Choosing the Future: Location Reinforcer Generalization across Spatial Dimension

Abstract

A growing body of evidence shows that both stimulus control and reinforcer control depend primarily on a discriminative function. Reinforcer generalization may contribute to discriminative control by reinforcer and non reinforcer stimuli. Such a generalization may cause the discriminated reinforcer differential to differ from the actual differential, resulting in imperfect control by the contingency. Reinforcers are mu lti dimensional, and thus generalization may occur not only across the t ime at which a reinforcer is obtained but a lso across responses that may produce reinforcers. The quantitative modeling approach suggests that the generalization of reinforcers across responses may take a similar form to generalization across time, but the response dimension has received relatively little attention. To assess reinforcer generalization across responses in an environment where an organism discriminates the contingency only by responses and obtained reinforcers, we conducted two experiments wi th humans and pigeons separately. In Experiment 1, human participants touched locations along a horizontal bar . T ouches to the left and right areas could produce reinforcers, but touches to the central area could not. Across components in a session, the reinforcer ratio between two areas varied Participants responded in the area that never produced reinforcers A ge neralization model that redistributed obtained reinforcers across location according to a normal distribution described the response distribution of participants who sho we d some degree of control by the location of reinforcers. The similarity between this model and models previously describe d generalization across time suggests that generalization across location and time may conform to the same rule. In Experiment 2, pigeon subjects could respond on any of three concurrently available keys. All three keys 4 were lit the same color, but only the left and right keys ever produced reinforcers. Across conditions, the left/right reinforcer ratio varied. Experiment 2 showed higher overall sensitivity and fewer unreinforced responses than Experiment 1, suggesting ge neralization was reduced under such highly discriminable contingencies. Taken together, Experiment 1 and 2 suggest that a generalization based approach may provide a viable explanation for the occur re nce of response s that are never reinforced. Thus, increa sing contingency discriminability and hence reducing generalization may be an important consideration for reinforcer based studies.