Source preprint posted October 5, 2026. Added to Neurofeedback.com October 11, 2026 after full-text review. This report has not undergone journal peer review.
Researchers randomized 48 adults with major depressive disorder to three sessions of real-time fMRI neurofeedback or yoked sham feedback. Participants viewed overlapping negative faces and neutral scenes while a classifier estimated where their attention was directed. In the real-feedback group, the display changed according to each person's own decoded brain pattern.
The preregistered outcome measured how likely participants were to remain stuck in the task's most negative attentional state. The change across sessions favored real feedback in the Bayesian model, with a posterior mean interaction of -0.223 and a 95% highest-density interval from -0.446 to -0.014. Offline analyses also found reduced activity in precuneus/posterior cingulate and medial prefrontal regions during successful attention to the neutral scene.
The result is promising because it tests a specific mechanism with a double-blind, yoked-sham design. It is not evidence that the intervention improved depression. The groups did not differ on overall depression severity or a separate behavioral gaze task. The primary measure was undefined when participants never entered the most negative state, within-session fatigue was substantial and sham feedback may itself influence learning.
The next step is to determine whether this task-specific learning transfers to everyday attention, persists over time and improves clinically meaningful symptoms. Replication after peer review is also needed.
The study was supported by the National Institute of Mental Health. The authors declared no competing interests.
AI-assisted source review and summary; no human clinical review is implied. The image is an AI-generated conceptual attention-training scene. It does not depict a study participant, actual clinic, fMRI procedure, measured brain data or treatment result.