Abstract
Face identity can be represented in a multi-dimensional space with the average at the center. Considerable evidence suggests that this average acts as a perceptual norm, yet it is not clear how this norm is neurally coded. It has been proposed that each dimension of face space is coded by only two, oppositely-tuned, channels, each responding most strongly to one end of the dimension. On this view the average is coded implicitly by equal activation of both pools. An alternative account posits a third channel, tuned explicitly to the average (i.e., center of a dimension). We used two face identity aftereffect paradigms to distinguish between these two possibilities and to rule out a third possibility, narrow-band multichannel coding, in which there is no norm. In Experiment 1 we show that as adaptors become more extreme along a dimension aftereffects initially increase sharply and then plateau. Crucially there is no decrease, even for very extreme adaptors, ruling out a narrow-band multichannel account but consistent with both two-channel and three-channel accounts. In Experiment 2 we distinguish between these latter two models. To do so we measured how the range of faces identified as the average was affected by two adapting conditions - adapting to alternating images from opposite ends of an identity trajectory (e.g., Dan, AntiDan) and adapting to the average (i.e. the center of the trajectory). A two-channel model predicts similar effects for both adapting conditions, whereas a three-channel model predicts opposite effects for the two conditions. We found opposite effects: Alternating adaptation widened the range of faces identified as the average whereas adaptation to the average narrowed the range. These data provide evidence for explicit coding of the norm.
Meeting abstract presented at VSS 2016