Theoretical Note: All·-or-none Learning and Intertrial Forgetting

Several alternative interpretations of all-or~none processes for paired-associate learning and concept formation are examined. These models, along with a simple incremental process, are compared to data obtained in three-response and four-response paired-associate learning experiments. The results strongly favor a two~process model that postulates a distinction between long-term and short-term retention and assumes an intertrial forgetting process. -lIn recent issues of this journal Estes (1960), Bower (1962), and Suppes and Ginsberg (1963) have examined a wide array of data on pairedassociate learning and concept formation in terms of an all-or-none process. The particular model they consider represents a special case of more general models of Stimulus Sampling Theory, and has been frequently labeled as the one-element pattern model. In a paired-associate experiment the single stimulus element represents a stimulus item from a list of paired-associates; in a concept formation experiment the stimulus element represents a concept, or some aspect of a concept. The two assumptions of the model are as follows: (1) Until the stimulus element is conditioned, there is a constant probability g that the SUbject will respond correctly; (2) On each trial there is a probability c that the single element will become conditioned to the correct response. Thus, on trial n of an experiment the stimulus element can be regarded as being in one of two conditioning states: In state C the element is conditioned to the correct response; in state C the element is unconditioned. The element starts out in state C and moves between these two states as specified by the transition matrix