Implicature, Relevance and Default Pragmatic Inference

In his paper ‘Logic and conversation’ Grice (1989: 37) introduced a distinction between generalized and particularized conversational implicatures. His notion of a generalized conversational implicature (GCI) has been developed in two competing directions, by neo-Griceans such as Horn (1989) and Levinson (1983, 1987b, 1995, 2000) on the one hand, and relevance theorists such as Sperber & Wilson (1986) and Carston (1988, 1993, 1995, 1997, 1998a,b) on the other. Levinson defends the claim that GCIs are inferred on the basis of a set of default heuristics that are triggered by the presence of certain sorts of lexical items. These default inferences will be drawn unless something unusual in the context blocks them. Carston reconceives GCIs as contents that a speaker directly communicates, rather than as contents that are merely conversationally implicated. GCIs are treated as pragmatic developments of semantically underspecified logical forms. They are not the products of default inferences, since what is communicated depends heavily on the specific context, and not merely on the presence or absence of certain lexical items. We introduce two processing models, the Default Model and the Underspecified Model, that are inspired by these rival theoretical views. This paper describes an eye monitoring experiment that is intended to test the predictions of these two models. Our primary concern is to make a case for the claim that it is fruitful to apply an eye tracking methodology to an area of pragmatic research that has not previously been explored from a processing perspective.