+ All Categories
Home > Documents > Phases and Locality in Japanese

Phases and Locality in Japanese

Date post: 21-Dec-2015
Category:
Upload: ramlohani
View: 22 times
Download: 7 times
Share this document with a friend
Description:
Phases and locality in Japanese
Popular Tags:
30
PHASES AND LOCALITY CONSTRAINTS ON A-MOVEMENT IN JAPANESE* Hirohisa Kiguchi Abstract. This paper explores locality conditions on A-movement in Japanese and their interaction with Chomsky’s phases, based on McGinnisÕ (2001, 2004) proposal. Assuming morphologically derived causative constructions share the same structural status with McGinnisÕ (2001, 2004) applicative constructions in the spirit of Marantz (1993), this paper provides evidence that locality constraints on A-movement exist in Japanese and that they should interact with phrasal and ÔphasalÕ structures. Furthermore, this analysis can provide a unified explanation to the (un)availability of long passives both in causative and applicative constructions in various languages. 1. Introduction There has been debate over whether Japanese is immune to locality constraints on A-movement, i.e. Rizzi’s (1990) Relativized Minimality or Shortest Move/Minimal link Condition (Kuroda 1988, Miyagawa 1997, Takano 1998 among others). This paper, assuming McGinnisÕ (2001, 2004) applicative constructions (such as the double object constructions), attempts to provide evidence that locality constraints on A-movement exist in Japanese and that they should interact with phrasal and ÔphasalÕ structures. This paper is organized as follows. Section 2 reviews McGinnisÕ (1998, 1999) previous argument for locality constraints on A-movement in Japanese scrambling and points out that the evidence provided there is not conclusive. Section 3 introduces the asymmetry in the syntax of Applicative constructions within passivization and McGinnisÕ (2001, 2004) account for this asymmetry. Section 4 extends McGinnisÕ (2001, 2004) analysis of the Applicative constructions to other types of constructions, especially to causative constructions. In section 5, given that causative constructions have a structural regulation similar to McGinnisÕ (2001, 2004) proposal for applicative constructions, Japanese causative constructions are shown to obey locality constraints on * I thank Duk-ho An, Hiroshi Aoyagi, Ilhan Cagri, Youngmi Jeong, Heidi Harley, Chigusa Morita, and Tomokazu Takehisa for discussion. Special thanks go to Cedric Boeckx for his suggestions and encouragement. I am also grateful for feedback from the audience at NELS 35, Minimalist Theorizing Workshop at Indiana University, and the 6th Seoul International Conference on Generative Grammar, especially Sung-ho Ahn, Yoshihisa Kitagawa, and Hang-jin Yoon. I also thank two anonymous reviewers, whose comments made this a better paper. Abbreviations are as follows. ACC: accusative; AGR: agreement; APPL: applicative; ASP: aspect; CAUS: causative; DAT: dative; FOC: focus; FV: final vowel; GEN: genitive; HON: honorification; NOM: nominative; OP: object pronoun; PASS: passive; PR: present; SP: subject pronoun; TOP: topic. Studia Linguistica 60(1) 2006, pp. 34–63. ȑ The author 2006. Journal compilation ȑ The Editorial Board of Studia Linguistica 2006. Published by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA
Transcript

PHASES AND LOCALITY CONSTRAINTS ONA-MOVEMENT IN JAPANESE*

Hirohisa Kiguchi

Abstract. This paper explores locality conditions on A-movement in Japanese andtheir interaction with Chomsky’s phases, based on McGinnis� (2001, 2004)proposal. Assuming morphologically derived causative constructions share thesame structural status with McGinnis� (2001, 2004) applicative constructions inthe spirit of Marantz (1993), this paper provides evidence that locality constraintson A-movement exist in Japanese and that they should interact with phrasal and�phasal� structures. Furthermore, this analysis can provide a unified explanationto the (un)availability of long passives both in causative and applicativeconstructions in various languages.

1. Introduction

There has been debate over whether Japanese is immune to localityconstraints on A-movement, i.e. Rizzi’s (1990) Relativized Minimality orShortest Move/Minimal link Condition (Kuroda 1988, Miyagawa 1997,Takano 1998 among others). This paper, assuming McGinnis� (2001,2004) applicative constructions (such as the double object constructions),attempts to provide evidence that locality constraints on A-movementexist in Japanese and that they should interact with phrasal and �phasal�structures. This paper is organized as follows. Section 2 reviewsMcGinnis� (1998, 1999) previous argument for locality constraints onA-movement in Japanese scrambling and points out that the evidenceprovided there is not conclusive. Section 3 introduces the asymmetry in thesyntax of Applicative constructions within passivization and McGinnis�(2001, 2004) account for this asymmetry. Section 4 extends McGinnis�(2001, 2004) analysis of the Applicative constructions to other types ofconstructions, especially to causative constructions. In section 5, giventhat causative constructions have a structural regulation similar toMcGinnis� (2001, 2004) proposal for applicative constructions, Japanesecausative constructions are shown to obey locality constraints on

* I thank Duk-ho An, Hiroshi Aoyagi, Ilhan Cagri, Youngmi Jeong, Heidi Harley,Chigusa Morita, and Tomokazu Takehisa for discussion. Special thanks go to CedricBoeckx for his suggestions and encouragement. I am also grateful for feedback from theaudience at NELS 35, Minimalist Theorizing Workshop at Indiana University, and the 6thSeoul International Conference on Generative Grammar, especially Sung-ho Ahn,Yoshihisa Kitagawa, and Hang-jin Yoon. I also thank two anonymous reviewers, whosecomments made this a better paper.Abbreviations are as follows. ACC: accusative; AGR: agreement; APPL: applicative; ASP:aspect; CAUS: causative; DAT: dative; FOC: focus; FV: final vowel; GEN: genitive; HON:honorification; NOM: nominative; OP: object pronoun; PASS: passive; PR: present; SP:subject pronoun; TOP: topic.

Studia Linguistica 60(1) 2006, pp. 34–63. � The author 2006. Journal compilation� The Editorial Board of Studia Linguistica 2006. Published by Blackwell Publishing Ltd.,9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA

A-movement. Section 6 shows that the proposed phrase structure forJapanese causative construction is in fact compatible with Pesetsky’s(1995) previous analysis of causative constructions. Section 7 discussessome theoretical implications of the consequences of the article. Section 8is the conclusion.

2. A-movement in Japanese

2.1. A-scrambling

As is well known, Japanese has a scrambling operation which can beregarded as A-movement. With scrambling, WCO-effects are mended,and anaphoric binding is licensed as indicated in (1) and (2), respectively.These properties are characteristic of A-movement as observed in raisingconstructions.

(1) a. ?? John-ga [Mary-ga proi atta]-atode dono hitoi-niJohn-nom Mary-nom pro met after every man-dat

-mo atta.-too met

�John met every mani after Mary had met him (pro)i.�b. �John-ga [dono hitoi-ni-mo]j [Mary-ga proi atta]-atode tj atta.

(e.g. Saito 1992)�

(2) a. ?* [Masao-wa [[otagaii-no sensei]-ni [karerai-oMasao-top each.other-gen teacher-dat they-acc

syookaisita]]]introduced

�Masao introduced them to each other’s teacher.�b. [[karerai-o]j [Masao-wa [[otagaii-no sensei]-ni [tj syookaisita]]].

The question then is why the object can A-move over the subject as in (2b):If it is A-movement, the subject should be an intervener for the movementof the object. That is, the movement violates some type of localityconstraint on A-movement. A possible answer is that A-movement doesnot obey locality in Japanese (Kuroda 1988, Saito 1989, Takano 1998).This type of account treats A-scrambling as a kind of adjunctionoperations which need not obey locality constraints on movement.

