We report on a functional magnetic resonance imaging (fMRI) study of the perception of human and artificial agents. Participants viewed videos of familiar body movements enacted by the android Repliee Q2, the human after whom it was modeled, and the skinned version of Q2 revealing its mechanical parts. We used a neural adaptation (repetition suppression) analysis to reveal brain areas sensitive to body movements, and explored whether the identity of the perceived agents modulated these responses. We found significantly higher activity in a distributed network of brain areas for the android, most notably in anterior intraparietal cortex. The responses for the human and the robot with the mechanical appearance resembled each other. We interpret these results within the framework of predictive coding and suggest that the uncanny valley phenomenon may have its roots in processing conflicts within the brains action perception system.