *N aze John-ga [ dare -ga doko -ni itta to ] itta no w h y -N o m w h o -N o m where-to w ent C om p said Q ( ka oshiete k u d a sa i) '(Tell m e) w h y John said that [w ho went where].' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 62 (72) *[ P R O i naze ] John-w a [CP dare^-ga [ P R O i doko 1-ni itta to ] itta no The intended reading is the interpretation where naze originates in the matrix cla u se .46 If so construed, this sentence is excluded. The explanation is self- evident; in (72), the pronoun in naze fails to be in the scope dom ain o f dare. Likewise: (73) * What i did w h o say that Mary gave 11 to w hom . (74) *[ PRO l w h o ] say [ that Mary gave what [ to [ PRO i w hom ]] Rendering what as a generator, this sentence is excluded due to the fact that the functional w h o is not in the em bedded clause, which is the scope dom ain o f the generator w h-phrase. Finally let us consider exam ples from Saito (1992b), which seem problem atic for our W C O analysis (The judgm ents are Saito’s.): (75) ( = Saito's ( 100a) and ( 1 0 0 c ) ) a .* N aze dare-ga John-ni ( r p M ary-ga son o hon -o katta to ] itta no? w hy w h o-N om -to -N om that book-A cc bought C om p said Q 'W hy w h o said to John [ that M ary bought that book] ? ' 46 Another possible interpretation is that naze originates in the em bedded clause and gets scram bled to the matrix clause. On this reading, the sentence seem s to be acceptable as the analysis predicts. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 63 h .* N aze dare-ga John-ni Trp M arv-ga nani -o katta to ] itta no? w h y w h o-N om -to -N om w hat-A cc bought C om p said Q 'W hy w ho said to John [ that Mary bought what ]?' N eed less to say, (75a) is an anti-superiority violation. Saito claim s that a low er w h such as nani ‘w hat’ in (75b) cannot im prove a sentence o f this sort w hich is separated from an offending naze - M l sequence by a finite clause boundary. Our account, how ever, fails to explain this observation, sin ce the cancellation o f w eak crossover under consideration does not appear to be sensitive to such a constraint: (7 6 ) a. * S o itsu )-n o hahaoya-ga subeteno gakuseei-ni [ M ary-gaT om -ni he -G en m other-N om every student -to -N om -to L A -de atta to ] itta. in m et C om p said ':h H is i m other to every student i said [ that Mary m et T om in LA ].' b. Soitsu i-n o hahaoya-ga subeteno gakusee i -ni [ M ary-ga he -G en m other-N om every student -to -N om soitsu i-n o shidookyookan-ni LA -de atta to ] itta. he -G en advisor -to in met C om p said 'H isi m other to every student) said [ that Mary m et h isl advisor in LA A s indicated in (76b ), a low er bound pronoun soitsu significantly im proves (76a). H ence, our analysis predicts that a sim ilar im provem ent should be observed in (75). Even though (75b) does seem better than (75a) to m e, I w ould like to leave the factual evaluation open due to the subtlety o f judgm ent. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 64 2 .5 .3 . On Higher W h-effects and Related Exam ples T he w eak crosssover analysis naturally accounts for another suspension o f the anti-superiority effect. A s pointed out by Saito ( 1992b) and A. W atanabe (1991), a potential anti-superiority violation is rem edied by a "higher" w h-phrase. This is illustrated in the follow ing contrast: (77) a. *?John-w a naze nani-o tabetano? -Top w hy w hat-A cc ate Q 'John ate w h y what?' b. Dare -ga naze nani -o tabeta no? w h o-N om w hy w hat-A cc ate Q 'W ho ate w hy what?' The anti-superiority violation in (77a) is suspended in (77b) by replacing John-w a ‘Joh n -T op ’ by dare-ga ‘w h o -N o m .’ On the W CO analysis, (77a) and (77b) are respectively mapped to (78a) and (78b) at LF: (78) a. ( T he position o f the wa-phrase is not accurate, but irrelevant. ) *[C P [IP John-w a [ y p [ a D P P R O i naze ] [ y p n an it-o katta]]]no] b. [C P [IP darei-ga [ y p [a D P P R O i naze ] [ y p [N P P R O i nani 1-o katta]]]no] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 N aze being rendered as a function o f nani. (78a) is ruled out as a w eak crossover violation. On the other hand, (78b) is a legitim ate representation in terms o f the W CO constraint; the additional higher wh-phrase dare ‘who" is interpreted as a generator, whereas naze and nani are functionally interpreted. A problem seem s to arise. Saito ( 1992b, 24) claim s that the higher w h-effect is clause-bound. Consider the exam ple in (79), cited from Saito with his judgment; (79) ( Saito’s (60c)) ?*Dare -ga [John-ga naze nani -o katta to] om otteiru no? w ho -N om -N om w hy w hat-A cc bought C om p think Q 'W ho thinks that John bought w h y what.' This exam ple appears to suggest that a higher wh-phrase is not capable o f salvaging an anti-superiority sequence (naze - w h) which is not its clause mate. The W CO analysis associates (79) w ith the follow in g LF-representation on, sim plified for the illustrative purpose: (80) ( sim p lifie d ) lip darei-g a [CP [lp ...[P R O i nazelIPR O i nani ]-o ...katta... ]] ...] N ow this representation parallels the sentence in (81): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 66 (81) [C P [IP Dare i -ga [CP [IP soitsu i -no shidookyookan -ga soitsu i -no w ho -N om he -G en advisor -N om he -Gen ronbun -o kiniitta to ]] itta ] no ] article -A cc liked Com p said Comp 'W ho l said that his l advisor liked his l paper?' Since this sentence is perfectly acceptable, bound variable licensing is not clause- bound in the sense under discussion. Therefore there should be nothing deviant about the representation in (80). Indeed, it is not clear whether Saito's judgm ent concerning (79) is shared by many native speakers. The sentence seem s, at least to m e. acceptable or slightly marginal at w o r s t . W ith a clear presuppositional setting in mind that (79) should be taken as a question regarding each m em ber o f a d-linked set o f dare, this sentence appears as good as a sentence with a typical case o f the higher w h -effect as in (77b). The current consideration sheds new light on a higher w h -effect regarding superiority effects in English. C onsider the contrast in (82) from Lasnik and Saito (1 9 9 2 , Ch. 4 ): (82) a. (L& S's (6 7 )) *B ill w onders w h ati w ho? bought ti b. (L& S's (6 6 )) W ho- ^ w onders w hati w ho? bought ti Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 67 N eed less to say, (82a) is a canonical superiority configuration. Interestingly, the higher w h w h o? in (82b) appears to suspend this superiority violation w hen w hen takes scope over the matrix clause. The sentence is still ruled out if the scope o f w h en is confined to the em bedded clause. The present analysis handles this case rather straightforwardly. The LF- representation o f the tw o readings o f (82b) in question are represented at LF as in (38): (8 3 ). a. who^t w onders [CP w h ati [PR O 3 w h o ? l bought tj ] b. * w h o 3 w onders [CP w h ati [P R O i w h o - ?] bought tj] Let us keep to H om stein's assum ption that in a multiple w h-question an inherently non d-linked w h-elem ent such as w h o is rendered as a function unless it is m oved to [Spec, CP] before Spell-O ut. (83a) represents a matrix construal o f w hen; w ho') is paired up with w ho? and consequently it is rendered as a person-valued function o f w h o 3 . This LF can produce a pair-list reading o f <whc>3 , w h cn > without violating the w eak crossover constraint.47 On the other hand, (83b) corresponds to the narrow scope construal o f w h o ? . In the representation, w ho? is rendered as a function w hose argum ent is bound to what t and this bound variable anaphora yields a W CO violation. 47 As B. Schein (p.c.) suggested to m e, the fact that who? contains a variable bound to w h o i does not itself guarantee w ho?'s matrix interpretation. C onsider the exam ple in (i): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 2.5.4. A N ote on C lause-boundedness o f the Anti-superiority E ffects A. W atanabe (1991) correctly points out that naze can c-com m and another w h- phrase if their scope dom ains are distinct. This is observed in a sentence like (84): (84) N aze kim i-w a [ M ary-ga nani -o katta ka ] shitteiru no? w hy you-T op -Norn w hat-A cc bought Q know Q 'W hy do you know [ what M ary bought ]?’ W hich b o v i w onders w ho bought w hich picture o f hirm ? ' W hich boy i is such that hei w onders w ho bought which picture o f him i?' A s indicated, w hich pictures o f hirm can take scope over the indirect question despite that this N P contains a pronoun bound to w h ic h b o y i. G iven this fact, I slightly m odify (83) as follow s: (ii) Ic p w h o 3 [ c C j [ip t3 w onders [C P w h ati [PRO 3 w h o ? l bought t{ ]]]] T hough it is not entirely clear h ow the scope o f an in-situ w h-phrase is determ ined, I assum e, follow in g C .L .B aker (1970) and Pesetsky (1 9 8 7 ) that the coindexation betw een the matrix C? and w h o ? marks the root construal o f this w h-phrase. (83b ) is to be rewritten likew ise. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 69 The explanation o f this sentence appears rather obvious under our proposal. Since naze and nani are construed within separate questions, a pair-list interpretation is not generated in this exam ple. Technically, this m eans that naze is not a function o f nani in this sentence. Therefore (84) does not violate the weak crossover condition. 2.5.5. Suspension o f (Anti-)superiority Effects by Scram bling T he W CO analysis uniform ly explains another aspect o f suspension phenom ena regarding the anti-superiority effect in Japanese and the superiority effect in E nglish. Recall that an anti-superiority violation is circum vented by clause-internal scram bling. A relevant contrast is repeated in (85): (85) a. * A nata-w a naze dare -o naguttano? you-T op w hy w h o-A cc hit Q 'You w h y w ho hit ?' b. A nata-w a dare-o naze nagutta no ? 'You w h o w hy hit ?' Their LF-representations are also recapitulated (86) a. *[C P [IP anata-wa [ y p [PR O j naze ] [ y p dare-oi nagutta ]]] no ] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 b- r r p TtP an ata-w afvp dare-o] [ y p [PR O i naze ITvp 11 nagutta ]]] no ] (8 6 a) is in violation o f the W CO constraint, whereas the scram bling o f dare rem edies this anom aly in (8 6 b). A w ell-know n theoretical interpretation o f this phenom enon is to regard scram bling as an instance o f what is custom arily called 'A -m ovem ent' (e.g. Raising and Passive) G iven this, the above contrast is now assim ilated to Hornstein's superiority contrast in (87): (8 7 ) ( from H ornstein (1 9 9 4 , Ch. 7 ) ) a. *What did it seem to w hom fell o ff the shelf. b. W hat seem ed to w hom t to fall o ff the shelf. Their LF-representations are as follow s: ( 8 8 ) a. * f r p w h a t) did [ip it seem to fP R O iw h o m l [CP [ip what i fell o ff the sh e lf]]]] b. [CP w h a ti [ip what | seem to [PR O j w hom ] [ip t[ fell o ff the shelf]]] (87a), a ty p ica lsu p erio rity ca se, is reduced to weak crossover. On the other hand, this W CO violation is canceled by the A -m ovem ent o f what in (87b ). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.5.6. T he Anti-superiority Effect in English 71 The present analysis accounts for what appears to be an anti-superiority effect in English: (89) (=based on Kuno and T akam i's (97)) ?*W hv did you buy w hich b ook ? A s Kuno and Takam i (1993, Ch. 3), A . W atanabe (1991) and others point out, such multiple w h-questions as (89) are rather unnatural irrespective o f the fact that the traditional "ECP" account predicts that they are fully gram m atical.48 In view o f the pursued explanation o f the anti-superiority effect, (89) is m apped either to the LF-representation in (90a) or to the one in (90b): 48 It has been repeatedly pointed out in the literature that a multiple w h-question like (i) is absolutely unacceptable: (i) *W hich book did you buy w h y ? Though the factual claim itself seem s to need a further qualification, the deviance o f (i), if true, remains as a puzzle in the M inim alist approach. Kuno and Takam i (1993,, Ch. 3) and R izzi (1990) suggest a p ossibility w here w h y originates in a position higher than an IP. Should their claim be the case, (i) is readily excluded, for w h y m ust not occur in an IP-internal "in-situ" position. H ow ever a question still remains regarding (ii): (ii) *W hat did you fix h o w ? A lon gsid e o f (i), (ii) has been considered as a "hopeless" m ultiple w h-question. (A gain, based on m y inform ant work. I have som e reservations to adopt this acceptability judgm ent as the w ay it is.) A s far as I know , there is no evidence that how m ust not occur in an IP-internal position. I leave open this puzzle for the future research. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 (90) a. *'?fCp vvhvi [c did r1 P w h v i [|P you buy [PR O i w hich b ook llll b. *'’[CP w h y [c did [IP[PRO i w h v lll [IP you buy w hich b o o k i ]]] (90a) renders w hy as a generator, whereas w hich book as a book-valued function from the set that w hy ranges over. This LF-representation is expresses an unnatural cognitive act o f association that m aps reasons to books. Thus the (90a) elicits a pair-list answer, such as (91), where the relevant books are “sorted” w ith respect to the reasons that they were bought: (91) ??B ecause I want to know more about syntax, I bought LGB; because I w ant to know more about physics, I bought the book written by H aw king; because I want to know more about politics, I bought the book written by Rush Limbaugh. T hough (90a) is perfectly grammatical, it is pragmatically gibberish for the bound variable anaphora involved in the representation expresses such an unnatural association in the minds o f the questioner and the respondent. The representation in (90b ), n eed less to say, is ruled out as a w eak crossover violation. This anti-superiority effect in E nglish appears to behave on a par with that in Japanese as to how it is circum vented. C onsider the contrasts in (92) and (93): (92) a. (G u ess) ?*w hv w hich person m et Tom . b. (G u ess) ?w hv w hich person■ met his, new boss. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 (93) a. (G uess) ?*w h v w hich person bought a book. b. (G uess) ?w h v w hich person bought what. The marginal (a)-exam ples are cases o f the anti-superiority effect. (92a) is am eliorated in (92b) by h isi, a variable bound to w hich person. (93a) is rem edied in (93b) due to the “lower" wh-phrase w hat. Rendering w hy as a function, these sentences are mapped to the follow ing LF-representations: (9 2 ’) a. [CP w h y [|P[PR O i w hy] [Ip w hich b ook i met Tom ]]] b. [CP w h y [ip[P R O j w h v l r[P which b o o k i met [hiS| new boss]]]] (9 3 ’) a. [CP w h y [[pfPRO! w hy] [,p w hich person^ bought a book ]]] b. [CP w h y [IP[PR O i w h vl rIP which personi bought fPRCh what]]] The (a)-representations are in violation o f the W CO-constraint. On the other hand, (9 2 ’b) and (9 3 ’b) are saved by the “legitim ate” bound pronouns his and PRO , respectively. 2.5.7. A Superiority Effect in Japanese Another em pirical support o f our W CO analysis com es from what appears to be a superiority effect in Japanese. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 C onsider the exam ple in (94), w hich Pesetsky (1987) cites from Kitagawa (1984);w (94) (slightly m odified) a. [Ittai dare-ga ] B osuton-de nani-o tukamaeta no? the hell w h o-N om B oston -in w hat-A cc caught Q 'W ho the hell caught what in Boston?' b .* Dare -ga B osuton-de [ittai nani-ol tsukamaeta no? w ho-N om B oston-in the hell w hat-A cc caught Q '*W ho caught what the hell in Boston?' Pesetsky (1987, 124-125) takes (94) as a fam iliar superiority contrast by assum ing that a wh-phrase w ith ittai must m ove first at LF. Provided that ittai is a Japanese equivalent o f the hell, this assum ption is based on the fact in English that a w h - phrase with the hell is obliged to m ove before Spell-Out leaving other w h-phrases in-situ w hich subsequently raise at LF. Thus postulated, (94a), m oving the highest wh-phrase first at LF, conform s to the custom ary superiority condition (C hom sky (1973)). On the other hand, (94b) is ruled out, since the low er w h- phrase ittai nani ‘the hell w hat’ m oves first. Supposing that the deviance o f (94b) is indeed a superiority violation, the proposed W CO analysis associates the LF-representations in (95a) and (95b) with (94a) and (94b), respectively: 44 See D. Takahashi (1993) for a different type o f superiority effect in Japanese. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 75 (95) (Irrelevant portions are om itted.) a. [ip [ I ittai darel-ga [yp [PRO] nani-ol tsukamaetall b. *f|p rPROi darel-ga [yp [ 1 ittai nani-ol tsukamaetall O bviously, the W CO constraint rules out (95b) but not (95a). An additional assum ption here is that a wh-phrase with ittai is construed as a generator o f a pair- list answ er (or just a pair answer, a pair-list on w hich there is only one pair). This interpretation is presumably forced by ittai's function to focalize a wh-phrase that follow s it. Such a pragmatic trait o f this expression is evidenced by the fact that a wh-phrase w ith ittai is normally pronounced with a heavy em phatic stress. Intuitively, in a m ultiple w h-question. this focalization marked by ittai indicates that the speaker's primary interest is in inquiry about the identities o f the focalized wh-phrase. The other w h-phrases then assum e 'secondary' interpretations, so to speak, where their values are covariant with the primary one associated with the focalized wh-phrase. I assum e that these interpretations are nothing but their functional interpretations. This W C O -account o f the superiority effect correctly predicts the paradigm in (96): (96) a. *D arei-g a kinoo f ittai dare-»l-ni John-o shookaishita no? w h o-N om yesterday the hell w ho -to -A cc introduced Q 'W ho yesterday to [w hom the hell] John introduced.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 76 b .? D arei-ga kinoo rittai dare->l-ni dare?.o shookaishita no? w h o-N om yesterday the hell w ho -to w ho-A cc introduced Q 'W ho yesterday to [ w hom the hell ] w ho introduced ' c. ?D arei-ga kinoo fittai dare? 1-ni soitsu2-no atarashii w h o-N om yesterday the hell w ho -to he -Gen new kateekyooshi-o shookaishita no ? tutor -A cc introduced Q 'W hoi yesterday to [2 w hom the hell ] his2 tutor introduced ' T he superiority effect in (96a) is circum vented by a low er w h-phrase dare? in (96b) and a low er bound pronoun soitsu? in (96c). On the current analysis, (96b) and (9 6 c) are associated with (97a) and (97b) at LF, respectively: (97) (Irrelevant portions are deleted.) a - [IP [ PRO? dare 11-ga fy p fittai dare-d-ni [PRO2 dare?1-o shookaishita]] b. [ip [PRO? dare 11-ga rv p rittai dare?l-ni f\rpsoitsu->-n o k a te ek y o o sh i]-o sh ook aish ita]] In (9 7 a ), dare 1 and dare? are both rendered as functions o f ittai dare?. Even though the higher PRO? is not c-com m anded by this generator, a potential W CO violation is suspended by the low er PRO?. (97b) is gram m atical likew ise. T hese suspension effects are constrained in a manner predictable on our W CO analysis: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 77 (98) a .* Darei-ga [M ary-ga rittai dare->l-ni John-o shookaishita tol itta no? w h o-N om -Norn the hell w ho -to -A cc introduced C om p said Q ’*W ho said [ that M ary to w hom the hell John introduced ].' b.*D are i-ga [ M ary-ga [ ittai dare? ]-ni dare-r-o shookaishita to ] itta no ? w h o-N om -N om the hell w ho-to w ho-A cc introduced Com p said Q ’*W ho said [ that M ary to w hom the hell w ho introduced ].' c .* D arei-ga f M arv-ga f ittai dare? 1-ni soitsu-> no kateekvooshi-o w h o-N om -N om the hell w ho -to his -G en -tutor -A cc shookaishita to ] itta no ? introduced C om p said Q ’*W ho said [ that M ary to |?w hom the hell] his2 tutor introduced ].' U nlike the paradigm in (96), the superiority effect in (98a) is rem edied neither by the low er wh-phrase dare? in (98b) nor the low er bound pronoun soitsu? in (98c). T he LF-representations o f (98b) and (98c) w ould be (99a) and (99b), respectively: (99) (Irrelevant portions are not presented.) a-*tlP [ PRCb darei 1-ga [cp M ary-ga [yp[ ittai dare-; ]-ni [PRCb dareU -o shookaishita] to] itta] b .*[ip [ PR O2 dare 11-ga [cpM ary-ga [vp[ ittai dare? ]-ni [np soitsu? no kateekyooshi]-o shookaishita] to ] itta ] R ecall that a wh-phrase is m ulti-scopal in that it bears a scope o f question and a scop e as a quantifier, and that bound variable licensing is sensitive to the latter. N either (99a) nor (99b) are w ell-form ed, since the higher PRO? do not fall within the scope dom ains o f the generator w h-phrases, i.e. the em bedded clauses. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 78 Finally the W C O analysis naturally explains the follow ing suspension o f the Japanese superiority effect, w hich is basically the sam e in nature as the "higher" w h -effect that w e discussed in 2.5.3: ( 100) a.*John-ga f M arv-ga darei-ni kinoo \ ittai nann 1 -o -N o m -N om w ho-to yesterday the hell what -A cc puresento-shita ka 1 shitteiru. presented Q know ' John k n o w s [ Mary to w hom yesterday what the hell gave ]' b. D arei-ga fM arv-ga dare-ni i kinoo littai n an b l -o -N o m -N om w ho-to yesterday the hell what -A cc puresento-shita ka ] shitteiru no. presented Q know Q 'W ho k n ow s [ M ary to w hom yesterday what the hell gave ] ' ( 100a) is a superiority violation. T his anom aly is w eakened in ( 100b) by dare-* ' w h o .’ w h en dare i is construed w ith the matrix question. Such suspension, how ever, cannot be observed if darei takes scope over the em bedded question. T he LF-representations in (101a) and (101b) correspond to the matrix and subordinate interpretations o f darei. respectively : (101) (sim p lified ) a. [ip d arei-ga [cp M ary-ga [PRO3 dare-ni 1 ] kinoo [ ittai nani?! -o puresento-shita ka ] shitteiru ] b. *[ip darei-ga [cp M ary-ga [PRCb dare-ni 11 .kinoo [ ittai nanb ] -o puresento-shita ka ] shitteiru] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 79 In ( 101a). its scope being confined to the matrix clause, darei is rendered as a function o f dare^. Since the bound pronoun PRO3 contained in this function is in the c-com m and domain o f its antecedent, (101a) does not violate the W CO constraint. On the other hand, ( 101b) is excluded by the constraint, since darei is a function o f ittai nanb and as a result it is forced to contain an illicit bound pronoun. 2.6. C onclusion In this chapter, I argued that the anti-superiority effect, discussed initially in Saito (1982), is subsum ed under the general w eak crossover phenom ena. A ssum ing the analyses in Chierchia (1991, 1992-1993) and Hornstein (1 9 9 4 ), I claim ed that in a multiple w h-question. naze is rendered as a function o f another wh-phrase. H ence, the W C O constraint is violated if naze c-com m ands a w h- phrase. If correct, the significant set o f LF-ECP phenom ena are reanalyzed in terms o f the W CO constraint: ( i ) distributions o f functional / pair-list interpretations in w h-questions w ith quantifiers ( i i ) the superiority effects in English and Japanese ( i i i ) the anti-superiority effect in Japanese Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 80 Suppose further that other ECP facts such as relativized minim ality effects (R izzi, 1991) have natural accounts in the econ om y considerations o f C hom ksy (1992, 1994). B eing com bined with these lines o f inquiry, our analysis strongly suggests that m ost o f (or, all of) the em pirical dem ands for the convention o f ECP should be elim inated, and hence such an artifact m ay w ell be dispensed in the theory o f the grammar. T his consequence in turn lends strong support for the m inim alist theory, where the notion o f ECP is just a "name tag" o f a given bundle o f data with no theoretical significance. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 Chapter Three T w o M odes o f Scope A ssignm ent to W h-phrases 3.1. Introduction A s N ishigauchi (1 9 9 0 , 26 - 32) points out, a w h-question is m arginal, w hen a wh-phrase is located in an em bedded ves-n o question. C onsider the contrast betw een ( 1) and (2 ): (1) (based on (39), N ish igau ch i (1990), 31) ?*Satoo-kun-w a [Suzuki-kun-ga nani -o tabeta kadooka ] -Top -N om w hat-A cc ate w hether shiri-ta-gatte-i-m asu -ka? know -w ant -Polite - Q 'Sato w ants to know w hether Suzuki ate w h a t.' (2) Satoo-kun-w a [Suzuki-kun-ga nani -o tabeta to] ii -m ashi -ta ka? -T op -N om w hat-A cc ate Com p say-Polite-Past Q ‘Sato said that Suzuki ate what. ’ The marginal w h-question in (1) contains nani ‘w hat’ in an indirect ves-n o question. In the fully acceptable exam ple in (2), on the other hand, the wh-phrase is located within a non-interrogative subordinate clause. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 A custom ary account o f tnl contrast, w hich N ishigauchi h im self (1990) puts forward and I shall ar u;e in this chapter, associates (1) and (2) w ith the LF- representations in 3) and (4), respectively: (3) (sim plified) ?*lrp nanii-oiTr'TiP - [cp ••• II ••• kadooka] . ..]] ka] (4) (sim plified) ?*[CP n a n j|-o i[c '[ip - [c p ••• II ...to] . ..]] ka] G iven LF w h-m ovem ent. nani-o ‘w hat-A cc’ 1 in (3) is extracted from the em bedded v es-n o question. This operation violates the so-called "wh-island constraint,"2 w hich the classical argument o f bounding has located, w ith other various "island" constraints o f R oss (1967), in the realm o f Subjacency (C hom sky, 1973). In contrast, the L F-m ovem ent in (4) ob viou sly does not violate this constraint. In this chapter, I shall pay clo se attention to a certain rem edy o f sentences like (1). In particular, (5) appears to be acceptable or at least decidedly better than (1), although the both exam ples contain nani within w h -islan d s:1 1 (3) assum es that a case-particle such as <> undergoes LF m ovem ent along with the w h-phrase to w hich it is appended. T his assum ption is not crucial. 2 I shall consider the nature o f this condition in section 3.4. 2 This fact w as first brought to m y attention by Hajim e H oji (class lecture, Sum m er, 1992). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 (5) [cp Satoo-kun-w a [Suzuki-kun-ga nani -o tabeta kadooka ] -Top -N om w hat-A cc ate whether shiri-ta-gatte-iru -no ]-desu ka? know -want -C om p -is -Q 'It is that Sato w ants to know whether Suzuki ate what. ’ The "core" parts o f (1) are em bedded in a clause headed by the so-called "nominal complementizer" no, and this no-clause is im m ediately follow ed by a copula verb d esu. 4 B asically follow in g Kuno (1973, Ch. 5), I assum e a "D-structure" like (6 ) for (5):s - 6 4 T his w h-question exhibits som e stylistic variations; ka, or both desu and ka can be omitted in this case. T hese stylistic alternations do not affect an acceptability judgm ent o f the exam ple. See the next chapter for discussion. See also S. M iyagaw a (19 8 7 ) and U eyam a (1991, 1992). 5 This structure w ill be slightly revised later. A N P that is marked with the so-called “topic” particle w a m ay very w ell reside outside o f the no-clause. W herever irrelevant, I shall neglect this matter. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Sato w ants to know [whether Suzuki ate what] (The position o f the wa-phrase is not accurate.) The no-clause, as it seem s natural to assum e, originates as a sister o f d esu . and this assum ption may w ell hold for whatever predicate preceding a copula. A n irrelevant question w ill be set aside in this chapter w hat fills the matrix IP-spec position in (6 ) . 7 T he aim o f this chapter is to suggest that basically tw o options are available in Japanese to assign scop e to w h-phrases. Pursuing a rather traditional view , I shall argue that the scope o f a wh-phrase is represented either (i) by m ovem ent operation or (ii) through binding by an interrogative com plem entizer such as ka along the lines o f 7 For detailed pragmatic / sem antic properties o f this pattern, see K uno (ibid.). Kuroda (1 9 7 3 ). