Fig. 5: Neurophysiological and behavior underpinnings of FFA-based face-processing evidence during memory-relevant epochs in the feedback-based association learning task.

The support vector machines, which were trained on the 1-back localizer task of each participant (n = 28, two participants were excluded from cross-classification analyses based on low localizer task performance), predicted face processing during memory recall, encoding and inter-trial-interval in the feedback-based association learning task, based on the 14 most face-selective voxels in the cytoarchitectonic probability mask of a fusiform gyrus 4 region. a General linear model results on evidence for face-processing displayed increased hemodynamic response in regions among fusiform face area (FFA), posterior medial frontal cortex (pMFC), dorsolateral prefrontal cortex (dlPFC), anterior insula (Ins), premotor cortex (PMC), and a cluster overlapping with a cytoarchitectonic mask of the basal nucleus of Meynert (NBM) subregion of the cholinergic basal forebrain (cBF) and amygdala. b The level of evidence for face-processing was higher when there was a demand of memory improvement during recall and encoding, and significantly higher for subsequent recall success during encoding epochs, as found in the linear mixed model results.