When vague sentences inform: A model of assertability

A speaker often decides whether or not to say something based on his assessment of the impact it would have on his hearer's beliefs. If he thinks it would bring them more in line with the truth, he says it; otherwise he does not. In this paper, I develop a model of these judgments, focusing specifically on those of vague sentences. Under the simplifying assumption that an utterance only conveys a speaker's applicability judgments, I present a Bayesian model of an utterance's impact on a hearer's beliefs. From this model I derive a model of a speaker's judgment of whether or not an utterance would be informative. I illustrate it with several examples of judgments of vague and non-vague sentences. For instance, I show that it models the common judgment that asserting either "George is tall" or "George is not tall" would be misleading if George were borderline tall, but asserting "George is tall and he isn't tall" would not be.