Remarks on Japanese subjects

Remarks on Japanese subjects

Lingua 85 (1991) 3033319. North-Holland 303 Remarks on Japanese subjects Mineharu Nakayama and Masatoshi Depcrrimenr of East ASKWI Lunguages ...

952KB Sizes 8 Downloads 67 Views

Lingua

85 (1991) 3033319.

North-Holland

303

Remarks on Japanese subjects Mineharu

Nakayama

and

Masatoshi

Depcrrimenr of East ASKWI Lunguages

Koizumi*

and Lireratures,

The Ohio Stnie University.

are base-generated

under VP (VP internal

Columbus.

OH

43210. USA

Received

May 1991

It has been proposed sis). In this paper, base-generated pieces temporal

phrases;

and direct unaccusatives

subject

hypothesis

under VP and move above VP at S-structure

of evidence

of unergative

that subjects

we argue that the VP internal are

discussed:

and secondary

and transitive passives

floating

depictive

numeral

predicates.

verbs are base-generated

are base-generated

under

be above VP at S-structure

is not warranted

quantifiers;

in Japanese.

pseudo-cleft

These collectively

indicate

Four

constructions; that the subjects

above VP while the subjects of unaccusatives

VP (i.e., the object

and the direct passives are different

subject hypothe-

that allows the subject to be

in that the subjects

while the subject of the unaccusative

position)

in Japanese.

The

of the direct passives

must

can be at the object position.

1. Introduction In recent syntactic theories, it has been proposed that subjects are basegenerated under VP (e.g., Fukui 1986, Kitagawa 1986, Kuroda 1988, Koopman and Sportiche 1988, Sportiche 1988). In this paper, we argue that the VP internal subject hypothesis that allows the subject to be base-generated under VP (i.e., the SPEC of VP) and move above VP (i.e., SPEC of IP) at Sstructure is not warranted in Japanese. We will discuss floating numeral quantifiers, pseudo-clefts, temporal phrases, and secondary depictive predicates. We conclude that the subjects of unergative and transitive verbs are base-generated above VP while the subjects of unaccusatives are base* Different versions Wisconsin, University

of the content of this paper have been presented of Rochester, and the LSA annual meeting. We thank

at University the participants

those talks and in particular, Heedon Ahn, Hiroko Y. Butler, Kyle Johnson, Yuki Kuroda. Anoop Mahajan, Shigeru Miyagawa, and Kazuhiko Tajima

Yoshi Kitagawa, for their helpful

comments and discussions. Any shortcomings are, or course, ours. The second author the Department of Linguistics and Philosophy, MIT, Cambridge, MA 02139, USA.

0024-3841/91,‘$03.50

cm 1991 -

Elsevier Science Publishers

B.V. (North-Holland)

of of

is now at

304

M. Nukayama. M. Koizumi / Remarks on Japunese

SUBJECTS

generated under VP in Japanese. The unaccusatives and the direct passives are different in that the subjects of the direct passives must be above VP at S-structure while the subject of the unaccusative can be at the object position. The organization of the paper is the following: in section 2, we provide four pieces of evidence that suggest that the subjects in the transitives are base-generated above VP. In section 3, we discuss the subjects of unergatives, unaccusatives, and direct passives. Section 4 provides our concluding remarks including some theoretical implications of our hypothesis. Throughout the paper, we assume the theoretical framework of Government and Binding (e.g. Chomsky 198 1).

2. Subjects of transitive verbs In this section, we discuss the subjects of transitive verbs. First, we provide a theoretical assumption, and then we present four pieces of evidence that suggest that subjects are base-generated above VP.l It is well known that quantifiers in Japanese take scope according to their linear order. A quantifier takes wide scope over its c-commanding quantificational elements and it does not take narrow scope (Hoji 1985, cf. Kuroda 1970).” For instance, sentence (1) is not ambiguous: daremo ‘everyone’ takes wide scope over dareka ‘someone’, but it cannot have narrow scope. (1) Daremo-ga dareka-o everyone-Nom someone-Act ‘Everyone loves someone.’

aishiteiru. loves

Observing this phenomenon, Hoji (1985) claims the rigidity of the structure of Japanese. The subject in Japanese appears higher than the indirect and direct objects at base and the indirect object is higher than the direct object. The relevant

sentences

are in (2).

1 Nakayama et al. (1991) used VP-preposing constructions by Hoji (1989) as one piece of evidence to support our claim here. However, it has been pointed out that the construction could be compatible with the VP internal hypothesis. See Hasegawa (1989). * The definition of c-command we are assuming here is essentially that of Reinhart’s (1976). (i) a c-commands nor p dominate

p iff the first branching

the other.

node dominating

a also dominates

p. and neither a

M. Nakayama. M. Koizumi / Remarks on Japanese subjects

305

shookaishita. dareka-ni John-o -Act introduced everyone-Nom someone-Dat ‘Everyone introduced John to someone.’ shookaishita. daremo-ni dareka-o (2b) John-ga

(24 Daremo-ga

-Nom

(W

everyone-Dat

someone-Act

introduced

‘John introduced someone to everyone.’ shookaishita. John-ni dareka-o Daremo-ga -Dat someone-Ace introduced everyone-Nom ‘Everyone introduced someone to John.’

