The paper first gives a conceptual distinction of the first, second and third person perspectives in social cognition research and connects them to the major present theories of understanding others (simulation, interaction and theory theory). It then argues for a foundational role of second person interactions for the development of social perspectives. To support this thesis, the paper analyzes in detail how infants, in particular through triangular interactions with persons and objects, expand their understanding of perspectives and arrive at a self–other metaperspective. This allows them to grasp the other's as well as their own perspective as such, which is equivalent to an explicit third person perspective and to an explicit first person perspective or self-consciousness. The paper describes the major steps towards these perspectives, pointing to a close interdependence of both developments. It argues that embodied second person interactions are not only an enabling, but the constitutive condition for the development of an explicit first and third person perspective.