As social conflict is a leading stressor and health risk, it is crucial to study such stress under controlled conditions that can be adapted for physiological or neuroimaging research. Moreover, short conflicts would enable comparing or practicing multiple, same-session interventions for research, workplace or clinical settings, and be compatible with more neuroimaging methods than longer paradigms. Virtual Reality presents an opportunity for utilizing a more ecologically valid stressor, and here we compared the physiological response of 3-minute virtual reality (VR) vs face-to-face (F2F) interpersonal conflict, presented in randomized order. Our hypothesis was that, regardless of format, engaging in conflict would elicit elevated physiological responses compared to baseline rest, including heart rate (HR), blood pressure (BP), and heart rate variability (HRV) components. We also predicted that, as VR systems have a diminished sense of social-cue integrity, VR responses would be less robust compared to F2F delivery. Our results showed that both VR and F2F social stressors evoked significant elevated HR, BP (both systolic and diastolic), and decreased low frequency HRV. Repeated measures ANCOVAs revealed that only blood pressure measures showed a statistical difference for presentation format, with F2F conflict evoking higher BP measures than VR format. We also identified a potential confound of order effect in the HR data, with greater HR when the F2F condition was presented first. Overall, our study shows that VR can be a useful format for short ecologically-valid paradigms in basic research and clinical studies, with the caveats that it may not be as sensitive as a F2F encounter, such that incorporating a practice period may be necessary when using VR for research. Furthermore, as VR becomes more widely used for work and recreation people should be made aware that interpersonal stress from real or gameplay VR can evoke strain on the physiological system.