In sentence (2a), the subject daremo takes wide scope over the indirect object and in (2b), the indirect object daremo has wide scope over the direct object dareka. In sentence (2~) the subject takes wide scope over the direct object. Therefore, the subject c-commands the indirect object and the indirect object c-commands the direct object (see also Saito 1985 and Whitman 1987).3

dareka,

2.1. Floating numeral quantifiers A phenomenon called floating quantifiers suggests that there is no subject trace under VP in active transitives. Miyagawa (1989a) argues that for an NP to be construed with a numeral floating quantifier (NQ), they must ccommand each other. VP-internal elements such as an instrumental phrase, the indirect and direct objects cannot appear between the subject and the NQ with which it is construed. This is shown in (4).

3

(44 Gakusei-ga

3-nin kono naihu-de

students-Nom ‘Three students (4b) *Gakusei-ga students-Nom ‘Three students (4c) *Gakusei-ga students-Nom ‘Three students (4d) *Gakusei-ga students-Nom ‘Three students

3-CL this knife-with meat-Act cut cut the meat with this knife.’ kitta. kono naihu-de 3-nin niku-o this knife-with 3-CL meat-Act cut cut the meat with this knife.’ niku-o 3-nin kitta. meat-Act 3-CL cut cut the meat.’ John-ni 3-nin tegami-o kaita. -Dat 3-CL letter-Act wrote wrote letters to John.’

Hoji (1985) also uses weak crossover

phenomenon

niku-o

to support

kitta.

the rigidity

claim.

306

M. Nakayama, M. Koizumi 1 Remarks on Japanese subjects

Sentence (4a) contains the NQ that is associated with the subject. No elements intervene between the subject and the NQ and the sentence is wellformed. In sentence (4b), the instrumental phrase appears between the subject and the NQ while the direct object intervenes between the subject and the NQ in (4~). Both sentences are ill-formed. In sentence (4d), the indirect object is between the subject and the NQ. The ill-formed sentences (4bd) indicate that no element can intervene between the subject and the NQ that it is construed with. On the other hand, time adverbials such as kinoo ‘yesterday’ and kyoo ‘today’ may occur between the subjects and the NQ. This is because a time adverbial can be generated outside the VP as well (cf. Miyagawa 1989a: 30, Koizumi 1990b, McNulty 1988: 9, and Travis 1989, among others). 3-nin kinoo students-Nom 3-CL yesterday ‘Three students bought a book 3-nin Gakusei-ga kinoo students-Nom yesterday 3-CL

(54 Gakusei-ga

(5b)

hon-o book-Act yesterday.’ hon-o book-Act

katta. bought katta. bought

Suppose that the subject NP originates in the SPEC of VP, and is moved into the SPEC of IP for some reason (e.g., the Extended Projection Principle (Chomsky 1981), and/or the Case Filter (Chomsky 1981)) at S-structure. Then, (4b) and (4~) could have structures such as (6a) and (6b), respectively, assuming that the maximal projections are allowed to adjoin only to the maximal projections (e.g., scrambling (see Saito 1985)). (6a) [ip Gakusei,-ga (6b) [ir Gakusei,-ga

[VP kono naihu-de [vr ti 3-nin niku-o [vr niku,-o [vr ti 3-nin tj kitta]]

kitta]]

In these structures, the trace of the subject NP in the VP and the NQ ccommand each other. Therefore, the sentences should be well-formed. These are different from direct passive sentences such as (7) where the numeral quantifier appears after the instrumental phrase. basu-de 30-nin hakobareta. (7) Shuujin-ga prisoner-Nom bus by 30-CL carried were ‘Thirty prisoners were transported by bus.’ Since sentence

(4b) is ill-formed

whereas

(7) is well-formed,

the subjects

of the

M. Nakayama, M. Koizumi i Remarks on Japanese subjects

307

active transitives and the direct passives must be in different positions. The passive (7) is well-formed because there is a trace of the surface subject in the object position which is under VP and c-commands the numeral quantifier (see Miyagawa 1989a, b and section 3). However, the ill-formed sentences in (4) suggest that the subject is not base-generated under VP as in (6), but rather base-generated above VP in the transitives. 2.2.

