A Theoretical Examination of an Experiment on Differential Resistance to Extinction

Abstract

McClelland (4) points out that a major paradox in learning theory is the fact that laboratory learning shows none of the persistance which we must assume characterizes human learning, especially learned motivation. McClelland refers to the fact that avoidance learning is more difficult to extinguish than other learning, and he assumes that the reason for this is that the organism has difficulty in discriminating extinction conditions from acquisition conditions. Generalizing from what is known about avoidance learning, McClelland predicts (4) that the more disorderly and confused the original conditions of learning are, the harder it will be to set up conditions which are sufficiently different from them for the organism to perceive the difference and unlearn a response which is no longer appropriate. McClelland elaborates by saying that very few laboratory experiments are sufficiently messy or disorderly to make the discrimination between learning and extinction difficult for the organism. Usually there is one cue which is always relevant, one response which is always appropriate, and one particular set of time relations between events and the cueresponse reward sequence. However, in natural learning situations this sequence is seldom regular. Sometimes a response is rewarded, sometimes it is ignored, and sometimes it is punished. In fact, in natural situations learning often occurs under such irregular, changing, and inconsistent conditions that the experimenter who is absorbed in his consistent cueresponse-reward sequences might well wonder how learning ever occurs under such conditions. But, as McClelland says, responses are learned under such conditions, and when such learning takes place, it should be very resistant to change, because it is so general in the first place, so compounded of different cues, responses, rewards and punishments that the organism will have difficulty in discovering that conditions have changed, and that some general expectation which he has built up is no longer being confirmed. Although McClelland's suggestion is that motives acquired under irregular conditions of acquisition should be resistant to change, it seems reasonable to assume that his suggestion should apply equally to instru-