Stimulus repetition decreases both the magnitude of the fMRI BOLD response (e.g., Grill-Spector & Malach, 2001) as well as the activity of cells in macaque IT (e.g., Li, et al. 1993), an effect generally attributed to competitive learning. Such adaptation is proving itself to be an important technique for studying visual representations. If adapted neurons are presented with a new stimulus that varies from the adapting stimulus in some property, rebound from adaptation indicates that these neurons are sensitive to that property. Would the N170 EEG response show adaptation to individual faces? To separate adaptation from other possible serial position effects, sequences of faces were designed in which new faces were presented continuously throughout the experiment. EEG was recorded from a minimal montage ofT5 and T6. A repetition of an individual face resulted in a marked reduction of the magnitude of the N170 compared to that of a non-repeated face in an equivalent serial position (which showed no reduction), indicating that the N170 adapts to individual faces. This effect is in contrast to fMRI adaptation in the FFA, which does not show adaptation specific to individual faces. The information distinguishing faces can be roughly partitioned into variations of shape and pigmentation, with almost all the marked cost of contrast negation in face recognition attributable to the latter component. Which component(s) is (are) responsible for the adaptation of the N170? Faces were generated that differed solely in shape or pigmentation. The image differences were equivalent, by a Gabor jet measure, for the two types of changes. A shape change completely eliminated the adaptation. Pigmentation changes, in contrast, had no effect on the magnitude of the adaptation. To the extent that the tuning of a neural substrate can be inferred from its adaptation signature, all the tuning of the N170 is to the variation in face shape from individual to individual.

Zhong-lin Lu, Bosco Tjan