Poster F35, Tuesday, March 28, 8:00 – 10:00 am, Pacific Concourse
A role for the striatum in feedback contingency estimation during perceptual category learning
Lauren E. Vucovich1, F. Gregory Ashby1; 1University of California, Santa Barbara
Recent research suggests that the effects of reward prediction error (RPE) on dopamine neuron firing may be modulated by feedback contingency – that is, by the correlation between response confidence and feedback valence. For example, when feedback is not contingent on behavior there is nothing to learn, and dopamine fluctuations carry no adaptive value. This degradation of feedback contingency has been shown to dramatically impair learning in a perceptual category-learning task (Ashby & Vucovich, 2016). Recent evidence implicates the striatum and prefrontal cortex as two potential mechanisms that make up a network responsible for contingency detection and computation. The current study indirectly tested the involvement of the striatum in contingency estimation by employing a feedback delay during a rule-based category learning task under both high- and low-contingency conditions. Due to the nature of synaptic plasticity in the striatum, delaying feedback by more than a few seconds should eliminate any striatal contribution to contingency estimation. In both feedback conditions, the optimal Bayesian classifier would receive positive feedback on 80% of the trials. In high-contingency conditions, the probability of positive feedback decreased with categorization difficulty, whereas in low-contingency conditions, the probability of positive feedback was independent of categorization difficulty. Overall, results showed a surprising interaction between contingency and feedback delay, with learning improving when contingency was low and deteriorating when contingency was high. These results suggest the striatum plays an important role in contingency estimation and they provide significant insight into the nature of that role.
Topic Area: EXECUTIVE PROCESSES: Other