Little computational modeling work has analyzed the ability of Pavlovian learning models to simulate the effect of pairing a conditioned inhibitor with an outcome. In particular, models utilizing total error reduction (TER) learning rules predict that previously inhibitory stimuli will gain excitatory properties faster than neutral stimuli, contradicting some empirical evidence (Rescorla, 1969). In contrast, models of learning based on within-compound associations appear to make the opposite prediction. To test these predictions, we will compare the rate of learning about a neutral stimulus with the rate of learning about an inhibitor in an associative learning task (Larkin, Aitkin, and Dickinson, 1998). In addition, a second experiment will utilize a compound test procedure (Rescorla, 2000) designed to control for differences in the initial level of responding controlled by the stimuli. The fit of within-compound and TER models to the results will be compared.
|Presenter:||Ryan Hutchings (Undergraduate Student)|
|Location:||Lobby of Edwards|
|Time:||1:15 pm (Session III)
Please note that presentation times are approximate. If you are interested in attending sessions with multiple presentations, please be in the room at the start of the session.