Electrophysiological correlates of self-specific prediction errors in the human brain

Abstract

Recognising one's self, vs. others, is a key component of self-awareness, crucial for social interactions. Here we investigated whether processing self-face and self-body images can be explained by the brain's prediction of sensory events, based on regularities in the given context. We measured evoked cortical responses while participants observed alternating sequences of self-face or other-face images (experiment 1) and self-body or other-body images (experiment 2), which were embedded in an identity-irrelevant task. In experiment 1, the expected sequences were violated by deviant morphed images, which contained 33%, 66% or 100% of the self-face when the other's face was expected (and vice versa). In experiment 2, the anticipated sequences were violated by deviant images of the self when the other's image was expected (and vice versa), or by two deviant images composed of pictures of the self-face attached to the other's body, or the other's face attached to the self-body. This manipulation allowed control of the prediction error associated with the self or the other's image. Deviant self-images (but not deviant images of the other) elicited a visual mismatch response (vMMR) — a cortical index of violations of regularity. This was source localised to face and body related visual, sensorimotor and limbic areas and had amplitude proportional to the amount of deviance from the self-image. We provide novel evidence that self-processing can be described by the brain's prediction error system, which accounts for self-bias in visual processing. These findings are discussed in the light of recent predictive coding models of self-processing.