Best Memory Games for Adults: The Brain Science

The best memory games for adults are short, focused challenges that make you hold and recall information under a little pressure, like remembering a growing string of numbers, a flashed grid of squares, or a lengthening sequence of tiles. The brain science is genuinely interesting, and also more honest than the marketing. These games reliably improve the exact skill you practise, and reviews find small but real gains in working memory and processing speed. Whether that carries over to remembering names and to-do lists in daily life is still debated. So play them because they're quick, free, and enjoyable, and treat any wider benefit as a bonus.

What Are the Best Memory Games for Adults?

There's no single champion, but a few types show up again and again in the research because they push different parts of memory. Here are the ones worth your time, and you can play each one free right now:

• Number memory. A string of digits flashes up and you type it back. It grows each round, testing how much your working memory can hold.
• Visual memory. Squares light up on a grid and you have to remember which ones. This trains spatial, visual recall rather than verbal.
• Sequence memory. A pattern plays and you repeat it back, with one more step each time. It's the classic Simon-style challenge.
• The chimp test. Numbered squares appear briefly, then you tap them in order from memory. It's humbling, and famously hard for humans.

The best pick is honestly the one you'll come back to. A game you enjoy gets played consistently, and consistency is what any memory benefit depends on. Rotate a couple so your brain keeps facing something a little new.

How Do Memory Games Affect the Brain?

When you play a memory game, you're loading and manipulating information in working memory, the mental scratchpad that holds things for a few seconds while you use them. Repeatedly pushing that capacity is what the games train. And the effect on the brain is measurable. A National Institute on Aging supported study had adults aged 60 to 80 play for 30 to 45 minutes a day across four weeks, and found improvements in hippocampal-based memory that were still measurable four weeks after the daily play stopped. That points to real change in the memory systems of the brain rather than just a higher score.

Broader reviews back this up for healthy adults too. A 2020 meta-analysis published in Scientific Reports pooled 16 studies covering 1,543 people with a mean age of 70, all over 60 and none cognitively impaired. Bonnechère and colleagues found commercial cognitive games produced significant gains in four domains: processing speed, working memory, executive function, and verbal memory. To understand what's actually being exercised here, our guide to what working memory is breaks down the system these games lean on.

Do Memory Games Actually Improve Your Memory?

Here's the honest answer, and it comes in two parts. Yes, they improve the trained task and closely related ones. That part isn't in dispute. Practise digit recall daily and your digit recall climbs. The contested part is transfer, whether that gain spreads to everyday memory and general thinking.

There's a proper name for the two halves, and a definitive paper behind it. Simons, Boot, Charness, Gathercole, Chabris, Hambrick and Stine-Morrow reviewed the whole field for Psychological Science in the Public Interest in 2016, and their summary is the clearest statement anyone has managed. They found "extensive evidence that brain-training interventions improve performance on the trained tasks, less evidence that such interventions improve performance on closely related tasks," and "little evidence that training enhances performance on distantly related tasks or that training improves everyday cognitive performance." Near transfer is solid. Far transfer is where it thins out. They also noted that none of the studies they cited met every one of the best practices you'd want before drawing firm conclusions, which is the quiet part that explains why this argument keeps running.

And it does keep running, on both sides. In 2014, a group of researchers organised through the Stanford Center on Longevity and the Max Planck Institute cautioned that there's no compelling evidence brain games broadly reduce or reverse cognitive decline. That statement had more than 70 signatories. What's less often mentioned is that a few months later 133 scientists and practitioners signed a response, organised through Cognitive Training Data, agreeing with parts of the original but worried readers would wrongly conclude no cognitive training works at all. Worth knowing that this second letter came from the side of the debate with commercial interests in the answer, so weigh it accordingly. But two open letters, 70 names against 133, is a fair picture of how unsettled this was. Meta-analyses land in the middle, and the honest way to read them is to look at the actual effect sizes rather than the word "significant." In that same Bonnechère review, the gains are real but modest: processing speed came out strongest at a Hedges g of 0.40, while working memory and executive function both sat at 0.21 and verbal memory at 0.12. Two domains, attention and visuospatial ability, showed no significant benefit at all. So memory games help, just more modestly and more narrowly than the ads suggest. For a deeper dive into that debate, see our companion piece on whether brain training actually works.

Do the Gains Last After You Stop Playing?

