Event-related potentials (ERPs) were measured while subjects passively looked at a line drawing of a normal face or a line drawing in which parts of the face were scattered so as not to be recognized as a face (scattered face) in order to estimate the speed of face recognition in humans. Because the spatial frequencies of these two types of line drawings were very similar, one could minimize the potentials due to the structure coding stage of the face processing. In addition, passive viewing of these stimuli eliminates the contribution of other processing such as discrimination or memory retrieval. Comparing these two conditions, statistically significant positive potentials were observed in frontal areas from 135ms after stimulus onset for the normal face. These early positive differences were probably due to the face processing per se in human visual recognition. Although statistically significant differences were observed in frontal areas, these are probably related to fusiform or inferior temporal area activation that is often reported using neuro-imaging technologies such as PET or fMRI.