Readings - 350 z 乃 regWoofesonSyn れ tac4icM@ovement ぬ...

Info iconThis preview shows pages 1–13. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 350 z 乃 regWoofesonSyn れ tac4icM@ovement ぬ @口 pan JQ & M. Wescoat, eds.. Proceedings of the West Coast Conference on Formal Lin ツ guistics Volume 2, Stanford University, Stanford. Sait0, M. (1984) "On the Definition of C-command and Government", in C. Jones & P. Sells, eds.. Proceedings of the Fourteenth Annual Meeting of the North Eastern Linguistic Chapter 10 四e Society, University of Massachusetts, Amherst. I 『.O ra 亡互 Onah y para 田ete 工 S 木 亡 John Whitman Sait0, M. (1985) Some Asymmetries Japanese and Their Theoretical Implications, in Doctoral dissertation, MIT. Stowel1,T. (1981a) Origins of Phrase Structure, Doctoral dissertation, MIT. Stowel1, T. (1981b) "Complementizers and the Empty Category Principle", in V. Buike and J. Pustejovsky, eds.. Proceedingsof the Eleventh Annual Meeting of the North Eastern Linguistic Society, University of Massachusetts, Amherst. Stowel1, T. (1985) "Null Operators and the Theory of Proper Government", s., m UCLA. in Y. Otsu & A. Farmer, eds., Tonoike, S. (1980) "More on Intra-Subjectivization", MIT Working Papers in Linguistics Volume 2: Theoretical Issues in Japanese Linguistics, Dept. of Linguistics and Philosophy, MIT. Toneg0, E. (1984) "On Inversion in Spanish and Some of its Effects", Linguistic Inquiry 15.1. Wah1, A. (1985) "Two Types of Locality", ms., USC and University of Maryland. Whitman, J. (1979) "Scrambled, Over Easy, or Sunny Side Up?" in P. dyne, W. Hanks & C. Hofbauer, eds.. Papers from the Fifteenth Regional Meeting of the Chicago Linguistic Society, University of Chicag0, Chicag0. Whitman,J. (1982) "Configurationality Parameters", ms.. Harvard University [included in this volume], 1. INTRODUCTION Discussion of configurationality as a typological parameter in syntax has focused on two observed features of a putative class of 'nonconfigu- rational'languages: (1) a. b. The absence of an overt reflex of Move-a The occurrence of 'scrambling'-type free word order Recent work on Japanese, by far the best-studied examplar of this putative typological class, has cast doubt on the claim that (la) and (Ib) are in fact properties of Japanese or similar languages (see Kuroda 1980, Masunaga 1982, Saito 1983, 1985 for arguments that at least some instances of 'scrambling'-type free word order in Japanese are derived by a syntactic movement rule). Focus on this important issue has obscured the original 'configurationality' debate, which was concerned with a parameter for syntactic configurationality that actually antedates concern with the features in (1). This parameter is presence or absence ofVP. We will see below that the actual statement of this parameter can be made substantially more complicated, but let us embark on the assumption that it is not. As early as Schwartz (1972), the proposal was made that languages with verb external arrangements of surface sentence constituents lack a constituent VP. The motivation for such a proposal on the basis of purely superficial order of constituents is obvious in the case of the VSO word order type preponderant among verb-initial languages. Arguments that verb-final languages may lack a VP constituent were advanced by Hinds (1974). Discussion of the issue of nonconfigurationality in the early 1980's crystallized around a number of proposals identifying lack of a VP con ツ stituent as the structural feature to be associated with the properties in (1). For a description of the most explicit of these proposals, see Hale * This paper is an extensively revised version of an original written in 1982 Con折糾 atfo 刀 4h ひ四 w 仰gfgKs (1982). The central nature of the VP issue in these proposals is often misunderstood, and thus worth restating here: the features in (1) are naturally associated with a 'flat' phrase structural analysis of maximal pealizeshichosition re hichoubjectsEliminating rould bsuch ainalysis istersead. nalysis rojectionsoS.¥Qnethe suchVP ia'w all p f a s rediction a n adjuncts madesan flat' p w t s hby tructurally w e that there s in no are non-subjects, thusthat relative and their position indeteris ThE ument5 'scrambling'asyntactic aコ g FOI te deriving effects by move- 口而 打 ment rule demonstrated the incorrectness of the attempt to associate absence of VP with the features in (1). At the same time, examination of the details of grammatical function distribution in languages identified as nonconfigurational demonstrates the need for a constituent corres- syntactic representation. The subject-0bject asymmetries observed in Marantz (1981) appear to exist in nonconfigurational languages as well ^Eor'discussion of such asymmetries in Japanese see Hasegawa (1981) a Chomsky's (1981) treatment of 'nonconfigurational' languages anticipated that a configurational representation would be required at least at the level of argument structure. On his account, grainmaticaLfunclions_ are defined by dominancerelations and associated with particular thematic roles in the lexicon; i.e. subject is defined as [NP, S] and object as [NP, VP] as in Chomsky (1965), and these GPs are associated with the thematic roles assigned y particularverbsexactly as in English.This 'configurab tional' representation of GFs in the lexicon is paired with a 'flat' representation at D-structure (yvhere grammatical functions are assigned in random order to argument positions, all of which are sisters of V: (2) Lexicon subject -roles Q 岬, '] objectQ-roles The major motivation for including the flat D-structure representa ツ tion in (2) was to derive the surface effect of free complement order. As suggested at the outset of this section, if the argumentsfor generating free constituent order at D-structure are removed, then the argument for a non-configurational D-structure representation also substantially disappears. The purpose of this paper is to reconsider what aspects of phrase struc而 n司ration ty afeう au L parameters f universal rammar, to 戒 l託 o g "the extent that this can be determined by investigation of the syntactic propertiesof languages Japanese. Dwill commence by examining like W tu the issueof the role of VP, concentrating on syntacticphenomena other than those in (1) that crucially bear on this issue. Section 2 will deal with cases where the existence of a maximal projection dominating comple ツ ments of the verb but not