Abstract
After looking at a photograph of someone, an average face can appear ‘opposite’ in gender, identity, and other attributes – but what happens to the appearance of other faces? Face aftereffects have repeatedly been ascribed to perceptual re-normalisation. According to this account, the adapting face, and more exaggerated versions of it, should appear more neutral, a pattern linked to adaptation within norm-based opponent channels. Other spatial aftereffects (e.g. tilt and spatial frequency) have a locally-repulsive pattern, in which differences between adapting and test stimuli are exaggerated, with little change to the adaptor’s appearance. This pattern has been linked to multi-channel encoding schemes. Existing face aftereffect reports do not provide clear evidence for either pattern, because most quantify the aftereffect as a shift in a category boundary during a binary classification task. Such tasks are sensitive only to changes in the appearance of near-neutral stimuli, for which re-normalising and locally-repulsive aftereffects make the same prediction. We overcome this problem by using a spatial comparison task to measure the relative aftereffect induced between two differently-adapted retinal locations. By interleaving three standard stimuli, we are able to measure the relative aftereffect at three points: each of the two adapted values, and mid-way between them on the stimulus dimension. In our experiments, re-normalisation should manifest as the same magnitude and direction of relative aftereffect at all three standard values. Local repulsion should manifest as a larger relative aftereffect at the central standard value than either of the adapted standard values. In behaviourally-matched experiments we compared aftereffect patterns after adapting to tilt, facial identity, and facial gender. In all three experiments, data matched the predictions of a locally-repulsive but not a re-normalising aftereffect. Data are consistent with there being similar encoding strategies for tilt, identity, and gender, likely involving multiple channels with no unique norms.
Meeting abstract presented at VSS 2015