Escaping the Negative Attentional Bias in Depression with Real-Time Neurofeedback

Participants receiving feedback tied to their own decoded attention became less likely than the sham group to remain stuck in the task's most negative attentional state. The study did not find a group difference in overall depression severity, so this is an encouraging mechanism result rather than evidence of symptom improvement.

Lubianiker N, Woods B, Nitchie F, Batzdorf A, Mennen A, Lin Q, Norman KA, Turk-Browne NB, Sheline YI.
medRxiv (preprint) · Publication date: 2026-10-05
Added to site: 2026-10-11
DOI: 10.64898/2026.09.30.26363896

Preprint posted October 5, 2026. Added to Neurofeedback.com October 11, 2026 after full-text review; not peer reviewed.

Modalities

Real-time fMRI neurofeedback, Closed-loop neurofeedback

Clinical focus

Major depressive disorder, Negative attentional bias

Population

Adults with major depressive disorder who were not taking psychotropic medication or attending psychotherapy during the study

Sample

48 analyzed participants randomized to real neurofeedback or yoked sham, 24 per group; 71 were screened

Intervention / method

Three real-time fMRI neurofeedback sessions using a multivoxel classifier to adjust the visibility of overlapping negative faces and neutral scenes according to decoded attention

Comparator

Yoked sham feedback from a matched participant, with genuine feedback used in the first and last runs to measure the task-specific outcome

Clinician summary

The preregistered group-by-session interaction for task-specific attentional-state stickiness favored real feedback in a Bayesian mixed model (posterior mean beta -0.223, 95% highest-density interval -0.446 to -0.014). Offline analyses found reduced precuneus/posterior cingulate and medial prefrontal activity during successful scene attention. Secondary MADRS and behavioral gaze outcomes did not differ between groups.

How to interpret the findings

This is a non-peer-reviewed preprint with a moderate sample and a task-specific neural outcome. The primary stickiness measure was undefined when participants never entered the most negative state, although the model retained 45 participants with at least one time point. Within-session fatigue, possible discouragement from noncontingent sham feedback, missing transfer to depression symptoms or a separate behavioral task, and unknown durability limit clinical interpretation.

Equipment and disclosures

Real-time functional MRI and the open-source RT-Cloud framework were used. The record describes a research paradigm, not a clinically available product or a BrainMaster system.

The preprint reports National Institute of Mental Health support under grant R61MH128492. The authors declared no competing interests. Preprint findings have not undergone journal peer review.

Related evidence

Compare findings from the same clinical area, including mixed results.

• Targeting the dopaminergic midbrain with precision 7-Tesla biofeedback training in depression: A proof-of-principle randomized controlled trial. · Positive with limitations
• EEG Microstate Alterations in Anhedonia Subtypes of Major Depressive Disorder and Their Modulation by Intermittent Theta-Burst Stimulation. · Positive with limitations
• Machine learning-based source-level EEG functional network biomarkers of antidepressant treatment response in major depressive disorder. · Positive with limitations