Kazuko Inoue

 

Case marking vs. Case checking in Japanese generative grammar:
An alternative proposal

 

Kazuko Inoue
Kanda University of International Studies
Research Center of Excellence in
Language Sciences

< 0 > Introduction*

Marantz (1995) concludes the description of the Minimalist Program by saying, "Syntax reduces to a simple description of how constituents drawn from the lexicon can be combined and how movement is possible. The computational system, this simple system of composition, is constrained by a small set of economy principles, which Chomsky claims enforce the general requirement, 'do the most economical things to create structures that pass the interface conditions'... Chomsky's vision of the end of syntax should have the positive consequence of forcing syntacticians to renew their interface credentials by paying serious attention to the relevant work in phonology and semantics."

The reduction of syntax forces a reevaluation of the past important work in traditional syntax, which now looks like a machine loaded down with too many functions. It also urges syntacticians to see whether or not the new machine can deal with relevant research results adequately. An attempt to recast some of the important past findings into the new framework may reveal fundamental theoretical problems. What is important is not only to see that past findings can be described and explained by the new theory but also to check whether or not the new machinery leaves out important syntactic facts as irrelevant to computation.

Morphology, as part of phonology in the Minimalist Program (hereafter MP), carries a very much reduced function, although it had played an important function as a part of syntax or an interface of syntax and phonology in generative grammar up to the GB theory, called the Extended Standard Theory (EST). The nature of morphology in EST is both syntactic and phonological. As a case in point I would like to take up problems involved in its case system. This is a preliminary attempt to recast the case marking system proposed in the EST framework into that of the MP. It is not so much important to see whether or not this attempt is successful as to find out the relevance of the problems brought up by MP.

This work consists of three chapters; Chapter 1 gives the survey of the past works, which reveals that the traditional case marking system cannot be maintained in the MP framework, due to the condition of inclusiveness that "no new objects are added in the course of computation apart from rearrangements of lexical properties".1

In Chapter 1 I proceed to present alternative proposals in the MP framework, and make the claim that É∆-checking rather than Case2 checking be the universal NP licensing mechanism. The following two chapers give evidence from Japanese to support this claim. Specifically, Chapter 2 deals with Japanese “subjectless sentences”, which support the claim against Case checking, showing that the obligatory nominative Case checking currently employed in the MP program cannot be maintained. Chapter 3 gives arguments for the same claim , specifically pointing out that the accusative Case checking does not account for the facts about Japanese particle o, usually called the accusative case particle.

In conclusion, I claim that NPs are licensed by the É∆checking of NPs mediated by postpositions. The other feature checking involves the features [+D], [+Topic], and [+Focus], which serves to guarantee sentence-hood by the following two mechanisms: establishment of the subject-predicate relation and the topic/focus-comment relation.

 

 

< 1 > Chapter 1 Survey of the past works

Section 1. 1 of this chapter gives an outline of the relevant proposals made by EST with regard to the Japanese case system, together with the necessary minimum assumptions underlying them. Section 1.2 and 1.3 summarize recent works related to the MP Case checking system, the former containing arguments against the system and the latter a proposal for an alternative within the MP framework. Section 1.3 proposes another alternative to account for important findings made by EST.

 

< 1.1 > Case-marking in the traditional theory

1.1.1. The structural vs. the inherent case

The distinction between structural and inherent case has been maintained all the way from EST up to MP. Though all cases, both structural and inherent, are marked with postpostions in Japanese, noun phrases without postpositions (unmarked noun phrases) are inserted into the subject and object positions in the EST framework, following Kuroda (1965). Kuroda's linear case marking (Kuroda 1965, 1978) marks the first unmarked NP with ga and the remaining unmarked NP with o. Sentences with complex verbs and adjectives like kak-ase ('write-cause') and tabe-tak ('eat-want') are analyzed as involving embedded complement sentences with affixal predicates (verbs and adjectives) like sase, (Caus(ative)) rare, (Pass(ive)) tak (Desid(erative)) as matrix predicates. Linear case marking is proposed to apply cyclically following syntactic operations like Equi-NP Deletion (or Counter Equi- NP Deletion) and Verb Raising3. There have been alternative proposals for case marking, like the one applying case marking to S-structure on the basis of structural information (Inoue 1988, 1991), or the one using configurational case marking (Takezawa 1987). However, these are all based on the assumption that case marking involves the closely interrelated syntactic operations such as Equi- and Counter Equi-NP Deletion and Verb Raising. Thus the structural case marking is a strictly syntactic operation that introduces case particles in syntax. This is certainly counter to the MP condition of inclusiveness as well as to the traditional notion of universal abstract Case.4

 

1.1.2. Verb Raising

Verb Raising is one of the key syntactic operations in EST. This is a natural device to deal with the derivation of morphological complexes which cause alternations of grammatical relations. A similar operation, also called Verb Raising, employed by languages like English, which does not involve morphological derivations affecting grammatical relations, is different in its effect from the one proposed for Japanese grammar in the EST framework. Verb Raising is necessary in English for accusative Case checking and the subject verb agreement mediated by Tense. In this case there is no change of grammatical relations as a consequence of Verb Raising.

 

< 1.2 > Case checking and licensing of noun phrases5

In MP NPs bearing structural cases must be licensed by case feature checking triggered by functional heads. In languages like English in which there is subject-Tense agreement, nominative Case checking by the head (T) of the Tense Phrase (TP) is well motivated. In languages like Japanese without agreement of this type, agreement based Case checking is hardly motivated. Kuroda (1988) claims that "the parametric difference between English and Japanese consists simply of the following: Agreement is forced in English; it is not in Japanese." Kuroda further argues that "there is little direct evidence of Case theory in Japanese," and claims, "In English Case-marking is forced, but in Japanese it is not. It can be left unenforced in Japanese since (or, since it is not forced,) there is another lower-case case-marking mechanism to license Max(N)'s, which assigns ga and o to them." Kuroda thinks that this lower-case case-marking is applied to D-structure.

Marantz (1992) points out the weakness of Case checking as a case realization mechanism and argues, "The mapping between semantic roles and argument positions, augmented by the subject requirement of EPP6, is sufficient to license NPs in argument positions." Marantz proposes that actual cases, rather then abstract Cases be assigned in Morphology according to the disjunctively ordered case realization hierarchy.7

Watanabe (1996) on "Ga-No Conversion" , exemplified by the sentences in (1), is an interesting work in this respect.

(1) watasi-tati ga/no ason-da kooen ĎThe park where we played.í

we Nom/Acc play- Past park

According to his claim the No version is the result of the NP with the inherent case particle staying in the VP internal position without having its Case checked in the overt syntax. The Case checking of this NP takes place in the covert syntax in terms of the Spec-Head relation.

 

< 1.3 > DP licensing through Spell-Out by Fukui and Takano (1996)

Supporting the claim by MP that linear ordering does not play a role either in LF or in the computation from Numeration to LF, and that it is part of the phonological component, Fukui and Takano propose the operation of Demerge in the phonological component. Demerge applies to the topmost construct recursively and detaches the maximal projection first and start the concatenation from the first detached element. (See Fukui and Takano(1996) for a formal and detailed presentation of Demerge and Concatenate.) The result is the Specifier-Complement-Head order, namely the head final order. The Spec-Head-Comp order in the head initial languages is the result of Verb Raising triggered by the strong feature of the functional head v. The presence or absence of the functional head with the strong feature of attracting a verb to its specifier position accounts for the linear order variation in languages.

According to the above theory, Japanese does not undergo Verb Raising, since its head final order is base generated. As a consequence the standard analysis of accusative Case checking based on Verb Raising is not available for Japanese. This is a desirable consequence in view of the lack of evidence in Japanese for the relevance of the abstract Case system. Roughly speaking, Fukui and Takano have come up with a conclusion similar to Kuroda's (1988), that is, in Japanese, subject and object NPs are licensed by ga and o8, not by Case checking. The significance of their work lies in the account of the process of uninterpretable feature "elimination". The case feature carried by NP and the feature "Assign the accusative case" ([Ā{Acc]) carried by the transitive verb must be eliminated before the derivation terminates in LF, since they are not interpretable. F and T claim that "Case particles make Case features visible to Spell-Out, " which triggers the elimination of the case feature of an NP because Spell-Out takes away visible features relevant to the phonological component. Without the case feature, the NP proceeds to LF and the derivation successfully terminates in the covert component, if the other conditions are met. F and T further assume that the accusative case in Japanese is an inherent case and that it is linked to a particular É∆-role. This means that [+Acc] on the transitive verb can stay on in the covert component as an interpretable feature.

As for the nominative case, it is assumed that T in Japanese does not carry the case feature "Assign the nominative case". The case particle ga makes the case feature of its complement NP visible, which leads to the elimination of this case feature by Spell-Out. The case particle ga licenses the multiple subject NP in the same way.

This is a very rough outline of F and T's approach to the Japanese case system within the MP framework. This approach seems to be on the right track, but there are some serious problems yet unsolved. How to account for the limited and often syntactically determined distribution of ga is one of them, which will be discussed in the next section.

 

< 1.4 > Reanalysis of old issues of the Japanese case system

This section presents the result of an attempt to recast into the framework of MP some of the central issues about the case system in the generative grammar of Japanese in the EST framework.

The organization of this section is as follows: Section 1.4.1 describes how the verb and argument relations can be treated in MP. Assuming further that Fukui and Takano are correct in the process of the elimination of the [+Acc] feature, there are some problems involved in their NP licensing by the case particle ga. The following proposal is made in order to solve problems revealed thus far: The Japanese particles other than ga must be treated as inherent case bearers, namely postpositions. Some supporting evidence is given in this section, and arguments along this line are continued in Chapter 2 and 3. I further assume that all arguments, external and internal, originate in the VP internal positions, the so called “VP internal subject hypothesis”9. My crucial claim is that ga does not appear in the VP internal argument positions. It is taken to be a morphological realisation of the feature [- Acc], namely, the feature without the positaive case feature. Ga is a default case particle in this system. The discussion of the distribution of ga in this section is restricted to sentences with simple predicates. Section 1.4.2 extends the discussion to those with complex predicates10.

 

1.4.1. An account of the distribution of Japanese ga and o

The inclusiveness condition prohibits insertion of any element in the computation from Spell-Out to PF11. This means that the case marking of unmarked NPs in syntax in the sense of EST is not available. In my framework, postpositions are selected from the beginning and put into computation, and merge with NPs forming postpositional phrases (PP). Predicates (verbs and adjectives) have É∆features specified in the form of É∆grids. Postpositions carry their intrinsic semantic features including those linked to É∆features checked by predicates. In this way PPs assume É∆ features of predicates. Predicates merge with PPs using each of theirÉ∆features as the checkers of the É∆features of PPs. Predicates carry the structural case feature [Ā}Acc] as well12. Let us call the checking of É∆features carried by PPs by those of predicates É∆checking. Actually PPs are arguments and nonarguments of predicates in this framework, but I will once in a while use the term “NP” in place of “PP” as standing for an argument or a nonargument, in order to make the comparison with the standard framework easier.

If we decide, following F and T, that o is a postpostion (an inherent case particle in their term), then the NP licensing proceeds as they specify. However, some problems remain, first with their assumption about case features carried by NPs. They seem to assume that NPs are assigned case features at the time they are selected and put into the computation process, as generally assumed in the MP framework. If we assume that postpositions select NPs as their complements, they determine the inherent case features of NPs. Moreover, each postpostion carries an interpretable semantic feature linked to a É∆feature, so that the elimination of inherent case features is unnecessary. The second problem is with ga, since this particle is not the realization of a unique É∆feature. The É∆features linked toga can be [+ Agent], [+ Experiencer], [+ Theme], and so on. Ga is a structual case particle in this sense. Moreover, ga licenses the NP simply as its complement, not as the NP functioning as the subject. F and T seem to regard the NP licensing as complements of ga is enough. Furthermore, it is not the case that multiple subjects co-occur with any type of verbs or adjectives, as shown by the sentences in (2).

(2)a. Taroo ga ie ga eki kara too-i

Nom house Nom station from far-Pres

ĎIt is Taro whose house is far from the station.í

b. *Taroo ga hon ga kat-ta ĎTaro bought a book.í

Nom book Nom buy-Past

There is also the well-known case of the object of the stative verb carrying ga. as in (3).

(3) kare wa watasi no hanasi ga yoku wakar-u

he Top my story Nom well understand-Pres

“He understands me very well.”

All these facts need to be accounted for.

Now let us assume thatÉ∆features of predicates check those of PPs; [+ Agent] checks that of the PP with ni1, or [+ Experiencer] that of the PP with ni2, [+ Theme], [+ Path], [+ Passing Point] those of PPs with o1, o2, and o3 respectively. Assume further that ga is not a postposition but the realization of the feature [-Acc] transferred from the head with this feature (a stative predicate and T) to its complement or its specifier. Agreeing with Kuroda (1988) that Tense (T) is without a Case feature (in our terms [-Acc]), I maintain that an NP with a postposition gets the [-Acc] feature of T copied onto it, when it comes to occupy the position of the specifier of T as the result of attraction by the strong D feature of T. The D feature of T, which attracts the same feature of DP (NP is used in this paper for ease of exposition), is sometimes assumed to be weak in Japanese. However, the features assumed by T can be non-uniform. For example, it can have either [+finite] or [-finite]. [+finite] can be sub-divided into [Ā}D], [+D] representing the strong D feature, which attracts the same feature of the external argument of verbs and adjectives and moves it to the position of its specifier for feature checking, thus licensing this NP. The strong D feature of T is cancelled by the attachment of the same feature from the NP. The [-Acc] feature transferred from T to the NP is realized as ga in Morpholoy.

Let us follow this procedure using an example.

(4) Taroo ga hon o yon- da

Nom book Acc read-Past

'íTaro read this book.'

As the result of É∆-checking , the following combination is permitted.

(5) Taroo ni hon o yon-da (Ni 1 is the postposition for [+ Agent].)

Due to the feature [+D] 'Taroo ni' is raised to the position of the specifier of T. The cancellation of the D feature of T and the transference of [-Acc] to 'Taroo ni' is carried out. Now the NP with the postposition is sent to Morphology with the feature [-Acc], which is phonologically realized as ga. In EST the deletion of a case particle caused by adjunction of another case particle is one of the necessary operation under the hypothesis of the cyclic application of case marking. Postpositions before case particles are generally undeletable. However, the attachment of ga to postpositions is an exception to this generalization. (6b) is the result of deleting ni due to the ga attachment.

