Robert A. Rescorla - US grants

Dr. Robert A. Rescorla, of the University of Pennsylvania, and Dr. Ruth M. Colwill, of Brown University, are collaborating in research on the learned basis of elementary goal-directed behavior. A large proportion of learned behavior of humans and other animals involves such goal-direction, and its understanding is central to many psychological processes. Drs. Rescorla and Colwill are using a widely-accepted model behavioral system in which laboratory rats receive positive outcomes when they engage in designated behaviors (responses) in the presence of signals. Such an arrangement can potentially result in the learning of a variety of relations among signal, response, and outcome. The overall goal of the project is to explore the nature of the associative relations that are actually learned. Four types of associative relations have been indicated by the results of previous research by Drs. Rescorla and Colwill, and will be the special focus of this project: (a) an association between the response and its outcome; (b) an association between the signal and that outcome; (c) an association between the signal and the response; and finally (d) a hierarchical association between the signal and the response- outcome relation. The theoretical analysis to be conducted draws heavily on our currently well-developed understanding of a related learning process, Pavlovian conditioning. In the past twenty years, our understanding of Pavlovian conditioning has advanced and changed dramatically, partly as a result of previous NSF-supported research in Dr. Rescorla's laboratory. It is natural and timely to attempt an extension of our understanding of Pavlovian associations to the more complex and intrinsically more important case of goal-directed behaviors. The prospects for significant advancement are substantial and exciting.

The goal of the proposed research is to expand our knowledge of the elementary learning processes of Pavlovian conditioning and instrumental training. These learning processes are fundamental not only to much of animal behavior but also to human normal and pathological functioning. Instrumental learning, which involves learning about the consequences of one's actions, is commonly conceptualized as having three component elements: a behavioral response (R), an outcome (O) of that response, and a stimulus (S) in the presence of which the response produces that outcome. Prior work from this and other laboratories has identified the development of associations among each of the pairs of these elements as contributing to behavior. Thus, organisms learn associations of the form R- O, S-O, and S-R, and each contributes to behavior. But they also acquire more hierarchical structures of the form S-(R- O), in which a stimulus activates an association between R and O. The development of the techniques of transfer and devaluation has allowed the separate measurement of these associations. The devaluation technique involves the changing of the value of an outcome after it has been earned by a response. Such devaluation of the outcome produces substantial and immediate depression of the value of both stimuli and responses that have led to that outcome, thus allowing measurement of both R-O and S-O associations. The transfer technique involves the ability of a stimulus trained with one behavior to transfer to the control of another behavior based on their sharing a common outcome. It therefore allows one to detect both the S-O and R-O association. Recent evidence suggests that these associations show amazing persistence even when they no longer accurately reflect the consequences of responding, as in extinction. The present research project will use the devaluation and transfer techniques, together with others to be developed, to measure the persistence of those a ssociations. It will explore the possibility that associations with goals persist but are suppressed by superimposed inhibitory associations not involving the outcomes. This research will also explore the fate of hierarchical associations. Similar procedures will also be applied to Pavlovian conditioning, which involves the learning of relations between two stimuli in the world. New results suggest that there, too, previously learned associations persist when conditions change. Moreover, there too hierarchical modulatory structures seem to contribute to performance. These studies will influence our thinking about elementary associative learning processes. They will particularly illuminate the process of extinction, a process important in many implications for understanding learning at the neural level.