You are on page 1of 13

Coordination, Subject Raising, and AgrP in Japanese Author(s): Ichiro Hirata Reviewed work(s): Source: Linguistic Inquiry, Vol.

37, No. 2 (Spring, 2006), pp. 318-329 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/4179367 . Accessed: 30/07/2012 13:53
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry.

http://www.jstor.org

318

SQUIBS

AND

DISCUSSION

Mikkelsen, Line Hove. 2000. Expletive subjects in subject relative clauses. Ms., University of Californiaat Santa Cruz. Pesetsky, David. 1987. Wh-in-situ:Movement and unselective binding. In The representationof (in)definiteness,ed. by Eric Reuland and Alice G. B. ter Meulen, 98-129. Cambridge,Mass.: MIT Press. Postal, Paul. 1971. Crossoverphenomena.New York: Holt, Rinehart and Winston. of the left periphery.In Elements Rizzi, Luigi. 1997. The fine structure of grammar,ed. by Liliane Haegeman, 281-337. Dordrecht: Kluwer. Safir, Ken. 1984. Multiple variable binding. Linguistic Inquiry 15: 603-638. Safir, Ken. 2004. The syntax of (in)dependence.Cambridge,Mass.: MIT Press. movementof whWatanabe,Akira. 1992. Subjacencyand S-Structure in-situ. Journal of East Asian Linguistics 1:255-291. Wiltschko, Martina.1998. Superiorityin German.In Proceedings of the 16th West Coast Conference on Formal Linguistics, ed. by Emily Curtis,James Lyle, and GabrielWebster,431-446. Stanford,Calif.: CSLI Publications. Yusuf, Ore. 1990. Yoruba copular ni. Journal of WestAfrican Languages 20:83-93.

COORDINATION, SUBJECT RAISING, AND AGRP IN JAPANESE

Ichiro Hirata Senshu University

Takano (2004) convincingly shows, on the basis of interpretation of betubetuno'different',thatverbsin coordination structures in Japanese stay within vP in overt syntax. (More precisely, he claims that the tense morphemeis located in T in syntax.) His reasoning is readily extended to examples where subjects are contained in conjuncts, so that it can be shown that in additionto verbs, subjectsreside in vP in overt syntax in Japanese. On the other hand, Kishimoto (2001) elegantly accountsfor the subject-objectasymmetrywith regardto indeterminate pronoun binding in Japanese by assuming that subjects undergo raising in that language. Surprisingly, the subject-object asymmetryat stake is detected even in coordinationstructureswhere subjects seemingly occupy vP-internalpositions. The purposeof this squib is to propose an analysis that resolves the contradiction.The analysis revives Pollock's (1989) original Split-IP Hypothesis by

For valuable comments and suggestions on various aspects of this squib, I would like to thankJeffrey Fryckmanand two anonymousreviewers for LI. This researchwas supported in partby OpenResearchCenterProjectfor Private Universities:matchingfund subsidy from MEXT (Ministryof Education,Culture, Sports, Science and Technology), 2005-2010.

SQUIBS

AND

DISCUSSION

319

which IP is divided into TP and AgrP. Some theoreticalimplications of the revival are also explored. 1 Coordination and Raising of Verbs and Subjects English has a constructionin which finite verbs are conjoinedto share their arguments. (1) John [copied and filed] the paper. Two types of translationof sentence (1) are possible in Japanese.' (2) a. John-ga sono ronbun-o [kopiisi fairusi]-ta.
John-NoM that paper-ACC [copy file]-PAST

b. John-ga sono ronbun-o [kopiisita sosite fairusita].


John-NoM that paper-ACC [copied and filed]

In the first type (2a), the first of the juxtaposed verbs is realized as the bareform kopiisi 'copy', while the second is followed by the tense morphemeta.2 In the second type (2b), both verbs are accompanied by the tense morphemeta, and the coordinateconjunctionsosite 'and' conjoins them. Using a diagnosis of coordinationdiscussedby Carlson (1987), Takano (2004) proves that only (2a) involves coordinationof two verbs. The diagnosis involves interpretationof adjectives like differentand same. Consider (3). (3) Smith went to a differentplace on his vacation this year. The adjectivedifferentimplies that some comparisonis made. In (3), for instance, both the speaker and the hearers know that Smith had gone to some other place already,and the speakerasserts that Smith went to a place different from that. It can be said, therefore,that the basis of comparisonfor differentin (3) lies outside the sentence. Carlson calls this kind of interpretation of different a sentence-external reading. Next, consider (4a). (4) a. Bob and Alice attend differentclasses. b. Bob attends Biology 101 and Alice attends Philosophy 799. Like differentin (8), differentin (4a) has a sentence-external reading.

