Abstract
The perception of others' body movements is subserved by a network of lateral temporal, parietal, and premotor brain areas, here called the action perception system (APS). Using fMRI adaptation, we explored selectivity for biological motion and/or biological form in this network. Participants watched 2 s clips of recognizable actions of a human (biological motion and form), a humanoid robot (mechanical motion and form), or an android (mechanical motion, biological form). The latter conditions actually were of the same robot, with identical kinematics, videotaped with or without human-like skin. Each movie clip was preceded by the same movie or a different movie and we explored brain areas that showed adaptation. With the exception of left extrastriate body area, which showed adaptation for biological appearance (human and android), the APS was not selective for motion or form per se. Instead, specific responses were found to the mismatch between motion and form: Whereas fMRI adaptation results for the human and robot (the agents that differed in both motion and form) conditions were similar to each other, there were additional areas of adaptation for the android condition. Most notably, in bilateral anterior intraparietal sulcus, a key node in the APS, we found significantly more adaptation for the android than the other agents, indicating the congruence of form and motion is an important factor to consider. We interpret these data in the predictive coding framework (Rao and Ballard, 1999) and suggest that these additional responses to the android reflect increased prediction error as the brain negotiates an agent that looks biological, but does not move biologically. These results contribute to our goal of identifying the functional properties of the APS, and may also help demystify the “uncanny valley” hypothesis from robotics, whereby artificial agents that are too human-like can evoke negative reactions (Mori, 1970).
Kavli Institute for Brain and Mind. California Institute for Telecommunications and Information Technology and.