subjects appears to be motivated. Section 3 will introducethe traditiona1, largelynegative evidence againstthe existence of a constituent VP. Section 4 will propose a somewhat subtler account of the issueof phrase structural onfigurationality. c 2. EVIDENCE FROM BINDING THEORY FOR S-STRUCTURE CONFIGURATIONALITY Our purpose in the immediately followingsectionis to investigate what evidenceBinding Theory gives us about the propertiesof S-structure in Japanese,somewhatthe opposite of the usual approach in these matters. It is first necessary to establish that the binding conditions we i p are interestedn (primarilythoseaffecting ronominalsandr-expressions) applyat the levelof S-structuren thislanguage. ecan do this by appeal i W ツ ing to the evidence Chomsky 1981) citesfor English: that ( 5 ( ab HHM 欣 [NP,VP] D-structure g[NP^NP. V] (i,j arbitrary) The conflgurationally defined grammatical functions are also available at D-stnicture as input to the analogues of Move-a in nonconfigurational languages. To satisfy the Projection Principle (Chomsky 1981), it is as ツ sumed that a configurational representation must also be available at Sstructure. As a result, versions of the approachoutlined here have assumed that configurational flat phrasestructure representationsre paired and a both at D-structure and at S-structure. l卜、ト、 353 John Whitman 「 ・ 352 which oohn'soks bo Jf J ll ohn's o af books The impossibility f coreference between he andJohn in (5) shows that o condition of the BindingTheory of Chomsky(1981),which statesthat C r-expressions must be free (for our purposes, not c-commanded by a c0- indexed P), must applyat a levelof representation N otherthan LF. This テ sbecauseatL 耳 ttbe , P肚aseWhichofJohn@sboo 祐狙 d 山equ 甜 t 田er phraseall of John's books are presumed to be raisedto to an operator , position in which they are no longer c-commanded by he The same facts obtain in Japanese 354 (6) Configurationality parameters a. *Kare.-ga John-no dono hon-ga itibansuki ka GEN which bookNOMmost like 0 he NOM 也止 Lw 田S0 " c M w he M es whi h F 了 。而L ML 。 口 b ・ the following examples indicatehat a non-c-commanding t pronominal may readily precede its antecedent: (9) Kare.-no okusan-ga John.-0 yasinat-te i-m rasi-i him GEN wife NOM ACC support RO bePRES seem PRES P 'It seems that his.wife is supportingohnJ (10) Kare.-nokodom0-tati-ga oks *hK寸伊 John.-noGEN NOMde 打 GEN all hook like a subete-no suki b on-ga b e NOM 2J ・ , coreference between the kare and John investigate the effects relation that determines binding domains as a diag ツ structure at a level distinct from LF. er 亡 HO pronoun 乃 oぴれ $ It has been proposed by Mohanan (1981) that binding domains in nonconfigurational languages are not defined by reference to a conflgurational relation (such as c-command), but rather by linear precedence.Let us con ツ sider why this proposal makes some sense before we reject it out of hand ム for we shall see in section 3 that it is partially correct. If we were to take c-command as the relation for defining binding domains in a language which lacked VP as a maximal projection, the binding domain of any matrix constituent would be the entire S that contains it. This would predict, for example, that a matrix NP objects could never c0-0ccur with a coreferent r-expression in the same S. As Mohanan argues, this is clearly wrong: (7) 8 L JT卜 而 伊0MoC atb ぬT A L ・呵 u酔 N C0 P S A 。 John^-0 amarisuki zya na-ikot0-wa ACC very fond of be not PRES 'It is unmistakable thathis-i hildren on't like Johni' c d can, in Japanese as in English, ぬ な z 乙 DO 川g加Sw ぬL れ him GEN children-PLUR NOM maFg 石 -n 士テ COMP THEME mistake not SeMh L ) likes ど 。血 ok 川7 The impossibility of in (6) shows that we of the configurational nostic for hierarchical 355 John Whitman nad c b e agasit Sp 川 ES 「rPR It is just as easy to show, however, that the correct condition for pronominal binding cannot be defined in terms of precedence alone; How can we account for the acceptability f all of (7-10)?So far, all the o datawe haveseen pattern exactlythe corresponding datain English. he T most obviousoption is to reject the assumption that (7-10) have a flat phrasestructuralrepresentationt the levelat which conditionC of the a Binding Theory applies,and suppose insteadthat the relevantphrase structuralrepresentation (7-10) is equivalent o the one we would for t assignto the corresponding Englishsentences.f (7-8) are assigned I the configurational representation (11-12), then the pronouns in eachPX in ツ ample would cease to c-command their antecedents: (11) て二 S 7 丁 M NP I John,-n0-itiban sitasi-itomodati-ga kare.-0 V ura 可 t-ta 356 , (12) 「 「 ( 二 8) 357 John Whitman COn匝g ぴ 7口 Iのれ口 Z@ pdrg 川gfg $ ゆ The featurespecification f PRO as [+anaphor, pronominal] requires o + S M ly not governed y the verb in Japanese. b This fact would be consistent with the presence of VP as a maximal projection dominating object but not subject position. [John.-ni uwasa-0], kansu-ru kare.-ga t. nagasi-ta The analysis in (12), following Saito (1983, 85), assumes that the 'scramツ bled' accusative-markedNP John-ni kansu-ru iwasa-0 'rumors about John (ACC)' has been adjoined to S by Move-a and occupies a position not c-commanded by the subject. Examples like these are in fact among the strongest evidence that 'scrambling'-type rearrangements of unmarked word order constitute movement to a specific structural position in Japanese. In (11) as wel1, we would appear to be forced to assume the presence of a maximal projection between the direct object kare-0 and the subject John-no itiban sitasi-i tomodati. The most plausible candidate is of course VP. 2.2 ・ Dぬ か 妨ぴぬ o 『 oダ PRO Saito (1982) shows that PRO in Japanese, as in English and other configurational languages, may occur only in subject position. The arguments in Japanese are slightly complicated by the fact that the 'zero pronoun' empty category occurs rather freely in Japanese in object and other nonsubject positions. Non-subject pro drop, however, is always subject to discourse anaphoric contro1, while subject PRO may have arbitrary reference. Thus the object in (13) Boku-ga mi-te mo dame da I NOM see GER even bad COP 'It's no good even if I see (she/him/it etc.)' can only be interpreted as coreferent to something mentioned in prior discourse, while the subject in 2 may Honmonu- o mi-te mo dame da real thing ACC see GER even bad COP 'It's no good even if (PRO=ARB) sees the real