1 Introduction

One of the characteristic properties of Japanese is that it allows several cases of Case alternation. Accusative-Nominative alternation such as (1) is one such case. An object can be marked with nominative Case ga when a transitive verb is accompanied by the potential suffix (rar)e (Kuno 1973).

    1. (1)
    1. Taroo-ga
    2. Taroo-NOM
    1. eigo-o/ga
    2. English-ACC.NOM
    1. hanas-e-ru.
    2. speak-can-PRES
    1. ‘Taroo can speak English.’

The construction has received much attention, mainly due to the scope asymmetry given in (2). Tada (1992) observes that the nominative object (henceforth NO) takes wide scope over the potential suffix in (2b), unlike the accusative object (henceforth AO) in (2a). In order to capture the wide scope reading of the NO, various approaches have been proposed in the literature (Tada 1992; Koizumi 1994; Saito & Hoshi 1998; Takano 2003; Nomura 2005; Saito 2010; Takahashi 2010 among many others).

    1. (2)
    1. Tada (1992: 94; with his judgement)
    1.  
    1. a.
    1. John-ga
    2. John-NOM
    1. migime-dake-o
    2. right.eye-only-ACC
    1. tumur-e-ru.
    2. close-can-PRES
    1. ‘John can close only his right eye.’
    2. (i)      can > only (John can wink his right eye.)
    3. (ii) ?*only > can (It is only his right eye that he can close.)
    1.  
    1. b.
    1. John-ga
    2. John-NOM
    1. migime-dake-ga
    2. right.eye-only-NOM
    1. tumur-e-ru.
    2. close-can-PRES
    1. ‘John can close only his right eye.’ (i) *can > only (ii) only > can

This paper focuses on the following variant of the NO-construction, which exhibits a similar scope asymmetry.

    1. (3)
    1. Takano (2003: 825; with his judgement)
    1.  
    1. a.
    1. John-wa
    2. John-TOP
    1. migime-dake-o
    2. right.eye-only-ACC
    1. tumur-u
    2. close-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘John can close only his right eye.’ (i) can > only (ii) ?*only > can
    1.  
    1. b.
    1. John-wa
    2. John-TOP
    1. migime-dake-ga
    2. right.eye-only-NOM
    1. tumur-u
    2. close-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘John can close only his right eye.’ (i) *can > only (ii) only > can

Although this variant has been discussed by Kuno (2002), Takano (2003), and Hiraiwa (2005: 186), it has received less attention and its syntactic behavior remains to be relatively unexplored, compared to its complex predicate counterpart.1 In contrast to the complex predicate type such as (1) and (2), the matrix predicate is a morphologically independent verb deki ‘can’ and takes a verbal complement headed by the nominalizer koto in (3). The reason why this paper focuses on the variant given in (3) is that in the complex predicate type, the potential suffix and the verb constitute a complex predicate, which limits the application of some syntactic tests that otherwise could have been applied. Since the variant in (3) is free from such a morphological integrity problem, it will be possible to reveal any syntactic properties which we could not find based on the complex predicate type. In fact, taking a closer look at the variant in (3), this paper will provide novel pieces of evidence which show that the NO overtly undergoes movement from its base-generated position, as independently argued by Kuno (2002).

Investigating the nature of the overt movement of the NO is another goal of this study, which pursues the hypothesis that the relevant movement is scrambling, contrary to the general assumption that nominals have to retain their Case after scrambling. Departing from this general assumption, this paper proposes that the construction under investigation involves scrambling of an object to the position where nominative Case can be assigned. On the other hand, when scrambling does not take place, accusative Case is assigned to the object at the base-generated position. In other words, the relevant Case alternation is contingent on the application of scrambling. The precursor of the analysis is Fukui & Nishigauchi (1992) and Fukui (1995), where scrambling is involved in the so-called Nominative Genitive Conversion, wherein a genitive subject is scrambled to the point where genitive Case can be assigned. This paper revives this idea in analyzing NOs.

Before beginning a detailed discussion, an important issue will be mentioned here. Recall that Tada (1992) observes that the narrow scope reading of the NO is unavailable in (2b). Contrary to Tada, Nomura (2005) claims that the relevant narrow scope reading is allowed under an appropriate context (see Nomura 2005 for details). Following Nomura, this paper takes the position that the relevant narrow scope reading is available in (2b). Similarly, I find the narrow scope reading of the NO available in (3b) as well, contrary to Takano (2003).

The rest of this paper is organized as follows. Carefully examining syntactic behaviors of the variant given in (3), Section 2 argues that the NO overtly moves out of the koto-phrase. Section 3 critically examines the overt movement approaches previously undertaken by Tada (1992), Koizumi (1994) and Nomura (2005). Section 4 puts forward the hypothesis that the movement of a NO is scrambling, and then illustrates how to derive the Case alternation under investigation, with the discussion about why Case alternation takes place in a limited way. Section 5 summarizes the paper.

2 A closer look at the variant

2.1 Structural asymmetries between NOs and AOs

This section investigates the syntactic behavior of NOs and AOs, taking a closer look at cases like (3), which have been discussed in Kuno (2002), Takano (2003), and Hiraiwa (2005). Two novel pieces of evidence will be provided in favor of the claim that NOs are outside of the koto-phrase, unlike AOs. The first one comes from coordination. Let us consider the contrast between (4a) and (4b).

    1. (4)
    1. a.
    1. Watasi-wa
    2. I-TOP
    1. [eigo-o
    2. English-ACC
    1. ryuutyooni
    2. fluently
    1. hanasu
    2. speak
    1. koto]
    2. NMLZ
    1. to
    2. and
    1. [piano-o
    2. piano-ACC
    1. ryoote-de
    2. both.hands-with
    1. hiku
    2. play
    1. koto]-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘I can speak English fluently and play the piano with both hands.’
    1.  
    1. b.
    1. *Watasi-wa
    2.   I-TOP
    1. eigo-ga
    2. English-NOM
    1. ryuutyooni
    2. fluently
    1. hanasu
    2. speak
    1. koto
    2. NMLZ
    1. to
    2. and
    1. piano-ga
    2. piano-NOM
    1. ryoote-de
    2. both.hands-with
    1. hiku
    2. play
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1.    ‘I can speak English fluently and play the piano with both hands.’

In (4a), there is nothing wrong with coordinating two koto-phrases including the AOs. If the NOs could stay within the koto-phrase like the AOs, nothing would prevent the koto-phrases from being coordinated in (4b) as well. However, (4b) is ungrammatical, which suggests that the NOs are outside of the koto-phrases.

The other argument involves the adverb mattaku ‘at all,’ which should be licensed by the clause-mate negation. (5b) is ungrammatical because the adverb, which is in the matrix clause, cannot be licensed by negation in the koto-phrase. In (5a) and (5c), on the other hand, the clause-mate requirement is satisfied.

    1. (5)
    1. a.
    1.   Taroo-wa
    2.   Taroo-TOP
    1. [Hanako-ga
    2. Hanako-NOM
    1. gohan-o
    2. rice-ACC
    1. zenbu
    2. all
    1. tabetesimatta
    2. ate
    1. koto]-ni
    2. NMLZ-DAT
    1. mattaku
    2. at.all
    1. hara-o
    2. belly-ACC
    1. tate-nakat-ta.
    2. make.stand.up-not-past
    1.    ‘Taroo did not get angry at all at Hanako’s eating all the rice.’
    1.  
    1. b.
    1. *Taroo-wa
    2.   Taroo-TOP
    1. [Hanako-ga
    2. Hanako-NOM
    1. gohan-o
    2. rice-ACC
    1. tabe-nai
    2. eat-not
    1. koto]-ni
    2. NMLZ-DAT
    1. mattaku
    2. at.all
    1. hara-o
    2. belly-ACC
    1. tate-ta.
    2. make.stand.up-past
    1.    Lit. ‘Taroo got angry at all at Hanako’s not eating rice.’
    1.  
    1. c.
    1.   Taroo-wa
    2.   Taroo-TOP
    1. [Hanako-ga
    2. Hanako-NOM
    1. mattaku
    2. at-all
    1. gohan-o
    2. rice-ACC
    1. tabe-nai
    2. eat-not
    1. koto]-ni
    2. NMLZ-DAT
    1. hara-o
    2. belly-ACC
    1. tate-ta.
    2. make.stand.up-past
    1.    ‘Taroo got angry at Hanako’s not eating rice at all.’

