Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue...

32
Zibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun contrasts with English reflexives in several ways. First, zibun is subject oriented. Its antecedent is typically a subject, whereas English reflexives do not have such restriction. 1 Compare (1a) and (1b). (1) a. Takashi i -ga jooshi j -ni zibun i/*j -o suisenshita Takashi-NOM boss-DAT self- ACC recommended ‘Takashi i recommended himself i/*j to his boss j b. John i told Bill j about himself i/j While himself in (1b) can refer to either John or Bill , zibun in (1a) must refer to the subject, Takashi. However, it is also well known that there are exceptions like (2), which are called backward reflexives, as discussed in McCawley (1976), Kuno (1973) and Pesetsky (1995) among others. (2) 2 [Zibun i -ga gan kamo sirenai koto]-ga Kenji i -o nayam-ase-ta [self-NOM cancer may have fact]-NOM Kenji-ACC be.worried-cause-PAST ‘That fact that self i may have cancer worried Kenji i In cases where zibun is inside a sentential subject, it may refer to the surface object. This is a prima-facie violation of the subject orientation. Second, zibun can be a long distance reflexive. As shown in (3), zibun can refer to a long distance subject as well as a local subject. 3 1 However, Iida (1996) presented a number of counter examples. Among them, I do not agree with her judgment on some of her examples, but the sentences below seem to be real problems. (i) Taroo i -wa Hanako j -o zibun i/?j -no jitensha-ni noseta Taroo-TOPHanako-ACC self-GEN bicycle-on put ‘Taroo i put Hanako j on his i /her ?j bicycle’ (ii) Yamada-sensei i -wa Taroo j -ni-totte zibun j/*i -no oya-no yoona sonzai datta Yamada-teacher-TOP Taroo-for self-GEN parent-GEN like existence was ‘For Taroo j , Prof. Yamada i was like his j/*i own parent’ I do not have anything to say about these cases. However, subject orientation seems to hold whenever zibun appears in the argument position of a predicate. 2 (2) is from McCawley (1976) with minor changes. 3 The long distance reading is strongly preferred by most native speakers. In sentences with three clauses as in (i), it is possible to have the intermediate subject be the most preferred antecedent if it is given the right context. (i) Nonaka-shi i -wa [Mori-shusho j -ga [Kato-shi k -ga jibun i/j/?k -o uragitta-koto]-o Nonaka-Mr.-TOP Mori-prime-minister-NOM Kato-Mr.-NOM self-ACC betrayed-fact-COMP okotteiru]-to nobeta be-mad]-COMP stated ‘Mr. Nonaka stated that Prime Minister Moro is mad that Mr. Kato betrayed self’ Sportiche (1986) presents a case where the embedded subject as the antecedent is natural. His example is given in (ii). (ii) John-wa [daremo i -ga zibun i -o semeta-to] omotta John-TOP everyone-NOM self-ACC blamed-COMP thought ‘John thought that everyone i blamed himself i (ii) can be uttered in order to deny a report that John said only Bill blamed himself. As long as the local reading is available if not preferred, it is necessary to be explained. I abstract away the preference in this paper and treat both readings as equally available.

Transcript of Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue...

Page 1: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 1

Zibun As a Residue of Overt A-movement Mitsue Motomura

1. Introduction As is well known, Japanese zibun contrasts with English reflexives in several ways. First, zibun is subject oriented. Its antecedent is typically a subject, whereas English reflexives do not have such restriction.1 Compare (1a) and (1b). (1) a. Takashii-ga jooshij-ni zibuni/*j-o suisenshita

Takashi-NOM boss-DAT self- ACC recommended ‘Takashii recommended himself i/*j to his bossj’

b. Johni told Billj about himselfi/j While himself in (1b) can refer to either John or Bill, zibun in (1a) must refer to the subject, Takashi. However, it is also well known that there are exceptions like (2), which are called backward reflexives, as discussed in McCawley (1976), Kuno (1973) and Pesetsky (1995) among others. (2) 2 [Zibuni-ga gan kamo sirenai koto]-ga Kenjii-o nayam-ase-ta

[self-NOM cancer may have fact]-NOM Kenji-ACC be.worried-cause-PAST ‘That fact that selfi may have cancer worried Kenjii’

In cases where zibun is inside a sentential subject, it may refer to the surface object. This is a prima-facie violation of the subject orientation.

Second, zibun can be a long distance reflexive. As shown in (3), zibun can refer to a long distance subject as well as a local subject.3

1 However, Iida (1996) presented a number of counter examples. Among them, I do not agree with her judgment on some of her examples, but the sentences below seem to be real problems. (i) Tarooi-wa Hanako j-o zibuni /? j-no jitensha-ni noseta Taroo-TOPHanako-ACC self-GEN bicycle-on put ‘Tarooi put Hanako j on his i/her? j bicycle’ (ii) Yamada-senseii-wa Tarooj-ni-totte zibunj/*i-no oya-no yoona sonzai datta Yamada-teacher-TOP Taroo-for self-GEN parent-GEN like existence was ‘For Tarooj, Prof. Yamadai was like his j/*i own parent’ I do not have anything to say about these cases. However, subject orientation seems to hold whenever zibun appears in the argument position of a predicate. 2 (2) is from McCawley (1976) with minor changes. 3 The long distance reading is strongly preferred by most native speakers. In sentences with three clauses as in (i), it is possible to have the intermediate subject be the most preferred antecedent if it is given the right context. (i) Nonaka-shii-wa [Mori-shushoj-ga [Kato-shik-ga jibuni/j/?k-o uragitta-koto]-o Nonaka-Mr.-TOP Mori-prime-minister-NOM Kato-Mr.-NOM self-ACC betrayed-fact-COMP

okotteiru]-to nobeta be-mad]-COMP stated ‘Mr. Nonaka stated that Prime Minister Moro is mad that Mr. Kato betrayed self’

Sportiche (1986) presents a case where the embedded subject as the antecedent is natural. His example is given in (ii). (ii) John-wa [daremo i-ga zibuni-o semeta-to] omotta John-TOP everyone-NOM self-ACC blamed-COMP thought ‘John thought that everyonei blamed himselfi’ (ii) can be uttered in order to deny a report that John said only Bill blamed himself. As long as the local reading is available if not preferred, it is necessary to be explained. I abstract away the preference in this paper and treat both readings as equally available.

Page 2: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 2

(3) Takashii-ga [ Kenjij-ga zibuni/j-o suisenshita-to ] omotta Takashi-NOM Kenji-NOM self-ACC recommended-COMP thought ‘Takashii thought that Kenjij recommended selfi/j’

Third, a nominative reflexive is possible in Japanese unlike English. Zibun in (4a) can get a nominative Case in contrast to she-self in (4b). (4) a. Marikoi-ga [ zibun-gai ichiban moteru-to ] shinjiteiru.

Mariko-NOM self- NOM best be-popular-COMP ] believe ‘Marikoi believes that shei is popular the most’

b. *Mary believes that she-self is the most popular. This paper argues that zibun is a reflexive, created as a result of overt A-movement. This assumption gives us a unified account for the properties mentioned above. In section 2, I show that zibun exhibits properties typical of reflexives. In section 3, I present the assumptions made by Hornstein (1999, 2000), which my account of zibun is based on. The main proposal follows in section 4, where I account for the properties of zibun presented above. Here, I extend the analysis of long-distance reflexives to object control constructions. Section 5 discusses two issues that are relevant to the analysis of zibun,. One concerns the non-complementarity of reflexives and pronouns in Japanese and the other, the treatment of two other reflexives in Japanese, namely, zibun-zisin and kare-zisin. As concluding remarks, I will present problems for the previous accounts of reflexives, and show that the account presented in this paper does not raise those problems. 2. The reflexive properties of zibun Sportiche (1986) argues that zibun can be either a reflexive or a bound pronoun. His argument crucially depends on the fact that zibun can take a local antecedent as shown in (1), thus a reflexive, and Fukui’s (1984) observation that zibun allows split antecedents, thus, a bound pronoun. Sportiche is correct in his observation that zibun covers both slots of reflexives and bound pronouns. I will show below that zibun exhibits the properties of reflexives by applying the diagnostics discussed in Hornstein (2000). 4

First, (5) shows that zibun requires a c-commanding antecedent. (5) [Takashii-no chichioya]j-ga zibunj/*i-o hometa

Takashi-GEN father-NOM self-ACC praised ‘Takashii’s fatherj praised self j/*i’ Second, (6) shows that the split antecedent is not allowed, contrary to Fukui’s

observation. (6) Takashii-ga Marikoj-ni [ Kenjik-ga zibuni/*j/k/*i+j/*i+k-o suisenshita -to ] tsugeta

Takashi-NOM Mariko-DAT Kenji-NOM self-pl.-ACC recommended-COMP reported ‘Takashii reported to Marikoj that Kenjik recommended selfi/*j/k/*i+j/*i+k’

The antecedent of zibun can be either Takashi or Kenji, but not both Takashi and Mariko or Takashi and Kenji, as the indices indicate. The story is not so simple, however. Fukui (1984) argues that zibun allows split antecedents in (7). (7) Takashii-ga Marikoj-ni [Kenjik-ga zibun-tachii+j-o suisenshita–to ] tsugeta

Takashi-NOM Mariko-to Kenji-NOM self-pl.-ACC recommended-COMP reported ‘Takashii reported to Marikoj that Kenjik recommended selvesi+j’

4 Some of the diagnostics do not work on zibun due to the facts that zibun can be a long distance reflexive and that zibun does not exhibit person, gender, and number features. See Horstein (2000) for the discussion of these diagnostics.

Page 3: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 3

He explains that -tachi is added to zibun for stylistic naturalness. However, -tachi seems to do more than just adding naturalness since it is impossible to get the split antecedent reading in (6) without -tachi. Now, compare (6) with (8). (8) Sono onna-tachii-wa zibuni-o hometa

the woman-pl.-TOP self-ACC praised ‘The women praised themselves’

In (8), zibun can refer to the plural antecedent without the plural marker, -tachi on zibun. This suggests that -tachi is not a number agreement marker between the antecedent and zibun. One might argue that (8) seems to have only the distributive reading without -tachi, and that -tachi is required in order to obtain the collective reading. However, English requires number agreement between the antecedent and the reflexive regardless of the collective/distributive readings of the sentence as the English translation in (8) indicates.

