Adaptation in face animacy perception: An event-related potential study

Abstract

A real face differs from an artificial face mainly in the animacy. Nowadays, the perception boundaries between the real and artificial faces are becoming blurred in our life with the ubiquitous use of AI. Therefore, the perception of animacy causes increasing interests. Here, we used an adaptation paradigm to investigate the animacy perception in faces. We morphed a real and an artificial face to generate a continuum of face images, and asked participants to judge the animacy of those face images after they were exposed to a real face or an artificial face. We found that after adaptation to a real face, the subjects were apt to identify a subsequently ambiguous face to be inanimate, whereas after adaptation to an artificial face, the subjects were apt to identify a subsequently ambiguous face to be animate, i.e., the face animacy aftereffect (FAAE). We simultaneously recorded EEG during the task and analyzed the event-related potentials in response to the test faces, and found that adaptation to a face animacy suppressed the amplitude of LPP (late positive potential) and prolonged the latencies of N250r and LPP, in response to subsequent animacy-congruent faces. However, for subsequent animacy-incongruent faces, the amplitude was enhanced in LPP and the latencies were shortened in N250r and LPP. Those modulations of N250r and LPP activity act as a neural correlate of face animacy adaptation.