This is the question almost nobody asks, and it's the one with the least encouraging answer. Most studies measure you right after training ends. Far fewer come back a year later, and the ones that do change the picture.

The best evidence here comes from Cochrane, which produces the systematic reviews that medicine treats as the top of the evidence pile. Their review by Nicola Gates and colleagues, published in February 2020, pooled 8 randomised controlled trials covering 1,183 cognitively healthy older adults, with training running from 12 to 26 weeks. Three findings, in order of how much they should change your thinking:

• There is a benefit, and it's small. The reviewers found "small benefits of CCT may be seen for global cognitive function when compared with active controls" straight after training. Note the comparison group. Measured against people who were doing something else engaging, not against people doing nothing.
• It didn't survive the year. They found "no evidence that the effect on global cognitive function persisted 12 months later." So the gain was real, and then it wasn't there.
• Nobody should be confident about any of this. Their own summary is that "the overall quality of evidence was low or very low for all outcomes", that all the included studies had design problems, and that future research "might well find something different."

That last bullet cuts both ways, and it's worth being fair about it. Low-quality evidence isn't evidence that something doesn't work. It's evidence that we don't know. The honest reading is that the case for lasting benefit hasn't been made, not that it's been ruled out.

A 2024 paper puts a clean number on the ceiling. Maria Syed, Jarrad Lum, Linda Byrne and David Skvarc ran a second-order meta-analysis in the Journal of Intelligence, meaning they pooled six existing meta-analyses of randomised trials rather than individual studies. Working memory training in healthy adults produced a standardised mean difference of 0.335, with a confidence interval of 0.223 to 0.447. Their conclusion is worth reading twice: "the working memory capacity of healthy adults can be improved through training. However, the effect size is small." On broader benefits they're blunter, writing that "the utility of this type of training in real-life improvements in cognition may be minimal."

They also found no training method beat any other. So the hunt for the one really effective app is probably a waste of your time.

What to actually do with this. Two things follow, and neither is "give up."

First, if you stop, expect the gain to fade. That's not a flaw unique to brain games. It's how nearly every trainable capacity behaves, and it's the same reason you don't stay fit from a gym membership you cancelled. If you want the small benefit, treat it as ongoing rather than as a course you complete.

Second, judge a game by whether you'll still be playing it in six months, not by what its marketing claims. A modest effect you maintain beats a larger one you abandon in March. That's a lower bar than the ads set and a much more useful one, and it happens to be the same reason our daily brain game routine is built around short sessions you'll actually repeat.

Why Did a Brain Game Company Get Fined?

Because it sold the transfer that the research above hasn't established. This is the most concrete thing in the whole debate, and it's worth knowing before you pay for anything.

In January 2016 the US Federal Trade Commission announced that Lumos Labs, the company behind Lumosity, would pay 2 million dollars to settle deceptive advertising charges. The FTC's complaint was about claims that the games helped people perform better at work and school, and that they could reduce or delay the cognitive decline that comes with age and other serious conditions. The regulator's position was blunt: the company was trading on people's fear of losing their memory, and it didn't have the science to support what it was saying. The full judgment was 50 million dollars, suspended once the 2 million was paid.

What makes this useful rather than just gossip is the specifics. Lumosity was 40 games, advertised at 10 to 15 minutes a session, three or four times a week. That's a completely reasonable regime, and roughly what the research on task-specific improvement supports. Nothing about the product was fake. What crossed the line was the promise attached to it: not "you will get better at this game" but "you will do better at your job and hold off decline."

Hold that distinction and you can read any brain-training claim correctly. Getting better at a task you practise is real, cheap, and uncontroversial. Getting better at life because you practised a task is the part nobody has convincingly demonstrated, and it's the part that gets sold. When a product leads with dementia, work performance, or IQ, you're looking at the claim a regulator has already acted on.

Which is a decent argument for playing these free. There's no subscription to justify, so nothing has to be oversold to you.

What About the Supplements That Promise the Same Thing?

Same promise, same evidence problem, and in one case the same regulator. If you're standing in a pharmacy weighing a memory game against a bottle of something, this section is the shortcut.

Because the Lumosity story above has a sibling, and it's almost the same story told with a pill.

In January 2017 the FTC and the New York Attorney General charged the marketers of Prevagen, a supplement sold on memory claims, with making false and unsubstantiated claims that it improves memory and was "clinically shown" to work. The complaint said the marketers relied on a study that failed to show the product worked better than a placebo on any measure of cognitive function. Read that sentence next to the FTC's Lumosity complaint and the family resemblance is hard to miss.

