Epistemic possibilities and the sources of belie

I develop Stalnaker's ideas of a causal/pragmatic account of belief, refitting them to function in Situation Theory. Building on the semantics of perception reports, I make the case for the idealizing assumption of the veridicality of belief. I introduce as a requirement on a belief state a soundness condition adapted from information theory. The problem of error leads me to develop epistemic support as part of a revised soundness condition.;I argue that a psychologically adequate model of belief will include epistemic possibilities. This gives us a richer account of belief---one that will skirt the problem of logical omniscience and have interesting things to say about whether or not we know our own minds---and free us to put forward a naive semantics for belief reports---one that can remain faithful to basic principles of semantic interpretation and at the same time step out of the way of the classic problem examples.;Epistemic possibilities are the alternative scenarios against which agents understand claims about their world and out of which they develop their beliefs. A situation becomes epistemically possible as the result of the agent's cognitive action, rather than by an act of the theorist's logical deduction. So the requirement on qualifying as an epistemic possibility here is stricter than is commonly the case. It is, however, at the same time looser, for an impossible situation may have a life as an epistemic possibility.;The conditions on having a particular belief are very strong, requiring actual causal interaction with a source situation and genuine epistemic support by that situation of the infon indicated. The conditions on satisfying a belief report are not so severe. The agent need not be prepared to endorse the report in anything like our terms in order to count as having the belief we report. I argue that we ought to avoid abandoning the principles of Direct Reference, Semantic Innocence and, especially, Compositionality. This leaves us with a naive semantics that relies on a pragmatic account of failures of substitutivity.