Disentangling large-scale brain dynamics and their links to behavior during the emotional face matching task.

Emotion processing engages multiple large-scale brain networks. However, prior investigations relying on a priori, contrast-based models of brain activity obscure networks' distinct temporal dynamics and roles in task performance. Here, we performed tensor independent component analysis to identify and track concurrent brain processes, including those with non-canonical dynamics, during the emotional face matching task (EFMT) in healthy young adults (n = 413; n = 416 replication). Ten EFMT-recruited large-scale brain networks were identified, reflecting flexible recoupling of visual association cortex to diverse non-visual networks. These networks collectively engaged 74% of cortex and more strongly explained variability in cognition and a performance-based index of emotion interference than contrast-based amygdala activation/connectivity. Variability in EFMT-recruited network activity was more strongly linked to variability in cognition than affect. Findings reveal a rich landscape of brain activity under the surface of contrast-based fMRI analyses and deepen insights into the distinct brain processes underlying subcomponents of emotional face processing.