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 85 C. L. Baker (1970) and Pesetsky (1987). The primary assir iptions that I shall make and defend in the course o f discussion are given in (7) : a. W h-m ovem ent is a last-resort operation triggered solely for assigning scope to a wh-phrase. (It has nothing to do with "m orphological feature checking" in the sense custom arily assum ed in the m inim alist theory.) b. An interrogative com plem entizer such as ka can represent scope o f a w h - phrase only if it c-com m andss the w h-phrase. c. Japanese questions in volve V -(I)-C m ovem ent. d. A syntactic operation cannot detach the no-clause in no-desu from the follow in g copula by a syntactic operation. (7) w ill explain the contrast betw een (1) and (5) as follow s; First, the marginal exam ple in ( 1 ) is associated w ith (8 ) at a certain point o f derivation: (8 ) (sim plified) s Section 3.4 w ill replace "c-command" with "seg(m ent)-com m and", a notion adopted from A. W atanabe (1991). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (7) CP I' [y l shiritagatteiru] [c ka] V P I kadooka nani 86 The matrix verb raises and adjoins to ka. The resultant adjoined structure, due to the higher "segment" o f C, prevents this interrogative com plem entizer from c- com m anding nani. H ence ka fails to represent the scope o f nani in (8 ). C onsequently nani must undergo covert m ovem ent to the matrix C P-spec position to represent its scope. This operation obviously cannot proceed w ithout violating the "wh-island constraint." On the other hand, the acceptable exam ple in (5) is m apped to the LF-representation in (9): For the no-clause cannot be separated from the follow ing copula, the entire V P is raised and adjoined to ka. W ith ka c-com m anding nani in (9), the scope o f the w h - phrase is represented by this com plem entier without m ovem ent. This chapter w ill build up an analysis step-by-step w ith m otivating and clarifying technical details. The organization o f this chapter is roughly as follow s. Section 3.2 w ill argue that traditional LF w h-m ovem ent is a m ode o f w h -sco p e assignm ent. I shall assum e that this covert operation in Japanese is m otivated for a scopal reason on Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (9) CP [... [ ... nani ... kadooka] ... no] 87 a par with Q(uantifier) R (aising) o f M ay (1977, 1985). This conclusion, if correct, im plies a necessity to m odify the M inim alist theory (C hom sky (1992, 1994, 1995)), since in that particular framework m ovem ent is adm issible only if it is to satisfy a certain m orphological requirement. Furthermore the view on LF w h-m ovem ent lends independent support to Fox (1994), w ho convincingly argues for the existence o f QR in his discussion o f sem antic properties o f the V P-ellipsis construction in E nglish. In section 3.3, I shall support "unselective binding" approach given by C. L. Baker (1970) and Pesetsky (1987). H ow ever, against a com m on b elief, the relevant relation betw een an interrogative com plem entizer and a wh-phrase w ill be distinguished from the general cases o f unselective quantification betw een adverbs o f quantification and indefinite noun phrases (L ew is (1975)). I w ill suggest, in particular, that the "unselective binding" between a wh-phrase and an interrogative com plem entizer is actually an instance o f association with focus (Jackendoff (1972)) and that this relation holds only w hen a wh-phrase falls under a certain formal domain o f an interrogative com plem entizer that section 3.4 will define. Empirical and conceptual consequence w ill be show n in section 3.5. Section 3.6 w ill contain a brief summary and a concluding remark. 3.2. An A ssum ption C oncerning The Nature o f LF W h-m ovem ent In this section, I w ill present a specific view on the nature o f LF w h-m ovem ent in Japanese along w ith som e consequences. In particular I shall suggest, basically in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 88 line with S. Kim (1991) and Mahajan (1990, Ch. 3), that this covert operation, akin to M ay's (1977, 1985) Q(uantifier) R (aising), takes place in order to assign scope to a wh-phrase. The fundamental features o f the proposal are sum m arized in ( 10): (10) A wh-phrase m oves to the specifier position o f an interrogative com plem entizer after Spell-O ut only if its scope cannot otherwise be represented. First consider the sim ple constituent question in (11): 01) Dare-ga ano hon -o kai-m ashi -ta ka? w ho-N om that book-A cc buy-Polite-Past Q 'W ho bought that book'?' H ow can dare 'who' be assigned scope? T w o prom inent p ossib ilities, both o f w hich I shall assum e in what follo w s, have been extensively discussed in the literature. In the first proposal, w hich stem s from C hom sky (1976) and is developed by Kayne (1 9 7 9 ), A oun et al. (1 9 8 1 ), Huang (1982), Lasnik and Saito (1 9 8 4 , 1992) and m any others, an in-situ wh-phrase undergoes familiar LF w h-m ovem ent into a CP- spec position and its scope is determ ined by where the w h-phrase finds itself at LF. T he other on e, proposed and elaborated in C.L. Baker (1 9 7 0 ) and P esetsky (1987), is that an interrogative com plem entizer ( = ka in ( 1 1 )) binds a w h-phrase. w here the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 89 former is treated as an "unselective" operator in the sense o f L ew is (1975) w hile the latter a variable bound to it. I assum e that (12) generally holds, w hich w ill be deduced from syntactic considerations as w e proceed: ( 12 ) In a w h-question ending in V(erb)-m asu-ka. ka fails to represent the scope o f a w h-phrase. The com putational system , as C hom sky (1 9 9 2 , 1994) proposes, selects for a sentence the least costly convergent derivation. Thus, w henever available, the Baker - Pesetsky type w h -sco p e assignm ent is chosen over w h -m ovem en t.9 W ith ka failing to represent the scope o f the w h-phrase. dare in (11) m oves to the C P-spec position after Spell-O ut as a last resort. (11) is thus associated w ith (13a) at the point o f Spell-O ut and ( 13b) at LF: (13) a- [c p [c'tlP dare-gai ano hon-o kaim ashita ] ka ]] b. Irp dare-gat [c tip tj ano hon-o kaim ashita ] ka ]] N ote that since there is no reason for dare to raise prior to Spell-O ut and since LF operations are less costly than pre-Spell-O ut ones, this m ovem ent procrastinates until 9 S ee Fox (1 9 9 5 ), w here he convincingly argues that covert operations are restricted by econ om y conditions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 LF (cf. C hom ksy, 1992, 1994), unless a wh-phrase is "accidentally" m oved into a C P-spec position by scram bling. 10 A question seem s to arise w hy a wh-phrase needs to m ove to the specifier position o f an interrogative com plem entizer. If the operation under investigation is m erely to assign scope to a wh-phrase. adjunction should suffice alongside o f Q R :11 (14) rrplrT iP dare-gai fip ti ano hon-o kaim ashita ]] ka ]] I assum e, nonetheless, that a wh-phrase must take scope at the specifier position o f an interrogative com plem entizer, and that an LF-representation o f the sort in (14) is sem antically gibberish (or uninterpretable), and hence ruled out. I adopt the statement in (15), follow ing the w idely accepted sem antic analysis o f questions by Karttunen (1977), w hich is based on Hamblin (1 9 7 3 ):1 2 10 See Kuroda (1 992, Ch. 10) and Takahashi (1993). 11 This possibility is indeed pursued in K im (1991) and Mahajan (1990). 1: H igginbotham (1991) suggests an intriguing alternative sem antic theory o f interrogatives, where a question denotes a partition o f possible states o f nature into fam ilies o f mutually exclusive (and possibly jointly exhaustive) alternatives (or cells, in his term inology). N ote that in the present discussion, I have no intention to c h o o se H am blin - Karttunen approach over H igginbotham 's analysis. Indeed his analysis w orks equally w ell for m y purpose, if w e assum e that his "I(nterrogative) formula," the scope o f a w h-phrase in his fram ework, syntactically corresponds to a C' introduced by an interrogative com plem entizer. See H igginbotham (1 9 9 1 ) for details. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 91 (15) A wh-phrase is an existential quantifier13 designated to take scope over a "proto question." In his v iew , a w h-question denotes a set o f propositions that jointly constitute a true and com plete answ er to the question. T his interpretation is obtained on his analysis by applying a wh-phrase. which translates into an existential quantifier, to an abstract sem antic object o f "proto question."14 A "proto question" is obtained by (16), called Proto Question Rule by Kartunnen: (16) (Kartunnen's (23)) PR O TO -Q U ESTIO N RULE: If 0 e P t (that is, if 0 is a phrase o f category t ), then ?(be P q (that is ?0 is a phrase o f category Q (uestion)). w If 0 translates to 0', then 70 translates to Xp [ p & p = A0 ] C onsider (1 3 ), repeated as (17), with respect to (16): (17) frp dare-gai [ c t ip t l ano hon-o kaim ashita ] ka ]] 1 ? See Kats and Postal (1964) and Kuroda (1965), where they propose to derive w h o and w hat transformationally from w h + som eone and w h + som ething, respectively. Treating wh-phrases in Japanese as existential quantifiers is in apparent contradiction w ith the customary thesis that they introduce variables at LF (Kuroda (1 9 6 5 ), N ishigauchi (1990)). A s a reconciliation, one may assum e, basically along the lines o f C hom sky (1992), A . W atanabe (1991) and Tsai (1 994), that a w h - phrase is a variable and its associated existential force is a contribution from an operator appended to it. 14 In Karttunnen’s analysis, an LF-representation is com positionally translated to M ontague's (1 9 7 4 ) I(ntentional) L (ogic). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A s the syntactic description o f (16) states, ^ is a declarative sentence and ? is a function which maps < J ) to an abstract semantic object o proto question. In the LF- representation in (17), the IP and the interrogative com plem entizer ka correspond to < £ and ? o f (16), respectively. Consequently, the sem antic value o f the C' w ou ld be a proto question o f the form in (18): (18) Xp [ p & p = A( x bought that book) ] N ow in (17) dare, an existential quantifier, takes scope over this proto question in the specific sense that (the translation of) the wh-phrase and the proto question com positionally yield the translation o f the C P as in ( 19):1 5 (19) Xp Exfperson'(x) & p & p = A ( x bought that book) ] This denotes a set o f the true propositions that jointly constitutes a com plete answer to the w h-question in (11). In order for this sem antic interpretation to com positionally obtain, dare and the C' m ust be sisters and, hence, the wh-phrase is required to be in the C P-spec position. If correct, the LF-representation in (14) is sem antically illegitim ate, since dare, w hich is not a sister o f the C', cannot take scope over the proto question denoted by it. 1 :1 For details, see Kartunnen (p. 19) for his wh-quantification rule. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 T his account is consistent with a w ell-know n fact that a w h-phrase. an "interrogative polarity item" (e.g. Abe (1994)), may not be licensed in a declarative sentence, as evid en ced in (2 0 ): ( 20) *Dare-ga ano hon -o katta (koto) w ho-N om that b ook-A cc bought (the fact) '(the fact) w h o bought that book' Suppose that a w h-phrase is an existential quantifier designated to take scope over a proto question. Then the sem antic interpretation o f this sentence w ould turn out to be gibberish, sin ce the involved C', headed by a non-interrogative com plem entizer, fails to denote a proto question. C onsider the sentence in (1), provided that this sem antics-based analysis is correct: (1) (based on (3 9 ), N ishigauchi (1990), 31) ? * [ c p [ c Satoo-k u n -w a [cp Suzuki-kun-ga nani-o tabeta kadooka ] shiri-ta-gatte-i-m asu-ka]] 'Sato w ants to know w hether Suzuki ate w h a t.' In (1), an exam p le com parable with (11), nani ought to m ove to the higher C P -spec position. T his m ovem ent is a last resort "scope-assigning" operation, w hich, follow in g N ishigauchi (1990), violates the "w h-island constraint." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 An important question arises what is the exact nature o f this constraint, tacitly assum ed thus far as if it were an independent "principle" o f UG. Putting the question differently, w hy is it that in Japanese the derivation leading to the abstract LF- representation in (2 1 ) is odd? 16 (21) (The direction in w hich the CP-spec's project is irrelevant.) * ? [c p w ll! [iP [vp [cp w hether [ c [ip ... tj ... ] Q ]]]] Q ] The anom aly o f the derivation, I assum e, fo llow s from the fact that the LF- m ovem ent o f w h i proceeds skipping the low er C', a "potential landing site" in that the C' corresponds to a proto-question where M il could take sco p e17 I maintain the generalization in (2 2 ), presumably reducible to econom y considerations: 16 G iven the present analysis, (21) satisfies Full Interpretation, w ith the scop e o f w h i is properly indicated. 17 T his discussion ob viou sly entails that a wh-phrase can occur and take scope in a ves-n o question. A lthough sentences o f the sort are normally gibberish on pragmatic grounds, a perfect exam ple is given in (i), that w as suggested to m e by A yum i U eyam a (p.c.). (i) [K ondo-no senkyo-de Jim intoo-ga nangiseki kakutoku-dekiru this-tim e-G en election-in L D P -N om how -m any-seats w in- can kadookal-g a giron-no shooten da. w hether-N om debate-G en focus is 'W hether the Liberal Dem ocratic Party w ill w in how many seats in the upcom ing election is the issue o f the discussion.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 95 (22) Marginality results if a scope m ovem ent skips a position where the m oved quantifier can take scope. The contrast betw een (23a) and (23b) naturally follow s from this generalization: (23) a. ?*Anata-wa [dare -ga Aum Shinrikyo -no kuromaku da you-T op w h o -N o m Aum Suprem e Justice-G en 'kuromaku' is kadooka] shitteim asu ka? w hether know Q 'You know [w hether w ho is the head o f the Aum Suprem e Justice].' b. Anata-w a Idare -ga A um Shinrukyo -no kurom aku-da to] you-T op w h o -N o m Aum Suprem e Justice-G en kuromaku' is C om p om oim asu ka? think Q 'You think [that w h o is the head o f the Aum Suprem e Justice].' In (23a), dare 'who' is contained in an em bedded ves-no question, w hereas in (23b) it is within a clause subordinate to om ou 'think.' The generalization in (22) is therefore violated in the derivation associated with the ill-form ed exam ple in (23a), for dare raises skipping the em bedded C' that denotes a proto-question. On the other hand, (23b), w h ose em bedded C’ is a declarative statem ent, is fully gram m atical18 At this point, it is not clear to me what factors determine acceptability judgm ents on sentences o f this kind. 18 Consider the exam ple in (i): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96 The generalization also accounts for a w ell-know n observation made in English that is exem plified in (24): (24) ( from Epstein (1992, 245)) Som eone thinks that Mary solved every problem . (?* som e < every ) It is know n that the "finite clause boundary" is a (som ew hat w eak) scop e island. Thus in (24) the w ide-scope interpretation o f every problem with respect to som e one is only m arginally available, if not im possible. On this interpretation, (24) is m apped to the LF-representation in (25): (25) ? * r p every prob letm riP som eon ei ftp t| thinks [cpthat [ipMary solved t2]]]]] The derivation that maps (24) to (25) is deviant, for it involves the LF-m ovem ent o f every one skipping the em bedded IP, a "potential landing site" o f the universal quantifer. 19 (i) ?ripT o m -n ii [jpSatoo-kun-ga [c p Suzuki-kun-ga t2 atta kadooka ] -to -N om -N om met w hether shiri-ta-gatte-iru] (koto) w ant-to-know (the fact) T o r m . Sato w ants to know w hether Suzuki met 11 Here T om -ni is scram bled out o f the w h-island. This sentence, albeit slightly m arginal, appears to m e to be m uch better than (1). This fact, too, naturally Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 3.3. The "Domain" o f an Interrogative C om plem entizer and Suspension o f W h-island Effects Recall the contrast betw een (1) and (5), reproduced in (26) for the ease o f exposition: (26) a. ( = ( 1)) ?*Satoo-kun-w a [Suzuki-kun-ga nani -o tabeta kadooka ] -Top -N om w hat-A cc ate whether shir-i-ta-gatte-i-masu -ka? know -want -Polite - Q 'Sato wants to know w hether Suzuki ate what. ' b. (= ((5 » [ Satoo-kun-w a [Suzuki-kun-ga nani -o tabeta kadooka ] shir-i-ta-gatte iru no] desu ka? C om p is Q 'It is that Sato wants to know whether Suzuki ate what' follo w s from the present generalization. 19 For the brevity o f argument, I put aside the possibility where a quantifier takes scop e at a V P adjoined position. See C hom sky ( 1986b). Furthermore, the constraint under discussion is relaxed w hen a quantifier is contained in a non-fm ite clause. It is unclear to me w hy this is the case. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 98 T o review , the deviance o f (26a) is understood as a violation o f the "wh-island constraint." W e are assum ing (27) regarding the application o f LF w h-m ovem ent. a scopally m otivated operation: (2 7 )(= (1 0 )) A wh-phrase m oves to the specifier position o f an interrogative com plem entizer after Spell-O ut only if its scope cannot otherwise be represented. Nani in (26a) m ust m ove to a C P-spec position, since this w h-phrase is assigned scope only by such an operation. W hy is it then that (26b) is grammatical? I suggested earlier that this is because the type o f w h -scop e construal given by C. L. Baker (1 9 7 0 ) and P esetsky (1 9 8 7 ), i.e. "unselective binding" by an interrogative com plem entizer, is available in (26b). If so, nani in this exam ple does not have to m ove and hence, according to (27), m ust stay in-situ. N o m ovem ent being involved, the "wh-island constraint" is not violated in the derivation o f (26b) as desired. A question im m ediately arises w hy Baker and Pesetsky's w h -scop e assignm ent is not available in (26a). To account for the difference betw een (26a) and (26b), I shall first argue that this particular option requires o f a wh-phrase to be w ithin the “dom ain” o f an interrogative com plem entizer. I shall further suggest that nani in (26b) but not that in (26a) falls under this "domain” o f ka. C onsequently nani in (26a) cannot stay in-situ at LF, since m ovem ent is the only available option to represent its scope. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 99 3.3.1. Foci o f Yes - No Questions T o construct an analysis along these lines, let us introduce first the notion o f "domain" o f an interrogative com plem entizer, initially discussed in Kuno (1980, 1982) and later in Takubo (1985) in relation to a certain trait o f ves-n o questions. The descriptive generalization in (28) is based on Kuno (1980), but is sim plified: (28) (based on Kuno's (1980) (10)) The "Domain" o f an Interrogative C om plem entizer: Only the follow ing can fall under the "domain" o f an interrogative complementizer: (i) an elem ent left-adjacent to it (ii) an elem ent contained in the no-clause o f the no-desu construction G iven (28), let us discuss Kuno (1 9 8 0 , 1982) and T akubo’s (1985) intriguing observations regarding distributions o f foci o f Japanese v es-n o questions. Their finding is that the focus o f a ves-n o question, if any, must be within the "domain" o f an interrogative com plem entizer. This fact, I interpret, indicates that association with focus in the sense o f Jackendoff (1 9 7 2 , Ch. 6 )) obtains betw een an interrogative com plem entizer and its intended focus only if the latter is within the "domain" o f the former. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 100 First, con sid er (29): (2 9 ) (based on (5 ), Kuno (19 8 2 )) a. K im i-w a kono hon -o Y O N D A ka? you-T op this book-A cc read Q 'Did you read or did you not read this book?' b. John-w a M ary-ni okane -o Y A T T A ka? -Top -Dat m oney-A cc gave Q 'Did John g ive or did John not giv e any m oney to Mary?' c. N an cy-w a Joseph-to H A N A SH IT A ka? -T op -w ith spoke Q 'D id N ancy talk or did N ancy not talk to Joseph?’ T he capitalized verbs in (29), w hich bear a focal stress, are the foci o f these ves-n o q u estio n s/ 0 A ll these exam ples are grammatical, since the foci, the verbs left- adjacent to, and therefore within the "domain" of, ka, are questioned. 20 H ence (29a), for exam ple, can tolerate (ia) but not (ib) as its answer: (i) a. U n, yon d a yo. y e s read ’Y eah, I read it.' b. U n, boku da yo. yes I is 'Y es, I read it.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 Compare the sentences in (29) w ith the unacceptable ones in (30): (30) a. *K IM I-G A kono hon -o yonda ka? you -N om this book-A cc read Q 'Is it you that read this book?' b. *John-wa M A R Y -N I okane -o yatta ka? -Top -D at m oney-A cc gave Q 'Is it to Mary that you gave som e money?' c. *N ancy -w a JO SEPH -TO hanashita ka? -T op -w ith spoke Q ‘Is it Joseph that N ancy talked w ith?’ T hese exam ples are ungrammatical, for their foci are outside the "domain" o f ka. In (30a), for exam ple, the subject kim i-ga 'vou-Nom ' is assigned em phatic stress, and therefore intended as a focus o f this question. H ow ever, this scop e assignm ent is illicit, since kim i-ga. not adjacent to ka, is not located within the "domain" o f the interrogative com plem entizer ka. 21 W e can w itness this very point in (31) as well: (3 1 ) a. *?K IM I-G A kono hon -o v o m i-mashi -ta ka? you -N om this b ook-A cc read-Polite-Past Q 'Is it you that read this book?’ 21 See Kuno (1982) and Takubo (1985) for a possible com plication in relation to what they call "m ultiple-choice focus." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 b. *?John-w a M A R Y -N I okane-o vari-m ashi-ta ka? -T op -Dat m oney-A cc give-Polite-Past Q 'Is it to Mary that you gave som e m oney?' c. *?N ancy-w a JO SEPH -TO hanashi-m ashi-ta ka? -T op -with speak-Polite-Past Q ‘Is it Joseph that N ancy talked w ith?’ T hese sentences m inim ally differ from those in (30) in that the verbs are all replaced by their polite counterparts. This lexical change, as indicated, seem s to induce slight im provem ent in acceptability.22 The sentences, how ever, remains highly marginal and I judge them ungrammatical. 22 I do not have any explanation regarding the im provem ent o f acceptability observed here. Incidentally, consider the exam ples in (i): (i) ( based on (4), Kuno (1982): T he judgm ents are m ine. ) a. *Anata -w a T O K Y O -D E umareta ka ? you -T op T okyo -in born-were Q 'Is it T ok yo where you were b om ?’ b.??K obayashi sensei -w a T O K Y O -D E o-umare-<J)-ni nar-i- teacher-Top -in be-born (H onor.) m ashi-ta ka. polite-Past Q ‘Is it T ok yo w here Prof. K obayashi w ere born?’ A ccording to Kuno, the (b)-sentence is, m arginal, but better than the corresponding (a)-sentence. (ib) uses the honorific form o - ... ni nar- and is in the Polite speech style 'm asu.' Kuno suggests that the verb form here ("Renyookei," or the continuative form) is responsible for the expansion o f the "focus domain" o f the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 103 Interestingly, (32) illustrates that basically any elem ent in the no-clause o f no-da / desu can be the focus o f a ves-n o question:21 (32) a. [KIM I -G A kono hon -o yonda no] desu ka? you-N O M this book-A cc read Com p is Q 'It is that Y O U read this book?' b. [John -w a M A R Y -N I okane -o yatta no] desu ka? -Top -Dat m oney-A cc gave Com p is Q 'It is that John gave M A R Y som e money?' interrogative com plem entizer. In particular, he argues that the suffix “i” , w h ich is appended to a verb stem to create its continuative form (this suffix is in visib le though, w hen it is affixed to a vocalic verb stem such as umare 'to be born' “w ea k ly ” extends the "domain" o f ka beyond the preceding verbal elem ent. H ow ever, K uno’s claim seem s to me to be mistaken. C onsider the fo llo w in g unacceptable e x a m p le : (ii) *K obayashi sensei -w a T O K Y O -D E o-umare-<J> -ni nattaka? teacher-Top -in be-bom (Honor.) Q 'Is it T ok yo w here Prof. K obayashi were born?’ N o w the verb natta is in its neutral form. This sentence is otherwise exactly the sam e as the one in (ib). If the verb umare. a continuative form, had a function o f extending the "focus domain" o f ka, as suggested by Kuno, this sentence w ou ld be (un)acceptable to the sam e extent that (ib) is (un)acceptable. W e can thus reasonably conclude that a polite m asu-form . rather than the continuative form o f a verb, is responsible for the slight im provem ent o f acceptability. 21 N eedless to say, an elem ent receiving a focus construal m ust be "sem antically non-em pty." S o , for exam ple, it m akes no sense to interpret as a focus a ca se particle such as ga or a com plem entizer no, even if such an entity is located inside o f the no-clause. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 c. [N ancy -w a JO SEPH -TO hanashita no] desu ka? -Top -with spoke C om p is Q 'It is that N ancy talked to JO SEPH ?’ Here the "core" parts o f the sentences in (30) and (31) are em bedded in the no-clause. T hese exam ples are w ell-form ed with their foci being inside the "domain" o f ka according to (28(ii)). 3.3.2. E xtension to W h-questions: On "Unselective Binding" and Scope- A ssignm ent to W h-phrases G iven these observations regarding y es - no questions, I propose to extend the generalization in (28) to w h-questions. w hich is stated in (33): (33) An interrogative com plem entizer can represent the scope o f a wh-phrase only if the latter falls w ithin the "domain" o f the former. The idea is that the "unselective binding" in the sense o f Baker (1970) and Pesetsky (1987) is indeed a case o f association with focus between an interrogative com plem entizer and a wh-phrase. Therefore, this relation requires o f a w h-phrase. the focus o f a w h -question. so to speak, must be w ithin the "domain" o f an interrogative com plem entizer just like the focus o f a ves-n o question must be in such a domain. