The structure of the resultative light verb construction in Japanese
*
Mikinari MATSUOKA
1. Introduction
This paper focuses on the syntactic structure of the construction in Japanese that involves the verb su ‘do’ and a resultative adjective with the inflectional suffix –ku or –ni denoting a state of the referent of the accusative NP, as shown in (1):1
(1) a. Mary-ga teeburu-o kirei-ni su-ru.
Mary-Nom table-Acc clean-Aff do-Pres
‘Mary makes the table clean.’
b. John-ga heya-o atataka-ku si-ta.
John-Nom room-Acc warm-Aff do-Past
‘John made the room warm.’
Given that the verb su is a light verb and the adjective is interpreted as being predicated of the accusative NP, it is claimed in previous studies that the verb is not involved in the θ–role assignment to the NP at all and the NP is generated as the subject of the adjective within a small clause complement of the verb (Kikuchi and Takahashi 1991, Fukumitsu 2001, Sakai et al. 2004). However, the present study casts doubt on this view, observing that the accusative NP has a property that is typically found with the direct object of a lexical verb and not with the subject of a predicate adjective. It is then proposed that the construction involves the same structure as ordinary lexical transitive verbs, assuming that those verbs are syntactically decomposed into light verbs and a predicate adjective (Baker 2003). Hereafter, I refer to the construction in (1) as the resultative light verb (RLV) construction.
The organization of the following discussion is as follows. In section 2, we look at data indicating that the verb involved in the RLV construction is a light verb rather than a heavy verb, and that the accusative NP is assigned its θ-role from the adjective. In section 3, it is shown that the accusative NP in the construction has a property of the direct object of a lexical verb. In section 4, I propose the structure of the RLV construction on the basis of the decomposition analysis of lexical verbs. Finally, concluding remarks are given in section 5.
2. Su as a light verb
As discussed by Sakai et al. (2004: 350), the verb su in the RLV construction participates in a transitivity alternation with the unaccusative verb nar ‘become’, as shown in (2):
(2) a. Takasi-ga heya-o kirei-ni si-ta.
Takashi-Nom room-Acc clean-Aff do-Past
‘Takashi made the room clean.’
b. Heya-ga kirei-ni nat-ta.
room-Nom clean-Aff become-Past
The object NP heya ‘room’ in (2a) thematically corresponds to the subject NP in the unaccusative counterpart in (2b). Noting that both the transitive and the unaccusative counterpart involve the same predicate adjective, Sakai et al. argue that the adjective assigns a θ-role to the shared argument and su/nar are light verbs in this alternation.
By contrast, the heavy verb su does not alternate with nar in this way (Sakai et al. 2004: 357):
(3) a. Setsuko-ga doosookai-no kanji-o si-ta.
Setsuko-Nom class reunion-Gen coordinator-Acc do-Past
‘Setsuko worked as the coordinator of the class reunion.’
b.* Doosookai-no kanji-ga nat-ta.
class reunion-Gen coordinator-Nom become-Past
‘The class reunion coordinator became.’
The object NP of the transitive verb su in (3a) cannot become the subject of the intransitive verb nar in (3b). Sakai et al. suggest that the θ-role assigned to the object in (3a) inherently comes from su as a heavy verb and is unavailable for the subject of nar in (3b).
Moreover, as noted by Sakai et al. (2004: 358), the resultative adjective in the RLV construction cannot be elided, as shown in (4):
(4) Setsuko-ga benkyoobeya-o kirei-ni si-ta.
Setsuko-Nom study-Acc clean-Aff do-Past
* Kanojo-wa kodomobeya-mo si-ta.
she-Top nursery-also do-Past
‘Setsuko made the study clean. She also made the nursery clean.’
This is contrasted with the optional occurrence of a resultative adjective as a secondary predicate, which is shown in (5) (see Sakai et al. 2004: 362):
(5) Mary-ga teeburu-o (kirei-ni) hui-ta.
Mary-Nom table-Acc clean-Aff wipe-Past
‘Mary wiped the table (clean).’
