Brain training sharpens the skill practiced, and one type was tied to fewer dementia diagnoses

Brain-training games are sold on a simple promise: practice for a few minutes a day and the mind stays sharp. In January 2016 the US Federal Trade Commission charged that one widely advertised program had promised more than its makers could show. The creators of Lumosity agreed to settle the allegation that they deceived consumers with unfounded claims, among them that the games could delay cognitive decline in old age, and the company agreed to pay $2 million.

The settlement closed a legal case and left the scientific question open. The largest study in the US to assess different types of cognitive training began in 1998 and is still producing results. Its latest report, in 2026, found that one trained group showed a reduced likelihood of being diagnosed with dementia two decades later. Read alongside the reviews that surround it, the trial supports a much narrower claim than the advertising made, though a real one.

What brain training is supposed to do

Cognitive training uses repeated practice on set exercises that target one or more mental abilities, such as memory, reasoning or the speed at which visual information is taken in. Nobody disputes that practice improves the thing practiced. A person who spends a month on a pattern-spotting game will get better at that game.

The question that matters, as Adrian Owen and colleagues at the MRC Cognition and Brain Sciences Unit in Cambridge wrote in Nature in 2010, is whether those benefits transfer to other untrained tasks or lead to any general improvement in the level of cognitive functioning. A program that makes someone quicker at a game and no better at following a recipe, keeping track of bills or driving in traffic has delivered a skill at the game and nothing more. Everything that follows turns on how far the benefit travels.

How the ACTIVE brain-training trial was run

The trial is known as ACTIVE, short for Advanced Cognitive Training for Independent and Vital Elderly. With recruitment from March 1998 through October 1999 in six metropolitan areas of the United States, it enrolled 2,832 adults aged 65 and over who were living independently. Their average age was 73.6, about three quarters were women and 26% were African American. Anyone with significant cognitive problems or a diagnosis of Alzheimer disease was excluded, so this was a study of people who were well at the start.

Participants were randomly assigned to one of four groups. Three received training, in memory, in reasoning or in speed of processing, and the fourth served as a control group that had no contact with the trainers. Training was conducted in small groups in ten sessions of 60 to 75 minutes spread over five to six weeks.

The three courses differed in kind as well as in subject matter. Memory training taught strategies for remembering verbal material, and reasoning training focused on improving the ability to solve problems that contained a serial pattern. Speed training worked differently. It was administered on a computer in an adaptive manner, with visual information presented in successively shorter inspection times and the difficulty rising as each person improved. Booster training was provided at 11 and 35 months after training to a random subset of participants in each training group, 39% of them, drawn from those who had completed at least eight of the ten sessions. Each booster consisted of four further 75-minute sessions.

George Rebok of Johns Hopkins University and colleagues reported what had become of the participants ten years later in the Journal of the American Geriatrics Society in 2014.

What ten years of follow-up showed about brain training

All interventions produced immediate improvement in the trained cognitive ability. A decade on, when the participants were 82 on average, the gains from reasoning and speed training could still be measured. The gain from memory training had lasted to at least five years and was no longer maintained at ten.

Rebok’s team counted how many people still matched or beat their own starting level on the ability they had been trained in. In the reasoning group the figure was 73.6%, against 61.7% of the control group on the same tests. Among those given speed training, 70.7% were performing at or above their respective cognitive ability, against 48.8% of controls. Speed training had the largest effect of the three.

The gains stayed where they were made. The authors wrote that “the benefits of cognitive training are specific to the cognitive ability trained”, which is the same pattern Owen had described: improvement on the task, and little movement beyond it.

Everyday function was the outcome that mattered most, and here the result was split in two.

Group	At or above starting level on the trained ability	Same or less difficulty with daily tasks
Memory training	No clear difference from controls	61.6%
Reasoning training	73.6% (controls 61.7%)	60.2%
Speed training	70.7% (controls 48.8%)	59.5%
No training	Comparison group	49.3%

Participants rated their own difficulty with 19 daily tasks, among them preparing meals, housework, managing finances, shopping and travel. After ten years about 60% of each trained group reported the same or improved level of difficulty as at the start, compared with 49.3% of the control group. On the performance-based measures, in which participants were tested on everyday problems and timed on everyday tasks, there was no effect of training.

The authors’ explanation is that those tests have more in common with standard cognitive tests than with actual acts of daily living, and call on several abilities at once. A less generous reading is also available. The control group had no contact with the trainers, so everyone knew whether they had been trained, and the measure that differed was the one people reported about themselves.

Brain training and dementia after 20 years

Two later analyses asked whether the training changed who went on to develop dementia. In 2017 Jerri Edwards of the University of South Florida and colleagues reported that a total of 260 cases of dementia were identified during the follow-up of ten years, using interviews and test scores in place of a clinical diagnosis. Speed training was followed by a 29% lower risk than no training, a result at the very edge of what statisticians accept as unlikely to be chance. Memory and reasoning training showed no clear difference.

The second analysis used an outcome recorded outside the study. Norma Coe of the University of Pennsylvania and colleagues linked the trial records to Medicare claims from 1999 to 2019 and counted diagnoses of Alzheimer’s disease and related dementias made in the course of ordinary medical care. The analysis covered 2,021 participants, those enrolled in traditional Medicare when the trial began. Over the two decades 77% of them died, at an average age of 83.9.

