Cognitive engagement

Most of what's sold as "brain training" doesn't work outside the app it's played in. What holds up is education, complex work, and learning genuinely hard new things — and even that evidence is thinner than the marketing suggests.

Value education and lifelong learning, choose demanding work, keep learning things that are new and hard for you, and stay socially connected — ideally all at once, as in a group class. Skip brain-training apps. With the partial exception of education, none of this is established as preventing dementia; treat it as a set of cheap, sensible bets.

Reserve is the idea that holds the evidence together

The brain can carry a lot of damage before it shows. Autopsy studies routinely find extensive Alzheimer's-type damage in people who had no symptoms during life, and little damage in people who clearly had dementia. Cognitive reserve is the name for the spare capacity in between: lifelong intellectual challenge is thought to build redundant pathways that let the brain keep working as damage accumulates.

The 2024 report of the Lancet Commission on dementia, the field's benchmark synthesis, estimates that around 45% of dementia cases are potentially preventable by addressing fourteen risk factors. It lists less education as the single early-life factor, accounting for roughly 5% of cases, with reserve as the stated mechanism.[1] The Commission gives brain training, crosswords and hobbies no risk factor of their own and found only low-quality evidence of small, short-term gains from computerised training. It does advise staying cognitively, physically and socially active through midlife and later life, and notes that mentally demanding work keeps up that stimulation far longer than a course or a hobby does.

A 2024 meta-analysis of 27 long-term studies estimated how much reserve each stage of life contributes. Comparing people high and low on each marker, dementia risk was about 18% lower for more education (birth to 30), about 9% lower for more complex work (30 to 60), and about 19% lower for more cognitive activity and social connection (over 60).[2] The stages connect: pooling 10,195 people in six countries, about 28% of education's protective effect ran through the more complex jobs it led to, with a threshold around finishing high school.[3] This is moderate evidence, and all of it observational.

Reserve delays symptoms; it doesn't stop the disease. Once the spare capacity runs out, decline tends to be faster in people who had more of it, because more damage built up before symptoms appeared. The net effect is still good — more years of full function, fewer of severe impairment — but reserve buys time, not immunity.

Why so much of this evidence could run backwards

Dementia has a silent phase lasting years, and it quietly cuts how much people do. So "less engagement before dementia" can simply be early dementia reducing engagement. In 28 years of follow-up of 10,308 British civil servants, physical activity began falling up to nine years before a dementia diagnosis, and there was no link at all between activity and dementia risk; the authors concluded earlier findings likely reflected this reverse causation.[4]

The same thing shows up for mental activity. In a UK study of more than 850,000 women averaging 60 years old, not taking part in adult-education classes, arts and crafts groups or voluntary work was linked to more dementia only in the first decade of follow-up; in the second decade the link had largely vanished.[5] That matters for the late-life figure above, which pooled some studies with only a few years of follow-up.

Not every study dissolves this way. Following 469 adults over 75 for a median of five years, each extra day a week of reading, board games, playing a musical instrument or dancing went with about 7% lower dementia risk, and the link survived excluding people who may already have had early dementia.[6] A 2025 French study counted only dementia diagnosed after a seven-year gap, among 3,326 people aged 65 and over, and still found that moderate or high engagement in mental leisure activities — crosswords and artistic hobbies stood out — went with about half the risk, after accounting for past occupation and the APOE ε4 risk gene. High engagement looked no better than moderate.[7] Both are still observational, so whether late-life activity itself protects remains open.

The same caution applies to lifespan. In 10,477 Chinese adults with cognitive impairment and a median age of 95, doing three regular mental activities went with about 33% lower death rates than doing none, with a graded pattern in between.[8] In a population that old and already impaired, reverse causation — failing health curtailing activity — could explain much of it, so it is weak evidence and no proof that hobbies extend life.

Brain-training apps: better at the app, not in life

The most-marketed claim in this space — that daily play with digital puzzles improves general cognition or wards off dementia — does not survive scrutiny. In 2016 the makers of Lumosity agreed to pay $2 million to settle US Federal Trade Commission charges that they advertised these benefits without scientific backing.[9]

Brain-training apps produce small benefits right after training, with no evidence that they last. A Cochrane review of twelve weeks or more of computerised training in healthy older adults found low-quality evidence of small immediate gains, of uncertain clinical importance, and no evidence that the effect on overall cognition persisted twelve months later; only one small trial of 77 people looked that far.[10] The core problem is far transfer: people get reliably better at the game they play, but there is little evidence the gain spreads to broad mental abilities or to handling everyday life. A careful review of both sides of the scientists' dispute agreed: training improves the trained task, sometimes a close relative, but no study in the industry's own evidence base was designed well enough to show everyday benefit.[11] This skeptical conclusion is moderately well supported.

