Temporal dynamics of integrative processes that construct task representations

Poster Session A - Saturday, April 13, 2024, 2:30 – 4:30 pm EDT, Sheraton Hall ABC

Stephanie Leach¹ (scleach@uiowa.edu), Hannah Morrow¹, Jiefeng Jiang¹, Kai Hwang¹; ¹University of Iowa

The human brain consists of distributed subunits encoding diverse sources of information. To flexibly guide behavior, the brain must be capable of integrating multiple streams of information. We have previously proposed a framework where integration is achieved by creating a joint distribution that encodes multiple task-relevant features to guide goal-directed behavior. In this framework, the degree of integration can be quantified by the statistical properties of this joint distribution. The goal of the current study is to determine the temporal dynamics of this integration computation. We collected electrophysiology data from 18 participants performing a paradigm requiring them to integrate perceptual (color) information with non-observable state information to execute a task. Trials started with a red and yellow dot array, in which the dominant color informs the task (face or scene judgement) subjects should perform, and the mapping between color and task randomly switched between two non-observable cognitive states. We used a Bayesian generative model to investigate integrative processes. This model estimates, on a trial-by-trial basis, subjects’ probabilistic belief on the cognitive state and dominant color. Critically, the model integrates these two probabilistic distributions into a joint distribution encoding the correct task to perform (and outperforms non-probabilistic control models in terms of higher prediction accuracy of task and lower BIC score). Our results show that statistical properties of the joint distribution most strongly correlate with transient delta-theta band (2-8Hz), revealing the functional role these electrophysiology signals play in integrating information to construct task representations.

Topic Area: EXECUTIVE PROCESSES: Other