There was a second allegation that's worth knowing because of what it says about how these products get marketed. The company claimed its active ingredient, apoaequorin, "enters the human brain to supplement endogenous proteins". The FTC alleged that was false, and pointed at the company's own safety studies showing the protein is rapidly digested in the stomach.

The two cases then diverged in an interesting way. Lumosity settled. Prevagen's maker fought, and after seven years of litigation including a jury trial, the FTC announced in December 2024 that the court had ordered Quincy Bioscience to stop making claims about Prevagen that mislead people worried about memory loss, holding that the company violated the FTC Act. Seven years and a jury, for a claim printed on a box.

Now the supplement that's actually been tested properly, because one of them has.

Ginkgo biloba got the trial the category rarely gets. The Ginkgo Evaluation of Memory study, led by Steven T. DeKosky and run at four clinical sites over eight years, is described by the National Center for Complementary and Integrative Health as the largest clinical trial ever conducted on ginkgo's effect on the occurrence of dementia. Published in JAMA. The finding: 240 milligrams a day was ineffective at reducing the development of dementia and Alzheimer's disease in older people.

Not "modest benefit". Not "more research needed". Ineffective, in the biggest trial anyone has run.

But this isn't a section about supplements being uniformly worthless, because the evidence doesn't say that either, and one result deserves a fair hearing.

The National Institute on Aging reported on a randomised trial led by Adam Brickman at Columbia, covering more than 3,500 volunteers aged 60 and over, assigned to a daily multivitamin or a placebo and tested annually for three years. It ran in the American Journal of Clinical Nutrition in May 2023. After one year, the multivitamin group scored significantly higher on immediate recall than placebo. That's a real result from a real trial, and it deserves saying plainly before the caveats.

Here are the caveats, and they're the numbers rather than the interpretation. The test showed people 20 words, three seconds each, then asked how many they could type back. The multivitamin group went from about 7.1 words to 7.8 after a year. The placebo group went from about 7.2 to 7.6. So the gap between the two groups is roughly a third of a word out of twenty. The multivitamin advantage did not significantly grow in years two and three. And other cognitive tests in the same study showed no significant difference between the groups at all.

If that shape feels familiar, it should. It's the same shape as the brain-training literature this whole page has been picking through. Real, measurable, statistically sound, and far smaller than anything the advertising implies.

So what do you actually do with all that?

• Separate correcting a deficiency from taking a supplement. These are different questions. If something is genuinely low, that's worth knowing and worth treating, and the way you find out is a conversation with a doctor and a blood test, not a claim on a label.
• Treat "clinically shown" as marketing language. It isn't a regulatory standard. Both FTC cases on this page turned on exactly that phrase being attached to studies that didn't support it.
• Put the effect size next to the price. A third of a word out of twenty is a fair description of the best supplement result on this page. Whatever the bottle costs, decide what that result is worth to you with your eyes open.
• Notice who gets targeted. Both of these products were sold to people frightened about their memory. That's a market built on worry, and worry is a poor basis for a purchase.
• The game is free. Whatever you conclude about the bottle, a round of number recall costs nothing and the evidence behind it is no weaker.

Which is the honest bottom line for this page and the reason the section belongs on it. Memory games and memory supplements are competing for the same worry, they produce effects of roughly the same modest size, and only one of them arrives with a monthly price attached.

How Would You Know If It's Actually Working?

Harder than it sounds, and there's one experiment that explains why this whole field argues with itself.

Foroughi, Monfort, Paczynski, McKnight and Greenwood ran it at George Mason and published it in PNAS in 2016. They recruited 50 people using two different flyers. One promised that "numerous studies have shown working memory training can increase fluid intelligence." The other was plain and promised nothing. Everyone then did the same single one-hour training session.

The group recruited by the suggestive flyer improved by an amount the authors put at the equivalent of 5 to 10 points on a standard IQ test. The group recruited by the plain flyer showed no meaningful change.

Here's the detail that makes it land. Both groups performed equally well during the training itself. The training didn't differ. What differed was what people had been led to expect before they started. One hour, and belief alone moved the number that much.

The authors' own conclusion is the uncomfortable one: any study that recruited participants by advertising brain training has that effect baked into its results unless it specifically controlled for it. That's a lot of studies, and it's a large part of why the meta-analyses above land on such modest effects and why the 2014 and 2015 open letters happened at all.