thing' have arbitrary reference O万 L巧L O CO乃 かり ヴ て ・ 丁 Japanese, English,exhibitsobligatorycontrolof subjectpositionin like certain contexts: (15) John^-ga [PRO^ ik-0o NOM 'John tried to go' to si-ta go SUBJCOMPdoPAST (16) John-ga Maiy^-iu [PR0, ik-]-ase-ta DAT NOM 'JohnmadeMarygo/come' go CAUS PAST The embedded subjectpositionin (15-16) can only be occupied y an b empty category coindexed with a matrix controlleras indicated. t is I extremely significant note at this juncture that after considerable research n the phenomenon f controlin ostensibly o o nonconfigurational languagescf. Mohanan ( 1982on Malayalam;George andKornfilt 1981on Turkish),no instances haveemerged of obligatory controlof non-subject positions. It seems safe to conclude that obligatorycontrolis universa1ロロ ly restricted to this position This resultof course fallsout if we followthe conventionalassumption that only PRO is subjectto obligatory controland only subjectposition (among argument positions)may be ungoverned.Again,sucha fact would appear to have serious consequencesfor a proposal based on a noncon- figurational rrangementof subjectand objectin phrase a structure, sinceit normallyassumed to be a structural roperty of subject p positionthat it is be ungoverned can 24 , (14) ・タ , メ れ口 pAorB 加d加 L A debatein Japanese rammar which considerably antedates advent g the ofgeneratiYegrammarconce mS 日 血eeqUestion ofwhe 山町山 ereisjust 而Cation in Japanese a grammatical function 'subject' separate for from the categories f surfacecase marking (cf. Mikami 1953, 1972,Shibatani, o 358 Con匝四rofionGl@ 川 wmefex$ ひ 1978).The arguments for distinguishing subjectGF are basedon two a properties,control of SubjectHonorification and bindingof reflexives, which are associated with a class of NPs that appear with more than one surface case marker but which arguably represent a single GF, i.e. subject rould like to focus on the argument from binding of reflexives, sincethis is the argument that is most clearlyrelevanto the theoretical t framework under discussion. It is tempting to turn this argument into an positionassociated with the GF subjectdefined argument for a structural at all levels of representation. This would follow if we assumed that the arguments in 2 for applyingthe conditions f the BindingTheory in So structure were valid for all three conditions of the Binding Theory, in- subjects prior we can assume structure. and other anaphorsn S-structure. econd, the argument outlinedabove i S cannot explainwhy only subjectscan bind reflexives n S in Japanese. i 1ngenera1,entences like (17) with non-subject antecedents for reflexives s are unacceptablein Japanese.1 (17) *John-ga Mary.-ni NOM zibun.-niuite situmon si-ta t DAT self about question PAST 'Johnquestioned ary. about herselfM This observation and data is due to Akio Kamio (pc). Such cases appear to be restricted to genitive reflexives and even then are lexically highly constrained Since condition A is a structural condition, it is not all obvious why its application should be restricted in this way, i.e. why other c-commanding NPs in argument positionsmay not bind reflexives. ewill reW turn to this question in section 3 25 , ・ Lfnれ eurRrgecgdenceA 糾m ぬ 飢 In the literaturecited in section 1, two arguments haveappeared hich w appear to show that subjectsmust precedenon-subjects in underlying structure. Since both arguments show that subjectsmust precedenon- to NP fronting, an application of Move a in Japanese, that the level of underlying structure referred to is D- The first argument is due to Masunaga (1982). She shows that in 'scrambling'-type reordering f NPs in Japanese,subjectsmay precede o non-subjects regardless of their discourse status, while non-subject argu ツ ment NPs may precedesubjectsonly if they are discourse naphoric a (examplesn (19) are fromMasunaga 1982): i (18) a. Taro0-ga sonohon-0 kat-ta NOM that book ACC buy PAST 'Taroo bought that book' b. Hitori-no oimanohit0-ga sonohon-0 kat-ta one GEN woman NOM that book ACC buy PAST 'A woman ate that sandwich' a. Sono hon-0, Taro0-ga kat-ta that bookACC NOM buy PAST 'That book, Taroo bought' eludingconditionA, which is relevant reflexives otheranaphors. for and Condition A states that anaphors must be bound in their governing cate ツ gory. Defininga subjectpositionwhich c-commands ll otherargument a positions in S will insure that a subject will always bind non-subject re ツ flexives in S. However, there are two curious aspects to the argument turned this way. First, the argument in 2 for applying the conditions f the Binding o Theory in S-structureappliesto conditionsB (for pronominals) and C (for r-expressions), but not directly to condition A. As far as I 3@ aware, there is no direct argument for applying a binding condition for reflexives 359 John Whitman (19) b. ?/??Issatu-no hon-0, Taro0-ga kat-ta GEN book ACC 'A book, Taroo bought' one NOM buy PAST Masunaga argues that this generalization is most easily captured if scram ツ bling is view as proposing from a fixed underlying position, and that a general restriction will stipulate that only discourse anaphoric NPs may be proposed.2 The second argument is due to Kuroda (1980). Kuroda notes that whereasquantifiers binding object NPs may be separated from their bindee by an intervening subject, quantifiers binding subject NPs may not be separated from their bindee by an intervening object (examples from Kuroda1980): (20) a. Igirisuzin-ga utide-n0- kozuti -0 hutatu kat-ta EnglishmanNOM striking GEN mallet ACC 2 buy PAST 'An Englishman bought 2 "mallets of luck" ' b. Utide-n0-kozuti-0, 屯田ぽ血ga hutatukata striking GEN mallet ACC EnglishmanNOM 2 buy PAST 十 '(The) 2 "malletsof luck", an Englishmanbought 360 (21) Configurationality parameters a. Igirisuzin-ga saniun tide-n0-kozuti-o u ka-ta EnglishmanNOM3 strikingGEN malletACC buyPAST 'ThreeEnglishmen bought (the) "mallet of luck" ' b. *Igirisuzm-ga utide-n0-kozuti-0 sannin ka-ta EngUshman strikingGEN malletACC 3 buy PAST NOM These canbe explained if it is assumed subjects facts again that originate tnhepoftshat thatplaces surface io -initialt indee,non-subtect-subiect S right b 'quantifiera uantifiers t osition, float' q immediately i and order i derived y proposing f the non-subjectconstituent. b o Both of the arguments reviewed bove motivatea D-structure position a for subjectsin S-initialposition.Noticethat they do not explicitlymotivツ ate a structurallydefinedpositionfor the subjectGP, i.e. a positionas ツ sociated with the unique NP daughter f S. However, it is hardto seehow o a fixed linearposition(i.e. 