Anne Bezuidenhout
Stephen Levinson has recently been developing a theory of generalized conversational implicatures (GCIs). He proposes to treat GCIs as the output of default pragmatic inferences that may be cancelled by specific assumptions, but otherwise go through. He has identified a set of default heuristics that are used to generate GCIs. These default heuristics yield interpretations that represent a level of meaning that he calls utterance-type meaning, which is intermediate between sentence-type meaning and speaker meaning. In this paper I argue against the idea of a set of default inference rules that are attached to certain utterance-types. An account of utterance interpretation that appeals to cognitive strategies that are independent of any particular utterance type is to be preferred. I suggest that the mechanisms of language understanding used in the recovery of the interpretations in question are more flexible and more context-dependent than allowed for by Levinson.