(6)a. watasi wa kono isu ni suwari-tak-13 i

I Top this chair on sit- Desid-Present

ĎI want to sit on this chair.í

b. watasi wa kono isu ga suwari-yasu-i

I Top this chair sit- easy- Pres

ĎThis chair is easy for me to sit on.í

There is a certain amount of supporting evidence for my claim that the nominative Case checking by T is not applicable to Japanese. One is the fact that there are tensed sentences without ga, but only with a postposition in the subject position, as shown by the sentences in (7). This fact casts doubt about the obligatory nominative Case checing by T, the head of TP (Tense Phrase). Chapter 2 elaborates this point using sentences without ga phrases.

(7) a. Katoo-san kara hon o okut-te ki -taĀ@

Mr. from book Acc14 send- come-Past

'Mr. Kato sent (me) a book.'

b. watasi to otooto de mondai o syori si -ta

I and brother with problem Acc deal with do-Past

'My brother and I dealt with the problem.'

The second piece of evidence comes from examples where there is an optional choice of either ga or a postposition15is permitted, one example being the case of the Ga-No conversion discussed by Watanabe (1996).

(8) watasi ga/no kai -ta hon

I Nom/Gen16 write-Past book

'the book I wrote'

Another example of this optionality comes from a class of stative verbs, which permit both ni and ga for the subject.

(9) a. kono akanboo ni (wa) oya no kotoba ga wakar -u

this baby to Top parent Gen speech Nom understand-Pres

'This baby understands his parents' speech.'

b. kono akanboo ga oya no kotoba ga wakar -u

this baby Nom parent Gen speech Nom understand-Pres

ĎThis baby understands his parents' speech.'

The ni in (9a) is the postposition with the É∆feature [+ Experiencer]. The optionality attested in these constructions is naturally explained in the framework proposed here.

 

1.4.2. Sentences with complex predicates

The same optionality is attested in sentences with complex predicates.

(10) a. Taroo ni kono ji ga yom- e mas-u ka

to this letter Nom read-Pot Pol-Pres Q17

'Can Taro read this letter? Is this letter readable to Taro?'

b. Taroo ga kono ji ga yom-e mas-u ka

Nom

'Can Taro read this book?'

There are two possible readings for (10a), but as far as its first reading is concerned, the alternation of the particle is a kind of free variation. This is again accounted for naturally in this framework.

Sentences with complex predicates reveal another type of optionality in the object position, often treated as the case of optional Verb Raising followed by Case Absorption within the GB theory. Some examples follow:

(11) a. Taroo ga kono ji ga yom-e mas-u ka (=(10b))

b. Taroo ga kono ji o yom-e mas-u ka

In these sentences the matrix verb e (Potential) embeds the VP as its complement. The complement verb yom with [+Acc] is raised and adjoined to e with [-Acc]. Now the case feature [-Acc] of the matrix verb e is retained, cancelling [+Acc] of the complement verb yom. This is a version of Case absorption. Thus, (11a) is derived by Verb Raising followed by Case Absorption. Neither Verb Raising nor Case Absorption has affected (11b). These are the accounts permitted by the GB theory. The sentences with the desiderative morpheme tak show the same alternation, as in (12).

(12) watasi wa eiga ga / o mi- tak- i

I Top movie Nom/Acc see- want-Pres

'I want to see a movie.'

Since more detailed study is needed to account for the findings of the previous works, especially those related to Case Absorption, I will show only what can be done within our framework.

Affixes like the Potential rare/re and the Desiderative tak can optionally have the strong v feature which attracts the v feature of the complement verb. As the result of this feature attraction, [-Acc] of the matrix verb is transferred to the complement with the postposition o. The rest of the process is the same as that described in Section 1.4.1.

Except in the case of the feature attraction by the strong D feature of verbal and adjectival affixes like the Potential and Desiderative morphemes, I assume Verb Raising is no longer a syntactic operation, but a morphological operation, which can be called affixation.

Finally, the problem related to the multiple subject construction finds the following solution.

As is well-known, multiple subjects are permitted mostly in sentences with adjectives and adjectival verbs as their predicates. Sentences with the present tense denoting habitual aspectual meaning permit multiple subjects as well.

(13)a. nooson ga zyosei ga roodoo-zikan ga nagak-i

farming villages Nom women Nom working hours Nom long-Pres

'As for farming villages, womenís working hours are long.í (Adjective)

b. yasei-doobutu ga mesu ga karada ga ganzyoo da

wild animal Nom female Nom body Nom strong Cop-Pres

ĎAs for wild animals, females are strongly built.í (Adjectival Verb)

c. daitokai ga yoru ga zinkoo ga gekigen su-ru

big cities Nom night Nom population Nom drastic decrease do-Pres

ĎAs for big cities, at night its population decreases drastically.í

(T with habitual aspect)

To account for this distributional restriction, I propose that the [+D] feature of T be subdivided into [Ā}generic]. The feature [+generic] is not only responsible for the habitual aspectual meaning of the present tense but also compatible with stativity of adjectives and adjectival verbs. My claim is that the T with [+generic] projects multiple specifier positions. The feature [-Acc] carried by T is copied onto the NPs in the specifier positions of T. The NPs with [-Acc] gets the phonological features of ga in Morphology.

This is a reasonable solution within the framework proposed in this paper. However, an additional account may be necessary, since ga in these sentences invokes a focus reading on the leftmost NP with ga. It is plausible that Focus may play a role in this connection as a functional head. Furthermore, the sentences in (13) have counterparts with wa with the possibility of replacing ga all the way, which means that the idea of Focus as a functional head must be studied in relation to Topic phrases. I will come back to this problem in Chapter 2.

In sum, in this section the following assumptions have been presented to be tested in Chapter 2 and 3:

(a) VP internal subject hyposthesis: external and internal arguments of a predicate (a verb, an adjective, an affixal verb and adjective like sase (Caus(ative)), rare (Pass(ive)), tak (Desid(erative)), and so on) undergo É∆ checking by the É∆features carried by predicates. For example, [+ Agent] of PPs withni1 is checked by the same feature carried by the predicate, [+ Experiencer] of PPs with ni2, [+ Goal] of ni3 PPs, [+ Location] of ni4 PPs, [+ Patient] of o1 PPs, [+ Theme] of o2 PPs, [+ Path] of o3 PPs, [+ Instrument] of de1 PPs, [+Cause] of de2 PPs, [+ Place] of de3 PPs, and so on undergo the same checking procedure.

(b) Postpositions o, ni, de, kara, e, to, and so on merge with NPs (or DPs).

(c) Thus, NPs, arguments and nonarguments, are licensed by the É∆checking mediated by postpositions.

(d) Ga is a default case particle, more specifically, phonological realization of the feature [-Acc], carried by stative predicates and T, the head of TP (Tense Phrase).

(e) The external argument of a predicate has its [+D] feature checked by the same feature carried by T. T has either plus or minus D feature. This means that the D feature checking of the external argument of the predicate is optional and the external argument may stay within VP.

 

< 1.5 > Conclusion

It is to be noted that postpositions are all treated in this paper as phonological feature complexes with semantic features linked to É∆features, which are checked by the same feature carreied by each predicate. Ga is treated as a phonological realization of a syntactic feature [-Acc]. The syntactic operations by means of formal features proposed here more or less conform to the prescriptions given by MP. However, a structural feature [-Acc] is crucially used in this framework, which reveals intrinsic nature of distribution of case particles in the traditional sense. They are really a morphological realization of syntactic functions. This means that syntactic aspects of Morphology must be seriously restudied.

The particle o is treated in this paper as a postposition without a syntactic function. However, there are good reasons for regarding o in some contexts as a case particle in the traditional sense. Instead of resorting simply to the mechanism of accusative Case checking, syntactic functions of o must be thoroughly reexamined in the framework of MP. Chapter 3 deals with this problem, and comes up with a conclusion that there is no fast and hard line of demarcation between the structural and inherent case particle o in syntactic functions. A direction of future researches to account for this problem is suggested at the end of Chapter 3.

 

Notes to Chapter 1:

* R. Martin read all the three chapters of this paper, and gave valuable comments, some of which contributed to the revision of the paper. The rest will be taken into consideration in future studies. Comments from R. Desilva, K. Murayama, S. Tonosaki, and Y. Ueda helped me in revising a preliminary draft . This chapter is a revised version of “A Note on Relevant Questions Involvd in the Case System” (Inoue 1997a).

1 See Chomsky (1995), p. 228.

2 “Case” stands for “abstract case” assumed in the GB theory.

3 Verb Raising used to be formulated as adjunction of a head to another head. This operation has been used extensively in Japanese gererative grammars.

4 Martin used the expression “the traditional notion of abstract Case” in his comment. This implies that there have been a variety of proposals for the Case system. Fukui and Takanoís case system, introduced in Section 1.3, is a proposal for morphological Case checking.

5 NP is now analysed as complement of D (Determiner), the head of DP ( Determiner Phrase). In this paper NP is used in place of DP for convenience of exposition.

6 EPP stands for Extended Projection Principle, which states that the subject (the specifier of IP) is obligatory. This principle, together with Projection Principle are restated in terms of EPP/D features and feature-checking.

7 The case realization hierarchy Marantz proposes is as follows:

a. lexically governed case

b. "dependent" case (accusative and ergative)

c. unmarked case (sensitive to environment)

d. default case

8 F and T call ga and o case particles, though they treat o as a bearer of an inherent case. In this paper inherent case bearers are called postpositions, rather than case particles.

9 To my knowledge, most arguments supporting the VP internal subject hypothesis are theory internal, rather than factually based. For example, the possiblity of allowing the uniform procedure for É∆role assignment both to the external and internal arguments is often mentioned. Ueda (1997) gives an argument for this hypothesis using data from Japanese. In brief, the argument goes as follows: In English negative polarity items (NPI) such as “anyone, anything” are not permitted in the subject position due to the licensing condition that negative polarity items must be c-commanded by NEG. The sentences in (i) are examples.

(i)a. *Anyone didnít go there.

b.*Anything, John didnít eat.

c. *Any unicorns we never sighted. (Linebarger, 1987)

Japanese negative polarity items like daREMO and naNIMO appear in the subject position, as in (ii)

(ii) daREMO Taroo o tanzyoo kai ni syootai-si-nai

NPI Acc birthday party to invite- NEG-Pres

ĎNobody invites Taroo to his birthdayí

(daREMO with the stress on REMO is NPI.) (Ueda (32))

Unless we assume that the Japanese subject is in the VP internal position c-commanded by NEG, we cannot account for this fact.

As for the positive polarity items (PPI), the English Ďsomeoneí and Ďorí and the Japanese DAreka (with the stress on DA means Ďsomeoneí) or ka , the distributional facts are reversed.

(iii)a. Someone didnít say hello to John. (Ibid. (37a))

b. *kyoo wa DAreka koko ni ko-nai hazu da

today Top someone here to come-NEG-must-Copula

ĎToday someone cannot come here.í (Ibid. (36b))

There is a constraint on PPIs prohibiting them from appearing in a position c-commanded by NEG. Due to the nominative Case checking, the English subject occupies a position higher than NEG, so that (iiia) is not subject to this constraint.

The fact that (iiib) is ungrammatical indicates that in Japanese the subject stays in the VP internal position c-commanded by NEG.

10 The term "complex predicate" is used for verbal and adjectival complexes consisting of a verb and a verbal affix like kak-e ('can write') and nomi-tak ('want to drink').

11 To be more specific, Chomsky (1995) states, "Select is inoperative in the phonological component: no items can be selected from the Numeration in the computation from Spell-Out to PF. The operation Select is available to the covert component, however,...But if an item with phonological features are selected, the derivation will crash at LF."

12 In addition to their intrinsic semantic features, postpositions are assumed to carry the structural feature [+ D], [- D], or [Ā} D]. See the discussion in Chapter 3.

13 The final consonat of tak is changed to i when followed by the Present morpheme i and then deleted.

14 Even though I do not treat o as the accusative case particle, the abbreviated form "Acc" is used for convenience.

15 Postpositions are considered in this paper to be phonological feature complexes with É∆features. Avoiding the cumbersome expression “phonological realization of the postposition”, I use the simple term “postposition” for convenience.

16 Watanabe claims that no in Ga/No conversion is not the genetive case particle. However, "Gen" is used here for the same reason as that for "Acc".

17 Pot and Pol stand for Potential and Polite respectively.

 

 

 

< 2 > Chapter 2:Ā@A sentence without a nominative noun phrase*

It is generally assumed that, in a language with an accusative case system, a sentence with a finite predicate must have the subject NP marked with nominative case. The abstract Case system developed within the “principles and parameters” approach is based on this assumption, using the mechanism of the nominative and accusative Case assignment for licensing the NPs with the structural functions of sentence building, namely the subject and the object.

The same assumption was adopted for the analysis of Japanese, one of the accusative languages, already within the framework of the extended standard theory (EST). Kuroda (1978), for example, posits a case filter, screening out a structure without the nominative case ga at the end of each derivational cycle. However, there exist quite a few types of finite (or independent) sentences without any nominative case, (henceforth called “subjectless sentences”). The sentences in (1) are some examples.

(1)a. ketueki sentaa de ketueki-gata siiru o o- hari-si- mas-u

blood center at blood type seal Acc** Hon- paste-do Pol-Pres

ĎThe blood center will paste the seal of your blood type.í

(Haraguchi)1

b. mazu hotondo no baai uti no hoo de yat-te ori-mas-u (Imoto)

first most cases our part by do-Prog-Pol-Pres

ĎIn the first place, in most cases, we are doing (it) on our part.í

c. watasi-domo de uketamawaru ka, arui wa o-kyaku-sama no hoo de

we by take order Q or Top Hon-customerís part by

go- yooi nasaru ka. (Imoto)

Hon-prepare Hon Q

ĎWhether we take your order or you get it on your part...í

d. Taroo to Hanako de bokoo o otozure-ta (Hoshino)

and by alma mater Acc visit- Past

ĎTaro and Hanako visited their alma mater.í

There is another type of sentences in which the postposition kara is preferred to the case particle ni marking the subject of the complement sentence, as in (2b).

(2)a. Ueno-san ga watasi ni Katoo-san o syookai-si-ta

Nom I to Acc introduce- Past

ĎMr. Ueno introduced Mr. Kato to me.í

b. Katoo-san ga Ueno-san kara/??ni watasi ni syookai-s-are-ta

Nom from by to introduce-Pass-Past

ĎMr. Kato was introduced to me by Mr. Ueno.í

These facts support my assumptions presented in 1.4.2, which are repeated below. (a) VP internal subject hyposthesis: external and internal arguments of a predicate (a verb, an adjective, an affixal verb and adjective like sase (Caus(ative)), rare (Pass(ive)), tak (Desid(erative)), and so on) undergo É∆ checking by the É∆features carried by predicates. For example, [+ Agent] of PPs with ni1 is checked by the same feature carried by the predicate; [+ Experiencer] of PPs with ni2, [+ Goal] of ni3 PPs, [+ Location] of ni4 PPs, [+ Patient] of o1 PPs, [+ Theme] of o2 PPs, [+ Path] of o3 PPs, [+ Instrument] of de1 PPs, [+Cause] of de2 PPs, [+ Place] of de3 PPs, and so on undergo the same checking procedure.

