Abstract

This paper compares two approaches to the semantics of modal expressions such as ‘might’. Both approaches define the conditions under which sentences of a language (not only modals) are acceptable relative to sets of possible worlds. Both approaches say that the sentence ‘Vivian might be in Vienna’ is acceptable relative to such a set if and only if ‘Vivian is in Vienna’ is true in some world in the set. One of these approaches treats the pertinent sets of worlds as information states of speakers. The other treats these sets as models of the statements that are objectively relevant in a conversation. This paper argues that the second of these approaches can better explain how statements made by means of modal sentences can be informative. It also suggests alterations to some common assumptions about the taxonomy of modals.