VP-cleft construction

There is a construction in Japanese similar to the English pseudo-cleft sentence. We are particularly interested in pseudo-clefting of VP, i.e. the one involving the verb suru ‘do’ at the presupposed position as in @a). Let us call it the VP-cleft construction. In VP-cleft constructions, the indirect object cannot remain with the subject. It must be clefted with the object and the verb. Observe the sentences in (8): (8a) John-ga -Nom ‘What John (8b) *John-ga -Nom ‘What John

shita-no-wa kaisha-ni tegami-o okuru-koto-da. did-N-Top company-Dat letter-Act send-N-Cop did was send a letter to the company.’ shita-no-wa tegami-o okuru-koto-da. kaisha-ni company-Dat did-N-Top letter-Act send-N-Cop did to the company was send a letter.’

In sentence (8a), the indirect object appears in the ‘focus’ position with the direct object and the verb whereas in (8b), the indirect object occurs with suru ‘do’ in the ‘presupposition’ position. Ill-formed sentence (8b) indicates that the indirect object forms a constituent with the direct object and the verb, and it is an unbreakable unit. Assuming that ‘breakable’ units are not single but rather maximal projections (cf. bar or intermediate projections, Chomsky 1986) this suggests that there is a maximal projection above the indirect object, but not between the indirect object and the direct object. Incidentally, VP-adjuncts like an instrumental phrase may, but need not be clefted along with the other VP-internal elements, as shown below. (9a) Gakusei-ga [vp kono naihu-de [vp niku-o this knife-with students-Nom meat-Act ‘Three students cut the meat with this knife.’

kitta]]. cut

308

M. Nakapmu,

(9b) Gakusei-ga students-Nom ‘What students (SC) Gakusei-ga students-Nom ‘What students

M. Koizumi

/ Remarks

on

Japanese subjects

shita-no-wa kono naihu-de niku-o kiru-koto-da. did-N-Top this knife-with meat-Act cut-N-Cop did was cut the meat with this knife.’ kono naihu-de shita-no-wa niku-o kiru-koto-da. this knife-with did-N-Top meat-Act cut-N-Cop did with this knife was cut the meat.’

In (9b), the upper VP is clefted while in (SC), the lower VP is clefted. The NQ associated with the subject may not be clefted along with VP. (1 Oa) Gakusei-ga 3-nin shita-no-wa kono naihu-de niku-o students-Nom 3-CL did-N-Top this knife-with meat-Act kiru-koto-da. cut-N-Cop ‘What three students did was cut the meat with this knife.’ (lob) * Gakusei-ga shita-no-wa 3-nin kono naihu-de niku-o students-Nom did-N-Top 3-CL this knife-with meat-Ace kiru-koto-da. cut-N-Cop If VP contains the empty category that is coindexed with the NQ, sentence (lob) should be well-formed. The ungrammaticality indicates that there is no empty category that holds the mutual c-command relationship with the NQ. Therefore, this construction also suggests that the subject is generated above VP. 2.3. Temporal phrases Temporal phrases such as 2-j-ni ‘at 2 o’clock’ may optionally undergo clefting along with the object and the verb (Koizumi 1990b). This is shown in (11). shita-no-wa did-N-Top

terebi-o TV-Act

( 11a) John-ga -Nom

2-ji-ni 2-o’clock-at

‘What John (1 I b) John-ga -Nom ‘What John

did at 2 o’clock was watch TV.’ shita-no-wa 2-ji-ni terebi-o did-N-Top 2-o’clock-at TV-Act did was watch TV at 2 o’clock.’

miru-koto-da. watch-N-Cop miru-koto-da. watch-N-Cop

In (1 la), the temporal phrase 2-ji-ni ‘at 2 o’clock’ remains with the subject, being a part of ‘presupposition’, while in (1 lb) it is pseudo-clefted along with the object and the verb, being a part of ‘focus’.

M, Nakayama,

M. Koizumi / Remarks on Japanese subjects

309

The quantificational scope interactions of the temporal phrase and subject/object suggest that the temporal phrase is base-generated above object, but under the subject. This is shown in (12).

the the

(12a)

John-ga [2-ji ka 3-ji]-ni [subete-no shoruil-o -Nom 2-o’clock or 3-o’clock -at all-Gen documents-Ace teeshutsushita. handed in ‘John handed in all the documents at two or three o’clock.’ (12b) Daremo-ga [2-ji ka 3-ji] -ni shorui-o everyone-Nom 2-o’clock or 3-o’clock-at documents-Act teeshutsusita. handed in ‘Everyone handed in the document at two or three o’clock.’