(b) Postpositions o, ni, de, kara, e, to, and so on merge with NPs (or DPs).

(c) Thus, NPs, arguments and nonarguments, are licensed by É∆checking mediated by postpositions.

(d) Ga is a default case particle, more specifically, phonological realization of the feature [-Acc], carried by stative predicates and T, the head of TP (Tense Phrase).

(e) The external argument of a predicate has its [+D] feature checked by the same feature carried by T. T has either plus or minus D feature. This means that the checking of the D feature of the external argument of the predicate is optional and the external argument may stay within VP.

The purpose of this chapter is to single out true subjectless sentences from among sentences with de and kara phrases in the subject position, in order to support the assumptions given above.

The findings about subjectless sentences have led to the conclusion that in Japanese NPs are licensed by É∆checking and that there are two sentence building mechanisms, one is based on the subject - predicate relation, and the other based on the topic /focus - comment relation, and that the subjectless sentences belong to the latter type.

Section 2.1 gives a preliminary classification by checking seemingly subjectless sentences to see whether these are results of deletion of underlying subjects. Four classes emerge as candidates for subjectless sentences, one with agentive NPs followed by the postposition de, next with agentive NPs with kara, the third with agentive NPs with delimiters like mo (also), sae (even), demo (for example), made (even), and the last with topicalized NPs. Section 2.2 gives a full account of subjectless sentences with NPs followed by de and kara in place of ga. Section 2.3 is devoted to the analysis of subjectless sentences with NPs followed by delimiters. Section 2.4 takes up topicalized sentences not only with wa but also some other forms like to ie ba , -tte (speaking of), nante, nara (as for) that play a functional role similar to the topic marker wa. Section 2.5 gives arguments accounting for the difference between the agentive de and kara phrases. Section 2.6, concludes the discussion.

 

< 2.1 > Subjectless sentences in Japanese

2.1.1. De and ga alternation

Let us first look at the sentences in (1), repeated here as (3).

(3)a. ketueki sentaa de ketueki-gata siiru o o-hari-si-mas-u (Haraguchi)

blood center at blood type seal Hon-paste-do Pol-Pres

Ā@ ĎThe blood center will paste the seal of your blood type.í

b. mazu hotondo no baai uti no hoo de yat-te ori-mas-u (Imoto)

first most cases our part by do- Prog-Pol-Pres

ĎIn the first place, in most cases, we are doing (it) on our part.í

c. watasi-domo de uketamawaru ka, arui wa o-kyaku-sama no hoo de

we by take order or or Top Hon-customer Gen part by

go-yooi nasaru ka. (Imoto)

Hon-prepare Hon or

ĎWhether we take your order or you get it on your part...í

d. Taroo to Hanako de bokoo o otozure-ta (Hoshino)

and by alma mater Acc visit- Past

ĎTaro and Hanako visited their alma mater.í

The sentences in (3) undergo either the degradation in acceptability or change in semantic content if de is replaced by ga. (3a) with ga in place of de makes this sentence sound awkward, and low in acceptability in some context. In the case of (3b) this replacement causes the loss of the meaning “it is us not some one else that is doing the job.” De in this context means delimitation to the exclusion of other possibilities. This semantic content makes clear the contrast “ we or you” in (3c). If de is replaced by ga in (3d), the meaning “Taroo and Hanako alone excluding others” is lost.

Now the question arises as to the possibility of these sentences being the results of deletion of underlying subjects. However, this is not the case. Addition of “wareware” (we) to (3a) gives an entirely different meaning to the sentence.

(3)aí. wareware ga ketueki sentaa de ketueki-gata siiru o o-hari-si-mas-u

ĎWe will paste the seal of your blood type at the blood center.”

The de phrase is now simply an adjunct denoting the place of action, losing its original agentive reading. It has often been argued that NPs standing for organizations imply people working in them. However, it is hard to guarantee that “wareware” (we) stand for those belonging to the organaization. Some volunteers may come and do the job.

The situation is a little different for the remaining three sentences. Addition of the same NPs as those followed by de as the underlying subjects will make the sentence redundant, and deletion of the subjects seems obligatory, as is clear in (3bí) through (3dí).

(3)bí. uti no hoo ga uti no hoo de...”

(3)cí. watasi-domo ga watasi-domo de uketamawaru ka, arui wa

o-kyaku-sama ga o-kyaku-sama no hoo de ...

(3)dí. Taroo to Hanako ga Taroo to Hanako de ...

The relevant questions to ask about these sentences are: (a) whether it is appropriate to assume the existence of empty subjects in these sentences, and (b) why the subject NP is deleted instead of the de phrase. My answer to the first question is negative, with the result that (b) loses its significance. A thorough discussion is given in Section 2.2.

Let us tentatively conclude that the sentences in (3) are not the results of deletion of the underlying subjects. They are a type of bona fide subjectless sentences. The characteristics of the de phrase of this type are: (a) It always appears in the subject position, (b) It takes only an agentive plural4 human NP as its complement. Let us call this type of phrase “agentive de phrase”.

 

2.1.2. De phrases compatible with ga phrases

(4)a. kazoku de umi ni it-te, otto to futari de umi ni moguru

family by sea to going, husband with two by sea in dive-Pres

ĎMy family go to the sea, and we, my husband and I, dive into the sea.í

(Watanabe)

b. minna de atumat-te pan o tabe-ru ...

all by getting together bread Acc eat- Pres

ĎAll of us get together and eat bread...í (Nagamatuya)

c. Toohoo gekizyoo kara kanai to futari de kaeroo to sitan- desu

theater from wife with two by go home-be about to-Pol

ĎFrom the Toho Theater, we two, my wife and I, was about to go homeí

(Park)

d. kyoo wa Vivian ga tyuugoku no katei-ryoori o tukut-te

today Top Nom Chinese home dishes cooking

sannin de wake-at-te tabe-yoo to iu koto ni nat- ta

three by sharing eat-will that to decide-Past

ĎIt was decided that today Vivian cook chinese home dishes and we, three, share and eat them.í (Imoto)

The de phrases in the sentences in (4) delimit members of those denoted by the putative underlying subject NPs. These phrases carry the sense of Ďpartitiveí. The following sentences, (4aí) through (4dí), are the versions with the underlying subjects.

(4)aí. wareware ga kazoku de umi ni it-te, otto to futari de umi ni mogur-u

ĎWe, the whole family, go to the sea, and we, my husband and I, dive into the sea.í

bí. wareware ga minna de atumat-te ... ĎWe all get together...í

cí. watasi to kanai ga futari de ... ĎMy wife and I, two of us, ..í

dí. watasitati ga sannin de ... ĎWe, three of us, ...í

If the de phrases in (4) are replaced by ga, the sentences no longer carry the sense of Ďpartitiveí. The putative underlying subjects in (4aí) through (4dí), have the following characteristics: (a) they are always plural NPs, (b) they are easily recoverable, (c) their deletion is optional.

These characteristics prevent us from treating them as subjectless sentences.

De phrases of this type share the same characteristics as the ones in the first group (the sentences in (1)), in addition to its partitive sense. Let us call this type of phrase “partitive de phrase”.

 

2.1.3. Kara in place of ga

(5)a. Taroo kara kuzi o hii-ta (Yamada)

from lot Acc draw-Past

ĎTaro started to draw a lot.í

b. saisyo ni yuuzaa daihyoo no Ueno-sensei kara kansoo to

first users representative teacher from impression and

situmon-bun o

question-sentence Acc (Park)

ĎFirst, from Ms. Ueno, the representative of users, please (start giving) your impression and questions...í

c. watasi kara renraku o tor-anaku nat- ta (Koizumi)

I from correspondence Acc become-Past

ĎI ceased to make correspondence from myself.í

d. zikka kara kome o okut-te ki-ta (Furukawa)

home from rice Acc send-come-Past

ĎMy family sent me some rice.í

e. seifu kara zaidan ni enzyo o okut-ta

government from foundation to financial support Acc send-Past

ĎThe government sent financial support to the foundation.í

First, observe that the replacement of kara by ga causes loss of the “initiating” meaning in (5aí) through (5cí), but this is not the case with (5dí) and (5eí).

(5)aí. Taroo ga kuzi o hii-ta ĎTaro drew a lot.í

bí. saisyo ni yuuzaa daihyoo no Ueno-sensei ga kansoo to

situmon-bun o

ĎFirst Ms. Ueno, the representative of users, please (give) your impression and questions ...í

cí. watasi ga renraku o tor-anaku nat-ta

ĎI ceased to make correspondence.í

dí. jikka ga kome o okut-te ki-ta (the same meaning as (5d))

eí. seifu ga zaidan ni enzyo o okut-ta (the same meaning as (5e))

Addition of the same NPs followed by ga do not make (5a) through (5c) sound totally redundant, and their deletion is optional, just as the case of the sentences in (4). Now we can assume that the added ga phrases are underlying subjects, and that (5a) through (5c) are not subjectless sentences. Let us call this type “initiating kara phrases”.

The same argument does not obtain with (5d) and (5e). Replacement of kara by ga does not cause a semantic change, and addition of putative underlying subjects brings about redundancy, making their deletion obligatory. This type of kara is obligatory in passive sentences like (2b), repeated below.

(2)a. Ueno-san ga watasi ni Katoo-san o syookai-si-ta

Nom I to Acc introduce- Past

ĎMr. Ueno introduced Mr. Kato to me.í

b. Katoo-san ga Ueno-san kara/??ni watasi ni syookai-s-are-ta5

Nom from by to introduce-Pass-Past

ĎMr. Kato was introduced to me by Mr. Ueno.í

On the basis of these facts, I tentatively assume that (5d) and (5e) are subjectless sentences. Let us call these phrases Ďagentive kara phrasesí.

 

2.1.4. Sentences with delimiters like demo, made, sae (even); dake, sika (only)

(6)a. kodomo demo sit-te-i-ru (Hani)

children even know- Pres

ĎEven children know that.í

b. sekinin-sya made kaet-te simat-ta ( “ )

person in charge even go home Perf-Past

ĎEven the person in charge has gone home.í

c. sinyuu sae uragit-ta ( “ )

best friend even betray-Past

ĎEven his best friend betrayed him.í

d. kanai dake sanka si- mas-u ( “ )

wife only participate-Pol- Pres

ĎOnly my wife will participate.í

It is well known that the so called structural case particles ga and o are deleted when followed by delimiters. The sentences in (6) are examples of this type. In the EST it was natural to posit syntactic rules marking structural cases, which are eventually deleted if delimiters or particles are adjoined to them. This type of operation is not compatible with the Minimalist approach, which suggests the necessity of treating the delimiters as base generated. With this assumption, I treat sentences like those in (6) as subjectless. Now the question is how to license these NPs with delimiters, called Ďdelimiter phrasesí. A detailed discussion is given in Section 2.2.

The important difference between these delimiters and de is that the former can be attached to not only the subject but also the object and any other NPs with postpositions, while the latter can attach only to the agentive subject. Postpositions other than ga and o are retained when followed by delimiters, as in “Sendai kara sae (even from Sendai)”, “tomodati to demo (possibly with a friend)”, and “okaasan dake ni (only to Mother)”. Made and dake permit postpositions to appear either before or after them, as in “tomodati ni made/ tomodati made ni (even to a friend)” and “tomodati ni dake/ tomodati dake ni (only to a friend).”

 

2.1.5. Variety of topic markers

(7)a. anata nara kono kanozyo no kokoro no uti to kooi o

you if this her Gen heart Gen inside and behavior Acc

doo yomi-toku no daroo ka (Watanabe)

how decipher will Q

ĎI wonder how you will decipher her feeling and behavior.í

b. zinsei tte tuneni jibun o sagas-u tabi mitaina mono desyoo

life speaking of always self Acc seek-Pres journey like thing Presum

(Watanabe)

ĎSpeaking of life, it is like a journey seeking for oneís real self, isnít it?í

c. zyagaimo to ieba niku-zyaga da (Park)

potatoes speaking beef-potatoe Cop

ĎSpeaking of potatoes, the best is those cooked with beef.í

As is shown by the sentences in (7), Japanese has a variety of topic markers other than wa, mostly with the meaning “speaking of”. Lasnik and Saito (1992) state that there is a notable difference in English between a genuine topicalized sentence and one with a variety of expressions appearing in the sentence initial position, such as “speaking of”, “as for”, and so on. Let us call them topic markers. In English the phrases with these topic markers have correferential pronouns inside the sentence. This is a characteristic shared by sentences derived by the operation called Left Dislocation. One difference between a topicalized and a left dislocated phrase is, as is well known, the fact that the former obeys the subjacency condition, and the latter does not. In Japanese the distribution of left dislocated sentences without any topic marker is very much restricted, which requires a lengthy discussion beyond the scope of this paper. I assume tentatively that NPs followed by a variety of topic markers as well as wa occupy the Specifier position of Topic Phrase (Top P).

In the Minimalist approach, the computation using the feature [+Top] must project Top P with its Specifier position. An NP with a topic marker is attracted by [+Top] of the head of Top P in order to get its [+Top] feature checked and deleted. Any NP can be merged with a topic marker, and attracted by the [+Top] feature. But only the NP in the VP internal subject position with a topic marker forms a subjectless sentence. The sentences in (7) are subjectless in this sense.

 

< 2.2 > The agentive de and kara phrases

I claim that the agentive de and kara phrases constitute a type of subjectless sentences. The arguments are based on the assumptions (a) through (e) given at the beginning of this chapter.

It is shown in this section that the VP internal subjects with the postposition ni and kara carry the [Ā}D] feature6, while the other postpositions are [- D]. [Ā}D] expresses the optionality of checking this feature.

 

2.2.1. Distinction between the two types of de phrases

First we have to argue for our claim that agentive de phrases are to be distinguished from partitive de phrases.

Observe the contrast between sentences with agentive de phrases and those with partitive de phrases under causativization and passivization.

(8)a. seifu ga ketueki sentaa de / ni ketueki-gata siiru o har-ase-ta

ĎThe government had the blood center paste the blood seal.í

b. seifu ga wareware ni ketueki sentaa de ketueki siiru o har-ase-ta

ĎThe government had us paste the blood seal at the blood center.í

(agentive de phrase, causativization)

(9)a. ketueki-gata siiru ga ketueki sentaa de/ ?niyotte har-are-ta

ĎThe blood seal was pasted at the blood center.í

b. ketueki-gata siiru ga wareware ni yotte ketueki sentaa de har-are-ta

ĎThe blood seal was pasted by us at the blood center.í

(agentive de phrase, passivization)

(10)a. okaasan ga watasi-tati ni minna de pan o tabe-sase-ta

ĎMother let us all eat bread.í

b. okaasan ga minna de pan o tabe-sase-ta

ĎMother let all eat bread.í

(partitive de phrase, causativization)

(11)a. ? sono ryoori ga wareware ni yotte sannin de tabe-rare-ta

ĎThe food was eaten by us three.í

b. ?sono ryoori ga sannin de tabe-rare-ta

ĎThe food was eaten by three.í

(partitive de phrase, passivization)

As I pointed out in Section 2.1.1, the agentive de phrase does not appear as a nonargument; more specifically the de phrase as an adjunct has only the meaning “location”. In (8b) and (9b) it appears in the nonargument position, so that it is no longer the agentive de phrase. It has the meaning “location”, not “agent”.