Sakai et al. argue that the adjective must be present in (4) because the object NP cannot be assigned a θ-role from the light verb. On the other hand, the adjective can be elided in (5) because the object NP is assigned a θ-role from the lexical verb.
Given these facts, Sakai et al. (2004: 368) claim that the RLV construction involves the structure in (6) (see also Kikuchi and Takahashi 1991, Fukumitsu 2001):
(6) [vP NP-Nom [AP NP-Acc Adj] suv]
The accusative NP and the adjective constitute an AP small clause, within which the NP is assigned a θ-role from the adjective. The AP is merged with the verb su as the v head, and the subject NP is generated in Spec,vP.
3. The accusative NP as a direct object
As discussed above, it is claimed in previous studies of the RLV construction that the accusative NP is generated as the subject of the adjective in the complement clause and the verb su as a light verb is not involved in θ-role assignment to the NP. In this section, however, it is shown that the accusative NP has a property that is normally found with the direct object of a lexical verb and not with the subject of an adjective, which would not be predicted under the analysis in previous studies.
It is observed by Kishimoto (2005) that the quantificational adverb ippai ‘a lot’ in Japanese can be interpreted as modifying an internal argument of a verb by specifying the quantity of the referent of the argument. However, the adverb cannot be construed with an external argument. For example, in (7), the adverb occurs in a sentence involving a lexical transitive verb and modifies the object argument, not the subject (Kishimoto 2005: 122):
(7) Gakusei-ga kabin-o kyoositu-de ippai kowasi-ta.
student-Nom vase-Acc classroom-in a.lot break-Past
‘The student(s) broke a lot of vases in the classroom.’ NOT: ‘A lot of students broke vases in the classroom.’
Kishimoto (2005: 122-124) also provides examples showing that ippai can be construed with the subject of an unaccusative verb or a passive, but not with the subject of an unergative verb.2
With these facts, Kishimoto (2005) suggests that the argument modified by ippai corresponds to a delimited expression in the sense of Tenny (1994) and is defined by aspectual terms. Baker (1997: 99) notes that the same kind of adverb is found in Mohawk and modifies the event argument of the VP it attaches to. Thus, if the modified verb has an argument that “measures out” the event (“incremental theme”), many events correspond to many tokens of the kind denoted by the argument. Given this analysis, we can understand why ippai can be associated with the internal argument of a verb, and not with its external argument.
Now note that when ippai occurs in the RLV construction, it can modify the accusative NP by specifying the quantity of its referent, as shown in (8) (Matsuoka 2010: 302):
(8) a. John-ga kami-o ippai maru-ku si-ta.
John-Nom paper-Acc a.lot round-Aff do-Past
‘John made a lot of paper round.’
b. Mary-ga koppu-o ippai kirei-ni si-ta.
Mary-Nom glass-Acc a.lot clean-Aff do-Past
‘Mary made a lot of glasses clean.’
These are contrasted with the following examples in (9), where ippai occurs in a sentence involving one of the adjectives in (8) as the primary predicate. Note that the subject NP cannot be modified by the adverb and the sentences are ungrammatical ((9a) from Kishimoto 2005: 129):3
(9) a. * Kami-ga (nazeka) ippai maru-katta.
paper-Nom somehow a.lot round-Pred.Past
‘A lot of paper was round (for some reason).’
b. * Koppu-ga ippai kirei-datta.
glass-Nom a.lot clean-Pred.Past
‘A lot of glasses were clean.’
The examples in (9) suggest that the theme arguments of these adjectives, unlike those of verbs, are external arguments, as argued by Baker (2003) and Kishimoto (2005). They also indicate that the accusative NPs in (8) are not simply generated as the subjects of these adjectives, contrary to the analysis of the RLV construction proposed in previous studies (see (6)). Rather, the NPs seem to have the same grammatical status as internal arguments of lexical verbs.4,5
4. The structure of the RLV construction
The data presented above indicate that the RLV construction has a puzzling property: although the predicate consists of a light verb and an adjective, the accusative NP behaves like the direct object of a lexical verb rather than the subject of an adjective. In this section I propose a solution to this puzzle on the basis of the view that lexical verbs are decomposed into light verbs and a predicate adjective in the syntax.