1 The coordinatingconjunction sosite 'and' can optionally be inserted between predicatesin (2a). See Takano2004:fn. 9 for some comments on the issue. Native speakers of Japanesehave a clear intuition that an example like (2b) consists of two independentsentences, ratherthan a single sentence with a coordinationstructure,as Takano(2004:172) observes. Let us assume nonetheless, for the sake of discussion, that it is a single sentence. 2 A tense morpheme obligatorilyoccurs sentence-finallyin Japanese:this, coupled with the language's canonical OV order,makes it ratherhardto know whetherverb raising takes place or not, by looking at the right conjunct of a coordinationstructure.I will thereforeconcentrateon left conjuncts in what follows, when considering verb raising in Japanese.

320

SQUIBS

AND

DISCUSSION

In addition, it has a reading on which the class that Bob attends and the class that Alice attends are comparedsentence-internally,giving rise to an interpretation like (4b). Carlsoncalls this type of interpretation a sentence-internalreading. A sentence-internalreading is obtained when the sentence contains a coordinationstructure,as is the case in (4a). Like conjoined DPs, as in (4a), conjoined verbs license sentence-internal readingsof different,as demonstrated by the examples in (5). (5) a. John [saw and reviewed] differentfilms. b. Different dogs [chased and bit] the cat. Thus, availability of a sentence-internalreading tells us whether a given sentence contains coordination. Takano,making use of this diagnosis, establishes that the juxtaposed verbs in (2a) but not (2b) instantiatecoordinationof verbs (i.e., (2b) consists of two complete sentences ratherthan one sentence involving coordination).Consider(6a-b), in which the adjectivebetubetuno 'differentfrom each other' modifies the object sharedby juxtaposed verbs. (6) a. John-ga betubetunoronbun-o [kopiisi fairusi]-ta. John-NoM different paper-Acc [copy file]-PAsT 'John copied and filed differentpapers.' b. *John-ga betubetunoronbun-o [kopiisita sosite John-NoM different paper-Acc [copied and
fairusita].

filed] 'John copied and filed differentpapers.' Unlike differentin English, betubetunoin Japaneseyields only a sentence-internalreading. (6a), where the first verb appears as a bare form, allows a sentence-internal reading,namely, thatJohncopied one paper and filed a differentpaper. Such a readingis excluded in (6b), where the first verb is followed by the tense morphemeta. In fact, the sequence ronbun-okopiisitasositefairusita in (6b) can only mean thatJohncopied and filed the same paper,an interpretation thatcontradicts the intrinsic meaning of betubetuno.Since betubetunolacks a sentence-external reading,sentence (6b), with no felicitous interpretation, is unacceptable.3 Fromthese observations,Takanoconcludesthat Japaneseverbs must be conjoinedin bareforms to producea structure like (7), where verbs necessarily stay within vP.4
I The first predicatein (6a) can take a gerundiveform, producinga string like kopiisi-tefairusita 'copying and filed', which Takanoalso shows is not a case of coordination.

(i) *John-ga betubetunoronbun-o [kopiisi-te fairusil-ta. John-NoM different paper-Acc [copy-ing file]-PAsT 'John copied and filed differentpapers.'
4

Takano is not explicit on the matterof V-to-v raising. I will take the

issue up in section 2.

SQUIBS

AND

DISCUSSION

321

(7)

T'

(vP)

VP

(v)

DP

&

This line of reasoningis easily carriedover to examplesof coordination of subject-predicate sequences. First, let us make sure that the diagnosis mentioned above works for subject-predicate sequences in English. (8) a. In differentrooms, John listened to the radio and Mary wrote letters. b. John listened to the radio in room 335 and Mary wrote letters in room 175. (9) a. On different days, John went fishing and Mary went shopping. b. John went fishing on Monday and Mary went shopping on Thursday. In (8a), wheretwo subject-predicate sequencesareconjoined,different has a sentence-internal reading,so we can deduce somethinglike (8b) from (8a). Likewise, (9a) yields a sentence-internal readinglike (9b). Now, let us examine coordinationof subject-predicate sequences in Japanesein this light. (10) a. Betubetunoheya-de [John-ga razio-o kiki different room-in [John-NoM radio-Acc listen.to Mary-ga tegami-o kail-ta.
Mary-NOMletter-Acc write]-PAST

'In differentrooms, John listenedto the radioand Mary wrote letters.' b. *Betubetunoheya-de [John-ga razio-o kiita different room-in [John-NoM radio-Acc listened sosite Mary-ga tegami-o kaita]. and Mary-NoM letter-Accwrote] 'In differentrooms, Johnlistened to the radioandMary wrote letters.'
Fukui and Sakai (2003:344-345) offer almost the same argumentas Takano, on the basis of facts regardingpredicatecoordinationin terms of katu 'and'.

322

SQUIBS AND DISCUSSION

(11) a. Betubetunohi-ni [John-ga turini iki Mary-ga different day-on [John-NoM fishing go Mary-NOM kaimononi it]-ta. shopping gO]-PAST 'On different days, John went fishing and Mary went shopping.' b. *Betubetunohi-ni [John-ga turini itta sosite different day-on [John-NoM fishing went and Mary-ga kaimononi itta]. Mary-NOM shopping went] 'On different days, John went fishing and Mary went shopping.' As the grammaticality of (lOa) and (1la) indicates, betubetunois licensed when the first verb in the juxtaposed elements assumes a bare form. When the first verb is accompanied by the tense morpheme ta, as in (lOb) and (1 Ib), the sentence becomes unacceptable,with betubetunocontradictorily modifying a single event. It is reasonable, therefore,to considerexamples like (1Oa)and (11a) to be cases where subject-predicate sequences,to the exclusion of tense-related elements, are conjoined as a complement of T, as illustratedin (12). (12) T'

vP

vP

vP

subject

V-v

subject

V-v

An interestingconclusion to be deduced from this is that subjects, as well as verbs, do not overtly evacuate vP, at least in coordination structuresin Japanese. 2 Kishimoto's Argument for Subject Raising Kishimoto (2001) offers an illuminating explanation for a subjectobjectasymmetrywith regardto indeterminate pronounbinding,based
5 While I do not have a ready explanationfor the fact that Japanesedoes not allow finite TPs to be conjoinedby sosite 'and', as in (lOb) and (1 lb), it is worthpointingout that such a structure can be createdby a differentcoordinate conjunction,ga 'but'.

(i) Betubetunokuruma-de[John-wa turini itta ga Mary-wa


different car-by [John-ToP fishing went but Mary-TOP

kaimononi itta]. shopping went] 'In different cars, John went fishing but Mary went shopping.'

SQUIBS

AND

DISCUSSION

323

on the assumptionthat subjects overtly move to Spec,TP in Japanese. Japanese indeterminatepronouns like nani 'what', dare 'who', and itu 'when' functionas negative polarityitems only when appropriately bound by the Q particle mo. Kishimoto assumes that mo can bind those elementsthatareincludedin the maximalprojectionwhose head contains it. Thus, if mo is contained in a V-v amalgam, then every element under vP is potentially bound by it. Observe the relevant examples in (13). kai-mo si-nakat-ta. (13) a. Taroo-wa nani-o Taroo-ToP anything-Accbuy-Q dO-NEG-PAST 'Taroo did not buy anything.' b. Taroo-wa dare-ni ai-mo si-nakat-ta. Taroo-ToP anyone-DAT meet-Qdo-NEG-PAST 'Taroo did not meet anyone.' In each sentence in (13), mo is attachedto the V-v amalgam(afterVto-v raising),so thatthe accusativeobject in (13a) andthe dativeobject in (13b) are properlylicensed by mo within vP. Given that Japanese lacks overt V-to-T raising,as we saw above, it is quite understandable that nominative subjects fail to be bound by mo, as in (14a), under the assumptionthat subjectsovertly move out of vP to Spec,TP.6The argumentcan be extendedto cases involving nonnominativesubjects. It is well known that dative-marked argumentsof a cluster of stative predicatesin Japanese,such as uta-e-ru 'can sing' andwaka-ru'understand',exhibit subjecthoodwith respectto some grammatical phenomena like reflexivizationand honorification.Kishimotocorrectlypoints out that such dative subjects also cannot be bound by mo attachedto lexical verbs, as (14b) shows.7