(10) and (11), on the contrary, show that subject NPs can optionally appear together with the partitive de phrase. It is true that the sentences in (11) are unnatural, but it is due to the inanimate subjects in the passive sentences, which is not generally felicitous in Japanese. The (b) sentences with the animate subjects in (12), and (13), the passive counterparts of the (a) sentences, are acceptable.7

(12)a. (kodomo-tati ga) minna de kono inu o kawai-gat-te i-ta

children all by this dog Acc love- State-Past

ĎAll (the children) loved this dog.í

b. kono inu ga (kodomo-tati ni) minna de kawai-gar-are-te i-ta

ĎThis dog was loved by all (the children).í

(13)a. (yuusi ga) suu-nin de ryuugakusei o syootai-si-ta

volunteers several persons by foreign students Acc invite- Past

ĎSeveral (volunteers) invited foreign students.í

b. ? ryuugakusei ga (yuusi ni) suu-nin de syootai-s-are-ta

foreign students Nom volunteers by several by invite- Pass-Past

ĎForeign students were invited by several volunteers.í

Regardless of whether or not the underlying subjects are deleted, de phrases of this type in both the active and passive sentences retain their original partitive sense.

Thus, the causative and passive tests justify our claim that there is a clear distinction between agentive and partitive de phrases, and that sentences with partitive de phrases have underlying subjects.

Next, let us test this claim further subjecting the agentive de to nominalization test. Under nominalization, argument and nonargument NPs get no adjoined to them. Ga , o, and ni followed by no are deleted, whereas the other postpositions stay in their original positions.

(14)a. ketueki sentaa de no ketueki-gata siiru no tenpu

blood center at Nm blood type seal Nm pasting

(Nm = Nominal marker)

Ďpasting the seal of the blood type at the blood centerí

b. ketueki sentaa no ketueki-gata siiru no tenpu

Ďblood centerís putting the seal of the blood typeí

(15)a. *Taroo to Hanako de no bokoo hoomon

and by Nm alma mater visit

ĎTaro and Hanakoís visit to their alma materí

b. Taroo to Hanako no bokoo hoomon

ĎTaro and Hanakoís visit to their alma materí

In (14) the agentive reading is retained only in the (b) sentence without de, while the de phrase in (a) has the sense of Ďlocationí only. The contrast in (15) is also very instructive, that is, the de in (a) makes the sentence uninterpretable. These facts suggest that the agentive de does not survive under nominalization.

The partitive de phrase is, on the other hand, permitted to stay before no, as shown by (16).

(16)a. otto to futari de no sensui

husband and two by Nm diving

Ďdiving by two, husband and Ií

b. otto to futari no sensui

De in (16a) retains its original sense of “partitive”, as well as the sense of “exclusion”. (16b), without de, loses both these senses entirely. The nominal expressions in (17) show the same contrast.

(17)a. (wareware) san-nin de no syokuji

we three by Nm eating

Ďeating exactly by (us) three”

b. (wareware) san-nin no syokuji

Ďeating by (us) threeí

The three tests, causativization, passivization, and nominalization have supported my claim that the agentive and partitive de phrases are distinct. These tests have also shown that the agentive de can never appear in internal argument or nonargument positions. This is a crucial characteristic which distinguishes agentive de from agentive kara. This difference is accounted for in Section 2.4.

 

2.2.2. Two types of kara phrases

I have made a claim that the agentive and initiating kara phrases are distinct in their syntactic behaviors. The same tests as those in Section 2.2.1 are applied in the following.

(18)a. titioya ga zikka kara /ni kome o okur-ase- ta

father Nom family from /Dat rice Acc send-Caus-Past

ĎFather made his family send some rice.í

(agentive kara phrase, causativization)

b. kome ga zikka kara/??ni yotte/*ni okur-are- ta

rice Nom family from by by send-Pass-Past

ĎSome rice was sent by/from the family.í

(agentive kara phrase, passivization)

(19)a. sensei ga (karera ni) Taroo kara kuzi o hik- ase- ta

teacher Nom (them Dat) from lot Acc draw-Caus-Past

ĎThe teacher let them draw lots from Taro.í

(initiating kara phrase, causativization)

b. kuzi ga karera niyotte Taroo kara hik- are- ta

lot Nom them by from draw-Pass-Past

ĎThe lots were drawn by them starting with Taro.í

(initiating kara phrase, passivizataion)

As is shown by the causative sentences in (8a) and (18a), agentive kara as well as de can be replaced by ni, which is the postposition linked to the É∆ features [+ Agent] and [+ Experiencer]. In case de is retained in the causative and passive sentences, as in (8a) and (9a), the de phrase loses its original agentive reading, keeping only the sense of “location”. However, in the case of the agentive kara, there is no loss of its original agentive meaning. This suggests a difference between the agentive de and kara with respect to their syntactic function. This point is taken up in Section 5.

Addition of the putative underlying subject “zikka ni” makes the sentences in (18) very redundant.

(18)aí. *titioya ga zikka ni zikka kara kome o okur-ase-ta

Next, let us apply the nominalization test to (5e), with the agentive kara phrase, and (5a), with the initiating kara phrase.

(20) seifu kara no/ no zaidan e no enzyo no soofu

government from Nm8/ Nm foundation to Nm financial aid Nm delivery

Ďthe governmentís delivery of financial aid to the foundationí

Ďthe delivery of financial aid to the foundation from the governmentí

(21)a. Taroo kara no kuzibiki

from Nm lotting

Ďlotting from Taroí

b. Taroo no kuzibiki ĎTaroís lottingí

Nm

c. kodomo-tati no (ni yoru) Taroo kara no kuzibiki

children Gen by from Gen lotting

Ďthe childrenís lotting from Taroí

Ďlotting by the children from Taroí

In (20), with an agentive kara phrase, kara no and no alternate, retaining the sense of agentivity, while in (21), with the initiating kara phrase, the sense of “initiating” is lost from the (b) sentence containing only no. The addition of the subject to (21c) yields a perfect nominal expression.

Thus, the three tests amply justify my claim that agentive and initiating kara are two distinct types, and the latter appears in sentences with underlying subjects. Only agentive kara appears in subjectless sentences.

 

< 2.3 > Sentences with delimiters

Delimiters have been extensively described by Japanese grammarians in various descriptive frameworks, but they have rarely been cast into the framework of generative grammar.

Kuroda made a semantic analysis of delimiters as early as 1970 in relation to the subject-predicate relation viewed in logico-semantic terms. In this work, Kuroda argues for the logical subjecthood of the topic phrase. Topic phrases followed by delimiters also function as logical subjects in this sense.

Kurodaís findings have relevance to this work in the following respects: (a) In logico-semantic terms, the topic-comment relation functions as the expression of “categorial judgment” called a predication. . (b) The grammatical subject-predicate relation serves another function of expressing “thetic (non-predicational) judgment”9, called description. (c) In a sentence with two delimiters, the wider scope resides with the subject phrase with a delimiter, if the sentence keeps the basic word order. This means that there is no scope ambiguity unless scrambling changes the basic word order.

Another work relevant to this study is König (1997), which discusses a typology of intensifiers focusing on emphatic reflexives.

König classifies uses of emphatic reflexives into adnominal and adverbial uses.

(22)a. I have myself swept this court. (adnominal)

b. I have swept this court myself. (adverbial)

The emphatic reflexive, an intensifier or a focus marker according to König, “can be paraphrased by Ďalsoí or Ďtooí in (22a), whereas Ďaloneí, Ďon my owní or Ďwithout helpí seem to be possible paraphrases in the case of” (22b). The former meaning is termed as “inclusive”, and the latter “exclusive”. He further notes that “exclusive intensifiers may only relate to agentive subjects.”

Going back to our findings about the agentive de, Königís characterization of intensifiers perfectly fits this type of de, because it can be adjoined only to agentive subjects and has the meaning of exclusiveness. Starting with emphatic reflexives, König goes on investigating typological variations of intensifiers of all kinds found in quite a number of languages. On the basis of these facts I propose that agentive de be treated as a delimiter, and as one of the intensifiers. Konig further notes that “inclusive intensifiers may also relate to non-subject arguments.” This is also one of the characteristics of Japanese delimiters other than agentive de. They can be adjoined to all arguments and nonarguments.

It is reasonable to treat intensifiers, including delimiters, as focus markers, as König insists. Within my framework, NPs with delimiters carry the feature [+Focus], which is attracted by the feature [+Focus] of the head of the focus phrase (FP), the projection of this head. In case more than one delimiter phrase is chosen, the [+ Focus] feature of the highest one in the thematic hierarchy is attracted and raised to the [Spec, Focus] position.

Agentive de, which merges only with agentive external argument, has its [+Focus] feature attracted to FP. In this way de phrase is licensed as a focus marker, and the preceding NP, its focus, plays the role of the logical subject in Kurodaís sense. Thus sentences containing agentive de phrases fulfill the function of predication.

The same obtains with external arguments marked with delimiters other than de. They occupy the FP-Spec position playing the role of the logical subject and taking a wider scope than the other phrases with delimiters, if there are any. Thus, all the other NPs with delimiters stay in the VP internal positions.

Delimiter phrases cooccur with topic phrases (Top P), as in the sentences in (23) given below.

(23)a. kono sigoto wa Taroo to Hanako de katazuke-mas-u

this task Top and finish- Pol- Pres

ĎThis task, Taro and Hanako will finish.í

Ā@Ā@ b. ano ziken wa kodomo demo sit-te i-ru

that incident Top children even know- Pres

ĎThat incident, even children know.í

c. koozi genba kara wa sekinin-sya made kaet-te simat- ta

construction site from Top person in charge even leave Compl-Past

ĎFrom the construction site, even the person in charge has left.í

d. asu no kaigi ni wa kanai dake sanka- si- mas-u

tomorrow Gen meeting in Top wife only participat-do Pol-Pres

ĎTomorrowís meeting, only my wife will participate in.í Ā@

The relative order of Top P followed by FP is generally assumed, which is accounted for by generating the former in the Spec-CP position and the latter adjoined to IP. However, this is not a conclusive solution. Lasnik and Saito (1992), for example, generate Top P by adjunction to IP. There has been another proposal that two types of Top P be distinguished with one base generated in the CP-Spec position and the other adjoined to IP. Since the purpose of this paper is to argue that the Top P and the FP are responsible for subjectless sentences in Japanese, the question of the relative order of these two types of phrases is set aside in this work.

 

< 2.4 > Topic phrases

As was explained in Section 2.1.5, there are a number of topic markers serving similar functions as wa. Diverse in morphological makeup though they are, their functions as topic markers are uniform, and they rarely leave behind correferencial pronouns, as opposed to the corresponding English expressions like “speaking of”,”as for”, and so on. In view of these facts I will treat them together with the standard topic marker wa.

Another characteristic they share is that they attach to any NP, as in (24) - (26).

(24)a. kono gakusei wa moo ronbun o dasi- masi-ta

this student Top already paper Acc submit- Pol-Past

ĎThis student has already submitted his paperí

b. kono gakusei nara moo ronbun o dasi- masi-ta

this student speaking of already paper Acc submit-Pol-Past

ĎSpeaking of this student, (he) has already submitted his paper.í

(25)a. watasi no kimono wa okaasan ga itumo erabi- mas-u

I Gen Top Mother Nom always select-Pol-Pres

ĎMy kimonos, Mother always select.

b. watasi no kimono tte okaasan ga itumo erab-u no desu

as for

ĎAs for my komonos, Mother always select them.í

(26)a. Tokyoo de wa ii sigoto ga mitukari-mas-u yo

in Top good job Nom be found-Pol-Pres

ĎIn Tokyo good jobs can be found.í

b. Tokyoo tte ii sigoto ga mitukari-mas-u yo

speaking of

ĎSpeaking of Tokyo, good jobs can be found (there).í

There are a few differences between wa phrases and those marked by the other topic markers. First, a sentence may contain multiple wa phrases, whereas a sentence permits only one occurrence of the other topic markers.

(27)a. kodomo-tati wa hon wa10 kai-masi-ta

children Top book Top buy-Pol-Past

ĎAs for the chilren, the books they bought.í

b. ?? kodomo-tati tte hon nara kai- masi-ta

speaking of book as for buy-Pol- Past

??ĎSpeaking of the children, as for the books, they bought.í

Second, both types of topic markers can cooccur in a sentence, as shown by the sentences in (28).

(28)a. zissyo tte kodomo-tati wa kat-te i- masen yo

dictionary speaking of children Top buy-State-Pol- Neg SF Par

ĎSpeaking of a dictionary, the children havenít bought (one).í

b. kodomo-tati wa omotya nara kai-mas-u yo

children Top toys as for buy-Pol- Pres SF Par

??ĎChildren, as for toys, (they) will buy (them).í

However, there are restrictions for their cooccurrence, and it is preferred to place topic phrases with topic markers other than wa before wa phrases. The situation is similar to the distributional difference between the English left dislocated phrases and topics. The former is restricted to the matrix sentence, whereas the latter can appear in both the matrix and embedded sentence. To account for this fact, Lasnik and Saito (1992) generate the former in the CP-Spec position, and the latter by adjunction to IP. They restrict each type to only one occurrence. Though further characterization of topic phrases in general is needed, I cannot go into the details of these matters here, since it will take me too far afield. Here, I will treat them all as topic phrases.

Topic phrases get their [+Top] feature checked by the same feature borne by the head of TP. Thus, they are licensed as the logical subjects in Kurodaís sense, and sentences containing them fulfill their function of predication.

To summarize the discussion so far, agentive de and kara phrases are responsible for subjectless sentences. Of these two, de can be treated as a delimiter, playing the function as one of the focus markers. De, a focus marker for an agentive subject, is responsible for a subjectless sentence as well as the other delimiters following agentive subjects. Topic markers accompanying subjects also mark subjectless sentences.

The only remaining problem is the difference between the agentive de and kara phrases, noted in Section 2.2.2 , under the tests of causativization, passivization, and nominalization; namely, the de phrase loses its agentive meaning in causative and passive sentences as well as nominalized versions, as shown by the sentences in (8), (9), (14), and (15), whereas the kara phrase retains its original agentive meaning under the same tests, as shown by (18) and (20). This question is taken up in Section 2.5, as one piece of crucial evidence supporting my assumption of the É∆ checking as the means of NP licensing, and VP internal argument and nonargument NPs being marked with postpositions (the assumptions in (a) - (e) at the beginning of this chapter).