In sentence (12a), the temporal phrase takes wide scope over all documents and the object does not take wide scope over the temporal phrase. In sentence (12b), the subject daremo’s ‘everyone’ takes wide scope over the temporal phrase and the temporal phrase cannot take wide scope over the subject. These indicate that the subject NP asymmetrically c-commands the temporal phrase and the temporal phrase c-commands the object. The internal subject hypothesis, under which there is a trace of the subject in the VP, cannot naturally account for unambiguous sentences like (l2b). This is because the scrambled sentence like (13) that contains the trace has ambiguous readings. [2-ji ka 3-ji]-ni (13) John-ga [subete-no shoruil-o, -Nom all-Gen documents-Act 2-o’clock or 3-o’clock-at ti teeshutsushita. handed in ‘John handed in all the documents at two or three o’clock.’ Therefore, these examples illustrate that there is no subject that is c-commanded by the temporal phrase.

trace under

VP

2.4. Depictive predicates There is a de-phrase which describes the state of the referent of an NP at the time when the action denoted by the verb occurs. A couple of examples are given below.

310

M. Nakayama. M. Koizumi / Remarks on Japanese suhjecrs

(14a)

John-ga katuo-o nama-de -Nom bonito-Act raw ‘John ate the bonito raw.’ (14b) John-ga hadaka-de hon-o -Nom naked book-Act ‘John read the book naked.’

tabeta. ate yonda. read

In (14a), de-phrase nama-de ‘raw’ is associated with the object NP, while in (14b), de-phrase haduka-de ‘naked’ is construed with the subject NP. Following Koizumi (1990a), we will refer to the de-phrases in (14a) and (14b) as the object-oriented depictive predicate (ODP) and the subject-oriented depictive predicate (SDP), respectively. Koizumi (1990a) shows that, at D-structure, the ODP is sister to the object, while the SDP is base-generated above the maximal projection which the object and the verb form. The difference between the two kinds of depictive predicates can be observed, for instance, in the following cleft sentences. (15a)

John-ga -Nom ‘What John (15b) *John-ga -Nom ‘What John

shita-no-wa nama-de katsuo-o taberu-koto-da. bonito-ACC eat-N-Cop did-N-Top raw did was eat the bonito raw.’ nama-de shita-no-wa katsuo-o taberu-koto-da. did-N-Top bonito-Acceat-N-Cop raw did raw was eat the bonito.’

(16a)

shita-no-wa hadaka-de biiru-o beer-Act did-N-Top naked did was drink beer naked.’ hadaka-de shita-no-wa biiru-o did-N-Top beer-Act naked did naked was drink beer.’

John-ga -Nom ‘What John (16b) John-ga -Nom ‘What John

nomu-koto-da. drink-N-Cop nomu-koto-da. drink-N-Cop

In (Isa), the ODP has been pseudo-clefted along with the object and the verb whereas in (15b), the ODP did not undergo pseudo-clefting, yielding the illformed sentence. This is explained if we consider that there is a trace of namude between the object and the verb (LF reconstruction). On the other hand, the sentences in (16) are both grammatical, because the SDP is dominated by a node high enough to be associated with the subject. Now, consider the examples in (17).

M. Nukayama,

(17a)

M. Koizumi / Remarks on Japanesr subjects

311

John-wa

[miso -aji ka shooyu-aji]-de subete-no -Top soy bean paste-taste or soy sauce-taste all-Gen sakana-o tabeta. fish-Act ate ‘John ate every fish seasoned with soy bean paste or seasoned with soy sauce.’ (17b) John-wa subete-no sakana-o [miso-aji ka shooyu-aji]-de tabeta. -Top all-Gen fish-Act miso-taste or shooyu-taste ate ‘John ate every fish seasoned with soy bean paste or seasoned with soy sauce.’ These sentences are both two-ways ambiguous with respect to quantifier scope: either the quantified depictive predicate or the quantified object NP may take wide scope. This is because the ODP and the object NP c-command each other (cf. Koizumi 1990). On the other hand, the sentences in (18) are not ambiguous: only the quantified subject can take wide scope. (18) Daremo-ga [hadaka ka kimono-sugatal-de everyone-Nom naked or in kimono ‘Everyone read a book naked or in kimono.’

hon-o book-Act

yonda. read

This sentence is unambiguous because the subject NP asymmetrically ccommands the SDP. If we adopt the VP-internal subject hypothesis where there is a trace of the subject in VP, it would wrongly predict that the SDP may take wide scope in the sentence in (18) similar to the ODP. In sum, the above four pieces of evidence collectively suggest that the subject of the transitive sentence is base-generated above VP.

