Knowledge closure spells out the intuition that agents can extend their knowledge by performing competent deductions from what they know. Alas, this principle leads to the puzzle of epistemic immodesty: If agents used deduction across the board, they could acquire knowledge that they do not seem able to acquire by those means. A traditional response is to reject knowledge closure. A less radical solution is to preserve knowledge closure but impose limits on the knowledge that can be acquired by deduction. Contextualism and contrastivism exemplify the compatibilist strategy. I shall defend a new form of compatibilism. My starting point will be the emerging consensus that competent deduction must figure in the antecedent of knowledge closure. I suggest that competent deduction requires the selection of premise-beliefs that are relevant to answering the question at hand, and I submit that this requirement is not fulfilled in alleged cases of epistemic immodesty.