I examine the role of mutual gaze in social cognition. I start by discussing recent studies of joint visual attention in order to show that social cognition is operative in infancy prior to the emergence of theoretical skills required to make judgments about other people's states of mind. Such social cognition depends on the communicative potential inherent in human bodies. I proceed to examine this embodied social cognition in the context of Merleau-Ponty's views on vision. I expose some inner difficulties within Merleau-Ponty's position as well as to point out the ways of resolving them by means of combined insights from developmental psychology and the analyses of self-other relations from philosophies of dialogue.