Orientation-sensitivity of face identity aftereffects

Abstract

Our ability to recognize faces despite their similarity as visual patterns depends on high-level face-coding mechanisms that are strongly tuned to upright faces. If face aftereffects reflect adaptation of these mechanisms, as widely assumed, then they should be sensitive to face orientation. Previous studies have not supported this hypothesis, but have generally used a figural aftereffect paradigm, which may not optimally engage expert face-coding mechanisms. Here, we used an identity aftereffect paradigm, which requires identification of target faces, to provide a stronger test of the hypothesis. We measured identity aftereffects for upright and inverted faces, with and without eliminating low-level retinotopic adaptation. Baseline identification performance was substantially better for upright than inverted faces, confirming that our task tapped orientation-selective face expertise. With orientation varied between participants, aftereffects were almost twice as large for upright as inverted faces, on three different aftereffect measures (change in threshold, change in overall proportion correct, change in perceived identity of the average face). With orientation varied within participants, the results were less clear. We suggest that adaptation of expert face-coding mechanisms can contribute to face identity aftereffects, although the effect may not be very robust.