Abstract
Face adaptation research — whereby prolonged viewing of a manipulated face systematically alters the otherwise normal percept of the next presented face — has the potential to bridge psychological face theories and their neural correlates. For instance, it has been shown that upright face shape is encoded in a norm-based fashion, implicating opponent coding mechanism at the neural level (Robbins et al., 2007). The current study tested whether inverted face shape is encoded in a norm-based or exemplar-based fashion. Norm-based coding (two oppositely and broadly tuned neural pools) predicts aftereffects will increase in size as the distortion level of the adaptor becomes more extreme; exemplar-based coding (multiple narrow band pools) predicts a decrease. Our distortion shifted eyes up or down relative to other features. Results show a greater shift in the face perceived as normal following adaptation to extreme adaptors (e.g., eyes up by 50 pixels to near hairline) than closer-to-average adaptors (e.g., eyes shifted up by 5 pixels), supporting norm-based coding for both upright and inverted faces. Further, the aftereffects were also used to model the neural tuning curve for face shape in both orientations. This study suggests that norm-based strategy might be applicable for any class of objects whose members share the same basic configuration, although the neural details might differ.