Description:

There is a growing body of psycholinguistics work arguing how comprehenders are capable of integrating interlocutor-based pragmatic reasoning into sentence comprehension. Under the traditional Gricean approach and the more recently formalized bayesian version of it, a listener's interpretation of a sentence is the outcome of a rational reasoning process that takes into account a speaker's intentions and production choices in a particular communicative context. Acknowledging the success of this approach,  in this talk, I want to revisit the question of how much comprehension should be explained by a speaker-dependent pragmatic reasoning process, and how much should be handled by a listener's independently acquired knowledge of semantic conventions. The case studies I will discuss involve the interpretation of gradable adjectives. Gradable adjectives conveniently lend themselves to the current investigation, because their interpretations are highly context dependent. Gradable adjectives denote properties that are relativized to contextual thresholds of application, e.g. how long an object must be in order to count as long in a context of utterance depends on what the threshold is in that context. But thresholds are variable across contexts and adjectives, and are in general uncertain. In the first part of this talk, using both human judgment data and computational modeling results, I will explore whether Bayesian pragmatics can completely replace the need to postulate lexical semantic conventions for contextually-sensitive thresholds. Our results showed that the best model performance is achieved when Bayesian pragmatic reasoning is supplemented with the traditionally assumed semantic conventions for thresholds. In the second part of this talk, using the eye-tracking data from a visual-world paradigm, I will demonstrate that the well-known referential contrast effect in the literature (Sedivy et al., 1999), which standardly receives a Gricean reasoning based explanation, may actually be better explained with a semantic-threshold-based explanation. Both case studies raise questions whether the field has overextended the role of pragmatic reasoning in sentence comprehension.