Raising-to-object in Japanese as two operations

Raising‑to‑object in Japanese as two operations

著者（英） Naoyuki Akaso

journal or

publication title

Doshisha literature

number 51

page range 1‑28

year 2008‑03‑15

権利（英） English Literary Society of Doshisha University

URL http://doi.org/10.14988/pa.2017.0000014855

Raising-to-Object in Japanese as Two Operations

NAOYUKI AKASO

1. Introduction

More than three decades have passed since Kuno (1976) first identified Raising-to-Object (henceforth RTO) in Japanese, such as we find (lb).l

(1) a. John-ga Bill-ga tensai-da to omot-teiru.

John-NOM Bill-NOM genius-COP COMP think-PROG 'John thinks that Bill is a genius.'

b. John-ga Bill-o tensai-da to omot-teiru.

John-NOM Bill-ACC genius-COP COMP think-PROG 'John thinks Bill to be a genius.'

Although numerous attempts have been made to pin down the properties of this syntactic phenomenon, there is no agreement on how the RTO construction is derived. What is certain, however, from previous efforts, is what aspects should be considered in analyzing this construction.

Based upon previous research, this paper proposes that RTO in Japanese consists of two distinct operations, not a single operation as traditionally assumed. To be more specific, it consists of two operations at different levels: one is in overt syntax and the other is at LF.2 The former opera- tion is the T-to-C movement in the sense of Pesetsky and Torrego (2001, 2004) and the latter the C-to-V incorporation advocated by Ogawa (2007).

The reason for this 'two-operation' analysis will be explained in what follows. In Section 2, some representative arguments for RTO from Kuno

Cl]

(1976) will be reviewed. In Section 3, after reviewing the interaction of RTO and indeterminate pronoun binding (in the sense of Kishimoto, 2001), which makes the behavior of RTO more mysterious, I will discuss the work of Ogawa (2007) which appears to explain this complicated data successfully. After presenting some problematic cases for Ogawa's analy- sis in Section 4, an alternative will be proposed in Section 5. Section 6 consists of a brief summary of the entire discussion.

2. Some representative arguments for Raising-to-Object

In this section, let me first review some of the arguments for RTO in Japanese from Kuno (1976) and verify that the accusative-cased DP is in the matrix clause.³

Kuno (1976) presents some arguments that the embedded DP in ques- tion is raised to the matrix clause. Let me start with the matrix adverb.

The matrix adverb may be placed after the accusative DP in (2a), but not after the embedded subject in (2b).

(2) a. John-ga Bill-o orokanimo tensai-da to omot-teiru.

John-NOM Bill-ACC stupidly genius-COP COMP think-PROG 'Stupidly, John thinks of Bill as a genius.'

b. *John-ga [Bill-ga orokanimo tensai-da to] omot-teiru.

John-NOM Bill-NOM stupidly genius-COP COMP think-PROG 'Stupidly, John thinks that Bill is a genius.'

This contrast indicates that the accusative DP in (2a) belongs to the matrix clause, while the embedded subject DP in (2b) does not. Since the adverb orokanimo is in the matrix clause, the subject DP Bill-ga must be outside of the embedded clause, and so the sentence is ungrammatical.

The second piece of evidence concerns scopal ambiguity. (3a) is scopally ambiguous, while (3b) is not.

(3) a. Dareka-ga minna-o tensai-da to omot-teiru.

Someone-NOM all-ACC genius-COP COMP think-PROG 'Someone thinks all to be geniuses.'

(some>every; every>some)

b. Dareka-ga minna-ga tensai-da to omot-teiru.

Someone-NOM all-NOM genius-COP COMP think-PROG 'Someone thinks all are geniuses.'

(some>every; *every>some)

The ambiguity in (3a) illustrates how the raised universal quantifier (i.e., accusative-marked one) can take a wider scope than the matrix subject (i.e., existential quantifier). In (3b) the universal quantifier cannot take a wider scope than the existential quantifier, since these two quantifiers belong to different clauses and cannot interact with each other.

Finally, it is generally accepted that pronouns in Japanese are subject to Binding Condition B, as shown in (4).4

(4) *Johni-ga karei-o hihansi-ta.

Johlli-NOM hei-ACC criticize-PAST '*Johni criticized himi.'

Keeping this in mind, let me consider the following sentences.

(5) a. ?Johm-ga [karei-ga tensai-da to] omot-teiru.

John-NOM he-NOM genius-COP-COMP think-PROG 'John; thinks that hei is a genius.'

b. *Johlli-ga karei-o tensai-da-to omot-teiru.

John-NOM he-ACC genius-COP-COMP think-PROG '*Johlli thinks ofhimi as a genius.'

