How should we understand validity? A standard way to characterize validity is in terms of the preservation of truth (or truth in a model). But there are several problems facing such characterizations. An alternative approach is to characterize validity epistemically, for instance in terms of the preservation of an epistemic status. In this paper, I raise a problem for such views. First, I argue that if the relevant epistemic status is factive, such as being in a position to know or having conclusive evidence for, then the account runs into trouble if we endorse certain familiar logical principles. Second, I argue that if the relevant epistemic status is non-factive, such as is rationally committed to or has justification for believing, then a similar problem arises if we endorse the logical principles as well as a sufficiently strong epistemic "level-bridging" principle. Finally, I argue that an analogous problem arises for the most natural characterization of validity in terms of rational credence.* Forthcoming in Philosophical Studies. 1 On some views, including those inspired by Harman's (1986) emphasis on the distinction between inference and implication, validity concerns an abstract relation among propositions, not any kind of reasoning. So it is a mistake to think of inferences as valid or invalid. If that is right, then the issues raised here should be put in terms of whether a purported entailment is valid or invalid. This would only require minor changes in the discussion to follow.