3. Subjects of unergatives, unaccusatives, Next, let us consider unergatives, sentences do not contain accusative sentences. (19a)

John-ga geragera waratta. -Nom loudly laughed ‘John laughed loudly.’

and direct passives

unaccusatives, Case-marked

and direct passives. These NPs. Observe the following

312

M. Nakayuma. M. Koizumi / Remarks on Japanrsr subjuts

(19b)

John-ga attoiumani shinda. -Nom momently died ‘John died suddenly.’ (19~) John-ga Mary-no kawari-ni shikarareta. scolded was -Nom -Gen behalf ‘John was scolded on behalf of Mary.’ Sentence (19a) is unergative while (19b) is unaccusative. Sentence (19~) is a direct passive sentence. First, let us consider the VP-cleft counterparts of unergative (19a), unaccusative (19b), and passive (19~) as in (20) (21) and (22) respectively. (20a)

(20b)

(21a)

(21b)

(22a)

(22b)

John-ga shitta-no-wa geragera -Nom did-N-Top loudly ‘What John did was laugh loudly.’ * John-ga geragera shita-no-wa -Nom loudly did-N-Top ‘What John did was laugh loudly.’

warau-koto-da. laugh-N-Cop warau-koto-da. laugh-N-Cop

* John-ga shita-no-wa attoiumani -Nom did-N-Top momentarily ‘What John did was die suddenly.’ * John-ga attoiumani shita-no-wa -Nom momentarily did-N-Top ‘What John momentarily did was die.’

shinu-koto-da. die-N-Cop shinu-koto-da. die-N-Cop

shita-no-wa Mary -no kawari-ni shikarareru-koto-da. John-ga -Gen behalf scolded be-N-Cop -Nom did-N-Top ‘What John did was be scolded on behalf of Mary.’ John-ga Mary-no kawari-ni shita-no-wa shikarareru-koto-da. did-N-Top scolded be-N-Cop -Nom -Gen behalf ‘What John did on behalf of Mary was be scolded.’

In sentence (20a), the onomatopoeic phrase geragera ‘loudly’ appears with the verb ‘laugh’ while it is with ‘did’ in (20b). The ill-formed sentence (20b) suggests that the onomatopoeic phrase and the verb form a single constituent. The sentences in (21) are both ill-formed. This may be due to the incompatibility of the verbs ‘do’ and ‘die’. In other words, the verb ‘do’ in Japanese seems to require an animate subject and some intentionality in the focused verb. Since the unaccusative verb ‘die’ does not have intentionality, the

M. Nakavama, hf. Koizumi i Remarks on Japanese subjects

pseudo-cleft

sentences

are ill-formed.

The sentences

in (22) are both

313

well-

formed. If the sentences do not have ‘on behalf of Mary’, the grammaticality is degraded because of a lack of intention. We observed some aspects of pseudo-cleftability of the three types of sentences above. However, those sentences do not clearly show us where the subjects are, i.e., under VP or above VP. The following NQ sentences indicate some structural differences among the three sentences (cf. Miyagawa 1989a,b). geragera 3-nin waratta. * Gakusei-ga 3-CL laughed -Nom loudly ‘Three students laughed loudly.’ (23b) Gakusei-ga attoiumani 3-nin shinda. -Nom momently 3-CL died ‘Three students died suddenly.’ (23~) Gakusei-ga Mary-no kawari-ni 3-nin shikarareta. 3-CL scolded was -Nom -Gen behalf ‘Three students were scolded on behalf of Mary.’ (23a)

The three sentences in (23) contain a numeral quantifier that is associated with the subject. The ill-formedness of (23a) suggests that the subject and the NQ do not hold a mutual c-command relationship. The ill-formed (20b) also suggests that the onomatopoeic phrase and the verb form a single constituent. Therefore, the unergative sentence seems to have the subject base-generated above VP, similarly to the transitive sentences. On the other hand, both the unaccusative (23b) and passive (23~) are well-formed. According to Miyagawa (1989 a, b), unaccusatives and direct passives have traces at the object position that c-command the numeral quantifiers. In other words, the surface subjects in those sentences are actually the objects at D-structure. Since their traces and the numeral quantifiers c-command each other, the sentences are well-formed. The above NQ sentences indicate that nominative Case marked NPs (i.e., subject) are base-generated above the onomatopoeic phrase in unergative sentence (19a), but under attoyuumani in unaccusative sentence (19b) and Mar~l-no kawarini in direct passive sentence (19~). This is further supported by the following sentences with quantifiers (cf. Hoji et al. 1989, Nakayama 1990). (24a)

2-ji ka 3-ji-ni waratta. Daremo-ga everyone-Nom 2 or 3 o’clock-at laughed ‘Everyone laughed at 2 or 3’oclock’

314

M. Nakayama, M. Koizumi 1 Remarks on Japanese subjects

shinda. Daremo-ga 2-ji-ka 3-ji-ni everyone-Nom 2 or 3 o’clock at died ‘Everyone died at 2 or 3 o’clock.’ shookaisareta. (24~) Daremo-ga John- ka Mary-ni everyone-Nom or -Dat introduced was (24b)

‘Everyone

was introduced

to John or Mary.’

