ELSEVIER Cognitive Brain Research 4 (1996) 65-76 COGNITIVE BRAIN RESEARCH Research report Brain events related to normal and moderately scrambled faces Nathalie George a,*, Julie Evans b, Nicole Fiori a, Jules Davidoff c, Bernard Renault a a LENA, CNRS URA 654 / UPMC, HOpital de la Salp&ri~re, 47 Bd de l'H3pital, 75651 Paris Cedex 13, France b Department of Psychology, London Guildhall University, Old Castle Street, London E1 7NT, UK c Department of Psychology, University of Essex, Wivenhoe Park, Colchester C04 3SQ, UK Accepted 24 October 1995 Abstract The neural basis of normal and scrambled face processing was investigated by recording evoked potentials from 21 electrodes at standard EEG sites, with respect to a nose reference. Temporal negativities were found that result from two overlapping phenomena: they arise from the polarity reversal on temporal electrodes of the vertex P2, a positive wave peaking about 170-200 ms after the onset of a face stimulus, and also from an overlapping 'processing negativity' of long duration associated with the processing difficulty of the scrambled face stimulus. The comparisons of scalp potential and current density mappings support the proposal that some neuronal networks are active both for faces and scrambled faces and are compatible with the involvement of the superior temporal sulcus, the inferotemporal cortex and the parahippocampal and fusiform gyri, whereas the processing negativity would only involve the deepest generators of this network. Furthermore, the encoding of both faces and scrambled faces seems to take place predominantly in the right hemisphere. Keywords: Face; Evoked potential; Scalp current density mapping; Vertex P2; Processing negativity 1. Introduction Human faces form a group of rather homogeneous complex figures made from a set of loosely defined fea- tures [9,15,29]. The spatial arrangement of these features is crucial for face processing. For example, Homa et al. [19] demonstrated that facial features, e.g. nose, mouth or eyes, are better recognized when they are included in a normal facial context than when included in a disorganized, or 'scrambled', face (the face superiority effect). The role of facial configuration is also observed for the presentation of inverted faces; typically, inversion drastically disturbs recognition [49,50]. Recognition of inverted faces seems only to be possible on the basis of component analysis [39]. Recently, Davidoff and Donnelly [10] have adapted the face superiority effect [19] for use with non-tachistoscopic presentations of target and probe stimuli. Normal adults made more errors in recognizing scrambled unfamiliar faces (obtained by reversing the position of the eyes and nose) than in recognizing normal unfamiliar faces. In their * Corresponding author. Fax: (33) (1) 44243954; E-mail: lenang@ext.jussieu.fr 0926-6410/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved SSDI 0926-64 10(95)00045-3 first experiment, a centrally located target stimulus was presented for a 250-ms exposure duration and then imme- diately followed by a two-alternative forced-choice probe either in the form of a complete or a part probe. Complete probes were in the same form as the target, i.e. they were faces when the target was a face and scrambled faces when the target was a scrambled face. Part probes were individ- ual face parts (eyes, noses or mouths). The importance of the face configuration was emphasized in the results; despite the fact that complete probes contained redundant information, it was only with complete probes that faces were better recognized than scrambled faces. A second experiment confirmed this result using two different expo- sure durations of the target stimulus (250 and 2000 ms) and a different response choice procedure. We adapted the first experiment of Davidoff and Don- nelly [10] in order to record event-related potentials (ERPs) to the presentation of faces and scrambled faces. The assumption was that face compared to scrambled face processing could be correlated with different spatio-tem- poral ERP organizations. More precisely, face stimuli (in- cluding schematic drawings) should evoke a positive ver- tex P2 wave of 170-200 ms peak latency which reverses in polarity on temporal electrodes [21]. It is known that the