I explore the traditional contention that a priori epistemic warrants enjoy some sort of higher epistemic security than empirical warrants. By focusing on warrants that might plausibly be called 'basic', and by availing myself of an original taxonomy of epistemic defeaters, I defend a claim in the vicinity of the traditional contention. By discussing some examples, I argue that basic a priori warrants are immune to some sort of empirical defeaters, which I describe in detail. An important by-product of my investigation is a novel theory of epistemic defeaters, according to which only agents able to engage in higher-order epistemic thinking can suffer undermining defeat, while wholly unreflective agents can, in principle, suffer overriding defeat.