A model for psychophysical learning is constructed by imposing some conditioning principles on concepts derived from the theory of signal detectability. The effects of a priori probability, feedback, and practice are derived in part by Monte Carlo simulation and in part by analysis. The theory makes some novel predictions for the effects of these variables, all of which find support in the literature. Some theoretical results are: (a) performance improves with practice; (b) feedback can be detrimental to performance in a psycho-physical task; (c) when the a priori probabilities of the stimuli are unequal and feedback is provided, the response criterion moves in the direction of optimality, but does not move far enough to reach optimality; however, when no feedback is provided, the criterion moves in the opposite direction.