In contrast to individual tasks, a specific social setting is created when two partners work together on a task. How does such a social setting affect memory for task-related information? We addressed this issue in a distributed joint-action paradigm, where two team partners respond to different types of information within the same task. Previous work has shown that joint action in such a task enhances memory for items that are relevant to the partner's task but not to the own task. By removing critical, non-social confounds, we wanted to pinpoint the social nature of this selective memory advantage. Specifically, we created joint task conditions in which participants were aware of the shared nature of the concurrent task but could not perceive sensory cues to the other's responses. For a differentiated analysis of the social parameters, we also varied the distance between partners. We found that the joint action effect emerged even without sensory cues from the partner, and it declined with increasing distance between partners. These results support the notion that the joint-action effect on memory is in its core driven by the experience of social co-presence, and does not simply emerge as a by-product of partner-generated sensory cues.