Exercise for Schizophrenia: Real Cognitive Gains, Real Limits

Exercise can improve cognition in schizophrenia. That is a bigger deal than it sounds.

Schizophrenia is one of the harder places to find clean wins. Cognitive symptoms are stubborn. Negative symptoms are stubborn. The drugs that help psychosis often bring weight gain, insulin resistance, and sedation to the party like very bad guests. So when a 2026 meta-analysis by Greca et al. pooled 17 randomized controlled trials and found that structured exercise improved cognition, symptoms, quality of life, and social functioning, that was not a cute wellness side note. It was a serious result.

The pooled effect sizes were moderate to large by psychiatric standards: cognition improved with a standardized mean difference of 0.59, overall symptom severity with an SMD of 0.71, quality of life with an SMD of 0.60, social functioning with an SMD of 0.55, and overall well-being with an SMD of 0.68. For well-being, the reported 95% confidence interval was 0.43 to 0.93, with p<0.001. The review used a PRISMA framework and random-effects modeling across trials conducted from 2009 to 2024. That does not make exercise a miracle. It does make it hard to dismiss.

The practical point is simple: structured exercise is now an evidence-based adjunctive treatment for schizophrenia spectrum disorders. Adjunctive matters here. These studies did not show that exercise replaces antipsychotic treatment. They show it can add measurable benefit on top of usual care.

What the evidence actually says

The strongest anchor is still Greca et al. 2026, because it synthesizes 17 RCTs rather than betting everything on one small trial. An independent 2025 scoping review by Vila-Barrios et al. points in the same direction across psychopathology, cognition, and quality of life. That kind of replication does not eliminate bias, but it does move the field past pure pilot-study optimism.

The largest individual trial in this set is the multicenter adjunctive exercise trial by Maurus et al. 2023, which tested aerobic endurance training in people with schizophrenia. The abstract available here does not provide the domain-specific cognitive numbers, so it cannot carry the whole argument by itself. Its value is different: it shows that exercise protocols can be run across multiple centers rather than only in one unusually motivated lab.

A companion methodology paper by Maurus et al. 2021 explains why this matters. The authors described the older exercise literature in schizophrenia as small, single-center, and heterogeneous. The new meta-analytic signal looks more convincing precisely because it sits on more trials and a more mature evidence base than the field had a few years ago.

Mode and intensity are not interchangeable

One mistake in exercise reporting is to talk as if all movement works the same way. It does not. A 2024 meta-analysis by Rißmayer et al. specifically examined exercise intensity, mindfulness components, and other moderators in psychotic disorders. Their review makes a useful corrective point: exercise is not one intervention. Aerobic training, combined aerobic-plus-resistance programs, and other formats may not produce the same pattern of effects across symptoms, functioning, and cardiometabolic health.

That matters because schizophrenia care has two overlapping targets. One is the illness itself: cognition, symptoms, daily functioning. The other is the metabolic damage that often comes with treatment. Second-generation antipsychotics can drive weight gain, dyslipidemia, and insulin resistance. If one exercise protocol helps cognition while another does more for cardiometabolic health, the choice is not cosmetic. It is part of the prescription.

The 17-trial pooled results in Greca et al. should therefore be read as an average across heterogeneous programs, not as proof that any random gym plan will do. Average effects are useful. They are not a training manual.

The mechanism is plausible. It is not nailed down.

The main biological story in this literature is that exercise may increase brain-derived neurotrophic factor, or BDNF, while also reducing inflammatory signaling such as IL-6, TNF-alpha, and C-reactive protein. That framework appears in Greca et al. 2026, the narrative review by Makris et al. 2021, and the trial design paper by Maurus et al. 2021. The proposed downstream effect is better hippocampal neuroplasticity, which makes sense given the cognitive deficits and hippocampal abnormalities associated with schizophrenia.