Keeping this in mind, let us consider (6a) and (6b).

    1. (6)
    1. a.
    1.   Taroo-wa
    2.   Taroo-TOP
    1. [mattaku
    2. at.all
    1. migime-o
    2. right.eye-ACC
    1. tumur-anai
    2. close-not
    1. koto]-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘Taroo is able not to close his right eye at all.’
    1.  
    1. b.
    1. *Taroo-wa
    2.   Taroo-TOP
    1. [mattaku
    2. at.all
    1. migime-ga
    2. right.eye-NOM
    1. tumur-anai
    2. close-not
    1. koto]-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1.    ‘Taroo is able not to close his right eye at all.’
    1.  
    1. c.
    1.   Taroo-wa
    2.   Taroo-TOP
    1. migime-ga
    2. right.eye-NOM
    1. [mattaku
    2. at.all
    1. tumur-anai
    2. close-not
    1. koto]-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1.    ‘Taroo is able not to close his right eye at all.’

Since the adverb mattaku should be licensed by negation in the embedded clause, the adverb is a hallmark of the left edge of the koto-phrase, which guarantees that the AO and the NO are in the koto-phrase in (6a) and (6b), respectively. The ungrammaticality of (6b) supports the claim that NOs are not included in the koto-phrase. As shown in (6c), when the NO precedes the adverb, the example becomes grammatical, which shows that the NO is outside of the koto-phrase.2

The novel pieces of evidence presented in this section indicate that the variant under investigation cannot be handled under one previous analysis, which was proposed by Takahashi (2010). He proposes that the wide scope interpretation of the NO dake-phrase in (2b) results from an application of the covert operation Quantifier Raising (see Saito 2010 for a similar approach). His analysis is crucially based on the assumption that the NO stays at its underlying position like the AO at overt syntax, but this assumption cannot be maintained with the variant, as shown in this section.

2.2 Base-generation or movement?

Kuno (2002) also investigates the relevant variant and claims that the NO undergoes overt movement out of the koto-phrase. The novel data presented in Section 2.1 is compatible with the overt movement approach but the relevant data can be also captured by base-generating the NO as an element of the matrix clause, without appealing to movement. The latter approach is pursued by Takano (2003). He proposes that (i) a NO is base-generated as a proleptic object in the matrix clause and (ii) that the object position of the embedded predicate is occupied by pro, as illustrated in (7).

    1. (7)
    1. a.
    1. John-wa
    2. John-TOP
    1. migime-dake-ga1 [PRO pro1
    2. right.eye-only-NOM
    1. tumur]-e-ru.
    2. close-can-PRES
    1. ‘John can close only his right eye.’
    1.  
    1. b.
    1. John-wa
    2. John-TOP
    1. migime-dake-ga1 [PRO pro1
    2. right.eye-only-NOM
    1. tumur-u]
    2. close-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘John can close only his right eye.’

Kuno (2002) does not discuss the base-generation approach but it is still important to examine which approach is more plausible. This is the topic of the rest of this section.

As mentioned in Section 1, this paper takes the position that the narrow scope reading of the NO is available in (7b), contrary to Takano (2003). This observation is readily captured under the movement approach because the lower copy of the NO can partake in the scope calculation. On the other hand, under the base-generation approach, it is difficult to expect the relevant narrow scope reading. In what follows, I will present another argument for the movement approach. Let us consider the example in (8), where the anaphor zibun in the NO can be bound by the causee Hanako-ni.34

    1. (8)
    1. a.
    1. Taroo-ga
    2. Taroo-NOM
    1. Hanako-ni1
    2. Hanako-DAT
    1. zibun1-no
    2. self-GEN
    1. migite-dake-ga
    2. right.hand-only-NOM
    1. age-sase-ru
    2. raise-cause-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ta.
    2. can-past
    1. Lit. ‘Taroo could make Hanako1 raise only self1’s right hand.’
    1.  
    1. b.
    1. Taroo-ga [zibun1-no migite-dake-ga]2 [Hanako1-ni pro2 age-sase-ru koto]-ga deki-ta.
    1.  
    1. c.
    1. Taroo-ga [Hanako1-ni]3 [zibun1-no migite-dake-ga]2 [t3 pro2 age-sase-ru koto]-ga deki-ta.

Under Takano’s approach, the NO is base-generated as a proleptic object in the matrix clause, which is higher than the binder Hanako-ni, as illustrated in (8b). It is necessary to scramble Hanako-ni across the NO in order to obtain the surface order, as illustrated in (8c). As (8b) shows, zibun is not c-commanded by the antecedent at the base-generated position, which leads to the expectation that (8a) would involve binding failure. One might suggest that binding relation between zibun and its antecedent can be established after the binder moves to a higher position than zibun in (8c). However, this suggestion is not maintainable. Let us consider (9).

    1. (9)
    1. a.
    1. *Zibun1-no
    2.   self1-GEN
    1. sensei-wa
    2. teacher-TOP
    1. Hanako1-ni
    2. Hanako1-DAT
    1. mondai-o
    2. question-ACC
    1. tok-ase-ta.
    2. answer-cause-past
    1.    Lit. ‘Self1’s teacher made Hanako1 answer the question.’
    1.  
    1. b.
    1. *[Hanako1-ni]2 zibun1-no sensei-wa t2 mondai-o tok-ase-ta.
    1.  
    1. c.
    1. [Hanako-ni]1
    2.     Hanako-DAT
    1. sensei-wa
    2. teacher-TOP
    1. t1
    2.  
    1. mondai-o
    2. question-ACC
    1. tok-ase-ta.
    2. answer-cause-past
    1.     ‘The teacher made Hanako answer the question.’

In (9a), zibun is not c-commanded by its antecedent Hanako-ni. Even if Hanako-ni undergoes scrambling, the binding failure cannot be repaired, as shown in (9b). If scrambling could create a new binding relation between zibun and its antecedent, (9b) would be grammatical, contrary to fact. (9c) confirms that the scrambling of Hanako-ni poses no problem if binding relation is not involved. The availability of zibun-binding in (8a) suggests that zibun in the NO is bound by its antecedent at the base-generated position and that the NO moves out of the koto-phrase followed by the movement of Hanako-ni, as illustrated below.

(10) Taroo-ga [Hanako-ni1]2 [zibun1-no migite-dake-ga]3 [t2 t3 age-sase-ru koto]-ga deki-ta.

This section has presented novel pieces of evidence in favor of the claim that the NO is outside of the koto-phrase at overt syntax, and then addressed the issue as to whether the relevant dislocation results from overt movement or base-generation, which has not been addressed by Kuno (2002). The section concludes that it is difficult to maintain the base-generation approach, based on the behavior of zibun-binding.

3 Overt movement approaches

In Section 2, it was shown that the relevant variant can be handled through neither Takahashi’s (2010) nor Takano’s (2003) approaches. The aim of this section is to critically examine another type of approach, where the NO undergoes overt movement for the purpose of Case checking, pursued by Tada (1992) and Koizumi (1994), among others. As reviewed later, Koizumi’s analysis is more plausible than Tada’s analysis, but it will be argued that the former analysis is also difficult to maintain. This is because Koizumi’s analysis is crucially based on the assumption that T’s EPP is obligatorily satisfied in Japanese like English, but this assumption needs to be reconsidered, along the lines of Fukui (1986) and Kuroda (1988).

Let us start the discussion with Tada’s analysis (1992). He argues that the NO moves to [Spec, AGRoP] for Case checking. Under his analysis, the potential predicate moves to AGRo and its [+stative] is responsible for nominative Case checking of the NO at [Spec, AGRoP], as illustrated below.

