Neural Correlates of Learning of Priors in Perceptual Decision-Making

Poster Session D - Monday, April 15, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Stephanie Wert¹ (stephaniewert@g.ucla.edu), Julia Schorn¹, Jesse Rissman¹, Barbara Knowlton¹; ¹University of California, Los Angeles

In a Bayesian framework of perceptual decision-making, sensory evidence and prior expectations integrate during the decision-making process. Here, we examine the neural correlates of acquiring base-rate priors in a perceptual decision-making task. Behavioral and neural data were collected from 22 adult participants in a two-alternative forced-choice motion discrimination task. Participants were exposed to colored dot stimuli with varying levels of motion coherence (0%, 13%, 35%, or 100%). For each participant, one motion direction (left or right) was more frequent than the other. The majority of participants became sensitive to the prior across learning, reflected in a bias to choose the more frequently occurring direction on 0% coherence trials in which no diagnostic sensory information was present. Analysis of the 3T fMRI data revealed a robust time by motion coherence interaction in the left caudate nucleus. In the 100% coherence condition, this region’s activation decreased across runs, suggesting neural adaptation as the task became more predictable. In the 0% coherence condition, we observed the opposite effect, indicating that left caudate is more activated during these uncertain trials as the participant gains experience with the base-rate prior. This pattern may reflect implementation of the prior to judge 0% coherent trials. This approach is novel in its application of fMRI to study learning of base-rate priors through experience. Future directions include employing drift-diffusion model parameters to correlate behavioral data with neural activity, offering a more comprehensive understanding of the interplay between learned priors and evidence in decision-making.

Topic Area: THINKING: Decision making