Lower your cognitive age

The Brain and Body study (Western University with the Science and Industry Museum, Manchester) ran about 1,000 adults through the Creyos battery of memory, reasoning, verbal and attention tests alongside a lifestyle survey. People who played five or more hours a week of a single kind of game performed, on average, like non-gamers 13.7 years younger; people who played less than five hours a week across all games, like people 5.2 years younger [8] (Paleczny et al. [1], which reports the same pattern with 3 hours a week as its frequent cut-off). The paper is explicit that the design cannot show cause: people whose reasoning is already sharp may be the ones who play, and the sample skewed older and UK-based. Exercise, measured against the WHO 150-minutes-a-week line, had no relationship with any cognitive score in the same people, but did go with 12% better odds of no depression symptoms and 9% for anxiety.

Wild, Nichols, Battista, Stojanoski and Owen [2] tested 10,886 people on the same Creyos battery against their usual sleep. Reasoning and verbal ability followed an inverted U, best at about 7.2 to 7.4 hours and worse with both less and more; short-term memory did not move. Translated into age, a typical 4-hour night was equivalent to ageing 8 years on overall cognition, 9 years on reasoning and 17 on verbal ability. Sleeping well over the optimum impaired the same abilities, though the paper prints no figure for a 9-hour-plus band and the reasons are probably different (illness, low mood).

Two brain-scan findings beside the cognitive-test curve for hours. Paz and colleagues [9] took 92 gene variants that make people more likely to nap and, in 378,932 UK Biobank adults, found that genetic liability to habitual napping went with 15.8 cm³ more total brain volume, which they put at 2.6 to 6.5 years less brain ageing (the chart shows the low end); hippocampal volume and the cognitive tests did not move. Miao and colleagues [10] scored 27,500 UK Biobank adults on five healthy sleep traits (early chronotype, 7 to 8 hours, no insomnia, no snoring, no daytime sleepiness) about nine years before an MRI: each trait lost added 0.48 years to the brain age gap, an intermediate pattern read 0.62 years older than a healthy one and a poor pattern 0.99, with low-grade inflammation carrying about a tenth of it. Small next to the 8 years a 4-hour night costs on a test, because a scan measures long-run structure, not a tired day.

Luders, Cherbuin and Gaser [3] took the BrainAGE model, an MRI pattern trained on more than 650 people to guess age from brain structure, and ran it over 50 long-term meditators (19.8 years of practice on average, range 4 to 46, Zen, Vipassana, Shamatha and others) and 50 controls matched for age and sex, all aged 24 to 77. At age 50, the meditators’ brains were estimated 7.5 years younger than the controls’, and past 50 the gap widened by about a month and three weeks with every year. It is a snapshot of people who kept a daily practice for decades, so it cannot say what a year of practice does, and people who meditate for twenty years may differ in other ways; the chart puts the figure on the most-days answer and gives an occasional practice a projected place, starred.

In the Baltimore Longitudinal Study of Aging, 347 adults aged 55 and over with no dementia took audiometry and cognitive tests [4]. Greater hearing loss went with lower scores on mental status, memory and executive function, and a 25 dB loss, roughly the step from normal hearing to mild loss, was equivalent to an age difference of 6.8 years. Hearing loss is also the largest single midlife dementia factor in the Lancet Commission model, and the ACHIEVE trial [11] found hearing aids roughly halved three-year cognitive decline in higher-risk older adults, which is why the aided answer sits partway back toward the reference on the chart, starred.

Daviet and colleagues [7] looked at brain MRI in 36,678 UK Biobank adults against their reported drinking. Grey and white matter shrank with each extra daily unit, and the authors translated the loss into years of ageing: going from one to two units a day looked like ageing 2 years (1.7 excluding heavy drinkers), and from two to three units, 3.5 more years. Zero to one unit made little difference. This is a brain-scan age, not a test score, and again a snapshot.

The BrainAGE method trains an MRI pattern to guess a person’s age from brain structure, then reports the gap between the guess and the calendar. In 185 older adults, the 98 with type 2 diabetes scored 4.6 years older than their age, while controls scored their age; over a follow-up averaging 3.8 years, the diabetic group’s gap grew about 0.2 years per year [6]. The study did not split well-controlled from poorly controlled diabetes.