    1. (11)

In the structure above, [Spec, AGRoP] is higher than the potential predicate, which leads to the wide scope reading of the NO. The narrow scope reading of the AO is successfully captured because the AO stays within VP. The scope asymmetry between the NO and the AO in (2) is thus explained.

Although Tada’s analysis successfully captures the wide scope nature of the NO, his analysis faces some problems, pointed out by Koizumi (1994). One of them is given below.

(12) Koizumi (1994: 220)
  Under his (=Tada’s) analysis, nominative Case is licensed by two different categories: nominative Case of subjects is licensed by Tense (+AGRs), and nominative Case of (nominative) objects is licensed by [+stative] predicates (+AGRo). Tense is a functional category, while [+stative] predicates are lexical categories (verbs, adjectives, and adjectival nouns). These two categories have nothing in common, except for the alleged nominative Case licensing ability. It is not clear at all why nominative Case should be licensed by two distinct sets of categories as different as Tense and stative predicates. This fact alone, of course, does not render the analysis untenable, but it surely makes it dubious.

Koizumi (1994) solves this problem by proposing that the Case of the NO is licensed by T in the same way as nominative subjects. This proposal is technically implemented by arguing that the NO overtly moves to [Spec, TP]. One of his arguments is based on scope interaction with negation. Let us consider the following example.

    1. (13)
    1. Koizumi (1994: 222)
    1. John-ga
    2. John-NOM
    1. migime-dake-ga
    2. right.eye-only-NOM
    1. tumur-e-na-i.
    2. close-can-not-PRES
    1. ‘John cannot close only his right eye.’ (only > not > can)

In the example above, the NO takes wide scope over negation, which is correctly predicted under his analysis because [Spec, TP] is higher than negation. Koizumi points out that (13) is problematic for Tada’s analysis, where the landing site of the NO under his analysis (i.e. [Spec, AGRoP]) is lower than negation. As just reviewed, it seems that Koizumi’s (1994) analysis is more plausible than Tada’s analysis conceptually and empirically but it will be argued below that it is also difficult to maintain Koizumi’s analysis.

Koizumi’s analysis crucially assumes that Japanese T plays a significant role in syntactic computation in a way similar to English T, which induces φ-feature agreement and triggers A-movement for the EPP requirement. However, this assumption should be carefully examined. It has been controversial whether Japanese has A-movement such as passivization and raising in English and other languages which exhibit φ-feature agreement. Concerning passivization, one could analyze the so-called passive construction in Japanese such as (14a) in the same way as its English counterpart, which is widely analyzed in terms of A-movement, as illustrated in (14b).

    1. (14)
    1. a.
    1. Taroo-ga
    2. Taroo-NOM
    1. Hanko-ni
    2. Hanako-DAT
    1. hihans-are-ta.
    2. criticize-PASS-past
    1. ‘Taroo was criticized by Hanako.’
    1.  
    1. b.
    1. Taroo-ga1 Hanko-ni t1 hihans-are-ta.

However, this is not the only possible analysis of the Japanese passive. Some researches argue that examples such as (14a) involve a clausal complement and that the dislocated subject is the argument selected by the passive morpheme (Kuroda 1965/79, among many others). Additionally, given that scrambling is widely available in Japanese, it will be also plausible to analyze the dislocation of the nominative subject is scrambling. In fact, carefully examining reconstruction effects, Hoji (2008) argues that the relevant dislocation in Japanese passive is analyzed as scrambling. The optionality of the dislocation of Taroo-ga in (14) is straightforwardly explained under the scrambling analysis because scrambling is also optional.

From a conceptual perspective, there is no reason to claim that T’s EPP is obligatorily satisfied in Japanese either. Recall that EPP was originally proposed as a structural requirement in order to capture the appearance of expletive elements in some languages like English (Chomsky 1981; see also Bever 2009; Chomsky 2009 for relevant discussion). As one of the reviewers points out, since EPP is just the name of a phenomenon in those languages, not a principle, nothing conceptually forces this structural requirement to hold universally. In fact, McCloskey (2001) and Bobaljik & Wurmbrand (2005), among others, argue that EPP does not hold universally. Recently, Chomsky (2015) also suggested that Italian lacks EPP. Turning to Japanese, given the absence of expletive elements, it is plausible to argue that T has no EPP requirement in the language. In fact, Fukui (1986) and Kuroda (1988) take a similar stand and pursue the hypothesis that Japanese subjects do not have to move to [Spec, TP] but can stay in-situ, unlike English.

One of the arguments for subject raising in Japanese in the literature is presented by Kishimoto (2001). His argument is based on the syntactic distribution of indeterminate pronouns in cases like (15), where they concur with the particle mo, which is originally studied by Kuroda (1965/79).

    1. (15)
    1. Hanako-wa
    2. Hanako-TOP
    1. [Taroo-ga
    2. Taroo-NOM
    1. nani-o
    2. what-ACC
    1. kat-ta
    2. buy-past
    1. to
    2. that
    1. mo]
    2. PART
    1. omowa-nakat-ta.
    2. think-not-past
    1. ‘Hanako did not think that Taroo bought anything.’

Kishimoto (2001: 600) observes that there is a subject/object asymmetry when mo is attached to a verb, as illustrated below.

    1. (16)
    1. a.
    1. *Dare-ga
    2.   who-NOM
    1. warai-mo
    2. laugh-PART
    1. si-nakat-ta.
    2. do-not-past
    1.    ‘Nobody laughed.’
    1.  
    1. b.
    1.   Taroo-wa
    2.   Taroo-TOP
    1. dare-ni
    2. who-DAT
    1. ai-mo
    2. see-PART
    1. si-nakat-ta.
    2. do-not-past
    1.    ‘Taroo did not meet anyone.’

Kishimoto claims (i) that mo is combined with a verb and they move to v together and (ii) that the scope of mo is the vP. He argues that the ungrammaticality of (16a) results from the existence of subject raising to [Spec, TP] out of the domain of mo.

However, it is necessary to reconsider the generalization that subject indeterminate pronouns cannot be licensed by mo attached to the verb. In fact, Kuroda (1965/79) provides the following example, where the indeterminate pronoun occupies the subject position.

    1. (17)
    1. Kuroda (1979: 93)
    1. Koremade
    2. this.till
    1. dare-ga
    2. who-NOM
    1. kangae-mo
    2. think-PART
    1. si-nakat-ta
    2. do-not-past
    1. aidia
    2. idea
    1. ‘the idea which nobody has ever thought of’

The grammaticality of the following example also casts doubt on the structural condition proposed by Kishimoto.

    1. (18)
    1. Takano (2003: 803)
    1. ?Watasi-wa
    2.   I-TOP
    1. dare-ni
    2. who-DAT
    1. [koi
    2. come
    1. to
    2. that
    1. mo]
    2. PART
    1. itte-inai.
    2. said-have.not
    1.    ‘I haven’t said to anyone to come.’

In (18), dare-ni is an object of the matrix verb while mo is attached to the complement. It is obvious that the former does not fall under the scope of mo. Nevertheless, the example is acceptable. To the extent that the nature of indeterminate pronouns is not clearly understood, it seems to be difficult to employ them as a diagnostic for exploring the syntactic position of a subject in Japanese (see also Yamashita 2009 for a prosodic analysis of the relevant construction).

Following Fukui (1986) and Kuroda (1988), this paper argues that subjects can stay in-situ in Japanese (see also Kato 2006 for relevant discussion). Let us consider the following example. In (19), the verb in the first conjunct is bare and lacks a tense morpheme.

    1. (19)
    1. Taroo-ga
    2. Taroo-NOM
    1. nattoo-o
    2. nattoo-ACC
    1. tabe,
    2. eat
    1. Jiroo-ga
    2. Jiroo-NOM
    1. koohii-o
    2. coffee-ACC
    1. nom-u
    2. drink-PRES
    1. yooni
    2. COMP
    1. natta.
    2. happened
    1. ‘It happened that Taroo ate nattoo and Jiroo drank coffee.’

