Exercise and Mental Health: What the Evidence Shows

Exercise reliably improves mood and reduces symptoms of depression and anxiety, with an effect size that in short-term trials sits in the same range as some psychological treatments. The evidence behind that sentence is also messier than the headlines suggest: most trials are small, almost none are properly blinded, and publication bias is a live concern. The one thing that is not in dispute is the boundary — in moderate to severe depression, physical activity supports treatment rather than replacing it.

Start with the numbers#

Every few months a headline announces that exercise beats antidepressants. Here is what sits underneath it.

The Cochrane review "Exercise for depression", updated in 2026, included 73 randomised trials and at least 4,985 participants. Against no treatment or a control intervention, exercise reduced depressive symptoms with a standardised mean difference of −0.67 (95% CI −0.82 to −0.52). On the conventional reading, that is a medium-to-large effect.

The next sentence in the abstract is the one that rarely makes it into coverage. When the reviewers restricted the analysis to the seven trials that had adequate allocation concealment, intention-to-treat analysis and blinded outcome assessment, the pooled effect dropped to −0.46 (95% CI −0.88 to −0.04) — a confidence interval that nearly touches zero. The pattern is familiar across medicine: the tighter the methods, the smaller the result.

Head-to-head comparisons look like this. Ten trials (414 participants) compared exercise with psychological therapy and found essentially no difference (SMD 0.03). Five trials (330 participants) compared it with antidepressants and again found no meaningful difference (SMD −0.11). "As good as therapy" is a defensible statement — but it rests on fifteen small trials, not on a mountain of data.

Comparison	Trials	Effect (SMD)	Certainty
Exercise vs no treatment or control	57 (2,189 people)	−0.67	Low
Same, restricted to well-conducted trials	7 (447 people)	−0.46	Lower; interval nearly crosses zero
Exercise vs psychological therapy	10 (414 people)	0.03 — no difference	Moderate
Exercise vs antidepressants	5 (330 people)	−0.11 — no difference	Low
Effect at long-term follow-up	9 (405 people)	−0.53, interval includes zero	Very low

Why to hold these numbers loosely#

The trials are not badly done so much as structurally hard to do well.

Nobody can be blinded. A participant always knows whether they spent twelve weeks in a gym or on a waiting list. Every trial in the Cochrane review was judged at high risk of performance bias, and the primary outcome is usually a self-report depression scale — precisely the kind of measurement that expectation moves most.

The comparator is usually passive. Against a waiting list, the exercise group gets more than exercise: structure, a place to be, contact with a supervisor, and the sense that something is being done about the problem. Some of the measured benefit belongs to those non-specific ingredients.

Samples are small. The median trial in this field enrols a few dozen people. Small trials scatter, and the ones that scatter upward are the ones that get written up.

Durability is unclear. Cochrane rated the long-term follow-up evidence as very low certainty, with a confidence interval spanning zero. Whether the improvement survives once the programme ends has not been established.

The largest synthesis makes the same point from the other direction. The 2024 BMJ network meta-analysis by Noetel and colleagues pulled together 218 studies and 14,170 participants — and reported that exactly one of them met the Cochrane criteria for low risk of bias. Confidence was graded low for walking or jogging and very low for everything else, though the results did appear robust to tests for publication bias.

The boundary that matters most#

Physical activity is not a substitute for treatment in moderate or severe depression. No dataset supports telling someone to stop their medication and take up running. The pharmacological comparison amounts to five trials and just over 300 mostly mild-to-moderate cases, and "no difference detected" at that sample size is not the same claim as "no difference exists".

There is also a practical problem specific to severe illness: the condition itself removes the capacity to act. Anergia, anhedonia and psychomotor slowing are symptoms, and asking someone in that state to train four times a week is close to asking them to cheer up. Exercise becomes realistic once treatment has restored enough function for action to be possible at all.

The defensible framing: movement is an addition that strengthens primary treatment, improves sleep and physical health, and offsets some medication side effects. Not an alternative.

Mechanisms, minus the mythology#

The endorphin problem#

"Endorphin rush" is the most repeated and least supported explanation on offer. It dates to 1980s observations of raised beta-endorphin in runners' blood, and it has two serious problems.

Beta-endorphin is a large molecule that crosses the blood-brain barrier poorly, so a blood measurement says little about the brain. And when researchers blocked opioid receptors with naltrexone, the post-exercise mood improvement largely persisted. If endorphins were doing the work, it should have disappeared.

The candidates that have replaced it#

Endocannabinoids. Anandamide rises after moderate aerobic exercise, these molecules cross into the brain readily, and their receptors are involved in anxiety and pain regulation. This is currently the more plausible account of the "runner's high".

BDNF. Brain-derived neurotrophic factor supports neuronal survival and synaptic plasticity, particularly in the hippocampus. It is reduced in depression and rises with regular aerobic training. The biology is well described, though the link between circulating BDNF and how someone actually feels remains indirect.

Lower-grade inflammation. A subset of people with depression have elevated inflammatory markers, and regular activity lowers them. This may explain some of the heterogeneity in response — larger benefits in the inflammatory phenotype, smaller elsewhere.

Behavioural activation. The least glamorous mechanism and possibly the most important. Training imposes structure: leave the house, be around people, finish something. That is precisely what behavioural activation does, and behavioural activation is an evidence-based depression treatment in its own right. A meaningful share of the "exercise effect" may not be physiological at all.

Self-efficacy. Watching your own capacity increase — the same route feels easier, the same weight moves faster — directly contradicts the depressive belief that you are incapable. It is a measurable predictor of improvement, not a motivational flourish.