Given a definition of -tachi in Martin (1988; 145), Muromatsu (1998; 91) assumes that -tachi is a classifier rather than a plural marker. She observes that predicate nominals resist the plural classifier, -tachi, when the subject is plural as in (9). (9) 5 Jiro-to Taro-wa gaka (*-tachi) dearu

Jiro-and Taro-TOP painter (-pl.) be ‘Jiro and Taro are painters’ Martin defines that ‘the meaning of [-tachi] is not plurality of the noun itself, but rather the reference is to a collective that includes -or centers on- the noun.’ In other words, -tachi adds other people to the referent of the head noun. This may suggest that in (7), although zibun itself refers to its antecedent (Takashi), -tachi adds Mariko as others to the antecedent of zibun, thereby allowing Mariko in addition to Takashi as the antecedent of zibun. Aikawa (1993; 31) also reports Hoji’s (1990) skepticism about the status of -tachi as a pure plural suffix. As she notes, it is very common to attach -tachi to names, as in Kenji-tachi, meaning Kenji and others. This fact is consistent with Martin’s claim. Given these observations, I conclude that -tachi is not a grammatical agreement suffix, but a type of classifiers, therefore, zibun does not allow split antecedents.

Third, (10) should have only the sloppy reading if zibun is an element created by movement. (10) Takashii-ga zibuni-o home, Kenji-mo soo-shita

Takashi-NOM self-ACC praised, Kenji-too so-did ‘Takashi praised himself, and Kenji did so too’

The judgment is not so straightforward. Some of my informants allow the strict reading although everyone agrees that the sloppy reading is strongly preferred. This seems true in English as well. Thus, Sag (1976) reports some of his informants allow the strict reading for anaphors in VP ellipsis constructions. The judgment seems clearer with a long distance reflexive as in (11). (11) Takashii-ga [zibuni-ga ichibanda-to] shuchooshi, Kenji-mo soo-shita

Takashi-NOM self-NOM the.best-COMP claim, Kenji-too so-did ‘Takashi claimed that self is the best, and Kenji did so too’ I will simply take the judgment of those who accept the sloppy reading as evidence for the reflexive status of zibun here, and leave those who accept the strict reading aside.

Fourth, zibun only allows the de se reading unlike pronouns as the contrast in (12) shows. 5 (9) is from Muromatsu (1998; 91).

Page 4: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 4

(12) Marikoi-wa [zibuni/kanojo?i/j-ga Kenji-ni erabareru-to ] shinjiteiru . Mariko-TOP self/she-NOM Kenji-by is-chosen-COMP believe ‘Marikoi believes that selfi/she?i/j will be chosen by Kenji’

Suppose Mariko is looking at a photo of herself, but she doesn’t know the person in that photo is Mariko herself for some reason, say because the photo was shot by a candid camera and was manipulated in order to look better. In this situation, the sentence with kanojo is true, but false with zibun.

Finally, it is known that a pronoun in (13a) is ambiguous presenting two meanings as given (13b) and (13c). (13) a. Only John thinks that the teacher praised him

b. [Only x: x=John] x thinks that the teacher praised x c. [Only x: x=John] x thinks that the teacher praised John

Thus, in a situation where John, Mary, and Bill all think that the teacher praised John, (13b) is true but (13c) is false. If zibun is a pronoun, it would show the same ambiguity as the English pronoun. The prediction is not born out, however. Compare (14a) and (14b). (14) a. Takashi-dakei-ga [sensei-ga zibuni-o hometa-to] omotteiru

Takashi-only-NOM teacher-NOM self-ACC praised-COMP think ‘[Only Takashi]i thinks that the teacher praised selfi’

b. Takashii-dake-ga [sensei-ga karei-o hometa-to] omotteiru Takashi-only-NOM teacher-NOM him-ACC praised-COMP think ‘Only [Takashi]i thinks that the teacher praised himi’

(14a) only has the meaning of (13b) and (14b) only shows the meaning of (13c).6 This is expected if zibun is a reflexive since it requires a c-commanding antecedent, but not if it is a pronoun.

The properties presented above suggest that zibun is a reflexive rather than a pronoun. Following Lebeaux (1985), Hornstein (2000) argues that behaviors of a locally bound reflexive and those of obligatory control PROs (henceforth, OC PRO) are identical. He concludes that the English reflexives are residue of movement on a par with his analysis of OC PRO.7 In the next section, I introduce Hornstein’s assumptions about the English reflexives. In section 3, I will argue that zibun is also a residue of movement. 3 Background assumptions Hornstein (1999, 2000) argues that OC PRO should be treated as an NP-trace.8 9 In GB, the theoretical motivation for distinguishing OC PRO from NP-traces was the biuniqueness condition (15c) of the Theta-Criterion. (15) Theta Criterion a. every argument must receive a theta-role

b. every theta-role must be assigned c. every argument has at most one theta-role.

6 It is well known that Japanese pronouns cannot be bound. See Noguchi (1997), and the reference cited there. See also the discussion in section 5.1 7 See Hornstein (1999, 2000) for the complete analysis of OC PRO and Hornstein (2000) for that of reflexive as well as the diagnostics adopted here. 8 Hornstein treats PRO in nonobligatory control configuration as a null pronoun, pro , unlike the standard assumption that English does not possess pro . 9 Lidz and Idsardi (1998) also proposed a similar idea.

Page 5: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 5

Under the framework of the Minimalist Program (henceforth MP), (15a) and (15b) are required by the principle of Full Interpretation. Given that the MP abandoned the level of D-structure, (15c) lost its motivation. Hornstein thus argues that (14c) should be abandoned as well. This opens the possibility that NPs get more than one θ-role. In other words, NPs can move from one θ-position to another. Since movement is triggered by the Principle of Greed, θ-roles should be treated as a formal feature on a par with Case, φ-features and EPP.10

The immediate question is; why is (16) not allowed? This approach to θ-roles should allow (16), where John gets both the internal and external θ-roles of the verb, contrary to fact. (16) *John likes. (meaning: John likes himself). Hornstein attributes the ungrammaticality of (16) to the inability of a single NP to check more than one Case. Thus, (16) is ungrammatical because John cannot check both nominative and accusative Cases at the same time. However, the derivation is saved by inserting self whose sole job is to check Case. He assumes that self is a grammatical formative inserted to save a derivation that would otherwise crash. The derivation for the grammatical sentence (17a) is given in (17b).11 (17) a. John likes himself

b. [TP John T0 [vP ssseeelllfff [v’ John v0 [VP likes John+self]]]] [θθ,ünom] [üacc] [θθ,nom] [θ,nom][acc]

c. [TP John T0 [vP ssseeelllfff [v’ John v0 [VP likes himself]]]] The derivation starts with merging John and self. This complex merges with the verb thereby obtaining the internal θ-role. John moves to Spec.vP, where it gets the external θ-role, and then, to Spec.TP, where it checks the nominative Case, φ-features and EPP. Self moves at LF to the outer Spec.vP to check the accusative Case. Finally, him is inserted in (17c) to support the bound morpheme, self. Hornstein (1999, 2000) thus concludes that English reflexives should also be treated as NP-traces. (18) is the summary of his assumptions for English reflexives. (18) a. DPs can have multiple theta-roles

b. Reflexives are not lexical items, but grammatical formatives introduced to save the derivation, which would otherwise crash

c. Self has only Case-features d. A pronominal element is inserted before Spell-Out to support the morphological

deficiency of self e. Formal features have to be checked by the same element at once 12

In the next section, I will present an account of zibun parallel to English reflexives based on the assumptions in (18). 4 Proposal 13

10 This idea has already been proposed by Lasnik (1995), Boškoviæ (1994), and Boškoviæ and Takahashi (1998). 11 Throughout this paper, double-lined elements are unpronounced copies, and engraved elements indicate its LF position. ü indicates the feature is checked. 12 This assumption is relevant to the lack of nominative reflexives in English, discussed in Section 4.4. 13 The analysis of zibun here is basically the same as that of a Chinese reflexive, ziji, proposed by Chuang (1997). She does not extend her analysis to zibun in Japanese, however, because of the absence of intervention effect observed in Chinese. (i) illustrates the effect in Chinese. (i) Zhangsani renwei [ Malij xiangzin ziji i/j ].

Page 6: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 6

4.1 Subject orientation The property of subject orientation in (1) repeated as (19a) is derived as shown in (19b) as a consequence of the preference for Merge over Move (Chomsky, 1995). (19) a. Takashii-ga jooshij-ni zibuni/*j-o suisenshita.

Takashi-NOM boss-DAT self- ACC recommended ‘Takashii recommended himself i/*j to his bossj’

b.14 TP

7Takashi T’ [θθ,ünom]

vP T 6 jjjoooooossshhhiii v’ ta [üdat]

5 zzziiibbbuuunnn v’ [üacc]

4 Takashi v’ [θθ, nom] VP v

3 è jooshi V’

[θ, dat] 2 è Takashi+zibun V

[θ, nom][acc] suisensi

1Takashi ↔ zibun [nom] [acc] The derivation goes as follows. Takashi and zibun merge (1). The complex, Takashi+zibun, merges with the verb, suisenshi, and gets the internal θ-role (2). Jooshi is inserted to Spec.VP and gets the Goal θ-role (3). Takashi moves to Spec.TP through the inner Spec.vP, where it

Zhangsan think Mali believe self Zhangsani thought that Malij believed selfi/j’

(ii) Zhangsani renwei [ wo j xiangzin ziji*i/j ]. Zhangsan think I believe self ‘Zhangsani thought that Ij believed self*i/j’

In (i), ziji can refer to either the matrix subject or the embedded one, but in (ii) the antecedent of ziji must be the embedded subject. Chuang correlates the intervention effect with lack of subject-verb agreement in Chinese/Korean type languages. Japanese is on a par with Chinese/Korean with respect to subject-verb agreement, yet does not exhibit the intervention effect according to her Japanese informants. Thus, she concludes that zibun in Japanese is a pronoun. It is not clear, however, if Japanese does not exhibit this effect. Although most of my informants accept the long distance reading for the Japanese equivalent of (ii), the preference is reversed. Thus, eight out of eleven people prefer the local reading for (ii). On the other hand, all of the eight people prefer the long distance reading for the Japanese equivalent of (i). It is not clear why the preference reversed in Japanese. However, as long as both readings are available in Japanese, I believe that this fact about the reversed preference does not threaten my account. I leave the blocking effect open now. Cole, Hermon, and Lee (2001) proposed that the blocking effect of Chinese reflexives is partially attributed to a discourse restriction that the antecedent of the reflexive must be the center of deixis in the sentence, which is called PIVOT. See there and the reference cited. 14 è indicates the insertion of a D/NP.

Page 7: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 7

gets the external θ-role (47). Meanwhile, zibun as well as jooshi moves at LF to check Case (67). This analysis adopts the assumptions (a-c) of (18). The difference between English self and Japanese zibun is that while self is a bound morpheme, hence requires an insertion of him for the morphological reason (18d), zibun is a free morpheme, hence, does not require any element to support it. Therefore, (18d) is irrelevant for zibun. 15 4.2 Backward Reflexives Backward reflexives have been a puzzle since they do not obey the subject orientation at least on its surface. There are at least three patterns in which backward reflexives appear. First, it has been noticed in the literature that backward reflexives appear in constructions that involve psychological state predicates, more specifically, in object experiencer constructions (henceforth OEPs),16 and that OECs involve causative morpheme, sase, in Japanese as shown in (2) repeated in (20). (20) [Zibuni-ga gan kamo sirenai koto]-ga Kenjii-o nayam-ase-ta.