Two Framingham-scale snapshots that print their finding in years. Gu and colleagues [5] scanned 674 older adults without dementia (mean age 80) in the Washington Heights–Inwood cohort: the half with higher Mediterranean-diet scores (5 to 9 of 9) had 13.1 ml more total brain volume than the half with lower scores, “similar to 5 years of aging”, with more fish and less meat carrying most of it. The MIND-diet result people quote (Morris et al. [12]: the top third declined as if 7.5 years younger) is a rate of decline rather than an offset, so it stays in the text. In the Framingham Heart Study itself (Pase, [13]; 3,846 adults with MRI, mean age 54), one to two sugary drinks a day went with less total brain volume, the equivalent of 1.6 years of brain ageing (2.0 above two a day), and delayed recall on a memory test 5.8 years older (11 above two); a daily glass of juice, 1.5 years on its own. The chart uses the diet quiz answers you have already given and stars the middling ones.

One pill, a level per answer, and treatment moves it. Pooling MRI from 19 samples (2,675 people with major depression) [14], depression carried a brain age 1.08 years older, and in that pooled sample it did not differ by medication, remission or severity: the in-treatment answer carries that figure. In the Dutch NESDA cohort (220 patients, 65 controls) [15] depression read 2.78 years older once antidepressant use was corrected for, anxiety disorders 2.91, and patients not on an antidepressant read 2.58 years older than those on one, so the untreated answers sit higher. Anxiety is contested: the far larger ENIGMA-Anxiety analysis (1,595 people with generalised anxiety) [16] found no average shift, only more spread, so the 2.9 is the upper reading. ADHD has no brain-age figure of its own and shares the depression answers. Bipolar disorder read 2.78 years older across 2,919 people in ENIGMA [17], with lithium users no different from controls, and schizophrenia 3.55 years across 2,803 [18], with no difference by antipsychotic use or dose; the quiz pools the two, so the treated answer is placed at the bipolar figure and the untreated at the schizophrenia one, both starred, and the star means that split is the spread between two conditions, not a measured treatment effect.

Steffener and colleagues [19] estimated brain age from grey matter in 331 healthy adults aged 19 to 79: each year of education went with a brain 0.95 years younger, and each flight of stairs climbed on a typical day with 0.58 years younger. These are slopes, not bands, so the chart turns each answer into slope times a typical count and marks the arithmetic: two years short of finishing school, four years beyond it, two, four or six flights a day. Three flights a day in that model is worth roughly as much as two years of school.

Two brain-age models grade it. In the UK Biobank model (Kang et al. [20], a preprint), overweight adults had a brain age 0.59 years older than normal-weight adults and obese adults 1.7 years older. The iSTAGING consortium (Anagnostakis et al. [21]; 46,288 adults across 15 studies) prints the same ladder for men, 8 months and about 2 years, with a smaller gap in women and one that fades with age. The older Cam-CAN white-matter finding (Ronan et al. [22]: overweight and obese together, 10 years older than lean) pooled the two bands and measured one tissue, and the same cohort’s whole-brain model puts obesity at about 3 years [23], so the chart uses the graded figures. No paper separates a BMI of 40 and over, so that answer shares the obese figure, and none prints a figure for underweight. All are snapshots, so none can say which came first.

Four MRI brain-age findings, none large, all published. In the Dunedin cohort (855 people followed from age 5) [24], being socially isolated only in adulthood went with a brain age 2.16 years older at 45 than never being isolated; isolation only in childhood made no difference. In 12,115 UK Biobank adults [25], smoking on most or all days went with a brain 0.6 years older, plus 0.03 years per pack-year (the chart adds 20 pack-years for the 10-a-day answer, about 1.2 years), while former and occasional smokers did not differ from never-smokers. In 686 adults scanned over 12 years [26], every 10 mmHg of blood pressure above optimal added about two months of brain age, so 135/85 reads more than 6 months older than 110/70. And in 33,381 UK Biobank adults [27], a resting pulse under 60 read 0.23 years younger than 60 to 69, 70 to 79 read 0.26 older and 80 or more 0.64 older, which is mostly the fitness pill wearing a different hat.

