The stickiness of face adaptation aftereffects

Abstract

Face adaptation aftereffects have been assumed to disappear rapidly with time and/or interference, as do standard adaptation effects in low-level vision. Our results challenge this assumption. We considered two adapter duration procedures: a 160 s adapter period with separate pre- and post-adaptation phases; and a 5 s adapter period intermixed with post-adaptation trials. Each of these is similar to procedures used in previous studies. Our distortion types involved either radially compressing-expanding a face (where adaptation to a compressed face makes a normal face look expanded), or altering only the spacing between features so that the eyes were shifted up or down the head (where adapting to an eyes-down face makes a normal face appear to have its eyes shifted up). Results showed that, following a 160 s adapter period, significant adaptation survived a 15 min delay filled with visual stimulation; moreover, this occurred when the visual input involved only words and objects, and when many normal undistorted faces were included. For the 160 s adapter, a 24 hr delay removed the aftereffect. Following a 5 s adapter period, significant adaptation survived a 10 s delay including one face for 200 ms and either a normal face or an XXX stimulus for 5 s; indeed, inclusion of the filled delay only approximately halved the effect compared to immediate test. For the 5 s adapter, a 20 s delay with twice the number of intervening stimuli removed the aftereffect. Overall, we conclude that face aftereffects are much ‘stickier’ than traditional aftereffects. This has important implications for both theory and methodology. In terms of theory, we consider implications for neural coding models of the face aftereffects, and for the way in which the norm of face-space changes in response to recent input. In terms of methodology, our results argue it cannot be assumed that a short delay between testing adapter conditions is sufficient to remove carryover from the previous condition.