Two honest limits before anyone over-reads this. It's 50 people, one session, one lab, and it does not show that every training gain is placebo. What it shows is that expectation is powerful enough to contaminate any result that doesn't guard against it, which is a claim about research methods rather than about whether your brain changed.

So how do you judge it for yourself? Three habits help.

• A rising score proves almost nothing. Getting better at the game is the near-transfer effect from earlier, and it's essentially guaranteed. It's the one part everybody agrees on, which makes it useless as evidence of anything wider.
• Watch the thing you actually care about. Do you lose your keys less? Walk into fewer rooms having forgotten why? Follow a recipe without re-reading it? Those are the outcomes worth tracking, and they're the ones the research says are hardest to shift.
• Be most suspicious right after reading something encouraging. Which includes this article. If a piece about brain science leaves you feeling sharper before you've played anything, you've just experienced the flyer.

None of that is a reason to stop playing. The games are free, quick, and enjoyable, and enjoying something is a perfectly good reason to do it. It's a reason to hold your conclusions loosely, especially the flattering ones.

Which Types of Memory Do These Games Train?

Memory isn't one thing, and different games hit different parts of it. Knowing which is which helps you pick a useful mix:

• Working memory. Holding and juggling information for a few seconds. Number and sequence games lean hard on this.
• Visual and spatial memory. Remembering where things were. Grid-based visual memory games target this directly.
• Short-term recall. Keeping a small amount of information available right now, which nearly every memory game touches.
• Processing speed. Not memory itself, but how fast you take in and respond to information, which timed games sharpen alongside recall.

Because the brain adapts to whatever you repeat, mixing types keeps the challenge fresh and works a wider slice of your memory. If you want a baseline, our memory test guide shows how to measure where you stand before and after a few weeks.

How Often Should You Play to See a Benefit?

Short and regular wins. Most studies that find an effect use sessions of about 15 to 30 minutes a day, several times a week. Going longer doesn't buy you proportionally more, and it tends to make people quit, which is the real killer of any training effect. A daily 15-minute round you actually keep up beats an hour-long session you do once a month.

Give it weeks, not days. Task-specific improvement often shows within a couple of weeks, but it builds gradually. Track your scores so you can see the trend, and don't be discouraged by an off day, since sleep, stress, and caffeine all swing your numbers around more than you'd think.

Does Learning a Technique Beat Playing a Game?

Yes, and by a distance that's genuinely startling. This is the answer to the question the last two sections kept circling, which is what you'd actually do if you want to remember names and lists rather than climb a leaderboard.

Dresler and colleagues ran the study that settles it, published in Neuron in March 2017. They started by scanning 23 of the world's top memory athletes alongside 23 people matched for age, health, and intelligence. Then they took 51 novices and split them three ways for 40 days of daily 30-minute sessions: one group learned the method of loci, one did short-term memory training, and one did nothing.

That middle group is the important one. Short-term memory training is essentially what the games on this page are. So this isn't technique versus nothing, it's technique versus the thing you came here to read about, run head to head.

The result wasn't close. Starting from an average of 26 words recalled from a list of 72, the method-of-loci group finished on 62. The memory-training group gained 11 words. The do-nothing group gained 7.

Then the part that matters more. Four months later, with no one still practising, only the technique group held on to it, still recalling around 22 words more than they had at the start. The other two had essentially nothing left to show. The brain imaging matched the behaviour: training reconfigured connectivity across memory and visual-spatial networks, shifting the novices' patterns toward those of the memory athletes.

So why does technique beat practice so heavily here? Because they're doing different jobs. A memory game pushes on how much raw capacity you can hold, which is largely fixed and which is exactly why the gains stay stuck to the trained task. A mnemonic doesn't expand capacity at all. It changes what you're storing, converting a meaningless list into vivid images placed along a route you already know by heart, so recall becomes a walk rather than a retrieval. You're not getting a bigger scratchpad. You're writing on it in a format your brain was already good at holding.

The method itself takes about a minute to explain. Pick a route you know cold, your walk to work or a path through your home. Turn each item you need to remember into an exaggerated image and leave it at a specific spot along that route. To recall, walk the route in your head and collect them. That's the whole technique, and it's roughly 2,500 years old.

