Meaning and Use of Gradable Adjectives: Formal Modeling Meets Empirical Data

The meaning of gradable adjectives is highly contextdependent, and is notoriously difficult to capture precisely. Recent work in theoretical linguistics suggests that the way we use gradable adjectives can be explained in terms of optimal language use. To test this hypothesis we formulate a probabilistic speaker model that combines ideas from Bayesian approaches to pragmatic reasoning as social cognition with broader optimality considerations, as suggested by evolutionary linguistics. We demonstrate that, despite its simplicity, the model explains empirical data on the applicability of adjectives in context astonishingly well.