|
Seeking to discover fundamental principles of learning in animals, we have specialized in behavioral studies of Pavlovian conditioning. Our early work focused on the issue of whether Pavlovian conditioned stimuli (CSs) lose their neutrality because (1) they provide information about impending biologically important unconditioned stimuli (USs) versus (2) occur close in time to those USs. This interest led us to ask whether conditioning might occur in truly random control procedures. There, CSs and USs are presented randomly and independently in time so that CSs, though occurring close to USs, can provide no reliable information about them. Because of this feature, truly random controls were proposed as procedures that would leave CSs neutral. We found, however, that they could make the CS excitatory and later showed that a crucial factor was the order in which chance CS-US pairings occurred relative to unpaired USs. |
|
Having found that conditioning can occur even when CSs provide no information about USs, we turned to models that stress temporal relations among events. We used "real time" models to test computer-simulated predictions. More recently, we have addressed the question of whether deficits in performance caused by blocking, overshadowing, or by CS preexposure (latent inhibition) are due to learning failures or to failures to express a well learned association. Our work has favored the learning-failure view. At present we are shifting our attention to clinical problems. We are using animal models of Pavlovian fear conditioning to search for optimal ways of eliminating human phobias.
|