Background: Emotion dysregulation has been suggested as an endophenotype of bipolar disorder (BD). Neuroimaging studies show aberrant neural activity during emotion regulation in remitted patients with BD and their unaffected first-degree relatives (UR) compared to healthy controls (HC). However, behavioural studies produce conflicting - generally negative findings - possibly due to limited sensitivity and ecological validity of current behavioural paradigms. Methods: This study aimed to explore emotion regulation in BD (n = 30) and UR (n = 26) relative to HC (n = 47) by using a novel emotion regulation task in virtual reality (VR). Participants were instructed to either react naturally to, or dampen, their emotional response to highly positive or highly negative scenarios presented in first-person 360-degree spherically camera-recorded VR environments. Participants also completed a more traditional computerised task of emotion regulation for comparison purposes. Results: Patients with BD exhibited difficulties with down-regulating their negative emotions in the VR paradigm compared to HC and UR (ps ≤ .04), whereas UR did not differ from HC (p = .97). There was no emotion regulation difference between groups in the more traditional computerised task (ps ≥ .40). Limitations: The small sample size limits generalisability. Conclusions: The results suggest trait-related reduced ability to down-regulate negative emotions in BD patients compared to HC in the VR paradigm, but not in the more traditional task of emotion regulation. This may indicate that VR provides a more sensitive measure relative to traditional paradigms. The findings provided no support for aberrant emotional regulation as an endophenotype of BD given the normal emotion regulation performance in UR.