The phenomenon of coreference, where two linguistic expressions refer to the same entity in the outside world, is basic to discourse processing and has been a central topic in language studies. It has been found that pronominal expressions, such as the pronoun he, promote discourse coherence and coreference because they provide critical links to integrate separate utterances into a model of discourse (Gordon, Grosz & Gilliom, 1993; Grosz, Joshi & Weinstein, 1995). Furthermore, this integration process has been found to be modulated by the accessibility of the discourse referent as determined by its syntactic prominence and by the lexical features of a pronominal expression such as gender (Gordon, et al., 1993; Gordon & Hendrick, 1998; Yang, Gordon, Hendrick, & Wu, 1999a, 1999b). The present work consists of eight reading-time studies in Chinese to systematically examine further how these factors interact with one another in this integration process. The use of Chinese in the current study not only affords specific linguistic properties for the purpose of this investigation, but it also provides a valuable opportunity to address the issue of generalization: To examine whether the cognitive processes found in the comprehension of coreference in English can be generalized to other language systems as well. The results show that the accessibility of a discourse referent in a model of discourse is determined by the syntactic prominence of the linguistic expression that introduces such discourse entity in the preceding utterances. Such syntactically-derived accessibility for discourse referents is the primary contributor to the comprehension of coreference for pronominal expressions in the subsequent processing of discourse integration because it modulates the case of accessing discourse referents from a model of discourse for later retrieval. They provide support for the characterization that the initial steps in language comprehension are driven by syntactic and sequential factors that influence how linguistic expressions are mapped onto discourse representations (Gordon & Hendrick, 1998). The results also indicate that a pronominal expression is used as a critical link to guide the integration of the semantics of the predicate following the pronominal expression with the associated semantics of a preceding referent that matches its gender. Different types of pronominal expressions induce different levels of cognitive processes in this semantic integration. The Gordon and Hendrick model (1998), discourse prominence theory, is employed to provide an explanation of these results and thus further our understanding of the general cognitive mechanisms used in the comprehension of coreference during discourse integration. In all, these results shed light on our understanding of how discourse entities are processed and represented in a model of discourse during the dynamic process of integrating utterances incrementally into a model of discourse.
展开▼