Cole and colleagues [28] trained a brain-age model on healthy adults and applied it to 99 people with persisting problems after a traumatic brain injury (mean age 38, scanned a year or more after; 83% moderate or severe by the Mayo classification, 24% road accidents, 23% assaults, 6% sport) and 113 controls. Grey matter read 4.66 years older than the calendar and white matter 5.97, the gap grew with time since injury, and it predicted cognitive impairment. The moderate and severe group carried the whole effect; the mild (probable) group did not differ from controls, which matters because the quiz question counts concussions.

The only trial on the page. Wan and colleagues [29] randomised 130 healthy adults aged 26 to 58 to a year of aerobic exercise, two supervised hour-long sessions a week plus home sessions to reach about 150 minutes, or to no change. The exercisers’ MRI brain age fell 0.6 years while the controls’ rose 0.35, a gap of about one year. This is a scan finding: on cognitive tests, the Brain and Body study found exercise moved mood and anxiety, not scores, so the chart puts the trial’s dose on the 5-plus-days answer and gives 3 to 4 days a projected place, starred. Running, cycling, swimming, rowing and any sport that keeps you breathing hard count as that dose: no paper prints a separate brain-age or cognitive-age figure in years for runners or for a particular sport, so they count through this pill and the fitness and strength ones below; hours of sitting, sleep apnoea, air pollution and grip strength have no years figure and no pill.

The running, cycling and sport lever, measured as the fitness it builds. At the baseline of the same eBACH trial [29], 130 adults aged 26 to 58 did a treadmill test alongside their MRI, and every standard deviation of peak oxygen uptake (about 7 ml/kg/min, roughly average to fit for your age) went with a brain age 1.83 years lower; the chart turns that slope into one step either side of average and marks the arithmetic. Strength has a trial of its own: in LISA [30], 309 adults aged 62 to 70 were randomised to a year of heavy resistance training three times a week, moderate training (one supervised session, two at home) or no change, and a functional MRI brain-age model read the heavy group 1.42 years younger at one year and 1.85 at two, the moderate group 1.39 and 2.26, against 0.66 and 0.82 for controls that were not significant; both arms differed from control. No paper prints a separate figure for runners or for one sport, so they count through these two pills and the aerobic one.

Walking, counted rather than reported. Spartano and colleagues [31] put a step counter on 2,354 Framingham adults (mean age 53) for a week and scanned them: every extra hour a day of light activity went with 0.22% more total brain volume, “approximately 1.1 years less brain aging”, and people taking 10,000 or more steps a day had 0.35% more brain volume than those under 5,000, about 1.75 years less brain ageing. Once light activity was in the model, harder exercise added nothing to volume. Among people below the 150-minute guideline, simply crossing 7,500 steps was worth about 2.2 years. A snapshot, so it cannot say which came first, and the 7,500-to-9,999 band is a projected place on the chart because its own figure just missed significance.

Rogenmoser, Kernbach, Schlaug and Gaser [32] ran an MRI brain-age model over 125 age-matched adults: 45 amateur musicians averaged a BrainAGE of −4.51 years, 42 professionals −3.7, against 38 non-musicians at about their calendar age. Among professionals, more years of playing went with a smaller benefit; among amateurs, years made no difference. The same team’s 2025 study of 81 twenty-year-olds found no link between musical sophistication and brain age, so the figure is for adults with years of playing behind them. It fits the ranking in A design for life [33]: in 9,443 adults on the Creyos tests, playing an instrument, piano above all, sat in the top tier of 20 lifestyle predictors with video games and puzzles, though that paper prints no years and explained only 7 to 9% of the variance in scores.

The second trial on the page, and the one with a cognitive-test age. COSMOS randomised more than 21,000 US adults aged 60 and over to a daily multivitamin-mineral tablet or placebo; 573 of them came in for cognitive testing after two years [34], and pooled with the trial’s two earlier cognition arms (over 5,000 non-overlapping people, two to three years of treatment) the multivitamin group’s global cognition was better by 0.06 standard units, which the authors estimate as slowing cognitive ageing by the equivalent of two years, with a benefit on episodic memory too. It is small, it is one brand of tablet, it is people over 60, and the authors stop short of recommending it. It is also randomised, which nothing else on this page except the exercise trial can say.

