Time Course of the Face Identity Aftereffect

Exposure to a face can alter the perception of another subsequently presented face. Using a computational 3-D model derived from a database of 200 scanned faces [1,2], we have previously demonstrated the existence of face identity aftereffect (FIAE) [3]. In a multidimensional “face-space”, the average prototype face occupies the central position, while individual real faces are represented as points or vectors emanating from the center. In this context, by inverting a vector corresponding to an individual face, one can create a so-called “anti-face” [4], which served as the adapting stimulus in our previous study. While the FIAE has much in common with more traditional aftereffects, such as its negative sign and its storage during brief unstimulated periods [3], its dynamic aspects have not yet been studied. The present investigation aimed to determine the effects of adaptation and test durations on this phenomenon. Subjects learned to recognize four faces at different identity levels over a period of several sessions, until they could correctly name low-identity faces. We then tested the effects of adaptation and testing duration on the FIAE. In each trial, one of the four names was shown on the screen, followed by the presentation of an “adaptation” face (the anti-face of the named individual) for a period between 1.0 and 16.0 s. Immediately following adaptation, a “test” face (the average face) appeared for between 0.1 and 1.6 s. Subjects were required to rate, on a scale of 1 to 7, the degree to which the test face resembled the individual named at the beginning of the trial. They were specically told to restrict their judgment to the instant that the test stimulus disappeared. The results indicate that the FIAE resembles other aftereffects in that it is increased in magnitude by long adaptation times, as well as by short test times. Mean ratings showed good ts to the adaptation and test times with positive and negative logarithmic functions, respectively. However, despite this similarity, the magnitude of this proportionality in the FIAE was considerably smaller than in low-level aftereffects