The contemporary consensus in analytic philosophy concerning cultural relativism is: it is impossible to formulate relativism coherently, diversity does not provide good reason for accepting relativist conclusions anyway, and if relativism is false, or incoherent, then cross-cultural disagreement, however intractable, raises no important epistemological challenge. I challenge every aspect of this consensus in the light of contemporary theories of reference and knowledge, focussing on various traditional cultures' supernatural explanations of illness. ;I defend the coherence of relativism against standard objections, arguing (...) that: it is possible to make sense of the claim that a sentence has different truth-conditions in different frameworks, while, in significant respects, retaining the same meaning; cultural relativism does not rule out the possibility of rational discussion between distinct cultural traditions; even self-referential relativist theses are neither self-refuting nor self-undermining. I suggest a distinctive way of arguing from cross-cultural differences to relativism, which is plausible, though inconclusive. ;I then argue that the alternative to relativism about rationally irresolvable cross-cultural disagreement is skepticism. Far from undermining the anthropological challenge to the objective justification of our beliefs--as might be thought--naturalistic epistemology and semantics themselves provide reasons to appreciate the force of this challenge, though they require us to rethink its nature. Reliability of belief production or regulation is not sufficient to provide knowledge, or even justification, in cases in which rival methods of belief assessment pose salient alternatives and there is no "neutral" reason for supposing one method to be more reliable. If, as I argue, neutral reasons for rejecting traditional methods of witchcraft detection are not available, then we do not know that witches do not exist. The resulting skepticism, however, would be a fairly circumscribed one. Most scientific beliefs we ordinarily take to be well confirmed--including beliefs about unobservable entities--remain so on the standards I defend. What proves less epistemically secure are the deeper metaphysical commitment to materialism and naturalism associated with Western science, commitments underlying our rejection of supernatural explanations of illness. (shrink)