Epistemology has always been concerned with mental states, especially doxastic states such as belief, suspension of judgment, and the like. A signifi cant part of epistemology is the attempt to evaluate, appraise, or criticize alternative procedures for the formation of belief and other doxastic at titudes. In addressing itself to doxastic states, epistemology has usually employed our everyday mental concepts and language. Occasionally it has tried to systematize or precise these mental categories, e.g., by introducing the notion of subjective probabilities. But this is a minor refinement in our in tuitive notion of degrees of confidence. What epistemology has not done?at least 20th century analytic epistemology hasn't done it?is to seek help from experimental psychology in choosing its doxastic categories. Analytic epistemology has generally assumed that doxastic descriptions generated by casual and introspective thought are adequate for its purposes. Now to the extent that epistemology simply tries to analyze everyday terms or concepts, refinement of our doxastic classifications may be irrele vant. If our only aim is to understand the ordinary use of the term 'know', for example, we need only be concerned with our ordinary mentalistic scheme. But if epistemology is to be a normative enterprise as well, it may have an ap propriate concern with cognitive categories and models that are not a stan dard part of everyday language. If epistemology is in the business of saying what psychological states a cognizer should be in in various circumstances, or what states it would be rational or intelligent for him to be in, we need as good a specification as possible of the range of psychological states open to him. It is questionable whether such a specification can be guaranteed by the conceptual apparatus and methods of ordinary thought about the mental. One might well suspect that systematic pursuit of cognitive psychology might yield a richer and more accurate set of doxastic descriptions than intuitive thought alone can do. At a minimum, it is worth asking what light can be shed on doxastic states by psychological perspectives on cognition, and what bearing this may have on normative or regulative epistemology. That is the aim of the present paper.1 A radical attack on doxastic-attitude concepts has been launched by Quine, who challenges the legitimacy of any propositional-content notions in science. This is a challenge I shall not take up here. Nor do I plan to consider other radical proposals, e.g., those of behaviorists, for dispensing entirely with doxastic-attitude concepts. Rather, I shall try to see how cognitive psy chology might help enrich, refine, or supplement doxastic classifications, and