6 Notice thatV-to-v raisingis crucialfor the argument thatthe nominative subject in (14a) undergoes overt movement to Spec,TP. Without the former operation, the subject can stand outside the binding domain of mo even in Spec,vP. 7Kishimoto (2001:606) arguesthatthe following example of vP-fronting confirms the view that the dative subject in a sentence like (14b) is overtly raised to Spec,TP.

(i) ?[sono-uta-ga uta-e-mo]i Taroo-ni ti si-ta koto [that-song-NOM sing-can-Q]Taroo-DATdo-PAST fact '(the fact that) Taroo could also sing that song' A nominative object like that in (14b), which Kishimoto (2001:606) assumes to reside within vP in overt syntax, also fails to be bound by mo. (ii) *Taroo-ni nani-ga uta-e-mo si-na-i.
Taroo-DAT anything-NOM sing-can-QdO-NEG-PRES

'Taroo cannot sing anything.' He claims that the nominativeobject undergoescovert phrasalmovement out of vP, which explains why mo does not bind it.

324

SQUIBS AND DISCUSSION

(14) a. *Dare-ga

warai-mo si-nakat-ta.
dO-NEG-PAST

anyone-NOM laugh-Q

'Anyone did not laugh.' b. *Dare-ni sono uta-ga uta-e-mo si-na-i. anyone-DAT that song-NOM sing-can-Qdo-NEG-PRES 'Anyone cannot sing that song.' The sentences in (15) are additionalcases in point. (15) a. *Dare-ga kao-o arai-mo si-nakat-ta.
do-NEG-PAST anyone-NOM face-ACC wash-Q

'Anyone did not wash her/his face.' b. *Dare-ni gitaa-ga hik-e-mo si-nakat-ta.
anyone-DAT guitar-NOM play-can-Q do-NEG-PAST

'Anyone could not play the guitar.' Citing these examples, Kishimoto maintains that the category T in Japanesecontains an EPP feature and that subjects, including dative ones, are requiredto move overtly to Spec,TP to check it. One might argue alternativelythat the data in (14) and (15) can be explained with no recourse to overt subject raising if we assume, contraryto Kishimotobut in agreementwith FukuiandTakano(1998) and Watanabe(2000), that Japaneselacks V-to-v raising. In thatcase, mo would attach to V, and anything outside VP, like the subjects in (14a-b) and (1Sa-b) (which are presumablyintroducedinto phrase structureat the vP stage), would escape being bound by mo, whether overt subject raising takes place or not. This explanation,therefore, would invalidatethe argumentfor overt subject raising based on the examples in (14) and (15). However, additional examples indicate that the subject-raisingaccount is more promising in dealing with indeterminate pronounbinding. Considerthe sentences in (16) (from Kishimoto 2001:604). (16) a. *Nani-ga
anything-NoM

(Taroo-ni) yom-are-mo si-nakat-ta.


Taroo-DAT read-PASS-Q do-NEG-PAST

'Anything was not read (by Taroo).' b. *Nani-ga koware-mo si-nakat-ta.


anything-NOM break-Q
do-NEG-PAST

'Anything did not break down.' The passive subjectin (16a) andthe unaccusativeone in (16b), though originatingfrom VP-internal positions, arenot boundby mo, a fact that is explainedin terms of overt subjectraisingbut not by the alternative scenario. 3 A Problem and Its Possible Solution We saw in section 1 that subjects are not overtly raised out of vP in coordinationstructures in Japanese,andin section 2 thatthereis reason to believe thatsubjectsovertly move out of vP. Logically, these results are compatible:while subjectsstay in vP in some contexts (namely, in coordinationstructures), they move out of vP in others.Unfortunately,

