What the exercise meta-analyses actually measured

“Exercise is good for the brain” is true enough to be a slogan and vague enough to hide what the studies actually did. This post opens the box on the meta-analyses cognitive-fitness readers keep seeing cited — especially Northey et al. (2018) — and translates their outcomes into plain language.

Who was in the main synthesis?

Northey and colleagues (2018) reviewed exercise interventions for cognitive function in adults older than 50, pooling randomized and controlled trials into a meta-analysis published in the British Journal of Sports Medicine. The headline result: physical exercise improved cognitive function in this age group, including across different baseline cognitive statuses in the included literature.

Source (confirmed 2026-07-21): PubMed 28438770 · BJSM.

That population frame matters. Results in healthy older adults do not automatically export to every teenager, every clinical dementia stage, or every elite athlete protocol.

What “cognitive function” meant in practice

Meta-analyses do not measure a single magical Brain Score. They average effects across neuropsychological tests and composite domains. Domains that frequently appear:

Executive function — inhibition, switching, planning (trail-making B, Stroop-like tasks, verbal fluency variants, etc.).
Processing speed — how quickly you complete simple symbol or cancellation tasks under time pressure.
Memory — immediate and delayed recall, recognition; often smaller or noisier effects than executive/speed in exercise literature.
Global / composite scores — mini-mental style screens or averaged batteries in some trials.

When a paper says “exercise improved cognition,” ask: which domain drove the average? Executive and speed effects are often the most consistent in older-adult exercise trials; memory is not always the star.

Dose patterns that show up

Across effective protocols in this literature, common ingredients include:

Session lengths often in the 45–60 minute neighborhood.
Moderate to vigorous intensity for aerobic work.
Both aerobic and resistance modalities appearing in positive trials.
Multi-week interventions (acute one-shot exercise studies answer different questions).

Public-health targets of ≥150 minutes/week moderate aerobic activity plus two muscle-strengthening days remain a practical scaffold that lines up with both cardiovascular guidelines and much of the cognitive trial dosing — not because 149 minutes is useless, but because round public numbers aid adherence messaging.

Effect sizes without the hype

A small-to-moderate average effect that replicates is more valuable than a huge effect from one unreplicated lab. For a reader:

You should not expect exercise to feel like a stimulant every morning.
You should expect benefits that accumulate with weeks of training and that sit alongside heart, metabolic, and mood benefits.
You should not read a domain-average improvement as “dementia cancelled.”

Resistance training is not a footnote

Separate from pure aerobic meta-results, progressive resistance training trials (including programs such as SMART and related work) report executive benefits in older adults with roughly twice-weekly dosing. If your plan is only walking, you are still doing something useful — and you may still be leaving a supported modality on the table.

What these meta-analyses did not measure

Far transfer from a smartphone game (different literature).
Guaranteed personal protection from Alzheimer’s disease.
Optimal exact heart-rate zone for every human.
Psychedelic or clinic-based interventions (out of scope for this site).

Practical takeaway

If you want the intervention class with the cleanest cognitive trial/meta-analytic footprint among lifestyle levers, structured exercise still leads commercial brain-training as a category. Read the long form on Exercise and cognition, then compare the weaker far-transfer record for apps on Do brain-training apps work?.

Heterogeneity: why two trials disagree

Even within exercise–cognition research, trials differ on:

Age and baseline cognition (healthy aging vs mild impairment)
Supervised gym sessions vs home walking logs
Which executive tests were chosen
Duration (8 weeks vs 12 months)
Whether the control group socialized (social contact alone can move some scores)

Meta-analyses average across that mess. They can conclude “exercise helps on average” while a single underpowered trial finds nothing. That is not a conspiracy; it is statistics. For personal decisions, prefer the direction of the literature plus the non-cognitive benefits of exercise over any one news story.

Mechanisms without magical thinking

Researchers discuss BDNF, angiogenesis, reduced inflammation, and improved sleep as candidate pathways. Those are mechanistic hypotheses. They explain how benefits might occur; they do not replace outcome trials. Marketing that says “boosts BDNF by 200%” without a clear human cognitive endpoint is selling a biomarker story, not a cognition story.

How this should change your week (not your identity)

You do not need to become a marathoner. A realistic shift:

Add two resistance sessions if you currently do zero.
Convert two sedentary evenings into brisk 30-minute walks.
Protect sleep so training adaptations stick (sleep and memory).

Reassess in eight weeks using life metrics you care about (energy, mood, ability to focus at work) rather than an in-app leaderboard. For the full page with sources, see Exercise and cognition.

Exercise and cognition — full pillar with Northey 2018 links.
Modifiable dementia risk factors — physical inactivity in the life-course model.
What the exercise meta-analyses actually measured is the short form; the pillar is the long form.

Northey et al. 2018 PubMed record confirmed 2026-07-21.

This content is for educational purposes only and does not constitute medical advice. Check with a clinician before changing exercise habits if you have medical conditions.

This post was drafted by AI and reviewed by our editorial team. Sources checked 2026-07-21.