2.2. Previous research for A-locality in Japanese: McGinnis (1998, 1999)

McGinnis (1998, 1999) challenges the view that A-scrambling is free fromlocality constraints on A-movement. She first introduces Richards� (1997)observation that an idiomatic interpretation is lost when the lower objectis scrambled. For example, the sentence in (3) has two readings: an

m

m

A-movement in Japanese 35

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

idiomatic and a non-idiomatic one. However, the idiomatic interpret-ation is available only when the indirect object precedes the direct object.When the lower object is scrambled over the higher object, the idiomaticreading is absent, as in (4).1

(3) Taro-ga hi-ni abura-o sosoida.Taro-nom fire-dat oil-acc pouredReading 1: �Taro made things worse.� (Idiomatic reading)Reading 2: �Taro poured some oil into the fire.� (Non-idiomaticreading)

(4) a. Taro-ga abura-oi hi-ni ti sosoida.

Taro-nom oil-acc fire-dat poured*Reading 1: �Taro made things worse.� (Idiomatic reading)Reading 2: �Taro poured some oil into the fire.� (Non-idiomatic reading)

b. abura-oi Taro-ga ti hi-ni ti sosoida.

oil-acc Taro-nom fire-dat poured*Reading 1: �Taro made things worse.� (Idiomatic reading)Reading 2: �Taro poured some oil into the fire.� (Non-Idiomatic reading)

McGinnis (1998, 1999) suggests that the movement in (4) is blockedbecause the idiomatic interpretation must be available only when theindirect object is a DP, following the argument of Miyagawa (1997). Onthe other hand, when the direct object scrambles over the indirect object,the indirect object must be a PP, which does not serve as an intervener forA-movement. In this case, the idiomatic interpretation is absent.However, as McGinnis (1998, 1999) admits, some idioms do maintain

their meaning after the lower object is scrambled.

(5) a. Taro-ga te-oi hoterugyou-ni ti nobashita.

Taro-nom hand-acc hotel-business-dat extended�Taro became involved in the hotel business.� (Idiomatic reading)

b. Te-oi Taro-ga ti hoterugyou-ni ti nobashita.

Hand-acc Taro-nom hote-business-dat extended�Taro became involved in the hotel business.� (Idiomatic reading)

McGinnis points out that in the latter case only the direct object isinvolved with the idiom chunk, and she proposes that the indirect object

m

m m

mm

m

1 In this case, only the literal meaning of the sentence is maintained, i.e., �Taro poured theoil into the fire.�

36 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

in the latter case is a PP. Since it is not a DP, it does not block theA-movement of the lower object. However, we can find cases where asimilar type of idiom does lose the idiomatic interpretation when thedirect object is scrambled. (6) is another idiom where only the directobject is involved with the idiom chunk and which loses the idiomaticinterpretation when the direct object is scrambled.

(6) a. Isha-ga sono kanjya-ni saji-o nageta.doctor-nom the patient-dat spoon-acc threwReading 1: �A doctor gave up on the patient.� (Idiomatic reading)Reading 2: �A doctor threw a spoon at the patient.�(Non-idiomatic reading)

b. Isha-ga saji-oi sono kanjya-ni ti nageta.

doctor-nom spoon-acc the patient-dat threw*Reading 1: �A doctor gave up on the patient.� (Idiomatic reading)Reading 2: �A doctor threw a spoon at the patient.�(Non-idiomatic reading)

c. Saji-oi isha-ga ti sono kanjya-ni ti nageta.

spoon-acc doctor-nom the patient-dat threw*Reading 1: �A doctor gave up on the patient.� (Idiomatic reading)Reading 2: �A doctor threw a spoon at the patient.�(Non-idiomatic reading)

The existence of the idiom chunk as in (6) is unexpected by McGinnis�account. According to her account, the idiomatic interpretation of thesentence in (6) must have a PP as the indirect object because the indirectobject is not involvedwith the idiomatic interpretation just like the examplein (5). Therefore, the scrambling of the direct object should not affect theavailability of the idiomatic reading, contrary to the fact. As discussed inMiyagawa & Tsujioka (2004), there seem to be two types of idioms. One isthe type of idioms in which no chunk of the idiom can move, such as theexample in (6). The other allows a chunk to move via scrambling as in theexample in (5). If this distinction is real,McGinnis� account for the contrastof the (un)availability of idiomatic reading in scrambling is notmaintained.McGinnis� (1998, 1999) second argument for A-scrambling is based on

quantifier floating. Miyagawa (1997) observes that a direct object cannotprecede an indirect object when the indirect object is associated with afloating quantifier as in (7a). On the other hand, the sentence is acceptablewhen the direct object follows the indirect object associated with a floatingquantifier as in (7b) or when the direct object precedes the indirect objectwithout a floating quantifier as in (7c). Miyagawa (1997) analyzes thedative marker in (7a) as a postposition and the dative marker in (7b) as acase marker. He further claims that the NP associated with a floatingquantifier must have a case marker. Thus, (7a) is ruled out because TheNP with a postposition cannot be associated with a floating quantifier.

m

m m

A-movement in Japanese 37

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

(7) a. ?? Mary-ga CD-oi [tomodachi-ni futari] ti okutta.Mary-nom CD-acc friends-dat two sent.�Mary sent two friends a CD.�

b. Mary-ga [tomodachi-ni futari] CD-o okutta.Mary-nom friends-dat two CD-acc sent.�Mary sent two friends a CD.�

c. Mary-ga CD-oi [tomodachi-ni] ti okutta.Mary-nom CD-acc friends-dat sent.�Mary sent friends a CD.�

McGinnis (1998, 1999) argues that Miyagawa’s observations must beinterpreted as locality constraints on A-movement. Assuming that apostpositional phrase does not serve as an intervener for A-movement, sheshows that when the indirect object is without a floating quantifier, bindingof reciprocal anaphor is licensed via scrambling of the direct object as in(8a). It means that this scrambling is A-movement. Since the indirect objectwithout a floating quantifier can be a postpositional phrase, it does notblock A-scrambling. On the other hand, when the indirect object is with afloating quantifier, binding of the reciprocal anaphor via direct objectscrambling is not licensed as in (8b). Since the indirect object with a floatingquantifier must be a DP, it serves as an intervener for A-scrambling.

(8) a. Taroo-ga [gakusei-oi [otagai-noi sensei-ni] [ti

Taroo-nom student-acc each other-gen teacher-datsyookaisita]].introduced

�Taro introduced the student to each other’s teacher.�b. *Taroo-ga [gakusei-oi [otagai-noi sensei-nifutari] [ti

Taroo-nom student-acc each other-gen teacher-dat twosyookaisita]].introduced

(�Taro introduced the student to two of each other’s teacher.�)

However, it seems that her argument is not very strong. First, there seemsto be some difficulty to bind a reciprocal anaphor in a dative DP with afloating quantifier. Even though the antecedent is base-generated in thematrix subject position which is higher than a reciprocal element with afloating quantifier, the status of the sentence is not improved.

(9) *Taroo-to Mary-gai [[otagai-noi sensei-ni futari]Taro-and Mary-nom each-other-gen teacher-dat two

ryoushin-o syookaisita].parents-acc introduced.

�Taro and Mary introduced their parents to two of each other’steacher.�

m

m

38 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

The example in (9) indicates that something is wrong with licensing areciprocal anaphor in a dative DP with a floating quantifier, [otagai-nosensei-ni futari]. More specifically, the problem of the sentence seems tostem from the inability of a reciprocal taking scope over a floatingquantifier in a dative DP. It is therefore unclear that the unavailability ofa reciprocal binding relation in (8b) is due to the unavailability ofA-movement of the direct object.Second, the marginality observed in (7a) disappears when the DO

is to the left of a manner adverbial (e.g., Haig 1980, Miyagawa1997).2

(10) ? Mary-ga CD-o1 isoide tomodachi-ni futari t1 okutta.Mary-nom CD-acc in.haste friend-dat two sent�Mary sent two friends a CD in haste.�

The examples in (10), in which the DO appears to scramble over adative DP with a floating quantifier, cast doubt on McGinnis� claim thatthe marginality observed in (7a) is due to locality constraints onA-movement.

2 An anonymous reviewer points out that the example in (10) does not have to be aproblem for McGinnis� analysis because the movement may involve A¢-movement. This isbasically Miyagawa’s (1997) treatment of the example in (10). Miyagawa (1997) argues thatthe movement observed in (10) is focus movement but not A-movement. Miyagawa (1997)claims that the following example lends a support to his argument.

