Exercise and Cognition: Part Three: The Long View

Abstract

Parts One and Two reported that exercise improves cognitive test scores, at effect sizes larger than anything in the supplement literature. This part asks the question those trials cannot answer.

Does being physically active across decades change how cognition ages? A 2024 systematic review and meta-analysis in JAMA Network Open pooled 104 studies with 341,471 participants to find out.

Its answer is weak association, no dose–response, and — in the authors' own summing up — postponement of cognitive decline at a population level, but only to a very small extent.

1.The Problem the Study Was Built Around

Paula Iso-Markku, Katja Waller and colleagues at Helsinki and Jyväskylä begin with a methodological objection to their own field.

Much of the evidence linking physical activity to cognitive outcomes comes from studies with short follow-ups, and short follow-ups are prone to reverse-causation bias.

The mechanism is straightforward and easy to miss. Cognitive decline begins years before diagnosis, and one of its early consequences is doing less. A person in the prodromal phase of dementia becomes less physically active because of the disease. Measure activity at baseline and cognition five years later, and low activity will predict decline — not because inactivity caused it, but because the disease caused both.

Their design addresses this by requiring at least a year of prospective follow-up and by testing whether follow-up length moderates the association. If reverse causation drives the effect, it should shrink as follow-up lengthens.

2.What They Found

A weak association between baseline physical activity and follow-up global cognition, evident also in episodic memory and verbal fluency.

Then the set of null moderator findings, which is where the real information sits. Neither study quality, nor follow-up length, nor baseline age, nor adjustment for preceding level of cognition moderated the association.

And there was no clear dose–response association between the amount of physical activity and global cognition.

The absence of a dose–response relationship is the most important negative here. In a genuine causal effect, more of the exposure generally produces more of the outcome. Its absence does not disprove causation — thresholds and ceilings exist — but it removes one of the strongest arguments for it.

3.Reconciling the Two Answers

So randomised trials show exercise improving cognitive test scores by a standardised mean difference of 0.42, while decades of observational data show physical activity weakly associated with cognitive trajectory and without a dose gradient. Both are large, careful syntheses. How do they fit?

The most likely reconciliation is that they measure different things. A trial measures performance on a test after an intervention — a state, improved while the intervention runs. The observational studies measure the slope of decline over years, which is a different quantity.

An analogy that holds reasonably well: exercise may raise how well the brain currently performs without much changing the rate at which it ages. Those are separable, and only the first is what a cognitive test in a three-month trial can detect.

The alternative reconciliation is less comfortable. Trials of exercise cannot be blinded, and unblinded trials with unblinded outcome assessment systematically overestimate effects on subjective and performance measures. Part Two flagged a standardised mean difference above 1.0 in that literature as implausible for the same reason.

We do not think either explanation is established, and the honest position is that both are probably contributing.

4.What This Does Not Say

It does not say exercise is unimportant. It says the association between physical activity and long-term cognitive trajectory is weak, and that anyone promising to prevent dementia through exercise is promising something this dataset does not support.

Exercise has among the strongest evidence in medicine for cardiovascular disease, type 2 diabetes, mortality, mood and functional independence in old age. None of that is diminished by a weak association with cognitive decline specifically.

The authors' phrasing is careful and worth adopting: at a population health level, and only to a very small extent. That is a statement about populations, which is what epidemiology produces, and it does not predict what happens to any individual.

5.Where Three Parts Leave It

Across 2,724 randomised trials, exercise improves general cognition, memory and executive function, most in children and adolescents and in people with ADHD, with a pattern of larger effects in shorter and gentler interventions that may reflect the literature rather than the biology.

Across 71 trials in cognitive impairment, every modality beats nothing, resistance training ranks first overall at an effect size large enough to distrust, and multicomponent programmes rank first in the milder population.

Across 104 prospective studies and a third of a million people, being active is weakly associated with cognitive trajectory, without a dose gradient.

That is the best-evidenced non-pharmacological lever available for cognitive performance, stated accurately. It is still much better evidenced than anything in a bottle.

Editorial Comment

MindHeaven® makes no claim relating to exercise or physical activity, and nothing we sell substitutes for either.

We wrote this series because a customer asking how to think more clearly deserves the evidence ranked honestly, and honest ranking puts exercise above every compound in our catalogue. We also wrote Part Three because ranking it honestly means not overselling it either — and "exercise prevents cognitive decline" is a claim this dataset weakens rather than supports.

If cognitive ageing is the concern, the interventions with the strongest evidence remain the ones the Lancet Commission identified and which we cover separately: hearing, blood pressure, smoking, alcohol, education, social contact. Exercise belongs on that list. It does not lead it.

And none of it is a reason to skip a conversation with a doctor if memory changes have become noticeable, because some causes of cognitive decline are treatable and identifying them matters more than any of this.