SQUIBS

AND

DISCUSSION

325

however, the case is not thatsimple. The examples thatsupportsubject raising, such as (15a-b), can be put in coordinationstructures,but they remain ungrammatical. (17) a. *[Dare-ga
[anyone-NOM

kao-o
face-Acc

arai-mo se-zu
wash-Q

dare-ga

do-NEG anyone-NOM

ha-o
teeth-Acc

migaki-mo si-nakat]-ta.
brush-Q do-NEG]-PAST

'Anyone did not wash her/his face and anyone did not brushher/his teeth.' gitaa-ga hik-e-mo se-zu b. *[Dare-ni
[anyone-DAT guitar-NOMplay-can-Q do-NEG

dare-ni
anyone-DAT

sakkusu-ga
saxophone-NOM

huk-e-mo si-nakat]-ta.
play-can-Q do-NEG]-PAST

'Anyone could not play the guitar and anyone could not play the saxophone.' pronounsin (17a-b) with normal When we replace the indeterminate DPs, and insert the resulting sentences in betubetunocontexts, we obtain sentence-internalreadings, a sure indication that (17a-b) involve coordinationstructures.8 (18) a. Betubetunohi-ni [John-ga kao-o arai-mose-zu face-Acc wash-Q do-NEG different day-on [John-NoM
Mary-ga ha-o migaki-mo si-nakat]-ta.
brush-Q do-NEG]-PAST Mary-NOM teeth-Acc

'On differentdays, John did not wash his face and Mary did not brush her teeth.' b. Betubetunokonsaato-de [John-ni gitaa-ga different concert-at [John-DAT guitar-NOM hik-e-mo se-zu Mary-ni sakkusu-ga
play-can-Q play-can-Q do-NEG Mary-DAT saxophone-NOM do-NEG]-PAST

huk-e-mo si-nakat]-ta. 'At differentconcerts,Johncould not play the guitarand Mary could not play the saxophone.' In (17a-b), each conjunctcontains a subject-predicate sequence without a tense morpheme,but still the subject fails to be bound by mo
amalgamated with the lexical verb. It is quite reasonable to assume

that neither conjunct projects up to TP, because no tense element is


8 As an anonymousreviewer points out, some examples with zu 'not' in the first conjunct, as in (i), lack a sentence-internal reading.

(i) *John-ga betubetunoronbun-o [kopii-mo se-zu fairu-mo John-NoM different paper-Acc [copy-Q do-NEG file-Q si-nakat]-ta.
dO-NEG]-PAST

The reason might be that the bare and gerundive forms of zu 'not' are the same, and the sentence is therefore ambiguous between a coordinate and a noncoordinatestructure.For a related matter,see footnote 3.

326

SQUIBS

AND

DISCUSSION

detected within either of them. This in turn means that the subject does not move out of vP, no landingsite being availablein the conjunct. It is, at the same time, sensible to suppose that each subject overtly moves out of vP, since neitheris boundby mo. Clearly, these conclusions are contradictory,if we stick to a TP-vP type of approachto To overcome the problem,let me revive and employ phrasestructure. the Split-IPHypothesisoriginallyproposedby Pollock (1989), according to which IP is divided into TP and AgrP. The hypothesis allows us to represent(17a-b) schematicallyas in (19). (19) AgrP T

AgrP

AgrP

subject

Agr'

subject

Agr'

vP

Agr

vP

Agr

V-v-mo-'do'-Neg

V-v-mo-'do'-Neg

In (19), each conjunctprojectsup to AgrP, and T takes the conjoined AgrP as its complement.The subjects move out of vP to Spec,AgrP, which explains why mo does not bind them. Notice also that T lies outside the conjuncts,correctlypredictingthat no tense morphemeis allowed to occur inside of them. The analysis in (19) thereforeguarantees a derivationin which subjectsmove out of vP in nontensedconjuncts.9
9 It is necessary to assume that subjects,but not verbs, move up to AgrP, to ensure that subjects escape being bound by mo attachedto lexical verbs. If verbs as well moved up to Agr, then anythingwithin AgrP, including raised be bound by mo. We can articulatethis by subjects, would inappropriately stating that Agr in Japanesehas a [ + N] feature, which overtly attractsa DP to its Spec, and a [- V] feature,which merely checks a relevantmorphological manifestationon the verb in vP. This propertyof Agr in Japaneseis analogous to that of T in English. As correctly pointed out by an anonymous reviewer, (i) is also a workable structurein this regard. subject V-v] & [vPsubject V-v]] Agr T (i) [vP[vP In (i), TP and AgrP are projectedoutside the conjoined vP, and the head Agr checks relevant morphologicalforms on the V-v amalgamsin an across-theboardfashion.I do not use the structure in (i), however, since it does not provide landing sites for subjectraising in conjuncts,leaving the problemaddressedin this squib unsolved.

