The ACTIVE trial

ACTIVE is a large, long-running randomised trial of brain training in older adults, and the source of the widely repeated claim that a speed-training game cut dementia by 25%. That figure comes from a comparison inside the trial that was not randomised. In the pre-specified comparisons of each whole training group, none significantly reduced dementia over 20 years, and the one randomised booster comparison sat right on the line.

Don't take up speed-of-processing training as dementia prevention on the strength of this trial. Reasoning and speed training durably improved the skill each trained (memory training's gain had faded by ten years); beyond a fragile drop in at-fault car crashes, nothing else measured objectively clearly moved.

What the trial tested

The Advanced Cognitive Training for Independent and Vital Elderly (ACTIVE) study randomised 2,832 cognitively healthy adults aged 65 to 94 to one of four groups: memory training, reasoning training, visual speed-of-processing training, or a no-contact control group. The core intervention was ten group sessions over five to six weeks in 1998–99. Booster training was then offered to a randomly chosen 60% of the trained participants, at 11 months and again at 35 months.^[1]^[2] Only trainees who had completed at least eight of the first ten sessions went into that draw, which matters below.

Speed-of-processing training asks you to identify several visual targets quickly under time pressure and divided attention — picking a target shape out of distractors at the edge of vision, at ever shorter presentation times. The idea is that it works on one specific bottleneck: how fast the brain can split attention across visual targets.

Ten years on: two trained skills lasted, little else did

At ten years, the reasoning and speed groups still outperformed controls on the ability each had trained, but memory training's effect on memory was gone. All three trained groups reported less difficulty with everyday activities, but that benefit was self-reported, not measured by performance.^[3] Training moved the trained thing, durably. It did not obviously move anything else.

Driving: fewer at-fault crashes, on fragile grounds

The trial's most concrete real-world result is on the road. Among 908 participants who were still driving, those in the speed group had about 43% fewer at-fault crashes per mile driven over roughly six years than controls (rate ratio 0.57). Reasoning training did about as well (0.50), which argues against anything specific to visual processing, and memory training did nothing. Crashes of any kind did not fall significantly. The 908 drivers were not a random slice of the trial: they came from four of the six sites, all drove at least about 1,600 km (1,000 miles) a year, people who had done booster sessions were left out, and the mileage was self-reported. The lead author owns stock in the company that sells the training and advises it, as the paper discloses.^[4] A 20-year follow-up, so far only a conference abstract, found the speed-training benefit only in rural drivers (urban drivers showed no clear difference), and the reasoning effect had gone.^[5] How other visual training fares is covered in Vision training.

Twenty years on: where "25% less dementia" comes from

Twenty years after the training, researchers linked the 2,021 participants enrolled in traditional Medicare, the US public health insurance for over-65s, to their insurance claims and tracked dementia diagnoses. Nearly half the control group — 239 of 491 people (48.7%) — was diagnosed with dementia over that period.^[6]

The headline subgroup. Speed trainees who also did booster sessions had about a quarter lower hazard of a dementia diagnosis than the whole control group: hazard ratio 0.75, 95% confidence interval 0.59 to 0.95. A hazard ratio below 1 means a lower rate; the bracketed range is where the true effect most plausibly lies, and when it crosses 1.0 the result is compatible with no effect at all. But this comparison is not randomised. The boosted group had already been filtered for sticking with training; the control group had not. Speed trainees who completed fewer than eight of the first ten sessions were diagnosed with dementia more often than controls (26 of 47, against 239 of 491).
Without boosters, nothing. Speed training without boosters gave a hazard ratio of 1.01 (0.81 to 1.27), effectively identical to control.
The fair test was null. In the pre-specified comparisons of each whole training group against control, every group trended slightly better but none reached statistical significance: memory 0.85 (0.70–1.02), speed 0.87 (0.72–1.05), reasoning 0.88 (0.73–1.07).
The one randomised booster comparison sat on the line. Boosted speed trainees against equally adherent speed trainees who were not drawn for boosters gave 0.81 (0.66–1.00), right on the boundary of no effect.

A reanalysis of the paper's published figures, corrected for the many comparisons made, found that none held up on its own. It noted that 29 of the 30 estimates leaned slightly in favour of training, which points, if anywhere, to a small general effect of any of the three trainings, mostly in participants aged 65 to 69, rather than to anything special about speed training.^[7]

Why to hold the result loosely

Dementia came from insurance claims, not from clinical assessment.
It is a single, unreplicated result, and the investigators themselves call for replication.
The mechanism is unproven. Whether the speed bottleneck explains the conditional signal is not known.
There is a commercial interest. The trial-grade speed-of-processing protocol is sold commercially as BrainHQ, and a co-author holds stock in the vendor. That doesn't make the finding wrong; it is a reason to weight independent replication heavily.

A possible harm in people with less schooling

In a 20-year mortality analysis of the same trial, cognitive training had no effect on mortality overall. But among participants with 12 years of education or less, memory and speed training were associated with higher mortality (hazard ratios 1.64 and 1.60, both intervals excluding no effect), with no such signal above 12 years of education.^[8] The authors caution that the estimates lack precision, and the finding has not been replicated. It raises the possibility that intensive late-life cognitive training behaves differently depending on the reserve a person already has.

What to do

Read "brain training cuts dementia by 25%" as a non-randomised subgroup result, not as a proven effect of the training.
If you enjoy speed-training games, expect to get better at them — that much ACTIVE shows durably — and don't expect protection beyond that.
Put the effort where reserve is better supported — education and demanding work — and into learning genuinely new skills, which is plausible but less proven; both are covered on Cognitive engagement. The wider case against generic apps is on Brain-training apps.