Does brain training increase your IQ?

Last updated 1 August 2026

This is one of the most thoroughly investigated claims in cognitive psychology, and the answer has been stable for over a decade. The interesting part is not the verdict. It is why the improvement people genuinely experience while training does not turn into the thing they were training for.

The distinction the whole question turns on

Psychologists separate two kinds of improvement, and almost every argument about brain training collapses once you hold them apart.

Near transfer is improvement on the trained task and on tasks that closely resemble it. Practise a digit-span exercise and your digit span improves. This is real, measurable, and it happens quickly.

Far transfer is improvement on abilities the training did not directly exercise: reasoning about unfamiliar problems, comprehension, arithmetic, general intelligence. This is what "increase your IQ" means, and it is the claim that has not survived testing.

The commercial confusion is that near transfer feels exactly like far transfer from the inside. Your scores climb week after week and the games get easier. It is genuinely difficult to experience that and conclude nothing general is happening.

The Owen study: 11,430 participants, six weeks, zero far transfer

The most quoted piece of evidence is a 2010 paper in Nature by Adrian Owen and colleagues, Putting brain training to the test. It recruited through a BBC television programme and ended with 11,430 participants who completed the protocol, a sample size that dwarfs almost anything else in the field.

Participants trained online for six weeks on tasks targeting reasoning, memory, planning, visuospatial skills and attention, with benchmark cognitive tests before and after.

Everyone improved on the training tasks. On the benchmark tests measuring general cognitive ability, the study found no transfer effects at all. Not small effects, not effects surfacing in some subgroups. The improvements stayed entirely inside the trained activities.

The paper has been cited more than 500 times and it set the terms of the debate. Its scale matters because the standard objection to a null result is insufficient statistical power, and with eleven thousand participants that objection does not survive.

The meta-analysis: 87 studies, and a control group that matters

Working memory training is the version of the claim with the most theoretical support. Working memory correlates strongly with fluid intelligence, so training capacity ought to raise the correlated ability. It is a reasonable hypothesis, and it is the one Melby-Lervag, Redick and Hulme tested in 2016 in Perspectives on Psychological Science.

They pooled 87 publications containing 145 experimental comparisons, substantially more than any earlier review in the field.

Immediately after training they found reliable improvements on measures of intermediate transfer: verbal and visuospatial working memory. Training working memory improves working memory tasks.

On far transfer — nonverbal ability, verbal ability, word decoding, reading comprehension, arithmetic — there was no convincing evidence of any reliable improvement when training was compared against a treated control condition.

That last clause carries the weight of the entire literature.

Treated controls versus untreated controls

An untreated control group does nothing. A treated control group does something else of comparable intensity and duration: a different set of games, a different activity, the same amount of time and attention.

Compare trainees against people who did nothing and you measure the training plus everything that comes with being in a study — expectation, motivation, familiarity with being tested, the effect of simply showing up. Compare them against a treated control and you isolate the training itself.

Far-transfer effects in this literature tend to appear against untreated controls and vanish against treated ones. That pattern is the signature of an expectancy effect rather than a cognitive one.

What the FTC established about the advertising

In January 2016 the Federal Trade Commission filed a complaint against Lumos Labs, the company behind Lumosity, in the Northern District of California. It concerned marketing claims that the product delayed age-related cognitive decline, reduced the risk of dementia and Alzheimer's, and improved performance at school, work and in sport.

The FTC's position was that the company lacked adequate scientific evidence for those claims. The complaint additionally alleged that some consumer testimonials on the site had been solicited through prize contests, including a free iPad and a trip to San Francisco, without that being disclosed.

Lumos Labs settled. The order required two million dollars in consumer redress, notification of subscribers who joined between 2009 and 2014, and an easy route to cancel auto-renewal. A fifty million dollar judgment was entered and then suspended on the basis of the company's financial condition. Going forward the company was required to hold competent and reliable human clinical evidence before making cognition claims.

Two years earlier, more than seventy cognitive psychologists and neuroscientists organised by the Stanford Center on Longevity and the Max Planck Institute for Human Development had signed a public statement describing the marketed benefits of brain-training products as exaggerated and at times misleading.

What does have evidence behind it

The honest picture is not that nothing affects cognitive performance. It is that the things which do are unglamorous, slow, and mostly not sold as apps.

Education. Additional years of schooling are associated with measurable IQ gains, and the effect survives designs addressing the obvious confound of more able students staying in education longer. It is the best-supported intervention in the field and it works on a timescale of years.

Specific training for a specific outcome. The ACTIVE trial's ten-year follow-up reported reduced dementia risk in a speed-of-processing training arm. Note what that is and is not: a health outcome in older adults from a structured clinical protocol, not an increase in general intelligence from a consumer app.

Removing what depresses performance. Poor sleep, untreated illness, alcohol and acute stress all reduce test performance. Fixing them does not raise your ceiling; it stops you performing below it. Worth doing, but a different claim.

Cardiovascular health, over decades. Aerobic fitness and vascular health show consistent associations with maintained cognition in later life. The effect is on the trajectory of decline rather than on peak ability.

Why your score can still go up

People who retest often score higher. This is genuine but it is not what it looks like.

• Practice effects. Repeat sittings of any cognitive test produce higher scores through familiarity with format, instructions and pacing. The gain is largest between the first and second attempt.
• Test-taking strategy. Learning to skip a stuck item rather than burn four minutes on it can move a timed score several points with no change in ability.
• Condition. Rested and unhurried beats tired and rushed, reliably.
• Measurement error. Every test has a confidence band of several points. Two sittings differing by five points may represent no change whatsoever.

None of these are cheating and none are improvement. They are the gap between your measured score and your actual standing, which is why a single result should be read as an estimate with a band around it rather than a fixed number.

What to do with all this

If your goal is a higher number on an IQ test, practice will get you a modest way there and then stop, and the gain will not follow you to any other test or task.

If your goal is to think better, the evidence points at learning things that are genuinely difficult and genuinely new — a language, an instrument, a technical field — not because it raises a score, but because the accumulated knowledge is itself the useful thing. That is the crystallised side of intelligence, it keeps growing into your sixties, and nobody has ever needed to sell it as a subscription.

Frequently asked questions