Dose: what the data support#

WHO recommends 150–300 minutes of moderate aerobic activity per week, or 75–150 minutes of vigorous, plus muscle-strengthening work on two or more days. That is a general health target. For mood specifically, the emphasis shifts.

Frequency outranks intensity and volume. Mood is a day-to-day variable, and short sessions spread across four or five days do more than one long session of equal total volume. A workable target is three sessions minimum, four to five ideally, of 20–40 minutes.

Walking counts. In the Noetel network meta-analysis, walking or jogging produced the largest effect of any modality (g = −0.62, 51 trials). For someone who has not exercised in years, that means no gym membership, no equipment and no learning curve stand between them and the best-supported option.

Intensity adds something, at a cost. Effects in that analysis were proportional to prescribed intensity. But harder programmes are abandoned more often, and an abandoned programme has an effect size of zero. Moderate intensity — where you can talk but not sing — is the sensible trade.

Give it four weeks. A single session can lift mood for a few hours; a durable shift takes consistency. Judging the experiment before a month of regular sessions tells you nothing.

Читайте также: How Many Steps a Day Do You Actually Need?

Resistance training deserves its own paragraph#

For years "exercise for depression" implicitly meant cardio. That has changed.

In the 2024 network meta-analysis, strength training produced an effect of −0.49 on depression across 22 trials and 643 participants — comparable to yoga (−0.55) and mixed aerobic programmes (−0.43). The more useful finding is acceptability: strength training and yoga were the best-tolerated modalities, meaning people stayed with them. Since adherence is the binding constraint in real life, tolerability may matter more than a marginal difference in effect size.

Resistance work also suits anxious people: it delivers visible progress without long stretches of elevated heart rate, which is often the sticking point.

The minimum that does the job: two sessions a week, five or six compound movements, two to three sets of 8–12 repetitions. No advanced programming required.

Читайте также: Strength Training for Beginners: Your First 3 Months

Exercise and anxiety, including the awkward part#

The 2023 umbrella review by Singh and colleagues in the British Journal of Sports Medicine synthesised 97 systematic reviews covering 1,039 trials and 128,119 participants. Median effect sizes were −0.43 for depression, −0.42 for anxiety and −0.60 for psychological distress. Anxiety responds roughly as well as depression does.

What that summary omits is that for some people the first few weeks go the other way.

Vigorous effort produces a pounding heart, breathlessness, sweating and light-headedness on standing. Those are the exact bodily sensations of a panic attack. For someone with high anxiety sensitivity — a tendency to interpret bodily arousal as dangerous — a hard session can read as an impending catastrophe rather than a workout. This is real, it is common, and it is the main reason anxious people quit in week two.

The resolution is slightly paradoxical: repeated safe exposure to those sensations is itself therapeutic. Experiencing a racing heart repeatedly in a context where nothing bad happens is interoceptive exposure, a standard component of CBT for panic. The mechanism only works if the dose is graded.

The two-way street#

Observational data are consistent: active people develop depression less often. Schuch and colleagues, writing in the American Journal of Psychiatry in 2018, pooled 49 prospective cohorts covering 266,939 people and 1,837,794 person-years. People with high activity levels had 17% lower odds of developing depression (adjusted OR 0.83), with protective associations holding in adolescents, adults and older people, and across Asia, Europe, North America and Oceania.

The temptation is to read that as "move and you will not get depressed". But causation runs in both directions. Depression reduces activity: low energy, loss of interest, disrupted sleep and slowing are symptoms of the illness, not consequences of a sedentary lifestyle. Part of any observed correlation is reverse causation — activity dropping in the months before a diagnosis is made, not before the illness began.

The meta-analysis authors addressed this by excluding early-onset outcomes, and the association survived. It cannot be fully resolved in an observational design, though. The honest reading is that activity probably does lower risk, and that the size of its contribution is smaller than raw correlations imply.

There is an unexpectedly kind corollary. If you find it nearly impossible to make yourself move while depressed, that is a symptom of the condition rather than a defect of character — which is exactly why the entry threshold has to be set absurdly low.

A starting protocol for when you have nothing#

"150 minutes a week" is useless advice for someone who struggles to get out of the door. The working goal at that point is not accumulating volume; it is breaking immobility.

1. Ten minutes, not thirty. Walk out for ten minutes and come back. The target is the act of leaving, not the session. A bar you cannot fail to clear beats a correct bar you cannot clear at all.
2. Same time, same route. Deciding whether to go consumes a resource you do not have. Remove the decision: fixed slot, familiar path.
3. No measurement. No pace, no step counts, no calories in the first weeks. Numbers turn the walk into another task that can be failed.
4. Daylight instead of intensity. A morning walk outdoors pairs movement with light exposure, which supports circadian timing and, through it, sleep and mood.
5. Track attempts, not outcomes. Record only went / did not go. After a month that tally becomes the main source of the feeling that you are managing.
6. A missed day is a missed day. It is not the end of the programme. Returning tomorrow matters more than an unbroken streak.

Review after four weeks, and not by asking how you feel today. Ask three concrete questions: is falling asleep easier, is irritability less frequent, has the ability to concentrate returned even briefly.

The bottom line#

Physical activity is one of very few interventions with a measurable effect on mood, a short list of side effects, and substantial collateral benefits for sleep, cardiovascular health and metabolism. That combination is enough to recommend it to almost everyone.

It is equally true that the effect shrinks in the best-conducted trials, that durability after a programme ends is poorly established, and that in significant depression it works with treatment rather than instead of it. The defensible position is to treat movement as a dependable addition to things already known to work — and to give "exercise instead of medication" the same scepticism you would give any simple answer to a complicated question.

Читайте также: How to Manage Stress: Evidence-Based Techniques

Читайте также: Sleep Hygiene: 12 Rules That Actually Work