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 105 Supposin g (33) and (27), let us consider (34) and (35). w hich are to be com pared with (31) and (32), respectively:24 (34) Suzuki-kun-w a nani -o tabe-m ashi-ta ka? -Top w hat-A cc eat-Polite-Past Q 'Suzuki ate w h a t.' (35) [Suzuki-kun -w a nani -o tabeta no] desu ka? -Top w hat-A cc ate C om p is Q 'It is that Suzuki ate w hat. ' 2 4 N otice that the verb in (34) is in the polite masu-form . T his verb form is obligatory here, since use o f its neutral counterpart results in ungrammaticality: (i) *Suzuki-kun -w a nani -o tabe-ta ka? -Top w hat-A cc eat-Past Q 'Suzuki ate w hat?' T o the best o f m y know ledge, this fact w as initially reported by Takubo (1985, 111- 112) and later discussed in S. M iyagaw a (19 8 7 ) and U eyam a (1 991, 1992). Incidentally, this sentence can be made fully grammatical by deleting ka: (ii) Suzuki-kun -w a nani -o tabe-ta < ))? -Top w hat-A cc eat-Past Q 'Suzuki ate w hat?' I set aside the theoretical issues regarding these data for the tim e. For discussion, see S. M iyagaw a (1 9 8 7 ), U eyam a (1 991, 1992) and Chapter 4 o f this thesis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 106 N ote that nani 'what' falls under the "domain" o f ka in (35) but not in (34). H ence, in the proposed view , (34) and (35) are mapped to the LF-representations in (36) and (37), respectively: (36) [cp nan i-oi [c tip Suzuki-kun-w a ti tabe-m ashi-ta ] ka]] (The position o f the wa-phrase is not accurate, but irrelevant.) (37) [CP [c'tlP [CP Suzuki-kun -w a nani-oi tabeta no] desu] ka<i>]] In (36), w ith its original position not being within the "domain" o f ka, nani 'what' raises to the C P -spec position to take scope. In contrast, nani in (37) is assigned scope by ka. A ssum ing the general last-resort property o f m ovem ent, as stated in (2 7 ), this w h -scop e representation, w hich in volves no m ovem ent operation, is obligatory for the w h-phrase in (3 5 ).2 5 2 5 Provided that the interrogative com plem entizer ka is a wh-operator. one may argue that the ban against vacuous quantification is violated in (38), where ka fails to bind any variable. O ne w ay out o f this potential problem is to assum e that vacuous quantification is in fact admitted in natural languages just as in an artificial language o f predicate calculus. A sentence such as (i) seem s problematic under this assum ption w hich has been ruled out by the restriction in question: (i) *(gu ess) w h o John bought that book. H ow ever, it appears plausible to exclude this sentence on com pletely independent grounds. For instance, the noun phrase w h o is not assigned any 0-role in this sentence and therefore it cannot be licensed as an argument, yielding a violation o f F(ull) I(nterpretation). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 107 Provided that the present approach is correct, the contrast betw een (26a) and (26b) are readily accounted for. Their respective LF-representations are given in (38a) and (38b): (38) a. *?frp nan i-oi [ c [ l P ••• [CP ••• M ... kadooka ] ... ] ka]] (= (26a)) b. [c p [c tlP [CP [CP - nan i-oi ... kadooka] ... no] desu] ka<i>]] (= (26b )) The difference is again whether or not (the original position of) nani is within the "domain" o f ka. In (3 8 a ), nani. w hich originates outside such a dom ain, must raise to the matrix C P -spec position. This operation, needless to say. necessarily violates the "wh-island constraint." Nani in (38b). on the other hand, falls under the "domain" o f ka. T hus, the w h- phrase is assigned a scop e by this interrogative com plem entizer without m ovem ent. G iven the relation betw een nani and ka in (38b) is association with fo cu s, this unbound relation across the w h-island is rather expected. This is because general cases o f association w ith focu s are independently known to be not subject to island constraints: (39) a. I only w onder w hether John met C hom sky (and not whether John met H alle). b. John only w anted to m eet the friend o f the m other-in-law o f the person that sh ook the hand o f the POPE, (and not the friend o f the m other-in-law o f the person that shook the hand o f the President ). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 108 c. You can do lots o f things with bananas; I even know a guv w ho SM O K E S them . ( Rooth (1985), 35 )) In (39a), for exam ple, association with focus obtains between only and C h om sk y. Just like in (38b), this relation is established across the w h-island. R ecall Pesetsky's (1987) claim that the w h -scop e construal in question is a case o f unselective quantification, com m only assum ed to hold between a non- presuppositional indefinite and unselective binders such as an adverb o f quantification and an existential closure operator (cf. D iesing (1 9 9 2 ), H eim (1982), H om m a et al. (1991), Kamp (1981) L ew is (1975) and Tsai (1 9 9 4 ), am ong many others). This approach, if not im possible, may face an em pirical problem in explaining the present property o f unboundedness o f the w h-scope representation under discussion. It is reported in Berman (1989), Ishii (1991) and Tsai (1994) that general cases o f unselective quantification are bounded far more strictly than m ovem ent. For instance, even a mere finite clause boundary is a "barrier" for such a relation:26- 2 7 2 6 For attempts to derive this "clause boundedness effect" o f unselective quantification, I w ould like to refer the reader to Berman (1989) and T sai (1994, Ch 3). 2 7 The “clause boundedness” needs to be relaxed in cases where an unselective quantifier ranges over H eim ’s (1990) “m inim al situation” (or, a D avidsonian event). C onsider (i), a standard donkey sentence: (i) M ostly [C P if John ow ns a caq], [Cp he (=John) thinks that shei is intelligent] Supposing that an unselective quantifier cannot bind an indefinite N P across a sentence boundary, a cat cannot be bound by m ostly in this exam ple. T his is not problem atic, since this indefinite N P can be bound by a closure operator inside the if-clause as in (ii): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 (40) (based on an exam ple in Ishii (1 9 9 1 , Ch 3, 216)) John seldom rem em bers that regular custom ers tip big. (41) a. Regular custom ers generally tip big, and John seldom remembers this. b. O f few regular custom ers w ho tip big, John rem em bers they did tip big. In (40) the u nselective binder seldom m ay not bind the nonpresuppositional indefinite regular custom ers, since they are separated by the finite clause boundary. (40) thus obtains the interpretation in (41a) but not the one in (41b). Likewise: (ii) M ostly [c p if John E (x) [vp ow ns a cat(x) ]], [cp he (=John) thinks that sh ei] is intelligent. T his LF-representation is not satisfactory in the follow in g tw o respects. First, it does not represent what m ostly ranges over. Second, she cannot be licensed as a bound pronoun, for a cat fails to c-com m and this pronoun. Then, the LF o f (i) should be as in (iii), letting M be a variable ranging over m inimal situations: (iii) M ostlyM [cp if John E (x) [vp ow n s a cat(x) ] in M ], [cp he (=John) think that [the cat that he(=John) ow ns in M ] is intelligent. ‘For m ost m inim al situation M , if there is a cat such that John ow n s it in M, John thinks that [the cat that he (=John) ow ns in M] is intelligent.’ In this representation, m ostly ranges over m inim al situations restricted by the antecedent clause, and she receives an interpretation as a traditional E-type pronoun. (Evans, 1980). T h is E-type pronoun contains a variable M bound to the unselective operator m ostly. T his dependency is by no m eans “clause-bound.” H ence w e need to relax “ clause-boundedness,” when an unselective binder ranges over m inim al situations. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 (42) (Tsai's (1 9 9 4 ) (13)) [ i p I [vp E [vp think [cp firem en are adm irable]]] (43) E x (I think [x: fireman (x)] are admirable ) 'There is som e x such that x is a fireman and I think x is admirable.' The existential closure operator E , w hich adjoins to the matrix V P (D iesin g( 1992)), cannot bind firemen in the em bedded sentence, failing to yield the interpretation in (43). Due to these observations, it appears em pirically preferable to leave the m ode o f w h -scop e representation in question out o f the realm o f the unselective quantification. 3.4. A Formal D efinition o f the "Domain" o f an Interrogative C om plem entizer T he rem aining questions are, o f course, what w ould be the exact form ulation o f the "domain" o f an interrogative com plem entizer and w hy it is that Kuno and Takubo's generalization, repeated in (4 4 ), obtains: (44) The "Domain" o f an Interrogative C om plem entizer: O nly the follow ing can fall under the "domain" o f an interrogative com plem entizer: (i) an elem ent left-adjacent to it (ii) an elem ent contained in the no-clause o f the no-desu construction Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I l l In this section, I w ish to argue, as in Takubo (1985), that this notion o f the "domain" o f an interrogative com plem entizer is definable in purely syntactic terms. In particular, I shall assum e that this "domain" is formulated in terms o f the notion o f segm ent-com m and. adopted from A . W atanabe (1991) with a slight m odification. 3.4.1. A D efinition and V erb-R aising First consider the sentence in (45), focusing on (44(i)): (45) K im i-w a kono hon -o yonda ka? you-T op this book-A cc read Q N othing but yonda ’read’ can be a focu s o f this ves-n o question, for the verb is the only elem ent that falls under the "domain" o f the interrogative com plem entizer. Rephrasing this observation in Jack en d offs (1 9 7 2 , C h.6) term inology, kacan enter into association with focus only w ith vonda in (45). A s Jackendoff (1972, Ch. 6) initially points out, general cases o f association with focus m ust m eet a certain structural condition. F ollow ing A oyagi (1994) and Tancredi (1 9 9 0 ), let us assum e as a first approxim ation that a "focus operator" such as sae ’even ’ m ust c-com m and its "focalized associate." T his is illustrated in (46): (46) (based on A oyagi's (6)) a. John -w a [M A R Y -G A shihonron -o yonda to] -sae shinziteiru. -T op -N om D as K apital-Acc read C om p-even b elieve 'It is even Mary that John b elieves read D as Kapital.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 b. John-w a [M ary-ga SH IH O N R O N -O yonda to]-sae shinziteiru. 'It is even Das Kapital that John b elieves Mary read.' c. John-w a [M ary-ga shihonron-o Y O N D A to]-sae shinziteiru. 'It is even read (it) that John b elieves w hat Mary did to Das Kapital is.' d.*John-w a [M ary-ga shihonron-o yonda to]-sae SH INZITEIRU. 'It is even b elieve (it) that what John did to the proposition o f "Mary read D as Kapital" is.' Plainly, sae. right-adjoined to the em bedded sentences,2 8 can only c-com m and29 the elem ents inside these com plem ent clauses. H ence in (46a), for exam ple, M ary-ga 'M ary-Nom ' and sae can successfully enter into association with fo cu s. H ow ever, shinziteiru 'believe.' not within the c-com m and dom ain o f sae. is never attracted to the focus operator as in (46d). G iven this restriction on association w ith focu s, it is not inconceivable that in (4 5 ), ka can only c-com m and vonda. T his in turn im plicates that ka and the verb form a constituent as in (47): 2 8 Or, one may assum e, with changing nothing essential, that the focus operator itself projects to, say, a F(ocus) P(hrase). 2 9 I speculate a segm ent o f a category (in this case, a segm ent o f the em bedded CP) forces the c-com m and domain o f sae confined within the adjunction structure. See the discussion below regarding the notion o f "segment command." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. yonda ka T his constituency, if correct, keeps ka from c-com m anding any elem ent outside this am algam as w e desire. The next question is why it should be that vonda and ka thus constitute a constituent. K eeping (i) to the assum ption that ka is a com plem entizer1 0 and (ii) to the prohibition against low ering operations in general, w e appear to be left only with 10 A lthough this view is by and large standard, I do not have any d ecisive argument for this categorial status o f ka. Should it be plausible that ka is not a com plem entizer, other possibilities rather than the one in the text may arise to obtain the intended constituency. For instance, as Takubo (1 9 8 5 , 101) alludes to, it seem s feasible prima facie that (4 5 ) has a structure like (i): (i) [|p K im i-w a [yp hon-o [y [y yonda ] ka ]]] I sim ply stipulate here that ka is a verbal suffix that adjoins to yonda and that due to the higher segm ent o f V, can c-com m and nothing but yonda. In so far as the present d iscussion o f the "domain" o f ka is concerned, this analysis appears to work equally w ell as the one pursued in the text. I tentatively ch oose the latter analysis partly because it has som e im plications for English and French that w ill be briefly d iscu ssed b elow , and I w ant to m inim ize "parametric variations" between Japanese and other languages unless firm ly m otivated on em pirical grounds. Incidentally, F u k u i(1 9 9 5 , Ch. 4) proposes (a) that unlike E nglish, Japanese entirely lacks C's, and (b) that ka in a question and to in an em bedded declarative sentence, representative Japanese "complementizers," are indeed a noun and a postposition, respectively. C onsider the exam ples in (ii) and (iii): (ii) (Fukui's (37a, d), 117) a. [[John-ga nani -o katta ] ka] -ga mondai da. -N om w hat-A cc bought Q -N om problem is 'The problem is w hether John bought it.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 114 verb raising to achieve this constituency. A ssum ing this is indeed the case, I w ish to associate (45) w ith the LF-representation in (48): 1 2 b. B ok u -w a [[B ill-g a nani -o katta ] ka]-o shiritai. -T om -N o m w hat-A cc bought Q -A cc want-to-know 'I w ant to know w hat John bought.' (iii) (Fukui's (41a, c), 119) a. *[[John-ga M ary-o nagutta ] to ] -ga odoroki da. -N o m -A cc hit C om p -N om surprising is 'It is surprising that John hit Mary.' b. *John-w a [[B ill-ga M ary-o nagutta ] to] -o shitteiru. -T op -N om -A cc hit C om p-A cc know 'John k n ow s that B ill hit Mary.' The sentences in (ii) illustrate that a case particle such as ^ga and ^o may follow an interrogative sentence ending in ka. Supposing that these case particles attach only to a noun. Fukui con clu d es that ka is a noun. The sentences in (ii), on the other hand, indicate that a declarative sentence w ith to at its end cannot be follow ed by a case particle. A ccording to Fukui, this is because to is not a noun (but a postposition for an independent reason.) A t this point, I am not quite sure how m uch this argument should w eigh. One factor that m akes m e reluctant to accept his claim as it stands is that observations sim ilar to those in (ii) and (iii) can be m ade in English as w ell. C onsider (iv) and (v): (iv) a. W e w ere talking [about [w ho w e should help]]. b. I con sid er [[w h o you d ecid e to work with] to be unimportant]. c. John explained [w hy the sky is blue] to his children. (Stow ell (1 9 8 1 ), 3 9 2 )) (v) a. *W e w ere talking [about [(that) w e should help them ]]. b. *1 consider [[that you w ork w ith Roger] to be unimportant]. c. *John explained [[(that) the sky is blue] to his children]. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 115 (48) (sim plified) [CP [c tlP tr [v p kono hon-o ti ] ti-I2 ] [c yon d a|-k a ]]] ror expository purposes, this representation contains (the history of) a series o f head m ovem ent V - 1 - C. First, the verb vonda 'read' adjoins to I(nfl). if one wants to adhere to the M inim alist theory, to knock o ff its tense feature. Then the V further (S tow ell (1981, 393)) Interrogative sentences m ay occur in typical "Case" positions as in (iv), w hereas declarative sentences m ay not as in (v). G iven Fukui's proposal, it then appears to be natural that a covert "complementizer" in (iv) is a noun, w hile that in (v) is not. Is this w hat w e want to assum e? Probably not. Stow ell (1 9 8 1 , Ch. 6), for exam ple, explains the distributional properties in question in term s o f his C (A SE ) R (esistance) P(rinciple), keeping to the custom ary assumption that an em bedded interrogative clause in E nglish is headed by a covert com plem entizer. Thus, to the extent that an explanation, whatever it is, is possible along the lines o f Stow ell, the contrast betw een (ii) and (iii) does not necessarily constitute evid en ce for Fukui's claim . M M any languages involve head m ovem ent o f this sort in interrogative sentences. One exam ple is drawn from French in (i): (i) C onnaissez-vous Jean-Roger? know you 'Do you know Jean-Roger?' See K oizum i (1995, Ch. 7) for m uch relevant discussion concerning verb raising in Japanese. The discussion is unaffected by w hether this agglutination takes place before or after Spell-O ut. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 116 raises to the position o f the com plem entizer for checking a yet unknown feature, possibly related to question formation.'*'' Let me enlarge the amalgam thus formed for the convenience o f illustration: W e must make sure that the follow ing obtains: (50) a. Ka c-com m ands yonda but nothing outside the higher segm ent o f C. b. Y ondai c-com m ands its trace. (50a) is o f course necessary to define the "domain" o f ka. W e need (50b) in order for the raised verb to satisfy the P(roper) B (inding) C (ondition) (F iengo (1977)), stating that a trace must be bound.'*4 ■ * * I am entertaining a possibility o f "successive cyclic head movement" or "excorporation," w hich M. Baker (1 9 8 8 ) prohibits to insure the effects o f the H (ead) M (ovem ent) C(onstraint). H ow ever, see Takano (1 9 9 4 ), where it is argued that X °-m ovem ent may proceed su ccessive cyclically, deriving the effects o f the H M C from econom y considerations; in particular, the last resort condition or Greed. 1 4 Or. as in Chom sky (1995, Ch. 4 ), it may be that M ove / Attract is defined in terms o f the notion o f c-com m and. For related discussion on the matter, see C ollins (1 9 9 4 ) and K oizum i (1995, Ch. 7). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (49) C 117 It seem s how ever rather difficult to obtain (50a) and (50b) at the sam e tim e with a conventional definition o f the notion o f c-com m and given, say, in C hom sky (1986b): (51) (C hom sky's (1986b ) (13)) C -com m and a c-com m ands P iff a does not dom inate P and every y that dom inates C X d om inates p . In order to yield (50a), the tw o segm ents o f C in (49) ought to be treated as independent categories (cf. Lasnik and Saito (19 9 2 )) so that the c-com m and dom ain o f ka does not extend beyond this am algam . H ow ever this assum ption clearly fails to capture (50b ), since the segm ent C dom inating vonda w ould prevent the latter from c-com m anding its trace. A tentative technical solution that I propose is that the "domains" o f an interrogative com plem entizer and focus operators in general are defined in terms o f seg(m ent)-com m and rather than c-com m and. Here is the definition o f s e g -com m and: (52) (partly based on A . W atanabe's (1991) (2 0 8 )) Seg( m ent)-com m and a seg( m ent)-com m ands P iff C L does not dom inate p and every segm ent that dom inates C L dom inates p. N ow , consider the schem atic representations in (53): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 118 (53) b Let J3 be an arbitrary focus operator. In (53a), com parable w ith the exam ples in (4 6 ), J3 is adjoined to O C , d ividing O t into tw o segm ents. A ccording to (52), can only seg-com m and the low er Ct and elem ents, if any, contained inside o f it. The paradigm in (46) follow s from the assum ption that association with focus demands that a focus operator seg-com m and its "focalized associate." O n the other hand, in (53b ). w hich is on a par w ith (4 9 ), |3 is adjoined to by a creating tw o segm ents o f P- A necessary additional assum ption is that the "domain" o f P. for association w ith focu s corresponds to the seg-com m and dom ain o f the low er J3 not that o f the entire category. If so, J3 can enter into association with focus on ly w ith a or the elem ents that it contains. N ote further that if C X is adjoined to J3 due to m ovem ent, as vonda in (4 9 ), O C can satisfy the PBC keeping to the assumption that its defining notion is c- com m and. On this account, let us consider the w h-question in (54): (54) Suzuki-kun-w a nani -o tabe-m ashi-ta ka? -Top w hat-A cc eat-Polite-Past Q 'Suzuki ate w h a t.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I 19 It is crucial that the question in (54) is not em bedded in a no-clause o f no da / d esu . G iven the discussion in the above sections, a wh-phrase within a question o f this form, being outside the "domain" o f ka^ must undergo last resort m ovem ent to a C P-spec position for scope representation, and if the wh-phrase originates within a w h-island, the "w h-island constraint" is necessarily violated. N ow (54) is associated with the structure in (55) at the point where the proposed V -m ovem ent applies: [n p nani-o] tj N otice in (55) that for nani 'what' is not in the "domain," i.e. seg-com m and dom ain, o f ka, the scop e o f the wh-phrase fails to be represented in this structure. A s a consequence, (5 5 ) further converts into (56): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (55) CP I' [v i tabem ashita] 120 ( 56) CP [n p 2 nani-o] ' [v l tabemashita] VP t- > ti In (56), nani raises to the CP-spec position to indicate scope as a last reso rt. An im m ediate prediction o f this analysis is that if an interrogative sentence involves V - C m ovem ent in a given language, the focus o f a ves - no question should be confined to the raised verb exactly like Japanese. Although I need to leave for the future research the task o f submitting this sim ple analogy to thorough cross linguistic considerations, a French contrast such as the one in (57) appears to provide us with an initial confirm ation o f this prediction:3 5 (57) a. PARLEZ vous fran^ais? speak you French 'Do you S P E A K French?' b. *Parlez vou s F R A N £ A IS ? speak you French? 'Is it French that you speak?' 3 5 I ow e this observation to Jean-Roger Vergnaud. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 In (5 7 ), where parlez 'speak' is raised and adjoined to the covert com plem entizer, nothing but this verb seem s to be licensed as a focus o f this question. This fact fo llo w s from the proposed analysis, w hich associates (57) with (58) at LF: (58) [ c [c Parlez I C[qj ]] [ip vous [vp ti fran^ais ] ] ] Since parlez is raised and adjoined to the interrogative com plem entizer, the C can only seg-com m and parlez. Consequently this verb and only this elem ent can enter into association with focus with the C. Apparent counterexam ples are found in English. C onsider (59): (59) D oes FR ED do sixty sit-ups (every day)? 'Is it FR ED that d oes not sixty sit-ups?' A s Jackendoff (1 9 7 2 , Ch. 6) am ply observes, a ves-n o question o f English licenses virtually any elem en t as its focus w hich is contained in the sentence. Thus, in (59), FR ED is fully allow ed as the focus o f this ves-n o question. T he present theory, as it stands, w ould falsely predict ill-form edness o f (59) by assigning (60) to this exam ple at LF: (60) [e tc D oes i C[Q] ]] [ i p FR ED [vp t) do sixty sit-ups ] ] ] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 Fred fails to be seg-com m anded by the C_since the com plem entizer is adjoined by d o es. Fred therefore cannot enter into association with focus with the C. I take the w ell-form edness o f (59) to suggest that d oes is deleted from the LF- representation. T hus, (61) instead o f (60) is the LF-representation o f (59) in this view: (61) [ c C|Q] C lP FRED [v p do sixty sit-ups ] ] ] I sim ply assum e here that does is a sem antically vacuous elem ent that is inserted only to let the C check o ff a certain (yet unknown) m orphological feature o f this "dummy" verb. Lacking a sem antic interpretation, does is an illegitim ate LF object and, hence, must be om itted. A s w e desire, the resultant LF-representation in (61) allow s the C to seg-com m and Fred, w hich licen ses this N P to be the focus o f the q uestion.3 6 36 A sentence such as (i) poses a blatant conceptual and empirical problem for the present claim: (i) ?Can FR E D do sixty pushups? 'Is it F R E D that can do sixty pushups?' A ccording to m y informant work, (i) appears to be acceptable on the reading where Fred is the focus o f this question. Can raises and adjoins to the C, form ing the configuration in (ii). (ii) [C'[C Can i C [Q ] ]] [ip Fred [VP ti do sixty pushups ] ] ] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 123 3.4.2. On Copula Sentences Let us now turn to (44(ii)). G iven the analysis in the last subsection, this generalization im plicates that in a question, the no-clause o f no-desu is in the seg - com m and dom ain o f an interrogative com plem entizer. Let us first consider a sim ple ves-n o question in (62), w hich is highly su ggestive for our d iscussion o f exam ples w ith no-d esu : (62) A no hito -w a [n p John-no otoosan] desu ka? that person-T op -Gen father is Q ‘Is that person John’s father?’ Kuno (ibid), T akubo (ibid.) and K. M iyagaw a (p.c.) point out that the "domain" o f ka , analogous to cases with no-desu. can freely extend to any elem ent in the predicate im m ediately preceding a copula. Thus in (62) either John or otoosan 'father' can be a focu s o f this question, as evid en ced in (62'): (6 2 ’) a. A no hito-w a [n p JO H N -no otoosan] desu ka? ‘Is that person JO H N ’S father?’ N otice that can, not a sem antically null elem ent, may not disappear at LF. W ere (ii) the LF-representation o f (i), Fred falsely w ould not be licensed as a focus, for the C fails to seg-com m and this NP. T w o possibilities im m ediately com e to m ind. The first one is to “undone” the head m ovem ent o f can after Spell-O ut as in (iii): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 b. A no hito-w a [n p John-no O T O O SA N ] desu ka? Ts that person John’s FA TH ER ?’ (6 2 ’a) and (6 2 ’b) can tolerate (63a) and (63b). respectively, as an answer: (63) a. lie. M ary-no otoosan-desu. N o -G en father is 'N o, he is M A R Y ’s father.' b. lie, John-no oniisan-desu. N o -G en brother is ‘N o, he is John’s BR O TH ER .' In contrast, a subject noun phrase outside a nominal predicate m ay not be a focus: (64) ?*A N O HITO -ga [n p John-no otoosan] desu ka? that person-N om -G en father is Q 'Is T H A T PE R SO N John's father. ' Ano hito 'that person,' w hich I assum e is in the IP-spec position, cannot be a focus o f this ves-n o question. The present analysis given, this observation leads us to conclude that the entire predicate nom inal preceding a copula, but not the subject N P, is in the seg-com m and dom ain o f an interrogative com plem entizer. T o obtain this result, I w ould like to propose a som ew hat "non-standard" structure for a copula sentence. In particular, I propose that (65a), an affirm ative counterpart o f (62), is associated with (65b) at a certain point o f derivation: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 125 (65) a. A no hito -ga John -no otoosan desu. that person-N om -G en father is 'That person is John’s father.' b. (sim plified) IP r ano hito-ga N P [v desu ] [n p John-no ] [n otoosan] The nominal predicate John-no otoosan ’John-G en father’ is directly adjoined to d esu . These elem ents, I suggest, constitute a "com plex predicate," and "freeze" in the sense that the am algam is syntactically inseparable, as evidenced in (66): (66) (based on (9), K u n o (1 9 8 0 )) *[n p i John -no otoosan] ano hito-ga ti desu. ’John’s father, that person is ’ (66) illustrates that scram bling, a syntactic operation, cannot displace the nom inal predicate apart from the copula. This fact naturally follow s if John-no otoosan desu John’s father is’ is a "syntactically frozen unit.",7 (iii) [ c C[Q ] [ip Fred [V P can do sixty pushups ]]] The question is w hy Japanese does not permit "LF-undone" o f this kind. T he answ er Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 might be that an interrogative com plem entizer such as ka in this language is a bound m orphem e. G iven that this elem ent must be "supported" even after Spell-O ut, a verb attached to ka cannot be low ered to its original position at LF. Regarding E nglish, c a n in (ii), for exam ple, raises solely for the checking purpose. A fter checking takes place, this elem ent can be freely undone. The second possibility is to assum e that can in (i) does not raise all the w ay up to the C, but to som e X °-category betw een the C and the I: (iv) [c . C [Q ] [XP [x can ,-X ] [,p Fred [vp t, do sixty pushups ]]]] The C can seg-com m and Fred in this representation, since nothing adjoins to it. The putative X may w ell be a category “hosting” an auxiliary raised by Subject-O bject inversion as in (v ) (cf. C ulicover (1991)): (v ) (K oizu m i's (1 9 9 5 , Ch. 6)) a. B ecky said that at no tim e w o u ld she agree to visit Marty. b. John sw ore that under no circum stances w o u ld he accept their offer. A gain, as for Japanese, a verb must raise to the position o f C in Japanese due to the affixal nature o f com plem entizers. I leave exact w orkings o f the above tw o possibilities for future discussion. I also leave it open how to deal with French exam ples like that in (57) in these considerations. 