The evidence, including several randomised trials and a recent meta-analysis, shows possible improvement in positive and negative symptoms, cognition and functioning — always in addition to antipsychotic treatment, never in place of it.
Updated :
What the research shows
EEG neurofeedback — standard amplitude training
(SMR/beta, theta, gamma, SCP, qEEG-guided protocols), in combination with antipsychotic treatment
AAPB3Probably efficaciousNeuroLogic3Probably efficacioussame vs AAPB
Level 3 maintained (AAPB level assigned jointly to biofeedback and neurofeedback.) The AAPB's base is amplitude training as an adjunct to antipsychotics: Naimijoo 2015 (n = 30 inpatients, SMR up with theta/delta inhibition, executive functioning improved vs control), Schneider 1992 (SCP), Singh 2020 (gamma, 31 patients, 12 weeks: frontal gamma power, working memory and PANSS improved) and the qEEG-guided series of Sürmeli 2012 (47 of 48 completers improved on the PANSS, mean score from 110.24 to 19.56) and Bolea 2010. The 2022-2026 window adds randomised superiority over waitlist, treatment as usual and sham from independent sites: Li 2024 (n = 80, single-blind sham, 30 sessions plus risperidone: MOAS aggression and PANSS excited, positive, cognitive and depressive components reduced), Narang 2026 (n = 68, 15 theta/beta sessions vs waitlist: vocational task performance and attention improved), Markiewicz 2024 (n = 37, 3-month add-on vs standard support: negative and general symptoms reduced, d2 attention improved) and Duan 2025 (meta-analysis of 14 RCTs, n = 1371: positive symptoms SMD −0.87, negative SMD −1.28, best with ≥ 8 weeks, ≥ 4 sessions/week and SMR/beta protocols). Level 4 is not yet assigned: protocols are unspecified in several abstracts, primary trials are low quality with publication bias, outcomes differ from trial to trial, and nothing has been tested without antipsychotics.
Other neurofeedback methods
Real-time fMRI neurofeedback (superior temporal gyrus, language network, anterior cingulate, frontoparietal connectivity) — auditory hallucinations and cognition
Ages
Adult (18+)
Techniques
fMRI neurofeedback
AAPB3Probably efficaciousNeuroLogic3Probably efficacioussame vs AAPB
Level 3 (AAPB level assigned jointly to biofeedback and neurofeedback; disaggregated here.) Multiple clinical studies from independent groups: Zweerings 2019 (double-blind sham-controlled crossover, language-network regulation; per AAPB 77 % of patients improved in daily activities and 63 % had fewer hallucinations), Orlov 2018 (11 analysed, all learned to down-regulate the left STG, hallucination scores reduced), Okano 2020 and Bauer 2020 (single sessions, AHRS reductions), Cordes 2015 (ACC control). 2022-2026: Bauer 2025 (n = 23, randomised, STG vs motor-cortex sham with mindfulness: hallucinations reduced in both groups without group difference; real NF produced greater reductions in secondary auditory cortex activation and auditory-DLPFC/ACC connectivity), Kobayashi 2025 (non-randomised pilot, n = 22, four sessions of frontoparietal connectivity NF vs N-back training: group-by-time interaction on digit span backward). Target engagement is consistent; clinical superiority over sham is not shown — under the Parsons 2026 framework a sham-null is not read as absence of a specific effect, but the base remains single-session and small: no raise.
Frontal gamma-coherence neurofeedback — working memory
Ages
Adult (18+)
Techniques
Coherence / connectivity
AAPB0Not ratedNeuroLogic2Possibly efficacious
Level 2. A single research programme: Singh 2020's open trial of frontal gamma neurofeedback (31 patients, 25 completing 12 weeks at two sessions a week: frontal gamma power, working memory, processing speed and PANSS improved, gains held at 4 weeks) followed by a double-blind placebo-controlled RCT targeting gamma coherence (Shu 2025: all active participants improved working memory relative to placebo, with coherence responses diverging by baseline working memory; sample size not stated in the abstract). No independent replication; cognitive rather than symptomatic outcome. Surface coherence training also carries the measurement caveats (volume conduction) noted elsewhere.
Level 1. A single case report (Nestoros 2022): a 38-year-old man with antipsychotic-resistant auditory hallucinations and delusions, rated on the SCL-90 and visual analogue scales immediately before and after a first 40-minute ILF session, with marked self-reported improvement. No controlled study, no series, no follow-up; nothing separates the session's effect from expectancy and the therapeutic setting.
