Résumé : Ten healthy volunteers took part in this event-related potential (ERP) study aimed at examining the electrophysiological correlates of the cross-modal audio-visual interactions in an identification task. Participants were confronted either to the simultaneous presentation of previously learned faces and voices (audio-visual condition; AV), either to the separate presentation of faces (visual, V) or voices (auditive, A). As expected, an interference effect of audition on vision was observed at a behavioral level, as the bimodal condition was performed more slowly than the visual condition. At the electrophysiological level, the subtraction (AV - (A + V)) gave prominence to three distinct cerebral activities: (1) a central positive/posterior negative wave around 110 ms, (2) a central negative/posterior positive wave around 170 ms, AND (3) a central positive wave around 270 ms. These data suggest that cross-modal cerebral interactions could be independent of behavioral facilitation or interference effects. Moreover, the implication of unimodal and multisensory convergence regions in these results, as suggested by a source localization analysis, is discussed.