Unergative sentence (24a) allows only one interpretation where duremo takes wide scope over the temporal phrase. Therefore, this suggests (24a) has the base word order. Since the temporal phrase appears at VP as discussed above, the subject appears above VP in the unergatives. In unaccusative (24b), two readings are possible where the universal quantifier takes either wide or narrow scope over the temporal phrase. Similarly to (24b), passive (24~) has two readings where the universal quantifier takes either wide or narrow scope over the indirect object (cf. Oka 1988, Hoji et al. 1989). Therefore, both unaccusatives and direct passives contain traces under the temporal phrase and the indirect object, respectively. This suggests that the subjects in those sentences are base-generated under VP. Furthermore, consider the following sentences: waratta. 2-ji ka 3-ji-ni Daremo-ga 2 or 3 o’clock-at everyone-Nom laughed ‘Everyone laughed at 2 or 3 o’clock.’ shinda. (25b) 2-ji-ka 3-ji-ni Daremo-ga 2 or 3 o’clock at everyone-Nom died ‘Everyone died at 2 or 3 o’clock.’ shookaisareta. (25~) John-ka Mary -ni Daremo-ga or -Dat everyone-Nom introduced was ‘Everyone was introduced to John or Mary.’ John-ni shookaisareta. 2-ji-ka 3-ji-ni (25d) Daremo-ga -Dat introduced was everyone-Nom 2 or 3 o’clock at ‘Everyone was introduced to John at 2 or 3 o’clock.’ John-ni shookaisareta. (25e) 2-ji-ka 3-ji-ni Daremo-ga -Dat introduced was 2 or 3 o’clock at everyone-Nom ‘Everyone was introduced to John at 2 or 3 o’clock.’ (25a)

In unergative (25a), there are two possible readings: the temporal phrase takes either wide or narrow scope over the universal quantifier. This suggests that the temporal phrase is base-generated under the universal quantifier as a

M. Nakayama, M. Koizumi i Remarks on Japanese subjects

315

VP-adjunct and then it is scrambled. Since there is a trace, the sentence has ambiguity. On the other hand, sentence (25b) has only one interpretation where the temporal phrase takes wide scope over daremo.4 Since the temporal phrase appears at VP, (25b) suggests that daremo appears under VP. This means that the subject in unaccusatives can appear in situ in Japanese. Moreover, it means that nominative Case is assigned to the NP in situ without NP-movement. The ambiguity of the sentences (25-e) however, suggests that all passive sentences (25c-e) contain traces and they are ccommanded by other quantificational phrases. This means that unlike the unaccusatives, the surface subject is above VP in direct passives. It cannot stay at the base position (the object position). Our question here is why passives require movement while the unaccusatives do not. According to Miyagawa (1989 a, b), both unaccusatives and direct passives undergo NP-movement. However, as we saw, only passives seem to require NP-movement. NP-movement occurs for a Case reason. In other words, the verb loses its accusative Case assigning ability in passives. Therefore, the internal argument moves to the surface subject position (presumably the SPEC of Tense Phrase) to receive Case so that it does not violate the Case Filter (Chomsky 1981). A similar explanation should apply to the unaccusatives since unaccusative verbs do not have Case assigning ability. However, as we saw, NP can receive nominative Case under VP in unaccusatives. Why is this possible for the unaccusatives and not for the passives? We propose the following account: Suppose that for the verb to assign Case, it must be licensed by INFL (e.g., Tense) (cf. Miyagawa 1990). Then, since the passive morpheme appears between the verb and Tense, Tense cannot license the verb directly. That is, the category of the passive morpheme becomes a sort of blocking category for INFL (Tense) to assign Case to the internal argument (cf. Nakayama 1990 and Nakayama and Tawa 1991). Therefore, the Caseless NP is raised to the position where it can receive Case, i.e., outside of VP. On the other hand, the unaccusative verbs do not have Case assigning ability from the beginning. Yet, nothing intervenes between tense and the verb. 4 Perhaps. the following are clearer examples of unaccusative (i)

Subete-no all-Gen

gakusei-ga students-Nom

‘All students (ii) Tokyo-ka

‘All students other.

arrived

Kanda-ni

-0r

(i) is scopally

Tokyo-ka -OF

at Tokyo subeteno

-at all-Gen arrived at Tokyo

ambiguous:

On the other hand,

sentences

(cf. Nakayama

1990).

Kanda-nitsuita. -atarrived or Kanda.’ gakusei-ga

tsuita.

students-Nom or Kanda.’

arrived

either the subject (ii) is unambiguous:

QP or the Goal

QP may take wide scope over the

only the Goal phrase

may take wide scope.

