Presuppositions, probabilities and belief

Abstract

The proviso problem concerns presuppositions of complex sentences that seem stronger than predicted by satisfaction-type thoeries of projection such as Heim's (1983). We propose a probabilistic account, similar to Lassiters's (2012) for coordinated and conditional sentences, for the case of embedded attitude reports. For example, in (1), the satisfaction theory predicts that only (1a) is presupposed, but (1b) is usually inferred as well.<br /><br />(1) John believes that it stopped raining<br />a. John believes that it was raining.<br />b. It was raining.<br /><br />Our account includes a standard semantics for belief, but employs a probabilistic information state parameter whose value is dynamically updated.