[self-NOM cancer may have fact]-NOM Kenji-ACC be.worried-cause-PAST ‘That fact that selfi may have cancer worried Kenjii’

However, McCawley (1976) notes Kuroda’s observation that backward reflexives also appear in sentences which do not involve sase. (21) is an example of such sentences. (21) [Zibuni-ga gan kamo sirenai koto]-ga Kenjii-o uchinomeshita

[self-NOM cancer may have fact]-NOM Kenji-ACC bowled-over ‘That fact that selfi may have cancer bowled over Kenjii’

Moreover, backward reflexives seem to appear in causative constructions that are not OECs as in (22).17 (22) [ Zibuni/*j/*i+j-ga gan kamoshirenai koto ]-ga Kenjii-ni oyaj-o itawar-ase-ta self-NOM cancer may-have fact -NOM Kenji-DAT parent-ACC care-cause-past

‘The fact that self may have cancer caused Kenji to care for (his) parent’

15 Norvin Richards pointed out to me the case of multiple subject constructions as in (i). (i) Takashii-ga imootoj-ga zibuni/j-o suisenshita Takashi-NOM sister-NOM self-ACC recommended ‘Takashii’s sisterj recommended selfi/j’ As Ura (1996) observed, zibun is ambiguous in (i) in that it can refer to either the possessor or the possessee. In the analysis proposed here, the possessee reading can be derived from (ii) by the possessor raising proposed by Ura (1996). (ii) [Takashii-no imooto]j-ga zibunj/*i-o suis enshita Takashi-GEN sister-NOM self-ACC recommended ‘Takashii’s siserj recommended selfj/*i’ However, the possessor reading is a potential problem for the analysis proposed above. Two possibilities were suggested by Norbert Hornstein. First is to assume that Takashi in the DP, Takashi-no imooto, sideward moves to form Takashi+zibun and inserted to the internal è-position of the verb, and imooto alone merges into the external è-position. Takashi+zibun overtly moves to the outer Spec.vP, checking the accusative Case, and then, moves to the outer Spec.TP over imooto in the inner Spec.TP. The other is to assume that Takashi starts with zibun, and sideward moves to form a DP, Takashi-no imooto. This DP, inserted into the external è-position, and moves to Spec.TP. Finally, Takashi in this DP undergoes the possessor raising. The question is why genitive Case prevents the possessor raising. I leave this issue for future research. 16 I use the term, experiencer, descriptively, not as a technical term, used in the discussion of thematic hierarchy, unless so stated. 17 (22) is suggested to me by Hirohisa Kiguchi as evidence that zibun involves movement. I will come back to this point below.

Page 8: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 8

I will treat (20) on a par with (21). That is, I assume that the object experiencer verb, nayam-ase, is a simple transitive verb unlike its appearance as a complex morpheme. On the other hand, I assume (22), the causative construction that does not express psychological state, involves a complex structure. I will come back to this issue after I present my analysis of backward reflexives below.

Before doing that, however, let us examine if zibun in (20)-(22) exhibits the properties of reflexives as we saw in section 2. First, the sentences in (23)-(25) have only one meaning. (23) [Zibuni-ga gan kamo sirenai koto]-ga Kenji-dakei-o nayam-ase-ta.

self-NOM cancer may have fact-NOM Kenji-only-ACC be.worried-cause-past ‘That fact that selfi may have cancer worried only Kenjii’

(24) [Zibuni-ga gan kamo sirenai koto]-ga Kenji-dakei-o uchinomeshi-ta. self-NOM cancer may have fact-NOM Kenji-only-ACC bowled-over ‘That fact that selfi may have cancer bowled over only Kenjii’

(25) [ Zibuni-ga gan kamoshirenai-koto ]-ga Kenji-dakei-ni oyak-o itawar-ase-ta self-NOM cancer may-have-fact -NOM Kenji-only-DAT parent-ACC care-caused

‘The fact that selfi may have cancer caused only Kenjii to care for (his) parent’ Thus, in a situation where, Kenji, Takashi and Mariko all got worried by the fact that Kenji may have cancer, (23) is true, but not in a situation where in addition to Kenji got worried by the fact that Kenji may have cancer, Takashi also got worried by the fact that Takashi may have cancer. The same result obtains in (24) and (25). This is identical to the situation where zibun is c-commanded by its antecedent. Second, zibun does not allow split antecedent as shown in (22).18 Finally, only de se reading is possible in (20)-(22). Imagine a situation where Kenji heard his doctor talking about a patient who is about Kenji’s age, and may have cancer. Kenji did not know his doctor was talking about Kenji himself. (20)-(22) cannot describe this situation. (20)-(22) can only describe the situation where Kenji knows that the doctor was talking about him, thus the de se. Given these results, it does not seem unreasonable to conclude that backward reflexives are also derived by movement.

In order to account for the backward reflexive constructions, I adopt the analysis of backward binding control constructions proposed by Kiguchi (forthcoming). Let us see first how his system works. The relevant data is given (26). (26) [PRO1 to have to feed himself/*oneself] annoys John1. Chomsky (1986) observes that PRO in (31) must be obligatorily controlled by John. Kiguchi accounts for this phenomenon by adopting Hornstein’s (1999, 2000) claim that OC PROs are residue of movement, and sideward movement proposed by Nunes (1995). Nunes notes when the computational system builds a sentence like ‘the man kissed Mary,’ two subtrees are constructed in parallel as shown in (27).

18 The diagnostic of split antecedence cannot apply to (21) since the verb is a simple transitive, and does not have more than one argument as the antecedent of zibun. I could not find a ditransitive verb that can select a proposition as its subject and two other arguments that can serve as antecedents of zibun. For (20), one might think it possible to construct a sentence with two possible antecedents as in (i). (i) [ Zibuni/*j/*i+j-ga ototteiru-koto]-ga Kenjii-ni Takashij-o nayam-ase-ta self-NOM inferior-fact -NOM Kenji-DAT Takashi-ACC be.worried-cause-past

‘The fact that self is inferior caused Kenji to worry Takashi’ However, we should treat (i) as another case of (22). Although (i) involves the same predicate, nayam-ase, as (20), it involves a hidden internal argument of the verb, nayam, which is different from the external argument of sase. (20) is a case of OECs, where the internal argument and the external argument are identical.

Page 9: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 9

(27)

the man kissed Mary He claims that it is possible to move an element from one subtree to another at this point since two subtrees are not merged yet. This type of movement is called sideward movement. He applies this idea to a parasitic gap construction in (28a) as illustrated below. (28) a. Which paper did you file without reading? b. c. without file PRO

reading which paperi which paperi The derivation starts building the adjunct clause with the wh-element in it shown in (28b). When it starts the main clause, the two subtrees are built up in parallel as illustrated in (28b) and (28c). When it starts the main clause, which paper in the adjunct clause is copied, and the copy sideward-moves to the internal è-position of the main verb, file. Eventually, the two subtrees are merged and which paper in adjunct is deleted.

Now, going back to Kiguchi’s analysis, he postulates the following derivation for (26). (29) a. [John to have to feed himself] b. [John1 to have to feed himself] [annoy [John1]] c. [[John1 to have to feed himself] [annoy [John1]]] d. [[John1 to have to feed himself] [annoy [John1]]] John starts as an argument internal to the sentential subject in (29a). The derivation starts building another subtree, which will eventually become the main clause. John sideward moves to the internal θ-position of the main verb (29b). Then, the two sub-trees merge (29c). Finally, the copy of John inside the sentential subject gets deleted (29d).

The analysis in (29) can be carried on to the analysis of the backward reflexive constructions in (20) repeated in (30a) as illustrated below. (30) a. [Zibuni-ga gan kamo sirenai koto]-ga Kenjii-o nayam-ase-ta.

[self-NOM cancer may have fact]-NOM Kenji-ACC be.worried-cause-PAST ‘That fact that selfi may have cancer worried Kenjii’

Page 10: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 10

b.19 CP

TP C

3 Kenji+zibun T’ koto [θ, dat][ünom]

vP T

2 è Kenji+zibun v’ [θ, dat][nom]

VP v

gan kamoshirenai 1Kenji ↔ zibun

c. TP

7 CP T’ [θ,ünom]

vP T

6 KKKeeennnjjjiii v’ ta [θθ,üacc] è 5 CP v’ [θ, nom]

� Kenji V [θθ, acc]

nayam-ase The derivation starts with the subject CP (30b). Next, the derivation builds up the main clause keeping the CP parallel in its work space. Kenji sideward moves from the CP in (30b) to the internal è-position of the main verb (�). Then, the CP merges with the main verb as its external argument (5). The question is why the CP does not merge with the verb before Kenji moves in step �. Moving Kenji before merging CP violates Merge over Move. Following Kiguchi, I assume that the selectional restriction of the verb ‘nayam-ase’ prevents it from selecting a proposition as its complement.

Our analysis predicts that if there is more than one DP in the numeration, the antecedent of zibun is always the highest DP in the tree. Sells (1987, 474) suggests as a possibility that “a more appropriate syntactic condition for Japanese (zibun) would be ‘highest available argument”

19 I treat koto as a complementizer rather than a noun, following Uchibori (1996, in progress) and Watanabe (1996). However, treating koto as a noun seems to work as well. It has been noted that Japanese does not obey the complex NP island constraint as pointed out by Hasegawa (1985) among others. If so, movement out of the DP is possible in Japanese. For instance, scrambling out of a relative clause is possible as shown in (i). (i) Honi-o [CP Kenji-wa [ ej ei ni-satsu yonda] gakuseij-ni atta]

book-ACC Kenji-TOP 2-NQ read student-DAT met ‘(Lit.) Books, Kenji met a student that read two =(Kenji met a student that read two books)’

Note that the floating quantifier in the relative clause indicates the scrambled element moves from that position. The exact mechanism of how movement takes place from inside a DP needs to be spelled out, however.

Page 11: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 11

following Bickerton (1986). (22), repeated as (31), shows that his suggestion is correct. Now, compare (31) with (32).20 (31) [ Zibuni/*j/*i+j-ga gan kamoshirenai koto ]-ga Kenjii-ni oyaj-o itawar-ase-ta self-NOM cancer may-have fact -NOM Kenji-DAT parent-ACC care-cause-past

‘The fact that self may have cancer caused Kenji to care for (his) parent’ (32) [ proi/j gan kamoshirenai-koto ]-ga Kenjii-ni oyaj-o itawar-ase-ta

self-NOM cancer may-have-fact -NOM Kenji-DAT parent-ACC care-CAUSE-PAST ‘The fact that pro may have cancer caused Kenji to care for (his) parent’

When zibun appears in the sentential subject in (31), it has to refer to Kenji rather than oya. On the other hand, when pro appears in (32), the sentence becomes ambiguous in that pro can refer to either Kenji or oya. This follows if zibun, unlike pro, is the residue of overt movement.