316

M. Nakuyama, M. Koizumi / Remarks on Japanese subject.\

Therefore, Tense can license the verb directly. Thus, tive Case marked without movement. (See Nakayama structure of passives.)

the NP can be nominaand Tawa 1991 for the

Our next question is, then, why other languages like Italian have to have NP-movement in unacussatives (unlike Japanese). We answer this question by saying that these languages require subject-verb agreement, which is a licensing condition of the subject. Therefore, the subject must move to the position where it can have the agreement (the SPEC of AgrP (cf. Chomsky 1989)). On the other hand, since Japanese does not require agreement, the subject does not have to move out of VP. As long as it satisfies the Case requirement and predication, it can remain in situ. Therefore, the word order in (24b) is possible as a base order.

4. Concluding remarks We have looked at the subjects of transitive, unergative, unaccusative, and direct passive sentences in Japanese. Unlike a VP internal subject hypothesis, the subjects are base-generated above VP in the transitives and unergatives while they are base-generated at the object position in both unaccusative and direct passive sentences. The unaccusatives and the direct passives are different in that the subjects of the direct passives must be above VP at Sstructure. The subject of the unaccusative can be at the object position at S-structure because it can receive an internal theta-role from the verb and can have Case in situ. On the other hand, the subject is base-generated at the object position in the direct passives, but it must move out of VP in order to receive Case. This NP-movement occurs because the passive morpheme absorbs Case. That is, although Case assignment requires the direct licensing of the verb by Tense, the verb cannot be licensed in the direct passives because the passive morpheme intervenes between Tense and the verb. Finally, our question is where exactly is the base-position of the subject in transitives and unergatives. We suggest that it is the SPEC of the closest maximal projection (XP) to VP (cf. Ahn 1991).

(26) LxpSubjectivpVI Xl This strict locality between the subject and VP is in part imposed by the theory of predication (e.g., Williams 1980, Rothstein 1983, McNulty 1988,

317

M. Nukayama. M. Koizumi / Remarks on Japanese subjects

Koizumi 1990a). 5 The category XP may differ from language to language: it is Tense Phrase in Japanese while it is Agreement Phrase in English if we adopt Pollock’s (1989) phrase structure (cf. Chomsky 1989, Ouhalla 1990). Since the SPEC of TP is a Case position, subject in Japanese remains in situ. On the other hand, the SPEC of AgrP is not a Case position, thus subject in a language like English moves up to the SPEC of TP in order to satisfy the Case Filter (Chomsky 1981).6 Among others, the two principal motivations for the VP Internal Subject Hypothesis are; (i) simplification of theta-role assignment, and (ii) long distance dependency between subject and subject-oriented phrases such as numeral quantifier and VP. As for (i), Fukui (1986) argues that VP internal subject makes it possible for theta-marking to take place within the projection of a lexical head, which is conceptually attractive (see also Kitagawa 1986). However, this claim has some empirical difficulties. For example, it cannot naturally account for the subject/object asymmetries observed in Marantz (1984). In this paper, we have challenged this claim by showing that there exists a maximal projection between subject and object at Dstructure. As for (ii), Sportiche (1988) observes that in a language like French a subject-oriented numeral quantifier is separated from the subject by a verb or an auxiliary verb in INFL at S-structure, from which he concludes that subject is generated at VP and is moved to the SPEC of IP (see also Koopman and Sportiche 1988). This derived subject hypothesis in Sportiche (1988) and Koopman and Sportiche (1988) is also preferable considering that the surface subject position and VP (a primary predicate of the subject) are separated ‘too far’ in recent theories of phrase structure such as Pollock’s (1988) and Chomsky’s (1989). However, as Sportiche and Koopman themselves seem to be aware of, (ii) does not necessarily require the subject NP to be base-generated within VP: the subject could be generated near (but outside) VP and raised to the surface subject position. Our analysis in this paper readily accommodates (ii) while maintaining that the subjects of the transitive and the unergative sentences are base-generated outside of VP.

5

Although

each theory

respect to the subject over the other here.

of predication

position

is different,

we posit. Therefore,

all of them bring the same consequence we will not commit

ourselves

with

to one theory

6 If there is Aspect Phrase between Tense Phrase and Verb Phrase in Japanese, subject is generated under Aspect Phrase. Since this position is not a Case position, it moves to the SPEC of TP in order to receive Case.

M. Nakayama. M. Koizumi

318

I Remarks on Japanese suhjecis

References

Ahn,

H..

1991. Light

dissertation.

verbs.

University

negation,

and

Chomsky,

N.. 1981. Lectures

on government

Chomsky,

N., 1986. Barriers.

Cambridge,

Chomsky,

N., 1989. Some notes on economy

Mahajan Fukui,

(eds.), MIT Working

N.,

1986. A theory

Massachusetts Hasegawa,

Institute

Nagoya:

Papers

of category

Daigaku

Hoji. H.. 1989. VP preposing

projection

H.. S. Miyagawa,

Southern

California,

Kitdgawa.

Y..

English.

Doctoral

Foris.

and representation.

In: I. Laka,

A.

10. 43374.

and its applications.

Doctoral

dissertation,

subject

hypothesis.

