Metonymy and Anaphora

Abstract. The starting point for this paper is the observation that a substantial majority of 'indirect' examples of anaphora in a corpus of naturally occurring discourse involve relations which would independently be categorised as 'metonymica'. Data presented from this corpus indicate that such examples occur quite regularly and tend to be overwhelmingly of a few single types; as indicated also in the psycholinguistic literature, they apparently present no special problems of interpretation. Furthermore, not just definite NPs are involved in such relations, but also pronouns. Considerations of the implications of this observation involve constructing a model of possible types of metonymical anaphora and elaborating on a paradigm of constructed examples provided by Fauconnier to explore some of the constraints on such examples. In particular, it is found that considerations of animacy (or perhaps discourse status) affect the acceptability judgments of these examples.