Do Puzzle Games Improve Your Brain? What the Research Says

Do puzzle games improve your brain? The honest answer from the research is narrower than either the advertising or the backlash suggests. Playing a puzzle game makes you better at that puzzle and often at closely related ones. Whether that improvement spreads to memory, attention or reasoning in daily life is the question that matters, and the largest controlled test to date found no such spread.

This article walks through that evidence in order: what "improve" has to mean for the claim to be testable, the 11,430-person trial that tested it, the U.S. regulator's action against the most famous brain-training company, the positive results that do exist, the two abilities where training shows something measurable (spatial skill and working memory), and the specific claims that currently lack support. It ends with what all of this means for an arrow puzzle like the one on this site.

The short answer: better at puzzles, unproven for everything else#

Three findings anchor the field. First, practice on a cognitive task reliably improves performance on that task. Nobody disputes this, and it is why any game feels like it is working. Second, improvement on closely related tasks is smaller and less consistent. Third, improvement on distant tasks, such as the memory, attention and reasoning you use at work or school, has little support in well-designed studies.

That three-step pattern comes from the most thorough review of the brain-training literature, published in 2016 by Simons and six colleagues. They examined every peer-reviewed study cited by the leading brain-training companies as evidence, and found extensive evidence that training improves the trained tasks, less for closely related tasks, and little for distant tasks or everyday performance. None of the cited studies met all the best practices the authors set out.

The review also explains why the public argument has been so confused. In 2014 two open letters appeared: a consensus statement from more than 70 scientists saying brain games offered no scientifically grounded way to improve cognition or stave off decline, and a reply from 133 scientists and practitioners saying the literature was full of demonstrated benefits. Both groups read the same studies. The difference, the review argues, was the standard of evidence each side applied.

The practical translation is simple. If the goal is to get good at a specific kind of puzzle, puzzle games work, and they work quickly. If the goal is a generally sharper brain, the evidence is not there yet, and several claims to the contrary have been penalized. The rest of this article is the detail behind that sentence.

Do puzzle games improve your brain? What "improve" has to mean#

The word "improve" hides the whole argument. A person who plays a sliding-arrow puzzle for a month will score higher on sliding-arrow puzzles. That is learning, and it is real. The brain-training claim is stronger: that the practice changes an underlying ability, so that scores rise on tasks the person never practiced. Researchers call that transfer, and they split it into near transfer, to similar tasks, and far transfer, to different ones.

Far transfer is what the advertising sells, because far transfer is what would make a game worth playing for reasons other than fun. It is also what most studies fail to find. A study that reports a gain only on the trained task, or on a test that closely resembles it, has shown learning, not enhancement. Reading any headline about a game "boosting" a mental ability, the first question is which tests were used and how different they were from the game.

The second question is what the comparison group did. A group that played nothing will improve less than a group that did anything at all, simply from expecting to improve and from getting used to the tests. Well-designed studies use an active control, a group that does a comparable but different activity, so that the only difference is the training itself. Many studies in this field did not, which is part of why the 2016 review found their conclusions premature.

The best practices the review lays out can be turned into a short checklist for reading any claim about a game and the brain. A study that fails two or three of these can still be interesting, but it cannot support a claim of general improvement, and by the review's count none of the studies cited by the industry met all of its best practices:

1. Was there an active control group doing something comparable, or did the controls do nothing?
2. Were the outcome tests different in kind from the training, or close cousins of it?
3. Was the sample large enough that a modest effect could be distinguished from noise? Groups of twenty rarely are.
4. Were participants tested again months later, or only at the end of training?
5. Did the people running the study know which group each participant was in when they scored the tests?

The largest test: 11,430 people, six weeks, no transfer#

In 2010 Adrian Owen and colleagues ran a six-week online study that remains the largest direct test of brain training. Participants trained several times a week for six weeks on tasks built to improve reasoning, memory, planning, visuospatial skill and attention. They were tested before and after on separate benchmark tasks measuring the same abilities.

The result, published in Nature, is the sentence the whole field keeps returning to: improvements were observed in every one of the trained tasks, but no evidence was found for transfer to untrained tasks, even when those tasks were cognitively closely related. People got better at what they practiced, by a lot, and that improvement stayed inside the practiced tasks.