 

< 2.5 > Conclusion

First I will argue that the difference between the agentive de and kara phrases finds a natural account under my assumptions (a) - (e), and proceed to argue that the findings of this work support my claim that argument and nonargument NPs are licensed by É∆ checking, and VP internal NPs are all marked by postpostions. Finally, it is proposed that there are two mechanisms for guaranteeing the sentence-hood, subject-predicate relation and topic/focus-comment relation, and languages choose between these two or adopt both of them. Japanese is viewed as belonging to the latter group. To support the basic sentence building mechanisms, checking of the features [+D],[+Topic], and [+Focus] is employed. For NP licensing only É∆checking is obligatory. Thus, sentence with VP internal subjects, such as the sentences in (1) are pemitted.

Let us limit our discussion to that of the external argument of a predicate dominated by VP, the so called VP internal subject. The postposition ni linked to the É∆feature [+ Agent] merges with NP with the feature [+ Animate]. A predicate carrying the argument structure with Agent as its external argument projects VP with its Specifier position (VP-Spec position). The NP followed by ni and with [+ Agent] fills this position. The agentive ni phrase with a strong D feature moves to the TP-Spec position attracted by the strong D feature of T, the head of TP, unless this position is already filled by some other NP.

In this framework, (29a) has the VP internal form (29b) after É∆ checking.

(29)a. Taroo ga hon o yon- da

Nom book Acc read-Past

ĎTaro read a book.í

b. Taroo ni hon o yon-da (Ni is the postposition linked to [+ Agent].)

Since the TP-Spec position is not filled in a simple sentence, the ni phrase moves to this position, getting the [-Acc] feature from T. This feature triggers attachment of the default case particle ga to the ni phrase in morphology, eventually deleting ni.

In a complex sentence, the embedded ni stays on, as in the case of the causative sentence (30), with (29a) as its complement.

(30) sensei ga Taroo ni hon o yom-ase- ta

teacher Nom book Acc read-Caus-Past

ĎThe teacher made Taro read a book.í

The situation is the same with the indirect passive (31)11.

(31) watasi wa Taroo ni hon o yom-are- ta

I Top book Acc read-Pass- Past

ĎI had Taro read a book.í

The experiencer ni phrase is assumed either to stay under VP or to move to the TP-Spec position. In other words, it has either plus or minus D feature. This assumption is supported by sentences with psychological verbs and the potential affix re, in which ni and ga alternate.

(32)a. Hanako ni/ga ane ga urayam-asi-i koto12

Nom sister Nom envious- Pres that

ĎHanako is envious of her sister.í

b. Katoo-san ni/ga komakai zi ga yom-e- ru koto

small letter Nom read-Pot-Pres that

ĎMr. Kato can read small letters.í

Agentive kara is the same. First it can alternate with ni without loss of its original agentive meaning, as is shown by (18a), repeated below as (33).

(33) titioya ga zikka kara /ni kome o okur-ase- ta

father Nom family from /Dat rice Acc send-Caus-Past

ĎFather made his family send some rice.í

Second, kara remains in passive sentences, as in (18b), repeated below as (34). This means that the agentive kara phrase stays in the VP internal subject position.

(34) kome ga zikka kara/??ni yotte/*ni okur-are- ta

rice Nom family from by by send-Pass-Past

ĎSome rice was sent by/from the family.í

The same applies to (2b) repeated below.

(35) Katoo-san ga Ueno-san kara/??ni watasi ni syookai-s-are-ta

Nom from by to introduce-Pass-Past

ĎMr. Kato was introduced to me by Mr. Ueno.í

(34) and (35) show that kara has to stay in the VP internal position, when the TP-Spec position is filled. With the unfilled TP-Spec position kara optionally stays in the VP internal subject position. This is the case of agentive kara. More specifically, ditransitive verbs with the É∆features [+ Source] and [+ Goal] permit agentive kara to remain as the VP internal subject, even when e or ni (Goal) is not selected. Thus, agentive kara has plus and minus D feature13. With the plus value, it moves to the TP-Spec position and eventually receives the default case particle ga. Kara in a regular transitive sentence without the É∆feature [+ Goal] has to move to the TP-Spec position, if it is unfilled. The sentences in (36) show this restriction on sentences without NPs with the feature [+ Goal].

(36)a. Katoo-san *kara/ga Hanako o aisi-te i-ru

from/Nom Acc love-Prog-Pres

ĎMr. Kato loves Hanako.í

b. Hanako ga Katoo-san kara/ni ais- are- te i- ru

from/by love-Pass-Prog-Pres

ĎHanako is loved by Mr. Kato.í

(36b) shows that in the embedded sentence kara as well as ni stays in the VP-Spec position, which shows that aisu-ru has the option of choosing between the É∆features [+ Source] and [+ Agent] for checking agentive kara and ni phrases.

The difference between the agentive de and kara phrases finds a natural account under my assumption; namely, kara phrases stays in the VP-Spec position, where their É∆ feature [+ Agent] is checked by the predicate, whereas agentive de phrases have to move after the É∆ checking to the FP-Spec position to have their feature [+ Focus] checked by the same feature borne by the head of FP. De phrases other than those with agentive de stay on in the VP internal positions. This is why only agentive de keeps the feature [+ Agent] in the FP-Spec position.

Unless agentive de and kara phrases get their É∆ feature [+ Agent] checked by the predicate in the VP internal position, there is no way of licensing them as arguments. Thus my assumption is supported by the facts about subjectless sentences with agentive de and kara phrases. The analysis of agentive de as a delimiter involves assumptions about Ďfocusí, which in turn induces assumptions about Ďtopicí. However, this work followed general assumptions regarding the syntactic statuses of Ďfocusí and Ďtopicí without much discussion. Needless to say, these are topics that require further thorough studies.

 

Notes to Chapter 2:

* Comments by T Ito, and Y Ueda helped revision of the preliminary version of this chapter.

** The abbreatory symbols used in this chapter: Acc = accusative, Caus = causative, Compl = completive aspect, Cop = copula, Dat = dative, Gen = genetive, Hon = honorific, Nom = nominative, Nm = nominal marker, Neg = negative, Pass = passive, Past = past tense, Perf = perfective aspect, Pres = present tense, Presum = presumptive, Prog = pregressive aspect, SF Par = sentence final particle, State = stative, Top = topic

1 The names of the graduate students who contributed the data are given in these parentheses.

2 The feature [-Acc] is carried by predicates without the ability of checking the accusative case. Stative verbs and adjectives as well as T, the head of TP carry this feature.

3 For details, see Section 2. 5 in particular.

4 The use of plural NPs, especially first person NPs, as subjects is a kind of “hedge”, softening the tone of the discourse. De also meets the requirement of softening the tone of self insistence carried by the use of ga.

5 The consecutive use of ni phrases are avoided by using the original postposition kara. Unless we assume the existence of VP internal postpositon kara, we cannot account for this fact.

6 To be more accurate the PP with the postposition ni linked to [+ Agent] has the [+D] feature, which is attracted by the same feature carried by T in a simplex sentence. In a sentence with a complex predicate with its embedded subject, the subject retains the original postposition ni, since the matrix position of TP-Spec is filled with the matrix subject. The PP with ni linked to [+ Experiencer] carries the [ Ā}D] feature, permitting free variation between ni and ga in a simple sentence. With a complex predicate the same situation holds with both Agent and Experiencer ni phrases. (See, for example, the sentences (9) and (10) in Chapter 1.) Within this framework, it is regarded that in a sentence with the embedded subject the choice of [+ D] leads to derivation crash because this uninterpretable feature remains at LF.

7 The sentences in (12) and (13) without the deletion of the subjects sound more natural.

8 The symbol Nm (nominal marker) is used to distinguish this particle from the homophonous genitive marker.

9 According to Kuroda the logical subject-predicate structure expresses “categorial judgement“ (in Japanese “dangen-teki handan”). The logical subject-predicate relation is the topic/focus-comment relation in this paper. The grammatical subject-predicate relation expresses what is called “thetic judgment” by Kuroda, which is rendered as “kotei-teki handan” in Japanese. The term “sentence-hood” is used in this paper as a semantic term standing for the predicational and descriptive statements in the sense given by Kuroda, which is different from a syntactic definition.

10 It is well known that only the first wa phrase carries the sense of topic, the rest of wa phrases bearing the sense of contrast. I set this problem aside, treating them all as topic phrases.

11 The direct passive has the underlying adjunct marked with ni linked to [+Agent].

12 “Koto” is added to make these sentences sound natural.

13 There are two classes of verbs with the É∆ feature [+ Source], one with the optional [+ Goal] and the other without this feature. Ditransitive verbs, sometimes called “giving verbs” like okur-u (send), atae-ru (give), age-ru (give), takus-u (entrust) syookai-su-ru (introduce) and so on, belong to the first class. Mental or emotional activity verbs such as ais-u (love), sonkei-su-ru (respect), sinyoo-su-ru (trust) are in the second class. Only the former permits the agentive kara to stay in VP in a simplex sentence.

14 It has generally been assumed in the EST that in the subject position double theta role assignment of Source and Agent or Goal and Agent is necessary, as in (i) and (ii) below.

(i) Katoo-san ga/kara Hanako ni hana o age- ta

Nom from to flower Acc give-Past

ĎMr. Kato gave some flowers to Hanako.í

(ii) Hanako ga Katoo-san kara hana o morat-ta

Nom from flowers Acc get- Past

ĎHanako got some flowers from Mr. Kato.í

In (i), both ga and kara have the double theta roles Agent and Source, while in (ii) ga has Agent and Goal. It may be appropriate to posit a single theta role [Agent-Source] and [Agent-Goal]. I leave this problem open.

 

 

< 3 > Chapter 3 On the Japanese particle o *

The Japanese particle o has been treated as the accusative case particle and also as a postposition standing for an inherent case with semantic features linked to the É∆ features [+ Path], [+ Passing Point], and so on. The status of o as a structural case has recently been called into question in relation to Case checking in the framework of the “Principles and Parameters” approach. Kuroda (1988) regards Case checking as dependent on a parameter, distinguishing languages with forced Case checking from those without it. Marantz (1992) discusses this mechanism as a means of NP licensing, pointing out the weakness of Case checking as a case realization mechanism and the fact that the NP licensing function can be achieved by Extended Projection Principle. Fukui and Takano (1996-unpublished) claim that Case checking involving Verb Raising is counter to their plausible hypothesis concerning the Japanese surface linear order, S O V, for which Verb Raising should not be applied in the process of its derivation. Discarding Case checking altogether, they propose that the Japanese particles ga and o license their complement NPs. Their claim led them to regard both the structural and inherent o as an inherent case marker, a postposition.

Chapter 1 presented my claim that all the VP internal NPs are marked with postpositions whose semantic features linked to É∆features are checked by predicates (É∆checking) ; ni 1 and ni2 phrases are checked by the É∆features [+ Agent] and [+ Experiencer] respectively, and o1, o2, and o3 phrases are checked by [+ Theme], [+ Patient], [+Path], and so on. In Chapter 2 additional supporting evidence for this claim was given, using as data sentences without ga phrases.

This chapter starts with a descriptive generalization of the functions of the accusative case marked by the particle o. Section 3.1 gives a survey of the distribution of the so called structural case particle o , in comparison with the inherent case particle (or postposition) o. Section 3.2 gives a tentative list of É∆ roles involved in the checking of o , together with the argument structure borne by each subclass of transitive verbs. Various tests traditionally used to examine the status of o as a structural case are given in Section 3.3. The results show that there is no fast and hard line between the structural and inherent case particle o. This point is further investigated using data showing the interchangeability of o and some other postpositions in Section 3.4.

Section 3.5 gives evidence supporting my claim that the accusative Case checking is not involved in the Japanese case system, introducing the work by Ueda (1997) on scopal interaction in Japanese.

The findings of this study suggest that accusative, and possibly dative, Case checking, employed by the GB theory, must at least be called into question. In sum, Case checking does not account for the distribution and functions of o.

 

< 3.1 > Distribution of the so called structural case particle o

It is generally assumed that structural cases stand for various semantic roles, while inherent cases preserve one to one correspondence to semantic roles. To see whether this assumption can be supported, I will start characterizing both the structural and inherent particles o.

 

3.1.1. Characterization of the structural case particle o

The sentences in (1) are typical transitive sentences with the so called structural case particle o , which is hereafter given the traditional name Accusative case (Acc).

(1)a. Akira ga futa o tot- ta ĎAkira took off the lid.í

Nom** lid Acc take off-Past (Agentive Transitive -AT)

b. Akira ga unten-menkyo-syoo o nakusi-ta ĎAkira lost his driving license.í

Nom driverís license Acc lose- Past (Non-agentive Transitive-NAT)

c. kado no yasin ga kare no zyumyoo o tizime- ta

excessive ambition Nom his life Acc shorten-Past

ĎExcessive ambition shortened his life.í (Causal Transitive-CT)

d. tunami ga kaihin no buraku o osot-ta

tidal waves Nom beachside Gen village Acc hit-Past

ĎTidal waves hit the village by the beach.í (Natural Force Transitive-NFT)

e. siroi nuno ga tukue o oot- te i- ta

white cloth Nom desk Acc cover-State-Past

ĎA piece of white cloth covered the table.í (Instrumental Transitive-IT)

f. kodomo-tati ga to o tatai-ta ĎThe children pounded on the door.í

children Nom door Acc pound-Past (Non-Causal Transitive-NCT)

g. watasi wa 30-pun mo basu o mat- ta

I Top minutes as long as bus Acc wait-Past

ĎI waited for the bus as long as 30 minutes.í (Neutral Transitive-NT)

h. titi ga honbako o tukut- ta ĎFather made a book case.í

Father Nom bookcase Acc make- Past (Creation Transitive-CRT)

i. wakamono-tati wa nakama no seikoo o yorokon-de i-ru

young people Top friends Gen success Acc be pleased- Pres

ĎThe young people are pleased with their friendís success.í

(Transitive Psychological Verb-TPV)

j. kyoo no teian wa ikutuka no mondai o hukun-de i-ru

today Gen proposal Top several problem Acc involve State-Pres

ĎTodayís proposal involves several problems.í (Relational Transitive-RT)2

For the characterization of the subclasses of transitive verbs, I used the following four features in Inoue (1994): (a) intentional action (b) change on the part of the patient (c) direct action (d) stativity. The result is shown by (2)

 

(2) A B C D E F G H I J

(AT) (NAT) (CT) (NFT) (IT) (NCT) (NT) (CRT)(TPV)(RT)

(a) Intentional Action + Ā@ Ā| Ā| Ā| Ā| + + + Ā| Ā| Ā@

(b) Change of Patient + + + Ā| Ā| Ā| Ā| Ā| Ā| Ā|

(c) Direct Action + Ā| Ā| + + + Ā| + Ā| Ā|

(d) Stativity1 Ā| Ā| Ā| Ā| Ā| Ā| Ā| Ā| Ā| +

The chart (2) shows the transitivity scale, with Agentive T(ransitive) (AT) at the top of the scale, and Relational T (RT) at the bottom. Note that Non-Agentive T (NAT) and Causal T (CT) share the same features with the same plus or minus value. They show an identical syntactic behavior under passivization; namely, they do not allow the agentive ni phrase.

