Schedule of Events | Symposia

Distinct fMRI Pattern Effects of Error- and Uncertainty-Based Event Model Updating

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Tan Nguyen¹ (n.tan@wustl.edu), Jo Etzel¹, Matthew Bezdek², Jeffrey Zacks¹; ¹Washington University in St. Louis, ²Elder Research

The human mind segments continuous sensory input into discrete events. Our study investigated two causal signals for segmentation and the neural mechanisms underlying these processes. Using a Finite Impulse Response (FIR) model to analyze BOLD signals, we observed activity changes around human event boundaries: Areas in the visual, dorsal, and control networks showed an increase in BOLD, while areas in the default network showed a decrease in BOLD around event boundaries. Using a FIR model to analyze successive pattern correlation, we observed successive pattern correlation changes around human boundaries across multiple brain regions belonging to control, default, dorsal attention, and visual networks. To understand the distinct mechanisms involved, we analyzed two types of boundaries derived from computational models: error-based and uncertainty-based. Both error-based and uncertainty-based boundaries uniquely predicted human event boundaries. FIR analysis revealed dissociable neural correlates for these boundary types. Error-based boundaries were strongly associated with successive pattern correlation changes in the ventrolateral prefrontal cortex, while uncertainty-based boundaries were strongly associated with pattern correlation changes in the post central gyrus, dorsal attention network, mid cingulate cortex, and visual networks. Networks associated with both boundary types included medial prefrontal cortex and temporal regions within the default network, which overlapped with regions showing pattern correlation changes during human boundaries. The distinct networks associated with error-based and uncertainty-based boundaries suggest that multiple cognitive mechanisms are involved with event segmentation. Together, our behavioral and neural findings suggest that humans utilize both error and uncertainty signals when segmenting continuous experience into discrete events.

Topic Area: PERCEPTION & ACTION: Vision