Brain and perceptual representations of faces, voices, and person identity

Abstract

Specialised brain regions have been identified that selectively process faces (Kanwisher, McDermott, & Chun, 1997) and voices (Belin, Zatorre, Lafaille, Ahad, & Pike, 2000). However, little is known regarding how information from the face and voice is integrated to represent person identity. According to two distinct models, person identity representations could exist in multimodal brain regions that process both faces and voices, or in face-selective and voice-selective regions through functional connections (Campanella & Belin, 2007). This thesis tested these predictions using functional magnetic resonance imaging and representational similarity analysis to directly compare representations of familiar faces and voices in different brain regions. A representation of person identity was found in the multimodal right posterior superior temporal sulcus, providing support for the notion that face and voice information is integrated in multimodal regions. This thesis also showed evidence of representations of face identity in face-selective regions and voice identity in voice-selective regions that could both ‘tell apart’ different identities and ‘tell together’ different, naturalistically varying tokens of each person’s face and voice (Burton, 2013). To investigate the information processed in these regions, brain representations of faces and voices were compared with multiple models of face and voice information. Face-selective regions and voiceselective regions were found to process information regarding the perceived and objective visual/auditory similarity between faces and voices, respectively. These findings provide novel insights into the computations of these regions. Lastly, this thesis investigated the relationship between information that is perceived from the face and the voice, and how this relationship compares between familiar and unfamiliar people. Information on social traits and perceived similarity was consistent across faces and voices, and more so for familiar compared with unfamiliar people. This finding suggests that having prior semantic knowledge about a person leads to similar judgements of their face and voice. Moreover, it suggests that some concordant information may be available even in the faces and voices of unfamiliar people.