An urn model for instrumental conditioning

Instrumental conditioning phenomena are fundamental to understanding human and animal learning. The learning process in conditioning phenomena is identified with updating of the action probabilities of the subjects. To account for the learning process, mathematical learning theory simply assumes certain closed form expressions for the updating of the action probabilities. However, there are no biological, psychological and physical grounds for these assumptions. In this paper we reexamine a classical paradigm of instrumental conditioning from a new perspective, and thereby abstract an urn model, which may provide a physical interpretation for the learning process. The concepts such as ''motivation'', ''reward'', ''punishment'', etc. which previous theories might utilize in order to make themselves appear to be plausible, are not needed to account for the learning process.