(i) a. ?? John-ga isoide hon-wa katta.John-nom quickly book-FOC bought�John quickly bought a book.�

b. John-ga hon-wa isoide katta.John-nom book-FOC quickly bought.

The objects in (i) bear a contrastive focus marker, -wa. They should therefore be construedas a focus. Miyagawa (1997) attributes the contrast in acceptability in (i a) to the availabilityof the position for a focus element. That is, (i b) indicates that there is a focus positionbetween the subject and the VP, marked by the manner adverb, and the unacceptability of(i a) suggests that a focus element must overtly move to the focus position. Based on thisobservation, Miyagawa (1997) concludes that the scrambling observed in (10) is movementto this focus position.However, Takano (1998) contends that this is not the case. He claims that these types ofsentences can be uttered in a completely neutral context. That is, the scrambled DO need notto be construed as a focus. Furthermore, he points out that example (ii), in which thescrambled DO does not move out of the VP to the focus position under Miyagawa’sassumption, is acceptable.

(ii) John-ga isoide gakusei-oi Bill-ni hutari ti syookaisita.John-nom quickly students-acc Bill-dat two introduced�John introduced two students to Bill quickly.�

In this example, the floating quantifier must be associated with the trace of the DO. Hence,the DO should scramble over the IO. But the landing site is not the focus position. Fromthis, Takano (1998) claims that the scrambling of the DO over the IO is not involved withfocalization. Given Takano’s (1998) claim, Miyagawa’s (1997) treatment of scramblingobserved in (10) as focus movement is untenable.

A-movement in Japanese 39

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

In sum, these two main arguments for locality in A-movement inJapanese scrambling made by McGinnis (1998, 1999) are significantlyweakened.3 In my view, it is very hard to show that locality constraintson A-movement exist in Japanese scrambling.4 In order to show localityconstraints on A-movement in Japanese, I will examine a differentsyntactic operation: passivization, whereby I will claim that localityconstraints on A-movement in Japanese interact with phases.

3. Applicatives and phases

There is an asymmetry in the syntax of applicative constructions withinpassivization (Baker 1988, Bresnan & Moshi 1990, Marantz 1993). Someapplicative constructions like Kinyarwanda Benefactives allow not only�short� passives in which the higher object (Benefactive) moves to thematrix subject position as in (11a) but also �long� passives in which thelower object (Theme) is raised to the subject position as in (11b).

(11) a. Umukoobwai a-ra-andik-ir-w-a tigirl sp-pr-write-appl-pass-asp

ibaruwa n�umuhuungu.letter by.boy

�The girl is having the letter written for her by the boy.�b. Ibaruwai i-ra-andik-ir-w-a umukoobwa ti

letter sp-pr-write-appl-pass-asp girln�umuhuungu.by.boy

�The letter is written for the girl by the boy.� (Kimenyi 1980)

3 An anonymous reviewer points out that McGinnis� (1998, 2004) lethal ambiguity servesas another argument for locality in A-scrambling. The gist of lethal ambiguity is that twoDPs which merge with the same head cannot be linked with each other. McGinnis (1998,2004) shows that Japanese A-scrambling is subject to lethal ambiguity. The scrambled objectcannot bind the subject itself.

(i) *Hiroshi-oi karejishin-gai ti waratta.Hiroshi-acc self-nom laughed(�Hiroshi laughed at himself.�)

However, it is unclear at all whether this suggests that Japanese A-scrambling obeys localityconstraints on A-movement, as the scrambled DP can bind the anaphor inside the subject asshown in (ii).

(ii) Hiroshi-oi [karejishin-noi hahaoya]-ga ti waratta.Hiroshi-acc self-gen mother-nom laughed�His own mother laughed at Hiroshi.�

Thus, lethal ambiguity appears to be a constraint on anaphoric licensing but not onA-movement. The example in (i) at most indicates that the landing site of the scrambledobject is under the same projection of the head that takes the subject as its specifier

4 Notice that all I did here is to show that the evidence McGinnis (1998, 1999) uses for herclaim that A-scrambling obeys locality is inconclusive because it involves different factorsthan locality. The whole claim in this paper does not rule out the possibility thatA-scrambling obeys locality constraints.

40 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

Others, such as the double object construction in English only allow theIO (¼indirect object) to be passivized as shown in (12).

(12) a. Alicei was baked ti a cake.b. *A cakei was baked Alice ti. (Oehrle 1976)

McGinnis (2001, 2004), who follows Pylkkanen’s (2001, 2002) proposalthat some Appl heads take VP as their complements (¼High Appl) andother appear as complements toV (¼LowAppl), argues that the contrast in(12) is accounted for ifHighAppls are phases but not LowAppls as in (13).5

(13) a. High applicative b. Low applicative

ApplP VP

IO Appl′ V ApplP

Phase Appl VP IO Appl′

V DO NOT phase Appl DO

Why can a High Appl head a phase while a Low Appl cannot? McGinnis(2001, 2004) suggests that what can count as a phasal head depends on itsphrasal status in a given phrase marker. The gist of her proposal is shownin (14).

(14) i) Different XPs may count as phases depending on their statusin the phrase structure.

ii) X is a phase if the head X is the sister of VP and an argumentis generated in its specifier.

According to (14ii), High Appl is a phase because it is the sister of VP andgenerates an argument in its specifier. On the other hand, Low Appl is nota phase because it is not the sister of VP. That is, whether a given Applhead can be a phase or not depends on its status in the phrase structure.Given McGinnis� (2001, 2004) proposals, the availability of both long

passive and short passive in (11) follows. Once the high applicative is aphase, which Chomsky (1998) assumes to have an EPP-feature, the DOcan move to the outer specifier of the high ApplP, leapfrogging over theIO because there is no intervener between the high applicative head(¼Probe) and the DO (¼Goal), as in (15a). Subsequently, the DO can

5 See Pylkkanen (2001, 2004) for a detailed distinction between these two types of ap-plicatives. I will however diverge from Pylkkanen’s (2001, 2004) definition of low Appli-cative and adapt the phrase structure proposed by Pesetsky (1995) instead. See thediscussion in section 6.

A-movement in Japanese 41

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

move from the outer specifier of the high ApplP to the matrix subjectposition as shown in (15b). Thus, �long� passive as in (11b) is available ina High Appl construction such as Kinyarwanda Benefactive.

(15) Long passive in High Appl construction via Leapfrog

a. DO’s movement to Spec ApplP b. DO’s movement to Spec TP

ApplP TP

DOi Appl′ DOi vP

IO Appl′ v ApplP

Appl VP ti Appl′(Phase)

V ti IO …..

…..

On the other hand, since Low Appl is not a phase, the DO must movedirectly from its base position to Spec TP. Since the IO is closer to T, itserves as an intervener to the movement. This explains why DO cannot bepassivized in Low Appl constructions such as English ditransitives.

(16) *DO’s movement to Spec TP

TP

DOi vP

v VP

V ApplP

the intervener Appl′

Appl ti

IO

In sum, the nature of the �phrasal/phasal� structure accounts for the(un)availability of �long� passive operation in an applicative construction.

4. The extension

4.1. Raising constructions as �Applicatives�

McGinnis (2001, 2004) extends this analysis to the availability of raisingover Experiencers. As shown in (17a), Icelandic allows the Experiencer to

42 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

move to the subject position of a raising verb but does not allow theembedded subject to be raised to the matrix subject position over theExperiencer as in (17b). On the other hand, such an operation is possiblein Italian as in (18).

