Brain training that holds up under peer review
The cognitive-fitness practices that actually have evidence behind them — and a skeptical look at the ones that don't.
What worksAbout
NeuroGym Club is a reference for cognitive fitness curated against the published neuroscience and behavioral-science literature. The commercial brain-training industry is large and largely unregulated; most consumer brain-game apps have struggled to show transfer to general cognition outside the tasks the user practiced. We focus on the practices that have replicated under peer review — physical exercise, sleep, novel motor learning, deliberate practice — and we call out the ones that haven't. Not medical advice.
What the evidence shows
The strongest evidence for protecting and improving cognitive function across the lifespan does not come from brain-training apps. It comes from a small set of practices that have replicated across many randomized trials and population studies:
- Aerobic and resistance exercise — meta-analyses (Northey et al., 2018; Falck et al., 2019) report meaningful effect sizes on executive function and processing speed in older adults, with dose-response relationships.
- Sleep — both quantity (7–9 h for most adults) and quality. Slow-wave sleep is required for declarative memory consolidation; REM is involved in procedural and emotional learning.
- Novel motor and cognitive learning — learning a new instrument, language, or complex skill drives structural and functional plasticity in a way that practicing the same task does not.
- Cardiovascular and metabolic health — vascular disease, type 2 diabetes, and uncontrolled hypertension are among the largest modifiable risk factors for late-life cognitive decline (Lancet Commission on Dementia Prevention, 2024 update).
- Social engagement — observational data link social isolation to dementia risk; experimental data on intervention are thinner but consistent.
None of these is a hack. They are unglamorous and replicable.
Daily practices with the strongest support
- 150+ minutes per week of moderate aerobic activity (the same target as for cardiovascular health). Higher-intensity intervals show particular promise for executive function.
- Two sessions per week of resistance training. Independent of aerobic effect, resistance training shows cognitive benefit in trials such as the SMART study.
- Consistent sleep schedule — same wake time across days protects sleep architecture more than total hours alone.
- Deliberate practice on a meaningful skill — language, music, chess, programming, sport — where you operate at the edge of your current ability with feedback.
- Treat what's treatable — hearing loss is one of the largest single modifiable risk factors for dementia; correct it.
What replicates, briefly
Five intervention categories with the most robust cognitive evidence
-
Aerobic exercise
Multiple meta-analyses across age groups. Effect on executive function and processing speed; smaller effect on memory. Dose-response above 150 min/week.
-
Resistance training
SMART trial and follow-ups show independent cognitive benefit, particularly for executive function in older adults. Twice weekly is the typical dose.
-
Sleep quality
Consolidation of declarative memory depends on slow-wave sleep. Sleep restriction below 6 hours impairs working memory and attention measurably.
-
Novel skill acquisition
Structural MRI shows gray-matter change with learning a new motor or language skill. Practicing what you already know does not produce the same effect.
-
Treating hearing loss
Lancet Commission (2024 update) ranks untreated mid-life hearing loss as the single largest modifiable dementia risk factor. Hearing-aid trials (ACHIEVE, 2023) show cognitive benefit in at-risk groups.
Skeptic's guide to brain-training products
Commercial brain-training apps — Lumosity, Elevate, Peak, BrainHQ, CogniFit, and many others — typically claim to improve general cognition. The peer-reviewed evidence base is mixed at best, and the claims have been the subject of regulatory action.
- Lumosity / Lumos Labs (2016 FTC settlement): the FTC ordered Lumos Labs to pay $2 million and barred the company from making unsubstantiated claims about cognitive improvement, age-related decline, and academic performance. The case is the standard cautionary citation for the field.
- 2014 Stanford / Max Planck consensus statement: ~70 scientists signed a public statement that the brain-training industry's claims were not supported by the literature, prompting a counter-statement from another group of researchers. The dispute is itself instructive — the question is not settled the way marketing pages suggest.
- Transfer is the central problem. Most apps measurably improve performance on the task practiced ("near transfer") but do not reliably produce gains in untrained, real-world cognition ("far transfer"). Far transfer is what consumers actually want.
- Dual n-back as a working-memory intervention enjoyed a wave of attention after Jaeggi et al. (2008). Replication attempts have been mixed; meta-analyses (Soveri et al., 2017) find modest near-transfer and limited far-transfer.
None of this means brain-training apps are useless — practicing reaction-time tasks does make you faster at reaction-time tasks, which is genuinely satisfying. It means the strong general-cognition claims do not hold up.
Frequently asked
Does Lumosity work?
It depends what you mean by "work." Practicing the games makes you better at the games. The FTC sanctioned Lumos Labs in 2016 for making unsubstantiated claims that the games improved general cognition, slowed age-related cognitive decline, or improved academic performance. The peer-reviewed literature has not since closed that gap.
What about dual n-back?
Dual n-back was popularized by Jaeggi and colleagues (2008) as a working-memory training task that might transfer to fluid intelligence. The replication record is mixed; recent meta-analyses find modest near-transfer effects and limited far-transfer. If you enjoy it, it is unlikely to be harmful. As a path to becoming generally smarter, the evidence is weaker than marketing suggests.
Is exercise really better than brain games?
Yes, when measured by cognitive outcomes in trials. Aerobic and resistance exercise have shown reproducible effects on executive function, processing speed, and (in older adults) memory. The effect sizes are not enormous, but they replicate, and they show dose-response. No commercial brain-training app has matched that profile.
What about meditation or mindfulness apps?
Mindfulness training has reasonable evidence for stress reduction and attention regulation; the cognitive-enhancement claims are more modest. Independent reviews tend to find smaller and noisier effects than app marketing suggests. The non-cognitive benefits (stress, mood, sleep onset) are better-supported than the cognitive ones.
Why do I keep seeing 'neuroplasticity' in marketing?
Because neuroplasticity is real — the brain does adapt to use across the lifespan — and that real phenomenon is used to back claims that the brand's specific product harnesses it. Real neuroplasticity does not require a subscription. Novel skill acquisition, learning a language, picking up a musical instrument, or training for a sport all engage the same plastic mechanisms.
Want the primary sources?
Start with the FTC v. Lumos Labs press release, the Stanford/Max Planck 2014 consensus statement, and the Lancet Commission on Dementia Prevention (2024 update).
Recent Posts
-
Near transfer vs far transfer, plainly
Why getting better at a brain game is not the same as getting smarter — the transfer distinction that decides whether training claims hold up.
-
What the exercise meta-analyses actually measured
Behind the claim that exercise helps cognition: which tests moved, who was studied, and how to read an effect size without marketing glasses.
-
Is dual n-back worth your time?
Jaeggi 2008 made dual n-back famous. The replication and meta-analytic record is more modest. Who might still practice it — and who should pick something else.