Poster C137, Sunday, March 26, 5:00 – 7:00 pm, Pacific Concourse
Activation of Paired Associates Predicts Cue Revaluation in Causal Learning
Sean O'Bryan1, Evan Livesey2, Tyler Davis1; 1Texas Tech University, 2University of Sydney
Instances of retrospective revaluation in human causal learning suggest that the predictive ratings of unpresented stimuli may be updated via associative retrieval (AR). In this experiment, we tested the predictions of the AR hypothesis using fMRI and a multi-stage allergy prediction task that encourages cue revaluation. Participants were first trained to predict whether an allergic reaction would occur in a hypothetical patient exposed to different animal-food stimulus pairs (e.g., Cat + Strawberry = Allergic Reaction). In a second learning stage, animals and foods from the initial stage were presented individually, with outcomes that either agreed or conflicted with previous learning (Strawberry = No Reaction). Revaluation was then assessed during a test phase where participants rated the likelihood of a reaction to the unpresented but associated stimuli from phase two (Cat). To measure the reactivation of associated cues, independent localizer scans were collected to distinguish between multi-voxel patterns associated with each object class. We found substantial individual differences in behavior, with half of participants showing classic revaluation effects. Consistent with the AR hypothesis, neural similarity to the unpresent, associated feature class during phase two was predictive of the degree to which participants modified their causal ratings at test. These results are the first to provide direct neural evidence for the long-held theory that associative memory retrieval is responsible for retrospective revaluation.
Topic Area: THINKING: Reasoning