The absence of subject raising in Japanese enables (19) to be analyzed in terms of the coordination of the embedded vPs, as illustrated in (20).

(20) [[vP Taroo-ga nattoo-o tabe] & [vP Jiroo-ga koohii-o nom]u] yooni natta.

On the other hand, under the hypothesis that T’s EPP is obligatorily satisfied in Japanese as in English, (19) is supposed to be derived from coordination of the matrix TPs with the ellipsis taking place in the first conjunct, as illustrated in (21).

(21) [Taroo-ga nattoo-o tabe-ru yooni natta] & [Jiroo-ga koohii-o nom-u yooni natta].

However, such ellipsis analysis is not plausible. Let us consider the ungrammaticality of (22a).

    1. (22)
    1. a.
    1. *Taroo-ga
    2.   Taroo-NOM
    1. Hanako-ni
    2. Hanako-DAT
    1. nattoo-o
    2. nattoo-ACC
    1. tabe,
    2. eat
    1. John-ga
    2. John-NOM
    1. Mary-ni
    2. Mary-DAT
    1. koohii-o
    2. coffee-ACC
    1. nomu
    2. drink
    1. yooni
    2. COMP
    1. itta.
    2. told
    1.    ‘Taroo told Hanako to eat nattoo and John told Mary to drink coffee.’
    1.  
    1. b.
    1. Taroo-ga Hanako-ni nattoo-o tabe-ru yooni itta, John-ga Mary-ni koohii-o nomu yooni itta.

If it were possible to derive (19) via ellipsis as illustrated in (21), (22a) would be analyzed as (22b) in a way similar to (21). It is not clear why the ellipsis strategy available for (21) is not available for (22b). On the other hand, under the non-ellipsis approach, the ungrammaticality of (22a) is easily captured in terms of coordination of non-constituents. It is obvious that Taroo-ga, Hanako-ni, nattoo-o and tabe do not make a constituent in (22a). Thus, it is more plausible to claim that (19) should be analyzed as (20) without recourse to ellipsis, which indicates that Japanese subjects do not have to move to [Spec, TP].

A similar argument holds with NOs. The following example shows that the NO does not undergo movement to [Spec, TP], contrary to Koizumi (1994).

    1. (23)
    1. Hanako-wa
    2. Hanako-TOP
    1. gakuhu-ga
    2. score-NOM
    1. surasurato
    2. easily
    1. yom-u
    2. read-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki,
    2. can
    1. piano-ga
    2. piano-NOM
    1. ryooote-de
    2. both.hands-with
    1. hik-u
    2. play-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru
    2. can-PRES
    1. yooni
    2. COMP
    1. natta.
    2. happened
    1. ‘It happened that Hanako was able to read scores easily and play the piano by both hands.’

It is plausible to claim that (23) involves vP-coordination as illustrated in (24a), in a way similar to (19). The analysis in (24a) is incompatible with Koizumi’s analysis because each conjunct lacks [Spec, TP], which is a landing site of the NO under his analysis.

(24) a. [vP Hanako-wa1 gakuhu-ga surasurato yom-u koto-ga deki] & [vP pro1 piano-ga ryooote-de hik-u koto-ga deki] ru yooni natta.
  b. [TP Hanako1-wa gakuhu-ga surasurato yom-u koto-ga deki-ru yooni natta] & [TP pro1 piano-ga ryooote-de hik-u koto-ga deki-ru yooni natta].

Alternatively, (23) could be analyzed as (24b), which involves TP coordination with ellipsis taking place on the right edge of the first conjunct. However, recall that such an ellipsis analysis is not plausible, as discussed in (22a).5

Nomura (2005) also pursues the analysis where the NO is licensed by T. Deviating from Koizumi, he proposes that the NO is allowed to stay at its base-generated position, in addition to the option of moving to [Spec, TP]. Under Nomura’s analysis, the grammaticality of (23) can be accommodated but given the data presented in Section 2, where it was shown that the NO overtly moves out of the koto-phrase from its underlying position, it is not plausible to allow the NO to stay in its base-generated position.

4 Proposal

The aim of this section is to provide an alternative analysis of the construction under discussion. Before presenting details of the proposal, the gist of it is provided below. Given that Japanese does not exhibit φ-feature agreement, this paper adopts the approach where a nominal is assigned Case on the basis of its structural position without appealing to agreement (see Kuroda 1978; Saito 1982; Fukui 1986; Zushi 2016 among others), instead of the Agree-based approach to Case (Chomsky 2000). Investigation of the internal structure of the koto-phrase is also an important step for a better understanding of the relevant construction. It will be argued that the koto-phrase taken by deki involves restructuring in the sense of Wurmbrand (2001). That is, the verb phrase within the relevant koto-phrase is just VP, not vP. This point will be explored in terms of the Negative Concord Items. The most important point of the proposed analysis is to challenge the general assumption that scrambling does not affect Case valuation. It will be argued that scrambled nominals can receive Case at the landing site of scrambling in principle and that this is indeed involved in the relevant construction.

4.1 Case

Under the standard approach to Case in the minimalist program, Case valuation is obtained as a by-product of φ-feature agreement (Chomsky 2000). It is controversial whether this Agree-based approach is plausible to some languages which do not exhibit φ-feature agreement such as Japanese. The literature includes another view of Case, whereby a nominal is assigned Case based on its structural position without appealing to agreement (see Kuroda 1978; Saito 1982; Fukui 1986 among others). Zushi (2016) investigates the latter approach under the minimalist program and proposes that Case valuation is executed in non-agreement languages such as Japanese via the following mechanisms, which are adopted in this paper.

(25) Zushi (2016: 48)
  a. When a nominal is merged with a lexical head, its Case feature is valued as accusative.
  b. When a nominal is merged with a phase head (v or n), its Case feature is valued as nominative or genitive.
  c. Otherwise, the Case feature of a nominal is valued as dative.

This paper also follows Zushi (2016) with respect to the structure of stative predicates, most of which are adjectives and nominal adjectives. Assuming that adjectives and nominal adjectives must combine with a phase head to take arguments, which is originally due to Baker (2003) and Kayne (2009), she proposes that a theme argument of a stative predicate occupies an edge position of v, not the complement position of the predicate. Thus, (26a) has the structure given in (26b). The internal argument is assigned nominative Case ga under the mechanism given in (25b).

    1. (26)
    1. Zushi (2016: 54)
    1.  
    1. a.
    1. Taroo-ga
    2. Taroo-NOM
    1. okane-ga
    2. money-NOM
    1. hosii/hituyoo-da/aru.
    2. want/need-be/have
    1. ‘Taroo {wants/needs/has} some money.’
    1.  
    1. b.

This paper claims that the stative verb deki has a similar structure, where its internal argument (i.e. the koto-phrase) occupies an edge position of v, as illustrated below.

    1. (27)

Before showing how the construction under investigation is derived, let us take a closer look at the internal structure of the koto-phrase in the next subsection.

4.2 On the internal structure of the koto-phrase

The nominalizer koto can be employed not only with the stative verb deki but also with non-stative verbs such as kokoromi ‘try’ as shown in (28a), where the koto-phrase is marked with accusative Case o, not nominative Case ga, in contrast to the construction under investigation, repeated as (28b).6

    1. (28)
    1. a.
    1. Taroo-wa
    2. Taroo-TOP
    1. issyookenmei
    2. hard
    1. benkyoos-uru
    2. study-PRES
    1. koto-o
    2. NMLZ-ACC
    1. kokoromi-ta.
    2. try-past
    1. ‘Taroo tried to work hard.’
    1.  
    1. b.
    1. Taroo-wa
    2. Taroo-TOP
    1. migime-dake-ga
    2. right.eye-only-NOM
    1. tumur-u
    2. close-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘Taroo can close only his right eye.’

