Coordinated social interactions are supported by integrated neural representations.

Joint actions are defined as coordinated interactions of two or more agents towards a shared goal, often requiring different and complementary individual contributions. However, how humans can successfully act together without the interfering effects of observing incongruent movements is still largely unknown. It has been proposed that interpersonal predictive processes are at play to allow the formation of a Dyadic Motor Plan, encompassing both agents' shares. Yet, direct empirical support for such an integrated motor plan is still limited. In this study, we aimed at testing the properties of these anticipated representations. We collected EEG data while human participants (N = 36; 27 females) drew shapes simultaneously to a virtual partner, in two social contexts: either they had to synchronize and act jointly, or they performed the movements alongside, but independently. We adopted a multivariate approach to show that the social context influenced how the upcoming action of the partner is anticipated during the interval preceding the movement. We found evidence that acting jointly induces an encoding of the partner's action that is strongly intertwined with the participant's action, supporting the hypothesis of an integrative motor plan in joint but not in parallel actions.