Dismantling the mechanism of VR self-compassion training: A two-session controlled trial with active controls

While prior research suggests that self-compassion training delivered via virtual reality (VR) is effective, it is limited by the absence of appropriate control conditions. Using a two-session, sequentially controlled trial design, this study compared a conventional VR self-compassion training paradigm to two active control conditions: a non-VR, computer-based self-compassion training, and a VR intervention that replaced the self-to-self component with a generic third-party interaction (VR Generic).

Eighty-nine participants completed the training. Primary (self-compassion, self-criticism) and secondary (depression, anxiety, stress) outcomes were assessed pre- and post-intervention. Exploratory analyses examined virtual embodiment (agency, ownership, self-location) as a mechanism of change.

No significant group differences emerged for anxiety, stress, or self-compassion. However, distinct efficacy patterns emerged for specific symptoms. For depression, the externally focused VR Generic condition demonstrated a preliminary advantage for symptom reduction compared to the computer control. Conversely, for self-criticism, the self-focused VR Self-Compassion training was significantly more effective than both controls. Both VR conditions induced higher agency than the computer control, but only the self-focused VR condition increased body ownership. This heightened embodiment, however, did not universally translate into superior wellbeing benefits across all domains.

The findings suggest that specific VR formats may be better suited for targeting distinct psychological symptoms. While externally focused VR shows promise for alleviating depressive symptoms, the embodied "self-to-self" simulation appears critical for dismantling self-criticism. Rather than acting as a universally superior intervention, VR embodiment may offer a targeted approach tailored to specific symptom presentations.