Abstract

Polysemy, understood as the phenomenon of a single linguistic expression having multiple related senses, is not a homogenous phenomenon. There are regular (apparently, rule-based) cases and irregular (resemblance-based) cases, which have different processing profiles, and there are several kinds of relations that may hold between the senses making up a polysemy family: narrowing/broadening, metaphor, metonymy. Focusing on the irregular cases, I consider whether polysemy is best treated as a semantic or a pragmatic phenomenon, that is, a set of stable conventionalised senses or a matter of occasion-specific context-sensitive inference. Drawing on relevance theory, I maintain that many instances of well-established (hence, arguably, semantic) polysemy have their origins in online pragmatic processes of ad hoc concept construction, including cases where a new word (not just a new sense) is created, based on an existing word, and giving rise to cross-categorial polysemy.

Finally, the question of what linguistic constraints there are on these apparently very flexible pragmatic processes of sense creation is considered. Here I attempt to situate polysemy in a broad picture of the human language faculty, one that respects the distinction between the narrow linguistic faculty (syntax and its interfaces) and the wider, more unruly ‘systems’ of sense and usage conventions that grow up in a language community.