In what follows, it will be shown that despite the absence of the apparent difference, the koto-phrase in (28a) has a different internal structure from that in (28b), based on the Japanese wh-mo expression appearing under a negative context such as dare-mo ‘who-PART’ and nani-mo ‘what-PART.’ Following Watanabe (2004), let us call these items negative concord items (NCIs) in this paper. The contrast between (29a) and (29b) shows that the NCI requires negation.

    1. (29)
    1. a.
    1.   Taroo-wa
    2.   Taroo-TOP
    1. dare-ni-mo
    2. who-DAT-PART
    1. awa-nakat-ta.
    2. see-not-past
    1.    ‘Taroo did not see anyone.’
    1.  
    1. b.
    1. *Taroo-wa
    2.   Taroo-TOP
    1. dare-ni-mo
    2. who-DAT-PART
    1. at-ta.
    2. see-past
    1.    Lit. ‘Taroo saw anyone.’
    1.  
    1. c.
    1. *Taroo-wa
    2.   Taroo-TOP
    1. [Hanako-ga
    2. Hanako-NOM
    1. dare-ni-mo
    2. who-DAT-PART
    1. au
    2. see
    1. to]
    2. that
    1. iwa-nakat-ta.
    2. say-not-past
    1.    ‘Taroo did not say that Hanako would see anyone.’

It has also been observed that the NCI and negation should be involved in the same clause, as shown in (29c).

As has been discussed, this paper takes the view that Japanese does not employ the Agree-based approach to Case valuation. However, this leaves the possibility that Agree still plays a role in other syntactic dependencies in Japanese where φ-feature is not involved. This paper adopts the hypothesis that the dependency between the relevant NCI and negation is captured via Agree. Along the lines of Yamashita (2003) and Maeda (2004), this paper also argues that the clause-mate requirement in question is derived from the Phase Impenetrability Condition (PIC), proposed by Chomsky (2000; 2001).7 Given that phases play an important role in the syntactic computation under the current minimalist program, it is desirable to reduce the clause-mate requirement to the PIC, originally proposed by Chomsky (2000), as given in (30).

(30) Chomsky (2000: 108)
  In phase α with head H, the domain of H is not accessible to operations outside α, only H and its edge are accessible to such operations.

Later, Chomsky (2001) relaxes the PIC as shown in (31), where ZP is a next higher phase.

(31) Chomsky (2001: 14)
  The domain of H is not accessible to operations at ZP; only H and its edge are accessible to such operations.

Let us consider how (30) and (31) differ under the schematically illustrated structure given in (32), where there is a non-phase head X between Z and H.

(32) [ZP Z [XP X [HP H YP]]]

Under (30), the non-phase head X cannot probe into the domain of H (i.e. YP) because the domain of H becomes inaccessible to any elements outside of HP. Under (31), on the other hand, the non-phase head X is allowed to probe into the domain of H (i.e. YP), although YP is still inaccessible to the next higher phase head Z. In fact, the latter definition is supported by the so-called quirky NO in Icelandic such as in (33). “X” in (32) is T in the following example, where the NO agrees with T.

    1. (33)
    1. Sigurðsson (2002: 692)
    1. Henni
    2. her.DAT
    1. leiddust
    2. bored
    1. strakarnir.
    2. boys.the.NOM
    1. ‘She found the boys boring.’

Yamashita (2003) and Maeda (2004) argue that the PIC in (31) is also more plausible in terms of NCI licensing. Let us observe how this is by taking as an example (34a), which includes an object NCI.

    1. (34)
    1. a.
    1. Taroo-ga
    2. Taroo-NOM
    1. nani-mo
    2. what-PART
    1. yoma-nakat-ta.
    2. read-not-past
    1. ‘Taroo did not read anything.’
    1.  
    1. b.
    1. [C [T [Neg [vP Taroo v [VP V NCI]]]]] (order irrelevant)

Let us adopt the assumption widely adopted in the literature that negation (Neg) is merged with vP. Under the definition of the PIC given in (30), Neg cannot probe into the domain of v (i.e. VP) because Neg is outside of the phase vP. On the other hand, under the definition given in (31), VP is not accessible to C, which is a next higher strong phase head, but nothing prevents Neg, which is below C, from probing into VP. Thus, Neg can successfully undergo Agree with the NCI (see Yamashita 2003 and Maeda 2004 for more empirical arguments).

Bearing the discussion so far in mind, let us consider the following contrast.

    1. (35)
    1. a.
    1. *Taroo-wa
    2.   Taroo-TOP
    1. [dare-ni-mo
    2. who-DAT-PART
    1. purezento-o
    2. present-ACC
    1. age-ru
    2. give-PRES
    1. koto]-o
    2. NMLZ-ACC
    1. kokoromi-nakat-ta.
    2. try-not-past
    1.    ‘Taroo didn’t try to give a present to anyone.’
    1.  
    1. b.
    1.   Taroo-wa
    2.   Taroo-TOP
    1. [dare-ni-mo
    2. who-DAT-PART
    1. purezento-o
    2. present-ACC
    1. age-ru
    2. give-PRES
    1. koto]-ga
    2. NMLZ-NOM
    1. deki-nai.
    2. can-not
    1.   ‘Taroo can’t give a present to anyone.’

In (35a), the NCI within the koto-phrase cannot be licensed by negation in the matrix clause. This is because the matrix negation fails to probe into the lower VP within the embedded vP, as illustrated in (36).

(36) [vP Taroo-wa [[VP [[vP PRO [[VPdare-ni-mo purezento-o age-ru]v]] koto-o]kokoromi]v]]-nakat-ta.

In contrast, interestingly, (35b) is grammatical. The NCI included in the koto-phrase can be licensed by the matrix negation. Based on the structure in (27), suppose that (35b) has the following structure.

    1. (37)

Let us consider the timing of Transfer of the VP within the koto-phrase. Given the PIC formulated in (31), it is reasonable to assume that when a given phase becomes part of another phase, the complement of the lower phase head gets transferred and becomes inaccessible. In (37), when the koto-phrase is merged to an edge position of the matrix vP, the VP complement gets transferred, which makes the VP inaccessible. The koto-phrase in (37) itself is still accessible to further syntactic computation because it occupies an edge position of the matrix vP. However, the VP within the koto-phrase is no longer accessible at the matrix vP-phase level. Thus, negation, which is outside of the matrix vP, is expected to fail to probe into the VP containing the relevant NCI, as is the case in (36), which calls for an alternative structure in order to capture the restructuring effect above.

Wurmbrand (2001) proposes that some restructuring predicates which she calls “lexical restructuring predicates” combine with a bare VP, which lacks an embedded syntactic subject. It is claimed in this paper that what is selected by koto in (35b) is also a bare VP without a vP-layer within the koto-phrase. Under this analysis, nai can probe into the VP within the koto-phrase due to the absence of a vP-layer within the koto-phrase. Given the absence of the vP-layer within the koto-phrase, since NegP involves only one phase, the NCI is accessible to negation, which leads to the welcome result, as illustrated in (38).

    1. (38)

The grammaticality of (35b) also suggests that the relevant koto is not a complementizer as a phase head. If so, the NCI in (35b) would fail to be licensed by negation, contrary to fact.

