Intuitive knowledge of linguistic co-reference.

Abstract

The research reported here is a systematic investigation of what competent, native speakers of English, native to contemporary syntactic theory, judge to be grammatically acceptable patterns of co-reference involving names and pronouns. Its central goal is the specification of syntactic factors that influence co-reference within and between sentences. The results show that naive subjects have consistent intuitions of grammaticality that agree with some principles of contemporary binding theory. The results also show that naive subjects diverge substantially from syntactic theorists in other judgments of grammaticality. In particular, subjects have strong intuitions that reflexives and pronouns are in complementary distribution, a fact that supports contemporary syntactic theory. Beyond that domain, subjects' judgments of co-reference in name-pronoun, name-name, and pronoun-name sequences are systematically influenced by syntactic structure in ways that are not consistent with syntactic theory. Co-reference in name-pronoun sequences is generally quite acceptable but becomes more acceptable as the syntactic prominence of the name increases. Co-reference in name-name sequences is only moderately acceptable and becomes less acceptable as the syntactic prominence of the first name increases. Co-reference in pronoun-name sequences is generally unacceptable and is only weakly influenced by the kinds of syntactic prominence that affect other relations of co-reference. We account for these results through the elaboration of a model of the process by which syntactic representations are mapped onto a representation of discourse capable of expressing generalizations about co-reference both intra-sententially and inter-sententially.