Poster Session E, Monday, March 25, 2:30 – 4:30 pm, Pacific Concourse
Revealing the Brain Network Structure of Individual Differences in Cognitive Control
Shelly R. Cooper1, Joshua J. Jackson1, Todd S. Braver1; 1Washington University in St. Louis
In cognitive neuroscience, fMRI is often used to impose specific task states in order to understand the neural underpinnings of cognitive functions. While individual variability may be characteristic of the brain network or region (task-independent), tasks may also have global effects (i.e., non-specific to brain network) that reveal individual differences. This is particularly salient for cognitive control, which is dependent upon the task at hand. We therefore hypothesized that specifying task states and brain networks as independent factors would best capture the underlying structure of individual differences. Structural equation models (SEM) were built from the Human Connectome Project dataset (based on the full release; n=1200), focusing on cognitive control-related tasks (N-back/Relational) and networks (frontoparietal/cinguloopercular). Four competing SEM models were tested that varied in whether both networks and tasks were treated as independent latent factors or not. Models were evaluated on goodness-of-fit indices and variance explained in a criterion outcome behavioral index of working memory function (List Sort task). The model with both tasks and networks specified as separate latent factors had the best fit index (chi-square, p<.001), and also explained the most total variance in working memory with the strongest predictor being the multi-network N-back factor. These findings suggest that considering tasks and networks as dissociable sources of cognitive individual difference is useful for interrogating brain-behavior relationships, especially in the cognitive control domain. Further work will explore the effects of adding additional brain networks (e.g., default mode) and tasks (e.g., language, social) to determine the generality of these conclusions.
Topic Area: METHODS: Neuroimaging