My general goal in this paper is to explore possible relations between a cognitive analysis of the notion of reference on the one hand, and computational models of speech acts (illo-cutionary acts, to be precise) on the other. On the cognitive side, we have the following problem: how can thoughts (and sentences that articulate them) be about objects? The deceptive simplicity of the question masks a century of lively philosophical debate but it is not the debate that I wish to focus on here. Instead, I will assume a framework of what I have called the deacriptive approach to the problem of reference I. The most important single principle of this framework is that the descriptive content of an individuating rep-resentation denoting an object is both necessary and sufficient for a belief (or any other propositional attitude) to be about that object. I have argued for this principle elsewhere and it will simply be assumed here.