Schedule of Events | Symposia

Cognitive control networks direct brain flows underlying rapid instructed learning of stimulus-response tasks

Poster Session A - Saturday, March 29, 2025, 3:00 – 5:00 pm EDT, Back Bay Ballroom/Republic Ballroom

Arun Aryal¹ (arun.aryal@rutgers.edu), Ravi D. Mill¹, Alexandros Tzalavras¹, Michael Cole¹, Nachshon Meiran², Inbar Amir²; ¹Rutgers University-Newark, ²Ben-Gurion University of the Negev

Rapid Instructed Task Learning (RITL) enables humans to quickly learn and execute new tasks based solely on instructions, demonstrating unmatched cognitive flexibility. While RITL is well-established, the neural mechanisms supporting immediate task execution without prior practice remain unclear. To address this, we combined behavioral testing, functional magnetic resonance imaging (fMRI), and computational modeling (n=36) using the NEXT cognitive paradigm. Participants performed both practiced and novel tasks during fMRI scanning. We modeled the generation of neurocognitive processes and resulting behavior using a multi-step activity flow modeling method. The model was data-driven, with a simulated flow of activity inputs from a subset of regions to downstream regions via empirical functional connectivity routes. Functional connectivity was estimated with graphical lasso at the region level and principal component regression vertexwise on regions surviving lasso thresholding, and trial-wise betas were estimated using ridge regression. Univariate fMRI analyses revealed increased activation in cognitive control networks and decreased activation in the default-mode network during novel tasks. The generative model simulated motor output patterns in somatomotor network with above-chance classification accuracy. The stepwise region-level flow graph highlighted the critical role of dorsal attention network and front parietal network in processing inputs from VIS1/VIS2 to bias appropriate motor responses. These findings provide a mechanistic framework for understanding how cognitive control networks enable RITL. Future work will simulate lesions within the model to further explore specific network contributions.

Topic Area: EXECUTIVE PROCESSES: Goal maintenance & switching