Research on mimicry has demonstrated that individuals imitate in-group members more strongly than out-group members. In the present study, we tested whether such top-down modulation also applies for more extreme forms of direct mapping, such as for cross-contextual imitation settings, in which individuals imitate others’ movements without sharing a common goal or context. Models on self-other control suggest that top-down modulations are based merely on a direct link between social sensory processing and imitation. That is, perceived similarities between oneself and another person is sufficient to amplify a shared representation between own and others’ actions, which then trigger imitation. However, motivational accounts explain such findings with the assumption that individuals are motivated to affiliate with others. Because imitation is linked to positive social consequences, individuals should imitate in-group members more strongly than out-group members. We tested these two theoretical accounts against each other by applying a cross-contextual imitation paradigm. The results demonstrate that in-group members are more strongly cross-contextually imitated than out-group members the higher individuals’ motivation to affiliate with the in-group is. This supports motivational models but not self-other control accounts. Further theoretical implications are discussed.