Topic-Comment Articulation in Japanese: A Categorial Approach

Topic-Comment Articulation in Japanese: A Categorial Approach Hiroaki NAKAMURA1 1Department of Liberal Arts, Japan Coast Guard Academy, 5-1, Wakaba-cho, Kure City, Hiroshima, 737-0832, JAPAN {Hiroaki Nakamura, nakamura@jcga.ac.jp) Abstract. This paper deals with the topic-comment articulation of information structures conveyed by sentences. In Japanese, the topic marker WA is attached not only to an N(P) but also to a PP or a clause, forming the information structure of a sentence, where the topicalized part represents a restriction and the remaining part of the sentence represents a nuclear scope. I propose the type-raised category for WA which embodies flexible constituency to realize divergent topic-comment structures. With our categorial definition for the topic marker and the combinatory rules in Combinatory Categorial Grammar, which derive the tripartite representation of the information state of a sentence, our grammar architecture can dispense with an independent level representing the information structure. Keywords: information structure, topic, comment, topicalization. 0. Introduction The study of Japanese topic marker “WA” has attracted much attention, especially in comparative studies with the nominative case marker “GA.” For instance, it has been assumed that WA-marked constituents represent old information while GA-marked ones new information, or WA marks the subjects of kind/individual-level predicates while GA the subjects of stage-level predicates. Consider the examples in (1): (1) a. Uma-wa kusa-o taber-u. horse-TOP grass-ACC feed.on-PRES ‘Horses feed on grass.’ (kind/individual-level sentence) b. Uma-ga kusa-o tabe-teir-u. horse-NOM grass-ACC eat-PROG-PRES ‘A horse is eating grass.’ (stage-level sentence) (2) a. Uma-ga kusa-o taber-u. horse-NOM grass-ACC feed.on-PRES ‘Horses feed on grass.’ (kind/individual-level sentence) b. Uma-wa kusa-o tabe-teir-u. horse-TOP grass-ACC eat-PROG-PRES ‘The horse is eating grass.’ (stage-level sentence) When the subject of the kind/individual-level predicate taberu 'feed on' is marked with the nominative case GA in a matrix clause, as in (2a), an exhaustive listing reading is forced in some marked context, and hence construed as (a part of) FOCUS. When the referent of the subject of a stage-level predicate is already given in discourse, it can (and should) be marked with the topic marker (see Kuno (1973), Shirai (1985), Uechi (1996), Yabushita (2005) for discussion). These observations often lead to the Diesing style analysis of individual/stage-level subjects in Japanese in which the subjects of generic/individual-level predicates should be marked with the topic marker WA, whereas those of stage-level predicates with the nominative marker GA as default. In this kind of analysis, the difference in marking of subjects is reduced to structural differences in syntax. The subjects of generic sentences are projected onto [Spec, IP] position, while those of episodic sentences are projected onto 198 [Spec, VP] position. But their account immediately runs into a problem because we easily find many counterexamples to this simple generalization about WA/GA-distinction in terms of genericity of sentences. Observe some of the examples shown in (3). (3)a. Kodomo-demo [uma-ga/?uma-wa usa-o taberu-koto]-o sit-teir-u. Children-even [horse-NOM/horse-TOP grass-ACC eat-COMP]-ACC know-PRES 'Even children know that horses feed on grass.' b. [Uma-ga/*Uma-wa taber-u] kusa-o osie-te-kudasai. horse-NOM/ -TOP eat-PRES] grass-ACC teach-PRES please ‘Please tell me the kind of grass which horses eat.’ …

Content truncated to 500 words