Effects of a Random versus Pattern Instructional Set in a Contingent Partial Reinforcement Situation

The characteristic response curves of Ss operating under partial reinforcement are well-known. In a two-choice situation, for instance, they start off responding equally to both choices and then increase the proportion of their responses to the most profitable choice. Ordinarily, the final level of response is less than 100% preference for the alternative associated with the most frequent reward. Possibly, Ss try to match the occurrence of the reinforcement, assuming incorrectly that its schedule is patterned. Hence, they will vary their choices between the two alternatives in the expectation of the occurrence of reward at both of them. Flood (6) has hypothesized that the pure type of behavior (100% responding to the most profitable side) will be found in Ss who are convinced of randomness while the mixed type of behavior (trying to anticipate reward on the least profitable side as well) would be obtained when Ss believed there was a patterned arrangement in the occurrence of reinforcement. To test Flood's hypothesis we have varied the instructional sets given to Ss. One group was told that the reinforcements were being manipulated according to a pattern and it was their task to ascertain what this pattern was to get as many rewarded responses as possible. A second group was told that the reinforcements were being manipulated in a completely random fashion, and it was their task to get as many rewarded responses as they could. Accordingly, one would expect the second group to respond more often to the most profitable side, if Ss believed the instructions given to them.