Visual perception is attuned to detect stimuli with high social value, such as faces, bodies and biological motion. I will discuss new findings from functional neuroimaging, eye tracking and behavioral methods on healthy adults and preverbal infants. Based on those findings, I will argue that human visual perception is further prepared to represent socially relevant (spatial) relations among multiple entities. This so-far uncharted property of human perception substantiates the construct of social vision, whereby the result of the perceptual analysis is not independent shape recognition, but an integrative structuring of information for higher-level inferential operations in social cognition.