According to Chomsky (1998, 1999), the computational system has multiple accesses to a lexical array, and this reduces computational complexity. Each sub-array has to be exhausted and all of the uninterpretable features in the sub-array must disappear before moving on to the next sub-array. Chomsky argues that CP and vP are phases that correspond to these sub-arrays. Suppose Kenji is in the sub-array corresponding to the sentential subject, and another DP, namely oya, is in the sub-array corresponding to the main clause. (33) illustrates the derivation of (31).21 (33) a. CP

TP C

3 Kenji+zibun T’ koto [θ, dat][ünom]

vP T

2 è Kenji+zibun v’ [θ, dat][nom]

VP v

gan kamoshirenai 1Kenji ↔ zibun

20 The contrast between (31) and (32) was suggested by Hirohisa Kiguchi as evidence that zibun involves movement. 21 I ignore the vP phase here since it does not seem to matter if we assume vP1 and vP2 compose one phase. However, It does matter if we do not assume that, or if we assume the causative morpheme, sase, heads a lexical VP, rather than a light verb, vP. If it heads a VP, the sentential subject CP is not introduced in the derivation when vP1 is built. Then, it is impossible to move Kenji to a è-position in vP1 phase simply because Kenji is not available in the derivation. Kiguchi, facing the same problem, concludes that vP is not a phase. I will leave this issue open.

Page 12: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 12

b. TP

� CP T’ [θ,ünom]

vP2 T

8 KKKeeennnjjjiii v2’ ta [θθ,üdat] è7 CP v2’ [θ, nom]

vP1 v2

6 oooyyyaaa v1’ ase [θθ,üacc]

5 Kenji v’ [θθ, dat]

VP v1

� è oya V [θθ, acc]

itawar The derivation starts with the sentential subject CP and Kenji starts in this CP as before. Then, the derivation starts building the matrix clause. The main verb, itawar, is a transitive verb, thus needs two arguments. Since a DP, oya, is available in the sub-array corresponding to the matrix CP phase, oya is inserted to the internal è-position (�). Kenji then sideward moves to the external θ-position of itawar (5), and the CP built in (33a) merges into the external θ-position of the causative, (s)ase (7). The CP moves into Spec.TP (�), checking the nominative Case and EPP, while Kenji and oya moves at LF to check their Case. If other DPs are available in the numeration for insertion, insertion takes precedence over the sideward movement due to the condition of Merge over Move. Thus, Sell’s suggestion is derived.

So far, I treat the causative predicate of OECs as a simple transitive unlike its appearance while the causative predicate of non-OECs as complex without argument. However, this raises a serious question of why the causative morpheme, sase, starts its derivational life already attached to another morpheme in one case, while it starts independently in another.

Alternatively, it may be possible to postulate the same complex structure for both types of causatives given the multiple è-role approach. In fact, this line of argument was proposed by Kuroda (1965). He postulates (34b) as the underlying structure of (34a).22 (34) a. Ongaku-ga John-o tanoshim-ase-ru music-NOM John-ACC be.amused.at-CAUSE-PRESENT ‘Music amuses John’ b. Ongaku-ga (John-ga ongaku-o tanoshim) sase-ru The underlying structure has two ongaku ‘music,’ one for the Theme of the embedded predicate, and the other for the Agent/Causer of the matrix predicate, sase.23 Then, the two sentences in

22 (34) is from Kuroda (1965) with minor modification. 23 I do not distinguish Agent and Causer following Baker (1997) with respect to its structural position.

Page 13: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 13

question repeated in (35) and (36) would have very similar derivations shown in (37) and (38) respectively, with its argument structure schematized in (39). (35) [Zibuni-ga gan kamo sirenai koto]-ga Kenjii-o nayam-ase-ta

[self-NOM cancer may have fact]-NOM Kenji-ACC be.worried-cause-PAST ‘That fact that selfi may have cancer worried Kenjii’

(36) [ Zibuni/*j/*i+j-ga gan kamoshirenai koto ]-ga Kenjii-ni oyaj-o itawar-ase-ta self-NOM cancer may-have fact -NOM Kenji-DAT parent-ACC care-cause-past

‘The fact that self may have cancer caused Kenji to care for (his) parent’ (37) [TP [CP Zibuni-ga gan kamo sirenai koto]-ga T0 [vP2 KKKeeennnjjjiii---ooo [v2’ CP [vP1 Kenji

[VP CP nayam]]-ase]]-ta] (38) [TP [CP Zibuni-ga gan kamo sirenai koto]-ga T0 [vP2 KKKeeennnjjjiii---nnniii [v2’ CP [vP1 oooyyyaaa---ooo [vP1 Kenji

[VP oya itawar]]]-ase]]-ta] (39) [vP2 Causer [vP1 Experiencer [VP Theme nayam ]]-ase]24 This approach also raises a couple of questions, however. First, it is not clear why the causative morpheme sase checks the accusative Case in (37) and the dative Case in (38). Second, we need to explain why it must be the case in (37) that one argument takes both the internal è-role of the main predicate (Theme) and the external è-role of sase (Causer) but no such restriction holds in (38). McCawley (1976) made an interesting observation in the following paradigm. (40) a. * Hiroshi-wa Michiko-o/ni nayan-da Hiroshi-TOP Michiko-ACC/DAT be.worried-PAST ‘Hiroshi was worried about Michiko’ b. Hiroshi-wa Michiko-no kokorogawari-ni nayan-da Hiroshi-TOP Michiko-GEN change.of.mind-DAT was.worried-PAST ‘Hiroshi was worried about Michiko’s change of mind’ (41) a. Michiko-wa Hiroshi-o nayam-ase-ta Michiko-TOP Hiroshi-ACC be.worried-CAUSE-PAST ‘Michiko worried Hiroshi’ b. Michiko-no kokorogawari-ga Hiroshi-o nayam-ase-ta Michiko-GEN change of mind-NOM Hiroshi-ACC was.worried-CAUSE-PAST ‘Michiko’s change of mind worried Hiroshi’ (40) and (41) are examples of subject experiencer constructions (henceforth, SECs), and OECs respectively. McCawley noticed that SECs has a semantic restriction on its choice of the internal argument. It cannot be a human DP as shown in (40a) but must be a non-human expression as in (40b). On the other hand, OECs allow both human and non-human expressions as its external argument in (41). If the OECs is derived from SECs as schematized in (42), the contrast in (40)-(41) would be surprising. (42) a. SECs [vP1 Experiencer [VP Theme nayam ]] b. OECs [vP2 Causer [vP1 Experiencer [VP Theme nayam ]]-ase] The question is: why does the semantic restriction on the Theme argument in (42a) disappear in (42b)? If we assume that the two constructions are unrelated as schematized in (43), the contrast in (40)-(41) is simply a matter of lexical choice. Since the Causer of OECs is unrelated to the Theme of SECs, we do not need to explain the contrast.25

24 Follow Baker (1997) and Dowty (1991), I assume that the subject of the OECs is a Causer, which is a prototypical agent, thus occupying the external è-position in the argument structure. 25 I am not claiming here that the object of OECs is Theme as opposed to say, Goal. I simply follow Baker (1997) and Dowty (1991) in that the internal argument of OECs is a prototypical theme.

Page 14: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 14

(43) a. SECs [vP1 Experiencer [VP Theme nayam ]] b. OECs [vP1 Causer [VP Theme nayam-ase]]

In summary, each of the alternatives presented above has problems of its own. Needless to say, thorough investigation is required to choose one over the other, and I leave it for future research. 4.3 Long-distance reflexives The long-distance dependency can also be explained by the notion of phases discussed in the previous section. In the following, I first account for the ambiguity of complex sentences with zibun. I then extend the analysis to the case of object control constructions in Japanese. 4.3.1 Ambiguity of complex sentences with zibun (3) is the relevant sentence, repeated as (44). (44) [CP2 Takashii-ga [CP1 Kenjij-ga zibuni/j-o suisenshita-to ] omotta

Takashi-NOM Kenji-NOM self- ACC recommended-COMP thought ‘Takashii thought that Kenjij recommended selfi/j’ This sentence is ambiguous in that zibun can refer to either the the embedded subject,

Kenji, or the matrix Takashi. Let us start with the embedded subject reading illustrated in (45).26 (45) zibun = Kenji a. CP1: {Kenji, suisenshi, ta (PAST), to (COMP), …} CP2: {Takashi, omou, ta, …}

26 I ignore the vP phase here again, since it does not matter.

Page 15: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 15

b. CP2 TP 7 Takashi T’ [θ,ünom] vP T 6è Takashi v’ ta [θ, nom] VP v

CP1 V

TP C omot

5 Kenji T’ to [θθ,ünom]

vP T

4 zzziiibbbuuunnn v’ ta [üacc]

3 Kenji v’ [θθ, nom]

VP v

2 è Kenji+zibun V [θ, nom][acc]

1Kenji ↔ zibun suisensi As shown in (45a), the sub-array corresponding to CP1 includes only one DP, Kenji. The derivation starts with merging Kenji and zibun in CP1 (1). This complex merges with the verb, suisenshi (2). Since no other DP is available in CP1, Kenji moves to Spec.vP, where it gets the Agent θ-role (3), while zibun moves at LF to the outer Spec.vP to check the accusative Case (4). Kenji continues to move to Spec.TP and checks nominative Case as well as other formal features (5). C0 is inserted. Since the sub-array of CP1 is exhausted, and all the uninterpretable features are checked, the derivation now starts the next phase, CP2. Another DP, Takeshi, is available, and is inserted to take the matrix θ-role (6). The derivation continues accordingly and converges.