(3)a. *unten-menkyo-shoo ga Akira niyotte nakus-are-ta

driverís license Nom by lose- Pass-Past

??íThe driverís license was lost by Akira.í (Passive of (1b))

b. *kare no zyumyoo ga kado no yasin niyotte tizime-rare-ta

his life Nom excessive ambition by shorten-Pass-Past

?? ĎHis life was shortened by excessive ambition.í (Passive of (1c))

If niyotte (by) is replaced by de (due to) in (3b), the sentence becomes grammatical. This fact suggests that they may be classified together as Causal T (CT), which does not allow the agentive ni or niyotte phrase.

Next, Natural Force T (NFT) and Instrumental T (IT) share the same distiribution of the plus and minus value. It seems reasonable to class them together as Instrumental T (IT). The passive test given in Section 3.3 will show their identical syntactic behavior, revealing that the agentive ni (by) phrase appears in the passive sentences derived from both these types of transitive sentences regardless of whether the passive subject is an inanimate or animate NP.

(4)a. kaihin no buraku ga tunami ni osow-are-ta

beachside village Nom tidal wave by hit- pass-Past

ĎThe village by the beach was hit by tidal waves.í (Passive of (1d))

b. tukue ga siroi nuno ni oow- are- te i- ta

desk Nom white cloth by cover-Pass-State-Past

ĎThe desk was covered by a piece of white cloth.í (Passive of (1e))

Thus far, the five classes (from A to E) have been reclassified into three, Agentive T (AT), Causal T (CT), and Instrumental T (IT).

Going down the scale we find that Non-causal T (NCT) and Creation T (CRT) share the same feature specifications. However, these two classes cannot be merged, because NCT does not cause Change of Patient, while CRT requires a created object for its accusative complement, not the Patient undergoing change. The remaining three classes (NT, TPV, and RT) are all distinct. Transitive Psychological Verb (TPV) and Relational T (RT) are unique in that they lack the three main Transitivity features, (a) through (c). The plus value of the stativity of RT reflects its least “transitive” characteristic.

The distribution of the so called structural case particle o shown in the transitivity characterization of (2) indicates that the structural o phrases stand for various semantic roles, as the general assumption goes.

 

3.1.2. The so called inherent case particle o

Inherent case particles are generally assumed to have one to one correspondence to É∆ roles. However, the situation is not so simple.

Sugimoto (1986) classifies this type of o phrases into two subclasses, one is called “Movement Complement”, and the other “Circumstantial Complement”. The former can be subclassified into the following subgroups.

(5)a. Katoo-san ga hodoo o arui- te i- ru

Nom sidewalk P walk-Prog-Pres (P = postposition)

ĎMr. Kato is walking along the sidewalk.í (Path)

b. basu ga koosaten o sugi-ta ĎThe bus passed the crossing.í

bus Nom crossing P pass-Past (Passing Point)

c. kodomo-tati ga koomon o de- ta

children Nom school gate P come out-Past (Source)

ĎThe children came out of the school gate.í

The o phrase with the meaning “Path” contrasts sharply with the de phrase meaning “Location”, as shown by the sentences in (6).

(6)a. gakusei-tati ga guraundo o hasit-te i- ru

students Nom playground Acc run- Prog-Pres

ĎThe students are running in the playground.í

b. gakusei-tati ga guraundo de hasit-te i-ru

ĎThe students are running in the playground.í

Even though the English rendition does not reveal the semantic difference, the o version implies that the running is taking place in a large portion of the playground, while the latter has no such implication, simply specifying the place of the action.3 (5b), on the other hand, means just the point which the moving object passes. Verbs like oi-kos-u (pass), koe-ru, kos-u (pass over) watar-u, (cross) and so on take the o phrase with the meaning “Passing Point”.

The o phrase in (5c) alternates with the kara phrase, as in (7).

(7) kodomo-tati ga koomon o/kara de-ta (the same meaning as (5c)4.

Sugimotoís second class “Circumstantial Complement” has unique characteristics. The sentences in (8) are some examples.

(8)a. 109-bin wa arasi no naka o kyookoo tyakuriku si-ta (Sugimoto, 1986 (1)) flight number Top storm midst P reckless landing do-Past

ĎFlight number 109 landed recklessly in the midst of the storm.í

b. sono paatii no ikkoo wa hubuki no naka o

that party of people Top snow storm of midst P

one o samayot-ta

mountain ridge Acc wander- Past (Ibid. (2)

ĎThat party wandered along the mountain ridge in the midst of a snow storm.í

c. dosya-buri no ame no naka o guraundo o hasit-ta (Ibid, (39b))

cats and dogs rain of midst P playground P run- Past

ĎWhile it was raining cats and dogs, (we) ran in the playground.í5

The sentences (8b, c) have two o phrases, with the first one Circumstantial and the second one Path. Sugimoto argues that this cooccurring possibility indicates the distinctness of the two types of inherent o phrases, since the repetition of the same type of particles in a simplex sentence is disallowed in Japanese. Detailed argument is given in Section 3. 3.

Thus, the facts revealed in Secton 3.1.1. and 3.1.2 refute the claim that structural cases stand for farious semantic roles but the inherent cases correspond to a single semantic role. The fact is that both structural and inherent cases stand for various semantic roles. Now the question is how to represent the argument structures reflecting these distributional facts of the so called structural case paraticle o.

 

< 3.2 > Argument structures of subclasses of transitive verbs

In this section I will specify the argument structure for each subclass of transitive verbs given in Section 2.1, using the traditional labels of É∆roles given below. 6 The external arguments are underlined.7

(9) É∆ Roles:

Agent: an intentional actor

Experiencer: one who experiences psychological state or feels influence from outside

Patient: an animate or inanimate entity that gets a direct action inflicted on itself and undergoes a change

Theme: an animate or inanimate entity that gets a direct action inflicted on itself but does not undergo a change, or an entity in a certain state

Instrument: an instrument used for an action

Source: an originating point of an action or change of state

Target: a target of psychological and physical action

Goal: an endpoint of an action, or a result of an action

Cause: a cause of an event or a psychological state

(10) Argument structures of transitive verbs

a. Agentive Transitive (AT): [Agent, Patient]

b. Non-agentive Transitive (NAT) and Causal Transitive (CT):

[Cause, Patient]

c. Natural Force Transitive (NFT) and Instrumental Transitive (IT):

[Instrument, Patient]

d. Non-causal Transitive (NCT): [Agent, Theme]

e. Neutral Transitive (NT): [Agent, Target]

f. Creation Transitive: [Agent, Goal]

g. Transitive Psychological Verb (TPV): [Experiencer, Target]

h. Relational Transitive: (RT) [Theme, Theme]

Transitive verbs are characterized by the argument structures they carry.

Now we face a problem involved in the characterization of the “movement complement” described in Section 3.1.2. It is obligatorily selected by the movement verbs like watar-u (cross), koe-ru, kos-u, oikos-u (pass over), and so on, but is optional for verbs like hasir-u (run), aruk-u (walk), kake-ru (run), and such. Movement complements, on the other hand, always select motion verbs. This fact makes it plausible to assign specifictions similar to argument structures in order to make clear the distictions among these verb groups.

(11) a. Verbs like aruk-u and hasir-u: [Agent, Path]

b. Verbs like watar-u, kos-u: [Agent, Passing Point]

c. Verbs like de-ru, hanare-ru: [Agent, Source]

 

< 3.3 > Structural tests

Sugimoto (Ibid.) uses the following four structural tests while trying to justify the distinction between the structural case particle o and the Movement Complement o.

(12) (a) possibility of doubling the particle o

(b) quantifier floating

(c) alternation between o and ga in a transitive stative sentence

(d) passivization

I will add to the list the test of unaccusatiave alternation as (12e).

(12) (e) unaccusative alternation.

The test (a) seems to be valid, but the remaining four reveal the murkiness of distinction instead of justifying it.

Sections 3.3.1 through 3.3.4 give an outline of these tests with comments. In Section 3.3.4, the transitive sentences with the so called structural o, (10a) - (10h) in Section 3.1, are subjected to the passivization test. The result is that the syntactic behaviors of these eight subclasses are not at all uniform, but vary depending on their argument structures. In the same section, inherent o phrases are subjected to the passivization test.

 

3.3.1. Possibility of doubling o

There is in Japanese a well known constraint on doubling o, called “Double o Constraint.”8 The “O-causative” sentence with the transitive complement sentence has the surface ni-o sequence rather than the expected o-o sequence, as shown by (13).

(13) a. Satoo-kun ga zisyo o kat- ta ĎSato bought a dictionary.í

Nom dictionary Acc-buy Past (complement transitive sentence)

b. *sensei ga Satoo-kun o zisyo o kaw-ase-ta

teacher Nom Acc dictionary Acc buy-Caus-Past

ĎThe teacher made Sato buy a dictionary.

c. sensei ga Satoo-kun ni zisyo o kaw-ase-ta

Dat Acc

ĎThe teacher made Sato buy a dictionary.í/ ĎThe teacher let Sato buy a dictionary.í

The consecutive use of the structural o and the “movement complement” (hereafter abbreviated as MC) o seems to be under the same constraint.

(14) a. kodomo-tati ga kono hasi o watat-ta ĎThe children crossed this bridge.í children Nom this bridge MC cross- Past

b. *watasi wa kodomo-tati o kono hasi o watar-ase- ta

I Top children Acc this bridge MC cross-Caus-Past

ĎI made the children cross this bridge.í

However, this unacceptability can be mitigated by adding intervening phrases, as in (15a), in sharp contrast with (15b), which has an intervening phrase in the sequence of the structural o. (See Sugimoto (Ibid.) for more examples.)

(15) a. ?watasi wa kodomo-tati o yukkuri kono hasi o watar-ase-ta

Acc slowly MC

ĎI made the children cross the bridge slowly.í

b. *sensei wa Satoo-kun o yatto zisyo o kaw-ase-ta

Acc with great effort Acc

ĎThe teacher made Sato buy a dictionary with great effort.í

Even though a slight unnaturalness remains in (15a), it is far better than (15b). This fact supports the claim that the multiple use of structural case is subject to the double o constraint, while the combination of the structural and inherent o is permitted if they do not appear consecutively. This is the strongest argument Sugimoto gives to support the distinction between the structural and inherent cases.

 

3.3.2. Quantifier Floating as a structural test

It is well known that the possibility of quantifier floating is not a valid test to see the argument-hood. Sugimotoís data amply show that the MC o phrase as well as the structural o phrase permits quantifier floating , as shown by (16) given below.

(16)a. gakkoo ga suunin no onna no ko o daihyoo ni eran-da

school Nom several young girl Acc representative as select-Past

ĎThe school selected several young girls as representatives.í

b. gakkoo ga onna no ko o suuninn daihyoo ni eran-da

young girl Acc several

c. sono mura ni ik-u ni wa, mittu no hasi o watar- anakereba narana-i

that village to go in order to three bridge MC cross have to Pres

ĎTo go to that village we have to cross three bridges.í

d. sono mura ni iku ni wa, hasi o mittu watar-anakereba narana-i

bridge MC three

In (16b) the quantifier suunin (several) has floated out of the structural o phrase, just as mittu (three) in (16d) has come out of the MC o phrase. This means that the facts about quantifier floating does not support the claim for the distinction between structural and inherent cases.

 

3.3.3. O and ga alternation in a transitive stative sentence

In a transitive stative sentence, the object is marked with ga rather than o.

(17) Katoo-san ga rosiago ga/ *o wakar- u ĎMr. Kato understand Russian.í

Nom Russian Nom/*Acc understand-Pres

In a sentence with a complex predicate, the o-ga alternation is permitted, as in (18).

(18) watasi wa gohan ga / o tabe-tak-i ĎI want to eat rice.í

I Top rice Nom/Acc eat-want-Pres

This has been taken as a structural case alternation. However, this is not exactly the case. The MC o permits this alternation as well, as Sugimotoís example (19) shows.

(19) watasi wa oyoi- de Dover kaikyoo ga/ o watari-ta- i

I Top swimming by strait Nom/MC cross- want-Pres

ĎI want to swim across the strait of Dover.í (Ibid. (51))

Similar examples follow:

(20)a. kono kuruma wa ano torakku ga/ o oikos- e- na-i

this car Top that truck Nom/MC outstrip-Pot-Neg-Pres

ĎThis car cannot outstrip that truck.í

b. watasi wa kono miti ga/ *o aruki-zura-i

I Top this road Nom/*MC walk-hard-Pres

ĎI find this road hard to walk along.í

Note that in (20b) o is not even allowed. All these data show that the o/ga alternation occurs in transitive stative sentences with both the structural and inherent o phrases, which makes the validity of this test dubious.

 

3.3.4. Passivization Test

3.3.4.1. The case of the structural o phrase

This test is first applied to the eight types of the standard transitive sentences with the structural o phrases (given in (10)), in order to see whether all of them behave the same way under this test.

(21)i. Agentive Transitive (AT) [Agent, Patient]

a. Akira ga to o ake- ta ĎAkira opened the door.í

Nom door Acc open-Pas”

b. to ga Akira niyotte ake- rare-ta ĎThe door was opened by Akira.í

door Nom by open-Pass-Past

(22)ii. Non-agentive (NAT) and Causal (CT) Transitives [Cause, Patient]

a. kado no yasin ga kare no zyumyoo o tizime- ta

excessive ambition Nom his life Acc shorten-Past

ĎExcessive ambition shortened his life.í

b. kare no zyumyoo ga kado no yasin ??ni yotte/de tizime-rare-ta

his life Nom excessive ambition by/ due to shorten-Pass-Past

ĎHis life was shortened by/due to excessive ambition.í

(23)iii. Natural Force (NFT) and Instrumental (IT) Transitives.