Although both of these sentences are ungrammatical, (5a) sounds much better than (5b) because in the latter case the accusative case-marked DP is in the matrix clause and this violates the Binding Condition B. From these facts and others, Kuno (1976) claims that Japanese has RTO, as does English (Postal, 1974).5

It is a fact, however, that a few researchers who are opposed to the raisingIDP-movement analysis of RTO have proposed the involvement of other mechanisms in (1b). Some claim that the DP in question is base- generated in some position of the matrix clause (referred to as the prolep- sis analysis in Takano (2002», and others claim that the DP is still in the embedded clause and case is checked through Agree (e.g., Hiraiwa (2001». For these non-RTO analyses, the reader may refer to the cited works. So far as I can determine, arguments have accumulated to demon- strate that the accusative-marked DP, originally base-generated in the embedded finite clause, has been displaced to some position in the domain ofthe matrix clause. The process is (partially) schematized as fol- lows:

(6) [ ... DPi (ACC) [Gp [TP ti T] COMP] matrix V ]

In the following sections of this paper, without further scrutinizing the non-RTO analyses, I will accept the view that raising/DP-movement occurs in (1b), as widely assumed, and proceed with my analysis.⁶

3. Indeterminate Pronouns and C-to-V Incorporation

In this section I will consider the interaction between RTO and indeter- minate pronoun binding, which has made the analysis of RTO more com- plicated and challenging.

In Japanese, indeterminate pronouns such as dare ('anyone') and nani ('anything'), when they precede the Q particle mo, act as negative polari- ty items (henceforth NPI). (McGloin, 1976)

(7) a. Taro-wa nani-mo Taro-TOP anything-Q

kawa-nakat-ta.

buy-NEG-PAST 'Taro did not buy anything.'

b. Dare-mo sono hon-o kawa-nakat-ta.

anyone-Q that-book-ACC buy-NEG-PAST 'No one bought that book.'

The Q particle mo, however, is not necessarily preceded directly by inde- terminate pronouns, though there are some kinds of restrictions, as illus- trated below.

(8) a. Taro-wa nani-o kai-mo si-nakat-ta.

Taro-TOP anything-ACC buy-Q do-NEG-PAST 'Taro did not buy anything.'

b.*Dare-ga warai-mo si-naka-ta.

Anyone-NOM laugh-Q do-NEG-PAST 'No one-laughed.'

c. *Dare-ga Hanako-o home-mo si-nakat-ta.

Anyone-NOM Hanako-ACC praise-Q do-NEG-PAST 'No one praised Hanako.'

When the indeterminate pronoun is accusative case-marked and mo is attached to V, as in (8a), the sentence is acceptable. On the other hand, when the subject is an indeterminate pronoun and mo is attached to V, the sentence is ruled out, as shown in (8b) and (8c). This amounts to showing that the object in a simple clause, but not the subject, is within the domain of mo, which is a condition of licensing NPI. In addition, when mo is attached to the complementizer to in the embedded CP, both subject and object in the embedded clause can be indeterminate pro- nouns, as illustrated below.

(9) a. Hanako-wa Taro-ga nani-o kat-ta to-mo Hanako-TOP Taro-NOM anything-ACC buy-PAST COMP-Q omowa-nakat-ta.

think-NEG-PAST

'Hanako did not think that Taro bought anything.' b. Taro-niwa dare-ga Masao-o home-ta to-mo

Taro DAT-TOP anyone-NOM Masao-ACC admire-PAST COMP-Q omo-e-nakat-ta.

think-can-NEG-PAST

'Taro could not think anyone admired Masao.'

This is because the domain of mo covers the entire embedded clause.

Based on these facts, Kishimoto (2001) proposes that the scope of mo is defined by the notion of domain given below:

(10) Y is in the domain of a head X if it is contained in Max (X), where Maxex) is the least full-category maximal projection dominating X.⁷

This means that, when mo is attached to V, the scope of mo is defined as

the VP. If V head-moves into v, the scope of mo is extended up to vP.

Furthermore, if mo is attached to the embedded C, the embedded clause is within the scope of moo

Keeping this (10) in mind, let us examine RTO with indeterminate pro- noun binding.

(11) a. Masao-ga [dare-ga tensai-da to-mol omot-tei-na-i.

Masao-NOM anyone-NOM geniusl-COP COMP-Q think-PROG-NEG 'Masao does not believe that anyone is a genius.'

b. *Dare-ga [Masao-ga tensai-da to-mol omot-tei- Anyone-NOM Masao-NOM genius-COP COMP-Q think-PROG- na-i.

NEG

'No one believes Masao to be a genius.'

In (11a) mo is attached to the embedded C and the scope of mo should be extended to cover the embedded CP. That is why (11a) is grammatical.

(l1b), on the other hand, is ruled out because the subject DP in the matrix clause cannot be within the scope of moo Thus (10) works well.

