Abstract
Different regions of interest (ROI) of the brain have been hypothesized to encode distinct emotion categories, most notably the amygdala in fear and sadness, the insula in disgust, the nucleus accumbens in anger, and the cingulate in happiness. An alternative model hypothesizes that the hippocampus and the postcentral gyrus encode these five emotion categories plus surprise. Each brain lesion, traumatic event and psychiatric disorder favors one of these models. We explore 33 hypothesized ROI with MVPA and show that most hypotheses hold true. Most categories are in fact consistently and differentially represented in multiple regions of the brain, suggesting there is a large network of areas responsible for the analysis and categorization of facial expressions of emotion. For example, happiness is consistently and differentially represented in the cingulate, hippocampus, postcentral gyrus, frontal pole, amygdala, thalamus and insula to name but a few, sadness in the parahippocampal gyrus, temporal pole, and superior temporal gyrus among many others. More specifically, we examined the brain activation patterns of 24 subjects using fMRI. During the scan, participants viewed samples of these six categories of facial expressions plus neutral. Each facial expression was displayed by 120 faces with distinct identities. The experiment consisted of ten 6-min runs. Each run had fourteen 12.5s face blocks alternating with fifteen 12.5s resting and fixation periods. A passive viewing paradigm was used on half of the subjects while a two-alternative-force-choice task was given to the other half. We applied Linear Discriminant Analysis in 33 hypothesized ROI. Significant classification accuracies (p<0.05) were obtained for the hypothesized emotion versus the other emotions in most ROI. To conclude, our results support the notion that categorical representation of these six emotion categories resides in a large network of brain areas and further suggest that this information is not uniformly distributed in the networks.
Meeting abstract presented at VSS 2014