(17) Raising over the experiencer in Icelandic: Out ¼ Low Appl

a. Jon telur [mer1 virðast t1 [HaraldurJon-nom believes me-dat seem Haraldur-nom

hafa gert þetta vel]]have done this well

�Jon believes Harald to seem to me to have done this well.�b. *Jon telur [Haralduri virðast mer [ti

Jon-nom believes Haraldur-nom seem me-dathafa gert þetta vel]].have done this well

(Thrainsson 1979)

(18) Raising over the experiencer in Italian: OK ¼ High Appl

Giannii non gli sembra [ti fare il suo dovere].Gianni not him-dat seemed do the his duty�Gianni seemed not to do his duty.� (Rizzi 1986)

McGinnis suggests that this contrast follows if the Experiencer construc-tion is a low applicative in Icelandic but a high Applicative in Italian.Since High Appl is a phase, the embedded subject in Italian raisingconstructions can move to the outer specifier of a HighApplP. It can thenmove to the matrix subject position. On the other hand, if the raisingconstruction in Icelandic is a low applicative, the embedded subject has tomove directly to the matrix subject position because LowApplP does nothave a phase-EPP feature. In this case, the Experiencer serves as anintervener for the movement.

4.2. Causatives as �Applicatives�

The previous subsection demonstrated that High/Low applicative con-structions can be extended to raising constructions.6 In fact, there is aprecedent to extend the applicative construction analysis to another kindof construction. Marantz (1993) analyzes double object constructionssuch as (19) as basically high applicative constructions in which the extraverbal head (¼V1) appears as the sister of VP (¼VP2), as in (20).7

6 See Boeckx (2003) for another argument that the Icelandic Experiencer constructionsare an applicative in a different perspective.

7 In fact, Marantz (1993) assumes that the matrix subject is generated in Spec IP. In theexample in (18), following recent theories of VP structure, the matrix subject is generated inthe specifier of the small v, which heads the VP. See Bruening (2001) for the same modifi-cation.

A-movement in Japanese 43

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

(19) John gave me a chocolate.

(20) vP

Subject v′John

v VP1

[Goal] V′1me

V1 VP2

V2 [Theme]gave a chocolate

Marantz (1993) applies the structure proposed in (20) to complexpredicates including causative structures, with the extra V introducinganother argument, namely an affected object, as in (21).

(21) vP

Subject v′

v VP1

[Affected] V′1Obj

V1 VP2

“CAUSE”V2 X

The motivation for Marantz’s (1993) proposal is that morphologicallyderived causatives in languages like Kinyarwanda appear to have asimilar construction to double object constructions with applicative affixor an affix verb. As in (22a, b), an affected object and the causee inKinyarwanda appear as the higher object. From this, Marantz (1993)analyzes the phrase structure of an affixal causative like Kinyarwanda in(22b) as the one in (23) in which the successive verb movement shouldoccur, to derive the surface verbal compound.

(22) a. Umugore a-ra-som-er-a umuhuungu igitabo.woman sp-pr-read-appl-asp boy book�The woman is reading the book for the boy.�

b. Umugabo a-r-uubak-iish-a abakozi inzu.man sp-pr-build-caus-asp workers house�The man is making the workers build a house.� (Kimenyi 1980)

44 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

(23) vP

Subjectumugabo

v VP1

[Affected] V′1Obj

abakoziV1 VP2

-iishi-V2

uubakr- inzu

v′

X

Given this structure, Marantz (1993) argues that only difference betweendouble object construction and the causative construction is the type ofthe extra V, V1 in (23). On the one hand, double object constructionshave the applied verb, while causative constructions have the causativeaffix on the other.8

The view that double object constructions and causatives are differentsides of the same coin is supported by Pesetsky (1995) and Bruening(2001), who point out that the IO in a double object construction inEnglish receives a causative reading as in (24a). That is, the Goal in thedouble object construction in (24a) is treated as an affected object. Thedative construction of the sentence in (24b) does not have a causativereading but a transferring reading, hence the sentence is awkward(Oehrle 1976). This indicates that a double object construction has anextra (null) causative verb, which its counterpart in a dative construc-tion lacks.

(24) a. The lighting gives me a headache.b. #The lighting gives a headache to me.

Given Marantz’s (1993) proposal, I suggest that morphologically derivedcausative constructions have a structural regulation similar to McGinnis�proposal for applicative constructions. That is, just like applicativemorphemes, some causative morphemes project a causative phrase as acomplement of V, corresponding to low applicatives. Some take VP astheir complement, corresponding to high applicatives.

5. Japanese A-locality in morphologically derived causatives

Given the proposal that morphologically derived causative constructionshave a structural regulation similar to McGinnis� proposal for applicative

8 From this, Marantz (1993) further claims that double object constructions are merelyone of a general class of complex predicate constructions.

A-movement in Japanese 45

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

constructions, this section will show that locality constraints onA-movement exist in Japanese �applicative� constructions and that, moreimportantly, these constraints interact not only with phrasal structuresbut also with �phasal� structures.9

5.1. Morphologically derived causatives in Japanese

Japanese causatives employ a causative affix –(s)ase just like theirKinyarwanda counterpart, as in (25).

(25) a. Umugore a-r-uubak-iish-a abakozi inzu.man sp-pr-build-caus-asp workers house�The man is making the workers build a house.� (Kinyarwanda)

b. Sono hito-ga daiku-ni ie-o tate-sase-ta.the man-nom carpenter-dat house-acc build-caus-pst.�The man made carpenters build a house.� (Japanese)

There has been debate over whether this type of causative construc-tions is derived in the syntax or the lexicon. I assume here that thesemorphologically derived causatives are generated in the syntax wherethe causative morpheme is independently introduced in the phrasemarker and the causative form of the verb is derived throughhead-movement. There is one piece of evidence that -(s)ase causativesare generated in the syntax. Shibatani (1976) observes that thesemorphologically derived causatives can have ambiguous adverbialscope.

(26) Taro-wa Hanako-o heya-ni damatte hair-ase-ta.Taro-top Hanako-acc room-to silently enter-caus-pst�Taro silently made Hanako enter the room.�Interpretation 1: Taro was silent when he caused her to enter the room.Interpretation 2: Hanako silently entered the room.

In one interpretation, Taro was silent when he caused her to enter theroom. In the other interpretation, Hanako silently entered the room. Byway of contrast as shown below, a lexical causative can have only oneinterpretation. In (27) it was Taro who was silent when he put her into theroom. That is, only the causer’s activity can be modified by the adverbialphrase.

9 See Kitagawa (1994) for another argument that the Japanese ditransitive verbs are atype of causative predicates selecting a verbal complement, in a different perspective.

46 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

(27) Taro-wa Hanako-o heya-ni damatte ire-ta.Taro-top Hanako-acc room-to silently put-pst�Taro silently put her into the room.�Reading 1: Taro was silent when he caused her to enter the room.

*Reading 2: Hanako silently entered the room.

This difference follows if the morphological causative projects independ-ently from the caused event. The adverbial phrase can modify either acaused event or a causative morpheme head in a phrase marker. This iswhy the two interpretations are available in (26). On the other hand, alexical causative construction does not provide a phrase marker thatdenotes the caused event. Therefore, only the matrix predicate can bemodified by the adverbial phrase in (27).

5.2. Long passives in causatives

Consider the contrast in (28), which shows that Japanese causatives onlyallow the affected object, daiku in (28b), to be passivized. This indicates thatthe Japanese causative is not High Appl but Low Appl. The unavailabilityof �long� passive in (28c) is attributed to the intervention effect of theaffected object that was seen in Low Appl (Marantz 1984 Inoue 1976).10

(28) a. Taro-ga daiku-ni ie-o tate-sase-ta.Taro-nom carpenter-dat house-acc build-caus-pst�Taro made carpenters build a house.�

b. Daikui-ga ti ie-o tate-sase-rare-ta.carpenter-nom house-acc build-caus-pass-pst�Carpenters were made to build a house�

c. *Iei-ga daiku-ni ti tate-sase-rare-ta.house-nom carpenter-dat build-caus-pass-pst(‘Lit.: �A house was made carpenters to build.�)

This line of thought allows one to expect the existence of languages withmorphologically derived causative constructions which allow �long�passive as well as �short� passive. As was seen in previous sections, thereis an asymmetry in the availability of �long� A-movement when anelement A-moves over an apparent intervener. In applicative construc-tions, Kinyarwanda Benefactives allow �long� passives in which the lowerobject can be passivized, stepping over the higher object while Englishditransitives do not. In raising constructions, Italian allows the raising of

10 In fact, Inoue (1976) and Kageyama (1999) have already tried to explain this fact interms of locality effects of A-movement. However, I will show that locality constraints onA-movement have to interact with phrase and phasal structure as I argue here in order toaccount for the contrast between causative constructions and control constructions in termsof the (un)availability of long passives.

