A lot of work in social virtual reality, including our own group's, has focused on effectiveness of specific social behaviours such as eye-gaze, turn taking, gestures and other verbal and non-verbal cues. We have built upon these to look at emergent phenomena such as co-presence, leadership and trust. These give us good information about the usability issues of specific social VR systems, but they don't give us much information about the requirements for such systems going forward. In this short paper we discuss how we are broadening the scope of our work on social systems, to move out of the laboratory to more ecologically valid situations and to study groups using social VR for longer periods of time.