The derivation for the other reading in which zibun refers to the matrix subject is illustrated in (46). (46) zibun = Takashi a. CP1: {Takashi, Kenji, suisenshi, ta (PAST), to (COMP), …} CP2: {omou, ta, …}

Page 16: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 16

b. TP 8 Takashi T’ [θθ,ünom] vP T 7 Takashi v’ ta [θθ, nom] VP v CP1 V TP C omot

6 Takashi TP to [θ, nom]

5 Kenji T’ [θ,ünom]

vP T

4 zzziiibbbuuunnn v’ ta [üacc]

3è Kenji v’ [θ, nom]

VP v

2 è Takashi+zibun V [θ, nom][acc]

1Takashi ↔ zibun suisensi This time, the sub-array corresponding to CP1 includes both Takashi and Kenji. The derivation starts as before with merging Takashi and zibun (1), then the complex merging with the verb thereby obtaining the internal θ-role (2). Since another DP, Kenji, is available in the array, the condition of Merge over Move forces Kenji to be inserted to the Spec.vP, where it obtains the external θ-role (3). Kenji moves to the Spec.TP to check the EPP and the nominative Case features (5), and zibun moves tat LF to check the accusative Case (4). Meanwhile, Takashi moves up to the matrix Spec.TP through adjunction to the lower TP (6, 7, 8). Adjunction of Takashi into the embedded TP in step6, which I assume to be an instance of scrambling, is crucial because this movement obviates the violation of the Minimal Link Condition (henceforth MLC)/Shortest Movement Condition (henceforth SMC). The definition of MLC/SMC is given in (47). Closeness is defined in terms of c-command and equidistance following Chomsky (1995, 296). (48) defines equidistance, and the definitions of the relevant domains are given in (49) (Chomsky, 1995; 178).27 It will become clear in the next subsection

27 The notions of domination and containment are defined as; (i) a category á dominates â if every segment of á dominates â (ii) a category á contains â if some segment of á dominates â

Page 17: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 17

why I adopt the definitions of the domain and the minimal domain in (49). (47) á can raise to target K only if there is no legitimate operation Move â targeting K, where

â is closer to K. (48) If á, â are in the same minimal domain, they are equidistant from ã. (Chomsky,

1995;184) (49) a. Domain of a head α

The set of nodes contained in Max(α) that are distinct and do not contain α, where Max (α) is the least full-category maximal projection dominating α

b. Minimal domain of α The smallest subset K of S (= the domain of α) such that for any γ∈S, some β∈K reflexively dominates γ

Given these definitions, Takashi and Kenji in the lower Spec.TP are equidistant to the target, matrix Spec.vP, because they are in the same minimal domain of the head of the embedded TP. Therefore, movement of Takashi from that position to the target does not violate the MLC/SMC.

I assume that scrambling into the embedded TP creates an escape hatch thereby allowing further movement. This is not unreasonable since Japanese is one of those languages that allows scrambling, and a standard assumption is that scrambling takes place by adjunction to TP (Saito,1992, among others). Takano (1998) argues that scrambling may become obligatory when a derivation would crash without scrambling. In normal cases, scrambling is optional. However, it appears to become obligatory since no other derivation would converge. Thus, following Takano, I assume that (46b) is one of those instances where only derivations with scrambling converge.28 29

28 There is a problem concerning the phase impenetrability condition in (i) proposed in Chomsky (1998). (i) In phase á with head H, the domain of H is not accessible to operations outside á, but only H and its edge In the CP phase, only C0 and Spec.CP are accessible. However, my analysis requires that Spec.TP and adjuncts to TP be also accessible from a probe in the next higher phase. I do not have anything to say about this, and leave it open now. 29 Colin Phillips pointed out to me that this analysis does not prevent two zibuns with different antecedents from appearing in a single sentence, which is ungrammatical as shown in (i). (i) Takashii-ga [zibuni/*j-ga satta-ato] Mariko j-ni zibuni-o suisenshita Takashi-NOM self-NOM left-after Mariko-DAT self-ACC recommended ‘Takashii recommended himselfi to Mariko j after selfi/*j left’ The derivation we need to prevent is the following; Mariko starts in the adjunct clause with ziun, Takashi with zibun is inserted to the lowest internal è-position of the main verb, and then Mariko sideward moves to the second internal è-position. Takashi moves to the external è-position, and other movements follow for the feature checking. One possibility suggested by Norbert Horstein is to assume that when two movements are about to take place, movement within a single subtree is more economical than movement across the subtrees. In the above case, sideward movement of Mariko from the adjunct clause to the main clause is more expensive than Takashi moving within the main clause. When Takashi’s movement in the main clause is done, Mariko looses its target, so the derivation crashes. The question is if this condition can be violated for convergence. Suppose the answer is ‘no.’ We still need to be able to derive the grammatical case of (i), where the antecedent of zibun in both the adjunct and the main clause is Takashi, and (ii), which is also grammatical. (ii) Takashii-ga [zibuni-ga satta-ato] zibuni-ni Mariko j-o suisenshita Takashi-NOM self-NOM left-after self-DAT Mariko-ACC recommended ‘Takashii recommended Mariko j to himselfi after selfi left’ In both cases, Takashi starts in the adjunct. In these cases, Takashi sideward moves, and Mariko is inserted. Thus, Merge over Move condition will derive (ii) but prevent the grammatical case of (i). In order to derive both cases, we

Page 18: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 18

The analysis in (46) is basically identical to the one proposed by Kimura (1994). He proposed that a feature-bundle of zibun can move at LF out of the embedded clause through the extra Spec. TP, which is allowed in Japanese according to Ura (1994), thereby voiding the MLC violation. My analysis differs minimally from Kimura’s in two respects. I assume adjunction to TP rather than substitution to Spec.TP as illustrated above, and overt movement instead of LF movement. I believe my analysis has an advantage in that it does not raise one problem that Ura’s (1994) proposal does.

Let us briefly summarize Ura’s argument. He found a correlation between the existence of multiple subject constructions and superraising constructions as in (50a) and (50c) respectively. (50) a. Zoo-ga hana-ga nagai elephant-NOM nose-NOM long

‘Elephant’s noses are long’ b. Boku-wa [John-ni piano-ga/-o hikeru-to ] omou

I-TOP John-DAT piano-NOM/ACC play-can-COMP think ‘I think that John can play the piano’

c. John-ni piano-ga/*-o hikeru John-DAT piano-NOM/*ACC play-can ‘John can play the piano’

As shown in (50a), Japanese allows multiple subjects. Ura proposes that specifiers of functional heads are licensed by the feature checking between the head and the element in its specifier. He claims that Agrs0, that is responsible for checking of Case and φ-features, has the ability to check more than one set of φ-features and Case in the languages which allow multiple specifiers like Japanese. Therefore, more than one DP can move to the AgrsP and check its features thereby licensing multiple specifiers. This accounts for the multiple subject constructions such as (50a). Now, compare (50b) and (50c). The internal argument of stative predicates like, hikeru, gets a nominative Case in a simplex sentence as in (50c). When it is embedded, however, the internal argument optionally takes an accusative Case as in (50b). Ura argues that the embedded object, piano, in (50b) optionally moves to the matrix clause at LF to check the accusative Case feature. This movement is possible through the extra Spec.AgrsP of the embedded clause because Japanese allows multiple specifiers. That is why the accusative Case on piano is possible only when it is the object of the embedded clause. It is not possible for the object of a simplex sentence in (50c) simply because there is no such position available to check the accusative Case. Thus, he accounts for the difference in grammaticality between (50b) and (50c). It is puzzling, however, why the embedded object in (50b) can move further if it checks its Case feature at the embedded Spec.AgrsP. In the case of multiple subject constructions in (50a), the extra subject, zoo ‘elephant’ stays at the outer Spec.AgrsP after checking the nominative Case. Thus, when we embed the multiple subject construction into another verb, it becomes ungrammatical as shown in (51).

need further to assume that either operation can take precedence over the other when both targets are in the same domain, namely, VP.

Another possibility is to assume that syntax can produce (i), but it violates some discourse constraint. Iida (1996) claims that this restriction is due to the “fact that within a specifiable domain, there can be only one perspective (Fillmore 1975, Partee 1989), hence only one potential referent for zibun.”

Page 19: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 19

(51) John-wa [zoo-ga/?*-o hana-ga mijikakatta-to ] shinjiteiru I-TOP elephant-NOM/ACC nose-NOM was.short-COMP believe ‘John believes that that elephant’s noses were short’

Thus, Ura’s proposal is inconsistent with respect to the mobility of DPs after the feature checking, hence requires an explanation. In the adjunction approach proposed here, on the other hand, the optionality of movement can be attributed to the optionality of scrambling.30 Therefore, it is consistent in the instances of long distance reflexives and scrambling. 4.3.2 Zibun in object control Let us extend our discussion to object control constructions. (52b) shows the property of object control as its English counterpart in (52a). Thus, PRO must refer to the matrix object rather than the subject. (52) a. Johni persuaded Maryj [PRO*i/j to leave ]

b. Takashii-ga Kenjij-o [PRO*i/j gakko-e iku-yoo(ni) ] settokushita 31 Takashi-NOM Kenji-ACC school-to go-SUBJ persuaded ‘Takashi

persuaded Kenji to go to school’

As mentioned in section 2, Hornstein claims that OC PRO in (52a) is an NP-trace left by the overt movement of Mary. The difference between OC PRO and reflexives is that OC PRO appears in non-Case position while reflexives appear in Case position. Aoshima (forthcoming) argues that OC PRO in Japanese should also be analyzed as a residue of movement.32 Adopting these claims, let us see how the derivation goes with zibun in an object control sentence as in (53). (53) Takashii-ga Kenjij-o [ tj zibuni/j-o suisensuru-yoo(ni)] settokushita

Takashi-NOM Kenji-ACC self- ACC recommend- SUBJ persuaded ‘Takashii persuaded Kenjij to recommend selfi/j’

Two interpretations are possible just like the regular complex constructions we saw above; one in which zibun refers to the embedded subject, Kenji, which turns out to be the matrix object, and the other in which zibun refers to the matrix subject, Takashi. The derivation for the former is illustrated in (54). (54) zibun = Kenji

a. CP1: {Kenji, suisensuru, v10, yooni(T0), C0, …)

CP2: {Takashi, settokusi, v20, ta(T0), …}

30 I will discuss more about the adjunction below. 31 I follow Uchibori (1996) in that this type of control complements in Japanese are subjunctive CP. Alternatively, the embedded clause may be a vP as proposed by Aoshima (forthcoming). I will discuss this issue below. 32 Zibun in the embedded subject position as in (i) refers to the matrix object. (i) sounds very odd unless zibun is used emphatically, but the distinction seems clear between the object reading which is somewhat marginal and the subject reading which is impossible.

(i) Takashii-ga Kenjij-o [zibun*i/??j-ga gakko-e iku-yoo(ni)] settokushita Takashi-NOM Kenji-ACC self-NOM school-to go-SUBJ persuaded

‘Takashii persuaded Kenjij for himself*i/?j to go to school’ I do not address (i) in this paper, but the analysis here can be extended to (i).

Page 20: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 20

b. CP2 TP2

8 Takashi T2’

[θ,ünom] vP2 T2

7 KKKeeennnjjjiii v2’ ta [θθθ,üacc]

6 è Takashi v2’ [θ, nom]

VP2 v2

5 Kenji V2’ [θθθ, acc]

CP1 V2

TP1 C1 settokusi 4 Kenji T1’ [θθ, acc]

vP1 T1

3 zzziiibbbuuunnn v1’ yooni

[üacc] 2 Kenji v1

’ [θθ, acc]

VP1 v1

1 è Kenji+zibun V1 [θ, acc][acc]

suisensuru Since the lower CP phase (CP1) does not include any other DP to insert, Kenji moves to the external è-position (1, 2, 3), and further to the matrix object position (4, 5), obtaining three θ-roles. In the next phase (CP2), Takashi is inserted at the external è-position of the matrix clause (6) and moves to Spec.TP2 (8).