A-movement in Japanese 47

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

the embedded subject over the Experiencer while Icelandic does not. Ifthis cross-linguistic contrast found in both applicative and raisingconstructions is due to the nature of phrasal and �phasal� status ofapplicative heads, we should see cross-linguistic contrasts in themorphologically derived causative constructions as well since I claimthese involve applicatives. The paradigm is demonstrated in the table in(29), where it is expected that there be cross-linguistic evidence for LongA-movement causatives.

The prediction is borne out; Kichaga is one of those languages. Kichagacausative constructions allow not only �short� passives but also �long�passives as shown in (30).

(30) a. Ndesambulro n-a-i-zrem-ilr-a mana muinda.Ndesambulro foc-agr-pr-cultivate-caus-fv child farm�Ndesambulro is causing that child to cultivate the farm.�

b. Mana n-a-le-zrem-ilro-o muinda.child foc-agr-pst-cultivate-caus-pass farm�The child was caused to cultivate the farm.�

c. Muinda u-i-m-zrem-ilr-ofarm agr-pr-him-cultivate-caus-passLit.: �The farm is caused him to cultivate.� (Alsina 1992)

Another example is Korean, which also allows both short and longpassive operations in causative constructions (Youngmi Jeong p.c.).Kichaga and Korean causative constructions can be analyzed to be HighAppl.

(31) a. Haksengdul-i notebook-ul sa-kae/torokstudents-nom notebook.computer-acc buy-caus

ha-o-chi-et-da.do-pass-pst

�Students were caused to buy a notebook computer.�b. Notebook-i haksengdul-ekey sa-kae/torok

notebook.computer-nom students-dat buy-causha-o-chi-et-da.do-pass-pst

Lit.: �A notebook computer was caused students to buy.�

Applicatives Raising Causative

Long A-movement Kinyarwanda Italian ?

*Long A-movement English Icelandic Japanese

(29)

48 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

5.3. On long passives in Japanese: Nishigauchi (1993)

In the previous section, I claimed that the unavailability of a long passiveoperation in Japanese morphologically derived causatives is due tolocality constraints on A-movement. However, there is a possiblealternative explanation of the phenomena, Case theory. One could saythat due to case absorption by the passive morpheme, the higher objectdoes not get its Case-feature checked. This would be a Case-filterviolation, and the lower verb cannot check its uninterpretable Casefeature, i.e., Inverse Case-filter violation as in (32).

(32) The computeri was made students to buy ti

*case *Inversecase

Note that this account should not be general because there are languageswhich allow this long passive operation in applicative constructions asshown above. Specifically, the direct object can manifest object agree-ment with the verb via �pronoun incorporation�, i.e., the phenomena inwhich the object pronoun that is indicated �OP� in the gloss can beincorporated with the verb stem, in Chaga long passives as in (33b).

(33) a. N-a-i-ki-m-lyi-i-a.foc-sp-pr-op-op-eat-appl-fv�He is eating it for him/her.�

b. K-i-m-lyi-i-o.sp-pr-op-eat-appl-pass�It is being eaten for/on him/her.� (Bresnan & Moshi 1990)

Assuming that the Case property is a reflex of agreement, (33b) indicatesthat the Case �assigner� of the higher object, i.e., v, maintains its Case-feature even after (long) passivization. Furthermore, the lower objectbeing raised to the subject shows that it gets its case checked by T. Thismeans that the Case-feature of the Case �assigner� of the lower object in thebase position is suppressed. That is, in the case of long passives, thepassivized verb retains its Case feature but the lower Case �assigner� gets itsCase feature suppressed, as proposed by Ura (1996) andMcGinnis (1998).It has been shown that in Japanese, Case suppression can apply to the

Case �assigner� of the lower object in a complex construction. Nishigauchi(1993) observes that long passives in Japanese do exist in controlconstructions, or to put it more precisely, in aspectual verbal construc-tions, which involve control structures.

(34) Kono hon-ga yomi-hajime-rare-ta.this book-nom read-begin-pass-pstLit.: �This book was begun to read.�

m m

A-movement in Japanese 49

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

This shows that Case suppression by Japanese passive morpheme canapply to the Case �assigner� of the lower object, penetrating the passivizedverb.11 From this, I assume that in Japanese long passives, the Case�assigner� of the higher object should maintain its Case-feature afterpassivization and the lower object has to move to the matrix subjectposition for Case-checking. If this is true, it excludes a Case-theoreticexplanation of the unavailability of long passives in Japanese causativeswhich says that the lower verb cannot check its uninterpretable Casefeature and the higher object does not get its Case-feature checked.The next question is why causative constructions do not allow long

passives while control constructions do. If both Japanese control andcausative constructions take a verb phrase as their complement, as isgenerally assumed, it would be predicted that both should allow longpassives as well. Given Chomsky’s (1998) phases, the object in the lowerpredicate in control constructions can move to the matrix subject positionvia multiple Specs of the lower vP, which is a phase.

(35) Long passive in Japanese control predicate.

TP

DPi T′

T vP

v VP

V vP

ti v′

PRO v′

v VP

V ti

11 Specifically, in order to explain these long passive phenomena, Nishigauchi (1993)proposes that Case-absorption requires morphological integrity. That is, in order for thepassive morpheme to absorb a verb’s ability to assign Case, it must be integrated with theverb. Given this, in the case of (34), successive head-movement forms the complex, verb, i.e.,yomi-hajime-rare (read-begin-pass). This enables the passive morpheme to absorb the Caseof the verb, yomi (read). On the other hand, in the English counterpart of (34), the passivemorpheme is attached to the matrix verb and the head of the lower verb does not move tothe higher verb.

(i) *This book was begun to read.

The passive morpheme cannot absorb the lower verb’s Case in (i). The complement of thelower verb can check its Case with the verb. Hence it does not move to the matrix subjectposition. The long passive construction is excluded.

50 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

If a causative morpheme heads the lower clause like control constructionsdo, the causative construction should allow a long passive operation thesame way control constructions do, contrary to fact as shown in (36):

(36) Long passive in Japanese causative-available. A wrong prediction

TP

DPi T′

T CausP

Cause vP

ti v′

DP v′

v VP

V ti

It is therefore unclear why causatives do not allow long passives whilecontrol constructions do, if both Japanese control constructions andcausative constructions head the lower clause.12 The current proposal cancapture the difference between control and causative constructions interms of the availability of long passive.

(37)

TPTP

DPi T′ T′

T vPT vP

v VP v VP

V vPV CausP

ti v′The intervener=DP

PRO v′ Caus DPv VP

V ti

b. *Long passive in causativesa. Long passive in control predicates

Caus′

12 Kageyama (1999) also attempts to account for this contrast due to locality constraintson A-movement. His account diverges from mine as to why control predicates allow longpassives in that, in his analysis, there is no PRO in the specifier position when controlpredicates allow a long passive operation. Hence no intervention effects arise. Given hisanalysis, however, it is unclear how the external theta role of the embedded verb is dis-charged. Furthermore, Aoshima (2001) shows that in Japanese, there is a PRO in Spec vPand that it has properties of Obligatory Control PRO. Given Aoshima’s (2001) analysis,Kageyama’s (1999) account for why control predicates allow long passives is untenable.

A-movement in Japanese 51

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

As defined in (14ii), a given head is a phase if it is the sister of VP,generating an external argument. In (37a), the lower v is a sister of VP,generating an external argument; hence it is a phase. The object in thelower predicate in a control construction can move to the matrix subjectposition via the multiple Spec of the lower vP. On the other hand, the factthat a long passive is not available in Japanese causative constructionsfollows if their structural status is similar to a low applicative construc-tion, as suggested in the previous section. As in (37b), a causative head isnot the sister of VP but the causative phrase that it projects appears as thecomplement of V. It is not a phase. The lower object has to move directlyto Spec TP. The higher object serves as an intervener for the movement.In sum, the unavailability of the long passive in Japanese causatives is

due to locality constraints on A-movement. Furthermore, these localityconstraints on A-movement interact with �phrasal/phasal structure� asproposed by McGinnis (2001, 2004).