[Instrument, Patient]

a. tunami ga kaihin no buraku o osot-ta

tidal wave Nom beachside Gen village Acc hit- Past

ĎTidal waves hit the village by the beach.í

b. kaihin no buraku ga tunami ni osow-are- ta

beachside Gen village Nom tidal wave by hit- Pass-Past

ĎThe village by the beach was hit by tidal waves.í

(24)iv. Non-causal Transitive (NCT) [Agent, Theme]

a. kodomo-tati ga inu o tatai-ta ĎThe children hit the dog.í

children Nom dog Acc hit- Past

b. inu ga kodomo-tati ni tatak-are- ta ĎThe dog was hit by the children.í

dog Nom children by hit- Pass-Past

(25)v. Neutral Transitive (NT) [Agent, Target]

a. watasi wa 30-pun mo tomodati o mat-ta

I Top as long as friend Acc wait-Past

ĎI waited for my friend as long as 30 minutes.í

b. *tomodati ga watasi ni 30-pun mo mat- are- ta

friend Nom I by minutes as long as wait-Pass-Past

??íMy friend was waited for by myself as long as 30 minutes.í

(26)vi. Creation Transitive (CRT) [Agent, Goal]

a. titi ga kono hon-bako o tukut-ta ĎFather made this bookcase.í

Father Nom this bookcase Acc make-Past

b. kono hon-bako ga titi niyotte tukur-are- ta

this bookcase Nom Father by make-Pass-Past

ĎThis bookcase was made by Father.í

(27)vii. Transitive Psychological Verb (TPV) [ Experiencer, Target]

a. wakamono-tati wa nakama no seikoo o yorokon-de i-ru

young people Top friends Gen success Acc be pleased- Pres

ĎThe young people are pleased with their friendsí success.í

b. nakama no seikoo ga wakamono-tati ni yorokob-are-te i- ru

friend Gen success Nom young people by please-Pass-State-Pres

ĎThe success of their friends pleases the young people.í

(28)viii. Relational Transitive (RT) [Theme, Theme]

a. kyoo no teian wa ikutuka no mondai o fukun-de i- ru

today Gen proposal Top several problem Acc involve-State-Pres

ĎTodayís proposal involves several problems.í

b. ikutuka no mondai ga kyoo no teian ni fukum-are- te i- ru

several problems Nom todayís proposal in involve-Pass-State-Pres

ĎSeveral problems are involved in todayís proposal.í

In brief it is clear that the structural o phrases in these eight types of transitive sentences undergo passivization, with Neutral Transitive (25) as the only exception. However, a closer examination reveals that (25) is excluded due to semantic and pragmatic factors. The sentences in (29), for example, are reasonably acceptable, despite the bookish or archaic flavor they convey.

(29)a. titioya no kaeri ga totemo mat- are- te i- ru

Father Gen return Nom strongly wait-Pass-State-Pres

ĎFatherís return is strongly waited for.í

b. sakkon wa tuyoi riidaasyippu ga mat-are- ru

nowadays strong leadership Nom wait-Pass-Pres

ĎNowadays a strong leadership is waited for.í

The other members of this class, nozom-u (wish), sagas-u (seek, search), kitai-su-ru (expect, hope for) mir-u (see, look at), kik-u (hear, listen to), and so on, undergo passivization freely, if semantic conditions are met.

(30)a. kono mokuhyoo no tassei ga kitai-s- are- te i- ru

this goal Gen achievement Nom hope for-Pass-State-Pres

ĎThe achievement of this goal is hoped for.í

b. gunsyuu no naka ni kare-ra no misuborasi-i sugata ga mi-rare-ta

crowd Gen midst in their shabby figure Nom see-Pass-Past

ĎTheir shabby figures were seen among the crowd.í

Even though the example (28b) does not pose any problem, the Relational Transitive class is problematic. Its membership consists of verbs lacking all of the distinctive features of transitivity, as the chart (2) shows. It is natural that many of them such as nas-u (constitute, form, make), mot-u (contain, have), kak-u (lack), resist passivization.

(31)a. kimi no setumei wa imi o nas- ana-i

your explanation Top sense Acc make-Neg-Pres

ĎYour explanation does not make sense.í

b. *imi ga kimi no setumei ni yotte nas- are- na- i

sense Nom your explanation by make-Pass-Neg-Pres

?íNo sense is made out of your explanation.í

(32)a. kare-ra wa subarasi-i kai o mot-i-masi-ta

they Top wonderful party Acc have-Pol- Past

ĎThey had a wonderful meeting.í

b. ??subarasi-i kai ga kare-ra niyotte mot-are- masi-ta

wonderful party Nom they by have-Pass-Pol-Past

??íA wonderful party was had by them.í

(33)a. kono hito wa zyoosiki o kai-te i- ru

this man Top common sense Acc lack-State-Pres

ĎThis man does not have common sense.í

b. *zyoosiki ga kono hito ni(yotte) kak-are- te i- ru

common sense Nom this man by lack-Pass-State-Pres

*ĎCommon sense is lacked by this man.í

Kak-u (lack) has the corresponding intransitive kake-ru (be lacking in), which forms a perfect sentence with the passive sense the ungrammatical version (33b) is intended to convey.

(34) zyoosiki ga kono hito ni kake-te i- ru

common sense Nom this man in lack-State-Pres

ĎCommon sense is lacking in this man.í

The result of the passivization test reveals that this test is valid for most of the transitive types, with the exception of Relational Transitive. However, the existence of RT makes the boundary between the structural and inherent statuses of the particle o fuzzy. This aspect is strengthened if it is found out that some of the inherent o phrases undergo passivization. This is in fact the case as is shown in the next section.

Before closing this section, we should pay special attention to the ni phrase (by phrase) in the passive. I assumed in Chapter 1 that the postposition ni marking inherent cases is linked to the É∆features [+ Agent], [+ Experiencer] and so on which are checked by the same É∆features of the predicates. All the VP internal argument and non-argument NPs are merged with the postpositions linked to É∆features. The passive test given above supports this assumption, namely, the original postpositions are kept in the passive sentences. The sentences of AT (21), NCT (24), NT (25), CRT(26), TPV (27) are good examples. The É∆features of external arguments of NAT and CT are checked by the É∆features [+ Cause], with the result that instead of ni or its alternative niyotte , de (Ďdue toí, Ďbecause ofí) or no tame ni (Ďbecause ofí) appears. An interesting fact related to the distribution of postpositions is that the É∆ features of the external argument of NFT and IT is checked by [+ Instrument], so that it is marked by de (Ďby means ofí) as expected, but at the same time ni can appear in its place. This reveals that [+ Agent] and [+ Instrument] are closely related in that [+ Instument] can never appear without the implicit existence of [+ Agent]. This is an example of interchangeability of postpositions, one of the characteristics of Japanese postpositions.

(35)a. siroi nuno ga tukue o oot- te i- ta

white cloth Nom desk Acc cover-State-Past

ĎA piece of white cloth covered the desk.í

b. tukue ga siroi nuno ni / de oow-are-te i- ta

desk Nom white cloth by/with cover-Pass-State-Past

ĎThe desk was covered by/with a piece of white cloth.í

Another point to be noted in this connection is that the ni in (28) is linked to the É∆feature [+ Location], rather than [+ Agent]. This suggests the possibility of the external argument of RT having [+ Location] instead of [+ Theme].

The use of the passive sentence is heavily constrained by semantic or pragmatic conditions, and the use of so called “by phrase” in passive sentences is mostly avoided. Under these circumstances, the passive sentences often resist sharp judgments as to their grammaticality.

 

3.3.4.2. The case of the inherent o phrase

The (b) sentences of the following show that passivization of the inherent o phrases are not totally unacceptable.

(36)a. minkan-ki ga sudeni Rosia no ryookuu o ton-de i- ru (Path)

civil airplane Nom already Russia Gen territorial sky MC fly State-Pres

ĎCivil airplanes have already been flying over the Russian territorial sky.í

b. ?Rosia no ryookuu ga sudeni minkan-ki ni tob-are-te i-ru

Russia Gen territorial sky Nom already civil airplane by fly-Pass-State-Pres

ĎThe Russian territorial sky has already been flown over by civil airplanes.í

(37)a. Hirosi ga taizyuu de Osamu o kosi- ta (Passing Point)

Nom weight in MC surpass-Past

ĎHirosi surpassed Osamu in weight.í

b. Osamu ga taizyuu de Hirosi ni kos- are- ta

Nom weight in by surpass-Pass-Past

ĎOsamu was surpassed by Hirosi in weight.í

(38)a. hotondo no gakusei ga kono gesyuku o de- ta (Source)

most student Nom this boarding house MC leave-Past

ĎMost students left this boarding house.í

b. kono gesyuku ga hotondo no gakusei ni de- rare- ta

this boarding house Nom most student by leave-Pass-Past

*íThis boarding house was left by most students.í

It is clear now that the MC o phrase undergoes passivization deriving grammatical sentences. The fact that some bona fide transitive sentences reject passivization, as some of the Relational Transitives do, and that the passivizability of some of the MCo phrases is undeniable, must make us highly suspicious of how well this test serves to distinguish structural from inherent o. To be blunt, passivization is not a valid test for distinguishing between the structural and inherent case.

 

3.3.5. Unaccusativity

Since unaccusativity concerns lexical relations, it is hard to use it as a decisive factor for judging between the structural and inherent case. Japanese has only a very limited number of verbs with the same forms for transitive and intransitive counterparts. Tozi-ru (close), hirak-u (open), mas-u (increase) are typical verbs, whose intransitive use corresponds to unaccusatives in English and some other languages. There are a fairly large number of Japanese verbs whose transitive and intransitive counterparts share the same stems and carry transitive or intransitive affixes. In syntactic terms, those with intransitive suffixes behave the same way as the unaccusatives in English.

Since unaccusativity is a lexically determined relation, not all the examples in (1) can undergo the accusative/unaccusative alternation. Adding appropriate examples wherever possible, I will try the test of unaccusativity, checking whether there is the o /ga correspondence in each example. Let us tentatively assume that the o phrase corresponding to the ga phrase in the unaccusative is the realization of the structural case.

First, we can predict that transitive sentences with the argument structures including Patient as one of their internal arguments have unaccusative counterparts with Patient as their subjects, since Patient is the undergoer of the change caused by the action denoted by the transitive verb. Our prediction is borne out, as the following examples show

(39)a. Akira ga futa o tot- ta (=(1a)) (AT) ĎAkira took off the lid.í

Nom lid Acc take off-Past [Agent, Patient]

b. futa ga tor- e- ta ĎThe lid came off.í

lid Nom take off-Int. Af-Past (Int. Af = intransitive affix)

(40)a. kono neko ga kabin o yuka ni ot-osi-ta

this cat Nom vase Acc floor to drop-Tr. Af-Past (NAT) [Cause, Patient] (Tr. Af = transitive affix)

ĎThis cat dropped the vase to the floor.í

b. kabin ga yuka ni oti- ta ĎThe vase dropped to the floor.í

vase Nom floor to drop-Past

(41)a. kado no yasin ga kare no zyumyoo o tizim-e-ta (=(1c))

excessive ambition Nom his life Acc shorten- Tr.Af-Past

ĎExcessive ambition shortened his life.í (CT) [Cause, Patient]

b. kado no yasin de kare no zyumyoo ga tizim-at- ta

excessive ambition due to his life Nom shorten-Int.Af-Past

ĎDue to excessive ambition his life shortened.í

(42)a. boohuu ga mado garasu o kowa-si- ta

storm Nom window pane Acc break-Tr. Af-Past

ĎThe storm broke the window pane.í (NFT) [Instrument, Patient]

b. boohuu de mado garasu ga kowa-re- ta

storm because of window pane Nom break-Int. Af-Past

ĎThe window pane broke because of the storm.í

(43)a. maruta ga dooro o husai-de i- ta “A log was blocking the road.”

log Nom road Acc block-State-Past (IT) [Instrument, Patient]

b. maruta de dooro ga husag- at- te i- ta

log with road Nom block-Int. Af-State- Past

ĎThe road was blocked with a log.í

The test has proved valid to the extent that transitive sentences with theÉ∆ role Patient undergo the o/ga alternation caused by unaccusativization. On the other hand, the Non-causal Transitive and Neutral Transitive whose argument structures do not include Patient do not permit unaccusativization.

(44)a. kodomo-tati ga to o tatai-ta (=1f) (NCT) [Agent, Theme]

children Nom door Acc hit- Past

ĎThe children hit the door.í

b. *to ga tatak-at- ta *íThe door hit.í

door Nom hit- Int.Af-Past

(45)a. hahaoya ga kodomo no kaeri o mat-te i-ru (NT) [Agent, Theme]

mother Nom child Gen return Acc wait-State-Pres

ĎMother is waiting for her child to come home.í

b. *kodomo no kaeri ga mat- at- te i- ru *íChildís return is waited for.í

child Gen return Nom wait-Int.Af-State-Pres

There are, however, exceptions. First, the Creation Transitive permits this alternation.

(46)a. imooto ga oyu o wak-asi- ta ĎMy sister boiled water.í

sister Nom hot water Acc boil-Tr. Af-Past (CRT) [Agent, Goal]

b. oyu ga wai-ta ĎThe water boiled.í

hot water Nom boil-Past

Since the o phrase in CRT is the end result of the action (Goal), it can function in the same way as Patient which is the undergoer of the change. This is not an unreasonal extention of the hypothesis, since the addition of a Goal phrase to an unergative sentence changes it to an unaccusative.9

The following is a genuine counterexample to our assumption that Patient is involved in the accusative and unaccusative correspondence.

(47)a. ano hito wa zyoosiki o kai- te i- ru (=(32))

that man Top common sense Acc lack- State-Pres

ĎThis man does not have a common sense.í (RT) [Theme, Theme]

b. zyoosiki ga ano hito ni kak-e-te i-ru

common sense Nom that man in lack-Int.Af-Pres

“Common sense is lacking in that man.”

This poses a problem to the analysis of the Relational Transitive, but I leave this matter for future studies.

Next, we have to see how the three subclasses of inherent o phrases behave under the unaccusativity test.

Since the first group, consisting of those like aruk-u (walk) and hasir-u (run), is made up of so called unergatives, the test has no meaning. The situation is the same with the remaining two subclasses. The sentences in (47) and (48) contain the ga/o correspondence, but the relation is entirely different from the accusative/unaccusative pair, since ga marking the transitive subject corresponds to the inherent o in (48) and (49).

(48)a. senzityuu ni kono hasi ga heitai o mukoogisi e watasi- ta

wartime during this bridge Nom soldier Acc opposite bank to transport -Past

ĎDuring the wartime this bridge let the soldiers go to the opposite bank.í

b. senjityuu ni heitai ga mukoogisi e kono hasi o watat-te it-ta

wartime during soldier Nom opposite bank to this bridge MC cross-go- Past

c. ĎDuring the wartime solders went to the opposite bank crossing this bridge.í10

(49)a. kono gesyuku ga hotondo no gakusei o dasi- ta

this boarding house Nom most of student Acc put out-Past

ĎThis boarding house put out most of the students.í

b. hotondo no gakusei ga kono gesyuku o de- ta

most of student Nom this boarding house MC leave-Past

ĎMost of the students left this boarding house.í (= (38b))

Even though there is a strict lexical limitation, the test of accusative and unaccusative correspondence reveals at least that the inherent o phrase cannot take part in this alternation.