Two points need clarification. First, the embedded T0 does not check Kenji’s Case, thereby allowing further movement of Kenji to the matrix clause (5). Uchibori (1996) argues that a clause headed by -yoo(ni) such as in (54) is a subjunctive CP, and that the subjunctive CP does not create an opaque domain for a local reflexive to be bound outside of the CP. Given the mechanism of feature checking under the MP I am assuming here, her assumptions can be interpreted as follows; the embedded T0 head in the subjunctive CP optionally lacks Case feature, thus allowing further movement of DP from its Spec. position. This accounts for the movement of Kenji from the embedded Spec.TP1 to the matrix Spec.VP2. Alternatively, it may be that the embedded clause is a vP, proposed by Aoshima (forthcoming). Then, Kenji moves

Page 21: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 21

directly to the matrix è-position. The choice between two alternatives requires consideration that is beyond the scope of this paper, and I leave it open here.

Second, note that movement of Kenji from TP1 to the matrix Spec.VP2 in step 5 violates the Merge over Move condition because Takashi could have been inserted to this position instead. However, the derivation would not converge in that case since movement of Kenji from TP1 to the inner Spec.vP2 over Takashi in Spec.VP2 would violate the MLC. Chomsky (1995; 348) states that ‘we select Attract/Move [rather than Merge]…if that is necessary for convergence.’ Thus, the less economical derivation is allowed in this case.

Before moving on to the second interpretation of (53), let us consider the definition of the minimal domain introduced in (49) repeated in (55). Watanabe (1996) notes that according to this definition of the minimal domain in Chomsky (1995, 178), an adjunct, ZP in (56), belongs to both the minimal domain of XP and that of YP at the same time.33 (55) a. Domain of a head α

The set of nodes contained in Max(α) that are distinct and do not contain α, where Max (α) is the least full-category maximal projection dominating α

b. Minimal domain of α

The smallest subset K of S (= the domain of α) such that for any γ∈S, some β∈K reflexively dominates γ

(56) YP Y0 XP

ZP XP

WP …X0… This means that ZP can satisfy checking requirement of X0 and Y0 at the same time. Given this assumption about adjunction, let us go back to the second interpretation of (53) repeated here as (57a) with its derivation in (57b). (57) a. Takashii-ga Kenjij-o [ tj zibuni-o suisensuru-yoo(ni)] settokushita

33 This was also noted by Nunes and Kato (in progress), Tada (1993), and Ura (1993).

Page 22: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 22

b. TP2 9 Takashi T2’

[θθ,ünom] vP2 T2

8 KKKeeennnjjjiii v2’ ta [θθ,üacc]

7 Takashi v2’ [θθ, nom]

VP2 v2

6 Kenji V2’ [θθ, acc]

CP1 V2

TP1 C1 settokusi 5 Takeshi TP1 [θ, nom]

4 Kenji T1’ [θ, acc, üD]

vP1 T1

3 zzziiibbbuuunnn v1’ yooni

[üacc] 2è Kenji v1

’ [θ, acc]

VP1 v1

1è Takeshi+zibun V1 [θ, nom][acc]

suisensuru In this case, both Takashi and Kenji are available in the lower CP phase. Takashi starts with zibun (1). Kenji is inserted as the Agent of the embedded clause (2), and moves to the matrix object position through Spec.TP1 (4, 6, 8). Takashi moves to the matrix subject position through adjunction to TP1 (5, 7, 9). Two assumptions are in order for why Takashi in TP1 can move over Kenji in VP2 in step 7. First, I assume that C1 incorporates into V2 in control constructions thereby absorbing the CP node.34 Second, adopting the definition of minimal domain in (55), I assume Takashi adjoined to TP1 is in the minimal domain of both T1 and V2 after the absorption of CP1. Note that Kenji in Spec.VP2 is also within the minimal domain of V2. Therefore, movement of Takashi over Kenji does not violate the MLC because they are equidistant with respect to the target, hence the movement is licit. The derivation converges as

34 I do not have any story to justify why C incorporates into the matrix verb. If we adopt Aoshima’s (forthcoming) assumption that the embedded clause of control constructions in Japanese is a vP, this problem would go away.

Page 23: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 23

desired. Notice that if the adjunction did not take place, the derivation would not converge, and there is no way to converge without adjunction as I discussed in section 4.3.1. 4.4 Nominative anaphors A nominative anaphor is allowed in Japanese as shown in (4a) repeated here in (58). (58) a. Marikoi-ga [ zibun-gai ichiban moteru-to ] shinjiteiru

Mariko-NOM self- NOM best be-popular-COMP ] believe ‘Marikoi believes that shei is popular the most’

b. *Maryi believes that she-selfi is the most popular Woolford (1999) presents evidence that a nominative anaphor is allowed in languages like Japanese, which do not exhibit subject-verb agreement supporting Rizzi’s (1990) anaphor agreement effect, which states that anaphors do not occur in syntactic positions construed with agreement. On the other hand, languages like English do not allow nominative anaphors. Hornstein (2000) explains this correlation between nominative anaphors and lack of subject verb agreement as follows. In English (58b), when the complex DP, Mary+self, reaches the lower Spec.TP, it must check φ-features, the nominative Case and EPP features of T0. However, these features must be checked by a single DP at once if possible. In other words, it is not possible that Mary checks φ-features and the EPP of T0, and self checks the Case. Rather, Mary has to check all those features at once. If Mary checks all of them, however, it cannot move further to the matrix subject. Thus the derivation crashes since the Case feature on self remains unchecked, and no DP can get the matrix θ-role, and check the uninterpretable features on matrix T0. Languages like Japanese, on the other hand, do not exhibit subject-verb agreement. When the complex NP, Mariko+zibun in (58a) moves to the embedded Spec.TP, what needs to be checked are the nominative Case and EPP features. Assuming that no φ-feature is involved in T0, zibun can check those features.35 Thus, zibun successfully checks all of the relevant features of T0 and of itself. Thus, Mariko can move up to the matrix clause, and the derivation converges. As Hornstein (2000) notes, his analysis of nominative reflexives predicts that reflexives cannot appear in the object position in languages that show object agreement. That seems correct. Thus, Woolford (1999), as further support for the anaphor agreement effect, presents ample evidence that languages that exhibit object agreement do not have anphors in object position. For instance, in Inuit, transitive verbs agree with subject and object in normal cases, but the object agreement with a reflexive object is not allowed as the paradigm in (59) indicates. (59) 36 a. Angutip arnaq taku-vaa man(ERG) woman(ABS) see-IND.3SG.3SG ‘The man sees the woman’ b. * Hansiup immi asap-pup Hansi(ERG) himself(ABS) wash-IND.3SG.3SG ‘Hansi washed himself’ In order to express the anaphoric relation, Inuit employs either the intransitive constructions with no overt object (60a) or the oblique Case marked reflexive object without object agreement (60b). (60) a. Asap-pup wash-IND.3SG He washed himself‘ 35 I have to assume that zibun has a D-feature in addition to Case. 36 Data in (59) and (60) are originally from Bok-Bennema (1991).

Page 24: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 24

b. Angut immi-nut taku-vuq man himself-DAT see-IND.3SG ‘The man sees himself’ Thus, Horstein’s prediction is confirmed in those languages. 5 Remaining issues I have presented a new account of zibun in the previous sections. For the complete analysis of zibun, however, it is inevitable to take other anaphoric expressions into consideration. I would like to address two issues that are related to zibun. One concerns the noncomplementary distribution of reflexives and pronouns in Japanese. The other is how to treat two other reflexives in Japanese, namely, zibun-zisin and kare-zisin. Obviously, these questions are beyond the scope of this paper, and I present ideas below simply as suggestions that need to be scrutinized in the future research.37 5.1 Noncomplementary distribution of reflexives and pronouns Japanese differs from English with respect to the complementarity of reflexives and pronouns. Reflexives and pronouns in English are in (near) complementary distribution as in (61).38 (61) a. Johni likes himselfi/*himi b. John thinks hei/*himselfi is a genius c. John believed that Mary likes himi/*himselfi In Japanese, on the other hand, the complementarity is restricted to the local anaphoric relations as shown in (62). (62) a. Takashii-ga zibuni/*karei-o suisenshita

Takashi-NOM self/him-ACC recommended ‘Takashii recommended himselfi’

b. Takashii-ga [ zibuni/karei/j-ga ichiban yuushuuda-to] omotteiru Takashi-NOM self/he-NOM best smart-COMP think ‘Takashii thinks that selfi/hei/j is the smartest.’

c. Takashii-ga [ Kenjij-ga zibuni/j/karei/*j/k-o suisenshita-to ] omotta Takashi-NOM Kenji-NOM self/him-ACC recommended-COMP thought ‘Takashii thought that Kenjij recommended selfi/j/himi/*j/k.’

In section 2, I discussed that zibun exhibits the properties of reflexives contrasting with pronouns which do not exhibit those properties. The generalization has been made in the literature as that Japanese personal pronouns can only refer. Noguchi (1997) argues that personal pronouns in Japanese are N-pronouns as opposed to the English personal pronouns which he claims are D-pronouns and that only D-pronouns are subject to both binding and coreference in the sense of Grodzinsky and Reinhart (1993) while N-pronouns are only subject to coreference.

37 It is important to note that there is another anaphoric element, namely, pro , in Japanese. However, I will not discuss about it since introducing pro will complicate the picture drastically. 38 It is not an absolute complementary distribution since a reflexive and a pronoun can co-occur in (i), as is well known. (i) a. Johni likes stories about himselfi/himi. b. Mary i thinks [pictures of herselfi/heri] are on display]

Page 25: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 25

Let us briefly summarize Grodzinsky and Reinhart’s proposal for the separation of binding and coreference. Following Reinhart (1983), they clam that only anaphoric elements that behave as a bound variable, whether lexical or not, are subject to binding, and those that do not are subject to coreference. Thus, A-bound anaphors, pronouns, and OC PRO in (63) are all bound variables subject to the binding theory.39 40 (63) a. Luciei adores herselfi

b. Luciei adores heri friends c. Alfredi thinks hei is a great cook

d. Everyonei bores himselfi e. Every actressi adores heri friends

f. Every scholari thinks hei is a great cook g. Everyonei thinks that Lucie admires himi h. Johni tried PROi to sneeze i. Everyonei tried PROi to sneeze

Pronouns in (64), on the other hand, are not subject to binding, but to coreference, which interprets anaphoric expressions that fail to be bound. 41 42 (64) a. Most of her friends adore Lucie b. A party without Lucie annoys her

Note that all of the anaphoric elements in (63) are derivable as a result of (failed attempts at) movement according to Hornstein (2000). He argues that a bound pronoun is not a lexical element in the numeration just like reflexives. It is rather introduced by the computational system to the derivation where a DP fails to move. For instance, the derivation of (63c) goes as follows. The numeration includes {Alfred, think, a, great, cook, ….}. Alfred starts in the derivation as the subject of the embedded clause. Alfred cannot undergo A-movement to the matrix clause given that the embedded clause is finite and that English does not allow nominative anaphors as we saw in section 4.4. In order to save this derivation, Alfred is unmerged, replaced by a pronoun, he, thereby allowing Alfred to remerge into the matrix clause. Hornstein calls this derivation as the Non-Movement Alternative (NMA). The NMA is also derived by movement in a sense that the numeration does not include the pronoun. Thus, we can reinterpret all of the binding constructions in (63) as involving movement. Hornstein assumes that the insertion of a pronoun is more costly than reflexives, which explains the complementary distribution of reflexives and pronouns. Thus, bound pronouns appear only where movement is not possible.