No study prints how many years younger bilinguals score on a test battery or read on a scan, so this pill borrows a different measurement and says so with a star. In 184 Toronto memory-clinic patients, lifelong bilinguals showed their first dementia symptoms 4 years later than monolinguals with equivalent education, occupation and immigration history [35]. A Hyderabad clinic found a 4.5-year gap in 648 patients, independent of literacy [36]. Cohorts of people who learnt a second language later, or use it less, have often found no delay. An onset delay is not an MRI brain age; the chart adds the 4 years to your total and flags it as a proxy.

The PROTECT online cohort asked 19,078 adults aged 50 to 93 how often they do word and number puzzles and ran them through tests of attention, reasoning and memory [37, 38]. People who did word puzzles performed like people about 10 years younger on grammatical reasoning and 8 years younger on short-term memory; on problem-solving, regular puzzlers of either kind averaged 8 years younger than non-puzzlers, and number puzzles more than once a day read about 8 years younger, with a dose response across frequency. The chart gives the daily answer 8 years and a less-than-daily habit a projected place, starred. Board and card games have no such figure: in A design for life [33] they sat third for memory in a random-forest ranking of 9,443 adults, which prints no coefficient to turn into years. Cross-sectional: the sharp may be the ones who do puzzles.

Three snapshots with year figures and small samples. Navarro-González and colleagues [39] ran a brain-age model over 82 controls, 91 people with episodic migraine and 74 with chronic migraine (15 or more headache days a month): the chronic group’s brain age gap was 4.16 years against −0.56 in controls (P = 0.01), the episodic group’s 1.21 did not reach significance, so it sits as a starred projection. Cruz-Almeida and colleagues [40] scanned 47 adults aged 60 to 83: the 33 with chronic pain read 2 years older than their calendar age while the 14 without read 4 years younger, and inside the pain group the treated (medication, physiotherapy, even cold packs) read 3.9 years younger, almost the pain-free figure, while the untreated read 5.6 years older (P = 0.002), a 9.6-year gap against the pain-free that ranks near the top of the chart on the strength of 47 people. Kim and colleagues [41] scanned 33 shift-working and 80 fixed-day healthcare workers with four brain-age models: the shift workers sat 3.6 to 4.2 years older across models, two of which kept significance after adjustment (3.7 years on the one shown).

The one paper with year figures for dance and art measures brain age differently from the rest of this page. Coronel-Oliveros and colleagues [42] trained a brain-age model on EEG and MEG connectivity from 1,240 healthy adults, then applied it to matched expert and non-expert groups from earlier studies: 23 tango dancers with at least a year of formal instruction read 7.1 years younger than 23 matched non-dancers (95% CI 1.5 to 12.8), 15 trained visual artists 6.2 years younger than 15 non-artists (1.6 to 10.8), musicians 5.4 and video gamers 4.1, with the more skilled sitting younger in every domain, and a 30-hour course of learning a strategy game moved novices 3.1 years younger against no change in an active control. Small groups, wide intervals, snapshots, and an electrical brain clock rather than an MRI one: the chart puts both pills on their own ruler and stars any total that mixes them with the scan pills. The for-fun answers get projected places, starred. No MRI dance study prints years: the Rehfeld and Burzynska trials report hippocampal volume and white-matter integrity.

Perceived stress has no years figure: the SHIP cohort prints a risk ratio for advanced brain ageing (1.21), the ENIGMA anxiety brain-age study is null, and the cognitive-decline cohorts (Chicago, REGARDS) print betas and odds. What has a slope is the habit of worrying. Karim and colleagues [43] ran a grey-matter brain-age model over 78 adults aged 50 and over recruited to span the range from no worry to generalised anxiety: each point on the Penn State Worry Questionnaire went with a brain 0.11 years older (P < 0.05), rumination the same (P = 0.057), brooding carrying it, while anxiety and depression scores on their own did not. The chart turns the slope into bands with the sample’s own spread, one and two standard deviations of worry (14.7 and 29 points, about 1.6 and 3.1 years), so both are derived rather than printed. Small, cross-sectional, anxiety-heavy, and the worry result held in one of the paper’s two analyses: a direction, not a measurement.

