Proceedings of the Annual Meeting of the Cognitive Science Society

Contrastive inferences, whereby a listener pragmatically infersa speaker’s referential intention of a partial referring expres-sion like the yellow by reasoning about other objects in thecontext, are notoriously unstable. We report a production-centric model of interpretation couched within the RationalSpeech Act framework. Adjective production probabilities alistener expects for objects in a context drive the size of con-trastive inferences: the greater the asymmetry in expectationfor a speaker to use a pre-nominal adjective for the target ratherthan for competitors, the greater the listener’s resulting targetpreference. Modifier production probabilities were collected(Exp. 1) and used to make predictions about comprehensionin an incremental decision task (Exp. 2). The model’s inter-pretation predictions are supported by the data. This accounthas the potential to explain the fluctuating appearance of con-trastive inferences and shifts the explanatory focus away fromcontrastive inference towards online interpretation of referringexpressions more broadly.