The famous exception is weaker than it sounds. In the ACTIVE trial, 2,832 healthy adults aged 65 to 94 were randomised to memory, reasoning or visual speed-of-processing training, or to no contact.[12] After 20 years, speed trainees who also did booster sessions had about a quarter lower rate of dementia diagnosis than controls — the source of "brain training cuts dementia by 25%". But boosters were drawn only from trainees who had stuck with the first sessions, so that comparison is not randomised. Speed training without boosters showed nothing, and in the fair comparisons of each whole training group against control, none significantly reduced dementia.[13] It is one unreplicated, preliminary result, and the protocol is now sold commercially.

ACTIVE also carries a possible harm. Training had no effect on 20-year mortality overall, but among participants with 12 years of education or less, memory and speed training were linked to higher mortality (about 60% higher); the estimates are imprecise and unreplicated.[14]

Hard new learning looks better than familiar pleasures

What seems to count is effortful, novel learning. In a small trial, adults aged 60 to 90 spent about 16 hours a week for three months either learning entirely new skills (digital photography, quilting), doing familiar things (music, familiar games, films), or socialising about familiar topics. Episodic memory improved in the new-skill groups compared with both familiar-activity groups.[15] The dose that worked is the kind that frustrates you and makes you feel slow for a while. With small groups, three months and one main outcome, it shows a plausible mechanism, not that familiar or social activity is worthless.

The field's central skeptic is worth keeping in view. Reviewing the whole "use it or lose it" idea, Salthouse found very few studies showing that mentally active people actually decline more slowly, and little support for the claim that mental activity changes the rate of mental aging.[16] The trials on this page settle that debate only partway.

Crosswords versus computer games. In one trial of 107 people who already had mild cognitive impairment (the stage between normal aging and dementia), 78 weeks of web-based crosswords beat intensive computerised training slightly on the main memory-and-thinking test — about 1.4 points on a 70-point scale, only just statistically significant.[17] The investigators had expected the games to win, and the trial's brain-scan report found no difference in its own main measure.[18] There was no do-nothing group, and a larger replication with a proper control group is under way.[19] The evidence here is weak and specific to people with mild impairment; it does not show that crosswords prevent dementia.

Observational data show the reserve trade-off in action. Among 101 older adults who went on to develop dementia, regular crossword-doers started their accelerated memory decline about two and a half years later — but then declined faster once it began.[20] Later onset, steeper slope, same disease. In middle-aged adults, frequent games and puzzles went with more grey matter in regions vulnerable to Alzheimer's, but that study was a single snapshot that can't say which came first.[21]

Music may deserve more credit than "familiar pleasure" suggests. In 10,893 Australians aged 70 and over, those who always listened to music had about 39% lower dementia risk than those who did so sometimes or less; the analysis adjusted only for age, sex and education, so it cannot show cause.[22]

Don't write off socialising

Social connection is one of the strongest reserve signals there is. When the life-course meta-analysis split its late-life category, social connection came out strongest — roughly 30% lower dementia risk, against about 9% for cognitive activity.[23] That cuts against the intuitive "puzzles over people" story.

Trials point the same way, weakly. A 2025 review of 15 trials of social-activity programmes — reminiscence groups, group crafts, music — pooled 10 of them and found a sizeable improvement in overall thinking scores. But 14 of the 15 were at high risk of bias, and some enrolled people who already had cognitive impairment, so the true effect in healthy older adults is probably smaller; the result for executive function was too uncertain to call.[24]

A related idea, sense of purpose, tracks with lower Alzheimer's risk. Among 951 adults averaging 80 years old followed for about four years, people in the top tenth of purpose scores were about 2.4 times more likely to stay free of Alzheimer's than those in the bottom tenth — an association that reverse causation could also explain.[25]

The short trial above found that socialising about familiar topics didn't produce the memory gains new skills did; the wider literature has social connection near the top. The two are not substitutes. Combining them — group classes, conversation circles, collaborative crafts, ensemble music — has support from both directions. More on social connection and the other risk factors is on Dementia prevention.