But when it comes to the following sentence (12), a serious problem occurs.

(12) Masao-ga dare-Oi [ti tensai-da to-mol omot-tei-na-i.

Masao-NOM anyone-ACC genius-COP COMP-Q think-PROG-NEG 'Masao does not believe anyone to be a genius.'

As seen in Section 2, an indeterminate pronoun with accusative case dare-o is in the matrix clause according to RTO. At the same time, mo is attached to C and the scope of mo is extended to the embedded clause.

This means that dare-o cannot be bound by mo and the sentence would be predicted to be ungrammatical. Nevertheless, the sentence (12) is grammatical. Sakai (1998) argues that the trace of a raised DP left in the embedded clause suffices for the indeterminate pronoun (i.e., dare-o) to be licensed by the particle mo in (12). But this is not convincing. In (8b) and (8c) the subject is base-generated in Spec-vP and raises to Spec-TP, leaving its trace in the original position. Although this trace is in the scope of mo, the sentence is ruled out. Ogawa (2007) concludes from this observation that an NPI indeterminate pronoun must be in the scope of the particle mo, not only in overt syntax, but also at LF.

To resolve the paradoxical situation that the DP in question must be raised out of the embedded clause, and is still not out of the scope of mo, Ogawa (2007) claims that the head of the complement CP in the RTO construction in Japanese undergoes incorporation into the verb in the matrix clause. The partial structure is schematized as below:

(13)

vP

~

dare-o V'

~

V v

~ ~

CP tv V C-mo

TP

~

The complementizer to with the particle mo incorporates into the select- ing verb in the matrix clause. Then V raises to v, which helps to extend the scope ofmo up to the matrix vP. Ogawa (2007) presents the following contrast to support his proposal.

(14) a. Taro-wa darei-o kashikoi to-mo Taro-TOP anyone-ACC smart COMP-Q

soitsUi-no gendoo-kara kanzi-nak-atta.

he-GEN speech and behavior-from feel-NEG-PAST

'Taro didn't feel anyonei to be smart, based on hisi speech and behavior.'

b. ?*Taro-wa darei-ga kashikoi to-mo Taro-TOP anyone-NOM smart COMP-Q

soitsUi-no gendoo-kara kanzi-nak-atta.

he-GEN speech and behavior-from feel-NEG-PAST

'Taro didn't feel anyonei to be smart, based on hisi speech and behavior.'

It is known that the bound pronoun soitsu 'that guy/he' must be c-com- manded by its antecedent.s This contrast demonstrates that in (14a) soit- su has a proper antecedent as a result of RTO, while in (14b) it does not, as illustrated below.

(15)

a. vP

VP v

dare-o

~ ~

V'

~~

CP~~

~

TP tc

b. vP

~P V~v vp~

/~ /\ /~

CP tv ~ V C-mo

TP

/~

dare-ga

~

Thus, Ogawa claims that the contrast in (14) can be explained by the C- to-V incorporation.

4. Problems

Ogawa's proposal on C-to-V incorporation seems to be valid for the empirical data of indeterminate pronoun binding as in (12). However, it is not at all clear whether the grammaticality contrast in (14) is indeed recognizable. Although Ogawa argues that (14b) is ungrammatical, this judgment seems to be very subtle, especially when the matrix verb is combined with a non-RTO verb (e.g., kiku 'hear'), in which case the sen- tence seems to be generally acceptable. Actually, the informants I asked all agreed that the following sound quite natural. ⁹

(16) a. Taro-wa [darei-ga tensai-da to-mol Taro-TOP anyone-NOM genius-COP COMP-Q soitsUi-no hahaoya-kara kika-nak-atta.

he-GEN mother-from hear-NEG-PAST

'Taro didn't hear from anyonei's mother that hei was a genius.' b. Taro-wa [darei-ga hannin-da to-mol

Taro-TOP anyone-NOM criminal-COP COMP-Q soitsUi-no gendoo-kara wakara-nak-atta.

he-GEN speech and behavior-from understand-NEG-PAST 'Taro was not able to predict that anyonei was a criminal based on hisi speech and behavior.'

Furthermore, when another RTO verb, dantei-suru 'decide', is used as a matrix verb, it is very difficult to identify the grammaticality contrast.

(17) a. Taro-wa [darei-ga hannin-da to-mol

b.