6. Semantic interpretation – Pesetsky’s (1995) causatives in Japanese

In the section above, I proposed that affixal causative constructions inJapanese share the same phrasal status as low applicatives. That is, thecausative morpheme projects in a position lower than the verbal phrase.This means that the causative morpheme head does not take an eventargument semantically. In fact, the caused �event� is not denoted as a unitin the phrase structure proposed above. Thus, the phrase structure in thecurrent proposal has a serious flaw in the semantic component. In orderto solve this semantic problem, we can adapt Pesetsky’s (1995) Japanesecausative structures. Pesetsky (1995) also suggests that Japanese causat-ive morphemes are base-generated below the verbal phrase. As discussedin (24) repeated as (38), Pesetsky (1995) points out that a causativeinterpretation disappears in the dative counterparts of ditransitiveconstructions.

(38) a. The lighting gives me a headache.b. #The lighting gives a headache to me.

Since generally assumed causative structures in which a causative verbembeds the caused event, fails to capture these correlations, Pesetsky(1995) proposes:

(39) (i) A (Zero) Causative morpheme head is generated belowLow ApplP (G affix in his sense).13

13 Based on the following examples, Pesetsky (1995) posits Cause (zero-morpheme), apreposition, which represents causation at the bottom of the phrase structure (See Pesetsky1995:196).

52 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

(ii) A Causative morpheme must move to V after adjunctionto G.14

(iii) Phrase structures are provided by Cascade theory inwhich internal h selection is established not only by therelation �sister of a� but also by the relation �specifier of thesister of a� (Phillips 1996, 2003).

Taken together, Pesetsky’s (1995) proposals in (39) generate the followingcausative structure in (40). As indicated, the causative morpheme head isfirst adjoined to the Applicative head (¼G). Subsequently, it attaches tothe main verbal head.

(40) VP

V′

V ApplP

DP Appl′

G CausP

DP Caus′

Caus DP

The reason why dative constructions lose the causative interpretation isthat the preposition, to, which is not an affix, intervenes with themovement of the causative morpheme to the verbal head. That is, theHead Movement Constraint blocks the movement of the causative headto the verbal head.

(i) a. Sue yelled out of frustration.b. Mary objected to the show because of Bill’s remarks.c. Mary jumped for joy.

14 Pesetsky’s (1995) motivation to assume that Cause is an affix, and it must be attachedto the main verb by PF comes from the causative morpheme, en–, in enlarge, embitter,endear and so on. See Pesetsky (1995:179–189) for a detailed explication of internal selectionfor Cascade structures.

A-movement in Japanese 53

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

(41) VP

V′

V ApplP

DP Appl′ *

to CausP

DP Caus′

Caus DP

Notice that in the tree diagram (40, 41), Pesetsky (1995) proposesthat the Causer is the complement of the Causative morpheme head.15

The backward binding facts in both English ditransitive and Japanesecausative constructions suggest this to be the case. As in (42a) and(43a), the anaphoric element, each other, and the bound pronominalelement, soitsu, which are apparently out of their binder’sc-commanding domain can be licensed in their base position, i.e., ascomplements of Cause as illustrated in (42b) and (43b).16

(42) a. Each otheri�s remarks gave [John and Mary]i a book.b. gave [John and Mary]i a book [Caus [Each otheri�s remarks].

(43) a. [Mary-ga soitsu-no warukuchi-o itteru] –koto-gaMary-nom that.man-gen abuse-acc says –fact-nom

daremoi-o kowagar-ase-ta.everyone-acc frighten-caus-pst

�The fact that Mary speaks ill of him made everyone frightened.�b. Daremoi-o kowagar-[Caus-ase[Mary-ga soitsui-no warukuchi-o

itteru –koto]].

15 One might wonder how the Causer generated as the complement of Cause can move upto the matrix subject position, skipping intervening DPs. Pesetsky (1995) suggests that theCauser moves to the external argument position of the main verb, where another Causer-role is licensed, and that the movement is �novel� in a sense that movement is among thesame theta-positions, i.e., one Causer-role position to another. If so, the movement is notakin to ordinary A-movement. Presumably, the main verbal head probes the nominal withCauser-role as its external argument. Given Pesetsky’s (1995) suggestion, the interveningDPs with other kind of theta role would not serve an intervener for this operation. SeeBoeckx (2002), Boeckx & Jeong (2003) and the discussion in section 7 for Agree/Move via astrict view on feature type.

16 I am indebt to Tomokazu Takehisa for providing example (43).

54 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

Pesetsky’s (1995) proposal is compatible with the analysis here. TheCausP is selected by an applicative head in Pesetsky’s (1995) analysis.On the other hand, I analyzed a Causative head as an applicative head.This difference is trivial in terms of the interaction between anintervention effect and phases. In Pesetsky’s (1995) phrase structure,the higher object in Spec Low ApplP serves as an intervener for theraising of the lower object in Spec CausP as in (44) because neither thehead of Low applicative nor of Cause internally selects the VP. Recallthat McGinnis (2001, 2004) defines a phasal head as the sister of the VP.To be the sister of the VP requires that a (phasal) head internally selectsthe VP. Thus, neither Low applicative nor Cause in (44) can be a phasewhich it is assumed would offer an escape hatch for leapfrog. Thus, thelower object must move directly to Spec TP in passivization. GivenPesetsky’s (1995) causative structure, the current analysis predictsthat there should be an intervention effect in Japanese causativeconstructions.

(44) * V′

V ApplP

DP ApplP′

Appl CausP

DP Caus′

Caus …..

Furthermore, Duk-ho An (p.c.) points out that a Korean counterpartof (43a) does not allow the intended backward variable binding, contraryto Japanese.

(45) *Mary-ka ku1-lul helttutnun kes-i motunMary-nom he-acc speak.ill comp-nom every

saram1-ul nolla-key-hay-ss-ta.people-acc surprise-caus-do-pst

�That Mary speaks ill of him surprised everyone.�

This difference follows from the current analysis. I claim that, since itdoes not allow long passive operation, Korean morphologically derivedcausatives must be a High Applicative, which takes VP, but not the

A-movement in Japanese 55

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

causer, as its complement. If so, its structure should be in line with whatMarantz’s (1993) proposes in (21) repeated as (46) where the subject,namely the causer, is base-generated in the specifier of vP (see also Jung(2004) for a similar analysis). Given this type of structure, the causer isnever c-commanded by the lower object in the course of the derivation.Therefore, the bound pronoun in the subject cannot be licensed inKorean causative constructions like (45).

(46) vP

Subject

v VP1

[Affected] V′1Obj

V1 VP2‘CAUSE’

V2 X

v′

In this section, I showed that Pesetsky’s (1995) proposal on Japanesecausative constructions is consistent with my proposal. The availabilityof backward variable binding in Japanese causative constructionssuggests that the causative morpheme head is base-generated lowerthan a verbal head. Furthermore, the difference between Japanese andKorean backward variable binding phenomena indicates that thesecausative constructions are different from one other. This correspondsto the (un)availability of a long passive operation in causativeconstructions in both languages. These two differences follow if Japanesecausatives are Low applicatives and Korean causatives are Highapplicatives.