3 7 One m ight explain the ungram m atically o f (66) by assum ing that like to 'and' in (i), desu is a "phonological clitic" (K oizum i. 1995, 204-207) that cliticizes to the non-case marked N P im m ediately preceding it: (i) a. K eiko to Yuki -ga kita. and -N om cam e 'K eiko and Y uki came.' b. *K eik o-ga to Yuki -ga kita. -N o m and -N om cam e 'K eiko and Yuki came.' G iven that to cannot cliticize to K eiko-ga 'K eiko-N om .' (ib) is ruled out on phonological grounds. Should this property o f to be shared by d esu . the deviance o f (66) could be phonological as w ell. H ow ever, consider (ii): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 127 G iven this analysis, I assum e that the sentence in (62) is mapped to the LF- representation in (67): (67) V , Ic k a ] I N P [v desu] [NP-John-no] [n otoosan] I maintain that ka is a com plem entizer, and that verb-m ovem ent takes place to ka. W hat raises and adjoins to the C in this case ought to be the entire com plex predicate (ii) A: Dare -ga ano hon -o katta no? w ho- nom that book-A cc bought Q ’W ho bought that book?' B: K eiko-ga desu. -N om is 'K eiko does.' A s the gram m aticality o f (iiB ) illustrates, a sequence o f a N P marked w ith the nom inative marker ga and desu is phonologically sound. H ence w e cannot sim ply assum e that desu is a phonological clitic that is obliged to attach to a preceding non case marked NP. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 I v 1nP2Inp John-nol t i.Uvfy j otoosanlfy _d esu lll. for syntactic operations cannot decom pose this am algam . Consequently, at LF ka seg-com m ands this com p lex and any elem ent inside o f it, but not the elem ent in the IP-spec, as desired. Let us now turn to (44(ii)). This generalization appears to follow if the no-clause forms a com plex predicate with the follow ing copula, and this w hole am algam adjoins to an interrogative com plem entizer. If so, the ves-no question in (68a) is associated w ith (68b) and (68c) before and after the proposed predicate raising, respectively: (68) a. [T om -ga kono hon -o yonda no] desu ka? -N om this book-A cc read Com p is Q 'Is it that T om read this book?' b. (sim plified) IP Y=VP) [y desu CP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 129 c. (sim plified) CP [C ka [v desu] A s in the above argument concerning nom inal predicates, the no-clause forms a "unit" with the follow in g copula that cannot be separated by a syntactic operation. This is evidenced in (69b): (69) a. B oku-w a [[John-ga kita no] da to] om ou (koto) -T op -N om cam e Com p is Com p thought 'I think that it is that John cam e.' b. * [John-ga kita no] i B oku-w a [ tj da to] om otta (koto) -N om cam e C om p -Top is C om p thought 'That John cam e M ary thought that it is.' N ot being syntactically separated, this w hole "complex" predicate raises to ka in (68c). A s a result, the no-clause and the elem ents therein are all relocated in the seg- com m and dom ain o f ka. T his is, o f course, w hat w e want to have. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 130 Finally, consider the w h-question in (70): (70) [Suzuki-kun-w a nani -o tabeta no] desu ka? -T op w hat-A cc ate Com p is Q 'It is that Suzuki ate w hat. ' Nani 'what' is em bedded in no-d esu . Provided that the earlier argument is correct, the scope o f this w h-phrase. falling within the "domain" o f ka, is represented by this interrogative com plem entizer without m ovem ent. W hat is presently assum ed associates (70) with the LF-representation in (71): B eing seg-com m anded by ka, the scope o f nani is represented by this interrogative com plem entizer. C onsequently this wh-phrase does not raise to the C P-spec position, sim ply because such an operation is not necessary for convergence. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (71) CP I tv desu] Suzuki-kun-w a nani-o tabeta no 131 3.4.3. Negation and Focus Em pirical evidence for the proposal that association with focus requires o f a "focus operator" to seg-com m and its "associate” may w ell com e from a certain property o f negative sentences. C onsider Jackendoff s (1972, 2 57) observation that the English negation not optionally induces association with focus:3 8 (72) John did not kick M A R Y . T w o contrasting interpretations are given in (73) am ong plausible readings o f this ; i ) .ii i sentence: , 38 U nlike negation, focus operators like even and only obligatorily induce association with fo cu s. For m uch relevant discussion, see Jackendoff (1 9 7 2 , Ch. 6). 3 9 For exam ple, this sentence allow s interpretations where the V P kick M A R Y is construed as a focus: (i) a. N ot [kick Mary] but som e other action w as what John did. b. W hat John did not do w as [kick M ary], G iven the English negation optionally induces association with focus , kick M A R Y is attracted to not in (ia) but not in (ib). 40 Jackendoff (1972. Ch. 6) observes that the readings in (73) can be distinguished by intonation contours. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 132 (73) a. It is not Mary that John kicked. focus presupposition b. It is Mary that John did not kick . focus presupposition In (73a), the type o f interpretation that I shall discuss regarding Japanese, the negative expression and Mary enter into association with focu s. Like the other instances o f association with focu s, this interpretation obtains only if the focus operator not contains its "focalized associate" within its dom ain. Thus the reading indicated by (73a) is unavailable for (74): (74) M A R Y , John did not kick. In (72), if M ARY is not attracted to the negation, the sentence receives the interpretation in (73b ), where riot is rendered sim ply as part o f the presupposition. M oreover this reading appears to be the only interpretation available for (74). Turning to Japanese, Kuno (1980, 1982) and Takubo (1985) observe that the behavior o f the focus o f a negative sentence is basically on a par with that o f a ves-n o question. C onsider the sentences in (75): (75) a. John -w a M ary -o KER-ana-katta. -T op -A cc kick neg-past ‘W hat John did to Mary is not K IC K .’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 b. *John -w a M A R Y -0 ker-anakatta.4 1 ‘It is not Mary that John kick.' c. [John-w a M A R Y -O ketta no] de ??(-wa) na-i. -T op -A cc kicked Com p is-FO C U S neg-present ‘It is not Mary that John kicked.’ The contrast betw een (75a) and (75b) indicates that the "domain" o f ana42 is limited only to the elem ent left-adjacent to the negative elem ent. (74a) is w ell-form ed, since the focus ker 'kick' is left-adjacent to, therefore within the dom ain of, (alna. On the other hand, (75b) is excluded due to the fact that MARY , outside the dom ain o f ana, is intended as a focus o f this negative sentence. Interestingly, (75c) illustrates that exactly as in the cases o f questions, (75b) is salvageable by em bedding the sentence excluding its negation part in the no-clause o f no-desu. 41 T his sentence is perfectly acceptable on the reading without association with focus betw een Mary and arm 'not' as in (i): (i) It is M ary that John did not kick. (i) naturally obtains when (75b) is uttered as an answer to a w h-question with a negative presupposition such as (ii): (ii) John -w a dare -o ker-ana-katta? -T op w h o-A cc ker-neg -past ‘W ho didn’t John kick?’ 42 The initial a is an epenthetic vow el that is inserted to avoid a consonant cluster w hen the negative morpheme follow s a verbal stem ending w ith a consonant. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 The contrast betw een (75a) and (75b) can be accounted for alongside o f questions. In particular, let us assum e that (75a, b) are mapped to the LF representation in (76): (76) [iP" [i [a p [v p M ary-o n ] [ A k e n [ A ana-katta]]] I ]] F ollow in g Cho (1993) and many others, I take the negative m orphem e ana to be an adjective. This m aybe is not so controversial since its alternation in terms o f tense (a)na-i 'not-present' / (a)na-katta is exactly the sam e as that o f a regular adjective such as aka-i 'red-present' / aka-katta 'read-past.' T he key feature o f (76) is that given an affixal nature o f the negative m orphem e, the verb ker 'kick' is adjoined to ana-katta. Due do this operation, the seg-com m and dom ain o f anakatta must be confined within the resultant am algam , yielding the contrast betw een (75a) and (7 5 b ).4' A s for the gram m aticality o f (75c), I w ould like to sketch a sim ple account leaving details for the future research. N otice that in this sentence the negative m orphem e stands all by itself, apparently contradicting the thesis that (a)na is an adjectival affix to be incorporated by a verbal stem . F ollow ing Kato (1985, Ch. 3), how ever, I assum e that the "independent" nai in (75c) is actually a PF suppletive 44 (76) contains one additional stipulation that the adjective, which is appended by a tense morpheme katta, does not have to go up to the position o f the L T his postulation, not crucial at this point, w ill be significant in the follow ing argument regarding (75c). Incidentally, if such an assum ption is justifiable, it m ight w ell be so that there is no formal operation o f checking required between an I and a tensed predicate in Japanese. O bviously, then, w e need to make an adjustment regarding the explanation given right below (48). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 form o f ar-anai ("formal verb" + neg), where the negative m orphem e is indeed supported by a "dummy verb" ar. G iven this view , (75c) has a structure as in (77) at a certain point o f derivation: (77) (Irrelevant parts and operations are om itted.) I further stipulate that the ban against a free standing affix is a condition operative in the M orphology com ponent (C hom sky (1994), located betw een Spell-O ut and the PF-interface. If so , the F(ull) I(nterpretation) must delete the dum m y verb ar at LF, since it has no role at that interface level. This deletion operation "simplifies," so to speak, the adjoined structure com prising ar and anai. w hich enables the negative m orphem e to seg-com m and the no-clause and any elem ent inside. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A P [Aanai] 136 3.5. Conceptual and Empirical Implications In this section, let us consider w hat conceptual and em pirical consequences the proposed theory o f w h-scope representation may bring about. In sum , w e are now assum ing that Japanese involves at least tw o distinct kinds o f w h -sco p e assignm ent, schem atized in (78): (78) (The linear order irrelevant.) (78a) presents LF w h-m ovem ent, w hich I argued is a last resort operation for scope assignm ent. The Baker-Pesetsky type w h-scope construal, n ow understood as an instance o f association with focus, is given in (78b). T his option is available only if a wh-phrase generates within the seg-com m and domain o f an interrogative com plem entizer. 3 .5 .1 . T he C om plex N P Constraint and "Pied-piping" M y analysis, just review ed, appears to lend support to a "pied-piping" analysis along the lines o f N ishigauchi (1990). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 Lasnik and Saito (1984, 1992) and N ishigauchi (1990) point out that Japanese, analogous to C hinese (Huang (1982), allow s a wh-phrase in a com p lex NP. This observation is exem plified in (7 9 ):44 (79) (based on N ishigauchi's (57)) a. K im i-w a [n p [ s u dare -o hihanshiteiru] ronbunl-o -Top w ho-A cc criticize article-Acc yom im ashita ka? read Q 'You read an article that criticizes w h o ?' b. [K im i-w a [n p [ eci dare -o hihanshiteiru] ronbun]-o -Top w ho-A cc criticize article-Acc yonda no ] desu ka? read Com p is Q 'It is that you read an article that criticizes w h o ?' T hese questions minim ally differ in that (79a) ends in V(erb)- m asu-ka, whereas (79b) is in no-desu. A problem arises regarding (79a); For dare 'who' is not seg- 44 N aze 'why' must be distinguished. It has been argued in the literature that a w h - question is odd to varying degrees if it contains naze in a com p lex N P. (i) a.?* Anata-w a [N P [John-ga naze kaita] hon]-o yom im ashita ka? you -Top -Top w hy wrote book-A cc read Q ‘Y ou read [a book that John read w h y )] .’ b.?*[A nata-w a [N P [John-ga naze kaita ] hon]-o yonda no] desu ka? ‘It is that you read [a book that John read w h y ]].’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 138 com m anded by ka at LF, the wh-phrase must raise to the matrix C P-spec position. H ow ever, this m ovem ent, if occurred, w ould falsely violate R oss's (1967) “C om plex N P Constraint," w hich I assum e is a condition on m ovem ent operations in general.45 N otice that this problem does not arise in (79b), sin ce dare, seg- com m anded by ka, rem ains in-situ throughout the derivation o f this sentence. The acceptability o f (79a), I assum e, indicates that N ishigauchi's pied-piping operation is indeed available in the language. G iven the theory in chapter 3 o f N ishigauchi (1990) with slight m odifications for expediency, the derivation o f (79a) w ould be roughly as in (80): (80) (Sim plified) a- [CP[C'[IP •••[NP [cP tc'tlP •••[NPtwhl dare] ...]]] N P] ... ]]] b. [cPtc'tlP - [ n p [cP[wh|[Npdarei][c[ip ... ti ...]]] N P] ... ]]] c. [ c p t c t l P - [ N P [vvhl [c p [N P d a r e |][ c [ ip ... tj ...]]] N P] ... ]]] d. [c p [N P l[wh| [CP[NP dareiH c'tip ... t\ • ]]] N P ] [ c [ i p ... ti ... ]]] T hough I have not secured firm judgm ents on these exam ples, I assum e, basically follow in g N ishigauchi, that these exam ples are unacceptable for naze alw ays m oves and it fails to “pied-pipe” a com plex NP. This suggestion is far from innocent on em pirical grounds. For m uch relevant discussion, see Fukui (1 9 8 8 ), Kuno and T akam i (1 993), N ishigauchi (1990) and Lasnik and Saito (1 9 8 4 , 1992). A lso, K oizum i (1991) presents an intriguing argument regarding how deviant exam ples such as those in (i) can be alleviated. 45 A question is left open for the future research w hy this constraint obtains. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 N ishigauchi assum es that a Iw hl-feature “percolates up” to a m axim al projection, w hen its specifier position contains the w h-phrase that bears this feature.46 Left unclear are (i) the exact m echanism o f this “percolation” operation and (ii) the formal nature o f the [wh]-feature. I leave the former matter open for the future research. I shall provide a su ggestion as to the sem antics o f N ishigauchi’s w h-feature. In (80), I take liberties o f accepting N ishigauchi's claim that a "target" o f w h-m ovem ent must be a bearer o f a [w h l-feature. G iven these assum ptions, (80b) first m oves dare to the spec position o f the relative clause CP, and then the [wh]-feature o f dare percolates to this CP. For the relative clause itself is in a specifier position o f the N P, the [w h]-feature further percolates to this N P in (80c). Finally this N P, the bearer o f the fw h l-phrase. undergoes w h-m ovem ent in (80d). Even though dare m oves within the relative clau se, it is not extracted out o f the com plex N P in the course o f derivation, as desired. I need to push this analysis a little further. N otice that (80d) should not be an LF- representation o f (79a) in the light o f my assum ption that w h-m ovem ent is to assign scope to a w h-phrase. G iven a custom ary conception o f quantifier scope, it is not so unnatural to rephrase this assum ption as follow s: (81) W h-m ovem ent is to construct an operator-variable relation. 46 S ee N ishigauchi (1 9 9 0 , 75 - 83) for evid en ce o f this assum ption. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 A problem is that (80d) does not appear to represent the scope o f dare: That is to say, dare, failing to bind a variable, does not enter into an operator-variable relation in this representation. I propose that (80d) is further elaborated as in (82): (82) (= (2 d)) CP [n p i [wh| [c p [n p d a r e iH c t ip ... [NP[wh| darei] ...]]] NP] ■Lnp 1 lwh| [c p [n p d a r e i][c [lP - InpIw 1>i darei I ...III NP (82) assum es C hom sky's (1 9 9 2 ) "copy theory" o f m ovem ent, w here m ovem ent leaves behind a full copy o f the m oved elem ent. I further posit that the [w h]-feature itself may serve as a w h-operator.47 N ow , (82) further coverts to (83): (83) [wh|] •••[n p i [ c p t c t i p - [n p darei] •••]]] NP] ... 47 In this interpretation, the [wh]-feature may be regarded as A . W atanabe (1991) and Tsai's (1 9 9 4 ) covert question operator. A difference is that I am assum ing that the [wh]-feature originates strictly with a w h-phrase. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 141 F ollow ing C hom sky (1992), only the [wh]-feature, an operator, survives in the C P-spec, whereas the other [wh]-features and the copy in the specifier position o f the relative clause are elim inated. Rendering dare as a variable bound to the [wh]- feature. the LF-representation in (83) roughly means (84): (84) For what x, you read [an article that criticizes [x: person (x)]] This seem s to rightly represent the meaning o f (79a). The ungrammaticality o f the exam ple in (85), w hose schem atic pre-Spell-Out structure is given in (8 6 ), naturally follow s from this analysis: (85) ?*Anata-wa [n p [ e c | [Yuki-ga nani -o katta kadooka] shitteiru] you-T op -N om w hat-A cc bought whether know hitoi J -ni aimashita ka? person -to met Q 'You m et a person that knew whether Mary bought what? ' (86) [CP - [CNP ••• [wh-island - M O l ••• ]■•■]•••] In order to pied-pipe the com plex N P, nani ought to m ove to the specifier position o f the relative clause. T his L F-m ovem ent, how ever, necessarily crosses the w h- island, inducing the oddity o f (85).48 48 Consider (i): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 142 Finally, w e m ay m odify, changing nothing essential, the theory o f w h-scope representation pursued in this chapter on the basis o f this "pied-piping" analysis. For instance, it is not so unreasonable that the sim ple w h-questions in (87) and (8 8 ) are associated with the derivations in (89) and (90), respectively: (87) Anata-wa nani -o kaimashita ka? you -T op w hat-A cc bought Q 'You bought w hat?’ (88) Anata -w a nani -o katta no desu ka? you -T op w hat-A cc bought Com p is Q 'It is that John bought what?' (89) a. [c p [C'tlP ••• [[wh] nani] ... ]ka]] b- [c p [(w hl nani] [c '[ip ••• [fwhl nani] ... ]ka]] c. [cp [w h i] [ c [ i p ... [n an ii] ... ]ka]] (90) a. [c p [C'tlP — [^whl nani] ... ]ka]] b .[c p [C'tlP - [n a n ii] ... ]k ai]] (i) ?IAnata-wa Im p! e c i fY uki-ga nani -o katta kadooka] shitteiru] h itoi]-n i atta no] desu ka? 'You m et a person that knew whether Mary bought w hat? ' This exam ple, in no-d esu . parallels (85) in that nani occurs in a structure as in (8 6 ). This exam ple d oes not violate the "wh-island" constraint, for nani d oes not sim ply m ove; Ka, seg-com m anding nani at LF, can represent the scope o f the w h-phrase. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 Plainly nani 'what' is in p ossession o f a [w h]-feature as its idiosyncratic lexical property. In (8 9 ), ka does not seg-com m and nani. H ence this w h-phrase. along w ith its [w h]-feature,49 m oves to the C P -spec position as in (89b). In (8 9 c), only the w h-feature survives at the C P-spec position, w hereas the other [wh]-feature gets d eleted to create an operator-variable relation. In (9 0 ), ka seg-com m ands nani. H ence the scope o f nani is represented w ithout m ovem ent. The wh-feature o f nani is sim ply om itted in (90c). 3 .5 .2 . Extractions o f C onjoined W h-phrases Out o f a W h-island C onsider the follow in g contrast:50 (91) ?rD are-ni1 keisatsu-w a [T om -ga t| fsoitsu i-no im a no koibito]-o w ho-to p olice-T op -N om he -G en now -G en lover-A cc shookaishita kadooka] shirabeteiru no] desu ka? introduced w hether investigate C om p is Q 'T o w hom i the police is investigating [whether T om t| [his i current lover] introduced.1 49 If one w ish to adhere to C hom sky's (1 9 9 5 ) theory o f Attract / M ove, w here a feature, not a category m oves after Spell-O ut, one can assum e only this feature m oves at LF. 50 T his contrast has been suggested to m e by H ajim e Hoji (p.c.). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 (92) * * fDare -to dare -n il keisatsu-w a [T om -ga t, [soitsu r no im a no koibitoj-o shookaishita kadooka] shirabeteiru no] desu ka? T hese exam ples are identical except for the forms o f the w h-phrases extracted from the w h-islands by scram bling; dare-ni 'who-to' in (91) and dare-to dare-ni 'w ho and w ho-to' in (92). T hese w h-phrases bind a pronoun soitsu in the both exam ples. (91), albeit slightly m arginal, falls on the side o f acceptability. This is expected, since the "wh-island constraint" only regulates m ovem ent operation triggered for a scopal reason. On the other hand, (92) is hopelessly unacceptable, sharply contrasting w ith (91). (93) generally holds: (93) A sentence is absolutely unacceptable, w hen conjoined w h-phrases that bind a pronoun are extracted out o f a w h -islan d .51 51 A s Hajim e Hoji pointed out to me (p.c.), this generalization may very w ell reduce to Cinque's (1 9 9 1 , Ch. 1) observation that a "non-referential" expression cannot be extracted from "weak islands" including w h-islands. Dare-to-dare in (92) is purely quantificational, and hence non-referential. Thus it cannot be extracted from the w h-island. N ote that a group interpretation o f dare-to-dare is excluded in (92). for the N P licen ses a singular bound pronoun soitsu . If this is correct, it should be that dare in (91) is referential, and it "corefers with" soitsu in som e sense. I leave details for the future research. For m uch relevant discussion, see Cinque (1 9 9 1 , Ch. 1) and Saito (1994). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 145 G iven this generalization, consider the contrast between (94) and (95): (94) ?[K eisatsu-w a [T om -ga dare-to dare-ni■ soitsupno ima no koibito-o shookaishita kadooka] shirabeteiru no] desu ka? 'It is that [the p olice is investigating (whether T om to whorrn [his i current lover] introduced].' (95) **K eisatsu-w a [T om -ga dare-to dare-ni. [soitsu r no ima no koibito]-o shookaishita kadooka] shirabeteimasu ka? 'The police is investigating (whether Tom to w h o m ; [hisj current lover] introduced.' (94) is acceptable, w hereas (95) is far worse than a mere violation o f the "wh-island constraint." In the both exam ples, dare-to-dare 'who and who,' binding soitsu 'he,' is within a w h-island. T he acceptability o f (94), w hich is in no-d esu . follo w s from the proposed theory o f w h -scop e representation. For dare-to dare is in the seg- com m and dom ain o f ka at LF, this conjoined N P does not m ove to the matrix CP- spec position. T herefore the generalization in (93) is just irrelevant. In (95), on the other hand, dare-to-dare. falling outside the seg-com m and dom ain o f ka at LF, must m ove across the w h-island. H ence (95) is ruled out on a par w ith (92). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 146 3.5.3. Some Copula Sentences A s it w as discussed earlier, besides the no-clause o f no- d esu . virtually any predicate im m ediately preceding a copula and the elem ents therein may fall under the seg-com m and dom ain o f an interrogative com plem entizer at LF. Thus, the contrast betw een (96) and (9 7 ) obtains: (96) * [cp John-ga T O M -ni shookaishita no] -ga [n p N ancy-no okaasan] -N om -to introduced C om p -N om -G en m other desu ka? is Q 'Is it N ancy’s mother that John introduced to TOM?' (9 7 ) [cp John-ga T om -ni shookaishita n o]-w a [n p N A N C Y -no okaasan] -N om -to introduced C om p-T op -Gen m other desu ka? is Q 'Is it N A N C Y 's m other that John introduced to Tom?' T om , outside the seg-com m and dom ain o f ka at LF, is focused in (96) and, hence, this sentence is ruled out. In contrast, (97) is grammatical, since the intended focus N ancy contained in the N P right before the copula, falls within such a dom ain. G iven this observation, the proposed theory o f w h-scope assignm ent can readily explain the contrast between (98) and (99): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 147 (98) *?[cp John-ga [ec][cp M ary-ga nani-o katta kadooka] tazuneta -N om -N om w hat-A cc bought w hether asked no] -ga T om desu ka? C om p-N om is Q 'It is T om that John asked w hether Mary bought what? ' (99) [cp John-ga T om -ni [ec] tazuneta no] -w a [cp M ary-ga -N om -to asked C om p-T op -N om nani -o katta kadooka] desu ka? w hat-A cc bought whether is Q 'It is whether Mary bought what John Although the both exam ples em bed nani 'what' in the w h-islands. only (98) appears to be in violation o f the "wh-island condition." On the view that I esp ou se here, this is naturally expected, for nani falls under the seg-com m and dom ain o f ka at LF in (9 9 ) but not in (98). Nani in (9 9 ) d oes not have to m ove, sin ce the B aker-Pesetsky type w h -scop e assignm ent is available in this exam ple. On the other hand, in (98), nani. outside such a dom ain, raises to the matrix C P-spec position to take scope, yielding a violation o f the "wh-island constraint." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 148 3.5.4. Another Type o f Wh-island Violation In no-desu the scope o f a wh-phrase can go beyond another type o f w h-island that w as not d iscussed in the preceding (sub-)sections C onsider the exam ples in ( 100):52 ( 100) a. A nata-w a [dare__-ga nani -o katta ka] shitteim asu ka? you-T op w h o-N om w hat-A cc bought Q know Q b. [A nata-w a [dare -ga nani -o katta ka ] shitteiru no ] you-T op w h o-N om w hat-A cc bought Q know C om p desu ka? is Q T he novelty in this observation is that the low est clauses, headed by ka, contain tw o w h-phrases dare 'who' and nani 'what.' T he matrix clauses being also headed by the interrogative com plem entizer, the w h-phrases. in principle, can take scop e over either the em bedded clause or the matrix clause. ( 1 0 0 a,b) are distinguished only by the fact that the latter is in no-desu. w hile the former is not. On the present account, this by-now fam iliar syntactic difference should result in significant sem antic disparities betw een ( 100a, b). T his prediction seem s to be borne out. 52 For m uch related discussion, see Saito (1989a), N ishigauchi (1 9 9 0 , Ch. 1 and Ch. 2) and A . W atanabe (1 9 9 1 ), to nam e just a few . Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 ( 1 0 1 ) sum m arizes the interpretive possibilities o f the exam ples by dotted lines. The schem e should be understood in such a way that the scope o f a vvh-phrase associated with the "inner" ka is the em bedded clause, w hereas that o f a w h linked w ith the "outer" ka is the matrix: ( 1 0 1 ) ( 1 0 0 a) a. dure nani ka ka ( 1 0 0 b) V b. dare nani ka k; a ^ n i c. dare nani ka k« am ka 1_ 1 d. dare nani ka k. ant ka kp ( 1 0 0 a) is acceptable only on the reading in ( 1 0 1 a), where the both w h-phrases are construed with the em bedded clause to yield a matrix ves-n o question interpretation ( 1 0 0 b), on the other hand, allow s all o f the interpretive possib ilities except for the reading in ( 1 0 Id). Under m y analysis o f w h -scop e construal, the LF-representations o f (100a) w ould be as in ( 1 0 2 ): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 150 (102) a. [cp [c'[ip - -[cp dare I-nani-) [ c l i p ••• li (2 ••• ]ka]]...]ka]] (=( 1 0 1 a)) b. *?