According to the main tradition in epistemology, knowledge is a variety of justified true belief, justification is an undefinable normative concept, and epistemic principles (principles about what justifies what) are necessary truths. According to the leading contemporary rival of the tradition, justification may be defined or explained in terms of reliability, thus permitting one to say that knowledge is reliable true belief and that epistemic principles are contingent. My aim here is to show that either of these approaches will yield a solution to the problem of induction. In particular, either of them makes it possible to ascertain the reliability of induction through induction itself. Such a procedure is usually dismissed as circular, but I shall argue that it cannot be so dismissed if either approach is correct.

As one would expect, the solution based on the traditional approach differs from the one based on the reliabilist alternative, but they have important features in common. The common elements are presented in sections I, II, and III, the elements specific to the reliabilist approach in sections IV, V, and VI, and those specific to the traditional approach in section VII.