A network analysis of audiovisual affective speech perception

In this study we were interested in the neural system supporting the audiovisual (AV) integration of emotional expression and emotional prosody. To this end normal participants were exposed to short videos of a computer-animated face voicing emotionally positive or negative words with the appropriate prosody. Facial expression of the face was either neutral or emotionally appropriate. To reveal the neural network involved in affective AV integration, standard univariate analysis of functional magnetic resonance (fMRI) data was followed by a random-effects Granger causality mapping (RFX-GCM). The regions that distinguished emotional from neutral facial expressions in the univariate analysis were taken as seed regions. In trials showing emotional expressions compared to neutral trials univariate analysis showed activation primarily in bilateral amygdala, fusiform gyrus, middle temporal gyrus/superior temporal sulcus and inferior occipital gyrus. When employing either the left amygdala or the right amygdala as a seed region in RFX-GCM we found connectivity with the right hemispheric fusiform gyrus, with the indication that the fusiform gyrus sends information to the Amygdala. These results led to a working model for face perception in general and for AV-affective integration in particular which is an elaborated adaptation of existing models.