The aim of the present review consists in reviewing data inconsistent with assumptions made by modular cognitive models of familiar people recognition. In particular, some of these inconsistencies are due to the failure to consider hemispheric specialization as an important variable in familiar people recognition. Indeed, hemispheric asymmetries exist between familiar faces and voices, underpinned by the right hemisphere, and names, subsumed by the left hemisphere. Furthermore, before the level of the person identity nodes (PINs), cross-communication exists between the perceptual channels for faces and voices, but not the channel for faces. Additionally, familiarity judgements are generated at the level of the modality-specific recognition units, with a right hemisphere dominance in the generation of face and voice familiarity feelings and PINs should not be considered as a simple gateway to a semantic system, storing information about people, but as structures involved in person-specific information retrieval processes. These data show that person-specific representations are mainly based on perceptual (face and voice) information in the right hemisphere and on verbal information in the left hemisphere.