Logo image
Task representation and cognitive flexibility: neural mechanisms of hierarchical cognitive control and working memory integration
Dissertation   Open access

Task representation and cognitive flexibility: neural mechanisms of hierarchical cognitive control and working memory integration

Xitong Chen
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Summer 2025
DOI: 10.25820/etd.008115
pdf
Xitong Chen Thesis5.23 MBDownloadView
Open Access Free to read and download

Abstract

Human behavior is remarkably flexible. People adapt their actions and decisions to changing contexts while integrating multiple sources of task-relevant information. This flexibility depends on task representations, mental representations that encode contexts, stimuli, and responses, to align actions with goals. However, how the human brain constructs and utilizes these representations remains unclear. This dissertation investigates several major questions to determine the neurocognitive architecture supporting task representations. First, how does the brain maintain hierarchically organized task representations, in which higher-level context representations influence actions and decisions, and how does it flexibly update these representations when contexts change? Second, how does working memory integrate multiple sources of information within task representations, including task content (the “what”) with the appropriate rules (the “how”), to generate goal-compatible actions and decisions? Across three studies, this work combines functional neuroimaging, computational modeling, and EEG to examine neural circuits, computational mechanisms, and temporal dynamics underlying task content (the “what”) and task rule (the “how”) representations. First, using fMRI and a hierarchical cognitive control task, we demonstrated that the thalamus encoded context representations within a hierarchically organized task rule, and updated cortical contextual representations via thalamocortical connectivity. Second, computational modeling revealed that the multidimensional “what” representations were integrated in a joint probabilistic manner, and that their formats and properties remained stable across different task rules. This suggests that the “what” system operates independently from the “how” system within working memory. The influence of task rule appears to occur through modulation of the readout of these joint probabilistic representations, rather than altering their encoding formats. Third, EEG evidence further showed that task rules did not affect how the task content was initially encoded but instead modulated neural representations of how that content was maintained and accessed over time, serving as a control signal that routed the “what” representation toward context-appropriate actions and decisions. Together, these findings delineate several specific neural and computational processes—thalamocortical contextual updating, working-memory integration of “what” and “how”, and rule-directed information routing—that translate abstract goals into specific, goal-directed actions.

Details

Metrics

7 File views/ downloads
60 Record Views
Logo image