The concept of epistemic defeat, or some surrogate for it, is essential for any fallibilistic epistemology. If knowledge requires infallibility, then the epistemic grounds of belief have to be strong enough that no further information could be made available to the cognizer to undermine these grounds of belief. When knowledge requires no such infallibility, however, grounds of belief can be undermined by further information, information that defeats the power of the original information to put one in a position to know that the claim in question is true. Even if some combinations of conditions for knowledge are sufficient for truth, if there is a nonpsychological condition for knowledge that is not sufficient for truth, that condition will need to appeal to some concept of defeat (or a surrogate of it).

I mention here the notion of a surrogate for the concept of defeat only to ignore it in what follows, for the following reason. Reliabilists, such as Alvin Goldman, recognize that a belief can be produced by a reliable mechanism, without putting one in a position to know. For example, one may form a perceptual belief in circumstances that one has good reason to believe are deceptive. This further information defeats the confirming power of the perceptual experience. Since reliabilists wish to construe talk of reasons and confirmation in terms of reliable processes and methods, they cannot be satisfied simply to note that these reasons defeat the confirming power of one's perceptual experience.