The authors examined face perception models with regard to the functional and temporal organization of facial identity and expression analysis. Participants performed a manual 2-choice go/no-go task to classify faces, where response hand depended on facial familiarity (famous vs. unfamiliar) and response execution depended on facial expression (happy vs. angry). Behavioral and electrophysiological markers of information processing--in particular, the lateralized readiness potential (LRP)--were recorded to assess the time course of facial identity and expression processing. The duration of facial identity and expression processes was manipulated in separate experiments, which allowed testing the differential predictions of alternative face perception models. Together, the reaction time and LRP findings indicate a parallel architecture of facial identity and expression analysis in which the analysis of facial expression relies on information about identity. (Contains 2 footnotes, 2 tables, and 15 figures.)