The Effect of Partial and Delayed Reinforcement on Resistance to Extinction

Abstract

It is now a well-established fact that partial or intermittent reinforcement produces greater resistance to extinction than does consistent reinforcement.1 Although this finding may seem to present a considerable challenge to any theory of learning based upon the assumption that a responsetendency is inevitably strengthened by need-reduction and weakened by the lack of it, a quite reasonable solution to the problem has been advanced by reinforcement theorists. The solution hinges upon the concept of stimulus-generalization. For consistently reinforced animals, the argument runs, afferent components arising from the lingering after-effects of reinforcement administered on previous trials (such as food-particles in the mouth) are part of the afferent pattern present at the start of each training trial. These after-effects are absent during the extinction series, and the relatively rapid rate of extinction found in consistently reinforced animals must be attributed, in part, to an alteration in the afferent pattern. Partially reinforced animals, on the other hand, are frequently rewarded during training for responses to an afferent pattern which does not include the after-effects of reinforcement and which more closely resembles, therefore, the stimulating conditions encountered during extinction. It follows that these animals should extinguish less rapidly. This interpretation is to some extent supported in a recent experiment by Sheffield in which it was demonstrated that partially and consistently reinforced animals extinguish at the same rate following spaced training.2