Two honest caveats. The 40-day commitment was real work, not a weekend, and 51 novices is a modest sample even with the imaging behind it. And a mnemonic is domain-specific by design: it's superb for ordered lists, names, and facts you deliberately choose to encode, and it does nothing for where you left your keys, because you never encoded that on purpose in the first place.

None of which is a reason to close the tab. Play the games because they're free, fast, and fun, which this guide has said from the top. Just know that if the actual goal is better everyday recall, half an hour learning a technique will outrun half an hour of practice, and it'll still be there in four months.

Which Memory Techniques Have the Strongest Evidence?

Testing yourself, and spreading it out. Those two beat everything else that's been properly measured, and neither one is a game or a mnemonic.

The section above ends on an important limit: a mnemonic is domain-specific by design. So what's the general-purpose answer? John Dunlosky, Katherine Rawson, Elizabeth Marsh, Mitchell Nathan and Daniel Willingham went looking for it, reviewing 10 common learning techniques for a 2013 report in Psychological Science in the Public Interest. Two came out on top. Practice testing and distributed practice, in their words, "received high utility assessments because they benefit learners of different ages and abilities".

Five landed at the bottom: summarisation, highlighting, imagery use for text learning, rereading, and the keyword mnemonic. Highlighting and rereading are the two things almost everyone does, which is the uncomfortable part.

One honest note before you read that as contradicting the last section. The low rating there is for the keyword mnemonic applied to text learning, not the method of loci that Dresler's group tested on ordered word lists. Different tool, different job, and it actually reinforces the point already made above. Mnemonics are narrow. These two are not.

So what are they, in plain terms?

• Practice testing. Trying to retrieve something is itself the thing that strengthens the memory. It isn't a check on whether learning happened, it's where the learning happens. Close the page and try to say it back before you look again.
• Distributed practice. The same total minutes, spread across days instead of piled into one sitting. Nothing extra to buy, nothing extra to do, just arranged differently.

Here's why this belongs in a guide about memory games rather than a study-skills article. Everything else on this page runs into the same wall, which is transfer. You train a game and hope the benefit shows up in your life. It mostly doesn't. Retrieval practice sidesteps that entirely, because the thing you're practising is the actual material you want to remember. There's no transfer gap to cross. The name you're trying to recall is the name you rehearsed.

And you can make it feel like a game if that's what keeps you doing it. Self-quizzing with cards, a spaced-repetition app, covering the page and reciting. Same mechanism, and it's closer to a memory game than anything sold as one, with far better evidence underneath it. If you want the background on the system this is loading, our guide to working memory covers it, and whether brain training works goes deeper on the transfer problem.

Does a Plain Crossword Beat a Brain-Training App?

In the one trial that put them head to head, yes. And the result is awkward enough for the brain-training industry that it's worth walking through properly.

D.P. Devanand, Terry Goldberg and colleagues ran the COG-IT trial, published in NEJM Evidence in October 2022. They took 107 people with mild cognitive impairment and randomly assigned them to one of two things: web-based crossword puzzles, or web-based computerised brain-training games of the kind this guide has been discussing. Fifty-six got crosswords, 51 got the games. Both groups did four 30-minute sessions a week for 12 weeks, then booster weeks at 20, 32, 42, 52, 64 and 78.

The crosswords won. On the ADAS-Cog, the standard cognitive scale used in Alzheimer's research, the difference at 78 weeks was 1.44 points in favour of crosswords (95 percent CI, minus 2.83 to minus 0.06, P = 0.04). The games group declined slightly over that period. The crossword group improved.

Then the finding that made people sit up. The researchers also scanned brains, and decreases in hippocampal volume and cortical thickness were greater in the games group than the crossword group. The hippocampus is the structure most central to forming new memories, and the low-tech option was the one associated with holding on to more of it.

Three caveats before you cancel any subscriptions, because this article has been careful about evidence and shouldn't stop now.

• These weren't healthy adults. Everyone in the trial had mild cognitive impairment. What helps a struggling memory isn't automatically what sharpens a working one, and this guide has made that point in the other direction too.
• The effects were modest. Cohen's d was 0.43 at week 12 and 0.34 at week 78, a medium effect fading to a small one. A 1.44 point ADAS-Cog difference is real and it is not dramatic.
• 107 people is one trial. A single study doesn't overturn a field, and the confidence interval at 78 weeks nearly touches zero. Treat it as a strong signal, not a verdict.