'leftmostNP argument')could be specifiedn i any other way in currenttheory; thus both arguments make a prima facie case for a structuraldistinctionbetween subjectand non-subject argumentsat the levelof D-structure in Japanese. 2 6 Review ofArgumen6sforooon ガ れぬ Rep esentafion f 九W 戸別川 o o か ・ ・ 日e 川川 $eF " , 「 , C化 加 In 2.1-2.5we saw a numberof arguments for a confieurationalepresenr tation of Japanese phrasestructure. n 2 we sawevidencehat the condiI t Onsof 血eeBtndingTheoryofChom y(1981)― L p 血 iCul COndition 曲 打 C, the conditionthat restrictsthe distributionof r-expressions must ム applyto S-structure.n 2.1 we sawthat the application f condition to I o C sentences involvingr-expressions nd the overt pronoun kare givesthe a resultswe wouldpredicton the assumption that preposed NPsareoutside of the c-command omainof otherNPs in the sentence, nd subiects d a are outside of the c-command omain of non-subjects. his appearsto d T indicatethe existenceof somenode which 'dominates on-subjects but n which does not dominate subjects. e suggested hat VP was the most W t likelydesignationfor thisnode. 1n 2.2 and 2.3 we saw that the distributionof PRO appears to be restrictedo subject t positionn Japanese, i suggesting onlysubject that positionmay be ungoverned. 1n 2.4 we reviewed ne of the traditionalarguments for a structural o distinction between subject ndnon-subject a arguments Japanese, in that only subjects ay bindreflexives. m 十テ , KFinally,afreviewed lof position urodaor e underlying asunaga (1980)ixedtherelative (osubject inf .5 2w arguments f and inear 1982) M and non-subject arguments. 361 John Whitman These arguments all refer to syntactic phenomena that appear to be most naturally characterized on the assumption of a structural representation that distinguishes between subjects and non-subjects. The data in 2.1 argue for such a representation at the level of S-structure. The data in 2.2-2.3 argue for such a representation throughout the syntax, since there is no evidence that subject position where PRO appears in Japanese is governed at any leve1. The data in 2.4 argue for a configurational repre ツ sentation at the level where condition A of the Binding Theory, which refers to reflexives and reciprocals, applies. The data in 2.5 argue for a configurational representationt the levelofD-structure, beforeproposing a applies. 1n the next section we will examine the evidence against a configurational distinction between subjects and non-subjects. We will then consider a solution to the apparently conflicting conclusions drawn from the data in sections 2 and 3. 3. ARGUMENTS AGAINST A CONFIGURATIONAL JAPANESE PHRASE STRUCTURE REPRESENTATION OF Clearly the facts discussed in section 2 present a challenge to the asso ツ ciation of a flat or nonconfigurational representation with a language like Japanese. The purpose of this section is to review some of the arguments againsta configurational phrasestructure representation Japanese. for Again, we will focus on the issue of the presence or absence of a maximal A projection dominating non-subjects but not subjects. gain,we will conツ fine ourselves to evidence unconnected with the putative correlations in (I): apparent absence of movementules and typical occurrence of r 'scrambling' type free word order. 3.1. Absence of VP Movement Rules Phenomena of the sort in (22) provide evidence of a VP fronting process in English. (22) Brad promised to come at 10, and come at 10 he did In English this process involves detachment of the VP from the subject and auxiliary,includingtense. In Japanese,ense and modal suffixes t may not be detached from the lexical verbal head, except by intervention of a focus or emphatic particle, as in (23): 362 Con巾g ぴ 「 gzfo 伽服 a w (23) (25) g なひタ「 g川e陀川 d sぬ y0m服 / mo -na-kat PAST H E eve read T ME doNEG m Eb 虹 "E@M k L)that , ( 。 (24) れ tc e0 ), VP ・ 。 テ t田 Ld ・ dbd Sonohon-m0, John-wa *John-wa yom-i - mo THEME read *Yom-i read -mo even / even -wa THEME / John- - wa kozut de THEME mallet with -wa テー - mo soo but汰o¥a5i-ta; JO ACC smash PAST -O ・ 卜a THEME si - ta 3.3. Absence of Adjacency Requirement/or ase C Assignment its clause final si - na - katta THEME that book ACC do NEG PAST sono hon - o si - na - katta THEME that book ACC do NEG PAST wa Directly related to the fact in 3.1 is the observation of Hinds (1974) that Japanese and similar languages give no evidence of VP Deletion cor ツ responding theprocess exemplified for Englishby (27): to I'll come at 10 and Brad will too Japanese does have an anaphodc verbal construction which appears to correspond to s0-called do so Pronominalization in English: Mary - wa M also thus do PAST *'Mary smashedthe TV with a hammer, and John did so the stereo' 3.2. Absence of VP Deletion (28) *Mary smashedthe TV with a hammer; John did so the stereo stereo read PAST iiono hon-0 that the object be replaced However, Japanese freely allows retention of non-anaphoricdirect objects yon-da position: (27) (29) that book even THEME read PAST 'Even that book John read' Sono hon - wa, John-wa that book THEME CONTRAST 'That book, John read' 363 English do so Pronominalization requires along with the rest of the VP: (3o) yon-da However, the verbal head may not be scrambled from (26) John Whitman TV - o mi - ta; John-mo soo si - ta THEME ACC see PAST also thus do PAST 'Mary watched TV, and John did so too' Nakau (1973) identifiesthis construction as the sole evidence the for existence of a constituent VP in Japanese. The Japanese construction is, however,crucially different from its ostensibleEnglish counterpart. In this and the succeeding sections we will examine a somewhat subtler class of evidence against VP as a maximal projection in representations of Japanese phrase structure. Referringto the observation of Schwartz (1972) that in English-type languages adverbs may not normally intervene between verb and direct object. Hinds (1974) notes that this observation does not apply to Japanese similarlanguages. hus,in English,placement of an adverb and T between verb and directobject is possibleonly with Heavy NP Shift constructions, (31) b. as in (31) e山ately 血eebook John read yesterday/immediatelythe new book that had been recommended by In Japanese, placement of adverbs between verb and direct object is un ツ objectionable: (32) John - wa sono hon - o kinoo / sugu yon - da THEME that book ACC yesterday immediately read PAST (=31a) There is a temptation to analyze (32) as merely a reflex of the Scrambling phenomenon- However the direct object in (32) does not appear in the position of a scrambled constituent, i.e. adjoined to S. Furthermore, a Scrambling analysis ould miss a cleargeneralizationere,namely, that w h adverb placement is quite free across languages, except where it is restrictcd by constraints on the interpretation of different kinds of adverbs, or 364 Conれ rQripnQ44 pg ゆ where intervention 「口所 efer ざ of an adverb between an argument and the lexical head that assigns Case to it violatesthe adjacencyequirement for Case r assignment (Stowell 1981). Thereforeit seems unlikely that Universal Grammar contains anything correspondingto a rule of Adverb Placement; instead,the position of adverbss left unspecified by the principles f i o grammar that fix linear order and their linearposition is 'visible' only when they interact with some other principle, such as Case assignment. Thus (32) is accounted for most readily if we conclude that Case marking is not restricted by adjacency in Japanese. The same point can be made by the relative order of non-subject ar ツ guments ofV: (33) a. L Mary - ga sono hon - o NOMthat bookACC 'Mary gave that book to John' ユ。血 M 365 乃加川肋伽伽 John - ni yat - ta DAT gjYePAST m 士 sono hb -k0A ya 血 CC gjve ST M DAT that Again, the direct object in (33a) does not appear in the characteristic Scrambling position but is not adjacent to the verb and is clearly Case quired to define structural Case positions. Thus, in English, the position for assignment f objectiveCase by V (corresponding to the position of o the GF 'object') is defined as the NP which is a daughter of V and sister paIameter 血aatwe haVe Fent 田e.(edsdL 雄ひ如叩 eCf L: K ofV.The etc. A negative value for this parameter would be achieved when the phrase structure representation of a language disallowed such a specification, by, for example, representing the projection ofV as indistinct. Such a representation would have the effect of eliminating (VH1") as a distinct projection, although it might retain some of the effects of hierarchical structure within S. In the remaining sections we will explore more evidence that supports a phrase structure representation of this type for テ , Japanese. 3.4. Behavior of Range-BearingElements Elements such and only behavedifferently in English with respect of association with focus, as noted by Jackendoff (1972). Thus, while both even and only may be associated with focus on the object in (35): as even to the phenomenon 血 。 (5 3 w ab 了丁 血 。 w 山川 read read BB 0 O wked We can explain the absence of an adjacency requirement for Case i assignment in (32-33) if we assume that Case in Japaneses in general inherent.Chomsky(1981) distinguishes inherentCase in instances like 血at of the patienta bookfmom John in (34) (34) only 口6 not be associated may ab St山C r司 Casein jnstancesljke 山atof 加 Mary gave John a book A language hich lacksDativeIncorporation structures like (34), suchas w Japanese, ight be supposed to haveonly either inherentor structural m Case;in the latter eventwe shouldexpect that Caseassignment wouldbe directlylinked to fixed structural positions.Sincethat is clearlynot the case in Japanese, e must conclude w that Japanese only inherent has Case for argumentswhose Case is assigned by the verb Let us now consider how the preceding discussion relates to our central concern: whether or not we are justified in positing a maximal projection that dominates just V andits complements. featuresuchas absence f A o structuralCasemust be parameterized;that is, its distribution cross lana guages cannot be random. A plausible parameter for presence or absence of structuralCase is presence or absence f the categorial o projectionser 研 丁工 0 川 0 with focus on the subject in (36) 刷川 十a h 亡 十 re ddaa a亡bb kko0 h w re A reasonable account of this data might proceed from the assumption that even in pre-verbal position is inserted under INFL, and thus has a c-conunand domain comprising the entire S, while only is inserted under VP, and thus has a c-command domain consisting only of VP. Alternatively, it might simply be stated that the range of even is S in this position, while that of only is VP; this is Jackendoffs account. On either account, the distinction between these two elements is predicated on the existence of a constituent VP in English. 1n Japanese, the postpositions correspondingto even and only are suf ツ fixed to a conjugational base form of the verbal head, as we saw in (23): (37) a. John-wa sonohon-0 yomi-mosi-ta THEME that book ACC read even do PAST 'John even READ that book' 366 Co れぬ b. John - sono hon - o wa yomi - dake - wa si - ta THEME that book ACC read only THEME do PAST 'JohnonlyREAD that book' Unliketheir Englishcounterparts, however, the sentences in (37) may be interpreted only as assigning focus to the verb 'read'. This fact parallels the discoveryof Kuno (1980) that the scope of suffixal negationin Japaneseis restrictedto the verb to which the negativemorphemes i suffixed. Thus we must conclude that the domain of the negative morpheme, as well as the postpositions in (37), consists only of V. How can we account for this conclusion in a principled way? A general fact about range or scope bearing elements such as those under discussion seems to be that their domain extends over the maximal projection of the head to which they are adjoined. This would lead us to expect that in Japanese the domain of the elements correspondingto even and only would be VP when they are adjoined to V; we find that this is not the case. Again, the explanation that suggests itself is that the maximal projection VP is not distinct, and therefore unavailable for the specification of a domain 3.5. Distribution of the Empty Category Pronominal The empty category pronominal in Japanese appears to be subject to a distribution relative to its antecedentsquite different from the distribution that we discovered for overt pronouns and r-expressions in 2.1. Like overt pronominals,the null pronominal may have a discourse antecedent: 3 ( 8) か伏而 0 C 肛而0 b. John - ga NOM nd 曲 0ymF MST R read PAST However, the null pronominal may not have an antecedent contained in an argument in its clause. In contrast to kare in (7-8), the null pronominal in (38)-(40)may not have John as its antecedent.3 39 () no GEN most Hd (40)John. about uwasa e*inagaSe"S -DAT kansuru L dL pTead t? ni -ACC o s te ib GI3R rumor M L .1 ubt H 也 * D a gr C 町 M ST ・ 'The rumors about John, [e] is spreading' テー ー工 ナノ , It is not the case that the null pronominal allows no antecedent within its clause. Antecedents inside adjuncts are acceptable: (41) [John.