Soft Presuppositions as Scalar Implicatures in Signaling Games

In this paper, I will discuss why soft presuppositions behave differently from hard presuppositions: the former are easily defeasible and project nonuniformly in quantificational sentences. I assume that soft triggers should be associated with alternatives, and thus share many similarities with scalar implicatures. As a generalization of Parikh’s games of partial information, I develop game-theoretic models, which provide a unified account for both scalar implicatures and soft presuppositions. I argue that iterated best response (IBR) reasoning allows us to analyze the behaviors of soft presuppositions in accordance with rational inferences. The models yield the following predictions of soft presuppositions: projection happens unless it is common knowledge that the speaker is ignorant about it, in which case the presupposition is defeasible; projection depends on the type of quantifiers, which may lead to nonuniform behaviors.