Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz Sessions
Characterizing Novelty-evoked Prediction Errors across the Mesolimbic System
Poster Session C - Sunday, March 30, 2025, 5:00 – 7:00 pm EDT, Back Bay Ballroom/Republic Ballroom
Yifang Liu1 (yifangl@uoregon.edu), Ian C. Ballard2, J. Benjamin Hutchinson1, Vishnu P. Murty1; 1University of Oregon, 2University of California, Riverside
Novelty signals a violation of predictions and a need to update our models of the world. These novelty-evoked prediction errors play a significant role in shaping learning and memory, yet may reflect different learning signals across brain regions underlying motivation. While some regions prioritize rapid adaptation to unexpected events, others focus on gradually integrating novel information into stable memory networks, a process that can be characterized as learning rates in reinforcement learning models. We hypothesize that mnemonic structures like the anterior hippocampus exhibit slower learning rates compared to non-mnemonic structures such as the VTA and nucleus accumbens. To address this, we will use the Natural Scenes Dataset, which provides high-resolution fMRI data from 8 participants exposed to thousands of novel and familiar natural scene images. We will fit a reinforcement learning model with neuroimaging data from novelty-evoked prediction error responses to estimate learning rates across the anterior hippocampus, VTA, and nucleus accumbens during novelty processing. We will then compare these learning rates across regions using a multi-level modeling approach to capture region-specific dynamics in responding to novelty. These findings will provide insights into how different brain regions contribute to novelty-driven learning and will refine our understanding of how the brain balances adaptability and memory stability.
Topic Area: LONG-TERM MEMORY: Episodic