The infant's participation in sequences of joint activity that require visual attention is usually seen as an outcome of and evidence for the existence of particular infant psychological competencies. In a review of the relevant literature, we suggest that what is presupposed in most theories of joint attention is the role that shared social practices play in understanding the mind. It is, in fact, with recourse to such practices that researchers theorize about the infant's understanding of mind in the first instance. We argue: (1) the mind is not an entity that is separable from human activity; (2) knowledge of shared practices is what the developing agent requires to come to an understanding of their own mind and that of others; and (3) rather than searching for the best indicator of a true competence lying behind and necessary for joint attention, we should consider the various forms of interaction involving shared attention as constitutive of varying degrees of understanding. We consider the relevance of these arguments for contemporary social developmental theory.