INTRODUCTION

Despite the amount of experimental and theoretical contributions to the study of face recognition in recent years, the questions about the cognitive mechanisms and neural substrates subtending the perception of age and gender on human faces have remained little investigated. According to a most influential model of face processing (Bruce and Young, 1986), age and gender discrimination judgements are mediated by one same system, referred to as the Directed Visual encoding module. This module would be activated following the Structural encoding module of facial features, and would act independently of the face systems supposed to process the movements of the mouth and tongue in speech (lipreading), the emotional expression of face, and/or the person’s identity.

Existence of the Structural encoding module fits with the repeated findings of face-specific brain responses from neuropsychological (Benton and Van Allen, 1972; de Renzi et al., 1991), neuroimaging (review in Haxby et al., 2000) and electrophysiological studies in both humans (Allison et al., 1994a, 1999; McCarthy et al., 1999) and monkeys (Perrett et al., 1982, 1984; Desimone, 1991). Specifically, it has been assumed that the electrophysiological committant of structural facial encoding may be reflected, in scalp event-related potentials (ERPs), by the occipito-temporal N170 component recorded around 170 msec latency (Bentin et al., 1996; George et al., 1996; Jeffreys, 1996).

There are however little experimental data supporting the existence of the Directed Visual encoding module and/or its relationship to the Structural encoding module. While several studies have shown that prosopagnosic patients with a right posterior brain damage may be impaired in classifying faces according to age or sex (Newcombe and Russell, 1969; de Renzi et al., 1989; Bruyer and Schweich, 1991; Carlesimo and Caltagirone, 1995), the observations did not allow for dissociating the brain processes involved in age or gender detection from those supposed to be strictly associated with the structural encoding of facial features.

In a previous ERP study, we found that explicit as well as implicit processing of gender from faces was manifested by fronto-central scalp activities around 145-185 msec, well-delimited from the posterior N170 component (Mouchetant-Rostaing et al., 2000a). These effects, which were not found in control conditions requiring gender processing from hands, were interpreted as evidence for face-specific processes for gender perception, distinct from those involved in structural encoding of facial features. A primary goal of this study was therefore to extend the electrophysiological investigation of the Bruce and Young’s model (1986) with the following predictions: if age and gender processing from faces are achieved through a same functional (Directed Visual encoding) module, then one should observe (1) no effect of age processing on the N170 ERP component, (2) similar ERP activities (140-190 msec based on our previous findings) for age and gender processing.

In addition, recent findings both in our group and in others have reported evidence for very early signs of visual categorization (around 100 msec) in experiments, manipulating complex visual stimuli (Thorpe et al., 1996; Fabre-Thorpe et al., 2001; Keysers et al., 2001; VanRullen and Thorpe, 2001) including facial features (George et al., 1997; Seeck et al., 1997; Debruille et al., 1998; Linkenkaer-Hansen et al., 1998; Pizzagalli et al., 1999; Halgren et al., 2000; Mouchetant-Rostaing et al., 2000a, 2000b). Such early categorization effects based on low-level visual processes would challenge the current view not only on the cognitive models of face recognition, but more generally on the theories of visual information processing. A second issue addressed here was therefore to test whether such early categorization processes could be observed between two wide classes of age on faces.