Abstract
Visual adaptation is a powerful tool for understanding perception. Most studies have focused on the effects of adaptation on low-level features such as local orientation, as in the tilt aftereffect. Adaptation to faces on the other hand can produce significant aftereffects in identity, expression, and ethnicity etc, which are high-level traits. Our recent findings on curve adaptation suggest that the curvature aftereffect can be generated by an incomplete curve (Xu and Liu, VSS 2011), suggesting that missing information about the curve is filled-in. In the current study, we aim to investigate whether aftereffects can be generated by partially visible faces. We first generated partially visible faces using the bubbles technique, in which the face is seen through randomly-positioned circular apertures, and tested whether subjects were able to identify the facial expression through the bubbles. We then selected 9 faces whose facial expressions the subjects could not clearly identify. When we adapted the subjects to a static display such that each trial one of these 9 faces was randomly selected for adaptation, we did not find significant facial expression aftereffect. However, when we changed the adapting pattern to a dynamic video display of these faces, we found a significant facial expression aftereffect. In both conditions, subjects cannot tell facial expression from individual faces. It therefore suggests that our vision system can integrate these unrecognizable faces over a short period of time and this integrated percept will affect our judgment on subsequently presented faces. We conclude that face aftereffects can be generated by partial face features with little facial expression cue, implying that our cognitive system fills-in the missing parts during adaptation.
Meeting abstract presented at VSS 2013