Greenspace and the brain has a volume literature (UK Biobank: about 630 mm³ more grey matter per 10% more natural land within 300 m, Noseworthy et al. [44]) but no paper prints an atrophy rate to convert it, and no brain-age paper prints years for greenery, so the pill is a cognitive-test age. In the Nurses’ Health Study II, 13,594 women (mean age 61) took the Cogstate battery online while the vegetation within about 270 m of their address was scored from satellite images: an interquartile range more greenery went with better psychomotor speed and attention, a difference the authors put at being 1.2 years younger [45]; the global score moved the same amount, learning and working memory did not, depression explained part of it and air pollution and exercise none. The same cohort later found greener midlife addresses slowed later cognitive decline by about 8 months, a rate rather than an offset. Cross-sectional, women only, and it measures where you live, not time outside; the middle answer is a starred projection.

Supplements mostly come up empty on this ruler: cocoa flavanols did nothing for cognition in the COSMOS trial, the VITACOG B-vitamin trial prints atrophy rates (0.76% a year against 1.08% on placebo) but no years, vitamin D has a brain-age association with no printed figure, and creatine, curcumin and the rest have effect sizes only. Omega-3 has a years figure because Framingham measured it in the blood. Tan and colleagues [46] took red-blood-cell DHA and an MRI from 1,575 adults without dementia (mean age 67): the quarter with the lowest DHA had less total brain volume than everyone else, “equivalent to approximately 2 years of structural brain aging”, and lower visual memory and executive scores. It is a deficiency signal, not a dose response; above the bottom quarter is the reference and eating more still has no printed figure. The chart maps rarely eating oily fish or taking fish oil to that low quarter, an approximation, and the daily multivitamin remains the only supplement on this page with a trial behind it.

Talking, visiting and reading have no cognitive-age or brain-age figure in years: the UK Biobank leisure paper prints standardised betas for occipital volume with no atrophy rate, laughter has a hazard ratio (0.84 for dementia in 12,165 Japanese over-65s) and nothing else, and no study puts conversation in years. What the Rush Memory and Aging Project prints, twice, is when dementia shows up. Chen and colleagues [47] followed 1,923 dementia-free older adults for a mean of 6.7 years: the least socially active (meals out, events, trips, volunteering, visiting friends and family) developed dementia at a predicted mean age of 87.7, the most active at 92.2, about 5 years later, with the same gap for mild cognitive impairment. Wilson and colleagues [48] followed 1,903 for 6.8 years: the 10th percentile of cognitive activity (reading, writing, games, library visits) developed Alzheimer’s dementia at 88.6, the 90th at 93.6, again 5 years, unrelated to the amount of Alzheimer’s pathology at autopsy. Both are the ends of a scale, observational, and open to reverse causation (people in early decline go out and read less). Like the second language, both sit on the onset ruler, starred, and the social pill overlaps the isolation pill from the other direction.

In Whitehall II (5,099 men and 2,137 women followed for 10 years from a mean age of 56) [49], male current smokers lost global cognition and executive function faster than never-smokers, a difference the authors describe as similar to the effect of 10 years of age on decline. Women showed no smoking effect, men who had quit within the decade still showed faster executive decline, and men who had quit 10 or more years earlier showed none. It is a rate of decline rather than an age offset, which is why the smoking pill uses the UK Biobank brain-age figure instead and this finding sits in "everything else".

McDonough and colleagues [50] ran a BrainAGE model over 86 adults aged 20 to 74 in the Alabama Brain Study: pet owners read 15.10 years younger than non-owners (t(73.28) = 3.22, adjusted P = 0.058), dog owners specifically 12.95 years younger (adjusted P = 0.025), owners of more than one pet 12.56 years younger (adjusted P = 0.035). Eighty-six people, cross-sectional, and only two of the three comparisons survive correction for multiple testing — real limits, published as printed rather than cut down. Three longer studies back the direction without the size. The Baltimore Longitudinal Study of Aging (637 adults, up to 13 years) [51] found pet and dog owners declined more slowly on recall and Trail Making tests (effect sizes 0.13 to 0.29, all P ≤ 0.05). SHARE (16,582 adults, 18 years) [52] found dog owners’ memory and cat owners’ verbal fluency each declined more slowly (both P < 0.05); fish and birds did nothing, which is why a fish- or bird-only home gets no number rather than a borrowed one. ELSA (7,945 adults) [53] found a pet erased the usual cognitive cost of living alone, with no effect in people living with others. A cat alone has no McDonough MRI figure of its own; this page projects it from the BLSA and SHARE effect sizes at about −3.5 years rather than reading it off the dog number.