THE ROLE OF PROBABILITY OF REINFORCEMENT IN MODELS OF CHOICE

A general account of choice behavior in animals, the cumulative effect s model, has been proposed by Davis, Staddon, Machado, and Palmer (199 3). Its basic assumptions are that choice occurs in an all-or-none fas hion for the response alternative with the highest probability of rein forcement and that the probability of reinforcement for each response alternative is calculated from the entire history of training (total n umber of reinforced responses/total number of reinforced and nonreinfo rced responses). The model's reliance on probability of reinforcement as the fundamental variable controlling choice behavior subjects the c umulative effects model to the same criticisms as have been directed t oward other related models of choice, notably melioration theory. Seve ral different data sets show that the relative value of a response alt ernative is not predicted by the obtained probability of reinforcement associated with that alternative. Alternative approaches to choice th eory are considered.