One might say that koto takes a TP because it is sometimes assumed in the literature that the morpheme attached to the verbal stem (i.e. (r)u) is a realization of present tense. In fact, this paper has glossed and continues to gloss the morpheme as “present” for the sake of convenience. However, this is not the only way of analyzing the morpheme in question. Alternatively, it is also plausible to analyze the morpheme as part of a verbal conjugation. In other words, the morpheme is just required for morpho-phonological reasons, without any semantic meaning (see Kusumoto 1999 for relevant discussion). Under this view, the relevant morpheme is required to make the verb an adnominal form when it appears before the noun koto. This paper adopts the latter view, which is compatible with Wurmbrand’s claim that the relevant restructuring predicates take a bare VP.8

Under the proposal, it is obvious that what is directly selected by the relevant verbs (kokoromi and deki) is a koto-phrase. One of the reviewers raises a question regarding how to ensure that the koto-phrase selected by kokoromi involves a vP while that selected by deki involves a VP, and not vice versa. As both of the verbs take a koto-phrase, it is reasonable that they are lexically specified in such a way that they syntactically take a nominal argument (i.e. an argument with [+N]). In addition to this syntactic selection, they are also lexically specified concerning semantic selection. For example, the verb kokoromi is lexically specified in such a way that it takes a proposition as a semantic type. VP corresponds to a predicative event while vP, which involves an embedded subject, corresponds to a proposition. Let us also assume that the nominalizer koto itself does not make any semantic contribution: if a given syntactic object X is merged with koto, the larger syntactic object {koto, X} has the same semantic type as X. For example, the semantic type of vP (i.e. a proposition) is inherited to the koto phrase where koto takes a vP. Suppose that kokoromi is merged with such a koto-phrase. Since kokoromi needs to take a proposition, no type mismatch will arise between the verb and the koto-phrase. On the other hand, if kokoromi is merged with the koto-phrase where koto takes a VP, type mismatch will arise because the relevant koto-phrase is interpreted as a predicative event, which kokoromi does not want.

Note that the koto-phrase in (35a) involves o while that in (35b) involves ga, which indicates that the former stays at the complement of the verb and the latter occupies an edge position of vP. One might say that the asymmetry in (35) is somehow due to the positional differences of the koto-phrases in the structure. However, this is not the case. Interestingly, the passive counterpart of (35a), whose koto-phrase is marked with ga, is also ungrammatical, as shown in (39).

    1. (39)
    1. *[Dare-ni-mo
    2.   who-DAT-PART
    1. purezento-o
    2. present-ACC
    1. age-ru
    2. give-PRES
    1. koto]-ga
    2. NMLZ-NOM
    1. kokoromi-rare-nakat-ta.
    2. try-PASS-not-past
    1.    Lit. ‘It was not tried to give a present to anyone.’

The existence of nominative Case suggests that the subject of the passive construction in Japanese is supposed to occupy an edge position of vP in a way similar to the koto-phrase in (35b). The ungrammaticality of (39) confirms that the contrast given in (35) is due to the internal structure of the koto-phrases, not their positional differences.

4.3 Case valuation after scrambling

It has been widely assumed that a noun phrase never changes its Case marker as a result of scrambling. That is, a scrambled phrase retains its Case marker assigned before scrambling. It is true that this is well-established empirically but theoretically speaking, nothing prevents nominals from receiving Case after scrambling in principle. In fact, Fukui & Nishigauchi (1992) and Fukui (1995) pursue the hypothesis that an application of scrambling can change its Case marker of a noun phrase, on the basis of Nominative/Genitive Conversion such as (40a) and (40b), which was originally discussed by Harada (1971).

    1. (40)
    1. a.
    1. Taroo-ga
    2. Taroo-NOM
    1. katta
    2. bought
    1. hon
    2. book
    1. ‘the book Taroo bought’
    1.  
    1. b.
    1. Taroo-no
    2. Taroo-GEN
    1. katta
    2. bought
    1. hon
    2. book
    1. ‘the book Taroo bought’
    1.  
    1. c.
    1. [Taroo-no1 [[relative clause t1 katta] hon]]

It has been argued that nominative Case marker ga can alternate with genitive Case marker no under several syntactic contexts such as relative clauses and noun complements (see Maki & Uchibori 2008 for an overview). Particularly, under Fukui’s (1995) analysis, the converted form in (40b) is derived in such a way that the subject of the relative clause undergoes scrambling out of that embedded clause and is merged to the nominal projection, where it receives genitive Case no, as illustrated in (40c). Under the mechanism of Case valuation adopted in this paper given in (25b), the relevant landing site is the edge of nP.

A similar analysis has been suggested to the construction explored in this paper by Kuno (2002). As briefly mentioned earlier, Kuno claims that the NO overtly undergoes movement (“focus raising” in his words) out of the koto-phrase, as shown in (41), where the moved object is merged with the VP with [+stative].

    1. (41)

The moved object receives nominative Case at the landing site under the assumption that the stativity of deki percolates onto the VP node. The percolated stativity is responsible for assigning nominative Case, even though the moved object is not a sister of deki.

Given that the theoretical status of percolation is not clear enough under the current syntactic theory, it would be desirable to eliminate the mechanism of percolation. In addition, Kuno does not address the issue on how to regulate Case alternation due to an application of scrambling. For example, even if an accusative object undergoes scrambling, it retains accusative Case at the landing site, as shown below.

    1. (42)
    1. Kono-hon-o/*ga
    2. this-book-ACC.NOM
    1. Taroo-ga
    2. Taroo-NOM
    1. yon-da.
    2. read-past
    1.               ‘Taroo read this book.’

In what follows, I also would like to pursue the hypothesis that scrambled phrases can receive Case at the landing site in a more sophisticated way. In doing so, I address the issue as to why scrambling can affect Case alternation in such a restricted way under the conceptions of minimalist program, without appealing to the mechanism of percolation.

Let us now take a close look at the derivation of the construction under investigation, which is repeated as (43).

    1. (43)
    1. Taroo-wa
    2. Taroo-TOP
    1. migime-dake-ga
    2. right.eye-only-NOM
    1. tumur-u
    2. close-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘Taroo can close only his right eye.’

The derivation of (43) is shown in (44), where the object migime-dake is first base-generated at the complement of the verb.

    1. (44)

This paper takes the view that the operation of valuation can freely apply. Although the application itself is optional, if unvalued features remain at the end of a derivation, the derivation will be illegitimate. Even if a nominal occurs in the appropriate configuration of Case valuation, Case valuation does not have to apply. If Case feature of the nominal is valued later in the course of a derivation, the Case feature will not make the derivation illegitimate. Thus, the object does not have to receive accusative Case at the base-generated position, although it is possible. The object can undergo scrambling with its Case feature unvalued, as illustrated in (44). Then, the scrambled phrase receives nominative Case ga at the landing site of scrambling (i.e. an edge position of vP) because the structural requirement for Case valuation is appropriately satisfied.9

One of the advantages of the proposed analysis is that nominative subjects and nominative objects are both licensed in a structurally unified way: both of them are licensed at an edge position of vP. Recall Koizumi’s criticism of Tada’s analysis given in (12). Under the latter, there are two sources for nominative Case assignment: stative predicates and T. The proposed approach overcomes this problem by unifying the two modes of nominative Case assignment.

Since the landing site of scrambling is higher than the predicate deki ‘can,’ it is correctly expected that the NO can take wide scope over deki. Since scrambling is optional, the object can remain at the base-generated position in (44). Since the base-generated position is the only position for Case valuation with the object, it should receive accusative Case there, which leads to the accusative counterpart given in (45).10

    1. (45)
    1. Taroo-ga
    2. Taroo-NOM
    1. migime-dake-o
    2. right.eye-only-ACC
    1. tumur-u
    2. close-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘Taroo can close only his right eye.’

The analysis given in (44) is extended to the complex predicate counterpart such as (2b), repeated as (46a), whose derivation is given in (46b), where the potential suffix e occupies the complement position of v, in a way similar to deki in (44).1112

    1. (46)
    1. a.
    1. John-ga
    2. John-NOM
    1. migime-dake-ga
    2. right.eye-only-NOM
    1. tumur-e-ru.
    2. close-can-PRES
    1. ‘John can close only his right eye.’
    1.  
    1. b.
    1. [TP [vP John [vmigime-dake1 [v[VP t1 tumur] [ve v]]]] T]

Under the proposed approach, the Case alternation is contingent on application of scrambling. Thus, the optionality of Case alternation is due to that of scrambling. The optionality of scrambling itself is not a problem. It is true that the optionality of scrambling used to be a problem under the early minimalist program, where the application of movement is restricted in such a way that it is subject to Last Resort: movement is obligatory. However, under the recent “Free Merge” theory of movement (Chomsky 2013; 2015), where movement/Internal Merge freely applies, the optionality of scrambling ceases to be a problem. Rather Japanese scrambling is a reflection of the notion of Free Merge in a straightforward way, as suggested by one of the reviewers.