7. Implications: On ditransitive structures in Japanese, intervention

effects: Icelandic, and Agree vs Move

This paper claims that Japanese A-movement obeys locality con-straints and should interact with �phrasal/phasal structure� as McGinnis(2001, 2004) proposes for applicative constructions. It also supportsMarantz’s (1993) proposal that causatives and double object construc-tions share the same structure with the affected or indirect object as thehigher object. If this is the case, it suggests that IO sits higher than DOin Japanese. This implication is in favor of Hoji (1985) and Boeckx &Niinuma (2004), who claim that <IO;DO> be the base order inJapanese on independent grounds. Further, Boeckx & Niinuma (2004)likened the constraints on Japanese honorification to the intervention

56 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

effects in Icelandic Quirky subject constructions in which dativesubjects prevent the verb from agreeing with a nominative element.In Japanese, the ditransitive verb cannot agree with the DO inhonorification as in (47b) (Harada 1976). In this sentence, only thedative nominal, Mary is honorified. Boeckx & Niinuma (2004) attributethe failure of object honorification in (47b) to the intervention effectof the dative element in Chomsky’s (1998) Agree relation as well asthe intervention effects in Icelandic Quirky subject constructions in(47a).

(47) a. Mer fannst/*fundust henni leiðast þeir.me-dat seemed 3sg/3pl her-dat bore they-nom�I thought she was bored with them�[v dat nom] (Number agreement is blocked by the

dative nominal.)b. *Taro-ga Mary-ni Tanaka sensei-o go-syookai-si-ta.

Taro-nomMary-dat Prof.Tanaka-acc hon-introduce-pst�Taro introduced Prof. Tanaka to Mary.�[v dat acc] (Object honorification is blocked by the dative

nominal.)

In both Japanese and Icelandic constructions, a similar intervention effectcan be observed. As in (28c) repeated here in (48a), a dative elementserves as an intervener for the movement of the lower object in Japanesecausatives, now analyzed as a low Appl construction. Recall thatIcelandic does not allow the embedded subject to be raised to the matrixsubject position over the Experiencer, another dative element, as in (17b)repeated here in (48b).

(48) a. *Iei-ga daiku-ni ti tate-sase-rare-ta.house-nom carpenter-dat build-caus-pass-pstLit.: �A house was made carpenters to build.�[v dat acc] (Passivization of acc is blocked by the

dative nominal.)b. *Jon telur [Haralduri virðast mer [ti hafa

Jon-nom believes Haraldur-nom seem me-dat havegert þetta vel]]done this well�Jon believes Harald to seem to me to have done this well.�[v dat nom] (Raising of the embedded subject is blocked

by the dative nominal.)

m*

*m

m*

m

*

A-movement in Japanese 57

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

Further, recall that dative elements DO NOT serve as an intervener incausative constructions in Korean in (31) and in the raising of theembedded subject in raising constructions in Italian in (18). These cross-linguistic variations are accounted for in terms of a high/low applicativedistinction.Taken together with Boeckx & Niinuma’s examples, the same

intervention effect seems to be at work in both Japanese and Icelandic.That is, this paper confirms Boeckx & Niinuma’s claim that Japanese(honorific) agreement system is not different from the (U-) agreementsystem observed in other languages. Given that the similar interventionmechanism is involved in both Japanese (honorific) agreement systemand the passivization of Japanese causatives, the only differences betweenobject honorification and the passivization of causatives in Japanese are(i) that the probe of the former is a small v and the probe of the latter isT; and (ii) that the latter further requires the EPP. That is, the movementof the lower object is not available because an Agree relation fails due toan intervention effect.One thing to note is that unlike causative constructions, Japanese

ditransitive constructions do allow long passives.17

(49) a. John-ga Mary-ni hon-o okutta.John-nom Mary-dat book-acc sent�John sent Mary a book.�

b. Hon-ga Mary-ni okur-are-taBook-nom Mary-dat send-pass-pst.Lit: �A book was sent Mary�

Here I would like to assume that Japanese ditransitives are Highapplicatives. The lower object step over the higher object via the multipleSpec of HighApplP, following Ura’s (2000) proposal, which refers toHighApplP as VmidP.The question that follows is why passivization of the DO over the IO is

possible, when the DO cannot agree with v due to the intervention effectin honorification. I would like to suggest that the involvement of Casefeatures resolves this tension. As generally assumed, if A-movement isCase-driven, in the case of long passives, the DO is motivated to move toget its Case feature checked eventually with T. To do so, the DO mustleapfrog over the IO, moving to the outer specifier of high ApplP bymeans of a phasal EPP. On the other hand, if Honorific feature checkingis done solely by an Agree operation as Boeckx & Niinuma assume, thecommand relation between the IO and the DO never changes in

17 This confirms my claim in the section 5.3 that in Japanese the Case �assigner� of thehigher object should maintain its Case-feature after the passivization and the lower objecthas to move to the matrix subject position for its Case-checking.

58 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

ditransitive constructions when the honorification-feature of v probes itstarget.18 Since Agree is an antisymmetric operation in that a probe needsits unvalued feature to be valued by a goal’s interpretable feature, there isno need for the DO, a goal, to move up to the outer specifier ofHighApplP. The difference is schematized in (50) and (51). In (50), v canagree only with the IO because the IO is closer to v in terms c-command.On the other hand in (51a), the outer specifier of HighApplP offers theDO an escape hatch, allowing the DO to step over the IO. The DO canthen agree with T and move to Spec TP as in (51b).

(50) Honorific feature Agree

a. [v [HighApplP IO [High Appl [VP [V DO]]]]] (Agree between v and IO; OK)

Agreeb. [v [HighApplP IO [High Appl [VP [V DO]]]]] (IO serves as an intervener)

*

(51) Long passive

a. [HighApplP DO1 [IO [High Appl [VP [V t1]]]]] (DO first leapfrogs over IO)

b. [TP DO1 [T[vP[v[HighApplP t1 [IO [High Appl [VP [V t1]]]]] (DO moves to Spec TP)

In fact, Boeckx (2002) and Boeckx & Jeong (2003) observe similarasymmetry between Move and Agree in English raising constructions:

(52) a. Johni seems to Mary [ti to be nice].b. There seem to be men in the room.c. ?*There seem to Mary to be men in the room.

As in (52a), English allows overt raising over an Experiencer. On theother hand, the presence of an Experiencer blocks Agree as in (52c).Boeckx (2002) and Boeckx & Jeong (2002) claim that the feature whichenters into Agree at a distance and the one which is involved withmovement are different. Specifically, in English raising above, a [Person]feature is responsible for the movement to Spec TP and a [Number]

18 There is one more technical issue here. Since High Appl is a phase, its complement isnot accessible to the next phasal head (i.e. the Phase Impenetrability Condition). However,Chomsky (2001) suggests that the complement of a phase is still accessible to some higherhead before the next phase is established. Given this, in the case of ditransitives, v is not aphase because v is not a sister of VP (e.g. 16ii). Therefore the complement of High Appl, aphase is still accessible to v.

A-movement in Japanese 59

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

feature only enters into Agree at a distance. Given these assumptions,the Move/Agree puzzle found in raising over an Experiencer isaccounted for. The situation found with raising over an Experiencerin English is described as in (53) where a corresponds to [Number] andb to [Person].

(53) a. a…a…ab. ab…a…ab

Since the intervening element is eligible for [Number]-checking only,there should be no intervention effects when the relevant checkingoperation involves a [Person] feature. That is, a strict view of featuretype can resolve the Move/Agree puzzle found in English raisingconstructions. A different view can be taken to resolve the observedissue above. It could be assumed that English raising constructionsare a High Applicative construction just like Italian as in (18). Whenthe embedded subject moves overtly as in (52a), it leapfrogs anExperiencer so that it can agree with the head T there and eventuallymoves to Spec TP.19 On the other hand, when the relevant con-struction needs solely to Agree with [Number] with the expletiveinserted in the matrix subject position as in (52bc), the embeddedsubject stays in the base position. In this case, an Experiencer servesas an intervener for Agree.To sum up, while ditransitives unlike causative constructions do allow

long passives, I conclude that the same locality constraints are at work inboth honorification agreement in ditransitive constructions and inpassivazation of causative constructions. The phasal EPP feature onHighAppl enables the DO to step over the IO in long passives inditransitives. On the other hand, since honorification involves onlyAgree, no movement is implicated with the operation.

