Abstract

A model of metaphor processing is suggested based on the application of pragmatic principles to the type of semantic information easy to access. It is argued that, with metaphor, higher-order categorical knowledge is given processing preference over instance-specific knowledge in an attempt to recover likely intended meaning. Instance-specific information is used more often when the higher-order knowledge is taken to violate conversational postulates. One such violation occurs when the categories implicated by metaphor topic and vehicle are similar, and thus unlikely to provide new or relevant information. It is argued further that these differences could, in part, explain at least one condition that produces the asymmetry observed in metaphor when topic and vehicle are reversed. Predictions supportive of the model were obtained in three studies, employing different methodologies: feature listing, recognition memory and a vehicle choice task.