What it does puncture is a specific assumption, and it's the one the whole industry rests on: that a game engineered by cognitive scientists must beat an ordinary puzzle from a newspaper. Here it didn't. The researchers themselves suggested the crossword's advantage might come from engagement and from the way it draws on language and general knowledge you already have, rather than drilling an isolated capacity.

Which is the same shape as the mnemonic result above. The things that work tend to hook into knowledge you already carry, instead of trying to enlarge the container.

Is There a Case for the Games After All?

There is, and leaving it out would make the section above tidier than the evidence actually is. A different research group ran roughly the same head to head and got the opposite winner. Both results can be true, and understanding why is the most useful thing on this page.

Li and colleagues published a randomised trial in Scientific Reports in 2021, splitting 148 healthy young adults into three groups: 50 trained on dual n-back, 50 learned the method of loci, and 48 did nothing. Everyone in the training arms did 30 minutes a day, five days a week, for four weeks.

Dual n-back is worth a sentence, because it's the task most of this research runs on and it sounds harder than it is. A stream of items comes at you and you say whether the current one matches the one from n steps back, with two streams running at once. It gets harder as you improve. Most of the memory games you'll find online are cousins of it.

Here's what they found on the tasks nobody trained. On digit span, both groups improved. On a change detection test, only the n-back group got faster, and clearly so at the highest load. The method of loci group improved on the thing it was built for, abstract word memory, and mostly stayed put elsewhere.

So which study is right? Both, because they asked different questions. Dresler measured how many words off a list you can recall, which is precisely what a memory palace is engineered to do, and technique won by a mile. Li measured whether anything carried over to untrained tasks, and the game-style training moved more of those. A mnemonic is a tool for material you deliberately choose to encode. Process training pushes on the machinery underneath. Neither result tells you the other one is wrong.

Now the caveat, and full credit to the authors for putting it in their own paper. They note that both untrained tests resemble dual n-back in format, which could tilt the comparison toward the process-based group before anyone starts. They call for assessments that cover both training domains properly, and for a wider age range than a sample averaging 21 years old.

That caveat is the whole field in miniature. Pick tests that look like the training and the training looks transferable. Pick tests that don't and it often doesn't. It's the same measurement problem behind every disagreement in this guide, including the duelling expert consensus statements further up.

Where that leaves you, honestly:

• If you want to remember a specific thing, learn the technique. Names, a speech, a shopping list, exam material. The Dresler result holds and it lasted four months.
• If you want the general workout, the games are defensible. Not proven to make you sharper at life, but this is a real randomised trial showing movement on tasks nobody practised.
• Twenty to thirty minutes is the dose that keeps appearing. Both trials landed near it independently, which is at least a reasonable place to start.
• Judge it on whether you keep doing it. A protocol you abandon in week two beats nothing by nothing, whichever arm of whichever study it came from.

What Works Even Better Than Memory Games?

This is the part the honest sources keep coming back to. If your real goal is a sharper, healthier brain, a few things have stronger evidence than any game:

• Aerobic exercise. It has larger, more consistent effects on working memory and processing speed than digital training, partly by boosting blood flow and brain-derived growth factors.
• Sleep. Seven to nine hours is when memories get consolidated. No game offsets chronic sleep loss.
• Learning something genuinely new. A language, an instrument, a complex skill. Novelty challenges the brain more than repeating a task you're already good at.
• Social connection. Regular, meaningful interaction is linked to slower cognitive decline in aging.

None of that makes memory games pointless. They're free, quick, low-risk, and genuinely fun. And the large NIH-funded ACTIVE trial, which followed 2,802 older adults for a decade, found speed-of-processing training cut dementia risk by 29 percent. Read the fine print on that one, though: the same trial tested memory training and reasoning training too, and neither showed the effect. It was specifically the speed work that moved the needle, and even then the result sat right on the edge of significance. Then, in 2026, the same team came back with twenty years of data. Norma Coe and colleagues linked ACTIVE participants to their Medicare records from 1999 to 2019 and published the result in Alzheimer's & Dementia: Translational Research & Clinical Interventions. Among the 2,021 people who could be tracked that way, speed training cut the risk of a dementia diagnosis by about a quarter (hazard ratio 0.75), but only for those who came back for at least one booster session. Speed training without boosters did nothing (hazard ratio 1.01). And once again, there was no effect from memory or reasoning training. So twenty years on, the one kind of training with a dementia signal still isn't the kind most memory games offer. Just use games as one ingredient in a broader routine rather than the whole recipe. Our roundup of the benefits of puzzle games covers where the wider evidence is strongest.

