Faces and voices are key features in human recognition. If the cerebral regions involved in the processing of both information are well known, the way the brain links them together is still largely unknown. Here we measured brain activity using fMRI while participants were recognizing previously learned faces, voices and voice–face associations.
14 healthy participants took part in this study. All were right-handed, french native speakers and had normal vision and audition. Stimuli consisted in 4 associations between a female face, a female voice saying the french word “Bonjour�, and a belgian family name. Prior to the fMRI session, a training session served to familiarize the participants with the associations. During the fMRI session, 3 conditions were presented : faces alone (F), voices alone (V), and face-voice associations (VF). Participants had to categorize each stimulus according to its identity (its family name).
Using a subtraction method between bimodal and unimodal conditions [ (VF-(V+F) ], we observed that face-voice associations activated both unimodal visual and auditory areas, and 2 multimodal convergence regions located in the left angular gyrus and the right hippocampus. An enhanced connectivity of the hippocampus with unimodal areas was confirmed by a functional connectivity analysis, while the left angular gyrus showed an enhanced connectivity with premotor and frontal regions.
The present results demonstrate that binding faces and voices relies on the activation of a cerebral network implying the unimodal face and voices areas as well as specific integration regions, such as the left angular gyrus possibly involved in some processes of divided attention between visual and auditory inputs, and the right hippocampus in which inputs from visual and auditory regions may be integrated into a coherent face-voice representation which could be compared with those stored in memory.