[cp d a re|- nani2 [ c ' [ i p . [ c p [ c [ i p ••• l l - 12 ... ]ka]]...]ka]] (=( 10 1 b)) c. * ? [c p d a r e i[c ,[lP.. [cPIlM i2[C'[lP ••• t|... t? ... ]ka]]...]ka]] (=( 101c)) d *[cpnani2 [ c [ ip ...[ c p d arei [ c [ i p ... t |... t2 ... ]ka]]...]ka]](=( 101d)) O bviously ( 102a) is gram m atical, involving no extraction from the w h -island. In this representation, dare and nani are m oved into the low er C P-spec position for scope assignm ent. N otice that these m ovem ent operations are necessary, since the w h-phrases. not in the no-com plem ent, fall outside the seg-com m and dom ain o f ka at LF. In ( 102b), where the w h-phrases are interpreted in the matrix clause, the em bedded sentence is regarded as a ves-n o question. On this reading, the low er ka is an abbreviated form o f kadooka 'whether.' Crucially, the w h-phrases raise to the matrix C P-spec, violating the "wh-island constraint." Sim ilarly, (102c, d) are excluded due to the illicit m ovem ent o f either one o f the w h-phrases. (100b) are associated with the LF-representation in (103) on the account: (103) (sim plified) a . [ c p [ c ' ••[cp .[cpdarei-nani2[C '[lP ••• !l - 1 2 lk a ]]...n o ]...]k a ]](= ( 101a>) b .? [cp [C ' .[c p .[c p [c '[lP ••• darei...n ani2 ... ]k a]]...n o]...]k a< 1,2]](= ( 101b )) c .? [ c p [ c ' .[c p .[cp n am 2 [ c [ i p - dam i ... t2 ... ] k a ]]...n o ]...]k a i]](= (io ic )) d . * [ c p [ c •••[cp. .[cp darei [c tlP ••• U - nani? ...]k a]]...n o]...]ka 2]](= ( 101d )) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 i ( 103a), a w ell-form ed matrix ves-n o question, is exactly the sam e as ( 102a) in the respects o f current relevance. (100b), (103b) illustrates, allow s an interpretation w here the both wh-phrases are construed with the matrix sentence, rendering the em bedded clause as a ves-no question. O bviously, the matrix ka m akes this construal available by virtue o f representing the scopes o f dare and nani. in its seg - com m and dom ain at LF, without m ovem ent. (103c) is also gram m atical reoccurring to the sam e strategy; nani takes scope over the matrix sentence by utilizing the matrix interrogative com plem entizer. A puzzle is w hy it is that ( 103d) should be ruled out. O nce the non-m ovem ent option o f w h -scop e marking is admitted, there should be nothing deviant about this representation. A tentative account can be drawn from Saito (1989a) and A . W atanabe (1991, Ch. 1),” w h o invoke the P(ath) C (ontainm ent) C (ondition) o f Pesetsky (1982) but retreating to the older definition in terms o f linearity, (cf. Kuno and Robinson (1 9 7 2 ) and Fodor (1978)). Let us assum e the informal definition in (1 0 4 ), cited from A W atanabe (1991, 19): (1 0 4 ) If tw o paths overlap, one must contain the other. ^ For m uch relevant discussion, see Kurata (1991). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The account further requires a speculation that a specifier projects rightward in Japanese. On these assum ptions, (1 0 3 c, d) are rewritten as follow s: (1 0 3 ’) C .? [c p [c ---[c p -- [c p [c '[ip - darei ... ]ka] nanj23-..n o ]...]k a i ]](= ( 1 0 1 c)) d . * [ c p [ c ’- - - [ c p - [ c p [c tlP -j-tl- na.ni.2 ... ]k a]d arg|]...n o]...]tja2 ]](= ( 1 0 1 d)) W hat ought to be considered here is the "linear association lines" betw een the w h - operators (darei and ka2 ) and the variables (t| and nani-;). Suppose that such a line, a conventional notation, counts as a path in the sense o f (104). N o w the difference betw een (1 03c') and ( 103d’) exp licitly surfaces; the former conform s to (104), w hile the latter, containing crossing paths, is ruled out by the con d ition .54 3.6. C on clu sion T his chapter argued that depending on w here a wh-phrase originates, its scope is represented by either one o f the tw o distinct options schem atically sh ow n in (105): 54 T his approach, where a liner relation plays such a pivotal role at LF, is not congenial to the M inim alist theory. C hom sky (1994) argues that K ayne's (1993) L(inear) C(orrespondence) A (xiom ), w hich states that asym m etric c-com m and im poses a linear ordering o f terminal elem ents, applies only at PF, leavin g LF blind to linear relations. O bviously, if m y argument is correct, it is necessary to assum e the LCA at LF so as to refer to linear relations at the interface level. T h is extension, how ever, aw aits further conceptual and em pirical justification. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 153 (1 0 5 ) a, w h b Traditional LF w h-m ovem ent. em ployed in ( 105a), applies only w hen this is the last resort option to represent the scope o f a wh-phrase. E xplicitly assum ed is that w h -m ovem en t in Japanese is solely m otivated by a scopal n ecessity, and, therefore, its application has no bearing on m orphological checking. T his proposal, if correct, indicates that unlike the recent attempt in Hornstein (1994, Ch. 8 ), w e may not totally deprive the theory o f grammar o f operations such as Q R , thereby giving support to Fox (1994). S econ d , the scope o f a w h-elem ent is representable through "unselective binding" by an interrogative com plem entizer without m ovem ent (= ( 105b)). This scope representation is possible only if a wh-phrase falls within the seg-com m and dom ain o f an interrogative com plem entizer. This structural condition indicates that the relation betw een a wh-phrase and an interrogative com plem entizer is association with focu s in the sense o f Jackendoff (1972). N eed less to say, this analysis strongly supports C. L. Baker (1 9 7 0 ) and Pesetsky (1987). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 154 Chapter Four "Stylistic" Restrictions on Q uestions in Japanese: Descriptions and a Possible A venue Toward Explanation 4 .1 . Introduction The last chapter discussed how w h-phrases in Japanese are assigned scope in different linguistic contexts. W e observed that the language allow s questions in (at least) tw o distinct constructions, i.e. a "bare form" as in ( la) and n o-desu as in (lb ): ( 1) a. John-w a nani -o kaimashita ka? -T op w hat-A cc boughtpoiite Q 'John bought w hat? ' b. [John-w a nani -o katta no] desu ka? -T op w hat-A cc bought C om p ispoiite Q 'It is that John bought w hat? ' Putting aside sem antic / pragmatic details, these questions are analogous in their essential functions as w h-questions. That is to say, the questioner asks the respondent to nam e the object(s) that John bought, (la ) and ( lb ) are associated with the structures (2a) and (2b), respectively, before Attract / M ove applies: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 155 (2a) (The position o f the wa-phrase is inaccurate, but irrelevant.) CP [ c ka [n p John-w a] I' V P I [n p nani-o l [y kaimashita] (2 b) CP V P John-w a nani-o katta no Provided that verb / predicate raising takes place in interrogative sentences, 1 kaimashita 'bought' in (2 a) and the "com plex predicate" in (2 b) m ove to the 1 T hough I leave open w hether or not verb-raising o f this sort is lim ited to questions, it is likely that such an operation may occur even in declarative sentences. C onsider the exam ple in (i), w hich in volves "Gapping:" (i) John -ga koohii -o soshite, T om -ga koocha -o nonda. -N om co ffee-A cc and -N om blacktea -A cc drank 'Tom drank tea, and John c o ffe e .’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 156 position o f ka and adjoin to this com plem entizer. Due to this operation, nani 'what' in (2a) ends up outside the seg-com m and dom ain o f ka. In contrast, nani in (2b), contained in the no-clause, falls within this dom ain o f ka. A ssum ing that (i) LF w h-m ovem ent is a last resort operation to assign scope to a w h-phrase. and (ii) ka can only mark the scope o f a wh-phrase that it seg-com m ands, the last chapter proposed that LF w h-m ovem ent applies in (2a) but not in (2b). Japanese questions must satisfy a set o f fairly com plicated but rather intriguing "stylistic" restrictions. O ne striking fact, w hich S. M iyagaw a (1987) initially notes and U eyam a (1 9 9 1 , 1992) later discusses in depth, is that the polite verb forms used in ( 1) cannot be replaced by their neutral counterparts: In the first conjunct o f (i), nonda 'drank' is gapped "stranding" the subject and the object N Fs. F o llo w in g Johnson's (19 9 4 ) analysis o f E nglish "Gapping" and K oizum i's (1 9 9 5 , Ch. 7) suggestion regarding a sim ilar phenom enon in Japanese, it is conceivable that this sentence has a Pre-Spell-O ut structure like (ii): (ii) [Np John-ga] [n p koohii-o] ^ I P ^ t v i nonda] [NP T om -ga] I' VP [N P koocna-o] The verb vonda undergoes A (cross) T (he) B(oard) head-m ovem ent, crucially here, to the C. T o the extent that this analysis is viable, a verb at least can raise and adjoin to the C in a declarative sentence. D etails are open for future discussion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 157 (3) a. (based on S. M iyagaw a's (4)) *John-w a nani -o katta ka? -Top w hat-A cc boughtneutral Q 'John bought w hat? ' b. * [John-w a nani -o katta no] da ka? -T op w hat-A cc bought Com p i s n e u tral Q 'It is that John bought w hat? ' T hese lexical changes equally render the tw o types o f questions rather m arginal.2 2 T w o com m ents are in order. First a w h-question o f the form in (3a) can be acceptable in a rather peculiar situation. In particular. (3a) is felicitous, w hen, as in playing a riddle (= (i)), the speaker know s the answ er o f the question and asks the respondent(s) to guess what the correct answ er m ight be: (i) John -w a naaani -o katta kaa? ( Atete-te g o ra n .) -T om w hat -A cc bought Q ( G u e s s .) 'What did John buy?' This context appears to require o f (i) to have an unusual rising intonation on ka, and lengthening on the m edial vow els o f ka and nani. I shall exclude as an exception this uncustom ary use o f w h-questions o f the kind. Second, the "stylistic restriction" in (3b) is observed in a sim pler "copula sentence" as in (ii): (ii) * Dare -ga [John-no koibito] da ka? w h o-N om -G en lover isneutral Q 'W ho is John's girlfriend?' W hatever restrictions hold for questions in no-da / desu. they basically regulate questions o f any forms w ith copulas as w ell. I shall, nonetheless, make relevant observations in questions in no-da / desu so as to com pare them with those in "bare” form. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 158 Extending the discussion to the arena o f ves-no questions, a conspicuous disparity reveals itself betw een the tw o types o f questions: (4) a. John-w a hon -o katta ka? -T op book-A cc boughtneutral Q 'John bought a book?' b. *[John -w a hon -o katta no] da ka? -Top book -A cc bought Com p isneutral Q 'It is that John bought a book?' O bviously, these sentences are distinguished from those in (3) only by nani 'what' being replaced by hon 'book' so as to obtain ves-n o questions. The result is that (4a) is acceptable, w hile (4b) is still out. A ssum ing that these traits o f Japanese interrogative sentences, am ong many others presented b elow , are to be explained in syntax,’ the purpose o f this chapter is three-fold. First, I shall su ggest that S. M iyagaw a's (1 8 9 7 ) "governm ent restriction" on ka explains parts o f intricate structural restrictions on Japanese w h - questions. In particular, I w ill argue that ka in a w h-question must be "governed" by an X °-category. Second, unlike S. M iyagaw a, I w ill su ggest that ves-no questions are im m une from this constraint. The relevant em pirical departure from S. M iyagaw a w ill be accounted for by invoking the notion o f the "domain" o f an 3 U eyam a (1992. 1 9 - 2 0 ) proposes to exclude sentences like (4a) on pragmatic grounds. Her argument h ow ever is not entirely clear to m e at this point, w hich keeps me from reproducing it here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 159 interrogative com plem entizer discussed in the last chapter. That is to say, I shall maintain that deviant exam ples that are excluded by the "government" restriction in S. M iyagaw a (ibid.) are cases where foci fall outside the seg-com m and dom ains o f interrogative com plem entizers. Third, I shall suggest that a ves-n o question is obedient to a rather different constraint from the "government" restriction on w h - questions. In brief, it w ill be maintained, that the deviance o f such a sentence as that in (4b) follow s from an assum ption that a copula verb da 'is' is "incompatible," in a sense to be defined, with an interrogative com plem entizer.4 The organization o f this chapter is roughly as follow s. Section 4. 2 w ill show fundamental properties o f w h-questions in the "bare" form. Section 4.3 suggests, follow ing S. M iyagaw a, that an overt com plem entizer ka in a w h-question ought to be "governed" roughly in the sense that it must be em bedded in a larger constituent. Section 4. 4 w ill point out that ka in a ves-n o question does not have to satisfy this structural requirement. The data w ill be treated on independent grounds that apparently indicate that ka in a ves-n o question is subject to the sam e constraint as the one in a w h-question. Section 4. 5 w ill provide further em pirical evidence o f these claim s with interrogative sentences in no-da / d esu . S ection 4. 6 w ill show a peculiar behavior o f a copula da 'is' in ves-n o questions, and Section 4. 7 w ill provide a tentative solution. Section 4. 8 w ill be a short additional note on ves-no questions. Section 4 .9 w ill contain a brief summary and a concluding remark. 4 I w ill later argue, how ever, that da is not actually a copula. Until that point o f discussion, I shall term this elem ent as a "neutral copula" for the sake o f reference. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 160 4.2. Basic Facts F ollow in g S. M iyagaw a (1987) in essen ce, this section w ill present fundamental observations concerning w h-questions in Japanese in the "bare" form. Let us start w ith considering the deviant wh-question in (4a), repeated b elow in (5): (5) *John -w a nani -o katta ka? - T o p w hat-A cc b o u g h t n e u tral Q 'John bought w hat?' For the purpose o f exposition, I generalize this observation as in (6 ) : 5 (6 ) A matrix w h-question with ka is ill-form ed, w hen it ends with verb is in its neutral form. (6 ) holds for a w ide range o f w h-questions. Som e exam ples are given in (i): (i) * Dare -ga hon -o katta ka? w h o-N om book-A cc bought Q 'W ho bought a book?' (ii) * Dare -ga nani -o shita ka? w h o-N om w hat-A cc did Q 'W ho did what? (iii) *John-w a dare-ni yubia-o ageta ka? -T op w ho-to ring-A cc gave Q T o w hom did John g ive a ring?' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 161 A s is partly im plied in (6 ), the exam ple (5) is rem edied in several w ays. First, (5) can be made perfect sim ply by deleting ka: (7) (based on S. M iyagaw a's (1)) John-w a nani -o katta < j > ? -Top w hat-A cc bought Q 'John bought w hat? ' T his sentence, as it seem s natural to assum e, contains a phonologically null interrogative com plem entizer. N otice that this im provem ent strongly indicates that ka is closely related to the deviance o f (5). Second, em bedding (5) in a larger constituent saves this exam p le:6 (B) a. (based on M iyagaw a's (6 )) [John-ga nani -o katta ka] (-o) shiri -tai. -N om w hat-A cc bought Q ](-A cc) know -w ant 'I want to know [John bought w h at!.' b. [John -ga nani -o katta ka] -ga mondai desu. -N om w hat-A cc bought Q -N om problem is '[John bought what] is an issue.' 6 John-w a 'John-Top' is replaced with John-ga 'John-N o m .' for a wa-m arked N P is only m arginally allow ed in an em bedded clause. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission 162 (8 a) em beds (5) as a subordinate clause o f shir 'to know.' A lthough S. M iyagaw a does not discuss a case where a w h-question o f the sort is em bedded as a sentential subject, this strategy too salvages (5), w hich (8 b) indicates. Third, w e can rem edy (5) by adding what S. M iyagaw a calls "sentential particles" such as naa 'I wonder' and nee 'I wonder.' (9) (based on S. M iyagaw a's (19)) [John-w a nani -o katta ka] naa / nee? -Top w hat-A cc bought Q I-wonder 'I w onder [John bought what]?' A ssim ilating this phenom enon to that in (8 ), I regard attachment o f these particles as a special case o f em bedding. That is to say, the sentential particles are considered verbal elem ents taking w h-questions as com plem ents. T h is assum ption does not seem so unreasonable, based on the meanings o f these particles indicated in the g loss. Fourth, (5) can be rem edied by sw itching katta for its polite counterpart kaim ashita: (1 0 ) ((= 1 ), based on M iyagaw a's (5)) John -w a nani -o kai-m ashita ka? -T op w hat-A cc buy-poiite-Past Q 'John bought w hat?' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 163 A s Hurada (1976) d iscusses in considerable depth, a polite verb form has a perform ative function as an addressee honorific. Later, I w ill turn to this semantic property in its relation to syntactic representations. 4.3. A "Government" Requirem ent The observations in the last section indicate that S. M iyagawa's (1987) generalization in (11) holds for interrogative sentences in Japanese: (11) (based on (8 ), S. M iyagaw a (1987)) Ka. an interrogative com plem entizer, must be governed. Borrow ing only the essen ce o f "government" from the standard G B -fram ew ork (C hom sky (1 981), (1986b )), let us take this generalization to state, in effect, that ka must occur in a structure som ething like that in ( 1 2 ): There should be an X °-category that c-com m ands ka, and any m axim al projection intervening X and ka (in (12), only the CP) must not be a "barrier." Let us sim ply postulate that a m aximal projection is a "barrier" if it prohibits m ovem ent out o f it. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ( 12) X P 164 W hy is it that (11) has to hold?7 At this m om ent, I am not capable o f submitting any definitive answer to this question, and. I shall reluctantly leave (11) as a descriptive generalization in this chapter. One stipulative, but perhaps plausible, approach is to assum e that for an unknown reason, ka must raise and adjoin to a higher X ()-category, 8 converting (12), for exam ple, to (13): (13) E c i ka] The CP must not be a "barrier," or this m ovem ent operation w ould be blocked. W hether or not this is a correct w ay to deduce (11), I hope to show that the generalization in ( 1 1 ) itself is firmly confirm ed on empirical grounds. S. M iyagaw a proposes that (i) and (ii) are accounted for by the sam e condition or principle: (i) *[ [c 0 ] John is happy ] is obvious. (ii) ( = (5 ) ) * John-w a nani-o katta ka? (i) exem p lifies a w ell-know n fact in English that a null declarative com plem entizer may not occur in an “ungovem ed” position. The m ost influential account o f this phenom enon, given by Stow ell (1981), is to rule out this sentence as an ECP violation by assum ing that the null com plem entizer must be “lexically governed,” for it is covert. It appears, how ever, unlikely to extend this analysis to (ii), sim ply because ka is by no m eans covert. 8 If this operation indeed applies, it m ay, at least, procrastinate until LF, since an indirect w h-question freely scram bles as in (i): (i) (based on (8 a)) [ c p i John-ga nani -o katta ka] boku-w a t( shiritai. -N om w hat-A cc bought Q I -Top want-to-know Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 165 First consider the ill-form ed sentence in (5). repeated in (14): (14) (= (5 )) *John -w a nani -o katta ka? -T op w hat-A cc bought Q 'John bought w hat?' G iven (11), the unacceptability o f this sentence is sim ply due to the fact that there is no X ()-category "governing" ka. Second, the im provem ent in acceptability in (8 ), repeated in (15), is rather expected with ( 1 1 ) being assumed: ( 1 5 ) (= (8 a )) [John-ga nani -o katta ka] (-o) shiri-tai. -N om w hat-A cc bought Q ](-A cc) know -w ant 'I w ant to know [ John bought what ].' (1 6 ) (= (8 b)) [John-ga nani -o katta ka] -ga m ondai desu. -N om w hat-A cc bought Q -N om problem is '[ John bought w hat ] is an issue.’ In (15) ka is "governed" either by shir 'know' or shiritai 'w ant-to-know ,' depending on how one analyzes the latter verbal com p lex.9 In such a case as (16), ‘[W hat John bought] I want to k n ow .’ In this case, ka raises after the scram bling o f the indirect w h-question is undone at LF. 9 S. M iyagaw a (1 9 8 7 ) contrasts such sen ten ces as (15) w ith (i), w hich he finds unacceptable: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 166 where an "illicit" w h-question is em bedded as a sentential subject, a likely "governor" is a covert com plem entizer or an I(nfl) (if the nom inative case marker ga is assigned V P-internally).'H (i) (B ased on S. M iyagaw a's (11), The judgm ent is M iyagaw a's.) *B ill-ga [cpJohn-ga dare-ni atta ka] donnata/sasayaita/tsubuyaita (koto) -N om -N om w ho-to met Q shouted/whispered/m urm ured 'Bill shouted / w hispered / murmured [ w h o John m et ]' He argues that (i) is ungrammatical, because the "barrierhood" o f the em bedded CP keeps ka from being "governed" by the m anner-of-speaking-verbs. T he C P is a "barrier," since clauses subordinate to this class o f verbs in general do not allow extraction from them: (ii) ‘ ??T om -nii B ill-ga [cpJohn-ga t] atta to] donatta/sasayaita/tsubuyaita(koto) -to -N om -N om m et C om p shouted/whispered/m urm ured 'T om i . Bill shouted / whispered / murmured [ that John met tj ]' What puzzles me at this point is that (i) is not so bad as S. M iyagawa's acceptability judgm ent o f this sentence. Although rather marginal, it appears to me that (i) falls on the side o f acceptability. If this is indeed the case, the generalization in ( 1 1) is too strong. 10 N otice that (8 b) illustrates that a subject, as w ell as an object, does not constitute a barrier in this language. This appears unproblematic, since extraction is possible either from an object or a subject: (i) (based on Saito's ( 1992b) (85)) a. [ip John-ga [N P tlP M ary-ga ano hon -o katta ] koto] -o mondai -ni -N om -N om that book-A cc bought fact-A cc problem - to shiteiru] (koto) making (the fact) 'John is m aking an issue out o f [the fact [that Mary bought that book! ]' b.?rtP A no h o n -o i [ip John-ga [N P tlP M ary-ga tj katta ] koto] -o m ondai w hat-A cc -N om -N om bought fact -A cc problem -ni shiteiru] (koto) -to making (the fact) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 Third, let us turn to (17), w hich requires an additional assum ptions: (17) (= (1 0 )) John -w a nani -o kai -m ashita ka? -T op w hat-A cc buy-p0iite-Past Q John bought what ?' The use o f the polite form kaimashita. instead o f katta. drastically im proves (14). T his fact indicates, in the present discussion, that the "government" requirement is 'That book l . John is m aking an issue out o f [the fact [that Mary bought 11 ]]' (ii) (based on S aito’s ( 1992b) (8 6 )) a. [jpJohn-ga [C P [N P [lP M ary-ga ano hon -o katta] koto]-ga m ondai-da -Norn -Norn that book-A cc bought fact-N om problem is to] om otteiru] (koto) C om p think (the fact) 'John thinks [that [the fact [that Mary bought that bookll is a problem ].' b. ?rTPA n o h o n - o i [ipJohn-ga [C PtN PtlPM ary-ga t( katta] koto]-ga that book -A cc -N om -N om bought fact-N om m ondai-da to] omotteiru] (koto) problem is C om p think (the fact) 'That b o o k i . John thinks [that [the fact [that Mary bought t[] ] is a problem].' (ib) in volves scram bling out o f a com plex N P which is an object. T his sentence is marginal, since the adnom inal m odifier clause, taken as an IP follow in g Murasugi (1 9 9 1 ), is an adjunct. A case o f interest is (iib), extracting ano hon-o (that book- A cc) from a com p lex noun phrase that is a subject. A gain, the operation results in m arginality induced by the barrierhood o f the relative clause. N otice, how ever, that there appears to be no detectable difference in acceptability betw een (ib) and (iib). T his fact strongly indicates that a subject is not a barrier in Japanese. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 168 satisfied in (17). H ence w e need to assum e that (17) is associated w ith a pre-Spell- Out structure som ething like (18): (18) [flp tc P - tc tlP John-w a nani-o kaim ashita ] ka ]] Q ] (18) em beds a w h-question as a com plem ent o f a provisional covert functional category Q . A s S. M iyagaw a (1 9 8 7 , 365) points ou t, 11 w ho o w es the suggestion to Harada (1 9 7 6 ), a polite verb form such as kaimashita is associated w ith a perform ative function o f marking the w hole sentence w ith "politeness," and I speculate that it is Q , c-com m anding the entire sentence except for its ow n projection, that syntactically represents this interpretation. I take a polite verb form as a polarity expression w hich ought to occur in the scope o f Q at LF. Provided that this argum ent is correct, the "government" requirement is satisfied in (1 8 ) due to the existence o f Q . 4 .4 . Y es - N o Q uestions On descriptive grounds, I w ill suggest in this section that Japanese ves-no questions do not obey the "government" restriction discussed in the last section. In " S. M iyagaw a claim s that (17) is grammatical, because the politeness affix masu raises at LF to the position that "governs" ka. T his assum ption, how ever, crucially relies on P esetsky's (19 8 5 ) proposal o f "LF affix raising," w h ich I do not intend to defend here. For the sake o f fairness, there is no difference betw een S. M iyagaw a’s proposal and m ine that is o f substance. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 169 this very respect. I w ill depart from S. M iyagaw a, w ho proposes that the given structural requirement must be m et by y es - no questions as w ell as w h -questions. The data that seem ingly support his claim , I w ill argue, should be treated independently o f the "government" requirement. Let us begin w ith the exam ples in (19) and (20): (19) * Dare -ga hon -o katta ka? w h o-N om b ook -A cc boughtneuiral Q 'W ho bought a book?' (20) *John-ga hon -o katta ka? -N om b ook-A cc boughtneutral Q 'Is it John that bought a book? ’12 T hese sentences, w hich I find both unacceptable, are distinguished o n ly by dare 'who' in (1 9 ) being replaced by John in (20). Based on the deviance o f (20), one might argue that ka in a y es - qq question is subject to the "government" requirement on ka. H ow ever this view does not seem to be em pirically correct. C onsider the contrast betw een (2 1 ) and (2 2 ), w hich w ould be rather surprising if (2 0 ) were ruled out on a par with (19) by the "government" restriction: 12 I w ill shortly explain w hy John is translated as a focus here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 (2 1 ) ('0 ' stands for a null com plem entizer.) Dare-ga hon-o katta 0 ? 'W ho bought a book' (22 ) *John-ga hon-o katta 0 ? (21) and (22) m inim ally differ from (19) and (20), respectively, in that ka is replaced w ith its covert counterpart. O bviously, the w ell-form edness o f (21) indicates that this covert com plem entizer does not need to meet the "government" requirement under discussion. H ence the ungrammaticality o f (2 2 ) should not follow from the fact that the null com plem entizer is "ungovemed" in this exam ple. It seem s then natural that whatever renders (22) ungrammatical is.also responsible for the oddity o f (2 0 ). I w ould like to suggest here that the exam ples in (20) and (22) are marginal, for they contain a focus outside the seg-com m and dom ain o f an interrogative com plem entizer. Recall our discussion from the last chapter that a Japanese interrogative sentence raises a verb and adjoins it to ka or its null counterpart, and that due to the resultant adjoined structure, the com plem entizers can seg-com m and only the verb adjoined to them . In this view , the LF-representation o f (20) and (22) is as in (23) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 ( 23) C ’ IP C [Np John-ga] I' [v i katta] { [c ka] / [c 0 ]} V P [n p hon-o] ti T he further assum ption is that John-ga. marked with the nom inative case marker ga, is an "obligatory focus" (K uroda (1992. Ch. 1)). This is no so unreasonable, because this ga-m arked N P is necessarily rendered as a focus in (24), a declarative counterpart o f (2 0 ) and (2 2 ): (24) (K inoo) John -ga hon -o katta. (Y esterday) -N om book-A cc bought 'It is John that bought a b ook yesterday.’ G iven that John-ga is an obligatory focus, (23) (hence (20) and (22)) is ungram m atical, since the representation contains a focus outside the seg-com m and dom ain o f the interrogative com plem entizers. 1 3 13 This discussion o f the "obligatory focus" o f N P -g a is rather rough and far from com plete. N ote that the contrast between (i) and (ii) is problem atic for what has been just claim ed in the text: (i) (based on Kuroda (1 992, Ch. 1)) Tam a -ga isu-no-ue-de nem utte iru. -T om chair-on sleeping be 'Tama is sleeping on a chair.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 172 O nce vonda 'read' is legitim ately interpreted as a focus. (20) and (22) becom e perfectly acceptable: (25) John -w a hon -o KATTA ka? -T op b ook-A cc bought Q 'Did John B U Y a book?’ (26) John-w a hon -o K A TTA 0 ? -T op b ook-A cc bought Q ’D id John B U Y a book?’ The "offending" nom inative phrase is replaced by the topic phrase John-w a in these cases, and KATTA should be read w ith an em phatic stress to attract a focus o f question. Lacking a nom inative phrase, this focus interpretation, in consonant w ith (23), is thus allow ed. N otice, in particular, the gram m aticality o f (25) w ould be totally m ysterious if the com plem entizer ka in a ves-n o question were subject to the "government" requirement. S. M iyagaw a (1987, 3 6 5 -3 6 6 ), w h o argues that ka in a ves-n o question must also be "governed," show s the follow in g paradigm: (27) (S . M iyagaw a’s (2 1 ), slightly m odified) a.*H anako -ga kuru ka? -N o m com e Q 'W ill H A N A K O com e?' (ii) ??Tam a -ga isu-no- ue-de nemutte iru ka? 'Is Tam a sleeping on the chair?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 173 b. [Hanako -ga kuru no] 0 ka? -N om com e C om p is Q 'Is it that Hanako w ill com e?' c. [H anako -ga kuru no] 0 0 ? -N om com e Q is Q 'Is it that Hanako w ill com e ? ' A ccording to S. M iyagaw a, (27a) is ruled out due to the "ungoverned" ka, w hile (27c) is ruled in, lacking an overt com plem entizer. M oreover, he speculates that (27b) is rescued, since no serves as a governor o f k a .u On our account, this paradigm can be readily explained as follow s. (27a) is excluded, since H anako-ga. not in the seg-com m and dom ain o f ka at LF, is Although Tam a-ga 'Tam a-Nom ' in (i) does not appear to receive a focus interpretation, (ii), w h ich is a y es - no question corresponding to (i), is rather marginal. G iven the present argument, it should be that Tam a-ga m ust be a focus in (ii). W hy is it then that Tam a-ga is necessarily a focus in (ii), but in not(i)? A s Kuroda (ibid.) points out, a matrix clause allow s for a ga-m arked N P to receive a "non-focus" interpretation, only w hen the clause sim ply expresses the speaker's recognition o f a particular event that takes place at the m om ent o f utterance and to w h ich the speaker is cognitively related, m ost likely, by visual perception. (S ee the page 22 o f Kuroda.) Thus, in (i), Tam a-ga can escape a focus interpretation, for in this sentence the speaker sim ply expresses his recognition o f a present event in w hich Tama is sleeping on the chair. On the other hand, (ii), not being a declarative sentence, does not sim ply express this event. M oreover, it is likely that in a felicitous context o f discourse, the speaker sim ply does not have any k n ow led ge w hether or not Tama is sleeping on the chair at the tim e o f utterance. T his ignorance may very w ell lead the speaker to utter the question in (ii). Then it is not so unreasonable that Tam a-ga in (ii) is rendered as a focus, sin ce this sentence does not express the questioner's recognition o f a present event. 14 In brief, S. M iyagaw a (ibid. 366) assum es that no can govern ka, for no occurs in the sam e "Comp" as ka. T his is im possible under my assum ption concerning the structure o f no-da / desu. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 174 intended as a focus o f this question. Indeed, if focus is shifted to the verb adjacent to ka, the sentence becom es acceptable as in (28): (28) H anako-w a K U R U ka? -T op com e Q Is H anako COM ING?' A gain, the nom inative marker is replaced with the topic marker and the verb, that falls within the seg-com m and domain o f ka at LF, is focalized. In (27b ) and (27c), H anako-ga is em bedded in the no-clauses. H ence this N P d oes not have to be rendered as a focus as in (2 9 ):1 5 (29) [ H anako-ga kuru no ] desu. -N om com e C om p is 'It is that John is com ing.' 15 It is w idely know n that the "obligatory focus" is a root phenom enon. Thus, iruka-ga 'dolphin-Nom ' in (i), but not that in (ii), must be rendered as a focus: (i) Iruka -ga tensai desu. dolphin-N om genius is 'It is dolphins that are genius.' (ii) Hruka -ga tensai da to yuu] shooko dolphins-N om genius is that evidence 'evidence that dolphins are genius' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Even if H anako-ga receives a focus interpretation. (27b) and (27c) are still grammatical. T his is because the NP is in the no-clauses that are seg-com m anded by the interrogative com plem entizers at LF as in (30): (30) (The LF-representation o f (27b) and (27c)) CP [ V 0 ] John-ga kuru no A no-clause cannot be separated from the follow in g copula by a syntactic operation. In (30), the no-clause raises along with a null copula and adjoins to ka or its null counterpart. A s a result, the no-clause and any elem ents within this clause end up within the seg-com m and dom ain o f the interrogative com plem entizers. 4. 5. Empirical Justification T his section w ill present em pirical support to the discussion so far. In b r ie f , I am assum ing that (i) ka in a w h-question. but not that in a ves-n o question, m ust be "governed," and (ii) the focus o f a ves-n o question must be within the seg- com m and dom ain o f an interrogative com plem entizer at LF. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 4.5.1. A Paradigm in the No-da / desu Construction C onsider the follow in g ill-form ed w h-question in no - d esu:16 (31) * [John-w a nani-o katta no] 0 ka? -T op w hat-A cc bought Com p is Q 'It is that John bought w hat? ' G iven the d iscussion thus far. it appears natural that the deviance o f this sentence is due to the "ungovem ed" ka. If this is indeed the case, the rem edies for (5) should im prove this exam ple as w ell. T o review how the deviant sentence in (5) can be cured, som e exam ples are repeated in (32): (32) a. (= (5 )) *John-w a nani -o katta ka? -T op w hat-A cc boughtneutralQ 'John bought w hat? ' b. (= (4 )) John-w a hon -o katta ka? -T op b ook -A cc boughtneutral Q ’ John bought a book?' c - (= (7 )) John-w a nani -o katta 0 ? -T op w hat-A cc bought Q 'John bought what? ' 16 This exam ple contains a null copula . Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 177 d (= (8 a)) [John-ga nani -o katta ka] (-o) shiri -tai. -N om w hat-A cc bought Q ](-A cc) know -w ant 'I w ant to know [John bought w hatl . ’ e. (= (8 b)) [John -ga nani -o katta ka] -ga m ondai desu. -N om w hat-A cc bought Q -N om problem is ’[John bought whatl is an issu e.’ f. (= (9 )) [John-w a nani -o katta ka] naa / nee? -T op w hat-A cc bought Q I-wonder ’I w onder [John bought w h atl? 1 g. (= (1 0 )) J o h n -w a nani -o kai-m ashita ka? -T op w hat-A cc buy-p0 |jte-Past Q ’John bought what? 1 (32a) is ruled out, for this w h-question contains an "ungovem ed" ka- (32a) and (32b) differ only in that nani ’w hat’ in the former exam ple is replaced by hon ’a b ook.’ The gram m aticality o f (32b) indicates that ka in a v es - no question m ay be left ungoverned unlike that in a w h-question. The null counterpart o f ka, as (3 2 c) illustrates, is im m une from the "government" restriction. (32a) is cured in (32d), (32e) and (3 2 0 by em bedding it in larger constituents. (32g) rem edies (32a) by changing katta into its polite counterpart kaimashita. A ssum ing a functional category Q , I subsum ed this cure under the last "embedding" strategy. N ow let us exam ine if these rem edies im prove (31). First, consider the exam ple in (33): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 (33) [John-wa hon-o katta no] 0 ka? -Top book-A cc bought C om p is Q 'It is that John bought a book ?' Nani 'what' in (31) is replaced by hon 'book.' T his resultant ves-n o question is fully acceptable, w hich is another confirmation o f the claim that ka in a ves-n o question does not have to be "governed." N ote, further, that any elem ent in the no-clause is focused in principle. H ence the sentence successfully renders hon. for exam ple, as the focus o f this question. Second, the "deletion" o f ka cures (31): (34) [John-wa nani-o katta no] 0 0 ? -Top w hat-A cc bought C om p is Q 'It is that John bought w hat? ' T his im provem ent readily fo llow s from the assumption that the covert interrogative com plem entizer does not have to satisfy the "government" requirement. Third, the deviance o f (31) is alleviated by em bedding it in a larger constituent: (35) a. [[John-ga nani-o katta no] 0 ka] -ga mondai da. -N om w hat-A cc bought C om p is Q -N om problem is '[It is that John bought w hatl is a problem.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 179 b. [[John-ga nani -o katta no] 0 ka] (-o) shiri-tai. -N om w hat-A cc bought C om p is Q (-A cc) know -w ant 'I want to know [it is that John bought w h a tl' (36) [[John-ga nani -o katta no] 0 ka] n e e /n a a ? -N om w hat-A cc bought C om p is Q I-w onder 'I wonder [it is that John bought what] ' (37) [John-ga nani-o katta no] desu ka? -N om w hat-A cc bought C om p ispoiite Q 'It is that John bought what. Recall that attachment o f a sentential particle such as nee or naa is a special case o f em bedding. A s for d esu . I assum e that this polite verb form occurs within the dom ain o f a functional category that marks the entire sentence with a performative function o f politeness. G iven these, the sentences in (3 5 )-(3 7 ) all satisfy the "government" requirement on ka. 4.5 .2 . Another Paradigm The "government" requirement on ka in w h-questions also accounts for the ungrammaticality o f (38): (38) * [John-wa nani -o katta no] da ka? -Top w hat-A cc bought C om p isneutral Q 'It is that John bought w hat. ' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 180 T his sentence is exactly the same as (37) except that desu is replaced by its neutral counterpart da. Ka in this exam ple does not satisfy the "government" requirement, evidenced by the fact below that the by-now familiar rem edies can im prove this sentence. First, consider the w ell-form ed exam ple in (3 9):17 (39) [[John-w a nani -o katta no] da 0 ? -Top w hat-A cc bought C om p i s n e u tral Q 'It is that John bought w hat. ' (39) minimally, differs from (38) in that the null interrogative com plem entizer appears in the place o f ka. This im provem ent further supports the proposal that ka, but not its covert counterpart, obeys the "government" restriction. 17 For an unknow n reason, this datum sounds sligh tly odd, w hich is cured by reducing the com plem entizer no to n: (i) [[John-w a nani -o katta n] da 0 ? I have no idea at this point as to the exact nature o f this phenom enon. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 181 Second, (38) is im proved by em bedding it in a larger structure:ls (40) a [[John-ga nani -o katta no] da ka] - ga mondai da. -N om w hat-A cc bought C om p i s n e u tra i Q -N om problem is '[It is that John bought w hatl is a problem .' b. ?[[John-ga nani -o katta no] da ka] (-o) shir-itai. -N om w hat-A cc bought C om p isneutral Q (-A cc) know -w ant 'I want to know [it is that John bought what] ' (41) ?[[John-ga nani -o katta no] da ka] nee / naa? 19 -N om w hat-A cc bought C om p i s n e u tra l Q I wonder 'I w onder [it is that John bought w hatl ' T hese sentences, w ith ka being "governed," are all grammatical. IS A ll o f the exam ples below are slightly marginal. Again, the reduction o f no to n rem edies the oddity, as in the last footnote. 19 U e y a m a (l9 9 2 , 20), how ever, finds unacceptable a sentence that resem bles (41). '(i) is her exam ple with her ow n judgm ent: (i) * Dare -ga kita no da ka ne? (U eyam a's (1 9 9 2 ), (51a)) w h o-N om cam e C om p is(neutral) Q (I wonder) 'I w onder it is that w h o cam e.’ (i) how ever appears to me far better than (ii), where ka is "ungoverned:" (i) *Dare-ga kita no da ka? ’It is that w h o cam e?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 4.6. A Puzzle in Yes-N o Questions The ungram m atically o f the ves-n o question in (42) co m es as a new twist in the w hole picture: (42) *[ John-ga hon -o katta no ] da ka? -N om book-A cc bought C om p isneutrai Q 'Is it that John bought a book?' So far w e have observed that tw o distinct factors determ ine the grammaticality o f Japanese questions. One is the "government" requirem ent on ka in a w h-question and the other is that the focus o f a v es-n o question must occu r within the seg- com m and dom ain o f an interrogative com plem entizer at LF. O bviously, the former requirement is irrelevant for the deviance o f (42). If it w ere, the w ell-form edness o f (43) and (44) w ould be a total mystery: (4 3 )(= (2 5 )) John -w a hon -o K A TTA ka? -T op b ook-A cc bought Q 'D id John B O U G H T a book?' (4 4 ) ((= 3 3 )) [John -ga hon -o katta no ] 0 ka? -N om b ook-A cc bought C om p Q 'It is that John bought a book?' M oreover, consider the contrasts in (45) - (47): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 (45) a.*John -w a nani -o katta ka? -Top w hat-A cc bought Q 'John bought what?' b. John -w a nani-o katta 0 ? (4 6 ) a.*[John-w a nani -o katta no] 0 ka? -Top w hat-A cc bought C om p Q 'It is that John bought w hat? ’ b. [John-w a nani -o katta no] 0 0 ? (47) a. *[John-w a nani -o katta no] da ka? -Top what-A cc bought Com p isneutral Q 'It is that John bought w hat?' b. [John -w a nani -o katta no] da 0 ? (45) - (47) indicate that a violation o f the "government" restriction is remedied sim ply by sw itching ka for its null counterpart. H ow ever, this strategy does not alleviate the oddity o f (42): (48) *[ John -ga hon -o katta no] da 0 ? -N om b ook-A cc bought C om p isneutral Q 'It is that John bought a book?' S o there are at least tw o pieces o f evidence that (42) is not ruled out because o f the "ungoverned" ka. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 184 N otice that (42) cannot be explained by assum ing that its focus is outside the seg-com m and domain o f ka at LF, either, since any elem ent can be focused in the no-clause o f no desu as a result o f predicate raising. Provided that the given tw o conditions / requirements fail to account for (42). there should be another condition on Japanese ves - no questions yet to be d iscovered. Let us inquire about what it may be. 4.7 . A Speculation U eyam a (1992) claim s that the existence o f da is responsible for the unacceptability o f a ves-no question like (42). This suggestion m ay w ell be confirm ed by the fact that (42) becom es acceptable when da is deleted, or equivalently in the current discussion, w hen da is replaced by the null copula: (49) (= (4 2 )) * [John-ga hon-o katta no] da ka? (50) [John-ga hon -o katta no] 0 ka? If the oddity o f (49) is due to the presence o f da, this im provem ent is readily understood. The question is, o f course, w hy da cannot appear in v es-n o q u estion s. First o f all, let us suppose, as stated in Chom sky's (1 9 9 4 ) G reed or any o f its variants. Attract / M ove applies only when the operation is necessary for feature checking. Then, X °-m ovem ent in ves-n o questions ought to be triggered in order to ch eck a certain formal feature, call it [Q ], associated with a verb and an Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 interrogative com plem entizer. On this view , the exam ple in (51). which involves Subject-A ux Inversion, has the derivation in (52):'"- 2 1 (51) W as John arrested? (52) (sim plified; The irrelevant N P m ovem ent is om itted.) a. [c 'C [q ] [ ip John [pi [v p i w a s[Q ] [yp2 arrested ]]]]] b. [ c [ c w a s|[Q ]-C [Q ] ] [ ip John [pi [y p i t| [vp2 a rrested ]]]]] c. [ e t c w a si-C ] [ip John [pi [vpi ti [yp2 arrested ]]]]] After w as raises and adjoins to the C, these elem ents enter into a checking relation so as to elim inate the [Q]-features. Returning to da, I assum e that this elem ent cannot be associated with a [Q]- feature. Schem atically, I consider (53) as an illicit linguistic object: (5 3 ) *da — [Q] Suppose, follow in g C hom sky (1 9 9 5 , Ch. 4, Section 2.2), that the lexicon provides an "optimal" coding o f the idiosyncratic features o f a lexical item, i.e. w hatever do not follow from principles o f UG and those o f a specific language. Thus, the lexical entry o f w as in (51) contains idiosyncratic (hence, unpredictable) 2 l) The [Q] feature is strong, since Subject-A ux inversion is overt in English. 2 1 S ince Subject-A ux Inversion is obligatory in matrix questions, an interrogative com plem entizer necessarily has a [Q]-feature. On the other hand, this feature is associated with a verb only optionally. Otherwise, the derivation o f a declarative sentence with a verb w ith [Q] w ould falsely crash. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 features such as phonological metrices, sem antic features, a categorial feature (= [V ]) and so forth. A feature [Q], since its optional presence on verbs is predictable from the categorial feature o f [V], is not listed in the lexical entry o f w a s. Rather, this feature is added to its lexical entry, when w as is introduced from the lex ico n into the com putational system . G iven these specific assum ptions, m y proposal is that da disallow s for a [Q]-feature to be added to its lexical entry. O f cou rse, this gives rise to a new question w hy a [Q]-feature cannot be added to the lexical entry o f da. A custom ary list o f the Japanese copulas in (54), w hich I have tacitly assum ed thus far, appears suggestive in an effort to answ er this question: (5 4 ) (b ased on U eyam a's (38) and (39)) N E U T R A L POLITE P R E SE N T da / 0 desu P A ST / PE R FE C T IV E datta deshita N otice that (54) assum es tw o alternative forms (da and the covert 0 ) only in the "slot" o f present-neutral copula. It follow s from this assum ption that (55a) and (55b) are basically synonym ous: (5 5 ) a. B ill-w a gakusee da. -T op student is 'John is a student.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 187 b. B ill-w a gakusee 0 . 'John is a student.' H ow ever, basically follow in g U eyam a (1992), I w ould like to assum e that Q_ is the only neutral present copula in Japanese.” Although I presently do not have strong em pirical evidence o f this proposal, the exam ples in (5 6 ), how ever, at least indicate that da has a usage w hich cannot be accounted for if it is necessarily a copula: (5 6 ) a. ?Y osaku-w a mada ano toki gakusee datta da. till then w asneuiral 'Yosaku w as still a student then.' b. ?Y osaku-w a im a gakusee desu da. lspolite 'Yosaku is now a student.' c. ?Y osaku-w a mada ano toki gakusee deshita da. waspolite 'Yosaku w as still a student then.' Although highly dialectal and possibly archaic, da follow s the "other" copulas in these sentences. O bviously, da is by no m eans a copula here.2 3 22 If so, (55a) should be represented as in (i): (i) B ill-w a gakusee 0 da. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 E ven though em pirical evidence is rather slim , let us entertain a possibility that da is not indeed a copula, and, hence, not a verb. R ecall that a [Q]-feature is added only to the lexical entries o f a verb and an interrogative com plem entizer. It then follow s that da cannot have this particular feature. Suppose that da is not indeed a copula, (49) is associated w ith a structure som ething like (57) before Attract / M ove applies: (57) (sim plified) ZP [?da] V P [v[Q] 0 ] John-ga no Ka is obligatorily assigned a [Q]-feature, as it is drawn from the lexicon into the com putational system . 24 In an absence o f relevant evidence, I sim ply speculate in (57) that da originates as a sister o f V P and projects to "ZP" for the concreteness 23 It appears to me that da in these exam ples is akin to y o , w hich roughly m eans "I tell you...." 24 T his assum ption is m aybe independently needed, since I am assum ing that a Japanese question raises a verb and adjoins it to an interrogative C. G iven that an interrogative C has a [Q] feature, this C must attract a verb with a [Q] feature. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 189 sake. M oreover I assum e that the [Q] feature is "strong" in that it is visible at PF. N otice that as far as the linear order o f the lexical item s are concerned, (57) directly corresponds to (49). H ow ever, since the [Q ]-features are yet to be checked in this configuration, the derivation crashes, if Spell-O ut applies to (57). For P F -convergence, Attracts / M ove converts (57) to (58): (58) (The [Q ]-features are deleted, assum ing that checking has applied.) V P i [C ka] CP [v 0 ] YP [ ?da] John-ga kita no K eeping to an assum ption from the last chapter that a copula forms a "complex" predicate with the preceding n o-clausc. (58) raises 0 along with the no-clause. (58) yield s the follow ing illicit surface string: (59) *D a [John-ga kita no] ka? (5 9 ) is ruled out, presumably because a afftxal nature o f da is not satisfied. Turning to (50), this w ell-form ed exam ple is assigned a convergent derivation in (60): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 190 (60) a- [ c [iP [v p [CP ••• n°] 0 [Q | 1 1 k a [Q] ] (before Attract / M ove) b. [ c [iP M 1 t d v p i [c p ••• no] 0[Q] ]-ka[Qj]] (Attract/ M ove) c. [ c [IP 1 1 ] [c tv P l [CP no] 0 ]-ka]] (checking) Ka[QI attracts the null copula 0i[Q] along with the no-clause and the [Q ]-features are elim inated after checking. N ote that (49) is rem edied by em bedding the sentence in a larger constituent, as (61) and (62) illustrate: (61) a.?[[ John -ga hon -o katta no ] da ka(doo ka) ]-ga m ondai-da. -N om book-A cc bought C om p isneutrai w hether -N om problem -is 'W hether it is that John bought a book is an issue.' b. [[John -ga hon -o katta no] da k a (d o o k a )] (-o) shiri -tai. -N om book-A cc bought C om p isneutral Q (how Q) (-A cc) know -w ant 'I w ould like to know w hether it is that John bought a book or not.' (62) ?[John -ga hon -o katta no] da ka(dooka) naa/nee? -N om book-A cc bought C om p isneutral whether (I w onder) '1 w onder whether it is that John bought a book.’ (61a) and (61b) em bed (49) as a sentential subject and as a clause subordinate to the verb shir 'know.' respectively. R egarding (62), I keep to the assum ption that attachment o f such a particle as naa and nee is a special case o f em bedding. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 191 The present account o f the deviance o f (49) naturally relates this phenom enon to a w ell-know n fact in English that the Subject-Aux Inversion d oes not occur in em bedded questions. C onsider (63): (63) I w onder whether John w as arrested? U nlike (51), w as does not raise to the em bedded C. In the g iv en fram ework, this fact indicates that the [Q]-feature o f the em bedded C is checked o ff w ithout raising w a s. I assum e that raising o f w as is circum vented in (63), because the em bedded C enters into a checking relation with w onder, w hich I take to be associated with [Q] as it enters a numeration. If correct, (63) undergoes the derivation in (64): (64) (Irrelevant parts are not show n.) a. [vp [v w o n d e rjQ ]] [cp w h e th e r [ c C [q j] [ip ... w as ... ]]] b. [v p [v [c l C[Q j][vw onder[Q ]]] [cp whether ti [ip ... w as ... ]]] b. [v p [v [c i C ][vw on d er]] [cp w hether ti [ip ... w as ... ]]] Here w as is not associated with an optional [Q]-feature. Therefore it does not raise and adjoin to the C. Rather, w on d eq n] attracts C[Q], and for convergence they enter into a checking relation to delete these [Q]-features. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 192 An analogous explanation is possible for the sentences in (61) and (62). (61b), for exam ple, is w ell-form ed, since the exam ple is assigned the convergent derivation in (65):2 5 (65) (sim plified) a. [v p (CP [C'[lP [CP ••• no] da ] [cC(Q ]]] ka(doo ka) ] [vshiqQ j] ] b. [v p [ C P [ c ' [ l P [ C P - no] da ] t) ]] ka(doo ka) ] [v [ciC [Q ]][vshir[Q j]] ] c. [v p [CP [C'[lP [CP - no] da ] ti ]] ka(doo ka) ] [v [c \C ][v sh ir ]] ] A gain, da cannot have a [Q]-feature. Like w onder in (64), shir 'know' is associated w ith a [Q]-feature as it is drawn from the lexicon into the com putational system . T his verb attracts C (=65b), and the [Q]-features are deleted from these lexical item s after checking. This derivation converges. It is independent that (48) cannot be saved by em bedding the sentence in a larger constituent: (66) a.*[[ John -g a hon -o katta no ] da 0 ]-ga m ondai-da. -N om book-A cc bought C om p isneutrai Q -N om problem -is 'W hether it is that John bought a book is an issue.' b.*[[John -g a hon -o katta no] da 0 ] (-o) shiri -tai. -N o m b ook -A cc bought C om p isneutral Q (-A cc) know -w ant 'I w ould like to know whether it is that John bought a book or not.' 25 For sim plicity, I take ka appearing in an em bedded question to be a reduced form o f kadooka 'whether.' The analysis is unaffected, if ka is actually a C. A gain, I om it the raising o f the no-clause from (65) for expediency. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 (67) *[[John-ga hon -o katta no] da 0 ] n a a /n e e ? 2 h -N om b ook -A cc bought C om p isneutral Q (I wonder) ’I w onder w hether it is that John bought a book.’ Though (48) is em bedded, these sentences are all ungrammatical. T his is due to a general ban against em bedded questions without ka or kadooka at the clause-final positions:2 7 (68) a. M ary-wa kita? -Top cam e 'Mary cam e.' b. *Taroo-w a [M ary -ga kita] shitteiru. -T op -N om cam e know 'I know w hether M ary cam e.' The nature o f this constraint is not clear at this point. 26 This sentence is acceptable on a reading where naa and nee are taken as "intensifying particles," attached to the end o f a declarative sentence. 2 7 This restriction holds for w h-questions as well: (i) a. Dare-ga kita? w h o -N o m cam e 'W ho cam e.' b. *T aroo-w a [M ary -ga kita] shitteiru. -T op w h o -N o m cam e know 'Taro k n ow s w h o cam e.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 194 4.8. More on Yes-NO Questions A noticeable fact, as w e observed in the last tw o sections, is that da* unlike its apparent null counterpart 0 . never appears in a matrix v es-n o question. B esid es da and 0 , the custom ary list o f Japanese copulas contains datta (the neutral past / perfective copula), desu (the polite present copula), and deshita (the past / perfective form o f d esu ). Com bining these five elem ents w ith the tw o types o f interrogative com plem entizers (ka and the phonologically null one), the follow in g table exhausts 10 logically possible sequences: (69) (a) *da ka (b) *da 0 (c) datta ka (d) datta 0 (e) 0 ka (f) 0 0 (g) desu ka (h) *desu 0 (i) deshita ka 0 ) deshita 0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 195 Exam ples involving these sequences are given in (70): (70) (a) *[ John-ga hon -o katta no] da ka? -N om book-A cc bought Com p i s n e u tral Q 'It is that John bought a book?' (b) *[ John -ga hon -o katta no] da 0 ? (c)(based on U eyam a's (1992) (47b)) ??