SQUIBS AND DISCUSSION

327

The structure proposedin (19) not only solves the problemabout raising of subjects but also receives support from a morphological consideration. The structure in (19) predicts that agreement morphemes can be manifested on verbs inside of, and independentlyof, the tense morpheme.'0The prediction is in fact fulfilled by subject one of the diagnosesfor subjecthoodin Japanese,which honorification, is extensively discussed in Shibatani 1978. Each Japanesepredicate has an honorific form as its alternativemorphologicalmanifestation. An honorific form is used when the subject of the predicaterefers to a person for whom one wants to show respect. In (20), for example, the predicate sikaru 'scold' is replaced by its alternativehonorific form osikarininaru,in honor of the referentof the nominativesubject, Tanaka-sensei'Prof. Tanaka'. (20) Tanaka-sensei-ga
Tanaka-teacher-NOM

Taroo-o
Taroo-Acc

osikarininat-ta.
scold(HON)-PAST

'Prof. Tanakascolded Taroo.' In contrast, accusative objects do not induce honorification,as the awkwardnessof (21) indicates. (21) *Taroo-ga Tanaka-sensei-o ayamatte mistakenly Tanaka-teacher-Acc Taroo-NoM onagurininat-ta.
hit(HoN)-PAST

'Taroo mistakenlyhit Prof. Tanaka.' The same holds true of dative objects. (22) *Taroo-ga Tanaka-sensei-ni 'Taroo saw Prof. Tanaka.' Dative subjects,on the otherhand,do triggerhonorification,as demonstratedin (23), which is one of the reasons why they are widely believed to be subjects. (23) Tanaka-sensei-ni girisyago-ga owakarininar-u.
Tanaka-teacher-DAT Greek-NOM understand(HoN)-PRES

oaininat-ta.

TarOo-NoMTanaka-teacher-DAT see(HON)-PAST

'Prof. TanakaunderstandsGreek.' Thus, we can regardhonorificationof this kind as the morphological honormanifestationof subjectagreementon predicates.Interestingly, ific forms of verbs can occur in conjunctsof coordinationstructures, where subject-predicatesequences with no tense elements are conjoined.

Admittedly, the data presented below cannot count as evidence that AgrP (or an extra maximalprojection)is containedin each conjunct(see footnote 9). Still, in my opinion, they provideevidence thatagreementmorphemes are realized inside of, and independentlyof, T, which is also predictedby the structurein (19).

10

328

SQUIBS

AND

DISCUSSION

(24) a. [Tanaka-sensei-ga Suzuki-sensei-ga


Suzuki-teacher-NoM

biiru-o wain-o
wine-AcC

onomininari onomininat]-ta.
drink(HoN)]-PAST

[Tanaka-teacher-NoM beer-Acc drink(HON)

'Prof. Tanakadrankbeer and Prof. Suzuki drankwine.' b. [Tanaka-sensei-ni girisyago-ga owakarininari


[Tanaka-teacher-DAT Greek-NOM understand(HoN)

