Empirical evidence on scalar implicature processing at the behavioural and neural levels

The Default hypothesis on implicature processing suggests that a rapid, automatic mechanism is used to process utterances such as “some of his family are attending the wedding” to infer that “not all of them are attending”, an inference subject to cancellation if additional contextual information is provided (e.g. “actually, they are all attending”). In contrast, the Relevance hypothesis suggests that only context-dependent inferences are computed and this process is cognitively effortful. This article reviews findings on behavioural and neural processing of scalar implicatures to clarify the cognitive effort involved.