Taro-TOP anyone-NOM criminal-COP COMP-Q

soitsUi-no gendoo-kara dantei-deki-nak-atta.

he-GEN speech and behavior-from decide-can-NEG-PAST 'Taro was not able to determine that anyonei is a criminal, based on hisi speech and behavior'

Taro-wa darei-o hannin-da to-mo Taro-TOP anyone-ACC criminal-COP COMP-Q

soitsUi-no gendoo-kara dantei-deki-nak-atta.

he-GEN speech and behavior-from decide-can-NEG-PAST 'Taro was not able to determine that anyonei is a criminal, based on hisi speech and behavior'

Wherever the differences in judgments may be, it seems that there might be some other factors that need to be considered for the data Ogawa pre- sents. Even if I accept his judgment, the following data may cause a seri- ous problem for his proposal that C-to-V Incorporation takes place in overt syntax. Let us consider the following sentence carefully.

12

(18) *Taro-wa Jiro-o kashikoi to-mo dare-no Taro-TOP Jiro-ACC smart COMP-Q anyone-GEN gendoo-kara kanzi-nak-atta.

speech and behavior from feel-NEG-PAST

'Taro did not feel Jiro to be smart, based on anyone's speech and behavior.'

(18) is completely ungrammatical. If I adopt Ogawa's analysis that the C- to-V movement takes place in overt syntax, however, it would be predict- ed that (18) would be grammatical, for the scope of mo is extended to the matrix verb, by which the indeterminate pronoun, dare in PP, is within the scope of mo, as shown in ^(19).

(19)

vP

V

~

V

~

TP tc dare ...

This observation makes it clear where the problem originates. That is, it originates with the assumption that C-to-V Incorporation should occur as overt syntactic movement.

5. Toward a solution

In the preceding section, I have seen that Ogawa's syntactic C-to-V incorporation, which is essential in triggering the overt movement (i.e., raising) of the DP to the matrix Spec-V, has serious empirical problems.

In spite of these problems, however, Ogawa's analysis is attractive in that it can successfully explain (12), which no previous analysis could handle in a convincing way. Therefore, what I need is to seek a mecha- nism for (18), while maintaining the effective portion of Ogawa's analy- sis: the effect of C-to-V incorporation.

As a possible solution to this problem, I propose that RTO in Japanese consists of two distinct operations: one is in overt syntax and the other is at LF. The latter is the very same operation that Ogawa proposed: C-to-V Incorporation. Ogawa claims that this incorporation takes place in overt syntax, but as observed in the previous section, a serious problem remains. Therefore, my idea is to retain his C-to-V Incorporation analy- sis, but as an operation at LF.

It should be noted that if! adopted Ogawa's generalization that an NPI indeterminate pronoun must be within the scope of the Q particle mo both in overt syntax and at LF, I would fail to explain (12) (repeated below as (20», since the raised DP is out of the scope ofmo in syntax, as illustrated in (21).

(20) Masao-ga dare-o tensai da to-mo omot-tei-na-i.

Masao-NOM anyone-ACC genius-COP COMP-Q think-PROG-NEG 'Masao does not believe anyone to be a genius.'

14 (21)

VP

dar~'

CP

/~ v

TP

~

vP

v

If I assume, instead, that an indeterminate pronoun must be within the scope of mo at LF, not in overt syntax, then I can successfully explain (20) with LF Incorporation. Actually, this assumption (i.e., the legitimacy of indeterminate pronoun binding is determined at LF) is the same as that proposed on the basis of the so-called Nominative Object construc- tions in Kishimoto (2001).

With the assumption that the binding of indeterminate pronouns by mo is determined in LF configurations, however, my analysis faces the same problem as Ogawa did; that is, (18) is left unexplained. The indeter- minate pronoun in pp is bound by mo and this incorrectly predicts that the sentence should be ruled in.

To avoid this catch-22 situation, I should carefully examine the differ- ences between (20) and (18). Close scrutiny enables us to recognize that the indeterminate pronoun in the former (i.e., dare-o) is an argument, while that in the latter (i.e., dare-no gendou-kara) is an adjunct. This is what is called argument/adjunct asymmetry, and this observation is sup- ported by the following data.

15 (22) a. Taro-wa sono nyusu-ga/o honto-da to omowa-nak-atta.

Taro-TOP that news-NOM/ACC true-COP COMP think-NEG-PAST 'Taro didn't think that that news was true.'

b. Taro-wa sono nyusu-ga/o honto-da to Taro-TOP that news-NOM/ACC true-COP COMP dono bangumi-kara-mo omowa-nak -atta.

any program-from-Q think-NEG-PAST

'Taro didn't think from any TV program that that news was true.' c.*Taro-wa sono nyusu-ga/o honto-da to-mo

Taro-TOP that news-NOM/ACC true-COP-COMP-Q dono bangumi-kara omowa-nak-atta.

any program-from think-NEG-P AST

'Taro didn't think from any TV program that that news was true.'

The contrast between (20) and (22c) demonstrates that an argument indeterminate pronoun must be within the scope of mo at LF, and an adjunct indeterminate pronoun must be within the scope of mo both in overt syntax and at LF. To put it differently, the legitimacy of indetermi- nate pronoun binding is determined on the basis of syntactic or LF con- figurations, depending on the grammatical status of the indeterminate pronoun in question: argument or adjunct.