-ga hikoozyo0-ni tui-ta NOM airport mukai-ni 。 told, Mary- ga LOC arrive PAST time NOM [e.] it - ta meet LOC goPAST W@henJo iar「 iyedatthe 血 po 『 ,L4ary wenttomeetLe R ・ , The facts in (39-41) are most readily accounted for if we assume that the c-command domain of the null pronominal in (39-40) contains its antecedent John, while in (41) it does not. Crucially, this will require that the subfect NP in (39) is contained in the c-command domain of the object null pronomina1.This is consistent with the argument that we have been developing in this section against a representation containing VP as a distinct maximal projection, but it is in direct conflict with our conclu ツ sions from 2.1. Let us therefore re-examine the evidence for a configurational representation including VP that we found in our consideration of binding domainsin section 2. ヌ6 ・ [e] yon - da 'John readpro' 367 John Whitman r 加われ 0ガり川MMerer$ , Fyf 口 e乃 CeAom BEndingDomainsReconsfdered The data in 3.5 at first appear to suggest that binding domains involving null pronominals and binding domains involving other nominal categories must be defined over entirely different representations of phrase structure. Before accepting such a radical conclusion, let us first consider just what subtypes of configurational structure were motivated by the data in 2.12.4. The data in 2.2-2.3 on the distribution of PRO and the data in 2.4 on the binding properties of reflexives argued for a structurally defined posi ツ tion for just a single GF: subjects. The data in 2.1 seemed to argue for structurally distinct positions in a more general sense: both the position of the subject in (7) and the proposed object in (8) had to be outside of the binding domain of the pronoun. Thus 2.1 represents the best case for a domain defined over a conflgurational representation independent of particular grammaticalfunctions. 1f this case is valid, we expect to find evidence for c-command domains inside the constituent VP as wel1. However, when we inspect the relevant evidence, WO find that the leftmost argument of the verb always binds argumentsto its left:4 368 (43)John- itwantb-ea zyosei]-0ni [ -ga a -eetteER AST ACCDAT NOMagat i P womanare. m -G t k im htootoo syookai si - te yat - ta at last introduce GER give PAST 'pro at lastintroduced woman John. hadbeenwanting to meet the to him,' (44) *Kare.-ni [John- - ga ai - tagat - te i - ta zyosei] - o him DAT NOM meet want GER be PAST woman ACC tootoo syookai si - te yat - ta at last introduce GER give PAST (glossas in (43)) (45) [John. - ga ai - tagat - te si - te yat i - ta - 。 *Kare. - o [John. himACC tootoo syookai at last introduce (gloss as in (45)) - ga ai - tagat - te i - ta ihadbeenw 狙 t 而g - te yat - (47) John-s brotherandhe./him, ent fishing w b. *y.e.1*ldm. andJohn's brotherwent fishing a. is relevant to the correct statement of binding conditions (48) JohnshowedMary, herself^n the mirror i (48) the r-expression ary must bind the reflexive,but it cannot be M boundby the reflexive. Both this case andthat of (47) can be character- ized as potential as zyosei] - ni NOM meet want GER be PAST woman DAT si Linear precedence has of course frequently been included in the statement of the 'command' relationdefining bindingdomains. any original M apparent reflexesof precedence (suchas the fact that in English, ubjects s bind non-subject arguments but not vice-versa) sincebeenattributed have to the more general (and non-directional) relation of c-command. How ツ ever, it is possible to discern precedence asymmetries with respect to binding domains even in English. An example is coordinate structures: In ta GER give PAST * proatIastintroduced ito 血eewomanJo to meet' (46) is more general;however, we also find that linearprecedences crucial i for determiningthe former type of binding omain. d Kuno (forthcoming) alsonotes cases like the following, here precedence w zyosei] - ni kare. - o NOM meet want GER be PAST woman DAT him ACC tootoo syookai at least introduce 369 John Whitman Co ゆ糾 rorfondお tVPdr な川 efex$ ta GER give PAST In (43 - 46), the examples where the pronoun precedes antecedent its (44 and46) are uniformlyunacceptable on the coreferentnterpretation. i This has nothing to do with the grammatical function(or surfacecase marking) of the NPs involved, since both dative ni- and accusative -0 marked pronouns are unacceptable when they precede their antecedents. This makes any effort to account for (43-46) by positinghierarchical structure within VP essentially vacuous, since such an account would have to assume that the leftmost NP within VP is always in a c-commanding positionwith respect to other NPs in VP irrespective f its grammatical o function or surface case. Such an account simply translates a linear pre ツ cedence relation into a hierarchical dominancerelation. We thus find that the distinction between the type of binding domain referred to in 2.1 and those discussed in 2.2-2.4 is indeed that the former in (49) instances of mutual c-command.This can be represented (49) Nand (=48)...' a. P, Pib.NIP^NP^ (=47) N theseinstances, that is, instances where the status of NPi or NP^ pronominal or r-expression ould requiredisjoint referencewith w anythingcontainedin the dominating maximal projectionXP-n,linear precedenceppears to becomerelevant: a NPn is containedn the binding i domain of NPi, but NPi does not containNPi in its bindingdomain In just as a Let us express this generalization with the following statement: (50) 附 ere two ove 『 NP5 市aare 血e same binding dom 血 血e binding t domain is associated only with the leftmost NP , Note that (50) does not require that the two NPs which share the same binding omain be coindexed; are not, for example,n (47). d they i Application of (50) to the facts of (43-46) eliminates apparent the difference between the binding domains of the overt pronoun kare 370 Co ゆ糾 rgffonQ4f ひ四 M川ergr$ John Whitman and the null pronomina1.We saw in (43-46) that precedence appears to determine the binding domain of the overt pronoun kare. This suggests that (50) is applying,that is, that the matrix argument NPs in (43-46) sharethesameC-Co 日日川ddoma ぬ vWesawin33n3.5thatthebindingdom 由n of the null pronominal appears to include all argument positions in S We have now reached the conclusion that the binding domains for overt and null pronominals in Japanese are in fact the same (as we might other ツ wise expect), and that in each case the binding domain for a pronominal argument includes all other argument positions. ・ 4- RECAPITULATION RATIONALITY OF EVIDENCE FOR PHRASE STRUCTURE CONFIGU- We set out to examine the evidence from phenomena other than surface word order for phrase structure configurationality in Japanese; in par ツ ticular, we were concerned with the issue of justifying a maximal projecツ tion dominating complements of V but not its subject, i.e. VP. Section 2 laid out the arguments for a representation including VP. Only one of these arguments appeared to directly require a representation including VP: the argument in 2.1 demonstrating asymmetry of subject and objectbinding domatnsfor 匹Onom 而亦 Thhe other WmentsL 2 were arguments for distinguishing a structural position for subjects; these arguments were consistent with a representation including VP, but not direct arguments for such a representation Section 3 laid out the arguments against a representation of Japanese phrase structure including VP. The arguments in 3.1-3.4 were based on negative evidence: absence of characteristic phenomena that we might ex ツ pect to find in languages that include VP as a maximal projection. These missing characteristic phenomena were VP movement, VF deletion, adツ jacency for Case assignment, and scopal domains corresponding to VP. Since this evidence is negative, it is necessarily quite weak. However, the evidence in 3.5 was of a more indicative nature: the evidence presented there suggested that the binding domain of null pronominals in Japanese ・ must include all other argument positions in S. This motivated a re-ex ツ amination of the binding domain of overt pronominals in 3.6, where we concluded that their binding domain was the same as that of null pronom ツ inals, but subject to the principle in (50), which assigns shared binding domains to the leftmost of two NPs. T.et us now re-examine the remainine data discussed in 2. As remarked above, these data constitute evidence for a distinct structural position for subject, but not for other structural relations. This is particularly clear in the case of reflexives discussed in 2.4. If the binding condition for 371 anaphors is defined over a configurational representation of phrase struc ツ ture in Japanese, we expect to find cases of non-subject binding of ana ツ phors wherever such a situation is structurally possible. As we saw in (17), however, non-subject binding of reflexives in Japanese does not occur. This fact has long been known, but I am unaware of any principled explanation for it. An analysis is suggested by our discoveries about the bindingdomain of pronominals in 3.6. We found therethat the binding condition on pronominals and r-expressions was the expected one: dis ツ joint reference within a binding domain. However, because of the phrase structuralpropertiesof the language, rguments within S all sharedthe a same bindingdomain, i.e. S. In this situation,the evidencen (43-46) i indicated that the grammar associated the binding domain with what we may call the 'most prominent NP', in this case the one distinguished by precedence. What this suggests is that while the current conditions of the Binding Theory may be in large measure correct, their interpretation is tied too closely to particular phrasestructurerepresentations.hus, ratherthan T identifying relations like 'c-command' and 'SUBJECT of as invariant components of the Binding Theory, we should identify 'obligatorily bound', 'obligatorily free' as its basic features, with the structural specifi ツ cation of the domains over which these features are defined subject to parametric variation. Applying this approach to the problem of Japanese reflexives, e find w ourselves with the standard featurefor anaphors, 'obligatorilyboundin some domain to be specified'. As was the case with pronominals, we will assume that we cannot avail ourselves of hierarchical configurations at S- or D-structure to specify the relevant domain. Precedence,the choice made to specify the domain for pronominal binding, is an inappropriate choice in the case of anaphors. We saw in 2. evidence that the binding conditionswhich restrict pronominals and r-expressions o not apply d at LF, but rather at a level where it makes sense to refer to linear order. NO such evidence is available in the case of anaphors. If the domain for the binding condition on anaphors were specified at a level of representation where linear order is irrelevant (such as LF), we would have to refer to another relation.The only relationobviously available that of "external is argument" in the sense of Williams (1980). Remember that we are accepting the model for "non-configurational" languages f Chomsky(1981) o describedn 1 where subjectandnon-subject arguments are distinguished i in the lexiconand in argument stmcture in LF. It is at precisely level this that externalarguments constitute "most prominent NP" with which the to associate the binding domain for anaphors. Of the remaining rguments for distinguishing structural a a position for subjects, the one based on the distribution of PRO in 2.2 and 2.3 is similar 372 to the argument based on subject control of reflexives: there is a clear motivation a structural for positionassociated PR0, but it is lessclear with at which level this position is defined. The cross-1inguistic evidence that PRO must be associated subject with positonis quitegood,but the current account of when PRO may appear in subject position is quite poor. Saito (1982) shows that PRO in Japanese ay occur in the subjectpositionof m a tensed clause. Japanese has no agreement, so an account of the S- or D-structure propertiesof the subtypesof subjectpositonswhich allow PRO in Japanese seems quite unpromising.On the other hand, an account which refersto subjectpositionin clauses hich lackindependenttense in w LF seems to characterize the distribution of PRO quite wel1. The final argumentsfor distinguishingstructural a positionfor subjects, those presented in 2.5, were based on linear precedence of this position over other argument