The study has limits worth stating. Participants self-selected, training was unsupervised, and six weeks is a short window. Supporters of brain training have argued that longer or more intensive programs might do better, and the authors themselves noted modest effects reported in older adults and preschoolers in earlier work. But no later study of comparable size has overturned the central finding, and a game claiming general benefit has to explain why its transfer should succeed where this trial's did not.

The scale is the point. Small studies produce scattered results in both directions, and a field built on small studies can argue forever, as the 2014 open letters show. A trial with more than eleven thousand participants has the statistical power to detect an effect a fraction the size of what the products promise. It found the trained-task improvement easily. It did not find the transfer.

When a brain-training company was fined for the claim#

Lumosity was the best-known brain-training product of the 2010s, and its advertising made the strongest version of the claim: that playing its games would improve performance at work and school and protect against the effects of aging. In January 2016 the U.S. Federal Trade Commission announced a settlement under which the company agreed to pay $2 million and stop making claims it could not substantiate.

The regulator's language was direct. In its words, the company had preyed on consumers' fears about age-related cognitive decline, suggesting its games could stave off memory loss, dementia and even Alzheimer's disease. The order also required the company to notify subscribers and offer an easy way to cancel.

This matters for anyone evaluating a puzzle game, including the one hosted here, because it sets the standard for what a game may honestly say about itself. A game can say it is fun, that it is hard, that you will get better at it. It cannot say it prevents dementia, raises intelligence or makes you better at your job, because no game has demonstrated any of those things to the standard a regulator or a peer reviewer would accept.

What the positive evidence actually shows#

The negative findings do not mean cognitive training never does anything, and a fair account has to include the results that held up. The strongest is the ACTIVE trial, a randomized study of 2,832 older adults, mean age 73.6, who received ten sessions of memory, reasoning or speed-of-processing training, with booster sessions 11 and 35 months later.

At the ten-year follow-up, reported in 2014 by Rebok and colleagues in the Journal of the American Geriatrics Society, the reasoning and speed groups still showed gains on the trained abilities, and trained participants reported less difficulty with instrumental activities of daily living, the everyday tasks that support independent living. Memory training did not hold.

Spatial skills: where training does transfer#

Spatial skill is the clearest exception to the no-transfer pattern. In 2013 Uttal and six colleagues published in Psychological Bulletin a meta-analysis of 217 studies of spatial skill training. After removing outliers, the average effect of training relative to control was 0.47 in standardized units, a moderate effect by the usual conventions. Training effects were stable across delays between training and testing, and training transferred to other spatial tasks that were not directly trained.

A likely reason is that spatial skill is less a single capacity than a set of learnable procedures. Mentally rotating a shape, holding a viewpoint steady while imagining movement, or reading a two-dimensional drawing as a three-dimensional object are strategies as much as talents, and strategies can be taught.

For games specifically, a 2018 systematic review in Frontiers in Psychology by Pallavicini and colleagues covered 35 studies of video game training in healthy adults. For mental spatial rotation, effect sizes ranged from 0.3 to 3.2, and the authors concluded that both purpose-built and commercial games showed benefits, particularly for young adults. The lower bound is a modest but real effect; the upper bound is implausibly large for a lasting change in ability and almost certainly reflects small samples or tests that closely resembled the game.

The Owen trial trained visuospatial tasks as well, and they did not transfer. Both results can stand because of what was trained. The studies in the meta-analysis often taught explicit strategies over sessions designed by researchers, while the online trial used short computerized tasks and measured transfer with benchmarks deliberately different from the training. Spatial reasoning can be trained under conditions that look like teaching; whether casual play trains it is not established.

Working memory: better scores, no spillover#

Working memory is the small mental workspace where information sits while it is being used, measured with span tasks and n-back tasks. Using it hard is not the same as increasing its capacity, in the way that carrying groceries is not the same as strength training.

