Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Cognitive Modes Detectable by fMRI during the Sternberg Task

Poster Session C - Sunday, March 8, 2026, 5:00 – 7:00 pm PDT, Fairview/Kitsilano Ballroom

Erica Zeng^1,2 (), Linda Chen¹, Eva Feredoes³, John Shahki², Todd S. Woodward^1,2; ¹University of British Columbia, Canada, ²BC Mental Health and Substance Use Services, Canada, ³University of Reading, United Kingdom

Background: Functional magnetic resonance imaging (fMRI) studies of working memory (WM) have focused on individual brain regions such as the dorsolateral prefrontal cortex. However, in past work we showed that five cognitive modes can be consistently retrieved from fMRI data for WM tasks, each with their own spatial pattern and empirically derived function. This more complete set of fMRI results for WM provides a richer set of measures for input for machine learning algorithms, valuable for classifying individuals on natural abilities, and subtyping patients with brain disease for precision medicine. In the present study, we confirm these precise anatomical configurations and WM-induced BOLD signal patterns on a novel sample. Methods: 38 adults completed the Sternberg WM task, whereby participants remembered a string of letters over a 4-second delay. Constrained principal component analysis for fMRI (fMRI-CPCA) was used to extract cognitive modes and compare to past published work. Results: The previously identified cognitive modes of Initiation (initiating WM cognitive processes, peaking early trial), Focus on Visual Features (deactivated throughout trial since visual details are not required to complete tasks), Maintaining Internal Attention (load-dependent activation during maintenance period), and Default Mode (load-dependent deactivation throughout trial) were retrieved. Conclusions: These results illustrate the stability and replicability of spatial and temporal patterns, and their empirically derived cognitive functions, captured by cognitive modes involved in WM. The advantages of using this rich set of measures for input for machine learning algorithms for classifying individuals and subtyping patients for precision medicine are discussed.

Topic Area: EXECUTIVE PROCESSES: Working memory