If this is on the right track, if C-to-V incorporation indeed takes place at LF, I can correctly predict (18) to be ungrammatical, since the indeter- minate pronoun in PP is not within the scope of mo in overt syntax, which violates the licensing condition of the adjunct indeterminate pro- noun: the adjunct indeterminate pronoun must be within the scope of mo both in overt syntax and at LF. Although my analysis has succeeded in overcoming one difficulty, another problem awaits it. That is, I need to look for some syntactic mechanism to move the DP out of an embedded

clause.

Generally, displacing any argument out of a finite clause into any argu- ment position of the matrix clause is not allowed, since movement from A-position through A'-position back to an A-position inevitably leads to improper movement. To deal with Japanese RTO, one should be careful to avoid this improper movement (sometimes referred to as the illegiti- mate chain: A-A'-A, or Tensed-S island effect). Let me consider two exam- ples along with Ogawa's analysis for their evasive mechanisms. One example is from Tanaka (2002), which evades the problem with a stipula- tion assuming that the position (the edge of the CP) at which the moved phrase lands before moving out of the CP is an A-position as an escape hatch for A-movement. Because of this, accusative-marked DP can move without improper movement. Another example is in Kawai (2006), which proposes that the RTO construction contains a small clause. Small claus- es are generally categorized as non-finite clausal projections which do not include COMP(lementizer) and T(ense), or I(nflection). This means that there is no CP boundary to be crossed, so that the raised DP does not cause a problem of improper movement.

In the case of Ogawa's C-to-V Incorporation, an A-movement across the CP-boundary is possible, since the embedded C is incorporated into the matrix V and the CP boundary disappears theoretically. That is, RTO is a single-step movement to the Spec of the matrix V.

In one way or another, various proposals have been offered to circum- vent the improper movement. Since I interpret C-to-V incorporation as an LF operation as illustrated above, I need to look for some syntactic mechanism for the DP movement portion of RTO, which guarantees that the accusative marked DP will be able to move out of the embedded clause without improper movement. In this paper, I propose that the rel- evant mechanism should be the T-to-C head movement in the sense of

Pesetsky & Torrego (2001, 2004). Below, I will explain how T-to-C move- ment is relevant to the avoidance of improper movement. My explanation is based on two independent studies of English syntax: Hirai (2004) and Pesetsky & Torrego (2004).

Let me start with Hirai's (2004) study on Heavy NP Shift (henceforth HNPS) in English.

Hirai demonstrates the following contrast of HNPS.

(23) a. I have attemptedlhoped/sought [to answer ti] for many years [the most difficult questions that Chomsky presented]i.

b. I have demanded ^[to know ti] for many years [exactly what hap- pened to Rosa Luxemburgh.

c. I have desired [to know ti] for many years [exactly what hap- pened to Rosa Luxemburgh. (Nakajima, 1986: 15)

d. I have expected [to find ti] since 1986 [the treasure said to have been buried on that islandh. (Postal, 1974:93)

e.?! intended/planned [to read ti] last night [a harshly critical review of Chomsky's exciting book];.

f. The editor has wanted [to publish ti] for many years [a harshly crit- ical review of Ch om sky's exciting book]i. (Matsuyama, 1998:102) g. I have wished [to find t;] since 1986 [the treasure said to have

been buried on that island]i.

(24) a.*John bothered [to meet ti] last night [the president of IBM's Texas Branch);.

b. *The editor has claimed [to hate ti] for many years [a harshly crit- ical review of Ch om sky's exciting bookh. (Matsuyama, 1998:10) c. *?George has declined [to accept til for many years [the most

famous award of country music in Americah.

d. *The editor has hated/loved [to publish ti] for many years [a

18

harshly critical review of Chomsky's exciting book]i.

e.*John has loathed [to ride ti] for many years [the black horse named "Speed" from Texas);.

Hirai (2004) claims that the difference between (23) and (24) lies in the irrealis/realis distinction of the infinitival complement.^IO In the case of the former (i.e., irrealis) HNPS is allowed, while in the latter it is not.

Hirai assumes that irrealis infinitival clauses are weak phases (his nota- tion is CP) which allow a heavy NP to move out of them, and that realis clauses are strong phases (C*P) which do not, paralleling with Chomsky's proposal about the distinction between vP (weak phase) and v*P (strong phase).

Hirai speculates that the irrealis/realis distinction of the infinitival clauses comes from the presence/absence of Spec-C, following Melvold (1991) in his analysis of finite complements to factive verbs: A realis CP has an operator in Spec-C.l! Hirai assumes that there is no Spec-C in irre- alis CPs. Whatever the reason may be, it is a fact that the irrealis/realis distinction parallels the difference between weak CPs and strong CPs.