39 Data in (63) are from Grodzinsky and Reinhart (1993). 40 The binding theory that Grodzinsky and Reinhart (1993) assume is given in (i). (i) Definition: a node á is bound by a node â iff á and â are coindexed and â c-commands á. (ii) Conditions A: An anaphor is bound in its GC

B: A pronoun is free in its GC (iii) Translation definition

An NP is a variable iff either (a)it is empty and A’-bound, or (b) it is A-bound and lacks lexical content. Other cases of NP coindexation are uninterpretable

41 The condition on coreference is given as; Rule I: Intrasentential Coreference

NP A cannot corefer with NP B if replacing A with C, C a variable A-bound by B, yields an indistinguishable interpretation

42 From now on, I use the convention of italics to indicate coreference following Grodzinsky and Reinhart.

Page 26: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 26

On the other hand, coreference takes place between a referential expression and a pronoun that fails to be bound as in (64). However, the kind of condition on coreference that Grodzinsky and Reinhart proposed is not necessary. As Hornstein (2000; 185-186) assumes, suppose that English possesses two types of pronouns, one that is introduced by the computational system in order to save the derivation and the one that is a lexical element, which starts in the numeration just like other nominal expressions. Suppose further that two DPs in the same sub-array corresponding to a phase are obviative in nature, whether referential or not.43 This will explain why coreference cannot take place within the same clause.44 45 In the cases of anaphoric relations in (63), this constraint is irrelevant since there is only one DP in the sub-array to begin with. The assumption that there are two types of pronouns in English explains the ambiguous status of pronouns in (63b, c, e, f, g). Since two pronouns are homophones, the pronoun in those sentences can be interpreted as either one.

Following Noguchi (1997), I assume that Japanese pronouns are only subject to coreference. In other words, Japanese possesses only one type of pronouns unlike English. This makes sense since the reflexive, zibun, can refer to either a local or a long distance antecedent, there is no need to introduce a more costly pronoun, if Hornstein is correct. Thus, Huang (1991:74, fn12) suggests that for Japanese pronouns, an overt pronoun should be excluded in a position where a reflexive is available extending the Avoid Pronoun Principle of Chomsky (1981)/Overt Pronoun Constraint of Montalbetti (1984). However, this reasoning is not correct as Noguchi points out. Consider (65).46 (65) a. Mary-ga John-ni [kare/*zibun-ga tensaida-to] itta

Mary-NOM John-DAT he/self-NOM genius-COMP said ‘Mary told John that he is a genius’

b. *Mary-ga dono hitoi-ni-mo [karei-ga tensaida-to] itta Mary-NOM every person-DAT-also he-NOM genius-COMP said ‘Mary told every personi that hei is a genius’

Zibun is not possible as the subject of the embedded clause referring to the matrix object because of its subject oriented nature. Instead, the pronoun, kare, appears to convey the intended meaning. However, this pronoun cannot be a bound pronoun since it fails to be bound by a quantificational expression as (65b) shows. This confirms the generalization mentioned above.47

43 This idea comes from Uriagereka’s (1995) claim that structurally case marked DPs are obviative. I will come back to his claim in the next section. 44 However, Paul Pietroski pointed out to me that this condition should not rule out the case in (i), discussed in Higginbotham (1985). (i) He put on John’s coat. But only John would do that, so he is John I have nothing to say about this case here, and leave it open. 45 In order to allow coreference in (64) however, it is necessary to assume D/NP is also a phase. See Raposo (1999) for the defense of DP/PP phases. 46 (65) is from Noguchi (1997; 775). 47 It is worth mentioning that Martin (1988; 1075) notes, “two modern pronouns, kare ‘he/him’ and kano-zyo ‘she/her’ were created originally to translate the sex-insistent pronouns of English; their use has spread so rapidly that they can now be considered a part of the colloquial language, though most people will not use them in semantically ‘unmarked’ situations. …in many unmarked situations, the appropriate translation of an English pronoun is either zero or a repetition of the noun.”

Page 27: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 27

In summary, the noncomplementarity of reflexives and pronouns in Japanese seems attributable to the facts that zibun can take a long distance antecedent and that pronouns are subject to coreference. 5.2 Treatment of zibun-zisin, & kare-zisin In this section, I would like to present a possible explanation for two other reflexives in Japanese, zibun-zisin and kare-zisin, that is in accord with the analysis of zibun presented in this paper.

Katada (1991) proposed the following classification for the three reflexives in Japanese; (66) Three-way classification of reflexive binding

a. zibun: long-distance reflexives with subject orientation b. zibun-zisin: local reflexives with subject orientation c. kare-zisin: local reflexives with no particular orientation

Katada (1991) obtained (66b) and (66c) based on the observations that zibun-zisin must have a local antecedent which is a subject as in (67a) while kare-zisin must also have a local antecedent but the antecedent does not have to be a subject as in (67b). (67) a. Masarui-ga [ Takashij-ga Kenjik-ni zibun-zisin*i/j/*k-o suisenshita-to

Masaru-NOM Takashi-NOM Kenji-DAT self-self-ACC recommended-COMP] omotta thought ‘Masarui thought that Takashij recommended himself*i/j/*k to Kenjik’

b. Masarui-ga [ Takashij-ga Kenjik-ni kare-zisin*i/j/k-o suisenshita-to] Masaru-NOM Takashi-NOM Kenji-DAT self-self-ACC recommended-COMP] omotta thought ‘Masarui thought that Takashij recommended himself*i/j/k to Kenjik’

Richards (1997) presents Aikawa (1995)’s observation that zibun-zisin behaves as a bound variable while kare-zisin does not. Katada (1991) also made the same point. (68) a. John-dake-ga kare-zisin-o hihanshita.

John-only-NOM him-self- ACC criticized ‘Only John criticized himself (no one else criticized John)’

b. John-dake-ga zibun-zisin-o hihanshita. John-only-NOM self-self- ACC criticized ‘Only John criticized himself (no one else performed self- criticism)’

As the English translations indicate, two sentences are used in different contexts. Thus, Richards postulates (69a) and (69b) as the semantic representation for (68a) and (68b) respectively. (69) a. λx [x criticized John]

b. λx [x criticized x] This is consistent with the proposals in this paper that zibun is a residue of movement and that kare is not. Difference between zibun and zibun-zisin/kare-zisin is whether they allow long distance reading or not. Now, compare (67) with (70). (70) a. Masarui-ga [CP Takashij-ga Kenjik-ni zibuni/?j/*k-o suisenshita-to]

Masaru-NOM Takashi-NOM Kenji-DAT self-ACC recommended-COMP] omotta. thought ‘Masarui thought that Takashij recommended himselfi/?j/*k to Kenjik’

Page 28: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 28

b. Masarui-ga [CP Takashij-ga Kenjik-ni karei/*j/*k-o suisenshita-to] Masaru-NOM Takashi-NOM Kenji-DAT him-ACC recommended-COMP] omotta. thought ‘Masarui thought that Takashij recommended himi/*j/*k to Kenjik’

Zibun-zisin and kare-zisin in (67) must have its antecedent in the local domain while zibun in (70a) need not, and kare in (70b) must not. The crucial difference is the existence of zisin.

Based on the Danish reflexive construction as in (71), Uriagereka (1995) argues that structurally case marked DPs are syntactically obviative in their local domains, but the morpheme, selv, imposes the semantic restriction of sameness. (71) a. Peter fortalte Anne om hende selv /* ham selv.

Peter told Anne about her self him self b. [the x: Peter(x) & X(x)] [the y: Anne(y) & Y(y)] [the z: one(z) & same-as-before(z)] x told y about z He states, “selv makes the context variable predicated of the variable of the definite description to be the same context variable as a previously introduced one. …the closest one,” assuming names introduce a context variable. Thus, the pronoun to which selv is attached has to refer to the same individual with its closest DP. Ham selv in (71) leads to a contradiction since the closest DP is Anne, which is feminine, while ham is masculine.

Returning to Japanese, suppose that zisin is a sameness restrictor parallel to the Danish selv, and that when zisin is attached to zibun/kare as in (67), the range of the context variable predicated of the variable introduced by zibun/kare is confined to the range of context variable introduced by the DPs in the same local domain (instead of the closest one in this case for some reason). The exact definition of the local domain has to be clarified, but let us assume tentatively, it is the CP phase. Then, although zibun by itself can refer to the long distance antecedent, zisin adds the sameness ristriction, thus zibun-zisin in (67a) must refer to the embedded subject, Takashi.48 Similarly, although kare is a lexical element thus subject to the ristriction on the lexical array, kare-zisin in (67b) must refer to either one of the DPs in the same clause (the same sub-array). I do not know why both DPs in the local domain can be the antecedent of kare-zisin in Japanese unlike Danish in (71), where the antecedent must be the closest DP, Anne, even though Peter is also in the same domain. Recall that zibun-zisin and kare-zisin are different in that only the former behaves as a bound-variable as discussed with the data in (68). The account here is consistent with this observation since zibun-zisin involves movement of zibun but kare-zisin does not. 6 Concluding remarks The LF movement approach to anaphors has been standard since Lebeaux (1983). Chomsky (1986, 1995), following Lebeaux (1983), proposed that a reflexive moves to I0(Infl) at LF leaving a trace, which is subject to the Empty Category Principle (ECP). Based on the observation that non-compound reflexives found in Danish and Icelandic are subject oriented and can allow a long distant antecedent while compound reflexives such as himself in English is local non-subject oriented, Pica (1987) argues that the non-compound reflexives are X0 reflexives and the compound reflexives are X”. He proposes that X0 reflexives head-move to I0 thereby deriving the subject orientation whereas X” reflexives adjoin to X” elements such as VP and PP

48 I assume that zibun in zibun-zisin is also a result of movement just like the monomorphemic zibun we have discussed in this paper.