4.1 Decomposition of lexical transitive verbs
It is claimed in some recent studies that predication relationships in general are mediated by a functional category in phrase structure (Bowers 1993, Baker 2003, den Dikken 2006). Bowers (1993) proposes that there is a specific category called Pred that serves for this purpose. According to this view, a sentence involving an adjective as the primary predicate like (10a) is assigned a structure in (10b) (see Baker 2003: 35):
(10) a. Chris is hungry.
b. [TP Chris T [PredP <Chris> Pred [AP hungry]]]
The predicate AP is merged with a Pred head, and the combination is merged with the theme argument of the adjective, which becomes Spec,PredP and then raises to Spec,TP for EPP reasons. Furthermore, it is assumed that the θ-role assigned to the argument in Spec,PredP is created out of the AP by the Pred (Baker 2003: 36). This is comparable to the θ-role assignment to the external argument in v-VP configuration discussed in Chomsky (1995: 315).
Noting that the theme θ–role is assigned to the direct object of a transitive verb as well as to the subject of a predicate adjective, Baker (2003: 77ff.) suggests that all verbs are derived from adjectives in the syntax. In particular, he claims that the structure of the predication of an adjective in (10b) is embedded within the structure of an ordinary transitive verb. Given the decomposition analysis of verbs proposed in work on lexical semantics (Dowty 1979, Jackendoff 1983), he assumes that a transitive verb consists of three elements, CAUSE, BE, and ADJECTIVE, as illustrated in (11), where x is defined to be the agent argument and y is the theme:
(11) [x CAUSE [y BE [ADJECTIVE]]]
Baker proposes that CAUSE is identified with a v head (Chomsky 1995), BE with a Pred head, and ADJECTIVE with an A head denoting a property. The surface verb is derived by combining these three heads into a single item by successive head movement. For example, the verb wipe in (12a) is assumed to be derived in a structure like (12b) under this analysis (see Baker 2003: 81, 221):
(12) a. I wipe the table.
b. [TP I T [vP <I> CAUSEv [V/PredP the table BEV/Pred [AP WIPEDA]]]]
It is noted by Baker (2003: 87) that Pred is inherently a functional category but becomes a lexical category, that is, a verb, in (12b) as a result of acquiring encyclopedic content by combining with ADJECTIVE. The direct object of the verb in Spec,V/PredP is assigned the theme θ-role in the same configuration as the subject of the adjective in (10b).
Furthermore, Baker claims that certain syntactic differences between the theme subject of an adjective and the theme object of a transitive verb that are found across languages can still be accounted for under this analysis. In general, the subject of an adjective patterns with an external argument, as we saw with modification by ippai in (9), whereas the object of a verb behaves like an internal argument (e.g., ne-cliticization in Italian (Cinque 1990)). Baker attributes this contrast to whether the argument in its original position is c-commanded by a lexical head in the minimal domain of the head.6 The object in Spec,V/PredP in (12b) is c-commanded by the lexical verb, which
is derived by conflating A+Pred and raised to v, being inside the maximal projection of the verb, that is, V/PredP. On the other hand, the subject of the adjective in Spec,PredP in (10b) is never c-commanded by a lexical head in its minimal domain since A is lexicalized without moving to Pred.
4.2 Decomposition in the RLV construction
I would like to propose that the RLV construction involves an underlying structure comparable to that in (12b). In particular, as shown in (13b), V/Pred is merged with AP headed by the resultative adjective, as its complement. The accusative NP as the theme argument is generated in Spec,V/PredP. Then, the V/PredP is merged with the v head, and the nominative NP as the agent argument is generated in Spec,vP. Moreover, T is added and the nominative NP raises to Spec,TP for EPP reasons:
(13) a. Mary-ga teeburu-o kirei-ni su-ru.
