Philosophers are keenly aware that there are at least several (much-discussed) arguments that free will and moral responsibility are incompatible with determinism. But whereas the standard arguments for incompatibilism are well-known, arguments of the opposite sort -positive arguments for compatibilism -are correspondingly harder to come by. In general, it seems that the main way that philosophers defend compatibilism is simply by trying to show that the arguments for incompatibilism fail. And perhaps this makes sense. After all, one might think that there is, in general, some sort of presumption of compatibility; if someone asks for a reason to think that the fact that there is water on Earth is compatible with the fact that there is also water on Mars, the natural reaction is presumably to say: well, why shouldn't these things be compatible?Similarly, to pick two philosophical positions at random, if one were asked for some sort of argument that consequentialism in ethics is compatible with reliabilism in epistemology, presumably one would find it difficult to know what to say. Proving that two things are compatible sometimes can be nothing more than rebutting any argument that they aren't.otherwise compatibilism (what we shall here mean by "free will"), the key responses we develop can, we believe, be applied to epistemic arguments of other sorts.Call the following two premises "the data" underlying the argument:(1) We know we are free.(2) For all we know everything is predetermined.Of course, the data stands in need of interpretation -for instance, to whom does "we" refer? As a first approximation, however, the epistemic argument in question somehow moves from the data to a compatibilist conclusion. Now, there will be, of course, responses to this style of argument that outright deny what its proponents consider "the data". For instance, skeptics about freedom will simply deny (1). And certain libertarians will outright deny (2). In this paper, however, we set aside simplistic data-denying responses. Instead, we aim to show that compatibilism doesn't follow from a natural interpretation of the data in question.We proceed as follows. We begin by articulating Lewis' epistemic argument, highlighting its crucial features, and avoiding certain key interpretive pitfalls, while setting aside what we take to be weak responses to the argument. We then turn to our preferred assessment of the argument. In doing so, we will draw out a connection between the epistemic argument and parallel issues that have arisen in the physicalism/dualism debates. If a dualist concedes (as it seems they should) that for all they know an oracle might tell them that the world is merely physical, then, since they know they are conscious, this concession would seem to force them into abandoning their commitment to dualism. But we will show that, properly understood, this concession is no threat to the dualist. And, in a parallel fashion, we will argue that an incompatibilist needn't be threatened by the concession underwriting (2), viz., tha...