Work that stretches you, and what happens when it stops

Overall, a mentally complex job is moderately well-supported as protective. Across about 108,000 adults in seven cohorts followed for 14 to 30 years, people with the most mentally stimulating jobs had about a fifth lower dementia risk than those with the least, even after adjusting for education, and the link held for dementia diagnosed ten or more years later, which makes reverse causation less likely, though the evidence remains observational.[26] In a Canadian cohort of 3,557 adults aged 65 and over, complex work with people went with about a third lower dementia risk and complex work with things with about a quarter lower — but mainly for vascular dementia, not Alzheimer's disease.[27] A Norwegian registry study of 7,003 people pointed the same way, though its dementia estimate lost significance once income and health were accounted for.[28]

Which kind of complexity matters most is not settled. The popular "people beats data beats things" ranking is inconsistent across studies. In 225 Swedish adults, complex work with data went with better working-memory updating and work with people with better task-switching.[29] In a one-time survey of 355 older adults in Indiana, complex work with people went with a lower chance of already having impairment, while data and things showed little — a snapshot that can't rule out reverse causation.[30] The plausible reason people-work might lead is that conversation is endlessly unpredictable, the kind of load the brain can't automate away.

Retirement. Whether leaving work harms thinking is unsettled. In one US cohort of 732 adults, retiring went with steeper memory decline only in people who tend to let go of demanding goals, and only among women.[31] A German pension-reform study of women aged 63 to 67 found reasoning scores fell with each year of retirement while knowledge-based scores did not, and diagnosed cognitive disorders did not rise.[32] An analysis across 19 countries found retirees remembered slightly more words than people still working.[33] A review of 22 long-term studies found retirement generally linked to lower cognitive function, shaped by how demanding the job had been.[34] As a hedge rather than a proven fix: replacing demanding work with 40 hours of unstructured leisure is probably not neutral, so plan something serious to take its place.

Bilingualism: a later start of symptoms at most

Bilingualism is often credited with delaying dementia by four to five years, which overstates even the supportive evidence. Pooling 18 studies of age at symptom onset, bilinguals were about three and a half years older when symptoms began, and the studies that counted new cases found no robust evidence that bilingualism prevents dementia.[35] In prospective studies, which follow people forward, bilinguals had essentially the same dementia risk as monolinguals, and the positive retrospective studies are prone to confounding by education and culture.[36] A 2026 snapshot of older Mexican adults found no cognitive advantage,[37] and a 2026 US cohort found bilinguals no less likely to develop dementia, though their memory declined slightly more slowly.[38] The evidence is weak and disputed. Learn a language for its own rewards, not as insurance.

Exercise and the multidomain trials: real but small

The brain runs on blood flow and the growth factors the cardiovascular system supplies. Aerobic exercise raises brain-derived neurotrophic factor (BDNF, a protein that supports neuron survival) and improves blood-vessel function in the brain — see VO₂ max and Zone 2 training.

Layering mental demand onto exercise — dance, tai chi, racquet sports, "exergaming" — is appealing, but there is only weak evidence that it adds anything. A review in healthy older adults found exergaming's cognitive benefit small and very inconsistent, and about as good as other forms of exercise, not better.[39] In Parkinson's disease, exergaming beat standard physiotherapy on several mobility scales, but only the walking-distance gain was large enough to matter clinically.[40]

The strongest trials bundle several lifestyle levers, and their effects are small; there is moderate evidence for the bundle, but none of them can show what the mental part contributes.

  • Finland. Two years of diet, exercise, cognitive training and vascular-risk checks in 1,260 at-risk adults aged 60 to 77 improved overall cognition slightly more than health advice, by a margin that only just cleared no effect.[41]
  • United States. In 2,111 adults averaging 68, a structured programme beat a self-guided version of the same components on overall cognition, modestly. Both groups improved; the edge was in executive function, and memory — the domain that matters most for dementia — showed essentially nothing. There was no true no-intervention group.[42] The accompanying editorial asked, pointedly, "When Is Enough Enough?"[43]
  • Australia and Japan. Three years of online coaching in about 6,100 adults with risk factors gave a small gain in overall cognition;[44] an 18-month programme in 531 people who already had mild cognitive impairment found no clear benefit.[45]

All of these measured test scores, not dementia. Pooling 109 trials in about 23,000 healthy older adults, exercise combined with cognitive training gave the largest — still small — improvement, cognitive training alone came next, and adding more lifestyle areas did not automatically add benefit; 40% of the trials were at high risk of bias.[46]

The link also runs the other way. Among 5,683 adults aged 70 to 82 at raised cardiovascular risk, those in the lowest third of cognitive test scores had 85% higher all-cause mortality than those in the top third.[47] The brain's decline tracks, and probably contributes to, the body's.

