Neural Coding of Facial Emotions in the Human Brain

Abstract

A distributed network of brain regions have previously been implicated in the neural processing of facial emotions (see Fusar-Polli et al., 2009). Such studies primarily used univariate methods of analysis; however, multivariate approaches allow new questions to be asked regarding the neural codes underlying categorization of facial emotions (e.g. Raizada & Kriegeskorte, 2009). Here we employed multivariate pattern analysis to investigate two questions concerning the neural coding of facial expressions: 1) do occipito-temporal brain regions contain emotion specific activity patterns? 2) If so, how do such activity patterns relate to perceptual categorization? We presented participants with each of the six basic facial expressions (plus neutral) in a block design while concurrently recording the fMRI BOLD signal. We trained a linear pattern classifier to discriminate the brain activity patterns generated by each expression. Voxels input to the classifier were selected from occipito-temporal regions that had high sensitivity to visual stimulation in an independent set of data. Significant decoding of facial emotions was found in each participant tested. Moreover, the errors in the neural classification of emotions significantly correlated with the errors made in a completely independent behavioral categorization experiment (that of Smith & Schyns, 2009). Thus information pertinent to perceptual categorization of facial emotions is present throughout occipito-temporal cortex.