At this point I could insist that predicates allowing RTO should be weak CPs, so that the raised DP can move out of the embedded clause without violation of improper movement. But this is just a stipulation and it would not allow us to explore the reason why irrealis CPs are weak. In the following paragraphs I will argue-on the basis of Pesetsky

& Torrego's analysis on the infinitival complements to nouns-that T-to- C movement, which takes place in irrealis CPs, but not in realis CPs, is a trigger for making a clause boundary disappear and eventually to mak- ing the DP move out of the CP without causing improper movement.

In a quite different context, Pesetsky & Torrego (2004) propose that the head movement of T to C occurs in irrealis CPs, while it does not occur in

19 realis CPs, based on the syntactic behavior of infinitival clauses. See the following contrast of the infinitival complements to nouns.

(25) a. *Mary's hatelhatred to ride in the car b. *Sue's love to solve problems

c. *Bill's luckiness to win a prize d. *Mary's dare to defy the government e. *Bill's help to understand things f. *John's neglect to turn offthe light g. *Mary's venture to ring the doorbell h. *John's dislike to hear rumors about them

(Pesetsky & Torrego 2004: 520) (26) a. Mary's desire to win

b. Bill's agreement to ride in the car c. John's wish to win a prize

d. Sue's attempt to defy the government e. Mary's consent to undergo the operation f. Bill's demand to be taken to the king g. John's hope to understand things h. John's learning to play the piano

(Pesetsky & Torrego, 2004: 521)

The nouns are divided into two groups: those that can take an infinitival complement and those that cannot. Pesetsky & Torrego claim that the former group takes a realis complement, and the latter an irrealis one.

The reason why irrealis infinitival clauses can act as complements to N is, according to Pesetsky & Torrego, that a null complementizer has the interpretative TENSE feature (= iT), which triggers T-movement to C.

Putting aside technical details, let me assume that their analysis is on

the right track.

Combining Pesetsky & Torrego with Hiari, it may be concluded that T- to-C movement helps CPs to be weak, which means that the clause boundary disappears, allowing an embedded DP to be raised to an argu- ment position in the matrix clause.

I believe that the very same mechanism works in the case of Japanese RTO. In other words, the T-to-C head movement makes Japanese RTO possible, without improper movement. The belief is supported by the fact that irrealis can be categorized to be non-past [-PAST] (putting aside a fine semantic analysis for now). As is generally reported (see Kuno (1976) and Ohta (1997), among others), Japanese RTO is disallowed for a [+PAST] complement. Examine the sentence (27)Y

(27) *? Kanojo-wa sono otoko-o sagishi da-tta to shinji-teiru.

she-TOP that man-ACC swindler be-PAST COMP believe-PROG 'She believes that man to have been a swindler.'

Thus, T-to-C movement, as observed in an irrealis complement, is a plau- sible mechanism responisble for the raising/DP-movement portion of Japanese RTO constructions.

One bit of evidence for the T-to-C movement in Japanese can be found in Uchibori (2001), which argues that a certain kind of subjunctive (i.e., irrealis) CP complement acts as a weak CP, and that raising to the matrix subject position is possible. Observe the following example.

(28) a. Johm-ga saikin [ti motto benkyoosu-ru yooni] nat-ta.

John-NOM recently more study-NON-PAST SBJ-COMPhappen-PAST 'John has come to study harder recently.'

b. *Johm-ga saikin [ ti motto benkyoosi-ta yooni] nat-ta.

21 John-NOM recently more study-PAST SBJ-COMP happen-PAST 'It recently happened (as a natural result) that John studied harder.'

According to Uchibori (2001), a nominative subject can undergo raising from a subjunctive (irrealis) embedded clause, but not from a realis clause.¹³

Thus, T-to-C movement is responsible for causing embedded irrealis complements to be weak CPs, which eventually allow raising out of the CPs without improper movement.

Summarizing, the main gist of my discussion in this section can be dia- grammatically schematized as below.

(29)

I

I

vP

VP v

DP-o

~

CP

~

TP

~

v~ T~(-mO)

D

tDP

22

~

vP

VP v

DP-a

~

v~

CP

~

~

T CC-maY

TP

~

vP

~

6

6. Conclusion

In this paper I have proposed that Japanese RTO is not a single opera- tion as traditionally analyzed, but the result of two distinct operations:

one is the displacement operation in overt syntax (e.g., T-to-C move- ment), and the other C-to-V Incorporation at LF. The former enables the accusative marked DP to move out of the embedded clause without illicit processes involved, and the latter can correctly deal with the interaction between RTO and indeterminate pronoun binding, some cases of which reject the single operation approaches to RTO including Ogawa's (2007) C-to-V Incorporation. Mter the presentation of some problematic data for Ogawa's analysis, which assumes that incorporation takes place in overt syntax, I proposed that incorporation should occur at LF. In Ogawa's analysis, the movement in question must take place in overt syntax, for it allows the embedded CP to be weak and helps the accusative marked DP move into an argument position in the matrix clause. To guarantee the syntactic movement of the raised DP out of the embedded clause, I