Page 29: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 29

thus non-subject orientation. Furthermore, since adjunction of the X” reflexive to the embedded CP is not allowed, X” reflexive cannot move out of the embedded clause, and ends up as a local reflexive. Cole and Sung (1994) and Cole and Wang (1996) also defend the LF head movement approach applying to Chinese reflexive, ziji. Katada (1991) and Huang and Tang (1991) proposed that long distance reflexives undergo successive cyclic A’-adjunction at LF, applying to Japanese zibun and Chinese ziji respectively. However, none of these analyses cover the disjunctive properties of reflexives, namely the subject orientation and backward reflexives as far as I can tell.49

In addition, two more questions arise in the LF movement approach. Chomksy (1995: 211) wrote that “Condition A may be dispensable if the approach … is correct and the effects of Condition A follow from the theory of movement (which is not obvious),… All indexing could then be abandoned, another welcome result.” However, all of the proposals above still retain binding condition A as well as the indices, contrary to Chomsky’s intention. Secondly, from the minimalist standpoint, it is not clear why reflexives move as these proposals claim they do. In other words, what triggers the LF movement?

On the other hand, overt A-movement approach presented in this paper provides a unified account for the disjunctive properties of reflexives that previous accounts failed to explain. I have argued that the subject orientation follows from Merge over Move. The apparent exception of backward reflexives follows naturally by adopting phases and sideward movement. Long distance reflexives in complex sentences as well as in object control constructions are also accounted for in a natural way if one assumes that scrambling is adjunction, and adjunction affords a way of escaping MLC/SMC violation. Nominative anaphor is possible in Japanese given that Japanese lacks φ-features in T0. Although there are questions and problems that I swept under the footnotes in various places, if this account is on the right truck, this supports the notion of phases, and Hornstein’s (2000) claims that binding Condition B is dismissable and that movement is triggered by è-features.

49 I did not address the intervention effect in this paper. As mentioned in footnote 8, if Japanese also exhibits this effect, that must be accounted for. Some of the previous analyses mentioned above explain that the effect is due to the mismatch between the matrix and the embedded subject’s person agreement features. This account is not available in the analysis proposed here since it relies on the assumption that Japanese does not instantiate the subject verb agreement. One possibility is to treat this effect as a consequence of discourse factors. A growing number of papers in recent literature proposed accounts for long distance reflexives that take both grammatical restrictions and discourse factors into consideration. For instance, Cole, Hermon, and Lee (2001) claim that the blocking effect should be at least partially attributed to the discourse condition. See Cole, Hermon, and Lee (2001).

Page 30: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 30

References Aikawa, T. (1993) Reflexivity in Japanese and LF analysis of zibun-binding. Doctoral

dissertation, Ohio State University. Aikawa, T. (1995) Remarks and replies: “Reflexivity” by Reinhart and Reuland (1993). ms. MIT Aoshima, S. (forthcoming) Mono-clausal control and scrambling. University of Maryland

Working Papers in Linguistics, vol. 11. Baker, M. (1997) Thematic roles and syntactic structure. In L. Haegeman (ed), Elements of

Grammar Bickerton, D. (1986) Binding parameters and argument structure. ms. University of Hawaii,

Manoa. Bok-Bennema, R. (1991) Case and agreement in Inuit. Dordrecht: Foris. Boškoviæ, Z. (1994) D-Structure, Theta Criterion, and movement into theta positions. Linguistic

Analysis, 24. Boškoviæ, Z. and D. Takahashi. 1998. Scrambling and last resort. LI, 29. Chomsky, N. (1981) Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1986) Knowledge of Language. New York: Praeger. Chomsky, N. (1995) A minimalist program for linguistic theory. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. (1998) Minimalist inquiries: the framework. MIT Occasional Papers in

Linguistics 15. Cambridge, MA: MITWPL. Chomsky, N. (1999) Derivation by phase. MIT Occasional Papers inLinguistics 18. Cambridge,

MA: MITWPL. Chuang, L. (1997) Long distance anaphora & multiple feature checking: a minimalist

approach. Doctoral dissertation, University of Maryland, College Park. Cole, P., G. Hermon, and C. L. Lee (2001) Grammatical and discourse conditions on long

distance reflexives in two Chinese dialects. In: Peter Cole, et al. (eds) Long Distance Reflexives, Syntax and Semantics. Academic Press.

Cole, P. and L. Sung (1994) Head movement and long-distance Reflexives. LI 25: 355-406. Cole, P. and Wang (1996) Antecedents and blockers of long-distance reflexives: the case of

Chinese ziji. LI, 27: 357-390. Dowty, D. (1991) Thematic Proto-roles and Argument Selection. Language, 67:547-619. Fillmore, (1975) Santa Cruz Lectures on Deixis. Bloomington: Indiana University Linguistic

Club. Fukui, N. (1984) Studies in Japanese anaphora. ms. MIT. Grodzinsky, Y. and T. Reinhart (1993) The innateness of binding and coreference. LI, 24: 69

-101. Hasegawa, N. (1985) On the so-called “zero pronouns” in Japanese. The Linguistic Review, 4:

289-341. Higginbotham, J. (1985) On semantics. LI, 16: 547-593. Hoji, H. (1990) Theories of anaphora and aspects of Japanese syntax. ms. University of

Southern California. Hornstein, N. (1999) Movement and control. LI, 30:69-96. Hornstein, N. (2000) Move! A Minimalist Theory of Construal. Oxford: Blackwell. Huang, C. -T. J. (1991) Remarks on the status of the null object. In R. Freidin (ed.), Principles

and Parameters in Comparative Grammar: 56-76. Huang, C.-T. J. and C.-C. J. Tang (1991) The local nature of the long-distance reflexive in

Page 31: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 31

Chinese. In: J. Koster and E. Reuland (eds), Long-distance anaphora. Cambridge, MA: Cambridge University Press.

Iida, M. (1996) Context and Binding in Japanese. Stanford, CA: CSLI Publications. Katada, F. (1991) The LF representation of anaphors. LI, 22: 287-313. Kato, M. & J. Nunes. (in progress) Adjunction configuration and structural ambiguity.

Universidade Estadual de Campinas, Brazil. Kiguchi, H. (forthcoming) PRO gate and movement. University of Maryland Working Papers in

Linguistics, vol. 11. Kimura, N. (1994) Multiple specifiers and long distance anaphora. In: M.Koizumi & H.Ura

(eds.), Formal Approaches to Japanese Linguistics 1. MITWPL, 24. Kuno, S. (1973) The Structure of the Japanese Language. Cambridge, MA: MIT Press. Kuroda, S. -Y. (1965) Causative forms in Japanese. Foundation of Language, 1:30-50. Kuroda, S. -Y. (1986) Movement of noun phrases in Japanese. In T. Imai and M. Saito (eds.),

Issues in Japanese Linguistics: 229-271. Dordrecht: Foris. Lasnik, H. (1995) Last resort and attract F. Proceedings of FLSM 6. Indiana University

Linguistics Club. Bloomington. Lebeaux, D. (1983) A distributional difference between reciprocals and reflexives. LI, 14:

723-730. Lebeaux, D. (1985) Locality and anaphoric binding. The Linguistic Review, 4: 343-363. Lidz, and Idsardi (1998) Chains and phono-logical form. U. Penn Working Papers in

Linguistics, vol. 5. McCawley, N. 1976. Reflexivization: a transformational approach. In M. Shibatani (ed.), Syntax

and Semantics 5: Japanese Generative Grammar:51-116. New York: Academic Press. Martin, S. E. (1988) A Reference Grammar of Japanese. Tokyo: Charles E. Tuttle. (First

published in 1975 by Yale University) Montalbetti, M. (1984) After binding: on the interpretation of pronouns. Doctoral dissertation,

MIT. Muromatsu, K. (1998) On the Syntax of Classifiers. Doctoral dissertation, University of

Maryland, College Park. Noguchi, T. (1997) Two types of pronouns and variable binding. Language, vol. 73: 770-797. Nunes, J. (1995) The copy theory of movement & linearization of chains in the minimalist

program. Doctoral dissertation, University of Maryland, College Park. Partee, B. (1989) Binding implicit variables in quantified contexts. Papers from the 25th

Regional Meeting of the Chicago Linguistic Society: 342-365. Chicago: Chicago Linguistics Society.

Pesetsky, D. (1995) Zero Syntax. Cambridge, MA: MIT Press. Pica, P. (1987) On the nature of the reflexivization cycle. Proceedings of NELS, 17, vol. 2. Reinhart, T. (1983) Anaphora and semantic interpretation. London: Croom Helm. Richards, N. (1997) Competition and disjoint reference. LI, 28: 178-187. Rizzi, R. (1990) On the anaphor-agreement effect. Rivista di Linguistica, 2:27-42. Sag, I. (1976) Deletion and logical form. Doctoral dissertation, MIT. Saito, M. (1992) Long distance scrambling in Japanese. Journal of East Asian Linguistics 1. Sells, P. (1987) Aspects of logophoricity. LI, 18: 445-479. Sportiche, D. (1986) Zibun. LI, 17: 368-374. Tada, H. (1993) A/A-bar partition in derivation. Doctoral dissertation, MIT. Takano, Y. (1998) Object shift and scrambling. Natural Language and Linguistic Theory, 16:

Page 32: Zibun As a Residue of Overt A-movement Mitsue MotomuraZibun, Mitsue Motomura, 1 Zibun As a Residue of Overt A-movement Mitsue Motomura 1. Introduction As is well known, Japanese zibun

Zibun, Mitsue Motomura, 32

817-889. Tateishi, K. (1991) The syntax of subjects. Distributed by GLSA (published from CSLI, 1995). Uchibori, A. (1996) Opacity and subjunctive complements in Japanese. In: H. Shon & J. Haig

(eds.), Japanese Korean Linguistics 6: 399-414. Uchibori, A. (In progress) The syntax of subjunctive complements. Doctoral dissertation,

University of Connecticut. Ura, H. (1993) On feature checking for wh-traces. In: J. D. Bobaljik and C. Phillips (eds.),

Papers on Case and Agreement 2. MITWPL 19. Ura, H. (1994) Varieties of raising and the feature-based bare phrase structure theory. MIT

Occasioal Papers in Linguistics 7, distributed by MITWPL. Uriagereka, J. (1995) Formal and substantive elegance in the minimalist program. In: C.

Wilder, H. Gartner and M. Bierwisch (eds.), The Role of Economy Principles in Linguistic Theory: 171-204. Berlin: Akademie Verlag.

Watanabe, A. (1996) Nominative-genitive conversion and agreement in Japanese: A Cross- Linguistic Perspective. Journal of East Asian Linguistics 5.

Woolford, E. 1999. More on the anaphor agreement effect. LI, 30: 257-287.