As has been discussed so far, scrambling plays a crucial role in triggering Case alternation in Japanese. In what follows, it will be argued that employing the structure-based approach to Case valuation is also important to the availability of Case alternation. As one of the reviewers points out, even though some languages such as Italian and German also have many restructuring phenomena, Case alternation is not allowed in those languages. Let us take German as an example. Recall that Wurmbrand (2001) argues that restructuring involves a configuration where the embedded predicate involves bare VP-structure, not full vP-structure. The so-called easy-to-please construction is one of them, given below.

    1. (47)
    1. Wurmbrand (2001: 27)
    1. Dieser
    2. this
    1. Text
    2. text.NOM
    1. ist
    2. is
    1. schwer
    2. hard
    1. zu
    2. to
    1. lessen.
    2. read
    1. ‘This text is hard to read.’

There is no case assigner within the embedded VP. Dieser Text has to undergo Agree with T, followed by movement to [Spec, TP], and nominative Case is assigned to Dieser Text. It is crucial that the object cannot appear with accusative Case, as shown below.

    1. (48)
    1. Wurmbrand (2001: 37)
    1. *weil
    2.   since
    1. den
    2. the
    1. Traktor
    2. tractor.ACC
    1. leicht
    2. easy
    1. zu
    2. to
    1. riparieren
    2. repair
    1. ist
    2. is
    1.    ‘since the tractor is easy to repair’

The ungrammaticality of (48) shows that Case alternation is not allowed even under the restructuring context in German, unlike Japanese. As the reviewer above points out, the difference with respect to the (un)availability of Case alternation boils down to the difference concerning the way of case valuation: the structure-based approach vs the Agree-based approach. If German adopted the former approach, the object could receive accusative Case in-situ in (48) and the example would be grammatical, contrary to fact.

4.4 A constraint on Case alternation

Let us consider the important issue which Kuno (2002) does not address: why scrambling does not affect Case marking in standard cases like (49), where the scrambled phrase has to retain accusative Case, although it moves to the edge of vP.

    1. (49)
    1. [TP [vP
    2.  
    1. Kono-hon-o/*ga 1 [vP
    2. this-book-ACC/NOM
    1. Taroo-ga [VP t1
    2. Taroo-NOM
    1. yon]]]-da].
    2. read-past
    1.             ‘Taroo read this book.’

Under the proposed analysis so far, nothing would prevent kono-hon from receiving nominative Case, contrary to fact. Let us assume that once an unvalued feature is assigned a value, it cannot undergo another process of valuation. On this assumption, in order for the scrambled object to receive nominative Case at the edge of vP, the Case feature of the object should remain unvalued when the object stays at the base-generated position. When Transfer applies to the complement of v, the unvalued Case feature (uCase) of the object will be transferred to the interfaces, as illustrated below.

    1. (50)

I would like to propose that the interfaces check whether a Case feature is valued or not in each transferred domain. Under this proposal, once a nominal is sent to the interfaces with its Case feature unvalued, the Case feature is not legible at the interfaces, even though the Case feature of the nominal is valued at the next transferred domain. Thus, the unvalued Case feature within the VP makes the derivation illegitimate. To circumvent this problem, the object has to be given a value of its Case feature at the base-generated position. In this case, the object cannot receive nominative Case after it undergoes scrambling because it has already received Case. It is thus guaranteed that the scrambled nominal does not receive Case after scrambling in (49).

Let us turn to the issue as to how Transfer applies to (44). As discussed in Section 4.2, Transfer does not take place within the koto-phrase because of the absence of a vP-layer. At the matrix CP-phase level the scrambled dake-phrase and the kokto-phrase are transferred together. Since the interfaces check whether a Case feature is valued or not in each transferred domain, as proposed earlier, it is confirmed at the interfaces that the Case-feature of migime-dake has been valued.13

The proposed analysis also captures the unavailability of Accusative-Nominative alternation in the koto-phrase under kokoromi, as shown below.

    1. (51)
    1. Taroo-wa
    2. Taroo-TOP
    1. migime-dake-o/*ga
    2. right.eye-only-ACC/NOM
    1. tumur-u
    2. close-PRES
    1. koto-o
    2. NMLZ-ACC
    1. kokoromi-ta.
    2. try-past
    1. ‘Taroo tried to close only his right eye.’

In order for the argument of the lower verb to receive nominative Case, it has to move to the edge of vP. One possibility is moving to the edge of vP within the koto-phrase. Recall that under the proposed approach, the koto-phrase selected by kokoromi has a vP, in contrast to deki. In this case, the unvalued Case feature makes the derivation illegitimate in the same way as (50). Even if the relevant argument moves to the edge of vP in the matrix clause, the problem cannot be circumvented because the base-generated position and the landing site are transferred separately.

The discussion about the locality constraint on the NCI leads to the conclusion that koto is not a C as a phase head, as argued in Section 4.2. If koto were a phase head, the underlying position of a nominal and its landing site would be transferred separately. One reviewer points out that the koto-phrase under discussion behaves differently from the koto-phrase where tense is realized in (52b).

    1. (52)
    1. a.
    1. Taroo-wa
    2. Taroo-TOP
    1. [eigo-ga/*no
    2. English-NOM.GEN
    1. hanas-u
    2. speak-PRES
    1. koto]-ga
    2. NMLZ.NOM
    1. deki-ru.
    2. can-PRES
    1. ‘Taroo can speak English.’
    1.  
    1. b.
    1. Taroo-wa
    2. Taroo-TOP
    1. [Hanako-ga/no
    2. Hanako-NOM.GEN
    1. kaetta
    2. left
    1. koto]-o
    2. NMLZ-ACC
    1. sitta.
    2. knew
    1. ‘Taroo knew that Hanako left.’

As shown in (52a), the koto-phrase selected by deki does not allow Nominative-Genitive Conversion, in contrast to (52b). Since n is responsible for assigning Genitive case, the koto in (52a) is not n as a phase head either.

Another crucial point of the analysis is that the koto-phrase occupies an edge position of vP, which enables the koto-phrase to escape from the transferred domain at the matrix vP-phase. If the koto-phrase were at the complement of v, it would undergo Transfer separately from the edges of vP and the same problem discussed in (50) would arise. Since this paper adopts the idea that base-generating a theme argument at an edge position of vP is available for stative predicates, not for non-stative predicates, the proposed analysis correctly captures the generalization that the investigated Case alternation is available only for the former.

One of the theoretically important points of this paper is to adopt the assumption that the transferred domain is the complement of a phase head, following Chomsky (2001) among others. This assumption plays a crucial role in the above discussion in that the base-generated position and the landing site of scrambling are transferred separately and the unwanted Case alternation in (49) is correctly excluded. Alternatively, Chomsky (2000) and Bošković (2016) among others have argued that the phase itself, not the complement of a phase head, is transferred. Under the latter approach, the base-generated position and the landing site of scrambling are included in the same transferred domain, which undermines an explanation for the unwanted Case alternation in (49). To the extent that the proposed analysis is on the right track, this paper lends support to the former approach.

As has been argued, Case alternation is available only when the underlying position of a nominal and its landing site of scrambling are included in the same transferred domain. In what follows, the derivation of other cases of Case alternation will be examined in terms of this constraint. Let us consider the derivation of Nominative-Genitive Conversion. Recall that under the analysis put forward in this paper along the lines of Fukui & Nishigauchi (1992) and Fukui (1995), the subject of a relative clause moves out of that clause via scrambling and is merged with nP. The transfer-domain-mate restriction has two implications for the analysis of relative clauses. One is that relative clauses cannot be merged with the complement of n. If the relative clause were merged with the complement of n, the former would be included in the transferred domain at nP. Thus, the landing site and the original position would be in different transferred domains, which blocks the relevant Case alternation.

