Preliminary Investigations in the Effect of Continually Changing Reinforcement on Learning and Extinction

A great deal of research has been done in the gener­al area of manipulating reinforcement parameters. However, reinforcement per se has rarely been varied systematically from trial to trial. The research exceptions have been indicated. In these experiments , the nutritive value, the delay, or the units of reinforcement have been systematical­ly varied and the results are not always conclusive as re­gards the experimental variable itself. The present study attempts to control for these factors and yet vary systema­tically properties of the reinforcing agent alone. Pigeons were the experimental subjects and they worked for food while they were at a controlled drive level. Since the caloric value in the food reward was kept constant from trial to trial, a Hullian prediction would be that the groups would show no difference in either conditioning or extinc­tion as all groups would be drive-reduced similarly (Hull, 1943). On the basis of most of the stimulus variation ex­periments it is possible to predict that the more the cue change in reinforcement, the slower the rate of acquisition and the lower the subjects' performance level. It is pos­sible to predict further that the groups with more ex­tensive reinforcement variation should resist extinction longer much the same as is shown in the stimulus change and also in the partially-delayed reinforcement experiments. In other words, they have been conditioned to changing cues and extinction can be viewed as another cue change. There is of course a third possibility. The more reinforcement variation, the higher the level of performance due to a sort of "grab­ bag" motivating effect. This third alternative would fit in with the amount or "bits" research and the novel stimu­lation research. The more stimulus change groups should be similar to getting more units for their efforts in terms of more cue-change. Specific directional predictions were not made although on the basis of previous pilot study work (see Chapter II) a trend could be predicted.