Mary-Nom table-Acc clean-Aff do-Pres
‘Mary makes the table clean.’
b. [TP Mary-ga [vP <Mary-ga> [V/PredP teeburu-o [AP kirei-ni] V/Pred] v] T]
I assume that the V/Pred head incorporates into the v head in overt syntax and is spelled out as the light verb su, whereas the adjective undergoes the same operation by way of V/Pred in LF (see Fukumitsu 2001), as shown in (13b).
There is a piece of data supporting the claim that the adjective undergoes LF incorporation into the verb. It is known that indeterminate pronouns such as dare ‘anyone’ and nani ‘anything’ in Japanese can serve as negative polarity items when they are bound by the Q particle mo (Kuroda 1965). Sakai et al. (2004: 363) observe that an indeterminate pronoun occurring as the accusative NP in the RLV construction receives the negative polarity interpretation when the particle mo is attached to the adjective, as shown in (14):
(14) Kenji-wa dare-o siawase-ni-mo si-na-katta.
Kenji-Top anyone-Acc happy-Aff-Q do-Neg-Pred.Past
‘Kenji did not make anyone happy.’
According to Kishimoto (2001), when mo is attached to a head, they form a complex head due to head adjunction. Given this, the example in (14) is assumed to involve the following LF structure under the present analysis: (15) Kenjii-wa [vP ei [V/PredP dare-o [AP<A-Aff-Q>] <V/Pred>] A-Aff-Q+V/Pred+v]-na-katta.
It is argued by Kishimoto (2001) that indeterminate pronouns can be bound by the particle mo only if they are contained in the least maximal projection dominating the particle in LF. In (15), the least maximal projection dominating mo is vP, and the indeterminate pronoun is contained in it, being in Spec,V/PredP. Thus, it is correctly predicted that the pronoun receives the negative polarity interpretation in (14).
The properties of the RLV construction discussed in sections 2 and 3 are also accounted for under the present analysis. First, the verb su behaves as a light verb rather than a heavy or lexical verb, participating in a transitivity alternation (see (2)). This is expected because the verb is a spell-out of conflating the two verbal functional heads, v and V/Pred, without involving the lexical component of the predicate, that is, the adjective. Second, the resultative adjective cannot be elided (see (4)). This fact is also predictable given that the theme θ-role assigned to the accusative NP is created out of the AP headed by the adjective in (13b) in the same way as the theme role assigned to the direct object of the lexical verb in (12b). Then, the NP could not appear in the construction without the adjective. Third, the accusative NP behaves like the direct object of a verb rather than the subject of an adjective, being modified by ippai (see (8)). This fact is attributed to a structural property that the NP shares with the direct object of a lexical verb. In particular, it is in Spec,V/PredP and c-commanded by the adjective that incorporates into v by way of V/Pred. Although the incorporation is assumed to occur in LF rather in overt syntax in the RLV construction, it is conceivable that the NP acquires the same property as the object of a lexical verb due to the operation.
5. Concluding remarks
In this paper, I have examined the structure of a light verb construction in Japanese in which a resultative adjective is predicted of the accusative NP. It has been proposed that the construction involves the same underlying structure as ordinary transitive verbs, being decomposed into two functional heads and a lexical head of adjective. Although the three heads are assumed to be conflated into one head in overt syntax in the structure of ordinary transitive verbs, I have argued that in the light verb construction, the two functional heads are combined in overt syntax and spelled out as the light verb, while the adjective incorporates into the verb in LF. If the proposed analysis is on the right track, it will motivate reconsideration of the traditional distinction between lexical
and functional categories, which I leave for future research.
Notes
* I would like to thank Hiromu Sakai for helpful discussion of some of the material presented in this paper. This research is supported in part by JSPS KAKENHI Grant Number 26370447. All remaining errors and inadequacies are my own.
1 Following abbreviations are used in the glosses: Acc (accusative case), Aff (inflectional affix), Comp (complementizer), Gen (genitive case), Neg (negative), Nom (nominative case), Pred (predicative head), Pres (present), Q (quantificational particle), Top (Topic).
2 Kishimoto (2005) also observes that the adverb ippai cannot specify the quantity of the referent of an indirect object or that of an adjunct.