But plausible is not the same as measured. The available evidence here does not report schizophrenia-specific BDNF change values or hippocampal volume changes tied directly to the clinical improvements. So the mechanism is a reasonable working model, not confirmed proof of causation. Readers should keep that distinction clean. Behavioral benefits are the main result. The biomarker story still trails behind.

Exercise helping a damaged system does strengthen the case that it supports brain function more generally. It does not prove that the same effect size will show up in healthy people with normal baseline cognition.

That last point matters because the temptation to over-extrapolate is strong. If structured exercise improves cognition in a population with severe psychiatric illness, that does make exercise-as-brain-tool harder to wave away. Still, the data reviewed here come from schizophrenia populations, not healthy adults, and effect sizes in people with intact baseline function will probably be smaller.

The catch is not efficacy. It is adherence.

If the literature had a villain, it would not be lack of benefit. It would be implementation. A protocol paper by Martland et al. lays out the real-world barriers in inpatient serious mental illness settings: people with schizophrenia often have very low baseline activity levels, antipsychotic-related sedation, antipsychotic-related weight gain, cognitive deficits, and motivational impairments. Those are not side issues. They directly interfere with the ability to start and sustain exercise.

This is where the nice pooled SMDs run into actual human life. Trials usually retain the people who can keep showing up. If the participants who complete exercise RCTs are more adherent, less sedated, or less impaired than the broader clinical population, trial efficacy will overshoot routine-care effectiveness. The available abstracts do not report the adherence and dropout numbers needed to judge how large that gap may be. That omission is a real limitation, not paperwork trivia.

There are other reasons to stay cautious. In the Greca et al. abstract, confidence intervals are reported for only one of the five main pooled outcomes. The well-being interval alone spans 0.43 to 0.93, which is compatible with a useful effect but also tells you the estimates are not razor-precise. The remaining four key outcomes in the abstract do not include confidence intervals, so the precision of those pooled estimates cannot be fully assessed from the published summary.

Subgroup findings also need a leash. The meta-analysis reported stronger apparent responses in males and in the 36 to 45 age group. Interesting, yes. Ready for clinical precision prescribing, no. Those look exploratory, and the available evidence does not test a mechanism for them or rule out confounding by medication type, illness duration, or baseline metabolic burden.

Exercise is not the only non-drug option

Exercise also should not be sold as the lone clever intervention in the room. A 2025 systematic review by Stratou et al. found that occupational therapy interventions can improve cognitive, social, and functional outcomes in schizophrenia. A 2025 pilot follow-up study by Golas et al. reported persistent cognitive gains after cognitive remediation therapy in late-life schizophrenia.

What is missing are head-to-head randomized comparisons of exercise versus cognitive remediation therapy or occupational therapy for cognitive outcomes. Without those trials, nobody can honestly claim that exercise is uniquely superior. It has a good evidence base. So do other non-pharmacological approaches, at least in narrower contexts.

What should readers do with this?

For schizophrenia care, the conclusion is stronger than many people realize. Structured exercise looks effective enough to treat as standard adjunctive care, especially because its side-effect profile points in the opposite direction from many pharmacologic add-ons. It can improve symptoms while helping the cardiometabolic problems that antipsychotics often worsen. That is a rare alignment in psychiatry.

But the prescription cannot be lazy. Programs need supervision, low-friction entry, progression that matches a patient's capacity, and support for motivation and attendance. Otherwise the intervention that works on paper turns into the intervention that lives forever in a discharge summary.

For healthy readers chasing better cognition, the schizophrenia literature supports plausibility, not a direct estimate. The honest lesson is not that exercise turns everyone into a genius. It is that one of the most treatment-resistant conditions in medicine still shows measurable cognitive benefit from structured exercise. That should lower the burden of proof for taking exercise seriously as a brain intervention. It should not license inflated promises about how much it will do in people who are already well.

The evidence is good enough to retire one old assumption: exercise in schizophrenia is not just general wellness advice wearing a lab coat. It is treatment. The remaining question is whether clinics can deliver it in a form patients can actually stick with.