There is a paradox about how our social understanding develops if we take seriously both theory theory and the cognitivist dictum that all skilful interaction has robust conceptual underpinnings. On the one hand, it is clear that young infants demonstrate a capacity to reliably detect and respond to other’s intentions. For example, recent experimental evidence confirms that they have the capacity to appropriately parse what would otherwise be an undifferentiated behaviour stream at its mentalistic joints. If we follow the cognitivist trend in thinking that having the appropriate concepts is an antecedent requirement for such recognitional capacities then it follows that we should ascribe to these infants the concept of intention (or, at least, a concept of intention). But if we also hold that mentalistic concepts are constituted by their links with other mentalistic concepts, such as belief and desire, as assumed by proponents of theory theory, then we ought to conclude that infants lack certain concepts, such as intention, because at their tender age they lack other concepts, such as belief, which are required to constitute their contents. I will consider two possible atomistic ways of squaring this circle, in the process ultimately rejecting cognitivism in favor of the idea that various nonconceptual capacities best account for our initial abilities for recognizing and responding to intentional agency. I hold that it is what these responses are directed at that remains common throughout the developments of our mentalistic concepts, enabling us to make sense of talk of conceptual change. Not only does this resolve the paradox with which theory theorists must content, it shows how we can start our folk psychological careers without a theory or even any concepts.