3 As noted by Kishimoto (2005), there are a few adjectives such as hituyoo ‘need’ and hosi ‘want’ that seem to select an internal argument modified by ippai. Kishimoto suggests that these adjectives are distinguished from the adjectives in (8) and (9) in not denoting the properties or attributes of their argument (cf. Cinque (1990) for adjectives in Italian that select an internal argument). What is important in the present study is that the arguments of adjectives that are generated as external arguments in primary predication (see (9)) behave like internal arguments in the RLV construction (see (8)).
4 The fact that the accusative NP in the RLV construction can be modified by ippai, as shown in (8), would not be attributed to its accusative-Case marker. In the following example of a Raising-to-Object construction involving the same adjective as (8b) and (9b) in the embedded clause, the accusative NP, which is assumed to originate as the subject of the adjective (Kuno 1976), cannot be modified by the adverb:
(i) Taroo-ga koppu-o (*ippai) kirei-da to omot-ta. Taro-Nom glass-Acc a.lot clean-Pred.Pres Comp think-Past ‘Taro thought the glass to be clean.’, NOT: ‘Taro thought a lot of glasses to be clean.’
5 Matsuoka (2010) argues that the accusative NP in the RLV construction is generated as the direct object of the verb rather than as the subject of the adjective by providing two kinds of data, one of which is concerned with modification by ippai.
6 The notion of minimal domain is defined as follows (Baker 2003: 67 fn. 21):
(i) X is in the minimal domain of Y if and only if X is inside the maximal projection YP of Y and there is no maximal projection ZP that properly contains X but not Y or a trace bound by Y.
References
Baker, Mark. 1997. Thematic roles and syntactic structure. In Elements of grammar, ed. by Liliane Haegeman, 73-137. Dordrecht: Kluwer.
Baker, Mark. 2003. Lexical categories: Verbs, nouns, and adjectives. Cambridge: Cambridge University Press. Bowers, John. 1993. The syntax of predication. Linguistic Inquiry 24: 591-656.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Cinque, Guglielmo. 1990. Ergative adjectives and the lexicalist hypothesis. Natural Language and Linguistic Theory 8: 1-40. Dikken, Marcel den. 2006. Relators and linkers: The syntax of predication, predicate inversion, and copulas. Cambridge, MA:
Dowty, David. 1979. Word meaning and Montague grammar. Dordrecht: Reidel.
Fukumitsu, Yuichiro. 2001. Covert incorporation of small clause predicates in Japanese. In Formal approaches to Japanese linguistics 3: MIT working papers in linguistics 41, ed. by M. C. Cuervo, D. Harbour, K. Hiraiwa, and S. Ishihara, 251-266. Cambridge MA: MIT Press.
Jackendoff, Ray. 1983. Semantics and cognition. Cambridge, MA: MIT Press.
Kikuchi, Akira, and Daiko Takahashi. 1991. Agreement and small clauses. In Topics in small clauses, ed. by Heizo Nakajima and Shigeo Tonoike, 75-105. Tokyo: Kurosio.
Kishimoto, Hideki. 2001. Binding of indeterminate pronouns and clause structure in Japanese. Linguistic Inquiry 32: 597-633. Kishimoto, Hideki. 2005. Toogokoozoo to bunpookankei [Syntactic structure and grammatical relations]. Tokyo: Kurosio
Publishers.
Kuno, Susumu. 1976. Subject raising. In Syntax and semantics 5: Japanese generative grammar, ed. by Masayoshi Shibatani, 17-49. New York: Academic Press.
Kuroda, Shige-Yuki. 1965. Generative grammatical studies in the Japanese language. Ph.D. dissertation, MIT.
Matsuoka, Mikinari. 2010. A note on the structure of a suru causative in Japanese. In Bulletin of the Faculty of Education and Human Sciences, the University of Yamanashi, Vol. 11, pp. 298-305.
Sakai, Hiromu, Adrian Ivana, and Chao Zhang. 2004. The role of light verb projection in transitivity alternation. English Linguistics 21: 348-375.
