Scalar diversity: relevance and prior probability

Successful comprehension involves much more than access to word meaning and syntactic structure. It essentially involves making inferences about speakers’ intended meaning. This raises the question of what principles and knowledge are involved in listeners’ inferential processes. This project approaches this broad question by investigating a specific type of pragmatic inference: Scalar Implicatures (SIs).

Scalar terms like some or possible often give rise to SIs like ‘some and not all’ or ‘possible and not certain’. Recent studies show that different scalar terms give rise to SIs at different rates (known as scalar diversity). The inferential process of SIs, namely whether it is Gricean reasoning at sentential level or local enrichment at sub-sentential level, has been intensively investigated. Yet, it is puzzling why applying the same inferential process (global or local) to these scalar terms yields different implicature rates.

This project argues that SI is likely to be the result of multiple mechanisms. The primary aim of this project is to demonstrate a role for different mechanisms in explain scalar diversity and to model the effect of different mechanisms on implicature rates using a Bayesian, Rational Speech Act framework. In the literature, the most widely assumed mechanism to explain SI is to exhaustify with respect to alternatives. For an inference to be drawn, the stronger alternative of the scalar term must be relevant in the context. Thus, this project further aims to investigate the effects of relevance on SI derivation and to account for scalar diversity using an extended RSA framework with the relevance prior.