Sudoku is fine, but it isn’t the point

The transfer problem. What the big brain-training trials actually found. And the six interventions with real evidence — walking, sleep, hearing aids, treated hypertension, social contact, and genuinely learning something new.

Every morning I see the same set of people in the day room working through their Sudoku. Some have been at it for years. It calms them, and it’s a pleasant way to start the day. So it always makes me a little uncomfortable when a family member asks me whether it’s actually keeping their grandmother’s mind sharp. The honest answer is more complicated than either “yes” or “no.”

Brain training, in the specific sense of pencil puzzles and app-based cognitive games, has been oversold for about two decades. The consumer industry runs on an assumption that hasn’t fully held up in the research — that training on one task will transfer to broad cognitive improvement. What we’ve actually learned is that Sudoku makes you better at Sudoku. It does not, on its own, protect against dementia in the way the marketing has suggested.

The good news is that we now have a fairly clear picture of what does work. Some of it involves your brain directly. Most of it doesn’t.

Where dementia actually starts

Alzheimer’s disease is a slow accumulation of an abnormal protein called amyloid beta in the brain, along with tau tangles inside the nerve cells themselves. The accumulation begins roughly twenty to thirty years before symptoms appear. Someone in their sixties who feels fine may have been building deposits since their forties.

Not all dementia is Alzheimer’s — vascular dementia, Lewy body, frontotemporal — but the shared feature is that once symptoms begin, they progress on a schedule the individual doesn’t fully control. Prevention efforts have to start before symptoms, when there’s still tissue to preserve.

The 2024 Lancet Commission on dementia estimated that roughly 40 to 45 percent of dementia risk is attributable to modifiable factors. That’s the ceiling of what prevention can plausibly deliver.

The “transfer” problem

Cognitive scientists talk about near transfer and far transfer. Near transfer means the training helps you at very similar tasks. Far transfer means it helps at unrelated tasks. Puzzles produce near transfer reliably — you get better at that specific puzzle. They produce far transfer only weakly, and not consistently across trials.

The 2010 BBC study with 11,000 participants — the largest of its kind — found no meaningful transfer from six weeks of brain training to broader cognitive function. The ACTIVE trial, an American study, did find some real benefits from reasoning training that persisted for a decade, but the effects were narrower than the popular press implied.

None of this means puzzles are bad. They keep the mind engaged, provide a sense of accomplishment, and for many older adults are one of the pleasant rituals of the morning. What they don’t do is substitute for the interventions that actually matter.

What has real evidence behind it

Aerobic movement

This is the single strongest piece of evidence in the dementia prevention literature. Regular aerobic activity — the kind that lifts your heart rate — actually increases the volume of the hippocampus, the brain region most responsible for memory. Erickson and colleagues at the University of Pittsburgh showed measurable hippocampal enlargement after a year of moderate aerobic training in older adults. The WHO recommendation is 150 minutes a week. Walking counts. Swimming counts. Aqua walking counts. Whatever gets a mild sweat going for 20 to 30 minutes a day.

Sleep

Sleep is when the brain does its housekeeping. The glymphatic system, described by Nedergaard’s group in Science in 2013, opens up during deep sleep and flushes metabolic debris, including amyloid beta. Chronic short sleep leaves that debris in place. The Whitehall II cohort found that people sleeping six hours or less in their fifties had 22 percent higher dementia risk decades later, compared to seven-hour sleepers.

Seven to eight hours, most nights. Not perfect — nobody has that — but reliable enough that the housekeeping happens.

Social contact

The brain treats conversation as complex cognitive work. Listening, tracking what somebody means, retrieving the right word, keeping context — all of it engages many networks at once, in a way that a solo puzzle doesn’t. Cohort studies consistently link social contact with reduced dementia risk. The Holt-Lunstad meta-analyses put loneliness in the same risk category as smoking.

What this looks like practically: at least one face-to-face contact per week, ideally more. Hobbies with other people. Volunteering. Family that actually shows up. If those are hard to come by, digital contact is better than nothing, but voice-plus-video beats text-only.

Hearing aids, actually

One of the most surprising findings of recent years. Untreated hearing loss is a substantial dementia risk factor, and correcting it with hearing aids appears to reduce risk. The ACHIEVE trial, published in 2023, showed a 48 percent reduction in three-year cognitive decline in high-risk older adults given hearing aids compared to controls. If you or a parent has hearing loss, do not put off getting fitted for aids. This intervention is more powerful than any brain training app.

Managing blood pressure, diabetes, and depression

Midlife hypertension is a major risk factor for both vascular dementia and Alzheimer’s. So is diabetes. So is chronic depression. Treating these — with medication if needed, with lifestyle changes always — reduces long-term dementia risk. This is why the “read the annual health check paper” article is really a dementia prevention article in disguise.

Learning something new

If puzzles have a role, it is here — as part of a broader pattern of cognitive novelty. Not repeating the same crossword variant every day, but actually learning things. A new language. A new instrument. A new craft. A course at a community center. The mechanism seems to involve building “cognitive reserve,” the brain’s capacity to compensate for damage that has already begun.

A practical synthesis

If you had to design an actual dementia-prevention routine for a person in their sixties or seventies, based on what the evidence supports, it would look something like this:

• 30 minutes of walking most days.
• Seven to eight hours of sleep, protected as a priority.
• One meaningful conversation a day, ideally in person.
• Blood pressure and blood sugar checked and treated.
• Hearing tested and corrected if needed.
• One new skill or subject being learned, at whatever pace fits.
• The Mediterranean-plus-washoku diet described elsewhere on this site.

Sudoku is fine as decoration on top of this. It is not a replacement for any part of it.

The apps that sell you “personalized brain training” for a monthly subscription are, mostly, not doing what they claim. The morning walk, the phone call to your sister, the choir practice, and the hearing aid appointment are doing much more. That is the ordinary, unpleasant news. It’s also, in a real sense, good news: none of it requires a subscription.