The Discourse Basis of Constructions: Some Evidence from Korean Acquisition

1. Introduction The " return of constructions " to linguistic theory has introduced a new approach for analyzing argument structure and its acquisition (Tomasello, 1998). In this approach (Goldberg, 1995), argument structure is conceptualized as involving clause-level semantic-syntactic templates: A construction accommodates verbs of a particular semantic type taking arguments that bear certain semantic and syntactic relationships to those verbs. The prototypical transitive construction, for example, involves a volitional agent encoded as the subject of the verb, which causally affects an inanimate patient encoded as the direct object. This construction-level meaning holds across specific verbs and arguments, and is extended to incorporate other less prototypical meanings (Goldberg, 1995: 117-118). If we assume that " constructions are the basic units of language " (Goldberg 1995), it follows that the child's task is to acquire constructions. What does this entail? Research within a Construction Grammar framework has emphasized the central role of the verb and the semantic relations between arguments and verb. What is the role of the semantic prototype 'human Agent-inanimate Patient' in the acquisition of transitivity (Slobin, 1985; Bowerman, 1990)? Since Construction Grammar was originally based on individual, invented sentences, usage-based information about the pragmatic status of arguments is not fundamental to the theory. In English, the core arguments of verbs must be overt in most contexts, and sentences analyzed in a construction framework typically feature lexical arguments. So languages that permit pervasive ellipsis of arguments in discourse raise important questions—theoretical and methodological—for a construction-based approach to the acquisition of argument structure. From a methodological point of view, languages with ellipsis pose a challenge: How can we measure 'acquisition' of a construction when adult usage neither requires, nor typically exhibits, the full form of that construction? How should we conceptualize the mental representation of a construction, such as the transitive construction, in languages where it takes various surface shapes? Does its full form, i.e. verb plus two overt arguments, have special cognitive status? Discourse-functional research on grammar provides a useful complement to Construction Grammar. It is based on discourse data and permits a focus on the discourse status of arguments and their realization. For example, cross-linguistic research on Preferred Argument Structure (Du Bois, 1987) has shown that two arguments of transitive verbs—A (subject of a transitive verb) and O (direct object)—are rarely both lexical. Instead, the A argument, usually given information, tends to be pronominal or elliptical, while the O …