Poster B30, Sunday, March 26, 8:00 – 10:00 am, Pacific Concourse
Activity flows over task-evoked networks shape cognitive task activations across task switches
Michael Cole1, Takuya Ito1, Douglas Schultz1, Ravi Mill1; 1Rutgers University-Newark
We recently found that activity flow – the movement of activations (e.g., from visual to motor regions in a visual-motor task) – over resting-state networks accurately predicts held-out task activations. This suggests that activity flow over the brain’s stable intrinsic network architecture shapes task activations. This result is surprising, however, given that routes of activity flow are thought to change across task context. For instance, the need to switch tasks – from a given visual stimulus causing a left button press to causing a right button press – would require the relevant neural population in visual cortex to switch where it sends its activity. We hypothesized that such shifts in activity flow are reflected in task-evoked functional connectivity changes to the intrinsic network architecture. Supporting this, we found enhanced prediction accuracies of held-out task activations when activity flow was estimated over task-evoked networks (relative to resting-state networks). This suggests that task-evoked functional connectivity describes the task-related network updates underlying shifts in activity flow. Further, using a computational model we found this effect could be explained in terms of “population thresholds” (nonlinear sigmoid functions). Specifically, incoming activity flow during tasks placed a given region’s activity closer to its threshold, such that the probability that any other incoming activity would produce a response in the region was increased (i.e., a functional connectivity shift). Overall, these results demonstrate a role for task-evoked functional network changes in shaping task activations and therefore cognitive processing, with nonlinearities in neural populations supporting this network mechanism.
Topic Area: EXECUTIVE PROCESSES: Goal maintenance & switching