In 2013 Melby-Lervåg and Hulme published a meta-analysis in Developmental Psychology of 23 studies with 30 group comparisons. The programs produced reliable short-term improvements in working memory skills, but there was no convincing evidence of generalization to other skills, including nonverbal and verbal ability, inhibitory processes in attention, word decoding and arithmetic. For verbal working memory the gains were not even sustained at follow-up; for visuospatial working memory, limited evidence suggested they might be.

The 2018 review of video games reported memory effect sizes from 0.06 to 1.82. The bottom of that range is nothing and the top is larger than most drug effects; ranges that wide usually mean small studies, different games, different memory tests and no common control condition. The safe reading matches the meta-analysis: scores on the trained kind of task go up, and the step from a better score to a better mind has not been shown.

Notice what these results have in common. They are specific: a trained ability, measured with related tests, in a defined group. None of them is "playing games makes you smarter." The ACTIVE result in particular is about structured sessions with boosters, not about a commercial app, and its memory arm failed. The evidence that exists is narrower and more conditional than any product page would suggest.

Claims that the evidence does not currently support#

Here is a plain list of claims that appear in game marketing and do not have support in well-designed studies as of this writing. This is not a list of claims that have been disproved in every case; it is a list of claims that nobody has demonstrated to an adequate standard.

1. That playing a puzzle game prevents, delays or reduces the risk of dementia or Alzheimer's disease. No trial has shown this for any game.
2. That a game raises general intelligence. Transfer to intelligence tests has been reported in individual studies and has repeatedly failed to replicate at scale.
3. That gains from a game transfer to school grades, job performance or everyday memory. The 2016 review found little evidence for distant transfer and no cited study that met all best practices.
4. That a few minutes a day is a meaningful dose. The trials with any lasting effect used structured sessions with boosters; nothing supports a specific daily minimum.

What health authorities do say is more careful and worth quoting accurately. The World Health Organization's 2026 dementia risk-reduction guidelines recommend cognitive training and cognitive stimulation, and engagement in social activities, for adults with normal cognition or mild cognitive impairment, alongside physical activity, stopping tobacco, diet, and managing blood pressure and diabetes. That is a recommendation about cognitive activity in general, given as one item in a long list. It is not an endorsement of any game, and the trials behind it did not test casual puzzle apps.

A useful habit when a game page cites a study is to find the study and check three things: whether the game in the study is the game on the page, whether the people in the study resemble you, and whether the outcome measured is the outcome promised. Most citations fail at least one. A study of structured reasoning training in adults with a mean age of 73 says nothing about a phone puzzle played by a 30-year-old, and the FTC order exists because that kind of borrowing was the industry norm.

Claim	Status in the research
Practice improves the trained puzzle	Consistently supported
Gains transfer to similar puzzles	Sometimes, smaller
Gains transfer to memory or reasoning in daily life	Little evidence; largest trial found none
Games raise intelligence	Not replicated at scale
Games prevent or delay dementia	No evidence; claim penalized by regulators
Cognitive activity in general is worthwhile	Recommended by WHO alongside exercise, diet, social contact

What this means for an arrow puzzle#

Arrow Escape is a puzzle about reading lines: an arrow leaves the board only if the straight path in front of its head is empty, and every board is packed, so the game is about finding which arrow can move now and predicting what its exit uncovers. Playing it will make you better at exactly that. Players who follow the complete level guide get faster at scanning edges and tracing lines, and the improvement is visible within a few levels.

Whether that skill carries anywhere else is unknown, and this site will not claim that it does. The visuospatial tasks in the Owen trial are the closest cousins to this kind of puzzle, and they improved without transferring. The honest description of what happens when you play Arrow Escape is that you learn a specific, satisfying skill, you spend attention on something with clear rules and immediate feedback, and you stop when the hearts run out. Whether that counts as improving your brain depends on which definition you started with.

That framing is deliberate. The 2016 FTC order is the clearest statement of where the line sits, and it sits well short of the claims most games make. A puzzle that is honest about being a puzzle has nothing to apologize for, and a reader who has followed the evidence this far knows exactly what they are getting: a skill, a pastime, and no promises.

What to read next#

If you would rather form your own opinion at the board, the game is a click away, level 1 has no clock, and the first honest measurement is how much better you are at it after ten levels. The complete level guide shows what those first ten levels contain.