Biofeedback
HRV, EMG, electrodermal and thermal biofeedback — psychotic symptoms, anxiety and functioning, as an adjunct to antipsychotics
AAPB3Probably efficaciousNeuroLogic3Probably efficacioussame vs AAPB
Level 3 maintained (AAPB level assigned jointly to biofeedback and neurofeedback.) Randomised trials: Clamor 2016 (n = 84 with attenuated psychotic symptoms; HRV biofeedback vs active relaxation vs waitlist: larger reductions in state paranoia than both controls, greater perceived control than waitlist), Pharr & Coursey 1989 (n = 30 medicated inpatients, EMG biofeedback vs progressive relaxation vs control: social functioning improved on nurse ratings, EMG reduced by 40 %), Markiewicz & Dobrowolska 2021 (n = 43, GSR biofeedback vs standard rehabilitation: BDNF and self-efficacy increased more). Older controlled EMG and thermal work (Nigl & Jackson 1979, Hawkins 1980) and pilots (Trousselard 2016: anxiety reduced, PANSS unchanged). No 2022-2026 primary trial; Schmidt 2025's scoping review (20 inpatient biofeedback studies) notes schizophrenia is a small heterogeneous subset. Level 4 is not reached: a single small trial per modality, and inconsistent clinical-symptom outcomes.
In short
Clinical reading
AAPB Level 3, joint biofeedback/neurofeedback, confirmed by NeuroLogic for amplitude neurofeedback (SMR/beta, theta, qEEG-guided; Li 2024, Narang 2026, Markiewicz 2024, Duan 2025), for fMRI neurofeedback (STG, language network) and for HRV/EMG/EDA biofeedback (Clamor 2016); gamma coherence 2, ILF 1. Explicit caveat: every study is an adjunct to an antipsychotic regimen; standalone efficacy has never been examined.
Protocols
SMR/beta reward with theta inhibition (Cz, F3/F4), qEEG-guided protocols matched to the individual profile, SCP; long series (≥ 8 weeks, ≥ 4 sessions/week per Duan 2025) in a supervised setting; resonance-frequency HRV, EMG or EDA for anxiety and social functioning.
Limits
No evaluation as a standalone intervention; protocols often undescribed in recent trials; meta-analysis dominated by low-quality trials with publication bias; heterogeneous populations and outcomes; no paediatric data (the only trial in high-risk youth used pupillometer-based neurofeedback).
Study base
Several RCTs, one meta-analysis of 14 RCTs (Duan 2025) and large qEEG-guided case series; 2022-2026 base: 11 publications indexed in the archive.
Brendan's perspective
Level 3, unchanged from the AAPB, and I would keep it there. The 2022-2026 window is markedly better than the old base: Li 2024 randomised 80 patients against sham, Markiewicz 2024 and Narang 2026 add independent sites, and Duan 2025 pools 14 RCTs with large effects on positive and negative symptoms. Level 4 fails on the basics — protocols undescribed in several papers, low-quality primary trials, publication bias in the meta-analysis, outcomes that change from study to study. Compared with what, measured how, reproduced by whom else: the answers here are still thin. Two things need saying before any of the clinical detail. This is a vulnerable population, and neurofeedback here belongs inside a structure that already includes psychological and psychiatric support — not alongside it, and certainly not instead of it. And most of the research is done on inpatients, which is a different setting from the one most of us work in. That said, I do think neurofeedback has a place here, particularly with medicated patients. It has to be done with the kid gloves on: protocols chosen conservatively, changes made slowly, arousal watched closely. It is entirely possible to destabilise someone in this population, and a practitioner working with them needs specialised training and needs to understand that risk properly rather than nominally. None of which is a reason to stay away — it is a reason to be careful and well prepared. Where I have worked here, it has been SMR/beta with theta inhibition, qEEG-guided, in long supervised courses. The line a clinician has to hold with the client: every single study is an add-on to antipsychotics. Standalone efficacy has never been tested, so it cannot be offered.
Clamor et al. (2016) A randomized-controlled trial of heart rate variability biofeedback for psychotic symptoms doi:10.1016/j.brat.2016.10.003
Sürmeli et al. (2012) Schizophrenia and the efficacy of qEEG-guided neurofeedback treatment: A clinical case series doi:10.1177/1550059411429531
Duan et al. (2025) Systematic review and meta-analysis of the effects of EEG neurofeedback combined with pharmacological treatment on the positive and negative symptoms in patients with schizophrenia doi:10.3389/fpsyt.2025.1537329
Li et al. (2024) Neurofeedback technique for treating male schizophrenia patients with impulsive behavior: a randomized controlled study doi:10.3389/fpsyt.2024.1472671
Narang et al. (2026) Does improvement in attention through theta/beta neurofeedback training result in better entry-level vocational task performance in individuals with schizophrenia? An open-label randomised controlled trial doi:10.1016/j.ajp.2026.104880
Schilirò et al. (2026) Neurocognitive Effects of Neurofeedback: A Systematic Review of the Applicability and Therapeutic Effect in Patients with Schizophrenia Spectrum Disorders, Psychosis or Clinical High Risks for Psychosis doi:10.1007/s10484-026-09773-x
Bauer et al. (2025) Real-time fMRI neurofeedback modulates auditory cortex activity and connectivity in schizophrenia patients with auditory hallucinations: A controlled study doi:10.1016/j.pscychresns.2025.112050