23 have proposed the possibility ofT-to-C movement which can make a CP weak. This portion (i.e., T-to-C Movement) is truly just a possibility, and one of several other possible candidates (or mechanisms) suggested so far for Japanese RTO, but I believe that the mechanism may contribute to a deeper understanding of the often-reported generalization that irrealis clauses (infinitival clauses or small clauses) lack Cor/and T. In this sense, my proposal of T-to-C Movement may be plausible and worth examining further. Of course, this remains to be confirmed by future research.

Abbreviations

Ace: accusative COMP: complementizer COP: copula DAT: dative GEN: genitive NEG: negation NOM: nominative NON-PAST: present PAST: past PROG: progressive Q: Q-particle SBJ: subjunctive TOP: topic CP: Complementizer Phrase

HNPS: Heavy NP Shift LF: Logical Form RTO: Raising-to-Obeject

DP: Determiner Phrase NPI: Negative Polarity Item

*1 would like to thank Matazo Izutani and two anonymous DL referees for helpful comments on an earlier draft. I also thank Phillip Morrow and William Herlofsky for suggesting stylistic improvements. All remaining inadequacies are my own.

Notes

1 Some of the previously published sentences cited in this paper may contain slight changes made silently in order to avoid possibly insulting expressions like 'baka (a fooD'. Of course, I have made sure these changes did not alter grammaticality. I thank the referee for this thoughtful suggestion.

2 LF is the abbreviation for Logical Form. It is the component of a grammar

which converts the syntactic structures into the initial representation of sen- tence meaning.

3 In the. Government and Binding Theory, the Projection Principle and the

x-

bar theory presented difficulties for any movement analysis of RTO.

Therefore, other mechanisms were proposed. For example, Saito (1983) sug- gested that the RTO construction might involve a null pronominal in the embedded clause, which led to Takano's (2002) 'prolepsis' analysis.

4 For convenience, I adopt the term 'the Binding Condition B,' of the Government and Binding Theory, which was not available when Kuno wrote his paper in 1976. It roughly means that pronouns cannot be coreferential with subjects in the same clauses.

5 Tanaka (2002) presents additional data supporting RTO in Japanese:

Proper Binding Condition, long-distance and local scrambling, Multiple-Wh questions, Cleft constructions, Negative Polarity Items and adverbs. See Tanaka (2002) for details.

6 For example, Tanaka (2002) demonstrates that the prolepsis analysis can- not explain the ungrammaticality of the following, since the preposed CP con- taining PRO would not cause any problem.

(i) *[cp PRO tensai da to); Taro-wa Hanako-o (orokanimo) ti genius COP COMP Taro-TOP Hanako-ACC (stupidly) omot-ta.

think-PAST.

'(Stupidly,) Taro considered Hanako to be a genius.'

(ii) *[cp ti tensai da to]j Taro-wa Hanako-oi (orokanimo) tj genius COP COMP Taro-TOP Hanako-ACC (stupidly) omot-ta.

think-PAST.

'(Stupidly,) Taro considered Hanako to be a genius.'

Under the movement analysis, the ungrammaticality of (ii) can be explained because the trace in the preposed CP cannot be bound, which contravenes the Proper Binding Condition: trace must be bound. (Fiengo, 1977). Tanaka (2002) also points out that Hiraiwa's Agree (optional RTO) analysis fails to explain the grammatical differences in (5). Thus, I believe that RTO is a movement to

25

the matrix object position.

7 This definition is from Kishimoto (2001: 601).

S Informally, 'c-command' is the relationship between an element and another element it is superior to in the syntactic structure.

9 I asked three linguists for these judgments.

10 According to Crystal (1997:321), these terms are explained as "a term used in the study of EPISTEMIC MODALITY: in a realis ('real') assertion, a propo- sition is strongly asserted to be true, the speaker being ready to back up the assertion with evidence or argument. It is opposed to an irrealis ('unreal') assertion, where the proposition is weakly asserted to be true, but the speaker is not ready to support the assertion."

11 An anonymous referee points out that Kiparsky & Kiparsky (1970) state that only non-factive predicates allow what is called the ECM construction, as illustrated below.

(i) Non-factive

a. I believe Mary to have been the one who did it.

b. He fancies himselfto be an expert in pottery.

c. I supposed there to have been a mistake somewhere.

