During communication, the interpretation of utterances is sensitive to a listener's probabilistic prior beliefs. In this paper, we focus on the influence of prior beliefs on so‐called exhaustivity interpretations, whereby a sentence such as Mary came is understood to mean that only Mary came. Two theoretical origins for exhaustivity effects have been proposed in the previous literature. On the one hand are perspectives that view these inferences as the result of a purely pragmatic process (as in the classical Gricean view, (...) and more recent Bayesian approaches); on the other hand are proposals that treat them as the result of an encapsulated semantic mechanism (Chierchia, Fox & Spector 2012). We gain traction on adjudicating between these two approaches with new theoretical and experimental evidence, focusing on the behavior of different models for exhaustivity effects, all of which fit under the Rational Speech Act modeling framework (RSA, Frank & Goodman, 2012). Some (but not all!) of these models include an encapsulated semantic mechanism. Theoretically, we demonstrate that many RSA models predict not only exhaustivity, but also anti‐exhaustivity, whereby “Mary came” would convey that Mary and someone else came. We evaluate these models against data obtained in a new study which tested the effects of prior beliefs on both production and comprehension, improving on previous empirical work. We find that the models which have the best fit to human behavior include an encapsulated exhaustivity mechanism. We conclude that, on the one hand, in the division of labor between semantics and pragmatics, semantics plays a larger role than is often thought, but, on the other hand, the tradeoff between informativity and cost which characterizes all RSA models does play a central role for genuine pragmatic effects. (shrink)