IIF Certified Assessment
Start IQ Test
IIF Certified Assessment
A follow-up published in February 2026 tracked 2,802 adults for two decades after five weeks of cognitive training. One of the three training arms showed a lower rate of dementia diagnosis. The other two showed nothing.
The brain-training industry spent fifteen years making a promise the evidence would not support, and the correction has been just as blunt: play the games and you get better at the games. So a result published in February 2026 is worth reading carefully, because for the first time a randomised trial has followed people for twenty years and found something on the far side of that gap – for exactly one of the three things it tested.
ACTIVE – Advanced Cognitive Training for Independent and Vital Elderly – enrolled 2,802 older adults across six American sites in 1998 and 1999. Each person was randomly assigned to one of four groups: memory training, reasoning training, speed-of-processing training, or a no-contact control group that received nothing. The training itself was short. Up to ten sessions of sixty to seventy-five minutes, spread over five to six weeks. Half of each training group was then randomised again to receive up to four booster sessions, delivered eleven and thirty-five months after the original block.
The twenty-year follow-up, published in Alzheimer's & Dementia: Translational Research & Clinical Interventions, linked those participants to claims-based dementia diagnoses. The headline number belongs to one cell of that design:
That is a genuinely unusual finding and it deserves to be reported as one. It is also a single result inside a trial with many arms, measured through an administrative proxy rather than a clinical assessment, in a cohort that has now been followed long enough for a great deal to have happened to it. Treat it as the strongest signal the field has produced, not as the settled answer.
Every argument about brain training is really an argument about one distinction. Near transfer is improvement on the trained task and on tasks that look almost identical to it. Far transfer is improvement on something the training never touched – your memory for where you left your keys, your reasoning at work, your score on a general ability test. Near transfer is easy to produce and has never been in dispute. Far transfer is the thing being sold, and it is the thing that keeps failing to appear.
Near transfer was never in dispute. Far transfer is what is being sold, and it is what keeps failing to appear.
The most thorough review of the question, published by Simons and colleagues in Psychological Science in the Public Interest in 2016, worked through the published trials and reached a conclusion that has aged well: there is solid evidence that people improve on the tasks they practise, considerably weaker evidence that this spreads to closely related tasks, and little compelling evidence that it improves everyday cognitive performance. Later meta-analytic work has kept landing in the same place. It is also why practice on a test raises the score without raising the ability the score is meant to estimate – the mechanism set out in our piece on what repeat testing does to a cognitive ability score.
In January 2016 the Federal Trade Commission announced a settlement with Lumos Labs, the maker of Lumosity. The company had advertised that training for ten to fifteen minutes, three or four times a week, would help users perform better at work and at school, and would reduce or delay the cognitive impairment associated with ageing and with serious health conditions. The Commission's position was that the company did not have the evidence to say any of it.
Lumos Labs paid $2 million; a $50 million judgment was entered and then suspended because of the company's finances. The part of the settlement that mattered most was not the money but the requirement attached to it – human clinical testing before making that class of claim again. The case remains the cleanest illustration of the distance between what the category advertises and what it can demonstrate. It is worth noting that the one arm of ACTIVE which produced a twenty-year effect was not a commercial product at all.
One further caution is specific to dementia research. A claims-based diagnosis records the moment a doctor coded a condition, and that moment depends on who goes to the doctor, how often, and how alert their clinician is. Groups that differ in engagement with the health system can differ in recorded incidence without differing in underlying disease. The authors are working within the limits of a twenty-year follow-up on a trial designed in the 1990s, and those limits are real ones.
If you want to know what your own reasoning looks like right now, take a properly timed test once and read the result with its confidence range rather than chasing a higher number through practice. Our percentile calculator converts a score into a position in the population, and the tools page explains what the range around it means.
Find your IQ score now! →The questions that separate a real finding from marketing are short, and the ACTIVE follow-up answers all of them, which is why it earns attention. Ask them of anything else you are shown. What exactly was trained, and for how long? What was measured afterwards, and was it something other than the trained task? Was there a control group, and were people randomly assigned to it? How long was the follow-up? And has it been replicated, or is this the first time anyone has seen it?
The honest summary in September 2026 is narrower than either side would like. Commercial brain-training products reliably make you better at their own exercises, and there is still no good evidence that this raises general ability. One tightly specified laboratory intervention, reinforced with boosters, is now associated with a lower rate of recorded dementia twenty years later. Both things are true at once, and the distance between them is the whole story. If you are curious about the underlying trait rather than the training, our piece on when different abilities actually peak is the better guide to what changes with age – and you can take the test or open the percentile calculator to see where your reasoning sits today.
They reliably improve your performance on their own exercises. The evidence that this improvement spreads to untrained everyday abilities – what researchers call far transfer – remains weak, which is what the major review by Simons and colleagues concluded in 2016 and what later meta-analytic work has continued to find. The twenty-year ACTIVE follow-up published in February 2026 is the strongest counter-signal so far, but it involved a specific laboratory speed task with booster sessions, not a commercial app.
There is no good evidence that it can. Training raises scores on the trained task, and repeated practice on a test raises the test score, but neither is the same as raising the underlying general ability an IQ score estimates. No IQ measure was an endpoint in the ACTIVE dementia analysis.
Among participants randomised to speed-of-processing training with booster sessions, dementia was diagnosed in 105 of 264 people (40 per cent), against 239 of 491 in the no-training control group (49 per cent) – roughly a 25 per cent lower incidence. It was the only arm of the trial to show a statistically significant difference; memory training and reasoning training did not.
In January 2016 the Federal Trade Commission settled charges that Lumos Labs had advertised, without adequate supporting evidence, that its games improved performance at work and school and reduced or delayed age-related cognitive impairment. The company paid $2 million, a $50 million judgment was suspended, and it was required to have human clinical testing before making such claims again.
Corrections: spotted an error? Email corrections@iqmetrics.org and we will update this story and note the change here.
Sitting the same test twice usually produces a higher second score. Almost none of that gain is ability — which is why employers use alternate forms, and why the second number is a worse estimate than the first.
Processing speed peaks in your twenties. Vocabulary and knowledge keep climbing into your sixties. A synthesis published in Intelligence puts overall psychological functioning at its highest between 55 and 60 — and your IQ score cannot tell you where you sit on any of those curves.
Every analogy item asks the same thing twice. Name the relation between the first pair in a single sentence, then find the pair that takes the same sentence — and the plausible-looking wrong answers stop being tempting.
Our IIF-certified assessment reports your score with its scale, percentile and confidence range — and a breakdown of the cognitive domains behind it.
Start IQ Test →
