Abstract
Uncanny experience refers to the phenomenon that human-like objects, such as android robots or animated cartoons, elicit eerie feeling of unfamiliarity even if they considerably resemble humans (Mori, 1970). To understand the cognitive mechanism of this phenomenon, we used an oddball task for three face categories, for which their event-related potentials(ERPs) were simultaneously recorded. Uncanny faces were constructed by synthesizing facial contours and volumes sampled from a set of cartoon faces with skin textures obtained from another set of human faces. In the oddball task, participants reported human faces as an infrequent target (20% probability of trial occurrence), cartoon faces as a frequent standard (60%), and uncanny faces as an infrequent deviant (20%). After the ERPs recording session, the participants rated familiarity and human-likeness on a Likert scale. While the P3 amplitude measured from the target trials (i.e., human) was greater, the P3 latencies were notably shorter than the deviant trials (i.e., uncanny). Participants also rated lowest familiarly with the uncanny faces, nevertheless rating identical human-likeness between the uncanny and cartoon faces. These results suggest that the uncanny faces as a non-target deviant initially attracted less attention than the human face target; however, the subsequent perception of their uncanniness may have further sustained the participant's attention than the target until their categorical decision was complete.
Meeting abstract presented at VSS 2018