Suzuki-sensei-ni
Suzuki-teacher-DAT

ratengo-gaowakarininar]-u.
Latin-NOM understand(HoN)]-PRES

'Prof.Tanakaunderstands GreekandProf. Suzukiunderstands Latin.' Observethat onomininaruand owakarininaru, the honorificforms for nomu 'drink'andwakaru'understand', are realizedin theirbareforms in the left conjuncts of the examples in (24). This fact clearly shows that agreementmorphemescan be realized separatelyfrom the tense morphemeinside of the latter,as the structure in (19) correctlypredicts. Thus, the structural analysis in (19), proposedto ensure that subjects move out of vP in nontensed conjuncts, receives quite independent supportfrom morphology. 4 Concluding Remarks To conclude this squib,I would like to considersome generaltheoretical implicationsthat it offers for linguistic theory. The original SplitIP Hypothesis advocated by Pollock (1989), which decomposes IP into TP and its complementAgrP, was soon replacedby the approach that assumes the opposite, AgrP-TP structure.As Haegeman (1994: 599) argues, the reason is that agreement-related morphemesappear outside of tense-relatedones in inflected verbalelements in some languages like French and German.Yet the ordercannot be universally constant, because we have seen here that agreement-relatedmorphemes are found inside of tense-related ones in Japanese, as in (24a-b). We can conjecture that Agr, which plays a minor role in is subject to parametricvariations with regard to its interpretation, realizationaroundT. What is peculiar and intriguingabout the examples in (17) and (24) is that EPP effects are detected inside TP. We can understand this state of affairsin the following way. Supposethat Agr ratherthan T has an EPP feature crosslinguistically.What varies from language to languageis where AgrP is locatedin phrasestructure. If it is located above TP, then we have a language like French or German, where agreement-related morphemesappearoutsideof tense-related ones and a familiar type of EPP effect is produced. On the other hand, if it is located below TP, then we have a language like Japanese, where agreement-related morphemesare realizedinside of tense-relatedones and where EPP effects, caused by Agr, manifest themselves inside TP, as observed in (17) and (24).

SQUIBS

AND

DISCUSSION

329

References Carlson,Greg. 1987. Same and different:Some consequencesfor syntax and semantics. Linguisticsand Philosophy 10:531-565. Fukui, Naoki, and Hiromu Sakai. 2003. The visibility guideline for functionalcategories:Verb raising in Japaneseand relatedissues. Lingua 113:321-375. Fukui, Naoki, and Yuji Takano. 1998. Symmetry in syntax: Merge and Demerge. Journal of East Asian Linguistics 7:27-86. Haegeman, Liliane. 1994. Introductionto Governmentand Binding Theory.Oxford:Blackwell. Kishimoto, Hideki. 2001. Binding of indeterminatepronouns and clause structurein Japanese.LinguisticInquiry32:597-633. Pollock, Jean-Yves. 1989. Verb movement, Universal Grammar, and the structureof IP. LinguisticInquiry20:365-424. Shibatani,Masayoshi. 1978. Nihongo no bunseki.Tokyo: Taishukan. Takano, Yuji. 2004. Coordinationof verbs and two types of verbal inflection. LinguisticInquiry35:168-178. Watanabe,Akira. 2000. Remarkson head movement within VP shell. In Proceedings of the Nanzan GLOW: The Second GLOW Meeting in Asia, ed. by Yasuaki Abe, Hiroshi Aoyagi, Masatake Arimoto, Keiko Murasugi, Mamoru Saito, and Tatsuyo Suzuki, 461-485. Nagoya: NanzanUniversity,Departmentof Japanese.

RAISING TO OBJECT IN JAPANESE:

1 Introduction The Japanese exceptional Case-marking(ECM) construction,illustratedin (lb), has been most successfully analyzed as an instance of subjectraisingto object (RO), as in (lc) (e.g., Kuno 1976, Sakai 1996, 1998, Ohta 1997, Bruening 2001, Tanaka2002). (1) a. Kanojo-wa[sono otoko-ga sagishi da to] shinjiteiru. she-TOP the man-NOM swindleris QUOTbelieves 'She believes that the man is a swindler.' b. Kanojo-wa[sono otoko-o sagishi da to] shinjiteiru. she-ToP the man-Accswindleris QUOTbelieves 'She believes the man to be a swindler.' (Kitagawa 1986) c. Kanojo-wa sono otokol-o [t1 sagishi da to] shinjiteiru. JapaneseRO appearsto be optional, given that the embeddedsubject
I would like to thankHowardLasnik, TimothyJ. Vance, Yoshihisa Kitagawa, SusumuKuno, KaoruOhta,Wako Tawa, OxanaBorzdyko,David Heap, Hajime Hoji, Jacques Lamarche,David Michaels, and Yehor Tsedryk. I also thank two anonymous reviewers for their valuable comments. I am indebted to Jami B. Kawai for editorialimprovement.As always, remainingerrorsare my own.

A SMALL CLAUSE ANALYSIS Michiya Kawai Huron UniversityCollegel Universityof WesternOntario

You might also like