In: T. Sakamoto,

Sooritsu

Kinen

Nihongo

Y. Abe (eds.).

Kyooiku

Kokusai

University.

in Japanese

on VP, held in conjunction Hoji.

Dordrecht:

of derivation

Nihongo-gakka

Nanzan

of California

and binding

in Linguistics

Hoji, H., 1985. Logical form constraints dissertation, University of Washington.

University

and

of Technology.

of Nanzan

Shinpojiumu.

in Korean

MA: MIT Press.

N., 1989. On the VP internal

Proceedings

VP-movement

of Wisconsin.

and configurational (and Korean).

with Southern

structure

in Japanese.

Paper presented

California

Doctoral

at the Syntax Workshop

Korean/Japanese

Linguistic

Conference.

at Los Angeles. and

H. Tada,

1989. NP-movement

The Ohio State University,

1986. Subject

in Japanese

in Japanese.

and Massachusetts

and

English.

Doctoral

Ms..

University

of

Institute

of Technology.

dissertation,

University

of

Massachusetts. Koizumi,

M.. 1990a, Secondary

Koizumi,

M.,

1990b.

depictive

Distribution

predicates

of temporal

in Japanese.

NI-phrases

Ms., The Ohio State University.

in Japanese.

Ms.. The Ohio

State

University. Koizumi. N., 1991. Syntax University. Koopman, Kuroda,

H. and D. Sportiche, S.-Y., 1970. Remarks

only’. Illustrating linguistic

and phrase

1988. Subjects. on the notion

manners

structures

Ms., University

Bulletin

Ms., The Ohio State

of California

of subject with reference

in which formal

Part 2. Annual

of Japanese.

at Los Angeles.

to words like crl.so, eren, or

systems are employed

4, 127~ 152. (Reprinted

as auxiliary in Papers

devices in in Japanese

11 (1986), 121-156.)

S.-Y..

1988, Whether

In: W. Poser CSLI. Marantz.

certain

descriptions:

Linguistics Kuroda.

of adjuncts

A.,

(ed.), Papers 1984. On

the

we agree or not: A comparative

syntax

from

Workshop

the Second

nature

International

of grammatical

relations.

of English

and Japanese.

on Japanese

Cambridge.

MA:

Syntax. The

MIT

Press. McNulty, E.. 1988. Syntax of adjunct predicates. Doctoral dissertation. University of Connecticut. Miyagawa, S.. 1989a. Structure and Case marking in Japanese. New York: Academic Press. Miyagawa, Miyagawa.

S., 1989b. Light verbs and the ergative hypothesis. Linguistic Inquiry 20. 647-65X. S., 1990. Case realization and scrambling. Ms.. The Ohio State University.

Nakayama.

M., 1990. Accessibility

The Ohio State University. Nakayama, M. and K. Tajima, State University

to the antecedents

in Japanese

1991. Four types of purposive

and University

of Wisconsin.

sentence

comprehension.

clauses in Japanese.

Ms.,

Ms., The Ohio

M. Nakayama, M. Koizumi / Remarks on Japanese subjects

Nakayama,

M. and W. Tawa,

Ohio State University Nakayama,

M., M. Koizumi

Paper presented

1991. Implicit

and Amherst and

J.-Y., 1988. Verb movement,

Reinhart,

T., 1976. The syntactic of Technology. S., 1983. The syntactic

Saito,

M.,

1985. Some

Sportiche,

Inquiry

a VP internal

subject

hypothesis.

minimality Tsukuba

domain forms

and

English

the

aspectual

status

of

Studies 7, 187-227.

of IP. Linguistic

Inquiry

20, 365424.

of anaphora.

Doctoral

dissertation,

Massachusetts

of predication.

Doctoral

dissertation,

Massachusetts

and

their

theoretical

implications.

Doctoral

for constituent

structure.

of Technology.

of floating

quantifiers

of adverbs.

J., 1987. Configurationality

Linguistics,

in Japanese

Institute

and its corollaries

19, 425449.

L.. 1989. The syntax

Williams,

1991. Against

UC and the structure

asymmetries

Massachusetts

D., 1988. Theory

Linguistic Travis,

ms. The

of Technology.

dissertation,

Whitman,

in Japanese.

Chicago.

Case and empty pronouns.

Pollock,

Institute

M. Ogino,

at the LSA meeting,

Oka. T., 1988. Abstract

in direct passives

College.

relativized 1990. Sentential negation, Linguistic Review 7, 183-23 1.

Ouhalla. J., auxiliaries.

Institute Rothstein,

arguments

319

351-371.

Dordrecht:

E.. 1980. Predication.

Ms., McGill parameter.

University.

In: T. Imai, M. Saito (eds.), Issues in Japanese

Foris.

Linguistic

Inquiry

15, 131-153.