[ John -ga hon -o katta no] datta ka. -N om book-A cc bought Com p w asneutral Q 'It w as that John bought a book?' (d) ?'?[ John -ga hon -o katta no] datta 0 . (e) [John-ga hon -o katta no] 0 ka? ( 0 [John-ga hon -o katta no] 0 0 ? (g) [John -ga hon -o katta no ] desu ka? -N om b ook-A cc bought C om p ispoiite Q (h)*[John -ga hon -o katta no ] desu 0 ? (i) (based on U eyam a's (1992) (47a)) ??[ John -ga hon -o katta no] deshita ka. -N om b ook-A cc bought Com p wasp0|ite Q 'It w as that John bought a book?' (j)??[ John -ga hon -o katta no] deshita 0 . The exam ples in (70c), (70d), (70i), and (70j) are som ew hat m arginal. The m arginality, how ever, does not indicate that these exam ples contain illicit Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 196 com binations o f copulas and interrogative com plem entizers. T hose cco p u la - C> pairs are indeed fully acceptable, as the much simpler sentences in (70) illustrate: (71) (c) John-w a mukashi keisatuskan datta ka? -T op before policem an w as Q 'W as John a policem an before?' (d) John-w a m ukashi keisatsukan datta 0 ? (i) John-w a m ukashi keisatsukan deshita ka? (j) John-w a m ukashi keisatsukan deshita 0 ? The m arginality o f (70c), (70d), (70i), and (70j) appear to be independent o f what concerns us here. C onsider (72): (72) ??[John -ga hon -o katta no ] datta / deshita. -N o m book-A cc bought C om p w a sneutral/polite 'It w as that John bought a book.' This exam ple illustrates that the past tense copulas do not fit w ell in the no-copula construction. The reason(s) is (are) not entirely clear to me this m om ent.2 8 2 8 O ne pragmatic factor may be relevant. It appears to me that a sentence that contains a no-clause im m ediately follow ed by datta / deshita m ay be and perhaps m ust be uttered in a certain discourse context. C onsider (72), repeated in (i), again: (i) ( = (7 2 )) [ John-ga hon-o katta no] datta / deshita. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 197 What is puzzling is the fact that (70h) is hopelessly unacceptable. A matrix v es- no question ending in desu - 0 , is not adm issible, as further evidenced by (73): (73) John -w a keisatsukan desu 0 ? -T op policem an isp0|jte Q 'Is John a policem an?' I leave a question open w hy this < d e su -0 > pair is illicit. 4.9. Sum m ary and C onclusion B ased on S. M iyagaw a (1987) and U eyam a (1 9 9 1 , 1992), this chapter presented a prelim inary study regarding w hat determ ine the w ell-form edness o f interrogative sentences in Japanese. C onsidering both w h-questions and v es-n o questions, and those in the "bare" form and in no-da / desu. the chapter suggested generalizations in (74) to predict the range o f w ell-form ed Japanese questions: A context that allow s (i) is where this sentence is narrated by an om niscient narrator (Kuroda, 1973, 3 82) o f a story. W hen this "narrative / non-reportive" use is felicitous, (7 0 c ), (70d), (70i), and (70j) appear to be also acceptable, (i) is rather odd, w here this exam ple is uttered in a usual conversational context as in (ii): (ii)*[John-ga hon -o katta no datta / deshita 1 yo. -N om b ook -A cc bought Com p w a sneutral/polite I-tell-you 'I am tellin g you that it was that John bought a book.' The sentence final particle y o , translated as I am telling vou. requires o f the sentence to have a "non-narrative / reportive" interpretation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 198 (74) a. Ka in a w h-question must be governed. b. The focus o f a v es-n o question must be seg-com m anded by an interrogative com plem entizer at LF. c. Da cannot be associated with an optional [Q]-feature. R eview in g, w e first observed that (74a) accounts for (75) - (77) am ong others: (75) *John-w a nani -o katta ka? (= (5)) -T op w hat-A cc boughtneutral Q 'What did John buy?' (76) fJohn -ga hon -o katta no] 0 ka? -N om b ook -A cc bought Com p is Q 'Is it that JO H N bought that book?' (77) John-w a nani -o katta 0 ? (=(7)) -T op w hat-A cc bought Q 'John bought w hat? ' Ka in a w h-question and that in a ves-no question are, in effect, hom ophones in that only the former m ust be "governed." This descriptive generalization im m ediately explains the contrast between (75) and (76). (77) further illustrates that the null interrogative com plem entizer does not have to be "governed. An important conceptual question remains w hy it is that ka in a w h-question must be "governed." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 199 Second, the generalization in (74b) accounts for the ungrammaticality o f (78): (78) *John-ga hon -o katta ka? (=(20)) -N om b ook -A cc bought Q 'Is it John that bought a book?' The nom inative phrase, not seg-com m anded by ka at LF, is an 'illegitim ate' focus o f this question. On the other hand, (79) and (80) are grammatical, since their foci are seg-com m anded by ka at LF. (79) John-w a hon-o KATTA ka? 'Did John B U Y books?’ (80) [JO H N -ga hon -o katta no] desu ka? -N om b ook -A cc bought Com p is Q ’Is it that JO H N bought that book?1 Importantly, due to these grammatical exam ples, the deviance o f (78) must not be taken as indicating that ka in a ves-n o question obeys the governm ent restriction alongside o f the one in a w h-question. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 Finally, given (74c), (81) is excluded: ( 81) *[ John-ga hon -o katta n o ] da ka. (= (42)) -N om book-A cc bought C om p i s n e u trai Q 'Is it that John bought a book?' Since da is not associated with a [Q]-feature, the [Q]-feature on ka fails to be checked off, w hich renders the relevant derivation crash at LF. A s a closin g remark, I once again note that (74a) is nothing but a descriptive generalization in that it associates to ka a peculiar formal property to "explain" its peculiar behavior in certain linguistic contexts. Even if it turns out that ka indeed has a property as postulated, I have little to say, at the present m om ent, about what enables a speaker o f Japanese to acquire such a peculiarity o f this elem ent. Furthermore, from a view point o f parametric syntax, I have no account for the fact that Korean, w hich m ostly behaves on a par with Japanese, d oes not have anything like (74a): (82) I-kes-nun m w ues -i -ni? this-Top w h a t -iS n e u tra l Q 'What is this?' (82) indicates that unlike ka in Japanese, ni, an overt interrogative com plem entizer in Korean, does not need to be "governed." I hope the future research w ould system atically explain this intriguing difference betw een Japanese and Korean. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 REFERENCES A b e, Jun (1 9 9 3 ) Binding C onditions and Scram bling without A / A' distinction. Ph. D dissertation, the U niversity o f C onnecticut. A b n ey, Steven P (1 9 8 7 ) T he English N oun Phrase in its Sentential A sp ect. Ph D dissertation, M IT, C am bridge, M ass. A oun, Joseph and Y en-hui Audrey Li ( 1993a) "W h-elem ent in situ: Syntax or LF, " Linguistic Inquiry 2 4 , 199-238. A oun, Joseph and Y en-hui Audrey Li (1993b) "On Som e D ifferences betw een C hinese and Japanese W h-elem ents." Linguistic Inquiry 2 0 , 2 6 5 -2 7 2 . A oun, Joseph and Y en-hui Audrey Li ( 1993c) Syntax o f S cop e, the M IT Press. A oun, Joseph, N . H ornstein, and D . Sportiche (1981) "Aspects o f W ide Scope Quantification," Journal o f Linguistic Research 1, 67 - 9 5 . A oy a g i, H iroshi (1994) "On A ssociation with Focus and Scope o f F ocus Particles in Japanese," M s., U niversity o f Southern C alifornia. A ron off, M . (1976) W ord Formation in G enerative Grammar. M IT Press, C am bridge, M ass. Baker, C. L. (1970) "Note on the D escription o f English Q uestions: the Role o f an Abstract Q uestion M orpheme," Foundations o f Language 6 , 197 - 219. B aker, Mark (1988) Incorporation: A Theory o f Grammatical Function C hanging. C hicago, U niversity o f C hicago Press. B arw ise, Jon and Robin C ooper (1 9 8 1 )" G eneralized Quantifiers and Natural Language." L in g u istics and Philosophy 4. 159-219. B elletti, Adriana (1988) "The Case o f Unaccusatives," Linguistic Inquiry 19, I - 34. B elnap, N u el (1982) "Q uestions and A nsw ers in M ontague Grammar," In S. Peters and E. Saarinen (ed s.), Processes. B eliefs, and Q u estio n s. 165- 198. Dordrecht, D . R eidel. B elnap, N u el, an d T . Steel (1 9 7 6 ) T he L ogic o f Q uestions and A n sw ers. N ew H aven, Y ale U niversity Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 202 Berm an, S. (1989) "On Certain D ifferences betw een W h-phrases and Indefinites," in J. Cater, et al. (ed s.), N ELS 20, G S L A , U niversity o f M assachusettes, Am herst, M ass., 31-45. BUring D aniel and Katharina Hartmann (1994) "The Dark Side o f W H - M ovem ent," Linguistische Berichte 149, 56-74. C hierchia, G ennaro (1 9 9 1 ) "Functional W H and W eak Crossover." in D. Bates (ed.), Proceedings o f W C C FL 10, C SLI, Stanford, C alifornia. Chierchia. Gennaro (1 9 9 2 -1 9 9 3 ) "Questions with Quantifiers," Natural Language Sem antics. 1, 181- 234. C ho, D ong-In (1993) "Functional Projections and V erb M ovem ent," M s., U niversity o f Southern C alifornia, Los A ngeles. C hom sky, N oam (1973) "Conditions on Transformation," in A nderson and Kiparsky (ed s.), A Festschrift for Morris H alle." H olt, Rinehart and W inston, N Y . C hom sky, N oam (1981) Lectures on G overnm ent and B inding. Dordrecht, Foris. C hom sky, N oam U 9 8 6 a ) K now ledge o f Language. N ew York, Praeger. C hom sky, N oam (1986b ) Barriers, the MIT Press, C am bridge, M A . C hom sky, N oam (1991) "Som e N o tes on E conom y o f D erivation and Representation," in R. Freidin (ed.), Principles and Parameters in Comparative Grammar, the M IT Press, Cam bridge, M assachusetts. C hom sky, N oam (1992) "A M inim alist Program for L inguistics Theory," MIT O ccasional Papers in L inguistics 1. Cam bridge, M assachusetts. C hom sky, N oam (1994) "Bare Phrase Structure," MIT O ccasional Papers in L inguistics. N um ber 5 . C hom sky, N oam (1995) M inim al Syntax, the M IT press, Cam bridge, M assachusetts. C hom sky, N oam and Howard Lasnik (1991) "Principles and Parameters Theory," to appear in J. Jacobs, A. van Stechow , W . Sternefeld, and T. V ennem ann (eds.) Syntax : An International H andbook o f Contemporary Research. W alter de Gruyter, Berlin. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 203 C inque, G u glielm o (1990) T ypes o f A ' D ependencies. C am bridge, M A , the M IT press. C inque, G u glielm o (1993) "A N ull Theory o f Phrase and C om pound Stress," Linguistic Inquiry. 24, 2 3 9 -2 9 7 . C ollins, Chris (1994) "Econom y o f D erivation and the G eneralized Proper B inding Condition," L inguistic Inquiry 25. 45 - 6 1 . C ulicover, Peter W . (1992) "Polarity, Inversion, and Focus in English," In Proceedings o f the Eastern States C onference on L inguistics'91. 46- 68 . C ulicover, Peter W . (1992) "A N ote on Quantifier," Linguistic Inquiry 23. 6 5 9 - 6 6 3 . D iesin g, M olly (1992) Indefinite. M IT Press, Cam bridge, M A . En£, M . (1 9 9 1 ) "Semantics o f Specificity," Linguistic Inquiry 2 2 , 1- 25. Engdahl, E. (1 9 8 6 ) Constituent Q uestions. R eidel, Dordrecht. Epstein, Sam uel David (1992) “Derivational Constraints on A ’-chain Form ation,” Linguistic Inquiry 23, 235 - 260. Erteschik, N (1973) On the Nature o f Island Constraints. Ph. D dissertation, M IT, Cam bridge, M assachusetts. F iengo, Robert (1977) "On T race Theory," Linguistic Inquiry. 3 5 -6 1 . F iengo, Robert and James H igginbotham (1981) "Opacity in NP," Linguistic A nalysis. 7. F iengo, Robert, C-T Jam es H uang, Howard Lasnik and T anya Reinhart (1 9 8 8 ) "The Syntax o f W h-in-situ." Proceedings o f the Seventh W est C oast C onference on Form al L inguistics, 81 - 98. Fox, D anny (1995) "Econom y, Scope and Sem antic Interpretation - Evidence from V P Ellipsis," N E L S 25. Fukui, N aoki (1986) A Theory o f Category Projection and Its A pplication. Ph. D dissertation, MIT. Fukui, N aoki (1988) "LF Extraction o f N aze: Som e Theoretical Implications," Natural Language & L inguistic Theory 6, 445 - 501. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 204 Fukui, N aoki (1995) Theory o f Projection in Syntax. C SLI Publications and K uroshio Publishers. G rim shaw , Jane (1990) Argument Structure. M IT Press, C am bridge. G roenendijk, J. and M . Stok h of (1 9 8 4 ) Studies on the Sem antics o f Q uestions and the Pragmatics o f A nsw ers. Ph. D dissertation, A cadem isch Proefschrift, Amsterdam . G roenendijk, J. M. S tokhof (1989) "Type-shifting R ules and the Sem antics o f Interrogatives,” in G. C hierchia, B. H. Partee, and R. Turner (eds.), Properties. T ypes and M eaning. V ol. 2: Sem antic Issu e. K luw er, Dordrecht. H ale, K. and S. J. K eyser (1993) "On Argum ent Structure and the L exical E xpression o f Syntactic Relations," in K. Hale and S.J.K eyser (eds.), The V iew From the B uilding 20: Essays in Linguistics in Honor o f S vlyain Bromherger. M IT Press, Cam bridge. H am blin, C.L. (1958) "Questions," The Autralasian Journal o f Philosophy 3 6 , 159-168. H am lin. C .L . (1973) "Questions in M ontague English," Foundations o f Languages 10, 41 - 53. H arada, S. I. (1976) "Honorifics." in M . Shibatani (ed .). Syntax and Sem antics 5. Japanese G enerative Grammar. A cadem ic Press, N e w York. H endrick, R., and M . R ochem ont (1 9 8 2 ) "Com plem entation, M ultiple W H, and E cho Questions," M s., U niversity o f North C arolina and U niversity o f California at Irvine. H eim , Irene (1982) The Sem antics o f D efinite and Indefinite N oun Phrases. Ph. D dissertation, U niversity o f M assachusettes, A m herst. H eim , Irene (1990) "E-type Pronouns and D onkey Anaphora," L inguistics and P h ilosop h y 13, 137-178. H igginbotham , James (19 8 5 ) "On Sem antics," Linguistic Inquiry 16. 5 4 7 -593. H igginbotham , James (1991) "Interrogatives I," M IT W orking Papers in L inguistics, V ol. 15, M IT, Cam bridge, M assachusetts. H oji, H ajim e (1983) "Multiple W h-O uestions in Japanese." M s., U W . Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 205 H oji, Hajim e (1985) L ogical Form Constraints and Configurational Structures in Japanese. Ph. D dissertation. U niversity o f W ashington. Hoji, Hajim e (1986) "Scope Interpretation in Japanese and Its Theoretical Implications," Proceedings o f the W est Coast C onference on Formal Linguistics." 5. H oji, Hajim e (1990) Theories o f Anaphora and A spects o f Japanese Syntax. M s., USC. H oji, Hajim e (1991) "Raising-to-Object, ECM and the Major Object in Japanese," paper presented at the R ochester W orkshop on Japanese L inguistics, held at U niversity o f Rochester on M ay 9 - 12, 1991. H oji, Hajim e and Audrey Li (1994) "Comments on the Syntax Papers," a handout distributed at the 5th. Japanese / Korean Linguistic C onference, L os A n geles, California. H onm a. Shinsuke et al. (1992) "Semantic Properties o f the Floated Q uantifier Construction in Japanese," Proceedings o f the 5th Sum m er C onference. T o k y o Linguistics Forum, T okyo, Japan. H ornstein, Norbert (1994) "The Grammar o f LF : From G B to M inim alism ," M s.. U niversity o f M aryland. Huang, CT. J (1982) L ogical R elations in C hinese and the Theory o f Grammar, Ph. D dissertation, M IT, Cam bridge. Ishii, Y a su o (1 9 9 1 ) Operators and Empty Categories in Japanese. Ph. D dissertation, T he U niversity o f C onnecticut. Jackendoff, Ray (1972) Sem antic Interpretation in Generative Grammar. T he M IT Press, Cam bridge. Johnson, K yle (1 9 9 4 ) B ridging the G ap. M s.. U niversity o f M assachusetts- Am herst. Kamp, J. A. W . (1 9 8 1 ) "A Theory o f Truth and Sem antic Representation," In J. G oroenendijk, T . Janssen, and M. S tok h of (eds.) Formal M ethods in the Study o f Language. 27 7 -3 2 1 , M athem atical Center, Am sterdam . Karttunen, Lauri (1977) "Syntax and Sem antics o f Questions," L inguistics and P hilosophy 1, 3-44. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 206 Kato, Y asuhiko (1985) N egative Sentences in Japanese. Sophia L inguistica. W orking Papers in L inguistics 19. Sophia U niversity, T ok yo, Japan. Katz, JerroldJ. and Paul M . Postal (1 9 6 4 ) An Integrated Theory o f Linguistic D escriptions. The M IT press, Cam bridge, M ass. K ayne S, Richard (1984) C onnectedness and Binary Branching. F oris. Dordrecht. K ayne, Richard (1993) "The A ntisym m etry o f Syntax," M s. C U N Y , graduate K im , S o o W on (1 9 9 1 ) Chain Scope and Quantification Structure. Ph. D dissertation, Brandeis U niversity, W altham, M A. K ikuchi, Akira (1989) "Comparative D eletion in Japanese," M s., Y am agata University. K itagawa, Y oshihisa (1984) "Superiority Effects in Japanese: An argument for LF," M s., U niversity o f M assachusetts, Amherst. Kitahara, H isatsugu (1992) "Checking Theory and Scope Interpretation W ithout Q uantifier Raising," In Susum u Kuno (ed.k Harvard W orking Papers in L inguistics. V ol. 1. K oizum i. M asatoshi (1991) Syntax o f Adjuncts and the Phrase Structure o f Japanese. M A thesis, the O hio State University. K oizum i, M asatoshi (1995) Phrase Structure in M inim alist Syntax. Ph. D dissertation, MIT. Krifka, M anfred (1992) "Definite N Ps aren't Quantifiers." Linguistic Inquiry 23 , 1 5 6 - 163. Kuno, Susum u (1 9 7 3 ) The structure o f the Japanese Language. T he M IT Press, Cam bridge. K uno, Susum u (1980) "The scop e o f the question and negation in som e verb-final languages," Papers from the Sixteenth Regional M eeting o f the C hicago Linguistic S ociety. J, Kreiman and A . Ojeda (eds). C hicago : C hicago Linguistic Society. K uno. Susum u (1982) "The focus o f the question and the focus o f the answer." CLS: Papers from the parasession on nondeclaratives. U niversity o f C hicago, C hicago. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 207 Kuno. Susum u and J. R obinson (1972) "Multiple w h -q u estio n s.” Linguistic Inquiry 3, 4 6 3 - 487. K uno, Susum u and K. M asunaga (1986) "Questions w ith W h-phrases in islands," U M O P 11, 1139 - 166. K uno, Susum u and K en-ichi Takam i (1993) Grammar and D iscourse Principle : Functional Syntax and G B Theory." The U niversity o f C hicago Press, C hicago. Kurata, K iyoshi (1991) The Syntax o f Dependent E lem ents. Ph. D dissertation, U niversity o f M assachusetts, Am herst. Kuroda, Shigeyuki (1 9 7 1 ) "Remarks on the N otion o f Subject with R eference to W ords like A lso . Even or Only." The Annual Bulletin o f the Research Institute o f Logopedics and Phoniatrics 4 . U niversity o f T okyo. Kuroda, Shige-Y uki (1973) "Where E pistem ology, S tyle, and Grammar meet: C ase Study From Japanese," Festschirift for M orris H alle. N ew York: H o lt. Kuroda, S h ige-Y uki (1 9 9 2 ) Japanese Syntax and Sem antics. Dordrecht, R eidel. L adlow , Peter (1 9 9 0 ) "C onditionals, Events, and D iscou rse Pronouns," M s., S U N Y , Stony Brook. Larson, R. K. (1 9 8 5 ) "On the Syntax o f D isjunction Scope," Natural Language and Linguistic Theory 3. 217 - 264. Lasnik, Howard (1992) "Case and E xpletives : N otes toward a Parametric Account," Linguistic Inquiry 23, 381 - 405. Lasnik, H oward and M amoru Saito (1984) "On The Nature o f Proper Governm ent," Linguistic Inquiry 14, 235 - 289. Lasnik, Howard and M amoru Saito (1992) M ove a ; C onditions on Its Application and Output. M IT press. L ew is, D avid (1975) "Adverbs o f Quantification" In E K eenan, ed., Formal Sem antics o f Natural Languages. Cam bridge, M ass. M ahajan, A noop (19901 The A / A-bar Distinction and M ovem ent Theory, Ph.D dissertation, MIT. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 208 M ay. Robert (1977) The Grammar o f Quantification. Ph. D dissertation, MIT. M ay, Robert (1981) “M ovem ent and B inding,'4 Linguistic Inquiry 12. M ay, Robert (1985) L ogical Form: Its Structure and D erivation. M IT Press. M ay, Robert (1988) "Am biguities o f Quantification and Wh: A R eply to W illiam s." Linguistic Inquiry. 1 9 ,1 1 8 - 1 3 5 . M cC aw ley, Jam es D. (1 9 8 1 ) Everything That Linguists H ave A lw ays W anted to K now about L ogic but W ere A sham ed to A sk. U niversity o f C hicago Press. M cD aniel, D. (1989) "Partial and M ultiple W h-m ovem ent," Natural Language and Linguistic Theory 7, 565 - 604. M ilsark, G . (1 9 7 4 ) "Existential Sentences in English," Ph. D. dissertation, M IT. M iyagaw a, K eiko (1992) "Som e N otes on N'-deletion in Japanese," M s., U S C , L os A n geles, C A . M iyagaw a, K eiko (1994) " A N ote on S cop e Interaction in Japanese," M s., U S C , California. M iyagaw a, K eiko (1994) "A N ote on Specificity and Prenominal vs. F loating Q uantifiers in Japanese," M s., USC. M iyagaw a, Shigeru. (1 9 8 7 ) "LF A ffix R aising in Japanese," Linguistic Inquiry 1 8 ,3 6 2 - 3 6 7 . M ontague, Richard. (1974) "The Proper Treatment o f Quantification in Ordinary English," In R. T hom ason, ed., Formal P h ilosop h y. N ew Y ale U niversity Press. M urasugi, K eiko (1991) N ou n Phrases in Japanese and English: A Study in Syntax, L eam ability and A quisition, Ph.D. Dissertation, U niversity o f C onnecticut. M urasugi, K eiko and M amoru Saito (1992) "Quasi-adjuncts as Sentential Argum ents," W EC O L 22. N ish igu ch i, T aisuke (1 9 8 6 ) Quantification in Syntax. Doctoral dissertation. U niversity o f M assachussets, Am herst. N ish igu ch i, Taisuke (1 9 9 0 ) Quantification in the Theory o f Grammar. K luw er. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Nishiyam a, K unio et al. (1994) "Syntactic M ovem ent o f Overt W h-phrases in Japanese and Korean," M s., C ornell U niversity. Ohno, Yutaka (1 9 8 9 ) “ M o, “ Papers in Ouantificaiton. N SF , Grant B N S 8 7 1 9 9 9 9 , U niversity o f M assachusetts, Am herst. Oka. T oshifusa {1988) "Abstract Case and Em pty Pronouns," Tsukuba English Studies. V ol. 7, 187-227. Okazaki, M asao (1 9 9 0 ) "Specificity and Accent: The Case o f English W h- questions." Tskuba English Studies 9, 77-1 0 5 . Partee. Barbara H (1 9 8 9 ) "Binding Impicit Variables in Quantified Contexts," Papers from the T w enty Fifth R egional M eeting, C hicago Linguistic S ociety, 3 4 2 - 365. Pesetsky, D avid (1 9 8 2 ) "Path and Categories," Ph. D dissertation, M IT. Pesetsky, D avid (1 9 8 7 ) "Wh-in-situ," M ovem ent and unselective binding," E.J. R euland and A . ter M eulen (eds.) T he Representation o f (in) D efiniteness. T he M IT Press. P ollock . Jean - Y v es (1989) "Verb m ovem ent. Universal Grammar and the Structure o f IP." Linguistic Inquiry 20, 36 5 -4 2 4 . Reinhart, Tanya (1 9 9 2 ) "Interpreting W h-in-situ." M s. R izzi, Luigi (1 9 9 0 ) Relativized M inimality. M IT Press, C am bridge, M ass. Rizzi, Luigi (1 9 9 1 ) "Residual Verb Second and the W h-Criterion." M s., U niversite de G eneve, R izzi, Luigi and Roberts Ian (1989) "Com plex Inversion in French," Probus 1, 1-30. R ooth, M ats (1 9 8 4 ) A ssociation with F ocus. Ph. D. dissertation, U niversity o f M assachusetts. Rooth, M ats and Barbara Partee (1982) "Conjunction, T ype A m biguity and W ide Scope Or," Proceedings o f the First W est C oast C onference on Formal L inguistics. L inguistics D ept., Stanford U niversity. R oss, John Robert (1 9 6 7 ) Constraints on V ariables in Syntax, Ph. D. dissertation, M IT. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 R oss, John Robert (1 9 6 9 ) "Guess W ho," Papers from the Fifth Regional M eetings o f the C hicago Linguistic S ociety. 2 5 2 -2 8 6 . Safir, K. (1982) Syntactic Chains and the D efiniteness E ffect, Ph. D. dissertation, M IT, C am bridge, M assachusetts. Saito, M amoru (1982) "Scram bling, T opicalization and Strong Crossover," M s., MIT. Saito, M amoru (1985) Som e Asym m etries in Japanese and Their Theoretical C on seq u en ces. Ph. D dissertation, MIT. Saito, M amoru (1989a) "Crossing and W h-O binding," Handout for the talk at the O hio State U niversity. Saito, M amoru ( 1989b) "Scrambling as sem antically vacuous A'- m ovem ent," in M . R. Baltin and A. S. Kroch (eds.) Alternative C onceptions o f Phrase Structure. The U niversity o f C hicago Press, 182 - 200. Saito, M amoru (1992a) "Long D istance Scram bling in Japanese," Journal o f East A sian Linguistics I. Saito. Mamoru (1992b) "The A dditional-W H Effects and the Adjunct Site Theory," M s., U niversity o f Connecticut. Saito. M amoru (1994a) "Improper Adjunction," M s. U niversity o f C onnecticut. Saito, M amoru (1994b) "Scrambling and the Functional Interpretation o f W H - phrases," M s., U niversity o f Connecticut. Saito, M am oru (1 9 9 4 c) "Barriers for A-adjunction," M s., U niversity o f C onnecticut. Saito, M amoru and K eiko M urasugi (1990) “N ’-deletion in Japanese: A Prelim inary Study, in Japanese / Korean Linguistics 1. C SL I, Stanford. Sakai, H irom u (1 9 9 4 ) "The Uniform ity M easure for Derivations," M s., U niversity o f California.. Sano, T etsuya (1 9 9 4 ) "On the So-called Japanese E el-sentences." M s., U C L A . Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. } 211 Scha, Rem ko (1 9 8 1 ) "Distributive, C ollective, and Cum ulative Q uantification," in G roenendijik, Janssenm and S tokhof (eds.), Formal M ethods in the Study o f Language. Amsterdam: Mathematical Centre Tract, nos. 135, 136. Schein, Barry (1 9 9 2 ) "Conjunction R eduction Redux," M s., U SC . Schein, Barry (1 9 9 3 ) Plurals and E vents. M IT Press. Srivastav, V eneeta (19 9 4 ) "Scope Marking as Indirect W h-dependencv." Natural Language Sem antics 2: 137-170. S tow ell, T im (1 9 8 1 ) Origins o f Phrase Structure. Ph. D dissertation, M IT. Szabolcsi, A nna (19 9 3 ) "Quantifiers in Pair-list Readings and the N on uniform ity o f Quantification," M s., U C LA . S zab olcsi, A nna and Frans Zwarts (1 9 9 3 )" W eak Island and an A lgebraic Sem antics for Scope Taking," N A L S. Takahashi, D aiko (1993a) "M ovem ent o f W h-phrases in Japanese," Natural Language and Linguistic Theory 11. 6 5 5 -6 7 8 , K luwer, N etherlands. Takahashi. D aiko (1993b ) "Sluicing in Japanese," "Journal o f East A sian L inguistics. 3: 265 - 300. Takano, Yuji (1 9 9 4 ) " E conom y o f Derivation, relativized m inim ality, and Proper Binding," M s., UC Irvine, C alifornia. Takubo, Y ukinori (1 9 8 5 )" On the scope o f negation and question in Japanese," Papers in Japanese L inguistics. 10, 87 - 115. Tancredi. C. (1 9 9 0 ) "Syntactic A ssociation w ith Focus," FL SM 1. T sai, W ei-T ien (1 9 9 4 ) On E conom izing the Theory o f A-bar D ependencies. Ph. D dissertation, M assachusettes Institutes o f T echnology. U eyam a, A yum i (1 9 9 1 ) "On the L icensing o f Functional Heads: the W P- analysis o f Interrogative Clauses," M s., K yoto G aikokugo Daigaku ( = K yoto U niversity o f Studies), K yoto, Japan. U eyam a, A yum i (19 9 2 ) "I-to-C m ovem ent as a Last Resort o f L icensing [+W h]," M s., K yoto U niversity o f Foreign Studies. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 212 W atanabe, Akira (1991) "W h-in-situ. Subjacency and Chain Formation." M s., MIT. W atanabe. Akira (1992) "Subjacency and S-structure M ovem ent o f W h-in- situ." Journal o f East Asian Linguistics 1, 255 - 291, 1992. W atanabe, Akira (1 9 9 3 ) Agr-based Case Theory and Its Interaction with the A-bar S y stem . Ph. D dissertation, MIT. W atanabe. Shin (1993) "Focus and Additional W h-effects: Preliminary Study," M s., U niversity o f Southern C alifornia. W ebelhuth, G. (1 9 8 9 ) Syntactic Saturation Phenom ena and the M odem Germanic L anguages. Ph.D . dissertation, U M ass. W illiam s, Edw in (1 9 7 8 ) "Across-the-board Rule Application," Linguistic Inquiry. 9, 31-43. W illiam s, Edw in (1 9 8 6 )" A R eassignm ent o f the Functions o f LF," Linguistic Inquiry 17, 2 65-299. W illiam s, Edw in (1988) "Is LF Distinct from S-structure? A R eply to May," Linguistic Inquiry. 19, 135-146. Y oshim ura, N oriko (1 9 9 2 ) Scram bling and Anaphora in Japanese. Ph. D. dissertation, U niversity o f Southern C alifornia. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Grammaticalization and the development of functional categories in Chinese