Does Learning a New Skill Beat Any of This?

On the best evidence available, yes. And the study that shows it didn't use anything resembling a brain game. It used quilting and digital photography.

That "learn something genuinely new" bullet in the list above deserves more than a bullet, because there's a proper randomized trial sitting behind it. Denise Park and her team at the University of Texas at Dallas ran the Synapse Project, published in Psychological Science in 2014. Older adults were randomly assigned to high-challenge conditions, meaning learning digital photography, learning to quilt, or both, for 15 hours a week over three months. Others went into low-challenge conditions: socializing, or simple cognitive tasks.

Only the high-challenge group improved. The lab reports gains in episodic memory and neural function for those participants, no cognitive improvement in the low-challenge groups, and some benefits still measurable up to a year later.

Two details in that design matter more than the headline:

• The comparison group wasn't doing nothing. They were socializing and completing simple cognitive tasks. So this isn't the usual "activity beats sitting still" finding. Both groups were busy. Only sustained difficulty moved anything, which is a quietly uncomfortable result for the crossword section further up this page.
• Nothing in the winning condition was a memory exercise. Nobody drilled recall or chased a score. They were programming complicated sewing machines and fighting with camera settings and unfamiliar software. They were, for 15 hours a week, bad at something and required to keep going anyway.

That points at a mechanism worth understanding, because it changes what you'd actually do. The active ingredient doesn't look like memory practice. It looks like sustained effortful learning in a domain where you're incompetent and can't coast. Which is precisely the thing a memory game stops being the moment you get good at it. The game gets easier by design. That's what makes it fun, and it's also what takes it out of the category that produced these results.

Now the honest caveats, because this article has been hard on other people's overselling and shouldn't do its own. Fifteen hours a week for three months is a serious commitment, wildly more than anyone puts into a memory game on their phone. It's one trial. And it doesn't show games are worthless, it shows the ceiling sits somewhere other than where the app stores suggest.

If you want the effect rather than the trivia, the practical version is simple enough. Pick something you're genuinely bad at, that gives you feedback when you get it wrong, and that you can't get through on autopilot. An instrument, a language, a craft with a technical component. Then switch it up when it starts feeling comfortable, because comfortable is the point at which it stops counting.

Does Sleep Do More for Memory Than Any Game?

On the evidence, yes, and it isn't close. The list above gives sleep a single line, which undersells it, because sleep is the only item there with a mechanism this precisely mapped.

And it reframes everything else on this page. The games are the encoding step. Sleep is the step that decides how much of that encoding you still have next week.

What your brain is actually doing at 3am

Filing, not resting. Leanna Keeble, Padraic Monaghan, Edwin Robertson and Sana Hannan reviewed the human EEG evidence in Frontiers in Behavioral Neuroscience in 2025, and their summary is blunt: "SWS plays a pivotal role in memory consolidation."

SWS is slow-wave sleep, the deep stage. What happens in it is a three-way timing act between slow oscillations, sleep spindles and bursts of hippocampal activity called sharp-wave ripples. The review describes "the precise temporal coupling of these oscillatory events" facilitating "memory transfer from the hippocampus to the neocortex, enhancing neuronal reactivation and optimizing long-term memory consolidation."

The coordination is startlingly exact. They note that "SWRs in the hippocampus are nested within spindle troughs, and spindles are thought to coordinate the timing of these ripples to facilitate information transfer."

Read the word transfer again, because it's the important one. What you learned today is sitting in a fragile, hippocampus-dependent form. Deep sleep replays it and moves it somewhere more durable. Skip the sleep and you skip the move.

You can even steer it, which is the genuinely strange part

The technique is called targeted memory reactivation, and it sounds like science fiction until you see how simple it is. The same review describes it as pairing "learning with auditory or olfactory stimuli, with re-exposure to cues during SWS shown to increase hippocampal activation, spindle rates and duration."

Learn something with a particular sound or smell present. Reintroduce that cue quietly while the person is in deep sleep. Recall for that specific material improves.

This has been pooled properly. Xiaoqing Hu, Larry Cheng, Man Hey Chiu and Ken Paller published a meta-analysis in Psychological Bulletin in 2020 covering 91 experiments and 212 effect sizes across roughly 2,004 participants. They report an overall effect of Hedges' g = 0.29, with slow-wave sleep at 0.27 and NREM stage 2 at 0.32.