In the control arm, 239 out of 491 participants were diagnosed with dementia within the 20 years, which is 48.7%. None of the three kinds of training, taken as a whole, changed that figure by more than chance could explain. The exception was a subgroup. People randomized to speed training who completed one or more booster sessions had a risk of diagnosis about 25% lower than the controls, while speed-trained participants with no booster training did not have a lower risk. Johns Hopkins Medicine gave the underlying counts: 105 out of 264 in the speed-training group with boosters were diagnosed with dementia, or 40%.

Marilyn Albert, who directs the Alzheimer’s Disease Research Center at Johns Hopkins Medicine and is the corresponding author, said in the release: “Even small delays in the onset of dementia may have a large impact on public health and help reduce rising health care costs.”

The authors’ account of why boosted speed training stood out rests on how it was delivered. Both the memory and reasoning trainings primarily provided strategies for improving performance, whereas the speed training gave no explicit strategies and grew harder as the person got better. Whether that difference explains the result has not been tested.

Several features of the analysis call for caution, and the authors list them. The finding comes from a subgroup, and boosters were offered only to participants who had completed at least eight initial training sessions, so the boosted group was made up of people who had already turned up for most of the course. Coe’s team checked this by comparing those randomized to boosters with speed-trained participants who were not, and on that comparison the lower risk sat right at the boundary of what chance could produce. The 725 participants enrolled in Medicare Advantage plans were left out. Diagnoses in claims data depend on who reaches the healthcare system and whose family members notice cognitive decline. And the authors acknowledge that, given the large number of publications generated from the ACTIVE study, some results may be false positives.

What reviews of brain-training programs have found

ACTIVE delivered its training in small groups, in scheduled sessions. Commercial programs such as the one in the FTC case are sold as online and mobile app subscriptions, to be used alone, and the evidence for that kind of use is weaker.

In Owen’s 2010 study, 11,430 volunteers recruited through a BBC science show trained online several times each week for six weeks on tasks covering reasoning, memory, planning, visuospatial skills and attention. They improved on every task they trained on. On the benchmark tests no evidence was found for transfer effects to untrained tasks, even when those tasks were closely related to the ones practiced, and a control group that had spent the same weeks answering obscure questions from six different categories improved by about as much.

Daniel Simons of the University of Illinois and six colleagues took a different route in 2016. They examined the published peer-reviewed intervention studies cited on the websites of leading brain-training companies, on the reasoning that these were the studies the companies themselves considered their best support. The reviewers found extensive evidence that training improves performance on the trained tasks, less evidence for closely related tasks, and little evidence that training enhances performance on distantly related tasks or that training improves everyday cognitive performance. None of the cited studies met all of the standards the reviewers regarded as essential.

Pooled analyses of randomized trials give a similar account, with one practical detail added. Amit Lampit and colleagues at the University of Sydney combined 52 studies with 4,885 healthy older participants in 2014 and found a small overall improvement on cognitive tests, and none that could be told apart from chance for executive functions and attention. Home-based administration was ineffective compared to group-based training, and so was training more than three times a week.

A Cochrane review led by Nicola Gates in 2020 looked only at computer-based programs lasting at least 12 weeks in healthy people aged 65 or older, and identified eight trials with a total of 1,183 participants. The reviewers rated the evidence as low or very low quality for all outcomes. It pointed to a slight improvement in overall cognitive function at the end of training, which they described as of uncertain clinical importance, and they found no evidence that the effect persisted 12 months later. No trial in the review reported on quality of life or activities of daily living.

Why a regulator acted on brain-training advertising

The distance between results like these and the marketing is what brought the Federal Trade Commission in. According to the agency’s complaint, the company advertised that training for 10 to 15 minutes three or four times a week would help users reach their full potential, and claimed that scientific studies proved benefits that ran from better performance at work to protection against dementia and Alzheimer’s disease.

“Lumosity simply did not have the science to back up its ads,” said Jessica Rich, director of the agency’s Bureau of Consumer Protection. The proposed court order required the company to have competent and reliable scientific evidence before making future claims about benefits for real-world performance or age-related decline.

The case concerned one product and its advertising, and it was not a judgment on the training used in ACTIVE. What the case and the trial have in common is the standard they point to: a claim about dementia needs a trial that measured dementia.

Where the brain-training evidence runs out

The people in ACTIVE were volunteers well enough to live independently and attend classes. Only 44% were still in the study at ten years, with death the primary reason for leaving. Losses were similar across the four groups, which protects the comparison between them, though those who remained were younger and had fewer health problems at the start than those who dropped out.

The benefit for daily function was, in the authors’ own word, modest, and it rested on self-report. Memory training showed the least of the three courses: its gains had gone by ten years and it showed no link with dementia diagnoses. The dementia finding itself belongs to one arm of one trial, among people who had refresher sessions. Johns Hopkins Medicine describes the 2026 analysis as the first randomized clinical trial, and only study of its kind, to assess 20-year links with dementia, so there is nothing yet to set beside it.

Rebok, quoted in the same release, kept the claim in proportion: “It is possible that adding this cognitive training to lifestyle change interventions may delay dementia onset, but that remains to be studied.”

Practice at a mental task reliably improves that task, and a single trial, not yet repeated, has found fewer dementia diagnoses after one adaptive speed program with refresher sessions.