Daniel Rothschild

There is a longstanding debate in the literature about static versus dynamic approaches to meaning and conversation. A formal result due to van Benthem is often thought to be important for understanding what, conceptually speaking, is at issue in the debate. We introduce the concept of a conversation system, and we use it to clarify the import of van Benthem's result. We then distinguish two classes of conversation systems, corresponding to two concepts of staticness. The first class corresponds to a generalization of the class of systems that van Benthem's result concerns. These are the strongly static conversation systems. The second class is broader, and allows for a certain commonly recognized form of context sensitivity. These are the weakly static conversation systems. In the vein of van Benthem's result, we supply representation theorems which independently characterize these two varieties of conversation system. We observe that some canonically dynamic semantic systems correspond to weakly static conversation systems. We close by discussing some hazards that arise in trying to bring our formal results to bear on natural language phenomena, and on the debate about whether the compositional semantics for natural language should take a dynamic shape.

There are certain uses of and and or that cannot be explained by their normal meanings as truth-functional connectives, even with sophisticated pragmatic resources. These include examples such as The cops show up, and a fight will break out (‘If the cops show up, a fight will break out’), and I have no friends, or I would throw a party (‘I have no friends. If I did have friends, I would throw a party.’). We argue that these uses are indeed distinct from the more ordinary uses of and and or, but that they are nonetheless related in a principled way. To explain them we give an analysis of what we call the dynamic effects of connectives, which arise in all their uses. The special uses at issue are then argued to be instances where the connectives exhibit their dynamic effects without their truth-conditional meaning

Understanding the pattern by which complex sentences inherit the presuppositions of their parts (presupposition projection) has been a major topic in formal pragmatics since the 1970s. Heim’s classic paper “On the Projection Problem for Presuppositions” (1983) proposed a replacement of truth-conditional semantics with a dynamic semantics that treats meanings as instructions to update the common ground. Heim’s system predicts the basic pattern of presupposition projection quite accurately. The classic objection to this program (including other versions of dynamic semantics) is that the treatment of binary connectives is stipulative, and other, equally natural treatments fail to make the right predictions about presupposition projection. I give a variation on Heim’s system that is designed to escape this objection. I show that the most liberal possible version of this variant is equivalent to a strong-Kleene system in terms of its definedness conditions.