(ii) Factive

a. *1 resent Mary to have been the one who did it.

b. *He comprehends himself to be an expert in pottery.

c. *1 took into consideration there to have been a mistake somewhere.

12 The sentence (27) is excerpted from Kawai (2006), who admits that the degradation is mild. He also reports (through a reviewer of his paper) that some people find (27) quite acceptable, and attributes it to a possibility of left- dislocation, as well as idiolectalldialectal variations. As for the involvement of left-dislocation, he states as follows:

For the reviewer, Japanese ECM (=RTO) sentences "require a contrastive interpretation of the accusative" subject. Such a contrastive reading is optional for me (and for those 1 consulted with), however. The obligatory presence of the contrastive interpretation, which presumably is accompa- nied by contrastive stress, suggests the involvement of left-dislocation, rather than RO (=RTO). Perhaps, then, it might be the case that some

Japanese speakers have only the left-dislocation option of ECM (=RTO), lacking the RO (=RTO) option.

13 As Uchibori (2001) mentioned in her footnote 4, the result complement marked by -yooni is considered to be semantically factive. I leave this seman- tic issue untouched here and assume that -yooni requires an irrealis form, as does the purpose complement.

References

Boscovic,

Z.

Chomsky, N. (2001). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 1-52). Cambridge, MA: The MIT Press.

Crystal, D. (1997). A dictionary of linguistics and phonetics (4^th ed.).

Oxford: Blackwell.

Fukui, N. (1986). A theory of category projection and its applications.

Doctoral dissertation, MIT.

Hirai, D. (2004). Control infinitives and two types of CP phases. English Linguistics, 21, 241-264.

Hiraiwa, K. (2001). Multiple agree and the defective intervention con- straint in Japanese. MIT Working Papers in Linguistics, 40,67-80.

Kawai, M. (2006). Raising to object in Japanese: A small clause analysis.

Linguistic Inquiry, 37,329-339.

Kawamura, T. (2004). A feature-checking analysis of Japanese scram- bling. Journal of Linguistics, 40, 45-68.

Kiparsky, P., & Kiparsky, C. (1970). Fact. In M. Bierwisch & K. Heidolph (Eds.), Progress in linguistics (pp. 143-73). The Hague: Mouton.

Kishimoto, H. (2001). Binding of indeterminate pronouns and clause structure in Japanese. Linguistic Inquiry, 32, 597-633.

Kuno, S. (1976). Subject raising. In M. Shibatani (Ed.), Syntax and

semantics 5: Japanese Generative Grammar (pp. 17-49). New York:

Academic Press.

Matsuyama, T. (1998). Shugo tougyo doushi hobun no toumeisei no kouka ni tuite [On the effects of transparency of complements of subject control verbsl. Proceedings of JELS, 15, 101-110.

McGloin, N. (1976). Negation. In M. Shibatani (Ed.), Syntax and seman- tics 5: Japanese Generative Grammar (pp. 371-419). New York:

Academic Press.

Melvold, J. (1991). Factivity and definiteness. MIT Working Papers in Linguistics, 15,97-117.

Nakajima, H. (1986). Apparent unboundedness in the want-equi con- struction. Metropolitan Linguistics, 6, 14-30.

Ogawa, Y. (2007). C-to-V incorporation and subject raising across CP- boundary. English Linguistics, 24, 33-66.

Ohta, K. (1997). Tense in the subject raising constructions. Japanese / Korean Linguistics, 6, 353-368.

Pesetsky, D., & Torrego, E. (2001). T-to-C movement: Causes and conse- quences. In M. Kenstowicz (Ed.). Ken Hale: A life in language.

Cambridge, MA: The MIT Press.

Pesetsky, D., ^& Torrego, ^E. (2004). Tense, case, and the nature of syntac- tic categories. In J. Gueron & J. Lecarme (Eds.), The syntax of time.

Cambridge, MA: The MIT Press.

Postal, P. M. (1974). On raising. Cambridge, MA: The MIT Press.

Saito, M. (1983). Comments on the papers on generative syntax. In Y.

Otsu, S. Okamura, & K. Inoue (Eds.), Studies in Generative Grammar and language acquisition: A report on recent trends in lin- guistics (Report of the scientific research program, 1982, supported by the Ministry of Education, Japan, pp. 79-89).

Sakai, H. (1998). Raising asymmetry and improper movement.

Japanese/ Korean Linguistics, 7, 481-497.

Takano, Y. (2003). Nominative object in Japanese complex predicate con- struction: A prolepsis analysis. Natural Language and Linguistic Theory, 21, 779-834.

Tanaka, H. (2002). Raising to object out of CP. Linguistic Inquiry, 33, 637-652.

Uchibori, A. (2001) Raising out of CP and C-T relations. MIT Working Papers in Linguistics, 41, 145-162.