19 Alternatively, if the lack of the intervention effects of an Experiencer in English raisingconstructions is due to its nature of inherent Case but not to the phasal property of theApplicative head as McGinnis (1998, 2004) argues, it could be assumed that the embeddedsubject lands on the edge of the raising verb phrase, as claimed by Sauerland (2003),whereby the embedded subject can agree with the head T, and then moves to Spec TP asin (i).

(i) Every childi doesn’t [vP ti seem to his father to ti be smart].

This type of analysis is supported in terms of Sportiche’s (1988) stranding analysis offloating quantifier constructions, on which the quantifier is stranded in the trace position ofthe element which it modifies. Consider the example (ii) in which the quantifier is floatedbetween the tensed element and the raising verb. Given Sportiche’s (1988) stranding analysisof floating quantifier constructions, it shows that there is a landing site between the verb andthe head T in raising constructions.

(ii) The players have all seemed to gain weight.

60 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

8. Conclusion

Assuming morphologically derived causative constructions share thesame structural status with McGinnis� (2001, 2004) applicative construc-tions in the spirit of Marantz (1993), this paper provides evidence thatlocality constraints on A-movement exist in Japanese and that theyshould interact with �phrasal and phasal structures�. This analysisprovides a unified explanation to the (un)availability of long passives inboth causative and applicative constructions in various languages.Further, it is claimed that Japanese locality constraints, which are shownto correspond to the ones observed in Icelandic, are no different fromother languages. This conclusion is in accord with Boeckx & Niinuma(2004).

References

Alsina, A. 1992. On the argument structure of causatives. Linguistic Inquiry 23,517–555.

Aoshima, S. 2001. Mono-clausality in Japanese Obligatory Control Construc-tions. University of Maryland College Park. Ms.

Baker, M. 1988. Theta theory and the syntax of applicatives in Chichewa.Natural Language and Linguistic Theory 6, 353–389.

Boeckx, C. 2002. Patterns of subject raising across experiencers. University ofIllinois. Ms.

Boeckx, C. 2003. Intricacies of Icelandic agreement. Harvard University. Ms.Boeckx, C. & Niinuma, F. 2004. Conditions on agreement in Japanese. NaturalLanguage and Linguistic Theory 22, 453–480.

Boeckx, C. & Jeong, M. 2003. The fine structure of intervention in syntax. Issuesin Current Linguistic Theory: A Festschrift for Hong Bae Lee, ed. C. Kwon &W. Lee, 83–116. Seoul: Kyungchin.

Bresnan, J. & Moshi, L. 1990. Object asymmetries in comparative Bantu syntax.Linguistic Inquiry 21, 147–185.

Bruening, B. 2001. QR obeys superiority: frozen scope and ACD. LinguisticInquiry 32, 233–273.

Chomsky, N. 1998. Minimalist inquiries: The framework.MIT Occasional Papersin Linguistics 15, Cambridge, MA: Department of Linguistics and Philosophy,MIT.

Chomsky, N. 2001. Derivation by phase. Ken Hale: A life in language, ed. M.Kenstowicz, 1–52. Cambridge, MA: MIT Press.

Haig, J. 1980. Some observations on quantifier float in Japanese. Linguistics 18,1065–1083.

Harada, S.-I. 1976. Honorifics. Syntax and Semantics 5, ed. M. Shibatani, 499–561. New York: Academic Press.

Hoji, H. 1985. Logical form constraints and configurational structures inJapanese. University of Washington. PhD diss.

Jung, D. 2004. Conspiracy in Korean modal verbs in light of v. Minimalist viewson interfaces. Proceedings of the 6th Seoul International Conference onGenerative Grammar, ed. R. Kim, 71–96. Seoul: The Korean GenerativeGrammar Circle.

Kimenyi, A. 1980. A relational grammar of Kinyarwanda. Berkeley: University ofCalifornia Press.

A-movement in Japanese 61

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

Inoue, K. 1976. Henkei bunpoo to Nihongo [Transformational Grammar andJapanese]. Tokyo: Taishukan.

Kageyama, T. 1999. Word formation. Handbook of Japanese linguistics, ed. N.Tsujimura, 297–325. London: Blackwell.

Kitagawa, Y. 1994. Shells, yolks, and scrambled e.g.s. NELS 24, 221–239.Kuroda, S.-Y. 1988. Whether we agree or not: A comparative syntax of Englishand Japanese. Linguistic Investigations 12, 1–47.

Marantz, A. 1984. On the nature of grammatical relations. Cambridge, MA: MITPress.

Marantz, A. 1993. Implications of asymmetries in double object constructions.Theoretical aspects of Bantu grammar, ed. S. Mchombo, 113–151. Stanford CA:CSLI Publications.

McGinnis, M. 1998. Locality in A-movement. MIT. PhD diss.McGinnis, M. 1999. Evidence for feature-driven A-scrambling. WCCFL 18, 364–378.

McGinnis, M. 2001. Variation in the phase structure of applicatives. LinguisticVariations Yearbook 1, ed. P. Pica & J. Rooryck, 101–142. Amsterdam: JohnBenjamins.

McGinnis, M. 2004. Lethal ambiguity. Linguistic Inquiry 35, 47–95.Miyagawa, S. 1997. Against optional scrambling. Linguistic Inquiry 28, 1–25.Miyagawa, S. 1999. Causatives. Handbook of Japanese linguistics, ed. N.Tsujimura, 236–268. London: Blackwell.

Miyagawa, S & Tsujioka, T. 2004. Argument structure and ditransitive verbs inJapanese. Journal of East Asian Linguistics 13, 1–38.

Nishigauchi, T. 1993. Long distance passive. Japanese syntax in comparativegrammar, ed. N. Hasegawa, 79–114. Tokyo: Kuroshio Syobou.

Oehrle, R. 1976. The grammatical status of the English dative alternation. MIT.PhD diss.

Pesetsky, D. 1995. Zero Syntax: experiencers and cascades. Cambridge, MA:MIT Press.

Phillips, C. 1996. Order and structure. MIT. PhD diss.Phillips, C. 2003. Linear order and constituency. Linguistic Inquiry 34, 37–90.Pylkkanen, L. 2001. What applicative heads apply to. Proceedings of the 24thPLC. Department of Linguistics, University of Pennsylvania.

Pylkkanen, L. 2002. Introducing arguments. MIT. PhD diss.Richards, N. 1997. What moves where when in which language? MIT. PhD diss.Rizzi, L. 1986. On chain formation. The syntax of pronominal clitics: Syntax andsemantics 19, ed. H. Borer, New York: Academic Press.

Rizzi L. 1990. Relativized minimality. Cambridge, MA: MIT Press.Saito, M. 1989. Scrambling as semantically vacuous A¢-movement. Alternativeconceptions of phrase structure, ed. M. Baltin & A. Kroch, 182–200. Chicago:Chicago University Press.

Saito, M. 1992. Long distance scrambling in Japanese. Journal of East AsianLinguistics 1, 69–118.

Sauerland, U. 2003. Intermediate adjunction with A-movement. LinguisticInquiry 34, 308–314.

Shibatani, M. 1976. Causativization. The grammar of causative construction.Syntax and semantics 6, ed. M. Shibatani, 239–294. New York: Academic Press.

Sportiche, D. 1988. A theory of floating quantifiers and its corollaries for con-stituent structure. Linguistic Inquiry 19, 425–450.

Takano, Y. 1998. Object Shift and scrambling. Natural Language and LinguisticTheory 16, 817–889.

Thrainsson, H. 1979. On complementation in Icelandic. New York: Garland.

62 Hirohisa Kiguchi

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.

Ura, H. 1996. Multiple feature-checking: A theory of grammatical function-splitting. MIT. PhD diss.

Ura, H. 2000. Checking theory and grammatical functions in Universal Grammar.Oxford: Oxford University Press.

Received May 19, 2004Accepted Mar 21, 2005

Hirohisa KiguchiMiyagi Gakuin Women’s University

9-1-1 Sakuragaoka, Aoba, Sendai 981-8557Japan

[email protected]

A-movement in Japanese 63

� The author 2006. Journal compilation � The Editorial Board of Studia Linguistica 2006.


Recommended