PDF

Correspondence and faithfulness constraints in optimality theory: A study of Korean consonantal phonology

PDF

Ellipsis constructions in Chinese

PDF

Beyond words and phrases: A unified theory of predicate composition

PDF

Form and meaning: Negation and question in Chinese

PDF

A comparative study of focus constructions

PDF

The Japanese demonstratives ko, so and a

PDF

Causativity in Korean: Syntactic causative, control, and purpose

PDF

Asymmetry of scope taking in wh -questions

PDF

On the nature of particles in Japanese and its theoretical implications

PDF

'Beowulf' and 'the hobbit': elegy into fantasy in j. R. R. Tolkien's creative technique

PDF

Arthur Machen’s treatment of the occult and a consideration of its reception in England and America

PDF

Conditions on null objects in Basque Spanish and their relation to leismo and clitic doubling

PDF

Automated postediting of documents

PDF

A necessary epigone: The fantastic and "dvoeverie" in the works of A. K. Tolstoi

PDF

Borrowing and borrowability

PDF

An analysis of United States overseas English language policy, 1938-1990

PDF

Andrej Belyj's "Petersburg" and James Joyce's "Ulysses": A comparative study

PDF

A case grammar of the parker manuscript of the "Anglo-Saxon chronicle" from 734 to 891

PDF

A rhetoric of the short story: A study of the realistic narratives of Flaubert, Maupassant, Joyce, and Hyon Chin'gon

Asset Metadata

Creator Watanabe, Shin (author)

Core Title Aspects of questions in Japanese and their theoretical implications.

Degree Doctor of Philosophy

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)

Advisor Hoji, Hajime (committee chair), Kim, Nam-kil (committee member), Schein, Barry (committee member), Vergnaud, Jean-Roger (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-486523

Unique identifier UC11352647

Identifier 9621643.pdf (filename),usctheses-c17-486523 (legacy record id)

Legacy Identifier 9621643-0.pdf

Dmrecord 486523

Document Type Dissertation

Rights Watanabe, Shin

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