Novelty may matter more than dose

Once a skill is automatic, it may stop building anything. Models of brain plasticity predict that a demanding new skill first expands the regions it uses, which then shrink back once it is mastered; a small snapshot study of young bilinguals fits that inverted-U pattern.[48] That is thin support, but if it holds, building reserve means rotating into genuinely new domains rather than drilling one skill forever.

No hard-endpoint trial has established a dose. Short, frequent and effortful — something like 15–30 minutes a day — is more plausible than a two-hour weekly marathon, because prolonged mental effort produces fatigue. Specific promises such as "30 minutes a day slows brain aging" trace largely to vendor and institute press materials. Structural brain change from demanding new learning is documented well into older age, so starting something hard at 60 or 70 is worthwhile, but the late-life benefit is modest.

What to actually do

  1. Pick one genuinely hard, new pursuit — a language, an instrument, a craft you're bad at — and practise it most days in short, effortful sessions.
  2. Do it with other people where you can: a class, a choir, a conversation group, a team.
  3. When it becomes easy, move on to something else you're bad at.
  4. Before you retire, plan the replacement: mentoring, teaching, serious volunteering, leading something, or studying a new field.
  5. Keep the basics that gate all of this: aerobic fitness, resistance training, enough sleep, the MIND-pattern diet and blood pressure in range — the levers covered across the brain health pillar.

Further reading

  • Livingston G et al. Dementia prevention, intervention, and care — 2024 report of the Lancet standing Commission. Lancet 2024;404:572–628.[49]
  • Liu Y et al. Cognitive reserve over the life course and risk of dementia — systematic review and meta-analysis. Front Aging Neurosci 2024.[50]
  • Sabia S et al. Physical activity, cognitive decline, and risk of dementia — 28-year follow-up of the Whitehall II cohort. BMJ 2017;357:j2709.[51]
  • Gates NJ et al. Computerised cognitive training for 12 or more weeks for maintaining cognitive function in cognitively healthy people in late life. Cochrane Database Syst Rev 2020;2:CD012277.[52]
  • Simons DJ et al. Do "brain-training" programs work? Psychol Sci Public Interest 2016;17:103–186.[53]
  • Coe NB et al. Impact of cognitive training on claims-based diagnosed dementia over 20 years — evidence from the ACTIVE study. Alzheimers Dement (N Y) 2026;12:e70197.[54]
  • Park DC et al. The impact of sustained engagement on cognitive function in older adults — the Synapse Project. Psychol Sci 2014;25:103–112.[55]
  • Salthouse TA. Mental exercise and mental aging — evaluating the "use it or lose it" hypothesis. Perspect Psychol Sci 2006;1:68–87.[56]
  • Verghese J et al. Leisure activities and the risk of dementia in the elderly. N Engl J Med 2003;348:2508–2516.[57]
  • Siette J et al. A meta-analysis of behaviour change techniques in social interventions targeting improved cognitive function in older adults. BMC Public Health 2025.[58]
  • Kröger E et al. Is complexity of work associated with risk of dementia? — the Canadian Study of Health and Aging. Am J Epidemiol 2008;167:820–830.[59]
  • Mukadam N et al. The relationship of bilingualism compared to monolingualism to the risk of cognitive decline or dementia — systematic review and meta-analysis. J Alzheimers Dis 2017;58:45–54.[60]
  • Ngandu T et al. A 2-year multidomain intervention to prevent cognitive decline (FINGER) — randomised controlled trial. Lancet 2015;385:2255–2263.[61]
  • Baker LD et al. Structured vs self-guided multidomain lifestyle interventions for global cognitive function — the US POINTER randomized clinical trial. JAMA 2025;334:681–691.[62]

— § —