Implicatures and Discourse Structure

International audienceOne of the characteristic marks of Gricean implicatures in general, and scalar implicatures in particular, examples of which are given in (1), is that they are the result of a defeasible inference. (1a) John had some of the cookies (1b)John had some of the cookies. In fact he had them all. (1a) invites the inference that John didn't have all the cookies,an inference that can be defeated by additional information, as in (1b). Scalar inferences like that in (1a) thus depend upon some sort of nonmonotonic reasoning over semantic contents. They share this characteristic of defeasiblility with inferences that result in the presence of discourse relations that link discourse segments together into a discourse structure for a coherent text or dialogue---call these inferences discourse or D inferences. I have studied these inferences about discourse structure, their effects on content and how they are computed in the theory known as Segmented Discourse Representation Theory or SDRT. In this paper I investigate how the tools used to infer discourse relations apply to what Griceans and others call scalar or quantity implicatures. The benefits of this investigation are three fold: at the theoretical level, we have a unified and relatively simple framework for computing defeasible inferences both of the quantity and discourse structure varieties; further, we can capture what ' s right about the intuitions of so called "localist" views about scalar implicatures; finally, this framework permits us to investigate how D-inferences and scalar inferences might interact, in particular how discourse structure might trigger scalar inferences, thus explaining the variability(Chemla, 2008) or even non-existence of embedded implicatures noted recently (e.g., Geurts and Pouscoulous, 2009), and their occasional noncancellability. The view of scalar inferences that emerges from this study is also rather different from the way both localists and Neo- Griceans conceive of them. Both localists and Neo-Griceans view implicatures as emerging from pragmatic reasoning processes that are strictly separated from the calculation of semantic values; where they differ is at what level the pragmatic implicatures are calculated. Localists take them to be calculated in parallel with semantic composition, whereas Neo-Griceans take them to have as input the complete semantic content of the assertion. My view is that scalar inferences depend on discourse structure and large view of semantic content in which semantics and pragmatics interact in a complex way to produce an interpretation of an utterance or a discourse