Embodied approaches to mindreading have tended to define themselves in contrast to cognitive approaches to social mindreading. One side effect of this has been a lack of engagement with key areas in the study of social cognition—in particular the topic of how we gain an understanding of the referential nature of others' thoughts, and how that understanding develops from infancy. I argue that embodied accounts of mindreading are well equipped to enter into this debate, by making use of the notion of a joint mental state, but that doing so will require taking a less antagonistic attitude towards mainstream cognitive approach.