Now hold that next to the brain-training evidence earlier on this page. A small-to-moderate effect drawn from 91 experiments is firmer ground than most commercial memory-game claims have ever stood on.

Three honest limits before you buy a sleep speaker

An effect around 0.29 is modest. Real, replicated, and not a superpower.

It's also a laboratory technique. Cues get delivered under controlled conditions by researchers watching an EEG to catch the right sleep stage. Playing audio through your bedroom all night is not that experiment, and nobody has shown it works.

And it only strengthens material you already learned. Reactivation is not instruction. Nothing gets into your head that you didn't put there while awake.

What to actually change

• Protect the sleep rather than trying to hack it. The consolidation runs whether or not you engineer any cues. Losing the sleep is what costs you, and no cue replaces it.
• Learn in the evening when you can. If you're going to spend twenty minutes on a memory game or a language app anyway, doing it before bed puts a filing system directly downstream of the effort.
• Never trade sleep for practice. An extra hour of study bought with an hour less sleep is a bad exchange on this evidence, and it's the exchange students and shift workers make constantly.
• Watch what the game does to your bedtime. A puzzle before bed is fine. A bright, engaging one that pushes lights-out back by an hour has just cancelled its own benefit and then some.

Which lands in the same place as the rest of this guide. Games aren't useless, they're just a smaller lever than the free thing you already do every night and probably do badly.

Which Brain Risks Are Actually Worth Your Attention?

If the real worry underneath "should I play memory games" is protecting your memory for the long run, there's a ranked, evidence-based answer to that question. And no game appears anywhere on it.

The reference point is the standing Lancet Commission on dementia prevention, intervention and care. Its 2024 report, led by Gill Livingston and published in The Lancet in August 2024, identifies 14 modifiable risk factors and estimates that 45 percent of future dementia could potentially be prevented if all 14 were eliminated.

The full list: less education, hearing loss, hypertension, smoking, obesity, depression, physical inactivity, diabetes, excessive alcohol, traumatic brain injury, air pollution, and social isolation. The 2024 update added two more, high LDL cholesterol in midlife, contributing an estimated 7 percent, and untreated vision loss in later life at 2 percent.

Read that list slowly, because what's missing is as informative as what's there. Hearing loss is on it. Eyesight is on it. Your cholesterol is on it. "Doesn't play enough brain games" is not, and never has been.

The two newest entries are also the least glamorous and the most actionable. Getting your cholesterol checked and getting your eyes tested are ordinary appointments that most people are already overdue for. Hearing is the same story and has been on the list since 2020, yet hearing loss goes untreated for years as a matter of routine. None of that is as pleasant as a round of digit span, which is precisely why it's worth saying out loud on a page about games.

Two honest limits on that 45 percent, because it's a number that gets repeated carelessly. It's a population-level modelling estimate of what would happen if an entire population eliminated every one of those factors, which is not a thing any individual can do and not a personal guarantee. The Commission presents it as theoretical potential, not a promise. And several entries are associations rather than proven causes, so some may turn out to be early signals of a disease already underway rather than triggers of it.

None of which makes memory games pointless. They're free, quick, and genuinely fun, which this guide has said from the first paragraph and still means. But if you arrived here because you're worried about your memory, the highest-value twenty minutes available to you probably isn't another round of number recall. It's booking a hearing test.

There's a newer wrinkle here that's worth building into your week. Rather than treating the game and the exercise as two separate boxes to tick, the evidence increasingly favours doing both together. A 2025 network meta-analysis in Medicine pooled 16 randomised controlled trials covering 1,149 older adults and sorted the programmes by how many distinct exercise components they included.

The pattern was consistent. Programmes pairing cognitive work with two exercise components improved memory with a standardised mean difference of 0.26, and those with three or more reached 0.43. Single-component programmes showed no significant memory benefit at all. Executive function followed the same shape, going from 0.40 at two components to 0.55 at three or more.

Be careful how you read that, because it's easy to overstate. The review didn't test combined training head to head against playing a memory game on its own, so it can't tell you that adding squats to your Sunday crossword beats the crossword. What it does suggest is that if you're going to pair mental work with movement, variety in the movement matters. Walking plus balance work plus some resistance looks better than walking alone. And those effect sizes are still modest, which is the running theme of this whole page.

Word games make a great companion to memory training, so keep the mental workout going with the free puzzles at Instant Word Finder.