positions. Clearly, this is a result that we get for free from a representation including P; therefore,t is particularly V i important to consider what other alternatives might be available 373 John Whitman Configurationality parameters to account for the cleargeneralization subjects that precede on-subjects in Japanese n Remember that while we have found negative evidence against distinguishing a maximal projection VP in Japanese, and a case in 3.5 where recognizing such a category would make the wrong prediction about the binding domain of the null pronomina1,nowhere have we found evidence against hierarchical structure in S in Japanese; only evidence against a maximal projectionntermediate between S, V, andits arguments. i 0ne position which clearlymust be structurallydistinctfrom other positions is that of the W7-marked topic. Typically this position is re ツ presented as a position left-adjoined to S. A suggestion that takes advantage f the presence of this positionand also explainsthe characo teristic linear position of subjects in Japanese would be to state that toprealization, i.e. insertion of some element into topic position in S-structure, is obligatory in Japanese. In the absence of some other W4-marked NP argument in topic position, subject is the default choice for topicaliz ation. This treatment captures the frequently' oted 'topic prominent' n chaFacterof 山eelangUage,asweIlas 血eegeneralizatfon 山aatsubjectsare山 overwhelming unmarked selection for wa-marking as topic complementsmmediately dominated by S, or it could be represented i with S as the maximal projection f V, with all argumentsof V dominated o by intermediate projections. t is not clear that thesetwo alternatives I make differentpredictions, or that current theoriesof phrasestructure n ought to be ableto distinguishhem. t A much more substantive question is the nature of the basic parameter involvedn sucha representation. parameter ore naturalthan'presence i A m or absence ofVPasa maximal projection'is 'headof S'. Phrase tructure s representationf clearlyconfigurational o languages generallyelected have s INFL as the category whose maximal projection s the initialsymbol of i the grammar. Selection of INFL as the head of S in Japanese requires a great deal of empirically unmotivated abstraction from superficial struc ツ ture: tense and modal elements ormally associated n with INFL appear as suffixes on the verb; they do not agree in any visible sense with the subjectand thus cannot plausiblebe arguedto containAGR; andas we saw in 3.4, scopalelements normallyassociated with INFL restrictthe domain of their scope or range to the verb.We might imagine then that selection of the head of S is parameterized, and that languages f the o Japanesetype select V as the head of S. This then will constitutethe parameter for configurationality that hasbeenthe focusof thisdiscussion. NOTES 1. Some speakers exceptionally allow reflexives like the following PoKceK Dai hungsibun, utiniOC PAST Keisata irnTyuu, zelf-nohouse in -ta NOMDae -o CC GEN- L tozikome -ga Kim C A shut The police shut Kim Dae Chung in his own house' 2. This is a simplified version of the original argument. 3. This observation is originally due to Susumu Kun0, as is the discovery that intrasentential antecedentsfor the null pronominal are possible in adjuncts. 4. This discovery is due to Susumu Kun0. 。 4ソ・カ あ e川面 VeRep 化se れ血ぬ oれ s f The purpose of this paper has been to consider the argumentsfor presence or absence of a maximal projection VP as a parameter for phrase structure configurationality. e havespeculatedery little about the actualnature W v of the phrase structure representation to be associated with a language lackingVP as a maximal projection;n theory,the phrasestructure of S i in such a language could either be represented as flat, with V and all its REFERENCES Chomsky, N. (1965) Aspects of the Theory of Syntax, Cambridge, MA.: MIT Press. Chomsky, N. (1981) Lectures on Governmentand Binding, Dordrecht: Foris. George, L. & J. Kornfilt (1981) "Finiteness and Boundedness in Turkish", in F.Heny, ed.. Binding and Filtering, London: Croom Helm. Hale, K. (1982) "Preliminaly Remarks on Configurationality", in J. Pustejovsky and P. Sells, eds.. Proceedings of the Twelfth Annual Meeting of the North- eastern Linguistic Society [NELS} Hasegawa, N. (1981) A Lexical Interpretive Theory with Emphasis on the Role of Subject, Doctoral dissertation. University of Washington. Conが糾口 FOれぬ 川口 m eKざ ゆ が 374 Hinds, J. (1974) "On the Status of the VP Node in Japanese", Indiana University Linguistics Qub Jackendoff, R. (1972) Semantic Interpretation in Generative Grammar, Cambridge, MA.: MIT Press. Kun0, S. (1980) "The Scope of the Question and Negation in Some Verb-Final Languages", Papers from the 16th Annual Meeting of the Chicago Linguistics Society. Kun0, S. (forthcoming) Functional Syntax, Chicag0. Kuroda, S.Y. (1980) "Bunkoozoo no Hikaku", in T. Kunihir0, ed., Niti-eigo Hikaku Koala, Tokyo: Taisyuukan. Marantz, A. (1984) On the Nature of Grammatical Relations, Cambridge, MA.: MIT Press. Masunaga, K. (1982) "Bridging", in S. Hattori and K. Inoue, eds.. Proceedingsof the XIIIth International Congress of Linguists, Toky0. Mikami, A. (1953) Gendai Gohoo Zyosetu, Tokyo: Kurosio Syuppan. Mohanan,K.P. (1981) "On Pronounsand Their Antecedents", ms., MIT. Mohanan, K.P. (1982) "Infinitival Subjects, Government, and Abstract Case", Linguistic Inquiry 13. Nakau, M. (1973) Sentential Complementation in Japanese, Tokyo: Kaitakusya. Sait0, M. (1982) "On Case Markingin Japanese", ms., MIT. Sait0, M. (1983) "Case and Government in Japanese", in Proceedings of the 2nd West Coast Conference on Formal Linguistics. Sait0, M. (1984) "Three Notes on Syntactic Movement in Japanese", ms., MIT. SMbatmi M て 19 , , mLu , inJ MiCs,Tokyo: StovV卜 u , Williams ・ フ 8) wMikami HindsandI. T (1981)L E (1980) ・ , ・ takuS ノ a La 川dtheNotionof Howard,ed5 ・ , P@Prorg あ Sub6ect"in 了 apaneseCr 加九pク ne 比 $ノ nt 似 4nondFem ノ R。 ・ ノ rzgfnsofPh 卜 rQseSrrucrvrg, ソ Prediction",Linguisric 工丁 山刀 octoIaI auf ヴ l. 工 』 Fser ation 亡 , MIT. - ...
View Full Document

This note was uploaded on 10/07/2010 for the course LING 341 taught by Professor Masyayoshida during the Fall '10 term at Northwestern.

Page1 / 13

Readings - 350 z 乃 regWoofesonSyn れ tac4icM@ovement ぬ...

This preview shows document pages 1 - 13. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online