Excepting the double o phrase test, the four remaining tests have revealed that there is no fast and hard line of demarcation between the structural and inherent o phrases, which supports my claim that É∆ checking by the predicate, rather than Case checking generally assumed under the Minimalist Program, has the syntactic function of NP licensing. The fact that one of the important structural tests like passivization has not yielded decisive results has to be seriously taken into consideration in future studies.

 

< 3.4 > Interchangeability of o phrases with some other postpostional phrases

It is well known that there are rather extensive interchanges among inherent postpositional phrases with appropriate semantic changes. The sentences in (6) present one typical example.

(6)a. gakusei-tati ga guraundo o hasit-te i- ru

Nom playground MC run- Prog-Pres

Ď The students are running in the playground.í

b. gakusei-tati ga guraundo de hasit-te i-ru (de - “at”)

As was pointed out before, the o version means a more or less total coverage of the space, while the de version simply specifies the location of the action. The example (7) shows the o/kara interchangeability.

(7) kodomo-tati ga koomon o/ kara de- ta

children Nom school-gate MC/from come out-Past

ĎThe children came out of/from the school gate.í

The semantic distinction is not obvious in (7), but there is another pair of sentences in which the semantic difference becomes clear.

(50)a. kokusekifumei-ki ga Nihon no zyookuu o hanare-ta

unidentified plane Nom Japan Gen sky MC leave- Past

ĎAn unidentified plane left the Japanese sky.í (Sugimoto (1986). (54))

b. kokusekifumei-ki ga Nihon no zyookuu kara hanare- ta

unidentified plane Nom Japan Gen sky from go away-Past

ĎAn unidentified plane went away from the Japanese sky.í

ĎAn unidentified plane left the Japanese sky.í

Sugimoto gives the following as the illustration of the semantic difference between the o and kara versions. (Ibid. 53)

(51)a. the o/kara versions b. the kara version

 

 

The o version cannot be used in the sense of (50b), that is, the starting point is obligatorily implied in the o version, as shown by (50a), while the kara version is used in both these senses. As for details of the selectional restriction involving this alternation, see Sugimoto (1983).

It is natural that verbs can select alternative postpostions that carry different semantic nuances. The relevant question here is whether or not there are cases of free interchanges of structural and inherent o phrases. Larson (1988) derives the double object construction by Case absorption of the prepostition “to”. In our terms, this means that the preposition “to” loses its inherent case assigning capacity and gives its complement NP the chance of getting the structural case from the verb by moving to the position adjacent to the verb. This is in fact a switch of a postposition to a case particle in Japanese in the framwork of Extended Standard Theory. However, in view of the fuzziness of the structural and inherent distinction so far revealed in this study, this type of switch can be another piece of evidence for non-distinctness of structural and inherent o phrases. There is in fact a tendency of a switch from ni (to) to o without a substantial semantic difference.

(52) sinnen ni wa oozei no hito ga Meizi Zinguu ni/o sanpai su-ru

New Year in Top many people Nom Meizi Shrine to Acc worship- Pres

ĎAt the beginning of the New Year many people go and worship at the Meiji Shrine.í

There is an unignorable semantic difference in the following pair.

(53)a. watasi-tati wa koko de hikooki ni nori-tui- da

we Top this place at plane to make a connection-Past

ĎWe made a plane connection at this station.í

b. watasi-tati wa koko de hikooki o nori-tui- da

we Top this place at plane Acc make a connection-Past

ĎWe made plane connections at this place.í

(53a) means that the plane connection was made from some other means of transportation, while (53b) implies the changes of planes. If we derive (53b) from (53a) by some syntactic process following Larson, the meaning content of ni is lost. This is why this case is not treated as an example of the switch of the inherent ni to the structural o.

There is a case of reverse switch from the structural o to the inherent ni , without semantic change.

(54)a. wakamono-tati wa nakama no seikoo o yorokon- de i- ru

young people Top friends Gen success Acc be pleased-State-Past

ĎThe young people are pleased with their friendsí success.í

b. ?wakamono-tati wa nakama no seikoo ni yorokon-de i-ru

(54b) is not perfect, but similar examples are often attested in everyday speech.

The existence of the sentences like (52) and (54) means that the boundary between the so called structural and inherent o is not clear cut conceptually, with the result that they are often interchanged .

 

< 3.5 > Scope interpretation in Japanese as evidence against the accusative Case checking

Ueda (1997) gives evidence for the assumption of the VP internal subject, which is introduced in Note 8 of Chapter 1. She goes on arguing that Japanese facts about the distribution of the positive polarity item (PPI) ka (Ďorí) cannot be accounted for by the accusative Case checking involving movement. First note the difference in distributions of the English PPI Ďorí and the Japanese ka.

(55)a. Roger ate [susi or unagi].

ĎRoger ate susi or Roger ate eel.í

b. Roger didnít eat [susi or unagi].

*íRoger didnít eat susi or Roger didnít eat eel.í

ĎRoger ate neither susi nor eel.í (Ueda (3))

In the positive sentence (55a) the disjunctive reading is given to the conjoined object NP, while in the negative version (55b) the disjunctive reading is lost due to the constraint prohibiting PPI from appearing in the position c-commanded by NEG. This is not the case with Japanese.

(56)a. Taroo ga [susi ka unagi]o tabe-ta

Nom susi or eel Acc eat- Past

ĎTaro ate susi or eel.í

b. Taroo ga [susi ka unagi] o tabe-nak- atta

Nom susi or eel Acc eat- NEG-Past

ĎTaroo didnít eat susi or Taroo didnít eat eel.í

The disjunctive reading is given to the negative sentence (56b), showing that the above constraint is not operative in Japanese. To account for this fact, she assumes that the object NP with the disjunctive conjunction ka must be raised to the position higher than the Spec-v position where the accusative Case checking is supposed to be carried out. This means that the accusative Case checking is nothing to do with this fact. (See Ueda (1977) for details.)

 

< 3. 6 > Conclusion

My claim has been well supported by the result of this study. This claim is consonant with the basic idea of “economy” in the Minimalist Program. The É∆checking is necessay anyway in syntax. Beyond that, the Case checking is induced by a functional head whose feature must check the same Case feature carried by an appropriate NP. The existence of such functional heads is parametrized, a very natural conclusion.

It was shown that there is no fast and hard line of demarcation between the structural and iherent case particle o. However, there are good reasons for regarding o in some contexts as a case particle in the traditional sense. To account for this problem I will make a proposal that the argument structure specifying the external argument is to be augmented with the specification of the hierarchy of internal arguments, in particular the specification of the internal argument of the highest prominence. Since a thorough study of ni, the so called dative case particle, is needed to carry out researches along this line, I will leave this problem for future studies.

 

Notes to Chapter 3:

* This is a revised version of my paper “On the Japanese Particle o.” Comments from B Horton, T Ito and Y Ueda contributed to the revision of the preliminary version of the paper.

** The abbreviatory symbols used in this chapter are: Acc = accusative, AT = Agentive Transitive, Caus = causative, CT = Causal Transitive, Gen = genitive, Int. Af = intransitive affix, IT = Instrumental Transitive,

MC = movement complement, Nom = nominative, NAT = Non-agentive Transitive, NCT = Non-causal Transitive, NFT = Natural Force Transitive, Nom = nominative, NT = Neutral Transitive, Pass = Passive, Past = past tense, Pol = polite form, PP = postposition, Pres = present tense, Prog = progressive, RT = Relational Transitive, State = stative, Top = topic, TPV = Transitive Psychological Verb, Tr. Af = transitive affix

1 Stativity is one of the characteristics marking the transitive and intransitive distinction, with the former “low” and the latter “high” in this feature.

2 Jacobsen (1992) calls this type of relation “dominance relation”, saying that the nominative NP has a dominant status over the accusative NP.

3 Hopper and Thompson (1980) state that the total affectedness of the object NP is one of the features that strengthen transitivity. This semantic difference between o and de in (6) is caused by this transitivity feature.

4 Here is no significant semantic difference between o and kara versions. However, there are cases where this does not hold. See (49) and (50) in Section 5 for details.

5 The circumstantial o has long been discussed in relation to the case marking in the Extended Standard Theory. The discussion originated from Harada (1973).

6 There are proposals to derive abstract É∆ roles from lexical conceptual structures instead of using the traditional labels. These labels are used in this paper for the ease of exposition.

7 The external argument of each verb (or adjective) is in general automatically determined by the É∆ role hierarchy.

8 The term “Double o Constraint” is ascribed to Harada.

9 Kizu (unpublished ) gives a persuasive argument for the function of the Goal phrase in unergative sentences.

(i)a. *In the room ran the children. (Kizu (4))

b. Into the room ran the children. (Levin & Rappaport (1989))

As the locative inversion is one of the unaccusative diagnostics in English, Kizu argues that (ib) with the unergative “ran” has assumed unaccusativity due to the presence of the Goal phrase “into the room”.

(ii)a. ?*Gakusee-gai kooen-de isoide san-nini hashit-ta (Kizu (5))

student NOM park- in hurriedly threeCL runPAST

ĎThree students ran in the park hurriedly.

b. Gakusee-gaij kooen-made/?ni isoide tj san-nin hashit-ta

student NOM park-until / to hurriedly threeCL runPAST

ĎThree students ran to the park hurriedly.í

Kize explains the grammaticality contrast in (ii) as follows: “In (5a) (our (iia)), the ungrammaticality is readily explained by the fact that the numeral quantifier san-nin “three people” occurs within the VP, and thus cannot be linked with the antecedent gakusee-ga. On the other hand the same linking is fine in sentence (5b) (our (iib)) which contains a goal phrase.” The grammaticality of (iib) can be accounted for by taking this sentence as an unaccusative sentence and the derived subject gakusee originates in the position occupied by the trace.

10 I owe this example to Yukiko Ueda.

 

References

Chomsky, Noam. 1970. "Remarks on Nominalization," in Jacobs and Rosenbaum (eds.) Readings in English Transformational Grammar. Ginn and Co. 184-221.

. 1986. Barriers. The MIT Press.

. 1995. The Minimalist Program. The MIT Press.

Fukui, Naoki and Yuji Takano.1996. "Symmetry in Syntax: Merge and Demerge." Unpublished.

Harada, S. I. 1973. “Counter Equi NP Deletion,” Annual Bulletin. Research Institute of Logopedics and Phoniatrics. University of Tokyo, 7, 113-147.

Hopper, Paul J., and Sandra A. Thompson. 1980. “Transitivity in grammar and discourse,” Language 56, 251-99.

Inoue, Kazuko. 1988. “An Attempt to Explain canonical case alignments in Japanese,” Annual Reports of the Research Institute of Language and Culture, No. 3, 2-24. Tsuda College.

1992. “Interaction of Theta-marking and Case Assignment,” Studies in Linguistics and Language Teaching, No. 2, 1-32. Kanda University of International Studies.

1994. “Transitivity and causative sentences,” In: Tokunaga (ed.) Systematic Study of Language Variation and Its Application to the Japanese Language Teaching, 109-136. Kanda University of International Studies.

1997a. “A note on relevant questions involved in the case system,” In: Inoue (ed.) The Interim Reports of the Researches Suppoted by the Monbusho COE Grant:

1997b. "Reanalysis of the Japanese Case Particle Ni," in M Ukaji, T Nakao, M Kajita, and S Chiba (eds.) Studies in English Linguistics: A Festschrift f or Akira Ota on the Occasion of His Eightieth Birthday. 968-985. Taishuukan.

1997c. “On Japanese Particle o,” to appear in M Janse (ed.) A Festschrift for Professor Uhlenbeck on the Occasion of His Eighty-fifth Birthday.

Jacobsen, W. M. 1992. The Transitive Structure of Events in Japanese. Kurosio Publishers.

Kizu, Mika. “A Goal Phrase and Unaccusativity.” Unpublished. McGill University.

Konig, Ekkehard. 1997. “Towards a Typology of Intensifiers (Emphatic Reflexives),” Plenary Sessions Texts, XVIth International Congress of Linguists, 1-16.

Kuroda, S.-Y. 1965. Generative Grammatical Studies in the Japanese Language. Ph. D. dissertation. M. I. T.

1970. “Remarks on the Notion of Subject and Reference to Words Like also, even, and only,” in Kuroda (1992), 78-113.

1978 ĎCase marking, canonical sentence patterns, and counter equi in Japnanese. (A preliminary survey)í, in John Hinds and Irwin Howards (eds.) Problems in Japanese Syntax and Semantics. Kaitakusha. 30-51.

1988. "Whether We Agree or Not: Rough Ideas About the Comparative Syntax of English and Japanese," Unpublished.

1992, Japanese Syntax and Semantics. Kluwer Academic Publishes.

1992, “Judgment Forms and Sentence Forms,” in Kuroda (1992), 13-77.

Lasnik, Howard, and Mamoru Saito. 1992. Move ÉŅ. MIT Press.

Larson, Richard K. 1988. “On the double object construction,” L. I. 19, 335-391.

Levin, Beth and Malka Rappaport. 1989. “An approach to Unaccusative Mismatch,” NELS 19, Proceedings of the 19th Confereance of the North Eastern Linguistics Society, pp. 314-329.

Linebarger, 1987. “Negative Polarity and Grammatical Representation,” Linguistics and Philosophy, 10, 325-87.

Marantz, Alec. 1992. "Case and licensing," Proceedings of the Eighth EASTERN STATES CONFERENCE ON LINGUISTICS, German Westphal, B. Ao, and H.-R. Chae, 91, 234-252.

. 1995. "The Minimalist Program," in Gert Webelbuth (ed.) Government and Binding Theory and the Minimalist Program. Blackwell. 351-382.

. 1996. "Cat as a phrasal idiom: Consequences of late insertion in Distributed Morphology," unpublished. MIT.

Sugimoto, Takesi. 1983. “Hanareru, hazureru, toozakaru “ Studies of Japanese 6, Tokyo Metropolitan University.

Ā@Ā@ Ā@1986. “Case Particles,” In: Okutsu et al (eds.) Studies on the so called Japanese particles, 227-380. Tokyo: Bonjin-sha

Takezawa, Koichi. 1987. Configurational Approach to Case-Marking in Japanese, Ph.D. dissertation, University of Washington.

Ueda, Yukiko. 1997. “Some Arguments against A-Movement Analyses of Scopal Interaction in Japanese.” Unpublished.

Watanabe, Akira. 1996. "Nominative-Genitive Conversion and Agreement in Japanese: A Cross-Linguistic Perspective," Journal of East Asian Linguistics, Vol. 5, No. 4. 373-410.