    1. (53)
    1. a.
    1. Taroo-no
    2. Taroo-GEN
    1. katta
    2. bought
    1. hon
    2. book
    1. ‘the book Taroo bought’
    1.  
    1. b.

On the other hand, if the relative clause is exempt from the transferred domain at the nP-level, as illustrated in (53b), the base-position and the landing site are included in the same transferred domain.

Another implication is that Japanese relative clauses involve no CP-layer, which is independently argued by Murasugi (1990). Otherwise, Transfer would apply to the complement of C within the relative clause, including the base-position of the subject. Since the landing site of the scrambled subject would be transferred in the different domain, the relevant Case alternation would be blocked, contrary to fact.14

The proposed approach can be extended to the following Case alternation, where genitive Case can be converted into nominative Case.

    1. (54)
    1. Taroo-no/ga
    2. Taroo-GEN.NOM
    1. hone-ga
    2. bone-NOM
    1. ore-ta.
    2. break-past
    1. ‘Taroo’s bone broke.’

This alternation results from scrambling of Taroo from the edge of nP to that of vP, as illustrated below.

    1. (55)

The transferred domain at the nP-phase is hone, not including the edge of nP. The edge position of nP is not included in the transferred domain at the vP-phase either. The edge of nP is transferred, together with the landing site of scrambling, when Transfer applies at the CP-phase level, finally. The transfer-domain-mate restriction is thus satisfied.

4.5 A comparison with alternative analyses

Recall that Koizumi (1994) argues that the example given in (13), which is repeated as (56), is problematic for Tada’s (1992) analysis. He points out that Tada’s analysis cannot capture the wide scope of the dake-phrase because under his analysis, the landing site of NOs (i.e. [Spec, AGRoP]) is below negation.

    1. (56)
    1. Koizumi (1994: 222)
    1. John-ga
    2. John-NOM
    1. migime-dake-ga
    2. right.eye-only-NOM
    1. tumur-e-na-i.
    2. close-can-not-PRES
    1. ‘John cannot close only his right eye.’ (only > not > can)

The same criticism holds for the proposed analysis as well because the landing of NOs is an edge position of vP, which is below negation. In order to capture the interpretation in question, the NO has to undergo further movement from the edge position to a position higher than negation. Given that scrambling is an instance of IM and freely available, it is reasonable to suggest that the relevant movement is scrambling. The scrambling from an edge position across negation is independently motivated from the wide scope in (56). It is also responsible for the wide scope of the subject QP over negation in the following example.

    1. (57)
    1. Subete-no
    2. all-GEN
    1. gakusei-ga
    2. student-NOM
    1. sono-hon-o
    2. the-book-ACC
    1. yoma-nakat-ta.
    2. read-not-past
    1. ‘All the students did not read the book.’

Recall that under the proposed analysis, Japanese subjects do not obligatorily move to [Spec, TP] but can stay within vP, which is in the scope of negation. I suggest that the subject QP also undergoes scrambling from the base-generated position to a higher position than negation like (56), when it takes wide scope.

A question arises as to why a similar application of scrambling is not available with the AO in the following example. If available, the wide scope of the AO would be expected in (58), contrary to fact.

    1. (58)
    1. Koizumi (1994: 221)
    1. John-ga
    2. John-NOM
    1. migime-dake-o
    2. rightv.eye-only-ACC
    1. tumur-e-na-i.
    2. close-can-not-PRES
    1. ‘John cannot close only his right eye.’ (not > can > only)

Importantly, other quantified objects such as subete-no hon ‘all the books’ can take easily wide scope over negation under the relevant syntactic context, as shown in (59) (cf. Kato 1985; Kataoka 2006 among others).

    1. (59)
    1. John-ga
    2. John-NOM
    1. subete-no
    2. all-GEN
    1. hon-o
    2. book-ACC
    1. yom-e-na-i.
    2. read-can-not-PRES
    1. ‘John cannot read all the books.’

Given this, I speculate that the absence of the wide scope of the dake-phrase in (58) is sort of exceptional and it will be reasonable to reduce the absence of the relevant wide scope to the intrinsic property of the particle dake. At this point such property is not well understood and further investigation of it is left for future research.

Let us consider another apparent problem for the proposed analysis, under which (60b) involves two instances of movement: the NO moves to the edge of vP across dareka-ni, which is followed by the movement of dareka-ni, as illustrated in (60c). One of the reviewers points out that the proposed analysis would expect that (60b) would allow the NO to take wide scope over the dative argument because the NO c-commands the lower copy of the dative argument in the course of the derivation. In contrast, the relevant wide scope reading is not available in (60a), which has no such movement.15

    1. (60)
    1. a.
    1. Taroo-wa
    2. Taroo-TOP
    1. dareka-ni
    2. someone-DAT
    1. 40izyoo-no
    2. more.than.40-GEN
    1. sigoto-o
    2. job-ACC
    1. makase-ru
    2. leave-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘Taroo can leave someone more than 40 jobs.’
    1.  
    1. b.
    1. Taroo-wa
    2. Taroo-TOP
    1. dareka-ni
    2. someone-DAT
    1. 40izyoo-no
    2. more.than.40-GEN
    1. sigoto-ga
    2. job-NOM
    1. makase-ru
    2. leave-PRES
    1. koto-ga
    2. NMLZ-NOM
    1. deki-ru.
    2. can-PRES
    1. ‘Taroo can leave someone more than 40 jobs.’
    1.  
    1. c.
    1. Taroo-wa dareka-ni1 40 izyoo-no sigoto-ga2 [t1 t2 makase-ru] koto-ga deki-ru.

However, contrary to this expectation, there is no significant contrast between (60a) and (60b): it is difficult to obtain the wide scope reading of 40izyoo-no sigoto in (60b) in a way similar to (60a). It is true that the proposed analysis offers no clear answer for the difficulty to obtain the relevant wide scope reading in (60b); however, other alternative approaches also fail to do so. Under Koizumi (1994) and Nomura (2005), (60b) could have the derivation where the NO moves to [Spec, TP], which is followed by the movement of the dative argument, in the way illustrated in (60c). A-movement can yield a new scope relation that is otherwise unavailable, as shown in (61), where someone can take wide scope over likely. It is expected that the A-movement of the NO could yield the wide scope of it.

(61) Someone1 is likely to t1 win the race.

Tada’s (1992) analysis also has a similar prediction because the landing site of the NO ([Spec, AGRoP] under his analysis) is higher than the underlying position of the dative argument, in a way similar to the proposed analysis.

5 Conclusion

Taking a closer look at the variant given in (3), this paper has argued that the NO does not stay at its base-generated position but overtly undergoes movement. Extending Fukui & Nishigauchi’s (1992) and Fukui’s (1995) analysis of Nominative/Genitive Conversion to the Case alternation under investigation, it has been proposed that the overt movement is scrambling. Specifically, the derivation of a NO involves scrambling of an object from its base-generated position to an edge position of vP, where it receives nominative Case, as illustrated in (62). Although scrambling itself is optional, in order for the object to be assigned nominative Case, scrambling takes place obligatorily, because what can be assigned at the base-generated position is accusative Case, not nominative Case. This is an explanation for the obligatory nature of the movement.

    1. (62)

On the other hand, when scrambling does not take place, accusative Case is assigned at the base-generated position. Since the Case alternation is contingent on application of scrambling under the proposed analysis, the optionality of the Case alternation is captured in terms of that of scrambling. It has also been argued that the Case alternation is restricted to the case where the landing site and the base-generated position should be included in the same transferred domain.

The proposed analysis is crucially based on the assumption that the transferred domain is the complement of a phase head (Chomsky 2001). If the analysis is on the right track, this paper lends support to this characterization of transferred domains, contrary to an alternative characterization recently argued by Bošković (2016), where the transferred domain is the phase itself. Also, this paper has argued that the proposed analysis based on the structure-based Case valuation approach is more suitable than the analyses based on the Agree-based approach such as Tada (1992), Koizumi (1994), and Nomura (2005). To the extent that the proposed analysis is successful, the former approach is more plausible to Japanese Case phenomena.